From 57e252bfa9ffec9f583f486376e43c8b2c0b8809 Mon Sep 17 00:00:00 2001 From: Michael Neumann Date: Wed, 27 May 2015 10:35:36 +0300 Subject: [PATCH 1/1] drm/radeon: Sync to Linux 3.11 --- sys/dev/drm/include/asm/io.h | 5 + sys/dev/drm/include/drm/drm_fixed.h | 94 + sys/dev/drm/include/linux/io.h | 2 +- sys/dev/drm/include/linux/math64.h | 19 +- sys/dev/drm/include/uapi_drm/radeon_drm.h | 32 +- sys/dev/drm/radeon/Makefile | 20 +- sys/dev/drm/radeon/ObjectID.h | 40 + sys/dev/drm/radeon/atom-bits.h | 2 - sys/dev/drm/radeon/atom-names.h | 2 - sys/dev/drm/radeon/atom-types.h | 2 - sys/dev/drm/radeon/atom.c | 5 + sys/dev/drm/radeon/atom.h | 4 +- sys/dev/drm/radeon/atombios.h | 549 +- sys/dev/drm/radeon/atombios_crtc.c | 94 +- sys/dev/drm/radeon/atombios_dp.c | 43 +- sys/dev/drm/radeon/atombios_encoders.c | 51 +- sys/dev/drm/radeon/avivod.h | 2 - sys/dev/drm/radeon/btc_dpm.c | 2740 +++++++ sys/dev/drm/radeon/btc_dpm.h | 57 + sys/dev/drm/radeon/btcd.h | 181 + sys/dev/drm/radeon/cayman_blit_shaders.c | 2 - sys/dev/drm/radeon/cayman_blit_shaders.h | 1 - sys/dev/drm/radeon/cik.c | 7039 +++++++++++++++++ .../{si_blit_shaders.c => cik_blit_shaders.c} | 21 +- .../{si_blit_shaders.h => cik_blit_shaders.h} | 11 +- sys/dev/drm/radeon/cik_reg.h | 147 + sys/dev/drm/radeon/{sid.h => cikd.h} | 1002 ++- sys/dev/drm/radeon/clearstate_cayman.h | 1081 +++ .../{atom-types.h => clearstate_defs.h} | 36 +- sys/dev/drm/radeon/clearstate_evergreen.h | 1080 +++ sys/dev/drm/radeon/clearstate_si.h | 941 +++ sys/dev/drm/radeon/cypress_dpm.c | 2178 +++++ sys/dev/drm/radeon/cypress_dpm.h | 160 + sys/dev/drm/radeon/evergreen.c | 673 +- sys/dev/drm/radeon/evergreen_blit_kms.c | 2 - sys/dev/drm/radeon/evergreen_blit_shaders.c | 2 - sys/dev/drm/radeon/evergreen_blit_shaders.h | 1 - sys/dev/drm/radeon/evergreen_hdmi.c | 45 +- sys/dev/drm/radeon/evergreen_reg.h | 14 +- sys/dev/drm/radeon/evergreen_smc.h | 67 + sys/dev/drm/radeon/evergreend.h | 395 +- sys/dev/drm/radeon/ni.c | 395 +- sys/dev/drm/radeon/ni_dpm.c | 4355 ++++++++++ sys/dev/drm/radeon/ni_dpm.h | 250 + sys/dev/drm/radeon/ni_reg.h | 2 - sys/dev/drm/radeon/nid.h | 583 +- sys/dev/drm/radeon/nislands_smc.h | 328 + sys/dev/drm/radeon/ppsmc.h | 116 + sys/dev/drm/radeon/r100.c | 11 +- sys/dev/drm/radeon/r100_track.h | 1 - sys/dev/drm/radeon/r100d.h | 2 - sys/dev/drm/radeon/r200.c | 3 - sys/dev/drm/radeon/r300_reg.h | 2 - sys/dev/drm/radeon/r300d.h | 2 - sys/dev/drm/radeon/r420d.h | 3 - sys/dev/drm/radeon/r500_reg.h | 2 - sys/dev/drm/radeon/r520.c | 3 - sys/dev/drm/radeon/r520d.h | 2 - sys/dev/drm/radeon/r600.c | 268 +- sys/dev/drm/radeon/r600_blit_shaders.c | 2 - sys/dev/drm/radeon/r600_blit_shaders.h | 1 - sys/dev/drm/radeon/r600_dpm.c | 1049 +++ sys/dev/drm/radeon/r600_dpm.h | 227 + sys/dev/drm/radeon/r600_hdmi.c | 56 +- sys/dev/drm/radeon/r600_reg.h | 8 +- sys/dev/drm/radeon/r600d.h | 238 +- sys/dev/drm/radeon/radeon.h | 557 +- sys/dev/drm/radeon/radeon_acpi.c | 152 + sys/dev/drm/radeon/radeon_asic.c | 730 +- sys/dev/drm/radeon/radeon_asic.h | 220 +- sys/dev/drm/radeon/radeon_atombios.c | 885 ++- sys/dev/drm/radeon/radeon_benchmark.c | 3 - sys/dev/drm/radeon/radeon_blit_common.h | 1 - sys/dev/drm/radeon/radeon_combios.c | 159 +- sys/dev/drm/radeon/radeon_cs.c | 1 + sys/dev/drm/radeon/radeon_cursor.c | 10 +- sys/dev/drm/radeon/radeon_device.c | 107 +- sys/dev/drm/radeon/radeon_display.c | 21 +- sys/dev/drm/radeon/radeon_drv.c | 13 +- sys/dev/drm/radeon/radeon_drv.h | 2 - sys/dev/drm/radeon/radeon_encoders.c | 3 - sys/dev/drm/radeon/radeon_family.h | 5 +- sys/dev/drm/radeon/radeon_fb.c | 2 +- sys/dev/drm/radeon/radeon_gart.c | 6 +- sys/dev/drm/radeon/radeon_irq.c | 2 - sys/dev/drm/radeon/radeon_irq_kms.c | 25 +- sys/dev/drm/radeon/radeon_kms.c | 41 +- sys/dev/drm/radeon/radeon_legacy_crtc.c | 3 - sys/dev/drm/radeon/radeon_legacy_tv.c | 2 - sys/dev/drm/radeon/radeon_mode.h | 94 + sys/dev/drm/radeon/radeon_object.c | 5 +- sys/dev/drm/radeon/radeon_object.h | 2 +- sys/dev/drm/radeon/radeon_pm.c | 740 +- sys/dev/drm/radeon/radeon_prime.c | 21 +- sys/dev/drm/radeon/radeon_reg.h | 4 +- sys/dev/drm/radeon/radeon_ring.c | 52 +- sys/dev/drm/radeon/radeon_sa.c | 9 +- sys/dev/drm/radeon/radeon_trace.h | 2 - sys/dev/drm/radeon/radeon_trace_points.c | 3 - sys/dev/drm/radeon/radeon_ucode.h | 129 + sys/dev/drm/radeon/radeon_uvd.c | 208 +- sys/dev/drm/radeon/rs100d.h | 2 - sys/dev/drm/radeon/rs400d.h | 2 - sys/dev/drm/radeon/rs600d.h | 2 - sys/dev/drm/radeon/rs690.c | 291 +- sys/dev/drm/radeon/rs690d.h | 2 - sys/dev/drm/radeon/rs780_dpm.c | 989 +++ sys/dev/drm/radeon/rs780_dpm.h | 109 + sys/dev/drm/radeon/rs780d.h | 171 + sys/dev/drm/radeon/rv200d.h | 2 - sys/dev/drm/radeon/rv250d.h | 2 - sys/dev/drm/radeon/rv350d.h | 2 - sys/dev/drm/radeon/rv515.c | 224 +- sys/dev/drm/radeon/rv515d.h | 2 - sys/dev/drm/radeon/rv6xx_dpm.c | 2123 +++++ sys/dev/drm/radeon/rv6xx_dpm.h | 95 + sys/dev/drm/radeon/rv6xxd.h | 246 + sys/dev/drm/radeon/rv730_dpm.c | 508 ++ sys/dev/drm/radeon/rv730d.h | 165 + sys/dev/drm/radeon/rv740_dpm.c | 416 + sys/dev/drm/radeon/rv740d.h | 117 + sys/dev/drm/radeon/rv770.c | 17 +- sys/dev/drm/radeon/rv770_dpm.c | 2532 ++++++ sys/dev/drm/radeon/rv770_dpm.h | 290 + sys/dev/drm/radeon/rv770_smc.c | 621 ++ sys/dev/drm/radeon/rv770_smc.h | 209 + sys/dev/drm/radeon/rv770d.h | 285 +- sys/dev/drm/radeon/si.c | 1686 +++- sys/dev/drm/radeon/si_blit_shaders.c | 2 - sys/dev/drm/radeon/si_blit_shaders.h | 1 - sys/dev/drm/radeon/si_dpm.c | 6441 +++++++++++++++ sys/dev/drm/radeon/si_dpm.h | 227 + sys/dev/drm/radeon/si_reg.h | 2 - sys/dev/drm/radeon/si_smc.c | 303 + sys/dev/drm/radeon/sid.h | 619 +- sys/dev/drm/radeon/sislands_smc.h | 396 + sys/dev/drm/radeon/sumo_dpm.c | 1893 +++++ sys/dev/drm/radeon/sumo_dpm.h | 220 + sys/dev/drm/radeon/sumo_smc.c | 221 + sys/dev/drm/radeon/sumod.h | 372 + sys/dev/drm/radeon/trinity_dpm.c | 1953 +++++ sys/dev/drm/radeon/trinity_dpm.h | 132 + sys/dev/drm/radeon/trinity_smc.c | 122 + sys/dev/drm/radeon/trinityd.h | 228 + 144 files changed, 53804 insertions(+), 1408 deletions(-) create mode 100644 sys/dev/drm/radeon/btc_dpm.c create mode 100644 sys/dev/drm/radeon/btc_dpm.h create mode 100644 sys/dev/drm/radeon/btcd.h create mode 100644 sys/dev/drm/radeon/cik.c copy sys/dev/drm/radeon/{si_blit_shaders.c => cik_blit_shaders.c} (93%) copy sys/dev/drm/radeon/{si_blit_shaders.h => cik_blit_shaders.h} (81%) create mode 100644 sys/dev/drm/radeon/cik_reg.h copy sys/dev/drm/radeon/{sid.h => cikd.h} (57%) create mode 100644 sys/dev/drm/radeon/clearstate_cayman.h copy sys/dev/drm/radeon/{atom-types.h => clearstate_defs.h} (72%) create mode 100644 sys/dev/drm/radeon/clearstate_evergreen.h create mode 100644 sys/dev/drm/radeon/clearstate_si.h create mode 100644 sys/dev/drm/radeon/cypress_dpm.c create mode 100644 sys/dev/drm/radeon/cypress_dpm.h create mode 100644 sys/dev/drm/radeon/evergreen_smc.h create mode 100644 sys/dev/drm/radeon/ni_dpm.c create mode 100644 sys/dev/drm/radeon/ni_dpm.h create mode 100644 sys/dev/drm/radeon/nislands_smc.h create mode 100644 sys/dev/drm/radeon/ppsmc.h create mode 100644 sys/dev/drm/radeon/r600_dpm.c create mode 100644 sys/dev/drm/radeon/r600_dpm.h create mode 100644 sys/dev/drm/radeon/radeon_ucode.h create mode 100644 sys/dev/drm/radeon/rs780_dpm.c create mode 100644 sys/dev/drm/radeon/rs780_dpm.h create mode 100644 sys/dev/drm/radeon/rs780d.h create mode 100644 sys/dev/drm/radeon/rv6xx_dpm.c create mode 100644 sys/dev/drm/radeon/rv6xx_dpm.h create mode 100644 sys/dev/drm/radeon/rv6xxd.h create mode 100644 sys/dev/drm/radeon/rv730_dpm.c create mode 100644 sys/dev/drm/radeon/rv730d.h create mode 100644 sys/dev/drm/radeon/rv740_dpm.c create mode 100644 sys/dev/drm/radeon/rv740d.h create mode 100644 sys/dev/drm/radeon/rv770_dpm.c create mode 100644 sys/dev/drm/radeon/rv770_dpm.h create mode 100644 sys/dev/drm/radeon/rv770_smc.c create mode 100644 sys/dev/drm/radeon/rv770_smc.h create mode 100644 sys/dev/drm/radeon/si_dpm.c create mode 100644 sys/dev/drm/radeon/si_dpm.h create mode 100644 sys/dev/drm/radeon/si_smc.c create mode 100644 sys/dev/drm/radeon/sislands_smc.h create mode 100644 sys/dev/drm/radeon/sumo_dpm.c create mode 100644 sys/dev/drm/radeon/sumo_dpm.h create mode 100644 sys/dev/drm/radeon/sumo_smc.c create mode 100644 sys/dev/drm/radeon/sumod.h create mode 100644 sys/dev/drm/radeon/trinity_dpm.c create mode 100644 sys/dev/drm/radeon/trinity_dpm.h create mode 100644 sys/dev/drm/radeon/trinity_smc.c create mode 100644 sys/dev/drm/radeon/trinityd.h diff --git a/sys/dev/drm/include/asm/io.h b/sys/dev/drm/include/asm/io.h index a28bf02374..191df3a833 100644 --- a/sys/dev/drm/include/asm/io.h +++ b/sys/dev/drm/include/asm/io.h @@ -51,6 +51,11 @@ static inline void __iomem *ioremap_wc(resource_size_t phys_addr, unsigned long return pmap_mapdev_attr(phys_addr, size, VM_MEMATTR_WRITE_COMBINING); } +static inline void iounmap(void __iomem *ptr, unsigned long size) +{ + pmap_unmapdev((vm_offset_t) ptr, size); +} + #define mmiowb cpu_sfence #endif /* _ASM_IO_H_ */ diff --git a/sys/dev/drm/include/drm/drm_fixed.h b/sys/dev/drm/include/drm/drm_fixed.h index 0ead502e17..d639049a61 100644 --- a/sys/dev/drm/include/drm/drm_fixed.h +++ b/sys/dev/drm/include/drm/drm_fixed.h @@ -20,10 +20,13 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Dave Airlie + * Christian König */ #ifndef DRM_FIXED_H #define DRM_FIXED_H +#include + typedef union dfixed { u32 full; } fixed20_12; @@ -65,4 +68,95 @@ static inline u32 dfixed_div(fixed20_12 A, fixed20_12 B) tmp /= 2; return lower_32_bits(tmp); } + +#define DRM_FIXED_POINT 32 +#define DRM_FIXED_ONE (1ULL << DRM_FIXED_POINT) +#define DRM_FIXED_DECIMAL_MASK (DRM_FIXED_ONE - 1) +#define DRM_FIXED_DIGITS_MASK (~DRM_FIXED_DECIMAL_MASK) + +static inline s64 drm_int2fixp(int a) +{ + return ((s64)a) << DRM_FIXED_POINT; +} + +static inline int drm_fixp2int(int64_t a) +{ + return ((s64)a) >> DRM_FIXED_POINT; +} + +static inline unsigned drm_fixp_msbset(int64_t a) +{ + unsigned shift, sign = (a >> 63) & 1; + + for (shift = 62; shift > 0; --shift) + if (((a >> shift) & 1) != sign) + return shift; + + return 0; +} + +static inline s64 drm_fixp_mul(s64 a, s64 b) +{ + unsigned shift = drm_fixp_msbset(a) + drm_fixp_msbset(b); + s64 result; + + if (shift > 61) { + shift = shift - 61; + a >>= (shift >> 1) + (shift & 1); + b >>= shift >> 1; + } else + shift = 0; + + result = a * b; + + if (shift > DRM_FIXED_POINT) + return result << (shift - DRM_FIXED_POINT); + + if (shift < DRM_FIXED_POINT) + return result >> (DRM_FIXED_POINT - shift); + + return result; +} + +static inline s64 drm_fixp_div(s64 a, s64 b) +{ + unsigned shift = 62 - drm_fixp_msbset(a); + s64 result; + + a <<= shift; + + if (shift < DRM_FIXED_POINT) + b >>= (DRM_FIXED_POINT - shift); + + result = div64_s64(a, b); + + if (shift > DRM_FIXED_POINT) + return result >> (shift - DRM_FIXED_POINT); + + return result; +} + +static inline s64 drm_fixp_exp(s64 x) +{ + s64 tolerance = div64_s64(DRM_FIXED_ONE, 1000000); + s64 sum = DRM_FIXED_ONE, term, y = x; + u64 count = 1; + + if (x < 0) + y = -1 * x; + + term = y; + + while (term >= tolerance) { + sum = sum + term; + count = count + 1; + term = drm_fixp_mul(term, div64_s64(y, count)); + } + + if (x < 0) + sum = drm_fixp_div(DRM_FIXED_ONE, sum); + + return sum; +} + #endif diff --git a/sys/dev/drm/include/linux/io.h b/sys/dev/drm/include/linux/io.h index f8b99e9479..02c00eb07f 100644 --- a/sys/dev/drm/include/linux/io.h +++ b/sys/dev/drm/include/linux/io.h @@ -36,7 +36,7 @@ #endif static inline void -writel(uint32_t value, void *addr) +writel(uint32_t value, volatile void *addr) { *(volatile uint32_t *)addr = value; } diff --git a/sys/dev/drm/include/linux/math64.h b/sys/dev/drm/include/linux/math64.h index 94650ca4a9..73764e28bd 100644 --- a/sys/dev/drm/include/linux/math64.h +++ b/sys/dev/drm/include/linux/math64.h @@ -49,10 +49,25 @@ */ static inline u64 div_u64_rem(u64 dividend, u32 divisor, u32 *remainder) { - *remainder = dividend % divisor; - return dividend / divisor; + *remainder = dividend % divisor; + return dividend / divisor; } +/** + * div64_u64 - unsigned 64bit divide with 64bit divisor + */ +static inline u64 div64_u64(u64 dividend, u64 divisor) +{ + return dividend / divisor; +} + +/** + * div64_s64 - signed 64bit divide with 64bit divisor + */ +static inline s64 div64_s64(s64 dividend, s64 divisor) +{ + return dividend / divisor; +} #elif BITS_PER_LONG == 32 diff --git a/sys/dev/drm/include/uapi_drm/radeon_drm.h b/sys/dev/drm/include/uapi_drm/radeon_drm.h index 6f884210ce..ca82852c5f 100644 --- a/sys/dev/drm/include/uapi_drm/radeon_drm.h +++ b/sys/dev/drm/include/uapi_drm/radeon_drm.h @@ -974,9 +974,9 @@ struct drm_radeon_cs { /* max SH per SE */ #define RADEON_INFO_MAX_SH_PER_SE 0x13 /* fast fb access is enabled */ -#define RADEON_INFO_FASTFB_WORKING 0x14 +#define RADEON_INFO_FASTFB_WORKING 0x14 /* query if a RADEON_CS_RING_* submission is supported */ -#define RADEON_INFO_RING_WORKING 0x15 +#define RADEON_INFO_RING_WORKING 0x15 /* SI tile mode array */ #define RADEON_INFO_SI_TILE_MODE_ARRAY 0x16 @@ -990,19 +990,19 @@ struct drm_radeon_info { /* Those correspond to the tile index to use, this is to explicitly state * the API that is implicitly defined by the tile mode array. */ -#define SI_TILE_MODE_COLOR_LINEAR_ALIGNED 8 -#define SI_TILE_MODE_COLOR_1D 13 -#define SI_TILE_MODE_COLOR_1D_SCANOUT 9 -#define SI_TILE_MODE_COLOR_2D_8BPP 14 -#define SI_TILE_MODE_COLOR_2D_16BPP 15 -#define SI_TILE_MODE_COLOR_2D_32BPP 16 -#define SI_TILE_MODE_COLOR_2D_64BPP 17 -#define SI_TILE_MODE_COLOR_2D_SCANOUT_16BPP 11 -#define SI_TILE_MODE_COLOR_2D_SCANOUT_32BPP 12 -#define SI_TILE_MODE_DEPTH_STENCIL_1D 4 -#define SI_TILE_MODE_DEPTH_STENCIL_2D 0 -#define SI_TILE_MODE_DEPTH_STENCIL_2D_2AA 3 -#define SI_TILE_MODE_DEPTH_STENCIL_2D_4AA 3 -#define SI_TILE_MODE_DEPTH_STENCIL_2D_8AA 2 +#define SI_TILE_MODE_COLOR_LINEAR_ALIGNED 8 +#define SI_TILE_MODE_COLOR_1D 13 +#define SI_TILE_MODE_COLOR_1D_SCANOUT 9 +#define SI_TILE_MODE_COLOR_2D_8BPP 14 +#define SI_TILE_MODE_COLOR_2D_16BPP 15 +#define SI_TILE_MODE_COLOR_2D_32BPP 16 +#define SI_TILE_MODE_COLOR_2D_64BPP 17 +#define SI_TILE_MODE_COLOR_2D_SCANOUT_16BPP 11 +#define SI_TILE_MODE_COLOR_2D_SCANOUT_32BPP 12 +#define SI_TILE_MODE_DEPTH_STENCIL_1D 4 +#define SI_TILE_MODE_DEPTH_STENCIL_2D 0 +#define SI_TILE_MODE_DEPTH_STENCIL_2D_2AA 3 +#define SI_TILE_MODE_DEPTH_STENCIL_2D_4AA 3 +#define SI_TILE_MODE_DEPTH_STENCIL_2D_8AA 2 #endif diff --git a/sys/dev/drm/radeon/Makefile b/sys/dev/drm/radeon/Makefile index 504e11729e..77acde2fa7 100644 --- a/sys/dev/drm/radeon/Makefile +++ b/sys/dev/drm/radeon/Makefile @@ -75,7 +75,25 @@ SRCS += \ ni.c \ si.c \ si_blit_shaders.c \ - radeon_uvd.c + radeon_uvd.c \ + cik.c \ + cik_blit_shaders.c \ + r600_dpm.c \ + rs780_dpm.c \ + rv6xx_dpm.c \ + rv770_dpm.c \ + rv730_dpm.c \ + rv740_dpm.c \ + rv770_smc.c \ + cypress_dpm.c \ + btc_dpm.c \ + sumo_dpm.c \ + sumo_smc.c \ + trinity_dpm.c \ + trinity_smc.c \ + ni_dpm.c \ + si_smc.c \ + si_dpm.c # UMS driver diff --git a/sys/dev/drm/radeon/ObjectID.h b/sys/dev/drm/radeon/ObjectID.h index 37abd7e2af..73a915531a 100644 --- a/sys/dev/drm/radeon/ObjectID.h +++ b/sys/dev/drm/radeon/ObjectID.h @@ -71,6 +71,8 @@ #define ENCODER_OBJECT_ID_ALMOND 0x22 #define ENCODER_OBJECT_ID_TRAVIS 0x23 #define ENCODER_OBJECT_ID_NUTMEG 0x22 +#define ENCODER_OBJECT_ID_HDMI_ANX9805 0x26 + /* Kaleidoscope (KLDSCP) Class Display Hardware (internal) */ #define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_TMDS1 0x13 #define ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1 0x14 @@ -88,6 +90,8 @@ #define ENCODER_OBJECT_ID_INTERNAL_UNIPHY1 0x20 #define ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 0x21 #define ENCODER_OBJECT_ID_INTERNAL_VCE 0x24 +#define ENCODER_OBJECT_ID_INTERNAL_UNIPHY3 0x25 +#define ENCODER_OBJECT_ID_INTERNAL_AMCLK 0x27 #define ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO 0xFF @@ -366,6 +370,14 @@ GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\ ENCODER_OBJECT_ID_INTERNAL_UNIPHY2 << OBJECT_ID_SHIFT) +#define ENCODER_INTERNAL_UNIPHY3_ENUM_ID1 ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ + ENCODER_OBJECT_ID_INTERNAL_UNIPHY3 << OBJECT_ID_SHIFT) + +#define ENCODER_INTERNAL_UNIPHY3_ENUM_ID2 ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\ + ENCODER_OBJECT_ID_INTERNAL_UNIPHY3 << OBJECT_ID_SHIFT) + #define ENCODER_GENERAL_EXTERNAL_DVO_ENUM_ID1 ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ ENCODER_OBJECT_ID_GENERAL_EXTERNAL_DVO << OBJECT_ID_SHIFT) @@ -394,6 +406,10 @@ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ ENCODER_OBJECT_ID_INTERNAL_VCE << OBJECT_ID_SHIFT) +#define ENCODER_HDMI_ANX9805_ENUM_ID1 ( GRAPH_OBJECT_TYPE_ENCODER << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ + ENCODER_OBJECT_ID_HDMI_ANX9805 << OBJECT_ID_SHIFT) + /****************************************************/ /* Connector Object ID definition - Shared with BIOS */ /****************************************************/ @@ -463,6 +479,14 @@ GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT) +#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID5 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID5 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT) + +#define CONNECTOR_SINGLE_LINK_DVI_D_ENUM_ID6 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID6 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_SINGLE_LINK_DVI_D << OBJECT_ID_SHIFT) + #define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID1 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT) @@ -475,6 +499,10 @@ GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT) +#define CONNECTOR_DUAL_LINK_DVI_D_ENUM_ID4 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_DUAL_LINK_DVI_D << OBJECT_ID_SHIFT) + #define CONNECTOR_VGA_ENUM_ID1 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_VGA << OBJECT_ID_SHIFT) @@ -543,6 +571,18 @@ GRAPH_OBJECT_ENUM_ID3 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT) +#define CONNECTOR_HDMI_TYPE_A_ENUM_ID4 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID4 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT) + +#define CONNECTOR_HDMI_TYPE_A_ENUM_ID5 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID5 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT) + +#define CONNECTOR_HDMI_TYPE_A_ENUM_ID6 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID6 << ENUM_ID_SHIFT |\ + CONNECTOR_OBJECT_ID_HDMI_TYPE_A << OBJECT_ID_SHIFT) + #define CONNECTOR_HDMI_TYPE_B_ENUM_ID1 ( GRAPH_OBJECT_TYPE_CONNECTOR << OBJECT_TYPE_SHIFT |\ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ CONNECTOR_OBJECT_ID_HDMI_TYPE_B << OBJECT_ID_SHIFT) diff --git a/sys/dev/drm/radeon/atom-bits.h b/sys/dev/drm/radeon/atom-bits.h index b106ada8f8..e8fae5c775 100644 --- a/sys/dev/drm/radeon/atom-bits.h +++ b/sys/dev/drm/radeon/atom-bits.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Author: Stanislaw Skowronek - * - * $FreeBSD: head/sys/dev/drm2/radeon/atom-bits.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef ATOM_BITS_H diff --git a/sys/dev/drm/radeon/atom-names.h b/sys/dev/drm/radeon/atom-names.h index 22d17dc68d..6f907a5ffa 100644 --- a/sys/dev/drm/radeon/atom-names.h +++ b/sys/dev/drm/radeon/atom-names.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Author: Stanislaw Skowronek - * - * $FreeBSD: head/sys/dev/drm2/radeon/atom-names.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef ATOM_NAMES_H diff --git a/sys/dev/drm/radeon/atom-types.h b/sys/dev/drm/radeon/atom-types.h index 2a8ae9c5ba..1125b866cd 100644 --- a/sys/dev/drm/radeon/atom-types.h +++ b/sys/dev/drm/radeon/atom-types.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Author: Dave Airlie - * - * $FreeBSD: head/sys/dev/drm2/radeon/atom-types.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef ATOM_TYPES_H diff --git a/sys/dev/drm/radeon/atom.c b/sys/dev/drm/radeon/atom.c index 28bd0b0adc..1b0aaf99c4 100644 --- a/sys/dev/drm/radeon/atom.c +++ b/sys/dev/drm/radeon/atom.c @@ -1219,12 +1219,17 @@ int atom_execute_table(struct atom_context *ctx, int index, uint32_t * params) int r; lockmgr(&ctx->mutex, LK_EXCLUSIVE); + /* reset data block */ + ctx->data_block = 0; /* reset reg block */ ctx->reg_block = 0; /* reset fb window */ ctx->fb_base = 0; /* reset io mode */ ctx->io_mode = ATOM_IO_MM; + /* reset divmul */ + ctx->divmul[0] = 0; + ctx->divmul[1] = 0; r = atom_execute_table_locked(ctx, index, params); lockmgr(&ctx->mutex, LK_RELEASE); return r; diff --git a/sys/dev/drm/radeon/atom.h b/sys/dev/drm/radeon/atom.h index 2553c0c8db..11cb5e211b 100644 --- a/sys/dev/drm/radeon/atom.h +++ b/sys/dev/drm/radeon/atom.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Author: Stanislaw Skowronek - * - * $FreeBSD: head/sys/dev/drm2/radeon/atom.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef ATOM_H @@ -126,7 +124,7 @@ struct card_info { struct atom_context { struct card_info *card; struct lock mutex; - void *bios; + uint8_t *bios; uint32_t cmd_table, data_table; uint16_t *iio; diff --git a/sys/dev/drm/radeon/atombios.h b/sys/dev/drm/radeon/atombios.h index 2d52d030a2..9105d2cc07 100644 --- a/sys/dev/drm/radeon/atombios.h +++ b/sys/dev/drm/radeon/atombios.h @@ -18,8 +18,6 @@ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. - * - * $FreeBSD: head/sys/dev/drm2/radeon/atombios.h 254885 2013-08-25 19:37:15Z dumbbell $ */ @@ -76,6 +74,8 @@ #define ATOM_PPLL2 1 #define ATOM_DCPLL 2 #define ATOM_PPLL0 2 +#define ATOM_PPLL3 3 + #define ATOM_EXT_PLL1 8 #define ATOM_EXT_PLL2 9 #define ATOM_EXT_CLOCK 10 @@ -261,7 +261,7 @@ typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES{ USHORT AdjustDisplayPll; //Atomic Table, used by various SW componentes. USHORT AdjustMemoryController; //Atomic Table, indirectly used by various SW components,called from SetMemoryClock USHORT EnableASIC_StaticPwrMgt; //Atomic Table, only used by Bios - USHORT ASIC_StaticPwrMgtStatusChange; //Obsolete , only used by Bios + USHORT SetUniphyInstance; //Atomic Table, only used by Bios USHORT DAC_LoadDetection; //Atomic Table, directly used by various SW components,latest version 1.2 USHORT LVTMAEncoderControl; //Atomic Table,directly used by various SW components,latest version 1.3 USHORT HW_Misc_Operation; //Atomic Table, directly used by various SW components,latest version 1.1 @@ -273,7 +273,7 @@ typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES{ USHORT TVEncoderControl; //Function Table,directly used by various SW components,latest version 1.1 USHORT PatchMCSetting; //only used by BIOS USHORT MC_SEQ_Control; //only used by BIOS - USHORT TV1OutputControl; //Atomic Table, Obsolete from Ry6xx, use DAC2 Output instead + USHORT Gfx_Harvesting; //Atomic Table, Obsolete from Ry6xx, Now only used by BIOS for GFX harvesting USHORT EnableScaler; //Atomic Table, used only by Bios USHORT BlankCRTC; //Atomic Table, directly used by various SW components,latest version 1.1 USHORT EnableCRTC; //Atomic Table, directly used by various SW components,latest version 1.1 @@ -330,7 +330,7 @@ typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES{ #define UNIPHYTransmitterControl DIG1TransmitterControl #define LVTMATransmitterControl DIG2TransmitterControl #define SetCRTC_DPM_State GetConditionalGoldenSetting -#define SetUniphyInstance ASIC_StaticPwrMgtStatusChange +#define ASIC_StaticPwrMgtStatusChange SetUniphyInstance #define HPDInterruptService ReadHWAssistedI2CStatus #define EnableVGA_Access GetSCLKOverMCLKRatio #define EnableYUV GetDispObjectInfo @@ -340,7 +340,7 @@ typedef struct _ATOM_MASTER_LIST_OF_COMMAND_TABLES{ #define TMDSAEncoderControl PatchMCSetting #define LVDSEncoderControl MC_SEQ_Control #define LCD1OutputControl HW_Misc_Operation - +#define TV1OutputControl Gfx_Harvesting typedef struct _ATOM_MASTER_COMMAND_TABLE { @@ -480,11 +480,11 @@ typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3 typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4 { #if ATOM_BIG_ENDIAN - ULONG ucPostDiv; //return parameter: post divider which is used to program to register directly + ULONG ucPostDiv:8; //return parameter: post divider which is used to program to register directly ULONG ulClock:24; //Input= target clock, output = actual clock #else ULONG ulClock:24; //Input= target clock, output = actual clock - ULONG ucPostDiv; //return parameter: post divider which is used to program to register directly + ULONG ucPostDiv:8; //return parameter: post divider which is used to program to register directly #endif }COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4; @@ -506,6 +506,32 @@ typedef struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V5 UCHAR ucReserved; }COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V5; + +typedef struct _COMPUTE_GPU_CLOCK_INPUT_PARAMETERS_V1_6 +{ + ATOM_COMPUTE_CLOCK_FREQ ulClock; //Input Parameter + ULONG ulReserved[2]; +}COMPUTE_GPU_CLOCK_INPUT_PARAMETERS_V1_6; + +//ATOM_COMPUTE_CLOCK_FREQ.ulComputeClockFlag +#define COMPUTE_GPUCLK_INPUT_FLAG_CLK_TYPE_MASK 0x0f +#define COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK 0x00 +#define COMPUTE_GPUCLK_INPUT_FLAG_SCLK 0x01 + +typedef struct _COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_6 +{ + COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4 ulClock; //Output Parameter: ucPostDiv=DFS divider + ATOM_S_MPLL_FB_DIVIDER ulFbDiv; //Output Parameter: PLL FB divider + UCHAR ucPllRefDiv; //Output Parameter: PLL ref divider + UCHAR ucPllPostDiv; //Output Parameter: PLL post divider + UCHAR ucPllCntlFlag; //Output Flags: control flag + UCHAR ucReserved; +}COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_6; + +//ucPllCntlFlag +#define SPLL_CNTL_FLAG_VCO_MODE_MASK 0x03 + + // ucInputFlag #define ATOM_PLL_INPUT_FLAG_PLL_STROBE_MODE_EN 1 // 1-StrobeMode, 0-PerformanceMode @@ -1688,6 +1714,7 @@ typedef struct _PIXEL_CLOCK_PARAMETERS_V6 #define PIXEL_CLOCK_V6_MISC_HDMI_30BPP 0x08 #define PIXEL_CLOCK_V6_MISC_HDMI_48BPP 0x0c #define PIXEL_CLOCK_V6_MISC_REF_DIV_SRC 0x10 +#define PIXEL_CLOCK_V6_MISC_GEN_DPREFCLK 0x40 typedef struct _GET_DISP_PLL_STATUS_INPUT_PARAMETERS_V2 { @@ -2104,6 +2131,17 @@ typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3 }DVO_ENCODER_CONTROL_PARAMETERS_V3; #define DVO_ENCODER_CONTROL_PS_ALLOCATION_V3 DVO_ENCODER_CONTROL_PARAMETERS_V3 +typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V1_4 +{ + USHORT usPixelClock; + UCHAR ucDVOConfig; + UCHAR ucAction; //ATOM_ENABLE/ATOM_DISABLE/ATOM_HPD_INIT + UCHAR ucBitPerColor; //please refer to definition of PANEL_xBIT_PER_COLOR + UCHAR ucReseved[3]; +}DVO_ENCODER_CONTROL_PARAMETERS_V1_4; +#define DVO_ENCODER_CONTROL_PS_ALLOCATION_V1_4 DVO_ENCODER_CONTROL_PARAMETERS_V1_4 + + //ucTableFormatRevision=1 //ucTableContentRevision=3 structure is not changed but usMisc add bit 1 as another input for // bit1=0: non-coherent mode @@ -2167,7 +2205,7 @@ typedef struct _DVO_ENCODER_CONTROL_PARAMETERS_V3 #define SET_ASIC_VOLTAGE_MODE_SOURCE_B 0x4 #define SET_ASIC_VOLTAGE_MODE_SET_VOLTAGE 0x0 -#define SET_ASIC_VOLTAGE_MODE_GET_GPIOVAL 0x1 +#define SET_ASIC_VOLTAGE_MODE_GET_GPIOVAL 0x1 #define SET_ASIC_VOLTAGE_MODE_GET_GPIOMASK 0x2 typedef struct _SET_VOLTAGE_PARAMETERS @@ -2202,15 +2240,20 @@ typedef struct _SET_VOLTAGE_PARAMETERS_V1_3 //SET_VOLTAGE_PARAMETERS_V3.ucVoltageMode #define ATOM_SET_VOLTAGE 0 //Set voltage Level #define ATOM_INIT_VOLTAGE_REGULATOR 3 //Init Regulator -#define ATOM_SET_VOLTAGE_PHASE 4 //Set Vregulator Phase -#define ATOM_GET_MAX_VOLTAGE 6 //Get Max Voltage, not used in SetVoltageTable v1.3 -#define ATOM_GET_VOLTAGE_LEVEL 6 //Get Voltage level from vitual voltage ID +#define ATOM_SET_VOLTAGE_PHASE 4 //Set Vregulator Phase, only for SVID/PVID regulator +#define ATOM_GET_MAX_VOLTAGE 6 //Get Max Voltage, not used from SetVoltageTable v1.3 +#define ATOM_GET_VOLTAGE_LEVEL 6 //Get Voltage level from vitual voltage ID, not used for SetVoltage v1.4 +#define ATOM_GET_LEAKAGE_ID 8 //Get Leakage Voltage Id ( starting from SMU7x IP ), SetVoltage v1.4 // define vitual voltage id in usVoltageLevel #define ATOM_VIRTUAL_VOLTAGE_ID0 0xff01 #define ATOM_VIRTUAL_VOLTAGE_ID1 0xff02 #define ATOM_VIRTUAL_VOLTAGE_ID2 0xff03 #define ATOM_VIRTUAL_VOLTAGE_ID3 0xff04 +#define ATOM_VIRTUAL_VOLTAGE_ID4 0xff05 +#define ATOM_VIRTUAL_VOLTAGE_ID5 0xff06 +#define ATOM_VIRTUAL_VOLTAGE_ID6 0xff07 +#define ATOM_VIRTUAL_VOLTAGE_ID7 0xff08 typedef struct _SET_VOLTAGE_PS_ALLOCATION { @@ -2630,7 +2673,8 @@ typedef struct _ATOM_FIRMWARE_INFO_V2_2 ULONG ulFirmwareRevision; ULONG ulDefaultEngineClock; //In 10Khz unit ULONG ulDefaultMemoryClock; //In 10Khz unit - ULONG ulReserved[2]; + ULONG ulSPLL_OutputFreq; //In 10Khz unit + ULONG ulGPUPLL_OutputFreq; //In 10Khz unit ULONG ulReserved1; //Was ulMaxEngineClockPLL_Output; //In 10Khz unit* ULONG ulReserved2; //Was ulMaxMemoryClockPLL_Output; //In 10Khz unit* ULONG ulMaxPixelClockPLL_Output; //In 10Khz unit @@ -3815,6 +3859,12 @@ typedef struct _ATOM_GPIO_PIN_ASSIGNMENT UCHAR ucGPIO_ID; }ATOM_GPIO_PIN_ASSIGNMENT; +//ucGPIO_ID pre-define id for multiple usage +//from SMU7.x, if ucGPIO_ID=PP_AC_DC_SWITCH_GPIO_PINID in GPIO_LUTTable, AC/DC swithing feature is enable +#define PP_AC_DC_SWITCH_GPIO_PINID 60 +//from SMU7.x, if ucGPIO_ID=VDDC_REGULATOR_VRHOT_GPIO_PINID in GPIO_LUTable, VRHot feature is enable +#define VDDC_VRHOT_GPIO_PINID 61 + typedef struct _ATOM_GPIO_PIN_LUT { ATOM_COMMON_TABLE_HEADER sHeader; @@ -4076,17 +4126,19 @@ typedef struct _EXT_DISPLAY_PATH //usCaps #define EXT_DISPLAY_PATH_CAPS__HBR2_DISABLE 0x01 +#define EXT_DISPLAY_PATH_CAPS__DP_FIXED_VS_EN 0x02 typedef struct _ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO { ATOM_COMMON_TABLE_HEADER sHeader; UCHAR ucGuid [NUMBER_OF_UCHAR_FOR_GUID]; // a GUID is a 16 byte long string EXT_DISPLAY_PATH sPath[MAX_NUMBER_OF_EXT_DISPLAY_PATH]; // total of fixed 7 entries. - UCHAR ucChecksum; // a simple Checksum of the sum of whole structure equal to 0x0. + UCHAR ucChecksum; // a simple Checksum of the sum of whole structure equal to 0x0. UCHAR uc3DStereoPinId; // use for eDP panel UCHAR ucRemoteDisplayConfig; UCHAR uceDPToLVDSRxId; - UCHAR Reserved[4]; // for potential expansion + UCHAR ucFixDPVoltageSwing; // usCaps[1]=1, this indicate DP_LANE_SET value + UCHAR Reserved[3]; // for potential expansion }ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO; //Related definitions, all records are different but they have a commond header @@ -4418,6 +4470,13 @@ typedef struct _ATOM_VOLTAGE_CONTROL #define VOLTAGE_CONTROL_ID_CHL822x 0x08 #define VOLTAGE_CONTROL_ID_VT1586M 0x09 #define VOLTAGE_CONTROL_ID_UP1637 0x0A +#define VOLTAGE_CONTROL_ID_CHL8214 0x0B +#define VOLTAGE_CONTROL_ID_UP1801 0x0C +#define VOLTAGE_CONTROL_ID_ST6788A 0x0D +#define VOLTAGE_CONTROL_ID_CHLIR3564SVI2 0x0E +#define VOLTAGE_CONTROL_ID_AD527x 0x0F +#define VOLTAGE_CONTROL_ID_NCP81022 0x10 +#define VOLTAGE_CONTROL_ID_LTC2635 0x11 typedef struct _ATOM_VOLTAGE_OBJECT { @@ -4460,6 +4519,15 @@ typedef struct _ATOM_VOLTAGE_OBJECT_HEADER_V3{ USHORT usSize; //Size of Object }ATOM_VOLTAGE_OBJECT_HEADER_V3; +// ATOM_VOLTAGE_OBJECT_HEADER_V3.ucVoltageMode +#define VOLTAGE_OBJ_GPIO_LUT 0 //VOLTAGE and GPIO Lookup table ->ATOM_GPIO_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_VR_I2C_INIT_SEQ 3 //VOLTAGE REGULATOR INIT sequece through I2C -> ATOM_I2C_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_PHASE_LUT 4 //Set Vregulator Phase lookup table ->ATOM_GPIO_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_SVID2 7 //Indicate voltage control by SVID2 ->ATOM_SVID2_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_PWRBOOST_LEAKAGE_LUT 0x10 //Powerboost Voltage and LeakageId lookup table->ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_HIGH_STATE_LEAKAGE_LUT 0x11 //High voltage state Voltage and LeakageId lookup table->ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 +#define VOLTAGE_OBJ_HIGH1_STATE_LEAKAGE_LUT 0x12 //High1 voltage state Voltage and LeakageId lookup table->ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 + typedef struct _VOLTAGE_LUT_ENTRY_V2 { ULONG ulVoltageId; // The Voltage ID which is used to program GPIO register @@ -4475,7 +4543,7 @@ typedef struct _LEAKAGE_VOLTAGE_LUT_ENTRY_V2 typedef struct _ATOM_I2C_VOLTAGE_OBJECT_V3 { - ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; + ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; // voltage mode = VOLTAGE_OBJ_VR_I2C_INIT_SEQ UCHAR ucVoltageRegulatorId; //Indicate Voltage Regulator Id UCHAR ucVoltageControlI2cLine; UCHAR ucVoltageControlAddress; @@ -4486,7 +4554,7 @@ typedef struct _ATOM_I2C_VOLTAGE_OBJECT_V3 typedef struct _ATOM_GPIO_VOLTAGE_OBJECT_V3 { - ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; + ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; // voltage mode = VOLTAGE_OBJ_GPIO_LUT or VOLTAGE_OBJ_PHASE_LUT UCHAR ucVoltageGpioCntlId; // default is 0 which indicate control through CG VID mode UCHAR ucGpioEntryNum; // indiate the entry numbers of Votlage/Gpio value Look up table UCHAR ucPhaseDelay; // phase delay in unit of micro second @@ -4497,7 +4565,7 @@ typedef struct _ATOM_GPIO_VOLTAGE_OBJECT_V3 typedef struct _ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 { - ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; + ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; // voltage mode = 0x10/0x11/0x12 UCHAR ucLeakageCntlId; // default is 0 UCHAR ucLeakageEntryNum; // indicate the entry number of LeakageId/Voltage Lut table UCHAR ucReserved[2]; @@ -4505,10 +4573,26 @@ typedef struct _ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 LEAKAGE_VOLTAGE_LUT_ENTRY_V2 asLeakageIdLut[1]; }ATOM_LEAKAGE_VOLTAGE_OBJECT_V3; + +typedef struct _ATOM_SVID2_VOLTAGE_OBJECT_V3 +{ + ATOM_VOLTAGE_OBJECT_HEADER_V3 sHeader; // voltage mode = VOLTAGE_OBJ_SVID2 +// 14:7 – PSI0_VID +// 6 – PSI0_EN +// 5 – PSI1 +// 4:2 – load line slope trim. +// 1:0 – offset trim, + USHORT usLoadLine_PSI; +// GPU GPIO pin Id to SVID2 regulator VRHot pin. possible value 0~31. 0 means GPIO0, 31 means GPIO31 + UCHAR ucReserved[2]; + ULONG ulReserved; +}ATOM_SVID2_VOLTAGE_OBJECT_V3; + typedef union _ATOM_VOLTAGE_OBJECT_V3{ ATOM_GPIO_VOLTAGE_OBJECT_V3 asGpioVoltageObj; ATOM_I2C_VOLTAGE_OBJECT_V3 asI2cVoltageObj; ATOM_LEAKAGE_VOLTAGE_OBJECT_V3 asLeakageObj; + ATOM_SVID2_VOLTAGE_OBJECT_V3 asSVID2Obj; }ATOM_VOLTAGE_OBJECT_V3; typedef struct _ATOM_VOLTAGE_OBJECT_INFO_V3_1 @@ -4538,6 +4622,21 @@ typedef struct _ATOM_ASIC_PROFILING_INFO ATOM_ASIC_PROFILE_VOLTAGE asVoltage; }ATOM_ASIC_PROFILING_INFO; +typedef struct _ATOM_ASIC_PROFILING_INFO_V2_1 +{ + ATOM_COMMON_TABLE_HEADER asHeader; + UCHAR ucLeakageBinNum; // indicate the entry number of LeakageId/Voltage Lut table + USHORT usLeakageBinArrayOffset; // offset of USHORT Leakage Bin list array ( from lower LeakageId to higher) + + UCHAR ucElbVDDC_Num; + USHORT usElbVDDC_IdArrayOffset; // offset of USHORT virtual VDDC voltage id ( 0xff01~0xff08 ) + USHORT usElbVDDC_LevelArrayOffset; // offset of 2 dimension voltage level USHORT array + + UCHAR ucElbVDDCI_Num; + USHORT usElbVDDCI_IdArrayOffset; // offset of USHORT virtual VDDCI voltage id ( 0xff01~0xff08 ) + USHORT usElbVDDCI_LevelArrayOffset; // offset of 2 dimension voltage level USHORT array +}ATOM_ASIC_PROFILING_INFO_V2_1; + typedef struct _ATOM_POWER_SOURCE_OBJECT { UCHAR ucPwrSrcId; // Power source @@ -4654,6 +4753,8 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 #define SYS_INFO_LVDSMISC__888_BPC 0x04 #define SYS_INFO_LVDSMISC__OVERRIDE_EN 0x08 #define SYS_INFO_LVDSMISC__BLON_ACTIVE_LOW 0x10 +// new since Trinity +#define SYS_INFO_LVDSMISC__TRAVIS_LVDS_VOL_OVERRIDE_EN 0x20 // not used any more #define SYS_INFO_LVDSMISC__VSYNC_ACTIVE_LOW 0x04 @@ -4754,6 +4855,29 @@ typedef struct _ATOM_FUSION_SYSTEM_INFO_V1 ATOM_INTEGRATED_SYSTEM_INFO_V6 sIntegratedSysInfo; ULONG ulPowerplayTable[128]; }ATOM_FUSION_SYSTEM_INFO_V1; + + +typedef struct _ATOM_TDP_CONFIG_BITS +{ +#if ATOM_BIG_ENDIAN + ULONG uReserved:2; + ULONG uTDP_Value:14; // Original TDP value in tens of milli watts + ULONG uCTDP_Value:14; // Override value in tens of milli watts + ULONG uCTDP_Enable:2; // = (uCTDP_Value > uTDP_Value? 2: (uCTDP_Value < uTDP_Value)) +#else + ULONG uCTDP_Enable:2; // = (uCTDP_Value > uTDP_Value? 2: (uCTDP_Value < uTDP_Value)) + ULONG uCTDP_Value:14; // Override value in tens of milli watts + ULONG uTDP_Value:14; // Original TDP value in tens of milli watts + ULONG uReserved:2; +#endif +}ATOM_TDP_CONFIG_BITS; + +typedef union _ATOM_TDP_CONFIG +{ + ATOM_TDP_CONFIG_BITS TDP_config; + ULONG TDP_config_all; +}ATOM_TDP_CONFIG; + /********************************************************************************************************************** ATOM_FUSION_SYSTEM_INFO_V1 Description sIntegratedSysInfo: refer to ATOM_INTEGRATED_SYSTEM_INFO_V6 definition. @@ -4786,7 +4910,8 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 UCHAR ucMemoryType; UCHAR ucUMAChannelNumber; UCHAR strVBIOSMsg[40]; - ULONG ulReserved[20]; + ATOM_TDP_CONFIG asTdpConfig; + ULONG ulReserved[19]; ATOM_AVAILABLE_SCLK_LIST sAvail_SCLK[5]; ULONG ulGMCRestoreResetTime; ULONG ulMinimumNClk; @@ -4811,7 +4936,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 USHORT GnbTdpLimit; USHORT usMaxLVDSPclkFreqInSingleLink; UCHAR ucLvdsMisc; - UCHAR ucLVDSReserved; + UCHAR ucTravisLVDSVolAdjust; UCHAR ucLVDSPwrOnSeqDIGONtoDE_in4Ms; UCHAR ucLVDSPwrOnSeqDEtoVARY_BL_in4Ms; UCHAR ucLVDSPwrOffSeqVARY_BLtoDE_in4Ms; @@ -4819,7 +4944,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 UCHAR ucLVDSOffToOnDelay_in4Ms; UCHAR ucLVDSPwrOnSeqVARY_BLtoBLON_in4Ms; UCHAR ucLVDSPwrOffSeqBLONtoVARY_BL_in4Ms; - UCHAR ucLVDSReserved1; + UCHAR ucMinAllowedBL_Level; ULONG ulLCDBitDepthControlVal; ULONG ulNbpStateMemclkFreq[4]; USHORT usNBP2Voltage; @@ -4848,6 +4973,7 @@ typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 #define SYS_INFO_GPUCAPS__TMDSHDMI_COHERENT_SINGLEPLL_MODE 0x01 #define SYS_INFO_GPUCAPS__DP_SINGLEPLL_MODE 0x02 #define SYS_INFO_GPUCAPS__DISABLE_AUX_MODE_DETECT 0x08 +#define SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS 0x10 /********************************************************************************************************************** ATOM_INTEGRATED_SYSTEM_INFO_V1_7 Description @@ -4947,6 +5073,9 @@ ucLVDSMisc: [bit0] LVDS 888bit panel mode =0: LVDS 888 pan [bit2] LVDS 888bit per color mode =0: 666 bit per color =1:888 bit per color [bit3] LVDS parameter override enable =0: ucLvdsMisc parameter are not used =1: ucLvdsMisc parameter should be used [bit4] Polarity of signal sent to digital BLON output pin. =0: not inverted(active high) =1: inverted ( active low ) + [bit5] Travid LVDS output voltage override enable, when =1, use ucTravisLVDSVolAdjust value to overwrite Traivs register LVDS_CTRL_4 +ucTravisLVDSVolAdjust When ucLVDSMisc[5]=1,it means platform SBIOS want to overwrite TravisLVDSVoltage. Then VBIOS will use ucTravisLVDSVolAdjust + value to program Travis register LVDS_CTRL_4 ucLVDSPwrOnSeqDIGONtoDE_in4Ms: LVDS power up sequence time in unit of 4ms, time delay from DIGON signal active to data enable signal active( DE ). =0 mean use VBIOS default which is 8 ( 32ms ). The LVDS power up sequence is as following: DIGON->DE->VARY_BL->BLON. This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. @@ -4966,18 +5095,241 @@ ucLVDSOffToOnDelay_in4Ms: LVDS power down sequence time in unit of 4ms. =0 means to use VBIOS default delay which is 125 ( 500ms ). This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. -ucLVDSPwrOnVARY_BLtoBLON_in4Ms: LVDS power up sequence time in unit of 4ms. Time delay from VARY_BL signal on to DLON signal active. +ucLVDSPwrOnSeqVARY_BLtoBLON_in4Ms: + LVDS power up sequence time in unit of 4ms. Time delay from VARY_BL signal on to DLON signal active. =0 means to use VBIOS default delay which is 0 ( 0ms ). This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. -ucLVDSPwrOffBLONtoVARY_BL_in4Ms: LVDS power down sequence time in unit of 4ms. Time delay from BLON signal off to VARY_BL signal off. +ucLVDSPwrOffSeqBLONtoVARY_BL_in4Ms: + LVDS power down sequence time in unit of 4ms. Time delay from BLON signal off to VARY_BL signal off. =0 means to use VBIOS default delay which is 0 ( 0ms ). This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucMinAllowedBL_Level: Lowest LCD backlight PWM level. This is customer platform specific parameters. By default it is 0. + ulNbpStateMemclkFreq[4]: system memory clock frequncey in unit of 10Khz in different NB pstate. **********************************************************************************************************************/ +// this IntegrateSystemInfoTable is used for Kaveri & Kabini APU +typedef struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_8 +{ + ATOM_COMMON_TABLE_HEADER sHeader; + ULONG ulBootUpEngineClock; + ULONG ulDentistVCOFreq; + ULONG ulBootUpUMAClock; + ATOM_CLK_VOLT_CAPABILITY sDISPCLK_Voltage[4]; + ULONG ulBootUpReqDisplayVector; + ULONG ulVBIOSMisc; + ULONG ulGPUCapInfo; + ULONG ulDISP_CLK2Freq; + USHORT usRequestedPWMFreqInHz; + UCHAR ucHtcTmpLmt; + UCHAR ucHtcHystLmt; + ULONG ulReserved2; + ULONG ulSystemConfig; + ULONG ulCPUCapInfo; + ULONG ulReserved3; + USHORT usGPUReservedSysMemSize; + USHORT usExtDispConnInfoOffset; + USHORT usPanelRefreshRateRange; + UCHAR ucMemoryType; + UCHAR ucUMAChannelNumber; + UCHAR strVBIOSMsg[40]; + ATOM_TDP_CONFIG asTdpConfig; + ULONG ulReserved[19]; + ATOM_AVAILABLE_SCLK_LIST sAvail_SCLK[5]; + ULONG ulGMCRestoreResetTime; + ULONG ulReserved4; + ULONG ulIdleNClk; + ULONG ulDDR_DLL_PowerUpTime; + ULONG ulDDR_PLL_PowerUpTime; + USHORT usPCIEClkSSPercentage; + USHORT usPCIEClkSSType; + USHORT usLvdsSSPercentage; + USHORT usLvdsSSpreadRateIn10Hz; + USHORT usHDMISSPercentage; + USHORT usHDMISSpreadRateIn10Hz; + USHORT usDVISSPercentage; + USHORT usDVISSpreadRateIn10Hz; + ULONG ulGPUReservedSysMemBaseAddrLo; + ULONG ulGPUReservedSysMemBaseAddrHi; + ULONG ulReserved5[3]; + USHORT usMaxLVDSPclkFreqInSingleLink; + UCHAR ucLvdsMisc; + UCHAR ucTravisLVDSVolAdjust; + UCHAR ucLVDSPwrOnSeqDIGONtoDE_in4Ms; + UCHAR ucLVDSPwrOnSeqDEtoVARY_BL_in4Ms; + UCHAR ucLVDSPwrOffSeqVARY_BLtoDE_in4Ms; + UCHAR ucLVDSPwrOffSeqDEtoDIGON_in4Ms; + UCHAR ucLVDSOffToOnDelay_in4Ms; + UCHAR ucLVDSPwrOnSeqVARY_BLtoBLON_in4Ms; + UCHAR ucLVDSPwrOffSeqBLONtoVARY_BL_in4Ms; + UCHAR ucMinAllowedBL_Level; + ULONG ulLCDBitDepthControlVal; + ULONG ulNbpStateMemclkFreq[4]; + ULONG ulReserved6; + ULONG ulNbpStateNClkFreq[4]; + USHORT usNBPStateVoltage[4]; + USHORT usBootUpNBVoltage; + USHORT usReserved2; + ATOM_EXTERNAL_DISPLAY_CONNECTION_INFO sExtDispConnInfo; +}ATOM_INTEGRATED_SYSTEM_INFO_V1_8; + +/********************************************************************************************************************** + ATOM_INTEGRATED_SYSTEM_INFO_V1_8 Description +ulBootUpEngineClock: VBIOS bootup Engine clock frequency, in 10kHz unit. if it is equal 0, then VBIOS use pre-defined bootup engine clock +ulDentistVCOFreq: Dentist VCO clock in 10kHz unit. +ulBootUpUMAClock: System memory boot up clock frequency in 10Khz unit. +sDISPCLK_Voltage: Report Display clock frequency requirement on GNB voltage(up to 4 voltage levels). + +ulBootUpReqDisplayVector: VBIOS boot up display IDs, following are supported devices in Trinity projects: + ATOM_DEVICE_CRT1_SUPPORT 0x0001 + ATOM_DEVICE_DFP1_SUPPORT 0x0008 + ATOM_DEVICE_DFP6_SUPPORT 0x0040 + ATOM_DEVICE_DFP2_SUPPORT 0x0080 + ATOM_DEVICE_DFP3_SUPPORT 0x0200 + ATOM_DEVICE_DFP4_SUPPORT 0x0400 + ATOM_DEVICE_DFP5_SUPPORT 0x0800 + ATOM_DEVICE_LCD1_SUPPORT 0x0002 + +ulVBIOSMisc: Miscellenous flags for VBIOS requirement and interface + bit[0]=0: INT15 callback function Get LCD EDID ( ax=4e08, bl=1b ) is not supported by SBIOS. + =1: INT15 callback function Get LCD EDID ( ax=4e08, bl=1b ) is supported by SBIOS. + bit[1]=0: INT15 callback function Get boot display( ax=4e08, bl=01h) is not supported by SBIOS + =1: INT15 callback function Get boot display( ax=4e08, bl=01h) is supported by SBIOS + bit[2]=0: INT15 callback function Get panel Expansion ( ax=4e08, bl=02h) is not supported by SBIOS + =1: INT15 callback function Get panel Expansion ( ax=4e08, bl=02h) is supported by SBIOS + bit[3]=0: VBIOS fast boot is disable + =1: VBIOS fast boot is enable. ( VBIOS skip display device detection in every set mode if LCD panel is connect and LID is open) + +ulGPUCapInfo: bit[0~2]= Reserved + bit[3]=0: Enable AUX HW mode detection logic + =1: Disable AUX HW mode detection logic + bit[4]=0: Disable DFS bypass feature + =1: Enable DFS bypass feature + +usRequestedPWMFreqInHz: When it's set to 0x0 by SBIOS: the LCD BackLight is not controlled by GPU(SW). + Any attempt to change BL using VBIOS function or enable VariBri from PP table is not effective since ATOM_BIOS_INFO_BL_CONTROLLED_BY_GPU==0; + + When it's set to a non-zero frequency, the BackLight is controlled by GPU (SW) in one of two ways below: + 1. SW uses the GPU BL PWM output to control the BL, in chis case, this non-zero frequency determines what freq GPU should use; + VBIOS will set up proper PWM frequency and ATOM_BIOS_INFO_BL_CONTROLLED_BY_GPU==1,as the result, + Changing BL using VBIOS function is functional in both driver and non-driver present environment; + and enabling VariBri under the driver environment from PP table is optional. + + 2. SW uses other means to control BL (like DPCD),this non-zero frequency serves as a flag only indicating + that BL control from GPU is expected. + VBIOS will NOT set up PWM frequency but make ATOM_BIOS_INFO_BL_CONTROLLED_BY_GPU==1 + Changing BL using VBIOS function could be functional in both driver and non-driver present environment,but + it's per platform + and enabling VariBri under the driver environment from PP table is optional. + +ucHtcTmpLmt: Refer to D18F3x64 bit[22:16], HtcTmpLmt. Threshold on value to enter HTC_active state. +ucHtcHystLmt: Refer to D18F3x64 bit[27:24], HtcHystLmt. + To calculate threshold off value to exit HTC_active state, which is Threshold on vlaue minus ucHtcHystLmt. + +ulSystemConfig: Bit[0]=0: PCIE Power Gating Disabled + =1: PCIE Power Gating Enabled + Bit[1]=0: DDR-DLL shut-down feature disabled. + 1: DDR-DLL shut-down feature enabled. + Bit[2]=0: DDR-PLL Power down feature disabled. + 1: DDR-PLL Power down feature enabled. + Bit[3]=0: GNB DPM is disabled + =1: GNB DPM is enabled +ulCPUCapInfo: TBD + +usExtDispConnInfoOffset: Offset to sExtDispConnInfo inside the structure +usPanelRefreshRateRange: Bit vector for LCD supported refresh rate range. If DRR is requestd by the platform, at least two bits need to be set + to indicate a range. + SUPPORTED_LCD_REFRESHRATE_30Hz 0x0004 + SUPPORTED_LCD_REFRESHRATE_40Hz 0x0008 + SUPPORTED_LCD_REFRESHRATE_50Hz 0x0010 + SUPPORTED_LCD_REFRESHRATE_60Hz 0x0020 + +ucMemoryType: [3:0]=1:DDR1;=2:DDR2;=3:DDR3;=5:GDDR5; [7:4] is reserved. +ucUMAChannelNumber: System memory channel numbers. + +strVBIOSMsg[40]: VBIOS boot up customized message string + +sAvail_SCLK[5]: Arrays to provide availabe list of SLCK and corresponding voltage, order from low to high + +ulGMCRestoreResetTime: GMC power restore and GMC reset time to calculate data reconnection latency. Unit in ns. +ulIdleNClk: NCLK speed while memory runs in self-refresh state, used to calculate self-refresh latency. Unit in 10kHz. +ulDDR_DLL_PowerUpTime: DDR PHY DLL power up time. Unit in ns. +ulDDR_PLL_PowerUpTime: DDR PHY PLL power up time. Unit in ns. + +usPCIEClkSSPercentage: PCIE Clock Spread Spectrum Percentage in unit 0.01%; 100 mean 1%. +usPCIEClkSSType: PCIE Clock Spread Spectrum Type. 0 for Down spread(default); 1 for Center spread. +usLvdsSSPercentage: LVDS panel ( not include eDP ) Spread Spectrum Percentage in unit of 0.01%, =0, use VBIOS default setting. +usLvdsSSpreadRateIn10Hz: LVDS panel ( not include eDP ) Spread Spectrum frequency in unit of 10Hz, =0, use VBIOS default setting. +usHDMISSPercentage: HDMI Spread Spectrum Percentage in unit 0.01%; 100 mean 1%, =0, use VBIOS default setting. +usHDMISSpreadRateIn10Hz: HDMI Spread Spectrum frequency in unit of 10Hz, =0, use VBIOS default setting. +usDVISSPercentage: DVI Spread Spectrum Percentage in unit 0.01%; 100 mean 1%, =0, use VBIOS default setting. +usDVISSpreadRateIn10Hz: DVI Spread Spectrum frequency in unit of 10Hz, =0, use VBIOS default setting. + +usGPUReservedSysMemSize: Reserved system memory size for ACP engine in APU GNB, units in MB. 0/2/4MB based on CMOS options, current default could be 0MB. KV only, not on KB. +ulGPUReservedSysMemBaseAddrLo: Low 32 bits base address to the reserved system memory. +ulGPUReservedSysMemBaseAddrHi: High 32 bits base address to the reserved system memory. + +usMaxLVDSPclkFreqInSingleLink: Max pixel clock LVDS panel single link, if=0 means VBIOS use default threhold, right now it is 85Mhz +ucLVDSMisc: [bit0] LVDS 888bit panel mode =0: LVDS 888 panel in LDI mode, =1: LVDS 888 panel in FPDI mode + [bit1] LVDS panel lower and upper link mapping =0: lower link and upper link not swap, =1: lower link and upper link are swapped + [bit2] LVDS 888bit per color mode =0: 666 bit per color =1:888 bit per color + [bit3] LVDS parameter override enable =0: ucLvdsMisc parameter are not used =1: ucLvdsMisc parameter should be used + [bit4] Polarity of signal sent to digital BLON output pin. =0: not inverted(active high) =1: inverted ( active low ) + [bit5] Travid LVDS output voltage override enable, when =1, use ucTravisLVDSVolAdjust value to overwrite Traivs register LVDS_CTRL_4 +ucTravisLVDSVolAdjust When ucLVDSMisc[5]=1,it means platform SBIOS want to overwrite TravisLVDSVoltage. Then VBIOS will use ucTravisLVDSVolAdjust + value to program Travis register LVDS_CTRL_4 +ucLVDSPwrOnSeqDIGONtoDE_in4Ms: + LVDS power up sequence time in unit of 4ms, time delay from DIGON signal active to data enable signal active( DE ). + =0 mean use VBIOS default which is 8 ( 32ms ). The LVDS power up sequence is as following: DIGON->DE->VARY_BL->BLON. + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucLVDSPwrOnDEtoVARY_BL_in4Ms: + LVDS power up sequence time in unit of 4ms., time delay from DE( data enable ) active to Vary Brightness enable signal active( VARY_BL ). + =0 mean use VBIOS default which is 90 ( 360ms ). The LVDS power up sequence is as following: DIGON->DE->VARY_BL->BLON. + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucLVDSPwrOffVARY_BLtoDE_in4Ms: + LVDS power down sequence time in unit of 4ms, time delay from data enable ( DE ) signal off to LCDVCC (DIGON) off. + =0 mean use VBIOS default delay which is 8 ( 32ms ). The LVDS power down sequence is as following: BLON->VARY_BL->DE->DIGON + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucLVDSPwrOffDEtoDIGON_in4Ms: + LVDS power down sequence time in unit of 4ms, time delay from vary brightness enable signal( VARY_BL) off to data enable ( DE ) signal off. + =0 mean use VBIOS default which is 90 ( 360ms ). The LVDS power down sequence is as following: BLON->VARY_BL->DE->DIGON + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucLVDSOffToOnDelay_in4Ms: + LVDS power down sequence time in unit of 4ms. Time delay from DIGON signal off to DIGON signal active. + =0 means to use VBIOS default delay which is 125 ( 500ms ). + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucLVDSPwrOnSeqVARY_BLtoBLON_in4Ms: + LVDS power up sequence time in unit of 4ms. Time delay from VARY_BL signal on to DLON signal active. + =0 means to use VBIOS default delay which is 0 ( 0ms ). + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. + +ucLVDSPwrOffSeqBLONtoVARY_BL_in4Ms: + LVDS power down sequence time in unit of 4ms. Time delay from BLON signal off to VARY_BL signal off. + =0 means to use VBIOS default delay which is 0 ( 0ms ). + This parameter is used by VBIOS only. VBIOS will patch LVDS_InfoTable. +ucMinAllowedBL_Level: Lowest LCD backlight PWM level. This is customer platform specific parameters. By default it is 0. + +ulLCDBitDepthControlVal: GPU display control encoder bit dither control setting, used to program register mmFMT_BIT_DEPTH_CONTROL + +ulNbpStateMemclkFreq[4]: system memory clock frequncey in unit of 10Khz in different NB P-State(P0, P1, P2 & P3). +ulNbpStateNClkFreq[4]: NB P-State NClk frequency in different NB P-State +usNBPStateVoltage[4]: NB P-State (P0/P1 & P2/P3) voltage; NBP3 refers to lowes voltage +usBootUpNBVoltage: NB P-State voltage during boot up before driver loaded +sExtDispConnInfo: Display connector information table provided to VBIOS + +**********************************************************************************************************************/ + +// this Table is used for Kaveri/Kabini APU +typedef struct _ATOM_FUSION_SYSTEM_INFO_V2 +{ + ATOM_INTEGRATED_SYSTEM_INFO_V1_8 sIntegratedSysInfo; // refer to ATOM_INTEGRATED_SYSTEM_INFO_V1_8 definition + ULONG ulPowerplayTable[128]; // Update comments here to link new powerplay table definition structure +}ATOM_FUSION_SYSTEM_INFO_V2; + + /**************************************************************************/ // This portion is only used when ext thermal chip or engine/memory clock SS chip is populated on a design //Memory SS Info Table @@ -5028,22 +5380,24 @@ typedef struct _ATOM_ASIC_SS_ASSIGNMENT //Define ucClockIndication, SW uses the IDs below to search if the SS is required/enabled on a clock branch/signal type. //SS is not required or enabled if a match is not found. -#define ASIC_INTERNAL_MEMORY_SS 1 -#define ASIC_INTERNAL_ENGINE_SS 2 -#define ASIC_INTERNAL_UVD_SS 3 -#define ASIC_INTERNAL_SS_ON_TMDS 4 -#define ASIC_INTERNAL_SS_ON_HDMI 5 -#define ASIC_INTERNAL_SS_ON_LVDS 6 -#define ASIC_INTERNAL_SS_ON_DP 7 -#define ASIC_INTERNAL_SS_ON_DCPLL 8 -#define ASIC_EXTERNAL_SS_ON_DP_CLOCK 9 -#define ASIC_INTERNAL_VCE_SS 10 +#define ASIC_INTERNAL_MEMORY_SS 1 +#define ASIC_INTERNAL_ENGINE_SS 2 +#define ASIC_INTERNAL_UVD_SS 3 +#define ASIC_INTERNAL_SS_ON_TMDS 4 +#define ASIC_INTERNAL_SS_ON_HDMI 5 +#define ASIC_INTERNAL_SS_ON_LVDS 6 +#define ASIC_INTERNAL_SS_ON_DP 7 +#define ASIC_INTERNAL_SS_ON_DCPLL 8 +#define ASIC_EXTERNAL_SS_ON_DP_CLOCK 9 +#define ASIC_INTERNAL_VCE_SS 10 +#define ASIC_INTERNAL_GPUPLL_SS 11 + typedef struct _ATOM_ASIC_SS_ASSIGNMENT_V2 { ULONG ulTargetClockRange; //For mem/engine/uvd, Clock Out frequence (VCO ), in unit of 10Khz //For TMDS/HDMI/LVDS, it is pixel clock , for DP, it is link clock ( 27000 or 16200 ) - USHORT usSpreadSpectrumPercentage; //in unit of 0.01% + USHORT usSpreadSpectrumPercentage; //in unit of 0.01% or 0.001%, decided by ucSpreadSpectrumMode bit4 USHORT usSpreadRateIn10Hz; //in unit of 10Hz, modulation freq UCHAR ucClockIndication; //Indicate which clock source needs SS UCHAR ucSpreadSpectrumMode; //Bit0=0 Down Spread,=1 Center Spread, bit1=0: internal SS bit1=1: external SS @@ -5081,6 +5435,11 @@ typedef struct _ATOM_ASIC_SS_ASSIGNMENT_V3 UCHAR ucReserved[2]; }ATOM_ASIC_SS_ASSIGNMENT_V3; +//ATOM_ASIC_SS_ASSIGNMENT_V3.ucSpreadSpectrumMode +#define SS_MODE_V3_CENTRE_SPREAD_MASK 0x01 +#define SS_MODE_V3_EXTERNAL_SS_MASK 0x02 +#define SS_MODE_V3_PERCENTAGE_DIV_BY_1000_MASK 0x10 + typedef struct _ATOM_ASIC_INTERNAL_SS_INFO_V3 { ATOM_COMMON_TABLE_HEADER sHeader; @@ -5721,6 +6080,7 @@ typedef struct _INDIRECT_IO_ACCESS #define INDIRECT_IO_PCIE 3 #define INDIRECT_IO_PCIEP 4 #define INDIRECT_IO_NBMISC 5 +#define INDIRECT_IO_SMU 5 #define INDIRECT_IO_PLL_READ INDIRECT_IO_PLL | INDIRECT_READ #define INDIRECT_IO_PLL_WRITE INDIRECT_IO_PLL | INDIRECT_WRITE @@ -5732,6 +6092,8 @@ typedef struct _INDIRECT_IO_ACCESS #define INDIRECT_IO_PCIEP_WRITE INDIRECT_IO_PCIEP | INDIRECT_WRITE #define INDIRECT_IO_NBMISC_READ INDIRECT_IO_NBMISC | INDIRECT_READ #define INDIRECT_IO_NBMISC_WRITE INDIRECT_IO_NBMISC | INDIRECT_WRITE +#define INDIRECT_IO_SMU_READ INDIRECT_IO_SMU | INDIRECT_READ +#define INDIRECT_IO_SMU_WRITE INDIRECT_IO_SMU | INDIRECT_WRITE typedef struct _ATOM_OEM_INFO { @@ -5877,6 +6239,7 @@ typedef struct _ATOM_MC_INIT_PARAM_TABLE #define _64Mx32 0x43 #define _128Mx8 0x51 #define _128Mx16 0x52 +#define _128Mx32 0x53 #define _256Mx8 0x61 #define _256Mx16 0x62 @@ -5895,6 +6258,8 @@ typedef struct _ATOM_MC_INIT_PARAM_TABLE #define PROMOS MOSEL #define KRETON INFINEON #define ELIXIR NANYA +#define MEZZA ELPIDA + /////////////Support for GDDR5 MC uCode to reside in upper 64K of ROM///////////// @@ -6627,6 +6992,10 @@ typedef struct _ATOM_DISP_OUT_INFO_V3 ASIC_TRANSMITTER_INFO_V2 asTransmitterInfo[1]; // for alligment only }ATOM_DISP_OUT_INFO_V3; +//ucDispCaps +#define DISPLAY_CAPS__DP_PCLK_FROM_PPLL 0x01 +#define DISPLAY_CAPS__FORCE_DISPDEV_CONNECTED 0x02 + typedef enum CORE_REF_CLK_SOURCE{ CLOCK_SRC_XTALIN=0, CLOCK_SRC_XO_IN=1, @@ -6831,6 +7200,17 @@ typedef struct _DIG_TRANSMITTER_INFO_HEADER_V3_1{ USHORT usPhyPllSettingOffset; // offset of CLOCK_CONDITION_SETTING_ENTRY* with Phy Pll Settings }DIG_TRANSMITTER_INFO_HEADER_V3_1; +typedef struct _DIG_TRANSMITTER_INFO_HEADER_V3_2{ + ATOM_COMMON_TABLE_HEADER sHeader; + USHORT usDPVsPreEmphSettingOffset; // offset of PHY_ANALOG_SETTING_INFO * with DP Voltage Swing and Pre-Emphasis for each Link clock + USHORT usPhyAnalogRegListOffset; // offset of CLOCK_CONDITION_REGESTER_INFO* with None-DP mode Analog Setting's register Info + USHORT usPhyAnalogSettingOffset; // offset of CLOCK_CONDITION_SETTING_ENTRY* with None-DP mode Analog Setting for each link clock range + USHORT usPhyPllRegListOffset; // offset of CLOCK_CONDITION_REGESTER_INFO* with Phy Pll register Info + USHORT usPhyPllSettingOffset; // offset of CLOCK_CONDITION_SETTING_ENTRY* with Phy Pll Settings + USHORT usDPSSRegListOffset; // offset of CLOCK_CONDITION_REGESTER_INFO* with Phy SS Pll register Info + USHORT usDPSSSettingOffset; // offset of CLOCK_CONDITION_SETTING_ENTRY* with Phy SS Pll Settings +}DIG_TRANSMITTER_INFO_HEADER_V3_2; + typedef struct _CLOCK_CONDITION_REGESTER_INFO{ USHORT usRegisterIndex; UCHAR ucStartBit; @@ -6854,12 +7234,24 @@ typedef struct _PHY_CONDITION_REG_VAL{ ULONG ulRegVal; }PHY_CONDITION_REG_VAL; +typedef struct _PHY_CONDITION_REG_VAL_V2{ + ULONG ulCondition; + UCHAR ucCondition2; + ULONG ulRegVal; +}PHY_CONDITION_REG_VAL_V2; + typedef struct _PHY_CONDITION_REG_INFO{ USHORT usRegIndex; USHORT usSize; PHY_CONDITION_REG_VAL asRegVal[1]; }PHY_CONDITION_REG_INFO; +typedef struct _PHY_CONDITION_REG_INFO_V2{ + USHORT usRegIndex; + USHORT usSize; + PHY_CONDITION_REG_VAL_V2 asRegVal[1]; +}PHY_CONDITION_REG_INFO_V2; + typedef struct _PHY_ANALOG_SETTING_INFO{ UCHAR ucEncodeMode; UCHAR ucPhySel; @@ -6867,6 +7259,25 @@ typedef struct _PHY_ANALOG_SETTING_INFO{ PHY_CONDITION_REG_INFO asAnalogSetting[1]; }PHY_ANALOG_SETTING_INFO; +typedef struct _PHY_ANALOG_SETTING_INFO_V2{ + UCHAR ucEncodeMode; + UCHAR ucPhySel; + USHORT usSize; + PHY_CONDITION_REG_INFO_V2 asAnalogSetting[1]; +}PHY_ANALOG_SETTING_INFO_V2; + +typedef struct _GFX_HAVESTING_PARAMETERS { + UCHAR ucGfxBlkId; //GFX blk id to be harvested, like CU, RB or PRIM + UCHAR ucReserved; //reserved + UCHAR ucActiveUnitNumPerSH; //requested active CU/RB/PRIM number per shader array + UCHAR ucMaxUnitNumPerSH; //max CU/RB/PRIM number per shader array +} GFX_HAVESTING_PARAMETERS; + +//ucGfxBlkId +#define GFX_HARVESTING_CU_ID 0 +#define GFX_HARVESTING_RB_ID 1 +#define GFX_HARVESTING_PRIM_ID 2 + /****************************************************************************/ //Portion VI: Definitinos for vbios MC scratch registers that driver used /****************************************************************************/ @@ -6877,8 +7288,17 @@ typedef struct _PHY_ANALOG_SETTING_INFO{ #define MC_MISC0__MEMORY_TYPE__GDDR3 0x30000000 #define MC_MISC0__MEMORY_TYPE__GDDR4 0x40000000 #define MC_MISC0__MEMORY_TYPE__GDDR5 0x50000000 +#define MC_MISC0__MEMORY_TYPE__HBM 0x60000000 #define MC_MISC0__MEMORY_TYPE__DDR3 0xB0000000 +#define ATOM_MEM_TYPE_DDR_STRING "DDR" +#define ATOM_MEM_TYPE_DDR2_STRING "DDR2" +#define ATOM_MEM_TYPE_GDDR3_STRING "GDDR3" +#define ATOM_MEM_TYPE_GDDR4_STRING "GDDR4" +#define ATOM_MEM_TYPE_GDDR5_STRING "GDDR5" +#define ATOM_MEM_TYPE_HBM_STRING "HBM" +#define ATOM_MEM_TYPE_DDR3_STRING "DDR3" + /****************************************************************************/ //Portion VI: Definitinos being oboselete /****************************************************************************/ @@ -7276,6 +7696,7 @@ typedef struct _ATOM_PPLIB_THERMALCONTROLLER #define ATOM_PP_THERMALCONTROLLER_NISLANDS 15 #define ATOM_PP_THERMALCONTROLLER_SISLANDS 16 #define ATOM_PP_THERMALCONTROLLER_LM96163 17 +#define ATOM_PP_THERMALCONTROLLER_CISLANDS 18 // Thermal controller 'combo type' to use an external controller for Fan control and an internal controller for thermal. // We probably should reserve the bit 0x80 for this use. @@ -7318,6 +7739,8 @@ typedef struct _ATOM_PPLIB_EXTENDEDHEADER // Add extra system parameters here, always adjust size to include all fields. USHORT usVCETableOffset; //points to ATOM_PPLIB_VCE_Table USHORT usUVDTableOffset; //points to ATOM_PPLIB_UVD_Table + USHORT usSAMUTableOffset; //points to ATOM_PPLIB_SAMU_Table + USHORT usPPMTableOffset; //points to ATOM_PPLIB_PPM_Table } ATOM_PPLIB_EXTENDEDHEADER; //// ATOM_PPLIB_POWERPLAYTABLE::ulPlatformCaps @@ -7339,7 +7762,10 @@ typedef struct _ATOM_PPLIB_EXTENDEDHEADER #define ATOM_PP_PLATFORM_CAP_VDDCI_CONTROL 0x8000 // Does the driver control VDDCI independently from VDDC. #define ATOM_PP_PLATFORM_CAP_REGULATOR_HOT 0x00010000 // Enable the 'regulator hot' feature. #define ATOM_PP_PLATFORM_CAP_BACO 0x00020000 // Does the driver supports BACO state. - +#define ATOM_PP_PLATFORM_CAP_NEW_CAC_VOLTAGE 0x00040000 // Does the driver supports new CAC voltage table. +#define ATOM_PP_PLATFORM_CAP_REVERT_GPIO5_POLARITY 0x00080000 // Does the driver supports revert GPIO5 polarity. +#define ATOM_PP_PLATFORM_CAP_OUTPUT_THERMAL2GPIO17 0x00100000 // Does the driver supports thermal2GPIO17. +#define ATOM_PP_PLATFORM_CAP_VRHOT_GPIO_CONFIGURABLE 0x00200000 // Does the driver supports VR HOT GPIO Configurable. typedef struct _ATOM_PPLIB_POWERPLAYTABLE { @@ -7400,7 +7826,7 @@ typedef struct _ATOM_PPLIB_POWERPLAYTABLE4 USHORT usVddcDependencyOnMCLKOffset; USHORT usMaxClockVoltageOnDCOffset; USHORT usVddcPhaseShedLimitsTableOffset; // Points to ATOM_PPLIB_PhaseSheddingLimits_Table - USHORT usReserved; + USHORT usMvddDependencyOnMCLKOffset; } ATOM_PPLIB_POWERPLAYTABLE4, *LPATOM_PPLIB_POWERPLAYTABLE4; typedef struct _ATOM_PPLIB_POWERPLAYTABLE5 @@ -7565,6 +7991,17 @@ typedef struct _ATOM_PPLIB_SI_CLOCK_INFO } ATOM_PPLIB_SI_CLOCK_INFO; +typedef struct _ATOM_PPLIB_CI_CLOCK_INFO +{ + USHORT usEngineClockLow; + UCHAR ucEngineClockHigh; + + USHORT usMemoryClockLow; + UCHAR ucMemoryClockHigh; + + UCHAR ucPCIEGen; + USHORT usPCIELane; +} ATOM_PPLIB_CI_CLOCK_INFO; typedef struct _ATOM_PPLIB_RS780_CLOCK_INFO @@ -7682,8 +8119,8 @@ typedef struct _ATOM_PPLIB_Clock_Voltage_Limit_Table typedef struct _ATOM_PPLIB_CAC_Leakage_Record { - USHORT usVddc; // We use this field for the "fake" standardized VDDC for power calculations - ULONG ulLeakageValue; + USHORT usVddc; // We use this field for the "fake" standardized VDDC for power calculations; For CI and newer, we use this as the real VDDC value. + ULONG ulLeakageValue; // For CI and newer we use this as the "fake" standar VDDC value. }ATOM_PPLIB_CAC_Leakage_Record; typedef struct _ATOM_PPLIB_CAC_Leakage_Table @@ -7798,6 +8235,42 @@ typedef struct _ATOM_PPLIB_UVD_Table // ATOM_PPLIB_UVD_State_Table states; }ATOM_PPLIB_UVD_Table; + +typedef struct _ATOM_PPLIB_SAMClk_Voltage_Limit_Record +{ + USHORT usVoltage; + USHORT usSAMClockLow; + UCHAR ucSAMClockHigh; +}ATOM_PPLIB_SAMClk_Voltage_Limit_Record; + +typedef struct _ATOM_PPLIB_SAMClk_Voltage_Limit_Table{ + UCHAR numEntries; + ATOM_PPLIB_SAMClk_Voltage_Limit_Record entries[1]; +}ATOM_PPLIB_SAMClk_Voltage_Limit_Table; + +typedef struct _ATOM_PPLIB_SAMU_Table +{ + UCHAR revid; + ATOM_PPLIB_SAMClk_Voltage_Limit_Table limits; +}ATOM_PPLIB_SAMU_Table; + +#define ATOM_PPM_A_A 1 +#define ATOM_PPM_A_I 2 +typedef struct _ATOM_PPLIB_PPM_Table +{ + UCHAR ucRevId; + UCHAR ucPpmDesign; //A+I or A+A + USHORT usCpuCoreNumber; + ULONG ulPlatformTDP; + ULONG ulSmallACPlatformTDP; + ULONG ulPlatformTDC; + ULONG ulSmallACPlatformTDC; + ULONG ulApuTDP; + ULONG ulDGpuTDP; + ULONG ulDGpuUlvPower; + ULONG ulTjmax; +} ATOM_PPLIB_PPM_Table; + /**************************************************************************/ diff --git a/sys/dev/drm/radeon/atombios_crtc.c b/sys/dev/drm/radeon/atombios_crtc.c index b0b1b62121..0e8e4086aa 100644 --- a/sys/dev/drm/radeon/atombios_crtc.c +++ b/sys/dev/drm/radeon/atombios_crtc.c @@ -22,10 +22,7 @@ * * Authors: Dave Airlie * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/atombios_crtc.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include @@ -558,7 +555,7 @@ static u32 atombios_adjust_pll(struct drm_crtc *crtc, if (rdev->family < CHIP_RV770) radeon_crtc->pll_flags |= RADEON_PLL_PREFER_MINM_OVER_MAXP; /* use frac fb div on APUs */ - if (ASIC_IS_DCE41(rdev) || ASIC_IS_DCE61(rdev)) + if (ASIC_IS_DCE41(rdev) || ASIC_IS_DCE61(rdev) || ASIC_IS_DCE8(rdev)) radeon_crtc->pll_flags |= RADEON_PLL_USE_FRAC_FB_DIV; /* use frac fb div on RS780/RS880 */ if ((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880)) @@ -746,7 +743,7 @@ static void atombios_crtc_set_disp_eng_pll(struct radeon_device *rdev, * SetPixelClock provides the dividers */ args.v6.ulDispEngClkFreq = cpu_to_le32(dispclk); - if (ASIC_IS_DCE61(rdev)) + if (ASIC_IS_DCE61(rdev) || ASIC_IS_DCE8(rdev)) args.v6.ucPpll = ATOM_EXT_PLL1; else if (ASIC_IS_DCE6(rdev)) args.v6.ucPpll = ATOM_PPLL0; @@ -1146,7 +1143,9 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc, } if (tiling_flags & RADEON_TILING_MACRO) { - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + tmp = rdev->config.cik.tile_config; + else if (rdev->family >= CHIP_TAHITI) tmp = rdev->config.si.tile_config; else if (rdev->family >= CHIP_CAYMAN) tmp = rdev->config.cayman.tile_config; @@ -1173,11 +1172,29 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc, fb_format |= EVERGREEN_GRPH_BANK_WIDTH(bankw); fb_format |= EVERGREEN_GRPH_BANK_HEIGHT(bankh); fb_format |= EVERGREEN_GRPH_MACRO_TILE_ASPECT(mtaspect); + if (rdev->family >= CHIP_BONAIRE) { + /* XXX need to know more about the surface tiling mode */ + fb_format |= CIK_GRPH_MICRO_TILE_MODE(CIK_DISPLAY_MICRO_TILING); + } } else if (tiling_flags & RADEON_TILING_MICRO) fb_format |= EVERGREEN_GRPH_ARRAY_MODE(EVERGREEN_GRPH_ARRAY_1D_TILED_THIN1); - if ((rdev->family == CHIP_TAHITI) || - (rdev->family == CHIP_PITCAIRN)) + if (rdev->family >= CHIP_BONAIRE) { + u32 num_pipe_configs = rdev->config.cik.max_tile_pipes; + u32 num_rb = rdev->config.cik.max_backends_per_se; + if (num_pipe_configs > 8) + num_pipe_configs = 8; + if (num_pipe_configs == 8) + fb_format |= CIK_GRPH_PIPE_CONFIG(CIK_ADDR_SURF_P8_32x32_16x16); + else if (num_pipe_configs == 4) { + if (num_rb == 4) + fb_format |= CIK_GRPH_PIPE_CONFIG(CIK_ADDR_SURF_P4_16x16); + else if (num_rb < 4) + fb_format |= CIK_GRPH_PIPE_CONFIG(CIK_ADDR_SURF_P4_8x16); + } else if (num_pipe_configs == 2) + fb_format |= CIK_GRPH_PIPE_CONFIG(CIK_ADDR_SURF_P2); + } else if ((rdev->family == CHIP_TAHITI) || + (rdev->family == CHIP_PITCAIRN)) fb_format |= SI_GRPH_PIPE_CONFIG(SI_ADDR_SURF_P8_32x32_8x16); else if (rdev->family == CHIP_VERDE) fb_format |= SI_GRPH_PIPE_CONFIG(SI_ADDR_SURF_P4_8x16); @@ -1227,8 +1244,12 @@ static int dce4_crtc_do_set_base(struct drm_crtc *crtc, WREG32(EVERGREEN_GRPH_PITCH + radeon_crtc->crtc_offset, fb_pitch_pixels); WREG32(EVERGREEN_GRPH_ENABLE + radeon_crtc->crtc_offset, 1); - WREG32(EVERGREEN_DESKTOP_HEIGHT + radeon_crtc->crtc_offset, - target_fb->height); + if (rdev->family >= CHIP_BONAIRE) + WREG32(CIK_LB_DESKTOP_HEIGHT + radeon_crtc->crtc_offset, + target_fb->height); + else + WREG32(EVERGREEN_DESKTOP_HEIGHT + radeon_crtc->crtc_offset, + target_fb->height); x &= ~3; y &= ~1; WREG32(EVERGREEN_VIEWPORT_START + radeon_crtc->crtc_offset, @@ -1600,6 +1621,12 @@ static int radeon_get_shared_nondp_ppll(struct drm_crtc *crtc) * * Asic specific PLL information * + * DCE 8.x + * KB/KV + * - PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP) + * CI + * - PPLL0, PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP) and DAC + * * DCE 6.1 * - PPLL2 is only available to UNIPHYA (both DP and non-DP) * - PPLL0, PPLL1 are available for UNIPHYB/C/D/E/F (both DP and non-DP) @@ -1626,7 +1653,47 @@ static int radeon_atom_pick_pll(struct drm_crtc *crtc) u32 pll_in_use; int pll; - if (ASIC_IS_DCE61(rdev)) { + if (ASIC_IS_DCE8(rdev)) { + if (ENCODER_MODE_IS_DP(atombios_get_encoder_mode(radeon_crtc->encoder))) { + if (rdev->clock.dp_extclk) + /* skip PPLL programming if using ext clock */ + return ATOM_PPLL_INVALID; + else { + /* use the same PPLL for all DP monitors */ + pll = radeon_get_shared_dp_ppll(crtc); + if (pll != ATOM_PPLL_INVALID) + return pll; + } + } else { + /* use the same PPLL for all monitors with the same clock */ + pll = radeon_get_shared_nondp_ppll(crtc); + if (pll != ATOM_PPLL_INVALID) + return pll; + } + /* otherwise, pick one of the plls */ + if ((rdev->family == CHIP_KAVERI) || + (rdev->family == CHIP_KABINI)) { + /* KB/KV has PPLL1 and PPLL2 */ + pll_in_use = radeon_get_pll_use_mask(crtc); + if (!(pll_in_use & (1 << ATOM_PPLL2))) + return ATOM_PPLL2; + if (!(pll_in_use & (1 << ATOM_PPLL1))) + return ATOM_PPLL1; + DRM_ERROR("unable to allocate a PPLL\n"); + return ATOM_PPLL_INVALID; + } else { + /* CI has PPLL0, PPLL1, and PPLL2 */ + pll_in_use = radeon_get_pll_use_mask(crtc); + if (!(pll_in_use & (1 << ATOM_PPLL2))) + return ATOM_PPLL2; + if (!(pll_in_use & (1 << ATOM_PPLL1))) + return ATOM_PPLL1; + if (!(pll_in_use & (1 << ATOM_PPLL0))) + return ATOM_PPLL0; + DRM_ERROR("unable to allocate a PPLL\n"); + return ATOM_PPLL_INVALID; + } + } else if (ASIC_IS_DCE61(rdev)) { struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv; @@ -1774,6 +1841,9 @@ int atombios_crtc_mode_set(struct drm_crtc *crtc, atombios_crtc_set_base(crtc, x, y, old_fb); atombios_overscan_setup(crtc, mode, adjusted_mode); atombios_scaler_setup(crtc); + /* update the hw version fpr dpm */ + radeon_crtc->hw_mode = *adjusted_mode; + return 0; } @@ -1864,7 +1934,7 @@ static void atombios_crtc_disable(struct drm_crtc *crtc) break; case ATOM_PPLL0: /* disable the ppll */ - if (ASIC_IS_DCE61(rdev)) + if ((rdev->family == CHIP_ARUBA) || (rdev->family == CHIP_BONAIRE)) atombios_crtc_program_pll(crtc, radeon_crtc->crtc_id, radeon_crtc->pll_id, 0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss); break; diff --git a/sys/dev/drm/radeon/atombios_dp.c b/sys/dev/drm/radeon/atombios_dp.c index be8c84aeff..51a7428367 100644 --- a/sys/dev/drm/radeon/atombios_dp.c +++ b/sys/dev/drm/radeon/atombios_dp.c @@ -47,6 +47,41 @@ static char *pre_emph_names[] = { }; /***** radeon AUX functions *****/ + +/* Atom needs data in little endian format + * so swap as appropriate when copying data to + * or from atom. Note that atom operates on + * dw units. + */ +static void radeon_copy_swap(u8 *dst, u8 *src, u8 num_bytes, bool to_le) +{ +#ifdef __BIG_ENDIAN + u8 src_tmp[20], dst_tmp[20]; /* used for byteswapping */ + u32 *dst32, *src32; + int i; + + memcpy(src_tmp, src, num_bytes); + src32 = (u32 *)src_tmp; + dst32 = (u32 *)dst_tmp; + if (to_le) { + for (i = 0; i < ((num_bytes + 3) / 4); i++) + dst32[i] = cpu_to_le32(src32[i]); + memcpy(dst, dst_tmp, num_bytes); + } else { + u8 dws = num_bytes & ~3; + for (i = 0; i < ((num_bytes + 3) / 4); i++) + dst32[i] = le32_to_cpu(src32[i]); + memcpy(dst, dst_tmp, dws); + if (num_bytes % 4) { + for (i = 0; i < (num_bytes % 4); i++) + dst[dws+i] = dst_tmp[dws+i]; + } + } +#else + memcpy(dst, src, num_bytes); +#endif +} + union aux_channel_transaction { PROCESS_AUX_CHANNEL_TRANSACTION_PS_ALLOCATION v1; PROCESS_AUX_CHANNEL_TRANSACTION_PARAMETERS_V2 v2; @@ -68,10 +103,10 @@ static int radeon_process_aux_ch(struct radeon_i2c_chan *chan, base = (unsigned char *)(rdev->mode_info.atom_context->scratch + 1); - memcpy(base, send, send_bytes); + radeon_copy_swap(base, send, send_bytes, true); - args.v1.lpAuxRequest = 0 + 4; - args.v1.lpDataOut = 16 + 4; + args.v1.lpAuxRequest = cpu_to_le16((u16)(0 + 4)); + args.v1.lpDataOut = cpu_to_le16((u16)(16 + 4)); args.v1.ucDataOutLen = 0; args.v1.ucChannelID = chan->rec.i2c_id; args.v1.ucDelay = delay / 10; @@ -105,7 +140,7 @@ static int radeon_process_aux_ch(struct radeon_i2c_chan *chan, recv_bytes = recv_size; if (recv && recv_size) - memcpy(recv, base + 16, recv_bytes); + radeon_copy_swap(recv, base + 16, recv_bytes, false); return recv_bytes; } diff --git a/sys/dev/drm/radeon/atombios_encoders.c b/sys/dev/drm/radeon/atombios_encoders.c index 134ec68a57..91df95f157 100644 --- a/sys/dev/drm/radeon/atombios_encoders.c +++ b/sys/dev/drm/radeon/atombios_encoders.c @@ -187,6 +187,13 @@ void radeon_atom_backlight_init(struct radeon_encoder *radeon_encoder, u8 backlight_level; char bl_name[16]; + /* Mac laptops with multiple GPUs use the gmux driver for backlight + * so don't register a backlight device + */ + if ((rdev->pdev->subsystem_vendor == PCI_VENDOR_ID_APPLE) && + (rdev->ddev->pci_device == 0x6741)) + return; + if (!radeon_encoder->enc_priv) return; @@ -295,6 +302,7 @@ static inline bool radeon_encoder_is_digital(struct drm_encoder *encoder) case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: return true; default: return false; @@ -478,11 +486,11 @@ static u8 radeon_atom_get_bpc(struct drm_encoder *encoder) } } - union dvo_encoder_control { ENABLE_EXTERNAL_TMDS_ENCODER_PS_ALLOCATION ext_tmds; DVO_ENCODER_CONTROL_PS_ALLOCATION dvo; DVO_ENCODER_CONTROL_PS_ALLOCATION_V3 dvo_v3; + DVO_ENCODER_CONTROL_PS_ALLOCATION_V1_4 dvo_v4; }; void @@ -532,6 +540,13 @@ atombios_dvo_setup(struct drm_encoder *encoder, int action) args.dvo_v3.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10); args.dvo_v3.ucDVOConfig = 0; /* XXX */ break; + case 4: + /* DCE8 */ + args.dvo_v4.ucAction = action; + args.dvo_v4.usPixelClock = cpu_to_le16(radeon_encoder->pixel_clock / 10); + args.dvo_v4.ucDVOConfig = 0; /* XXX */ + args.dvo_v4.ucBitPerColor = radeon_atom_get_bpc(encoder); + break; default: DRM_ERROR("Unknown table version %d, %d\n", frev, crev); break; @@ -914,10 +929,14 @@ atombios_dig_encoder_setup(struct drm_encoder *encoder, int action, int panel_mo args.v4.ucLaneNum = 4; if (ENCODER_MODE_IS_DP(args.v4.ucEncoderMode)) { - if (dp_clock == 270000) - args.v1.ucConfig |= ATOM_ENCODER_CONFIG_V4_DPLINKRATE_2_70GHZ; - else if (dp_clock == 540000) + if (dp_clock == 540000) args.v1.ucConfig |= ATOM_ENCODER_CONFIG_V4_DPLINKRATE_5_40GHZ; + else if (dp_clock == 324000) + args.v1.ucConfig |= ATOM_ENCODER_CONFIG_V4_DPLINKRATE_3_24GHZ; + else if (dp_clock == 270000) + args.v1.ucConfig |= ATOM_ENCODER_CONFIG_V4_DPLINKRATE_2_70GHZ; + else + args.v1.ucConfig |= ATOM_ENCODER_CONFIG_V4_DPLINKRATE_1_62GHZ; } args.v4.acConfig.ucDigSel = dig->dig_encoder; args.v4.ucBitPerColor = radeon_atom_get_bpc(encoder); @@ -1011,6 +1030,7 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action, uint8_t case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: index = GetIndexIntoMasterTable(COMMAND, UNIPHYTransmitterControl); break; case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: @@ -1270,6 +1290,9 @@ atombios_dig_transmitter_setup(struct drm_encoder *encoder, int action, uint8_t else args.v5.ucPhyId = ATOM_PHY_ID_UNIPHYE; break; + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: + args.v5.ucPhyId = ATOM_PHY_ID_UNIPHYG; + break; } if (is_dp) args.v5.ucLaneNum = dp_lane_count; @@ -1734,6 +1757,7 @@ radeon_atom_encoder_dpms(struct drm_encoder *encoder, int mode) case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: radeon_atom_encoder_dpms_dig(encoder, mode); break; @@ -1871,6 +1895,7 @@ atombios_set_encoder_crtc_source(struct drm_encoder *encoder) case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: dig = radeon_encoder->enc_priv; switch (dig->dig_encoder) { @@ -1892,6 +1917,9 @@ atombios_set_encoder_crtc_source(struct drm_encoder *encoder) case 5: args.v2.ucEncoderID = ASIC_INT_DIG6_ENCODER_ID; break; + case 6: + args.v2.ucEncoderID = ASIC_INT_DIG7_ENCODER_ID; + break; } break; case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1: @@ -1954,7 +1982,13 @@ atombios_apply_encoder_quirks(struct drm_encoder *encoder, /* set scaler clears this on some chips */ if (ASIC_IS_AVIVO(rdev) && (!(radeon_encoder->active_device & (ATOM_DEVICE_TV_SUPPORT)))) { - if (ASIC_IS_DCE4(rdev)) { + if (ASIC_IS_DCE8(rdev)) { + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + WREG32(CIK_LB_DATA_FORMAT + radeon_crtc->crtc_offset, + CIK_INTERLEAVE_EN); + else + WREG32(CIK_LB_DATA_FORMAT + radeon_crtc->crtc_offset, 0); + } else if (ASIC_IS_DCE4(rdev)) { if (mode->flags & DRM_MODE_FLAG_INTERLACE) WREG32(EVERGREEN_DATA_FORMAT + radeon_crtc->crtc_offset, EVERGREEN_INTERLEAVE_EN); @@ -2001,6 +2035,9 @@ static int radeon_atom_pick_dig_encoder(struct drm_encoder *encoder) else return 4; break; + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: + return 6; + break; } } else if (ASIC_IS_DCE4(rdev)) { /* DCE4/5 */ @@ -2085,6 +2122,7 @@ radeon_atom_encoder_init(struct radeon_device *rdev) case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: atombios_dig_transmitter_setup(encoder, ATOM_TRANSMITTER_ACTION_INIT, 0, 0); break; @@ -2129,6 +2167,7 @@ radeon_atom_encoder_mode_set(struct drm_encoder *encoder, case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: /* handled in dpms */ break; @@ -2394,6 +2433,7 @@ static void radeon_atom_encoder_disable(struct drm_encoder *encoder) case ENCODER_OBJECT_ID_INTERNAL_UNIPHY: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: /* handled in dpms */ break; @@ -2630,6 +2670,7 @@ radeon_add_atom_encoder(struct drm_device *dev, case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_LVTMA: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1: case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2: + case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3: if (radeon_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) { radeon_encoder->rmx_type = RMX_FULL; drm_encoder_init(dev, encoder, &radeon_atom_enc_funcs, DRM_MODE_ENCODER_LVDS); diff --git a/sys/dev/drm/radeon/avivod.h b/sys/dev/drm/radeon/avivod.h index c46db9a3b9..3c391e7e9f 100644 --- a/sys/dev/drm/radeon/avivod.h +++ b/sys/dev/drm/radeon/avivod.h @@ -23,8 +23,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/avivod.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef AVIVOD_H #define AVIVOD_H diff --git a/sys/dev/drm/radeon/btc_dpm.c b/sys/dev/drm/radeon/btc_dpm.c new file mode 100644 index 0000000000..ee58a8b515 --- /dev/null +++ b/sys/dev/drm/radeon/btc_dpm.c @@ -0,0 +1,2740 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "btcd.h" +#include "r600_dpm.h" +#include "cypress_dpm.h" +#include "btc_dpm.h" +#include "atom.h" +#include "radeon_asic.h" + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define MC_CG_SEQ_DRAMCONF_S0 0x05 +#define MC_CG_SEQ_DRAMCONF_S1 0x06 +#define MC_CG_SEQ_YCLK_SUSPEND 0x04 +#define MC_CG_SEQ_YCLK_RESUME 0x0a + +#define SMC_RAM_END 0x8000 + +#ifndef BTC_MGCG_SEQUENCE +#define BTC_MGCG_SEQUENCE 300 + +struct rv7xx_ps *rv770_get_ps(struct radeon_ps *rps); +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev); +void btc_dpm_reset_asic(struct radeon_device *rdev); + + +//********* BARTS **************// +static const u32 barts_cgcg_cgls_default[] = +{ + /* Register, Value, Mask bits */ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define BARTS_CGCG_CGLS_DEFAULT_LENGTH sizeof(barts_cgcg_cgls_default) / (3 * sizeof(u32)) + +static const u32 barts_cgcg_cgls_disable[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00000644, 0x000f7912, 0x001f4180, + 0x00000644, 0x000f3812, 0x001f4180 +}; +#define BARTS_CGCG_CGLS_DISABLE_LENGTH sizeof(barts_cgcg_cgls_disable) / (3 * sizeof(u32)) + +static const u32 barts_cgcg_cgls_enable[] = +{ + /* 0x0000c124, 0x84180000, 0x00180000, */ + 0x00000644, 0x000f7892, 0x001f4080, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff +}; +#define BARTS_CGCG_CGLS_ENABLE_LENGTH sizeof(barts_cgcg_cgls_enable) / (3 * sizeof(u32)) + +static const u32 barts_mgcg_default[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00005448, 0x00000100, 0xffffffff, + 0x000055e4, 0x00600100, 0xffffffff, + 0x0000160c, 0x00000100, 0xffffffff, + 0x0000c164, 0x00000100, 0xffffffff, + 0x00008a18, 0x00000100, 0xffffffff, + 0x0000897c, 0x06000100, 0xffffffff, + 0x00008b28, 0x00000100, 0xffffffff, + 0x00009144, 0x00000100, 0xffffffff, + 0x00009a60, 0x00000100, 0xffffffff, + 0x00009868, 0x00000100, 0xffffffff, + 0x00008d58, 0x00000100, 0xffffffff, + 0x00009510, 0x00000100, 0xffffffff, + 0x0000949c, 0x00000100, 0xffffffff, + 0x00009654, 0x00000100, 0xffffffff, + 0x00009030, 0x00000100, 0xffffffff, + 0x00009034, 0x00000100, 0xffffffff, + 0x00009038, 0x00000100, 0xffffffff, + 0x0000903c, 0x00000100, 0xffffffff, + 0x00009040, 0x00000100, 0xffffffff, + 0x0000a200, 0x00000100, 0xffffffff, + 0x0000a204, 0x00000100, 0xffffffff, + 0x0000a208, 0x00000100, 0xffffffff, + 0x0000a20c, 0x00000100, 0xffffffff, + 0x0000977c, 0x00000100, 0xffffffff, + 0x00003f80, 0x00000100, 0xffffffff, + 0x0000a210, 0x00000100, 0xffffffff, + 0x0000a214, 0x00000100, 0xffffffff, + 0x000004d8, 0x00000100, 0xffffffff, + 0x00009784, 0x00000100, 0xffffffff, + 0x00009698, 0x00000100, 0xffffffff, + 0x000004d4, 0x00000200, 0xffffffff, + 0x000004d0, 0x00000000, 0xffffffff, + 0x000030cc, 0x00000100, 0xffffffff, + 0x0000d0c0, 0xff000100, 0xffffffff, + 0x0000802c, 0x40000000, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009224, 0x00070000, 0xffffffff, + 0x00009228, 0x00030002, 0xffffffff, + 0x0000922c, 0x00050004, 0xffffffff, + 0x00009238, 0x00010006, 0xffffffff, + 0x0000923c, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x0000802c, 0x40010000, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009224, 0x00070000, 0xffffffff, + 0x00009228, 0x00030002, 0xffffffff, + 0x0000922c, 0x00050004, 0xffffffff, + 0x00009238, 0x00010006, 0xffffffff, + 0x0000923c, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define BARTS_MGCG_DEFAULT_LENGTH sizeof(barts_mgcg_default) / (3 * sizeof(u32)) + +static const u32 barts_mgcg_disable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x00009150, 0x00600000, 0xffffffff +}; +#define BARTS_MGCG_DISABLE_LENGTH sizeof(barts_mgcg_disable) / (3 * sizeof(u32)) + +static const u32 barts_mgcg_enable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00009150, 0x81944000, 0xffffffff +}; +#define BARTS_MGCG_ENABLE_LENGTH sizeof(barts_mgcg_enable) / (3 * sizeof(u32)) + +//********* CAICOS **************// +static const u32 caicos_cgcg_cgls_default[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define CAICOS_CGCG_CGLS_DEFAULT_LENGTH sizeof(caicos_cgcg_cgls_default) / (3 * sizeof(u32)) + +static const u32 caicos_cgcg_cgls_disable[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00000644, 0x000f7912, 0x001f4180, + 0x00000644, 0x000f3812, 0x001f4180 +}; +#define CAICOS_CGCG_CGLS_DISABLE_LENGTH sizeof(caicos_cgcg_cgls_disable) / (3 * sizeof(u32)) + +static const u32 caicos_cgcg_cgls_enable[] = +{ + /* 0x0000c124, 0x84180000, 0x00180000, */ + 0x00000644, 0x000f7892, 0x001f4080, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff +}; +#define CAICOS_CGCG_CGLS_ENABLE_LENGTH sizeof(caicos_cgcg_cgls_enable) / (3 * sizeof(u32)) + +static const u32 caicos_mgcg_default[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00005448, 0x00000100, 0xffffffff, + 0x000055e4, 0x00600100, 0xffffffff, + 0x0000160c, 0x00000100, 0xffffffff, + 0x0000c164, 0x00000100, 0xffffffff, + 0x00008a18, 0x00000100, 0xffffffff, + 0x0000897c, 0x06000100, 0xffffffff, + 0x00008b28, 0x00000100, 0xffffffff, + 0x00009144, 0x00000100, 0xffffffff, + 0x00009a60, 0x00000100, 0xffffffff, + 0x00009868, 0x00000100, 0xffffffff, + 0x00008d58, 0x00000100, 0xffffffff, + 0x00009510, 0x00000100, 0xffffffff, + 0x0000949c, 0x00000100, 0xffffffff, + 0x00009654, 0x00000100, 0xffffffff, + 0x00009030, 0x00000100, 0xffffffff, + 0x00009034, 0x00000100, 0xffffffff, + 0x00009038, 0x00000100, 0xffffffff, + 0x0000903c, 0x00000100, 0xffffffff, + 0x00009040, 0x00000100, 0xffffffff, + 0x0000a200, 0x00000100, 0xffffffff, + 0x0000a204, 0x00000100, 0xffffffff, + 0x0000a208, 0x00000100, 0xffffffff, + 0x0000a20c, 0x00000100, 0xffffffff, + 0x0000977c, 0x00000100, 0xffffffff, + 0x00003f80, 0x00000100, 0xffffffff, + 0x0000a210, 0x00000100, 0xffffffff, + 0x0000a214, 0x00000100, 0xffffffff, + 0x000004d8, 0x00000100, 0xffffffff, + 0x00009784, 0x00000100, 0xffffffff, + 0x00009698, 0x00000100, 0xffffffff, + 0x000004d4, 0x00000200, 0xffffffff, + 0x000004d0, 0x00000000, 0xffffffff, + 0x000030cc, 0x00000100, 0xffffffff, + 0x0000d0c0, 0xff000100, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define CAICOS_MGCG_DEFAULT_LENGTH sizeof(caicos_mgcg_default) / (3 * sizeof(u32)) + +static const u32 caicos_mgcg_disable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x00009150, 0x00600000, 0xffffffff +}; +#define CAICOS_MGCG_DISABLE_LENGTH sizeof(caicos_mgcg_disable) / (3 * sizeof(u32)) + +static const u32 caicos_mgcg_enable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00009150, 0x46944040, 0xffffffff +}; +#define CAICOS_MGCG_ENABLE_LENGTH sizeof(caicos_mgcg_enable) / (3 * sizeof(u32)) + +//********* TURKS **************// +static const u32 turks_cgcg_cgls_default[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define TURKS_CGCG_CGLS_DEFAULT_LENGTH sizeof(turks_cgcg_cgls_default) / (3 * sizeof(u32)) + +static const u32 turks_cgcg_cgls_disable[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00000644, 0x000f7912, 0x001f4180, + 0x00000644, 0x000f3812, 0x001f4180 +}; +#define TURKS_CGCG_CGLS_DISABLE_LENGTH sizeof(turks_cgcg_cgls_disable) / (3 * sizeof(u32)) + +static const u32 turks_cgcg_cgls_enable[] = +{ + /* 0x0000c124, 0x84180000, 0x00180000, */ + 0x00000644, 0x000f7892, 0x001f4080, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff +}; +#define TURKS_CGCG_CGLS_ENABLE_LENGTH sizeof(turks_cgcg_cgls_enable) / (3 * sizeof(u32)) + +// These are the sequences for turks_mgcg_shls +static const u32 turks_mgcg_default[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00005448, 0x00000100, 0xffffffff, + 0x000055e4, 0x00600100, 0xffffffff, + 0x0000160c, 0x00000100, 0xffffffff, + 0x0000c164, 0x00000100, 0xffffffff, + 0x00008a18, 0x00000100, 0xffffffff, + 0x0000897c, 0x06000100, 0xffffffff, + 0x00008b28, 0x00000100, 0xffffffff, + 0x00009144, 0x00000100, 0xffffffff, + 0x00009a60, 0x00000100, 0xffffffff, + 0x00009868, 0x00000100, 0xffffffff, + 0x00008d58, 0x00000100, 0xffffffff, + 0x00009510, 0x00000100, 0xffffffff, + 0x0000949c, 0x00000100, 0xffffffff, + 0x00009654, 0x00000100, 0xffffffff, + 0x00009030, 0x00000100, 0xffffffff, + 0x00009034, 0x00000100, 0xffffffff, + 0x00009038, 0x00000100, 0xffffffff, + 0x0000903c, 0x00000100, 0xffffffff, + 0x00009040, 0x00000100, 0xffffffff, + 0x0000a200, 0x00000100, 0xffffffff, + 0x0000a204, 0x00000100, 0xffffffff, + 0x0000a208, 0x00000100, 0xffffffff, + 0x0000a20c, 0x00000100, 0xffffffff, + 0x0000977c, 0x00000100, 0xffffffff, + 0x00003f80, 0x00000100, 0xffffffff, + 0x0000a210, 0x00000100, 0xffffffff, + 0x0000a214, 0x00000100, 0xffffffff, + 0x000004d8, 0x00000100, 0xffffffff, + 0x00009784, 0x00000100, 0xffffffff, + 0x00009698, 0x00000100, 0xffffffff, + 0x000004d4, 0x00000200, 0xffffffff, + 0x000004d0, 0x00000000, 0xffffffff, + 0x000030cc, 0x00000100, 0xffffffff, + 0x0000d0c0, 0x00000100, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define TURKS_MGCG_DEFAULT_LENGTH sizeof(turks_mgcg_default) / (3 * sizeof(u32)) + +static const u32 turks_mgcg_disable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x00009150, 0x00600000, 0xffffffff +}; +#define TURKS_MGCG_DISABLE_LENGTH sizeof(turks_mgcg_disable) / (3 * sizeof(u32)) + +static const u32 turks_mgcg_enable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00009150, 0x6e944000, 0xffffffff +}; +#define TURKS_MGCG_ENABLE_LENGTH sizeof(turks_mgcg_enable) / (3 * sizeof(u32)) + +#endif + +#ifndef BTC_SYSLS_SEQUENCE +#define BTC_SYSLS_SEQUENCE 100 + + +//********* BARTS **************// +static const u32 barts_sysls_default[] = +{ + /* Register, Value, Mask bits */ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#define BARTS_SYSLS_DEFAULT_LENGTH sizeof(barts_sysls_default) / (3 * sizeof(u32)) + +static const u32 barts_sysls_disable[] = +{ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x00041401, 0xffffffff, + 0x0000264c, 0x00040400, 0xffffffff, + 0x00002648, 0x00040400, 0xffffffff, + 0x00002650, 0x00040400, 0xffffffff, + 0x000020b8, 0x00040400, 0xffffffff, + 0x000020bc, 0x00040400, 0xffffffff, + 0x000020c0, 0x00040c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680000, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00007ffd, 0xffffffff, + 0x00000c7c, 0x0000ff00, 0xffffffff, + 0x00006dfc, 0x0000007f, 0xffffffff +}; +#define BARTS_SYSLS_DISABLE_LENGTH sizeof(barts_sysls_disable) / (3 * sizeof(u32)) + +static const u32 barts_sysls_enable[] = +{ + 0x000055e8, 0x00000001, 0xffffffff, + 0x0000d0bc, 0x00000100, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000004c8, 0x00000000, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#define BARTS_SYSLS_ENABLE_LENGTH sizeof(barts_sysls_enable) / (3 * sizeof(u32)) + +//********* CAICOS **************// +static const u32 caicos_sysls_default[] = +{ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#define CAICOS_SYSLS_DEFAULT_LENGTH sizeof(caicos_sysls_default) / (3 * sizeof(u32)) + +static const u32 caicos_sysls_disable[] = +{ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x00041401, 0xffffffff, + 0x0000264c, 0x00040400, 0xffffffff, + 0x00002648, 0x00040400, 0xffffffff, + 0x00002650, 0x00040400, 0xffffffff, + 0x000020b8, 0x00040400, 0xffffffff, + 0x000020bc, 0x00040400, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680000, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00007ffd, 0xffffffff, + 0x00000c7c, 0x0000ff00, 0xffffffff, + 0x00006dfc, 0x0000007f, 0xffffffff +}; +#define CAICOS_SYSLS_DISABLE_LENGTH sizeof(caicos_sysls_disable) / (3 * sizeof(u32)) + +static const u32 caicos_sysls_enable[] = +{ + 0x000055e8, 0x00000001, 0xffffffff, + 0x0000d0bc, 0x00000100, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff, + 0x000004c8, 0x00000000, 0xffffffff +}; +#define CAICOS_SYSLS_ENABLE_LENGTH sizeof(caicos_sysls_enable) / (3 * sizeof(u32)) + +//********* TURKS **************// +static const u32 turks_sysls_default[] = +{ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#define TURKS_SYSLS_DEFAULT_LENGTH sizeof(turks_sysls_default) / (3 * sizeof(u32)) + +static const u32 turks_sysls_disable[] = +{ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x00041401, 0xffffffff, + 0x0000264c, 0x00040400, 0xffffffff, + 0x00002648, 0x00040400, 0xffffffff, + 0x00002650, 0x00040400, 0xffffffff, + 0x000020b8, 0x00040400, 0xffffffff, + 0x000020bc, 0x00040400, 0xffffffff, + 0x000020c0, 0x00040c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680000, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00007ffd, 0xffffffff, + 0x00000c7c, 0x0000ff00, 0xffffffff, + 0x00006dfc, 0x0000007f, 0xffffffff +}; +#define TURKS_SYSLS_DISABLE_LENGTH sizeof(turks_sysls_disable) / (3 * sizeof(u32)) + +static const u32 turks_sysls_enable[] = +{ + 0x000055e8, 0x00000001, 0xffffffff, + 0x0000d0bc, 0x00000100, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x000004c8, 0x00000000, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#define TURKS_SYSLS_ENABLE_LENGTH sizeof(turks_sysls_enable) / (3 * sizeof(u32)) + +#endif + +u32 btc_valid_sclk[40] = +{ + 5000, 10000, 15000, 20000, 25000, 30000, 35000, 40000, 45000, 50000, + 55000, 60000, 65000, 70000, 75000, 80000, 85000, 90000, 95000, 100000, + 105000, 110000, 11500, 120000, 125000, 130000, 135000, 140000, 145000, 150000, + 155000, 160000, 165000, 170000, 175000, 180000, 185000, 190000, 195000, 200000 +}; + +static const struct radeon_blacklist_clocks btc_blacklist_clocks[] = +{ + { 10000, 30000, RADEON_SCLK_UP }, + { 15000, 30000, RADEON_SCLK_UP }, + { 20000, 30000, RADEON_SCLK_UP }, + { 25000, 30000, RADEON_SCLK_UP } +}; + +void btc_apply_voltage_dependency_rules(struct radeon_clock_voltage_dependency_table *table, + u32 clock, u16 max_voltage, u16 *voltage) +{ + u32 i; + + if ((table == NULL) || (table->count == 0)) + return; + + for (i= 0; i < table->count; i++) { + if (clock <= table->entries[i].clk) { + if (*voltage < table->entries[i].v) + *voltage = (u16)((table->entries[i].v < max_voltage) ? + table->entries[i].v : max_voltage); + return; + } + } + + *voltage = (*voltage > max_voltage) ? *voltage : max_voltage; +} + +static u32 btc_find_valid_clock(struct radeon_clock_array *clocks, + u32 max_clock, u32 requested_clock) +{ + unsigned int i; + + if ((clocks == NULL) || (clocks->count == 0)) + return (requested_clock < max_clock) ? requested_clock : max_clock; + + for (i = 0; i < clocks->count; i++) { + if (clocks->values[i] >= requested_clock) + return (clocks->values[i] < max_clock) ? clocks->values[i] : max_clock; + } + + return (clocks->values[clocks->count - 1] < max_clock) ? + clocks->values[clocks->count - 1] : max_clock; +} + +static u32 btc_get_valid_mclk(struct radeon_device *rdev, + u32 max_mclk, u32 requested_mclk) +{ + return btc_find_valid_clock(&rdev->pm.dpm.dyn_state.valid_mclk_values, + max_mclk, requested_mclk); +} + +static u32 btc_get_valid_sclk(struct radeon_device *rdev, + u32 max_sclk, u32 requested_sclk) +{ + return btc_find_valid_clock(&rdev->pm.dpm.dyn_state.valid_sclk_values, + max_sclk, requested_sclk); +} + +void btc_skip_blacklist_clocks(struct radeon_device *rdev, + const u32 max_sclk, const u32 max_mclk, + u32 *sclk, u32 *mclk) +{ + int i, num_blacklist_clocks; + + if ((sclk == NULL) || (mclk == NULL)) + return; + + num_blacklist_clocks = ARRAY_SIZE(btc_blacklist_clocks); + + for (i = 0; i < num_blacklist_clocks; i++) { + if ((btc_blacklist_clocks[i].sclk == *sclk) && + (btc_blacklist_clocks[i].mclk == *mclk)) + break; + } + + if (i < num_blacklist_clocks) { + if (btc_blacklist_clocks[i].action == RADEON_SCLK_UP) { + *sclk = btc_get_valid_sclk(rdev, max_sclk, *sclk + 1); + + if (*sclk < max_sclk) + btc_skip_blacklist_clocks(rdev, max_sclk, max_mclk, sclk, mclk); + } + } +} + +void btc_adjust_clock_combinations(struct radeon_device *rdev, + const struct radeon_clock_and_voltage_limits *max_limits, + struct rv7xx_pl *pl) +{ + + if ((pl->mclk == 0) || (pl->sclk == 0)) + return; + + if (pl->mclk == pl->sclk) + return; + + if (pl->mclk > pl->sclk) { + if (((pl->mclk + (pl->sclk - 1)) / pl->sclk) > rdev->pm.dpm.dyn_state.mclk_sclk_ratio) + pl->sclk = btc_get_valid_sclk(rdev, + max_limits->sclk, + (pl->mclk + + (rdev->pm.dpm.dyn_state.mclk_sclk_ratio - 1)) / + rdev->pm.dpm.dyn_state.mclk_sclk_ratio); + } else { + if ((pl->sclk - pl->mclk) > rdev->pm.dpm.dyn_state.sclk_mclk_delta) + pl->mclk = btc_get_valid_mclk(rdev, + max_limits->mclk, + pl->sclk - + rdev->pm.dpm.dyn_state.sclk_mclk_delta); + } +} + +static u16 btc_find_voltage(struct atom_voltage_table *table, u16 voltage) +{ + unsigned int i; + + for (i = 0; i < table->count; i++) { + if (voltage <= table->entries[i].value) + return table->entries[i].value; + } + + return table->entries[table->count - 1].value; +} + +void btc_apply_voltage_delta_rules(struct radeon_device *rdev, + u16 max_vddc, u16 max_vddci, + u16 *vddc, u16 *vddci) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u16 new_voltage; + + if ((0 == *vddc) || (0 == *vddci)) + return; + + if (*vddc > *vddci) { + if ((*vddc - *vddci) > rdev->pm.dpm.dyn_state.vddc_vddci_delta) { + new_voltage = btc_find_voltage(&eg_pi->vddci_voltage_table, + (*vddc - rdev->pm.dpm.dyn_state.vddc_vddci_delta)); + *vddci = (new_voltage < max_vddci) ? new_voltage : max_vddci; + } + } else { + if ((*vddci - *vddc) > rdev->pm.dpm.dyn_state.vddc_vddci_delta) { + new_voltage = btc_find_voltage(&eg_pi->vddc_voltage_table, + (*vddci - rdev->pm.dpm.dyn_state.vddc_vddci_delta)); + *vddc = (new_voltage < max_vddc) ? new_voltage : max_vddc; + } + } +} + +static void btc_enable_bif_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp, bif; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + if (enable) { + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + if (!pi->boot_in_gen2) { + bif = RREG32(CG_BIF_REQ_AND_RSP) & ~CG_CLIENT_REQ_MASK; + bif |= CG_CLIENT_REQ(0xd); + WREG32(CG_BIF_REQ_AND_RSP, bif); + + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp |= LC_HW_VOLTAGE_IF_CONTROL(1); + tmp |= LC_GEN2_EN_STRAP; + + tmp |= LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + DRM_UDELAY(10); + tmp &= ~LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } + } + } else { + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) || + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + if (!pi->boot_in_gen2) { + bif = RREG32(CG_BIF_REQ_AND_RSP) & ~CG_CLIENT_REQ_MASK; + bif |= CG_CLIENT_REQ(0xd); + WREG32(CG_BIF_REQ_AND_RSP, bif); + + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp &= ~LC_GEN2_EN_STRAP; + } + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } + } +} + +static void btc_enable_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + btc_enable_bif_dynamic_pcie_gen2(rdev, enable); + + if (enable) + WREG32_P(GENERAL_PWRMGT, ENABLE_GEN2PCIE, ~ENABLE_GEN2PCIE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~ENABLE_GEN2PCIE); +} + +static int btc_disable_ulv(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (eg_pi->ulv.supported) { + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_DisableULV) != PPSMC_Result_OK) + return -EINVAL; + } + return 0; +} + +static int btc_populate_ulv_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + int ret = -EINVAL; + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_pl *ulv_pl = eg_pi->ulv.pl; + + if (ulv_pl->vddc) { + ret = cypress_convert_power_level_to_smc(rdev, + ulv_pl, + &table->ULVState.levels[0], + PPSMC_DISPLAY_WATERMARK_LOW); + if (ret == 0) { + table->ULVState.levels[0].arbValue = MC_CG_ARB_FREQ_F0; + table->ULVState.levels[0].ACIndex = 1; + + table->ULVState.levels[1] = table->ULVState.levels[0]; + table->ULVState.levels[2] = table->ULVState.levels[0]; + + table->ULVState.flags |= PPSMC_SWSTATE_FLAG_DC; + + WREG32(CG_ULV_CONTROL, BTC_CGULVCONTROL_DFLT); + WREG32(CG_ULV_PARAMETER, BTC_CGULVPARAMETER_DFLT); + } + } + + return ret; +} + +static int btc_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + int ret = cypress_populate_smc_acpi_state(rdev, table); + + if (ret == 0) { + table->ACPIState.levels[0].ACIndex = 0; + table->ACPIState.levels[1].ACIndex = 0; + table->ACPIState.levels[2].ACIndex = 0; + } + + return ret; +} + +void btc_program_mgcg_hw_sequence(struct radeon_device *rdev, + const u32 *sequence, u32 count) +{ + u32 i, length = count * 3; + u32 tmp; + + for (i = 0; i < length; i+=3) { + tmp = RREG32(sequence[i]); + tmp &= ~sequence[i+2]; + tmp |= sequence[i+1] & sequence[i+2]; + WREG32(sequence[i], tmp); + } +} + +static void btc_cg_clock_gating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *p = NULL; + + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_cgcg_cgls_default; + count = BARTS_CGCG_CGLS_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_cgcg_cgls_default; + count = TURKS_CGCG_CGLS_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_cgcg_cgls_default; + count = CAICOS_CGCG_CGLS_DEFAULT_LENGTH; + } else + return; + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +static void btc_cg_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *p = NULL; + + if (enable) { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_cgcg_cgls_enable; + count = BARTS_CGCG_CGLS_ENABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_cgcg_cgls_enable; + count = TURKS_CGCG_CGLS_ENABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_cgcg_cgls_enable; + count = CAICOS_CGCG_CGLS_ENABLE_LENGTH; + } else + return; + } else { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_cgcg_cgls_disable; + count = BARTS_CGCG_CGLS_DISABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_cgcg_cgls_disable; + count = TURKS_CGCG_CGLS_DISABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_cgcg_cgls_disable; + count = CAICOS_CGCG_CGLS_DISABLE_LENGTH; + } else + return; + } + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +static void btc_mg_clock_gating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *p = NULL; + + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_mgcg_default; + count = BARTS_MGCG_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_mgcg_default; + count = TURKS_MGCG_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_mgcg_default; + count = CAICOS_MGCG_DEFAULT_LENGTH; + } else + return; + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +static void btc_mg_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *p = NULL; + + if (enable) { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_mgcg_enable; + count = BARTS_MGCG_ENABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_mgcg_enable; + count = TURKS_MGCG_ENABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_mgcg_enable; + count = CAICOS_MGCG_ENABLE_LENGTH; + } else + return; + } else { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_mgcg_disable[0]; + count = BARTS_MGCG_DISABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_mgcg_disable[0]; + count = TURKS_MGCG_DISABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_mgcg_disable[0]; + count = CAICOS_MGCG_DISABLE_LENGTH; + } else + return; + } + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +static void btc_ls_clock_gating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *p = NULL; + + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_sysls_default; + count = BARTS_SYSLS_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_sysls_default; + count = TURKS_SYSLS_DEFAULT_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_sysls_default; + count = CAICOS_SYSLS_DEFAULT_LENGTH; + } else + return; + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +static void btc_ls_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *p = NULL; + + if (enable) { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_sysls_enable; + count = BARTS_SYSLS_ENABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_sysls_enable; + count = TURKS_SYSLS_ENABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_sysls_enable; + count = CAICOS_SYSLS_ENABLE_LENGTH; + } else + return; + } else { + if (rdev->family == CHIP_BARTS) { + p = (const u32 *)&barts_sysls_disable; + count = BARTS_SYSLS_DISABLE_LENGTH; + } else if (rdev->family == CHIP_TURKS) { + p = (const u32 *)&turks_sysls_disable; + count = TURKS_SYSLS_DISABLE_LENGTH; + } else if (rdev->family == CHIP_CAICOS) { + p = (const u32 *)&caicos_sysls_disable; + count = CAICOS_SYSLS_DISABLE_LENGTH; + } else + return; + } + + btc_program_mgcg_hw_sequence(rdev, p, count); +} + +bool btc_dpm_enabled(struct radeon_device *rdev) +{ + if (rv770_is_smc_running(rdev)) + return true; + else + return false; +} + +static int btc_init_smc_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + RV770_SMC_STATETABLE *table = &pi->smc_statetable; + int ret; + + memset(table, 0, sizeof(RV770_SMC_STATETABLE)); + + cypress_populate_smc_voltage_tables(rdev, table); + + switch (rdev->pm.int_thermal_type) { + case THERMAL_TYPE_EVERGREEN: + case THERMAL_TYPE_EMC2103_WITH_INTERNAL: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_INTERNAL; + break; + case THERMAL_TYPE_NONE: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_NONE; + break; + default: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL; + break; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REGULATOR_HOT) + table->systemFlags |= PPSMC_SYSTEMFLAG_REGULATOR_HOT; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (pi->mem_gddr5) + table->systemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + ret = cypress_populate_smc_initial_state(rdev, radeon_boot_state, table); + if (ret) + return ret; + + if (eg_pi->sclk_deep_sleep) + WREG32_P(SCLK_PSKIP_CNTL, PSKIP_ON_ALLOW_STOP_HI(32), + ~PSKIP_ON_ALLOW_STOP_HI_MASK); + + ret = btc_populate_smc_acpi_state(rdev, table); + if (ret) + return ret; + + if (eg_pi->ulv.supported) { + ret = btc_populate_ulv_state(rdev, table); + if (ret) + eg_pi->ulv.supported = false; + } + + table->driverState = table->initialState; + + return rv770_copy_bytes_to_smc(rdev, + pi->state_table_start, + (u8 *)table, + sizeof(RV770_SMC_STATETABLE), + pi->sram_end); +} + +static void btc_set_at_for_uvd(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + int idx = 0; + + if (r600_is_uvd_state(radeon_new_state->class, radeon_new_state->class2)) + idx = 1; + + if ((idx == 1) && !eg_pi->smu_uvd_hs) { + pi->rlp = 10; + pi->rmp = 100; + pi->lhp = 100; + pi->lmp = 10; + } else { + pi->rlp = eg_pi->ats[idx].rlp; + pi->rmp = eg_pi->ats[idx].rmp; + pi->lhp = eg_pi->ats[idx].lhp; + pi->lmp = eg_pi->ats[idx].lmp; + } + +} + +void btc_notify_uvd_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (r600_is_uvd_state(radeon_new_state->class, radeon_new_state->class2)) { + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_uvd_enabled, 1); + eg_pi->uvd_enabled = true; + } else { + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_uvd_enabled, 0); + eg_pi->uvd_enabled = false; + } +} + +int btc_reset_to_default(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_ResetToDefaults) != PPSMC_Result_OK) + return -EINVAL; + + return 0; +} + +static void btc_stop_smc(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (((RREG32(LB_SYNC_RESET_SEL) & LB_SYNC_RESET_SEL_MASK) >> LB_SYNC_RESET_SEL_SHIFT) != 1) + break; + DRM_UDELAY(1); + } + DRM_UDELAY(100); + + r7xx_stop_smc(rdev); +} + +void btc_read_arb_registers(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct evergreen_arb_registers *arb_registers = + &eg_pi->bootup_arb_registers; + + arb_registers->mc_arb_dram_timing = RREG32(MC_ARB_DRAM_TIMING); + arb_registers->mc_arb_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + arb_registers->mc_arb_rfsh_rate = RREG32(MC_ARB_RFSH_RATE); + arb_registers->mc_arb_burst_time = RREG32(MC_ARB_BURST_TIME); +} + + +static void btc_set_arb0_registers(struct radeon_device *rdev, + struct evergreen_arb_registers *arb_registers) +{ + u32 val; + + WREG32(MC_ARB_DRAM_TIMING, arb_registers->mc_arb_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2, arb_registers->mc_arb_dram_timing2); + + val = (arb_registers->mc_arb_rfsh_rate & POWERMODE0_MASK) >> + POWERMODE0_SHIFT; + WREG32_P(MC_ARB_RFSH_RATE, POWERMODE0(val), ~POWERMODE0_MASK); + + val = (arb_registers->mc_arb_burst_time & STATE0_MASK) >> + STATE0_SHIFT; + WREG32_P(MC_ARB_BURST_TIME, STATE0(val), ~STATE0_MASK); +} + +static void btc_set_boot_state_timing(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (eg_pi->ulv.supported) + btc_set_arb0_registers(rdev, &eg_pi->bootup_arb_registers); +} + +static bool btc_is_state_ulv_compatible(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_pl *ulv_pl = eg_pi->ulv.pl; + + if (state->low.mclk != ulv_pl->mclk) + return false; + + if (state->low.vddci != ulv_pl->vddci) + return false; + + /* XXX check minclocks, etc. */ + + return true; +} + + +static int btc_set_ulv_dram_timing(struct radeon_device *rdev) +{ + u32 val; + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_pl *ulv_pl = eg_pi->ulv.pl; + + radeon_atom_set_engine_dram_timings(rdev, + ulv_pl->sclk, + ulv_pl->mclk); + + val = rv770_calculate_memory_refresh_rate(rdev, ulv_pl->sclk); + WREG32_P(MC_ARB_RFSH_RATE, POWERMODE0(val), ~POWERMODE0_MASK); + + val = cypress_calculate_burst_time(rdev, ulv_pl->sclk, ulv_pl->mclk); + WREG32_P(MC_ARB_BURST_TIME, STATE0(val), ~STATE0_MASK); + + return 0; +} + +static int btc_enable_ulv(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableULV) != PPSMC_Result_OK) + return -EINVAL; + + return 0; +} + +static int btc_set_power_state_conditionally_enable_ulv(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + int ret = 0; + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (eg_pi->ulv.supported) { + if (btc_is_state_ulv_compatible(rdev, radeon_new_state)) { + // Set ARB[0] to reflect the DRAM timing needed for ULV. + ret = btc_set_ulv_dram_timing(rdev); + if (ret == 0) + ret = btc_enable_ulv(rdev); + } + } + + return ret; +} + +static bool btc_check_s0_mc_reg_index(u16 in_reg, u16 *out_reg) +{ + bool result = true; + + switch (in_reg) { + case MC_SEQ_RAS_TIMING >> 2: + *out_reg = MC_SEQ_RAS_TIMING_LP >> 2; + break; + case MC_SEQ_CAS_TIMING >> 2: + *out_reg = MC_SEQ_CAS_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING >> 2: + *out_reg = MC_SEQ_MISC_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING2 >> 2: + *out_reg = MC_SEQ_MISC_TIMING2_LP >> 2; + break; + case MC_SEQ_RD_CTL_D0 >> 2: + *out_reg = MC_SEQ_RD_CTL_D0_LP >> 2; + break; + case MC_SEQ_RD_CTL_D1 >> 2: + *out_reg = MC_SEQ_RD_CTL_D1_LP >> 2; + break; + case MC_SEQ_WR_CTL_D0 >> 2: + *out_reg = MC_SEQ_WR_CTL_D0_LP >> 2; + break; + case MC_SEQ_WR_CTL_D1 >> 2: + *out_reg = MC_SEQ_WR_CTL_D1_LP >> 2; + break; + case MC_PMG_CMD_EMRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + break; + case MC_PMG_CMD_MRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS_LP >> 2; + break; + case MC_PMG_CMD_MRS1 >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + break; + default: + result = false; + break; + } + + return result; +} + +static void btc_set_valid_flag(struct evergreen_mc_reg_table *table) +{ + u8 i, j; + + for (i = 0; i < table->last; i++) { + for (j = 1; j < table->num_entries; j++) { + if (table->mc_reg_table_entry[j-1].mc_data[i] != + table->mc_reg_table_entry[j].mc_data[i]) { + table->valid_flag |= (1 << i); + break; + } + } + } +} + +static int btc_set_mc_special_registers(struct radeon_device *rdev, + struct evergreen_mc_reg_table *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 i, j, k; + u32 tmp; + + for (i = 0, j = table->last; i < table->last; i++) { + switch (table->mc_reg_address[i].s1) { + case MC_SEQ_MISC1 >> 2: + tmp = RREG32(MC_PMG_CMD_EMRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_EMRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + ((tmp & 0xffff0000)) | + ((table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16); + } + j++; + + if (j > SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + + tmp = RREG32(MC_PMG_CMD_MRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS_LP >> 2; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (tmp & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + if (!pi->mem_gddr5) + table->mc_reg_table_entry[k].mc_data[j] |= 0x100; + } + j++; + + if (j > SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + break; + case MC_SEQ_RESERVE_M >> 2: + tmp = RREG32(MC_PMG_CMD_MRS1); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS1 >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (tmp & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + } + j++; + + if (j > SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + break; + default: + break; + } + } + + table->last = j; + + return 0; +} + +static void btc_set_s0_mc_reg_index(struct evergreen_mc_reg_table *table) +{ + u32 i; + u16 address; + + for (i = 0; i < table->last; i++) { + table->mc_reg_address[i].s0 = + btc_check_s0_mc_reg_index(table->mc_reg_address[i].s1, &address) ? + address : table->mc_reg_address[i].s1; + } +} + +static int btc_copy_vbios_mc_reg_table(struct atom_mc_reg_table *table, + struct evergreen_mc_reg_table *eg_table) +{ + u8 i, j; + + if (table->last > SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + + if (table->num_entries > MAX_AC_TIMING_ENTRIES) + return -EINVAL; + + for (i = 0; i < table->last; i++) + eg_table->mc_reg_address[i].s1 = table->mc_reg_address[i].s1; + eg_table->last = table->last; + + for (i = 0; i < table->num_entries; i++) { + eg_table->mc_reg_table_entry[i].mclk_max = + table->mc_reg_table_entry[i].mclk_max; + for(j = 0; j < table->last; j++) + eg_table->mc_reg_table_entry[i].mc_data[j] = + table->mc_reg_table_entry[i].mc_data[j]; + } + eg_table->num_entries = table->num_entries; + + return 0; +} + +static int btc_initialize_mc_reg_table(struct radeon_device *rdev) +{ + int ret; + struct atom_mc_reg_table *table; + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct evergreen_mc_reg_table *eg_table = &eg_pi->mc_reg_table; + u8 module_index = rv770_get_memory_module_index(rdev); + + table = kzalloc(sizeof(struct atom_mc_reg_table), GFP_KERNEL); + if (!table) + return -ENOMEM; + + /* Program additional LP registers that are no longer programmed by VBIOS */ + WREG32(MC_SEQ_RAS_TIMING_LP, RREG32(MC_SEQ_RAS_TIMING)); + WREG32(MC_SEQ_CAS_TIMING_LP, RREG32(MC_SEQ_CAS_TIMING)); + WREG32(MC_SEQ_MISC_TIMING_LP, RREG32(MC_SEQ_MISC_TIMING)); + WREG32(MC_SEQ_MISC_TIMING2_LP, RREG32(MC_SEQ_MISC_TIMING2)); + WREG32(MC_SEQ_RD_CTL_D0_LP, RREG32(MC_SEQ_RD_CTL_D0)); + WREG32(MC_SEQ_RD_CTL_D1_LP, RREG32(MC_SEQ_RD_CTL_D1)); + WREG32(MC_SEQ_WR_CTL_D0_LP, RREG32(MC_SEQ_WR_CTL_D0)); + WREG32(MC_SEQ_WR_CTL_D1_LP, RREG32(MC_SEQ_WR_CTL_D1)); + WREG32(MC_SEQ_PMG_CMD_EMRS_LP, RREG32(MC_PMG_CMD_EMRS)); + WREG32(MC_SEQ_PMG_CMD_MRS_LP, RREG32(MC_PMG_CMD_MRS)); + WREG32(MC_SEQ_PMG_CMD_MRS1_LP, RREG32(MC_PMG_CMD_MRS1)); + + ret = radeon_atom_init_mc_reg_table(rdev, module_index, table); + + if (ret) + goto init_mc_done; + + ret = btc_copy_vbios_mc_reg_table(table, eg_table); + + if (ret) + goto init_mc_done; + + btc_set_s0_mc_reg_index(eg_table); + ret = btc_set_mc_special_registers(rdev, eg_table); + + if (ret) + goto init_mc_done; + + btc_set_valid_flag(eg_table); + +init_mc_done: + kfree(table); + + return ret; +} + +static void btc_init_stutter_mode(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + + if (pi->mclk_stutter_mode_threshold) { + if (pi->mem_gddr5) { + tmp = RREG32(MC_PMG_AUTO_CFG); + if ((0x200 & tmp) == 0) { + tmp = (tmp & 0xfffffc0b) | 0x204; + WREG32(MC_PMG_AUTO_CFG, tmp); + } + } + } +} + +bool btc_dpm_vblank_too_short(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 vblank_time = r600_dpm_get_vblank_time(rdev); + u32 switch_limit = pi->mem_gddr5 ? 450 : 100; + + if (vblank_time < switch_limit) + return true; + else + return false; + +} + +static void btc_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct rv7xx_ps *ps = rv770_get_ps(rps); + struct radeon_clock_and_voltage_limits *max_limits; + bool disable_mclk_switching; + u32 mclk, sclk; + u16 vddc, vddci; + + if ((rdev->pm.dpm.new_active_crtc_count > 1) || + btc_dpm_vblank_too_short(rdev)) + disable_mclk_switching = true; + else + disable_mclk_switching = false; + + if (rdev->pm.dpm.ac_power) + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + else + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc; + + if (rdev->pm.dpm.ac_power == false) { + if (ps->high.mclk > max_limits->mclk) + ps->high.mclk = max_limits->mclk; + if (ps->high.sclk > max_limits->sclk) + ps->high.sclk = max_limits->sclk; + if (ps->high.vddc > max_limits->vddc) + ps->high.vddc = max_limits->vddc; + if (ps->high.vddci > max_limits->vddci) + ps->high.vddci = max_limits->vddci; + + if (ps->medium.mclk > max_limits->mclk) + ps->medium.mclk = max_limits->mclk; + if (ps->medium.sclk > max_limits->sclk) + ps->medium.sclk = max_limits->sclk; + if (ps->medium.vddc > max_limits->vddc) + ps->medium.vddc = max_limits->vddc; + if (ps->medium.vddci > max_limits->vddci) + ps->medium.vddci = max_limits->vddci; + + if (ps->low.mclk > max_limits->mclk) + ps->low.mclk = max_limits->mclk; + if (ps->low.sclk > max_limits->sclk) + ps->low.sclk = max_limits->sclk; + if (ps->low.vddc > max_limits->vddc) + ps->low.vddc = max_limits->vddc; + if (ps->low.vddci > max_limits->vddci) + ps->low.vddci = max_limits->vddci; + } + + /* XXX validate the min clocks required for display */ + + if (disable_mclk_switching) { + sclk = ps->low.sclk; + mclk = ps->high.mclk; + vddc = ps->low.vddc; + vddci = ps->high.vddci; + } else { + sclk = ps->low.sclk; + mclk = ps->low.mclk; + vddc = ps->low.vddc; + vddci = ps->low.vddci; + } + + /* adjusted low state */ + ps->low.sclk = sclk; + ps->low.mclk = mclk; + ps->low.vddc = vddc; + ps->low.vddci = vddci; + + btc_skip_blacklist_clocks(rdev, max_limits->sclk, max_limits->mclk, + &ps->low.sclk, &ps->low.mclk); + + /* adjusted medium, high states */ + if (ps->medium.sclk < ps->low.sclk) + ps->medium.sclk = ps->low.sclk; + if (ps->medium.vddc < ps->low.vddc) + ps->medium.vddc = ps->low.vddc; + if (ps->high.sclk < ps->medium.sclk) + ps->high.sclk = ps->medium.sclk; + if (ps->high.vddc < ps->medium.vddc) + ps->high.vddc = ps->medium.vddc; + + if (disable_mclk_switching) { + mclk = ps->low.mclk; + if (mclk < ps->medium.mclk) + mclk = ps->medium.mclk; + if (mclk < ps->high.mclk) + mclk = ps->high.mclk; + ps->low.mclk = mclk; + ps->low.vddci = vddci; + ps->medium.mclk = mclk; + ps->medium.vddci = vddci; + ps->high.mclk = mclk; + ps->high.vddci = vddci; + } else { + if (ps->medium.mclk < ps->low.mclk) + ps->medium.mclk = ps->low.mclk; + if (ps->medium.vddci < ps->low.vddci) + ps->medium.vddci = ps->low.vddci; + if (ps->high.mclk < ps->medium.mclk) + ps->high.mclk = ps->medium.mclk; + if (ps->high.vddci < ps->medium.vddci) + ps->high.vddci = ps->medium.vddci; + } + + btc_skip_blacklist_clocks(rdev, max_limits->sclk, max_limits->mclk, + &ps->medium.sclk, &ps->medium.mclk); + btc_skip_blacklist_clocks(rdev, max_limits->sclk, max_limits->mclk, + &ps->high.sclk, &ps->high.mclk); + + btc_adjust_clock_combinations(rdev, max_limits, &ps->low); + btc_adjust_clock_combinations(rdev, max_limits, &ps->medium); + btc_adjust_clock_combinations(rdev, max_limits, &ps->high); + + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + ps->low.sclk, max_limits->vddc, &ps->low.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + ps->low.mclk, max_limits->vddci, &ps->low.vddci); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + ps->low.mclk, max_limits->vddc, &ps->low.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk, + rdev->clock.current_dispclk, max_limits->vddc, &ps->low.vddc); + + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + ps->medium.sclk, max_limits->vddc, &ps->medium.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + ps->medium.mclk, max_limits->vddci, &ps->medium.vddci); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + ps->medium.mclk, max_limits->vddc, &ps->medium.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk, + rdev->clock.current_dispclk, max_limits->vddc, &ps->medium.vddc); + + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + ps->high.sclk, max_limits->vddc, &ps->high.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + ps->high.mclk, max_limits->vddci, &ps->high.vddci); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + ps->high.mclk, max_limits->vddc, &ps->high.vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk, + rdev->clock.current_dispclk, max_limits->vddc, &ps->high.vddc); + + btc_apply_voltage_delta_rules(rdev, max_limits->vddc, max_limits->vddci, + &ps->low.vddc, &ps->low.vddci); + btc_apply_voltage_delta_rules(rdev, max_limits->vddc, max_limits->vddci, + &ps->medium.vddc, &ps->medium.vddci); + btc_apply_voltage_delta_rules(rdev, max_limits->vddc, max_limits->vddci, + &ps->high.vddc, &ps->high.vddci); + + if ((ps->high.vddc <= rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc) && + (ps->medium.vddc <= rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc) && + (ps->low.vddc <= rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc)) + ps->dc_compatible = true; + else + ps->dc_compatible = false; + + if (ps->low.vddc < rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2) + ps->low.flags &= ~ATOM_PPLIB_R600_FLAGS_PCIEGEN2; + if (ps->medium.vddc < rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2) + ps->medium.flags &= ~ATOM_PPLIB_R600_FLAGS_PCIEGEN2; + if (ps->high.vddc < rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2) + ps->high.flags &= ~ATOM_PPLIB_R600_FLAGS_PCIEGEN2; +} + +static void btc_update_current_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct rv7xx_ps *new_ps = rv770_get_ps(rps); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + eg_pi->current_rps = *rps; + eg_pi->current_ps = *new_ps; + eg_pi->current_rps.ps_priv = &eg_pi->current_ps; +} + +static void btc_update_requested_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct rv7xx_ps *new_ps = rv770_get_ps(rps); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + eg_pi->requested_rps = *rps; + eg_pi->requested_ps = *new_ps; + eg_pi->requested_rps.ps_priv = &eg_pi->requested_ps; +} + +void btc_dpm_reset_asic(struct radeon_device *rdev) +{ + rv770_restrict_performance_levels_before_switch(rdev); + btc_disable_ulv(rdev); + btc_set_boot_state_timing(rdev); + rv770_set_boot_state(rdev); +} + +int btc_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps; + struct radeon_ps *new_ps = &requested_ps; + + btc_update_requested_ps(rdev, new_ps); + + btc_apply_state_adjust_rules(rdev, &eg_pi->requested_rps); + + return 0; +} + +int btc_dpm_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + struct radeon_ps *old_ps = &eg_pi->current_rps; + int ret; + + ret = btc_disable_ulv(rdev); + btc_set_boot_state_timing(rdev); + ret = rv770_restrict_performance_levels_before_switch(rdev); + if (ret) { + DRM_ERROR("rv770_restrict_performance_levels_before_switch failed\n"); + return ret; + } + if (eg_pi->pcie_performance_request) + cypress_notify_link_speed_change_before_state_change(rdev, new_ps, old_ps); + + rv770_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + ret = rv770_halt_smc(rdev); + if (ret) { + DRM_ERROR("rv770_halt_smc failed\n"); + return ret; + } + btc_set_at_for_uvd(rdev, new_ps); + if (eg_pi->smu_uvd_hs) + btc_notify_uvd_to_smc(rdev, new_ps); + ret = cypress_upload_sw_state(rdev, new_ps); + if (ret) { + DRM_ERROR("cypress_upload_sw_state failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = cypress_upload_mc_reg_table(rdev, new_ps); + if (ret) { + DRM_ERROR("cypress_upload_mc_reg_table failed\n"); + return ret; + } + } + + cypress_program_memory_timing_parameters(rdev, new_ps); + + ret = rv770_resume_smc(rdev); + if (ret) { + DRM_ERROR("rv770_resume_smc failed\n"); + return ret; + } + ret = rv770_set_sw_state(rdev); + if (ret) { + DRM_ERROR("rv770_set_sw_state failed\n"); + return ret; + } + rv770_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + if (eg_pi->pcie_performance_request) + cypress_notify_link_speed_change_after_state_change(rdev, new_ps, old_ps); + + ret = btc_set_power_state_conditionally_enable_ulv(rdev, new_ps); + if (ret) { + DRM_ERROR("btc_set_power_state_conditionally_enable_ulv failed\n"); + return ret; + } + + ret = rv770_dpm_force_performance_level(rdev, RADEON_DPM_FORCED_LEVEL_AUTO); + if (ret) { + DRM_ERROR("rv770_dpm_force_performance_level failed\n"); + return ret; + } + + return 0; +} + +void btc_dpm_post_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + + btc_update_current_ps(rdev, new_ps); +} + +int btc_dpm_enable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (pi->gfx_clock_gating) + btc_cg_clock_gating_default(rdev); + + if (btc_dpm_enabled(rdev)) + return -EINVAL; + + if (pi->mg_clock_gating) + btc_mg_clock_gating_default(rdev); + + if (eg_pi->ls_clock_gating) + btc_ls_clock_gating_default(rdev); + + if (pi->voltage_control) { + rv770_enable_voltage_control(rdev, true); + ret = cypress_construct_voltage_tables(rdev); + if (ret) { + DRM_ERROR("cypress_construct_voltage_tables failed\n"); + return ret; + } + } + + if (pi->mvdd_control) { + ret = cypress_get_mvdd_configuration(rdev); + if (ret) { + DRM_ERROR("cypress_get_mvdd_configuration failed\n"); + return ret; + } + } + + if (eg_pi->dynamic_ac_timing) { + ret = btc_initialize_mc_reg_table(rdev); + if (ret) + eg_pi->dynamic_ac_timing = false; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv770_enable_backbias(rdev, true); + + if (pi->dynamic_ss) + cypress_enable_spread_spectrum(rdev, true); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, true); + + rv770_setup_bsp(rdev); + rv770_program_git(rdev); + rv770_program_tp(rdev); + rv770_program_tpp(rdev); + rv770_program_sstp(rdev); + rv770_program_engine_speed_parameters(rdev); + cypress_enable_display_gap(rdev); + rv770_program_vc(rdev); + + if (pi->dynamic_pcie_gen2) + btc_enable_dynamic_pcie_gen2(rdev, true); + + ret = rv770_upload_firmware(rdev); + if (ret) { + DRM_ERROR("rv770_upload_firmware failed\n"); + return ret; + } + ret = cypress_get_table_locations(rdev); + if (ret) { + DRM_ERROR("cypress_get_table_locations failed\n"); + return ret; + } + ret = btc_init_smc_table(rdev, boot_ps); + if (ret) + return ret; + + if (eg_pi->dynamic_ac_timing) { + ret = cypress_populate_mc_reg_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("cypress_populate_mc_reg_table failed\n"); + return ret; + } + } + + cypress_program_response_times(rdev); + r7xx_start_smc(rdev); + ret = cypress_notify_smc_display_change(rdev, false); + if (ret) { + DRM_ERROR("cypress_notify_smc_display_change failed\n"); + return ret; + } + cypress_enable_sclk_control(rdev, true); + + if (eg_pi->memory_transition) + cypress_enable_mclk_control(rdev, true); + + cypress_start_dpm(rdev); + + if (pi->gfx_clock_gating) + btc_cg_clock_gating_enable(rdev, true); + + if (pi->mg_clock_gating) + btc_mg_clock_gating_enable(rdev, true); + + if (eg_pi->ls_clock_gating) + btc_ls_clock_gating_enable(rdev, true); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + PPSMC_Result result; + + ret = rv770_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + + if (result != PPSMC_Result_OK) + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + } + + rv770_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + btc_init_stutter_mode(rdev); + + btc_update_current_ps(rdev, rdev->pm.dpm.boot_ps); + + return 0; +}; + +void btc_dpm_disable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (!btc_dpm_enabled(rdev)) + return; + + rv770_clear_vc(rdev); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, false); + + if (pi->dynamic_pcie_gen2) + btc_enable_dynamic_pcie_gen2(rdev, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + if (pi->gfx_clock_gating) + btc_cg_clock_gating_enable(rdev, false); + + if (pi->mg_clock_gating) + btc_mg_clock_gating_enable(rdev, false); + + if (eg_pi->ls_clock_gating) + btc_ls_clock_gating_enable(rdev, false); + + rv770_stop_dpm(rdev); + btc_reset_to_default(rdev); + btc_stop_smc(rdev); + cypress_enable_spread_spectrum(rdev, false); + + btc_update_current_ps(rdev, rdev->pm.dpm.boot_ps); +} + +void btc_dpm_setup_asic(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + rv770_get_memory_type(rdev); + rv740_read_clock_registers(rdev); + btc_read_arb_registers(rdev); + rv770_read_voltage_smio_registers(rdev); + + if (eg_pi->pcie_performance_request) + cypress_advertise_gen2_capability(rdev); + + rv770_get_pcie_gen2_status(rdev); + rv770_enable_acpi_pm(rdev); +} + +int btc_dpm_init(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi; + struct evergreen_power_info *eg_pi; + struct atom_clock_dividers dividers; + int ret; + + eg_pi = kzalloc(sizeof(struct evergreen_power_info), GFP_KERNEL); + if (eg_pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = eg_pi; + pi = &eg_pi->rv7xx; + + rv770_get_max_vddc(rdev); + + eg_pi->ulv.supported = false; + pi->acpi_vddc = 0; + eg_pi->acpi_vddci = 0; + pi->min_vddc_in_table = 0; + pi->max_vddc_in_table = 0; + + ret = rv7xx_parse_power_table(rdev); + if (ret) + return ret; + ret = r600_parse_extended_power_table(rdev); + if (ret) + return ret; + + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries = + kzalloc(4 * sizeof(struct radeon_clock_voltage_dependency_entry), GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries) { + r600_free_extended_power_table(rdev); + return -ENOMEM; + } + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.count = 4; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].clk = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].v = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].clk = 36000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].v = 800; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].clk = 54000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].v = 800; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].clk = 72000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].v = 800; + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->ref_div = dividers.ref_div + 1; + else + pi->ref_div = R600_REFERENCEDIVIDER_DFLT; + + pi->mclk_strobe_mode_threshold = 40000; + pi->mclk_edc_enable_threshold = 40000; + eg_pi->mclk_edc_wr_enable_threshold = 40000; + + pi->rlp = RV770_RLP_DFLT; + pi->rmp = RV770_RMP_DFLT; + pi->lhp = RV770_LHP_DFLT; + pi->lmp = RV770_LMP_DFLT; + + eg_pi->ats[0].rlp = RV770_RLP_DFLT; + eg_pi->ats[0].rmp = RV770_RMP_DFLT; + eg_pi->ats[0].lhp = RV770_LHP_DFLT; + eg_pi->ats[0].lmp = RV770_LMP_DFLT; + + eg_pi->ats[1].rlp = BTC_RLP_UVD_DFLT; + eg_pi->ats[1].rmp = BTC_RMP_UVD_DFLT; + eg_pi->ats[1].lhp = BTC_LHP_UVD_DFLT; + eg_pi->ats[1].lmp = BTC_LMP_UVD_DFLT; + + eg_pi->smu_uvd_hs = true; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0); + + pi->mvdd_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_MVDDC, 0); + + eg_pi->vddci_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDCI, 0); + + rv770_get_engine_memory_ss(rdev); + + pi->asi = RV770_ASI_DFLT; + pi->pasi = CYPRESS_HASI_DFLT; + pi->vrc = CYPRESS_VRC_DFLT; + + pi->power_gating = false; + + pi->gfx_clock_gating = true; + + pi->mg_clock_gating = true; + pi->mgcgtssm = true; + eg_pi->ls_clock_gating = false; + eg_pi->sclk_deep_sleep = false; + + pi->dynamic_pcie_gen2 = true; + + if (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + pi->display_gap = true; + + if (rdev->flags & RADEON_IS_MOBILITY) + pi->dcodt = true; + else + pi->dcodt = false; + + pi->ulps = true; + + eg_pi->dynamic_ac_timing = true; + eg_pi->abm = true; + eg_pi->mcls = true; + eg_pi->light_sleep = true; + eg_pi->memory_transition = true; +#if defined(CONFIG_ACPI) + eg_pi->pcie_performance_request = + radeon_acpi_is_pcie_performance_request_supported(rdev); +#else + eg_pi->pcie_performance_request = false; +#endif + + if (rdev->family == CHIP_BARTS) + eg_pi->dll_default_on = true; + else + eg_pi->dll_default_on = false; + + eg_pi->sclk_deep_sleep = false; + if (ASIC_IS_LOMBOK(rdev)) + pi->mclk_stutter_mode_threshold = 30000; + else + pi->mclk_stutter_mode_threshold = 0; + + pi->sram_end = SMC_RAM_END; + + rdev->pm.dpm.dyn_state.mclk_sclk_ratio = 4; + rdev->pm.dpm.dyn_state.vddc_vddci_delta = 200; + rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2 = 900; + rdev->pm.dpm.dyn_state.valid_sclk_values.count = ARRAY_SIZE(btc_valid_sclk); + rdev->pm.dpm.dyn_state.valid_sclk_values.values = btc_valid_sclk; + rdev->pm.dpm.dyn_state.valid_mclk_values.count = 0; + rdev->pm.dpm.dyn_state.valid_mclk_values.values = NULL; + + if (rdev->family == CHIP_TURKS) + rdev->pm.dpm.dyn_state.sclk_mclk_delta = 15000; + else + rdev->pm.dpm.dyn_state.sclk_mclk_delta = 10000; + + return 0; +} + +void btc_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries); + r600_free_extended_power_table(rdev); +} + +u32 btc_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_ps *requested_state = rv770_get_ps(&eg_pi->requested_rps); + + if (low) + return requested_state->low.sclk; + else + return requested_state->high.sclk; +} + +u32 btc_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_ps *requested_state = rv770_get_ps(&eg_pi->requested_rps); + + if (low) + return requested_state->low.mclk; + else + return requested_state->high.mclk; +} diff --git a/sys/dev/drm/radeon/btc_dpm.h b/sys/dev/drm/radeon/btc_dpm.h new file mode 100644 index 0000000000..1a15e0e419 --- /dev/null +++ b/sys/dev/drm/radeon/btc_dpm.h @@ -0,0 +1,57 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __BTC_DPM_H__ +#define __BTC_DPM_H__ + +#define BTC_RLP_UVD_DFLT 20 +#define BTC_RMP_UVD_DFLT 50 +#define BTC_LHP_UVD_DFLT 50 +#define BTC_LMP_UVD_DFLT 20 +#define BARTS_MGCGCGTSSMCTRL_DFLT 0x81944000 +#define TURKS_MGCGCGTSSMCTRL_DFLT 0x6e944000 +#define CAICOS_MGCGCGTSSMCTRL_DFLT 0x46944040 +#define BTC_CGULVPARAMETER_DFLT 0x00040035 +#define BTC_CGULVCONTROL_DFLT 0x00001450 + +extern u32 btc_valid_sclk[40]; + +void btc_read_arb_registers(struct radeon_device *rdev); +void btc_program_mgcg_hw_sequence(struct radeon_device *rdev, + const u32 *sequence, u32 count); +void btc_skip_blacklist_clocks(struct radeon_device *rdev, + const u32 max_sclk, const u32 max_mclk, + u32 *sclk, u32 *mclk); +void btc_adjust_clock_combinations(struct radeon_device *rdev, + const struct radeon_clock_and_voltage_limits *max_limits, + struct rv7xx_pl *pl); +void btc_apply_voltage_dependency_rules(struct radeon_clock_voltage_dependency_table *table, + u32 clock, u16 max_voltage, u16 *voltage); +void btc_apply_voltage_delta_rules(struct radeon_device *rdev, + u16 max_vddc, u16 max_vddci, + u16 *vddc, u16 *vddci); +bool btc_dpm_enabled(struct radeon_device *rdev); +int btc_reset_to_default(struct radeon_device *rdev); +void btc_notify_uvd_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state); + +#endif diff --git a/sys/dev/drm/radeon/btcd.h b/sys/dev/drm/radeon/btcd.h new file mode 100644 index 0000000000..29e32de7e0 --- /dev/null +++ b/sys/dev/drm/radeon/btcd.h @@ -0,0 +1,181 @@ +/* + * Copyright 2010 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#ifndef _BTCD_H_ +#define _BTCD_H_ + +/* pm registers */ + +#define GENERAL_PWRMGT 0x63c +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define ENABLE_GEN2PCIE (1 << 4) +# define ENABLE_GEN2XSP (1 << 5) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (3 << 6) +# define SW_SMIO_INDEX_SHIFT 6 +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 18) +# define BACKBIAS_VALUE (1 << 19) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +# define AC_DC_SW (1 << 24) + +#define CG_BIF_REQ_AND_RSP 0x7f4 +#define CG_CLIENT_REQ(x) ((x) << 0) +#define CG_CLIENT_REQ_MASK (0xff << 0) +#define CG_CLIENT_REQ_SHIFT 0 +#define CG_CLIENT_RESP(x) ((x) << 8) +#define CG_CLIENT_RESP_MASK (0xff << 8) +#define CG_CLIENT_RESP_SHIFT 8 +#define CLIENT_CG_REQ(x) ((x) << 16) +#define CLIENT_CG_REQ_MASK (0xff << 16) +#define CLIENT_CG_REQ_SHIFT 16 +#define CLIENT_CG_RESP(x) ((x) << 24) +#define CLIENT_CG_RESP_MASK (0xff << 24) +#define CLIENT_CG_RESP_SHIFT 24 + +#define SCLK_PSKIP_CNTL 0x8c0 +#define PSKIP_ON_ALLOW_STOP_HI(x) ((x) << 16) +#define PSKIP_ON_ALLOW_STOP_HI_MASK (0xff << 16) +#define PSKIP_ON_ALLOW_STOP_HI_SHIFT 16 + +#define CG_ULV_CONTROL 0x8c8 +#define CG_ULV_PARAMETER 0x8cc + +#define MC_ARB_DRAM_TIMING 0x2774 +#define MC_ARB_DRAM_TIMING2 0x2778 + +#define MC_ARB_RFSH_RATE 0x27b0 +#define POWERMODE0(x) ((x) << 0) +#define POWERMODE0_MASK (0xff << 0) +#define POWERMODE0_SHIFT 0 +#define POWERMODE1(x) ((x) << 8) +#define POWERMODE1_MASK (0xff << 8) +#define POWERMODE1_SHIFT 8 +#define POWERMODE2(x) ((x) << 16) +#define POWERMODE2_MASK (0xff << 16) +#define POWERMODE2_SHIFT 16 +#define POWERMODE3(x) ((x) << 24) +#define POWERMODE3_MASK (0xff << 24) +#define POWERMODE3_SHIFT 24 + +#define MC_ARB_BURST_TIME 0x2808 +#define STATE0(x) ((x) << 0) +#define STATE0_MASK (0x1f << 0) +#define STATE0_SHIFT 0 +#define STATE1(x) ((x) << 5) +#define STATE1_MASK (0x1f << 5) +#define STATE1_SHIFT 5 +#define STATE2(x) ((x) << 10) +#define STATE2_MASK (0x1f << 10) +#define STATE2_SHIFT 10 +#define STATE3(x) ((x) << 15) +#define STATE3_MASK (0x1f << 15) +#define STATE3_SHIFT 15 + +#define MC_SEQ_RAS_TIMING 0x28a0 +#define MC_SEQ_CAS_TIMING 0x28a4 +#define MC_SEQ_MISC_TIMING 0x28a8 +#define MC_SEQ_MISC_TIMING2 0x28ac + +#define MC_SEQ_RD_CTL_D0 0x28b4 +#define MC_SEQ_RD_CTL_D1 0x28b8 +#define MC_SEQ_WR_CTL_D0 0x28bc +#define MC_SEQ_WR_CTL_D1 0x28c0 + +#define MC_PMG_AUTO_CFG 0x28d4 + +#define MC_SEQ_STATUS_M 0x29f4 +# define PMG_PWRSTATE (1 << 16) + +#define MC_SEQ_MISC0 0x2a00 +#define MC_SEQ_MISC0_GDDR5_SHIFT 28 +#define MC_SEQ_MISC0_GDDR5_MASK 0xf0000000 +#define MC_SEQ_MISC0_GDDR5_VALUE 5 +#define MC_SEQ_MISC1 0x2a04 +#define MC_SEQ_RESERVE_M 0x2a08 +#define MC_PMG_CMD_EMRS 0x2a0c + +#define MC_SEQ_MISC3 0x2a2c + +#define MC_SEQ_MISC5 0x2a54 +#define MC_SEQ_MISC6 0x2a58 + +#define MC_SEQ_MISC7 0x2a64 + +#define MC_SEQ_CG 0x2a68 +#define CG_SEQ_REQ(x) ((x) << 0) +#define CG_SEQ_REQ_MASK (0xff << 0) +#define CG_SEQ_REQ_SHIFT 0 +#define CG_SEQ_RESP(x) ((x) << 8) +#define CG_SEQ_RESP_MASK (0xff << 8) +#define CG_SEQ_RESP_SHIFT 8 +#define SEQ_CG_REQ(x) ((x) << 16) +#define SEQ_CG_REQ_MASK (0xff << 16) +#define SEQ_CG_REQ_SHIFT 16 +#define SEQ_CG_RESP(x) ((x) << 24) +#define SEQ_CG_RESP_MASK (0xff << 24) +#define SEQ_CG_RESP_SHIFT 24 +#define MC_SEQ_RAS_TIMING_LP 0x2a6c +#define MC_SEQ_CAS_TIMING_LP 0x2a70 +#define MC_SEQ_MISC_TIMING_LP 0x2a74 +#define MC_SEQ_MISC_TIMING2_LP 0x2a78 +#define MC_SEQ_WR_CTL_D0_LP 0x2a7c +#define MC_SEQ_WR_CTL_D1_LP 0x2a80 +#define MC_SEQ_PMG_CMD_EMRS_LP 0x2a84 +#define MC_SEQ_PMG_CMD_MRS_LP 0x2a88 + +#define MC_PMG_CMD_MRS 0x2aac + +#define MC_SEQ_RD_CTL_D0_LP 0x2b1c +#define MC_SEQ_RD_CTL_D1_LP 0x2b20 + +#define MC_PMG_CMD_MRS1 0x2b44 +#define MC_SEQ_PMG_CMD_MRS1_LP 0x2b48 + +#define LB_SYNC_RESET_SEL 0x6b28 +#define LB_SYNC_RESET_SEL_MASK (3 << 0) +#define LB_SYNC_RESET_SEL_SHIFT 0 + +/* PCIE link stuff */ +#define PCIE_LC_SPEED_CNTL 0xa4 /* PCIE_P */ +# define LC_GEN2_EN_STRAP (1 << 0) +# define LC_TARGET_LINK_SPEED_OVERRIDE_EN (1 << 1) +# define LC_FORCE_EN_HW_SPEED_CHANGE (1 << 5) +# define LC_FORCE_DIS_HW_SPEED_CHANGE (1 << 6) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK (0x3 << 8) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT 3 +# define LC_CURRENT_DATA_RATE (1 << 11) +# define LC_HW_VOLTAGE_IF_CONTROL(x) ((x) << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_MASK (3 << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_SHIFT 12 +# define LC_VOLTAGE_TIMER_SEL_MASK (0xf << 14) +# define LC_CLR_FAILED_SPD_CHANGE_CNT (1 << 21) +# define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 23) +# define LC_OTHER_SIDE_SUPPORTS_GEN2 (1 << 24) + +#endif diff --git a/sys/dev/drm/radeon/cayman_blit_shaders.c b/sys/dev/drm/radeon/cayman_blit_shaders.c index 9bc3657472..6b36b06e2e 100644 --- a/sys/dev/drm/radeon/cayman_blit_shaders.c +++ b/sys/dev/drm/radeon/cayman_blit_shaders.c @@ -22,8 +22,6 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/cayman_blit_shaders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/cayman_blit_shaders.h b/sys/dev/drm/radeon/cayman_blit_shaders.h index bc1a689bda..f5d0e9a602 100644 --- a/sys/dev/drm/radeon/cayman_blit_shaders.h +++ b/sys/dev/drm/radeon/cayman_blit_shaders.h @@ -20,7 +20,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/cayman_blit_shaders.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef CAYMAN_BLIT_SHADERS_H diff --git a/sys/dev/drm/radeon/cik.c b/sys/dev/drm/radeon/cik.c new file mode 100644 index 0000000000..deaa610d45 --- /dev/null +++ b/sys/dev/drm/radeon/cik.c @@ -0,0 +1,7039 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#include +#include +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "cikd.h" +#include "atom.h" +#include "cik_blit_shaders.h" + +/* GFX */ +#define CIK_PFP_UCODE_SIZE 2144 +#define CIK_ME_UCODE_SIZE 2144 +#define CIK_CE_UCODE_SIZE 2144 +/* compute */ +#define CIK_MEC_UCODE_SIZE 4192 +/* interrupts */ +#define BONAIRE_RLC_UCODE_SIZE 2048 +#define KB_RLC_UCODE_SIZE 2560 +#define KV_RLC_UCODE_SIZE 2560 +/* gddr controller */ +#define CIK_MC_UCODE_SIZE 7866 +/* sdma */ +#define CIK_SDMA_UCODE_SIZE 1050 +#define CIK_SDMA_UCODE_VERSION 64 + +MODULE_FIRMWARE("radeon/BONAIRE_pfp.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_me.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_ce.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_mec.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_mc.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_rlc.bin"); +MODULE_FIRMWARE("radeon/BONAIRE_sdma.bin"); +MODULE_FIRMWARE("radeon/KAVERI_pfp.bin"); +MODULE_FIRMWARE("radeon/KAVERI_me.bin"); +MODULE_FIRMWARE("radeon/KAVERI_ce.bin"); +MODULE_FIRMWARE("radeon/KAVERI_mec.bin"); +MODULE_FIRMWARE("radeon/KAVERI_rlc.bin"); +MODULE_FIRMWARE("radeon/KAVERI_sdma.bin"); +MODULE_FIRMWARE("radeon/KABINI_pfp.bin"); +MODULE_FIRMWARE("radeon/KABINI_me.bin"); +MODULE_FIRMWARE("radeon/KABINI_ce.bin"); +MODULE_FIRMWARE("radeon/KABINI_mec.bin"); +MODULE_FIRMWARE("radeon/KABINI_rlc.bin"); +MODULE_FIRMWARE("radeon/KABINI_sdma.bin"); + +static void cik_rlc_stop(struct radeon_device *rdev); + +/* + * Indirect registers accessor + */ +u32 cik_pciep_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(PCIE_INDEX, reg); + (void)RREG32(PCIE_INDEX); + r = RREG32(PCIE_DATA); + return r; +} + +void cik_pciep_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(PCIE_INDEX, reg); + (void)RREG32(PCIE_INDEX); + WREG32(PCIE_DATA, v); + (void)RREG32(PCIE_DATA); +} + +static const u32 bonaire_golden_spm_registers[] = +{ + 0x30800, 0xe0ffffff, 0xe0000000 +}; + +static const u32 bonaire_golden_common_registers[] = +{ + 0xc770, 0xffffffff, 0x00000800, + 0xc774, 0xffffffff, 0x00000800, + 0xc798, 0xffffffff, 0x00007fbf, + 0xc79c, 0xffffffff, 0x00007faf +}; + +static const u32 bonaire_golden_registers[] = +{ + 0x3354, 0x00000333, 0x00000333, + 0x3350, 0x000c0fc0, 0x00040200, + 0x9a10, 0x00010000, 0x00058208, + 0x3c000, 0xffff1fff, 0x00140000, + 0x3c200, 0xfdfc0fff, 0x00000100, + 0x3c234, 0x40000000, 0x40000200, + 0x9830, 0xffffffff, 0x00000000, + 0x9834, 0xf00fffff, 0x00000400, + 0x9838, 0x0002021c, 0x00020200, + 0xc78, 0x00000080, 0x00000000, + 0x5bb0, 0x000000f0, 0x00000070, + 0x5bc0, 0xf0311fff, 0x80300000, + 0x98f8, 0x73773777, 0x12010001, + 0x350c, 0x00810000, 0x408af000, + 0x7030, 0x31000111, 0x00000011, + 0x2f48, 0x73773777, 0x12010001, + 0x220c, 0x00007fb6, 0x0021a1b1, + 0x2210, 0x00007fb6, 0x002021b1, + 0x2180, 0x00007fb6, 0x00002191, + 0x2218, 0x00007fb6, 0x002121b1, + 0x221c, 0x00007fb6, 0x002021b1, + 0x21dc, 0x00007fb6, 0x00002191, + 0x21e0, 0x00007fb6, 0x00002191, + 0x3628, 0x0000003f, 0x0000000a, + 0x362c, 0x0000003f, 0x0000000a, + 0x2ae4, 0x00073ffe, 0x000022a2, + 0x240c, 0x000007ff, 0x00000000, + 0x8a14, 0xf000003f, 0x00000007, + 0x8bf0, 0x00002001, 0x00000001, + 0x8b24, 0xffffffff, 0x00ffffff, + 0x30a04, 0x0000ff0f, 0x00000000, + 0x28a4c, 0x07ffffff, 0x06000000, + 0x4d8, 0x00000fff, 0x00000100, + 0x3e78, 0x00000001, 0x00000002, + 0x9100, 0x03000000, 0x0362c688, + 0x8c00, 0x000000ff, 0x00000001, + 0xe40, 0x00001fff, 0x00001fff, + 0x9060, 0x0000007f, 0x00000020, + 0x9508, 0x00010000, 0x00010000, + 0xac14, 0x000003ff, 0x000000f3, + 0xac0c, 0xffffffff, 0x00001032 +}; + +static const u32 bonaire_mgcg_cgcg_init[] = +{ + 0xc420, 0xffffffff, 0xfffffffc, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c2a0, 0xffffffff, 0x00000100, + 0x3c208, 0xffffffff, 0x00000100, + 0x3c2c0, 0xffffffff, 0xc0000100, + 0x3c2c8, 0xffffffff, 0xc0000100, + 0x3c2c4, 0xffffffff, 0xc0000100, + 0x55e4, 0xffffffff, 0x00600100, + 0x3c280, 0xffffffff, 0x00000100, + 0x3c214, 0xffffffff, 0x06000100, + 0x3c220, 0xffffffff, 0x00000100, + 0x3c218, 0xffffffff, 0x06000100, + 0x3c204, 0xffffffff, 0x00000100, + 0x3c2e0, 0xffffffff, 0x00000100, + 0x3c224, 0xffffffff, 0x00000100, + 0x3c200, 0xffffffff, 0x00000100, + 0x3c230, 0xffffffff, 0x00000100, + 0x3c234, 0xffffffff, 0x00000100, + 0x3c250, 0xffffffff, 0x00000100, + 0x3c254, 0xffffffff, 0x00000100, + 0x3c258, 0xffffffff, 0x00000100, + 0x3c25c, 0xffffffff, 0x00000100, + 0x3c260, 0xffffffff, 0x00000100, + 0x3c27c, 0xffffffff, 0x00000100, + 0x3c278, 0xffffffff, 0x00000100, + 0x3c210, 0xffffffff, 0x06000100, + 0x3c290, 0xffffffff, 0x00000100, + 0x3c274, 0xffffffff, 0x00000100, + 0x3c2b4, 0xffffffff, 0x00000100, + 0x3c2b0, 0xffffffff, 0x00000100, + 0x3c270, 0xffffffff, 0x00000100, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c020, 0xffffffff, 0x00010000, + 0x3c024, 0xffffffff, 0x00030002, + 0x3c028, 0xffffffff, 0x00040007, + 0x3c02c, 0xffffffff, 0x00060005, + 0x3c030, 0xffffffff, 0x00090008, + 0x3c034, 0xffffffff, 0x00010000, + 0x3c038, 0xffffffff, 0x00030002, + 0x3c03c, 0xffffffff, 0x00040007, + 0x3c040, 0xffffffff, 0x00060005, + 0x3c044, 0xffffffff, 0x00090008, + 0x3c048, 0xffffffff, 0x00010000, + 0x3c04c, 0xffffffff, 0x00030002, + 0x3c050, 0xffffffff, 0x00040007, + 0x3c054, 0xffffffff, 0x00060005, + 0x3c058, 0xffffffff, 0x00090008, + 0x3c05c, 0xffffffff, 0x00010000, + 0x3c060, 0xffffffff, 0x00030002, + 0x3c064, 0xffffffff, 0x00040007, + 0x3c068, 0xffffffff, 0x00060005, + 0x3c06c, 0xffffffff, 0x00090008, + 0x3c070, 0xffffffff, 0x00010000, + 0x3c074, 0xffffffff, 0x00030002, + 0x3c078, 0xffffffff, 0x00040007, + 0x3c07c, 0xffffffff, 0x00060005, + 0x3c080, 0xffffffff, 0x00090008, + 0x3c084, 0xffffffff, 0x00010000, + 0x3c088, 0xffffffff, 0x00030002, + 0x3c08c, 0xffffffff, 0x00040007, + 0x3c090, 0xffffffff, 0x00060005, + 0x3c094, 0xffffffff, 0x00090008, + 0x3c098, 0xffffffff, 0x00010000, + 0x3c09c, 0xffffffff, 0x00030002, + 0x3c0a0, 0xffffffff, 0x00040007, + 0x3c0a4, 0xffffffff, 0x00060005, + 0x3c0a8, 0xffffffff, 0x00090008, + 0x3c000, 0xffffffff, 0x96e00200, + 0x8708, 0xffffffff, 0x00900100, + 0xc424, 0xffffffff, 0x0020003f, + 0x38, 0xffffffff, 0x0140001c, + 0x3c, 0x000f0000, 0x000f0000, + 0x220, 0xffffffff, 0xC060000C, + 0x224, 0xc0000fff, 0x00000100, + 0xf90, 0xffffffff, 0x00000100, + 0xf98, 0x00000101, 0x00000000, + 0x20a8, 0xffffffff, 0x00000104, + 0x55e4, 0xff000fff, 0x00000100, + 0x30cc, 0xc0000fff, 0x00000104, + 0xc1e4, 0x00000001, 0x00000001, + 0xd00c, 0xff000ff0, 0x00000100, + 0xd80c, 0xff000ff0, 0x00000100 +}; + +static const u32 spectre_golden_spm_registers[] = +{ + 0x30800, 0xe0ffffff, 0xe0000000 +}; + +static const u32 spectre_golden_common_registers[] = +{ + 0xc770, 0xffffffff, 0x00000800, + 0xc774, 0xffffffff, 0x00000800, + 0xc798, 0xffffffff, 0x00007fbf, + 0xc79c, 0xffffffff, 0x00007faf +}; + +static const u32 spectre_golden_registers[] = +{ + 0x3c000, 0xffff1fff, 0x96940200, + 0x3c00c, 0xffff0001, 0xff000000, + 0x3c200, 0xfffc0fff, 0x00000100, + 0x6ed8, 0x00010101, 0x00010000, + 0x9834, 0xf00fffff, 0x00000400, + 0x9838, 0xfffffffc, 0x00020200, + 0x5bb0, 0x000000f0, 0x00000070, + 0x5bc0, 0xf0311fff, 0x80300000, + 0x98f8, 0x73773777, 0x12010001, + 0x9b7c, 0x00ff0000, 0x00fc0000, + 0x2f48, 0x73773777, 0x12010001, + 0x8a14, 0xf000003f, 0x00000007, + 0x8b24, 0xffffffff, 0x00ffffff, + 0x28350, 0x3f3f3fff, 0x00000082, + 0x28355, 0x0000003f, 0x00000000, + 0x3e78, 0x00000001, 0x00000002, + 0x913c, 0xffff03df, 0x00000004, + 0xc768, 0x00000008, 0x00000008, + 0x8c00, 0x000008ff, 0x00000800, + 0x9508, 0x00010000, 0x00010000, + 0xac0c, 0xffffffff, 0x54763210, + 0x214f8, 0x01ff01ff, 0x00000002, + 0x21498, 0x007ff800, 0x00200000, + 0x2015c, 0xffffffff, 0x00000f40, + 0x30934, 0xffffffff, 0x00000001 +}; + +static const u32 spectre_mgcg_cgcg_init[] = +{ + 0xc420, 0xffffffff, 0xfffffffc, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c2a0, 0xffffffff, 0x00000100, + 0x3c208, 0xffffffff, 0x00000100, + 0x3c2c0, 0xffffffff, 0x00000100, + 0x3c2c8, 0xffffffff, 0x00000100, + 0x3c2c4, 0xffffffff, 0x00000100, + 0x55e4, 0xffffffff, 0x00600100, + 0x3c280, 0xffffffff, 0x00000100, + 0x3c214, 0xffffffff, 0x06000100, + 0x3c220, 0xffffffff, 0x00000100, + 0x3c218, 0xffffffff, 0x06000100, + 0x3c204, 0xffffffff, 0x00000100, + 0x3c2e0, 0xffffffff, 0x00000100, + 0x3c224, 0xffffffff, 0x00000100, + 0x3c200, 0xffffffff, 0x00000100, + 0x3c230, 0xffffffff, 0x00000100, + 0x3c234, 0xffffffff, 0x00000100, + 0x3c250, 0xffffffff, 0x00000100, + 0x3c254, 0xffffffff, 0x00000100, + 0x3c258, 0xffffffff, 0x00000100, + 0x3c25c, 0xffffffff, 0x00000100, + 0x3c260, 0xffffffff, 0x00000100, + 0x3c27c, 0xffffffff, 0x00000100, + 0x3c278, 0xffffffff, 0x00000100, + 0x3c210, 0xffffffff, 0x06000100, + 0x3c290, 0xffffffff, 0x00000100, + 0x3c274, 0xffffffff, 0x00000100, + 0x3c2b4, 0xffffffff, 0x00000100, + 0x3c2b0, 0xffffffff, 0x00000100, + 0x3c270, 0xffffffff, 0x00000100, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c020, 0xffffffff, 0x00010000, + 0x3c024, 0xffffffff, 0x00030002, + 0x3c028, 0xffffffff, 0x00040007, + 0x3c02c, 0xffffffff, 0x00060005, + 0x3c030, 0xffffffff, 0x00090008, + 0x3c034, 0xffffffff, 0x00010000, + 0x3c038, 0xffffffff, 0x00030002, + 0x3c03c, 0xffffffff, 0x00040007, + 0x3c040, 0xffffffff, 0x00060005, + 0x3c044, 0xffffffff, 0x00090008, + 0x3c048, 0xffffffff, 0x00010000, + 0x3c04c, 0xffffffff, 0x00030002, + 0x3c050, 0xffffffff, 0x00040007, + 0x3c054, 0xffffffff, 0x00060005, + 0x3c058, 0xffffffff, 0x00090008, + 0x3c05c, 0xffffffff, 0x00010000, + 0x3c060, 0xffffffff, 0x00030002, + 0x3c064, 0xffffffff, 0x00040007, + 0x3c068, 0xffffffff, 0x00060005, + 0x3c06c, 0xffffffff, 0x00090008, + 0x3c070, 0xffffffff, 0x00010000, + 0x3c074, 0xffffffff, 0x00030002, + 0x3c078, 0xffffffff, 0x00040007, + 0x3c07c, 0xffffffff, 0x00060005, + 0x3c080, 0xffffffff, 0x00090008, + 0x3c084, 0xffffffff, 0x00010000, + 0x3c088, 0xffffffff, 0x00030002, + 0x3c08c, 0xffffffff, 0x00040007, + 0x3c090, 0xffffffff, 0x00060005, + 0x3c094, 0xffffffff, 0x00090008, + 0x3c098, 0xffffffff, 0x00010000, + 0x3c09c, 0xffffffff, 0x00030002, + 0x3c0a0, 0xffffffff, 0x00040007, + 0x3c0a4, 0xffffffff, 0x00060005, + 0x3c0a8, 0xffffffff, 0x00090008, + 0x3c0ac, 0xffffffff, 0x00010000, + 0x3c0b0, 0xffffffff, 0x00030002, + 0x3c0b4, 0xffffffff, 0x00040007, + 0x3c0b8, 0xffffffff, 0x00060005, + 0x3c0bc, 0xffffffff, 0x00090008, + 0x3c000, 0xffffffff, 0x96e00200, + 0x8708, 0xffffffff, 0x00900100, + 0xc424, 0xffffffff, 0x0020003f, + 0x38, 0xffffffff, 0x0140001c, + 0x3c, 0x000f0000, 0x000f0000, + 0x220, 0xffffffff, 0xC060000C, + 0x224, 0xc0000fff, 0x00000100, + 0xf90, 0xffffffff, 0x00000100, + 0xf98, 0x00000101, 0x00000000, + 0x20a8, 0xffffffff, 0x00000104, + 0x55e4, 0xff000fff, 0x00000100, + 0x30cc, 0xc0000fff, 0x00000104, + 0xc1e4, 0x00000001, 0x00000001, + 0xd00c, 0xff000ff0, 0x00000100, + 0xd80c, 0xff000ff0, 0x00000100 +}; + +static const u32 kalindi_golden_spm_registers[] = +{ + 0x30800, 0xe0ffffff, 0xe0000000 +}; + +static const u32 kalindi_golden_common_registers[] = +{ + 0xc770, 0xffffffff, 0x00000800, + 0xc774, 0xffffffff, 0x00000800, + 0xc798, 0xffffffff, 0x00007fbf, + 0xc79c, 0xffffffff, 0x00007faf +}; + +static const u32 kalindi_golden_registers[] = +{ + 0x3c000, 0xffffdfff, 0x6e944040, + 0x55e4, 0xff607fff, 0xfc000100, + 0x3c220, 0xff000fff, 0x00000100, + 0x3c224, 0xff000fff, 0x00000100, + 0x3c200, 0xfffc0fff, 0x00000100, + 0x6ed8, 0x00010101, 0x00010000, + 0x9830, 0xffffffff, 0x00000000, + 0x9834, 0xf00fffff, 0x00000400, + 0x5bb0, 0x000000f0, 0x00000070, + 0x5bc0, 0xf0311fff, 0x80300000, + 0x98f8, 0x73773777, 0x12010001, + 0x98fc, 0xffffffff, 0x00000010, + 0x9b7c, 0x00ff0000, 0x00fc0000, + 0x8030, 0x00001f0f, 0x0000100a, + 0x2f48, 0x73773777, 0x12010001, + 0x2408, 0x000fffff, 0x000c007f, + 0x8a14, 0xf000003f, 0x00000007, + 0x8b24, 0x3fff3fff, 0x00ffcfff, + 0x30a04, 0x0000ff0f, 0x00000000, + 0x28a4c, 0x07ffffff, 0x06000000, + 0x4d8, 0x00000fff, 0x00000100, + 0x3e78, 0x00000001, 0x00000002, + 0xc768, 0x00000008, 0x00000008, + 0x8c00, 0x000000ff, 0x00000003, + 0x214f8, 0x01ff01ff, 0x00000002, + 0x21498, 0x007ff800, 0x00200000, + 0x2015c, 0xffffffff, 0x00000f40, + 0x88c4, 0x001f3ae3, 0x00000082, + 0x88d4, 0x0000001f, 0x00000010, + 0x30934, 0xffffffff, 0x00000000 +}; + +static const u32 kalindi_mgcg_cgcg_init[] = +{ + 0xc420, 0xffffffff, 0xfffffffc, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c2a0, 0xffffffff, 0x00000100, + 0x3c208, 0xffffffff, 0x00000100, + 0x3c2c0, 0xffffffff, 0x00000100, + 0x3c2c8, 0xffffffff, 0x00000100, + 0x3c2c4, 0xffffffff, 0x00000100, + 0x55e4, 0xffffffff, 0x00600100, + 0x3c280, 0xffffffff, 0x00000100, + 0x3c214, 0xffffffff, 0x06000100, + 0x3c220, 0xffffffff, 0x00000100, + 0x3c218, 0xffffffff, 0x06000100, + 0x3c204, 0xffffffff, 0x00000100, + 0x3c2e0, 0xffffffff, 0x00000100, + 0x3c224, 0xffffffff, 0x00000100, + 0x3c200, 0xffffffff, 0x00000100, + 0x3c230, 0xffffffff, 0x00000100, + 0x3c234, 0xffffffff, 0x00000100, + 0x3c250, 0xffffffff, 0x00000100, + 0x3c254, 0xffffffff, 0x00000100, + 0x3c258, 0xffffffff, 0x00000100, + 0x3c25c, 0xffffffff, 0x00000100, + 0x3c260, 0xffffffff, 0x00000100, + 0x3c27c, 0xffffffff, 0x00000100, + 0x3c278, 0xffffffff, 0x00000100, + 0x3c210, 0xffffffff, 0x06000100, + 0x3c290, 0xffffffff, 0x00000100, + 0x3c274, 0xffffffff, 0x00000100, + 0x3c2b4, 0xffffffff, 0x00000100, + 0x3c2b0, 0xffffffff, 0x00000100, + 0x3c270, 0xffffffff, 0x00000100, + 0x30800, 0xffffffff, 0xe0000000, + 0x3c020, 0xffffffff, 0x00010000, + 0x3c024, 0xffffffff, 0x00030002, + 0x3c028, 0xffffffff, 0x00040007, + 0x3c02c, 0xffffffff, 0x00060005, + 0x3c030, 0xffffffff, 0x00090008, + 0x3c034, 0xffffffff, 0x00010000, + 0x3c038, 0xffffffff, 0x00030002, + 0x3c03c, 0xffffffff, 0x00040007, + 0x3c040, 0xffffffff, 0x00060005, + 0x3c044, 0xffffffff, 0x00090008, + 0x3c000, 0xffffffff, 0x96e00200, + 0x8708, 0xffffffff, 0x00900100, + 0xc424, 0xffffffff, 0x0020003f, + 0x38, 0xffffffff, 0x0140001c, + 0x3c, 0x000f0000, 0x000f0000, + 0x220, 0xffffffff, 0xC060000C, + 0x224, 0xc0000fff, 0x00000100, + 0x20a8, 0xffffffff, 0x00000104, + 0x55e4, 0xff000fff, 0x00000100, + 0x30cc, 0xc0000fff, 0x00000104, + 0xc1e4, 0x00000001, 0x00000001, + 0xd00c, 0xff000ff0, 0x00000100, + 0xd80c, 0xff000ff0, 0x00000100 +}; + +static void cik_init_golden_registers(struct radeon_device *rdev) +{ + switch (rdev->family) { + case CHIP_BONAIRE: + radeon_program_register_sequence(rdev, + bonaire_mgcg_cgcg_init, + (const u32)ARRAY_SIZE(bonaire_mgcg_cgcg_init)); + radeon_program_register_sequence(rdev, + bonaire_golden_registers, + (const u32)ARRAY_SIZE(bonaire_golden_registers)); + radeon_program_register_sequence(rdev, + bonaire_golden_common_registers, + (const u32)ARRAY_SIZE(bonaire_golden_common_registers)); + radeon_program_register_sequence(rdev, + bonaire_golden_spm_registers, + (const u32)ARRAY_SIZE(bonaire_golden_spm_registers)); + break; + case CHIP_KABINI: + radeon_program_register_sequence(rdev, + kalindi_mgcg_cgcg_init, + (const u32)ARRAY_SIZE(kalindi_mgcg_cgcg_init)); + radeon_program_register_sequence(rdev, + kalindi_golden_registers, + (const u32)ARRAY_SIZE(kalindi_golden_registers)); + radeon_program_register_sequence(rdev, + kalindi_golden_common_registers, + (const u32)ARRAY_SIZE(kalindi_golden_common_registers)); + radeon_program_register_sequence(rdev, + kalindi_golden_spm_registers, + (const u32)ARRAY_SIZE(kalindi_golden_spm_registers)); + break; + case CHIP_KAVERI: + radeon_program_register_sequence(rdev, + spectre_mgcg_cgcg_init, + (const u32)ARRAY_SIZE(spectre_mgcg_cgcg_init)); + radeon_program_register_sequence(rdev, + spectre_golden_registers, + (const u32)ARRAY_SIZE(spectre_golden_registers)); + radeon_program_register_sequence(rdev, + spectre_golden_common_registers, + (const u32)ARRAY_SIZE(spectre_golden_common_registers)); + radeon_program_register_sequence(rdev, + spectre_golden_spm_registers, + (const u32)ARRAY_SIZE(spectre_golden_spm_registers)); + break; + default: + break; + } +} + +/** + * cik_get_xclk - get the xclk + * + * @rdev: radeon_device pointer + * + * Returns the reference clock used by the gfx engine + * (CIK). + */ +u32 cik_get_xclk(struct radeon_device *rdev) +{ + u32 reference_clock = rdev->clock.spll.reference_freq; + + if (rdev->flags & RADEON_IS_IGP) { + if (RREG32_SMC(GENERAL_PWRMGT) & GPU_COUNTER_CLK) + return reference_clock / 2; + } else { + if (RREG32_SMC(CG_CLKPIN_CNTL) & XTALIN_DIVIDE) + return reference_clock / 4; + } + return reference_clock; +} + +/** + * cik_mm_rdoorbell - read a doorbell dword + * + * @rdev: radeon_device pointer + * @offset: byte offset into the aperture + * + * Returns the value in the doorbell aperture at the + * requested offset (CIK). + */ +u32 cik_mm_rdoorbell(struct radeon_device *rdev, u32 offset) +{ + if (offset < rdev->doorbell.size) { + return readl(((uint8_t __iomem *)rdev->doorbell.ptr) + offset); + } else { + DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", offset); + return 0; + } +} + +/** + * cik_mm_wdoorbell - write a doorbell dword + * + * @rdev: radeon_device pointer + * @offset: byte offset into the aperture + * @v: value to write + * + * Writes @v to the doorbell aperture at the + * requested offset (CIK). + */ +void cik_mm_wdoorbell(struct radeon_device *rdev, u32 offset, u32 v) +{ + if (offset < rdev->doorbell.size) { + writel(v, ((uint8_t __iomem *)rdev->doorbell.ptr) + offset); + } else { + DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", offset); + } +} + +#define BONAIRE_IO_MC_REGS_SIZE 36 + +static const u32 bonaire_io_mc_regs[BONAIRE_IO_MC_REGS_SIZE][2] = +{ + {0x00000070, 0x04400000}, + {0x00000071, 0x80c01803}, + {0x00000072, 0x00004004}, + {0x00000073, 0x00000100}, + {0x00000074, 0x00ff0000}, + {0x00000075, 0x34000000}, + {0x00000076, 0x08000014}, + {0x00000077, 0x00cc08ec}, + {0x00000078, 0x00000400}, + {0x00000079, 0x00000000}, + {0x0000007a, 0x04090000}, + {0x0000007c, 0x00000000}, + {0x0000007e, 0x4408a8e8}, + {0x0000007f, 0x00000304}, + {0x00000080, 0x00000000}, + {0x00000082, 0x00000001}, + {0x00000083, 0x00000002}, + {0x00000084, 0xf3e4f400}, + {0x00000085, 0x052024e3}, + {0x00000087, 0x00000000}, + {0x00000088, 0x01000000}, + {0x0000008a, 0x1c0a0000}, + {0x0000008b, 0xff010000}, + {0x0000008d, 0xffffefff}, + {0x0000008e, 0xfff3efff}, + {0x0000008f, 0xfff3efbf}, + {0x00000092, 0xf7ffffff}, + {0x00000093, 0xffffff7f}, + {0x00000095, 0x00101101}, + {0x00000096, 0x00000fff}, + {0x00000097, 0x00116fff}, + {0x00000098, 0x60010000}, + {0x00000099, 0x10010000}, + {0x0000009a, 0x00006000}, + {0x0000009b, 0x00001000}, + {0x0000009f, 0x00b48000} +}; + +/** + * cik_srbm_select - select specific register instances + * + * @rdev: radeon_device pointer + * @me: selected ME (micro engine) + * @pipe: pipe + * @queue: queue + * @vmid: VMID + * + * Switches the currently active registers instances. Some + * registers are instanced per VMID, others are instanced per + * me/pipe/queue combination. + */ +static void cik_srbm_select(struct radeon_device *rdev, + u32 me, u32 pipe, u32 queue, u32 vmid) +{ + u32 srbm_gfx_cntl = (PIPEID(pipe & 0x3) | + MEID(me & 0x3) | + VMID(vmid & 0xf) | + QUEUEID(queue & 0x7)); + WREG32(SRBM_GFX_CNTL, srbm_gfx_cntl); +} + +/* ucode loading */ +/** + * ci_mc_load_microcode - load MC ucode into the hw + * + * @rdev: radeon_device pointer + * + * Load the GDDR MC ucode into the hw (CIK). + * Returns 0 on success, error on failure. + */ +static __unused int ci_mc_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + u32 running, blackout = 0; + u32 *io_mc_regs; + int i, ucode_size, regs_size; + + if (!rdev->mc_fw) + return -EINVAL; + + switch (rdev->family) { + case CHIP_BONAIRE: + default: + io_mc_regs = (u32 *)&bonaire_io_mc_regs; + ucode_size = CIK_MC_UCODE_SIZE; + regs_size = BONAIRE_IO_MC_REGS_SIZE; + break; + } + + running = RREG32(MC_SEQ_SUP_CNTL) & RUN_MASK; + + if (running == 0) { + if (running) { + blackout = RREG32(MC_SHARED_BLACKOUT_CNTL); + WREG32(MC_SHARED_BLACKOUT_CNTL, blackout | 1); + } + + /* reset the engine and set to writable */ + WREG32(MC_SEQ_SUP_CNTL, 0x00000008); + WREG32(MC_SEQ_SUP_CNTL, 0x00000010); + + /* load mc io regs */ + for (i = 0; i < regs_size; i++) { + WREG32(MC_SEQ_IO_DEBUG_INDEX, io_mc_regs[(i << 1)]); + WREG32(MC_SEQ_IO_DEBUG_DATA, io_mc_regs[(i << 1) + 1]); + } + /* load the MC ucode */ + fw_data = (const __be32 *)rdev->mc_fw->data; + for (i = 0; i < ucode_size; i++) + WREG32(MC_SEQ_SUP_PGM, be32_to_cpup(fw_data++)); + + /* put the engine back into the active state */ + WREG32(MC_SEQ_SUP_CNTL, 0x00000008); + WREG32(MC_SEQ_SUP_CNTL, 0x00000004); + WREG32(MC_SEQ_SUP_CNTL, 0x00000001); + + /* wait for training to complete */ + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D0) + break; + DRM_UDELAY(1); + } + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(MC_SEQ_TRAIN_WAKEUP_CNTL) & TRAIN_DONE_D1) + break; + DRM_UDELAY(1); + } + + if (running) + WREG32(MC_SHARED_BLACKOUT_CNTL, blackout); + } + + return 0; +} + +/** + * cik_init_microcode - load ucode images from disk + * + * @rdev: radeon_device pointer + * + * Use the firmware interface to load the ucode images into + * the driver (not loaded into hw). + * Returns 0 on success, error on failure. + */ +static int cik_init_microcode(struct radeon_device *rdev) +{ + const char *chip_name; + size_t pfp_req_size, me_req_size, ce_req_size, + mec_req_size, rlc_req_size, mc_req_size, + sdma_req_size; + char fw_name[30]; + int err; + + DRM_DEBUG("\n"); + + switch (rdev->family) { + case CHIP_BONAIRE: + chip_name = "BONAIRE"; + pfp_req_size = CIK_PFP_UCODE_SIZE * 4; + me_req_size = CIK_ME_UCODE_SIZE * 4; + ce_req_size = CIK_CE_UCODE_SIZE * 4; + mec_req_size = CIK_MEC_UCODE_SIZE * 4; + rlc_req_size = BONAIRE_RLC_UCODE_SIZE * 4; + mc_req_size = CIK_MC_UCODE_SIZE * 4; + sdma_req_size = CIK_SDMA_UCODE_SIZE * 4; + break; + case CHIP_KAVERI: + chip_name = "KAVERI"; + pfp_req_size = CIK_PFP_UCODE_SIZE * 4; + me_req_size = CIK_ME_UCODE_SIZE * 4; + ce_req_size = CIK_CE_UCODE_SIZE * 4; + mec_req_size = CIK_MEC_UCODE_SIZE * 4; + rlc_req_size = KV_RLC_UCODE_SIZE * 4; + sdma_req_size = CIK_SDMA_UCODE_SIZE * 4; + break; + case CHIP_KABINI: + chip_name = "KABINI"; + pfp_req_size = CIK_PFP_UCODE_SIZE * 4; + me_req_size = CIK_ME_UCODE_SIZE * 4; + ce_req_size = CIK_CE_UCODE_SIZE * 4; + mec_req_size = CIK_MEC_UCODE_SIZE * 4; + rlc_req_size = KB_RLC_UCODE_SIZE * 4; + sdma_req_size = CIK_SDMA_UCODE_SIZE * 4; + break; + default: BUG(); + } + + DRM_INFO("Loading %s Microcode\n", chip_name); + err = 0; + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_pfp", chip_name); + rdev->pfp_fw = firmware_get(fw_name); + if (rdev->pfp_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->pfp_fw->datasize != pfp_req_size) { + printk(KERN_ERR + "cik_cp: Bogus length %zu in firmware \"%s\"\n", + rdev->pfp_fw->datasize, fw_name); + err = -EINVAL; + goto out; + } + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_me", chip_name); + rdev->me_fw = firmware_get(fw_name); + if (rdev->me_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->me_fw->datasize != me_req_size) { + printk(KERN_ERR + "cik_cp: Bogus length %zu in firmware \"%s\"\n", + rdev->me_fw->datasize, fw_name); + err = -EINVAL; + } + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_ce", chip_name); + rdev->ce_fw = firmware_get(fw_name); + if (rdev->ce_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->ce_fw->datasize != ce_req_size) { + printk(KERN_ERR + "cik_cp: Bogus length %zu in firmware \"%s\"\n", + rdev->ce_fw->datasize, fw_name); + err = -EINVAL; + } + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_mec", chip_name); + rdev->mec_fw = firmware_get(fw_name); + if (rdev->mec_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->mec_fw->datasize != mec_req_size) { + printk(KERN_ERR + "cik_cp: Bogus length %zu in firmware \"%s\"\n", + rdev->mec_fw->datasize, fw_name); + err = -EINVAL; + } + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_rlc", chip_name); + rdev->rlc_fw = firmware_get(fw_name); + if (rdev->rlc_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->rlc_fw->datasize != rlc_req_size) { + printk(KERN_ERR + "cik_rlc: Bogus length %zu in firmware \"%s\"\n", + rdev->rlc_fw->datasize, fw_name); + err = -EINVAL; + } + + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_sdma", chip_name); + rdev->sdma_fw = firmware_get(fw_name); + if (rdev->sdma_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->sdma_fw->datasize != sdma_req_size) { + printk(KERN_ERR + "cik_sdma: Bogus length %zu in firmware \"%s\"\n", + rdev->sdma_fw->datasize, fw_name); + err = -EINVAL; + } + + /* No MC ucode on APUs */ + if (!(rdev->flags & RADEON_IS_IGP)) { + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_mc", chip_name); + rdev->mc_fw = firmware_get(fw_name); + if (rdev->mc_fw == NULL) { + err = -ENOENT; + goto out; + } + if (rdev->mc_fw->datasize != mc_req_size) { + printk(KERN_ERR + "cik_mc: Bogus length %zu in firmware \"%s\"\n", + rdev->mc_fw->datasize, fw_name); + err = -EINVAL; + } + } + +out: + if (err) { + if (err != -EINVAL) + printk(KERN_ERR + "cik_cp: Failed to load firmware \"%s\"\n", + fw_name); + if (rdev->pfp_fw != NULL) { + firmware_put(rdev->pfp_fw, FIRMWARE_UNLOAD); + rdev->pfp_fw = NULL; + } + if (rdev->me_fw != NULL) { + firmware_put(rdev->me_fw, FIRMWARE_UNLOAD); + rdev->me_fw = NULL; + } + if (rdev->ce_fw != NULL) { + firmware_put(rdev->ce_fw, FIRMWARE_UNLOAD); + rdev->ce_fw = NULL; + } + if (rdev->mec_fw != NULL) { + firmware_put(rdev->mec_fw, FIRMWARE_UNLOAD); + rdev->mec_fw = NULL; + } + if (rdev->rlc_fw != NULL) { + firmware_put(rdev->rlc_fw, FIRMWARE_UNLOAD); + rdev->rlc_fw = NULL; + } + if (rdev->sdma_fw != NULL) { + firmware_put(rdev->sdma_fw, FIRMWARE_UNLOAD); + rdev->sdma_fw = NULL; + } + if (rdev->mc_fw != NULL) { + firmware_put(rdev->mc_fw, FIRMWARE_UNLOAD); + rdev->mc_fw = NULL; + } + } + return err; +} + +/* + * Core functions + */ +/** + * cik_tiling_mode_table_init - init the hw tiling table + * + * @rdev: radeon_device pointer + * + * Starting with SI, the tiling setup is done globally in a + * set of 32 tiling modes. Rather than selecting each set of + * parameters per surface as on older asics, we just select + * which index in the tiling table we want to use, and the + * surface uses those parameters (CIK). + */ +static void cik_tiling_mode_table_init(struct radeon_device *rdev) +{ + const u32 num_tile_mode_states = 32; + const u32 num_secondary_tile_mode_states = 16; + u32 reg_offset, gb_tile_moden, split_equal_to_row_size; + u32 num_pipe_configs; + u32 num_rbs = rdev->config.cik.max_backends_per_se * + rdev->config.cik.max_shader_engines; + + switch (rdev->config.cik.mem_row_size_in_kb) { + case 1: + split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_1KB; + break; + case 2: + default: + split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_2KB; + break; + case 4: + split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_4KB; + break; + } + + num_pipe_configs = rdev->config.cik.max_tile_pipes; + if (num_pipe_configs > 8) + num_pipe_configs = 8; /* ??? */ + + if (num_pipe_configs == 8) { + for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B)); + break; + case 1: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B)); + break; + case 2: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 3: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B)); + break; + case 4: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 5: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING)); + break; + case 6: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 7: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 8: + gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16)); + break; + case 9: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING)); + break; + case 10: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 11: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 12: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 13: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING)); + break; + case 14: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 16: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 17: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 27: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING)); + break; + case 28: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 29: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 30: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P8_32x32_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + default: + gb_tile_moden = 0; + break; + } + rdev->config.cik.tile_mode_array[reg_offset] = gb_tile_moden; + WREG32(GB_TILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 1: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 2: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 3: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 4: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + case 5: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_4_BANK)); + break; + case 6: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_2_BANK)); + break; + case 8: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_8) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 9: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 10: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 11: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 12: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + case 13: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_4_BANK)); + break; + case 14: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_2_BANK)); + break; + default: + gb_tile_moden = 0; + break; + } + WREG32(GB_MACROTILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + } else if (num_pipe_configs == 4) { + if (num_rbs == 4) { + for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B)); + break; + case 1: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B)); + break; + case 2: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 3: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B)); + break; + case 4: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 5: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING)); + break; + case 6: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 7: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 8: + gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) | + PIPE_CONFIG(ADDR_SURF_P4_16x16)); + break; + case 9: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING)); + break; + case 10: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 11: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 12: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 13: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING)); + break; + case 14: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 16: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 17: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 27: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING)); + break; + case 28: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 29: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 30: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_16x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + default: + gb_tile_moden = 0; + break; + } + rdev->config.cik.tile_mode_array[reg_offset] = gb_tile_moden; + WREG32(GB_TILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + } else if (num_rbs < 4) { + for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B)); + break; + case 1: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B)); + break; + case 2: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 3: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B)); + break; + case 4: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 5: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING)); + break; + case 6: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 7: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 8: + gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) | + PIPE_CONFIG(ADDR_SURF_P4_8x16)); + break; + case 9: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING)); + break; + case 10: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 11: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 12: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 13: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING)); + break; + case 14: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 16: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 17: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 27: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING)); + break; + case 28: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 29: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 30: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P4_8x16) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + default: + gb_tile_moden = 0; + break; + } + rdev->config.cik.tile_mode_array[reg_offset] = gb_tile_moden; + WREG32(GB_TILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + } + for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 1: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 2: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 3: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 4: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 5: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + case 6: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_4_BANK)); + break; + case 8: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_8) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 9: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 10: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 11: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 12: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 13: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + case 14: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) | + NUM_BANKS(ADDR_SURF_4_BANK)); + break; + default: + gb_tile_moden = 0; + break; + } + WREG32(GB_MACROTILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + } else if (num_pipe_configs == 2) { + for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B)); + break; + case 1: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B)); + break; + case 2: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 3: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B)); + break; + case 4: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 5: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING)); + break; + case 6: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B)); + break; + case 7: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + TILE_SPLIT(split_equal_to_row_size)); + break; + case 8: + gb_tile_moden = ARRAY_MODE(ARRAY_LINEAR_ALIGNED); + break; + case 9: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING)); + break; + case 10: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 11: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 12: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 13: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING)); + break; + case 14: + gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 16: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 17: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 27: + gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING)); + break; + case 28: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 29: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + case 30: + gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) | + MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) | + PIPE_CONFIG(ADDR_SURF_P2) | + SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2)); + break; + default: + gb_tile_moden = 0; + break; + } + rdev->config.cik.tile_mode_array[reg_offset] = gb_tile_moden; + WREG32(GB_TILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) { + switch (reg_offset) { + case 0: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 1: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 2: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 3: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 4: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 5: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 6: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + case 8: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_4) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_8) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 9: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_4) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 10: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 11: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 12: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 13: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) | + NUM_BANKS(ADDR_SURF_16_BANK)); + break; + case 14: + gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) | + BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) | + MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) | + NUM_BANKS(ADDR_SURF_8_BANK)); + break; + default: + gb_tile_moden = 0; + break; + } + WREG32(GB_MACROTILE_MODE0 + (reg_offset * 4), gb_tile_moden); + } + } else + DRM_ERROR("unknown num pipe config: 0x%x\n", num_pipe_configs); +} + +/** + * cik_select_se_sh - select which SE, SH to address + * + * @rdev: radeon_device pointer + * @se_num: shader engine to address + * @sh_num: sh block to address + * + * Select which SE, SH combinations to address. Certain + * registers are instanced per SE or SH. 0xffffffff means + * broadcast to all SEs or SHs (CIK). + */ +static void cik_select_se_sh(struct radeon_device *rdev, + u32 se_num, u32 sh_num) +{ + u32 data = INSTANCE_BROADCAST_WRITES; + + if ((se_num == 0xffffffff) && (sh_num == 0xffffffff)) + data |= SH_BROADCAST_WRITES | SE_BROADCAST_WRITES; + else if (se_num == 0xffffffff) + data |= SE_BROADCAST_WRITES | SH_INDEX(sh_num); + else if (sh_num == 0xffffffff) + data |= SH_BROADCAST_WRITES | SE_INDEX(se_num); + else + data |= SH_INDEX(sh_num) | SE_INDEX(se_num); + WREG32(GRBM_GFX_INDEX, data); +} + +/** + * cik_create_bitmask - create a bitmask + * + * @bit_width: length of the mask + * + * create a variable length bit mask (CIK). + * Returns the bitmask. + */ +static u32 cik_create_bitmask(u32 bit_width) +{ + u32 i, mask = 0; + + for (i = 0; i < bit_width; i++) { + mask <<= 1; + mask |= 1; + } + return mask; +} + +/** + * cik_select_se_sh - select which SE, SH to address + * + * @rdev: radeon_device pointer + * @max_rb_num: max RBs (render backends) for the asic + * @se_num: number of SEs (shader engines) for the asic + * @sh_per_se: number of SH blocks per SE for the asic + * + * Calculates the bitmask of disabled RBs (CIK). + * Returns the disabled RB bitmask. + */ +static u32 cik_get_rb_disabled(struct radeon_device *rdev, + u32 max_rb_num, u32 se_num, + u32 sh_per_se) +{ + u32 data, mask; + + data = RREG32(CC_RB_BACKEND_DISABLE); + if (data & 1) + data &= BACKEND_DISABLE_MASK; + else + data = 0; + data |= RREG32(GC_USER_RB_BACKEND_DISABLE); + + data >>= BACKEND_DISABLE_SHIFT; + + mask = cik_create_bitmask(max_rb_num / se_num / sh_per_se); + + return data & mask; +} + +/** + * cik_setup_rb - setup the RBs on the asic + * + * @rdev: radeon_device pointer + * @se_num: number of SEs (shader engines) for the asic + * @sh_per_se: number of SH blocks per SE for the asic + * @max_rb_num: max RBs (render backends) for the asic + * + * Configures per-SE/SH RB registers (CIK). + */ +static void cik_setup_rb(struct radeon_device *rdev, + u32 se_num, u32 sh_per_se, + u32 max_rb_num) +{ + int i, j; + u32 data, mask; + u32 disabled_rbs = 0; + u32 enabled_rbs = 0; + + for (i = 0; i < se_num; i++) { + for (j = 0; j < sh_per_se; j++) { + cik_select_se_sh(rdev, i, j); + data = cik_get_rb_disabled(rdev, max_rb_num, se_num, sh_per_se); + disabled_rbs |= data << ((i * sh_per_se + j) * CIK_RB_BITMAP_WIDTH_PER_SH); + } + } + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + + mask = 1; + for (i = 0; i < max_rb_num; i++) { + if (!(disabled_rbs & mask)) + enabled_rbs |= mask; + mask <<= 1; + } + + for (i = 0; i < se_num; i++) { + cik_select_se_sh(rdev, i, 0xffffffff); + data = 0; + for (j = 0; j < sh_per_se; j++) { + switch (enabled_rbs & 3) { + case 1: + data |= (RASTER_CONFIG_RB_MAP_0 << (i * sh_per_se + j) * 2); + break; + case 2: + data |= (RASTER_CONFIG_RB_MAP_3 << (i * sh_per_se + j) * 2); + break; + case 3: + default: + data |= (RASTER_CONFIG_RB_MAP_2 << (i * sh_per_se + j) * 2); + break; + } + enabled_rbs >>= 2; + } + WREG32(PA_SC_RASTER_CONFIG, data); + } + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); +} + +/** + * cik_gpu_init - setup the 3D engine + * + * @rdev: radeon_device pointer + * + * Configures the 3D engine and tiling configuration + * registers so that the 3D engine is usable. + */ +static __unused void cik_gpu_init(struct radeon_device *rdev) +{ + u32 gb_addr_config = RREG32(GB_ADDR_CONFIG); + u32 mc_shared_chmap, mc_arb_ramcfg; + u32 hdp_host_path_cntl; + u32 tmp; + int i, j; + + switch (rdev->family) { + case CHIP_BONAIRE: + rdev->config.cik.max_shader_engines = 2; + rdev->config.cik.max_tile_pipes = 4; + rdev->config.cik.max_cu_per_sh = 7; + rdev->config.cik.max_sh_per_se = 1; + rdev->config.cik.max_backends_per_se = 2; + rdev->config.cik.max_texture_channel_caches = 4; + rdev->config.cik.max_gprs = 256; + rdev->config.cik.max_gs_threads = 32; + rdev->config.cik.max_hw_contexts = 8; + + rdev->config.cik.sc_prim_fifo_size_frontend = 0x20; + rdev->config.cik.sc_prim_fifo_size_backend = 0x100; + rdev->config.cik.sc_hiz_tile_fifo_size = 0x30; + rdev->config.cik.sc_earlyz_tile_fifo_size = 0x130; + gb_addr_config = BONAIRE_GB_ADDR_CONFIG_GOLDEN; + break; + case CHIP_KAVERI: + /* TODO */ + break; + case CHIP_KABINI: + default: + rdev->config.cik.max_shader_engines = 1; + rdev->config.cik.max_tile_pipes = 2; + rdev->config.cik.max_cu_per_sh = 2; + rdev->config.cik.max_sh_per_se = 1; + rdev->config.cik.max_backends_per_se = 1; + rdev->config.cik.max_texture_channel_caches = 2; + rdev->config.cik.max_gprs = 256; + rdev->config.cik.max_gs_threads = 16; + rdev->config.cik.max_hw_contexts = 8; + + rdev->config.cik.sc_prim_fifo_size_frontend = 0x20; + rdev->config.cik.sc_prim_fifo_size_backend = 0x100; + rdev->config.cik.sc_hiz_tile_fifo_size = 0x30; + rdev->config.cik.sc_earlyz_tile_fifo_size = 0x130; + gb_addr_config = BONAIRE_GB_ADDR_CONFIG_GOLDEN; + break; + } + + /* Initialize HDP */ + for (i = 0, j = 0; i < 32; i++, j += 0x18) { + WREG32((0x2c14 + j), 0x00000000); + WREG32((0x2c18 + j), 0x00000000); + WREG32((0x2c1c + j), 0x00000000); + WREG32((0x2c20 + j), 0x00000000); + WREG32((0x2c24 + j), 0x00000000); + } + + WREG32(GRBM_CNTL, GRBM_READ_TIMEOUT(0xff)); + + WREG32(BIF_FB_EN, FB_READ_EN | FB_WRITE_EN); + + mc_shared_chmap = RREG32(MC_SHARED_CHMAP); + mc_arb_ramcfg = RREG32(MC_ARB_RAMCFG); + + rdev->config.cik.num_tile_pipes = rdev->config.cik.max_tile_pipes; + rdev->config.cik.mem_max_burst_length_bytes = 256; + tmp = (mc_arb_ramcfg & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT; + rdev->config.cik.mem_row_size_in_kb = (4 * (1 << (8 + tmp))) / 1024; + if (rdev->config.cik.mem_row_size_in_kb > 4) + rdev->config.cik.mem_row_size_in_kb = 4; + /* XXX use MC settings? */ + rdev->config.cik.shader_engine_tile_size = 32; + rdev->config.cik.num_gpus = 1; + rdev->config.cik.multi_gpu_tile_size = 64; + + /* fix up row size */ + gb_addr_config &= ~ROW_SIZE_MASK; + switch (rdev->config.cik.mem_row_size_in_kb) { + case 1: + default: + gb_addr_config |= ROW_SIZE(0); + break; + case 2: + gb_addr_config |= ROW_SIZE(1); + break; + case 4: + gb_addr_config |= ROW_SIZE(2); + break; + } + + /* setup tiling info dword. gb_addr_config is not adequate since it does + * not have bank info, so create a custom tiling dword. + * bits 3:0 num_pipes + * bits 7:4 num_banks + * bits 11:8 group_size + * bits 15:12 row_size + */ + rdev->config.cik.tile_config = 0; + switch (rdev->config.cik.num_tile_pipes) { + case 1: + rdev->config.cik.tile_config |= (0 << 0); + break; + case 2: + rdev->config.cik.tile_config |= (1 << 0); + break; + case 4: + rdev->config.cik.tile_config |= (2 << 0); + break; + case 8: + default: + /* XXX what about 12? */ + rdev->config.cik.tile_config |= (3 << 0); + break; + } + if ((mc_arb_ramcfg & NOOFBANK_MASK) >> NOOFBANK_SHIFT) + rdev->config.cik.tile_config |= 1 << 4; + else + rdev->config.cik.tile_config |= 0 << 4; + rdev->config.cik.tile_config |= + ((gb_addr_config & PIPE_INTERLEAVE_SIZE_MASK) >> PIPE_INTERLEAVE_SIZE_SHIFT) << 8; + rdev->config.cik.tile_config |= + ((gb_addr_config & ROW_SIZE_MASK) >> ROW_SIZE_SHIFT) << 12; + + WREG32(GB_ADDR_CONFIG, gb_addr_config); + WREG32(HDP_ADDR_CONFIG, gb_addr_config); + WREG32(DMIF_ADDR_CALC, gb_addr_config); + WREG32(SDMA0_TILING_CONFIG + SDMA0_REGISTER_OFFSET, gb_addr_config & 0x70); + WREG32(SDMA0_TILING_CONFIG + SDMA1_REGISTER_OFFSET, gb_addr_config & 0x70); + WREG32(UVD_UDEC_ADDR_CONFIG, gb_addr_config); + WREG32(UVD_UDEC_DB_ADDR_CONFIG, gb_addr_config); + WREG32(UVD_UDEC_DBW_ADDR_CONFIG, gb_addr_config); + + cik_tiling_mode_table_init(rdev); + + cik_setup_rb(rdev, rdev->config.cik.max_shader_engines, + rdev->config.cik.max_sh_per_se, + rdev->config.cik.max_backends_per_se); + + /* set HW defaults for 3D engine */ + WREG32(CP_MEQ_THRESHOLDS, MEQ1_START(0x30) | MEQ2_START(0x60)); + + WREG32(SX_DEBUG_1, 0x20); + + WREG32(TA_CNTL_AUX, 0x00010000); + + tmp = RREG32(SPI_CONFIG_CNTL); + tmp |= 0x03000000; + WREG32(SPI_CONFIG_CNTL, tmp); + + WREG32(SQ_CONFIG, 1); + + WREG32(DB_DEBUG, 0); + + tmp = RREG32(DB_DEBUG2) & ~0xf00fffff; + tmp |= 0x00000400; + WREG32(DB_DEBUG2, tmp); + + tmp = RREG32(DB_DEBUG3) & ~0x0002021c; + tmp |= 0x00020200; + WREG32(DB_DEBUG3, tmp); + + tmp = RREG32(CB_HW_CONTROL) & ~0x00010000; + tmp |= 0x00018208; + WREG32(CB_HW_CONTROL, tmp); + + WREG32(SPI_CONFIG_CNTL_1, VTX_DONE_DELAY(4)); + + WREG32(PA_SC_FIFO_SIZE, (SC_FRONTEND_PRIM_FIFO_SIZE(rdev->config.cik.sc_prim_fifo_size_frontend) | + SC_BACKEND_PRIM_FIFO_SIZE(rdev->config.cik.sc_prim_fifo_size_backend) | + SC_HIZ_TILE_FIFO_SIZE(rdev->config.cik.sc_hiz_tile_fifo_size) | + SC_EARLYZ_TILE_FIFO_SIZE(rdev->config.cik.sc_earlyz_tile_fifo_size))); + + WREG32(VGT_NUM_INSTANCES, 1); + + WREG32(CP_PERFMON_CNTL, 0); + + WREG32(SQ_CONFIG, 0); + + WREG32(PA_SC_FORCE_EOV_MAX_CNTS, (FORCE_EOV_MAX_CLK_CNT(4095) | + FORCE_EOV_MAX_REZ_CNT(255))); + + WREG32(VGT_CACHE_INVALIDATION, CACHE_INVALIDATION(VC_AND_TC) | + AUTO_INVLD_EN(ES_AND_GS_AUTO)); + + WREG32(VGT_GS_VERTEX_REUSE, 16); + WREG32(PA_SC_LINE_STIPPLE_STATE, 0); + + tmp = RREG32(HDP_MISC_CNTL); + tmp |= HDP_FLUSH_INVALIDATE_CACHE; + WREG32(HDP_MISC_CNTL, tmp); + + hdp_host_path_cntl = RREG32(HDP_HOST_PATH_CNTL); + WREG32(HDP_HOST_PATH_CNTL, hdp_host_path_cntl); + + WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3)); + WREG32(PA_SC_ENHANCE, ENABLE_PA_SC_OUT_OF_ORDER); + + DRM_UDELAY(50); +} + +/* + * GPU scratch registers helpers function. + */ +/** + * cik_scratch_init - setup driver info for CP scratch regs + * + * @rdev: radeon_device pointer + * + * Set up the number and offset of the CP scratch registers. + * NOTE: use of CP scratch registers is a legacy inferface and + * is not used by default on newer asics (r6xx+). On newer asics, + * memory buffers are used for fences rather than scratch regs. + */ +static __unused void cik_scratch_init(struct radeon_device *rdev) +{ + int i; + + rdev->scratch.num_reg = 7; + rdev->scratch.reg_base = SCRATCH_REG0; + for (i = 0; i < rdev->scratch.num_reg; i++) { + rdev->scratch.free[i] = true; + rdev->scratch.reg[i] = rdev->scratch.reg_base + (i * 4); + } +} + +/** + * cik_ring_test - basic gfx ring test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Allocate a scratch register and write to it using the gfx ring (CIK). + * Provides a basic gfx ring test to verify that the ring is working. + * Used by cik_cp_gfx_resume(); + * Returns 0 on success, error on failure. + */ +int cik_ring_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + uint32_t scratch; + uint32_t tmp = 0; + unsigned i; + int r; + + r = radeon_scratch_get(rdev, &scratch); + if (r) { + DRM_ERROR("radeon: cp failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + r = radeon_ring_lock(rdev, ring, 3); + if (r) { + DRM_ERROR("radeon: cp failed to lock ring %d (%d).\n", ring->idx, r); + radeon_scratch_free(rdev, scratch); + return r; + } + radeon_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + radeon_ring_write(ring, ((scratch - PACKET3_SET_UCONFIG_REG_START) >> 2)); + radeon_ring_write(ring, 0xDEADBEEF); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed (scratch(0x%04X)=0x%08X)\n", + ring->idx, scratch, tmp); + r = -EINVAL; + } + radeon_scratch_free(rdev, scratch); + return r; +} + +/** + * cik_fence_gfx_ring_emit - emit a fence on the gfx ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Emits a fence sequnce number on the gfx ring and flushes + * GPU caches. + */ +void cik_fence_gfx_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + + /* EVENT_WRITE_EOP - flush caches, send int */ + radeon_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4)); + radeon_ring_write(ring, (EOP_TCL1_ACTION_EN | + EOP_TC_ACTION_EN | + EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5))); + radeon_ring_write(ring, addr & 0xfffffffc); + radeon_ring_write(ring, (upper_32_bits(addr) & 0xffff) | DATA_SEL(1) | INT_SEL(2)); + radeon_ring_write(ring, fence->seq); + radeon_ring_write(ring, 0); + /* HDP flush */ + /* We should be using the new WAIT_REG_MEM special op packet here + * but it causes the CP to hang + */ + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, HDP_MEM_COHERENCY_FLUSH_CNTL >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, 0); +} + +/** + * cik_fence_compute_ring_emit - emit a fence on the compute ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Emits a fence sequnce number on the compute ring and flushes + * GPU caches. + */ +void cik_fence_compute_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + + /* RELEASE_MEM - flush caches, send int */ + radeon_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 5)); + radeon_ring_write(ring, (EOP_TCL1_ACTION_EN | + EOP_TC_ACTION_EN | + EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | + EVENT_INDEX(5))); + radeon_ring_write(ring, DATA_SEL(1) | INT_SEL(2)); + radeon_ring_write(ring, addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(addr)); + radeon_ring_write(ring, fence->seq); + radeon_ring_write(ring, 0); + /* HDP flush */ + /* We should be using the new WAIT_REG_MEM special op packet here + * but it causes the CP to hang + */ + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, HDP_MEM_COHERENCY_FLUSH_CNTL >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, 0); +} + +void cik_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + uint64_t addr = semaphore->gpu_addr; + unsigned sel = emit_wait ? PACKET3_SEM_SEL_WAIT : PACKET3_SEM_SEL_SIGNAL; + + radeon_ring_write(ring, PACKET3(PACKET3_MEM_SEMAPHORE, 1)); + radeon_ring_write(ring, addr & 0xffffffff); + radeon_ring_write(ring, (upper_32_bits(addr) & 0xffff) | sel); +} + +/* + * IB stuff + */ +/** + * cik_ring_ib_execute - emit an IB (Indirect Buffer) on the gfx ring + * + * @rdev: radeon_device pointer + * @ib: radeon indirect buffer object + * + * Emits an DE (drawing engine) or CE (constant engine) IB + * on the gfx ring. IBs are usually generated by userspace + * acceleration drivers and submitted to the kernel for + * sheduling on the ring. This function schedules the IB + * on the gfx ring for execution by the GPU. + */ +void cik_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + u32 header, control = INDIRECT_BUFFER_VALID; + + if (ib->is_const_ib) { + /* set switch buffer packet before const IB */ + radeon_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0)); + radeon_ring_write(ring, 0); + + header = PACKET3(PACKET3_INDIRECT_BUFFER_CONST, 2); + } else { + u32 next_rptr; + if (ring->rptr_save_reg) { + next_rptr = ring->wptr + 3 + 4; + radeon_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + radeon_ring_write(ring, ((ring->rptr_save_reg - + PACKET3_SET_UCONFIG_REG_START) >> 2)); + radeon_ring_write(ring, next_rptr); + } else if (rdev->wb.enabled) { + next_rptr = ring->wptr + 5 + 4; + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, WRITE_DATA_DST_SEL(1)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + radeon_ring_write(ring, next_rptr); + } + + header = PACKET3(PACKET3_INDIRECT_BUFFER, 2); + } + + control |= ib->length_dw | + (ib->vm ? (ib->vm->id << 24) : 0); + + radeon_ring_write(ring, header); + radeon_ring_write(ring, +#ifdef __BIG_ENDIAN + (2 << 0) | +#endif + (ib->gpu_addr & 0xFFFFFFFC)); + radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF); + radeon_ring_write(ring, control); +} + +/** + * cik_ib_test - basic gfx ring IB test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Allocate an IB and execute it on the gfx ring (CIK). + * Provides a basic gfx ring test to verify that IBs are working. + * Returns 0 on success, error on failure. + */ +int cik_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_ib ib; + uint32_t scratch; + uint32_t tmp = 0; + unsigned i; + int r; + + r = radeon_scratch_get(rdev, &scratch); + if (r) { + DRM_ERROR("radeon: failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); + ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START) >> 2); + ib.ptr[2] = 0xDEADBEEF; + ib.length_dw = 3; + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + radeon_scratch_free(rdev, scratch); + radeon_ib_free(rdev, &ib); + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + return r; + } + r = radeon_fence_wait(ib.fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + return r; + } + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); + } else { + DRM_ERROR("radeon: ib test failed (scratch(0x%04X)=0x%08X)\n", + scratch, tmp); + r = -EINVAL; + } + radeon_scratch_free(rdev, scratch); + radeon_ib_free(rdev, &ib); + return r; +} + +/* + * CP. + * On CIK, gfx and compute now have independant command processors. + * + * GFX + * Gfx consists of a single ring and can process both gfx jobs and + * compute jobs. The gfx CP consists of three microengines (ME): + * PFP - Pre-Fetch Parser + * ME - Micro Engine + * CE - Constant Engine + * The PFP and ME make up what is considered the Drawing Engine (DE). + * The CE is an asynchronous engine used for updating buffer desciptors + * used by the DE so that they can be loaded into cache in parallel + * while the DE is processing state update packets. + * + * Compute + * The compute CP consists of two microengines (ME): + * MEC1 - Compute MicroEngine 1 + * MEC2 - Compute MicroEngine 2 + * Each MEC supports 4 compute pipes and each pipe supports 8 queues. + * The queues are exposed to userspace and are programmed directly + * by the compute runtime. + */ +/** + * cik_cp_gfx_enable - enable/disable the gfx CP MEs + * + * @rdev: radeon_device pointer + * @enable: enable or disable the MEs + * + * Halts or unhalts the gfx MEs. + */ +static void cik_cp_gfx_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32(CP_ME_CNTL, 0); + else { + WREG32(CP_ME_CNTL, (CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT)); + rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false; + } + DRM_UDELAY(50); +} + +/** + * cik_cp_gfx_load_microcode - load the gfx CP ME ucode + * + * @rdev: radeon_device pointer + * + * Loads the gfx PFP, ME, and CE ucode. + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static int cik_cp_gfx_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + int i; + + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw) + return -EINVAL; + + cik_cp_gfx_enable(rdev, false); + + /* PFP */ + fw_data = (const __be32 *)rdev->pfp_fw->data; + WREG32(CP_PFP_UCODE_ADDR, 0); + for (i = 0; i < CIK_PFP_UCODE_SIZE; i++) + WREG32(CP_PFP_UCODE_DATA, be32_to_cpup(fw_data++)); + WREG32(CP_PFP_UCODE_ADDR, 0); + + /* CE */ + fw_data = (const __be32 *)rdev->ce_fw->data; + WREG32(CP_CE_UCODE_ADDR, 0); + for (i = 0; i < CIK_CE_UCODE_SIZE; i++) + WREG32(CP_CE_UCODE_DATA, be32_to_cpup(fw_data++)); + WREG32(CP_CE_UCODE_ADDR, 0); + + /* ME */ + fw_data = (const __be32 *)rdev->me_fw->data; + WREG32(CP_ME_RAM_WADDR, 0); + for (i = 0; i < CIK_ME_UCODE_SIZE; i++) + WREG32(CP_ME_RAM_DATA, be32_to_cpup(fw_data++)); + WREG32(CP_ME_RAM_WADDR, 0); + + WREG32(CP_PFP_UCODE_ADDR, 0); + WREG32(CP_CE_UCODE_ADDR, 0); + WREG32(CP_ME_RAM_WADDR, 0); + WREG32(CP_ME_RAM_RADDR, 0); + return 0; +} + +/** + * cik_cp_gfx_start - start the gfx ring + * + * @rdev: radeon_device pointer + * + * Enables the ring and loads the clear state context and other + * packets required to init the ring. + * Returns 0 for success, error for failure. + */ +static int cik_cp_gfx_start(struct radeon_device *rdev) +{ + struct radeon_ring *ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; + int r, i; + + /* init the CP */ + WREG32(CP_MAX_CONTEXT, rdev->config.cik.max_hw_contexts - 1); + WREG32(CP_ENDIAN_SWAP, 0); + WREG32(CP_DEVICE_ID, 1); + + cik_cp_gfx_enable(rdev, true); + + r = radeon_ring_lock(rdev, ring, cik_default_size + 17); + if (r) { + DRM_ERROR("radeon: cp failed to lock ring (%d).\n", r); + return r; + } + + /* init the CE partitions. CE only used for gfx on CIK */ + radeon_ring_write(ring, PACKET3(PACKET3_SET_BASE, 2)); + radeon_ring_write(ring, PACKET3_BASE_INDEX(CE_PARTITION_BASE)); + radeon_ring_write(ring, 0xc000); + radeon_ring_write(ring, 0xc000); + + /* setup clear context state */ + radeon_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + radeon_ring_write(ring, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE); + + radeon_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1)); + radeon_ring_write(ring, 0x80000000); + radeon_ring_write(ring, 0x80000000); + + for (i = 0; i < cik_default_size; i++) + radeon_ring_write(ring, cik_default_state[i]); + + radeon_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0)); + radeon_ring_write(ring, PACKET3_PREAMBLE_END_CLEAR_STATE); + + /* set clear context state */ + radeon_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0)); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2)); + radeon_ring_write(ring, 0x00000316); + radeon_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */ + radeon_ring_write(ring, 0x00000010); /* VGT_OUT_DEALLOC_CNTL */ + + radeon_ring_unlock_commit(rdev, ring); + + return 0; +} + +/** + * cik_cp_gfx_fini - stop the gfx ring + * + * @rdev: radeon_device pointer + * + * Stop the gfx ring and tear down the driver ring + * info. + */ +static void cik_cp_gfx_fini(struct radeon_device *rdev) +{ + cik_cp_gfx_enable(rdev, false); + radeon_ring_fini(rdev, &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]); +} + +/** + * cik_cp_gfx_resume - setup the gfx ring buffer registers + * + * @rdev: radeon_device pointer + * + * Program the location and size of the gfx ring buffer + * and test it to make sure it's working. + * Returns 0 for success, error for failure. + */ +static int cik_cp_gfx_resume(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + u32 tmp; + u32 rb_bufsz; + u64 rb_addr; + int r; + + WREG32(CP_SEM_WAIT_TIMER, 0x0); + WREG32(CP_SEM_INCOMPLETE_TIMER_CNTL, 0x0); + + /* Set the write pointer delay */ + WREG32(CP_RB_WPTR_DELAY, 0); + + /* set the RB to use vmid 0 */ + WREG32(CP_RB_VMID, 0); + + WREG32(SCRATCH_ADDR, ((rdev->wb.gpu_addr + RADEON_WB_SCRATCH_OFFSET) >> 8) & 0xFFFFFFFF); + + /* ring 0 - compute and gfx */ + /* Set ring buffer size */ + ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; + rb_bufsz = drm_order(ring->ring_size / 8); + tmp = (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8) | rb_bufsz; +#ifdef __BIG_ENDIAN + tmp |= BUF_SWAP_32BIT; +#endif + WREG32(CP_RB0_CNTL, tmp); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(CP_RB0_CNTL, tmp | RB_RPTR_WR_ENA); + ring->wptr = 0; + WREG32(CP_RB0_WPTR, ring->wptr); + + /* set the wb address wether it's enabled or not */ + WREG32(CP_RB0_RPTR_ADDR, (rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFFFFFFFC); + WREG32(CP_RB0_RPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + RADEON_WB_CP_RPTR_OFFSET) & 0xFF); + + /* scratch register shadowing is no longer supported */ + WREG32(SCRATCH_UMSK, 0); + + if (!rdev->wb.enabled) + tmp |= RB_NO_UPDATE; + + DRM_MDELAY(1); + WREG32(CP_RB0_CNTL, tmp); + + rb_addr = ring->gpu_addr >> 8; + WREG32(CP_RB0_BASE, rb_addr); + WREG32(CP_RB0_BASE_HI, upper_32_bits(rb_addr)); + + ring->rptr = RREG32(CP_RB0_RPTR); + + /* start the ring */ + cik_cp_gfx_start(rdev); + rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = true; + r = radeon_ring_test(rdev, RADEON_RING_TYPE_GFX_INDEX, &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]); + if (r) { + rdev->ring[RADEON_RING_TYPE_GFX_INDEX].ready = false; + return r; + } + return 0; +} + +u32 cik_compute_ring_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + u32 rptr; + + + + if (rdev->wb.enabled) { + rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]); + } else { + spin_lock(&rdev->srbm_mutex); + cik_srbm_select(rdev, ring->me, ring->pipe, ring->queue, 0); + rptr = RREG32(CP_HQD_PQ_RPTR); + cik_srbm_select(rdev, 0, 0, 0, 0); + spin_unlock(&rdev->srbm_mutex); + } + rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + + return rptr; +} + +u32 cik_compute_ring_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + u32 wptr; + + if (rdev->wb.enabled) { + wptr = le32_to_cpu(rdev->wb.wb[ring->wptr_offs/4]); + } else { + spin_lock(&rdev->srbm_mutex); + cik_srbm_select(rdev, ring->me, ring->pipe, ring->queue, 0); + wptr = RREG32(CP_HQD_PQ_WPTR); + cik_srbm_select(rdev, 0, 0, 0, 0); + spin_unlock(&rdev->srbm_mutex); + } + wptr = (wptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + + return wptr; +} + +void cik_compute_ring_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + u32 wptr = (ring->wptr << ring->ptr_reg_shift) & ring->ptr_reg_mask; + + rdev->wb.wb[ring->wptr_offs/4] = cpu_to_le32(wptr); + WDOORBELL32(ring->doorbell_offset, wptr); +} + +/** + * cik_cp_compute_enable - enable/disable the compute CP MEs + * + * @rdev: radeon_device pointer + * @enable: enable or disable the MEs + * + * Halts or unhalts the compute MEs. + */ +static void cik_cp_compute_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32(CP_MEC_CNTL, 0); + else + WREG32(CP_MEC_CNTL, (MEC_ME1_HALT | MEC_ME2_HALT)); + DRM_UDELAY(50); +} + +/** + * cik_cp_compute_load_microcode - load the compute CP ME ucode + * + * @rdev: radeon_device pointer + * + * Loads the compute MEC1&2 ucode. + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static int cik_cp_compute_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + int i; + + if (!rdev->mec_fw) + return -EINVAL; + + cik_cp_compute_enable(rdev, false); + + /* MEC1 */ + fw_data = (const __be32 *)rdev->mec_fw->data; + WREG32(CP_MEC_ME1_UCODE_ADDR, 0); + for (i = 0; i < CIK_MEC_UCODE_SIZE; i++) + WREG32(CP_MEC_ME1_UCODE_DATA, be32_to_cpup(fw_data++)); + WREG32(CP_MEC_ME1_UCODE_ADDR, 0); + + if (rdev->family == CHIP_KAVERI) { + /* MEC2 */ + fw_data = (const __be32 *)rdev->mec_fw->data; + WREG32(CP_MEC_ME2_UCODE_ADDR, 0); + for (i = 0; i < CIK_MEC_UCODE_SIZE; i++) + WREG32(CP_MEC_ME2_UCODE_DATA, be32_to_cpup(fw_data++)); + WREG32(CP_MEC_ME2_UCODE_ADDR, 0); + } + + return 0; +} + +/** + * cik_cp_compute_start - start the compute queues + * + * @rdev: radeon_device pointer + * + * Enable the compute queues. + * Returns 0 for success, error for failure. + */ +static int cik_cp_compute_start(struct radeon_device *rdev) +{ + cik_cp_compute_enable(rdev, true); + + return 0; +} + +/** + * cik_cp_compute_fini - stop the compute queues + * + * @rdev: radeon_device pointer + * + * Stop the compute queues and tear down the driver queue + * info. + */ +static void cik_cp_compute_fini(struct radeon_device *rdev) +{ + int i, idx, r; + + cik_cp_compute_enable(rdev, false); + + for (i = 0; i < 2; i++) { + if (i == 0) + idx = CAYMAN_RING_TYPE_CP1_INDEX; + else + idx = CAYMAN_RING_TYPE_CP2_INDEX; + + if (rdev->ring[idx].mqd_obj) { + r = radeon_bo_reserve(rdev->ring[idx].mqd_obj, false); + if (unlikely(r != 0)) + dev_warn(rdev->dev, "(%d) reserve MQD bo failed\n", r); + + radeon_bo_unpin(rdev->ring[idx].mqd_obj); + radeon_bo_unreserve(rdev->ring[idx].mqd_obj); + + radeon_bo_unref(&rdev->ring[idx].mqd_obj); + rdev->ring[idx].mqd_obj = NULL; + } + } +} + +static void cik_mec_fini(struct radeon_device *rdev) +{ + int r; + + if (rdev->mec.hpd_eop_obj) { + r = radeon_bo_reserve(rdev->mec.hpd_eop_obj, false); + if (unlikely(r != 0)) + dev_warn(rdev->dev, "(%d) reserve HPD EOP bo failed\n", r); + radeon_bo_unpin(rdev->mec.hpd_eop_obj); + radeon_bo_unreserve(rdev->mec.hpd_eop_obj); + + radeon_bo_unref(&rdev->mec.hpd_eop_obj); + rdev->mec.hpd_eop_obj = NULL; + } +} + +#define MEC_HPD_SIZE 2048 + +static int cik_mec_init(struct radeon_device *rdev) +{ + int r; + u32 *hpd; + + /* + * KV: 2 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 64 Queues total + * CI/KB: 1 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 32 Queues total + */ + if (rdev->family == CHIP_KAVERI) + rdev->mec.num_mec = 2; + else + rdev->mec.num_mec = 1; + rdev->mec.num_pipe = 4; + rdev->mec.num_queue = rdev->mec.num_mec * rdev->mec.num_pipe * 8; + + if (rdev->mec.hpd_eop_obj == NULL) { + r = radeon_bo_create(rdev, + rdev->mec.num_mec *rdev->mec.num_pipe * MEC_HPD_SIZE * 2, + PAGE_SIZE, true, + RADEON_GEM_DOMAIN_GTT, NULL, + &rdev->mec.hpd_eop_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create HDP EOP bo failed\n", r); + return r; + } + } + + r = radeon_bo_reserve(rdev->mec.hpd_eop_obj, false); + if (unlikely(r != 0)) { + cik_mec_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->mec.hpd_eop_obj, RADEON_GEM_DOMAIN_GTT, + &rdev->mec.hpd_eop_gpu_addr); + if (r) { + dev_warn(rdev->dev, "(%d) pin HDP EOP bo failed\n", r); + cik_mec_fini(rdev); + return r; + } + r = radeon_bo_kmap(rdev->mec.hpd_eop_obj, (void **)&hpd); + if (r) { + dev_warn(rdev->dev, "(%d) map HDP EOP bo failed\n", r); + cik_mec_fini(rdev); + return r; + } + + /* clear memory. Not sure if this is required or not */ + memset(hpd, 0, rdev->mec.num_mec *rdev->mec.num_pipe * MEC_HPD_SIZE * 2); + + radeon_bo_kunmap(rdev->mec.hpd_eop_obj); + radeon_bo_unreserve(rdev->mec.hpd_eop_obj); + + return 0; +} + +struct hqd_registers +{ + u32 cp_mqd_base_addr; + u32 cp_mqd_base_addr_hi; + u32 cp_hqd_active; + u32 cp_hqd_vmid; + u32 cp_hqd_persistent_state; + u32 cp_hqd_pipe_priority; + u32 cp_hqd_queue_priority; + u32 cp_hqd_quantum; + u32 cp_hqd_pq_base; + u32 cp_hqd_pq_base_hi; + u32 cp_hqd_pq_rptr; + u32 cp_hqd_pq_rptr_report_addr; + u32 cp_hqd_pq_rptr_report_addr_hi; + u32 cp_hqd_pq_wptr_poll_addr; + u32 cp_hqd_pq_wptr_poll_addr_hi; + u32 cp_hqd_pq_doorbell_control; + u32 cp_hqd_pq_wptr; + u32 cp_hqd_pq_control; + u32 cp_hqd_ib_base_addr; + u32 cp_hqd_ib_base_addr_hi; + u32 cp_hqd_ib_rptr; + u32 cp_hqd_ib_control; + u32 cp_hqd_iq_timer; + u32 cp_hqd_iq_rptr; + u32 cp_hqd_dequeue_request; + u32 cp_hqd_dma_offload; + u32 cp_hqd_sema_cmd; + u32 cp_hqd_msg_type; + u32 cp_hqd_atomic0_preop_lo; + u32 cp_hqd_atomic0_preop_hi; + u32 cp_hqd_atomic1_preop_lo; + u32 cp_hqd_atomic1_preop_hi; + u32 cp_hqd_hq_scheduler0; + u32 cp_hqd_hq_scheduler1; + u32 cp_mqd_control; +}; + +struct bonaire_mqd +{ + u32 header; + u32 dispatch_initiator; + u32 dimensions[3]; + u32 start_idx[3]; + u32 num_threads[3]; + u32 pipeline_stat_enable; + u32 perf_counter_enable; + u32 pgm[2]; + u32 tba[2]; + u32 tma[2]; + u32 pgm_rsrc[2]; + u32 vmid; + u32 resource_limits; + u32 static_thread_mgmt01[2]; + u32 tmp_ring_size; + u32 static_thread_mgmt23[2]; + u32 restart[3]; + u32 thread_trace_enable; + u32 reserved1; + u32 user_data[16]; + u32 vgtcs_invoke_count[2]; + struct hqd_registers queue_state; + u32 dequeue_cntr; + u32 interrupt_queue[64]; +}; + +/** + * cik_cp_compute_resume - setup the compute queue registers + * + * @rdev: radeon_device pointer + * + * Program the compute queues and test them to make sure they + * are working. + * Returns 0 for success, error for failure. + */ +static int cik_cp_compute_resume(struct radeon_device *rdev) +{ + int r, i, idx; + u32 tmp; + bool use_doorbell = true; + u64 hqd_gpu_addr; + u64 mqd_gpu_addr; + u64 eop_gpu_addr; + u64 wb_gpu_addr; + u32 *buf; + struct bonaire_mqd *mqd; + + r = cik_cp_compute_start(rdev); + if (r) + return r; + + /* fix up chicken bits */ + tmp = RREG32(CP_CPF_DEBUG); + tmp |= (1 << 23); + WREG32(CP_CPF_DEBUG, tmp); + + /* init the pipes */ + spin_lock(&rdev->srbm_mutex); + for (i = 0; i < (rdev->mec.num_pipe * rdev->mec.num_mec); i++) { + int me = (i < 4) ? 1 : 2; + int pipe = (i < 4) ? i : (i - 4); + + eop_gpu_addr = rdev->mec.hpd_eop_gpu_addr + (i * MEC_HPD_SIZE * 2); + + cik_srbm_select(rdev, me, pipe, 0, 0); + + /* write the EOP addr */ + WREG32(CP_HPD_EOP_BASE_ADDR, eop_gpu_addr >> 8); + WREG32(CP_HPD_EOP_BASE_ADDR_HI, upper_32_bits(eop_gpu_addr) >> 8); + + /* set the VMID assigned */ + WREG32(CP_HPD_EOP_VMID, 0); + + /* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */ + tmp = RREG32(CP_HPD_EOP_CONTROL); + tmp &= ~EOP_SIZE_MASK; + tmp |= drm_order(MEC_HPD_SIZE / 8); + WREG32(CP_HPD_EOP_CONTROL, tmp); + } + cik_srbm_select(rdev, 0, 0, 0, 0); + spin_unlock(&rdev->srbm_mutex); + + /* init the queues. Just two for now. */ + for (i = 0; i < 2; i++) { + if (i == 0) + idx = CAYMAN_RING_TYPE_CP1_INDEX; + else + idx = CAYMAN_RING_TYPE_CP2_INDEX; + + if (rdev->ring[idx].mqd_obj == NULL) { + r = radeon_bo_create(rdev, + sizeof(struct bonaire_mqd), + PAGE_SIZE, true, + RADEON_GEM_DOMAIN_GTT, NULL, + &rdev->ring[idx].mqd_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create MQD bo failed\n", r); + return r; + } + } + + r = radeon_bo_reserve(rdev->ring[idx].mqd_obj, false); + if (unlikely(r != 0)) { + cik_cp_compute_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->ring[idx].mqd_obj, RADEON_GEM_DOMAIN_GTT, + &mqd_gpu_addr); + if (r) { + dev_warn(rdev->dev, "(%d) pin MQD bo failed\n", r); + cik_cp_compute_fini(rdev); + return r; + } + r = radeon_bo_kmap(rdev->ring[idx].mqd_obj, (void **)&buf); + if (r) { + dev_warn(rdev->dev, "(%d) map MQD bo failed\n", r); + cik_cp_compute_fini(rdev); + return r; + } + + /* doorbell offset */ + rdev->ring[idx].doorbell_offset = + (rdev->ring[idx].doorbell_page_num * PAGE_SIZE) + 0; + + /* init the mqd struct */ + memset(buf, 0, sizeof(struct bonaire_mqd)); + + mqd = (struct bonaire_mqd *)buf; + mqd->header = 0xC0310800; + mqd->static_thread_mgmt01[0] = 0xffffffff; + mqd->static_thread_mgmt01[1] = 0xffffffff; + mqd->static_thread_mgmt23[0] = 0xffffffff; + mqd->static_thread_mgmt23[1] = 0xffffffff; + + spin_lock(&rdev->srbm_mutex); + cik_srbm_select(rdev, rdev->ring[idx].me, + rdev->ring[idx].pipe, + rdev->ring[idx].queue, 0); + + /* disable wptr polling */ + tmp = RREG32(CP_PQ_WPTR_POLL_CNTL); + tmp &= ~WPTR_POLL_EN; + WREG32(CP_PQ_WPTR_POLL_CNTL, tmp); + + /* enable doorbell? */ + mqd->queue_state.cp_hqd_pq_doorbell_control = + RREG32(CP_HQD_PQ_DOORBELL_CONTROL); + if (use_doorbell) + mqd->queue_state.cp_hqd_pq_doorbell_control |= DOORBELL_EN; + else + mqd->queue_state.cp_hqd_pq_doorbell_control &= ~DOORBELL_EN; + WREG32(CP_HQD_PQ_DOORBELL_CONTROL, + mqd->queue_state.cp_hqd_pq_doorbell_control); + + /* disable the queue if it's active */ + mqd->queue_state.cp_hqd_dequeue_request = 0; + mqd->queue_state.cp_hqd_pq_rptr = 0; + mqd->queue_state.cp_hqd_pq_wptr= 0; + if (RREG32(CP_HQD_ACTIVE) & 1) { + WREG32(CP_HQD_DEQUEUE_REQUEST, 1); + for (i = 0; i < rdev->usec_timeout; i++) { + if (!(RREG32(CP_HQD_ACTIVE) & 1)) + break; + DRM_UDELAY(1); + } + WREG32(CP_HQD_DEQUEUE_REQUEST, mqd->queue_state.cp_hqd_dequeue_request); + WREG32(CP_HQD_PQ_RPTR, mqd->queue_state.cp_hqd_pq_rptr); + WREG32(CP_HQD_PQ_WPTR, mqd->queue_state.cp_hqd_pq_wptr); + } + + /* set the pointer to the MQD */ + mqd->queue_state.cp_mqd_base_addr = mqd_gpu_addr & 0xfffffffc; + mqd->queue_state.cp_mqd_base_addr_hi = upper_32_bits(mqd_gpu_addr); + WREG32(CP_MQD_BASE_ADDR, mqd->queue_state.cp_mqd_base_addr); + WREG32(CP_MQD_BASE_ADDR_HI, mqd->queue_state.cp_mqd_base_addr_hi); + /* set MQD vmid to 0 */ + mqd->queue_state.cp_mqd_control = RREG32(CP_MQD_CONTROL); + mqd->queue_state.cp_mqd_control &= ~MQD_VMID_MASK; + WREG32(CP_MQD_CONTROL, mqd->queue_state.cp_mqd_control); + + /* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */ + hqd_gpu_addr = rdev->ring[idx].gpu_addr >> 8; + mqd->queue_state.cp_hqd_pq_base = hqd_gpu_addr; + mqd->queue_state.cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr); + WREG32(CP_HQD_PQ_BASE, mqd->queue_state.cp_hqd_pq_base); + WREG32(CP_HQD_PQ_BASE_HI, mqd->queue_state.cp_hqd_pq_base_hi); + + /* set up the HQD, this is similar to CP_RB0_CNTL */ + mqd->queue_state.cp_hqd_pq_control = RREG32(CP_HQD_PQ_CONTROL); + mqd->queue_state.cp_hqd_pq_control &= + ~(QUEUE_SIZE_MASK | RPTR_BLOCK_SIZE_MASK); + + mqd->queue_state.cp_hqd_pq_control |= + drm_order(rdev->ring[idx].ring_size / 8); + mqd->queue_state.cp_hqd_pq_control |= + (drm_order(RADEON_GPU_PAGE_SIZE/8) << 8); +#ifdef __BIG_ENDIAN + mqd->queue_state.cp_hqd_pq_control |= BUF_SWAP_32BIT; +#endif + mqd->queue_state.cp_hqd_pq_control &= + ~(UNORD_DISPATCH | ROQ_PQ_IB_FLIP | PQ_VOLATILE); + mqd->queue_state.cp_hqd_pq_control |= + PRIV_STATE | KMD_QUEUE; /* assuming kernel queue control */ + WREG32(CP_HQD_PQ_CONTROL, mqd->queue_state.cp_hqd_pq_control); + + /* only used if CP_PQ_WPTR_POLL_CNTL.WPTR_POLL_EN=1 */ + if (i == 0) + wb_gpu_addr = rdev->wb.gpu_addr + CIK_WB_CP1_WPTR_OFFSET; + else + wb_gpu_addr = rdev->wb.gpu_addr + CIK_WB_CP2_WPTR_OFFSET; + mqd->queue_state.cp_hqd_pq_wptr_poll_addr = wb_gpu_addr & 0xfffffffc; + mqd->queue_state.cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff; + WREG32(CP_HQD_PQ_WPTR_POLL_ADDR, mqd->queue_state.cp_hqd_pq_wptr_poll_addr); + WREG32(CP_HQD_PQ_WPTR_POLL_ADDR_HI, + mqd->queue_state.cp_hqd_pq_wptr_poll_addr_hi); + + /* set the wb address wether it's enabled or not */ + if (i == 0) + wb_gpu_addr = rdev->wb.gpu_addr + RADEON_WB_CP1_RPTR_OFFSET; + else + wb_gpu_addr = rdev->wb.gpu_addr + RADEON_WB_CP2_RPTR_OFFSET; + mqd->queue_state.cp_hqd_pq_rptr_report_addr = wb_gpu_addr & 0xfffffffc; + mqd->queue_state.cp_hqd_pq_rptr_report_addr_hi = + upper_32_bits(wb_gpu_addr) & 0xffff; + WREG32(CP_HQD_PQ_RPTR_REPORT_ADDR, + mqd->queue_state.cp_hqd_pq_rptr_report_addr); + WREG32(CP_HQD_PQ_RPTR_REPORT_ADDR_HI, + mqd->queue_state.cp_hqd_pq_rptr_report_addr_hi); + + /* enable the doorbell if requested */ + if (use_doorbell) { + mqd->queue_state.cp_hqd_pq_doorbell_control = + RREG32(CP_HQD_PQ_DOORBELL_CONTROL); + mqd->queue_state.cp_hqd_pq_doorbell_control &= ~DOORBELL_OFFSET_MASK; + mqd->queue_state.cp_hqd_pq_doorbell_control |= + DOORBELL_OFFSET(rdev->ring[idx].doorbell_offset / 4); + mqd->queue_state.cp_hqd_pq_doorbell_control |= DOORBELL_EN; + mqd->queue_state.cp_hqd_pq_doorbell_control &= + ~(DOORBELL_SOURCE | DOORBELL_HIT); + + } else { + mqd->queue_state.cp_hqd_pq_doorbell_control = 0; + } + WREG32(CP_HQD_PQ_DOORBELL_CONTROL, + mqd->queue_state.cp_hqd_pq_doorbell_control); + + /* read and write pointers, similar to CP_RB0_WPTR/_RPTR */ + rdev->ring[idx].wptr = 0; + mqd->queue_state.cp_hqd_pq_wptr = rdev->ring[idx].wptr; + WREG32(CP_HQD_PQ_WPTR, mqd->queue_state.cp_hqd_pq_wptr); + rdev->ring[idx].rptr = RREG32(CP_HQD_PQ_RPTR); + mqd->queue_state.cp_hqd_pq_rptr = rdev->ring[idx].rptr; + + /* set the vmid for the queue */ + mqd->queue_state.cp_hqd_vmid = 0; + WREG32(CP_HQD_VMID, mqd->queue_state.cp_hqd_vmid); + + /* activate the queue */ + mqd->queue_state.cp_hqd_active = 1; + WREG32(CP_HQD_ACTIVE, mqd->queue_state.cp_hqd_active); + + cik_srbm_select(rdev, 0, 0, 0, 0); + spin_unlock(&rdev->srbm_mutex); + + radeon_bo_kunmap(rdev->ring[idx].mqd_obj); + radeon_bo_unreserve(rdev->ring[idx].mqd_obj); + + rdev->ring[idx].ready = true; + r = radeon_ring_test(rdev, idx, &rdev->ring[idx]); + if (r) + rdev->ring[idx].ready = false; + } + + return 0; +} + +static void cik_cp_enable(struct radeon_device *rdev, bool enable) +{ + cik_cp_gfx_enable(rdev, enable); + cik_cp_compute_enable(rdev, enable); +} + +static int cik_cp_load_microcode(struct radeon_device *rdev) +{ + int r; + + r = cik_cp_gfx_load_microcode(rdev); + if (r) + return r; + r = cik_cp_compute_load_microcode(rdev); + if (r) + return r; + + return 0; +} + +static void cik_cp_fini(struct radeon_device *rdev) +{ + cik_cp_gfx_fini(rdev); + cik_cp_compute_fini(rdev); +} + +static int cik_cp_resume(struct radeon_device *rdev) +{ + int r; + + /* Reset all cp blocks */ + WREG32(GRBM_SOFT_RESET, SOFT_RESET_CP); + RREG32(GRBM_SOFT_RESET); + DRM_MDELAY(15); + WREG32(GRBM_SOFT_RESET, 0); + RREG32(GRBM_SOFT_RESET); + + r = cik_cp_load_microcode(rdev); + if (r) + return r; + + r = cik_cp_gfx_resume(rdev); + if (r) + return r; + r = cik_cp_compute_resume(rdev); + if (r) + return r; + + return 0; +} + +/* + * sDMA - System DMA + * Starting with CIK, the GPU has new asynchronous + * DMA engines. These engines are used for compute + * and gfx. There are two DMA engines (SDMA0, SDMA1) + * and each one supports 1 ring buffer used for gfx + * and 2 queues used for compute. + * + * The programming model is very similar to the CP + * (ring buffer, IBs, etc.), but sDMA has it's own + * packet format that is different from the PM4 format + * used by the CP. sDMA supports copying data, writing + * embedded data, solid fills, and a number of other + * things. It also has support for tiling/detiling of + * buffers. + */ +/** + * cik_sdma_ring_ib_execute - Schedule an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ib: IB object to schedule + * + * Schedule an IB in the DMA ring (CIK). + */ +void cik_sdma_ring_ib_execute(struct radeon_device *rdev, + struct radeon_ib *ib) +{ + struct radeon_ring *ring = &rdev->ring[ib->ring]; + u32 extra_bits = (ib->vm ? ib->vm->id : 0) & 0xf; + + if (rdev->wb.enabled) { + u32 next_rptr = ring->wptr + 5; + while ((next_rptr & 7) != 4) + next_rptr++; + next_rptr += 4; + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff); + radeon_ring_write(ring, 1); /* number of DWs to follow */ + radeon_ring_write(ring, next_rptr); + } + + /* IB packet must end on a 8 DW boundary */ + while ((ring->wptr & 7) != 4) + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits)); + radeon_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */ + radeon_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xffffffff); + radeon_ring_write(ring, ib->length_dw); + +} + +/** + * cik_sdma_fence_ring_emit - emit a fence on the DMA ring + * + * @rdev: radeon_device pointer + * @fence: radeon fence object + * + * Add a DMA fence packet to the ring to write + * the fence seq number and DMA trap packet to generate + * an interrupt if needed (CIK). + */ +void cik_sdma_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence) +{ + struct radeon_ring *ring = &rdev->ring[fence->ring]; + u64 addr = rdev->fence_drv[fence->ring].gpu_addr; + u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | + SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ + u32 ref_and_mask; + + if (fence->ring == R600_RING_TYPE_DMA_INDEX) + ref_and_mask = SDMA0; + else + ref_and_mask = SDMA1; + + /* write the fence */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_FENCE, 0, 0)); + radeon_ring_write(ring, addr & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); + radeon_ring_write(ring, fence->seq); + /* generate an interrupt */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_TRAP, 0, 0)); + /* flush HDP */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); + radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); + radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); + radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ + radeon_ring_write(ring, ref_and_mask); /* MASK */ + radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ +} + +/** + * cik_sdma_semaphore_ring_emit - emit a semaphore on the dma ring + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * @semaphore: radeon semaphore object + * @emit_wait: wait or signal semaphore + * + * Add a DMA semaphore packet to the ring wait on or signal + * other rings (CIK). + */ +void cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait) +{ + u64 addr = semaphore->gpu_addr; + u32 extra_bits = emit_wait ? 0 : SDMA_SEMAPHORE_EXTRA_S; + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SEMAPHORE, 0, extra_bits)); + radeon_ring_write(ring, addr & 0xfffffff8); + radeon_ring_write(ring, upper_32_bits(addr) & 0xffffffff); +} + +/** + * cik_sdma_gfx_stop - stop the gfx async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the gfx async dma ring buffers (CIK). + */ +static void cik_sdma_gfx_stop(struct radeon_device *rdev) +{ + u32 rb_cntl, reg_offset; + int i; + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.visible_vram_size); + + for (i = 0; i < 2; i++) { + if (i == 0) + reg_offset = SDMA0_REGISTER_OFFSET; + else + reg_offset = SDMA1_REGISTER_OFFSET; + rb_cntl = RREG32(SDMA0_GFX_RB_CNTL + reg_offset); + rb_cntl &= ~SDMA_RB_ENABLE; + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); + WREG32(SDMA0_GFX_IB_CNTL + reg_offset, 0); + } +} + +/** + * cik_sdma_rlc_stop - stop the compute async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the compute async dma queues (CIK). + */ +static void cik_sdma_rlc_stop(struct radeon_device *rdev) +{ + /* XXX todo */ +} + +/** + * cik_sdma_enable - stop the async dma engines + * + * @rdev: radeon_device pointer + * @enable: enable/disable the DMA MEs. + * + * Halt or unhalt the async dma engines (CIK). + */ +static void cik_sdma_enable(struct radeon_device *rdev, bool enable) +{ + u32 me_cntl, reg_offset; + int i; + + for (i = 0; i < 2; i++) { + if (i == 0) + reg_offset = SDMA0_REGISTER_OFFSET; + else + reg_offset = SDMA1_REGISTER_OFFSET; + me_cntl = RREG32(SDMA0_ME_CNTL + reg_offset); + if (enable) + me_cntl &= ~SDMA_HALT; + else + me_cntl |= SDMA_HALT; + WREG32(SDMA0_ME_CNTL + reg_offset, me_cntl); + } +} + +/** + * cik_sdma_gfx_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the gfx DMA ring buffers and enable them (CIK). + * Returns 0 for success, error for failure. + */ +static int cik_sdma_gfx_resume(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + u32 rb_cntl, ib_cntl; + u32 rb_bufsz; + u32 reg_offset, wb_offset; + int i, r; + + for (i = 0; i < 2; i++) { + if (i == 0) { + ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + reg_offset = SDMA0_REGISTER_OFFSET; + wb_offset = R600_WB_DMA_RPTR_OFFSET; + } else { + ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; + reg_offset = SDMA1_REGISTER_OFFSET; + wb_offset = CAYMAN_WB_DMA1_RPTR_OFFSET; + } + + WREG32(SDMA0_SEM_INCOMPLETE_TIMER_CNTL + reg_offset, 0); + WREG32(SDMA0_SEM_WAIT_FAIL_TIMER_CNTL + reg_offset, 0); + + /* Set ring buffer size in dwords */ + rb_bufsz = drm_order(ring->ring_size / 4); + rb_cntl = rb_bufsz << 1; +#ifdef __BIG_ENDIAN + rb_cntl |= SDMA_RB_SWAP_ENABLE | SDMA_RPTR_WRITEBACK_SWAP_ENABLE; +#endif + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl); + + /* Initialize the ring buffer's read and write pointers */ + WREG32(SDMA0_GFX_RB_RPTR + reg_offset, 0); + WREG32(SDMA0_GFX_RB_WPTR + reg_offset, 0); + + /* set the wb address whether it's enabled or not */ + WREG32(SDMA0_GFX_RB_RPTR_ADDR_HI + reg_offset, + upper_32_bits(rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF); + WREG32(SDMA0_GFX_RB_RPTR_ADDR_LO + reg_offset, + ((rdev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC)); + + if (rdev->wb.enabled) + rb_cntl |= SDMA_RPTR_WRITEBACK_ENABLE; + + WREG32(SDMA0_GFX_RB_BASE + reg_offset, ring->gpu_addr >> 8); + WREG32(SDMA0_GFX_RB_BASE_HI + reg_offset, ring->gpu_addr >> 40); + + ring->wptr = 0; + WREG32(SDMA0_GFX_RB_WPTR + reg_offset, ring->wptr << 2); + + ring->rptr = RREG32(SDMA0_GFX_RB_RPTR + reg_offset) >> 2; + + /* enable DMA RB */ + WREG32(SDMA0_GFX_RB_CNTL + reg_offset, rb_cntl | SDMA_RB_ENABLE); + + ib_cntl = SDMA_IB_ENABLE; +#ifdef __BIG_ENDIAN + ib_cntl |= SDMA_IB_SWAP_ENABLE; +#endif + /* enable DMA IBs */ + WREG32(SDMA0_GFX_IB_CNTL + reg_offset, ib_cntl); + + ring->ready = true; + + r = radeon_ring_test(rdev, ring->idx, ring); + if (r) { + ring->ready = false; + return r; + } + } + + radeon_ttm_set_active_vram_size(rdev, rdev->mc.real_vram_size); + + return 0; +} + +/** + * cik_sdma_rlc_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the compute DMA queues and enable them (CIK). + * Returns 0 for success, error for failure. + */ +static int cik_sdma_rlc_resume(struct radeon_device *rdev) +{ + /* XXX todo */ + return 0; +} + +/** + * cik_sdma_load_microcode - load the sDMA ME ucode + * + * @rdev: radeon_device pointer + * + * Loads the sDMA0/1 ucode. + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static int cik_sdma_load_microcode(struct radeon_device *rdev) +{ + const __be32 *fw_data; + int i; + + if (!rdev->sdma_fw) + return -EINVAL; + + /* stop the gfx rings and rlc compute queues */ + cik_sdma_gfx_stop(rdev); + cik_sdma_rlc_stop(rdev); + + /* halt the MEs */ + cik_sdma_enable(rdev, false); + + /* sdma0 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA0_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + /* sdma1 */ + fw_data = (const __be32 *)rdev->sdma_fw->data; + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + for (i = 0; i < CIK_SDMA_UCODE_SIZE; i++) + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, be32_to_cpup(fw_data++)); + WREG32(SDMA0_UCODE_DATA + SDMA1_REGISTER_OFFSET, CIK_SDMA_UCODE_VERSION); + + WREG32(SDMA0_UCODE_ADDR + SDMA0_REGISTER_OFFSET, 0); + WREG32(SDMA0_UCODE_ADDR + SDMA1_REGISTER_OFFSET, 0); + return 0; +} + +/** + * cik_sdma_resume - setup and start the async dma engines + * + * @rdev: radeon_device pointer + * + * Set up the DMA engines and enable them (CIK). + * Returns 0 for success, error for failure. + */ +static __unused int cik_sdma_resume(struct radeon_device *rdev) +{ + int r; + + /* Reset dma */ + WREG32(SRBM_SOFT_RESET, SOFT_RESET_SDMA | SOFT_RESET_SDMA1); + RREG32(SRBM_SOFT_RESET); + DRM_UDELAY(50); + WREG32(SRBM_SOFT_RESET, 0); + RREG32(SRBM_SOFT_RESET); + + r = cik_sdma_load_microcode(rdev); + if (r) + return r; + + /* unhalt the MEs */ + cik_sdma_enable(rdev, true); + + /* start the gfx rings and rlc compute queues */ + r = cik_sdma_gfx_resume(rdev); + if (r) + return r; + r = cik_sdma_rlc_resume(rdev); + if (r) + return r; + + return 0; +} + +/** + * cik_sdma_fini - tear down the async dma engines + * + * @rdev: radeon_device pointer + * + * Stop the async dma engines and free the rings (CIK). + */ +static __unused void cik_sdma_fini(struct radeon_device *rdev) +{ + /* stop the gfx rings and rlc compute queues */ + cik_sdma_gfx_stop(rdev); + cik_sdma_rlc_stop(rdev); + /* halt the MEs */ + cik_sdma_enable(rdev, false); + radeon_ring_fini(rdev, &rdev->ring[R600_RING_TYPE_DMA_INDEX]); + radeon_ring_fini(rdev, &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]); + /* XXX - compute dma queue tear down */ +} + +/** + * cik_copy_dma - copy pages using the DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the DMA engine (CIK). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int cik_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.dma_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_bytes, cur_size_in_bytes; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); + num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); + r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + for (i = 0; i < num_loops; i++) { + cur_size_in_bytes = size_in_bytes; + if (cur_size_in_bytes > 0x1fffff) + cur_size_in_bytes = 0x1fffff; + size_in_bytes -= cur_size_in_bytes; + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, cur_size_in_bytes); + radeon_ring_write(ring, 0); /* src/dst endian swap */ + radeon_ring_write(ring, src_offset & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(src_offset) & 0xffffffff); + radeon_ring_write(ring, dst_offset & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xffffffff); + src_offset += cur_size_in_bytes; + dst_offset += cur_size_in_bytes; + } + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} + +/** + * cik_sdma_ring_test - simple async dma engine test + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test the DMA engine by writing using it to write an + * value to memory. (CIK). + * Returns 0 for success, error for failure. + */ +int cik_sdma_ring_test(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + unsigned i; + int r; + volatile void __iomem *ptr = (volatile void *)rdev->vram_scratch.ptr; + u32 tmp; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ring_lock(rdev, ring, 4); + if (r) { + DRM_ERROR("radeon: dma failed to lock ring %d (%d).\n", ring->idx, r); + return r; + } + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0)); + radeon_ring_write(ring, rdev->vram_scratch.gpu_addr & 0xfffffffc); + radeon_ring_write(ring, upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff); + radeon_ring_write(ring, 1); /* number of DWs to follow */ + radeon_ring_write(ring, 0xDEADBEEF); + radeon_ring_unlock_commit(rdev, ring); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < rdev->usec_timeout) { + DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i); + } else { + DRM_ERROR("radeon: ring %d test failed (0x%08X)\n", + ring->idx, tmp); + r = -EINVAL; + } + return r; +} + +/** + * cik_sdma_ib_test - test an IB on the DMA engine + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Test a simple IB in the DMA ring (CIK). + * Returns 0 on success, error on failure. + */ +int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring) +{ + struct radeon_ib ib; + unsigned i; + int r; + volatile void __iomem *ptr = (volatile void *)rdev->vram_scratch.ptr; + u32 tmp = 0; + + if (!ptr) { + DRM_ERROR("invalid vram scratch pointer\n"); + return -EINVAL; + } + + tmp = 0xCAFEDEAD; + writel(tmp, ptr); + + r = radeon_ib_get(rdev, ring->idx, &ib, NULL, 256); + if (r) { + DRM_ERROR("radeon: failed to get ib (%d).\n", r); + return r; + } + + ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib.ptr[1] = rdev->vram_scratch.gpu_addr & 0xfffffffc; + ib.ptr[2] = upper_32_bits(rdev->vram_scratch.gpu_addr) & 0xffffffff; + ib.ptr[3] = 1; + ib.ptr[4] = 0xDEADBEEF; + ib.length_dw = 5; + + r = radeon_ib_schedule(rdev, &ib, NULL); + if (r) { + radeon_ib_free(rdev, &ib); + DRM_ERROR("radeon: failed to schedule ib (%d).\n", r); + return r; + } + r = radeon_fence_wait(ib.fence, false); + if (r) { + DRM_ERROR("radeon: fence wait failed (%d).\n", r); + return r; + } + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = readl(ptr); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i < rdev->usec_timeout) { + DRM_INFO("ib test on ring %d succeeded in %u usecs\n", ib.fence->ring, i); + } else { + DRM_ERROR("radeon: ib test failed (0x%08X)\n", tmp); + r = -EINVAL; + } + radeon_ib_free(rdev, &ib); + return r; +} + + +static void cik_print_gpu_status_regs(struct radeon_device *rdev) +{ + dev_info(rdev->dev, " GRBM_STATUS=0x%08X\n", + RREG32(GRBM_STATUS)); + dev_info(rdev->dev, " GRBM_STATUS2=0x%08X\n", + RREG32(GRBM_STATUS2)); + dev_info(rdev->dev, " GRBM_STATUS_SE0=0x%08X\n", + RREG32(GRBM_STATUS_SE0)); + dev_info(rdev->dev, " GRBM_STATUS_SE1=0x%08X\n", + RREG32(GRBM_STATUS_SE1)); + dev_info(rdev->dev, " GRBM_STATUS_SE2=0x%08X\n", + RREG32(GRBM_STATUS_SE2)); + dev_info(rdev->dev, " GRBM_STATUS_SE3=0x%08X\n", + RREG32(GRBM_STATUS_SE3)); + dev_info(rdev->dev, " SRBM_STATUS=0x%08X\n", + RREG32(SRBM_STATUS)); + dev_info(rdev->dev, " SRBM_STATUS2=0x%08X\n", + RREG32(SRBM_STATUS2)); + dev_info(rdev->dev, " SDMA0_STATUS_REG = 0x%08X\n", + RREG32(SDMA0_STATUS_REG + SDMA0_REGISTER_OFFSET)); + dev_info(rdev->dev, " SDMA1_STATUS_REG = 0x%08X\n", + RREG32(SDMA0_STATUS_REG + SDMA1_REGISTER_OFFSET)); + dev_info(rdev->dev, " CP_STAT = 0x%08x\n", RREG32(CP_STAT)); + dev_info(rdev->dev, " CP_STALLED_STAT1 = 0x%08x\n", + RREG32(CP_STALLED_STAT1)); + dev_info(rdev->dev, " CP_STALLED_STAT2 = 0x%08x\n", + RREG32(CP_STALLED_STAT2)); + dev_info(rdev->dev, " CP_STALLED_STAT3 = 0x%08x\n", + RREG32(CP_STALLED_STAT3)); + dev_info(rdev->dev, " CP_CPF_BUSY_STAT = 0x%08x\n", + RREG32(CP_CPF_BUSY_STAT)); + dev_info(rdev->dev, " CP_CPF_STALLED_STAT1 = 0x%08x\n", + RREG32(CP_CPF_STALLED_STAT1)); + dev_info(rdev->dev, " CP_CPF_STATUS = 0x%08x\n", RREG32(CP_CPF_STATUS)); + dev_info(rdev->dev, " CP_CPC_BUSY_STAT = 0x%08x\n", RREG32(CP_CPC_BUSY_STAT)); + dev_info(rdev->dev, " CP_CPC_STALLED_STAT1 = 0x%08x\n", + RREG32(CP_CPC_STALLED_STAT1)); + dev_info(rdev->dev, " CP_CPC_STATUS = 0x%08x\n", RREG32(CP_CPC_STATUS)); +} + +/** + * cik_gpu_check_soft_reset - check which blocks are busy + * + * @rdev: radeon_device pointer + * + * Check which blocks are busy and return the relevant reset + * mask to be used by cik_gpu_soft_reset(). + * Returns a mask of the blocks to be reset. + */ +static u32 cik_gpu_check_soft_reset(struct radeon_device *rdev) +{ + u32 reset_mask = 0; + u32 tmp; + + /* GRBM_STATUS */ + tmp = RREG32(GRBM_STATUS); + if (tmp & (PA_BUSY | SC_BUSY | + BCI_BUSY | SX_BUSY | + TA_BUSY | VGT_BUSY | + DB_BUSY | CB_BUSY | + GDS_BUSY | SPI_BUSY | + IA_BUSY | IA_BUSY_NO_DMA)) + reset_mask |= RADEON_RESET_GFX; + + if (tmp & (CP_BUSY | CP_COHERENCY_BUSY)) + reset_mask |= RADEON_RESET_CP; + + /* GRBM_STATUS2 */ + tmp = RREG32(GRBM_STATUS2); + if (tmp & RLC_BUSY) + reset_mask |= RADEON_RESET_RLC; + + /* SDMA0_STATUS_REG */ + tmp = RREG32(SDMA0_STATUS_REG + SDMA0_REGISTER_OFFSET); + if (!(tmp & SDMA_IDLE)) + reset_mask |= RADEON_RESET_DMA; + + /* SDMA1_STATUS_REG */ + tmp = RREG32(SDMA0_STATUS_REG + SDMA1_REGISTER_OFFSET); + if (!(tmp & SDMA_IDLE)) + reset_mask |= RADEON_RESET_DMA1; + + /* SRBM_STATUS2 */ + tmp = RREG32(SRBM_STATUS2); + if (tmp & SDMA_BUSY) + reset_mask |= RADEON_RESET_DMA; + + if (tmp & SDMA1_BUSY) + reset_mask |= RADEON_RESET_DMA1; + + /* SRBM_STATUS */ + tmp = RREG32(SRBM_STATUS); + + if (tmp & IH_BUSY) + reset_mask |= RADEON_RESET_IH; + + if (tmp & SEM_BUSY) + reset_mask |= RADEON_RESET_SEM; + + if (tmp & GRBM_RQ_PENDING) + reset_mask |= RADEON_RESET_GRBM; + + if (tmp & VMC_BUSY) + reset_mask |= RADEON_RESET_VMC; + + if (tmp & (MCB_BUSY | MCB_NON_DISPLAY_BUSY | + MCC_BUSY | MCD_BUSY)) + reset_mask |= RADEON_RESET_MC; + + if (evergreen_is_display_hung(rdev)) + reset_mask |= RADEON_RESET_DISPLAY; + + /* Skip MC reset as it's mostly likely not hung, just busy */ + if (reset_mask & RADEON_RESET_MC) { + DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask); + reset_mask &= ~RADEON_RESET_MC; + } + + return reset_mask; +} + +/** + * cik_gpu_soft_reset - soft reset GPU + * + * @rdev: radeon_device pointer + * @reset_mask: mask of which blocks to reset + * + * Soft reset the blocks specified in @reset_mask. + */ +static void cik_gpu_soft_reset(struct radeon_device *rdev, u32 reset_mask) +{ + struct evergreen_mc_save save; + u32 grbm_soft_reset = 0, srbm_soft_reset = 0; + u32 tmp; + + if (reset_mask == 0) + return; + + dev_info(rdev->dev, "GPU softreset: 0x%08X\n", reset_mask); + + cik_print_gpu_status_regs(rdev); + dev_info(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n", + RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR)); + dev_info(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n", + RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS)); + + /* stop the rlc */ + cik_rlc_stop(rdev); + + /* Disable GFX parsing/prefetching */ + WREG32(CP_ME_CNTL, CP_ME_HALT | CP_PFP_HALT | CP_CE_HALT); + + /* Disable MEC parsing/prefetching */ + WREG32(CP_MEC_CNTL, MEC_ME1_HALT | MEC_ME2_HALT); + + if (reset_mask & RADEON_RESET_DMA) { + /* sdma0 */ + tmp = RREG32(SDMA0_ME_CNTL + SDMA0_REGISTER_OFFSET); + tmp |= SDMA_HALT; + WREG32(SDMA0_ME_CNTL + SDMA0_REGISTER_OFFSET, tmp); + } + if (reset_mask & RADEON_RESET_DMA1) { + /* sdma1 */ + tmp = RREG32(SDMA0_ME_CNTL + SDMA1_REGISTER_OFFSET); + tmp |= SDMA_HALT; + WREG32(SDMA0_ME_CNTL + SDMA1_REGISTER_OFFSET, tmp); + } + + evergreen_mc_stop(rdev, &save); + if (evergreen_mc_wait_for_idle(rdev)) { + dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); + } + if (reset_mask & (RADEON_RESET_GFX | RADEON_RESET_COMPUTE | RADEON_RESET_CP)) + grbm_soft_reset = SOFT_RESET_CP | SOFT_RESET_GFX; + + if (reset_mask & RADEON_RESET_CP) { + grbm_soft_reset |= SOFT_RESET_CP; + + srbm_soft_reset |= SOFT_RESET_GRBM; + } + + if (reset_mask & RADEON_RESET_DMA) + srbm_soft_reset |= SOFT_RESET_SDMA; + + if (reset_mask & RADEON_RESET_DMA1) + srbm_soft_reset |= SOFT_RESET_SDMA1; + + if (reset_mask & RADEON_RESET_DISPLAY) + srbm_soft_reset |= SOFT_RESET_DC; + + if (reset_mask & RADEON_RESET_RLC) + grbm_soft_reset |= SOFT_RESET_RLC; + + if (reset_mask & RADEON_RESET_SEM) + srbm_soft_reset |= SOFT_RESET_SEM; + + if (reset_mask & RADEON_RESET_IH) + srbm_soft_reset |= SOFT_RESET_IH; + + if (reset_mask & RADEON_RESET_GRBM) + srbm_soft_reset |= SOFT_RESET_GRBM; + + if (reset_mask & RADEON_RESET_VMC) + srbm_soft_reset |= SOFT_RESET_VMC; + + if (!(rdev->flags & RADEON_IS_IGP)) { + if (reset_mask & RADEON_RESET_MC) + srbm_soft_reset |= SOFT_RESET_MC; + } + + if (grbm_soft_reset) { + tmp = RREG32(GRBM_SOFT_RESET); + tmp |= grbm_soft_reset; + dev_info(rdev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(GRBM_SOFT_RESET, tmp); + tmp = RREG32(GRBM_SOFT_RESET); + + DRM_UDELAY(50); + + tmp &= ~grbm_soft_reset; + WREG32(GRBM_SOFT_RESET, tmp); + tmp = RREG32(GRBM_SOFT_RESET); + } + + if (srbm_soft_reset) { + tmp = RREG32(SRBM_SOFT_RESET); + tmp |= srbm_soft_reset; + dev_info(rdev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp); + WREG32(SRBM_SOFT_RESET, tmp); + tmp = RREG32(SRBM_SOFT_RESET); + + DRM_UDELAY(50); + + tmp &= ~srbm_soft_reset; + WREG32(SRBM_SOFT_RESET, tmp); + tmp = RREG32(SRBM_SOFT_RESET); + } + + /* Wait a little for things to settle down */ + DRM_UDELAY(50); + + evergreen_mc_resume(rdev, &save); + DRM_UDELAY(50); + + cik_print_gpu_status_regs(rdev); +} + +/** + * cik_asic_reset - soft reset GPU + * + * @rdev: radeon_device pointer + * + * Look up which blocks are hung and attempt + * to reset them. + * Returns 0 for success. + */ +int cik_asic_reset(struct radeon_device *rdev) +{ + u32 reset_mask; + + reset_mask = cik_gpu_check_soft_reset(rdev); + + if (reset_mask) + r600_set_bios_scratch_engine_hung(rdev, true); + + cik_gpu_soft_reset(rdev, reset_mask); + + reset_mask = cik_gpu_check_soft_reset(rdev); + + if (!reset_mask) + r600_set_bios_scratch_engine_hung(rdev, false); + + return 0; +} + +/** + * cik_gfx_is_lockup - check if the 3D engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the 3D engine is locked up (CIK). + * Returns true if the engine is locked, false if not. + */ +bool cik_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = cik_gpu_check_soft_reset(rdev); + + if (!(reset_mask & (RADEON_RESET_GFX | + RADEON_RESET_COMPUTE | + RADEON_RESET_CP))) { + radeon_ring_lockup_update(ring); + return false; + } + /* force CP activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + +/** + * cik_sdma_is_lockup - Check if the DMA engine is locked up + * + * @rdev: radeon_device pointer + * @ring: radeon_ring structure holding ring information + * + * Check if the async DMA engine is locked up (CIK). + * Returns true if the engine appears to be locked up, false if not. + */ +bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring) +{ + u32 reset_mask = cik_gpu_check_soft_reset(rdev); + u32 mask; + + if (ring->idx == R600_RING_TYPE_DMA_INDEX) + mask = RADEON_RESET_DMA; + else + mask = RADEON_RESET_DMA1; + + if (!(reset_mask & mask)) { + radeon_ring_lockup_update(ring); + return false; + } + /* force ring activities */ + radeon_ring_force_activity(rdev, ring); + return radeon_ring_test_lockup(rdev, ring); +} + +/* MC */ +/** + * cik_mc_program - program the GPU memory controller + * + * @rdev: radeon_device pointer + * + * Set the location of vram, gart, and AGP in the GPU's + * physical address space (CIK). + */ +static __unused void cik_mc_program(struct radeon_device *rdev) +{ + struct evergreen_mc_save save; + u32 tmp; + int i, j; + + /* Initialize HDP */ + for (i = 0, j = 0; i < 32; i++, j += 0x18) { + WREG32((0x2c14 + j), 0x00000000); + WREG32((0x2c18 + j), 0x00000000); + WREG32((0x2c1c + j), 0x00000000); + WREG32((0x2c20 + j), 0x00000000); + WREG32((0x2c24 + j), 0x00000000); + } + WREG32(HDP_REG_COHERENCY_FLUSH_CNTL, 0); + + evergreen_mc_stop(rdev, &save); + if (radeon_mc_wait_for_idle(rdev)) { + dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); + } + /* Lockout access through VGA aperture*/ + WREG32(VGA_HDP_CONTROL, VGA_MEMORY_DISABLE); + /* Update configuration */ + WREG32(MC_VM_SYSTEM_APERTURE_LOW_ADDR, + rdev->mc.vram_start >> 12); + WREG32(MC_VM_SYSTEM_APERTURE_HIGH_ADDR, + rdev->mc.vram_end >> 12); + WREG32(MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR, + rdev->vram_scratch.gpu_addr >> 12); + tmp = ((rdev->mc.vram_end >> 24) & 0xFFFF) << 16; + tmp |= ((rdev->mc.vram_start >> 24) & 0xFFFF); + WREG32(MC_VM_FB_LOCATION, tmp); + /* XXX double check these! */ + WREG32(HDP_NONSURFACE_BASE, (rdev->mc.vram_start >> 8)); + WREG32(HDP_NONSURFACE_INFO, (2 << 7) | (1 << 30)); + WREG32(HDP_NONSURFACE_SIZE, 0x3FFFFFFF); + WREG32(MC_VM_AGP_BASE, 0); + WREG32(MC_VM_AGP_TOP, 0x0FFFFFFF); + WREG32(MC_VM_AGP_BOT, 0x0FFFFFFF); + if (radeon_mc_wait_for_idle(rdev)) { + dev_warn(rdev->dev, "Wait for MC idle timedout !\n"); + } + evergreen_mc_resume(rdev, &save); + /* we need to own VRAM, so turn off the VGA renderer here + * to stop it overwriting our objects */ + rv515_vga_render_disable(rdev); +} + +/** + * cik_mc_init - initialize the memory controller driver params + * + * @rdev: radeon_device pointer + * + * Look up the amount of vram, vram width, and decide how to place + * vram and gart within the GPU's physical address space (CIK). + * Returns 0 for success. + */ +static __unused int cik_mc_init(struct radeon_device *rdev) +{ + u32 tmp; + int chansize, numchan; + + /* Get VRAM informations */ + rdev->mc.vram_is_ddr = true; + tmp = RREG32(MC_ARB_RAMCFG); + if (tmp & CHANSIZE_MASK) { + chansize = 64; + } else { + chansize = 32; + } + tmp = RREG32(MC_SHARED_CHMAP); + switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { + case 0: + default: + numchan = 1; + break; + case 1: + numchan = 2; + break; + case 2: + numchan = 4; + break; + case 3: + numchan = 8; + break; + case 4: + numchan = 3; + break; + case 5: + numchan = 6; + break; + case 6: + numchan = 10; + break; + case 7: + numchan = 12; + break; + case 8: + numchan = 16; + break; + } + rdev->mc.vram_width = numchan * chansize; + /* Could aper size report 0 ? */ + rdev->mc.aper_base = drm_get_resource_start(rdev->ddev, 0); + rdev->mc.aper_size = drm_get_resource_len(rdev->ddev, 0); + /* size in MB on si */ + rdev->mc.mc_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; + rdev->mc.real_vram_size = RREG32(CONFIG_MEMSIZE) * 1024 * 1024; + rdev->mc.visible_vram_size = rdev->mc.aper_size; + si_vram_gtt_location(rdev, &rdev->mc); + radeon_update_bandwidth_info(rdev); + + return 0; +} + +/* + * GART + * VMID 0 is the physical GPU addresses as used by the kernel. + * VMIDs 1-15 are used for userspace clients and are handled + * by the radeon vm/hsa code. + */ +/** + * cik_pcie_gart_tlb_flush - gart tlb flush callback + * + * @rdev: radeon_device pointer + * + * Flush the TLB for the VMID 0 page table (CIK). + */ +void cik_pcie_gart_tlb_flush(struct radeon_device *rdev) +{ + /* flush hdp cache */ + WREG32(HDP_MEM_COHERENCY_FLUSH_CNTL, 0); + + /* bits 0-15 are the VM contexts0-15 */ + WREG32(VM_INVALIDATE_REQUEST, 0x1); +} + +/** + * cik_pcie_gart_enable - gart enable + * + * @rdev: radeon_device pointer + * + * This sets up the TLBs, programs the page tables for VMID0, + * sets up the hw for VMIDs 1-15 which are allocated on + * demand, and sets up the global locations for the LDS, GDS, + * and GPUVM for FSA64 clients (CIK). + * Returns 0 for success, errors for failure. + */ +static __unused int cik_pcie_gart_enable(struct radeon_device *rdev) +{ + int r, i; + + if (rdev->gart.robj == NULL) { + dev_err(rdev->dev, "No VRAM object for PCIE GART.\n"); + return -EINVAL; + } + r = radeon_gart_table_vram_pin(rdev); + if (r) + return r; + radeon_gart_restore(rdev); + /* Setup TLB control */ + WREG32(MC_VM_MX_L1_TLB_CNTL, + (0xA << 7) | + ENABLE_L1_TLB | + SYSTEM_ACCESS_MODE_NOT_IN_SYS | + ENABLE_ADVANCED_DRIVER_MODEL | + SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU); + /* Setup L2 cache */ + WREG32(VM_L2_CNTL, ENABLE_L2_CACHE | + ENABLE_L2_FRAGMENT_PROCESSING | + ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE | + ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE | + EFFECTIVE_L2_QUEUE_SIZE(7) | + CONTEXT1_IDENTITY_ACCESS_MODE(1)); + WREG32(VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS | INVALIDATE_L2_CACHE); + WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY | + L2_CACHE_BIGK_FRAGMENT_SIZE(6)); + /* setup context0 */ + WREG32(VM_CONTEXT0_PAGE_TABLE_START_ADDR, rdev->mc.gtt_start >> 12); + WREG32(VM_CONTEXT0_PAGE_TABLE_END_ADDR, rdev->mc.gtt_end >> 12); + WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR, rdev->gart.table_addr >> 12); + WREG32(VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, + (u32)(rdev->dummy_page.addr >> 12)); + WREG32(VM_CONTEXT0_CNTL2, 0); + WREG32(VM_CONTEXT0_CNTL, (ENABLE_CONTEXT | PAGE_TABLE_DEPTH(0) | + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT)); + + WREG32(0x15D4, 0); + WREG32(0x15D8, 0); + WREG32(0x15DC, 0); + + /* empty context1-15 */ + /* FIXME start with 4G, once using 2 level pt switch to full + * vm size space + */ + /* set vm size, must be a multiple of 4 */ + WREG32(VM_CONTEXT1_PAGE_TABLE_START_ADDR, 0); + WREG32(VM_CONTEXT1_PAGE_TABLE_END_ADDR, rdev->vm_manager.max_pfn); + for (i = 1; i < 16; i++) { + if (i < 8) + WREG32(VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (i << 2), + rdev->gart.table_addr >> 12); + else + WREG32(VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((i - 8) << 2), + rdev->gart.table_addr >> 12); + } + + /* enable context1-15 */ + WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR, + (u32)(rdev->dummy_page.addr >> 12)); + WREG32(VM_CONTEXT1_CNTL2, 4); + WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) | + RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT | + RANGE_PROTECTION_FAULT_ENABLE_DEFAULT | + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT | + DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT | + PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT | + PDE0_PROTECTION_FAULT_ENABLE_DEFAULT | + VALID_PROTECTION_FAULT_ENABLE_INTERRUPT | + VALID_PROTECTION_FAULT_ENABLE_DEFAULT | + READ_PROTECTION_FAULT_ENABLE_INTERRUPT | + READ_PROTECTION_FAULT_ENABLE_DEFAULT | + WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT | + WRITE_PROTECTION_FAULT_ENABLE_DEFAULT); + + /* TC cache setup ??? */ + WREG32(TC_CFG_L1_LOAD_POLICY0, 0); + WREG32(TC_CFG_L1_LOAD_POLICY1, 0); + WREG32(TC_CFG_L1_STORE_POLICY, 0); + + WREG32(TC_CFG_L2_LOAD_POLICY0, 0); + WREG32(TC_CFG_L2_LOAD_POLICY1, 0); + WREG32(TC_CFG_L2_STORE_POLICY0, 0); + WREG32(TC_CFG_L2_STORE_POLICY1, 0); + WREG32(TC_CFG_L2_ATOMIC_POLICY, 0); + + WREG32(TC_CFG_L1_VOLATILE, 0); + WREG32(TC_CFG_L2_VOLATILE, 0); + + if (rdev->family == CHIP_KAVERI) { + u32 tmp = RREG32(CHUB_CONTROL); + tmp &= ~BYPASS_VM; + WREG32(CHUB_CONTROL, tmp); + } + + /* XXX SH_MEM regs */ + /* where to put LDS, scratch, GPUVM in FSA64 space */ + spin_lock(&rdev->srbm_mutex); + for (i = 0; i < 16; i++) { + cik_srbm_select(rdev, 0, 0, 0, i); + /* CP and shaders */ + WREG32(SH_MEM_CONFIG, 0); + WREG32(SH_MEM_APE1_BASE, 1); + WREG32(SH_MEM_APE1_LIMIT, 0); + WREG32(SH_MEM_BASES, 0); + /* SDMA GFX */ + WREG32(SDMA0_GFX_VIRTUAL_ADDR + SDMA0_REGISTER_OFFSET, 0); + WREG32(SDMA0_GFX_APE1_CNTL + SDMA0_REGISTER_OFFSET, 0); + WREG32(SDMA0_GFX_VIRTUAL_ADDR + SDMA1_REGISTER_OFFSET, 0); + WREG32(SDMA0_GFX_APE1_CNTL + SDMA1_REGISTER_OFFSET, 0); + /* XXX SDMA RLC - todo */ + } + cik_srbm_select(rdev, 0, 0, 0, 0); + spin_unlock(&rdev->srbm_mutex); + + cik_pcie_gart_tlb_flush(rdev); + DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", + (unsigned)(rdev->mc.gtt_size >> 20), + (unsigned long long)rdev->gart.table_addr); + rdev->gart.ready = true; + return 0; +} + +/** + * cik_pcie_gart_disable - gart disable + * + * @rdev: radeon_device pointer + * + * This disables all VM page table (CIK). + */ +static void cik_pcie_gart_disable(struct radeon_device *rdev) +{ + /* Disable all tables */ + WREG32(VM_CONTEXT0_CNTL, 0); + WREG32(VM_CONTEXT1_CNTL, 0); + /* Setup TLB control */ + WREG32(MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE_NOT_IN_SYS | + SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU); + /* Setup L2 cache */ + WREG32(VM_L2_CNTL, + ENABLE_L2_FRAGMENT_PROCESSING | + ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE | + ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE | + EFFECTIVE_L2_QUEUE_SIZE(7) | + CONTEXT1_IDENTITY_ACCESS_MODE(1)); + WREG32(VM_L2_CNTL2, 0); + WREG32(VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY | + L2_CACHE_BIGK_FRAGMENT_SIZE(6)); + radeon_gart_table_vram_unpin(rdev); +} + +/** + * cik_pcie_gart_fini - vm fini callback + * + * @rdev: radeon_device pointer + * + * Tears down the driver GART/VM setup (CIK). + */ +static __unused void cik_pcie_gart_fini(struct radeon_device *rdev) +{ + cik_pcie_gart_disable(rdev); + radeon_gart_table_vram_free(rdev); + radeon_gart_fini(rdev); +} + +/* vm parser */ +/** + * cik_ib_parse - vm ib_parse callback + * + * @rdev: radeon_device pointer + * @ib: indirect buffer pointer + * + * CIK uses hw IB checking so this is a nop (CIK). + */ +int cik_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib) +{ + return 0; +} + +/* + * vm + * VMID 0 is the physical GPU addresses as used by the kernel. + * VMIDs 1-15 are used for userspace clients and are handled + * by the radeon vm/hsa code. + */ +/** + * cik_vm_init - cik vm init callback + * + * @rdev: radeon_device pointer + * + * Inits cik specific vm parameters (number of VMs, base of vram for + * VMIDs 1-15) (CIK). + * Returns 0 for success. + */ +int cik_vm_init(struct radeon_device *rdev) +{ + /* number of VMs */ + rdev->vm_manager.nvm = 16; + /* base offset of vram pages */ + if (rdev->flags & RADEON_IS_IGP) { + u64 tmp = RREG32(MC_VM_FB_OFFSET); + tmp <<= 22; + rdev->vm_manager.vram_base_offset = tmp; + } else + rdev->vm_manager.vram_base_offset = 0; + + return 0; +} + +/** + * cik_vm_fini - cik vm fini callback + * + * @rdev: radeon_device pointer + * + * Tear down any asic specific VM setup (CIK). + */ +void cik_vm_fini(struct radeon_device *rdev) +{ +} + +/** + * cik_vm_decode_fault - print human readable fault info + * + * @rdev: radeon_device pointer + * @status: VM_CONTEXT1_PROTECTION_FAULT_STATUS register value + * @addr: VM_CONTEXT1_PROTECTION_FAULT_ADDR register value + * + * Print human readable fault information (CIK). + */ +static void cik_vm_decode_fault(struct radeon_device *rdev, + u32 status, u32 addr, u32 mc_client) +{ + u32 mc_id = (status & MEMORY_CLIENT_ID_MASK) >> MEMORY_CLIENT_ID_SHIFT; + u32 vmid = (status & FAULT_VMID_MASK) >> FAULT_VMID_SHIFT; + u32 protections = (status & PROTECTIONS_MASK) >> PROTECTIONS_SHIFT; + char *block = (char *)&mc_client; + + printk("VM fault (0x%02x, vmid %d) at page %u, %s from %s (%d)\n", + protections, vmid, addr, + (status & MEMORY_CLIENT_RW_MASK) ? "write" : "read", + block, mc_id); +} + +/** + * cik_vm_flush - cik vm flush using the CP + * + * @rdev: radeon_device pointer + * + * Update the page table base and flush the VM TLB + * using the CP (CIK). + */ +void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +{ + struct radeon_ring *ring = &rdev->ring[ridx]; + + if (vm == NULL) + return; + + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + if (vm->id < 8) { + radeon_ring_write(ring, + (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + } else { + radeon_ring_write(ring, + (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + } + radeon_ring_write(ring, 0); + radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + + /* update SH_MEM_* regs */ + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, VMID(vm->id)); + + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 6)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, SH_MEM_BASES >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, 0); /* SH_MEM_BASES */ + radeon_ring_write(ring, 0); /* SH_MEM_CONFIG */ + radeon_ring_write(ring, 1); /* SH_MEM_APE1_BASE */ + radeon_ring_write(ring, 0); /* SH_MEM_APE1_LIMIT */ + + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, VMID(0)); + + /* HDP flush */ + /* We should be using the WAIT_REG_MEM packet here like in + * cik_fence_ring_emit(), but it causes the CP to hang in this + * context... + */ + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, HDP_MEM_COHERENCY_FLUSH_CNTL >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, 0); + + /* bits 0-15 are the VM contexts0-15 */ + radeon_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + radeon_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(0))); + radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); + radeon_ring_write(ring, 0); + radeon_ring_write(ring, 1 << vm->id); + + /* compute doesn't have PFP */ + if (ridx == RADEON_RING_TYPE_GFX_INDEX) { + /* sync PFP to ME, otherwise we might get invalid PFP reads */ + radeon_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0)); + radeon_ring_write(ring, 0x0); + } +} + +/** + * cik_vm_set_page - update the page tables using sDMA + * + * @rdev: radeon_device pointer + * @ib: indirect buffer to fill with commands + * @pe: addr of the page entry + * @addr: dst addr to write into pe + * @count: number of page entries to update + * @incr: increase next addr by incr bytes + * @flags: access flags + * + * Update the page tables using CP or sDMA (CIK). + */ +void cik_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags) +{ + uint32_t r600_flags = cayman_vm_page_flags(rdev, flags); + uint64_t value; + unsigned ndw; + + if (rdev->asic->vm.pt_ring_index == RADEON_RING_TYPE_GFX_INDEX) { + /* CP */ + while (count) { + ndw = 2 + count * 2; + if (ndw > 0x3FFE) + ndw = 0x3FFE; + + ib->ptr[ib->length_dw++] = PACKET3(PACKET3_WRITE_DATA, ndw); + ib->ptr[ib->length_dw++] = (WRITE_DATA_ENGINE_SEL(0) | + WRITE_DATA_DST_SEL(1)); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + for (; ndw > 2; ndw -= 2, --count, pe += 8) { + if (flags & RADEON_VM_PAGE_SYSTEM) { + value = radeon_vm_map_gart(rdev, addr); + value &= 0xFFFFFFFFFFFFF000ULL; + } else if (flags & RADEON_VM_PAGE_VALID) { + value = addr; + } else { + value = 0; + } + addr += incr; + value |= r600_flags; + ib->ptr[ib->length_dw++] = value; + ib->ptr[ib->length_dw++] = upper_32_bits(value); + } + } + } else { + /* DMA */ + if (flags & RADEON_VM_PAGE_SYSTEM) { + while (count) { + ndw = count * 2; + if (ndw > 0xFFFFE) + ndw = 0xFFFFE; + + /* for non-physically contiguous pages (system) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0); + ib->ptr[ib->length_dw++] = pe; + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = ndw; + for (; ndw > 0; ndw -= 2, --count, pe += 8) { + if (flags & RADEON_VM_PAGE_SYSTEM) { + value = radeon_vm_map_gart(rdev, addr); + value &= 0xFFFFFFFFFFFFF000ULL; + } else if (flags & RADEON_VM_PAGE_VALID) { + value = addr; + } else { + value = 0; + } + addr += incr; + value |= r600_flags; + ib->ptr[ib->length_dw++] = value; + ib->ptr[ib->length_dw++] = upper_32_bits(value); + } + } + } else { + while (count) { + ndw = count; + if (ndw > 0x7FFFF) + ndw = 0x7FFFF; + + if (flags & RADEON_VM_PAGE_VALID) + value = addr; + else + value = 0; + /* for physically contiguous pages (vram) */ + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0); + ib->ptr[ib->length_dw++] = pe; /* dst addr */ + ib->ptr[ib->length_dw++] = upper_32_bits(pe); + ib->ptr[ib->length_dw++] = r600_flags; /* mask */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = value; /* value */ + ib->ptr[ib->length_dw++] = upper_32_bits(value); + ib->ptr[ib->length_dw++] = incr; /* increment size */ + ib->ptr[ib->length_dw++] = 0; + ib->ptr[ib->length_dw++] = ndw; /* number of entries */ + pe += ndw * 8; + addr += ndw * incr; + count -= ndw; + } + } + while (ib->length_dw & 0x7) + ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0); + } +} + +/** + * cik_dma_vm_flush - cik vm flush using sDMA + * + * @rdev: radeon_device pointer + * + * Update the page table base and flush the VM TLB + * using sDMA (CIK). + */ +void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) +{ + struct radeon_ring *ring = &rdev->ring[ridx]; + u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) | + SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ + u32 ref_and_mask; + + if (vm == NULL) + return; + + if (ridx == R600_RING_TYPE_DMA_INDEX) + ref_and_mask = SDMA0; + else + ref_and_mask = SDMA1; + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + if (vm->id < 8) { + radeon_ring_write(ring, (VM_CONTEXT0_PAGE_TABLE_BASE_ADDR + (vm->id << 2)) >> 2); + } else { + radeon_ring_write(ring, (VM_CONTEXT8_PAGE_TABLE_BASE_ADDR + ((vm->id - 8) << 2)) >> 2); + } + radeon_ring_write(ring, vm->pd_gpu_addr >> 12); + + /* update SH_MEM_* regs */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, VMID(vm->id)); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_BASES >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_CONFIG >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_APE1_BASE >> 2); + radeon_ring_write(ring, 1); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SH_MEM_APE1_LIMIT >> 2); + radeon_ring_write(ring, 0); + + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, SRBM_GFX_CNTL >> 2); + radeon_ring_write(ring, VMID(0)); + + /* flush HDP */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits)); + radeon_ring_write(ring, GPU_HDP_FLUSH_DONE); + radeon_ring_write(ring, GPU_HDP_FLUSH_REQ); + radeon_ring_write(ring, ref_and_mask); /* REFERENCE */ + radeon_ring_write(ring, ref_and_mask); /* MASK */ + radeon_ring_write(ring, (4 << 16) | 10); /* RETRY_COUNT, POLL_INTERVAL */ + + /* flush TLB */ + radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000)); + radeon_ring_write(ring, VM_INVALIDATE_REQUEST >> 2); + radeon_ring_write(ring, 1 << vm->id); +} + +/* + * RLC + * The RLC is a multi-purpose microengine that handles a + * variety of functions, the most important of which is + * the interrupt controller. + */ +/** + * cik_rlc_stop - stop the RLC ME + * + * @rdev: radeon_device pointer + * + * Halt the RLC ME (MicroEngine) (CIK). + */ +static void cik_rlc_stop(struct radeon_device *rdev) +{ + int i, j, k; + u32 mask, tmp; + + tmp = RREG32(CP_INT_CNTL_RING0); + tmp &= ~(CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); + WREG32(CP_INT_CNTL_RING0, tmp); + + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + + tmp = RREG32(RLC_CGCG_CGLS_CTRL) & 0xfffffffc; + WREG32(RLC_CGCG_CGLS_CTRL, tmp); + + WREG32(RLC_CNTL, 0); + + for (i = 0; i < rdev->config.cik.max_shader_engines; i++) { + for (j = 0; j < rdev->config.cik.max_sh_per_se; j++) { + cik_select_se_sh(rdev, i, j); + for (k = 0; k < rdev->usec_timeout; k++) { + if (RREG32(RLC_SERDES_CU_MASTER_BUSY) == 0) + break; + DRM_UDELAY(1); + } + } + } + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + + mask = SE_MASTER_BUSY_MASK | GC_MASTER_BUSY | TC0_MASTER_BUSY | TC1_MASTER_BUSY; + for (k = 0; k < rdev->usec_timeout; k++) { + if ((RREG32(RLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0) + break; + DRM_UDELAY(1); + } +} + +/** + * cik_rlc_start - start the RLC ME + * + * @rdev: radeon_device pointer + * + * Unhalt the RLC ME (MicroEngine) (CIK). + */ +static void cik_rlc_start(struct radeon_device *rdev) +{ + u32 tmp; + + WREG32(RLC_CNTL, RLC_ENABLE); + + tmp = RREG32(CP_INT_CNTL_RING0); + tmp |= (CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); + WREG32(CP_INT_CNTL_RING0, tmp); + + DRM_UDELAY(50); +} + +/** + * cik_rlc_resume - setup the RLC hw + * + * @rdev: radeon_device pointer + * + * Initialize the RLC registers, load the ucode, + * and start the RLC (CIK). + * Returns 0 for success, -EINVAL if the ucode is not available. + */ +static __unused int cik_rlc_resume(struct radeon_device *rdev) +{ + u32 i, size; + u32 clear_state_info[3]; + const __be32 *fw_data; + + if (!rdev->rlc_fw) + return -EINVAL; + + switch (rdev->family) { + case CHIP_BONAIRE: + default: + size = BONAIRE_RLC_UCODE_SIZE; + break; + case CHIP_KAVERI: + size = KV_RLC_UCODE_SIZE; + break; + case CHIP_KABINI: + size = KB_RLC_UCODE_SIZE; + break; + } + + cik_rlc_stop(rdev); + + WREG32(GRBM_SOFT_RESET, SOFT_RESET_RLC); + RREG32(GRBM_SOFT_RESET); + DRM_UDELAY(50); + WREG32(GRBM_SOFT_RESET, 0); + RREG32(GRBM_SOFT_RESET); + DRM_UDELAY(50); + + WREG32(RLC_LB_CNTR_INIT, 0); + WREG32(RLC_LB_CNTR_MAX, 0x00008000); + + cik_select_se_sh(rdev, 0xffffffff, 0xffffffff); + WREG32(RLC_LB_INIT_CU_MASK, 0xffffffff); + WREG32(RLC_LB_PARAMS, 0x00600408); + WREG32(RLC_LB_CNTL, 0x80000004); + + WREG32(RLC_MC_CNTL, 0); + WREG32(RLC_UCODE_CNTL, 0); + + fw_data = (const __be32 *)rdev->rlc_fw->data; + WREG32(RLC_GPM_UCODE_ADDR, 0); + for (i = 0; i < size; i++) + WREG32(RLC_GPM_UCODE_DATA, be32_to_cpup(fw_data++)); + WREG32(RLC_GPM_UCODE_ADDR, 0); + + /* XXX */ + clear_state_info[0] = 0;//upper_32_bits(rdev->rlc.save_restore_gpu_addr); + clear_state_info[1] = 0;//rdev->rlc.save_restore_gpu_addr; + clear_state_info[2] = 0;//cik_default_size; + WREG32(RLC_GPM_SCRATCH_ADDR, 0x3d); + for (i = 0; i < 3; i++) + WREG32(RLC_GPM_SCRATCH_DATA, clear_state_info[i]); + WREG32(RLC_DRIVER_DMA_STATUS, 0); + + cik_rlc_start(rdev); + + return 0; +} + +/* + * Interrupts + * Starting with r6xx, interrupts are handled via a ring buffer. + * Ring buffers are areas of GPU accessible memory that the GPU + * writes interrupt vectors into and the host reads vectors out of. + * There is a rptr (read pointer) that determines where the + * host is currently reading, and a wptr (write pointer) + * which determines where the GPU has written. When the + * pointers are equal, the ring is idle. When the GPU + * writes vectors to the ring buffer, it increments the + * wptr. When there is an interrupt, the host then starts + * fetching commands and processing them until the pointers are + * equal again at which point it updates the rptr. + */ + +/** + * cik_enable_interrupts - Enable the interrupt ring buffer + * + * @rdev: radeon_device pointer + * + * Enable the interrupt ring buffer (CIK). + */ +static void cik_enable_interrupts(struct radeon_device *rdev) +{ + u32 ih_cntl = RREG32(IH_CNTL); + u32 ih_rb_cntl = RREG32(IH_RB_CNTL); + + ih_cntl |= ENABLE_INTR; + ih_rb_cntl |= IH_RB_ENABLE; + WREG32(IH_CNTL, ih_cntl); + WREG32(IH_RB_CNTL, ih_rb_cntl); + rdev->ih.enabled = true; +} + +/** + * cik_disable_interrupts - Disable the interrupt ring buffer + * + * @rdev: radeon_device pointer + * + * Disable the interrupt ring buffer (CIK). + */ +static void cik_disable_interrupts(struct radeon_device *rdev) +{ + u32 ih_rb_cntl = RREG32(IH_RB_CNTL); + u32 ih_cntl = RREG32(IH_CNTL); + + ih_rb_cntl &= ~IH_RB_ENABLE; + ih_cntl &= ~ENABLE_INTR; + WREG32(IH_RB_CNTL, ih_rb_cntl); + WREG32(IH_CNTL, ih_cntl); + /* set rptr, wptr to 0 */ + WREG32(IH_RB_RPTR, 0); + WREG32(IH_RB_WPTR, 0); + rdev->ih.enabled = false; + rdev->ih.rptr = 0; +} + +/** + * cik_disable_interrupt_state - Disable all interrupt sources + * + * @rdev: radeon_device pointer + * + * Clear all interrupt enable bits used by the driver (CIK). + */ +static void cik_disable_interrupt_state(struct radeon_device *rdev) +{ + u32 tmp; + + /* gfx ring */ + WREG32(CP_INT_CNTL_RING0, CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); + /* sdma */ + tmp = RREG32(SDMA0_CNTL + SDMA0_REGISTER_OFFSET) & ~TRAP_ENABLE; + WREG32(SDMA0_CNTL + SDMA0_REGISTER_OFFSET, tmp); + tmp = RREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET) & ~TRAP_ENABLE; + WREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET, tmp); + /* compute queues */ + WREG32(CP_ME1_PIPE0_INT_CNTL, 0); + WREG32(CP_ME1_PIPE1_INT_CNTL, 0); + WREG32(CP_ME1_PIPE2_INT_CNTL, 0); + WREG32(CP_ME1_PIPE3_INT_CNTL, 0); + WREG32(CP_ME2_PIPE0_INT_CNTL, 0); + WREG32(CP_ME2_PIPE1_INT_CNTL, 0); + WREG32(CP_ME2_PIPE2_INT_CNTL, 0); + WREG32(CP_ME2_PIPE3_INT_CNTL, 0); + /* grbm */ + WREG32(GRBM_INT_CNTL, 0); + /* vline/vblank, etc. */ + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, 0); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, 0); + if (rdev->num_crtc >= 4) { + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, 0); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, 0); + } + if (rdev->num_crtc >= 6) { + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, 0); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, 0); + } + + /* dac hotplug */ + WREG32(DAC_AUTODETECT_INT_CONTROL, 0); + + /* digital hotplug */ + tmp = RREG32(DC_HPD1_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD1_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD2_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD2_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD3_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD3_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD4_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD4_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD5_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD5_INT_CONTROL, tmp); + tmp = RREG32(DC_HPD6_INT_CONTROL) & DC_HPDx_INT_POLARITY; + WREG32(DC_HPD6_INT_CONTROL, tmp); + +} + +/** + * cik_irq_init - init and enable the interrupt ring + * + * @rdev: radeon_device pointer + * + * Allocate a ring buffer for the interrupt controller, + * enable the RLC, disable interrupts, enable the IH + * ring buffer and enable it (CIK). + * Called at device load and reume. + * Returns 0 for success, errors for failure. + */ +static __unused int cik_irq_init(struct radeon_device *rdev) +{ + int ret = 0; + int rb_bufsz; + u32 interrupt_cntl, ih_cntl, ih_rb_cntl; + + /* allocate ring */ + ret = r600_ih_ring_alloc(rdev); + if (ret) + return ret; + + /* disable irqs */ + cik_disable_interrupts(rdev); + + /* init rlc */ + ret = cik_rlc_resume(rdev); + if (ret) { + r600_ih_ring_fini(rdev); + return ret; + } + + /* setup interrupt control */ + /* XXX this should actually be a bus address, not an MC address. same on older asics */ + WREG32(INTERRUPT_CNTL2, rdev->ih.gpu_addr >> 8); + interrupt_cntl = RREG32(INTERRUPT_CNTL); + /* IH_DUMMY_RD_OVERRIDE=0 - dummy read disabled with msi, enabled without msi + * IH_DUMMY_RD_OVERRIDE=1 - dummy read controlled by IH_DUMMY_RD_EN + */ + interrupt_cntl &= ~IH_DUMMY_RD_OVERRIDE; + /* IH_REQ_NONSNOOP_EN=1 if ring is in non-cacheable memory, e.g., vram */ + interrupt_cntl &= ~IH_REQ_NONSNOOP_EN; + WREG32(INTERRUPT_CNTL, interrupt_cntl); + + WREG32(IH_RB_BASE, rdev->ih.gpu_addr >> 8); + rb_bufsz = drm_order(rdev->ih.ring_size / 4); + + ih_rb_cntl = (IH_WPTR_OVERFLOW_ENABLE | + IH_WPTR_OVERFLOW_CLEAR | + (rb_bufsz << 1)); + + if (rdev->wb.enabled) + ih_rb_cntl |= IH_WPTR_WRITEBACK_ENABLE; + + /* set the writeback address whether it's enabled or not */ + WREG32(IH_RB_WPTR_ADDR_LO, (rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFFFFFFFC); + WREG32(IH_RB_WPTR_ADDR_HI, upper_32_bits(rdev->wb.gpu_addr + R600_WB_IH_WPTR_OFFSET) & 0xFF); + + WREG32(IH_RB_CNTL, ih_rb_cntl); + + /* set rptr, wptr to 0 */ + WREG32(IH_RB_RPTR, 0); + WREG32(IH_RB_WPTR, 0); + + /* Default settings for IH_CNTL (disabled at first) */ + ih_cntl = MC_WRREQ_CREDIT(0x10) | MC_WR_CLEAN_CNT(0x10) | MC_VMID(0); + /* RPTR_REARM only works if msi's are enabled */ + if (rdev->msi_enabled) + ih_cntl |= RPTR_REARM; + WREG32(IH_CNTL, ih_cntl); + + /* force the active interrupt state to all disabled */ + cik_disable_interrupt_state(rdev); + + pci_enable_busmaster(rdev->dev); + + /* enable irqs */ + cik_enable_interrupts(rdev); + + return ret; +} + +/** + * cik_irq_set - enable/disable interrupt sources + * + * @rdev: radeon_device pointer + * + * Enable interrupt sources on the GPU (vblanks, hpd, + * etc.) (CIK). + * Returns 0 for success, errors for failure. + */ +int cik_irq_set(struct radeon_device *rdev) +{ + u32 cp_int_cntl = CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE | + PRIV_INSTR_INT_ENABLE | PRIV_REG_INT_ENABLE; + u32 cp_m1p0, cp_m1p1, cp_m1p2, cp_m1p3; + u32 cp_m2p0, cp_m2p1, cp_m2p2, cp_m2p3; + u32 crtc1 = 0, crtc2 = 0, crtc3 = 0, crtc4 = 0, crtc5 = 0, crtc6 = 0; + u32 hpd1, hpd2, hpd3, hpd4, hpd5, hpd6; + u32 grbm_int_cntl = 0; + u32 dma_cntl, dma_cntl1; + + if (!rdev->irq.installed) { + WARN(1, "Can't enable IRQ/MSI because no handler is installed\n"); + return -EINVAL; + } + /* don't enable anything if the ih is disabled */ + if (!rdev->ih.enabled) { + cik_disable_interrupts(rdev); + /* force the active interrupt state to all disabled */ + cik_disable_interrupt_state(rdev); + return 0; + } + + hpd1 = RREG32(DC_HPD1_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd2 = RREG32(DC_HPD2_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd3 = RREG32(DC_HPD3_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd4 = RREG32(DC_HPD4_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~DC_HPDx_INT_EN; + hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN; + + dma_cntl = RREG32(SDMA0_CNTL + SDMA0_REGISTER_OFFSET) & ~TRAP_ENABLE; + dma_cntl1 = RREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET) & ~TRAP_ENABLE; + + cp_m1p0 = RREG32(CP_ME1_PIPE0_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m1p1 = RREG32(CP_ME1_PIPE1_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m1p2 = RREG32(CP_ME1_PIPE2_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m1p3 = RREG32(CP_ME1_PIPE3_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m2p0 = RREG32(CP_ME2_PIPE0_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m2p1 = RREG32(CP_ME2_PIPE1_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m2p2 = RREG32(CP_ME2_PIPE2_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + cp_m2p3 = RREG32(CP_ME2_PIPE3_INT_CNTL) & ~TIME_STAMP_INT_ENABLE; + + /* enable CP interrupts on all rings */ + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { + DRM_DEBUG("cik_irq_set: sw int gfx\n"); + cp_int_cntl |= TIME_STAMP_INT_ENABLE; + } + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP1_INDEX])) { + struct radeon_ring *ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; + DRM_DEBUG("si_irq_set: sw int cp1\n"); + if (ring->me == 1) { + switch (ring->pipe) { + case 0: + cp_m1p0 |= TIME_STAMP_INT_ENABLE; + break; + case 1: + cp_m1p1 |= TIME_STAMP_INT_ENABLE; + break; + case 2: + cp_m1p2 |= TIME_STAMP_INT_ENABLE; + break; + case 3: + cp_m1p2 |= TIME_STAMP_INT_ENABLE; + break; + default: + DRM_DEBUG("si_irq_set: sw int cp1 invalid pipe %d\n", ring->pipe); + break; + } + } else if (ring->me == 2) { + switch (ring->pipe) { + case 0: + cp_m2p0 |= TIME_STAMP_INT_ENABLE; + break; + case 1: + cp_m2p1 |= TIME_STAMP_INT_ENABLE; + break; + case 2: + cp_m2p2 |= TIME_STAMP_INT_ENABLE; + break; + case 3: + cp_m2p2 |= TIME_STAMP_INT_ENABLE; + break; + default: + DRM_DEBUG("si_irq_set: sw int cp1 invalid pipe %d\n", ring->pipe); + break; + } + } else { + DRM_DEBUG("si_irq_set: sw int cp1 invalid me %d\n", ring->me); + } + } + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_CP2_INDEX])) { + struct radeon_ring *ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; + DRM_DEBUG("si_irq_set: sw int cp2\n"); + if (ring->me == 1) { + switch (ring->pipe) { + case 0: + cp_m1p0 |= TIME_STAMP_INT_ENABLE; + break; + case 1: + cp_m1p1 |= TIME_STAMP_INT_ENABLE; + break; + case 2: + cp_m1p2 |= TIME_STAMP_INT_ENABLE; + break; + case 3: + cp_m1p2 |= TIME_STAMP_INT_ENABLE; + break; + default: + DRM_DEBUG("si_irq_set: sw int cp2 invalid pipe %d\n", ring->pipe); + break; + } + } else if (ring->me == 2) { + switch (ring->pipe) { + case 0: + cp_m2p0 |= TIME_STAMP_INT_ENABLE; + break; + case 1: + cp_m2p1 |= TIME_STAMP_INT_ENABLE; + break; + case 2: + cp_m2p2 |= TIME_STAMP_INT_ENABLE; + break; + case 3: + cp_m2p2 |= TIME_STAMP_INT_ENABLE; + break; + default: + DRM_DEBUG("si_irq_set: sw int cp2 invalid pipe %d\n", ring->pipe); + break; + } + } else { + DRM_DEBUG("si_irq_set: sw int cp2 invalid me %d\n", ring->me); + } + } + + if (atomic_read(&rdev->irq.ring_int[R600_RING_TYPE_DMA_INDEX])) { + DRM_DEBUG("cik_irq_set: sw int dma\n"); + dma_cntl |= TRAP_ENABLE; + } + + if (atomic_read(&rdev->irq.ring_int[CAYMAN_RING_TYPE_DMA1_INDEX])) { + DRM_DEBUG("cik_irq_set: sw int dma1\n"); + dma_cntl1 |= TRAP_ENABLE; + } + + if (rdev->irq.crtc_vblank_int[0] || + atomic_read(&rdev->irq.pflip[0])) { + DRM_DEBUG("cik_irq_set: vblank 0\n"); + crtc1 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.crtc_vblank_int[1] || + atomic_read(&rdev->irq.pflip[1])) { + DRM_DEBUG("cik_irq_set: vblank 1\n"); + crtc2 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.crtc_vblank_int[2] || + atomic_read(&rdev->irq.pflip[2])) { + DRM_DEBUG("cik_irq_set: vblank 2\n"); + crtc3 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.crtc_vblank_int[3] || + atomic_read(&rdev->irq.pflip[3])) { + DRM_DEBUG("cik_irq_set: vblank 3\n"); + crtc4 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.crtc_vblank_int[4] || + atomic_read(&rdev->irq.pflip[4])) { + DRM_DEBUG("cik_irq_set: vblank 4\n"); + crtc5 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.crtc_vblank_int[5] || + atomic_read(&rdev->irq.pflip[5])) { + DRM_DEBUG("cik_irq_set: vblank 5\n"); + crtc6 |= VBLANK_INTERRUPT_MASK; + } + if (rdev->irq.hpd[0]) { + DRM_DEBUG("cik_irq_set: hpd 1\n"); + hpd1 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[1]) { + DRM_DEBUG("cik_irq_set: hpd 2\n"); + hpd2 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[2]) { + DRM_DEBUG("cik_irq_set: hpd 3\n"); + hpd3 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[3]) { + DRM_DEBUG("cik_irq_set: hpd 4\n"); + hpd4 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[4]) { + DRM_DEBUG("cik_irq_set: hpd 5\n"); + hpd5 |= DC_HPDx_INT_EN; + } + if (rdev->irq.hpd[5]) { + DRM_DEBUG("cik_irq_set: hpd 6\n"); + hpd6 |= DC_HPDx_INT_EN; + } + + WREG32(CP_INT_CNTL_RING0, cp_int_cntl); + + WREG32(SDMA0_CNTL + SDMA0_REGISTER_OFFSET, dma_cntl); + WREG32(SDMA0_CNTL + SDMA1_REGISTER_OFFSET, dma_cntl1); + + WREG32(CP_ME1_PIPE0_INT_CNTL, cp_m1p0); + WREG32(CP_ME1_PIPE1_INT_CNTL, cp_m1p1); + WREG32(CP_ME1_PIPE2_INT_CNTL, cp_m1p2); + WREG32(CP_ME1_PIPE3_INT_CNTL, cp_m1p3); + WREG32(CP_ME2_PIPE0_INT_CNTL, cp_m2p0); + WREG32(CP_ME2_PIPE1_INT_CNTL, cp_m2p1); + WREG32(CP_ME2_PIPE2_INT_CNTL, cp_m2p2); + WREG32(CP_ME2_PIPE3_INT_CNTL, cp_m2p3); + + WREG32(GRBM_INT_CNTL, grbm_int_cntl); + + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, crtc1); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, crtc2); + if (rdev->num_crtc >= 4) { + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC2_REGISTER_OFFSET, crtc3); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC3_REGISTER_OFFSET, crtc4); + } + if (rdev->num_crtc >= 6) { + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC4_REGISTER_OFFSET, crtc5); + WREG32(LB_INTERRUPT_MASK + EVERGREEN_CRTC5_REGISTER_OFFSET, crtc6); + } + + WREG32(DC_HPD1_INT_CONTROL, hpd1); + WREG32(DC_HPD2_INT_CONTROL, hpd2); + WREG32(DC_HPD3_INT_CONTROL, hpd3); + WREG32(DC_HPD4_INT_CONTROL, hpd4); + WREG32(DC_HPD5_INT_CONTROL, hpd5); + WREG32(DC_HPD6_INT_CONTROL, hpd6); + + return 0; +} + +/** + * cik_irq_ack - ack interrupt sources + * + * @rdev: radeon_device pointer + * + * Ack interrupt sources on the GPU (vblanks, hpd, + * etc.) (CIK). Certain interrupts sources are sw + * generated and do not require an explicit ack. + */ +static inline void cik_irq_ack(struct radeon_device *rdev) +{ + u32 tmp; + + rdev->irq.stat_regs.cik.disp_int = RREG32(DISP_INTERRUPT_STATUS); + rdev->irq.stat_regs.cik.disp_int_cont = RREG32(DISP_INTERRUPT_STATUS_CONTINUE); + rdev->irq.stat_regs.cik.disp_int_cont2 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE2); + rdev->irq.stat_regs.cik.disp_int_cont3 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE3); + rdev->irq.stat_regs.cik.disp_int_cont4 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE4); + rdev->irq.stat_regs.cik.disp_int_cont5 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE5); + rdev->irq.stat_regs.cik.disp_int_cont6 = RREG32(DISP_INTERRUPT_STATUS_CONTINUE6); + + if (rdev->irq.stat_regs.cik.disp_int & LB_D1_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int & LB_D1_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC0_REGISTER_OFFSET, VLINE_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont & LB_D2_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont & LB_D2_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC1_REGISTER_OFFSET, VLINE_ACK); + + if (rdev->num_crtc >= 4) { + if (rdev->irq.stat_regs.cik.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont2 & LB_D3_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC2_REGISTER_OFFSET, VLINE_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont3 & LB_D4_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC3_REGISTER_OFFSET, VLINE_ACK); + } + + if (rdev->num_crtc >= 6) { + if (rdev->irq.stat_regs.cik.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont4 & LB_D5_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC4_REGISTER_OFFSET, VLINE_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) + WREG32(LB_VBLANK_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VBLANK_ACK); + if (rdev->irq.stat_regs.cik.disp_int_cont5 & LB_D6_VLINE_INTERRUPT) + WREG32(LB_VLINE_STATUS + EVERGREEN_CRTC5_REGISTER_OFFSET, VLINE_ACK); + } + + if (rdev->irq.stat_regs.cik.disp_int & DC_HPD1_INTERRUPT) { + tmp = RREG32(DC_HPD1_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD1_INT_CONTROL, tmp); + } + if (rdev->irq.stat_regs.cik.disp_int_cont & DC_HPD2_INTERRUPT) { + tmp = RREG32(DC_HPD2_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD2_INT_CONTROL, tmp); + } + if (rdev->irq.stat_regs.cik.disp_int_cont2 & DC_HPD3_INTERRUPT) { + tmp = RREG32(DC_HPD3_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD3_INT_CONTROL, tmp); + } + if (rdev->irq.stat_regs.cik.disp_int_cont3 & DC_HPD4_INTERRUPT) { + tmp = RREG32(DC_HPD4_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD4_INT_CONTROL, tmp); + } + if (rdev->irq.stat_regs.cik.disp_int_cont4 & DC_HPD5_INTERRUPT) { + tmp = RREG32(DC_HPD5_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD5_INT_CONTROL, tmp); + } + if (rdev->irq.stat_regs.cik.disp_int_cont5 & DC_HPD6_INTERRUPT) { + tmp = RREG32(DC_HPD5_INT_CONTROL); + tmp |= DC_HPDx_INT_ACK; + WREG32(DC_HPD6_INT_CONTROL, tmp); + } +} + +/** + * cik_irq_disable - disable interrupts + * + * @rdev: radeon_device pointer + * + * Disable interrupts on the hw (CIK). + */ +static void cik_irq_disable(struct radeon_device *rdev) +{ + cik_disable_interrupts(rdev); + /* Wait and acknowledge irq */ + DRM_MDELAY(1); + cik_irq_ack(rdev); + cik_disable_interrupt_state(rdev); +} + +/** + * cik_irq_disable - disable interrupts for suspend + * + * @rdev: radeon_device pointer + * + * Disable interrupts and stop the RLC (CIK). + * Used for suspend. + */ +static void cik_irq_suspend(struct radeon_device *rdev) +{ + cik_irq_disable(rdev); + cik_rlc_stop(rdev); +} + +/** + * cik_irq_fini - tear down interrupt support + * + * @rdev: radeon_device pointer + * + * Disable interrupts on the hw and free the IH ring + * buffer (CIK). + * Used for driver unload. + */ +static __unused void cik_irq_fini(struct radeon_device *rdev) +{ + cik_irq_suspend(rdev); + r600_ih_ring_fini(rdev); +} + +/** + * cik_get_ih_wptr - get the IH ring buffer wptr + * + * @rdev: radeon_device pointer + * + * Get the IH ring buffer wptr from either the register + * or the writeback memory buffer (CIK). Also check for + * ring buffer overflow and deal with it. + * Used by cik_irq_process(). + * Returns the value of the wptr. + */ +static inline u32 cik_get_ih_wptr(struct radeon_device *rdev) +{ + u32 wptr, tmp; + + if (rdev->wb.enabled) + wptr = le32_to_cpu(rdev->wb.wb[R600_WB_IH_WPTR_OFFSET/4]); + else + wptr = RREG32(IH_RB_WPTR); + + if (wptr & RB_OVERFLOW) { + /* When a ring buffer overflow happen start parsing interrupt + * from the last not overwritten vector (wptr + 16). Hopefully + * this should allow us to catchup. + */ + dev_warn(rdev->dev, "IH ring buffer overflow (0x%08X, %d, %d)\n", + wptr, rdev->ih.rptr, (wptr + 16) + rdev->ih.ptr_mask); + rdev->ih.rptr = (wptr + 16) & rdev->ih.ptr_mask; + tmp = RREG32(IH_RB_CNTL); + tmp |= IH_WPTR_OVERFLOW_CLEAR; + WREG32(IH_RB_CNTL, tmp); + } + return (wptr & rdev->ih.ptr_mask); +} + +/* CIK IV Ring + * Each IV ring entry is 128 bits: + * [7:0] - interrupt source id + * [31:8] - reserved + * [59:32] - interrupt source data + * [63:60] - reserved + * [71:64] - RINGID + * CP: + * ME_ID [1:0], PIPE_ID[1:0], QUEUE_ID[2:0] + * QUEUE_ID - for compute, which of the 8 queues owned by the dispatcher + * - for gfx, hw shader state (0=PS...5=LS, 6=CS) + * ME_ID - 0 = gfx, 1 = first 4 CS pipes, 2 = second 4 CS pipes + * PIPE_ID - ME0 0=3D + * - ME1&2 compute dispatcher (4 pipes each) + * SDMA: + * INSTANCE_ID [1:0], QUEUE_ID[1:0] + * INSTANCE_ID - 0 = sdma0, 1 = sdma1 + * QUEUE_ID - 0 = gfx, 1 = rlc0, 2 = rlc1 + * [79:72] - VMID + * [95:80] - PASID + * [127:96] - reserved + */ +/** + * cik_irq_process - interrupt handler + * + * @rdev: radeon_device pointer + * + * Interrupt hander (CIK). Walk the IH ring, + * ack interrupts and schedule work to handle + * interrupt events. + * Returns irq process return code. + */ +irqreturn_t cik_irq_process(struct radeon_device *rdev) +{ + struct radeon_ring *cp1_ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; + struct radeon_ring *cp2_ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; + u32 wptr; + u32 rptr; + u32 src_id, src_data, ring_id; + u8 me_id, pipe_id, queue_id; + u32 ring_index; + bool queue_hotplug = false; + bool queue_reset = false; + u32 addr, status, mc_client; + + if (!rdev->ih.enabled || rdev->shutdown) + return IRQ_NONE; + + wptr = cik_get_ih_wptr(rdev); + +restart_ih: + /* is somebody else already processing irqs? */ + if (atomic_xchg(&rdev->ih.lock, 1)) + return IRQ_NONE; + + rptr = rdev->ih.rptr; + DRM_DEBUG("cik_irq_process start: rptr %d, wptr %d\n", rptr, wptr); + + /* Order reading of wptr vs. reading of IH ring data */ + cpu_lfence(); + + /* display interrupts */ + cik_irq_ack(rdev); + + while (rptr != wptr) { + /* wptr/rptr are in bytes! */ + ring_index = rptr / 4; + src_id = le32_to_cpu(rdev->ih.ring[ring_index]) & 0xff; + src_data = le32_to_cpu(rdev->ih.ring[ring_index + 1]) & 0xfffffff; + ring_id = le32_to_cpu(rdev->ih.ring[ring_index + 2]) & 0xff; + + switch (src_id) { + case 1: /* D1 vblank/vline */ + switch (src_data) { + case 0: /* D1 vblank */ + if (rdev->irq.stat_regs.cik.disp_int & LB_D1_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[0]) { + drm_handle_vblank(rdev->ddev, 0); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[0])) + radeon_crtc_handle_flip(rdev, 0); + rdev->irq.stat_regs.cik.disp_int &= ~LB_D1_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D1 vblank\n"); + } + break; + case 1: /* D1 vline */ + if (rdev->irq.stat_regs.cik.disp_int & LB_D1_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int &= ~LB_D1_VLINE_INTERRUPT; + DRM_DEBUG("IH: D1 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 2: /* D2 vblank/vline */ + switch (src_data) { + case 0: /* D2 vblank */ + if (rdev->irq.stat_regs.cik.disp_int_cont & LB_D2_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[1]) { + drm_handle_vblank(rdev->ddev, 1); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[1])) + radeon_crtc_handle_flip(rdev, 1); + rdev->irq.stat_regs.cik.disp_int_cont &= ~LB_D2_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D2 vblank\n"); + } + break; + case 1: /* D2 vline */ + if (rdev->irq.stat_regs.cik.disp_int_cont & LB_D2_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont &= ~LB_D2_VLINE_INTERRUPT; + DRM_DEBUG("IH: D2 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 3: /* D3 vblank/vline */ + switch (src_data) { + case 0: /* D3 vblank */ + if (rdev->irq.stat_regs.cik.disp_int_cont2 & LB_D3_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[2]) { + drm_handle_vblank(rdev->ddev, 2); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[2])) + radeon_crtc_handle_flip(rdev, 2); + rdev->irq.stat_regs.cik.disp_int_cont2 &= ~LB_D3_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D3 vblank\n"); + } + break; + case 1: /* D3 vline */ + if (rdev->irq.stat_regs.cik.disp_int_cont2 & LB_D3_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont2 &= ~LB_D3_VLINE_INTERRUPT; + DRM_DEBUG("IH: D3 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 4: /* D4 vblank/vline */ + switch (src_data) { + case 0: /* D4 vblank */ + if (rdev->irq.stat_regs.cik.disp_int_cont3 & LB_D4_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[3]) { + drm_handle_vblank(rdev->ddev, 3); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[3])) + radeon_crtc_handle_flip(rdev, 3); + rdev->irq.stat_regs.cik.disp_int_cont3 &= ~LB_D4_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D4 vblank\n"); + } + break; + case 1: /* D4 vline */ + if (rdev->irq.stat_regs.cik.disp_int_cont3 & LB_D4_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont3 &= ~LB_D4_VLINE_INTERRUPT; + DRM_DEBUG("IH: D4 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 5: /* D5 vblank/vline */ + switch (src_data) { + case 0: /* D5 vblank */ + if (rdev->irq.stat_regs.cik.disp_int_cont4 & LB_D5_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[4]) { + drm_handle_vblank(rdev->ddev, 4); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[4])) + radeon_crtc_handle_flip(rdev, 4); + rdev->irq.stat_regs.cik.disp_int_cont4 &= ~LB_D5_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D5 vblank\n"); + } + break; + case 1: /* D5 vline */ + if (rdev->irq.stat_regs.cik.disp_int_cont4 & LB_D5_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont4 &= ~LB_D5_VLINE_INTERRUPT; + DRM_DEBUG("IH: D5 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 6: /* D6 vblank/vline */ + switch (src_data) { + case 0: /* D6 vblank */ + if (rdev->irq.stat_regs.cik.disp_int_cont5 & LB_D6_VBLANK_INTERRUPT) { + if (rdev->irq.crtc_vblank_int[5]) { + drm_handle_vblank(rdev->ddev, 5); + rdev->pm.vblank_sync = true; + wake_up(&rdev->irq.vblank_queue); + } + if (atomic_read(&rdev->irq.pflip[5])) + radeon_crtc_handle_flip(rdev, 5); + rdev->irq.stat_regs.cik.disp_int_cont5 &= ~LB_D6_VBLANK_INTERRUPT; + DRM_DEBUG("IH: D6 vblank\n"); + } + break; + case 1: /* D6 vline */ + if (rdev->irq.stat_regs.cik.disp_int_cont5 & LB_D6_VLINE_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont5 &= ~LB_D6_VLINE_INTERRUPT; + DRM_DEBUG("IH: D6 vline\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 42: /* HPD hotplug */ + switch (src_data) { + case 0: + if (rdev->irq.stat_regs.cik.disp_int & DC_HPD1_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int &= ~DC_HPD1_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD1\n"); + } + break; + case 1: + if (rdev->irq.stat_regs.cik.disp_int_cont & DC_HPD2_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont &= ~DC_HPD2_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD2\n"); + } + break; + case 2: + if (rdev->irq.stat_regs.cik.disp_int_cont2 & DC_HPD3_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont2 &= ~DC_HPD3_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD3\n"); + } + break; + case 3: + if (rdev->irq.stat_regs.cik.disp_int_cont3 & DC_HPD4_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont3 &= ~DC_HPD4_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD4\n"); + } + break; + case 4: + if (rdev->irq.stat_regs.cik.disp_int_cont4 & DC_HPD5_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont4 &= ~DC_HPD5_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD5\n"); + } + break; + case 5: + if (rdev->irq.stat_regs.cik.disp_int_cont5 & DC_HPD6_INTERRUPT) { + rdev->irq.stat_regs.cik.disp_int_cont5 &= ~DC_HPD6_INTERRUPT; + queue_hotplug = true; + DRM_DEBUG("IH: HPD6\n"); + } + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + break; + case 146: + case 147: + addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR); + status = RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS); + mc_client = RREG32(VM_CONTEXT1_PROTECTION_FAULT_MCCLIENT); + dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data); + dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n", + addr); + dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n", + status); + cik_vm_decode_fault(rdev, status, addr, mc_client); + /* reset addr and status */ + WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1); + break; + case 176: /* GFX RB CP_INT */ + case 177: /* GFX IB CP_INT */ + radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX); + break; + case 181: /* CP EOP event */ + DRM_DEBUG("IH: CP EOP\n"); + /* XXX check the bitfield order! */ + me_id = (ring_id & 0x60) >> 5; + pipe_id = (ring_id & 0x18) >> 3; + queue_id = (ring_id & 0x7) >> 0; + switch (me_id) { + case 0: + radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX); + break; + case 1: + case 2: + if ((cp1_ring->me == me_id) & (cp1_ring->pipe == pipe_id)) + radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP1_INDEX); + if ((cp2_ring->me == me_id) & (cp2_ring->pipe == pipe_id)) + radeon_fence_process(rdev, CAYMAN_RING_TYPE_CP2_INDEX); + break; + } + break; + case 184: /* CP Privileged reg access */ + DRM_ERROR("Illegal register access in command stream\n"); + /* XXX check the bitfield order! */ + me_id = (ring_id & 0x60) >> 5; + pipe_id = (ring_id & 0x18) >> 3; + queue_id = (ring_id & 0x7) >> 0; + switch (me_id) { + case 0: + /* This results in a full GPU reset, but all we need to do is soft + * reset the CP for gfx + */ + queue_reset = true; + break; + case 1: + /* XXX compute */ + queue_reset = true; + break; + case 2: + /* XXX compute */ + queue_reset = true; + break; + } + break; + case 185: /* CP Privileged inst */ + DRM_ERROR("Illegal instruction in command stream\n"); + /* XXX check the bitfield order! */ + me_id = (ring_id & 0x60) >> 5; + pipe_id = (ring_id & 0x18) >> 3; + queue_id = (ring_id & 0x7) >> 0; + switch (me_id) { + case 0: + /* This results in a full GPU reset, but all we need to do is soft + * reset the CP for gfx + */ + queue_reset = true; + break; + case 1: + /* XXX compute */ + queue_reset = true; + break; + case 2: + /* XXX compute */ + queue_reset = true; + break; + } + break; + case 224: /* SDMA trap event */ + /* XXX check the bitfield order! */ + me_id = (ring_id & 0x3) >> 0; + queue_id = (ring_id & 0xc) >> 2; + DRM_DEBUG("IH: SDMA trap\n"); + switch (me_id) { + case 0: + switch (queue_id) { + case 0: + radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX); + break; + case 1: + /* XXX compute */ + break; + case 2: + /* XXX compute */ + break; + } + break; + case 1: + switch (queue_id) { + case 0: + radeon_fence_process(rdev, CAYMAN_RING_TYPE_DMA1_INDEX); + break; + case 1: + /* XXX compute */ + break; + case 2: + /* XXX compute */ + break; + } + break; + } + break; + case 241: /* SDMA Privileged inst */ + case 247: /* SDMA Privileged inst */ + DRM_ERROR("Illegal instruction in SDMA command stream\n"); + /* XXX check the bitfield order! */ + me_id = (ring_id & 0x3) >> 0; + queue_id = (ring_id & 0xc) >> 2; + switch (me_id) { + case 0: + switch (queue_id) { + case 0: + queue_reset = true; + break; + case 1: + /* XXX compute */ + queue_reset = true; + break; + case 2: + /* XXX compute */ + queue_reset = true; + break; + } + break; + case 1: + switch (queue_id) { + case 0: + queue_reset = true; + break; + case 1: + /* XXX compute */ + queue_reset = true; + break; + case 2: + /* XXX compute */ + queue_reset = true; + break; + } + break; + } + break; + case 233: /* GUI IDLE */ + DRM_DEBUG("IH: GUI idle\n"); + break; + default: + DRM_DEBUG("Unhandled interrupt: %d %d\n", src_id, src_data); + break; + } + + /* wptr/rptr are in bytes! */ + rptr += 16; + rptr &= rdev->ih.ptr_mask; + } + if (queue_hotplug) + taskqueue_enqueue(rdev->tq, &rdev->hotplug_work); + if (queue_reset) + taskqueue_enqueue(rdev->tq, &rdev->reset_work); + rdev->ih.rptr = rptr; + WREG32(IH_RB_RPTR, rdev->ih.rptr); + atomic_set(&rdev->ih.lock, 0); + + /* make sure wptr hasn't changed while processing */ + wptr = cik_get_ih_wptr(rdev); + if (wptr != rptr) + goto restart_ih; + + return IRQ_HANDLED; +} + +/* + * startup/shutdown callbacks + */ +/** + * cik_startup - program the asic to a functional state + * + * @rdev: radeon_device pointer + * + * Programs the asic to a functional state (CIK). + * Called by cik_init() and cik_resume(). + * Returns 0 for success, error for failure. + */ +static int cik_startup(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int r; + + cik_mc_program(rdev); + + if (rdev->flags & RADEON_IS_IGP) { + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || + !rdev->mec_fw || !rdev->sdma_fw || !rdev->rlc_fw) { + r = cik_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + } else { + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || + !rdev->mec_fw || !rdev->sdma_fw || !rdev->rlc_fw || + !rdev->mc_fw) { + r = cik_init_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load firmware!\n"); + return r; + } + } + + r = ci_mc_load_microcode(rdev); + if (r) { + DRM_ERROR("Failed to load MC firmware!\n"); + return r; + } + } + + r = r600_vram_scratch_init(rdev); + if (r) + return r; + + r = cik_pcie_gart_enable(rdev); + if (r) + return r; + cik_gpu_init(rdev); + + /* allocate rlc buffers */ + r = si_rlc_init(rdev); + if (r) { + DRM_ERROR("Failed to init rlc BOs!\n"); + return r; + } + + /* allocate wb buffer */ + r = radeon_wb_init(rdev); + if (r) + return r; + + /* allocate mec buffers */ + r = cik_mec_init(rdev); + if (r) { + DRM_ERROR("Failed to init MEC BOs!\n"); + return r; + } + + r = radeon_fence_driver_start_ring(rdev, RADEON_RING_TYPE_GFX_INDEX); + if (r) { + dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r); + return r; + } + + r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_CP1_INDEX); + if (r) { + dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r); + return r; + } + + r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_CP2_INDEX); + if (r) { + dev_err(rdev->dev, "failed initializing CP fences (%d).\n", r); + return r; + } + + r = radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_DMA_INDEX); + if (r) { + dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r); + return r; + } + + r = radeon_fence_driver_start_ring(rdev, CAYMAN_RING_TYPE_DMA1_INDEX); + if (r) { + dev_err(rdev->dev, "failed initializing DMA fences (%d).\n", r); + return r; + } + + r = cik_uvd_resume(rdev); + if (!r) { + r = radeon_fence_driver_start_ring(rdev, + R600_RING_TYPE_UVD_INDEX); + if (r) + dev_err(rdev->dev, "UVD fences init error (%d).\n", r); + } + if (r) + rdev->ring[R600_RING_TYPE_UVD_INDEX].ring_size = 0; + + /* Enable IRQ */ + if (!rdev->irq.installed) { + r = radeon_irq_kms_init(rdev); + if (r) + return r; + } + + r = cik_irq_init(rdev); + if (r) { + DRM_ERROR("radeon: IH init failed (%d).\n", r); + radeon_irq_kms_fini(rdev); + return r; + } + cik_irq_set(rdev); + + ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; + r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP_RPTR_OFFSET, + CP_RB0_RPTR, CP_RB0_WPTR, + 0, 0xfffff, RADEON_CP_PACKET2); + if (r) + return r; + + /* set up the compute queues */ + /* type-2 packets are deprecated on MEC, use type-3 instead */ + ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; + r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP1_RPTR_OFFSET, + CP_HQD_PQ_RPTR, CP_HQD_PQ_WPTR, + 0, 0xfffff, PACKET3(PACKET3_NOP, 0x3FFF)); + if (r) + return r; + ring->me = 1; /* first MEC */ + ring->pipe = 0; /* first pipe */ + ring->queue = 0; /* first queue */ + ring->wptr_offs = CIK_WB_CP1_WPTR_OFFSET; + + /* type-2 packets are deprecated on MEC, use type-3 instead */ + ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; + r = radeon_ring_init(rdev, ring, ring->ring_size, RADEON_WB_CP2_RPTR_OFFSET, + CP_HQD_PQ_RPTR, CP_HQD_PQ_WPTR, + 0, 0xffffffff, PACKET3(PACKET3_NOP, 0x3FFF)); + if (r) + return r; + /* dGPU only have 1 MEC */ + ring->me = 1; /* first MEC */ + ring->pipe = 0; /* first pipe */ + ring->queue = 1; /* second queue */ + ring->wptr_offs = CIK_WB_CP2_WPTR_OFFSET; + + ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + r = radeon_ring_init(rdev, ring, ring->ring_size, R600_WB_DMA_RPTR_OFFSET, + SDMA0_GFX_RB_RPTR + SDMA0_REGISTER_OFFSET, + SDMA0_GFX_RB_WPTR + SDMA0_REGISTER_OFFSET, + 2, 0xfffffffc, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + if (r) + return r; + + ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; + r = radeon_ring_init(rdev, ring, ring->ring_size, CAYMAN_WB_DMA1_RPTR_OFFSET, + SDMA0_GFX_RB_RPTR + SDMA1_REGISTER_OFFSET, + SDMA0_GFX_RB_WPTR + SDMA1_REGISTER_OFFSET, + 2, 0xfffffffc, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0)); + if (r) + return r; + + r = cik_cp_resume(rdev); + if (r) + return r; + + r = cik_sdma_resume(rdev); + if (r) + return r; + + ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + if (ring->ring_size) { + r = radeon_ring_init(rdev, ring, ring->ring_size, + R600_WB_UVD_RPTR_OFFSET, + UVD_RBC_RB_RPTR, UVD_RBC_RB_WPTR, + 0, 0xfffff, RADEON_CP_PACKET2); + if (!r) + r = r600_uvd_init(rdev); + if (r) + DRM_ERROR("radeon: failed initializing UVD (%d).\n", r); + } + + r = radeon_ib_pool_init(rdev); + if (r) { + dev_err(rdev->dev, "IB initialization failed (%d).\n", r); + return r; + } + + r = radeon_vm_manager_init(rdev); + if (r) { + dev_err(rdev->dev, "vm manager initialization failed (%d).\n", r); + return r; + } + + return 0; +} + +/** + * cik_resume - resume the asic to a functional state + * + * @rdev: radeon_device pointer + * + * Programs the asic to a functional state (CIK). + * Called at resume. + * Returns 0 for success, error for failure. + */ +int cik_resume(struct radeon_device *rdev) +{ + int r; + + /* post card */ + atom_asic_init(rdev->mode_info.atom_context); + + /* init golden registers */ + cik_init_golden_registers(rdev); + + rdev->accel_working = true; + r = cik_startup(rdev); + if (r) { + DRM_ERROR("cik startup failed on resume\n"); + rdev->accel_working = false; + return r; + } + + return r; + +} + +/** + * cik_suspend - suspend the asic + * + * @rdev: radeon_device pointer + * + * Bring the chip into a state suitable for suspend (CIK). + * Called at suspend. + * Returns 0 for success. + */ +int cik_suspend(struct radeon_device *rdev) +{ + radeon_vm_manager_fini(rdev); + cik_cp_enable(rdev, false); + cik_sdma_enable(rdev, false); + r600_uvd_stop(rdev); + radeon_uvd_suspend(rdev); + cik_irq_suspend(rdev); + radeon_wb_disable(rdev); + cik_pcie_gart_disable(rdev); + return 0; +} + +/* Plan is to move initialization in that function and use + * helper function so that radeon_device_init pretty much + * do nothing more than calling asic specific function. This + * should also allow to remove a bunch of callback function + * like vram_info. + */ +/** + * cik_init - asic specific driver and hw init + * + * @rdev: radeon_device pointer + * + * Setup asic specific driver variables and program the hw + * to a functional state (CIK). + * Called at driver startup. + * Returns 0 for success, errors for failure. + */ +int cik_init(struct radeon_device *rdev) +{ + struct radeon_ring *ring; + int r; + + /* Read BIOS */ + if (!radeon_get_bios(rdev)) { + if (ASIC_IS_AVIVO(rdev)) + return -EINVAL; + } + /* Must be an ATOMBIOS */ + if (!rdev->is_atom_bios) { + dev_err(rdev->dev, "Expecting atombios for cayman GPU\n"); + return -EINVAL; + } + r = radeon_atombios_init(rdev); + if (r) + return r; + + /* Post card if necessary */ + if (!radeon_card_posted(rdev)) { + if (!rdev->bios) { + dev_err(rdev->dev, "Card not posted and no BIOS - ignoring\n"); + return -EINVAL; + } + DRM_INFO("GPU not posted. posting now...\n"); + atom_asic_init(rdev->mode_info.atom_context); + } + /* init golden registers */ + cik_init_golden_registers(rdev); + /* Initialize scratch registers */ + cik_scratch_init(rdev); + /* Initialize surface registers */ + radeon_surface_init(rdev); + /* Initialize clocks */ + radeon_get_clock_info(rdev->ddev); + + /* Fence driver */ + r = radeon_fence_driver_init(rdev); + if (r) + return r; + + /* initialize memory controller */ + r = cik_mc_init(rdev); + if (r) + return r; + /* Memory manager */ + r = radeon_bo_init(rdev); + if (r) + return r; + + ring = &rdev->ring[RADEON_RING_TYPE_GFX_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 1024 * 1024); + + ring = &rdev->ring[CAYMAN_RING_TYPE_CP1_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 1024 * 1024); + r = radeon_doorbell_get(rdev, &ring->doorbell_page_num); + if (r) + return r; + + ring = &rdev->ring[CAYMAN_RING_TYPE_CP2_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 1024 * 1024); + r = radeon_doorbell_get(rdev, &ring->doorbell_page_num); + if (r) + return r; + + ring = &rdev->ring[R600_RING_TYPE_DMA_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 256 * 1024); + + ring = &rdev->ring[CAYMAN_RING_TYPE_DMA1_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 256 * 1024); + + r = radeon_uvd_init(rdev); + if (!r) { + ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; + ring->ring_obj = NULL; + r600_ring_init(rdev, ring, 4096); + } + + rdev->ih.ring_obj = NULL; + r600_ih_ring_init(rdev, 64 * 1024); + + r = r600_pcie_gart_init(rdev); + if (r) + return r; + + rdev->accel_working = true; + r = cik_startup(rdev); + if (r) { + dev_err(rdev->dev, "disabling GPU acceleration\n"); + cik_cp_fini(rdev); + cik_sdma_fini(rdev); + cik_irq_fini(rdev); + si_rlc_fini(rdev); + cik_mec_fini(rdev); + radeon_wb_fini(rdev); + radeon_ib_pool_fini(rdev); + radeon_vm_manager_fini(rdev); + radeon_irq_kms_fini(rdev); + cik_pcie_gart_fini(rdev); + rdev->accel_working = false; + } + + /* Don't start up if the MC ucode is missing. + * The default clocks and voltages before the MC ucode + * is loaded are not suffient for advanced operations. + */ + if (!rdev->mc_fw && !(rdev->flags & RADEON_IS_IGP)) { + DRM_ERROR("radeon: MC ucode required for NI+.\n"); + return -EINVAL; + } + + return 0; +} + +/** + * cik_fini - asic specific driver and hw fini + * + * @rdev: radeon_device pointer + * + * Tear down the asic specific driver variables and program the hw + * to an idle state (CIK). + * Called at driver unload. + */ +void cik_fini(struct radeon_device *rdev) +{ + cik_cp_fini(rdev); + cik_sdma_fini(rdev); + cik_irq_fini(rdev); + si_rlc_fini(rdev); + cik_mec_fini(rdev); + radeon_wb_fini(rdev); + radeon_vm_manager_fini(rdev); + radeon_ib_pool_fini(rdev); + radeon_irq_kms_fini(rdev); + r600_uvd_stop(rdev); + radeon_uvd_fini(rdev); + cik_pcie_gart_fini(rdev); + r600_vram_scratch_fini(rdev); + radeon_gem_fini(rdev); + radeon_fence_driver_fini(rdev); + radeon_bo_fini(rdev); + radeon_atombios_fini(rdev); + kfree(rdev->bios); + rdev->bios = NULL; +} + +/* display watermark setup */ +/** + * dce8_line_buffer_adjust - Set up the line buffer + * + * @rdev: radeon_device pointer + * @radeon_crtc: the selected display controller + * @mode: the current display mode on the selected display + * controller + * + * Setup up the line buffer allocation for + * the selected display controller (CIK). + * Returns the line buffer size in pixels. + */ +static u32 dce8_line_buffer_adjust(struct radeon_device *rdev, + struct radeon_crtc *radeon_crtc, + struct drm_display_mode *mode) +{ + u32 tmp; + + /* + * Line Buffer Setup + * There are 6 line buffers, one for each display controllers. + * There are 3 partitions per LB. Select the number of partitions + * to enable based on the display width. For display widths larger + * than 4096, you need use to use 2 display controllers and combine + * them using the stereo blender. + */ + if (radeon_crtc->base.enabled && mode) { + if (mode->crtc_hdisplay < 1920) + tmp = 1; + else if (mode->crtc_hdisplay < 2560) + tmp = 2; + else if (mode->crtc_hdisplay < 4096) + tmp = 0; + else { + DRM_DEBUG_KMS("Mode too big for LB!\n"); + tmp = 0; + } + } else + tmp = 1; + + WREG32(LB_MEMORY_CTRL + radeon_crtc->crtc_offset, + LB_MEMORY_CONFIG(tmp) | LB_MEMORY_SIZE(0x6B0)); + + if (radeon_crtc->base.enabled && mode) { + switch (tmp) { + case 0: + default: + return 4096 * 2; + case 1: + return 1920 * 2; + case 2: + return 2560 * 2; + } + } + + /* controller not enabled, so no lb used */ + return 0; +} + +/** + * cik_get_number_of_dram_channels - get the number of dram channels + * + * @rdev: radeon_device pointer + * + * Look up the number of video ram channels (CIK). + * Used for display watermark bandwidth calculations + * Returns the number of dram channels + */ +static u32 cik_get_number_of_dram_channels(struct radeon_device *rdev) +{ + u32 tmp = RREG32(MC_SHARED_CHMAP); + + switch ((tmp & NOOFCHAN_MASK) >> NOOFCHAN_SHIFT) { + case 0: + default: + return 1; + case 1: + return 2; + case 2: + return 4; + case 3: + return 8; + case 4: + return 3; + case 5: + return 6; + case 6: + return 10; + case 7: + return 12; + case 8: + return 16; + } +} + +struct dce8_wm_params { + u32 dram_channels; /* number of dram channels */ + u32 yclk; /* bandwidth per dram data pin in kHz */ + u32 sclk; /* engine clock in kHz */ + u32 disp_clk; /* display clock in kHz */ + u32 src_width; /* viewport width */ + u32 active_time; /* active display time in ns */ + u32 blank_time; /* blank time in ns */ + bool interlaced; /* mode is interlaced */ + fixed20_12 vsc; /* vertical scale ratio */ + u32 num_heads; /* number of active crtcs */ + u32 bytes_per_pixel; /* bytes per pixel display + overlay */ + u32 lb_size; /* line buffer allocated to pipe */ + u32 vtaps; /* vertical scaler taps */ +}; + +/** + * dce8_dram_bandwidth - get the dram bandwidth + * + * @wm: watermark calculation data + * + * Calculate the raw dram bandwidth (CIK). + * Used for display watermark bandwidth calculations + * Returns the dram bandwidth in MBytes/s + */ +static u32 dce8_dram_bandwidth(struct dce8_wm_params *wm) +{ + /* Calculate raw DRAM Bandwidth */ + fixed20_12 dram_efficiency; /* 0.7 */ + fixed20_12 yclk, dram_channels, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + yclk.full = dfixed_const(wm->yclk); + yclk.full = dfixed_div(yclk, a); + dram_channels.full = dfixed_const(wm->dram_channels * 4); + a.full = dfixed_const(10); + dram_efficiency.full = dfixed_const(7); + dram_efficiency.full = dfixed_div(dram_efficiency, a); + bandwidth.full = dfixed_mul(dram_channels, yclk); + bandwidth.full = dfixed_mul(bandwidth, dram_efficiency); + + return dfixed_trunc(bandwidth); +} + +/** + * dce8_dram_bandwidth_for_display - get the dram bandwidth for display + * + * @wm: watermark calculation data + * + * Calculate the dram bandwidth used for display (CIK). + * Used for display watermark bandwidth calculations + * Returns the dram bandwidth for display in MBytes/s + */ +static u32 dce8_dram_bandwidth_for_display(struct dce8_wm_params *wm) +{ + /* Calculate DRAM Bandwidth and the part allocated to display. */ + fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */ + fixed20_12 yclk, dram_channels, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + yclk.full = dfixed_const(wm->yclk); + yclk.full = dfixed_div(yclk, a); + dram_channels.full = dfixed_const(wm->dram_channels * 4); + a.full = dfixed_const(10); + disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */ + disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a); + bandwidth.full = dfixed_mul(dram_channels, yclk); + bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation); + + return dfixed_trunc(bandwidth); +} + +/** + * dce8_data_return_bandwidth - get the data return bandwidth + * + * @wm: watermark calculation data + * + * Calculate the data return bandwidth used for display (CIK). + * Used for display watermark bandwidth calculations + * Returns the data return bandwidth in MBytes/s + */ +static u32 dce8_data_return_bandwidth(struct dce8_wm_params *wm) +{ + /* Calculate the display Data return Bandwidth */ + fixed20_12 return_efficiency; /* 0.8 */ + fixed20_12 sclk, bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + sclk.full = dfixed_const(wm->sclk); + sclk.full = dfixed_div(sclk, a); + a.full = dfixed_const(10); + return_efficiency.full = dfixed_const(8); + return_efficiency.full = dfixed_div(return_efficiency, a); + a.full = dfixed_const(32); + bandwidth.full = dfixed_mul(a, sclk); + bandwidth.full = dfixed_mul(bandwidth, return_efficiency); + + return dfixed_trunc(bandwidth); +} + +/** + * dce8_dmif_request_bandwidth - get the dmif bandwidth + * + * @wm: watermark calculation data + * + * Calculate the dmif bandwidth used for display (CIK). + * Used for display watermark bandwidth calculations + * Returns the dmif bandwidth in MBytes/s + */ +static u32 dce8_dmif_request_bandwidth(struct dce8_wm_params *wm) +{ + /* Calculate the DMIF Request Bandwidth */ + fixed20_12 disp_clk_request_efficiency; /* 0.8 */ + fixed20_12 disp_clk, bandwidth; + fixed20_12 a, b; + + a.full = dfixed_const(1000); + disp_clk.full = dfixed_const(wm->disp_clk); + disp_clk.full = dfixed_div(disp_clk, a); + a.full = dfixed_const(32); + b.full = dfixed_mul(a, disp_clk); + + a.full = dfixed_const(10); + disp_clk_request_efficiency.full = dfixed_const(8); + disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a); + + bandwidth.full = dfixed_mul(b, disp_clk_request_efficiency); + + return dfixed_trunc(bandwidth); +} + +/** + * dce8_available_bandwidth - get the min available bandwidth + * + * @wm: watermark calculation data + * + * Calculate the min available bandwidth used for display (CIK). + * Used for display watermark bandwidth calculations + * Returns the min available bandwidth in MBytes/s + */ +static u32 dce8_available_bandwidth(struct dce8_wm_params *wm) +{ + /* Calculate the Available bandwidth. Display can use this temporarily but not in average. */ + u32 dram_bandwidth = dce8_dram_bandwidth(wm); + u32 data_return_bandwidth = dce8_data_return_bandwidth(wm); + u32 dmif_req_bandwidth = dce8_dmif_request_bandwidth(wm); + + return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth)); +} + +/** + * dce8_average_bandwidth - get the average available bandwidth + * + * @wm: watermark calculation data + * + * Calculate the average available bandwidth used for display (CIK). + * Used for display watermark bandwidth calculations + * Returns the average available bandwidth in MBytes/s + */ +static u32 dce8_average_bandwidth(struct dce8_wm_params *wm) +{ + /* Calculate the display mode Average Bandwidth + * DisplayMode should contain the source and destination dimensions, + * timing, etc. + */ + fixed20_12 bpp; + fixed20_12 line_time; + fixed20_12 src_width; + fixed20_12 bandwidth; + fixed20_12 a; + + a.full = dfixed_const(1000); + line_time.full = dfixed_const(wm->active_time + wm->blank_time); + line_time.full = dfixed_div(line_time, a); + bpp.full = dfixed_const(wm->bytes_per_pixel); + src_width.full = dfixed_const(wm->src_width); + bandwidth.full = dfixed_mul(src_width, bpp); + bandwidth.full = dfixed_mul(bandwidth, wm->vsc); + bandwidth.full = dfixed_div(bandwidth, line_time); + + return dfixed_trunc(bandwidth); +} + +/** + * dce8_latency_watermark - get the latency watermark + * + * @wm: watermark calculation data + * + * Calculate the latency watermark (CIK). + * Used for display watermark bandwidth calculations + * Returns the latency watermark in ns + */ +static u32 dce8_latency_watermark(struct dce8_wm_params *wm) +{ + /* First calculate the latency in ns */ + u32 mc_latency = 2000; /* 2000 ns. */ + u32 available_bandwidth = dce8_available_bandwidth(wm); + u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth; + u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth; + u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */ + u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) + + (wm->num_heads * cursor_line_pair_return_time); + u32 latency = mc_latency + other_heads_data_return_time + dc_latency; + u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time; + u32 tmp, dmif_size = 12288; + fixed20_12 a, b, c; + + if (wm->num_heads == 0) + return 0; + + a.full = dfixed_const(2); + b.full = dfixed_const(1); + if ((wm->vsc.full > a.full) || + ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) || + (wm->vtaps >= 5) || + ((wm->vsc.full >= a.full) && wm->interlaced)) + max_src_lines_per_dst_line = 4; + else + max_src_lines_per_dst_line = 2; + + a.full = dfixed_const(available_bandwidth); + b.full = dfixed_const(wm->num_heads); + a.full = dfixed_div(a, b); + + b.full = dfixed_const(mc_latency + 512); + c.full = dfixed_const(wm->disp_clk); + b.full = dfixed_div(b, c); + + c.full = dfixed_const(dmif_size); + b.full = dfixed_div(c, b); + + tmp = min(dfixed_trunc(a), dfixed_trunc(b)); + + b.full = dfixed_const(1000); + c.full = dfixed_const(wm->disp_clk); + b.full = dfixed_div(c, b); + c.full = dfixed_const(wm->bytes_per_pixel); + b.full = dfixed_mul(b, c); + + lb_fill_bw = min(tmp, dfixed_trunc(b)); + + a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); + b.full = dfixed_const(1000); + c.full = dfixed_const(lb_fill_bw); + b.full = dfixed_div(c, b); + a.full = dfixed_div(a, b); + line_fill_time = dfixed_trunc(a); + + if (line_fill_time < wm->active_time) + return latency; + else + return latency + (line_fill_time - wm->active_time); + +} + +/** + * dce8_average_bandwidth_vs_dram_bandwidth_for_display - check + * average and available dram bandwidth + * + * @wm: watermark calculation data + * + * Check if the display average bandwidth fits in the display + * dram bandwidth (CIK). + * Used for display watermark bandwidth calculations + * Returns true if the display fits, false if not. + */ +static bool dce8_average_bandwidth_vs_dram_bandwidth_for_display(struct dce8_wm_params *wm) +{ + if (dce8_average_bandwidth(wm) <= + (dce8_dram_bandwidth_for_display(wm) / wm->num_heads)) + return true; + else + return false; +} + +/** + * dce8_average_bandwidth_vs_available_bandwidth - check + * average and available bandwidth + * + * @wm: watermark calculation data + * + * Check if the display average bandwidth fits in the display + * available bandwidth (CIK). + * Used for display watermark bandwidth calculations + * Returns true if the display fits, false if not. + */ +static bool dce8_average_bandwidth_vs_available_bandwidth(struct dce8_wm_params *wm) +{ + if (dce8_average_bandwidth(wm) <= + (dce8_available_bandwidth(wm) / wm->num_heads)) + return true; + else + return false; +} + +/** + * dce8_check_latency_hiding - check latency hiding + * + * @wm: watermark calculation data + * + * Check latency hiding (CIK). + * Used for display watermark bandwidth calculations + * Returns true if the display fits, false if not. + */ +static bool dce8_check_latency_hiding(struct dce8_wm_params *wm) +{ + u32 lb_partitions = wm->lb_size / wm->src_width; + u32 line_time = wm->active_time + wm->blank_time; + u32 latency_tolerant_lines; + u32 latency_hiding; + fixed20_12 a; + + a.full = dfixed_const(1); + if (wm->vsc.full > a.full) + latency_tolerant_lines = 1; + else { + if (lb_partitions <= (wm->vtaps + 1)) + latency_tolerant_lines = 1; + else + latency_tolerant_lines = 2; + } + + latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time); + + if (dce8_latency_watermark(wm) <= latency_hiding) + return true; + else + return false; +} + +/** + * dce8_program_watermarks - program display watermarks + * + * @rdev: radeon_device pointer + * @radeon_crtc: the selected display controller + * @lb_size: line buffer size + * @num_heads: number of display controllers in use + * + * Calculate and program the display watermarks for the + * selected display controller (CIK). + */ +static void dce8_program_watermarks(struct radeon_device *rdev, + struct radeon_crtc *radeon_crtc, + u32 lb_size, u32 num_heads) +{ + struct drm_display_mode *mode = &radeon_crtc->base.mode; + struct dce8_wm_params wm; + u32 pixel_period; + u32 line_time = 0; + u32 latency_watermark_a = 0, latency_watermark_b = 0; + u32 tmp, wm_mask; + + if (radeon_crtc->base.enabled && num_heads && mode) { + pixel_period = 1000000 / (u32)mode->clock; + line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535); + + wm.yclk = rdev->pm.current_mclk * 10; + wm.sclk = rdev->pm.current_sclk * 10; + wm.disp_clk = mode->clock; + wm.src_width = mode->crtc_hdisplay; + wm.active_time = mode->crtc_hdisplay * pixel_period; + wm.blank_time = line_time - wm.active_time; + wm.interlaced = false; + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + wm.interlaced = true; + wm.vsc = radeon_crtc->vsc; + wm.vtaps = 1; + if (radeon_crtc->rmx_type != RMX_OFF) + wm.vtaps = 2; + wm.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm.lb_size = lb_size; + wm.dram_channels = cik_get_number_of_dram_channels(rdev); + wm.num_heads = num_heads; + + /* set for high clocks */ + latency_watermark_a = min(dce8_latency_watermark(&wm), (u32)65535); + /* set for low clocks */ + /* wm.yclk = low clk; wm.sclk = low clk */ + latency_watermark_b = min(dce8_latency_watermark(&wm), (u32)65535); + + /* possibly force display priority to high */ + /* should really do this at mode validation time... */ + if (!dce8_average_bandwidth_vs_dram_bandwidth_for_display(&wm) || + !dce8_average_bandwidth_vs_available_bandwidth(&wm) || + !dce8_check_latency_hiding(&wm) || + (rdev->disp_priority == 2)) { + DRM_DEBUG_KMS("force priority to high\n"); + } + } + + /* select wm A */ + wm_mask = RREG32(DPG_WATERMARK_MASK_CONTROL + radeon_crtc->crtc_offset); + tmp = wm_mask; + tmp &= ~LATENCY_WATERMARK_MASK(3); + tmp |= LATENCY_WATERMARK_MASK(1); + WREG32(DPG_WATERMARK_MASK_CONTROL + radeon_crtc->crtc_offset, tmp); + WREG32(DPG_PIPE_LATENCY_CONTROL + radeon_crtc->crtc_offset, + (LATENCY_LOW_WATERMARK(latency_watermark_a) | + LATENCY_HIGH_WATERMARK(line_time))); + /* select wm B */ + tmp = RREG32(DPG_WATERMARK_MASK_CONTROL + radeon_crtc->crtc_offset); + tmp &= ~LATENCY_WATERMARK_MASK(3); + tmp |= LATENCY_WATERMARK_MASK(2); + WREG32(DPG_WATERMARK_MASK_CONTROL + radeon_crtc->crtc_offset, tmp); + WREG32(DPG_PIPE_LATENCY_CONTROL + radeon_crtc->crtc_offset, + (LATENCY_LOW_WATERMARK(latency_watermark_b) | + LATENCY_HIGH_WATERMARK(line_time))); + /* restore original selection */ + WREG32(DPG_WATERMARK_MASK_CONTROL + radeon_crtc->crtc_offset, wm_mask); +} + +/** + * dce8_bandwidth_update - program display watermarks + * + * @rdev: radeon_device pointer + * + * Calculate and program the display watermarks and line + * buffer allocation (CIK). + */ +void dce8_bandwidth_update(struct radeon_device *rdev) +{ + struct drm_display_mode *mode = NULL; + u32 num_heads = 0, lb_size; + int i; + + radeon_update_display_priority(rdev); + + for (i = 0; i < rdev->num_crtc; i++) { + if (rdev->mode_info.crtcs[i]->base.enabled) + num_heads++; + } + for (i = 0; i < rdev->num_crtc; i++) { + mode = &rdev->mode_info.crtcs[i]->base.mode; + lb_size = dce8_line_buffer_adjust(rdev, rdev->mode_info.crtcs[i], mode); + dce8_program_watermarks(rdev, rdev->mode_info.crtcs[i], lb_size, num_heads); + } +} + +/** + * cik_get_gpu_clock_counter - return GPU clock counter snapshot + * + * @rdev: radeon_device pointer + * + * Fetches a GPU clock counter snapshot (SI). + * Returns the 64 bit clock counter snapshot. + */ +uint64_t cik_get_gpu_clock_counter(struct radeon_device *rdev) +{ + uint64_t clock; + + spin_lock(&rdev->gpu_clock_mutex); + WREG32(RLC_CAPTURE_GPU_CLOCK_COUNT, 1); + clock = (uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_LSB) | + ((uint64_t)RREG32(RLC_GPU_CLOCK_COUNT_MSB) << 32ULL); + spin_unlock(&rdev->gpu_clock_mutex); + return clock; +} + +static int cik_set_uvd_clock(struct radeon_device *rdev, u32 clock, + u32 cntl_reg, u32 status_reg) +{ + int r, i; + struct atom_clock_dividers dividers; + uint32_t tmp; + + r = radeon_atom_get_clock_dividers(rdev, COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, + clock, false, ÷rs); + if (r) + return r; + + tmp = RREG32_SMC(cntl_reg); + tmp &= ~(DCLK_DIR_CNTL_EN|DCLK_DIVIDER_MASK); + tmp |= dividers.post_divider; + WREG32_SMC(cntl_reg, tmp); + + for (i = 0; i < 100; i++) { + if (RREG32_SMC(status_reg) & DCLK_STATUS) + break; + DRM_MDELAY(10); + } + if (i == 100) + return -ETIMEDOUT; + + return 0; +} + +int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) +{ + int r = 0; + + r = cik_set_uvd_clock(rdev, vclk, CG_VCLK_CNTL, CG_VCLK_STATUS); + if (r) + return r; + + r = cik_set_uvd_clock(rdev, dclk, CG_DCLK_CNTL, CG_DCLK_STATUS); + return r; +} + +int cik_uvd_resume(struct radeon_device *rdev) +{ + uint64_t addr; + uint32_t size; + int r; + + r = radeon_uvd_resume(rdev); + if (r) + return r; + + /* programm the VCPU memory controller bits 0-27 */ + addr = rdev->uvd.gpu_addr >> 3; + size = RADEON_GPU_PAGE_ALIGN(rdev->uvd_fw->datasize + 4) >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET0, addr); + WREG32(UVD_VCPU_CACHE_SIZE0, size); + + addr += size; + size = RADEON_UVD_STACK_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET1, addr); + WREG32(UVD_VCPU_CACHE_SIZE1, size); + + addr += size; + size = RADEON_UVD_HEAP_SIZE >> 3; + WREG32(UVD_VCPU_CACHE_OFFSET2, addr); + WREG32(UVD_VCPU_CACHE_SIZE2, size); + + /* bits 28-31 */ + addr = (rdev->uvd.gpu_addr >> 28) & 0xF; + WREG32(UVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); + + /* bits 32-39 */ + addr = (rdev->uvd.gpu_addr >> 32) & 0xFF; + WREG32(UVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); + + return 0; +} diff --git a/sys/dev/drm/radeon/si_blit_shaders.c b/sys/dev/drm/radeon/cik_blit_shaders.c similarity index 93% copy from sys/dev/drm/radeon/si_blit_shaders.c copy to sys/dev/drm/radeon/cik_blit_shaders.c index 6798cbdde4..ff1311806e 100644 --- a/sys/dev/drm/radeon/si_blit_shaders.c +++ b/sys/dev/drm/radeon/cik_blit_shaders.c @@ -1,5 +1,5 @@ /* - * Copyright 2011 Advanced Micro Devices, Inc. + * Copyright 2012 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -22,13 +22,13 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/si_blit_shaders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ -#include +#include +#include +#include -const u32 si_default_state[] = +const u32 cik_default_state[] = { 0xc0066900, 0x00000000, @@ -109,11 +109,6 @@ const u32 si_default_state[] = 0x00000000, /* PA_SC_VPORT_ZMIN_0 */ 0x3f800000, /* PA_SC_VPORT_ZMAX_0 */ - 0xc0026900, - 0x000000d9, - 0x00000000, /* CP_RINGID */ - 0x00000000, /* CP_VMID */ - 0xc0046900, 0x00000100, 0xffffffff, /* VGT_MAX_VTX_INDX */ @@ -132,7 +127,7 @@ const u32 si_default_state[] = 0x000001e0, 0x00000000, /* CB_BLEND0_CONTROL */ - 0xc00e6900, + 0xc00c6900, 0x00000200, 0x00000000, /* DB_DEPTH_CONTROL */ 0x00000000, /* DB_EQAA */ @@ -146,8 +141,6 @@ const u32 si_default_state[] = 0x00000000, /* PA_SU_LINE_STIPPLE_CNTL */ 0x00000000, /* PA_SU_LINE_STIPPLE_SCALE */ 0x00000000, /* PA_SU_PRIM_FILTER_CNTL */ - 0x00000000, /* */ - 0x00000000, /* */ 0xc0116900, 0x00000280, @@ -250,4 +243,4 @@ const u32 si_default_state[] = 0x00000010, /* */ }; -const u32 si_default_size = DRM_ARRAY_SIZE(si_default_state); +const u32 cik_default_size = ARRAY_SIZE(cik_default_state); diff --git a/sys/dev/drm/radeon/si_blit_shaders.h b/sys/dev/drm/radeon/cik_blit_shaders.h similarity index 81% copy from sys/dev/drm/radeon/si_blit_shaders.h copy to sys/dev/drm/radeon/cik_blit_shaders.h index 24de9c413f..dfe7314f9f 100644 --- a/sys/dev/drm/radeon/si_blit_shaders.h +++ b/sys/dev/drm/radeon/cik_blit_shaders.h @@ -1,5 +1,5 @@ /* - * Copyright 2011 Advanced Micro Devices, Inc. + * Copyright 2012 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,14 +20,13 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/si_blit_shaders.h 254885 2013-08-25 19:37:15Z dumbbell $ */ -#ifndef SI_BLIT_SHADERS_H -#define SI_BLIT_SHADERS_H +#ifndef CIK_BLIT_SHADERS_H +#define CIK_BLIT_SHADERS_H -extern const u32 si_default_state[]; +extern const u32 cik_default_state[]; -extern const u32 si_default_size; +extern const u32 cik_default_size; #endif diff --git a/sys/dev/drm/radeon/cik_reg.h b/sys/dev/drm/radeon/cik_reg.h new file mode 100644 index 0000000000..d71e46d571 --- /dev/null +++ b/sys/dev/drm/radeon/cik_reg.h @@ -0,0 +1,147 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#ifndef __CIK_REG_H__ +#define __CIK_REG_H__ + +#define CIK_DC_GPIO_HPD_MASK 0x65b0 +#define CIK_DC_GPIO_HPD_A 0x65b4 +#define CIK_DC_GPIO_HPD_EN 0x65b8 +#define CIK_DC_GPIO_HPD_Y 0x65bc + +#define CIK_GRPH_CONTROL 0x6804 +# define CIK_GRPH_DEPTH(x) (((x) & 0x3) << 0) +# define CIK_GRPH_DEPTH_8BPP 0 +# define CIK_GRPH_DEPTH_16BPP 1 +# define CIK_GRPH_DEPTH_32BPP 2 +# define CIK_GRPH_NUM_BANKS(x) (((x) & 0x3) << 2) +# define CIK_ADDR_SURF_2_BANK 0 +# define CIK_ADDR_SURF_4_BANK 1 +# define CIK_ADDR_SURF_8_BANK 2 +# define CIK_ADDR_SURF_16_BANK 3 +# define CIK_GRPH_Z(x) (((x) & 0x3) << 4) +# define CIK_GRPH_BANK_WIDTH(x) (((x) & 0x3) << 6) +# define CIK_ADDR_SURF_BANK_WIDTH_1 0 +# define CIK_ADDR_SURF_BANK_WIDTH_2 1 +# define CIK_ADDR_SURF_BANK_WIDTH_4 2 +# define CIK_ADDR_SURF_BANK_WIDTH_8 3 +# define CIK_GRPH_FORMAT(x) (((x) & 0x7) << 8) +/* 8 BPP */ +# define CIK_GRPH_FORMAT_INDEXED 0 +/* 16 BPP */ +# define CIK_GRPH_FORMAT_ARGB1555 0 +# define CIK_GRPH_FORMAT_ARGB565 1 +# define CIK_GRPH_FORMAT_ARGB4444 2 +# define CIK_GRPH_FORMAT_AI88 3 +# define CIK_GRPH_FORMAT_MONO16 4 +# define CIK_GRPH_FORMAT_BGRA5551 5 +/* 32 BPP */ +# define CIK_GRPH_FORMAT_ARGB8888 0 +# define CIK_GRPH_FORMAT_ARGB2101010 1 +# define CIK_GRPH_FORMAT_32BPP_DIG 2 +# define CIK_GRPH_FORMAT_8B_ARGB2101010 3 +# define CIK_GRPH_FORMAT_BGRA1010102 4 +# define CIK_GRPH_FORMAT_8B_BGRA1010102 5 +# define CIK_GRPH_FORMAT_RGB111110 6 +# define CIK_GRPH_FORMAT_BGR101111 7 +# define CIK_GRPH_BANK_HEIGHT(x) (((x) & 0x3) << 11) +# define CIK_ADDR_SURF_BANK_HEIGHT_1 0 +# define CIK_ADDR_SURF_BANK_HEIGHT_2 1 +# define CIK_ADDR_SURF_BANK_HEIGHT_4 2 +# define CIK_ADDR_SURF_BANK_HEIGHT_8 3 +# define CIK_GRPH_TILE_SPLIT(x) (((x) & 0x7) << 13) +# define CIK_ADDR_SURF_TILE_SPLIT_64B 0 +# define CIK_ADDR_SURF_TILE_SPLIT_128B 1 +# define CIK_ADDR_SURF_TILE_SPLIT_256B 2 +# define CIK_ADDR_SURF_TILE_SPLIT_512B 3 +# define CIK_ADDR_SURF_TILE_SPLIT_1KB 4 +# define CIK_ADDR_SURF_TILE_SPLIT_2KB 5 +# define CIK_ADDR_SURF_TILE_SPLIT_4KB 6 +# define CIK_GRPH_MACRO_TILE_ASPECT(x) (((x) & 0x3) << 18) +# define CIK_ADDR_SURF_MACRO_TILE_ASPECT_1 0 +# define CIK_ADDR_SURF_MACRO_TILE_ASPECT_2 1 +# define CIK_ADDR_SURF_MACRO_TILE_ASPECT_4 2 +# define CIK_ADDR_SURF_MACRO_TILE_ASPECT_8 3 +# define CIK_GRPH_ARRAY_MODE(x) (((x) & 0x7) << 20) +# define CIK_GRPH_ARRAY_LINEAR_GENERAL 0 +# define CIK_GRPH_ARRAY_LINEAR_ALIGNED 1 +# define CIK_GRPH_ARRAY_1D_TILED_THIN1 2 +# define CIK_GRPH_ARRAY_2D_TILED_THIN1 4 +# define CIK_GRPH_PIPE_CONFIG(x) (((x) & 0x1f) << 24) +# define CIK_ADDR_SURF_P2 0 +# define CIK_ADDR_SURF_P4_8x16 4 +# define CIK_ADDR_SURF_P4_16x16 5 +# define CIK_ADDR_SURF_P4_16x32 6 +# define CIK_ADDR_SURF_P4_32x32 7 +# define CIK_ADDR_SURF_P8_16x16_8x16 8 +# define CIK_ADDR_SURF_P8_16x32_8x16 9 +# define CIK_ADDR_SURF_P8_32x32_8x16 10 +# define CIK_ADDR_SURF_P8_16x32_16x16 11 +# define CIK_ADDR_SURF_P8_32x32_16x16 12 +# define CIK_ADDR_SURF_P8_32x32_16x32 13 +# define CIK_ADDR_SURF_P8_32x64_32x32 14 +# define CIK_GRPH_MICRO_TILE_MODE(x) (((x) & 0x7) << 29) +# define CIK_DISPLAY_MICRO_TILING 0 +# define CIK_THIN_MICRO_TILING 1 +# define CIK_DEPTH_MICRO_TILING 2 +# define CIK_ROTATED_MICRO_TILING 4 + +/* CUR blocks at 0x6998, 0x7598, 0x10198, 0x10d98, 0x11998, 0x12598 */ +#define CIK_CUR_CONTROL 0x6998 +# define CIK_CURSOR_EN (1 << 0) +# define CIK_CURSOR_MODE(x) (((x) & 0x3) << 8) +# define CIK_CURSOR_MONO 0 +# define CIK_CURSOR_24_1 1 +# define CIK_CURSOR_24_8_PRE_MULT 2 +# define CIK_CURSOR_24_8_UNPRE_MULT 3 +# define CIK_CURSOR_2X_MAGNIFY (1 << 16) +# define CIK_CURSOR_FORCE_MC_ON (1 << 20) +# define CIK_CURSOR_URGENT_CONTROL(x) (((x) & 0x7) << 24) +# define CIK_CURSOR_URGENT_ALWAYS 0 +# define CIK_CURSOR_URGENT_1_8 1 +# define CIK_CURSOR_URGENT_1_4 2 +# define CIK_CURSOR_URGENT_3_8 3 +# define CIK_CURSOR_URGENT_1_2 4 +#define CIK_CUR_SURFACE_ADDRESS 0x699c +# define CIK_CUR_SURFACE_ADDRESS_MASK 0xfffff000 +#define CIK_CUR_SIZE 0x69a0 +#define CIK_CUR_SURFACE_ADDRESS_HIGH 0x69a4 +#define CIK_CUR_POSITION 0x69a8 +#define CIK_CUR_HOT_SPOT 0x69ac +#define CIK_CUR_COLOR1 0x69b0 +#define CIK_CUR_COLOR2 0x69b4 +#define CIK_CUR_UPDATE 0x69b8 +# define CIK_CURSOR_UPDATE_PENDING (1 << 0) +# define CIK_CURSOR_UPDATE_TAKEN (1 << 1) +# define CIK_CURSOR_UPDATE_LOCK (1 << 16) +# define CIK_CURSOR_DISABLE_MULTIPLE_UPDATE (1 << 24) + +#define CIK_ALPHA_CONTROL 0x6af0 +# define CIK_CURSOR_ALPHA_BLND_ENA (1 << 1) + +#define CIK_LB_DATA_FORMAT 0x6b00 +# define CIK_INTERLEAVE_EN (1 << 3) + +#define CIK_LB_DESKTOP_HEIGHT 0x6b0c + +#endif diff --git a/sys/dev/drm/radeon/sid.h b/sys/dev/drm/radeon/cikd.h similarity index 57% copy from sys/dev/drm/radeon/sid.h copy to sys/dev/drm/radeon/cikd.h index b0003c1640..7e9275eaef 100644 --- a/sys/dev/drm/radeon/sid.h +++ b/sys/dev/drm/radeon/cikd.h @@ -1,5 +1,5 @@ /* - * Copyright 2011 Advanced Micro Devices, Inc. + * Copyright 2012 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,86 +20,40 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/sid.h 254885 2013-08-25 19:37:15Z dumbbell $ */ -#ifndef SI_H -#define SI_H - -#define TAHITI_RB_BITMAP_WIDTH_PER_SH 2 - -#define TAHITI_GB_ADDR_CONFIG_GOLDEN 0x12011003 -#define VERDE_GB_ADDR_CONFIG_GOLDEN 0x12010002 -#define HAINAN_GB_ADDR_CONFIG_GOLDEN 0x02010001 - -/* discrete uvd clocks */ -#define CG_UPLL_FUNC_CNTL 0x634 -# define UPLL_RESET_MASK 0x00000001 -# define UPLL_SLEEP_MASK 0x00000002 -# define UPLL_BYPASS_EN_MASK 0x00000004 -# define UPLL_CTLREQ_MASK 0x00000008 -# define UPLL_VCO_MODE_MASK 0x00000600 -# define UPLL_REF_DIV_MASK 0x003F0000 -# define UPLL_CTLACK_MASK 0x40000000 -# define UPLL_CTLACK2_MASK 0x80000000 -#define CG_UPLL_FUNC_CNTL_2 0x638 -# define UPLL_PDIV_A(x) ((x) << 0) -# define UPLL_PDIV_A_MASK 0x0000007F -# define UPLL_PDIV_B(x) ((x) << 8) -# define UPLL_PDIV_B_MASK 0x00007F00 -# define VCLK_SRC_SEL(x) ((x) << 20) -# define VCLK_SRC_SEL_MASK 0x01F00000 -# define DCLK_SRC_SEL(x) ((x) << 25) -# define DCLK_SRC_SEL_MASK 0x3E000000 -#define CG_UPLL_FUNC_CNTL_3 0x63C -# define UPLL_FB_DIV(x) ((x) << 0) -# define UPLL_FB_DIV_MASK 0x01FFFFFF -#define CG_UPLL_FUNC_CNTL_4 0x644 -# define UPLL_SPARE_ISPARE9 0x00020000 -#define CG_UPLL_FUNC_CNTL_5 0x648 -# define RESET_ANTI_MUX_MASK 0x00000200 -#define CG_UPLL_SPREAD_SPECTRUM 0x650 -# define SSEN_MASK 0x00000001 - -#define CG_MULT_THERMAL_STATUS 0x714 -#define ASIC_MAX_TEMP(x) ((x) << 0) -#define ASIC_MAX_TEMP_MASK 0x000001ff -#define ASIC_MAX_TEMP_SHIFT 0 -#define CTF_TEMP(x) ((x) << 9) -#define CTF_TEMP_MASK 0x0003fe00 -#define CTF_TEMP_SHIFT 9 - -#define SI_MAX_SH_GPRS 256 -#define SI_MAX_TEMP_GPRS 16 -#define SI_MAX_SH_THREADS 256 -#define SI_MAX_SH_STACK_ENTRIES 4096 -#define SI_MAX_FRC_EOV_CNT 16384 -#define SI_MAX_BACKENDS 8 -#define SI_MAX_BACKENDS_MASK 0xFF -#define SI_MAX_BACKENDS_PER_SE_MASK 0x0F -#define SI_MAX_SIMDS 12 -#define SI_MAX_SIMDS_MASK 0x0FFF -#define SI_MAX_SIMDS_PER_SE_MASK 0x00FF -#define SI_MAX_PIPES 8 -#define SI_MAX_PIPES_MASK 0xFF -#define SI_MAX_PIPES_PER_SIMD_MASK 0x3F -#define SI_MAX_LDS_NUM 0xFFFF -#define SI_MAX_TCC 16 -#define SI_MAX_TCC_MASK 0xFFFF +#ifndef CIK_H +#define CIK_H -#define VGA_HDP_CONTROL 0x328 -#define VGA_MEMORY_DISABLE (1 << 4) +#define BONAIRE_GB_ADDR_CONFIG_GOLDEN 0x12010001 + +#define CIK_RB_BITMAP_WIDTH_PER_SH 2 -#define CG_CLKPIN_CNTL 0x660 +/* SMC IND registers */ +#define GENERAL_PWRMGT 0xC0200000 +# define GPU_COUNTER_CLK (1 << 15) + +#define CG_CLKPIN_CNTL 0xC05001A0 # define XTALIN_DIVIDE (1 << 1) -#define CG_CLKPIN_CNTL_2 0x664 -# define MUX_TCLK_TO_XCLK (1 << 8) -#define DMIF_ADDR_CONFIG 0xBD4 +#define PCIE_INDEX 0x38 +#define PCIE_DATA 0x3C + +#define VGA_HDP_CONTROL 0x328 +#define VGA_MEMORY_DISABLE (1 << 4) #define DMIF_ADDR_CALC 0xC00 +#define SRBM_GFX_CNTL 0xE44 +#define PIPEID(x) ((x) << 0) +#define MEID(x) ((x) << 2) +#define VMID(x) ((x) << 4) +#define QUEUEID(x) ((x) << 8) + +#define SRBM_STATUS2 0xE4C +#define SDMA_BUSY (1 << 5) +#define SDMA1_BUSY (1 << 6) #define SRBM_STATUS 0xE50 +#define UVD_RQ_PENDING (1 << 1) #define GRBM_RQ_PENDING (1 << 5) #define VMC_BUSY (1 << 8) #define MCB_BUSY (1 << 9) @@ -108,11 +62,13 @@ #define MCD_BUSY (1 << 12) #define SEM_BUSY (1 << 14) #define IH_BUSY (1 << 17) +#define UVD_BUSY (1 << 19) -#define SRBM_SOFT_RESET 0x0E60 +#define SRBM_SOFT_RESET 0xE60 #define SOFT_RESET_BIF (1 << 1) +#define SOFT_RESET_R0PLL (1 << 4) #define SOFT_RESET_DC (1 << 5) -#define SOFT_RESET_DMA1 (1 << 6) +#define SOFT_RESET_SDMA1 (1 << 6) #define SOFT_RESET_GRBM (1 << 8) #define SOFT_RESET_HDP (1 << 9) #define SOFT_RESET_IH (1 << 10) @@ -120,17 +76,11 @@ #define SOFT_RESET_ROM (1 << 14) #define SOFT_RESET_SEM (1 << 15) #define SOFT_RESET_VMC (1 << 17) -#define SOFT_RESET_DMA (1 << 20) +#define SOFT_RESET_SDMA (1 << 20) #define SOFT_RESET_TST (1 << 21) -#define SOFT_RESET_REGBB (1 << 22) +#define SOFT_RESET_REGBB (1 << 22) #define SOFT_RESET_ORB (1 << 23) - -#define CC_SYS_RB_BACKEND_DISABLE 0xe80 -#define GC_USER_SYS_RB_BACKEND_DISABLE 0xe84 - -#define SRBM_STATUS2 0x0EC4 -#define DMA_BUSY (1 << 5) -#define DMA1_BUSY (1 << 6) +#define SOFT_RESET_VCE (1 << 24) #define VM_L2_CNTL 0x1400 #define ENABLE_L2_CACHE (1 << 0) @@ -182,12 +132,29 @@ #define VM_CONTEXT14_PAGE_TABLE_BASE_ADDR 0x1450 #define VM_CONTEXT15_PAGE_TABLE_BASE_ADDR 0x1454 -#define VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x14FC -#define VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x14DC - #define VM_INVALIDATE_REQUEST 0x1478 #define VM_INVALIDATE_RESPONSE 0x147c +#define VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x14DC +#define PROTECTIONS_MASK (0xf << 0) +#define PROTECTIONS_SHIFT 0 + /* bit 0: range + * bit 1: pde0 + * bit 2: valid + * bit 3: read + * bit 4: write + */ +#define MEMORY_CLIENT_ID_MASK (0xff << 12) +#define MEMORY_CLIENT_ID_SHIFT 12 +#define MEMORY_CLIENT_RW_MASK (1 << 24) +#define MEMORY_CLIENT_RW_SHIFT 24 +#define FAULT_VMID_MASK (0xf << 25) +#define FAULT_VMID_SHIFT 25 + +#define VM_CONTEXT1_PROTECTION_FAULT_MCCLIENT 0x14E4 + +#define VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x14FC + #define VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR 0x1518 #define VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR 0x151c @@ -210,6 +177,9 @@ #define NOOFCHAN_MASK 0x0000f000 #define MC_SHARED_CHREMAP 0x2008 +#define CHUB_CONTROL 0x1864 +#define BYPASS_VM (1 << 0) + #define MC_VM_FB_LOCATION 0x2024 #define MC_VM_AGP_TOP 0x2028 #define MC_VM_AGP_BOT 0x202C @@ -227,6 +197,7 @@ #define SYSTEM_ACCESS_MODE_NOT_IN_SYS (3 << 3) #define SYSTEM_APERTURE_UNMAPPED_ACCESS_PASS_THRU (0 << 5) #define ENABLE_ADVANCED_DRIVER_MODEL (1 << 6) +#define MC_VM_FB_OFFSET 0x2068 #define MC_SHARED_BLACKOUT_CNTL 0x20ac @@ -241,18 +212,17 @@ #define NOOFCOLS_MASK 0x000000C0 #define CHANSIZE_SHIFT 8 #define CHANSIZE_MASK 0x00000100 -#define CHANSIZE_OVERRIDE (1 << 11) #define NOOFGROUPS_SHIFT 12 #define NOOFGROUPS_MASK 0x00001000 -#define MC_SEQ_TRAIN_WAKEUP_CNTL 0x2808 -#define TRAIN_DONE_D0 (1 << 30) -#define TRAIN_DONE_D1 (1 << 31) - #define MC_SEQ_SUP_CNTL 0x28c8 #define RUN_MASK (1 << 0) #define MC_SEQ_SUP_PGM 0x28cc +#define MC_SEQ_TRAIN_WAKEUP_CNTL 0x28e8 +#define TRAIN_DONE_D0 (1 << 30) +#define TRAIN_DONE_D1 (1 << 31) + #define MC_IO_PAD_CNTL_D0 0x29d0 #define MEM_FALL_OUT_CMD (1 << 8) @@ -270,7 +240,7 @@ #define IH_RB_CNTL 0x3e00 # define IH_RB_ENABLE (1 << 0) -# define IH_IB_SIZE(x) ((x) << 1) /* log2 */ +# define IH_RB_SIZE(x) ((x) << 1) /* log2 */ # define IH_RB_FULL_DRAIN_ENABLE (1 << 6) # define IH_WPTR_WRITEBACK_ENABLE (1 << 8) # define IH_WPTR_WRITEBACK_TIMER(x) ((x) << 9) /* log2 */ @@ -312,40 +282,52 @@ #define HDP_REG_COHERENCY_FLUSH_CNTL 0x54A0 -#define DC_LB_MEMORY_SPLIT 0x6b0c -#define DC_LB_MEMORY_CONFIG(x) ((x) << 20) - -#define PRIORITY_A_CNT 0x6b18 -#define PRIORITY_MARK_MASK 0x7fff -#define PRIORITY_OFF (1 << 16) -#define PRIORITY_ALWAYS_ON (1 << 20) -#define PRIORITY_B_CNT 0x6b1c - -#define DPG_PIPE_ARBITRATION_CONTROL3 0x6cc8 -# define LATENCY_WATERMARK_MASK(x) ((x) << 16) -#define DPG_PIPE_LATENCY_CONTROL 0x6ccc -# define LATENCY_LOW_WATERMARK(x) ((x) << 0) -# define LATENCY_HIGH_WATERMARK(x) ((x) << 16) - -/* 0x6bb8, 0x77b8, 0x103b8, 0x10fb8, 0x11bb8, 0x127b8 */ -#define VLINE_STATUS 0x6bb8 +#define GPU_HDP_FLUSH_REQ 0x54DC +#define GPU_HDP_FLUSH_DONE 0x54E0 +#define CP0 (1 << 0) +#define CP1 (1 << 1) +#define CP2 (1 << 2) +#define CP3 (1 << 3) +#define CP4 (1 << 4) +#define CP5 (1 << 5) +#define CP6 (1 << 6) +#define CP7 (1 << 7) +#define CP8 (1 << 8) +#define CP9 (1 << 9) +#define SDMA0 (1 << 10) +#define SDMA1 (1 << 11) + +/* 0x6b04, 0x7704, 0x10304, 0x10f04, 0x11b04, 0x12704 */ +#define LB_MEMORY_CTRL 0x6b04 +#define LB_MEMORY_SIZE(x) ((x) << 0) +#define LB_MEMORY_CONFIG(x) ((x) << 20) + +#define DPG_WATERMARK_MASK_CONTROL 0x6cc8 +# define LATENCY_WATERMARK_MASK(x) ((x) << 8) +#define DPG_PIPE_LATENCY_CONTROL 0x6ccc +# define LATENCY_LOW_WATERMARK(x) ((x) << 0) +# define LATENCY_HIGH_WATERMARK(x) ((x) << 16) + +/* 0x6b24, 0x7724, 0x10324, 0x10f24, 0x11b24, 0x12724 */ +#define LB_VLINE_STATUS 0x6b24 # define VLINE_OCCURRED (1 << 0) # define VLINE_ACK (1 << 4) # define VLINE_STAT (1 << 12) # define VLINE_INTERRUPT (1 << 16) # define VLINE_INTERRUPT_TYPE (1 << 17) -/* 0x6bbc, 0x77bc, 0x103bc, 0x10fbc, 0x11bbc, 0x127bc */ -#define VBLANK_STATUS 0x6bbc +/* 0x6b2c, 0x772c, 0x1032c, 0x10f2c, 0x11b2c, 0x1272c */ +#define LB_VBLANK_STATUS 0x6b2c # define VBLANK_OCCURRED (1 << 0) # define VBLANK_ACK (1 << 4) # define VBLANK_STAT (1 << 12) # define VBLANK_INTERRUPT (1 << 16) # define VBLANK_INTERRUPT_TYPE (1 << 17) -/* 0x6b40, 0x7740, 0x10340, 0x10f40, 0x11b40, 0x12740 */ -#define INT_MASK 0x6b40 -# define VBLANK_INT_MASK (1 << 0) -# define VLINE_INT_MASK (1 << 4) +/* 0x6b20, 0x7720, 0x10320, 0x10f20, 0x11b20, 0x12720 */ +#define LB_INTERRUPT_MASK 0x6b20 +# define VBLANK_INTERRUPT_MASK (1 << 0) +# define VLINE_INTERRUPT_MASK (1 << 4) +# define VLINE2_INTERRUPT_MASK (1 << 8) #define DISP_INTERRUPT_STATUS 0x60f4 # define LB_D1_VLINE_INTERRUPT (1 << 2) @@ -382,17 +364,9 @@ # define LB_D6_VBLANK_INTERRUPT (1 << 3) # define DC_HPD6_INTERRUPT (1 << 17) # define DC_HPD6_RX_INTERRUPT (1 << 18) +#define DISP_INTERRUPT_STATUS_CONTINUE6 0x6780 -/* 0x6858, 0x7458, 0x10058, 0x10c58, 0x11858, 0x12458 */ -#define GRPH_INT_STATUS 0x6858 -# define GRPH_PFLIP_INT_OCCURRED (1 << 0) -# define GRPH_PFLIP_INT_CLEAR (1 << 8) -/* 0x685c, 0x745c, 0x1005c, 0x10c5c, 0x1185c, 0x1245c */ -#define GRPH_INT_CONTROL 0x685c -# define GRPH_PFLIP_INT_MASK (1 << 0) -# define GRPH_PFLIP_INT_TYPE (1 << 8) - -#define DACA_AUTODETECT_INT_CONTROL 0x66c8 +#define DAC_AUTODETECT_INT_CONTROL 0x67c8 #define DC_HPD1_INT_STATUS 0x601c #define DC_HPD2_INT_STATUS 0x6028 @@ -402,6 +376,7 @@ #define DC_HPD6_INT_STATUS 0x6058 # define DC_HPDx_INT_STATUS (1 << 0) # define DC_HPDx_SENSE (1 << 1) +# define DC_HPDx_SENSE_DELAYED (1 << 4) # define DC_HPDx_RX_INT_STATUS (1 << 8) #define DC_HPD1_INT_CONTROL 0x6020 @@ -426,34 +401,44 @@ # define DC_HPDx_RX_INT_TIMER(x) ((x) << 16) # define DC_HPDx_EN (1 << 28) -/* 0x6e98, 0x7a98, 0x10698, 0x11298, 0x11e98, 0x12a98 */ -#define CRTC_STATUS_FRAME_COUNT 0x6e98 - #define GRBM_CNTL 0x8000 #define GRBM_READ_TIMEOUT(x) ((x) << 0) #define GRBM_STATUS2 0x8008 -#define RLC_RQ_PENDING (1 << 0) -#define RLC_BUSY (1 << 8) -#define TC_BUSY (1 << 9) +#define ME0PIPE1_CMDFIFO_AVAIL_MASK 0x0000000F +#define ME0PIPE1_CF_RQ_PENDING (1 << 4) +#define ME0PIPE1_PF_RQ_PENDING (1 << 5) +#define ME1PIPE0_RQ_PENDING (1 << 6) +#define ME1PIPE1_RQ_PENDING (1 << 7) +#define ME1PIPE2_RQ_PENDING (1 << 8) +#define ME1PIPE3_RQ_PENDING (1 << 9) +#define ME2PIPE0_RQ_PENDING (1 << 10) +#define ME2PIPE1_RQ_PENDING (1 << 11) +#define ME2PIPE2_RQ_PENDING (1 << 12) +#define ME2PIPE3_RQ_PENDING (1 << 13) +#define RLC_RQ_PENDING (1 << 14) +#define RLC_BUSY (1 << 24) +#define TC_BUSY (1 << 25) +#define CPF_BUSY (1 << 28) +#define CPC_BUSY (1 << 29) +#define CPG_BUSY (1 << 30) #define GRBM_STATUS 0x8010 -#define CMDFIFO_AVAIL_MASK 0x0000000F -#define RING2_RQ_PENDING (1 << 4) +#define ME0PIPE0_CMDFIFO_AVAIL_MASK 0x0000000F #define SRBM_RQ_PENDING (1 << 5) -#define RING1_RQ_PENDING (1 << 6) -#define CF_RQ_PENDING (1 << 7) -#define PF_RQ_PENDING (1 << 8) +#define ME0PIPE0_CF_RQ_PENDING (1 << 7) +#define ME0PIPE0_PF_RQ_PENDING (1 << 8) #define GDS_DMA_RQ_PENDING (1 << 9) -#define GRBM_EE_BUSY (1 << 10) #define DB_CLEAN (1 << 12) #define CB_CLEAN (1 << 13) #define TA_BUSY (1 << 14) #define GDS_BUSY (1 << 15) +#define WD_BUSY_NO_DMA (1 << 16) #define VGT_BUSY (1 << 17) #define IA_BUSY_NO_DMA (1 << 18) #define IA_BUSY (1 << 19) #define SX_BUSY (1 << 20) +#define WD_BUSY (1 << 21) #define SPI_BUSY (1 << 22) #define BCI_BUSY (1 << 23) #define SC_BUSY (1 << 24) @@ -465,6 +450,8 @@ #define GUI_ACTIVE (1 << 31) #define GRBM_STATUS_SE0 0x8014 #define GRBM_STATUS_SE1 0x8018 +#define GRBM_STATUS_SE2 0x8038 +#define GRBM_STATUS_SE3 0x803C #define SE_DB_CLEAN (1 << 1) #define SE_CB_CLEAN (1 << 2) #define SE_BCI_BUSY (1 << 22) @@ -478,71 +465,50 @@ #define SE_CB_BUSY (1 << 31) #define GRBM_SOFT_RESET 0x8020 -#define SOFT_RESET_CP (1 << 0) -#define SOFT_RESET_CB (1 << 1) -#define SOFT_RESET_RLC (1 << 2) -#define SOFT_RESET_DB (1 << 3) -#define SOFT_RESET_GDS (1 << 4) -#define SOFT_RESET_PA (1 << 5) -#define SOFT_RESET_SC (1 << 6) -#define SOFT_RESET_BCI (1 << 7) -#define SOFT_RESET_SPI (1 << 8) -#define SOFT_RESET_SX (1 << 10) -#define SOFT_RESET_TC (1 << 11) -#define SOFT_RESET_TA (1 << 12) -#define SOFT_RESET_VGT (1 << 14) -#define SOFT_RESET_IA (1 << 15) - -#define GRBM_GFX_INDEX 0x802C -#define INSTANCE_INDEX(x) ((x) << 0) -#define SH_INDEX(x) ((x) << 8) -#define SE_INDEX(x) ((x) << 16) -#define SH_BROADCAST_WRITES (1 << 29) -#define INSTANCE_BROADCAST_WRITES (1 << 30) -#define SE_BROADCAST_WRITES (1 << 31) +#define SOFT_RESET_CP (1 << 0) /* All CP blocks */ +#define SOFT_RESET_RLC (1 << 2) /* RLC */ +#define SOFT_RESET_GFX (1 << 16) /* GFX */ +#define SOFT_RESET_CPF (1 << 17) /* CP fetcher shared by gfx and compute */ +#define SOFT_RESET_CPC (1 << 18) /* CP Compute (MEC1/2) */ +#define SOFT_RESET_CPG (1 << 19) /* CP GFX (PFP, ME, CE) */ #define GRBM_INT_CNTL 0x8060 # define RDERR_INT_ENABLE (1 << 0) # define GUI_IDLE_INT_ENABLE (1 << 19) -#define CP_STRMOUT_CNTL 0x84FC -#define SCRATCH_REG0 0x8500 -#define SCRATCH_REG1 0x8504 -#define SCRATCH_REG2 0x8508 -#define SCRATCH_REG3 0x850C -#define SCRATCH_REG4 0x8510 -#define SCRATCH_REG5 0x8514 -#define SCRATCH_REG6 0x8518 -#define SCRATCH_REG7 0x851C +#define CP_CPC_STATUS 0x8210 +#define CP_CPC_BUSY_STAT 0x8214 +#define CP_CPC_STALLED_STAT1 0x8218 +#define CP_CPF_STATUS 0x821c +#define CP_CPF_BUSY_STAT 0x8220 +#define CP_CPF_STALLED_STAT1 0x8224 -#define SCRATCH_UMSK 0x8540 -#define SCRATCH_ADDR 0x8544 +#define CP_MEC_CNTL 0x8234 +#define MEC_ME2_HALT (1 << 28) +#define MEC_ME1_HALT (1 << 30) -#define CP_SEM_WAIT_TIMER 0x85BC +#define CP_MEC_CNTL 0x8234 +#define MEC_ME2_HALT (1 << 28) +#define MEC_ME1_HALT (1 << 30) -#define CP_SEM_INCOMPLETE_TIMER_CNTL 0x85C8 +#define CP_STALLED_STAT3 0x8670 +#define CP_STALLED_STAT1 0x8674 +#define CP_STALLED_STAT2 0x8678 + +#define CP_STAT 0x8680 #define CP_ME_CNTL 0x86D8 #define CP_CE_HALT (1 << 24) #define CP_PFP_HALT (1 << 26) #define CP_ME_HALT (1 << 28) -#define CP_COHER_CNTL2 0x85E8 - -#define CP_RB2_RPTR 0x86f8 -#define CP_RB1_RPTR 0x86fc #define CP_RB0_RPTR 0x8700 #define CP_RB_WPTR_DELAY 0x8704 -#define CP_QUEUE_THRESHOLDS 0x8760 -#define ROQ_IB1_START(x) ((x) << 0) -#define ROQ_IB2_START(x) ((x) << 8) #define CP_MEQ_THRESHOLDS 0x8764 #define MEQ1_START(x) ((x) << 0) #define MEQ2_START(x) ((x) << 8) -#define CP_PERFMON_CNTL 0x87FC - #define VGT_VTX_VECT_EJECT_REG 0x88B0 #define VGT_CACHE_INVALIDATION 0x88C4 @@ -555,23 +521,9 @@ #define ES_AUTO 1 #define GS_AUTO 2 #define ES_AND_GS_AUTO 3 -#define VGT_ESGS_RING_SIZE 0x88C8 -#define VGT_GSVS_RING_SIZE 0x88CC #define VGT_GS_VERTEX_REUSE 0x88D4 -#define VGT_PRIMITIVE_TYPE 0x8958 -#define VGT_INDEX_TYPE 0x895C - -#define VGT_NUM_INDICES 0x8970 -#define VGT_NUM_INSTANCES 0x8974 - -#define VGT_TF_RING_SIZE 0x8988 - -#define VGT_HS_OFFCHIP_PARAM 0x89B0 - -#define VGT_TF_MEMORY_BASE 0x89B8 - #define CC_GC_SHADER_ARRAY_CONFIG 0x89bc #define INACTIVE_CUS_MASK 0xFFFF0000 #define INACTIVE_CUS_SHIFT 16 @@ -581,10 +533,6 @@ #define CLIP_VTX_REORDER_ENA (1 << 0) #define NUM_CLIP_SEQ(x) ((x) << 1) -#define PA_SU_LINE_STIPPLE_VALUE 0x8A60 - -#define PA_SC_LINE_STIPPLE_STATE 0x8B10 - #define PA_SC_FORCE_EOV_MAX_CNTS 0x8B24 #define FORCE_EOV_MAX_CLK_CNT(x) ((x) << 0) #define FORCE_EOV_MAX_REZ_CNT(x) ((x) << 16) @@ -596,31 +544,43 @@ #define SC_EARLYZ_TILE_FIFO_SIZE(x) ((x) << 23) #define PA_SC_ENHANCE 0x8BF0 +#define ENABLE_PA_SC_OUT_OF_ORDER (1 << 0) +#define DISABLE_PA_SC_GUIDANCE (1 << 13) #define SQ_CONFIG 0x8C00 -#define SQC_CACHES 0x8C08 +#define SH_MEM_BASES 0x8C28 +/* if PTR32, these are the bases for scratch and lds */ +#define PRIVATE_BASE(x) ((x) << 0) /* scratch */ +#define SHARED_BASE(x) ((x) << 16) /* LDS */ +#define SH_MEM_APE1_BASE 0x8C2C +/* if PTR32, this is the base location of GPUVM */ +#define SH_MEM_APE1_LIMIT 0x8C30 +/* if PTR32, this is the upper limit of GPUVM */ +#define SH_MEM_CONFIG 0x8C34 +#define PTR32 (1 << 0) +#define ALIGNMENT_MODE(x) ((x) << 2) +#define SH_MEM_ALIGNMENT_MODE_DWORD 0 +#define SH_MEM_ALIGNMENT_MODE_DWORD_STRICT 1 +#define SH_MEM_ALIGNMENT_MODE_STRICT 2 +#define SH_MEM_ALIGNMENT_MODE_UNALIGNED 3 +#define DEFAULT_MTYPE(x) ((x) << 4) +#define APE1_MTYPE(x) ((x) << 7) #define SX_DEBUG_1 0x9060 -#define SPI_STATIC_THREAD_MGMT_1 0x90E0 -#define SPI_STATIC_THREAD_MGMT_2 0x90E4 -#define SPI_STATIC_THREAD_MGMT_3 0x90E8 -#define SPI_PS_MAX_WAVE_ID 0x90EC - #define SPI_CONFIG_CNTL 0x9100 #define SPI_CONFIG_CNTL_1 0x913C #define VTX_DONE_DELAY(x) ((x) << 0) #define INTERP_ONE_PRIM_PER_ROW (1 << 4) -#define CGTS_TCC_DISABLE 0x9148 -#define CGTS_USER_TCC_DISABLE 0x914C -#define TCC_DISABLE_MASK 0xFFFF0000 -#define TCC_DISABLE_SHIFT 16 - #define TA_CNTL_AUX 0x9508 +#define DB_DEBUG 0x9830 +#define DB_DEBUG2 0x9834 +#define DB_DEBUG3 0x9838 + #define CC_RB_BACKEND_DISABLE 0x98F4 #define BACKEND_DISABLE(x) ((x) << 16) #define GB_ADDR_CONFIG 0x98F8 @@ -636,26 +596,18 @@ #define SHADER_ENGINE_TILE_SIZE(x) ((x) << 16) #define SHADER_ENGINE_TILE_SIZE_MASK 0x00070000 #define SHADER_ENGINE_TILE_SIZE_SHIFT 16 -#define NUM_GPUS(x) ((x) << 20) -#define NUM_GPUS_MASK 0x00700000 -#define NUM_GPUS_SHIFT 20 -#define MULTI_GPU_TILE_SIZE(x) ((x) << 24) -#define MULTI_GPU_TILE_SIZE_MASK 0x03000000 -#define MULTI_GPU_TILE_SIZE_SHIFT 24 #define ROW_SIZE(x) ((x) << 28) #define ROW_SIZE_MASK 0x30000000 #define ROW_SIZE_SHIFT 28 #define GB_TILE_MODE0 0x9910 -# define MICRO_TILE_MODE(x) ((x) << 0) -# define ADDR_SURF_DISPLAY_MICRO_TILING 0 -# define ADDR_SURF_THIN_MICRO_TILING 1 -# define ADDR_SURF_DEPTH_MICRO_TILING 2 # define ARRAY_MODE(x) ((x) << 2) # define ARRAY_LINEAR_GENERAL 0 # define ARRAY_LINEAR_ALIGNED 1 # define ARRAY_1D_TILED_THIN1 2 # define ARRAY_2D_TILED_THIN1 4 +# define ARRAY_PRT_TILED_THIN1 5 +# define ARRAY_PRT_2D_TILED_THIN1 6 # define PIPE_CONFIG(x) ((x) << 6) # define ADDR_SURF_P2 0 # define ADDR_SURF_P4_8x16 4 @@ -677,35 +629,40 @@ # define ADDR_SURF_TILE_SPLIT_1KB 4 # define ADDR_SURF_TILE_SPLIT_2KB 5 # define ADDR_SURF_TILE_SPLIT_4KB 6 -# define BANK_WIDTH(x) ((x) << 14) +# define MICRO_TILE_MODE_NEW(x) ((x) << 22) +# define ADDR_SURF_DISPLAY_MICRO_TILING 0 +# define ADDR_SURF_THIN_MICRO_TILING 1 +# define ADDR_SURF_DEPTH_MICRO_TILING 2 +# define ADDR_SURF_ROTATED_MICRO_TILING 3 +# define SAMPLE_SPLIT(x) ((x) << 25) +# define ADDR_SURF_SAMPLE_SPLIT_1 0 +# define ADDR_SURF_SAMPLE_SPLIT_2 1 +# define ADDR_SURF_SAMPLE_SPLIT_4 2 +# define ADDR_SURF_SAMPLE_SPLIT_8 3 + +#define GB_MACROTILE_MODE0 0x9990 +# define BANK_WIDTH(x) ((x) << 0) # define ADDR_SURF_BANK_WIDTH_1 0 # define ADDR_SURF_BANK_WIDTH_2 1 # define ADDR_SURF_BANK_WIDTH_4 2 # define ADDR_SURF_BANK_WIDTH_8 3 -# define BANK_HEIGHT(x) ((x) << 16) +# define BANK_HEIGHT(x) ((x) << 2) # define ADDR_SURF_BANK_HEIGHT_1 0 # define ADDR_SURF_BANK_HEIGHT_2 1 # define ADDR_SURF_BANK_HEIGHT_4 2 # define ADDR_SURF_BANK_HEIGHT_8 3 -# define MACRO_TILE_ASPECT(x) ((x) << 18) +# define MACRO_TILE_ASPECT(x) ((x) << 4) # define ADDR_SURF_MACRO_ASPECT_1 0 # define ADDR_SURF_MACRO_ASPECT_2 1 # define ADDR_SURF_MACRO_ASPECT_4 2 # define ADDR_SURF_MACRO_ASPECT_8 3 -# define NUM_BANKS(x) ((x) << 20) +# define NUM_BANKS(x) ((x) << 6) # define ADDR_SURF_2_BANK 0 # define ADDR_SURF_4_BANK 1 # define ADDR_SURF_8_BANK 2 # define ADDR_SURF_16_BANK 3 -#define CB_PERFCOUNTER0_SELECT0 0x9a20 -#define CB_PERFCOUNTER0_SELECT1 0x9a24 -#define CB_PERFCOUNTER1_SELECT0 0x9a28 -#define CB_PERFCOUNTER1_SELECT1 0x9a2c -#define CB_PERFCOUNTER2_SELECT0 0x9a30 -#define CB_PERFCOUNTER2_SELECT1 0x9a34 -#define CB_PERFCOUNTER3_SELECT0 0x9a38 -#define CB_PERFCOUNTER3_SELECT1 0x9a3c +#define CB_HW_CONTROL 0x9A10 #define GC_USER_RB_BACKEND_DISABLE 0x9B7C #define BACKEND_DISABLE_MASK 0x00FF0000 @@ -714,6 +671,17 @@ #define TCP_CHAN_STEER_LO 0xac0c #define TCP_CHAN_STEER_HI 0xac10 +#define TC_CFG_L1_LOAD_POLICY0 0xAC68 +#define TC_CFG_L1_LOAD_POLICY1 0xAC6C +#define TC_CFG_L1_STORE_POLICY 0xAC70 +#define TC_CFG_L2_LOAD_POLICY0 0xAC74 +#define TC_CFG_L2_LOAD_POLICY1 0xAC78 +#define TC_CFG_L2_STORE_POLICY0 0xAC7C +#define TC_CFG_L2_STORE_POLICY1 0xAC80 +#define TC_CFG_L2_ATOMIC_POLICY 0xAC84 +#define TC_CFG_L1_VOLATILE 0xAC88 +#define TC_CFG_L2_VOLATILE 0xAC8C + #define CP_RB0_BASE 0xC100 #define CP_RB0_CNTL 0xC104 #define RB_BUFSZ(x) ((x) << 0) @@ -723,9 +691,14 @@ #define RB_RPTR_WR_ENA (1 << 31) #define CP_RB0_RPTR_ADDR 0xC10C +#define RB_RPTR_SWAP_32BIT (2 << 0) #define CP_RB0_RPTR_ADDR_HI 0xC110 #define CP_RB0_WPTR 0xC114 +#define CP_DEVICE_ID 0xC12C +#define CP_ENDIAN_SWAP 0xC140 +#define CP_RB_VMID 0xC144 + #define CP_PFP_UCODE_ADDR 0xC150 #define CP_PFP_UCODE_DATA 0xC154 #define CP_ME_RAM_RADDR 0xC158 @@ -734,57 +707,149 @@ #define CP_CE_UCODE_ADDR 0xC168 #define CP_CE_UCODE_DATA 0xC16C +#define CP_MEC_ME1_UCODE_ADDR 0xC170 +#define CP_MEC_ME1_UCODE_DATA 0xC174 +#define CP_MEC_ME2_UCODE_ADDR 0xC178 +#define CP_MEC_ME2_UCODE_DATA 0xC17C -#define CP_RB1_BASE 0xC180 -#define CP_RB1_CNTL 0xC184 -#define CP_RB1_RPTR_ADDR 0xC188 -#define CP_RB1_RPTR_ADDR_HI 0xC18C -#define CP_RB1_WPTR 0xC190 -#define CP_RB2_BASE 0xC194 -#define CP_RB2_CNTL 0xC198 -#define CP_RB2_RPTR_ADDR 0xC19C -#define CP_RB2_RPTR_ADDR_HI 0xC1A0 -#define CP_RB2_WPTR 0xC1A4 #define CP_INT_CNTL_RING0 0xC1A8 -#define CP_INT_CNTL_RING1 0xC1AC -#define CP_INT_CNTL_RING2 0xC1B0 # define CNTX_BUSY_INT_ENABLE (1 << 19) # define CNTX_EMPTY_INT_ENABLE (1 << 20) -# define WAIT_MEM_SEM_INT_ENABLE (1 << 21) +# define PRIV_INSTR_INT_ENABLE (1 << 22) +# define PRIV_REG_INT_ENABLE (1 << 23) # define TIME_STAMP_INT_ENABLE (1 << 26) # define CP_RINGID2_INT_ENABLE (1 << 29) # define CP_RINGID1_INT_ENABLE (1 << 30) # define CP_RINGID0_INT_ENABLE (1 << 31) + #define CP_INT_STATUS_RING0 0xC1B4 -#define CP_INT_STATUS_RING1 0xC1B8 -#define CP_INT_STATUS_RING2 0xC1BC -# define WAIT_MEM_SEM_INT_STAT (1 << 21) +# define PRIV_INSTR_INT_STAT (1 << 22) +# define PRIV_REG_INT_STAT (1 << 23) # define TIME_STAMP_INT_STAT (1 << 26) # define CP_RINGID2_INT_STAT (1 << 29) # define CP_RINGID1_INT_STAT (1 << 30) # define CP_RINGID0_INT_STAT (1 << 31) -#define CP_DEBUG 0xC1FC +#define CP_CPF_DEBUG 0xC200 + +#define CP_PQ_WPTR_POLL_CNTL 0xC20C +#define WPTR_POLL_EN (1 << 31) + +#define CP_ME1_PIPE0_INT_CNTL 0xC214 +#define CP_ME1_PIPE1_INT_CNTL 0xC218 +#define CP_ME1_PIPE2_INT_CNTL 0xC21C +#define CP_ME1_PIPE3_INT_CNTL 0xC220 +#define CP_ME2_PIPE0_INT_CNTL 0xC224 +#define CP_ME2_PIPE1_INT_CNTL 0xC228 +#define CP_ME2_PIPE2_INT_CNTL 0xC22C +#define CP_ME2_PIPE3_INT_CNTL 0xC230 +# define DEQUEUE_REQUEST_INT_ENABLE (1 << 13) +# define WRM_POLL_TIMEOUT_INT_ENABLE (1 << 17) +# define PRIV_REG_INT_ENABLE (1 << 23) +# define TIME_STAMP_INT_ENABLE (1 << 26) +# define GENERIC2_INT_ENABLE (1 << 29) +# define GENERIC1_INT_ENABLE (1 << 30) +# define GENERIC0_INT_ENABLE (1 << 31) +#define CP_ME1_PIPE0_INT_STATUS 0xC214 +#define CP_ME1_PIPE1_INT_STATUS 0xC218 +#define CP_ME1_PIPE2_INT_STATUS 0xC21C +#define CP_ME1_PIPE3_INT_STATUS 0xC220 +#define CP_ME2_PIPE0_INT_STATUS 0xC224 +#define CP_ME2_PIPE1_INT_STATUS 0xC228 +#define CP_ME2_PIPE2_INT_STATUS 0xC22C +#define CP_ME2_PIPE3_INT_STATUS 0xC230 +# define DEQUEUE_REQUEST_INT_STATUS (1 << 13) +# define WRM_POLL_TIMEOUT_INT_STATUS (1 << 17) +# define PRIV_REG_INT_STATUS (1 << 23) +# define TIME_STAMP_INT_STATUS (1 << 26) +# define GENERIC2_INT_STATUS (1 << 29) +# define GENERIC1_INT_STATUS (1 << 30) +# define GENERIC0_INT_STATUS (1 << 31) + +#define CP_MAX_CONTEXT 0xC2B8 + +#define CP_RB0_BASE_HI 0xC2C4 #define RLC_CNTL 0xC300 # define RLC_ENABLE (1 << 0) -#define RLC_RL_BASE 0xC304 -#define RLC_RL_SIZE 0xC308 -#define RLC_LB_CNTL 0xC30C -#define RLC_SAVE_AND_RESTORE_BASE 0xC310 -#define RLC_LB_CNTR_MAX 0xC314 -#define RLC_LB_CNTR_INIT 0xC318 - -#define RLC_CLEAR_STATE_RESTORE_BASE 0xC320 - -#define RLC_UCODE_ADDR 0xC32C -#define RLC_UCODE_DATA 0xC330 -#define RLC_GPU_CLOCK_COUNT_LSB 0xC338 -#define RLC_GPU_CLOCK_COUNT_MSB 0xC33C -#define RLC_CAPTURE_GPU_CLOCK_COUNT 0xC340 -#define RLC_MC_CNTL 0xC344 -#define RLC_UCODE_CNTL 0xC348 +#define RLC_MC_CNTL 0xC30C + +#define RLC_LB_CNTR_MAX 0xC348 + +#define RLC_LB_CNTL 0xC364 + +#define RLC_LB_CNTR_INIT 0xC36C + +#define RLC_SAVE_AND_RESTORE_BASE 0xC374 +#define RLC_DRIVER_DMA_STATUS 0xC378 + +#define RLC_GPM_UCODE_ADDR 0xC388 +#define RLC_GPM_UCODE_DATA 0xC38C +#define RLC_GPU_CLOCK_COUNT_LSB 0xC390 +#define RLC_GPU_CLOCK_COUNT_MSB 0xC394 +#define RLC_CAPTURE_GPU_CLOCK_COUNT 0xC398 +#define RLC_UCODE_CNTL 0xC39C + +#define RLC_CGCG_CGLS_CTRL 0xC424 + +#define RLC_LB_INIT_CU_MASK 0xC43C + +#define RLC_LB_PARAMS 0xC444 + +#define RLC_SERDES_CU_MASTER_BUSY 0xC484 +#define RLC_SERDES_NONCU_MASTER_BUSY 0xC488 +# define SE_MASTER_BUSY_MASK 0x0000ffff +# define GC_MASTER_BUSY (1 << 16) +# define TC0_MASTER_BUSY (1 << 17) +# define TC1_MASTER_BUSY (1 << 18) + +#define RLC_GPM_SCRATCH_ADDR 0xC4B0 +#define RLC_GPM_SCRATCH_DATA 0xC4B4 + +#define CP_HPD_EOP_BASE_ADDR 0xC904 +#define CP_HPD_EOP_BASE_ADDR_HI 0xC908 +#define CP_HPD_EOP_VMID 0xC90C +#define CP_HPD_EOP_CONTROL 0xC910 +#define EOP_SIZE(x) ((x) << 0) +#define EOP_SIZE_MASK (0x3f << 0) +#define CP_MQD_BASE_ADDR 0xC914 +#define CP_MQD_BASE_ADDR_HI 0xC918 +#define CP_HQD_ACTIVE 0xC91C +#define CP_HQD_VMID 0xC920 + +#define CP_HQD_PQ_BASE 0xC934 +#define CP_HQD_PQ_BASE_HI 0xC938 +#define CP_HQD_PQ_RPTR 0xC93C +#define CP_HQD_PQ_RPTR_REPORT_ADDR 0xC940 +#define CP_HQD_PQ_RPTR_REPORT_ADDR_HI 0xC944 +#define CP_HQD_PQ_WPTR_POLL_ADDR 0xC948 +#define CP_HQD_PQ_WPTR_POLL_ADDR_HI 0xC94C +#define CP_HQD_PQ_DOORBELL_CONTROL 0xC950 +#define DOORBELL_OFFSET(x) ((x) << 2) +#define DOORBELL_OFFSET_MASK (0x1fffff << 2) +#define DOORBELL_SOURCE (1 << 28) +#define DOORBELL_SCHD_HIT (1 << 29) +#define DOORBELL_EN (1 << 30) +#define DOORBELL_HIT (1 << 31) +#define CP_HQD_PQ_WPTR 0xC954 +#define CP_HQD_PQ_CONTROL 0xC958 +#define QUEUE_SIZE(x) ((x) << 0) +#define QUEUE_SIZE_MASK (0x3f << 0) +#define RPTR_BLOCK_SIZE(x) ((x) << 8) +#define RPTR_BLOCK_SIZE_MASK (0x3f << 8) +#define PQ_VOLATILE (1 << 26) +#define NO_UPDATE_RPTR (1 << 27) +#define UNORD_DISPATCH (1 << 28) +#define ROQ_PQ_IB_FLIP (1 << 29) +#define PRIV_STATE (1 << 30) +#define KMD_QUEUE (1 << 31) + +#define CP_HQD_DEQUEUE_REQUEST 0xC974 + +#define CP_MQD_CONTROL 0xC99C +#define MQD_VMID(x) ((x) << 0) +#define MQD_VMID_MASK (0xf << 0) #define PA_SC_RASTER_CONFIG 0x28350 # define RASTER_CONFIG_RB_MAP_0 0 @@ -814,6 +879,7 @@ # define PIPELINESTAT_STOP (26 << 0) # define PERFCOUNTER_SAMPLE (27 << 0) # define SAMPLE_PIPELINESTAT (30 << 0) +# define SO_VGT_STREAMOUT_FLUSH (31 << 0) # define SAMPLE_STREAMOUTSTATS (32 << 0) # define RESET_VTX_CNT (33 << 0) # define VGT_FLUSH (36 << 0) @@ -830,20 +896,74 @@ # define THREAD_TRACE_STOP (52 << 0) # define THREAD_TRACE_FLUSH (54 << 0) # define THREAD_TRACE_FINISH (55 << 0) +# define PIXEL_PIPE_STAT_CONTROL (56 << 0) +# define PIXEL_PIPE_STAT_DUMP (57 << 0) +# define PIXEL_PIPE_STAT_RESET (58 << 0) -/* - * UVD - */ -#define UVD_UDEC_ADDR_CONFIG 0xEF4C -#define UVD_UDEC_DB_ADDR_CONFIG 0xEF50 -#define UVD_UDEC_DBW_ADDR_CONFIG 0xEF54 -#define UVD_RBC_RB_RPTR 0xF690 -#define UVD_RBC_RB_WPTR 0xF694 +#define SCRATCH_REG0 0x30100 +#define SCRATCH_REG1 0x30104 +#define SCRATCH_REG2 0x30108 +#define SCRATCH_REG3 0x3010C +#define SCRATCH_REG4 0x30110 +#define SCRATCH_REG5 0x30114 +#define SCRATCH_REG6 0x30118 +#define SCRATCH_REG7 0x3011C + +#define SCRATCH_UMSK 0x30140 +#define SCRATCH_ADDR 0x30144 + +#define CP_SEM_WAIT_TIMER 0x301BC + +#define CP_SEM_INCOMPLETE_TIMER_CNTL 0x301C8 + +#define CP_WAIT_REG_MEM_TIMEOUT 0x301D0 + +#define GRBM_GFX_INDEX 0x30800 +#define INSTANCE_INDEX(x) ((x) << 0) +#define SH_INDEX(x) ((x) << 8) +#define SE_INDEX(x) ((x) << 16) +#define SH_BROADCAST_WRITES (1 << 29) +#define INSTANCE_BROADCAST_WRITES (1 << 30) +#define SE_BROADCAST_WRITES (1 << 31) + +#define VGT_ESGS_RING_SIZE 0x30900 +#define VGT_GSVS_RING_SIZE 0x30904 +#define VGT_PRIMITIVE_TYPE 0x30908 +#define VGT_INDEX_TYPE 0x3090C + +#define VGT_NUM_INDICES 0x30930 +#define VGT_NUM_INSTANCES 0x30934 +#define VGT_TF_RING_SIZE 0x30938 +#define VGT_HS_OFFCHIP_PARAM 0x3093C +#define VGT_TF_MEMORY_BASE 0x30940 + +#define PA_SU_LINE_STIPPLE_VALUE 0x30a00 +#define PA_SC_LINE_STIPPLE_STATE 0x30a04 + +#define SQC_CACHES 0x30d20 + +#define CP_PERFMON_CNTL 0x36020 + +#define CGTS_TCC_DISABLE 0x3c00c +#define CGTS_USER_TCC_DISABLE 0x3c010 +#define TCC_DISABLE_MASK 0xFFFF0000 +#define TCC_DISABLE_SHIFT 16 + +#define CB_CGTT_SCLK_CTRL 0x3c2a0 /* * PM4 */ -#define PACKET0(reg, n) ((RADEON_PACKET_TYPE0 << 30) | \ +#define PACKET_TYPE0 0 +#define PACKET_TYPE1 1 +#define PACKET_TYPE2 2 +#define PACKET_TYPE3 3 + +#define CP_PACKET_GET_TYPE(h) (((h) >> 30) & 3) +#define CP_PACKET_GET_COUNT(h) (((h) >> 16) & 0x3FFF) +#define CP_PACKET0_GET_REG(h) (((h) & 0xFFFF) << 2) +#define CP_PACKET3_GET_OPCODE(h) (((h) >> 8) & 0xFF) +#define PACKET0(reg, n) ((PACKET_TYPE0 << 30) | \ (((reg) >> 2) & 0xFFFF) | \ ((n) & 0x3FFF) << 16) #define CP_PACKET2 0x80000000 @@ -852,7 +972,7 @@ #define PACKET2(v) (CP_PACKET2 | REG_SET(PACKET2_PAD, (v))) -#define PACKET3(op, n) ((RADEON_PACKET_TYPE3 << 30) | \ +#define PACKET3(op, n) ((PACKET_TYPE3 << 30) | \ (((op) & 0xFF) << 8) | \ ((n) & 0x3FFF) << 16) @@ -862,16 +982,13 @@ #define PACKET3_NOP 0x10 #define PACKET3_SET_BASE 0x11 #define PACKET3_BASE_INDEX(x) ((x) << 0) -#define GDS_PARTITION_BASE 2 #define CE_PARTITION_BASE 3 #define PACKET3_CLEAR_STATE 0x12 #define PACKET3_INDEX_BUFFER_SIZE 0x13 #define PACKET3_DISPATCH_DIRECT 0x15 #define PACKET3_DISPATCH_INDIRECT 0x16 -#define PACKET3_ALLOC_GDS 0x1B -#define PACKET3_WRITE_GDS_RAM 0x1C #define PACKET3_ATOMIC_GDS 0x1D -#define PACKET3_ATOMIC 0x1E +#define PACKET3_ATOMIC_MEM 0x1E #define PACKET3_OCCLUSION_QUERY 0x1F #define PACKET3_SET_PREDICATION 0x20 #define PACKET3_REG_RMW 0x21 @@ -885,25 +1002,27 @@ #define PACKET3_INDEX_TYPE 0x2A #define PACKET3_DRAW_INDIRECT_MULTI 0x2C #define PACKET3_DRAW_INDEX_AUTO 0x2D -#define PACKET3_DRAW_INDEX_IMMD 0x2E #define PACKET3_NUM_INSTANCES 0x2F #define PACKET3_DRAW_INDEX_MULTI_AUTO 0x30 -#define PACKET3_INDIRECT_BUFFER_CONST 0x31 -#define PACKET3_INDIRECT_BUFFER 0x32 +#define PACKET3_INDIRECT_BUFFER_CONST 0x33 #define PACKET3_STRMOUT_BUFFER_UPDATE 0x34 #define PACKET3_DRAW_INDEX_OFFSET_2 0x35 -#define PACKET3_DRAW_INDEX_MULTI_ELEMENT 0x36 +#define PACKET3_DRAW_PREAMBLE 0x36 #define PACKET3_WRITE_DATA 0x37 #define WRITE_DATA_DST_SEL(x) ((x) << 8) /* 0 - register * 1 - memory (sync - via GRBM) - * 2 - tc/l2 + * 2 - gl2 * 3 - gds * 4 - reserved * 5 - memory (async - direct) */ #define WR_ONE_ADDR (1 << 16) #define WR_CONFIRM (1 << 20) +#define WRITE_DATA_CACHE_POLICY(x) ((x) << 25) + /* 0 - LRU + * 1 - Stream + */ #define WRITE_DATA_ENGINE_SEL(x) ((x) << 30) /* 0 - me * 1 - pfp @@ -911,59 +1030,43 @@ */ #define PACKET3_DRAW_INDEX_INDIRECT_MULTI 0x38 #define PACKET3_MEM_SEMAPHORE 0x39 -#define PACKET3_MPEG_INDEX 0x3A +# define PACKET3_SEM_USE_MAILBOX (0x1 << 16) +# define PACKET3_SEM_SEL_SIGNAL_TYPE (0x1 << 20) /* 0 = increment, 1 = write 1 */ +# define PACKET3_SEM_CLIENT_CODE ((x) << 24) /* 0 = CP, 1 = CB, 2 = DB */ +# define PACKET3_SEM_SEL_SIGNAL (0x6 << 29) +# define PACKET3_SEM_SEL_WAIT (0x7 << 29) #define PACKET3_COPY_DW 0x3B #define PACKET3_WAIT_REG_MEM 0x3C -#define PACKET3_MEM_WRITE 0x3D -#define PACKET3_COPY_DATA 0x40 -#define PACKET3_CP_DMA 0x41 -/* 1. header - * 2. SRC_ADDR_LO or DATA [31:0] - * 3. CP_SYNC [31] | SRC_SEL [30:29] | ENGINE [27] | DST_SEL [21:20] | - * SRC_ADDR_HI [7:0] - * 4. DST_ADDR_LO [31:0] - * 5. DST_ADDR_HI [7:0] - * 6. COMMAND [30:21] | BYTE_COUNT [20:0] - */ -# define PACKET3_CP_DMA_DST_SEL(x) ((x) << 20) - /* 0 - SRC_ADDR - * 1 - GDS +#define WAIT_REG_MEM_FUNCTION(x) ((x) << 0) + /* 0 - always + * 1 - < + * 2 - <= + * 3 - == + * 4 - != + * 5 - >= + * 6 - > */ -# define PACKET3_CP_DMA_ENGINE(x) ((x) << 27) - /* 0 - ME - * 1 - PFP +#define WAIT_REG_MEM_MEM_SPACE(x) ((x) << 4) + /* 0 - reg + * 1 - mem */ -# define PACKET3_CP_DMA_SRC_SEL(x) ((x) << 29) - /* 0 - SRC_ADDR - * 1 - GDS - * 2 - DATA - */ -# define PACKET3_CP_DMA_CP_SYNC (1 << 31) -/* COMMAND */ -# define PACKET3_CP_DMA_DIS_WC (1 << 21) -# define PACKET3_CP_DMA_CMD_SRC_SWAP(x) ((x) << 23) - /* 0 - none - * 1 - 8 in 16 - * 2 - 8 in 32 - * 3 - 8 in 64 +#define WAIT_REG_MEM_OPERATION(x) ((x) << 6) + /* 0 - wait_reg_mem + * 1 - wr_wait_wr_reg */ -# define PACKET3_CP_DMA_CMD_DST_SWAP(x) ((x) << 24) - /* 0 - none - * 1 - 8 in 16 - * 2 - 8 in 32 - * 3 - 8 in 64 - */ -# define PACKET3_CP_DMA_CMD_SAS (1 << 26) - /* 0 - memory - * 1 - register +#define WAIT_REG_MEM_ENGINE(x) ((x) << 8) + /* 0 - me + * 1 - pfp */ -# define PACKET3_CP_DMA_CMD_DAS (1 << 27) - /* 0 - memory - * 1 - register +#define PACKET3_INDIRECT_BUFFER 0x3F +#define INDIRECT_BUFFER_TCL2_VOLATILE (1 << 22) +#define INDIRECT_BUFFER_VALID (1 << 23) +#define INDIRECT_BUFFER_CACHE_POLICY(x) ((x) << 28) + /* 0 - LRU + * 1 - Stream + * 2 - Bypass */ -# define PACKET3_CP_DMA_CMD_SAIC (1 << 28) -# define PACKET3_CP_DMA_CMD_DAIC (1 << 29) -# define PACKET3_CP_DMA_CMD_RAW_WAIT (1 << 30) +#define PACKET3_COPY_DATA 0x40 #define PACKET3_PFP_SYNC_ME 0x42 #define PACKET3_SURFACE_SYNC 0x43 # define PACKET3_DEST_BASE_0_ENA (1 << 0) @@ -977,51 +1080,70 @@ # define PACKET3_CB6_DEST_BASE_ENA (1 << 12) # define PACKET3_CB7_DEST_BASE_ENA (1 << 13) # define PACKET3_DB_DEST_BASE_ENA (1 << 14) +# define PACKET3_TCL1_VOL_ACTION_ENA (1 << 15) +# define PACKET3_TC_VOL_ACTION_ENA (1 << 16) /* L2 */ +# define PACKET3_TC_WB_ACTION_ENA (1 << 18) /* L2 */ # define PACKET3_DEST_BASE_2_ENA (1 << 19) # define PACKET3_DEST_BASE_3_ENA (1 << 21) # define PACKET3_TCL1_ACTION_ENA (1 << 22) -# define PACKET3_TC_ACTION_ENA (1 << 23) +# define PACKET3_TC_ACTION_ENA (1 << 23) /* L2 */ # define PACKET3_CB_ACTION_ENA (1 << 25) # define PACKET3_DB_ACTION_ENA (1 << 26) # define PACKET3_SH_KCACHE_ACTION_ENA (1 << 27) +# define PACKET3_SH_KCACHE_VOL_ACTION_ENA (1 << 28) # define PACKET3_SH_ICACHE_ACTION_ENA (1 << 29) -#define PACKET3_ME_INITIALIZE 0x44 -#define PACKET3_ME_INITIALIZE_DEVICE_ID(x) ((x) << 16) #define PACKET3_COND_WRITE 0x45 #define PACKET3_EVENT_WRITE 0x46 #define EVENT_TYPE(x) ((x) << 0) #define EVENT_INDEX(x) ((x) << 8) /* 0 - any non-TS event - * 1 - ZPASS_DONE + * 1 - ZPASS_DONE, PIXEL_PIPE_STAT_* * 2 - SAMPLE_PIPELINESTAT * 3 - SAMPLE_STREAMOUTSTAT* * 4 - *S_PARTIAL_FLUSH * 5 - EOP events * 6 - EOS events - * 7 - CACHE_FLUSH, CACHE_FLUSH_AND_INV_EVENT */ -#define INV_L2 (1 << 20) - /* INV TC L2 cache when EVENT_INDEX = 7 */ #define PACKET3_EVENT_WRITE_EOP 0x47 +#define EOP_TCL1_VOL_ACTION_EN (1 << 12) +#define EOP_TC_VOL_ACTION_EN (1 << 13) /* L2 */ +#define EOP_TC_WB_ACTION_EN (1 << 15) /* L2 */ +#define EOP_TCL1_ACTION_EN (1 << 16) +#define EOP_TC_ACTION_EN (1 << 17) /* L2 */ +#define EOP_CACHE_POLICY(x) ((x) << 25) + /* 0 - LRU + * 1 - Stream + * 2 - Bypass + */ +#define EOP_TCL2_VOLATILE (1 << 27) #define DATA_SEL(x) ((x) << 29) /* 0 - discard * 1 - send low 32bit data * 2 - send 64bit data - * 3 - send 64bit counter value + * 3 - send 64bit GPU counter value + * 4 - send 64bit sys counter value */ #define INT_SEL(x) ((x) << 24) /* 0 - none * 1 - interrupt only (DATA_SEL = 0) * 2 - interrupt when data write is confirmed */ +#define DST_SEL(x) ((x) << 16) + /* 0 - MC + * 1 - TC/L2 + */ #define PACKET3_EVENT_WRITE_EOS 0x48 +#define PACKET3_RELEASE_MEM 0x49 #define PACKET3_PREAMBLE_CNTL 0x4A # define PACKET3_PREAMBLE_BEGIN_CLEAR_STATE (2 << 28) # define PACKET3_PREAMBLE_END_CLEAR_STATE (3 << 28) -#define PACKET3_ONE_REG_WRITE 0x57 -#define PACKET3_LOAD_CONFIG_REG 0x5F -#define PACKET3_LOAD_CONTEXT_REG 0x60 -#define PACKET3_LOAD_SH_REG 0x61 +#define PACKET3_DMA_DATA 0x50 +#define PACKET3_AQUIRE_MEM 0x58 +#define PACKET3_REWIND 0x59 +#define PACKET3_LOAD_UCONFIG_REG 0x5E +#define PACKET3_LOAD_SH_REG 0x5F +#define PACKET3_LOAD_CONFIG_REG 0x60 +#define PACKET3_LOAD_CONTEXT_REG 0x61 #define PACKET3_SET_CONFIG_REG 0x68 #define PACKET3_SET_CONFIG_REG_START 0x00008000 #define PACKET3_SET_CONFIG_REG_END 0x0000b000 @@ -1029,85 +1151,163 @@ #define PACKET3_SET_CONTEXT_REG_START 0x00028000 #define PACKET3_SET_CONTEXT_REG_END 0x00029000 #define PACKET3_SET_CONTEXT_REG_INDIRECT 0x73 -#define PACKET3_SET_RESOURCE_INDIRECT 0x74 #define PACKET3_SET_SH_REG 0x76 #define PACKET3_SET_SH_REG_START 0x0000b000 #define PACKET3_SET_SH_REG_END 0x0000c000 #define PACKET3_SET_SH_REG_OFFSET 0x77 -#define PACKET3_ME_WRITE 0x7A +#define PACKET3_SET_QUEUE_REG 0x78 +#define PACKET3_SET_UCONFIG_REG 0x79 +#define PACKET3_SET_UCONFIG_REG_START 0x00030000 +#define PACKET3_SET_UCONFIG_REG_END 0x00031000 #define PACKET3_SCRATCH_RAM_WRITE 0x7D #define PACKET3_SCRATCH_RAM_READ 0x7E -#define PACKET3_CE_WRITE 0x7F #define PACKET3_LOAD_CONST_RAM 0x80 #define PACKET3_WRITE_CONST_RAM 0x81 -#define PACKET3_WRITE_CONST_RAM_OFFSET 0x82 #define PACKET3_DUMP_CONST_RAM 0x83 #define PACKET3_INCREMENT_CE_COUNTER 0x84 #define PACKET3_INCREMENT_DE_COUNTER 0x85 #define PACKET3_WAIT_ON_CE_COUNTER 0x86 -#define PACKET3_WAIT_ON_DE_COUNTER 0x87 #define PACKET3_WAIT_ON_DE_COUNTER_DIFF 0x88 -#define PACKET3_SET_CE_DE_COUNTERS 0x89 -#define PACKET3_WAIT_ON_AVAIL_BUFFER 0x8A #define PACKET3_SWITCH_BUFFER 0x8B -/* ASYNC DMA - first instance at 0xd000, second at 0xd800 */ -#define DMA0_REGISTER_OFFSET 0x0 /* not a register */ -#define DMA1_REGISTER_OFFSET 0x800 /* not a register */ - -#define DMA_RB_CNTL 0xd000 -# define DMA_RB_ENABLE (1 << 0) -# define DMA_RB_SIZE(x) ((x) << 1) /* log2 */ -# define DMA_RB_SWAP_ENABLE (1 << 9) /* 8IN32 */ -# define DMA_RPTR_WRITEBACK_ENABLE (1 << 12) -# define DMA_RPTR_WRITEBACK_SWAP_ENABLE (1 << 13) /* 8IN32 */ -# define DMA_RPTR_WRITEBACK_TIMER(x) ((x) << 16) /* log2 */ -#define DMA_RB_BASE 0xd004 -#define DMA_RB_RPTR 0xd008 -#define DMA_RB_WPTR 0xd00c - -#define DMA_RB_RPTR_ADDR_HI 0xd01c -#define DMA_RB_RPTR_ADDR_LO 0xd020 - -#define DMA_IB_CNTL 0xd024 -# define DMA_IB_ENABLE (1 << 0) -# define DMA_IB_SWAP_ENABLE (1 << 4) -#define DMA_IB_RPTR 0xd028 -#define DMA_CNTL 0xd02c +/* SDMA - first instance at 0xd000, second at 0xd800 */ +#define SDMA0_REGISTER_OFFSET 0x0 /* not a register */ +#define SDMA1_REGISTER_OFFSET 0x800 /* not a register */ + +#define SDMA0_UCODE_ADDR 0xD000 +#define SDMA0_UCODE_DATA 0xD004 + +#define SDMA0_CNTL 0xD010 # define TRAP_ENABLE (1 << 0) # define SEM_INCOMPLETE_INT_ENABLE (1 << 1) # define SEM_WAIT_INT_ENABLE (1 << 2) # define DATA_SWAP_ENABLE (1 << 3) # define FENCE_SWAP_ENABLE (1 << 4) +# define AUTO_CTXSW_ENABLE (1 << 18) # define CTXEMPTY_INT_ENABLE (1 << 28) -#define DMA_STATUS_REG 0xd034 -# define DMA_IDLE (1 << 0) -#define DMA_TILING_CONFIG 0xd0b8 - -#define DMA_PACKET(cmd, b, t, s, n) ((((cmd) & 0xF) << 28) | \ - (((b) & 0x1) << 26) | \ - (((t) & 0x1) << 23) | \ - (((s) & 0x1) << 22) | \ - (((n) & 0xFFFFF) << 0)) - -#define DMA_IB_PACKET(cmd, vmid, n) ((((cmd) & 0xF) << 28) | \ - (((vmid) & 0xF) << 20) | \ - (((n) & 0xFFFFF) << 0)) - -#define DMA_PTE_PDE_PACKET(n) ((2 << 28) | \ - (1 << 26) | \ - (1 << 21) | \ - (((n) & 0xFFFFF) << 0)) - -/* async DMA Packet types */ -#define DMA_PACKET_WRITE 0x2 -#define DMA_PACKET_COPY 0x3 -#define DMA_PACKET_INDIRECT_BUFFER 0x4 -#define DMA_PACKET_SEMAPHORE 0x5 -#define DMA_PACKET_FENCE 0x6 -#define DMA_PACKET_TRAP 0x7 -#define DMA_PACKET_SRBM_WRITE 0x9 -#define DMA_PACKET_CONSTANT_FILL 0xd -#define DMA_PACKET_NOP 0xf + +#define SDMA0_TILING_CONFIG 0xD018 + +#define SDMA0_SEM_INCOMPLETE_TIMER_CNTL 0xD020 +#define SDMA0_SEM_WAIT_FAIL_TIMER_CNTL 0xD024 + +#define SDMA0_STATUS_REG 0xd034 +# define SDMA_IDLE (1 << 0) + +#define SDMA0_ME_CNTL 0xD048 +# define SDMA_HALT (1 << 0) + +#define SDMA0_GFX_RB_CNTL 0xD200 +# define SDMA_RB_ENABLE (1 << 0) +# define SDMA_RB_SIZE(x) ((x) << 1) /* log2 */ +# define SDMA_RB_SWAP_ENABLE (1 << 9) /* 8IN32 */ +# define SDMA_RPTR_WRITEBACK_ENABLE (1 << 12) +# define SDMA_RPTR_WRITEBACK_SWAP_ENABLE (1 << 13) /* 8IN32 */ +# define SDMA_RPTR_WRITEBACK_TIMER(x) ((x) << 16) /* log2 */ +#define SDMA0_GFX_RB_BASE 0xD204 +#define SDMA0_GFX_RB_BASE_HI 0xD208 +#define SDMA0_GFX_RB_RPTR 0xD20C +#define SDMA0_GFX_RB_WPTR 0xD210 + +#define SDMA0_GFX_RB_RPTR_ADDR_HI 0xD220 +#define SDMA0_GFX_RB_RPTR_ADDR_LO 0xD224 +#define SDMA0_GFX_IB_CNTL 0xD228 +# define SDMA_IB_ENABLE (1 << 0) +# define SDMA_IB_SWAP_ENABLE (1 << 4) +# define SDMA_SWITCH_INSIDE_IB (1 << 8) +# define SDMA_CMD_VMID(x) ((x) << 16) + +#define SDMA0_GFX_VIRTUAL_ADDR 0xD29C +#define SDMA0_GFX_APE1_CNTL 0xD2A0 + +#define SDMA_PACKET(op, sub_op, e) ((((e) & 0xFFFF) << 16) | \ + (((sub_op) & 0xFF) << 8) | \ + (((op) & 0xFF) << 0)) +/* sDMA opcodes */ +#define SDMA_OPCODE_NOP 0 +#define SDMA_OPCODE_COPY 1 +# define SDMA_COPY_SUB_OPCODE_LINEAR 0 +# define SDMA_COPY_SUB_OPCODE_TILED 1 +# define SDMA_COPY_SUB_OPCODE_SOA 3 +# define SDMA_COPY_SUB_OPCODE_LINEAR_SUB_WINDOW 4 +# define SDMA_COPY_SUB_OPCODE_TILED_SUB_WINDOW 5 +# define SDMA_COPY_SUB_OPCODE_T2T_SUB_WINDOW 6 +#define SDMA_OPCODE_WRITE 2 +# define SDMA_WRITE_SUB_OPCODE_LINEAR 0 +# define SDMA_WRTIE_SUB_OPCODE_TILED 1 +#define SDMA_OPCODE_INDIRECT_BUFFER 4 +#define SDMA_OPCODE_FENCE 5 +#define SDMA_OPCODE_TRAP 6 +#define SDMA_OPCODE_SEMAPHORE 7 +# define SDMA_SEMAPHORE_EXTRA_O (1 << 13) + /* 0 - increment + * 1 - write 1 + */ +# define SDMA_SEMAPHORE_EXTRA_S (1 << 14) + /* 0 - wait + * 1 - signal + */ +# define SDMA_SEMAPHORE_EXTRA_M (1 << 15) + /* mailbox */ +#define SDMA_OPCODE_POLL_REG_MEM 8 +# define SDMA_POLL_REG_MEM_EXTRA_OP(x) ((x) << 10) + /* 0 - wait_reg_mem + * 1 - wr_wait_wr_reg + */ +# define SDMA_POLL_REG_MEM_EXTRA_FUNC(x) ((x) << 12) + /* 0 - always + * 1 - < + * 2 - <= + * 3 - == + * 4 - != + * 5 - >= + * 6 - > + */ +# define SDMA_POLL_REG_MEM_EXTRA_M (1 << 15) + /* 0 = register + * 1 = memory + */ +#define SDMA_OPCODE_COND_EXEC 9 +#define SDMA_OPCODE_CONSTANT_FILL 11 +# define SDMA_CONSTANT_FILL_EXTRA_SIZE(x) ((x) << 14) + /* 0 = byte fill + * 2 = DW fill + */ +#define SDMA_OPCODE_GENERATE_PTE_PDE 12 +#define SDMA_OPCODE_TIMESTAMP 13 +# define SDMA_TIMESTAMP_SUB_OPCODE_SET_LOCAL 0 +# define SDMA_TIMESTAMP_SUB_OPCODE_GET_LOCAL 1 +# define SDMA_TIMESTAMP_SUB_OPCODE_GET_GLOBAL 2 +#define SDMA_OPCODE_SRBM_WRITE 14 +# define SDMA_SRBM_WRITE_EXTRA_BYTE_ENABLE(x) ((x) << 12) + /* byte mask */ + +/* UVD */ + +#define UVD_UDEC_ADDR_CONFIG 0xef4c +#define UVD_UDEC_DB_ADDR_CONFIG 0xef50 +#define UVD_UDEC_DBW_ADDR_CONFIG 0xef54 + +#define UVD_LMI_EXT40_ADDR 0xf498 +#define UVD_LMI_ADDR_EXT 0xf594 +#define UVD_VCPU_CACHE_OFFSET0 0xf608 +#define UVD_VCPU_CACHE_SIZE0 0xf60c +#define UVD_VCPU_CACHE_OFFSET1 0xf610 +#define UVD_VCPU_CACHE_SIZE1 0xf614 +#define UVD_VCPU_CACHE_OFFSET2 0xf618 +#define UVD_VCPU_CACHE_SIZE2 0xf61c + +#define UVD_RBC_RB_RPTR 0xf690 +#define UVD_RBC_RB_WPTR 0xf694 + +/* UVD clocks */ + +#define CG_DCLK_CNTL 0xC050009C +# define DCLK_DIVIDER_MASK 0x7f +# define DCLK_DIR_CNTL_EN (1 << 8) +#define CG_DCLK_STATUS 0xC05000A0 +# define DCLK_STATUS (1 << 0) +#define CG_VCLK_CNTL 0xC05000A4 +#define CG_VCLK_STATUS 0xC05000A8 #endif diff --git a/sys/dev/drm/radeon/clearstate_cayman.h b/sys/dev/drm/radeon/clearstate_cayman.h new file mode 100644 index 0000000000..c00339440c --- /dev/null +++ b/sys/dev/drm/radeon/clearstate_cayman.h @@ -0,0 +1,1081 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +static const u32 SECT_CONTEXT_def_1[] = +{ + 0x00000000, // DB_RENDER_CONTROL + 0x00000000, // DB_COUNT_CONTROL + 0x00000000, // DB_DEPTH_VIEW + 0x00000000, // DB_RENDER_OVERRIDE + 0x00000000, // DB_RENDER_OVERRIDE2 + 0x00000000, // DB_HTILE_DATA_BASE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCIL_CLEAR + 0x00000000, // DB_DEPTH_CLEAR + 0x00000000, // PA_SC_SCREEN_SCISSOR_TL + 0x40004000, // PA_SC_SCREEN_SCISSOR_BR + 0, // HOLE + 0x00000000, // DB_DEPTH_INFO + 0x00000000, // DB_Z_INFO + 0x00000000, // DB_STENCIL_INFO + 0x00000000, // DB_Z_READ_BASE + 0x00000000, // DB_STENCIL_READ_BASE + 0x00000000, // DB_Z_WRITE_BASE + 0x00000000, // DB_STENCIL_WRITE_BASE + 0x00000000, // DB_DEPTH_SIZE + 0x00000000, // DB_DEPTH_SLICE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_15 + 0x00000000, // PA_SC_WINDOW_OFFSET + 0x80000000, // PA_SC_WINDOW_SCISSOR_TL + 0x40004000, // PA_SC_WINDOW_SCISSOR_BR + 0x0000ffff, // PA_SC_CLIPRECT_RULE + 0x00000000, // PA_SC_CLIPRECT_0_TL + 0x40004000, // PA_SC_CLIPRECT_0_BR + 0x00000000, // PA_SC_CLIPRECT_1_TL + 0x40004000, // PA_SC_CLIPRECT_1_BR + 0x00000000, // PA_SC_CLIPRECT_2_TL + 0x40004000, // PA_SC_CLIPRECT_2_BR + 0x00000000, // PA_SC_CLIPRECT_3_TL + 0x40004000, // PA_SC_CLIPRECT_3_BR + 0xaa99aaaa, // PA_SC_EDGERULE + 0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET + 0xffffffff, // CB_TARGET_MASK + 0xffffffff, // CB_SHADER_MASK + 0x80000000, // PA_SC_GENERIC_SCISSOR_TL + 0x40004000, // PA_SC_GENERIC_SCISSOR_BR + 0x00000000, // COHER_DEST_BASE_0 + 0x00000000, // COHER_DEST_BASE_1 + 0x80000000, // PA_SC_VPORT_SCISSOR_0_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_0_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_1_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_1_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_2_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_2_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_3_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_3_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_4_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_4_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_5_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_5_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_6_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_6_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_7_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_7_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_8_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_8_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_9_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_9_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_10_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_10_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_11_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_11_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_12_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_12_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_13_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_13_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_14_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_14_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_15_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_15_BR + 0x00000000, // PA_SC_VPORT_ZMIN_0 + 0x3f800000, // PA_SC_VPORT_ZMAX_0 + 0x00000000, // PA_SC_VPORT_ZMIN_1 + 0x3f800000, // PA_SC_VPORT_ZMAX_1 + 0x00000000, // PA_SC_VPORT_ZMIN_2 + 0x3f800000, // PA_SC_VPORT_ZMAX_2 + 0x00000000, // PA_SC_VPORT_ZMIN_3 + 0x3f800000, // PA_SC_VPORT_ZMAX_3 + 0x00000000, // PA_SC_VPORT_ZMIN_4 + 0x3f800000, // PA_SC_VPORT_ZMAX_4 + 0x00000000, // PA_SC_VPORT_ZMIN_5 + 0x3f800000, // PA_SC_VPORT_ZMAX_5 + 0x00000000, // PA_SC_VPORT_ZMIN_6 + 0x3f800000, // PA_SC_VPORT_ZMAX_6 + 0x00000000, // PA_SC_VPORT_ZMIN_7 + 0x3f800000, // PA_SC_VPORT_ZMAX_7 + 0x00000000, // PA_SC_VPORT_ZMIN_8 + 0x3f800000, // PA_SC_VPORT_ZMAX_8 + 0x00000000, // PA_SC_VPORT_ZMIN_9 + 0x3f800000, // PA_SC_VPORT_ZMAX_9 + 0x00000000, // PA_SC_VPORT_ZMIN_10 + 0x3f800000, // PA_SC_VPORT_ZMAX_10 + 0x00000000, // PA_SC_VPORT_ZMIN_11 + 0x3f800000, // PA_SC_VPORT_ZMAX_11 + 0x00000000, // PA_SC_VPORT_ZMIN_12 + 0x3f800000, // PA_SC_VPORT_ZMAX_12 + 0x00000000, // PA_SC_VPORT_ZMIN_13 + 0x3f800000, // PA_SC_VPORT_ZMAX_13 + 0x00000000, // PA_SC_VPORT_ZMIN_14 + 0x3f800000, // PA_SC_VPORT_ZMAX_14 + 0x00000000, // PA_SC_VPORT_ZMIN_15 + 0x3f800000, // PA_SC_VPORT_ZMAX_15 + 0x00000000, // SX_MISC + 0x00000000, // SX_SURFACE_SYNC + 0x00000000, // SX_SCATTER_EXPORT_BASE + 0x00000000, // SX_SCATTER_EXPORT_SIZE + 0x00000000, // CP_PERFMON_CNTX_CNTL + 0x00000000, // CP_RINGID + 0x00000000, // CP_VMID + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_VTX_SEMANTIC_0 + 0x00000000, // SQ_VTX_SEMANTIC_1 + 0x00000000, // SQ_VTX_SEMANTIC_2 + 0x00000000, // SQ_VTX_SEMANTIC_3 + 0x00000000, // SQ_VTX_SEMANTIC_4 + 0x00000000, // SQ_VTX_SEMANTIC_5 + 0x00000000, // SQ_VTX_SEMANTIC_6 + 0x00000000, // SQ_VTX_SEMANTIC_7 + 0x00000000, // SQ_VTX_SEMANTIC_8 + 0x00000000, // SQ_VTX_SEMANTIC_9 + 0x00000000, // SQ_VTX_SEMANTIC_10 + 0x00000000, // SQ_VTX_SEMANTIC_11 + 0x00000000, // SQ_VTX_SEMANTIC_12 + 0x00000000, // SQ_VTX_SEMANTIC_13 + 0x00000000, // SQ_VTX_SEMANTIC_14 + 0x00000000, // SQ_VTX_SEMANTIC_15 + 0x00000000, // SQ_VTX_SEMANTIC_16 + 0x00000000, // SQ_VTX_SEMANTIC_17 + 0x00000000, // SQ_VTX_SEMANTIC_18 + 0x00000000, // SQ_VTX_SEMANTIC_19 + 0x00000000, // SQ_VTX_SEMANTIC_20 + 0x00000000, // SQ_VTX_SEMANTIC_21 + 0x00000000, // SQ_VTX_SEMANTIC_22 + 0x00000000, // SQ_VTX_SEMANTIC_23 + 0x00000000, // SQ_VTX_SEMANTIC_24 + 0x00000000, // SQ_VTX_SEMANTIC_25 + 0x00000000, // SQ_VTX_SEMANTIC_26 + 0x00000000, // SQ_VTX_SEMANTIC_27 + 0x00000000, // SQ_VTX_SEMANTIC_28 + 0x00000000, // SQ_VTX_SEMANTIC_29 + 0x00000000, // SQ_VTX_SEMANTIC_30 + 0x00000000, // SQ_VTX_SEMANTIC_31 + 0xffffffff, // VGT_MAX_VTX_INDX + 0x00000000, // VGT_MIN_VTX_INDX + 0x00000000, // VGT_INDX_OFFSET + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX + 0x00000000, // SX_ALPHA_TEST_CONTROL + 0x00000000, // CB_BLEND_RED + 0x00000000, // CB_BLEND_GREEN + 0x00000000, // CB_BLEND_BLUE + 0x00000000, // CB_BLEND_ALPHA + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCILREFMASK + 0x00000000, // DB_STENCILREFMASK_BF + 0x00000000, // SX_ALPHA_REF + 0x00000000, // PA_CL_VPORT_XSCALE + 0x00000000, // PA_CL_VPORT_XOFFSET + 0x00000000, // PA_CL_VPORT_YSCALE + 0x00000000, // PA_CL_VPORT_YOFFSET + 0x00000000, // PA_CL_VPORT_ZSCALE + 0x00000000, // PA_CL_VPORT_ZOFFSET + 0x00000000, // PA_CL_VPORT_XSCALE_1 + 0x00000000, // PA_CL_VPORT_XOFFSET_1 + 0x00000000, // PA_CL_VPORT_YSCALE_1 + 0x00000000, // PA_CL_VPORT_YOFFSET_1 + 0x00000000, // PA_CL_VPORT_ZSCALE_1 + 0x00000000, // PA_CL_VPORT_ZOFFSET_1 + 0x00000000, // PA_CL_VPORT_XSCALE_2 + 0x00000000, // PA_CL_VPORT_XOFFSET_2 + 0x00000000, // PA_CL_VPORT_YSCALE_2 + 0x00000000, // PA_CL_VPORT_YOFFSET_2 + 0x00000000, // PA_CL_VPORT_ZSCALE_2 + 0x00000000, // PA_CL_VPORT_ZOFFSET_2 + 0x00000000, // PA_CL_VPORT_XSCALE_3 + 0x00000000, // PA_CL_VPORT_XOFFSET_3 + 0x00000000, // PA_CL_VPORT_YSCALE_3 + 0x00000000, // PA_CL_VPORT_YOFFSET_3 + 0x00000000, // PA_CL_VPORT_ZSCALE_3 + 0x00000000, // PA_CL_VPORT_ZOFFSET_3 + 0x00000000, // PA_CL_VPORT_XSCALE_4 + 0x00000000, // PA_CL_VPORT_XOFFSET_4 + 0x00000000, // PA_CL_VPORT_YSCALE_4 + 0x00000000, // PA_CL_VPORT_YOFFSET_4 + 0x00000000, // PA_CL_VPORT_ZSCALE_4 + 0x00000000, // PA_CL_VPORT_ZOFFSET_4 + 0x00000000, // PA_CL_VPORT_XSCALE_5 + 0x00000000, // PA_CL_VPORT_XOFFSET_5 + 0x00000000, // PA_CL_VPORT_YSCALE_5 + 0x00000000, // PA_CL_VPORT_YOFFSET_5 + 0x00000000, // PA_CL_VPORT_ZSCALE_5 + 0x00000000, // PA_CL_VPORT_ZOFFSET_5 + 0x00000000, // PA_CL_VPORT_XSCALE_6 + 0x00000000, // PA_CL_VPORT_XOFFSET_6 + 0x00000000, // PA_CL_VPORT_YSCALE_6 + 0x00000000, // PA_CL_VPORT_YOFFSET_6 + 0x00000000, // PA_CL_VPORT_ZSCALE_6 + 0x00000000, // PA_CL_VPORT_ZOFFSET_6 + 0x00000000, // PA_CL_VPORT_XSCALE_7 + 0x00000000, // PA_CL_VPORT_XOFFSET_7 + 0x00000000, // PA_CL_VPORT_YSCALE_7 + 0x00000000, // PA_CL_VPORT_YOFFSET_7 + 0x00000000, // PA_CL_VPORT_ZSCALE_7 + 0x00000000, // PA_CL_VPORT_ZOFFSET_7 + 0x00000000, // PA_CL_VPORT_XSCALE_8 + 0x00000000, // PA_CL_VPORT_XOFFSET_8 + 0x00000000, // PA_CL_VPORT_YSCALE_8 + 0x00000000, // PA_CL_VPORT_YOFFSET_8 + 0x00000000, // PA_CL_VPORT_ZSCALE_8 + 0x00000000, // PA_CL_VPORT_ZOFFSET_8 + 0x00000000, // PA_CL_VPORT_XSCALE_9 + 0x00000000, // PA_CL_VPORT_XOFFSET_9 + 0x00000000, // PA_CL_VPORT_YSCALE_9 + 0x00000000, // PA_CL_VPORT_YOFFSET_9 + 0x00000000, // PA_CL_VPORT_ZSCALE_9 + 0x00000000, // PA_CL_VPORT_ZOFFSET_9 + 0x00000000, // PA_CL_VPORT_XSCALE_10 + 0x00000000, // PA_CL_VPORT_XOFFSET_10 + 0x00000000, // PA_CL_VPORT_YSCALE_10 + 0x00000000, // PA_CL_VPORT_YOFFSET_10 + 0x00000000, // PA_CL_VPORT_ZSCALE_10 + 0x00000000, // PA_CL_VPORT_ZOFFSET_10 + 0x00000000, // PA_CL_VPORT_XSCALE_11 + 0x00000000, // PA_CL_VPORT_XOFFSET_11 + 0x00000000, // PA_CL_VPORT_YSCALE_11 + 0x00000000, // PA_CL_VPORT_YOFFSET_11 + 0x00000000, // PA_CL_VPORT_ZSCALE_11 + 0x00000000, // PA_CL_VPORT_ZOFFSET_11 + 0x00000000, // PA_CL_VPORT_XSCALE_12 + 0x00000000, // PA_CL_VPORT_XOFFSET_12 + 0x00000000, // PA_CL_VPORT_YSCALE_12 + 0x00000000, // PA_CL_VPORT_YOFFSET_12 + 0x00000000, // PA_CL_VPORT_ZSCALE_12 + 0x00000000, // PA_CL_VPORT_ZOFFSET_12 + 0x00000000, // PA_CL_VPORT_XSCALE_13 + 0x00000000, // PA_CL_VPORT_XOFFSET_13 + 0x00000000, // PA_CL_VPORT_YSCALE_13 + 0x00000000, // PA_CL_VPORT_YOFFSET_13 + 0x00000000, // PA_CL_VPORT_ZSCALE_13 + 0x00000000, // PA_CL_VPORT_ZOFFSET_13 + 0x00000000, // PA_CL_VPORT_XSCALE_14 + 0x00000000, // PA_CL_VPORT_XOFFSET_14 + 0x00000000, // PA_CL_VPORT_YSCALE_14 + 0x00000000, // PA_CL_VPORT_YOFFSET_14 + 0x00000000, // PA_CL_VPORT_ZSCALE_14 + 0x00000000, // PA_CL_VPORT_ZOFFSET_14 + 0x00000000, // PA_CL_VPORT_XSCALE_15 + 0x00000000, // PA_CL_VPORT_XOFFSET_15 + 0x00000000, // PA_CL_VPORT_YSCALE_15 + 0x00000000, // PA_CL_VPORT_YOFFSET_15 + 0x00000000, // PA_CL_VPORT_ZSCALE_15 + 0x00000000, // PA_CL_VPORT_ZOFFSET_15 + 0x00000000, // PA_CL_UCP_0_X + 0x00000000, // PA_CL_UCP_0_Y + 0x00000000, // PA_CL_UCP_0_Z + 0x00000000, // PA_CL_UCP_0_W + 0x00000000, // PA_CL_UCP_1_X + 0x00000000, // PA_CL_UCP_1_Y + 0x00000000, // PA_CL_UCP_1_Z + 0x00000000, // PA_CL_UCP_1_W + 0x00000000, // PA_CL_UCP_2_X + 0x00000000, // PA_CL_UCP_2_Y + 0x00000000, // PA_CL_UCP_2_Z + 0x00000000, // PA_CL_UCP_2_W + 0x00000000, // PA_CL_UCP_3_X + 0x00000000, // PA_CL_UCP_3_Y + 0x00000000, // PA_CL_UCP_3_Z + 0x00000000, // PA_CL_UCP_3_W + 0x00000000, // PA_CL_UCP_4_X + 0x00000000, // PA_CL_UCP_4_Y + 0x00000000, // PA_CL_UCP_4_Z + 0x00000000, // PA_CL_UCP_4_W + 0x00000000, // PA_CL_UCP_5_X + 0x00000000, // PA_CL_UCP_5_Y + 0x00000000, // PA_CL_UCP_5_Z + 0x00000000, // PA_CL_UCP_5_W + 0x00000000, // SPI_VS_OUT_ID_0 + 0x00000000, // SPI_VS_OUT_ID_1 + 0x00000000, // SPI_VS_OUT_ID_2 + 0x00000000, // SPI_VS_OUT_ID_3 + 0x00000000, // SPI_VS_OUT_ID_4 + 0x00000000, // SPI_VS_OUT_ID_5 + 0x00000000, // SPI_VS_OUT_ID_6 + 0x00000000, // SPI_VS_OUT_ID_7 + 0x00000000, // SPI_VS_OUT_ID_8 + 0x00000000, // SPI_VS_OUT_ID_9 + 0x00000000, // SPI_PS_INPUT_CNTL_0 + 0x00000000, // SPI_PS_INPUT_CNTL_1 + 0x00000000, // SPI_PS_INPUT_CNTL_2 + 0x00000000, // SPI_PS_INPUT_CNTL_3 + 0x00000000, // SPI_PS_INPUT_CNTL_4 + 0x00000000, // SPI_PS_INPUT_CNTL_5 + 0x00000000, // SPI_PS_INPUT_CNTL_6 + 0x00000000, // SPI_PS_INPUT_CNTL_7 + 0x00000000, // SPI_PS_INPUT_CNTL_8 + 0x00000000, // SPI_PS_INPUT_CNTL_9 + 0x00000000, // SPI_PS_INPUT_CNTL_10 + 0x00000000, // SPI_PS_INPUT_CNTL_11 + 0x00000000, // SPI_PS_INPUT_CNTL_12 + 0x00000000, // SPI_PS_INPUT_CNTL_13 + 0x00000000, // SPI_PS_INPUT_CNTL_14 + 0x00000000, // SPI_PS_INPUT_CNTL_15 + 0x00000000, // SPI_PS_INPUT_CNTL_16 + 0x00000000, // SPI_PS_INPUT_CNTL_17 + 0x00000000, // SPI_PS_INPUT_CNTL_18 + 0x00000000, // SPI_PS_INPUT_CNTL_19 + 0x00000000, // SPI_PS_INPUT_CNTL_20 + 0x00000000, // SPI_PS_INPUT_CNTL_21 + 0x00000000, // SPI_PS_INPUT_CNTL_22 + 0x00000000, // SPI_PS_INPUT_CNTL_23 + 0x00000000, // SPI_PS_INPUT_CNTL_24 + 0x00000000, // SPI_PS_INPUT_CNTL_25 + 0x00000000, // SPI_PS_INPUT_CNTL_26 + 0x00000000, // SPI_PS_INPUT_CNTL_27 + 0x00000000, // SPI_PS_INPUT_CNTL_28 + 0x00000000, // SPI_PS_INPUT_CNTL_29 + 0x00000000, // SPI_PS_INPUT_CNTL_30 + 0x00000000, // SPI_PS_INPUT_CNTL_31 + 0x00000000, // SPI_VS_OUT_CONFIG + 0x00000001, // SPI_THREAD_GROUPING + 0x00000002, // SPI_PS_IN_CONTROL_0 + 0x00000000, // SPI_PS_IN_CONTROL_1 + 0x00000000, // SPI_INTERP_CONTROL_0 + 0x00000000, // SPI_INPUT_Z + 0x00000000, // SPI_FOG_CNTL + 0x00000000, // SPI_BARYC_CNTL + 0x00000000, // SPI_PS_IN_CONTROL_2 + 0x00000000, // SPI_COMPUTE_INPUT_CNTL + 0x00000000, // SPI_COMPUTE_NUM_THREAD_X + 0x00000000, // SPI_COMPUTE_NUM_THREAD_Y + 0x00000000, // SPI_COMPUTE_NUM_THREAD_Z + 0x00000000, // SPI_GPR_MGMT + 0x00000000, // SPI_LDS_MGMT + 0x00000000, // SPI_STACK_MGMT + 0x00000000, // SPI_WAVE_MGMT_1 + 0x00000000, // SPI_WAVE_MGMT_2 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // GDS_ADDR_BASE + 0x00003fff, // GDS_ADDR_SIZE + 0, // HOLE + 0, // HOLE + 0x00000000, // GDS_ORDERED_COUNT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // GDS_APPEND_CONSUME_UAV0 + 0x00000000, // GDS_APPEND_CONSUME_UAV1 + 0x00000000, // GDS_APPEND_CONSUME_UAV2 + 0x00000000, // GDS_APPEND_CONSUME_UAV3 + 0x00000000, // GDS_APPEND_CONSUME_UAV4 + 0x00000000, // GDS_APPEND_CONSUME_UAV5 + 0x00000000, // GDS_APPEND_CONSUME_UAV6 + 0x00000000, // GDS_APPEND_CONSUME_UAV7 + 0x00000000, // GDS_APPEND_CONSUME_UAV8 + 0x00000000, // GDS_APPEND_CONSUME_UAV9 + 0x00000000, // GDS_APPEND_CONSUME_UAV10 + 0x00000000, // GDS_APPEND_CONSUME_UAV11 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_BLEND0_CONTROL + 0x00000000, // CB_BLEND1_CONTROL + 0x00000000, // CB_BLEND2_CONTROL + 0x00000000, // CB_BLEND3_CONTROL + 0x00000000, // CB_BLEND4_CONTROL + 0x00000000, // CB_BLEND5_CONTROL + 0x00000000, // CB_BLEND6_CONTROL + 0x00000000, // CB_BLEND7_CONTROL +}; +static const u32 SECT_CONTEXT_def_2[] = +{ + 0x00000000, // PA_CL_POINT_X_RAD + 0x00000000, // PA_CL_POINT_Y_RAD + 0x00000000, // PA_CL_POINT_SIZE + 0x00000000, // PA_CL_POINT_CULL_RAD + 0x00000000, // VGT_DMA_BASE_HI + 0x00000000, // VGT_DMA_BASE +}; +static const u32 SECT_CONTEXT_def_3[] = +{ + 0x00000000, // DB_DEPTH_CONTROL + 0x00000000, // DB_EQAA + 0x00000000, // CB_COLOR_CONTROL + 0x00000200, // DB_SHADER_CONTROL + 0x00000000, // PA_CL_CLIP_CNTL + 0x00000000, // PA_SU_SC_MODE_CNTL + 0x00000000, // PA_CL_VTE_CNTL + 0x00000000, // PA_CL_VS_OUT_CNTL + 0x00000000, // PA_CL_NANINF_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_SCALE + 0x00000000, // PA_SU_PRIM_FILTER_CNTL + 0x00000000, // SQ_LSTMP_RING_ITEMSIZE + 0x00000000, // SQ_HSTMP_RING_ITEMSIZE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_PS + 0x00000000, // SQ_PGM_RESOURCES_PS + 0x00000000, // SQ_PGM_RESOURCES_2_PS + 0x00000000, // SQ_PGM_EXPORTS_PS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_VS + 0x00000000, // SQ_PGM_RESOURCES_VS + 0x00000000, // SQ_PGM_RESOURCES_2_VS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_GS + 0x00000000, // SQ_PGM_RESOURCES_GS + 0x00000000, // SQ_PGM_RESOURCES_2_GS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_ES + 0x00000000, // SQ_PGM_RESOURCES_ES + 0x00000000, // SQ_PGM_RESOURCES_2_ES + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_FS + 0x00000000, // SQ_PGM_RESOURCES_FS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_HS + 0x00000000, // SQ_PGM_RESOURCES_HS + 0x00000000, // SQ_PGM_RESOURCES_2_HS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_LS + 0x00000000, // SQ_PGM_RESOURCES_LS + 0x00000000, // SQ_PGM_RESOURCES_2_LS +}; +static const u32 SECT_CONTEXT_def_4[] = +{ + 0x00000000, // SQ_LDS_ALLOC + 0x00000000, // SQ_LDS_ALLOC_PS + 0x00000000, // SQ_VTX_SEMANTIC_CLEAR + 0, // HOLE + 0x00000000, // SQ_THREAD_TRACE_CTRL + 0, // HOLE + 0x00000000, // SQ_ESGS_RING_ITEMSIZE + 0x00000000, // SQ_GSVS_RING_ITEMSIZE + 0x00000000, // SQ_ESTMP_RING_ITEMSIZE + 0x00000000, // SQ_GSTMP_RING_ITEMSIZE + 0x00000000, // SQ_VSTMP_RING_ITEMSIZE + 0x00000000, // SQ_PSTMP_RING_ITEMSIZE + 0, // HOLE + 0x00000000, // SQ_GS_VERT_ITEMSIZE + 0x00000000, // SQ_GS_VERT_ITEMSIZE_1 + 0x00000000, // SQ_GS_VERT_ITEMSIZE_2 + 0x00000000, // SQ_GS_VERT_ITEMSIZE_3 + 0x00000000, // SQ_GSVS_RING_OFFSET_1 + 0x00000000, // SQ_GSVS_RING_OFFSET_2 + 0x00000000, // SQ_GSVS_RING_OFFSET_3 + 0x00000000, // SQ_GWS_RING_OFFSET + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_CACHE_PS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_15 + 0x00000000, // PA_SU_POINT_SIZE + 0x00000000, // PA_SU_POINT_MINMAX + 0x00000000, // PA_SU_LINE_CNTL + 0x00000000, // PA_SC_LINE_STIPPLE + 0x00000000, // VGT_OUTPUT_PATH_CNTL + 0x00000000, // VGT_HOS_CNTL + 0x00000000, // VGT_HOS_MAX_TESS_LEVEL + 0x00000000, // VGT_HOS_MIN_TESS_LEVEL + 0x00000000, // VGT_HOS_REUSE_DEPTH + 0x00000000, // VGT_GROUP_PRIM_TYPE + 0x00000000, // VGT_GROUP_FIRST_DECR + 0x00000000, // VGT_GROUP_DECR + 0x00000000, // VGT_GROUP_VECT_0_CNTL + 0x00000000, // VGT_GROUP_VECT_1_CNTL + 0x00000000, // VGT_GROUP_VECT_0_FMT_CNTL + 0x00000000, // VGT_GROUP_VECT_1_FMT_CNTL + 0x00000000, // VGT_GS_MODE + 0, // HOLE + 0x00000000, // PA_SC_MODE_CNTL_0 + 0x00000000, // PA_SC_MODE_CNTL_1 + 0x00000000, // VGT_ENHANCE + 0x00000100, // VGT_GS_PER_ES + 0x00000080, // VGT_ES_PER_GS + 0x00000002, // VGT_GS_PER_VS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_GS_OUT_PRIM_TYPE + 0x00000000, // IA_ENHANCE +}; +static const u32 SECT_CONTEXT_def_5[] = +{ + 0x00000000, // VGT_DMA_MAX_SIZE + 0x00000000, // VGT_DMA_INDEX_TYPE + 0, // HOLE + 0x00000000, // VGT_PRIMITIVEID_EN + 0x00000000, // VGT_DMA_NUM_INSTANCES +}; +static const u32 SECT_CONTEXT_def_6[] = +{ + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_INSTANCE_STEP_RATE_0 + 0x00000000, // VGT_INSTANCE_STEP_RATE_1 + 0x000000ff, // IA_MULTI_VGT_PARAM + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_REUSE_OFF + 0x00000000, // VGT_VTX_CNT_EN + 0x00000000, // DB_HTILE_SURFACE + 0x00000000, // DB_SRESULTS_COMPARE_STATE0 + 0x00000000, // DB_SRESULTS_COMPARE_STATE1 + 0x00000000, // DB_PRELOAD_CONTROL + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_0 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_0 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_0 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_1 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_1 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_1 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_2 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_2 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_2 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_3 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_3 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_3 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_3 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_3 + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE + 0, // HOLE + 0x00000000, // VGT_GS_MAX_VERT_OUT + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_0 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_1 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_2 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_3 + 0x00000000, // VGT_SHADER_STAGES_EN + 0x00000000, // VGT_LS_HS_CONFIG + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_TF_PARAM + 0x00000000, // DB_ALPHA_TO_MASK +}; +static const u32 SECT_CONTEXT_def_7[] = +{ + 0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL + 0x00000000, // PA_SU_POLY_OFFSET_CLAMP + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET + 0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET + 0x00000000, // VGT_GS_INSTANCE_CNT + 0x00000000, // VGT_STRMOUT_CONFIG + 0x00000000, // VGT_STRMOUT_BUFFER_CONFIG + 0x00000000, // CB_IMMED0_BASE + 0x00000000, // CB_IMMED1_BASE + 0x00000000, // CB_IMMED2_BASE + 0x00000000, // CB_IMMED3_BASE + 0x00000000, // CB_IMMED4_BASE + 0x00000000, // CB_IMMED5_BASE + 0x00000000, // CB_IMMED6_BASE + 0x00000000, // CB_IMMED7_BASE + 0x00000000, // CB_IMMED8_BASE + 0x00000000, // CB_IMMED9_BASE + 0x00000000, // CB_IMMED10_BASE + 0x00000000, // CB_IMMED11_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_CENTROID_PRIORITY_0 + 0x00000000, // PA_SC_CENTROID_PRIORITY_1 + 0x00001000, // PA_SC_LINE_CNTL + 0x00000000, // PA_SC_AA_CONFIG + 0x00000005, // PA_SU_VTX_CNTL + 0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ + 0x3f800000, // PA_CL_GB_VERT_DISC_ADJ + 0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ + 0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3 + 0xffffffff, // PA_SC_AA_MASK_X0Y0_X1Y0 + 0xffffffff, // PA_SC_AA_MASK_X0Y1_X1Y1 + 0x00000000, // CB_CLRCMP_CONTROL + 0x00000000, // CB_CLRCMP_SRC + 0x00000000, // CB_CLRCMP_DST + 0x00000000, // CB_CLRCMP_MSK + 0, // HOLE + 0, // HOLE + 0x0000000e, // VGT_VERTEX_REUSE_BLOCK_CNTL + 0x00000010, // VGT_OUT_DEALLOC_CNTL + 0x00000000, // CB_COLOR0_BASE + 0x00000000, // CB_COLOR0_PITCH + 0x00000000, // CB_COLOR0_SLICE + 0x00000000, // CB_COLOR0_VIEW + 0x00000000, // CB_COLOR0_INFO + 0x00000000, // CB_COLOR0_ATTRIB + 0x00000000, // CB_COLOR0_DIM + 0x00000000, // CB_COLOR0_CMASK + 0x00000000, // CB_COLOR0_CMASK_SLICE + 0x00000000, // CB_COLOR0_FMASK + 0x00000000, // CB_COLOR0_FMASK_SLICE + 0x00000000, // CB_COLOR0_CLEAR_WORD0 + 0x00000000, // CB_COLOR0_CLEAR_WORD1 + 0x00000000, // CB_COLOR0_CLEAR_WORD2 + 0x00000000, // CB_COLOR0_CLEAR_WORD3 + 0x00000000, // CB_COLOR1_BASE + 0x00000000, // CB_COLOR1_PITCH + 0x00000000, // CB_COLOR1_SLICE + 0x00000000, // CB_COLOR1_VIEW + 0x00000000, // CB_COLOR1_INFO + 0x00000000, // CB_COLOR1_ATTRIB + 0x00000000, // CB_COLOR1_DIM + 0x00000000, // CB_COLOR1_CMASK + 0x00000000, // CB_COLOR1_CMASK_SLICE + 0x00000000, // CB_COLOR1_FMASK + 0x00000000, // CB_COLOR1_FMASK_SLICE + 0x00000000, // CB_COLOR1_CLEAR_WORD0 + 0x00000000, // CB_COLOR1_CLEAR_WORD1 + 0x00000000, // CB_COLOR1_CLEAR_WORD2 + 0x00000000, // CB_COLOR1_CLEAR_WORD3 + 0x00000000, // CB_COLOR2_BASE + 0x00000000, // CB_COLOR2_PITCH + 0x00000000, // CB_COLOR2_SLICE + 0x00000000, // CB_COLOR2_VIEW + 0x00000000, // CB_COLOR2_INFO + 0x00000000, // CB_COLOR2_ATTRIB + 0x00000000, // CB_COLOR2_DIM + 0x00000000, // CB_COLOR2_CMASK + 0x00000000, // CB_COLOR2_CMASK_SLICE + 0x00000000, // CB_COLOR2_FMASK + 0x00000000, // CB_COLOR2_FMASK_SLICE + 0x00000000, // CB_COLOR2_CLEAR_WORD0 + 0x00000000, // CB_COLOR2_CLEAR_WORD1 + 0x00000000, // CB_COLOR2_CLEAR_WORD2 + 0x00000000, // CB_COLOR2_CLEAR_WORD3 + 0x00000000, // CB_COLOR3_BASE + 0x00000000, // CB_COLOR3_PITCH + 0x00000000, // CB_COLOR3_SLICE + 0x00000000, // CB_COLOR3_VIEW + 0x00000000, // CB_COLOR3_INFO + 0x00000000, // CB_COLOR3_ATTRIB + 0x00000000, // CB_COLOR3_DIM + 0x00000000, // CB_COLOR3_CMASK + 0x00000000, // CB_COLOR3_CMASK_SLICE + 0x00000000, // CB_COLOR3_FMASK + 0x00000000, // CB_COLOR3_FMASK_SLICE + 0x00000000, // CB_COLOR3_CLEAR_WORD0 + 0x00000000, // CB_COLOR3_CLEAR_WORD1 + 0x00000000, // CB_COLOR3_CLEAR_WORD2 + 0x00000000, // CB_COLOR3_CLEAR_WORD3 + 0x00000000, // CB_COLOR4_BASE + 0x00000000, // CB_COLOR4_PITCH + 0x00000000, // CB_COLOR4_SLICE + 0x00000000, // CB_COLOR4_VIEW + 0x00000000, // CB_COLOR4_INFO + 0x00000000, // CB_COLOR4_ATTRIB + 0x00000000, // CB_COLOR4_DIM + 0x00000000, // CB_COLOR4_CMASK + 0x00000000, // CB_COLOR4_CMASK_SLICE + 0x00000000, // CB_COLOR4_FMASK + 0x00000000, // CB_COLOR4_FMASK_SLICE + 0x00000000, // CB_COLOR4_CLEAR_WORD0 + 0x00000000, // CB_COLOR4_CLEAR_WORD1 + 0x00000000, // CB_COLOR4_CLEAR_WORD2 + 0x00000000, // CB_COLOR4_CLEAR_WORD3 + 0x00000000, // CB_COLOR5_BASE + 0x00000000, // CB_COLOR5_PITCH + 0x00000000, // CB_COLOR5_SLICE + 0x00000000, // CB_COLOR5_VIEW + 0x00000000, // CB_COLOR5_INFO + 0x00000000, // CB_COLOR5_ATTRIB + 0x00000000, // CB_COLOR5_DIM + 0x00000000, // CB_COLOR5_CMASK + 0x00000000, // CB_COLOR5_CMASK_SLICE + 0x00000000, // CB_COLOR5_FMASK + 0x00000000, // CB_COLOR5_FMASK_SLICE + 0x00000000, // CB_COLOR5_CLEAR_WORD0 + 0x00000000, // CB_COLOR5_CLEAR_WORD1 + 0x00000000, // CB_COLOR5_CLEAR_WORD2 + 0x00000000, // CB_COLOR5_CLEAR_WORD3 + 0x00000000, // CB_COLOR6_BASE + 0x00000000, // CB_COLOR6_PITCH + 0x00000000, // CB_COLOR6_SLICE + 0x00000000, // CB_COLOR6_VIEW + 0x00000000, // CB_COLOR6_INFO + 0x00000000, // CB_COLOR6_ATTRIB + 0x00000000, // CB_COLOR6_DIM + 0x00000000, // CB_COLOR6_CMASK + 0x00000000, // CB_COLOR6_CMASK_SLICE + 0x00000000, // CB_COLOR6_FMASK + 0x00000000, // CB_COLOR6_FMASK_SLICE + 0x00000000, // CB_COLOR6_CLEAR_WORD0 + 0x00000000, // CB_COLOR6_CLEAR_WORD1 + 0x00000000, // CB_COLOR6_CLEAR_WORD2 + 0x00000000, // CB_COLOR6_CLEAR_WORD3 + 0x00000000, // CB_COLOR7_BASE + 0x00000000, // CB_COLOR7_PITCH + 0x00000000, // CB_COLOR7_SLICE + 0x00000000, // CB_COLOR7_VIEW + 0x00000000, // CB_COLOR7_INFO + 0x00000000, // CB_COLOR7_ATTRIB + 0x00000000, // CB_COLOR7_DIM + 0x00000000, // CB_COLOR7_CMASK + 0x00000000, // CB_COLOR7_CMASK_SLICE + 0x00000000, // CB_COLOR7_FMASK + 0x00000000, // CB_COLOR7_FMASK_SLICE + 0x00000000, // CB_COLOR7_CLEAR_WORD0 + 0x00000000, // CB_COLOR7_CLEAR_WORD1 + 0x00000000, // CB_COLOR7_CLEAR_WORD2 + 0x00000000, // CB_COLOR7_CLEAR_WORD3 + 0x00000000, // CB_COLOR8_BASE + 0x00000000, // CB_COLOR8_PITCH + 0x00000000, // CB_COLOR8_SLICE + 0x00000000, // CB_COLOR8_VIEW + 0x00000000, // CB_COLOR8_INFO + 0x00000000, // CB_COLOR8_ATTRIB + 0x00000000, // CB_COLOR8_DIM + 0x00000000, // CB_COLOR9_BASE + 0x00000000, // CB_COLOR9_PITCH + 0x00000000, // CB_COLOR9_SLICE + 0x00000000, // CB_COLOR9_VIEW + 0x00000000, // CB_COLOR9_INFO + 0x00000000, // CB_COLOR9_ATTRIB + 0x00000000, // CB_COLOR9_DIM + 0x00000000, // CB_COLOR10_BASE + 0x00000000, // CB_COLOR10_PITCH + 0x00000000, // CB_COLOR10_SLICE + 0x00000000, // CB_COLOR10_VIEW + 0x00000000, // CB_COLOR10_INFO + 0x00000000, // CB_COLOR10_ATTRIB + 0x00000000, // CB_COLOR10_DIM + 0x00000000, // CB_COLOR11_BASE + 0x00000000, // CB_COLOR11_PITCH + 0x00000000, // CB_COLOR11_SLICE + 0x00000000, // CB_COLOR11_VIEW + 0x00000000, // CB_COLOR11_INFO + 0x00000000, // CB_COLOR11_ATTRIB + 0x00000000, // CB_COLOR11_DIM + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_CACHE_HS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_15 +}; +static const struct cs_extent_def SECT_CONTEXT_defs[] = +{ + {SECT_CONTEXT_def_1, 0x0000a000, 488 }, + {SECT_CONTEXT_def_2, 0x0000a1f5, 6 }, + {SECT_CONTEXT_def_3, 0x0000a200, 55 }, + {SECT_CONTEXT_def_4, 0x0000a23a, 99 }, + {SECT_CONTEXT_def_5, 0x0000a29e, 5 }, + {SECT_CONTEXT_def_6, 0x0000a2a5, 56 }, + {SECT_CONTEXT_def_7, 0x0000a2de, 290 }, + { 0, 0, 0 } +}; +static const u32 SECT_CLEAR_def_1[] = +{ + 0xffffffff, // SQ_TEX_SAMPLER_CLEAR + 0xffffffff, // SQ_TEX_RESOURCE_CLEAR + 0xffffffff, // SQ_LOOP_BOOL_CLEAR +}; +static const struct cs_extent_def SECT_CLEAR_defs[] = +{ + {SECT_CLEAR_def_1, 0x0000ffc0, 3 }, + { 0, 0, 0 } +}; +static const u32 SECT_CTRLCONST_def_1[] = +{ + 0x00000000, // SQ_VTX_BASE_VTX_LOC + 0x00000000, // SQ_VTX_START_INST_LOC +}; +static const struct cs_extent_def SECT_CTRLCONST_defs[] = +{ + {SECT_CTRLCONST_def_1, 0x0000f3fc, 2 }, + { 0, 0, 0 } +}; +struct cs_section_def cayman_cs_data[] = { + { SECT_CONTEXT_defs, SECT_CONTEXT }, + { SECT_CLEAR_defs, SECT_CLEAR }, + { SECT_CTRLCONST_defs, SECT_CTRLCONST }, + { 0, SECT_NONE } +}; diff --git a/sys/dev/drm/radeon/atom-types.h b/sys/dev/drm/radeon/clearstate_defs.h similarity index 72% copy from sys/dev/drm/radeon/atom-types.h copy to sys/dev/drm/radeon/clearstate_defs.h index 2a8ae9c5ba..3eda707d73 100644 --- a/sys/dev/drm/radeon/atom-types.h +++ b/sys/dev/drm/radeon/clearstate_defs.h @@ -1,5 +1,5 @@ /* - * Copyright 2008 Red Hat Inc. + * Copyright 2012 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -19,26 +19,26 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR * OTHER DEALINGS IN THE SOFTWARE. * - * Author: Dave Airlie - * - * $FreeBSD: head/sys/dev/drm2/radeon/atom-types.h 254885 2013-08-25 19:37:15Z dumbbell $ */ +#ifndef CLEARSTATE_DEFS_H +#define CLEARSTATE_DEFS_H -#ifndef ATOM_TYPES_H -#define ATOM_TYPES_H - -/* sync atom types to kernel types */ +enum section_id { + SECT_NONE, + SECT_CONTEXT, + SECT_CLEAR, + SECT_CTRLCONST +}; -typedef uint16_t USHORT; -typedef uint32_t ULONG; -typedef uint8_t UCHAR; +struct cs_extent_def { + const unsigned int *extent; + const unsigned int reg_index; + const unsigned int reg_count; +}; +struct cs_section_def { + const struct cs_extent_def *section; + const enum section_id id; +}; -#ifndef ATOM_BIG_ENDIAN -#if defined(__BIG_ENDIAN) -#define ATOM_BIG_ENDIAN 1 -#else -#define ATOM_BIG_ENDIAN 0 -#endif -#endif #endif diff --git a/sys/dev/drm/radeon/clearstate_evergreen.h b/sys/dev/drm/radeon/clearstate_evergreen.h new file mode 100644 index 0000000000..4791d856b7 --- /dev/null +++ b/sys/dev/drm/radeon/clearstate_evergreen.h @@ -0,0 +1,1080 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +static const u32 SECT_CONTEXT_def_1[] = +{ + 0x00000000, // DB_RENDER_CONTROL + 0x00000000, // DB_COUNT_CONTROL + 0x00000000, // DB_DEPTH_VIEW + 0x00000000, // DB_RENDER_OVERRIDE + 0x00000000, // DB_RENDER_OVERRIDE2 + 0x00000000, // DB_HTILE_DATA_BASE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCIL_CLEAR + 0x00000000, // DB_DEPTH_CLEAR + 0x00000000, // PA_SC_SCREEN_SCISSOR_TL + 0x40004000, // PA_SC_SCREEN_SCISSOR_BR + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_Z_INFO + 0x00000000, // DB_STENCIL_INFO + 0x00000000, // DB_Z_READ_BASE + 0x00000000, // DB_STENCIL_READ_BASE + 0x00000000, // DB_Z_WRITE_BASE + 0x00000000, // DB_STENCIL_WRITE_BASE + 0x00000000, // DB_DEPTH_SIZE + 0x00000000, // DB_DEPTH_SLICE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_PS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_VS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_GS_15 + 0x00000000, // PA_SC_WINDOW_OFFSET + 0x80000000, // PA_SC_WINDOW_SCISSOR_TL + 0x40004000, // PA_SC_WINDOW_SCISSOR_BR + 0x0000ffff, // PA_SC_CLIPRECT_RULE + 0x00000000, // PA_SC_CLIPRECT_0_TL + 0x40004000, // PA_SC_CLIPRECT_0_BR + 0x00000000, // PA_SC_CLIPRECT_1_TL + 0x40004000, // PA_SC_CLIPRECT_1_BR + 0x00000000, // PA_SC_CLIPRECT_2_TL + 0x40004000, // PA_SC_CLIPRECT_2_BR + 0x00000000, // PA_SC_CLIPRECT_3_TL + 0x40004000, // PA_SC_CLIPRECT_3_BR + 0xaa99aaaa, // PA_SC_EDGERULE + 0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET + 0xffffffff, // CB_TARGET_MASK + 0xffffffff, // CB_SHADER_MASK + 0x80000000, // PA_SC_GENERIC_SCISSOR_TL + 0x40004000, // PA_SC_GENERIC_SCISSOR_BR + 0x00000000, // COHER_DEST_BASE_0 + 0x00000000, // COHER_DEST_BASE_1 + 0x80000000, // PA_SC_VPORT_SCISSOR_0_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_0_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_1_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_1_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_2_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_2_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_3_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_3_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_4_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_4_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_5_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_5_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_6_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_6_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_7_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_7_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_8_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_8_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_9_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_9_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_10_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_10_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_11_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_11_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_12_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_12_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_13_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_13_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_14_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_14_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_15_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_15_BR + 0x00000000, // PA_SC_VPORT_ZMIN_0 + 0x3f800000, // PA_SC_VPORT_ZMAX_0 + 0x00000000, // PA_SC_VPORT_ZMIN_1 + 0x3f800000, // PA_SC_VPORT_ZMAX_1 + 0x00000000, // PA_SC_VPORT_ZMIN_2 + 0x3f800000, // PA_SC_VPORT_ZMAX_2 + 0x00000000, // PA_SC_VPORT_ZMIN_3 + 0x3f800000, // PA_SC_VPORT_ZMAX_3 + 0x00000000, // PA_SC_VPORT_ZMIN_4 + 0x3f800000, // PA_SC_VPORT_ZMAX_4 + 0x00000000, // PA_SC_VPORT_ZMIN_5 + 0x3f800000, // PA_SC_VPORT_ZMAX_5 + 0x00000000, // PA_SC_VPORT_ZMIN_6 + 0x3f800000, // PA_SC_VPORT_ZMAX_6 + 0x00000000, // PA_SC_VPORT_ZMIN_7 + 0x3f800000, // PA_SC_VPORT_ZMAX_7 + 0x00000000, // PA_SC_VPORT_ZMIN_8 + 0x3f800000, // PA_SC_VPORT_ZMAX_8 + 0x00000000, // PA_SC_VPORT_ZMIN_9 + 0x3f800000, // PA_SC_VPORT_ZMAX_9 + 0x00000000, // PA_SC_VPORT_ZMIN_10 + 0x3f800000, // PA_SC_VPORT_ZMAX_10 + 0x00000000, // PA_SC_VPORT_ZMIN_11 + 0x3f800000, // PA_SC_VPORT_ZMAX_11 + 0x00000000, // PA_SC_VPORT_ZMIN_12 + 0x3f800000, // PA_SC_VPORT_ZMAX_12 + 0x00000000, // PA_SC_VPORT_ZMIN_13 + 0x3f800000, // PA_SC_VPORT_ZMAX_13 + 0x00000000, // PA_SC_VPORT_ZMIN_14 + 0x3f800000, // PA_SC_VPORT_ZMAX_14 + 0x00000000, // PA_SC_VPORT_ZMIN_15 + 0x3f800000, // PA_SC_VPORT_ZMAX_15 + 0x00000000, // SX_MISC + 0x00000000, // SX_SURFACE_SYNC + 0x00000000, // CP_PERFMON_CNTX_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_VTX_SEMANTIC_0 + 0x00000000, // SQ_VTX_SEMANTIC_1 + 0x00000000, // SQ_VTX_SEMANTIC_2 + 0x00000000, // SQ_VTX_SEMANTIC_3 + 0x00000000, // SQ_VTX_SEMANTIC_4 + 0x00000000, // SQ_VTX_SEMANTIC_5 + 0x00000000, // SQ_VTX_SEMANTIC_6 + 0x00000000, // SQ_VTX_SEMANTIC_7 + 0x00000000, // SQ_VTX_SEMANTIC_8 + 0x00000000, // SQ_VTX_SEMANTIC_9 + 0x00000000, // SQ_VTX_SEMANTIC_10 + 0x00000000, // SQ_VTX_SEMANTIC_11 + 0x00000000, // SQ_VTX_SEMANTIC_12 + 0x00000000, // SQ_VTX_SEMANTIC_13 + 0x00000000, // SQ_VTX_SEMANTIC_14 + 0x00000000, // SQ_VTX_SEMANTIC_15 + 0x00000000, // SQ_VTX_SEMANTIC_16 + 0x00000000, // SQ_VTX_SEMANTIC_17 + 0x00000000, // SQ_VTX_SEMANTIC_18 + 0x00000000, // SQ_VTX_SEMANTIC_19 + 0x00000000, // SQ_VTX_SEMANTIC_20 + 0x00000000, // SQ_VTX_SEMANTIC_21 + 0x00000000, // SQ_VTX_SEMANTIC_22 + 0x00000000, // SQ_VTX_SEMANTIC_23 + 0x00000000, // SQ_VTX_SEMANTIC_24 + 0x00000000, // SQ_VTX_SEMANTIC_25 + 0x00000000, // SQ_VTX_SEMANTIC_26 + 0x00000000, // SQ_VTX_SEMANTIC_27 + 0x00000000, // SQ_VTX_SEMANTIC_28 + 0x00000000, // SQ_VTX_SEMANTIC_29 + 0x00000000, // SQ_VTX_SEMANTIC_30 + 0x00000000, // SQ_VTX_SEMANTIC_31 + 0xffffffff, // VGT_MAX_VTX_INDX + 0x00000000, // VGT_MIN_VTX_INDX + 0x00000000, // VGT_INDX_OFFSET + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX + 0x00000000, // SX_ALPHA_TEST_CONTROL + 0x00000000, // CB_BLEND_RED + 0x00000000, // CB_BLEND_GREEN + 0x00000000, // CB_BLEND_BLUE + 0x00000000, // CB_BLEND_ALPHA + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCILREFMASK + 0x00000000, // DB_STENCILREFMASK_BF + 0x00000000, // SX_ALPHA_REF + 0x00000000, // PA_CL_VPORT_XSCALE + 0x00000000, // PA_CL_VPORT_XOFFSET + 0x00000000, // PA_CL_VPORT_YSCALE + 0x00000000, // PA_CL_VPORT_YOFFSET + 0x00000000, // PA_CL_VPORT_ZSCALE + 0x00000000, // PA_CL_VPORT_ZOFFSET + 0x00000000, // PA_CL_VPORT_XSCALE_1 + 0x00000000, // PA_CL_VPORT_XOFFSET_1 + 0x00000000, // PA_CL_VPORT_YSCALE_1 + 0x00000000, // PA_CL_VPORT_YOFFSET_1 + 0x00000000, // PA_CL_VPORT_ZSCALE_1 + 0x00000000, // PA_CL_VPORT_ZOFFSET_1 + 0x00000000, // PA_CL_VPORT_XSCALE_2 + 0x00000000, // PA_CL_VPORT_XOFFSET_2 + 0x00000000, // PA_CL_VPORT_YSCALE_2 + 0x00000000, // PA_CL_VPORT_YOFFSET_2 + 0x00000000, // PA_CL_VPORT_ZSCALE_2 + 0x00000000, // PA_CL_VPORT_ZOFFSET_2 + 0x00000000, // PA_CL_VPORT_XSCALE_3 + 0x00000000, // PA_CL_VPORT_XOFFSET_3 + 0x00000000, // PA_CL_VPORT_YSCALE_3 + 0x00000000, // PA_CL_VPORT_YOFFSET_3 + 0x00000000, // PA_CL_VPORT_ZSCALE_3 + 0x00000000, // PA_CL_VPORT_ZOFFSET_3 + 0x00000000, // PA_CL_VPORT_XSCALE_4 + 0x00000000, // PA_CL_VPORT_XOFFSET_4 + 0x00000000, // PA_CL_VPORT_YSCALE_4 + 0x00000000, // PA_CL_VPORT_YOFFSET_4 + 0x00000000, // PA_CL_VPORT_ZSCALE_4 + 0x00000000, // PA_CL_VPORT_ZOFFSET_4 + 0x00000000, // PA_CL_VPORT_XSCALE_5 + 0x00000000, // PA_CL_VPORT_XOFFSET_5 + 0x00000000, // PA_CL_VPORT_YSCALE_5 + 0x00000000, // PA_CL_VPORT_YOFFSET_5 + 0x00000000, // PA_CL_VPORT_ZSCALE_5 + 0x00000000, // PA_CL_VPORT_ZOFFSET_5 + 0x00000000, // PA_CL_VPORT_XSCALE_6 + 0x00000000, // PA_CL_VPORT_XOFFSET_6 + 0x00000000, // PA_CL_VPORT_YSCALE_6 + 0x00000000, // PA_CL_VPORT_YOFFSET_6 + 0x00000000, // PA_CL_VPORT_ZSCALE_6 + 0x00000000, // PA_CL_VPORT_ZOFFSET_6 + 0x00000000, // PA_CL_VPORT_XSCALE_7 + 0x00000000, // PA_CL_VPORT_XOFFSET_7 + 0x00000000, // PA_CL_VPORT_YSCALE_7 + 0x00000000, // PA_CL_VPORT_YOFFSET_7 + 0x00000000, // PA_CL_VPORT_ZSCALE_7 + 0x00000000, // PA_CL_VPORT_ZOFFSET_7 + 0x00000000, // PA_CL_VPORT_XSCALE_8 + 0x00000000, // PA_CL_VPORT_XOFFSET_8 + 0x00000000, // PA_CL_VPORT_YSCALE_8 + 0x00000000, // PA_CL_VPORT_YOFFSET_8 + 0x00000000, // PA_CL_VPORT_ZSCALE_8 + 0x00000000, // PA_CL_VPORT_ZOFFSET_8 + 0x00000000, // PA_CL_VPORT_XSCALE_9 + 0x00000000, // PA_CL_VPORT_XOFFSET_9 + 0x00000000, // PA_CL_VPORT_YSCALE_9 + 0x00000000, // PA_CL_VPORT_YOFFSET_9 + 0x00000000, // PA_CL_VPORT_ZSCALE_9 + 0x00000000, // PA_CL_VPORT_ZOFFSET_9 + 0x00000000, // PA_CL_VPORT_XSCALE_10 + 0x00000000, // PA_CL_VPORT_XOFFSET_10 + 0x00000000, // PA_CL_VPORT_YSCALE_10 + 0x00000000, // PA_CL_VPORT_YOFFSET_10 + 0x00000000, // PA_CL_VPORT_ZSCALE_10 + 0x00000000, // PA_CL_VPORT_ZOFFSET_10 + 0x00000000, // PA_CL_VPORT_XSCALE_11 + 0x00000000, // PA_CL_VPORT_XOFFSET_11 + 0x00000000, // PA_CL_VPORT_YSCALE_11 + 0x00000000, // PA_CL_VPORT_YOFFSET_11 + 0x00000000, // PA_CL_VPORT_ZSCALE_11 + 0x00000000, // PA_CL_VPORT_ZOFFSET_11 + 0x00000000, // PA_CL_VPORT_XSCALE_12 + 0x00000000, // PA_CL_VPORT_XOFFSET_12 + 0x00000000, // PA_CL_VPORT_YSCALE_12 + 0x00000000, // PA_CL_VPORT_YOFFSET_12 + 0x00000000, // PA_CL_VPORT_ZSCALE_12 + 0x00000000, // PA_CL_VPORT_ZOFFSET_12 + 0x00000000, // PA_CL_VPORT_XSCALE_13 + 0x00000000, // PA_CL_VPORT_XOFFSET_13 + 0x00000000, // PA_CL_VPORT_YSCALE_13 + 0x00000000, // PA_CL_VPORT_YOFFSET_13 + 0x00000000, // PA_CL_VPORT_ZSCALE_13 + 0x00000000, // PA_CL_VPORT_ZOFFSET_13 + 0x00000000, // PA_CL_VPORT_XSCALE_14 + 0x00000000, // PA_CL_VPORT_XOFFSET_14 + 0x00000000, // PA_CL_VPORT_YSCALE_14 + 0x00000000, // PA_CL_VPORT_YOFFSET_14 + 0x00000000, // PA_CL_VPORT_ZSCALE_14 + 0x00000000, // PA_CL_VPORT_ZOFFSET_14 + 0x00000000, // PA_CL_VPORT_XSCALE_15 + 0x00000000, // PA_CL_VPORT_XOFFSET_15 + 0x00000000, // PA_CL_VPORT_YSCALE_15 + 0x00000000, // PA_CL_VPORT_YOFFSET_15 + 0x00000000, // PA_CL_VPORT_ZSCALE_15 + 0x00000000, // PA_CL_VPORT_ZOFFSET_15 + 0x00000000, // PA_CL_UCP_0_X + 0x00000000, // PA_CL_UCP_0_Y + 0x00000000, // PA_CL_UCP_0_Z + 0x00000000, // PA_CL_UCP_0_W + 0x00000000, // PA_CL_UCP_1_X + 0x00000000, // PA_CL_UCP_1_Y + 0x00000000, // PA_CL_UCP_1_Z + 0x00000000, // PA_CL_UCP_1_W + 0x00000000, // PA_CL_UCP_2_X + 0x00000000, // PA_CL_UCP_2_Y + 0x00000000, // PA_CL_UCP_2_Z + 0x00000000, // PA_CL_UCP_2_W + 0x00000000, // PA_CL_UCP_3_X + 0x00000000, // PA_CL_UCP_3_Y + 0x00000000, // PA_CL_UCP_3_Z + 0x00000000, // PA_CL_UCP_3_W + 0x00000000, // PA_CL_UCP_4_X + 0x00000000, // PA_CL_UCP_4_Y + 0x00000000, // PA_CL_UCP_4_Z + 0x00000000, // PA_CL_UCP_4_W + 0x00000000, // PA_CL_UCP_5_X + 0x00000000, // PA_CL_UCP_5_Y + 0x00000000, // PA_CL_UCP_5_Z + 0x00000000, // PA_CL_UCP_5_W + 0x00000000, // SPI_VS_OUT_ID_0 + 0x00000000, // SPI_VS_OUT_ID_1 + 0x00000000, // SPI_VS_OUT_ID_2 + 0x00000000, // SPI_VS_OUT_ID_3 + 0x00000000, // SPI_VS_OUT_ID_4 + 0x00000000, // SPI_VS_OUT_ID_5 + 0x00000000, // SPI_VS_OUT_ID_6 + 0x00000000, // SPI_VS_OUT_ID_7 + 0x00000000, // SPI_VS_OUT_ID_8 + 0x00000000, // SPI_VS_OUT_ID_9 + 0x00000000, // SPI_PS_INPUT_CNTL_0 + 0x00000000, // SPI_PS_INPUT_CNTL_1 + 0x00000000, // SPI_PS_INPUT_CNTL_2 + 0x00000000, // SPI_PS_INPUT_CNTL_3 + 0x00000000, // SPI_PS_INPUT_CNTL_4 + 0x00000000, // SPI_PS_INPUT_CNTL_5 + 0x00000000, // SPI_PS_INPUT_CNTL_6 + 0x00000000, // SPI_PS_INPUT_CNTL_7 + 0x00000000, // SPI_PS_INPUT_CNTL_8 + 0x00000000, // SPI_PS_INPUT_CNTL_9 + 0x00000000, // SPI_PS_INPUT_CNTL_10 + 0x00000000, // SPI_PS_INPUT_CNTL_11 + 0x00000000, // SPI_PS_INPUT_CNTL_12 + 0x00000000, // SPI_PS_INPUT_CNTL_13 + 0x00000000, // SPI_PS_INPUT_CNTL_14 + 0x00000000, // SPI_PS_INPUT_CNTL_15 + 0x00000000, // SPI_PS_INPUT_CNTL_16 + 0x00000000, // SPI_PS_INPUT_CNTL_17 + 0x00000000, // SPI_PS_INPUT_CNTL_18 + 0x00000000, // SPI_PS_INPUT_CNTL_19 + 0x00000000, // SPI_PS_INPUT_CNTL_20 + 0x00000000, // SPI_PS_INPUT_CNTL_21 + 0x00000000, // SPI_PS_INPUT_CNTL_22 + 0x00000000, // SPI_PS_INPUT_CNTL_23 + 0x00000000, // SPI_PS_INPUT_CNTL_24 + 0x00000000, // SPI_PS_INPUT_CNTL_25 + 0x00000000, // SPI_PS_INPUT_CNTL_26 + 0x00000000, // SPI_PS_INPUT_CNTL_27 + 0x00000000, // SPI_PS_INPUT_CNTL_28 + 0x00000000, // SPI_PS_INPUT_CNTL_29 + 0x00000000, // SPI_PS_INPUT_CNTL_30 + 0x00000000, // SPI_PS_INPUT_CNTL_31 + 0x00000000, // SPI_VS_OUT_CONFIG + 0x00000001, // SPI_THREAD_GROUPING + 0x00000000, // SPI_PS_IN_CONTROL_0 + 0x00000000, // SPI_PS_IN_CONTROL_1 + 0x00000000, // SPI_INTERP_CONTROL_0 + 0x00000000, // SPI_INPUT_Z + 0x00000000, // SPI_FOG_CNTL + 0x00000000, // SPI_BARYC_CNTL + 0x00000000, // SPI_PS_IN_CONTROL_2 + 0x00000000, // SPI_COMPUTE_INPUT_CNTL + 0x00000000, // SPI_COMPUTE_NUM_THREAD_X + 0x00000000, // SPI_COMPUTE_NUM_THREAD_Y + 0x00000000, // SPI_COMPUTE_NUM_THREAD_Z + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // GDS_ADDR_BASE + 0x00003fff, // GDS_ADDR_SIZE + 0x00000001, // GDS_ORDERED_WAVE_PER_SE + 0x00000000, // GDS_APPEND_CONSUME_UAV0 + 0x00000000, // GDS_APPEND_CONSUME_UAV1 + 0x00000000, // GDS_APPEND_CONSUME_UAV2 + 0x00000000, // GDS_APPEND_CONSUME_UAV3 + 0x00000000, // GDS_APPEND_CONSUME_UAV4 + 0x00000000, // GDS_APPEND_CONSUME_UAV5 + 0x00000000, // GDS_APPEND_CONSUME_UAV6 + 0x00000000, // GDS_APPEND_CONSUME_UAV7 + 0x00000000, // GDS_APPEND_CONSUME_UAV8 + 0x00000000, // GDS_APPEND_CONSUME_UAV9 + 0x00000000, // GDS_APPEND_CONSUME_UAV10 + 0x00000000, // GDS_APPEND_CONSUME_UAV11 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_BLEND0_CONTROL + 0x00000000, // CB_BLEND1_CONTROL + 0x00000000, // CB_BLEND2_CONTROL + 0x00000000, // CB_BLEND3_CONTROL + 0x00000000, // CB_BLEND4_CONTROL + 0x00000000, // CB_BLEND5_CONTROL + 0x00000000, // CB_BLEND6_CONTROL + 0x00000000, // CB_BLEND7_CONTROL +}; +static const u32 SECT_CONTEXT_def_2[] = +{ + 0x00000000, // PA_CL_POINT_X_RAD + 0x00000000, // PA_CL_POINT_Y_RAD + 0x00000000, // PA_CL_POINT_SIZE + 0x00000000, // PA_CL_POINT_CULL_RAD + 0x00000000, // VGT_DMA_BASE_HI + 0x00000000, // VGT_DMA_BASE +}; +static const u32 SECT_CONTEXT_def_3[] = +{ + 0x00000000, // DB_DEPTH_CONTROL + 0, // HOLE + 0x00000000, // CB_COLOR_CONTROL + 0x00000200, // DB_SHADER_CONTROL + 0x00000000, // PA_CL_CLIP_CNTL + 0x00000000, // PA_SU_SC_MODE_CNTL + 0x00000000, // PA_CL_VTE_CNTL + 0x00000000, // PA_CL_VS_OUT_CNTL + 0x00000000, // PA_CL_NANINF_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_SCALE + 0x00000000, // PA_SU_PRIM_FILTER_CNTL + 0x00000000, // SQ_LSTMP_RING_ITEMSIZE + 0x00000000, // SQ_HSTMP_RING_ITEMSIZE + 0x00000000, // SQ_DYN_GPR_RESOURCE_LIMIT_1 + 0, // HOLE + 0x00000000, // SQ_PGM_START_PS + 0x00000000, // SQ_PGM_RESOURCES_PS + 0x00000000, // SQ_PGM_RESOURCES_2_PS + 0x00000000, // SQ_PGM_EXPORTS_PS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_VS + 0x00000000, // SQ_PGM_RESOURCES_VS + 0x00000000, // SQ_PGM_RESOURCES_2_VS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_GS + 0x00000000, // SQ_PGM_RESOURCES_GS + 0x00000000, // SQ_PGM_RESOURCES_2_GS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_ES + 0x00000000, // SQ_PGM_RESOURCES_ES + 0x00000000, // SQ_PGM_RESOURCES_2_ES + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_FS + 0x00000000, // SQ_PGM_RESOURCES_FS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_HS + 0x00000000, // SQ_PGM_RESOURCES_HS + 0x00000000, // SQ_PGM_RESOURCES_2_HS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_PGM_START_LS + 0x00000000, // SQ_PGM_RESOURCES_LS + 0x00000000, // SQ_PGM_RESOURCES_2_LS +}; +static const u32 SECT_CONTEXT_def_4[] = +{ + 0x00000000, // SQ_LDS_ALLOC + 0x00000000, // SQ_LDS_ALLOC_PS + 0x00000000, // SQ_VTX_SEMANTIC_CLEAR + 0, // HOLE + 0x00000000, // SQ_THREAD_TRACE_CTRL + 0, // HOLE + 0x00000000, // SQ_ESGS_RING_ITEMSIZE + 0x00000000, // SQ_GSVS_RING_ITEMSIZE + 0x00000000, // SQ_ESTMP_RING_ITEMSIZE + 0x00000000, // SQ_GSTMP_RING_ITEMSIZE + 0x00000000, // SQ_VSTMP_RING_ITEMSIZE + 0x00000000, // SQ_PSTMP_RING_ITEMSIZE + 0, // HOLE + 0x00000000, // SQ_GS_VERT_ITEMSIZE + 0x00000000, // SQ_GS_VERT_ITEMSIZE_1 + 0x00000000, // SQ_GS_VERT_ITEMSIZE_2 + 0x00000000, // SQ_GS_VERT_ITEMSIZE_3 + 0x00000000, // SQ_GSVS_RING_OFFSET_1 + 0x00000000, // SQ_GSVS_RING_OFFSET_2 + 0x00000000, // SQ_GSVS_RING_OFFSET_3 + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_CACHE_PS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_PS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_VS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_GS_15 + 0x00000000, // PA_SU_POINT_SIZE + 0x00000000, // PA_SU_POINT_MINMAX + 0x00000000, // PA_SU_LINE_CNTL + 0x00000000, // PA_SC_LINE_STIPPLE + 0x00000000, // VGT_OUTPUT_PATH_CNTL + 0x00000000, // VGT_HOS_CNTL + 0x00000000, // VGT_HOS_MAX_TESS_LEVEL + 0x00000000, // VGT_HOS_MIN_TESS_LEVEL + 0x00000000, // VGT_HOS_REUSE_DEPTH + 0x00000000, // VGT_GROUP_PRIM_TYPE + 0x00000000, // VGT_GROUP_FIRST_DECR + 0x00000000, // VGT_GROUP_DECR + 0x00000000, // VGT_GROUP_VECT_0_CNTL + 0x00000000, // VGT_GROUP_VECT_1_CNTL + 0x00000000, // VGT_GROUP_VECT_0_FMT_CNTL + 0x00000000, // VGT_GROUP_VECT_1_FMT_CNTL + 0x00000000, // VGT_GS_MODE + 0, // HOLE + 0x00000000, // PA_SC_MODE_CNTL_0 + 0x00000000, // PA_SC_MODE_CNTL_1 + 0x00000000, // VGT_ENHANCE + 0x00000000, // VGT_GS_PER_ES + 0x00000000, // VGT_ES_PER_GS + 0x00000000, // VGT_GS_PER_VS + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_GS_OUT_PRIM_TYPE +}; +static const u32 SECT_CONTEXT_def_5[] = +{ + 0x00000000, // VGT_DMA_MAX_SIZE + 0x00000000, // VGT_DMA_INDEX_TYPE + 0, // HOLE + 0x00000000, // VGT_PRIMITIVEID_EN + 0x00000000, // VGT_DMA_NUM_INSTANCES +}; +static const u32 SECT_CONTEXT_def_6[] = +{ + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_INSTANCE_STEP_RATE_0 + 0x00000000, // VGT_INSTANCE_STEP_RATE_1 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_REUSE_OFF + 0x00000000, // VGT_VTX_CNT_EN + 0x00000000, // DB_HTILE_SURFACE + 0x00000000, // DB_SRESULTS_COMPARE_STATE0 + 0x00000000, // DB_SRESULTS_COMPARE_STATE1 + 0x00000000, // DB_PRELOAD_CONTROL + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_0 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_0 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_0 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_1 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_1 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_1 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_2 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_2 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_2 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_3 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_3 + 0x00000000, // VGT_STRMOUT_BUFFER_BASE_3 + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_3 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_3 + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE + 0, // HOLE + 0x00000000, // VGT_GS_MAX_VERT_OUT + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_0 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_1 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_2 + 0x00000000, // VGT_STRMOUT_BASE_OFFSET_HI_3 + 0x00000000, // VGT_SHADER_STAGES_EN + 0x00000000, // VGT_LS_HS_CONFIG + 0x00000000, // VGT_LS_SIZE + 0x00000000, // VGT_HS_SIZE + 0x00000000, // VGT_LS_HS_ALLOC + 0x00000000, // VGT_HS_PATCH_CONST + 0x00000000, // VGT_TF_PARAM + 0x00000000, // DB_ALPHA_TO_MASK +}; +static const u32 SECT_CONTEXT_def_7[] = +{ + 0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL + 0x00000000, // PA_SU_POLY_OFFSET_CLAMP + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET + 0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET + 0x00000000, // VGT_GS_INSTANCE_CNT + 0x00000000, // VGT_STRMOUT_CONFIG + 0x00000000, // VGT_STRMOUT_BUFFER_CONFIG + 0x00000000, // CB_IMMED0_BASE + 0x00000000, // CB_IMMED1_BASE + 0x00000000, // CB_IMMED2_BASE + 0x00000000, // CB_IMMED3_BASE + 0x00000000, // CB_IMMED4_BASE + 0x00000000, // CB_IMMED5_BASE + 0x00000000, // CB_IMMED6_BASE + 0x00000000, // CB_IMMED7_BASE + 0x00000000, // CB_IMMED8_BASE + 0x00000000, // CB_IMMED9_BASE + 0x00000000, // CB_IMMED10_BASE + 0x00000000, // CB_IMMED11_BASE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00001000, // PA_SC_LINE_CNTL + 0x00000000, // PA_SC_AA_CONFIG + 0x00000005, // PA_SU_VTX_CNTL + 0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ + 0x3f800000, // PA_CL_GB_VERT_DISC_ADJ + 0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ + 0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_4 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_5 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_6 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_7 + 0xffffffff, // PA_SC_AA_MASK + 0x00000000, // CB_CLRCMP_CONTROL + 0x00000000, // CB_CLRCMP_SRC + 0x00000000, // CB_CLRCMP_DST + 0x00000000, // CB_CLRCMP_MSK + 0, // HOLE + 0, // HOLE + 0x0000000e, // VGT_VERTEX_REUSE_BLOCK_CNTL + 0x00000010, // VGT_OUT_DEALLOC_CNTL + 0x00000000, // CB_COLOR0_BASE + 0x00000000, // CB_COLOR0_PITCH + 0x00000000, // CB_COLOR0_SLICE + 0x00000000, // CB_COLOR0_VIEW + 0x00000000, // CB_COLOR0_INFO + 0x00000000, // CB_COLOR0_ATTRIB + 0x00000000, // CB_COLOR0_DIM + 0x00000000, // CB_COLOR0_CMASK + 0x00000000, // CB_COLOR0_CMASK_SLICE + 0x00000000, // CB_COLOR0_FMASK + 0x00000000, // CB_COLOR0_FMASK_SLICE + 0x00000000, // CB_COLOR0_CLEAR_WORD0 + 0x00000000, // CB_COLOR0_CLEAR_WORD1 + 0x00000000, // CB_COLOR0_CLEAR_WORD2 + 0x00000000, // CB_COLOR0_CLEAR_WORD3 + 0x00000000, // CB_COLOR1_BASE + 0x00000000, // CB_COLOR1_PITCH + 0x00000000, // CB_COLOR1_SLICE + 0x00000000, // CB_COLOR1_VIEW + 0x00000000, // CB_COLOR1_INFO + 0x00000000, // CB_COLOR1_ATTRIB + 0x00000000, // CB_COLOR1_DIM + 0x00000000, // CB_COLOR1_CMASK + 0x00000000, // CB_COLOR1_CMASK_SLICE + 0x00000000, // CB_COLOR1_FMASK + 0x00000000, // CB_COLOR1_FMASK_SLICE + 0x00000000, // CB_COLOR1_CLEAR_WORD0 + 0x00000000, // CB_COLOR1_CLEAR_WORD1 + 0x00000000, // CB_COLOR1_CLEAR_WORD2 + 0x00000000, // CB_COLOR1_CLEAR_WORD3 + 0x00000000, // CB_COLOR2_BASE + 0x00000000, // CB_COLOR2_PITCH + 0x00000000, // CB_COLOR2_SLICE + 0x00000000, // CB_COLOR2_VIEW + 0x00000000, // CB_COLOR2_INFO + 0x00000000, // CB_COLOR2_ATTRIB + 0x00000000, // CB_COLOR2_DIM + 0x00000000, // CB_COLOR2_CMASK + 0x00000000, // CB_COLOR2_CMASK_SLICE + 0x00000000, // CB_COLOR2_FMASK + 0x00000000, // CB_COLOR2_FMASK_SLICE + 0x00000000, // CB_COLOR2_CLEAR_WORD0 + 0x00000000, // CB_COLOR2_CLEAR_WORD1 + 0x00000000, // CB_COLOR2_CLEAR_WORD2 + 0x00000000, // CB_COLOR2_CLEAR_WORD3 + 0x00000000, // CB_COLOR3_BASE + 0x00000000, // CB_COLOR3_PITCH + 0x00000000, // CB_COLOR3_SLICE + 0x00000000, // CB_COLOR3_VIEW + 0x00000000, // CB_COLOR3_INFO + 0x00000000, // CB_COLOR3_ATTRIB + 0x00000000, // CB_COLOR3_DIM + 0x00000000, // CB_COLOR3_CMASK + 0x00000000, // CB_COLOR3_CMASK_SLICE + 0x00000000, // CB_COLOR3_FMASK + 0x00000000, // CB_COLOR3_FMASK_SLICE + 0x00000000, // CB_COLOR3_CLEAR_WORD0 + 0x00000000, // CB_COLOR3_CLEAR_WORD1 + 0x00000000, // CB_COLOR3_CLEAR_WORD2 + 0x00000000, // CB_COLOR3_CLEAR_WORD3 + 0x00000000, // CB_COLOR4_BASE + 0x00000000, // CB_COLOR4_PITCH + 0x00000000, // CB_COLOR4_SLICE + 0x00000000, // CB_COLOR4_VIEW + 0x00000000, // CB_COLOR4_INFO + 0x00000000, // CB_COLOR4_ATTRIB + 0x00000000, // CB_COLOR4_DIM + 0x00000000, // CB_COLOR4_CMASK + 0x00000000, // CB_COLOR4_CMASK_SLICE + 0x00000000, // CB_COLOR4_FMASK + 0x00000000, // CB_COLOR4_FMASK_SLICE + 0x00000000, // CB_COLOR4_CLEAR_WORD0 + 0x00000000, // CB_COLOR4_CLEAR_WORD1 + 0x00000000, // CB_COLOR4_CLEAR_WORD2 + 0x00000000, // CB_COLOR4_CLEAR_WORD3 + 0x00000000, // CB_COLOR5_BASE + 0x00000000, // CB_COLOR5_PITCH + 0x00000000, // CB_COLOR5_SLICE + 0x00000000, // CB_COLOR5_VIEW + 0x00000000, // CB_COLOR5_INFO + 0x00000000, // CB_COLOR5_ATTRIB + 0x00000000, // CB_COLOR5_DIM + 0x00000000, // CB_COLOR5_CMASK + 0x00000000, // CB_COLOR5_CMASK_SLICE + 0x00000000, // CB_COLOR5_FMASK + 0x00000000, // CB_COLOR5_FMASK_SLICE + 0x00000000, // CB_COLOR5_CLEAR_WORD0 + 0x00000000, // CB_COLOR5_CLEAR_WORD1 + 0x00000000, // CB_COLOR5_CLEAR_WORD2 + 0x00000000, // CB_COLOR5_CLEAR_WORD3 + 0x00000000, // CB_COLOR6_BASE + 0x00000000, // CB_COLOR6_PITCH + 0x00000000, // CB_COLOR6_SLICE + 0x00000000, // CB_COLOR6_VIEW + 0x00000000, // CB_COLOR6_INFO + 0x00000000, // CB_COLOR6_ATTRIB + 0x00000000, // CB_COLOR6_DIM + 0x00000000, // CB_COLOR6_CMASK + 0x00000000, // CB_COLOR6_CMASK_SLICE + 0x00000000, // CB_COLOR6_FMASK + 0x00000000, // CB_COLOR6_FMASK_SLICE + 0x00000000, // CB_COLOR6_CLEAR_WORD0 + 0x00000000, // CB_COLOR6_CLEAR_WORD1 + 0x00000000, // CB_COLOR6_CLEAR_WORD2 + 0x00000000, // CB_COLOR6_CLEAR_WORD3 + 0x00000000, // CB_COLOR7_BASE + 0x00000000, // CB_COLOR7_PITCH + 0x00000000, // CB_COLOR7_SLICE + 0x00000000, // CB_COLOR7_VIEW + 0x00000000, // CB_COLOR7_INFO + 0x00000000, // CB_COLOR7_ATTRIB + 0x00000000, // CB_COLOR7_DIM + 0x00000000, // CB_COLOR7_CMASK + 0x00000000, // CB_COLOR7_CMASK_SLICE + 0x00000000, // CB_COLOR7_FMASK + 0x00000000, // CB_COLOR7_FMASK_SLICE + 0x00000000, // CB_COLOR7_CLEAR_WORD0 + 0x00000000, // CB_COLOR7_CLEAR_WORD1 + 0x00000000, // CB_COLOR7_CLEAR_WORD2 + 0x00000000, // CB_COLOR7_CLEAR_WORD3 + 0x00000000, // CB_COLOR8_BASE + 0x00000000, // CB_COLOR8_PITCH + 0x00000000, // CB_COLOR8_SLICE + 0x00000000, // CB_COLOR8_VIEW + 0x00000000, // CB_COLOR8_INFO + 0x00000000, // CB_COLOR8_ATTRIB + 0x00000000, // CB_COLOR8_DIM + 0x00000000, // CB_COLOR9_BASE + 0x00000000, // CB_COLOR9_PITCH + 0x00000000, // CB_COLOR9_SLICE + 0x00000000, // CB_COLOR9_VIEW + 0x00000000, // CB_COLOR9_INFO + 0x00000000, // CB_COLOR9_ATTRIB + 0x00000000, // CB_COLOR9_DIM + 0x00000000, // CB_COLOR10_BASE + 0x00000000, // CB_COLOR10_PITCH + 0x00000000, // CB_COLOR10_SLICE + 0x00000000, // CB_COLOR10_VIEW + 0x00000000, // CB_COLOR10_INFO + 0x00000000, // CB_COLOR10_ATTRIB + 0x00000000, // CB_COLOR10_DIM + 0x00000000, // CB_COLOR11_BASE + 0x00000000, // CB_COLOR11_PITCH + 0x00000000, // CB_COLOR11_SLICE + 0x00000000, // CB_COLOR11_VIEW + 0x00000000, // CB_COLOR11_INFO + 0x00000000, // CB_COLOR11_ATTRIB + 0x00000000, // CB_COLOR11_DIM + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SQ_ALU_CONST_CACHE_HS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_HS_15 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_0 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_1 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_2 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_3 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_4 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_5 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_6 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_7 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_8 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_9 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_10 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_11 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_12 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_13 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_14 + 0x00000000, // SQ_ALU_CONST_CACHE_LS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_HS_15 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_0 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_1 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_2 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_3 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_4 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_5 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_6 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_7 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_8 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_9 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_10 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_11 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_12 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_13 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_14 + 0x00000000, // SQ_ALU_CONST_BUFFER_SIZE_LS_15 +}; +static const struct cs_extent_def SECT_CONTEXT_defs[] = +{ + {SECT_CONTEXT_def_1, 0x0000a000, 488 }, + {SECT_CONTEXT_def_2, 0x0000a1f5, 6 }, + {SECT_CONTEXT_def_3, 0x0000a200, 55 }, + {SECT_CONTEXT_def_4, 0x0000a23a, 98 }, + {SECT_CONTEXT_def_5, 0x0000a29e, 5 }, + {SECT_CONTEXT_def_6, 0x0000a2a5, 56 }, + {SECT_CONTEXT_def_7, 0x0000a2de, 290 }, + { 0, 0, 0 } +}; +static const u32 SECT_CLEAR_def_1[] = +{ + 0xffffffff, // SQ_TEX_SAMPLER_CLEAR + 0xffffffff, // SQ_TEX_RESOURCE_CLEAR + 0xffffffff, // SQ_LOOP_BOOL_CLEAR +}; +static const struct cs_extent_def SECT_CLEAR_defs[] = +{ + {SECT_CLEAR_def_1, 0x0000ffc0, 3 }, + { 0, 0, 0 } +}; +static const u32 SECT_CTRLCONST_def_1[] = +{ + 0x00000000, // SQ_VTX_BASE_VTX_LOC + 0x00000000, // SQ_VTX_START_INST_LOC +}; +static const struct cs_extent_def SECT_CTRLCONST_defs[] = +{ + {SECT_CTRLCONST_def_1, 0x0000f3fc, 2 }, + { 0, 0, 0 } +}; +struct cs_section_def evergreen_cs_data[] = { + { SECT_CONTEXT_defs, SECT_CONTEXT }, + { SECT_CLEAR_defs, SECT_CLEAR }, + { SECT_CTRLCONST_defs, SECT_CTRLCONST }, + { 0, SECT_NONE } +}; diff --git a/sys/dev/drm/radeon/clearstate_si.h b/sys/dev/drm/radeon/clearstate_si.h new file mode 100644 index 0000000000..b994cb2a35 --- /dev/null +++ b/sys/dev/drm/radeon/clearstate_si.h @@ -0,0 +1,941 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +static const u32 si_SECT_CONTEXT_def_1[] = +{ + 0x00000000, // DB_RENDER_CONTROL + 0x00000000, // DB_COUNT_CONTROL + 0x00000000, // DB_DEPTH_VIEW + 0x00000000, // DB_RENDER_OVERRIDE + 0x00000000, // DB_RENDER_OVERRIDE2 + 0x00000000, // DB_HTILE_DATA_BASE + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_DEPTH_BOUNDS_MIN + 0x00000000, // DB_DEPTH_BOUNDS_MAX + 0x00000000, // DB_STENCIL_CLEAR + 0x00000000, // DB_DEPTH_CLEAR + 0x00000000, // PA_SC_SCREEN_SCISSOR_TL + 0x40004000, // PA_SC_SCREEN_SCISSOR_BR + 0, // HOLE + 0x00000000, // DB_DEPTH_INFO + 0x00000000, // DB_Z_INFO + 0x00000000, // DB_STENCIL_INFO + 0x00000000, // DB_Z_READ_BASE + 0x00000000, // DB_STENCIL_READ_BASE + 0x00000000, // DB_Z_WRITE_BASE + 0x00000000, // DB_STENCIL_WRITE_BASE + 0x00000000, // DB_DEPTH_SIZE + 0x00000000, // DB_DEPTH_SLICE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // TA_BC_BASE_ADDR + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // COHER_DEST_BASE_2 + 0x00000000, // COHER_DEST_BASE_3 + 0x00000000, // PA_SC_WINDOW_OFFSET + 0x80000000, // PA_SC_WINDOW_SCISSOR_TL + 0x40004000, // PA_SC_WINDOW_SCISSOR_BR + 0x0000ffff, // PA_SC_CLIPRECT_RULE + 0x00000000, // PA_SC_CLIPRECT_0_TL + 0x40004000, // PA_SC_CLIPRECT_0_BR + 0x00000000, // PA_SC_CLIPRECT_1_TL + 0x40004000, // PA_SC_CLIPRECT_1_BR + 0x00000000, // PA_SC_CLIPRECT_2_TL + 0x40004000, // PA_SC_CLIPRECT_2_BR + 0x00000000, // PA_SC_CLIPRECT_3_TL + 0x40004000, // PA_SC_CLIPRECT_3_BR + 0xaa99aaaa, // PA_SC_EDGERULE + 0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET + 0xffffffff, // CB_TARGET_MASK + 0xffffffff, // CB_SHADER_MASK + 0x80000000, // PA_SC_GENERIC_SCISSOR_TL + 0x40004000, // PA_SC_GENERIC_SCISSOR_BR + 0x00000000, // COHER_DEST_BASE_0 + 0x00000000, // COHER_DEST_BASE_1 + 0x80000000, // PA_SC_VPORT_SCISSOR_0_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_0_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_1_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_1_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_2_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_2_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_3_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_3_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_4_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_4_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_5_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_5_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_6_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_6_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_7_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_7_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_8_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_8_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_9_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_9_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_10_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_10_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_11_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_11_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_12_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_12_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_13_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_13_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_14_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_14_BR + 0x80000000, // PA_SC_VPORT_SCISSOR_15_TL + 0x40004000, // PA_SC_VPORT_SCISSOR_15_BR + 0x00000000, // PA_SC_VPORT_ZMIN_0 + 0x3f800000, // PA_SC_VPORT_ZMAX_0 + 0x00000000, // PA_SC_VPORT_ZMIN_1 + 0x3f800000, // PA_SC_VPORT_ZMAX_1 + 0x00000000, // PA_SC_VPORT_ZMIN_2 + 0x3f800000, // PA_SC_VPORT_ZMAX_2 + 0x00000000, // PA_SC_VPORT_ZMIN_3 + 0x3f800000, // PA_SC_VPORT_ZMAX_3 + 0x00000000, // PA_SC_VPORT_ZMIN_4 + 0x3f800000, // PA_SC_VPORT_ZMAX_4 + 0x00000000, // PA_SC_VPORT_ZMIN_5 + 0x3f800000, // PA_SC_VPORT_ZMAX_5 + 0x00000000, // PA_SC_VPORT_ZMIN_6 + 0x3f800000, // PA_SC_VPORT_ZMAX_6 + 0x00000000, // PA_SC_VPORT_ZMIN_7 + 0x3f800000, // PA_SC_VPORT_ZMAX_7 + 0x00000000, // PA_SC_VPORT_ZMIN_8 + 0x3f800000, // PA_SC_VPORT_ZMAX_8 + 0x00000000, // PA_SC_VPORT_ZMIN_9 + 0x3f800000, // PA_SC_VPORT_ZMAX_9 + 0x00000000, // PA_SC_VPORT_ZMIN_10 + 0x3f800000, // PA_SC_VPORT_ZMAX_10 + 0x00000000, // PA_SC_VPORT_ZMIN_11 + 0x3f800000, // PA_SC_VPORT_ZMAX_11 + 0x00000000, // PA_SC_VPORT_ZMIN_12 + 0x3f800000, // PA_SC_VPORT_ZMAX_12 + 0x00000000, // PA_SC_VPORT_ZMIN_13 + 0x3f800000, // PA_SC_VPORT_ZMAX_13 + 0x00000000, // PA_SC_VPORT_ZMIN_14 + 0x3f800000, // PA_SC_VPORT_ZMAX_14 + 0x00000000, // PA_SC_VPORT_ZMIN_15 + 0x3f800000, // PA_SC_VPORT_ZMAX_15 +}; +static const u32 si_SECT_CONTEXT_def_2[] = +{ + 0x00000000, // CP_PERFMON_CNTX_CNTL + 0x00000000, // CP_RINGID + 0x00000000, // CP_VMID + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0xffffffff, // VGT_MAX_VTX_INDX + 0x00000000, // VGT_MIN_VTX_INDX + 0x00000000, // VGT_INDX_OFFSET + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX + 0, // HOLE + 0x00000000, // CB_BLEND_RED + 0x00000000, // CB_BLEND_GREEN + 0x00000000, // CB_BLEND_BLUE + 0x00000000, // CB_BLEND_ALPHA + 0, // HOLE + 0, // HOLE + 0x00000000, // DB_STENCIL_CONTROL + 0x00000000, // DB_STENCILREFMASK + 0x00000000, // DB_STENCILREFMASK_BF + 0, // HOLE + 0x00000000, // PA_CL_VPORT_XSCALE + 0x00000000, // PA_CL_VPORT_XOFFSET + 0x00000000, // PA_CL_VPORT_YSCALE + 0x00000000, // PA_CL_VPORT_YOFFSET + 0x00000000, // PA_CL_VPORT_ZSCALE + 0x00000000, // PA_CL_VPORT_ZOFFSET + 0x00000000, // PA_CL_VPORT_XSCALE_1 + 0x00000000, // PA_CL_VPORT_XOFFSET_1 + 0x00000000, // PA_CL_VPORT_YSCALE_1 + 0x00000000, // PA_CL_VPORT_YOFFSET_1 + 0x00000000, // PA_CL_VPORT_ZSCALE_1 + 0x00000000, // PA_CL_VPORT_ZOFFSET_1 + 0x00000000, // PA_CL_VPORT_XSCALE_2 + 0x00000000, // PA_CL_VPORT_XOFFSET_2 + 0x00000000, // PA_CL_VPORT_YSCALE_2 + 0x00000000, // PA_CL_VPORT_YOFFSET_2 + 0x00000000, // PA_CL_VPORT_ZSCALE_2 + 0x00000000, // PA_CL_VPORT_ZOFFSET_2 + 0x00000000, // PA_CL_VPORT_XSCALE_3 + 0x00000000, // PA_CL_VPORT_XOFFSET_3 + 0x00000000, // PA_CL_VPORT_YSCALE_3 + 0x00000000, // PA_CL_VPORT_YOFFSET_3 + 0x00000000, // PA_CL_VPORT_ZSCALE_3 + 0x00000000, // PA_CL_VPORT_ZOFFSET_3 + 0x00000000, // PA_CL_VPORT_XSCALE_4 + 0x00000000, // PA_CL_VPORT_XOFFSET_4 + 0x00000000, // PA_CL_VPORT_YSCALE_4 + 0x00000000, // PA_CL_VPORT_YOFFSET_4 + 0x00000000, // PA_CL_VPORT_ZSCALE_4 + 0x00000000, // PA_CL_VPORT_ZOFFSET_4 + 0x00000000, // PA_CL_VPORT_XSCALE_5 + 0x00000000, // PA_CL_VPORT_XOFFSET_5 + 0x00000000, // PA_CL_VPORT_YSCALE_5 + 0x00000000, // PA_CL_VPORT_YOFFSET_5 + 0x00000000, // PA_CL_VPORT_ZSCALE_5 + 0x00000000, // PA_CL_VPORT_ZOFFSET_5 + 0x00000000, // PA_CL_VPORT_XSCALE_6 + 0x00000000, // PA_CL_VPORT_XOFFSET_6 + 0x00000000, // PA_CL_VPORT_YSCALE_6 + 0x00000000, // PA_CL_VPORT_YOFFSET_6 + 0x00000000, // PA_CL_VPORT_ZSCALE_6 + 0x00000000, // PA_CL_VPORT_ZOFFSET_6 + 0x00000000, // PA_CL_VPORT_XSCALE_7 + 0x00000000, // PA_CL_VPORT_XOFFSET_7 + 0x00000000, // PA_CL_VPORT_YSCALE_7 + 0x00000000, // PA_CL_VPORT_YOFFSET_7 + 0x00000000, // PA_CL_VPORT_ZSCALE_7 + 0x00000000, // PA_CL_VPORT_ZOFFSET_7 + 0x00000000, // PA_CL_VPORT_XSCALE_8 + 0x00000000, // PA_CL_VPORT_XOFFSET_8 + 0x00000000, // PA_CL_VPORT_YSCALE_8 + 0x00000000, // PA_CL_VPORT_YOFFSET_8 + 0x00000000, // PA_CL_VPORT_ZSCALE_8 + 0x00000000, // PA_CL_VPORT_ZOFFSET_8 + 0x00000000, // PA_CL_VPORT_XSCALE_9 + 0x00000000, // PA_CL_VPORT_XOFFSET_9 + 0x00000000, // PA_CL_VPORT_YSCALE_9 + 0x00000000, // PA_CL_VPORT_YOFFSET_9 + 0x00000000, // PA_CL_VPORT_ZSCALE_9 + 0x00000000, // PA_CL_VPORT_ZOFFSET_9 + 0x00000000, // PA_CL_VPORT_XSCALE_10 + 0x00000000, // PA_CL_VPORT_XOFFSET_10 + 0x00000000, // PA_CL_VPORT_YSCALE_10 + 0x00000000, // PA_CL_VPORT_YOFFSET_10 + 0x00000000, // PA_CL_VPORT_ZSCALE_10 + 0x00000000, // PA_CL_VPORT_ZOFFSET_10 + 0x00000000, // PA_CL_VPORT_XSCALE_11 + 0x00000000, // PA_CL_VPORT_XOFFSET_11 + 0x00000000, // PA_CL_VPORT_YSCALE_11 + 0x00000000, // PA_CL_VPORT_YOFFSET_11 + 0x00000000, // PA_CL_VPORT_ZSCALE_11 + 0x00000000, // PA_CL_VPORT_ZOFFSET_11 + 0x00000000, // PA_CL_VPORT_XSCALE_12 + 0x00000000, // PA_CL_VPORT_XOFFSET_12 + 0x00000000, // PA_CL_VPORT_YSCALE_12 + 0x00000000, // PA_CL_VPORT_YOFFSET_12 + 0x00000000, // PA_CL_VPORT_ZSCALE_12 + 0x00000000, // PA_CL_VPORT_ZOFFSET_12 + 0x00000000, // PA_CL_VPORT_XSCALE_13 + 0x00000000, // PA_CL_VPORT_XOFFSET_13 + 0x00000000, // PA_CL_VPORT_YSCALE_13 + 0x00000000, // PA_CL_VPORT_YOFFSET_13 + 0x00000000, // PA_CL_VPORT_ZSCALE_13 + 0x00000000, // PA_CL_VPORT_ZOFFSET_13 + 0x00000000, // PA_CL_VPORT_XSCALE_14 + 0x00000000, // PA_CL_VPORT_XOFFSET_14 + 0x00000000, // PA_CL_VPORT_YSCALE_14 + 0x00000000, // PA_CL_VPORT_YOFFSET_14 + 0x00000000, // PA_CL_VPORT_ZSCALE_14 + 0x00000000, // PA_CL_VPORT_ZOFFSET_14 + 0x00000000, // PA_CL_VPORT_XSCALE_15 + 0x00000000, // PA_CL_VPORT_XOFFSET_15 + 0x00000000, // PA_CL_VPORT_YSCALE_15 + 0x00000000, // PA_CL_VPORT_YOFFSET_15 + 0x00000000, // PA_CL_VPORT_ZSCALE_15 + 0x00000000, // PA_CL_VPORT_ZOFFSET_15 + 0x00000000, // PA_CL_UCP_0_X + 0x00000000, // PA_CL_UCP_0_Y + 0x00000000, // PA_CL_UCP_0_Z + 0x00000000, // PA_CL_UCP_0_W + 0x00000000, // PA_CL_UCP_1_X + 0x00000000, // PA_CL_UCP_1_Y + 0x00000000, // PA_CL_UCP_1_Z + 0x00000000, // PA_CL_UCP_1_W + 0x00000000, // PA_CL_UCP_2_X + 0x00000000, // PA_CL_UCP_2_Y + 0x00000000, // PA_CL_UCP_2_Z + 0x00000000, // PA_CL_UCP_2_W + 0x00000000, // PA_CL_UCP_3_X + 0x00000000, // PA_CL_UCP_3_Y + 0x00000000, // PA_CL_UCP_3_Z + 0x00000000, // PA_CL_UCP_3_W + 0x00000000, // PA_CL_UCP_4_X + 0x00000000, // PA_CL_UCP_4_Y + 0x00000000, // PA_CL_UCP_4_Z + 0x00000000, // PA_CL_UCP_4_W + 0x00000000, // PA_CL_UCP_5_X + 0x00000000, // PA_CL_UCP_5_Y + 0x00000000, // PA_CL_UCP_5_Z + 0x00000000, // PA_CL_UCP_5_W + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_CNTL_0 + 0x00000000, // SPI_PS_INPUT_CNTL_1 + 0x00000000, // SPI_PS_INPUT_CNTL_2 + 0x00000000, // SPI_PS_INPUT_CNTL_3 + 0x00000000, // SPI_PS_INPUT_CNTL_4 + 0x00000000, // SPI_PS_INPUT_CNTL_5 + 0x00000000, // SPI_PS_INPUT_CNTL_6 + 0x00000000, // SPI_PS_INPUT_CNTL_7 + 0x00000000, // SPI_PS_INPUT_CNTL_8 + 0x00000000, // SPI_PS_INPUT_CNTL_9 + 0x00000000, // SPI_PS_INPUT_CNTL_10 + 0x00000000, // SPI_PS_INPUT_CNTL_11 + 0x00000000, // SPI_PS_INPUT_CNTL_12 + 0x00000000, // SPI_PS_INPUT_CNTL_13 + 0x00000000, // SPI_PS_INPUT_CNTL_14 + 0x00000000, // SPI_PS_INPUT_CNTL_15 + 0x00000000, // SPI_PS_INPUT_CNTL_16 + 0x00000000, // SPI_PS_INPUT_CNTL_17 + 0x00000000, // SPI_PS_INPUT_CNTL_18 + 0x00000000, // SPI_PS_INPUT_CNTL_19 + 0x00000000, // SPI_PS_INPUT_CNTL_20 + 0x00000000, // SPI_PS_INPUT_CNTL_21 + 0x00000000, // SPI_PS_INPUT_CNTL_22 + 0x00000000, // SPI_PS_INPUT_CNTL_23 + 0x00000000, // SPI_PS_INPUT_CNTL_24 + 0x00000000, // SPI_PS_INPUT_CNTL_25 + 0x00000000, // SPI_PS_INPUT_CNTL_26 + 0x00000000, // SPI_PS_INPUT_CNTL_27 + 0x00000000, // SPI_PS_INPUT_CNTL_28 + 0x00000000, // SPI_PS_INPUT_CNTL_29 + 0x00000000, // SPI_PS_INPUT_CNTL_30 + 0x00000000, // SPI_PS_INPUT_CNTL_31 + 0x00000000, // SPI_VS_OUT_CONFIG + 0, // HOLE + 0x00000000, // SPI_PS_INPUT_ENA + 0x00000000, // SPI_PS_INPUT_ADDR + 0x00000000, // SPI_INTERP_CONTROL_0 + 0x00000002, // SPI_PS_IN_CONTROL + 0, // HOLE + 0x00000000, // SPI_BARYC_CNTL + 0, // HOLE + 0x00000000, // SPI_TMPRING_SIZE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // SPI_WAVE_MGMT_1 + 0x00000000, // SPI_WAVE_MGMT_2 + 0x00000000, // SPI_SHADER_POS_FORMAT + 0x00000000, // SPI_SHADER_Z_FORMAT + 0x00000000, // SPI_SHADER_COL_FORMAT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_BLEND0_CONTROL + 0x00000000, // CB_BLEND1_CONTROL + 0x00000000, // CB_BLEND2_CONTROL + 0x00000000, // CB_BLEND3_CONTROL + 0x00000000, // CB_BLEND4_CONTROL + 0x00000000, // CB_BLEND5_CONTROL + 0x00000000, // CB_BLEND6_CONTROL + 0x00000000, // CB_BLEND7_CONTROL +}; +static const u32 si_SECT_CONTEXT_def_3[] = +{ + 0x00000000, // PA_CL_POINT_X_RAD + 0x00000000, // PA_CL_POINT_Y_RAD + 0x00000000, // PA_CL_POINT_SIZE + 0x00000000, // PA_CL_POINT_CULL_RAD + 0x00000000, // VGT_DMA_BASE_HI + 0x00000000, // VGT_DMA_BASE +}; +static const u32 si_SECT_CONTEXT_def_4[] = +{ + 0x00000000, // DB_DEPTH_CONTROL + 0x00000000, // DB_EQAA + 0x00000000, // CB_COLOR_CONTROL + 0x00000000, // DB_SHADER_CONTROL + 0x00090000, // PA_CL_CLIP_CNTL + 0x00000004, // PA_SU_SC_MODE_CNTL + 0x00000000, // PA_CL_VTE_CNTL + 0x00000000, // PA_CL_VS_OUT_CNTL + 0x00000000, // PA_CL_NANINF_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_CNTL + 0x00000000, // PA_SU_LINE_STIPPLE_SCALE + 0x00000000, // PA_SU_PRIM_FILTER_CNTL + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SU_POINT_SIZE + 0x00000000, // PA_SU_POINT_MINMAX + 0x00000000, // PA_SU_LINE_CNTL + 0x00000000, // PA_SC_LINE_STIPPLE + 0x00000000, // VGT_OUTPUT_PATH_CNTL + 0x00000000, // VGT_HOS_CNTL + 0x00000000, // VGT_HOS_MAX_TESS_LEVEL + 0x00000000, // VGT_HOS_MIN_TESS_LEVEL + 0x00000000, // VGT_HOS_REUSE_DEPTH + 0x00000000, // VGT_GROUP_PRIM_TYPE + 0x00000000, // VGT_GROUP_FIRST_DECR + 0x00000000, // VGT_GROUP_DECR + 0x00000000, // VGT_GROUP_VECT_0_CNTL + 0x00000000, // VGT_GROUP_VECT_1_CNTL + 0x00000000, // VGT_GROUP_VECT_0_FMT_CNTL + 0x00000000, // VGT_GROUP_VECT_1_FMT_CNTL + 0x00000000, // VGT_GS_MODE + 0, // HOLE + 0x00000000, // PA_SC_MODE_CNTL_0 + 0x00000000, // PA_SC_MODE_CNTL_1 + 0x00000000, // VGT_ENHANCE + 0x00000100, // VGT_GS_PER_ES + 0x00000080, // VGT_ES_PER_GS + 0x00000002, // VGT_GS_PER_VS + 0x00000000, // VGT_GSVS_RING_OFFSET_1 + 0x00000000, // VGT_GSVS_RING_OFFSET_2 + 0x00000000, // VGT_GSVS_RING_OFFSET_3 + 0x00000000, // VGT_GS_OUT_PRIM_TYPE + 0x00000000, // IA_ENHANCE +}; +static const u32 si_SECT_CONTEXT_def_5[] = +{ + 0x00000000, // VGT_PRIMITIVEID_EN +}; +static const u32 si_SECT_CONTEXT_def_6[] = +{ + 0x00000000, // VGT_PRIMITIVEID_RESET +}; +static const u32 si_SECT_CONTEXT_def_7[] = +{ + 0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_INSTANCE_STEP_RATE_0 + 0x00000000, // VGT_INSTANCE_STEP_RATE_1 + 0x000000ff, // IA_MULTI_VGT_PARAM + 0x00000000, // VGT_ESGS_RING_ITEMSIZE + 0x00000000, // VGT_GSVS_RING_ITEMSIZE + 0x00000000, // VGT_REUSE_OFF + 0x00000000, // VGT_VTX_CNT_EN + 0x00000000, // DB_HTILE_SURFACE + 0x00000000, // DB_SRESULTS_COMPARE_STATE0 + 0x00000000, // DB_SRESULTS_COMPARE_STATE1 + 0x00000000, // DB_PRELOAD_CONTROL + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_0 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_0 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_0 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_1 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_1 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_1 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_2 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_2 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_2 + 0x00000000, // VGT_STRMOUT_BUFFER_SIZE_3 + 0x00000000, // VGT_STRMOUT_VTX_STRIDE_3 + 0, // HOLE + 0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_3 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE + 0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE + 0, // HOLE + 0x00000000, // VGT_GS_MAX_VERT_OUT + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // VGT_SHADER_STAGES_EN + 0x00000000, // VGT_LS_HS_CONFIG + 0x00000000, // VGT_GS_VERT_ITEMSIZE + 0x00000000, // VGT_GS_VERT_ITEMSIZE_1 + 0x00000000, // VGT_GS_VERT_ITEMSIZE_2 + 0x00000000, // VGT_GS_VERT_ITEMSIZE_3 + 0x00000000, // VGT_TF_PARAM + 0x00000000, // DB_ALPHA_TO_MASK + 0, // HOLE + 0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL + 0x00000000, // PA_SU_POLY_OFFSET_CLAMP + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET + 0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE + 0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET + 0x00000000, // VGT_GS_INSTANCE_CNT + 0x00000000, // VGT_STRMOUT_CONFIG + 0x00000000, // VGT_STRMOUT_BUFFER_CONFIG + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x00000000, // PA_SC_CENTROID_PRIORITY_0 + 0x00000000, // PA_SC_CENTROID_PRIORITY_1 + 0x00001000, // PA_SC_LINE_CNTL + 0x00000000, // PA_SC_AA_CONFIG + 0x00000005, // PA_SU_VTX_CNTL + 0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ + 0x3f800000, // PA_CL_GB_VERT_DISC_ADJ + 0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ + 0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2 + 0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3 + 0xffffffff, // PA_SC_AA_MASK_X0Y0_X1Y0 + 0xffffffff, // PA_SC_AA_MASK_X0Y1_X1Y1 + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0, // HOLE + 0x0000000e, // VGT_VERTEX_REUSE_BLOCK_CNTL + 0x00000010, // VGT_OUT_DEALLOC_CNTL + 0x00000000, // CB_COLOR0_BASE + 0x00000000, // CB_COLOR0_PITCH + 0x00000000, // CB_COLOR0_SLICE + 0x00000000, // CB_COLOR0_VIEW + 0x00000000, // CB_COLOR0_INFO + 0x00000000, // CB_COLOR0_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR0_CMASK + 0x00000000, // CB_COLOR0_CMASK_SLICE + 0x00000000, // CB_COLOR0_FMASK + 0x00000000, // CB_COLOR0_FMASK_SLICE + 0x00000000, // CB_COLOR0_CLEAR_WORD0 + 0x00000000, // CB_COLOR0_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR1_BASE + 0x00000000, // CB_COLOR1_PITCH + 0x00000000, // CB_COLOR1_SLICE + 0x00000000, // CB_COLOR1_VIEW + 0x00000000, // CB_COLOR1_INFO + 0x00000000, // CB_COLOR1_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR1_CMASK + 0x00000000, // CB_COLOR1_CMASK_SLICE + 0x00000000, // CB_COLOR1_FMASK + 0x00000000, // CB_COLOR1_FMASK_SLICE + 0x00000000, // CB_COLOR1_CLEAR_WORD0 + 0x00000000, // CB_COLOR1_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR2_BASE + 0x00000000, // CB_COLOR2_PITCH + 0x00000000, // CB_COLOR2_SLICE + 0x00000000, // CB_COLOR2_VIEW + 0x00000000, // CB_COLOR2_INFO + 0x00000000, // CB_COLOR2_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR2_CMASK + 0x00000000, // CB_COLOR2_CMASK_SLICE + 0x00000000, // CB_COLOR2_FMASK + 0x00000000, // CB_COLOR2_FMASK_SLICE + 0x00000000, // CB_COLOR2_CLEAR_WORD0 + 0x00000000, // CB_COLOR2_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR3_BASE + 0x00000000, // CB_COLOR3_PITCH + 0x00000000, // CB_COLOR3_SLICE + 0x00000000, // CB_COLOR3_VIEW + 0x00000000, // CB_COLOR3_INFO + 0x00000000, // CB_COLOR3_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR3_CMASK + 0x00000000, // CB_COLOR3_CMASK_SLICE + 0x00000000, // CB_COLOR3_FMASK + 0x00000000, // CB_COLOR3_FMASK_SLICE + 0x00000000, // CB_COLOR3_CLEAR_WORD0 + 0x00000000, // CB_COLOR3_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR4_BASE + 0x00000000, // CB_COLOR4_PITCH + 0x00000000, // CB_COLOR4_SLICE + 0x00000000, // CB_COLOR4_VIEW + 0x00000000, // CB_COLOR4_INFO + 0x00000000, // CB_COLOR4_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR4_CMASK + 0x00000000, // CB_COLOR4_CMASK_SLICE + 0x00000000, // CB_COLOR4_FMASK + 0x00000000, // CB_COLOR4_FMASK_SLICE + 0x00000000, // CB_COLOR4_CLEAR_WORD0 + 0x00000000, // CB_COLOR4_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR5_BASE + 0x00000000, // CB_COLOR5_PITCH + 0x00000000, // CB_COLOR5_SLICE + 0x00000000, // CB_COLOR5_VIEW + 0x00000000, // CB_COLOR5_INFO + 0x00000000, // CB_COLOR5_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR5_CMASK + 0x00000000, // CB_COLOR5_CMASK_SLICE + 0x00000000, // CB_COLOR5_FMASK + 0x00000000, // CB_COLOR5_FMASK_SLICE + 0x00000000, // CB_COLOR5_CLEAR_WORD0 + 0x00000000, // CB_COLOR5_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR6_BASE + 0x00000000, // CB_COLOR6_PITCH + 0x00000000, // CB_COLOR6_SLICE + 0x00000000, // CB_COLOR6_VIEW + 0x00000000, // CB_COLOR6_INFO + 0x00000000, // CB_COLOR6_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR6_CMASK + 0x00000000, // CB_COLOR6_CMASK_SLICE + 0x00000000, // CB_COLOR6_FMASK + 0x00000000, // CB_COLOR6_FMASK_SLICE + 0x00000000, // CB_COLOR6_CLEAR_WORD0 + 0x00000000, // CB_COLOR6_CLEAR_WORD1 + 0, // HOLE + 0, // HOLE + 0x00000000, // CB_COLOR7_BASE + 0x00000000, // CB_COLOR7_PITCH + 0x00000000, // CB_COLOR7_SLICE + 0x00000000, // CB_COLOR7_VIEW + 0x00000000, // CB_COLOR7_INFO + 0x00000000, // CB_COLOR7_ATTRIB + 0, // HOLE + 0x00000000, // CB_COLOR7_CMASK + 0x00000000, // CB_COLOR7_CMASK_SLICE + 0x00000000, // CB_COLOR7_FMASK + 0x00000000, // CB_COLOR7_FMASK_SLICE + 0x00000000, // CB_COLOR7_CLEAR_WORD0 + 0x00000000, // CB_COLOR7_CLEAR_WORD1 +}; +static const struct cs_extent_def si_SECT_CONTEXT_defs[] = +{ + {si_SECT_CONTEXT_def_1, 0x0000a000, 212 }, + {si_SECT_CONTEXT_def_2, 0x0000a0d8, 272 }, + {si_SECT_CONTEXT_def_3, 0x0000a1f5, 6 }, + {si_SECT_CONTEXT_def_4, 0x0000a200, 157 }, + {si_SECT_CONTEXT_def_5, 0x0000a2a1, 1 }, + {si_SECT_CONTEXT_def_6, 0x0000a2a3, 1 }, + {si_SECT_CONTEXT_def_7, 0x0000a2a5, 233 }, + { 0, 0, 0 } +}; +static const struct cs_section_def si_cs_data[] = { + { si_SECT_CONTEXT_defs, SECT_CONTEXT }, + { 0, SECT_NONE } +}; diff --git a/sys/dev/drm/radeon/cypress_dpm.c b/sys/dev/drm/radeon/cypress_dpm.c new file mode 100644 index 0000000000..41ffdb1cfc --- /dev/null +++ b/sys/dev/drm/radeon/cypress_dpm.c @@ -0,0 +1,2178 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "evergreend.h" +#include "r600_dpm.h" +#include "cypress_dpm.h" +#include "atom.h" +#include "radeon_asic.h" + +#define SMC_RAM_END 0x8000 + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define MC_CG_SEQ_DRAMCONF_S0 0x05 +#define MC_CG_SEQ_DRAMCONF_S1 0x06 +#define MC_CG_SEQ_YCLK_SUSPEND 0x04 +#define MC_CG_SEQ_YCLK_RESUME 0x0a + +struct rv7xx_ps *rv770_get_ps(struct radeon_ps *rps); +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev); +void cypress_dpm_reset_asic(struct radeon_device *rdev); + +static void cypress_enable_bif_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp, bif; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + if (enable) { + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + if (!pi->boot_in_gen2) { + bif = RREG32(CG_BIF_REQ_AND_RSP) & ~CG_CLIENT_REQ_MASK; + bif |= CG_CLIENT_REQ(0xd); + WREG32(CG_BIF_REQ_AND_RSP, bif); + + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp |= LC_HW_VOLTAGE_IF_CONTROL(1); + tmp |= LC_GEN2_EN_STRAP; + + tmp |= LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + DRM_UDELAY(10); + tmp &= ~LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } + } + } else { + if (!pi->boot_in_gen2) { + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp &= ~LC_GEN2_EN_STRAP; + } + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) || + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } +} + +static void cypress_enable_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + cypress_enable_bif_dynamic_pcie_gen2(rdev, enable); + + if (enable) + WREG32_P(GENERAL_PWRMGT, ENABLE_GEN2PCIE, ~ENABLE_GEN2PCIE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~ENABLE_GEN2PCIE); +} + +#if 0 +static int cypress_enter_ulp_state(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (pi->gfx_clock_gating) { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + + RREG32(GB_ADDR_CONFIG); + } + + WREG32_P(SMC_MSG, HOST_SMC_MSG(PPSMC_MSG_SwitchToMinimumPower), + ~HOST_SMC_MSG_MASK); + + DRM_UDELAY(7000); + + return 0; +} +#endif + +static void cypress_gfx_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (enable) { + if (eg_pi->light_sleep) { + WREG32(GRBM_GFX_INDEX, 0xC0000000); + + WREG32_CG(CG_CGLS_TILE_0, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_1, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_2, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_3, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_4, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_5, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_6, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_7, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_8, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_9, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_10, 0xFFFFFFFF); + WREG32_CG(CG_CGLS_TILE_11, 0xFFFFFFFF); + + WREG32_P(SCLK_PWRMGT_CNTL, DYN_LIGHT_SLEEP_EN, ~DYN_LIGHT_SLEEP_EN); + } + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + } else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_ADDR_CONFIG); + + if (eg_pi->light_sleep) { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_LIGHT_SLEEP_EN); + + WREG32(GRBM_GFX_INDEX, 0xC0000000); + + WREG32_CG(CG_CGLS_TILE_0, 0); + WREG32_CG(CG_CGLS_TILE_1, 0); + WREG32_CG(CG_CGLS_TILE_2, 0); + WREG32_CG(CG_CGLS_TILE_3, 0); + WREG32_CG(CG_CGLS_TILE_4, 0); + WREG32_CG(CG_CGLS_TILE_5, 0); + WREG32_CG(CG_CGLS_TILE_6, 0); + WREG32_CG(CG_CGLS_TILE_7, 0); + WREG32_CG(CG_CGLS_TILE_8, 0); + WREG32_CG(CG_CGLS_TILE_9, 0); + WREG32_CG(CG_CGLS_TILE_10, 0); + WREG32_CG(CG_CGLS_TILE_11, 0); + } + } +} + +static void cypress_mg_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (enable) { + u32 cgts_sm_ctrl_reg; + + if (rdev->family == CHIP_CEDAR) + cgts_sm_ctrl_reg = CEDAR_MGCGCGTSSMCTRL_DFLT; + else if (rdev->family == CHIP_REDWOOD) + cgts_sm_ctrl_reg = REDWOOD_MGCGCGTSSMCTRL_DFLT; + else + cgts_sm_ctrl_reg = CYPRESS_MGCGCGTSSMCTRL_DFLT; + + WREG32(GRBM_GFX_INDEX, 0xC0000000); + + WREG32_CG(CG_CGTT_LOCAL_0, CYPRESS_MGCGTTLOCAL0_DFLT); + WREG32_CG(CG_CGTT_LOCAL_1, CYPRESS_MGCGTTLOCAL1_DFLT & 0xFFFFCFFF); + WREG32_CG(CG_CGTT_LOCAL_2, CYPRESS_MGCGTTLOCAL2_DFLT); + WREG32_CG(CG_CGTT_LOCAL_3, CYPRESS_MGCGTTLOCAL3_DFLT); + + if (pi->mgcgtssm) + WREG32(CGTS_SM_CTRL_REG, cgts_sm_ctrl_reg); + + if (eg_pi->mcls) { + WREG32_P(MC_CITF_MISC_RD_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_CITF_MISC_WR_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_CITF_MISC_VM_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_HUB_MISC_HUB_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_HUB_MISC_VM_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_HUB_MISC_SIP_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(MC_XPB_CLK_GAT, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + WREG32_P(VM_L2_CG, MEM_LS_ENABLE, ~MEM_LS_ENABLE); + } + } else { + WREG32(GRBM_GFX_INDEX, 0xC0000000); + + WREG32_CG(CG_CGTT_LOCAL_0, 0xFFFFFFFF); + WREG32_CG(CG_CGTT_LOCAL_1, 0xFFFFFFFF); + WREG32_CG(CG_CGTT_LOCAL_2, 0xFFFFFFFF); + WREG32_CG(CG_CGTT_LOCAL_3, 0xFFFFFFFF); + + if (pi->mgcgtssm) + WREG32(CGTS_SM_CTRL_REG, 0x81f44bc0); + } +} + +void cypress_enable_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + if (pi->sclk_ss) + WREG32_P(GENERAL_PWRMGT, DYN_SPREAD_SPECTRUM_EN, ~DYN_SPREAD_SPECTRUM_EN); + + if (pi->mclk_ss) + WREG32_P(MPLL_CNTL_MODE, SS_SSEN, ~SS_SSEN); + } else { + WREG32_P(CG_SPLL_SPREAD_SPECTRUM, 0, ~SSEN); + WREG32_P(GENERAL_PWRMGT, 0, ~DYN_SPREAD_SPECTRUM_EN); + WREG32_P(MPLL_CNTL_MODE, 0, ~SS_SSEN); + WREG32_P(MPLL_CNTL_MODE, 0, ~SS_DSMODE_EN); + } +} + +void cypress_start_dpm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); +} + +void cypress_enable_sclk_control(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~SCLK_PWRMGT_OFF); + else + WREG32_P(SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, ~SCLK_PWRMGT_OFF); +} + +void cypress_enable_mclk_control(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(MCLK_PWRMGT_CNTL, 0, ~MPLL_PWRMGT_OFF); + else + WREG32_P(MCLK_PWRMGT_CNTL, MPLL_PWRMGT_OFF, ~MPLL_PWRMGT_OFF); +} + +int cypress_notify_smc_display_change(struct radeon_device *rdev, + bool has_display) +{ + PPSMC_Msg msg = has_display ? + (PPSMC_Msg)PPSMC_MSG_HasDisplay : (PPSMC_Msg)PPSMC_MSG_NoDisplay; + + if (rv770_send_msg_to_smc(rdev, msg) != PPSMC_Result_OK) + return -EINVAL; + + return 0; +} + +void cypress_program_response_times(struct radeon_device *rdev) +{ + u32 reference_clock; + u32 mclk_switch_limit; + + reference_clock = radeon_get_xclk(rdev); + mclk_switch_limit = (460 * reference_clock) / 100; + + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_mclk_switch_lim, + mclk_switch_limit); + + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_mvdd_chg_time, 1); + + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_mc_block_delay, 0xAA); + + rv770_program_response_times(rdev); + + if (ASIC_IS_LOMBOK(rdev)) + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_is_asic_lombok, 1); + +} + +static int cypress_pcie_performance_request(struct radeon_device *rdev, + u8 perf_req, bool advertise) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 tmp; + + DRM_UDELAY(10); + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + if ((perf_req == PCIE_PERF_REQ_PECI_GEN1) && (tmp & LC_CURRENT_DATA_RATE)) + return 0; + +#if defined(CONFIG_ACPI) + if ((perf_req == PCIE_PERF_REQ_PECI_GEN1) || + (perf_req == PCIE_PERF_REQ_PECI_GEN2)) { + eg_pi->pcie_performance_request_registered = true; + return radeon_acpi_pcie_performance_request(rdev, perf_req, advertise); + } else if ((perf_req == PCIE_PERF_REQ_REMOVE_REGISTRY) && + eg_pi->pcie_performance_request_registered) { + eg_pi->pcie_performance_request_registered = false; + return radeon_acpi_pcie_performance_request(rdev, perf_req, advertise); + } +#endif + + return 0; +} + +void cypress_advertise_gen2_capability(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + +#if defined(CONFIG_ACPI) + radeon_acpi_pcie_notify_device_ready(rdev); +#endif + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) + pi->pcie_gen2 = true; + else + pi->pcie_gen2 = false; + + if (!pi->pcie_gen2) + cypress_pcie_performance_request(rdev, PCIE_PERF_REQ_PECI_GEN2, true); + +} + +static enum radeon_pcie_gen cypress_get_maximum_link_speed(struct radeon_ps *radeon_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + + if (state->high.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + return 1; + return 0; +} + +void cypress_notify_link_speed_change_after_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + enum radeon_pcie_gen pcie_link_speed_target = + cypress_get_maximum_link_speed(radeon_new_state); + enum radeon_pcie_gen pcie_link_speed_current = + cypress_get_maximum_link_speed(radeon_current_state); + u8 request; + + if (pcie_link_speed_target < pcie_link_speed_current) { + if (pcie_link_speed_target == RADEON_PCIE_GEN1) + request = PCIE_PERF_REQ_PECI_GEN1; + else if (pcie_link_speed_target == RADEON_PCIE_GEN2) + request = PCIE_PERF_REQ_PECI_GEN2; + else + request = PCIE_PERF_REQ_PECI_GEN3; + + cypress_pcie_performance_request(rdev, request, false); + } +} + +void cypress_notify_link_speed_change_before_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + enum radeon_pcie_gen pcie_link_speed_target = + cypress_get_maximum_link_speed(radeon_new_state); + enum radeon_pcie_gen pcie_link_speed_current = + cypress_get_maximum_link_speed(radeon_current_state); + u8 request; + + if (pcie_link_speed_target > pcie_link_speed_current) { + if (pcie_link_speed_target == RADEON_PCIE_GEN1) + request = PCIE_PERF_REQ_PECI_GEN1; + else if (pcie_link_speed_target == RADEON_PCIE_GEN2) + request = PCIE_PERF_REQ_PECI_GEN2; + else + request = PCIE_PERF_REQ_PECI_GEN3; + + cypress_pcie_performance_request(rdev, request, false); + } +} + +static int cypress_populate_voltage_value(struct radeon_device *rdev, + struct atom_voltage_table *table, + u16 value, RV770_SMC_VOLTAGE_VALUE *voltage) +{ + unsigned int i; + + for (i = 0; i < table->count; i++) { + if (value <= table->entries[i].value) { + voltage->index = (u8)i; + voltage->value = cpu_to_be16(table->entries[i].value); + break; + } + } + + if (i == table->count) + return -EINVAL; + + return 0; +} + +u8 cypress_get_strobe_mode_settings(struct radeon_device *rdev, u32 mclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 result = 0; + bool strobe_mode = false; + + if (pi->mem_gddr5) { + if (mclk <= pi->mclk_strobe_mode_threshold) + strobe_mode = true; + result = cypress_get_mclk_frequency_ratio(rdev, mclk, strobe_mode); + + if (strobe_mode) + result |= SMC_STROBE_ENABLE; + } + + return result; +} + +u32 cypress_map_clkf_to_ibias(struct radeon_device *rdev, u32 clkf) +{ + u32 ref_clk = rdev->clock.mpll.reference_freq; + u32 vco = clkf * ref_clk; + + /* 100 Mhz ref clk */ + if (ref_clk == 10000) { + if (vco > 500000) + return 0xC6; + if (vco > 400000) + return 0x9D; + if (vco > 330000) + return 0x6C; + if (vco > 250000) + return 0x2B; + if (vco > 160000) + return 0x5B; + if (vco > 120000) + return 0x0A; + return 0x4B; + } + + /* 27 Mhz ref clk */ + if (vco > 250000) + return 0x8B; + if (vco > 200000) + return 0xCC; + if (vco > 150000) + return 0x9B; + return 0x6B; +} + +static int cypress_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + RV7XX_SMC_MCLK_VALUE *mclk, + bool strobe_mode, bool dll_state_on) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + u32 mpll_ad_func_cntl = + pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = + pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = + pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = + pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 mclk_pwrmgt_cntl = + pi->clk_regs.rv770.mclk_pwrmgt_cntl; + u32 dll_cntl = + pi->clk_regs.rv770.dll_cntl; + u32 mpll_ss1 = pi->clk_regs.rv770.mpll_ss1; + u32 mpll_ss2 = pi->clk_regs.rv770.mpll_ss2; + struct atom_clock_dividers dividers; + u32 ibias; + u32 dll_speed; + int ret; + u32 mc_seq_misc7; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + memory_clock, strobe_mode, ÷rs); + if (ret) + return ret; + + if (!strobe_mode) { + mc_seq_misc7 = RREG32(MC_SEQ_MISC7); + + if(mc_seq_misc7 & 0x8000000) + dividers.post_div = 1; + } + + ibias = cypress_map_clkf_to_ibias(rdev, dividers.whole_fb_div); + + mpll_ad_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_ad_func_cntl |= CLKR(dividers.ref_div); + mpll_ad_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_ad_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_ad_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_ad_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_ad_func_cntl_2 |= VCO_MODE; + else + mpll_ad_func_cntl_2 &= ~VCO_MODE; + + if (pi->mem_gddr5) { + mpll_dq_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_dq_func_cntl |= CLKR(dividers.ref_div); + mpll_dq_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_dq_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_dq_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_dq_func_cntl |= IBIAS(ibias); + + if (strobe_mode) + mpll_dq_func_cntl &= ~PDNB; + else + mpll_dq_func_cntl |= PDNB; + + if (dividers.vco_mode) + mpll_dq_func_cntl_2 |= VCO_MODE; + else + mpll_dq_func_cntl_2 &= ~VCO_MODE; + } + + if (pi->mclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = memory_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, vco_freq)) { + u32 reference_clock = rdev->clock.mpll.reference_freq; + u32 decoded_ref = rv740_get_decoded_reference_divider(dividers.ref_div); + u32 clk_s = reference_clock * 5 / (decoded_ref * ss.rate); + u32 clk_v = ss.percentage * + (0x4000 * dividers.whole_fb_div + 0x800 * dividers.frac_fb_div) / (clk_s * 625); + + mpll_ss1 &= ~CLKV_MASK; + mpll_ss1 |= CLKV(clk_v); + + mpll_ss2 &= ~CLKS_MASK; + mpll_ss2 |= CLKS(clk_s); + } + } + + dll_speed = rv740_get_dll_speed(pi->mem_gddr5, + memory_clock); + + mclk_pwrmgt_cntl &= ~DLL_SPEED_MASK; + mclk_pwrmgt_cntl |= DLL_SPEED(dll_speed); + if (dll_state_on) + mclk_pwrmgt_cntl |= (MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + else + mclk_pwrmgt_cntl &= ~(MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + + mclk->mclk770.mclk_value = cpu_to_be32(memory_clock); + mclk->mclk770.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + mclk->mclk770.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + mclk->mclk770.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + mclk->mclk770.vMPLL_SS = cpu_to_be32(mpll_ss1); + mclk->mclk770.vMPLL_SS2 = cpu_to_be32(mpll_ss2); + + return 0; +} + +u8 cypress_get_mclk_frequency_ratio(struct radeon_device *rdev, + u32 memory_clock, bool strobe_mode) +{ + u8 mc_para_index; + + if (rdev->family >= CHIP_BARTS) { + if (strobe_mode) { + if (memory_clock < 10000) + mc_para_index = 0x00; + else if (memory_clock > 47500) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 10000) / 2500); + } else { + if (memory_clock < 65000) + mc_para_index = 0x00; + else if (memory_clock > 135000) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 60000) / 5000); + } + } else { + if (strobe_mode) { + if (memory_clock < 10000) + mc_para_index = 0x00; + else if (memory_clock > 47500) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 10000) / 2500); + } else { + if (memory_clock < 40000) + mc_para_index = 0x00; + else if (memory_clock > 115000) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 40000) / 5000); + } + } + return mc_para_index; +} + +static int cypress_populate_mvdd_value(struct radeon_device *rdev, + u32 mclk, + RV770_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (!pi->mvdd_control) { + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + return 0; + } + + if (mclk <= pi->mvdd_split_frequency) { + voltage->index = eg_pi->mvdd_low_index; + voltage->value = cpu_to_be16(MVDD_LOW_VALUE); + } else { + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + } + + return 0; +} + +int cypress_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + RV770_SMC_HW_PERFORMANCE_LEVEL *level, + u8 watermark_level) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + int ret; + bool dll_state_on; + + level->gen2PCIE = pi->pcie_gen2 ? + ((pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? 1 : 0) : 0; + level->gen2XSP = (pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? 1 : 0; + level->backbias = (pl->flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? 1 : 0; + level->displayWatermark = watermark_level; + + ret = rv740_populate_sclk_value(rdev, pl->sclk, &level->sclk); + if (ret) + return ret; + + level->mcFlags = 0; + if (pi->mclk_stutter_mode_threshold && + (pl->mclk <= pi->mclk_stutter_mode_threshold) && + !eg_pi->uvd_enabled) { + level->mcFlags |= SMC_MC_STUTTER_EN; + if (eg_pi->sclk_deep_sleep) + level->stateFlags |= PPSMC_STATEFLAG_AUTO_PULSE_SKIP; + else + level->stateFlags &= ~PPSMC_STATEFLAG_AUTO_PULSE_SKIP; + } + + if (pi->mem_gddr5) { + if (pl->mclk > pi->mclk_edc_enable_threshold) + level->mcFlags |= SMC_MC_EDC_RD_FLAG; + + if (pl->mclk > eg_pi->mclk_edc_wr_enable_threshold) + level->mcFlags |= SMC_MC_EDC_WR_FLAG; + + level->strobeMode = cypress_get_strobe_mode_settings(rdev, pl->mclk); + + if (level->strobeMode & SMC_STROBE_ENABLE) { + if (cypress_get_mclk_frequency_ratio(rdev, pl->mclk, true) >= + ((RREG32(MC_SEQ_MISC7) >> 16) & 0xf)) + dll_state_on = ((RREG32(MC_SEQ_MISC5) >> 1) & 0x1) ? true : false; + else + dll_state_on = ((RREG32(MC_SEQ_MISC6) >> 1) & 0x1) ? true : false; + } else + dll_state_on = eg_pi->dll_default_on; + + ret = cypress_populate_mclk_value(rdev, + pl->sclk, + pl->mclk, + &level->mclk, + (level->strobeMode & SMC_STROBE_ENABLE) != 0, + dll_state_on); + } else { + ret = cypress_populate_mclk_value(rdev, + pl->sclk, + pl->mclk, + &level->mclk, + true, + true); + } + if (ret) + return ret; + + ret = cypress_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pl->vddc, + &level->vddc); + if (ret) + return ret; + + if (eg_pi->vddci_control) { + ret = cypress_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + pl->vddci, + &level->vddci); + if (ret) + return ret; + } + + ret = cypress_populate_mvdd_value(rdev, pl->mclk, &level->mvdd); + + return ret; +} + +static int cypress_convert_power_state_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + int ret; + + if (!(radeon_state->caps & ATOM_PPLIB_DISALLOW_ON_DC)) + smc_state->flags |= PPSMC_SWSTATE_FLAG_DC; + + ret = cypress_convert_power_level_to_smc(rdev, + &state->low, + &smc_state->levels[0], + PPSMC_DISPLAY_WATERMARK_LOW); + if (ret) + return ret; + + ret = cypress_convert_power_level_to_smc(rdev, + &state->medium, + &smc_state->levels[1], + PPSMC_DISPLAY_WATERMARK_LOW); + if (ret) + return ret; + + ret = cypress_convert_power_level_to_smc(rdev, + &state->high, + &smc_state->levels[2], + PPSMC_DISPLAY_WATERMARK_HIGH); + if (ret) + return ret; + + smc_state->levels[0].arbValue = MC_CG_ARB_FREQ_F1; + smc_state->levels[1].arbValue = MC_CG_ARB_FREQ_F2; + smc_state->levels[2].arbValue = MC_CG_ARB_FREQ_F3; + + if (eg_pi->dynamic_ac_timing) { + smc_state->levels[0].ACIndex = 2; + smc_state->levels[1].ACIndex = 3; + smc_state->levels[2].ACIndex = 4; + } else { + smc_state->levels[0].ACIndex = 0; + smc_state->levels[1].ACIndex = 0; + smc_state->levels[2].ACIndex = 0; + } + + rv770_populate_smc_sp(rdev, radeon_state, smc_state); + + return rv770_populate_smc_t(rdev, radeon_state, smc_state); +} + +static void cypress_convert_mc_registers(struct evergreen_mc_reg_entry *entry, + SMC_Evergreen_MCRegisterSet *data, + u32 num_entries, u32 valid_flag) +{ + u32 i, j; + + for (i = 0, j = 0; j < num_entries; j++) { + if (valid_flag & (1 << j)) { + data->value[i] = cpu_to_be32(entry->mc_data[j]); + i++; + } + } +} + +static void cypress_convert_mc_reg_table_entry_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SMC_Evergreen_MCRegisterSet *mc_reg_table_data) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 i = 0; + + for (i = 0; i < eg_pi->mc_reg_table.num_entries; i++) { + if (pl->mclk <= + eg_pi->mc_reg_table.mc_reg_table_entry[i].mclk_max) + break; + } + + if ((i == eg_pi->mc_reg_table.num_entries) && (i > 0)) + --i; + + cypress_convert_mc_registers(&eg_pi->mc_reg_table.mc_reg_table_entry[i], + mc_reg_table_data, + eg_pi->mc_reg_table.last, + eg_pi->mc_reg_table.valid_flag); +} + +static void cypress_convert_mc_reg_table_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SMC_Evergreen_MCRegisters *mc_reg_table) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + + cypress_convert_mc_reg_table_entry_to_smc(rdev, + &state->low, + &mc_reg_table->data[2]); + cypress_convert_mc_reg_table_entry_to_smc(rdev, + &state->medium, + &mc_reg_table->data[3]); + cypress_convert_mc_reg_table_entry_to_smc(rdev, + &state->high, + &mc_reg_table->data[4]); +} + +int cypress_upload_sw_state(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u16 address = pi->state_table_start + + offsetof(RV770_SMC_STATETABLE, driverState); + RV770_SMC_SWSTATE state = { 0 }; + int ret; + + ret = cypress_convert_power_state_to_smc(rdev, radeon_new_state, &state); + if (ret) + return ret; + + return rv770_copy_bytes_to_smc(rdev, address, (u8 *)&state, + sizeof(RV770_SMC_SWSTATE), + pi->sram_end); +} + +int cypress_upload_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + SMC_Evergreen_MCRegisters mc_reg_table = { 0 }; + u16 address; + + cypress_convert_mc_reg_table_to_smc(rdev, radeon_new_state, &mc_reg_table); + + address = eg_pi->mc_reg_table_start + + (u16)offsetof(SMC_Evergreen_MCRegisters, data[2]); + + return rv770_copy_bytes_to_smc(rdev, address, + (u8 *)&mc_reg_table.data[2], + sizeof(SMC_Evergreen_MCRegisterSet) * 3, + pi->sram_end); +} + +u32 cypress_calculate_burst_time(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 multiplier = pi->mem_gddr5 ? 1 : 2; + u32 result = (4 * multiplier * engine_clock) / (memory_clock / 2); + u32 burst_time; + + if (result <= 4) + burst_time = 0; + else if (result < 8) + burst_time = result - 4; + else { + burst_time = result / 2 ; + if (burst_time > 18) + burst_time = 18; + } + + return burst_time; +} + +void cypress_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_ps *new_state = rv770_get_ps(radeon_new_state); + u32 mc_arb_burst_time = RREG32(MC_ARB_BURST_TIME); + + mc_arb_burst_time &= ~(STATE1_MASK | STATE2_MASK | STATE3_MASK); + + mc_arb_burst_time |= STATE1(cypress_calculate_burst_time(rdev, + new_state->low.sclk, + new_state->low.mclk)); + mc_arb_burst_time |= STATE2(cypress_calculate_burst_time(rdev, + new_state->medium.sclk, + new_state->medium.mclk)); + mc_arb_burst_time |= STATE3(cypress_calculate_burst_time(rdev, + new_state->high.sclk, + new_state->high.mclk)); + + rv730_program_memory_timing_parameters(rdev, radeon_new_state); + + WREG32(MC_ARB_BURST_TIME, mc_arb_burst_time); +} + +static void cypress_populate_mc_reg_addresses(struct radeon_device *rdev, + SMC_Evergreen_MCRegisters *mc_reg_table) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 i, j; + + for (i = 0, j = 0; j < eg_pi->mc_reg_table.last; j++) { + if (eg_pi->mc_reg_table.valid_flag & (1 << j)) { + mc_reg_table->address[i].s0 = + cpu_to_be16(eg_pi->mc_reg_table.mc_reg_address[j].s0); + mc_reg_table->address[i].s1 = + cpu_to_be16(eg_pi->mc_reg_table.mc_reg_address[j].s1); + i++; + } + } + + mc_reg_table->last = (u8)i; +} + +static void cypress_set_mc_reg_address_table(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 i = 0; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_RAS_TIMING_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_RAS_TIMING >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_CAS_TIMING_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_CAS_TIMING >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_MISC_TIMING_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_MISC_TIMING >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_MISC_TIMING2_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_MISC_TIMING2 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_RD_CTL_D0_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_RD_CTL_D0 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_RD_CTL_D1_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_RD_CTL_D1 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_WR_CTL_D0_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_WR_CTL_D0 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_WR_CTL_D1_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_WR_CTL_D1 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_PMG_CMD_EMRS >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_PMG_CMD_MRS_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_PMG_CMD_MRS >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_PMG_CMD_MRS1 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_MISC1 >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_MISC1 >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_RESERVE_M >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_RESERVE_M >> 2; + i++; + + eg_pi->mc_reg_table.mc_reg_address[i].s0 = MC_SEQ_MISC3 >> 2; + eg_pi->mc_reg_table.mc_reg_address[i].s1 = MC_SEQ_MISC3 >> 2; + i++; + + eg_pi->mc_reg_table.last = (u8)i; +} + +static void cypress_retrieve_ac_timing_for_one_entry(struct radeon_device *rdev, + struct evergreen_mc_reg_entry *entry) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 i; + + for (i = 0; i < eg_pi->mc_reg_table.last; i++) + entry->mc_data[i] = + RREG32(eg_pi->mc_reg_table.mc_reg_address[i].s1 << 2); + +} + +static void cypress_retrieve_ac_timing_for_all_ranges(struct radeon_device *rdev, + struct atom_memory_clock_range_table *range_table) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 i, j; + + for (i = 0; i < range_table->num_entries; i++) { + eg_pi->mc_reg_table.mc_reg_table_entry[i].mclk_max = + range_table->mclk[i]; + radeon_atom_set_ac_timing(rdev, range_table->mclk[i]); + cypress_retrieve_ac_timing_for_one_entry(rdev, + &eg_pi->mc_reg_table.mc_reg_table_entry[i]); + } + + eg_pi->mc_reg_table.num_entries = range_table->num_entries; + eg_pi->mc_reg_table.valid_flag = 0; + + for (i = 0; i < eg_pi->mc_reg_table.last; i++) { + for (j = 1; j < range_table->num_entries; j++) { + if (eg_pi->mc_reg_table.mc_reg_table_entry[j-1].mc_data[i] != + eg_pi->mc_reg_table.mc_reg_table_entry[j].mc_data[i]) { + eg_pi->mc_reg_table.valid_flag |= (1 << i); + break; + } + } + } +} + +static int cypress_initialize_mc_reg_table(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 module_index = rv770_get_memory_module_index(rdev); + struct atom_memory_clock_range_table range_table = { 0 }; + int ret; + + ret = radeon_atom_get_mclk_range_table(rdev, + pi->mem_gddr5, + module_index, &range_table); + if (ret) + return ret; + + cypress_retrieve_ac_timing_for_all_ranges(rdev, &range_table); + + return 0; +} + +static void cypress_wait_for_mc_sequencer(struct radeon_device *rdev, u8 value) +{ + u32 i, j; + u32 channels = 2; + + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK)) + channels = 4; + else if (rdev->family == CHIP_CEDAR) + channels = 1; + + for (i = 0; i < channels; i++) { + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK)) { + WREG32_P(MC_CONFIG_MCD, MC_RD_ENABLE_MCD(i), ~MC_RD_ENABLE_MCD_MASK); + WREG32_P(MC_CG_CONFIG_MCD, MC_RD_ENABLE_MCD(i), ~MC_RD_ENABLE_MCD_MASK); + } else { + WREG32_P(MC_CONFIG, MC_RD_ENABLE(i), ~MC_RD_ENABLE_MASK); + WREG32_P(MC_CG_CONFIG, MC_RD_ENABLE(i), ~MC_RD_ENABLE_MASK); + } + for (j = 0; j < rdev->usec_timeout; j++) { + if (((RREG32(MC_SEQ_CG) & CG_SEQ_RESP_MASK) >> CG_SEQ_RESP_SHIFT) == value) + break; + DRM_UDELAY(1); + } + } +} + +static void cypress_force_mc_use_s1(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_ps *boot_state = rv770_get_ps(radeon_boot_state); + u32 strobe_mode; + u32 mc_seq_cg; + int i; + + if (RREG32(MC_SEQ_STATUS_M) & PMG_PWRSTATE) + return; + + radeon_atom_set_ac_timing(rdev, boot_state->low.mclk); + radeon_mc_wait_for_idle(rdev); + + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK)) { + WREG32(MC_CONFIG_MCD, 0xf); + WREG32(MC_CG_CONFIG_MCD, 0xf); + } else { + WREG32(MC_CONFIG, 0xf); + WREG32(MC_CG_CONFIG, 0xf); + } + + for (i = 0; i < rdev->num_crtc; i++) + radeon_wait_for_vblank(rdev, i); + + WREG32(MC_SEQ_CG, MC_CG_SEQ_YCLK_SUSPEND); + cypress_wait_for_mc_sequencer(rdev, MC_CG_SEQ_YCLK_SUSPEND); + + strobe_mode = cypress_get_strobe_mode_settings(rdev, + boot_state->low.mclk); + + mc_seq_cg = CG_SEQ_REQ(MC_CG_SEQ_DRAMCONF_S1); + mc_seq_cg |= SEQ_CG_RESP(strobe_mode); + WREG32(MC_SEQ_CG, mc_seq_cg); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(MC_SEQ_STATUS_M) & PMG_PWRSTATE) + break; + DRM_UDELAY(1); + } + + mc_seq_cg &= ~CG_SEQ_REQ_MASK; + mc_seq_cg |= CG_SEQ_REQ(MC_CG_SEQ_YCLK_RESUME); + WREG32(MC_SEQ_CG, mc_seq_cg); + + cypress_wait_for_mc_sequencer(rdev, MC_CG_SEQ_YCLK_RESUME); +} + +static void cypress_copy_ac_timing_from_s1_to_s0(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 value; + u32 i; + + for (i = 0; i < eg_pi->mc_reg_table.last; i++) { + value = RREG32(eg_pi->mc_reg_table.mc_reg_address[i].s1 << 2); + WREG32(eg_pi->mc_reg_table.mc_reg_address[i].s0 << 2, value); + } +} + +static void cypress_force_mc_use_s0(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_ps *boot_state = rv770_get_ps(radeon_boot_state); + u32 strobe_mode; + u32 mc_seq_cg; + int i; + + cypress_copy_ac_timing_from_s1_to_s0(rdev); + radeon_mc_wait_for_idle(rdev); + + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK)) { + WREG32(MC_CONFIG_MCD, 0xf); + WREG32(MC_CG_CONFIG_MCD, 0xf); + } else { + WREG32(MC_CONFIG, 0xf); + WREG32(MC_CG_CONFIG, 0xf); + } + + for (i = 0; i < rdev->num_crtc; i++) + radeon_wait_for_vblank(rdev, i); + + WREG32(MC_SEQ_CG, MC_CG_SEQ_YCLK_SUSPEND); + cypress_wait_for_mc_sequencer(rdev, MC_CG_SEQ_YCLK_SUSPEND); + + strobe_mode = cypress_get_strobe_mode_settings(rdev, + boot_state->low.mclk); + + mc_seq_cg = CG_SEQ_REQ(MC_CG_SEQ_DRAMCONF_S0); + mc_seq_cg |= SEQ_CG_RESP(strobe_mode); + WREG32(MC_SEQ_CG, mc_seq_cg); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (!(RREG32(MC_SEQ_STATUS_M) & PMG_PWRSTATE)) + break; + DRM_UDELAY(1); + } + + mc_seq_cg &= ~CG_SEQ_REQ_MASK; + mc_seq_cg |= CG_SEQ_REQ(MC_CG_SEQ_YCLK_RESUME); + WREG32(MC_SEQ_CG, mc_seq_cg); + + cypress_wait_for_mc_sequencer(rdev, MC_CG_SEQ_YCLK_RESUME); +} + +static int cypress_populate_initial_mvdd_value(struct radeon_device *rdev, + RV770_SMC_VOLTAGE_VALUE *voltage) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + + return 0; +} + +int cypress_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_initial_state, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_ps *initial_state = rv770_get_ps(radeon_initial_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 a_t; + + table->initialState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mpll_ad_func_cntl); + table->initialState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_ad_func_cntl_2); + table->initialState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mpll_dq_func_cntl); + table->initialState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_dq_func_cntl_2); + table->initialState.levels[0].mclk.mclk770.vMCLK_PWRMGT_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mclk_pwrmgt_cntl); + table->initialState.levels[0].mclk.mclk770.vDLL_CNTL = + cpu_to_be32(pi->clk_regs.rv770.dll_cntl); + + table->initialState.levels[0].mclk.mclk770.vMPLL_SS = + cpu_to_be32(pi->clk_regs.rv770.mpll_ss1); + table->initialState.levels[0].mclk.mclk770.vMPLL_SS2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_ss2); + + table->initialState.levels[0].mclk.mclk770.mclk_value = + cpu_to_be32(initial_state->low.mclk); + + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl_2); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl_3); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_spread_spectrum); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM_2 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_spread_spectrum_2); + + table->initialState.levels[0].sclk.sclk_value = + cpu_to_be32(initial_state->low.sclk); + + table->initialState.levels[0].arbValue = MC_CG_ARB_FREQ_F0; + + table->initialState.levels[0].ACIndex = 0; + + cypress_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + initial_state->low.vddc, + &table->initialState.levels[0].vddc); + + if (eg_pi->vddci_control) + cypress_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + initial_state->low.vddci, + &table->initialState.levels[0].vddci); + + cypress_populate_initial_mvdd_value(rdev, + &table->initialState.levels[0].mvdd); + + a_t = CG_R(0xffff) | CG_L(0); + table->initialState.levels[0].aT = cpu_to_be32(a_t); + + table->initialState.levels[0].bSP = cpu_to_be32(pi->dsp); + + + if (pi->boot_in_gen2) + table->initialState.levels[0].gen2PCIE = 1; + else + table->initialState.levels[0].gen2PCIE = 0; + if (initial_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + table->initialState.levels[0].gen2XSP = 1; + else + table->initialState.levels[0].gen2XSP = 0; + + if (pi->mem_gddr5) { + table->initialState.levels[0].strobeMode = + cypress_get_strobe_mode_settings(rdev, + initial_state->low.mclk); + + if (initial_state->low.mclk > pi->mclk_edc_enable_threshold) + table->initialState.levels[0].mcFlags = SMC_MC_EDC_RD_FLAG | SMC_MC_EDC_WR_FLAG; + else + table->initialState.levels[0].mcFlags = 0; + } + + table->initialState.levels[1] = table->initialState.levels[0]; + table->initialState.levels[2] = table->initialState.levels[0]; + + table->initialState.flags |= PPSMC_SWSTATE_FLAG_DC; + + return 0; +} + +int cypress_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 mpll_ad_func_cntl = + pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = + pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = + pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = + pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 spll_func_cntl = + pi->clk_regs.rv770.cg_spll_func_cntl; + u32 spll_func_cntl_2 = + pi->clk_regs.rv770.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = + pi->clk_regs.rv770.cg_spll_func_cntl_3; + u32 mclk_pwrmgt_cntl = + pi->clk_regs.rv770.mclk_pwrmgt_cntl; + u32 dll_cntl = + pi->clk_regs.rv770.dll_cntl; + + table->ACPIState = table->initialState; + + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + cypress_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pi->acpi_vddc, + &table->ACPIState.levels[0].vddc); + if (pi->pcie_gen2) { + if (pi->acpi_pcie_gen2) + table->ACPIState.levels[0].gen2PCIE = 1; + else + table->ACPIState.levels[0].gen2PCIE = 0; + } else + table->ACPIState.levels[0].gen2PCIE = 0; + if (pi->acpi_pcie_gen2) + table->ACPIState.levels[0].gen2XSP = 1; + else + table->ACPIState.levels[0].gen2XSP = 0; + } else { + cypress_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pi->min_vddc_in_table, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = 0; + } + + if (eg_pi->acpi_vddci) { + if (eg_pi->vddci_control) { + cypress_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + eg_pi->acpi_vddci, + &table->ACPIState.levels[0].vddci); + } + } + + mpll_ad_func_cntl &= ~PDNB; + + mpll_ad_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN; + + if (pi->mem_gddr5) + mpll_dq_func_cntl &= ~PDNB; + mpll_dq_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN | BYPASS; + + mclk_pwrmgt_cntl |= (MRDCKA0_RESET | + MRDCKA1_RESET | + MRDCKB0_RESET | + MRDCKB1_RESET | + MRDCKC0_RESET | + MRDCKC1_RESET | + MRDCKD0_RESET | + MRDCKD1_RESET); + + mclk_pwrmgt_cntl &= ~(MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + + dll_cntl |= (MRDCKA0_BYPASS | + MRDCKA1_BYPASS | + MRDCKB0_BYPASS | + MRDCKB1_BYPASS | + MRDCKC0_BYPASS | + MRDCKC1_BYPASS | + MRDCKD0_BYPASS | + MRDCKD1_BYPASS); + + /* evergreen only */ + if (rdev->family <= CHIP_HEMLOCK) + spll_func_cntl |= SPLL_RESET | SPLL_SLEEP | SPLL_BYPASS_EN; + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(mpll_ad_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL_2 = + cpu_to_be32(mpll_ad_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(mpll_dq_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL_2 = + cpu_to_be32(mpll_dq_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk770.vMCLK_PWRMGT_CNTL = + cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + + table->ACPIState.levels[0].mclk.mclk770.mclk_value = 0; + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(spll_func_cntl_3); + + table->ACPIState.levels[0].sclk.sclk_value = 0; + + cypress_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + if (eg_pi->dynamic_ac_timing) + table->ACPIState.levels[0].ACIndex = 1; + + table->ACPIState.levels[1] = table->ACPIState.levels[0]; + table->ACPIState.levels[2] = table->ACPIState.levels[0]; + + return 0; +} + +static void cypress_trim_voltage_table_to_fit_state_table(struct radeon_device *rdev, + struct atom_voltage_table *voltage_table) +{ + unsigned int i, diff; + + if (voltage_table->count <= MAX_NO_VREG_STEPS) + return; + + diff = voltage_table->count - MAX_NO_VREG_STEPS; + + for (i= 0; i < MAX_NO_VREG_STEPS; i++) + voltage_table->entries[i] = voltage_table->entries[i + diff]; + + voltage_table->count = MAX_NO_VREG_STEPS; +} + +int cypress_construct_voltage_tables(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + int ret; + + ret = radeon_atom_get_voltage_table(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0, + &eg_pi->vddc_voltage_table); + if (ret) + return ret; + + if (eg_pi->vddc_voltage_table.count > MAX_NO_VREG_STEPS) + cypress_trim_voltage_table_to_fit_state_table(rdev, + &eg_pi->vddc_voltage_table); + + if (eg_pi->vddci_control) { + ret = radeon_atom_get_voltage_table(rdev, SET_VOLTAGE_TYPE_ASIC_VDDCI, 0, + &eg_pi->vddci_voltage_table); + if (ret) + return ret; + + if (eg_pi->vddci_voltage_table.count > MAX_NO_VREG_STEPS) + cypress_trim_voltage_table_to_fit_state_table(rdev, + &eg_pi->vddci_voltage_table); + } + + return 0; +} + +static void cypress_populate_smc_voltage_table(struct radeon_device *rdev, + struct atom_voltage_table *voltage_table, + RV770_SMC_STATETABLE *table) +{ + unsigned int i; + + for (i = 0; i < voltage_table->count; i++) { + table->highSMIO[i] = 0; + table->lowSMIO[i] |= cpu_to_be32(voltage_table->entries[i].smio_low); + } +} + +int cypress_populate_smc_voltage_tables(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + unsigned char i; + + if (eg_pi->vddc_voltage_table.count) { + cypress_populate_smc_voltage_table(rdev, + &eg_pi->vddc_voltage_table, + table); + + table->voltageMaskTable.highMask[RV770_SMC_VOLTAGEMASK_VDDC] = 0; + table->voltageMaskTable.lowMask[RV770_SMC_VOLTAGEMASK_VDDC] = + cpu_to_be32(eg_pi->vddc_voltage_table.mask_low); + + for (i = 0; i < eg_pi->vddc_voltage_table.count; i++) { + if (pi->max_vddc_in_table <= + eg_pi->vddc_voltage_table.entries[i].value) { + table->maxVDDCIndexInPPTable = i; + break; + } + } + } + + if (eg_pi->vddci_voltage_table.count) { + cypress_populate_smc_voltage_table(rdev, + &eg_pi->vddci_voltage_table, + table); + + table->voltageMaskTable.highMask[RV770_SMC_VOLTAGEMASK_VDDCI] = 0; + table->voltageMaskTable.lowMask[RV770_SMC_VOLTAGEMASK_VDDCI] = + cpu_to_be32(eg_pi->vddc_voltage_table.mask_low); + } + + return 0; +} + +static u32 cypress_get_mclk_split_point(struct atom_memory_info *memory_info) +{ + if ((memory_info->mem_type == MEM_TYPE_GDDR3) || + (memory_info->mem_type == MEM_TYPE_DDR3)) + return 30000; + + return 0; +} + +int cypress_get_mvdd_configuration(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u8 module_index; + struct atom_memory_info memory_info; + u32 tmp = RREG32(GENERAL_PWRMGT); + + if (!(tmp & BACKBIAS_PAD_EN)) { + eg_pi->mvdd_high_index = 0; + eg_pi->mvdd_low_index = 1; + pi->mvdd_control = false; + return 0; + } + + if (tmp & BACKBIAS_VALUE) + eg_pi->mvdd_high_index = 1; + else + eg_pi->mvdd_high_index = 0; + + eg_pi->mvdd_low_index = + (eg_pi->mvdd_high_index == 0) ? 1 : 0; + + module_index = rv770_get_memory_module_index(rdev); + + if (radeon_atom_get_memory_info(rdev, module_index, &memory_info)) { + pi->mvdd_control = false; + return 0; + } + + pi->mvdd_split_frequency = + cypress_get_mclk_split_point(&memory_info); + + if (pi->mvdd_split_frequency == 0) { + pi->mvdd_control = false; + return 0; + } + + return 0; +} + +static int cypress_init_smc_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + RV770_SMC_STATETABLE *table = &pi->smc_statetable; + int ret; + + memset(table, 0, sizeof(RV770_SMC_STATETABLE)); + + cypress_populate_smc_voltage_tables(rdev, table); + + switch (rdev->pm.int_thermal_type) { + case THERMAL_TYPE_EVERGREEN: + case THERMAL_TYPE_EMC2103_WITH_INTERNAL: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_INTERNAL; + break; + case THERMAL_TYPE_NONE: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_NONE; + break; + default: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL; + break; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REGULATOR_HOT) + table->systemFlags |= PPSMC_SYSTEMFLAG_REGULATOR_HOT; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (pi->mem_gddr5) + table->systemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + ret = cypress_populate_smc_initial_state(rdev, radeon_boot_state, table); + if (ret) + return ret; + + ret = cypress_populate_smc_acpi_state(rdev, table); + if (ret) + return ret; + + table->driverState = table->initialState; + + return rv770_copy_bytes_to_smc(rdev, + pi->state_table_start, + (u8 *)table, sizeof(RV770_SMC_STATETABLE), + pi->sram_end); +} + +int cypress_populate_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_ps *boot_state = rv770_get_ps(radeon_boot_state); + SMC_Evergreen_MCRegisters mc_reg_table = { 0 }; + + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_seq_index, 1); + + cypress_populate_mc_reg_addresses(rdev, &mc_reg_table); + + cypress_convert_mc_reg_table_entry_to_smc(rdev, + &boot_state->low, + &mc_reg_table.data[0]); + + cypress_convert_mc_registers(&eg_pi->mc_reg_table.mc_reg_table_entry[0], + &mc_reg_table.data[1], eg_pi->mc_reg_table.last, + eg_pi->mc_reg_table.valid_flag); + + cypress_convert_mc_reg_table_to_smc(rdev, radeon_boot_state, &mc_reg_table); + + return rv770_copy_bytes_to_smc(rdev, eg_pi->mc_reg_table_start, + (u8 *)&mc_reg_table, sizeof(SMC_Evergreen_MCRegisters), + pi->sram_end); +} + +int cypress_get_table_locations(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 tmp; + int ret; + + ret = rv770_read_smc_sram_dword(rdev, + EVERGREEN_SMC_FIRMWARE_HEADER_LOCATION + + EVERGREEN_SMC_FIRMWARE_HEADER_stateTable, + &tmp, pi->sram_end); + if (ret) + return ret; + + pi->state_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + EVERGREEN_SMC_FIRMWARE_HEADER_LOCATION + + EVERGREEN_SMC_FIRMWARE_HEADER_softRegisters, + &tmp, pi->sram_end); + if (ret) + return ret; + + pi->soft_regs_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + EVERGREEN_SMC_FIRMWARE_HEADER_LOCATION + + EVERGREEN_SMC_FIRMWARE_HEADER_mcRegisterTable, + &tmp, pi->sram_end); + if (ret) + return ret; + + eg_pi->mc_reg_table_start = (u16)tmp; + + return 0; +} + +void cypress_enable_display_gap(struct radeon_device *rdev) +{ + u32 tmp = RREG32(CG_DISPLAY_GAP_CNTL); + + tmp &= ~(DISP1_GAP_MASK | DISP2_GAP_MASK); + tmp |= (DISP1_GAP(R600_PM_DISPLAY_GAP_IGNORE) | + DISP2_GAP(R600_PM_DISPLAY_GAP_IGNORE)); + + tmp &= ~(DISP1_GAP_MCHG_MASK | DISP2_GAP_MCHG_MASK); + tmp |= (DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK) | + DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE)); + WREG32(CG_DISPLAY_GAP_CNTL, tmp); +} + +static void cypress_program_display_gap(struct radeon_device *rdev) +{ + u32 tmp, pipe; + int i; + + tmp = RREG32(CG_DISPLAY_GAP_CNTL) & ~(DISP1_GAP_MASK | DISP2_GAP_MASK); + if (rdev->pm.dpm.new_active_crtc_count > 0) + tmp |= DISP1_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM); + else + tmp |= DISP1_GAP(R600_PM_DISPLAY_GAP_IGNORE); + + if (rdev->pm.dpm.new_active_crtc_count > 1) + tmp |= DISP2_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM); + else + tmp |= DISP2_GAP(R600_PM_DISPLAY_GAP_IGNORE); + + WREG32(CG_DISPLAY_GAP_CNTL, tmp); + + tmp = RREG32(DCCG_DISP_SLOW_SELECT_REG); + pipe = (tmp & DCCG_DISP1_SLOW_SELECT_MASK) >> DCCG_DISP1_SLOW_SELECT_SHIFT; + + if ((rdev->pm.dpm.new_active_crtc_count > 0) && + (!(rdev->pm.dpm.new_active_crtcs & (1 << pipe)))) { + /* find the first active crtc */ + for (i = 0; i < rdev->num_crtc; i++) { + if (rdev->pm.dpm.new_active_crtcs & (1 << i)) + break; + } + if (i == rdev->num_crtc) + pipe = 0; + else + pipe = i; + + tmp &= ~DCCG_DISP1_SLOW_SELECT_MASK; + tmp |= DCCG_DISP1_SLOW_SELECT(pipe); + WREG32(DCCG_DISP_SLOW_SELECT_REG, tmp); + } + + cypress_notify_smc_display_change(rdev, rdev->pm.dpm.new_active_crtc_count > 0); +} + +void cypress_dpm_setup_asic(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + rv740_read_clock_registers(rdev); + rv770_read_voltage_smio_registers(rdev); + rv770_get_max_vddc(rdev); + rv770_get_memory_type(rdev); + + if (eg_pi->pcie_performance_request) + eg_pi->pcie_performance_request_registered = false; + + if (eg_pi->pcie_performance_request) + cypress_advertise_gen2_capability(rdev); + + rv770_get_pcie_gen2_status(rdev); + + rv770_enable_acpi_pm(rdev); +} + +int cypress_dpm_enable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (pi->gfx_clock_gating) + rv770_restore_cgcg(rdev); + + if (rv770_dpm_enabled(rdev)) + return -EINVAL; + + if (pi->voltage_control) { + rv770_enable_voltage_control(rdev, true); + ret = cypress_construct_voltage_tables(rdev); + if (ret) { + DRM_ERROR("cypress_construct_voltage_tables failed\n"); + return ret; + } + } + + if (pi->mvdd_control) { + ret = cypress_get_mvdd_configuration(rdev); + if (ret) { + DRM_ERROR("cypress_get_mvdd_configuration failed\n"); + return ret; + } + } + + if (eg_pi->dynamic_ac_timing) { + cypress_set_mc_reg_address_table(rdev); + cypress_force_mc_use_s0(rdev, boot_ps); + ret = cypress_initialize_mc_reg_table(rdev); + if (ret) + eg_pi->dynamic_ac_timing = false; + cypress_force_mc_use_s1(rdev, boot_ps); + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv770_enable_backbias(rdev, true); + + if (pi->dynamic_ss) + cypress_enable_spread_spectrum(rdev, true); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, true); + + rv770_setup_bsp(rdev); + rv770_program_git(rdev); + rv770_program_tp(rdev); + rv770_program_tpp(rdev); + rv770_program_sstp(rdev); + rv770_program_engine_speed_parameters(rdev); + cypress_enable_display_gap(rdev); + rv770_program_vc(rdev); + + if (pi->dynamic_pcie_gen2) + cypress_enable_dynamic_pcie_gen2(rdev, true); + + ret = rv770_upload_firmware(rdev); + if (ret) { + DRM_ERROR("rv770_upload_firmware failed\n"); + return ret; + } + + ret = cypress_get_table_locations(rdev); + if (ret) { + DRM_ERROR("cypress_get_table_locations failed\n"); + return ret; + } + ret = cypress_init_smc_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("cypress_init_smc_table failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = cypress_populate_mc_reg_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("cypress_populate_mc_reg_table failed\n"); + return ret; + } + } + + cypress_program_response_times(rdev); + + r7xx_start_smc(rdev); + + ret = cypress_notify_smc_display_change(rdev, false); + if (ret) { + DRM_ERROR("cypress_notify_smc_display_change failed\n"); + return ret; + } + cypress_enable_sclk_control(rdev, true); + + if (eg_pi->memory_transition) + cypress_enable_mclk_control(rdev, true); + + cypress_start_dpm(rdev); + + if (pi->gfx_clock_gating) + cypress_gfx_clock_gating_enable(rdev, true); + + if (pi->mg_clock_gating) + cypress_mg_clock_gating_enable(rdev, true); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + PPSMC_Result result; + + ret = rv770_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + + if (result != PPSMC_Result_OK) + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + } + + rv770_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + return 0; +} + +void cypress_dpm_disable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + + if (!rv770_dpm_enabled(rdev)) + return; + + rv770_clear_vc(rdev); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, false); + + if (pi->dynamic_pcie_gen2) + cypress_enable_dynamic_pcie_gen2(rdev, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + if (pi->gfx_clock_gating) + cypress_gfx_clock_gating_enable(rdev, false); + + if (pi->mg_clock_gating) + cypress_mg_clock_gating_enable(rdev, false); + + rv770_stop_dpm(rdev); + r7xx_stop_smc(rdev); + + cypress_enable_spread_spectrum(rdev, false); + + if (eg_pi->dynamic_ac_timing) + cypress_force_mc_use_s1(rdev, boot_ps); + + rv770_reset_smio_status(rdev); +} + +int cypress_dpm_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + struct radeon_ps *old_ps = rdev->pm.dpm.current_ps; + int ret; + + ret = rv770_restrict_performance_levels_before_switch(rdev); + if (ret) { + DRM_ERROR("rv770_restrict_performance_levels_before_switch failed\n"); + return ret; + } + if (eg_pi->pcie_performance_request) + cypress_notify_link_speed_change_before_state_change(rdev, new_ps, old_ps); + + rv770_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + ret = rv770_halt_smc(rdev); + if (ret) { + DRM_ERROR("rv770_halt_smc failed\n"); + return ret; + } + ret = cypress_upload_sw_state(rdev, new_ps); + if (ret) { + DRM_ERROR("cypress_upload_sw_state failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = cypress_upload_mc_reg_table(rdev, new_ps); + if (ret) { + DRM_ERROR("cypress_upload_mc_reg_table failed\n"); + return ret; + } + } + + cypress_program_memory_timing_parameters(rdev, new_ps); + + ret = rv770_resume_smc(rdev); + if (ret) { + DRM_ERROR("rv770_resume_smc failed\n"); + return ret; + } + ret = rv770_set_sw_state(rdev); + if (ret) { + DRM_ERROR("rv770_set_sw_state failed\n"); + return ret; + } + rv770_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + if (eg_pi->pcie_performance_request) + cypress_notify_link_speed_change_after_state_change(rdev, new_ps, old_ps); + + ret = rv770_dpm_force_performance_level(rdev, RADEON_DPM_FORCED_LEVEL_AUTO); + if (ret) { + DRM_ERROR("rv770_dpm_force_performance_level failed\n"); + return ret; + } + + return 0; +} + +void cypress_dpm_reset_asic(struct radeon_device *rdev) +{ + rv770_restrict_performance_levels_before_switch(rdev); + rv770_set_boot_state(rdev); +} + +void cypress_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + cypress_program_display_gap(rdev); +} + +int cypress_dpm_init(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi; + struct evergreen_power_info *eg_pi; + struct atom_clock_dividers dividers; + int ret; + + eg_pi = kzalloc(sizeof(struct evergreen_power_info), GFP_KERNEL); + if (eg_pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = eg_pi; + pi = &eg_pi->rv7xx; + + rv770_get_max_vddc(rdev); + + eg_pi->ulv.supported = false; + pi->acpi_vddc = 0; + eg_pi->acpi_vddci = 0; + pi->min_vddc_in_table = 0; + pi->max_vddc_in_table = 0; + + ret = rv7xx_parse_power_table(rdev); + if (ret) + return ret; + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->ref_div = dividers.ref_div + 1; + else + pi->ref_div = R600_REFERENCEDIVIDER_DFLT; + + pi->mclk_strobe_mode_threshold = 40000; + pi->mclk_edc_enable_threshold = 40000; + eg_pi->mclk_edc_wr_enable_threshold = 40000; + + pi->rlp = RV770_RLP_DFLT; + pi->rmp = RV770_RMP_DFLT; + pi->lhp = RV770_LHP_DFLT; + pi->lmp = RV770_LMP_DFLT; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0); + + pi->mvdd_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_MVDDC, 0); + + eg_pi->vddci_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDCI, 0); + + rv770_get_engine_memory_ss(rdev); + + pi->asi = RV770_ASI_DFLT; + pi->pasi = CYPRESS_HASI_DFLT; + pi->vrc = CYPRESS_VRC_DFLT; + + pi->power_gating = false; + + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK)) + pi->gfx_clock_gating = false; + else + pi->gfx_clock_gating = true; + + pi->mg_clock_gating = true; + pi->mgcgtssm = true; + eg_pi->ls_clock_gating = false; + eg_pi->sclk_deep_sleep = false; + + pi->dynamic_pcie_gen2 = true; + + if (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + pi->display_gap = true; + + if (rdev->flags & RADEON_IS_MOBILITY) + pi->dcodt = true; + else + pi->dcodt = false; + + pi->ulps = true; + + eg_pi->dynamic_ac_timing = true; + eg_pi->abm = true; + eg_pi->mcls = true; + eg_pi->light_sleep = true; + eg_pi->memory_transition = true; +#if defined(CONFIG_ACPI) + eg_pi->pcie_performance_request = + radeon_acpi_is_pcie_performance_request_supported(rdev); +#else + eg_pi->pcie_performance_request = false; +#endif + + if ((rdev->family == CHIP_CYPRESS) || + (rdev->family == CHIP_HEMLOCK) || + (rdev->family == CHIP_JUNIPER)) + eg_pi->dll_default_on = true; + else + eg_pi->dll_default_on = false; + + eg_pi->sclk_deep_sleep = false; + pi->mclk_stutter_mode_threshold = 0; + + pi->sram_end = SMC_RAM_END; + + return 0; +} + +void cypress_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +bool cypress_dpm_vblank_too_short(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 vblank_time = r600_dpm_get_vblank_time(rdev); + u32 switch_limit = pi->mem_gddr5 ? 450 : 300; + + if (vblank_time < switch_limit) + return true; + else + return false; + +} diff --git a/sys/dev/drm/radeon/cypress_dpm.h b/sys/dev/drm/radeon/cypress_dpm.h new file mode 100644 index 0000000000..4c3f18c69f --- /dev/null +++ b/sys/dev/drm/radeon/cypress_dpm.h @@ -0,0 +1,160 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __CYPRESS_DPM_H__ +#define __CYPRESS_DPM_H__ + +#include "rv770_dpm.h" +#include "evergreen_smc.h" + +struct evergreen_mc_reg_entry { + u32 mclk_max; + u32 mc_data[SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE]; +}; + +struct evergreen_mc_reg_table { + u8 last; + u8 num_entries; + u16 valid_flag; + struct evergreen_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES]; + SMC_Evergreen_MCRegisterAddress mc_reg_address[SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE]; +}; + +struct evergreen_ulv_param { + bool supported; + struct rv7xx_pl *pl; +}; + +struct evergreen_arb_registers { + u32 mc_arb_dram_timing; + u32 mc_arb_dram_timing2; + u32 mc_arb_rfsh_rate; + u32 mc_arb_burst_time; +}; + +struct at { + u32 rlp; + u32 rmp; + u32 lhp; + u32 lmp; +}; + +struct evergreen_power_info { + /* must be first! */ + struct rv7xx_power_info rv7xx; + /* flags */ + bool vddci_control; + bool dynamic_ac_timing; + bool abm; + bool mcls; + bool light_sleep; + bool memory_transition; + bool pcie_performance_request; + bool pcie_performance_request_registered; + bool sclk_deep_sleep; + bool dll_default_on; + bool ls_clock_gating; + bool smu_uvd_hs; + bool uvd_enabled; + /* stored values */ + u16 acpi_vddci; + u8 mvdd_high_index; + u8 mvdd_low_index; + u32 mclk_edc_wr_enable_threshold; + struct evergreen_mc_reg_table mc_reg_table; + struct atom_voltage_table vddc_voltage_table; + struct atom_voltage_table vddci_voltage_table; + struct evergreen_arb_registers bootup_arb_registers; + struct evergreen_ulv_param ulv; + struct at ats[2]; + /* smc offsets */ + u16 mc_reg_table_start; + struct radeon_ps current_rps; + struct rv7xx_ps current_ps; + struct radeon_ps requested_rps; + struct rv7xx_ps requested_ps; +}; + +#define CYPRESS_HASI_DFLT 400000 +#define CYPRESS_MGCGTTLOCAL0_DFLT 0x00000000 +#define CYPRESS_MGCGTTLOCAL1_DFLT 0x00000000 +#define CYPRESS_MGCGTTLOCAL2_DFLT 0x00000000 +#define CYPRESS_MGCGTTLOCAL3_DFLT 0x00000000 +#define CYPRESS_MGCGCGTSSMCTRL_DFLT 0x81944bc0 +#define REDWOOD_MGCGCGTSSMCTRL_DFLT 0x6e944040 +#define CEDAR_MGCGCGTSSMCTRL_DFLT 0x46944040 +#define CYPRESS_VRC_DFLT 0xC00033 + +#define PCIE_PERF_REQ_REMOVE_REGISTRY 0 +#define PCIE_PERF_REQ_FORCE_LOWPOWER 1 +#define PCIE_PERF_REQ_PECI_GEN1 2 +#define PCIE_PERF_REQ_PECI_GEN2 3 +#define PCIE_PERF_REQ_PECI_GEN3 4 + +int cypress_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + RV770_SMC_HW_PERFORMANCE_LEVEL *level, + u8 watermark_level); +int cypress_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table); +int cypress_populate_smc_voltage_tables(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table); +int cypress_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_initial_state, + RV770_SMC_STATETABLE *table); +u32 cypress_calculate_burst_time(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock); +void cypress_notify_link_speed_change_before_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state); +int cypress_upload_sw_state(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state); +int cypress_upload_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state); +void cypress_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state); +void cypress_notify_link_speed_change_after_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state); +int cypress_construct_voltage_tables(struct radeon_device *rdev); +int cypress_get_mvdd_configuration(struct radeon_device *rdev); +void cypress_enable_spread_spectrum(struct radeon_device *rdev, + bool enable); +void cypress_enable_display_gap(struct radeon_device *rdev); +int cypress_get_table_locations(struct radeon_device *rdev); +int cypress_populate_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state); +void cypress_program_response_times(struct radeon_device *rdev); +int cypress_notify_smc_display_change(struct radeon_device *rdev, + bool has_display); +void cypress_enable_sclk_control(struct radeon_device *rdev, + bool enable); +void cypress_enable_mclk_control(struct radeon_device *rdev, + bool enable); +void cypress_start_dpm(struct radeon_device *rdev); +void cypress_advertise_gen2_capability(struct radeon_device *rdev); +u32 cypress_map_clkf_to_ibias(struct radeon_device *rdev, u32 clkf); +u8 cypress_get_mclk_frequency_ratio(struct radeon_device *rdev, + u32 memory_clock, bool strobe_mode); +u8 cypress_get_strobe_mode_settings(struct radeon_device *rdev, u32 mclk); + +#endif diff --git a/sys/dev/drm/radeon/evergreen.c b/sys/dev/drm/radeon/evergreen.c index 3e98fb5538..1a8b0663d3 100644 --- a/sys/dev/drm/radeon/evergreen.c +++ b/sys/dev/drm/radeon/evergreen.c @@ -33,9 +33,7 @@ #include "avivod.h" #include "evergreen_reg.h" #include "evergreen_blit_shaders.h" - -#define EVERGREEN_PFP_UCODE_SIZE 1120 -#define EVERGREEN_PM4_UCODE_SIZE 1376 +#include "radeon_ucode.h" static const u32 crtc_offsets[6] = { @@ -47,10 +45,98 @@ static const u32 crtc_offsets[6] = EVERGREEN_CRTC5_REGISTER_OFFSET }; +#include "clearstate_evergreen.h" + +static u32 sumo_rlc_save_restore_register_list[] = +{ + 0x98fc, + 0x9830, + 0x9834, + 0x9838, + 0x9870, + 0x9874, + 0x8a14, + 0x8b24, + 0x8bcc, + 0x8b10, + 0x8d00, + 0x8d04, + 0x8c00, + 0x8c04, + 0x8c08, + 0x8c0c, + 0x8d8c, + 0x8c20, + 0x8c24, + 0x8c28, + 0x8c18, + 0x8c1c, + 0x8cf0, + 0x8e2c, + 0x8e38, + 0x8c30, + 0x9508, + 0x9688, + 0x9608, + 0x960c, + 0x9610, + 0x9614, + 0x88c4, + 0x88d4, + 0xa008, + 0x900c, + 0x9100, + 0x913c, + 0x98f8, + 0x98f4, + 0x9b7c, + 0x3f8c, + 0x8950, + 0x8954, + 0x8a18, + 0x8b28, + 0x9144, + 0x9148, + 0x914c, + 0x3f90, + 0x3f94, + 0x915c, + 0x9160, + 0x9178, + 0x917c, + 0x9180, + 0x918c, + 0x9190, + 0x9194, + 0x9198, + 0x919c, + 0x91a8, + 0x91ac, + 0x91b0, + 0x91b4, + 0x91b8, + 0x91c4, + 0x91c8, + 0x91cc, + 0x91d0, + 0x91d4, + 0x91e0, + 0x91e4, + 0x91ec, + 0x91f0, + 0x91f4, + 0x9200, + 0x9204, + 0x929c, + 0x9150, + 0x802c, +}; +static u32 sumo_rlc_save_restore_register_list_size = ARRAY_SIZE(sumo_rlc_save_restore_register_list); + static void evergreen_gpu_init(struct radeon_device *rdev); void evergreen_pcie_gen2_enable(struct radeon_device *rdev); void evergreen_print_gpu_status_regs(struct radeon_device *rdev); -bool evergreen_is_display_hung(struct radeon_device *rdev); +void evergreen_program_aspm(struct radeon_device *rdev); static const u32 evergreen_golden_registers[] = { @@ -1420,8 +1506,8 @@ void evergreen_pm_misc(struct radeon_device *rdev) struct radeon_voltage *voltage = &ps->clock_info[req_cm_idx].voltage; if (voltage->type == VOLTAGE_SW) { - /* 0xff01 is a flag rather then an actual voltage */ - if (voltage->voltage == 0xff01) + /* 0xff0x are flags rather then an actual voltage */ + if ((voltage->voltage & 0xff00) == 0xff00) return; if (voltage->voltage && (voltage->voltage != rdev->pm.current_vddc)) { radeon_atom_set_voltage(rdev, voltage->voltage, SET_VOLTAGE_TYPE_ASIC_VDDC); @@ -1441,8 +1527,8 @@ void evergreen_pm_misc(struct radeon_device *rdev) voltage = &rdev->pm.power_state[req_ps_idx]. clock_info[rdev->pm.profiles[PM_PROFILE_HIGH_MH_IDX].dpms_on_cm_idx].voltage; - /* 0xff01 is a flag rather then an actual voltage */ - if (voltage->vddci == 0xff01) + /* 0xff0x are flags rather then an actual voltage */ + if ((voltage->vddci & 0xff00) == 0xff00) return; if (voltage->vddci && (voltage->vddci != rdev->pm.current_vddci)) { radeon_atom_set_voltage(rdev, voltage->vddci, SET_VOLTAGE_TYPE_ASIC_VDDCI); @@ -2039,7 +2125,8 @@ static void evergreen_program_watermarks(struct radeon_device *rdev, u32 lb_size, u32 num_heads) { struct drm_display_mode *mode = &radeon_crtc->base.mode; - struct evergreen_wm_params wm; + struct evergreen_wm_params wm_low, wm_high; + u32 dram_channels; u32 pixel_period; u32 line_time = 0; u32 latency_watermark_a = 0, latency_watermark_b = 0; @@ -2055,39 +2142,81 @@ static void evergreen_program_watermarks(struct radeon_device *rdev, line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535); priority_a_cnt = 0; priority_b_cnt = 0; + dram_channels = evergreen_get_number_of_dram_channels(rdev); + + /* watermark for high clocks */ + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + wm_high.yclk = + radeon_dpm_get_mclk(rdev, false) * 10; + wm_high.sclk = + radeon_dpm_get_sclk(rdev, false) * 10; + } else { + wm_high.yclk = rdev->pm.current_mclk * 10; + wm_high.sclk = rdev->pm.current_sclk * 10; + } + + wm_high.disp_clk = mode->clock; + wm_high.src_width = mode->crtc_hdisplay; + wm_high.active_time = mode->crtc_hdisplay * pixel_period; + wm_high.blank_time = line_time - wm_high.active_time; + wm_high.interlaced = false; + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + wm_high.interlaced = true; + wm_high.vsc = radeon_crtc->vsc; + wm_high.vtaps = 1; + if (radeon_crtc->rmx_type != RMX_OFF) + wm_high.vtaps = 2; + wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm_high.lb_size = lb_size; + wm_high.dram_channels = dram_channels; + wm_high.num_heads = num_heads; + + /* watermark for low clocks */ + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + wm_low.yclk = + radeon_dpm_get_mclk(rdev, true) * 10; + wm_low.sclk = + radeon_dpm_get_sclk(rdev, true) * 10; + } else { + wm_low.yclk = rdev->pm.current_mclk * 10; + wm_low.sclk = rdev->pm.current_sclk * 10; + } - wm.yclk = rdev->pm.current_mclk * 10; - wm.sclk = rdev->pm.current_sclk * 10; - wm.disp_clk = mode->clock; - wm.src_width = mode->crtc_hdisplay; - wm.active_time = mode->crtc_hdisplay * pixel_period; - wm.blank_time = line_time - wm.active_time; - wm.interlaced = false; + wm_low.disp_clk = mode->clock; + wm_low.src_width = mode->crtc_hdisplay; + wm_low.active_time = mode->crtc_hdisplay * pixel_period; + wm_low.blank_time = line_time - wm_low.active_time; + wm_low.interlaced = false; if (mode->flags & DRM_MODE_FLAG_INTERLACE) - wm.interlaced = true; - wm.vsc = radeon_crtc->vsc; - wm.vtaps = 1; + wm_low.interlaced = true; + wm_low.vsc = radeon_crtc->vsc; + wm_low.vtaps = 1; if (radeon_crtc->rmx_type != RMX_OFF) - wm.vtaps = 2; - wm.bytes_per_pixel = 4; /* XXX: get this from fb config */ - wm.lb_size = lb_size; - wm.dram_channels = evergreen_get_number_of_dram_channels(rdev); - wm.num_heads = num_heads; + wm_low.vtaps = 2; + wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm_low.lb_size = lb_size; + wm_low.dram_channels = dram_channels; + wm_low.num_heads = num_heads; /* set for high clocks */ - latency_watermark_a = min(evergreen_latency_watermark(&wm), (u32)65535); + latency_watermark_a = min(evergreen_latency_watermark(&wm_high), (u32)65535); /* set for low clocks */ - /* wm.yclk = low clk; wm.sclk = low clk */ - latency_watermark_b = min(evergreen_latency_watermark(&wm), (u32)65535); + latency_watermark_b = min(evergreen_latency_watermark(&wm_low), (u32)65535); /* possibly force display priority to high */ /* should really do this at mode validation time... */ - if (!evergreen_average_bandwidth_vs_dram_bandwidth_for_display(&wm) || - !evergreen_average_bandwidth_vs_available_bandwidth(&wm) || - !evergreen_check_latency_hiding(&wm) || + if (!evergreen_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) || + !evergreen_average_bandwidth_vs_available_bandwidth(&wm_high) || + !evergreen_check_latency_hiding(&wm_high) || (rdev->disp_priority == 2)) { - DRM_DEBUG_KMS("force priority to high\n"); + DRM_DEBUG_KMS("force priority a to high\n"); priority_a_cnt |= PRIORITY_ALWAYS_ON; + } + if (!evergreen_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) || + !evergreen_average_bandwidth_vs_available_bandwidth(&wm_low) || + !evergreen_check_latency_hiding(&wm_low) || + (rdev->disp_priority == 2)) { + DRM_DEBUG_KMS("force priority b to high\n"); priority_b_cnt |= PRIORITY_ALWAYS_ON; } @@ -2140,6 +2269,10 @@ static void evergreen_program_watermarks(struct radeon_device *rdev, WREG32(PRIORITY_A_CNT + radeon_crtc->crtc_offset, priority_a_cnt); WREG32(PRIORITY_B_CNT + radeon_crtc->crtc_offset, priority_b_cnt); + /* save values for DPM */ + radeon_crtc->line_time = line_time; + radeon_crtc->wm_high = latency_watermark_a; + radeon_crtc->wm_low = latency_watermark_b; } /** @@ -3123,10 +3256,8 @@ static void evergreen_gpu_init(struct radeon_device *rdev) u32 efuse_straps_4; u32 efuse_straps_3; - WREG32(RCU_IND_INDEX, 0x204); - efuse_straps_4 = RREG32(RCU_IND_DATA); - WREG32(RCU_IND_INDEX, 0x203); - efuse_straps_3 = RREG32(RCU_IND_DATA); + efuse_straps_4 = RREG32_RCU(0x204); + efuse_straps_3 = RREG32_RCU(0x203); tmp = (((efuse_straps_4 & 0xf) << 4) | ((efuse_straps_3 & 0xf0000000) >> 28)); } else { @@ -3730,6 +3861,266 @@ bool evergreen_dma_is_lockup(struct radeon_device *rdev, struct radeon_ring *rin return radeon_ring_test_lockup(rdev, ring); } +/* + * RLC + */ +#define RLC_SAVE_RESTORE_LIST_END_MARKER 0x00000000 +#define RLC_CLEAR_STATE_END_MARKER 0x00000001 + +void sumo_rlc_fini(struct radeon_device *rdev) +{ + int r; + + /* save restore block */ + if (rdev->rlc.save_restore_obj) { + r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); + if (unlikely(r != 0)) + dev_warn(rdev->dev, "(%d) reserve RLC sr bo failed\n", r); + radeon_bo_unpin(rdev->rlc.save_restore_obj); + radeon_bo_unreserve(rdev->rlc.save_restore_obj); + + radeon_bo_unref(&rdev->rlc.save_restore_obj); + rdev->rlc.save_restore_obj = NULL; + } + + /* clear state block */ + if (rdev->rlc.clear_state_obj) { + r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); + if (unlikely(r != 0)) + dev_warn(rdev->dev, "(%d) reserve RLC c bo failed\n", r); + radeon_bo_unpin(rdev->rlc.clear_state_obj); + radeon_bo_unreserve(rdev->rlc.clear_state_obj); + + radeon_bo_unref(&rdev->rlc.clear_state_obj); + rdev->rlc.clear_state_obj = NULL; + } +} + +int sumo_rlc_init(struct radeon_device *rdev) +{ + u32 *src_ptr; + volatile u32 *dst_ptr; + u32 dws, data, i, j, k, reg_num; + u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index; + u64 reg_list_mc_addr; + struct cs_section_def *cs_data; + int r; + void *vptr; + + vptr = NULL; + src_ptr = rdev->rlc.reg_list; + dws = rdev->rlc.reg_list_size; + cs_data = rdev->rlc.cs_data; + + /* save restore block */ + if (rdev->rlc.save_restore_obj == NULL) { + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.save_restore_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create RLC sr bo failed\n", r); + return r; + } + } + + r = radeon_bo_reserve(rdev->rlc.save_restore_obj, false); + if (unlikely(r != 0)) { + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->rlc.save_restore_gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->rlc.save_restore_obj); + dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_kmap(rdev->rlc.save_restore_obj, (void**)&vptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC sr bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + rdev->rlc.sr_ptr = vptr; + /* write the sr buffer */ + dst_ptr = rdev->rlc.sr_ptr; + /* format: + * dw0: (reg2 << 16) | reg1 + * dw1: reg1 save space + * dw2: reg2 save space + */ + for (i = 0; i < dws; i++) { + data = src_ptr[i] >> 2; + i++; + if (i < dws) + data |= (src_ptr[i] >> 2) << 16; + j = (((i - 1) * 3) / 2); + dst_ptr[j] = data; + } + j = ((i * 3) / 2); + dst_ptr[j] = RLC_SAVE_RESTORE_LIST_END_MARKER; + + radeon_bo_kunmap(rdev->rlc.save_restore_obj); + radeon_bo_unreserve(rdev->rlc.save_restore_obj); + + /* clear state block */ + reg_list_num = 0; + dws = 0; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_list_num++; + dws += cs_data[i].section[j].reg_count; + } + } + reg_list_blk_index = (3 * reg_list_num + 2); + dws += reg_list_blk_index; + + if (rdev->rlc.clear_state_obj == NULL) { + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); + if (r) { + dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + } + r = radeon_bo_reserve(rdev->rlc.clear_state_obj, false); + if (unlikely(r != 0)) { + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM, + &rdev->rlc.clear_state_gpu_addr); + if (r) { + + radeon_bo_unreserve(rdev->rlc.clear_state_obj); + dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + r = radeon_bo_kmap(rdev->rlc.clear_state_obj, (void **)&vptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC c bo failed\n", r); + sumo_rlc_fini(rdev); + return r; + } + rdev->rlc.cs_ptr = vptr; + /* set up the cs buffer */ + dst_ptr = rdev->rlc.cs_ptr; + reg_list_hdr_blk_index = 0; + reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); + data = upper_32_bits(reg_list_mc_addr); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_num = cs_data[i].section[j].reg_count; + data = reg_list_mc_addr & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = 0x08000000 | (reg_num * 4); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + for (k = 0; k < reg_num; k++) { + data = cs_data[i].section[j].extent[k]; + dst_ptr[reg_list_blk_index + k] = data; + } + reg_list_mc_addr += reg_num * 4; + reg_list_blk_index += reg_num; + } + } + dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; + + radeon_bo_kunmap(rdev->rlc.clear_state_obj); + radeon_bo_unreserve(rdev->rlc.clear_state_obj); + + return 0; +} + +static void evergreen_rlc_start(struct radeon_device *rdev) +{ + u32 mask = RLC_ENABLE; + + if (rdev->flags & RADEON_IS_IGP) { + mask |= GFX_POWER_GATING_ENABLE | GFX_POWER_GATING_SRC; + } + + WREG32(RLC_CNTL, mask); +} + +int evergreen_rlc_resume(struct radeon_device *rdev) +{ + u32 i; + const __be32 *fw_data; + + if (!rdev->rlc_fw) + return -EINVAL; + + r600_rlc_stop(rdev); + + WREG32(RLC_HB_CNTL, 0); + + if (rdev->flags & RADEON_IS_IGP) { + if (rdev->family == CHIP_ARUBA) { + u32 always_on_bitmap = + 3 | (3 << (16 * rdev->config.cayman.max_shader_engines)); + /* find out the number of active simds */ + u32 tmp = (RREG32(CC_GC_SHADER_PIPE_CONFIG) & 0xffff0000) >> 16; + tmp |= 0xffffffff << rdev->config.cayman.max_simds_per_se; + tmp = hweight32(~tmp); + if (tmp == rdev->config.cayman.max_simds_per_se) { + WREG32(TN_RLC_LB_ALWAYS_ACTIVE_SIMD_MASK, always_on_bitmap); + WREG32(TN_RLC_LB_PARAMS, 0x00601004); + WREG32(TN_RLC_LB_INIT_SIMD_MASK, 0xffffffff); + WREG32(TN_RLC_LB_CNTR_INIT, 0x00000000); + WREG32(TN_RLC_LB_CNTR_MAX, 0x00002000); + } + } else { + WREG32(RLC_HB_WPTR_LSB_ADDR, 0); + WREG32(RLC_HB_WPTR_MSB_ADDR, 0); + } + WREG32(TN_RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); + WREG32(TN_RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8); + } else { + WREG32(RLC_HB_BASE, 0); + WREG32(RLC_HB_RPTR, 0); + WREG32(RLC_HB_WPTR, 0); + WREG32(RLC_HB_WPTR_LSB_ADDR, 0); + WREG32(RLC_HB_WPTR_MSB_ADDR, 0); + } + WREG32(RLC_MC_CNTL, 0); + WREG32(RLC_UCODE_CNTL, 0); + + fw_data = (const __be32 *)rdev->rlc_fw->data; + if (rdev->family >= CHIP_ARUBA) { + for (i = 0; i < ARUBA_RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } else if (rdev->family >= CHIP_CAYMAN) { + for (i = 0; i < CAYMAN_RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } else { + for (i = 0; i < EVERGREEN_RLC_UCODE_SIZE; i++) { + WREG32(RLC_UCODE_ADDR, i); + WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); + } + } + WREG32(RLC_UCODE_ADDR, 0); + + evergreen_rlc_start(rdev); + + return 0; +} + /* Interrupts */ u32 evergreen_get_vblank_counter(struct radeon_device *rdev, int crtc) @@ -3808,6 +4199,7 @@ int evergreen_irq_set(struct radeon_device *rdev) u32 grph1 = 0, grph2 = 0, grph3 = 0, grph4 = 0, grph5 = 0, grph6 = 0; u32 afmt1 = 0, afmt2 = 0, afmt3 = 0, afmt4 = 0, afmt5 = 0, afmt6 = 0; u32 dma_cntl, dma_cntl1 = 0; + u32 thermal_int = 0; if (!rdev->irq.installed) { dev_warn(rdev->dev, "Can't enable IRQ/MSI because no handler is installed\n"); @@ -3827,6 +4219,12 @@ int evergreen_irq_set(struct radeon_device *rdev) hpd4 = RREG32(DC_HPD4_INT_CONTROL) & ~DC_HPDx_INT_EN; hpd5 = RREG32(DC_HPD5_INT_CONTROL) & ~DC_HPDx_INT_EN; hpd6 = RREG32(DC_HPD6_INT_CONTROL) & ~DC_HPDx_INT_EN; + if (rdev->family == CHIP_ARUBA) + thermal_int = RREG32(TN_CG_THERMAL_INT_CTRL) & + ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + else + thermal_int = RREG32(CG_THERMAL_INT) & + ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); afmt1 = RREG32(AFMT_AUDIO_PACKET_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET) & ~AFMT_AZ_FORMAT_WTRIG_MASK; afmt2 = RREG32(AFMT_AUDIO_PACKET_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET) & ~AFMT_AZ_FORMAT_WTRIG_MASK; @@ -3872,6 +4270,11 @@ int evergreen_irq_set(struct radeon_device *rdev) } } + if (rdev->irq.dpm_thermal) { + DRM_DEBUG("dpm thermal\n"); + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + } + if (rdev->irq.crtc_vblank_int[0] || atomic_read(&rdev->irq.pflip[0])) { DRM_DEBUG("evergreen_irq_set: vblank 0\n"); @@ -3993,6 +4396,10 @@ int evergreen_irq_set(struct radeon_device *rdev) WREG32(DC_HPD4_INT_CONTROL, hpd4); WREG32(DC_HPD5_INT_CONTROL, hpd5); WREG32(DC_HPD6_INT_CONTROL, hpd6); + if (rdev->family == CHIP_ARUBA) + WREG32(TN_CG_THERMAL_INT_CTRL, thermal_int); + else + WREG32(CG_THERMAL_INT, thermal_int); WREG32(AFMT_AUDIO_PACKET_CONTROL + EVERGREEN_CRTC0_REGISTER_OFFSET, afmt1); WREG32(AFMT_AUDIO_PACKET_CONTROL + EVERGREEN_CRTC1_REGISTER_OFFSET, afmt2); @@ -4184,6 +4591,8 @@ irqreturn_t evergreen_irq_process(struct radeon_device *rdev) u32 ring_index; bool queue_hotplug = false; bool queue_hdmi = false; + bool queue_thermal = false; + u32 status, addr; if (!rdev->ih.enabled || rdev->shutdown) return IRQ_NONE; @@ -4470,11 +4879,14 @@ restart_ih: break; case 146: case 147: + addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR); + status = RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS); dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data); dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n", - RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR)); + addr); dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n", - RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS)); + status); + cayman_vm_decode_fault(rdev, status, addr); /* reset addr and status */ WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1); break; @@ -4505,6 +4917,16 @@ restart_ih: DRM_DEBUG("IH: DMA trap\n"); radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX); break; + case 230: /* thermal low to high */ + DRM_DEBUG("IH: thermal low to high\n"); + rdev->pm.dpm.thermal.high_to_low = false; + queue_thermal = true; + break; + case 231: /* thermal high to low */ + DRM_DEBUG("IH: thermal high to low\n"); + rdev->pm.dpm.thermal.high_to_low = true; + queue_thermal = true; + break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); break; @@ -4527,6 +4949,9 @@ restart_ih: taskqueue_enqueue(rdev->tq, &rdev->hotplug_work); if (queue_hdmi) taskqueue_enqueue(rdev->tq, &rdev->audio_work); + if (queue_thermal && rdev->pm.dpm_enabled) + taskqueue_enqueue(rdev->tq, &rdev->pm.dpm.thermal.work); + rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); atomic_set(&rdev->ih.lock, 0); @@ -4683,6 +5108,10 @@ static int evergreen_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ evergreen_pcie_gen2_enable(rdev); + /* enable aspm */ + evergreen_program_aspm(rdev); + + evergreen_mc_program(rdev); if (ASIC_IS_DCE5(rdev)) { if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw || !rdev->mc_fw) { @@ -4711,7 +5140,6 @@ static int evergreen_startup(struct radeon_device *rdev) if (r) return r; - evergreen_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { evergreen_agp_enable(rdev); } else { @@ -4728,6 +5156,18 @@ static int evergreen_startup(struct radeon_device *rdev) dev_warn(rdev->dev, "failed blitter (%d) falling back to memcpy\n", r); } + /* allocate rlc buffers */ + if (rdev->flags & RADEON_IS_IGP) { + rdev->rlc.reg_list = sumo_rlc_save_restore_register_list; + rdev->rlc.reg_list_size = sumo_rlc_save_restore_register_list_size; + rdev->rlc.cs_data = evergreen_cs_data; + r = sumo_rlc_init(rdev); + if (r) { + DRM_ERROR("Failed to init rlc BOs!\n"); + return r; + } + } + /* allocate wb buffer */ r = radeon_wb_init(rdev); if (r) @@ -4857,10 +5297,10 @@ int evergreen_resume(struct radeon_device *rdev) int evergreen_suspend(struct radeon_device *rdev) { r600_audio_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_suspend(rdev); r700_cp_stop(rdev); r600_dma_stop(rdev); - r600_uvd_rbc_stop(rdev); evergreen_irq_suspend(rdev); radeon_wb_disable(rdev); evergreen_pcie_gart_disable(rdev); @@ -4959,6 +5399,8 @@ int evergreen_init(struct radeon_device *rdev) r700_cp_fini(rdev); r600_dma_fini(rdev); r600_irq_fini(rdev); + if (rdev->flags & RADEON_IS_IGP) + sumo_rlc_fini(rdev); radeon_wb_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); @@ -4987,10 +5429,13 @@ void evergreen_fini(struct radeon_device *rdev) r700_cp_fini(rdev); r600_dma_fini(rdev); r600_irq_fini(rdev); + if (rdev->flags & RADEON_IS_IGP) + sumo_rlc_fini(rdev); radeon_wb_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); evergreen_pcie_gart_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); @@ -5072,3 +5517,153 @@ void evergreen_pcie_gen2_enable(struct radeon_device *rdev) WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, link_width_cntl); } } + +void evergreen_program_aspm(struct radeon_device *rdev) +{ + u32 data, orig; + u32 pcie_lc_cntl, pcie_lc_cntl_old; + bool disable_l0s, disable_l1 = false, disable_plloff_in_l1 = false; + /* fusion_platform = true + * if the system is a fusion system + * (APU or DGPU in a fusion system). + * todo: check if the system is a fusion platform. + */ + bool fusion_platform = false; + + if (radeon_aspm == 0) + return; + + if (!(rdev->flags & RADEON_IS_PCIE)) + return; + + switch (rdev->family) { + case CHIP_CYPRESS: + case CHIP_HEMLOCK: + case CHIP_JUNIPER: + case CHIP_REDWOOD: + case CHIP_CEDAR: + case CHIP_SUMO: + case CHIP_SUMO2: + case CHIP_PALM: + case CHIP_ARUBA: + disable_l0s = true; + break; + default: + disable_l0s = false; + break; + } + + if (rdev->flags & RADEON_IS_IGP) + fusion_platform = true; /* XXX also dGPUs in a fusion system */ + + data = orig = RREG32_PIF_PHY0(PB0_PIF_PAIRING); + if (fusion_platform) + data &= ~MULTI_PIF; + else + data |= MULTI_PIF; + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_PAIRING, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_PAIRING); + if (fusion_platform) + data &= ~MULTI_PIF; + else + data |= MULTI_PIF; + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_PAIRING, data); + + pcie_lc_cntl = pcie_lc_cntl_old = RREG32_PCIE_PORT(PCIE_LC_CNTL); + pcie_lc_cntl &= ~(LC_L0S_INACTIVITY_MASK | LC_L1_INACTIVITY_MASK); + if (!disable_l0s) { + if (rdev->family >= CHIP_BARTS) + pcie_lc_cntl |= LC_L0S_INACTIVITY(7); + else + pcie_lc_cntl |= LC_L0S_INACTIVITY(3); + } + + if (!disable_l1) { + if (rdev->family >= CHIP_BARTS) + pcie_lc_cntl |= LC_L1_INACTIVITY(7); + else + pcie_lc_cntl |= LC_L1_INACTIVITY(8); + + if (!disable_plloff_in_l1) { + data = orig = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0); + data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK); + data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7); + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data); + + data = orig = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1); + data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK); + data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7); + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0); + data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK); + data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7); + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1); + data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK); + data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7); + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data); + + if (rdev->family >= CHIP_BARTS) { + data = orig = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0); + data &= ~PLL_RAMP_UP_TIME_0_MASK; + data |= PLL_RAMP_UP_TIME_0(4); + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data); + + data = orig = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1); + data &= ~PLL_RAMP_UP_TIME_1_MASK; + data |= PLL_RAMP_UP_TIME_1(4); + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0); + data &= ~PLL_RAMP_UP_TIME_0_MASK; + data |= PLL_RAMP_UP_TIME_0(4); + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1); + data &= ~PLL_RAMP_UP_TIME_1_MASK; + data |= PLL_RAMP_UP_TIME_1(4); + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data); + } + + data = orig = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL); + data &= ~LC_DYN_LANES_PWR_STATE_MASK; + data |= LC_DYN_LANES_PWR_STATE(3); + if (data != orig) + WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, data); + + if (rdev->family >= CHIP_BARTS) { + data = orig = RREG32_PIF_PHY0(PB0_PIF_CNTL); + data &= ~LS2_EXIT_TIME_MASK; + data |= LS2_EXIT_TIME(1); + if (data != orig) + WREG32_PIF_PHY0(PB0_PIF_CNTL, data); + + data = orig = RREG32_PIF_PHY1(PB1_PIF_CNTL); + data &= ~LS2_EXIT_TIME_MASK; + data |= LS2_EXIT_TIME(1); + if (data != orig) + WREG32_PIF_PHY1(PB1_PIF_CNTL, data); + } + } + } + + /* evergreen parts only */ + if (rdev->family < CHIP_BARTS) + pcie_lc_cntl |= LC_PMI_TO_L1_DIS; + + if (pcie_lc_cntl != pcie_lc_cntl_old) + WREG32_PCIE_PORT(PCIE_LC_CNTL, pcie_lc_cntl); +} diff --git a/sys/dev/drm/radeon/evergreen_blit_kms.c b/sys/dev/drm/radeon/evergreen_blit_kms.c index 07ff532dcb..1b22ec3fe4 100644 --- a/sys/dev/drm/radeon/evergreen_blit_kms.c +++ b/sys/dev/drm/radeon/evergreen_blit_kms.c @@ -22,8 +22,6 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/evergreen_blit_kms.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/evergreen_blit_shaders.c b/sys/dev/drm/radeon/evergreen_blit_shaders.c index 0586bc3fbf..39a4b02dbb 100644 --- a/sys/dev/drm/radeon/evergreen_blit_shaders.c +++ b/sys/dev/drm/radeon/evergreen_blit_shaders.c @@ -22,8 +22,6 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/evergreen_blit_shaders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/evergreen_blit_shaders.h b/sys/dev/drm/radeon/evergreen_blit_shaders.h index fe7938ff59..bb8d6c7515 100644 --- a/sys/dev/drm/radeon/evergreen_blit_shaders.h +++ b/sys/dev/drm/radeon/evergreen_blit_shaders.h @@ -20,7 +20,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/evergreen_blit_shaders.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef EVERGREEN_BLIT_SHADERS_H diff --git a/sys/dev/drm/radeon/evergreen_hdmi.c b/sys/dev/drm/radeon/evergreen_hdmi.c index 3103ef4ec5..c45f862e6a 100644 --- a/sys/dev/drm/radeon/evergreen_hdmi.c +++ b/sys/dev/drm/radeon/evergreen_hdmi.c @@ -23,10 +23,7 @@ * * Authors: Christian König * Rafał Miłecki - * - * $FreeBSD: head/sys/dev/drm2/radeon/evergreen_hdmi.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include @@ -132,13 +129,7 @@ static void evergreen_hdmi_update_avi_infoframe(struct drm_encoder *encoder, uint32_t offset = dig->afmt->offset; uint8_t *frame = (uint8_t*)buffer + 3; - /* Our header values (type, version, length) should be alright, Intel - * is using the same. Checksum function also seems to be OK, it works - * fine for audio infoframe. However calculated value is always lower - * by 2 in comparison to fglrx. It breaks displaying anything in case - * of TVs that strictly check the checksum. Hack it manually here to - * workaround this issue. */ - frame[0x0] += 2; + uint8_t *header = buffer; WREG32(AFMT_AVI_INFO0 + offset, frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24)); @@ -147,7 +138,7 @@ static void evergreen_hdmi_update_avi_infoframe(struct drm_encoder *encoder, WREG32(AFMT_AVI_INFO2 + offset, frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24)); WREG32(AFMT_AVI_INFO3 + offset, - frame[0xC] | (frame[0xD] << 8)); + frame[0xC] | (frame[0xD] << 8) | (header[1] << 24)); } static void evergreen_audio_set_dto(struct drm_encoder *encoder, u32 clock) @@ -158,18 +149,40 @@ static void evergreen_audio_set_dto(struct drm_encoder *encoder, u32 clock) struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv; struct radeon_crtc *radeon_crtc = to_radeon_crtc(encoder->crtc); u32 base_rate = 24000; + u32 max_ratio = clock / base_rate; + u32 dto_phase; + u32 dto_modulo = clock; + u32 wallclock_ratio; + u32 dto_cntl; if (!dig || !dig->afmt) return; + if (max_ratio >= 8) { + dto_phase = 192 * 1000; + wallclock_ratio = 3; + } else if (max_ratio >= 4) { + dto_phase = 96 * 1000; + wallclock_ratio = 2; + } else if (max_ratio >= 2) { + dto_phase = 48 * 1000; + wallclock_ratio = 1; + } else { + dto_phase = 24 * 1000; + wallclock_ratio = 0; + } + dto_cntl = RREG32(DCCG_AUDIO_DTO0_CNTL) & ~DCCG_AUDIO_DTO_WALLCLOCK_RATIO_MASK; + dto_cntl |= DCCG_AUDIO_DTO_WALLCLOCK_RATIO(wallclock_ratio); + WREG32(DCCG_AUDIO_DTO0_CNTL, dto_cntl); + /* XXX two dtos; generally use dto0 for hdmi */ /* Express [24MHz / target pixel clock] as an exact rational * number (coefficient of two integer numbers. DCCG_AUDIO_DTOx_PHASE * is the numerator, DCCG_AUDIO_DTOx_MODULE is the denominator */ - WREG32(DCCG_AUDIO_DTO0_PHASE, base_rate * 100); - WREG32(DCCG_AUDIO_DTO0_MODULE, clock * 100); WREG32(DCCG_AUDIO_DTO_SOURCE, DCCG_AUDIO_DTO0_SOURCE_SEL(radeon_crtc->crtc_id)); + WREG32(DCCG_AUDIO_DTO0_PHASE, dto_phase); + WREG32(DCCG_AUDIO_DTO0_MODULE, dto_modulo); } @@ -187,6 +200,9 @@ void evergreen_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode uint32_t offset; ssize_t err; + if (!dig || !dig->afmt) + return; + /* Silent, r600_hdmi_enable will raise WARN for us */ if (!dig->afmt->enabled) return; @@ -290,6 +306,9 @@ void evergreen_hdmi_enable(struct drm_encoder *encoder, bool enable) struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv; + if (!dig || !dig->afmt) + return; + /* Silent, r600_hdmi_enable will raise WARN for us */ if (enable && dig->afmt->enabled) return; diff --git a/sys/dev/drm/radeon/evergreen_reg.h b/sys/dev/drm/radeon/evergreen_reg.h index f02fb8d703..8a4e641f0e 100644 --- a/sys/dev/drm/radeon/evergreen_reg.h +++ b/sys/dev/drm/radeon/evergreen_reg.h @@ -20,13 +20,20 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * $FreeBSD: head/sys/dev/drm2/radeon/evergreen_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ - #ifndef __EVERGREEN_REG_H__ #define __EVERGREEN_REG_H__ +/* trinity */ +#define TN_SMC_IND_INDEX_0 0x200 +#define TN_SMC_IND_DATA_0 0x204 + /* evergreen */ +#define EVERGREEN_PIF_PHY0_INDEX 0x8 +#define EVERGREEN_PIF_PHY0_DATA 0xc +#define EVERGREEN_PIF_PHY1_INDEX 0x10 +#define EVERGREEN_PIF_PHY1_DATA 0x14 + #define EVERGREEN_VGA_MEMORY_BASE_ADDRESS 0x310 #define EVERGREEN_VGA_MEMORY_BASE_ADDRESS_HIGH 0x324 #define EVERGREEN_D3VGA_CONTROL 0x3e0 @@ -42,6 +49,9 @@ #define EVERGREEN_AUDIO_PLL1_DIV 0x5b4 #define EVERGREEN_AUDIO_PLL1_UNK 0x5bc +#define EVERGREEN_CG_IND_ADDR 0x8f8 +#define EVERGREEN_CG_IND_DATA 0x8fc + #define EVERGREEN_AUDIO_ENABLE 0x5e78 #define EVERGREEN_AUDIO_VENDOR_ID 0x5ec0 diff --git a/sys/dev/drm/radeon/evergreen_smc.h b/sys/dev/drm/radeon/evergreen_smc.h new file mode 100644 index 0000000000..76ada8cfe9 --- /dev/null +++ b/sys/dev/drm/radeon/evergreen_smc.h @@ -0,0 +1,67 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __EVERGREEN_SMC_H__ +#define __EVERGREEN_SMC_H__ + +#include "rv770_smc.h" + +#pragma pack(push, 1) + +#define SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE 16 + +struct SMC_Evergreen_MCRegisterAddress +{ + uint16_t s0; + uint16_t s1; +}; + +typedef struct SMC_Evergreen_MCRegisterAddress SMC_Evergreen_MCRegisterAddress; + + +struct SMC_Evergreen_MCRegisterSet +{ + uint32_t value[SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE]; +}; + +typedef struct SMC_Evergreen_MCRegisterSet SMC_Evergreen_MCRegisterSet; + +struct SMC_Evergreen_MCRegisters +{ + uint8_t last; + uint8_t reserved[3]; + SMC_Evergreen_MCRegisterAddress address[SMC_EVERGREEN_MC_REGISTER_ARRAY_SIZE]; + SMC_Evergreen_MCRegisterSet data[5]; +}; + +typedef struct SMC_Evergreen_MCRegisters SMC_Evergreen_MCRegisters; + +#define EVERGREEN_SMC_FIRMWARE_HEADER_LOCATION 0x100 + +#define EVERGREEN_SMC_FIRMWARE_HEADER_softRegisters 0x0 +#define EVERGREEN_SMC_FIRMWARE_HEADER_stateTable 0xC +#define EVERGREEN_SMC_FIRMWARE_HEADER_mcRegisterTable 0x20 + + +#pragma pack(pop) + +#endif diff --git a/sys/dev/drm/radeon/evergreend.h b/sys/dev/drm/radeon/evergreend.h index 5cda8c8dbf..0d582ac1dc 100644 --- a/sys/dev/drm/radeon/evergreend.h +++ b/sys/dev/drm/radeon/evergreend.h @@ -20,10 +20,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/evergreend.h 254885 2013-08-25 19:37:15Z dumbbell $ */ - #ifndef EVERGREEND_H #define EVERGREEND_H @@ -51,6 +48,293 @@ #define SUMO_GB_ADDR_CONFIG_GOLDEN 0x02010002 #define SUMO2_GB_ADDR_CONFIG_GOLDEN 0x02010002 +/* pm registers */ +#define SMC_MSG 0x20c +#define HOST_SMC_MSG(x) ((x) << 0) +#define HOST_SMC_MSG_MASK (0xff << 0) +#define HOST_SMC_MSG_SHIFT 0 +#define HOST_SMC_RESP(x) ((x) << 8) +#define HOST_SMC_RESP_MASK (0xff << 8) +#define HOST_SMC_RESP_SHIFT 8 +#define SMC_HOST_MSG(x) ((x) << 16) +#define SMC_HOST_MSG_MASK (0xff << 16) +#define SMC_HOST_MSG_SHIFT 16 +#define SMC_HOST_RESP(x) ((x) << 24) +#define SMC_HOST_RESP_MASK (0xff << 24) +#define SMC_HOST_RESP_SHIFT 24 + +#define DCCG_DISP_SLOW_SELECT_REG 0x4fc +#define DCCG_DISP1_SLOW_SELECT(x) ((x) << 0) +#define DCCG_DISP1_SLOW_SELECT_MASK (7 << 0) +#define DCCG_DISP1_SLOW_SELECT_SHIFT 0 +#define DCCG_DISP2_SLOW_SELECT(x) ((x) << 4) +#define DCCG_DISP2_SLOW_SELECT_MASK (7 << 4) +#define DCCG_DISP2_SLOW_SELECT_SHIFT 4 + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_PDIV_A(x) ((x) << 20) +#define SPLL_PDIV_A_MASK (0x7f << 20) +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_DITHEN (1 << 28) + +#define MPLL_CNTL_MODE 0x61c +# define SS_SSEN (1 << 24) +# define SS_DSMODE_EN (1 << 25) + +#define MPLL_AD_FUNC_CNTL 0x624 +#define CLKF(x) ((x) << 0) +#define CLKF_MASK (0x7f << 0) +#define CLKR(x) ((x) << 7) +#define CLKR_MASK (0x1f << 7) +#define CLKFRAC(x) ((x) << 12) +#define CLKFRAC_MASK (0x1f << 12) +#define YCLK_POST_DIV(x) ((x) << 17) +#define YCLK_POST_DIV_MASK (3 << 17) +#define IBIAS(x) ((x) << 20) +#define IBIAS_MASK (0x3ff << 20) +#define RESET (1 << 30) +#define PDNB (1 << 31) +#define MPLL_AD_FUNC_CNTL_2 0x628 +#define BYPASS (1 << 19) +#define BIAS_GEN_PDNB (1 << 24) +#define RESET_EN (1 << 25) +#define VCO_MODE (1 << 29) +#define MPLL_DQ_FUNC_CNTL 0x62c +#define MPLL_DQ_FUNC_CNTL_2 0x630 + +#define GENERAL_PWRMGT 0x63c +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define ENABLE_GEN2PCIE (1 << 4) +# define ENABLE_GEN2XSP (1 << 5) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (3 << 6) +# define SW_SMIO_INDEX_SHIFT 6 +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 18) +# define BACKBIAS_VALUE (1 << 19) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +# define AC_DC_SW (1 << 24) + +#define SCLK_PWRMGT_CNTL 0x644 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) +# define DYN_LIGHT_SLEEP_EN (1 << 14) +#define MCLK_PWRMGT_CNTL 0x648 +# define DLL_SPEED(x) ((x) << 0) +# define DLL_SPEED_MASK (0x1f << 0) +# define MPLL_PWRMGT_OFF (1 << 5) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA0_PDNB (1 << 8) +# define MRDCKA1_PDNB (1 << 9) +# define MRDCKB0_PDNB (1 << 10) +# define MRDCKB1_PDNB (1 << 11) +# define MRDCKC0_PDNB (1 << 12) +# define MRDCKC1_PDNB (1 << 13) +# define MRDCKD0_PDNB (1 << 14) +# define MRDCKD1_PDNB (1 << 15) +# define MRDCKA0_RESET (1 << 16) +# define MRDCKA1_RESET (1 << 17) +# define MRDCKB0_RESET (1 << 18) +# define MRDCKB1_RESET (1 << 19) +# define MRDCKC0_RESET (1 << 20) +# define MRDCKC1_RESET (1 << 21) +# define MRDCKD0_RESET (1 << 22) +# define MRDCKD1_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define MPLL_TURNOFF_D2 (1 << 28) +#define DLL_CNTL 0x64c +# define MRDCKA0_BYPASS (1 << 24) +# define MRDCKA1_BYPASS (1 << 25) +# define MRDCKB0_BYPASS (1 << 26) +# define MRDCKB1_BYPASS (1 << 27) +# define MRDCKC0_BYPASS (1 << 28) +# define MRDCKC1_BYPASS (1 << 29) +# define MRDCKD0_BYPASS (1 << 30) +# define MRDCKD1_BYPASS (1 << 31) + +#define CG_AT 0x6d4 +# define CG_R(x) ((x) << 0) +# define CG_R_MASK (0xffff << 0) +# define CG_L(x) ((x) << 16) +# define CG_L_MASK (0xffff << 16) + +#define CG_DISPLAY_GAP_CNTL 0x714 +# define DISP1_GAP(x) ((x) << 0) +# define DISP1_GAP_MASK (3 << 0) +# define DISP2_GAP(x) ((x) << 2) +# define DISP2_GAP_MASK (3 << 2) +# define VBI_TIMER_COUNT(x) ((x) << 4) +# define VBI_TIMER_COUNT_MASK (0x3fff << 4) +# define VBI_TIMER_UNIT(x) ((x) << 20) +# define VBI_TIMER_UNIT_MASK (7 << 20) +# define DISP1_GAP_MCHG(x) ((x) << 24) +# define DISP1_GAP_MCHG_MASK (3 << 24) +# define DISP2_GAP_MCHG(x) ((x) << 26) +# define DISP2_GAP_MCHG_MASK (3 << 26) + +#define CG_BIF_REQ_AND_RSP 0x7f4 +#define CG_CLIENT_REQ(x) ((x) << 0) +#define CG_CLIENT_REQ_MASK (0xff << 0) +#define CG_CLIENT_REQ_SHIFT 0 +#define CG_CLIENT_RESP(x) ((x) << 8) +#define CG_CLIENT_RESP_MASK (0xff << 8) +#define CG_CLIENT_RESP_SHIFT 8 +#define CLIENT_CG_REQ(x) ((x) << 16) +#define CLIENT_CG_REQ_MASK (0xff << 16) +#define CLIENT_CG_REQ_SHIFT 16 +#define CLIENT_CG_RESP(x) ((x) << 24) +#define CLIENT_CG_RESP_MASK (0xff << 24) +#define CLIENT_CG_RESP_SHIFT 24 + +#define CG_SPLL_SPREAD_SPECTRUM 0x790 +#define SSEN (1 << 0) +#define CG_SPLL_SPREAD_SPECTRUM_2 0x794 + +#define MPLL_SS1 0x85c +#define CLKV(x) ((x) << 0) +#define CLKV_MASK (0x3ffffff << 0) +#define MPLL_SS2 0x860 +#define CLKS(x) ((x) << 0) +#define CLKS_MASK (0xfff << 0) + +#define CG_IND_ADDR 0x8f8 +#define CG_IND_DATA 0x8fc +/* CGIND regs */ +#define CG_CGTT_LOCAL_0 0x00 +#define CG_CGTT_LOCAL_1 0x01 +#define CG_CGTT_LOCAL_2 0x02 +#define CG_CGTT_LOCAL_3 0x03 +#define CG_CGLS_TILE_0 0x20 +#define CG_CGLS_TILE_1 0x21 +#define CG_CGLS_TILE_2 0x22 +#define CG_CGLS_TILE_3 0x23 +#define CG_CGLS_TILE_4 0x24 +#define CG_CGLS_TILE_5 0x25 +#define CG_CGLS_TILE_6 0x26 +#define CG_CGLS_TILE_7 0x27 +#define CG_CGLS_TILE_8 0x28 +#define CG_CGLS_TILE_9 0x29 +#define CG_CGLS_TILE_10 0x2a +#define CG_CGLS_TILE_11 0x2b + +#define VM_L2_CG 0x15c0 + +#define MC_CONFIG 0x2000 + +#define MC_CONFIG_MCD 0x20a0 +#define MC_CG_CONFIG_MCD 0x20a4 +#define MC_RD_ENABLE_MCD(x) ((x) << 8) +#define MC_RD_ENABLE_MCD_MASK (7 << 8) + +#define MC_HUB_MISC_HUB_CG 0x20b8 +#define MC_HUB_MISC_VM_CG 0x20bc +#define MC_HUB_MISC_SIP_CG 0x20c0 + +#define MC_XPB_CLK_GAT 0x2478 + +#define MC_CG_CONFIG 0x25bc +#define MC_RD_ENABLE(x) ((x) << 4) +#define MC_RD_ENABLE_MASK (3 << 4) + +#define MC_CITF_MISC_RD_CG 0x2648 +#define MC_CITF_MISC_WR_CG 0x264c +#define MC_CITF_MISC_VM_CG 0x2650 +# define MEM_LS_ENABLE (1 << 19) + +#define MC_ARB_BURST_TIME 0x2808 +#define STATE0(x) ((x) << 0) +#define STATE0_MASK (0x1f << 0) +#define STATE1(x) ((x) << 5) +#define STATE1_MASK (0x1f << 5) +#define STATE2(x) ((x) << 10) +#define STATE2_MASK (0x1f << 10) +#define STATE3(x) ((x) << 15) +#define STATE3_MASK (0x1f << 15) + +#define MC_SEQ_RAS_TIMING 0x28a0 +#define MC_SEQ_CAS_TIMING 0x28a4 +#define MC_SEQ_MISC_TIMING 0x28a8 +#define MC_SEQ_MISC_TIMING2 0x28ac + +#define MC_SEQ_RD_CTL_D0 0x28b4 +#define MC_SEQ_RD_CTL_D1 0x28b8 +#define MC_SEQ_WR_CTL_D0 0x28bc +#define MC_SEQ_WR_CTL_D1 0x28c0 + +#define MC_SEQ_STATUS_M 0x29f4 +# define PMG_PWRSTATE (1 << 16) + +#define MC_SEQ_MISC1 0x2a04 +#define MC_SEQ_RESERVE_M 0x2a08 +#define MC_PMG_CMD_EMRS 0x2a0c + +#define MC_SEQ_MISC3 0x2a2c + +#define MC_SEQ_MISC5 0x2a54 +#define MC_SEQ_MISC6 0x2a58 + +#define MC_SEQ_MISC7 0x2a64 + +#define MC_SEQ_CG 0x2a68 +#define CG_SEQ_REQ(x) ((x) << 0) +#define CG_SEQ_REQ_MASK (0xff << 0) +#define CG_SEQ_REQ_SHIFT 0 +#define CG_SEQ_RESP(x) ((x) << 8) +#define CG_SEQ_RESP_MASK (0xff << 8) +#define CG_SEQ_RESP_SHIFT 8 +#define SEQ_CG_REQ(x) ((x) << 16) +#define SEQ_CG_REQ_MASK (0xff << 16) +#define SEQ_CG_REQ_SHIFT 16 +#define SEQ_CG_RESP(x) ((x) << 24) +#define SEQ_CG_RESP_MASK (0xff << 24) +#define SEQ_CG_RESP_SHIFT 24 +#define MC_SEQ_RAS_TIMING_LP 0x2a6c +#define MC_SEQ_CAS_TIMING_LP 0x2a70 +#define MC_SEQ_MISC_TIMING_LP 0x2a74 +#define MC_SEQ_MISC_TIMING2_LP 0x2a78 +#define MC_SEQ_WR_CTL_D0_LP 0x2a7c +#define MC_SEQ_WR_CTL_D1_LP 0x2a80 +#define MC_SEQ_PMG_CMD_EMRS_LP 0x2a84 +#define MC_SEQ_PMG_CMD_MRS_LP 0x2a88 + +#define MC_PMG_CMD_MRS 0x2aac + +#define MC_SEQ_RD_CTL_D0_LP 0x2b1c +#define MC_SEQ_RD_CTL_D1_LP 0x2b20 + +#define MC_PMG_CMD_MRS1 0x2b44 +#define MC_SEQ_PMG_CMD_MRS1_LP 0x2b48 + +#define CGTS_SM_CTRL_REG 0x9150 + /* Registers */ #define RCU_IND_INDEX 0x100 @@ -93,6 +377,34 @@ #define CG_VCLK_STATUS 0x61c #define CG_SCRATCH1 0x820 +#define RLC_CNTL 0x3f00 +# define RLC_ENABLE (1 << 0) +# define GFX_POWER_GATING_ENABLE (1 << 7) +# define GFX_POWER_GATING_SRC (1 << 8) +# define DYN_PER_SIMD_PG_ENABLE (1 << 27) +# define LB_CNT_SPIM_ACTIVE (1 << 30) +# define LOAD_BALANCE_ENABLE (1 << 31) + +#define RLC_HB_BASE 0x3f10 +#define RLC_HB_CNTL 0x3f0c +#define RLC_HB_RPTR 0x3f20 +#define RLC_HB_WPTR 0x3f1c +#define RLC_HB_WPTR_LSB_ADDR 0x3f14 +#define RLC_HB_WPTR_MSB_ADDR 0x3f18 +#define RLC_MC_CNTL 0x3f44 +#define RLC_UCODE_CNTL 0x3f48 +#define RLC_UCODE_ADDR 0x3f2c +#define RLC_UCODE_DATA 0x3f30 + +/* new for TN */ +#define TN_RLC_SAVE_AND_RESTORE_BASE 0x3f10 +#define TN_RLC_LB_CNTR_MAX 0x3f14 +#define TN_RLC_LB_CNTR_INIT 0x3f18 +#define TN_RLC_CLEAR_STATE_RESTORE_BASE 0x3f20 +#define TN_RLC_LB_INIT_SIMD_MASK 0x3fe4 +#define TN_RLC_LB_ALWAYS_ACTIVE_SIMD_MASK 0x3fe8 +#define TN_RLC_LB_PARAMS 0x3fec + #define GRBM_GFX_INDEX 0x802C #define INSTANCE_INDEX(x) ((x) << 0) #define SE_INDEX(x) ((x) << 16) @@ -185,6 +497,9 @@ #define DCCG_AUDIO_DTO0_MODULE 0x05b4 #define DCCG_AUDIO_DTO0_LOAD 0x05b8 #define DCCG_AUDIO_DTO0_CNTL 0x05bc +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO(x) (((x) & 7) << 0) +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO_MASK 7 +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO_SHIFT 0 #define DCCG_AUDIO_DTO1_PHASE 0x05c0 #define DCCG_AUDIO_DTO1_MODULE 0x05c4 @@ -506,6 +821,30 @@ #define CG_THERMAL_CTRL 0x72c #define TOFFSET_MASK 0x00003FE0 #define TOFFSET_SHIFT 5 +#define DIG_THERM_DPM(x) ((x) << 14) +#define DIG_THERM_DPM_MASK 0x003FC000 +#define DIG_THERM_DPM_SHIFT 14 + +#define CG_THERMAL_INT 0x734 +#define DIG_THERM_INTH(x) ((x) << 8) +#define DIG_THERM_INTH_MASK 0x0000FF00 +#define DIG_THERM_INTH_SHIFT 8 +#define DIG_THERM_INTL(x) ((x) << 16) +#define DIG_THERM_INTL_MASK 0x00FF0000 +#define DIG_THERM_INTL_SHIFT 16 +#define THERM_INT_MASK_HIGH (1 << 24) +#define THERM_INT_MASK_LOW (1 << 25) + +#define TN_CG_THERMAL_INT_CTRL 0x738 +#define TN_DIG_THERM_INTH(x) ((x) << 0) +#define TN_DIG_THERM_INTH_MASK 0x000000FF +#define TN_DIG_THERM_INTH_SHIFT 0 +#define TN_DIG_THERM_INTL(x) ((x) << 8) +#define TN_DIG_THERM_INTL_MASK 0x0000FF00 +#define TN_DIG_THERM_INTL_SHIFT 8 +#define TN_THERM_INT_MASK_HIGH (1 << 24) +#define TN_THERM_INT_MASK_LOW (1 << 25) + #define CG_MULT_THERMAL_STATUS 0x740 #define ASIC_T(x) ((x) << 16) #define ASIC_T_MASK 0x07FF0000 @@ -513,6 +852,7 @@ #define CG_TS0_STATUS 0x760 #define TS0_ADC_DOUT_MASK 0x000003FF #define TS0_ADC_DOUT_SHIFT 0 + /* APU */ #define CG_THERMAL_STATUS 0x678 @@ -995,7 +1335,48 @@ #define DMA_PACKET_CONSTANT_FILL 0xd #define DMA_PACKET_NOP 0xf -/* PCIE link stuff */ +/* PIF PHY0 indirect regs */ +#define PB0_PIF_CNTL 0x10 +# define LS2_EXIT_TIME(x) ((x) << 17) +# define LS2_EXIT_TIME_MASK (0x7 << 17) +# define LS2_EXIT_TIME_SHIFT 17 +#define PB0_PIF_PAIRING 0x11 +# define MULTI_PIF (1 << 25) +#define PB0_PIF_PWRDOWN_0 0x12 +# define PLL_POWER_STATE_IN_TXS2_0(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_0_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_0_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_0(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_0_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_0_SHIFT 10 +# define PLL_RAMP_UP_TIME_0(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_0_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_0_SHIFT 24 +#define PB0_PIF_PWRDOWN_1 0x13 +# define PLL_POWER_STATE_IN_TXS2_1(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_1_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_1_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_1(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_1_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_1_SHIFT 10 +# define PLL_RAMP_UP_TIME_1(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_1_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_1_SHIFT 24 +/* PIF PHY1 indirect regs */ +#define PB1_PIF_CNTL 0x10 +#define PB1_PIF_PAIRING 0x11 +#define PB1_PIF_PWRDOWN_0 0x12 +#define PB1_PIF_PWRDOWN_1 0x13 +/* PCIE PORT indirect regs */ +#define PCIE_LC_CNTL 0xa0 +# define LC_L0S_INACTIVITY(x) ((x) << 8) +# define LC_L0S_INACTIVITY_MASK (0xf << 8) +# define LC_L0S_INACTIVITY_SHIFT 8 +# define LC_L1_INACTIVITY(x) ((x) << 12) +# define LC_L1_INACTIVITY_MASK (0xf << 12) +# define LC_L1_INACTIVITY_SHIFT 12 +# define LC_PMI_TO_L1_DIS (1 << 16) +# define LC_ASPM_TO_L1_DIS (1 << 24) #define PCIE_LC_TRAINING_CNTL 0xa1 /* PCIE_P */ #define PCIE_LC_LINK_WIDTH_CNTL 0xa2 /* PCIE_P */ # define LC_LINK_WIDTH_SHIFT 0 @@ -1015,6 +1396,9 @@ # define LC_SHORT_RECONFIG_EN (1 << 11) # define LC_UPCONFIGURE_SUPPORT (1 << 12) # define LC_UPCONFIGURE_DIS (1 << 13) +# define LC_DYN_LANES_PWR_STATE(x) ((x) << 21) +# define LC_DYN_LANES_PWR_STATE_MASK (0x3 << 21) +# define LC_DYN_LANES_PWR_STATE_SHIFT 21 #define PCIE_LC_SPEED_CNTL 0xa4 /* PCIE_P */ # define LC_GEN2_EN_STRAP (1 << 0) # define LC_TARGET_LINK_SPEED_OVERRIDE_EN (1 << 1) @@ -1023,6 +1407,9 @@ # define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK (0x3 << 8) # define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT 3 # define LC_CURRENT_DATA_RATE (1 << 11) +# define LC_HW_VOLTAGE_IF_CONTROL(x) ((x) << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_MASK (3 << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_SHIFT 12 # define LC_VOLTAGE_TIMER_SEL_MASK (0xf << 14) # define LC_CLR_FAILED_SPD_CHANGE_CNT (1 << 21) # define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 23) diff --git a/sys/dev/drm/radeon/ni.c b/sys/dev/drm/radeon/ni.c index 27272ceb40..f08ca2fa9f 100644 --- a/sys/dev/drm/radeon/ni.c +++ b/sys/dev/drm/radeon/ni.c @@ -24,6 +24,7 @@ */ #include +#include #include "radeon.h" #include "radeon_asic.h" #include @@ -31,23 +32,162 @@ #include "atom.h" #include "ni_reg.h" #include "cayman_blit_shaders.h" +#include "radeon_ucode.h" +#include "clearstate_cayman.h" + +static u32 tn_rlc_save_restore_register_list[] = +{ + 0x98fc, + 0x98f0, + 0x9834, + 0x9838, + 0x9870, + 0x9874, + 0x8a14, + 0x8b24, + 0x8bcc, + 0x8b10, + 0x8c30, + 0x8d00, + 0x8d04, + 0x8c00, + 0x8c04, + 0x8c10, + 0x8c14, + 0x8d8c, + 0x8cf0, + 0x8e38, + 0x9508, + 0x9688, + 0x9608, + 0x960c, + 0x9610, + 0x9614, + 0x88c4, + 0x8978, + 0x88d4, + 0x900c, + 0x9100, + 0x913c, + 0x90e8, + 0x9354, + 0xa008, + 0x98f8, + 0x9148, + 0x914c, + 0x3f94, + 0x98f4, + 0x9b7c, + 0x3f8c, + 0x8950, + 0x8954, + 0x8a18, + 0x8b28, + 0x9144, + 0x3f90, + 0x915c, + 0x9160, + 0x9178, + 0x917c, + 0x9180, + 0x918c, + 0x9190, + 0x9194, + 0x9198, + 0x919c, + 0x91a8, + 0x91ac, + 0x91b0, + 0x91b4, + 0x91b8, + 0x91c4, + 0x91c8, + 0x91cc, + 0x91d0, + 0x91d4, + 0x91e0, + 0x91e4, + 0x91ec, + 0x91f0, + 0x91f4, + 0x9200, + 0x9204, + 0x929c, + 0x8030, + 0x9150, + 0x9a60, + 0x920c, + 0x9210, + 0x9228, + 0x922c, + 0x9244, + 0x9248, + 0x91e8, + 0x9294, + 0x9208, + 0x9224, + 0x9240, + 0x9220, + 0x923c, + 0x9258, + 0x9744, + 0xa200, + 0xa204, + 0xa208, + 0xa20c, + 0x8d58, + 0x9030, + 0x9034, + 0x9038, + 0x903c, + 0x9040, + 0x9654, + 0x897c, + 0xa210, + 0xa214, + 0x9868, + 0xa02c, + 0x9664, + 0x9698, + 0x949c, + 0x8e10, + 0x8e18, + 0x8c50, + 0x8c58, + 0x8c60, + 0x8c68, + 0x89b4, + 0x9830, + 0x802c, +}; +static u32 tn_rlc_save_restore_register_list_size = ARRAY_SIZE(tn_rlc_save_restore_register_list); -extern bool evergreen_is_display_hung(struct radeon_device *rdev); extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev); extern void evergreen_pcie_gen2_enable(struct radeon_device *rdev); - -#define EVERGREEN_PFP_UCODE_SIZE 1120 -#define EVERGREEN_PM4_UCODE_SIZE 1376 -#define EVERGREEN_RLC_UCODE_SIZE 768 -#define BTC_MC_UCODE_SIZE 6024 - -#define CAYMAN_PFP_UCODE_SIZE 2176 -#define CAYMAN_PM4_UCODE_SIZE 2176 -#define CAYMAN_RLC_UCODE_SIZE 1024 -#define CAYMAN_MC_UCODE_SIZE 6037 - -#define ARUBA_RLC_UCODE_SIZE 1536 - +extern void evergreen_program_aspm(struct radeon_device *rdev); + +/* Firmware Names */ +MODULE_FIRMWARE("radeon/BARTS_pfp.bin"); +MODULE_FIRMWARE("radeon/BARTS_me.bin"); +MODULE_FIRMWARE("radeon/BARTS_mc.bin"); +MODULE_FIRMWARE("radeon/BARTS_smc.bin"); +MODULE_FIRMWARE("radeon/BTC_rlc.bin"); +MODULE_FIRMWARE("radeon/TURKS_pfp.bin"); +MODULE_FIRMWARE("radeon/TURKS_me.bin"); +MODULE_FIRMWARE("radeon/TURKS_mc.bin"); +MODULE_FIRMWARE("radeon/TURKS_smc.bin"); +MODULE_FIRMWARE("radeon/CAICOS_pfp.bin"); +MODULE_FIRMWARE("radeon/CAICOS_me.bin"); +MODULE_FIRMWARE("radeon/CAICOS_mc.bin"); +MODULE_FIRMWARE("radeon/CAICOS_smc.bin"); +MODULE_FIRMWARE("radeon/CAYMAN_pfp.bin"); +MODULE_FIRMWARE("radeon/CAYMAN_me.bin"); +MODULE_FIRMWARE("radeon/CAYMAN_mc.bin"); +MODULE_FIRMWARE("radeon/CAYMAN_rlc.bin"); +MODULE_FIRMWARE("radeon/CAYMAN_smc.bin"); +MODULE_FIRMWARE("radeon/ARUBA_pfp.bin"); +MODULE_FIRMWARE("radeon/ARUBA_me.bin"); +MODULE_FIRMWARE("radeon/ARUBA_rlc.bin"); static const u32 cayman_golden_registers2[] = { @@ -535,6 +675,7 @@ int ni_init_microcode(struct radeon_device *rdev) const char *chip_name; const char *rlc_chip_name; size_t pfp_req_size, me_req_size, rlc_req_size, mc_req_size; + size_t smc_req_size = 0; char fw_name[30]; int err; @@ -548,6 +689,7 @@ int ni_init_microcode(struct radeon_device *rdev) me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4; rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4; mc_req_size = BTC_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(BARTS_SMC_UCODE_SIZE, 4); break; case CHIP_TURKS: chip_name = "TURKS"; @@ -556,6 +698,7 @@ int ni_init_microcode(struct radeon_device *rdev) me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4; rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4; mc_req_size = BTC_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(TURKS_SMC_UCODE_SIZE, 4); break; case CHIP_CAICOS: chip_name = "CAICOS"; @@ -564,6 +707,7 @@ int ni_init_microcode(struct radeon_device *rdev) me_req_size = EVERGREEN_PM4_UCODE_SIZE * 4; rlc_req_size = EVERGREEN_RLC_UCODE_SIZE * 4; mc_req_size = BTC_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(CAICOS_SMC_UCODE_SIZE, 4); break; case CHIP_CAYMAN: chip_name = "CAYMAN"; @@ -572,6 +716,7 @@ int ni_init_microcode(struct radeon_device *rdev) me_req_size = CAYMAN_PM4_UCODE_SIZE * 4; rlc_req_size = CAYMAN_RLC_UCODE_SIZE * 4; mc_req_size = CAYMAN_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(CAYMAN_SMC_UCODE_SIZE, 4); break; case CHIP_ARUBA: chip_name = "ARUBA"; @@ -596,7 +741,7 @@ int ni_init_microcode(struct radeon_device *rdev) } if (rdev->pfp_fw->datasize != pfp_req_size) { DRM_ERROR( - "ni_cp: Bogus length %zu in firmware \"%s\"\n", + "ni_pfp: Bogus length %zu in firmware \"%s\"\n", rdev->pfp_fw->datasize, fw_name); err = -EINVAL; goto out; @@ -610,7 +755,7 @@ int ni_init_microcode(struct radeon_device *rdev) } if (rdev->me_fw->datasize != me_req_size) { DRM_ERROR( - "ni_cp: Bogus length %zu in firmware \"%s\"\n", + "ni_me: Bogus length %zu in firmware \"%s\"\n", rdev->me_fw->datasize, fw_name); err = -EINVAL; } @@ -645,6 +790,24 @@ int ni_init_microcode(struct radeon_device *rdev) err = -EINVAL; } } + + if ((rdev->family >= CHIP_BARTS) && (rdev->family <= CHIP_CAYMAN)) { + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_smc", chip_name); + err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev); + if (err) { + printk(KERN_ERR + "smc: error loading firmware \"%s\"\n", + fw_name); + release_firmware(rdev->smc_fw); + rdev->smc_fw = NULL; + } else if (rdev->smc_fw->datasize != smc_req_size) { + printk(KERN_ERR + "ni_smc: Bogus length %zu in firmware \"%s\"\n", + rdev->smc_fw->datasize, fw_name); + err = -EINVAL; + } + } + out: if (err) { if (err != -EINVAL) @@ -667,6 +830,10 @@ out: firmware_put(rdev->mc_fw, FIRMWARE_UNLOAD); rdev->mc_fw = NULL; } + if (rdev->smc_fw != NULL) { + firmware_put(rdev->smc_fw, FIRMWARE_UNLOAD); + rdev->smc_fw = NULL; + } } return err; } @@ -704,6 +871,14 @@ void ni_fini_microcode(struct radeon_device *rdev) } +int tn_get_temp(struct radeon_device *rdev) +{ + u32 temp = RREG32_SMC(TN_CURRENT_GNB_TEMP) & 0x7ff; + int actual_temp = (temp / 8) - 49; + + return actual_temp * 1000; +} + /* * Core functions */ @@ -1039,6 +1214,16 @@ static void cayman_gpu_init(struct radeon_device *rdev) WREG32(PA_CL_ENHANCE, CLIP_VTX_REORDER_ENA | NUM_CLIP_SEQ(3)); DRM_UDELAY(50); + + /* set clockgating golden values on TN */ + if (rdev->family == CHIP_ARUBA) { + tmp = RREG32_CG(CG_CGTT_LOCAL_0); + tmp &= ~0x00380000; + WREG32_CG(CG_CGTT_LOCAL_0, tmp); + tmp = RREG32_CG(CG_CGTT_LOCAL_1); + tmp &= ~0x0e000000; + WREG32_CG(CG_CGTT_LOCAL_1, tmp); + } } /* @@ -1940,6 +2125,10 @@ static int cayman_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ evergreen_pcie_gen2_enable(rdev); + /* enable aspm */ + evergreen_program_aspm(rdev); + + evergreen_mc_program(rdev); if (rdev->flags & RADEON_IS_IGP) { if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { @@ -1969,7 +2158,6 @@ static int cayman_startup(struct radeon_device *rdev) if (r) return r; - evergreen_mc_program(rdev); r = cayman_pcie_gart_enable(rdev); if (r) return r; @@ -1984,7 +2172,10 @@ static int cayman_startup(struct radeon_device *rdev) /* allocate rlc buffers */ if (rdev->flags & RADEON_IS_IGP) { - r = si_rlc_init(rdev); + rdev->rlc.reg_list = tn_rlc_save_restore_register_list; + rdev->rlc.reg_list_size = tn_rlc_save_restore_register_list_size; + rdev->rlc.cs_data = cayman_cs_data; + r = sumo_rlc_init(rdev); if (r) { DRM_ERROR("Failed to init rlc BOs!\n"); return r; @@ -2145,7 +2336,7 @@ int cayman_suspend(struct radeon_device *rdev) radeon_vm_manager_fini(rdev); cayman_cp_enable(rdev, false); cayman_dma_stop(rdev); - r600_uvd_rbc_stop(rdev); + r600_uvd_stop(rdev); radeon_uvd_suspend(rdev); evergreen_irq_suspend(rdev); radeon_wb_disable(rdev); @@ -2241,7 +2432,7 @@ int cayman_init(struct radeon_device *rdev) cayman_dma_fini(rdev); r600_irq_fini(rdev); if (rdev->flags & RADEON_IS_IGP) - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); radeon_wb_fini(rdev); radeon_ib_pool_fini(rdev); radeon_vm_manager_fini(rdev); @@ -2272,11 +2463,12 @@ void cayman_fini(struct radeon_device *rdev) cayman_dma_fini(rdev); r600_irq_fini(rdev); if (rdev->flags & RADEON_IS_IGP) - si_rlc_fini(rdev); + sumo_rlc_fini(rdev); radeon_wb_fini(rdev); radeon_vm_manager_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_fini(rdev); cayman_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); @@ -2310,6 +2502,167 @@ void cayman_vm_fini(struct radeon_device *rdev) { } +/** + * cayman_vm_decode_fault - print human readable fault info + * + * @rdev: radeon_device pointer + * @status: VM_CONTEXT1_PROTECTION_FAULT_STATUS register value + * @addr: VM_CONTEXT1_PROTECTION_FAULT_ADDR register value + * + * Print human readable fault information (cayman/TN). + */ +void cayman_vm_decode_fault(struct radeon_device *rdev, + u32 status, u32 addr) +{ + u32 mc_id = (status & MEMORY_CLIENT_ID_MASK) >> MEMORY_CLIENT_ID_SHIFT; + u32 vmid = (status & FAULT_VMID_MASK) >> FAULT_VMID_SHIFT; + u32 protections = (status & PROTECTIONS_MASK) >> PROTECTIONS_SHIFT; + char *block; + + switch (mc_id) { + case 32: + case 16: + case 96: + case 80: + case 160: + case 144: + case 224: + case 208: + block = "CB"; + break; + case 33: + case 17: + case 97: + case 81: + case 161: + case 145: + case 225: + case 209: + block = "CB_FMASK"; + break; + case 34: + case 18: + case 98: + case 82: + case 162: + case 146: + case 226: + case 210: + block = "CB_CMASK"; + break; + case 35: + case 19: + case 99: + case 83: + case 163: + case 147: + case 227: + case 211: + block = "CB_IMMED"; + break; + case 36: + case 20: + case 100: + case 84: + case 164: + case 148: + case 228: + case 212: + block = "DB"; + break; + case 37: + case 21: + case 101: + case 85: + case 165: + case 149: + case 229: + case 213: + block = "DB_HTILE"; + break; + case 38: + case 22: + case 102: + case 86: + case 166: + case 150: + case 230: + case 214: + block = "SX"; + break; + case 39: + case 23: + case 103: + case 87: + case 167: + case 151: + case 231: + case 215: + block = "DB_STEN"; + break; + case 40: + case 24: + case 104: + case 88: + case 232: + case 216: + case 168: + case 152: + block = "TC_TFETCH"; + break; + case 41: + case 25: + case 105: + case 89: + case 233: + case 217: + case 169: + case 153: + block = "TC_VFETCH"; + break; + case 42: + case 26: + case 106: + case 90: + case 234: + case 218: + case 170: + case 154: + block = "VC"; + break; + case 112: + block = "CP"; + break; + case 113: + case 114: + block = "SH"; + break; + case 115: + block = "VGT"; + break; + case 178: + block = "IH"; + break; + case 51: + block = "RLC"; + break; + case 55: + block = "DMA"; + break; + case 56: + block = "HDP"; + break; + default: + block = "unknown"; + break; + } + + printk("VM fault (0x%02x, vmid %d) at page %u, %s from %s (%d)\n", + protections, vmid, addr, + (status & MEMORY_CLIENT_RW_MASK) ? "write" : "read", + block, mc_id); +} + #define R600_ENTRY_VALID (1 << 0) #define R600_PTE_SYSTEM (1 << 1) #define R600_PTE_SNOOPED (1 << 2) diff --git a/sys/dev/drm/radeon/ni_dpm.c b/sys/dev/drm/radeon/ni_dpm.c new file mode 100644 index 0000000000..51e5839db1 --- /dev/null +++ b/sys/dev/drm/radeon/ni_dpm.c @@ -0,0 +1,4355 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "nid.h" +#include "r600_dpm.h" +#include "ni_dpm.h" +#include "atom.h" +#include +#include + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define SMC_RAM_END 0xC000 + +static const struct ni_cac_weights cac_weights_cayman_xt = +{ + 0x15, + 0x2, + 0x19, + 0x2, + 0x8, + 0x14, + 0x2, + 0x16, + 0xE, + 0x17, + 0x13, + 0x2B, + 0x10, + 0x7, + 0x5, + 0x5, + 0x5, + 0x2, + 0x3, + 0x9, + 0x10, + 0x10, + 0x2B, + 0xA, + 0x9, + 0x4, + 0xD, + 0xD, + 0x3E, + 0x18, + 0x14, + 0, + 0x3, + 0x3, + 0x5, + 0, + 0x2, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0x1CC, + 0, + 0x164, + 1, + 1, + 1, + 1, + 12, + 12, + 12, + 0x12, + 0x1F, + 132, + 5, + 7, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0 }, + true +}; + +static const struct ni_cac_weights cac_weights_cayman_pro = +{ + 0x16, + 0x4, + 0x10, + 0x2, + 0xA, + 0x16, + 0x2, + 0x18, + 0x10, + 0x1A, + 0x16, + 0x2D, + 0x12, + 0xA, + 0x6, + 0x6, + 0x6, + 0x2, + 0x4, + 0xB, + 0x11, + 0x11, + 0x2D, + 0xC, + 0xC, + 0x7, + 0x10, + 0x10, + 0x3F, + 0x1A, + 0x16, + 0, + 0x7, + 0x4, + 0x6, + 1, + 0x2, + 0x1, + 0, + 0, + 0, + 0, + 0, + 0, + 0x30, + 0, + 0x1CF, + 0, + 0x166, + 1, + 1, + 1, + 1, + 12, + 12, + 12, + 0x15, + 0x1F, + 132, + 6, + 6, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0 }, + true +}; + +static const struct ni_cac_weights cac_weights_cayman_le = +{ + 0x7, + 0xE, + 0x1, + 0xA, + 0x1, + 0x3F, + 0x2, + 0x18, + 0x10, + 0x1A, + 0x1, + 0x3F, + 0x1, + 0xE, + 0x6, + 0x6, + 0x6, + 0x2, + 0x4, + 0x9, + 0x1A, + 0x1A, + 0x2C, + 0xA, + 0x11, + 0x8, + 0x19, + 0x19, + 0x1, + 0x1, + 0x1A, + 0, + 0x8, + 0x5, + 0x8, + 0x1, + 0x3, + 0x1, + 0, + 0, + 0, + 0, + 0, + 0, + 0x38, + 0x38, + 0x239, + 0x3, + 0x18A, + 1, + 1, + 1, + 1, + 12, + 12, + 12, + 0x15, + 0x22, + 132, + 6, + 6, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0 }, + true +}; + +#define NISLANDS_MGCG_SEQUENCE 300 + +static const u32 cayman_cgcg_cgls_default[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define CAYMAN_CGCG_CGLS_DEFAULT_LENGTH sizeof(cayman_cgcg_cgls_default) / (3 * sizeof(u32)) + +static const u32 cayman_cgcg_cgls_disable[] = +{ + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00000644, 0x000f7902, 0x001f4180, + 0x00000644, 0x000f3802, 0x001f4180 +}; +#define CAYMAN_CGCG_CGLS_DISABLE_LENGTH sizeof(cayman_cgcg_cgls_disable) / (3 * sizeof(u32)) + +static const u32 cayman_cgcg_cgls_enable[] = +{ + 0x00000644, 0x000f7882, 0x001f4080, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000020, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000021, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000022, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000023, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000024, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000025, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000026, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000027, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000028, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000029, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002a, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x0000002b, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff +}; +#define CAYMAN_CGCG_CGLS_ENABLE_LENGTH sizeof(cayman_cgcg_cgls_enable) / (3 * sizeof(u32)) + +static const u32 cayman_mgcg_default[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00003fc4, 0xc0000000, 0xffffffff, + 0x00005448, 0x00000100, 0xffffffff, + 0x000055e4, 0x00000100, 0xffffffff, + 0x0000160c, 0x00000100, 0xffffffff, + 0x00008984, 0x06000100, 0xffffffff, + 0x0000c164, 0x00000100, 0xffffffff, + 0x00008a18, 0x00000100, 0xffffffff, + 0x0000897c, 0x06000100, 0xffffffff, + 0x00008b28, 0x00000100, 0xffffffff, + 0x00009144, 0x00800200, 0xffffffff, + 0x00009a60, 0x00000100, 0xffffffff, + 0x00009868, 0x00000100, 0xffffffff, + 0x00008d58, 0x00000100, 0xffffffff, + 0x00009510, 0x00000100, 0xffffffff, + 0x0000949c, 0x00000100, 0xffffffff, + 0x00009654, 0x00000100, 0xffffffff, + 0x00009030, 0x00000100, 0xffffffff, + 0x00009034, 0x00000100, 0xffffffff, + 0x00009038, 0x00000100, 0xffffffff, + 0x0000903c, 0x00000100, 0xffffffff, + 0x00009040, 0x00000100, 0xffffffff, + 0x0000a200, 0x00000100, 0xffffffff, + 0x0000a204, 0x00000100, 0xffffffff, + 0x0000a208, 0x00000100, 0xffffffff, + 0x0000a20c, 0x00000100, 0xffffffff, + 0x00009744, 0x00000100, 0xffffffff, + 0x00003f80, 0x00000100, 0xffffffff, + 0x0000a210, 0x00000100, 0xffffffff, + 0x0000a214, 0x00000100, 0xffffffff, + 0x000004d8, 0x00000100, 0xffffffff, + 0x00009664, 0x00000100, 0xffffffff, + 0x00009698, 0x00000100, 0xffffffff, + 0x000004d4, 0x00000200, 0xffffffff, + 0x000004d0, 0x00000000, 0xffffffff, + 0x000030cc, 0x00000104, 0xffffffff, + 0x0000d0c0, 0x00000100, 0xffffffff, + 0x0000d8c0, 0x00000100, 0xffffffff, + 0x0000802c, 0x40000000, 0xffffffff, + 0x00003fc4, 0x40000000, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009224, 0x00070000, 0xffffffff, + 0x00009228, 0x00030002, 0xffffffff, + 0x0000922c, 0x00050004, 0xffffffff, + 0x00009238, 0x00010006, 0xffffffff, + 0x0000923c, 0x00090008, 0xffffffff, + 0x00009240, 0x00070000, 0xffffffff, + 0x00009244, 0x00030002, 0xffffffff, + 0x00009248, 0x00050004, 0xffffffff, + 0x00009254, 0x00010006, 0xffffffff, + 0x00009258, 0x00090008, 0xffffffff, + 0x0000925c, 0x00070000, 0xffffffff, + 0x00009260, 0x00030002, 0xffffffff, + 0x00009264, 0x00050004, 0xffffffff, + 0x00009270, 0x00010006, 0xffffffff, + 0x00009274, 0x00090008, 0xffffffff, + 0x00009278, 0x00070000, 0xffffffff, + 0x0000927c, 0x00030002, 0xffffffff, + 0x00009280, 0x00050004, 0xffffffff, + 0x0000928c, 0x00010006, 0xffffffff, + 0x00009290, 0x00090008, 0xffffffff, + 0x000092a8, 0x00070000, 0xffffffff, + 0x000092ac, 0x00030002, 0xffffffff, + 0x000092b0, 0x00050004, 0xffffffff, + 0x000092bc, 0x00010006, 0xffffffff, + 0x000092c0, 0x00090008, 0xffffffff, + 0x000092c4, 0x00070000, 0xffffffff, + 0x000092c8, 0x00030002, 0xffffffff, + 0x000092cc, 0x00050004, 0xffffffff, + 0x000092d8, 0x00010006, 0xffffffff, + 0x000092dc, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x0000802c, 0x40010000, 0xffffffff, + 0x00003fc4, 0x40010000, 0xffffffff, + 0x0000915c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009224, 0x00070000, 0xffffffff, + 0x00009228, 0x00030002, 0xffffffff, + 0x0000922c, 0x00050004, 0xffffffff, + 0x00009238, 0x00010006, 0xffffffff, + 0x0000923c, 0x00090008, 0xffffffff, + 0x00009240, 0x00070000, 0xffffffff, + 0x00009244, 0x00030002, 0xffffffff, + 0x00009248, 0x00050004, 0xffffffff, + 0x00009254, 0x00010006, 0xffffffff, + 0x00009258, 0x00090008, 0xffffffff, + 0x0000925c, 0x00070000, 0xffffffff, + 0x00009260, 0x00030002, 0xffffffff, + 0x00009264, 0x00050004, 0xffffffff, + 0x00009270, 0x00010006, 0xffffffff, + 0x00009274, 0x00090008, 0xffffffff, + 0x00009278, 0x00070000, 0xffffffff, + 0x0000927c, 0x00030002, 0xffffffff, + 0x00009280, 0x00050004, 0xffffffff, + 0x0000928c, 0x00010006, 0xffffffff, + 0x00009290, 0x00090008, 0xffffffff, + 0x000092a8, 0x00070000, 0xffffffff, + 0x000092ac, 0x00030002, 0xffffffff, + 0x000092b0, 0x00050004, 0xffffffff, + 0x000092bc, 0x00010006, 0xffffffff, + 0x000092c0, 0x00090008, 0xffffffff, + 0x000092c4, 0x00070000, 0xffffffff, + 0x000092c8, 0x00030002, 0xffffffff, + 0x000092cc, 0x00050004, 0xffffffff, + 0x000092d8, 0x00010006, 0xffffffff, + 0x000092dc, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff, + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00003fc4, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000010, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000011, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000012, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000013, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000014, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000015, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000016, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000017, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000018, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000019, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001a, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x0000001b, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff +}; +#define CAYMAN_MGCG_DEFAULT_LENGTH sizeof(cayman_mgcg_default) / (3 * sizeof(u32)) + +static const u32 cayman_mgcg_disable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0xffffffff, 0xffffffff, + 0x00009150, 0x00600000, 0xffffffff +}; +#define CAYMAN_MGCG_DISABLE_LENGTH sizeof(cayman_mgcg_disable) / (3 * sizeof(u32)) + +static const u32 cayman_mgcg_enable[] = +{ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x000008f8, 0x00000002, 0xffffffff, + 0x000008fc, 0x00600000, 0xffffffff, + 0x000008f8, 0x00000003, 0xffffffff, + 0x000008fc, 0x00000000, 0xffffffff, + 0x00009150, 0x96944200, 0xffffffff +}; + +#define CAYMAN_MGCG_ENABLE_LENGTH sizeof(cayman_mgcg_enable) / (3 * sizeof(u32)) + +#define NISLANDS_SYSLS_SEQUENCE 100 + +static const u32 cayman_sysls_default[] = +{ + /* Register, Value, Mask bits */ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x0000d8bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x00002f50, 0x00000404, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00008dfc, 0x00000000, 0xffffffff +}; +#define CAYMAN_SYSLS_DEFAULT_LENGTH sizeof(cayman_sysls_default) / (3 * sizeof(u32)) + +static const u32 cayman_sysls_disable[] = +{ + /* Register, Value, Mask bits */ + 0x0000d0c0, 0x00000000, 0xffffffff, + 0x0000d8c0, 0x00000000, 0xffffffff, + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x0000d8bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x00041401, 0xffffffff, + 0x0000264c, 0x00040400, 0xffffffff, + 0x00002648, 0x00040400, 0xffffffff, + 0x00002650, 0x00040400, 0xffffffff, + 0x000020b8, 0x00040400, 0xffffffff, + 0x000020bc, 0x00040400, 0xffffffff, + 0x000020c0, 0x00040c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680000, 0xffffffff, + 0x00002f50, 0x00000404, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x000064ec, 0x00007ffd, 0xffffffff, + 0x00000c7c, 0x0000ff00, 0xffffffff, + 0x00008dfc, 0x0000007f, 0xffffffff +}; +#define CAYMAN_SYSLS_DISABLE_LENGTH sizeof(cayman_sysls_disable) / (3 * sizeof(u32)) + +static const u32 cayman_sysls_enable[] = +{ + /* Register, Value, Mask bits */ + 0x000055e8, 0x00000001, 0xffffffff, + 0x0000d0bc, 0x00000100, 0xffffffff, + 0x0000d8bc, 0x00000100, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x00002f50, 0x00000903, 0xffffffff, + 0x000004c8, 0x00000000, 0xffffffff, + 0x000064ec, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00008dfc, 0x00000000, 0xffffffff +}; +#define CAYMAN_SYSLS_ENABLE_LENGTH sizeof(cayman_sysls_enable) / (3 * sizeof(u32)) + +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev); +struct ni_power_info *ni_get_pi(struct radeon_device *rdev); +struct ni_ps *ni_get_ps(struct radeon_ps *rps); +void ni_dpm_reset_asic(struct radeon_device *rdev); + +struct ni_power_info *ni_get_pi(struct radeon_device *rdev) +{ + struct ni_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +struct ni_ps *ni_get_ps(struct radeon_ps *rps) +{ + struct ni_ps *ps = rps->ps_priv; + + return ps; +} + +static void ni_calculate_leakage_for_v_and_t_formula(const struct ni_leakage_coeffients *coeff, + u16 v, s32 t, + u32 ileakage, + u32 *leakage) +{ + s64 kt, kv, leakage_w, i_leakage, vddc, temperature; + + i_leakage = div64_s64(drm_int2fixp(ileakage), 1000); + vddc = div64_s64(drm_int2fixp(v), 1000); + temperature = div64_s64(drm_int2fixp(t), 1000); + + kt = drm_fixp_mul(div64_s64(drm_int2fixp(coeff->at), 1000), + drm_fixp_exp(drm_fixp_mul(div64_s64(drm_int2fixp(coeff->bt), 1000), temperature))); + kv = drm_fixp_mul(div64_s64(drm_int2fixp(coeff->av), 1000), + drm_fixp_exp(drm_fixp_mul(div64_s64(drm_int2fixp(coeff->bv), 1000), vddc))); + + leakage_w = drm_fixp_mul(drm_fixp_mul(drm_fixp_mul(i_leakage, kt), kv), vddc); + + *leakage = drm_fixp2int(leakage_w * 1000); +} + +static void ni_calculate_leakage_for_v_and_t(struct radeon_device *rdev, + const struct ni_leakage_coeffients *coeff, + u16 v, + s32 t, + u32 i_leakage, + u32 *leakage) +{ + ni_calculate_leakage_for_v_and_t_formula(coeff, v, t, i_leakage, leakage); +} + +bool ni_dpm_vblank_too_short(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 vblank_time = r600_dpm_get_vblank_time(rdev); + u32 switch_limit = pi->mem_gddr5 ? 450 : 300; + + if (vblank_time < switch_limit) + return true; + else + return false; + +} + +static void ni_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct ni_ps *ps = ni_get_ps(rps); + struct radeon_clock_and_voltage_limits *max_limits; + bool disable_mclk_switching; + u32 mclk, sclk; + u16 vddc, vddci; + int i; + + if ((rdev->pm.dpm.new_active_crtc_count > 1) || + ni_dpm_vblank_too_short(rdev)) + disable_mclk_switching = true; + else + disable_mclk_switching = false; + + if (rdev->pm.dpm.ac_power) + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + else + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc; + + if (rdev->pm.dpm.ac_power == false) { + for (i = 0; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].mclk > max_limits->mclk) + ps->performance_levels[i].mclk = max_limits->mclk; + if (ps->performance_levels[i].sclk > max_limits->sclk) + ps->performance_levels[i].sclk = max_limits->sclk; + if (ps->performance_levels[i].vddc > max_limits->vddc) + ps->performance_levels[i].vddc = max_limits->vddc; + if (ps->performance_levels[i].vddci > max_limits->vddci) + ps->performance_levels[i].vddci = max_limits->vddci; + } + } + + /* XXX validate the min clocks required for display */ + + if (disable_mclk_switching) { + mclk = ps->performance_levels[ps->performance_level_count - 1].mclk; + sclk = ps->performance_levels[0].sclk; + vddc = ps->performance_levels[0].vddc; + vddci = ps->performance_levels[ps->performance_level_count - 1].vddci; + } else { + sclk = ps->performance_levels[0].sclk; + mclk = ps->performance_levels[0].mclk; + vddc = ps->performance_levels[0].vddc; + vddci = ps->performance_levels[0].vddci; + } + + /* adjusted low state */ + ps->performance_levels[0].sclk = sclk; + ps->performance_levels[0].mclk = mclk; + ps->performance_levels[0].vddc = vddc; + ps->performance_levels[0].vddci = vddci; + + btc_skip_blacklist_clocks(rdev, max_limits->sclk, max_limits->mclk, + &ps->performance_levels[0].sclk, + &ps->performance_levels[0].mclk); + + for (i = 1; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].sclk < ps->performance_levels[i - 1].sclk) + ps->performance_levels[i].sclk = ps->performance_levels[i - 1].sclk; + if (ps->performance_levels[i].vddc < ps->performance_levels[i - 1].vddc) + ps->performance_levels[i].vddc = ps->performance_levels[i - 1].vddc; + } + + if (disable_mclk_switching) { + mclk = ps->performance_levels[0].mclk; + for (i = 1; i < ps->performance_level_count; i++) { + if (mclk < ps->performance_levels[i].mclk) + mclk = ps->performance_levels[i].mclk; + } + for (i = 0; i < ps->performance_level_count; i++) { + ps->performance_levels[i].mclk = mclk; + ps->performance_levels[i].vddci = vddci; + } + } else { + for (i = 1; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].mclk < ps->performance_levels[i - 1].mclk) + ps->performance_levels[i].mclk = ps->performance_levels[i - 1].mclk; + if (ps->performance_levels[i].vddci < ps->performance_levels[i - 1].vddci) + ps->performance_levels[i].vddci = ps->performance_levels[i - 1].vddci; + } + } + + for (i = 1; i < ps->performance_level_count; i++) + btc_skip_blacklist_clocks(rdev, max_limits->sclk, max_limits->mclk, + &ps->performance_levels[i].sclk, + &ps->performance_levels[i].mclk); + + for (i = 0; i < ps->performance_level_count; i++) + btc_adjust_clock_combinations(rdev, max_limits, + &ps->performance_levels[i]); + + for (i = 0; i < ps->performance_level_count; i++) { + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + ps->performance_levels[i].sclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + ps->performance_levels[i].mclk, + max_limits->vddci, &ps->performance_levels[i].vddci); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + ps->performance_levels[i].mclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk, + rdev->clock.current_dispclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + } + + for (i = 0; i < ps->performance_level_count; i++) { + btc_apply_voltage_delta_rules(rdev, + max_limits->vddc, max_limits->vddci, + &ps->performance_levels[i].vddc, + &ps->performance_levels[i].vddci); + } + + ps->dc_compatible = true; + for (i = 0; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].vddc > rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc) + ps->dc_compatible = false; + + if (ps->performance_levels[i].vddc < rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2) + ps->performance_levels[i].flags &= ~ATOM_PPLIB_R600_FLAGS_PCIEGEN2; + } +} + +static void ni_cg_clockgating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *ps = NULL; + + ps = (const u32 *)&cayman_cgcg_cgls_default; + count = CAYMAN_CGCG_CGLS_DEFAULT_LENGTH; + + btc_program_mgcg_hw_sequence(rdev, ps, count); +} + +static void ni_gfx_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *ps = NULL; + + if (enable) { + ps = (const u32 *)&cayman_cgcg_cgls_enable; + count = CAYMAN_CGCG_CGLS_ENABLE_LENGTH; + } else { + ps = (const u32 *)&cayman_cgcg_cgls_disable; + count = CAYMAN_CGCG_CGLS_DISABLE_LENGTH; + } + + btc_program_mgcg_hw_sequence(rdev, ps, count); +} + +static void ni_mg_clockgating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *ps = NULL; + + ps = (const u32 *)&cayman_mgcg_default; + count = CAYMAN_MGCG_DEFAULT_LENGTH; + + btc_program_mgcg_hw_sequence(rdev, ps, count); +} + +static void ni_mg_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *ps = NULL; + + if (enable) { + ps = (const u32 *)&cayman_mgcg_enable; + count = CAYMAN_MGCG_ENABLE_LENGTH; + } else { + ps = (const u32 *)&cayman_mgcg_disable; + count = CAYMAN_MGCG_DISABLE_LENGTH; + } + + btc_program_mgcg_hw_sequence(rdev, ps, count); +} + +static void ni_ls_clockgating_default(struct radeon_device *rdev) +{ + u32 count; + const u32 *ps = NULL; + + ps = (const u32 *)&cayman_sysls_default; + count = CAYMAN_SYSLS_DEFAULT_LENGTH; + + btc_program_mgcg_hw_sequence(rdev, ps, count); +} + +static void ni_ls_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *ps = NULL; + + if (enable) { + ps = (const u32 *)&cayman_sysls_enable; + count = CAYMAN_SYSLS_ENABLE_LENGTH; + } else { + ps = (const u32 *)&cayman_sysls_disable; + count = CAYMAN_SYSLS_DISABLE_LENGTH; + } + + btc_program_mgcg_hw_sequence(rdev, ps, count); + +} + +static int ni_patch_single_dependency_table_based_on_leakage(struct radeon_device *rdev, + struct radeon_clock_voltage_dependency_table *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 i; + + if (table) { + for (i = 0; i < table->count; i++) { + if (0xff01 == table->entries[i].v) { + if (pi->max_vddc == 0) + return -EINVAL; + table->entries[i].v = pi->max_vddc; + } + } + } + return 0; +} + +static int ni_patch_dependency_tables_based_on_leakage(struct radeon_device *rdev) +{ + int ret = 0; + + ret = ni_patch_single_dependency_table_based_on_leakage(rdev, + &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk); + + ret = ni_patch_single_dependency_table_based_on_leakage(rdev, + &rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk); + return ret; +} + +static void ni_stop_dpm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, 0, ~GLOBAL_PWRMGT_EN); +} + +#if 0 +static int ni_notify_hw_of_power_source(struct radeon_device *rdev, + bool ac_power) +{ + if (ac_power) + return (rv770_send_msg_to_smc(rdev, PPSMC_MSG_RunningOnAC) == PPSMC_Result_OK) ? + 0 : -EINVAL; + + return 0; +} +#endif + +static PPSMC_Result ni_send_msg_to_smc_with_parameter(struct radeon_device *rdev, + PPSMC_Msg msg, u32 parameter) +{ + WREG32(SMC_SCRATCH0, parameter); + return rv770_send_msg_to_smc(rdev, msg); +} + +static int ni_restrict_performance_levels_before_switch(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_NoForcedLevel) != PPSMC_Result_OK) + return -EINVAL; + + return (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 1) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +int ni_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 1) != PPSMC_Result_OK) + return -EINVAL; + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 1) != PPSMC_Result_OK) + return -EINVAL; + } else if (level == RADEON_DPM_FORCED_LEVEL_AUTO) { + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + + if (ni_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + } + + rdev->pm.dpm.forced_level = level; + + return 0; +} + +static void ni_stop_smc(struct radeon_device *rdev) +{ + u32 tmp; + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(LB_SYNC_RESET_SEL) & LB_SYNC_RESET_SEL_MASK; + if (tmp != 1) + break; + DRM_UDELAY(1); + } + + DRM_UDELAY(100); + + r7xx_stop_smc(rdev); +} + +static int ni_process_firmware_header(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 tmp; + int ret; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_stateTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + pi->state_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_softRegisters, + &tmp, pi->sram_end); + + if (ret) + return ret; + + pi->soft_regs_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_mcRegisterTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + eg_pi->mc_reg_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_fanTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + ni_pi->fan_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_mcArbDramAutoRefreshTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + ni_pi->arb_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_cacTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + ni_pi->cac_table_start = (u16)tmp; + + ret = rv770_read_smc_sram_dword(rdev, + NISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + NISLANDS_SMC_FIRMWARE_HEADER_spllTable, + &tmp, pi->sram_end); + + if (ret) + return ret; + + ni_pi->spll_table_start = (u16)tmp; + + + return ret; +} + +static void ni_read_clock_registers(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + + ni_pi->clock_registers.cg_spll_func_cntl = RREG32(CG_SPLL_FUNC_CNTL); + ni_pi->clock_registers.cg_spll_func_cntl_2 = RREG32(CG_SPLL_FUNC_CNTL_2); + ni_pi->clock_registers.cg_spll_func_cntl_3 = RREG32(CG_SPLL_FUNC_CNTL_3); + ni_pi->clock_registers.cg_spll_func_cntl_4 = RREG32(CG_SPLL_FUNC_CNTL_4); + ni_pi->clock_registers.cg_spll_spread_spectrum = RREG32(CG_SPLL_SPREAD_SPECTRUM); + ni_pi->clock_registers.cg_spll_spread_spectrum_2 = RREG32(CG_SPLL_SPREAD_SPECTRUM_2); + ni_pi->clock_registers.mpll_ad_func_cntl = RREG32(MPLL_AD_FUNC_CNTL); + ni_pi->clock_registers.mpll_ad_func_cntl_2 = RREG32(MPLL_AD_FUNC_CNTL_2); + ni_pi->clock_registers.mpll_dq_func_cntl = RREG32(MPLL_DQ_FUNC_CNTL); + ni_pi->clock_registers.mpll_dq_func_cntl_2 = RREG32(MPLL_DQ_FUNC_CNTL_2); + ni_pi->clock_registers.mclk_pwrmgt_cntl = RREG32(MCLK_PWRMGT_CNTL); + ni_pi->clock_registers.dll_cntl = RREG32(DLL_CNTL); + ni_pi->clock_registers.mpll_ss1 = RREG32(MPLL_SS1); + ni_pi->clock_registers.mpll_ss2 = RREG32(MPLL_SS2); +} + +#if 0 +static int ni_enter_ulp_state(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (pi->gfx_clock_gating) { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_ADDR_CONFIG); + } + + WREG32_P(SMC_MSG, HOST_SMC_MSG(PPSMC_MSG_SwitchToMinimumPower), + ~HOST_SMC_MSG_MASK); + + DRM_UDELAY(25000); + + return 0; +} +#endif + +static void ni_program_response_times(struct radeon_device *rdev) +{ + u32 voltage_response_time, backbias_response_time, acpi_delay_time, vbi_time_out; + u32 vddc_dly, bb_dly, acpi_dly, vbi_dly, mclk_switch_limit; + u32 reference_clock; + + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_mvdd_chg_time, 1); + + voltage_response_time = (u32)rdev->pm.dpm.voltage_response_time; + backbias_response_time = (u32)rdev->pm.dpm.backbias_response_time; + + if (voltage_response_time == 0) + voltage_response_time = 1000; + + if (backbias_response_time == 0) + backbias_response_time = 1000; + + acpi_delay_time = 15000; + vbi_time_out = 100000; + + reference_clock = radeon_get_xclk(rdev); + + vddc_dly = (voltage_response_time * reference_clock) / 1600; + bb_dly = (backbias_response_time * reference_clock) / 1600; + acpi_dly = (acpi_delay_time * reference_clock) / 1600; + vbi_dly = (vbi_time_out * reference_clock) / 1600; + + mclk_switch_limit = (460 * reference_clock) / 100; + + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_delay_vreg, vddc_dly); + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_delay_bbias, bb_dly); + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_delay_acpi, acpi_dly); + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_mclk_chg_timeout, vbi_dly); + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_mc_block_delay, 0xAA); + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_mclk_switch_lim, mclk_switch_limit); +} + +static void ni_populate_smc_voltage_table(struct radeon_device *rdev, + struct atom_voltage_table *voltage_table, + NISLANDS_SMC_STATETABLE *table) +{ + unsigned int i; + + for (i = 0; i < voltage_table->count; i++) { + table->highSMIO[i] = 0; + table->lowSMIO[i] |= cpu_to_be32(voltage_table->entries[i].smio_low); + } +} + +static void ni_populate_smc_voltage_tables(struct radeon_device *rdev, + NISLANDS_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + unsigned char i; + + if (eg_pi->vddc_voltage_table.count) { + ni_populate_smc_voltage_table(rdev, &eg_pi->vddc_voltage_table, table); + table->voltageMaskTable.highMask[NISLANDS_SMC_VOLTAGEMASK_VDDC] = 0; + table->voltageMaskTable.lowMask[NISLANDS_SMC_VOLTAGEMASK_VDDC] = + cpu_to_be32(eg_pi->vddc_voltage_table.mask_low); + + for (i = 0; i < eg_pi->vddc_voltage_table.count; i++) { + if (pi->max_vddc_in_table <= eg_pi->vddc_voltage_table.entries[i].value) { + table->maxVDDCIndexInPPTable = i; + break; + } + } + } + + if (eg_pi->vddci_voltage_table.count) { + ni_populate_smc_voltage_table(rdev, &eg_pi->vddci_voltage_table, table); + + table->voltageMaskTable.highMask[NISLANDS_SMC_VOLTAGEMASK_VDDCI] = 0; + table->voltageMaskTable.lowMask[NISLANDS_SMC_VOLTAGEMASK_VDDCI] = + cpu_to_be32(eg_pi->vddc_voltage_table.mask_low); + } +} + +static int ni_populate_voltage_value(struct radeon_device *rdev, + struct atom_voltage_table *table, + u16 value, + NISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + unsigned int i; + + for (i = 0; i < table->count; i++) { + if (value <= table->entries[i].value) { + voltage->index = (u8)i; + voltage->value = cpu_to_be16(table->entries[i].value); + break; + } + } + + if (i >= table->count) + return -EINVAL; + + return 0; +} + +static void ni_populate_mvdd_value(struct radeon_device *rdev, + u32 mclk, + NISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (!pi->mvdd_control) { + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + return; + } + + if (mclk <= pi->mvdd_split_frequency) { + voltage->index = eg_pi->mvdd_low_index; + voltage->value = cpu_to_be16(MVDD_LOW_VALUE); + } else { + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + } +} + +static int ni_get_std_voltage_value(struct radeon_device *rdev, + NISLANDS_SMC_VOLTAGE_VALUE *voltage, + u16 *std_voltage) +{ + if (rdev->pm.dpm.dyn_state.cac_leakage_table.entries && + ((u32)voltage->index < rdev->pm.dpm.dyn_state.cac_leakage_table.count)) + *std_voltage = rdev->pm.dpm.dyn_state.cac_leakage_table.entries[voltage->index].vddc; + else + *std_voltage = be16_to_cpu(voltage->value); + + return 0; +} + +static void ni_populate_std_voltage_value(struct radeon_device *rdev, + u16 value, u8 index, + NISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + voltage->index = index; + voltage->value = cpu_to_be16(value); +} + +static u32 ni_get_smc_power_scaling_factor(struct radeon_device *rdev) +{ + u32 xclk_period; + u32 xclk = radeon_get_xclk(rdev); + u32 tmp = RREG32(CG_CAC_CTRL) & TID_CNT_MASK; + + xclk_period = (1000000000UL / xclk); + xclk_period /= 10000UL; + + return tmp * xclk_period; +} + +static u32 ni_scale_power_for_smc(u32 power_in_watts, u32 scaling_factor) +{ + return (power_in_watts * scaling_factor) << 2; +} + +static u32 ni_calculate_power_boost_limit(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + u32 near_tdp_limit) +{ + struct ni_ps *state = ni_get_ps(radeon_state); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 power_boost_limit = 0; + int ret; + + if (ni_pi->enable_power_containment && + ni_pi->use_power_boost_limit) { + NISLANDS_SMC_VOLTAGE_VALUE vddc; + u16 std_vddc_med; + u16 std_vddc_high; + u64 tmp, n, d; + + if (state->performance_level_count < 3) + return 0; + + ret = ni_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + state->performance_levels[state->performance_level_count - 2].vddc, + &vddc); + if (ret) + return 0; + + ret = ni_get_std_voltage_value(rdev, &vddc, &std_vddc_med); + if (ret) + return 0; + + ret = ni_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + state->performance_levels[state->performance_level_count - 1].vddc, + &vddc); + if (ret) + return 0; + + ret = ni_get_std_voltage_value(rdev, &vddc, &std_vddc_high); + if (ret) + return 0; + + n = ((u64)near_tdp_limit * ((u64)std_vddc_med * (u64)std_vddc_med) * 90); + d = ((u64)std_vddc_high * (u64)std_vddc_high * 100); + tmp = div64_u64(n, d); + + if (tmp >> 32) + return 0; + power_boost_limit = (u32)tmp; + } + + return power_boost_limit; +} + +static int ni_calculate_adjusted_tdp_limits(struct radeon_device *rdev, + bool adjust_polarity, + u32 tdp_adjustment, + u32 *tdp_limit, + u32 *near_tdp_limit) +{ + if (tdp_adjustment > (u32)rdev->pm.dpm.tdp_od_limit) + return -EINVAL; + + if (adjust_polarity) { + *tdp_limit = ((100 + tdp_adjustment) * rdev->pm.dpm.tdp_limit) / 100; + *near_tdp_limit = rdev->pm.dpm.near_tdp_limit + (*tdp_limit - rdev->pm.dpm.tdp_limit); + } else { + *tdp_limit = ((100 - tdp_adjustment) * rdev->pm.dpm.tdp_limit) / 100; + *near_tdp_limit = rdev->pm.dpm.near_tdp_limit - (rdev->pm.dpm.tdp_limit - *tdp_limit); + } + + return 0; +} + +static int ni_populate_smc_tdp_limits(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + + if (ni_pi->enable_power_containment) { + NISLANDS_SMC_STATETABLE *smc_table = &ni_pi->smc_statetable; + u32 scaling_factor = ni_get_smc_power_scaling_factor(rdev); + u32 tdp_limit; + u32 near_tdp_limit; + u32 power_boost_limit; + int ret; + + if (scaling_factor == 0) + return -EINVAL; + + memset(smc_table, 0, sizeof(NISLANDS_SMC_STATETABLE)); + + ret = ni_calculate_adjusted_tdp_limits(rdev, + false, /* ??? */ + rdev->pm.dpm.tdp_adjustment, + &tdp_limit, + &near_tdp_limit); + if (ret) + return ret; + + power_boost_limit = ni_calculate_power_boost_limit(rdev, radeon_state, + near_tdp_limit); + + smc_table->dpm2Params.TDPLimit = + cpu_to_be32(ni_scale_power_for_smc(tdp_limit, scaling_factor)); + smc_table->dpm2Params.NearTDPLimit = + cpu_to_be32(ni_scale_power_for_smc(near_tdp_limit, scaling_factor)); + smc_table->dpm2Params.SafePowerLimit = + cpu_to_be32(ni_scale_power_for_smc((near_tdp_limit * NISLANDS_DPM2_TDP_SAFE_LIMIT_PERCENT) / 100, + scaling_factor)); + smc_table->dpm2Params.PowerBoostLimit = + cpu_to_be32(ni_scale_power_for_smc(power_boost_limit, scaling_factor)); + + ret = rv770_copy_bytes_to_smc(rdev, + (u16)(pi->state_table_start + offsetof(NISLANDS_SMC_STATETABLE, dpm2Params) + + offsetof(PP_NIslands_DPM2Parameters, TDPLimit)), + (u8 *)(&smc_table->dpm2Params.TDPLimit), + sizeof(u32) * 4, pi->sram_end); + if (ret) + return ret; + } + + return 0; +} + +int ni_copy_and_switch_arb_sets(struct radeon_device *rdev, + u32 arb_freq_src, u32 arb_freq_dest) +{ + u32 mc_arb_dram_timing; + u32 mc_arb_dram_timing2; + u32 burst_time; + u32 mc_cg_config; + + switch (arb_freq_src) { + case MC_CG_ARB_FREQ_F0: + mc_arb_dram_timing = RREG32(MC_ARB_DRAM_TIMING); + mc_arb_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + burst_time = (RREG32(MC_ARB_BURST_TIME) & STATE0_MASK) >> STATE0_SHIFT; + break; + case MC_CG_ARB_FREQ_F1: + mc_arb_dram_timing = RREG32(MC_ARB_DRAM_TIMING_1); + mc_arb_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2_1); + burst_time = (RREG32(MC_ARB_BURST_TIME) & STATE1_MASK) >> STATE1_SHIFT; + break; + case MC_CG_ARB_FREQ_F2: + mc_arb_dram_timing = RREG32(MC_ARB_DRAM_TIMING_2); + mc_arb_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2_2); + burst_time = (RREG32(MC_ARB_BURST_TIME) & STATE2_MASK) >> STATE2_SHIFT; + break; + case MC_CG_ARB_FREQ_F3: + mc_arb_dram_timing = RREG32(MC_ARB_DRAM_TIMING_3); + mc_arb_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2_3); + burst_time = (RREG32(MC_ARB_BURST_TIME) & STATE3_MASK) >> STATE3_SHIFT; + break; + default: + return -EINVAL; + } + + switch (arb_freq_dest) { + case MC_CG_ARB_FREQ_F0: + WREG32(MC_ARB_DRAM_TIMING, mc_arb_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2, mc_arb_dram_timing2); + WREG32_P(MC_ARB_BURST_TIME, STATE0(burst_time), ~STATE0_MASK); + break; + case MC_CG_ARB_FREQ_F1: + WREG32(MC_ARB_DRAM_TIMING_1, mc_arb_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_1, mc_arb_dram_timing2); + WREG32_P(MC_ARB_BURST_TIME, STATE1(burst_time), ~STATE1_MASK); + break; + case MC_CG_ARB_FREQ_F2: + WREG32(MC_ARB_DRAM_TIMING_2, mc_arb_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_2, mc_arb_dram_timing2); + WREG32_P(MC_ARB_BURST_TIME, STATE2(burst_time), ~STATE2_MASK); + break; + case MC_CG_ARB_FREQ_F3: + WREG32(MC_ARB_DRAM_TIMING_3, mc_arb_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_3, mc_arb_dram_timing2); + WREG32_P(MC_ARB_BURST_TIME, STATE3(burst_time), ~STATE3_MASK); + break; + default: + return -EINVAL; + } + + mc_cg_config = RREG32(MC_CG_CONFIG) | 0x0000000F; + WREG32(MC_CG_CONFIG, mc_cg_config); + WREG32_P(MC_ARB_CG, CG_ARB_REQ(arb_freq_dest), ~CG_ARB_REQ_MASK); + + return 0; +} + +static int ni_init_arb_table_index(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 tmp; + int ret; + + ret = rv770_read_smc_sram_dword(rdev, ni_pi->arb_table_start, + &tmp, pi->sram_end); + if (ret) + return ret; + + tmp &= 0x00FFFFFF; + tmp |= ((u32)MC_CG_ARB_FREQ_F1) << 24; + + return rv770_write_smc_sram_dword(rdev, ni_pi->arb_table_start, + tmp, pi->sram_end); +} + +static int ni_initial_switch_from_arb_f0_to_f1(struct radeon_device *rdev) +{ + return ni_copy_and_switch_arb_sets(rdev, MC_CG_ARB_FREQ_F0, MC_CG_ARB_FREQ_F1); +} + +static int ni_force_switch_to_arb_f0(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 tmp; + int ret; + + ret = rv770_read_smc_sram_dword(rdev, ni_pi->arb_table_start, + &tmp, pi->sram_end); + if (ret) + return ret; + + tmp = (tmp >> 24) & 0xff; + + if (tmp == MC_CG_ARB_FREQ_F0) + return 0; + + return ni_copy_and_switch_arb_sets(rdev, tmp, MC_CG_ARB_FREQ_F0); +} + +static int ni_populate_memory_timing_parameters(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SMC_NIslands_MCArbDramTimingRegisterSet *arb_regs) +{ + u32 dram_timing; + u32 dram_timing2; + + arb_regs->mc_arb_rfsh_rate = + (u8)rv770_calculate_memory_refresh_rate(rdev, pl->sclk); + + + radeon_atom_set_engine_dram_timings(rdev, + pl->sclk, + pl->mclk); + + dram_timing = RREG32(MC_ARB_DRAM_TIMING); + dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + + arb_regs->mc_arb_dram_timing = cpu_to_be32(dram_timing); + arb_regs->mc_arb_dram_timing2 = cpu_to_be32(dram_timing2); + + return 0; +} + +static int ni_do_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + unsigned int first_arb_set) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + SMC_NIslands_MCArbDramTimingRegisterSet arb_regs = { 0 }; + int i, ret = 0; + + for (i = 0; i < state->performance_level_count; i++) { + ret = ni_populate_memory_timing_parameters(rdev, &state->performance_levels[i], &arb_regs); + if (ret) + break; + + ret = rv770_copy_bytes_to_smc(rdev, + (u16)(ni_pi->arb_table_start + + offsetof(SMC_NIslands_MCArbDramTimingRegisters, data) + + sizeof(SMC_NIslands_MCArbDramTimingRegisterSet) * (first_arb_set + i)), + (u8 *)&arb_regs, + (u16)sizeof(SMC_NIslands_MCArbDramTimingRegisterSet), + pi->sram_end); + if (ret) + break; + } + return ret; +} + +static int ni_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + return ni_do_program_memory_timing_parameters(rdev, radeon_new_state, + NISLANDS_DRIVER_STATE_ARB_INDEX); +} + +static void ni_populate_initial_mvdd_value(struct radeon_device *rdev, + struct NISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + voltage->index = eg_pi->mvdd_high_index; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); +} + +static int ni_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_initial_state, + NISLANDS_SMC_STATETABLE *table) +{ + struct ni_ps *initial_state = ni_get_ps(radeon_initial_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 reg; + int ret; + + table->initialState.levels[0].mclk.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(ni_pi->clock_registers.mpll_ad_func_cntl); + table->initialState.levels[0].mclk.vMPLL_AD_FUNC_CNTL_2 = + cpu_to_be32(ni_pi->clock_registers.mpll_ad_func_cntl_2); + table->initialState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(ni_pi->clock_registers.mpll_dq_func_cntl); + table->initialState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL_2 = + cpu_to_be32(ni_pi->clock_registers.mpll_dq_func_cntl_2); + table->initialState.levels[0].mclk.vMCLK_PWRMGT_CNTL = + cpu_to_be32(ni_pi->clock_registers.mclk_pwrmgt_cntl); + table->initialState.levels[0].mclk.vDLL_CNTL = + cpu_to_be32(ni_pi->clock_registers.dll_cntl); + table->initialState.levels[0].mclk.vMPLL_SS = + cpu_to_be32(ni_pi->clock_registers.mpll_ss1); + table->initialState.levels[0].mclk.vMPLL_SS2 = + cpu_to_be32(ni_pi->clock_registers.mpll_ss2); + table->initialState.levels[0].mclk.mclk_value = + cpu_to_be32(initial_state->performance_levels[0].mclk); + + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(ni_pi->clock_registers.cg_spll_func_cntl); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(ni_pi->clock_registers.cg_spll_func_cntl_2); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(ni_pi->clock_registers.cg_spll_func_cntl_3); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_4 = + cpu_to_be32(ni_pi->clock_registers.cg_spll_func_cntl_4); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM = + cpu_to_be32(ni_pi->clock_registers.cg_spll_spread_spectrum); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM_2 = + cpu_to_be32(ni_pi->clock_registers.cg_spll_spread_spectrum_2); + table->initialState.levels[0].sclk.sclk_value = + cpu_to_be32(initial_state->performance_levels[0].sclk); + table->initialState.levels[0].arbRefreshState = + NISLANDS_INITIAL_STATE_ARB_INDEX; + + table->initialState.levels[0].ACIndex = 0; + + ret = ni_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + initial_state->performance_levels[0].vddc, + &table->initialState.levels[0].vddc); + if (!ret) { + u16 std_vddc; + + ret = ni_get_std_voltage_value(rdev, + &table->initialState.levels[0].vddc, + &std_vddc); + if (!ret) + ni_populate_std_voltage_value(rdev, std_vddc, + table->initialState.levels[0].vddc.index, + &table->initialState.levels[0].std_vddc); + } + + if (eg_pi->vddci_control) + ni_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + initial_state->performance_levels[0].vddci, + &table->initialState.levels[0].vddci); + + ni_populate_initial_mvdd_value(rdev, &table->initialState.levels[0].mvdd); + + reg = CG_R(0xffff) | CG_L(0); + table->initialState.levels[0].aT = cpu_to_be32(reg); + + table->initialState.levels[0].bSP = cpu_to_be32(pi->dsp); + + if (pi->boot_in_gen2) + table->initialState.levels[0].gen2PCIE = 1; + else + table->initialState.levels[0].gen2PCIE = 0; + + if (pi->mem_gddr5) { + table->initialState.levels[0].strobeMode = + cypress_get_strobe_mode_settings(rdev, + initial_state->performance_levels[0].mclk); + + if (initial_state->performance_levels[0].mclk > pi->mclk_edc_enable_threshold) + table->initialState.levels[0].mcFlags = NISLANDS_SMC_MC_EDC_RD_FLAG | NISLANDS_SMC_MC_EDC_WR_FLAG; + else + table->initialState.levels[0].mcFlags = 0; + } + + table->initialState.levelCount = 1; + + table->initialState.flags |= PPSMC_SWSTATE_FLAG_DC; + + table->initialState.levels[0].dpm2.MaxPS = 0; + table->initialState.levels[0].dpm2.NearTDPDec = 0; + table->initialState.levels[0].dpm2.AboveSafeInc = 0; + table->initialState.levels[0].dpm2.BelowSafeInc = 0; + + reg = MIN_POWER_MASK | MAX_POWER_MASK; + table->initialState.levels[0].SQPowerThrottle = cpu_to_be32(reg); + + reg = MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + table->initialState.levels[0].SQPowerThrottle_2 = cpu_to_be32(reg); + + return 0; +} + +static int ni_populate_smc_acpi_state(struct radeon_device *rdev, + NISLANDS_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 mpll_ad_func_cntl = ni_pi->clock_registers.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = ni_pi->clock_registers.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = ni_pi->clock_registers.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = ni_pi->clock_registers.mpll_dq_func_cntl_2; + u32 spll_func_cntl = ni_pi->clock_registers.cg_spll_func_cntl; + u32 spll_func_cntl_2 = ni_pi->clock_registers.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = ni_pi->clock_registers.cg_spll_func_cntl_3; + u32 spll_func_cntl_4 = ni_pi->clock_registers.cg_spll_func_cntl_4; + u32 mclk_pwrmgt_cntl = ni_pi->clock_registers.mclk_pwrmgt_cntl; + u32 dll_cntl = ni_pi->clock_registers.dll_cntl; + u32 reg; + int ret; + + table->ACPIState = table->initialState; + + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + ret = ni_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pi->acpi_vddc, &table->ACPIState.levels[0].vddc); + if (!ret) { + u16 std_vddc; + + ret = ni_get_std_voltage_value(rdev, + &table->ACPIState.levels[0].vddc, &std_vddc); + if (!ret) + ni_populate_std_voltage_value(rdev, std_vddc, + table->ACPIState.levels[0].vddc.index, + &table->ACPIState.levels[0].std_vddc); + } + + if (pi->pcie_gen2) { + if (pi->acpi_pcie_gen2) + table->ACPIState.levels[0].gen2PCIE = 1; + else + table->ACPIState.levels[0].gen2PCIE = 0; + } else { + table->ACPIState.levels[0].gen2PCIE = 0; + } + } else { + ret = ni_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pi->min_vddc_in_table, + &table->ACPIState.levels[0].vddc); + if (!ret) { + u16 std_vddc; + + ret = ni_get_std_voltage_value(rdev, + &table->ACPIState.levels[0].vddc, + &std_vddc); + if (!ret) + ni_populate_std_voltage_value(rdev, std_vddc, + table->ACPIState.levels[0].vddc.index, + &table->ACPIState.levels[0].std_vddc); + } + table->ACPIState.levels[0].gen2PCIE = 0; + } + + if (eg_pi->acpi_vddci) { + if (eg_pi->vddci_control) + ni_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + eg_pi->acpi_vddci, + &table->ACPIState.levels[0].vddci); + } + + + mpll_ad_func_cntl &= ~PDNB; + + mpll_ad_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN; + + if (pi->mem_gddr5) + mpll_dq_func_cntl &= ~PDNB; + mpll_dq_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN | BYPASS; + + + mclk_pwrmgt_cntl |= (MRDCKA0_RESET | + MRDCKA1_RESET | + MRDCKB0_RESET | + MRDCKB1_RESET | + MRDCKC0_RESET | + MRDCKC1_RESET | + MRDCKD0_RESET | + MRDCKD1_RESET); + + mclk_pwrmgt_cntl &= ~(MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + + dll_cntl |= (MRDCKA0_BYPASS | + MRDCKA1_BYPASS | + MRDCKB0_BYPASS | + MRDCKB1_BYPASS | + MRDCKC0_BYPASS | + MRDCKC1_BYPASS | + MRDCKD0_BYPASS | + MRDCKD1_BYPASS); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + table->ACPIState.levels[0].mclk.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + table->ACPIState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + table->ACPIState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + table->ACPIState.levels[0].mclk.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.vDLL_CNTL = cpu_to_be32(dll_cntl); + + table->ACPIState.levels[0].mclk.mclk_value = 0; + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_4 = cpu_to_be32(spll_func_cntl_4); + + table->ACPIState.levels[0].sclk.sclk_value = 0; + + ni_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + if (eg_pi->dynamic_ac_timing) + table->ACPIState.levels[0].ACIndex = 1; + + table->ACPIState.levels[0].dpm2.MaxPS = 0; + table->ACPIState.levels[0].dpm2.NearTDPDec = 0; + table->ACPIState.levels[0].dpm2.AboveSafeInc = 0; + table->ACPIState.levels[0].dpm2.BelowSafeInc = 0; + + reg = MIN_POWER_MASK | MAX_POWER_MASK; + table->ACPIState.levels[0].SQPowerThrottle = cpu_to_be32(reg); + + reg = MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + table->ACPIState.levels[0].SQPowerThrottle_2 = cpu_to_be32(reg); + + return 0; +} + +static int ni_init_smc_table(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + int ret; + struct radeon_ps *radeon_boot_state = rdev->pm.dpm.boot_ps; + NISLANDS_SMC_STATETABLE *table = &ni_pi->smc_statetable; + + memset(table, 0, sizeof(NISLANDS_SMC_STATETABLE)); + + ni_populate_smc_voltage_tables(rdev, table); + + switch (rdev->pm.int_thermal_type) { + case THERMAL_TYPE_NI: + case THERMAL_TYPE_EMC2103_WITH_INTERNAL: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_INTERNAL; + break; + case THERMAL_TYPE_NONE: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_NONE; + break; + default: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL; + break; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REGULATOR_HOT) + table->systemFlags |= PPSMC_SYSTEMFLAG_REGULATOR_HOT; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (pi->mem_gddr5) + table->systemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + ret = ni_populate_smc_initial_state(rdev, radeon_boot_state, table); + if (ret) + return ret; + + ret = ni_populate_smc_acpi_state(rdev, table); + if (ret) + return ret; + + table->driverState = table->initialState; + + table->ULVState = table->initialState; + + ret = ni_do_program_memory_timing_parameters(rdev, radeon_boot_state, + NISLANDS_INITIAL_STATE_ARB_INDEX); + if (ret) + return ret; + + return rv770_copy_bytes_to_smc(rdev, pi->state_table_start, (u8 *)table, + sizeof(NISLANDS_SMC_STATETABLE), pi->sram_end); +} + +static int ni_calculate_sclk_params(struct radeon_device *rdev, + u32 engine_clock, + NISLANDS_SMC_SCLK_VALUE *sclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct atom_clock_dividers dividers; + u32 spll_func_cntl = ni_pi->clock_registers.cg_spll_func_cntl; + u32 spll_func_cntl_2 = ni_pi->clock_registers.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = ni_pi->clock_registers.cg_spll_func_cntl_3; + u32 spll_func_cntl_4 = ni_pi->clock_registers.cg_spll_func_cntl_4; + u32 cg_spll_spread_spectrum = ni_pi->clock_registers.cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2 = ni_pi->clock_registers.cg_spll_spread_spectrum_2; + u64 tmp; + u32 reference_clock = rdev->clock.spll.reference_freq; + u32 reference_divider; + u32 fbdiv; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + engine_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = 1 + dividers.ref_div; + + + tmp = (u64) engine_clock * reference_divider * dividers.post_div * 16834; + do_div(tmp, reference_clock); + fbdiv = (u32) tmp; + + spll_func_cntl &= ~(SPLL_PDIV_A_MASK | SPLL_REF_DIV_MASK); + spll_func_cntl |= SPLL_REF_DIV(dividers.ref_div); + spll_func_cntl |= SPLL_PDIV_A(dividers.post_div); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(2); + + spll_func_cntl_3 &= ~SPLL_FB_DIV_MASK; + spll_func_cntl_3 |= SPLL_FB_DIV(fbdiv); + spll_func_cntl_3 |= SPLL_DITHEN; + + if (pi->sclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = engine_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = 4 * ss.percentage * fbdiv / (clk_s * 10000); + + cg_spll_spread_spectrum &= ~CLK_S_MASK; + cg_spll_spread_spectrum |= CLK_S(clk_s); + cg_spll_spread_spectrum |= SSEN; + + cg_spll_spread_spectrum_2 &= ~CLK_V_MASK; + cg_spll_spread_spectrum_2 |= CLK_V(clk_v); + } + } + + sclk->sclk_value = engine_clock; + sclk->vCG_SPLL_FUNC_CNTL = spll_func_cntl; + sclk->vCG_SPLL_FUNC_CNTL_2 = spll_func_cntl_2; + sclk->vCG_SPLL_FUNC_CNTL_3 = spll_func_cntl_3; + sclk->vCG_SPLL_FUNC_CNTL_4 = spll_func_cntl_4; + sclk->vCG_SPLL_SPREAD_SPECTRUM = cg_spll_spread_spectrum; + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cg_spll_spread_spectrum_2; + + return 0; +} + +static int ni_populate_sclk_value(struct radeon_device *rdev, + u32 engine_clock, + NISLANDS_SMC_SCLK_VALUE *sclk) +{ + NISLANDS_SMC_SCLK_VALUE sclk_tmp; + int ret; + + ret = ni_calculate_sclk_params(rdev, engine_clock, &sclk_tmp); + if (!ret) { + sclk->sclk_value = cpu_to_be32(sclk_tmp.sclk_value); + sclk->vCG_SPLL_FUNC_CNTL = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL); + sclk->vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_2); + sclk->vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_3); + sclk->vCG_SPLL_FUNC_CNTL_4 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_4); + sclk->vCG_SPLL_SPREAD_SPECTRUM = cpu_to_be32(sclk_tmp.vCG_SPLL_SPREAD_SPECTRUM); + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cpu_to_be32(sclk_tmp.vCG_SPLL_SPREAD_SPECTRUM_2); + } + + return ret; +} + +static int ni_init_smc_spll_table(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + SMC_NISLANDS_SPLL_DIV_TABLE *spll_table; + NISLANDS_SMC_SCLK_VALUE sclk_params; + u32 fb_div; + u32 p_div; + u32 clk_s; + u32 clk_v; + u32 sclk = 0; + int i, ret; + u32 tmp; + + if (ni_pi->spll_table_start == 0) + return -EINVAL; + + spll_table = kzalloc(sizeof(SMC_NISLANDS_SPLL_DIV_TABLE), GFP_KERNEL); + if (spll_table == NULL) + return -ENOMEM; + + for (i = 0; i < 256; i++) { + ret = ni_calculate_sclk_params(rdev, sclk, &sclk_params); + if (ret) + break; + + p_div = (sclk_params.vCG_SPLL_FUNC_CNTL & SPLL_PDIV_A_MASK) >> SPLL_PDIV_A_SHIFT; + fb_div = (sclk_params.vCG_SPLL_FUNC_CNTL_3 & SPLL_FB_DIV_MASK) >> SPLL_FB_DIV_SHIFT; + clk_s = (sclk_params.vCG_SPLL_SPREAD_SPECTRUM & CLK_S_MASK) >> CLK_S_SHIFT; + clk_v = (sclk_params.vCG_SPLL_SPREAD_SPECTRUM_2 & CLK_V_MASK) >> CLK_V_SHIFT; + + fb_div &= ~0x00001FFF; + fb_div >>= 1; + clk_v >>= 6; + + if (p_div & ~(SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_MASK >> SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT)) + ret = -EINVAL; + + if (clk_s & ~(SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_MASK >> SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT)) + ret = -EINVAL; + + if (clk_s & ~(SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_MASK >> SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT)) + ret = -EINVAL; + + if (clk_v & ~(SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_MASK >> SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT)) + ret = -EINVAL; + + if (ret) + break; + + tmp = ((fb_div << SMC_NISLANDS_SPLL_DIV_TABLE_FBDIV_SHIFT) & SMC_NISLANDS_SPLL_DIV_TABLE_FBDIV_MASK) | + ((p_div << SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT) & SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_MASK); + spll_table->freq[i] = cpu_to_be32(tmp); + + tmp = ((clk_v << SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT) & SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_MASK) | + ((clk_s << SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT) & SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_MASK); + spll_table->ss[i] = cpu_to_be32(tmp); + + sclk += 512; + } + + if (!ret) + ret = rv770_copy_bytes_to_smc(rdev, ni_pi->spll_table_start, (u8 *)spll_table, + sizeof(SMC_NISLANDS_SPLL_DIV_TABLE), pi->sram_end); + + kfree(spll_table); + + return ret; +} + +static int ni_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, + u32 memory_clock, + NISLANDS_SMC_MCLK_VALUE *mclk, + bool strobe_mode, + bool dll_state_on) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 mpll_ad_func_cntl = ni_pi->clock_registers.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = ni_pi->clock_registers.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = ni_pi->clock_registers.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = ni_pi->clock_registers.mpll_dq_func_cntl_2; + u32 mclk_pwrmgt_cntl = ni_pi->clock_registers.mclk_pwrmgt_cntl; + u32 dll_cntl = ni_pi->clock_registers.dll_cntl; + u32 mpll_ss1 = ni_pi->clock_registers.mpll_ss1; + u32 mpll_ss2 = ni_pi->clock_registers.mpll_ss2; + struct atom_clock_dividers dividers; + u32 ibias; + u32 dll_speed; + int ret; + u32 mc_seq_misc7; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + memory_clock, strobe_mode, ÷rs); + if (ret) + return ret; + + if (!strobe_mode) { + mc_seq_misc7 = RREG32(MC_SEQ_MISC7); + + if (mc_seq_misc7 & 0x8000000) + dividers.post_div = 1; + } + + ibias = cypress_map_clkf_to_ibias(rdev, dividers.whole_fb_div); + + mpll_ad_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_ad_func_cntl |= CLKR(dividers.ref_div); + mpll_ad_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_ad_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_ad_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_ad_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_ad_func_cntl_2 |= VCO_MODE; + else + mpll_ad_func_cntl_2 &= ~VCO_MODE; + + if (pi->mem_gddr5) { + mpll_dq_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_dq_func_cntl |= CLKR(dividers.ref_div); + mpll_dq_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_dq_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_dq_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_dq_func_cntl |= IBIAS(ibias); + + if (strobe_mode) + mpll_dq_func_cntl &= ~PDNB; + else + mpll_dq_func_cntl |= PDNB; + + if (dividers.vco_mode) + mpll_dq_func_cntl_2 |= VCO_MODE; + else + mpll_dq_func_cntl_2 &= ~VCO_MODE; + } + + if (pi->mclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = memory_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, vco_freq)) { + u32 reference_clock = rdev->clock.mpll.reference_freq; + u32 decoded_ref = rv740_get_decoded_reference_divider(dividers.ref_div); + u32 clk_s = reference_clock * 5 / (decoded_ref * ss.rate); + u32 clk_v = ss.percentage * + (0x4000 * dividers.whole_fb_div + 0x800 * dividers.frac_fb_div) / (clk_s * 625); + + mpll_ss1 &= ~CLKV_MASK; + mpll_ss1 |= CLKV(clk_v); + + mpll_ss2 &= ~CLKS_MASK; + mpll_ss2 |= CLKS(clk_s); + } + } + + dll_speed = rv740_get_dll_speed(pi->mem_gddr5, + memory_clock); + + mclk_pwrmgt_cntl &= ~DLL_SPEED_MASK; + mclk_pwrmgt_cntl |= DLL_SPEED(dll_speed); + if (dll_state_on) + mclk_pwrmgt_cntl |= (MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + else + mclk_pwrmgt_cntl &= ~(MRDCKA0_PDNB | + MRDCKA1_PDNB | + MRDCKB0_PDNB | + MRDCKB1_PDNB | + MRDCKC0_PDNB | + MRDCKC1_PDNB | + MRDCKD0_PDNB | + MRDCKD1_PDNB); + + + mclk->mclk_value = cpu_to_be32(memory_clock); + mclk->vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + mclk->vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + mclk->vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + mclk->vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + mclk->vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->vDLL_CNTL = cpu_to_be32(dll_cntl); + mclk->vMPLL_SS = cpu_to_be32(mpll_ss1); + mclk->vMPLL_SS2 = cpu_to_be32(mpll_ss2); + + return 0; +} + +static void ni_populate_smc_sp(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + NISLANDS_SMC_SWSTATE *smc_state) +{ + struct ni_ps *ps = ni_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + for (i = 0; i < ps->performance_level_count - 1; i++) + smc_state->levels[i].bSP = cpu_to_be32(pi->dsp); + + smc_state->levels[ps->performance_level_count - 1].bSP = + cpu_to_be32(pi->psp); +} + +static int ni_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + NISLANDS_SMC_HW_PERFORMANCE_LEVEL *level) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + int ret; + bool dll_state_on; + u16 std_vddc; + u32 tmp = RREG32(DC_STUTTER_CNTL); + + level->gen2PCIE = pi->pcie_gen2 ? + ((pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? 1 : 0) : 0; + + ret = ni_populate_sclk_value(rdev, pl->sclk, &level->sclk); + if (ret) + return ret; + + level->mcFlags = 0; + if (pi->mclk_stutter_mode_threshold && + (pl->mclk <= pi->mclk_stutter_mode_threshold) && + !eg_pi->uvd_enabled && + (tmp & DC_STUTTER_ENABLE_A) && + (tmp & DC_STUTTER_ENABLE_B)) + level->mcFlags |= NISLANDS_SMC_MC_STUTTER_EN; + + if (pi->mem_gddr5) { + if (pl->mclk > pi->mclk_edc_enable_threshold) + level->mcFlags |= NISLANDS_SMC_MC_EDC_RD_FLAG; + if (pl->mclk > eg_pi->mclk_edc_wr_enable_threshold) + level->mcFlags |= NISLANDS_SMC_MC_EDC_WR_FLAG; + + level->strobeMode = cypress_get_strobe_mode_settings(rdev, pl->mclk); + + if (level->strobeMode & NISLANDS_SMC_STROBE_ENABLE) { + if (cypress_get_mclk_frequency_ratio(rdev, pl->mclk, true) >= + ((RREG32(MC_SEQ_MISC7) >> 16) & 0xf)) + dll_state_on = ((RREG32(MC_SEQ_MISC5) >> 1) & 0x1) ? true : false; + else + dll_state_on = ((RREG32(MC_SEQ_MISC6) >> 1) & 0x1) ? true : false; + } else { + dll_state_on = false; + if (pl->mclk > ni_pi->mclk_rtt_mode_threshold) + level->mcFlags |= NISLANDS_SMC_MC_RTT_ENABLE; + } + + ret = ni_populate_mclk_value(rdev, pl->sclk, pl->mclk, + &level->mclk, + (level->strobeMode & NISLANDS_SMC_STROBE_ENABLE) != 0, + dll_state_on); + } else + ret = ni_populate_mclk_value(rdev, pl->sclk, pl->mclk, &level->mclk, 1, 1); + + if (ret) + return ret; + + ret = ni_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + pl->vddc, &level->vddc); + if (ret) + return ret; + + ret = ni_get_std_voltage_value(rdev, &level->vddc, &std_vddc); + if (ret) + return ret; + + ni_populate_std_voltage_value(rdev, std_vddc, + level->vddc.index, &level->std_vddc); + + if (eg_pi->vddci_control) { + ret = ni_populate_voltage_value(rdev, &eg_pi->vddci_voltage_table, + pl->vddci, &level->vddci); + if (ret) + return ret; + } + + ni_populate_mvdd_value(rdev, pl->mclk, &level->mvdd); + + return ret; +} + +static int ni_populate_smc_t(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + NISLANDS_SMC_SWSTATE *smc_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + u32 a_t; + u32 t_l, t_h; + u32 high_bsp; + int i, ret; + + if (state->performance_level_count >= 9) + return -EINVAL; + + if (state->performance_level_count < 2) { + a_t = CG_R(0xffff) | CG_L(0); + smc_state->levels[0].aT = cpu_to_be32(a_t); + return 0; + } + + smc_state->levels[0].aT = cpu_to_be32(0); + + for (i = 0; i <= state->performance_level_count - 2; i++) { + if (eg_pi->uvd_enabled) + ret = r600_calculate_at( + 1000 * (i * (eg_pi->smu_uvd_hs ? 2 : 8) + 2), + 100 * R600_AH_DFLT, + state->performance_levels[i + 1].sclk, + state->performance_levels[i].sclk, + &t_l, + &t_h); + else + ret = r600_calculate_at( + 1000 * (i + 1), + 100 * R600_AH_DFLT, + state->performance_levels[i + 1].sclk, + state->performance_levels[i].sclk, + &t_l, + &t_h); + + if (ret) { + t_h = (i + 1) * 1000 - 50 * R600_AH_DFLT; + t_l = (i + 1) * 1000 + 50 * R600_AH_DFLT; + } + + a_t = be32_to_cpu(smc_state->levels[i].aT) & ~CG_R_MASK; + a_t |= CG_R(t_l * pi->bsp / 20000); + smc_state->levels[i].aT = cpu_to_be32(a_t); + + high_bsp = (i == state->performance_level_count - 2) ? + pi->pbsp : pi->bsp; + + a_t = CG_R(0xffff) | CG_L(t_h * high_bsp / 20000); + smc_state->levels[i + 1].aT = cpu_to_be32(a_t); + } + + return 0; +} + +static int ni_populate_power_containment_values(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + NISLANDS_SMC_SWSTATE *smc_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + u32 prev_sclk; + u32 max_sclk; + u32 min_sclk; + int i, ret; + u32 tdp_limit; + u32 near_tdp_limit; + u32 power_boost_limit; + u8 max_ps_percent; + + if (ni_pi->enable_power_containment == false) + return 0; + + if (state->performance_level_count == 0) + return -EINVAL; + + if (smc_state->levelCount != state->performance_level_count) + return -EINVAL; + + ret = ni_calculate_adjusted_tdp_limits(rdev, + false, /* ??? */ + rdev->pm.dpm.tdp_adjustment, + &tdp_limit, + &near_tdp_limit); + if (ret) + return ret; + + power_boost_limit = ni_calculate_power_boost_limit(rdev, radeon_state, near_tdp_limit); + + ret = rv770_write_smc_sram_dword(rdev, + pi->state_table_start + + offsetof(NISLANDS_SMC_STATETABLE, dpm2Params) + + offsetof(PP_NIslands_DPM2Parameters, PowerBoostLimit), + ni_scale_power_for_smc(power_boost_limit, ni_get_smc_power_scaling_factor(rdev)), + pi->sram_end); + if (ret) + power_boost_limit = 0; + + smc_state->levels[0].dpm2.MaxPS = 0; + smc_state->levels[0].dpm2.NearTDPDec = 0; + smc_state->levels[0].dpm2.AboveSafeInc = 0; + smc_state->levels[0].dpm2.BelowSafeInc = 0; + smc_state->levels[0].stateFlags |= power_boost_limit ? PPSMC_STATEFLAG_POWERBOOST : 0; + + for (i = 1; i < state->performance_level_count; i++) { + prev_sclk = state->performance_levels[i-1].sclk; + max_sclk = state->performance_levels[i].sclk; + max_ps_percent = (i != (state->performance_level_count - 1)) ? + NISLANDS_DPM2_MAXPS_PERCENT_M : NISLANDS_DPM2_MAXPS_PERCENT_H; + + if (max_sclk < prev_sclk) + return -EINVAL; + + if ((max_ps_percent == 0) || (prev_sclk == max_sclk) || eg_pi->uvd_enabled) + min_sclk = max_sclk; + else if (1 == i) + min_sclk = prev_sclk; + else + min_sclk = (prev_sclk * (u32)max_ps_percent) / 100; + + if (min_sclk < state->performance_levels[0].sclk) + min_sclk = state->performance_levels[0].sclk; + + if (min_sclk == 0) + return -EINVAL; + + smc_state->levels[i].dpm2.MaxPS = + (u8)((NISLANDS_DPM2_MAX_PULSE_SKIP * (max_sclk - min_sclk)) / max_sclk); + smc_state->levels[i].dpm2.NearTDPDec = NISLANDS_DPM2_NEAR_TDP_DEC; + smc_state->levels[i].dpm2.AboveSafeInc = NISLANDS_DPM2_ABOVE_SAFE_INC; + smc_state->levels[i].dpm2.BelowSafeInc = NISLANDS_DPM2_BELOW_SAFE_INC; + smc_state->levels[i].stateFlags |= + ((i != (state->performance_level_count - 1)) && power_boost_limit) ? + PPSMC_STATEFLAG_POWERBOOST : 0; + } + + return 0; +} + +static int ni_populate_sq_ramping_values(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + NISLANDS_SMC_SWSTATE *smc_state) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + u32 sq_power_throttle; + u32 sq_power_throttle2; + bool enable_sq_ramping = ni_pi->enable_sq_ramping; + int i; + + if (state->performance_level_count == 0) + return -EINVAL; + + if (smc_state->levelCount != state->performance_level_count) + return -EINVAL; + + if (rdev->pm.dpm.sq_ramping_threshold == 0) + return -EINVAL; + + if (NISLANDS_DPM2_SQ_RAMP_MAX_POWER > (MAX_POWER_MASK >> MAX_POWER_SHIFT)) + enable_sq_ramping = false; + + if (NISLANDS_DPM2_SQ_RAMP_MIN_POWER > (MIN_POWER_MASK >> MIN_POWER_SHIFT)) + enable_sq_ramping = false; + + if (NISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA > (MAX_POWER_DELTA_MASK >> MAX_POWER_DELTA_SHIFT)) + enable_sq_ramping = false; + + if (NISLANDS_DPM2_SQ_RAMP_STI_SIZE > (STI_SIZE_MASK >> STI_SIZE_SHIFT)) + enable_sq_ramping = false; + + if (NISLANDS_DPM2_SQ_RAMP_LTI_RATIO <= (LTI_RATIO_MASK >> LTI_RATIO_SHIFT)) + enable_sq_ramping = false; + + for (i = 0; i < state->performance_level_count; i++) { + sq_power_throttle = 0; + sq_power_throttle2 = 0; + + if ((state->performance_levels[i].sclk >= rdev->pm.dpm.sq_ramping_threshold) && + enable_sq_ramping) { + sq_power_throttle |= MAX_POWER(NISLANDS_DPM2_SQ_RAMP_MAX_POWER); + sq_power_throttle |= MIN_POWER(NISLANDS_DPM2_SQ_RAMP_MIN_POWER); + sq_power_throttle2 |= MAX_POWER_DELTA(NISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA); + sq_power_throttle2 |= STI_SIZE(NISLANDS_DPM2_SQ_RAMP_STI_SIZE); + sq_power_throttle2 |= LTI_RATIO(NISLANDS_DPM2_SQ_RAMP_LTI_RATIO); + } else { + sq_power_throttle |= MAX_POWER_MASK | MIN_POWER_MASK; + sq_power_throttle2 |= MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + } + + smc_state->levels[i].SQPowerThrottle = cpu_to_be32(sq_power_throttle); + smc_state->levels[i].SQPowerThrottle_2 = cpu_to_be32(sq_power_throttle2); + } + + return 0; +} + +static int ni_enable_power_containment(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + bool enable) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + PPSMC_Result smc_result; + int ret = 0; + + if (ni_pi->enable_power_containment) { + if (enable) { + if (!r600_is_uvd_state(radeon_new_state->class, radeon_new_state->class2)) { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_TDPClampingActive); + if (smc_result != PPSMC_Result_OK) { + ret = -EINVAL; + ni_pi->pc_enabled = false; + } else { + ni_pi->pc_enabled = true; + } + } + } else { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_TDPClampingInactive); + if (smc_result != PPSMC_Result_OK) + ret = -EINVAL; + ni_pi->pc_enabled = false; + } + } + + return ret; +} + +static int ni_convert_power_state_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + NISLANDS_SMC_SWSTATE *smc_state) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + int i, ret; + u32 threshold = state->performance_levels[state->performance_level_count - 1].sclk * 100 / 100; + + if (!(radeon_state->caps & ATOM_PPLIB_DISALLOW_ON_DC)) + smc_state->flags |= PPSMC_SWSTATE_FLAG_DC; + + smc_state->levelCount = 0; + + if (state->performance_level_count > NISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE) + return -EINVAL; + + for (i = 0; i < state->performance_level_count; i++) { + ret = ni_convert_power_level_to_smc(rdev, &state->performance_levels[i], + &smc_state->levels[i]); + smc_state->levels[i].arbRefreshState = + (u8)(NISLANDS_DRIVER_STATE_ARB_INDEX + i); + + if (ret) + return ret; + + if (ni_pi->enable_power_containment) + smc_state->levels[i].displayWatermark = + (state->performance_levels[i].sclk < threshold) ? + PPSMC_DISPLAY_WATERMARK_LOW : PPSMC_DISPLAY_WATERMARK_HIGH; + else + smc_state->levels[i].displayWatermark = (i < 2) ? + PPSMC_DISPLAY_WATERMARK_LOW : PPSMC_DISPLAY_WATERMARK_HIGH; + + if (eg_pi->dynamic_ac_timing) + smc_state->levels[i].ACIndex = NISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT + i; + else + smc_state->levels[i].ACIndex = 0; + + smc_state->levelCount++; + } + + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_watermark_threshold, + cpu_to_be32(threshold / 512)); + + ni_populate_smc_sp(rdev, radeon_state, smc_state); + + ret = ni_populate_power_containment_values(rdev, radeon_state, smc_state); + if (ret) + ni_pi->enable_power_containment = false; + + ret = ni_populate_sq_ramping_values(rdev, radeon_state, smc_state); + if (ret) + ni_pi->enable_sq_ramping = false; + + return ni_populate_smc_t(rdev, radeon_state, smc_state); +} + +static int ni_upload_sw_state(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u16 address = pi->state_table_start + + offsetof(NISLANDS_SMC_STATETABLE, driverState); + u16 state_size = sizeof(NISLANDS_SMC_SWSTATE) + + ((NISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1) * sizeof(NISLANDS_SMC_HW_PERFORMANCE_LEVEL)); + int ret; + NISLANDS_SMC_SWSTATE *smc_state = kzalloc(state_size, GFP_KERNEL); + + if (smc_state == NULL) + return -ENOMEM; + + ret = ni_convert_power_state_to_smc(rdev, radeon_new_state, smc_state); + if (ret) + goto done; + + ret = rv770_copy_bytes_to_smc(rdev, address, (u8 *)smc_state, state_size, pi->sram_end); + +done: + kfree(smc_state); + + return ret; +} + +static int ni_set_mc_special_registers(struct radeon_device *rdev, + struct ni_mc_reg_table *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 i, j, k; + u32 temp_reg; + + for (i = 0, j = table->last; i < table->last; i++) { + switch (table->mc_reg_address[i].s1) { + case MC_SEQ_MISC1 >> 2: + if (j >= SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + temp_reg = RREG32(MC_PMG_CMD_EMRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_EMRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + for (k = 0; k < table->num_entries; k++) + table->mc_reg_table_entry[k].mc_data[j] = + ((temp_reg & 0xffff0000)) | + ((table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16); + j++; + if (j >= SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + + temp_reg = RREG32(MC_PMG_CMD_MRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS_LP >> 2; + for(k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + if (!pi->mem_gddr5) + table->mc_reg_table_entry[k].mc_data[j] |= 0x100; + } + j++; + if (j > SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + break; + case MC_SEQ_RESERVE_M >> 2: + temp_reg = RREG32(MC_PMG_CMD_MRS1); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS1 >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + for (k = 0; k < table->num_entries; k++) + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + j++; + if (j > SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + break; + default: + break; + } + } + + table->last = j; + + return 0; +} + +static bool ni_check_s0_mc_reg_index(u16 in_reg, u16 *out_reg) +{ + bool result = true; + + switch (in_reg) { + case MC_SEQ_RAS_TIMING >> 2: + *out_reg = MC_SEQ_RAS_TIMING_LP >> 2; + break; + case MC_SEQ_CAS_TIMING >> 2: + *out_reg = MC_SEQ_CAS_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING >> 2: + *out_reg = MC_SEQ_MISC_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING2 >> 2: + *out_reg = MC_SEQ_MISC_TIMING2_LP >> 2; + break; + case MC_SEQ_RD_CTL_D0 >> 2: + *out_reg = MC_SEQ_RD_CTL_D0_LP >> 2; + break; + case MC_SEQ_RD_CTL_D1 >> 2: + *out_reg = MC_SEQ_RD_CTL_D1_LP >> 2; + break; + case MC_SEQ_WR_CTL_D0 >> 2: + *out_reg = MC_SEQ_WR_CTL_D0_LP >> 2; + break; + case MC_SEQ_WR_CTL_D1 >> 2: + *out_reg = MC_SEQ_WR_CTL_D1_LP >> 2; + break; + case MC_PMG_CMD_EMRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + break; + case MC_PMG_CMD_MRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS_LP >> 2; + break; + case MC_PMG_CMD_MRS1 >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + break; + case MC_SEQ_PMG_TIMING >> 2: + *out_reg = MC_SEQ_PMG_TIMING_LP >> 2; + break; + case MC_PMG_CMD_MRS2 >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS2_LP >> 2; + break; + default: + result = false; + break; + } + + return result; +} + +static void ni_set_valid_flag(struct ni_mc_reg_table *table) +{ + u8 i, j; + + for (i = 0; i < table->last; i++) { + for (j = 1; j < table->num_entries; j++) { + if (table->mc_reg_table_entry[j-1].mc_data[i] != table->mc_reg_table_entry[j].mc_data[i]) { + table->valid_flag |= 1 << i; + break; + } + } + } +} + +static void ni_set_s0_mc_reg_index(struct ni_mc_reg_table *table) +{ + u32 i; + u16 address; + + for (i = 0; i < table->last; i++) + table->mc_reg_address[i].s0 = + ni_check_s0_mc_reg_index(table->mc_reg_address[i].s1, &address) ? + address : table->mc_reg_address[i].s1; +} + +static int ni_copy_vbios_mc_reg_table(struct atom_mc_reg_table *table, + struct ni_mc_reg_table *ni_table) +{ + u8 i, j; + + if (table->last > SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + if (table->num_entries > MAX_AC_TIMING_ENTRIES) + return -EINVAL; + + for (i = 0; i < table->last; i++) + ni_table->mc_reg_address[i].s1 = table->mc_reg_address[i].s1; + ni_table->last = table->last; + + for (i = 0; i < table->num_entries; i++) { + ni_table->mc_reg_table_entry[i].mclk_max = + table->mc_reg_table_entry[i].mclk_max; + for (j = 0; j < table->last; j++) + ni_table->mc_reg_table_entry[i].mc_data[j] = + table->mc_reg_table_entry[i].mc_data[j]; + } + ni_table->num_entries = table->num_entries; + + return 0; +} + +static int ni_initialize_mc_reg_table(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + int ret; + struct atom_mc_reg_table *table; + struct ni_mc_reg_table *ni_table = &ni_pi->mc_reg_table; + u8 module_index = rv770_get_memory_module_index(rdev); + + table = kzalloc(sizeof(struct atom_mc_reg_table), GFP_KERNEL); + if (!table) + return -ENOMEM; + + WREG32(MC_SEQ_RAS_TIMING_LP, RREG32(MC_SEQ_RAS_TIMING)); + WREG32(MC_SEQ_CAS_TIMING_LP, RREG32(MC_SEQ_CAS_TIMING)); + WREG32(MC_SEQ_MISC_TIMING_LP, RREG32(MC_SEQ_MISC_TIMING)); + WREG32(MC_SEQ_MISC_TIMING2_LP, RREG32(MC_SEQ_MISC_TIMING2)); + WREG32(MC_SEQ_PMG_CMD_EMRS_LP, RREG32(MC_PMG_CMD_EMRS)); + WREG32(MC_SEQ_PMG_CMD_MRS_LP, RREG32(MC_PMG_CMD_MRS)); + WREG32(MC_SEQ_PMG_CMD_MRS1_LP, RREG32(MC_PMG_CMD_MRS1)); + WREG32(MC_SEQ_WR_CTL_D0_LP, RREG32(MC_SEQ_WR_CTL_D0)); + WREG32(MC_SEQ_WR_CTL_D1_LP, RREG32(MC_SEQ_WR_CTL_D1)); + WREG32(MC_SEQ_RD_CTL_D0_LP, RREG32(MC_SEQ_RD_CTL_D0)); + WREG32(MC_SEQ_RD_CTL_D1_LP, RREG32(MC_SEQ_RD_CTL_D1)); + WREG32(MC_SEQ_PMG_TIMING_LP, RREG32(MC_SEQ_PMG_TIMING)); + WREG32(MC_SEQ_PMG_CMD_MRS2_LP, RREG32(MC_PMG_CMD_MRS2)); + + ret = radeon_atom_init_mc_reg_table(rdev, module_index, table); + + if (ret) + goto init_mc_done; + + ret = ni_copy_vbios_mc_reg_table(table, ni_table); + + if (ret) + goto init_mc_done; + + ni_set_s0_mc_reg_index(ni_table); + + ret = ni_set_mc_special_registers(rdev, ni_table); + + if (ret) + goto init_mc_done; + + ni_set_valid_flag(ni_table); + +init_mc_done: + kfree(table); + + return ret; +} + +static void ni_populate_mc_reg_addresses(struct radeon_device *rdev, + SMC_NIslands_MCRegisters *mc_reg_table) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 i, j; + + for (i = 0, j = 0; j < ni_pi->mc_reg_table.last; j++) { + if (ni_pi->mc_reg_table.valid_flag & (1 << j)) { + if (i >= SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + break; + mc_reg_table->address[i].s0 = + cpu_to_be16(ni_pi->mc_reg_table.mc_reg_address[j].s0); + mc_reg_table->address[i].s1 = + cpu_to_be16(ni_pi->mc_reg_table.mc_reg_address[j].s1); + i++; + } + } + mc_reg_table->last = (u8)i; +} + + +static void ni_convert_mc_registers(struct ni_mc_reg_entry *entry, + SMC_NIslands_MCRegisterSet *data, + u32 num_entries, u32 valid_flag) +{ + u32 i, j; + + for (i = 0, j = 0; j < num_entries; j++) { + if (valid_flag & (1 << j)) { + data->value[i] = cpu_to_be32(entry->mc_data[j]); + i++; + } + } +} + +static void ni_convert_mc_reg_table_entry_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SMC_NIslands_MCRegisterSet *mc_reg_table_data) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 i = 0; + + for (i = 0; i < ni_pi->mc_reg_table.num_entries; i++) { + if (pl->mclk <= ni_pi->mc_reg_table.mc_reg_table_entry[i].mclk_max) + break; + } + + if ((i == ni_pi->mc_reg_table.num_entries) && (i > 0)) + --i; + + ni_convert_mc_registers(&ni_pi->mc_reg_table.mc_reg_table_entry[i], + mc_reg_table_data, + ni_pi->mc_reg_table.last, + ni_pi->mc_reg_table.valid_flag); +} + +static void ni_convert_mc_reg_table_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SMC_NIslands_MCRegisters *mc_reg_table) +{ + struct ni_ps *state = ni_get_ps(radeon_state); + int i; + + for (i = 0; i < state->performance_level_count; i++) { + ni_convert_mc_reg_table_entry_to_smc(rdev, + &state->performance_levels[i], + &mc_reg_table->data[NISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT + i]); + } +} + +static int ni_populate_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *boot_state = ni_get_ps(radeon_boot_state); + SMC_NIslands_MCRegisters *mc_reg_table = &ni_pi->smc_mc_reg_table; + + memset(mc_reg_table, 0, sizeof(SMC_NIslands_MCRegisters)); + + rv770_write_smc_soft_register(rdev, NI_SMC_SOFT_REGISTER_seq_index, 1); + + ni_populate_mc_reg_addresses(rdev, mc_reg_table); + + ni_convert_mc_reg_table_entry_to_smc(rdev, &boot_state->performance_levels[0], + &mc_reg_table->data[0]); + + ni_convert_mc_registers(&ni_pi->mc_reg_table.mc_reg_table_entry[0], + &mc_reg_table->data[1], + ni_pi->mc_reg_table.last, + ni_pi->mc_reg_table.valid_flag); + + ni_convert_mc_reg_table_to_smc(rdev, radeon_boot_state, mc_reg_table); + + return rv770_copy_bytes_to_smc(rdev, eg_pi->mc_reg_table_start, + (u8 *)mc_reg_table, + sizeof(SMC_NIslands_MCRegisters), + pi->sram_end); +} + +static int ni_upload_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *ni_new_state = ni_get_ps(radeon_new_state); + SMC_NIslands_MCRegisters *mc_reg_table = &ni_pi->smc_mc_reg_table; + u16 address; + + memset(mc_reg_table, 0, sizeof(SMC_NIslands_MCRegisters)); + + ni_convert_mc_reg_table_to_smc(rdev, radeon_new_state, mc_reg_table); + + address = eg_pi->mc_reg_table_start + + (u16)offsetof(SMC_NIslands_MCRegisters, data[NISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT]); + + return rv770_copy_bytes_to_smc(rdev, address, + (u8 *)&mc_reg_table->data[NISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT], + sizeof(SMC_NIslands_MCRegisterSet) * ni_new_state->performance_level_count, + pi->sram_end); +} + +static int ni_init_driver_calculated_leakage_table(struct radeon_device *rdev, + PP_NIslands_CACTABLES *cac_tables) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 leakage = 0; + unsigned int i, j, table_size; + s32 t; + u32 smc_leakage, max_leakage = 0; + u32 scaling_factor; + + table_size = eg_pi->vddc_voltage_table.count; + + if (SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES < table_size) + table_size = SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; + + scaling_factor = ni_get_smc_power_scaling_factor(rdev); + + for (i = 0; i < SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES; i++) { + for (j = 0; j < table_size; j++) { + t = (1000 * ((i + 1) * 8)); + + if (t < ni_pi->cac_data.leakage_minimum_temperature) + t = ni_pi->cac_data.leakage_minimum_temperature; + + ni_calculate_leakage_for_v_and_t(rdev, + &ni_pi->cac_data.leakage_coefficients, + eg_pi->vddc_voltage_table.entries[j].value, + t, + ni_pi->cac_data.i_leakage, + &leakage); + + smc_leakage = ni_scale_power_for_smc(leakage, scaling_factor) / 1000; + if (smc_leakage > max_leakage) + max_leakage = smc_leakage; + + cac_tables->cac_lkge_lut[i][j] = cpu_to_be32(smc_leakage); + } + } + + for (j = table_size; j < SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; j++) { + for (i = 0; i < SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES; i++) + cac_tables->cac_lkge_lut[i][j] = cpu_to_be32(max_leakage); + } + return 0; +} + +static int ni_init_simplified_leakage_table(struct radeon_device *rdev, + PP_NIslands_CACTABLES *cac_tables) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_cac_leakage_table *leakage_table = + &rdev->pm.dpm.dyn_state.cac_leakage_table; + u32 i, j, table_size; + u32 smc_leakage, max_leakage = 0; + u32 scaling_factor; + + if (!leakage_table) + return -EINVAL; + + table_size = leakage_table->count; + + if (eg_pi->vddc_voltage_table.count != table_size) + table_size = (eg_pi->vddc_voltage_table.count < leakage_table->count) ? + eg_pi->vddc_voltage_table.count : leakage_table->count; + + if (SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES < table_size) + table_size = SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; + + if (table_size == 0) + return -EINVAL; + + scaling_factor = ni_get_smc_power_scaling_factor(rdev); + + for (j = 0; j < table_size; j++) { + smc_leakage = leakage_table->entries[j].leakage; + + if (smc_leakage > max_leakage) + max_leakage = smc_leakage; + + for (i = 0; i < SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES; i++) + cac_tables->cac_lkge_lut[i][j] = + cpu_to_be32(ni_scale_power_for_smc(smc_leakage, scaling_factor)); + } + + for (j = table_size; j < SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; j++) { + for (i = 0; i < SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES; i++) + cac_tables->cac_lkge_lut[i][j] = + cpu_to_be32(ni_scale_power_for_smc(max_leakage, scaling_factor)); + } + return 0; +} + +static int ni_initialize_smc_cac_tables(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + PP_NIslands_CACTABLES *cac_tables = NULL; + int i, ret; + u32 reg; + + if (ni_pi->enable_cac == false) + return 0; + + cac_tables = kzalloc(sizeof(PP_NIslands_CACTABLES), GFP_KERNEL); + if (!cac_tables) + return -ENOMEM; + + reg = RREG32(CG_CAC_CTRL) & ~(TID_CNT_MASK | TID_UNIT_MASK); + reg |= (TID_CNT(ni_pi->cac_weights->tid_cnt) | + TID_UNIT(ni_pi->cac_weights->tid_unit)); + WREG32(CG_CAC_CTRL, reg); + + for (i = 0; i < NISLANDS_DCCAC_MAX_LEVELS; i++) + ni_pi->dc_cac_table[i] = ni_pi->cac_weights->dc_cac[i]; + + for (i = 0; i < SMC_NISLANDS_BIF_LUT_NUM_OF_ENTRIES; i++) + cac_tables->cac_bif_lut[i] = ni_pi->cac_weights->pcie_cac[i]; + + ni_pi->cac_data.i_leakage = rdev->pm.dpm.cac_leakage; + ni_pi->cac_data.pwr_const = 0; + ni_pi->cac_data.dc_cac_value = ni_pi->dc_cac_table[NISLANDS_DCCAC_LEVEL_0]; + ni_pi->cac_data.bif_cac_value = 0; + ni_pi->cac_data.mc_wr_weight = ni_pi->cac_weights->mc_write_weight; + ni_pi->cac_data.mc_rd_weight = ni_pi->cac_weights->mc_read_weight; + ni_pi->cac_data.allow_ovrflw = 0; + ni_pi->cac_data.l2num_win_tdp = ni_pi->lta_window_size; + ni_pi->cac_data.num_win_tdp = 0; + ni_pi->cac_data.lts_truncate_n = ni_pi->lts_truncate; + + if (ni_pi->driver_calculate_cac_leakage) + ret = ni_init_driver_calculated_leakage_table(rdev, cac_tables); + else + ret = ni_init_simplified_leakage_table(rdev, cac_tables); + + if (ret) + goto done_free; + + cac_tables->pwr_const = cpu_to_be32(ni_pi->cac_data.pwr_const); + cac_tables->dc_cacValue = cpu_to_be32(ni_pi->cac_data.dc_cac_value); + cac_tables->bif_cacValue = cpu_to_be32(ni_pi->cac_data.bif_cac_value); + cac_tables->AllowOvrflw = ni_pi->cac_data.allow_ovrflw; + cac_tables->MCWrWeight = ni_pi->cac_data.mc_wr_weight; + cac_tables->MCRdWeight = ni_pi->cac_data.mc_rd_weight; + cac_tables->numWin_TDP = ni_pi->cac_data.num_win_tdp; + cac_tables->l2numWin_TDP = ni_pi->cac_data.l2num_win_tdp; + cac_tables->lts_truncate_n = ni_pi->cac_data.lts_truncate_n; + + ret = rv770_copy_bytes_to_smc(rdev, ni_pi->cac_table_start, (u8 *)cac_tables, + sizeof(PP_NIslands_CACTABLES), pi->sram_end); + +done_free: + if (ret) { + ni_pi->enable_cac = false; + ni_pi->enable_power_containment = false; + } + + kfree(cac_tables); + + return 0; +} + +static int ni_initialize_hardware_cac_manager(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + u32 reg; + + if (!ni_pi->enable_cac || + !ni_pi->cac_configuration_required) + return 0; + + if (ni_pi->cac_weights == NULL) + return -EINVAL; + + reg = RREG32_CG(CG_CAC_REGION_1_WEIGHT_0) & ~(WEIGHT_TCP_SIG0_MASK | + WEIGHT_TCP_SIG1_MASK | + WEIGHT_TA_SIG_MASK); + reg |= (WEIGHT_TCP_SIG0(ni_pi->cac_weights->weight_tcp_sig0) | + WEIGHT_TCP_SIG1(ni_pi->cac_weights->weight_tcp_sig1) | + WEIGHT_TA_SIG(ni_pi->cac_weights->weight_ta_sig)); + WREG32_CG(CG_CAC_REGION_1_WEIGHT_0, reg); + + reg = RREG32_CG(CG_CAC_REGION_1_WEIGHT_1) & ~(WEIGHT_TCC_EN0_MASK | + WEIGHT_TCC_EN1_MASK | + WEIGHT_TCC_EN2_MASK); + reg |= (WEIGHT_TCC_EN0(ni_pi->cac_weights->weight_tcc_en0) | + WEIGHT_TCC_EN1(ni_pi->cac_weights->weight_tcc_en1) | + WEIGHT_TCC_EN2(ni_pi->cac_weights->weight_tcc_en2)); + WREG32_CG(CG_CAC_REGION_1_WEIGHT_1, reg); + + reg = RREG32_CG(CG_CAC_REGION_2_WEIGHT_0) & ~(WEIGHT_CB_EN0_MASK | + WEIGHT_CB_EN1_MASK | + WEIGHT_CB_EN2_MASK | + WEIGHT_CB_EN3_MASK); + reg |= (WEIGHT_CB_EN0(ni_pi->cac_weights->weight_cb_en0) | + WEIGHT_CB_EN1(ni_pi->cac_weights->weight_cb_en1) | + WEIGHT_CB_EN2(ni_pi->cac_weights->weight_cb_en2) | + WEIGHT_CB_EN3(ni_pi->cac_weights->weight_cb_en3)); + WREG32_CG(CG_CAC_REGION_2_WEIGHT_0, reg); + + reg = RREG32_CG(CG_CAC_REGION_2_WEIGHT_1) & ~(WEIGHT_DB_SIG0_MASK | + WEIGHT_DB_SIG1_MASK | + WEIGHT_DB_SIG2_MASK | + WEIGHT_DB_SIG3_MASK); + reg |= (WEIGHT_DB_SIG0(ni_pi->cac_weights->weight_db_sig0) | + WEIGHT_DB_SIG1(ni_pi->cac_weights->weight_db_sig1) | + WEIGHT_DB_SIG2(ni_pi->cac_weights->weight_db_sig2) | + WEIGHT_DB_SIG3(ni_pi->cac_weights->weight_db_sig3)); + WREG32_CG(CG_CAC_REGION_2_WEIGHT_1, reg); + + reg = RREG32_CG(CG_CAC_REGION_2_WEIGHT_2) & ~(WEIGHT_SXM_SIG0_MASK | + WEIGHT_SXM_SIG1_MASK | + WEIGHT_SXM_SIG2_MASK | + WEIGHT_SXS_SIG0_MASK | + WEIGHT_SXS_SIG1_MASK); + reg |= (WEIGHT_SXM_SIG0(ni_pi->cac_weights->weight_sxm_sig0) | + WEIGHT_SXM_SIG1(ni_pi->cac_weights->weight_sxm_sig1) | + WEIGHT_SXM_SIG2(ni_pi->cac_weights->weight_sxm_sig2) | + WEIGHT_SXS_SIG0(ni_pi->cac_weights->weight_sxs_sig0) | + WEIGHT_SXS_SIG1(ni_pi->cac_weights->weight_sxs_sig1)); + WREG32_CG(CG_CAC_REGION_2_WEIGHT_2, reg); + + reg = RREG32_CG(CG_CAC_REGION_3_WEIGHT_0) & ~(WEIGHT_XBR_0_MASK | + WEIGHT_XBR_1_MASK | + WEIGHT_XBR_2_MASK | + WEIGHT_SPI_SIG0_MASK); + reg |= (WEIGHT_XBR_0(ni_pi->cac_weights->weight_xbr_0) | + WEIGHT_XBR_1(ni_pi->cac_weights->weight_xbr_1) | + WEIGHT_XBR_2(ni_pi->cac_weights->weight_xbr_2) | + WEIGHT_SPI_SIG0(ni_pi->cac_weights->weight_spi_sig0)); + WREG32_CG(CG_CAC_REGION_3_WEIGHT_0, reg); + + reg = RREG32_CG(CG_CAC_REGION_3_WEIGHT_1) & ~(WEIGHT_SPI_SIG1_MASK | + WEIGHT_SPI_SIG2_MASK | + WEIGHT_SPI_SIG3_MASK | + WEIGHT_SPI_SIG4_MASK | + WEIGHT_SPI_SIG5_MASK); + reg |= (WEIGHT_SPI_SIG1(ni_pi->cac_weights->weight_spi_sig1) | + WEIGHT_SPI_SIG2(ni_pi->cac_weights->weight_spi_sig2) | + WEIGHT_SPI_SIG3(ni_pi->cac_weights->weight_spi_sig3) | + WEIGHT_SPI_SIG4(ni_pi->cac_weights->weight_spi_sig4) | + WEIGHT_SPI_SIG5(ni_pi->cac_weights->weight_spi_sig5)); + WREG32_CG(CG_CAC_REGION_3_WEIGHT_1, reg); + + reg = RREG32_CG(CG_CAC_REGION_4_WEIGHT_0) & ~(WEIGHT_LDS_SIG0_MASK | + WEIGHT_LDS_SIG1_MASK | + WEIGHT_SC_MASK); + reg |= (WEIGHT_LDS_SIG0(ni_pi->cac_weights->weight_lds_sig0) | + WEIGHT_LDS_SIG1(ni_pi->cac_weights->weight_lds_sig1) | + WEIGHT_SC(ni_pi->cac_weights->weight_sc)); + WREG32_CG(CG_CAC_REGION_4_WEIGHT_0, reg); + + reg = RREG32_CG(CG_CAC_REGION_4_WEIGHT_1) & ~(WEIGHT_BIF_MASK | + WEIGHT_CP_MASK | + WEIGHT_PA_SIG0_MASK | + WEIGHT_PA_SIG1_MASK | + WEIGHT_VGT_SIG0_MASK); + reg |= (WEIGHT_BIF(ni_pi->cac_weights->weight_bif) | + WEIGHT_CP(ni_pi->cac_weights->weight_cp) | + WEIGHT_PA_SIG0(ni_pi->cac_weights->weight_pa_sig0) | + WEIGHT_PA_SIG1(ni_pi->cac_weights->weight_pa_sig1) | + WEIGHT_VGT_SIG0(ni_pi->cac_weights->weight_vgt_sig0)); + WREG32_CG(CG_CAC_REGION_4_WEIGHT_1, reg); + + reg = RREG32_CG(CG_CAC_REGION_4_WEIGHT_2) & ~(WEIGHT_VGT_SIG1_MASK | + WEIGHT_VGT_SIG2_MASK | + WEIGHT_DC_SIG0_MASK | + WEIGHT_DC_SIG1_MASK | + WEIGHT_DC_SIG2_MASK); + reg |= (WEIGHT_VGT_SIG1(ni_pi->cac_weights->weight_vgt_sig1) | + WEIGHT_VGT_SIG2(ni_pi->cac_weights->weight_vgt_sig2) | + WEIGHT_DC_SIG0(ni_pi->cac_weights->weight_dc_sig0) | + WEIGHT_DC_SIG1(ni_pi->cac_weights->weight_dc_sig1) | + WEIGHT_DC_SIG2(ni_pi->cac_weights->weight_dc_sig2)); + WREG32_CG(CG_CAC_REGION_4_WEIGHT_2, reg); + + reg = RREG32_CG(CG_CAC_REGION_4_WEIGHT_3) & ~(WEIGHT_DC_SIG3_MASK | + WEIGHT_UVD_SIG0_MASK | + WEIGHT_UVD_SIG1_MASK | + WEIGHT_SPARE0_MASK | + WEIGHT_SPARE1_MASK); + reg |= (WEIGHT_DC_SIG3(ni_pi->cac_weights->weight_dc_sig3) | + WEIGHT_UVD_SIG0(ni_pi->cac_weights->weight_uvd_sig0) | + WEIGHT_UVD_SIG1(ni_pi->cac_weights->weight_uvd_sig1) | + WEIGHT_SPARE0(ni_pi->cac_weights->weight_spare0) | + WEIGHT_SPARE1(ni_pi->cac_weights->weight_spare1)); + WREG32_CG(CG_CAC_REGION_4_WEIGHT_3, reg); + + reg = RREG32_CG(CG_CAC_REGION_5_WEIGHT_0) & ~(WEIGHT_SQ_VSP_MASK | + WEIGHT_SQ_VSP0_MASK); + reg |= (WEIGHT_SQ_VSP(ni_pi->cac_weights->weight_sq_vsp) | + WEIGHT_SQ_VSP0(ni_pi->cac_weights->weight_sq_vsp0)); + WREG32_CG(CG_CAC_REGION_5_WEIGHT_0, reg); + + reg = RREG32_CG(CG_CAC_REGION_5_WEIGHT_1) & ~(WEIGHT_SQ_GPR_MASK); + reg |= WEIGHT_SQ_GPR(ni_pi->cac_weights->weight_sq_gpr); + WREG32_CG(CG_CAC_REGION_5_WEIGHT_1, reg); + + reg = RREG32_CG(CG_CAC_REGION_4_OVERRIDE_4) & ~(OVR_MODE_SPARE_0_MASK | + OVR_VAL_SPARE_0_MASK | + OVR_MODE_SPARE_1_MASK | + OVR_VAL_SPARE_1_MASK); + reg |= (OVR_MODE_SPARE_0(ni_pi->cac_weights->ovr_mode_spare_0) | + OVR_VAL_SPARE_0(ni_pi->cac_weights->ovr_val_spare_0) | + OVR_MODE_SPARE_1(ni_pi->cac_weights->ovr_mode_spare_1) | + OVR_VAL_SPARE_1(ni_pi->cac_weights->ovr_val_spare_1)); + WREG32_CG(CG_CAC_REGION_4_OVERRIDE_4, reg); + + reg = RREG32(SQ_CAC_THRESHOLD) & ~(VSP_MASK | + VSP0_MASK | + GPR_MASK); + reg |= (VSP(ni_pi->cac_weights->vsp) | + VSP0(ni_pi->cac_weights->vsp0) | + GPR(ni_pi->cac_weights->gpr)); + WREG32(SQ_CAC_THRESHOLD, reg); + + reg = (MCDW_WR_ENABLE | + MCDX_WR_ENABLE | + MCDY_WR_ENABLE | + MCDZ_WR_ENABLE | + INDEX(0x09D4)); + WREG32(MC_CG_CONFIG, reg); + + reg = (READ_WEIGHT(ni_pi->cac_weights->mc_read_weight) | + WRITE_WEIGHT(ni_pi->cac_weights->mc_write_weight) | + ALLOW_OVERFLOW); + WREG32(MC_CG_DATAPORT, reg); + + return 0; +} + +static int ni_enable_smc_cac(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + bool enable) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + int ret = 0; + PPSMC_Result smc_result; + + if (ni_pi->enable_cac) { + if (enable) { + if (!r600_is_uvd_state(radeon_new_state->class, radeon_new_state->class2)) { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_CollectCAC_PowerCorreln); + + if (ni_pi->support_cac_long_term_average) { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_CACLongTermAvgEnable); + if (PPSMC_Result_OK != smc_result) + ni_pi->support_cac_long_term_average = false; + } + + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableCac); + if (PPSMC_Result_OK != smc_result) + ret = -EINVAL; + + ni_pi->cac_enabled = (PPSMC_Result_OK == smc_result) ? true : false; + } + } else if (ni_pi->cac_enabled) { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_DisableCac); + + ni_pi->cac_enabled = false; + + if (ni_pi->support_cac_long_term_average) { + smc_result = rv770_send_msg_to_smc(rdev, PPSMC_CACLongTermAvgDisable); + if (PPSMC_Result_OK != smc_result) + ni_pi->support_cac_long_term_average = false; + } + } + } + + return ret; +} + +static int ni_pcie_performance_request(struct radeon_device *rdev, + u8 perf_req, bool advertise) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + +#if defined(CONFIG_ACPI) + if ((perf_req == PCIE_PERF_REQ_PECI_GEN1) || + (perf_req == PCIE_PERF_REQ_PECI_GEN2)) { + if (eg_pi->pcie_performance_request_registered == false) + radeon_acpi_pcie_notify_device_ready(rdev); + eg_pi->pcie_performance_request_registered = true; + return radeon_acpi_pcie_performance_request(rdev, perf_req, advertise); + } else if ((perf_req == PCIE_PERF_REQ_REMOVE_REGISTRY) && + eg_pi->pcie_performance_request_registered) { + eg_pi->pcie_performance_request_registered = false; + return radeon_acpi_pcie_performance_request(rdev, perf_req, advertise); + } +#endif + return 0; +} + +static int ni_advertise_gen2_capability(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) + pi->pcie_gen2 = true; + else + pi->pcie_gen2 = false; + + if (!pi->pcie_gen2) + ni_pcie_performance_request(rdev, PCIE_PERF_REQ_PECI_GEN2, true); + + return 0; +} + +static void ni_enable_bif_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp, bif; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + if (enable) { + if (!pi->boot_in_gen2) { + bif = RREG32(CG_BIF_REQ_AND_RSP) & ~CG_CLIENT_REQ_MASK; + bif |= CG_CLIENT_REQ(0xd); + WREG32(CG_BIF_REQ_AND_RSP, bif); + } + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp |= LC_HW_VOLTAGE_IF_CONTROL(1); + tmp |= LC_GEN2_EN_STRAP; + + tmp |= LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + DRM_UDELAY(10); + tmp &= ~LC_CLR_FAILED_SPD_CHANGE_CNT; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } else { + if (!pi->boot_in_gen2) { + bif = RREG32(CG_BIF_REQ_AND_RSP) & ~CG_CLIENT_REQ_MASK; + bif |= CG_CLIENT_REQ(0xd); + WREG32(CG_BIF_REQ_AND_RSP, bif); + + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp &= ~LC_GEN2_EN_STRAP; + } + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } + } +} + +static void ni_enable_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + ni_enable_bif_dynamic_pcie_gen2(rdev, enable); + + if (enable) + WREG32_P(GENERAL_PWRMGT, ENABLE_GEN2PCIE, ~ENABLE_GEN2PCIE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~ENABLE_GEN2PCIE); +} + +void ni_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct ni_ps *new_state = ni_get_ps(new_ps); + struct ni_ps *current_state = ni_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->performance_levels[new_state->performance_level_count - 1].sclk >= + current_state->performance_levels[current_state->performance_level_count - 1].sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +void ni_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct ni_ps *new_state = ni_get_ps(new_ps); + struct ni_ps *current_state = ni_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->performance_levels[new_state->performance_level_count - 1].sclk < + current_state->performance_levels[current_state->performance_level_count - 1].sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +void ni_dpm_setup_asic(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + ni_read_clock_registers(rdev); + btc_read_arb_registers(rdev); + rv770_get_memory_type(rdev); + if (eg_pi->pcie_performance_request) + ni_advertise_gen2_capability(rdev); + rv770_get_pcie_gen2_status(rdev); + rv770_enable_acpi_pm(rdev); +} + +void ni_update_current_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct ni_ps *new_ps = ni_get_ps(rps); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + + eg_pi->current_rps = *rps; + ni_pi->current_ps = *new_ps; + eg_pi->current_rps.ps_priv = &ni_pi->current_ps; +} + +void ni_update_requested_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct ni_ps *new_ps = ni_get_ps(rps); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + + eg_pi->requested_rps = *rps; + ni_pi->requested_ps = *new_ps; + eg_pi->requested_rps.ps_priv = &ni_pi->requested_ps; +} + +int ni_dpm_enable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (pi->gfx_clock_gating) + ni_cg_clockgating_default(rdev); + if (btc_dpm_enabled(rdev)) + return -EINVAL; + if (pi->mg_clock_gating) + ni_mg_clockgating_default(rdev); + if (eg_pi->ls_clock_gating) + ni_ls_clockgating_default(rdev); + if (pi->voltage_control) { + rv770_enable_voltage_control(rdev, true); + ret = cypress_construct_voltage_tables(rdev); + if (ret) { + DRM_ERROR("cypress_construct_voltage_tables failed\n"); + return ret; + } + } + if (eg_pi->dynamic_ac_timing) { + ret = ni_initialize_mc_reg_table(rdev); + if (ret) + eg_pi->dynamic_ac_timing = false; + } + if (pi->dynamic_ss) + cypress_enable_spread_spectrum(rdev, true); + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, true); + rv770_setup_bsp(rdev); + rv770_program_git(rdev); + rv770_program_tp(rdev); + rv770_program_tpp(rdev); + rv770_program_sstp(rdev); + cypress_enable_display_gap(rdev); + rv770_program_vc(rdev); + if (pi->dynamic_pcie_gen2) + ni_enable_dynamic_pcie_gen2(rdev, true); + ret = rv770_upload_firmware(rdev); + if (ret) { + DRM_ERROR("rv770_upload_firmware failed\n"); + return ret; + } + ret = ni_process_firmware_header(rdev); + if (ret) { + DRM_ERROR("ni_process_firmware_header failed\n"); + return ret; + } + ret = ni_initial_switch_from_arb_f0_to_f1(rdev); + if (ret) { + DRM_ERROR("ni_initial_switch_from_arb_f0_to_f1 failed\n"); + return ret; + } + ret = ni_init_smc_table(rdev); + if (ret) { + DRM_ERROR("ni_init_smc_table failed\n"); + return ret; + } + ret = ni_init_smc_spll_table(rdev); + if (ret) { + DRM_ERROR("ni_init_smc_spll_table failed\n"); + return ret; + } + ret = ni_init_arb_table_index(rdev); + if (ret) { + DRM_ERROR("ni_init_arb_table_index failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = ni_populate_mc_reg_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("ni_populate_mc_reg_table failed\n"); + return ret; + } + } + ret = ni_initialize_smc_cac_tables(rdev); + if (ret) { + DRM_ERROR("ni_initialize_smc_cac_tables failed\n"); + return ret; + } + ret = ni_initialize_hardware_cac_manager(rdev); + if (ret) { + DRM_ERROR("ni_initialize_hardware_cac_manager failed\n"); + return ret; + } + ret = ni_populate_smc_tdp_limits(rdev, boot_ps); + if (ret) { + DRM_ERROR("ni_populate_smc_tdp_limits failed\n"); + return ret; + } + ni_program_response_times(rdev); + r7xx_start_smc(rdev); + ret = cypress_notify_smc_display_change(rdev, false); + if (ret) { + DRM_ERROR("cypress_notify_smc_display_change failed\n"); + return ret; + } + cypress_enable_sclk_control(rdev, true); + if (eg_pi->memory_transition) + cypress_enable_mclk_control(rdev, true); + cypress_start_dpm(rdev); + if (pi->gfx_clock_gating) + ni_gfx_clockgating_enable(rdev, true); + if (pi->mg_clock_gating) + ni_mg_clockgating_enable(rdev, true); + if (eg_pi->ls_clock_gating) + ni_ls_clockgating_enable(rdev, true); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + PPSMC_Result result; + + ret = rv770_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, 0xff * 1000); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + + if (result != PPSMC_Result_OK) + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + } + + rv770_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + ni_update_current_ps(rdev, boot_ps); + + return 0; +} + +void ni_dpm_disable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + + if (!btc_dpm_enabled(rdev)) + return; + rv770_clear_vc(rdev); + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, false); + ni_enable_power_containment(rdev, boot_ps, false); + ni_enable_smc_cac(rdev, boot_ps, false); + cypress_enable_spread_spectrum(rdev, false); + rv770_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, false); + if (pi->dynamic_pcie_gen2) + ni_enable_dynamic_pcie_gen2(rdev, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + if (pi->gfx_clock_gating) + ni_gfx_clockgating_enable(rdev, false); + if (pi->mg_clock_gating) + ni_mg_clockgating_enable(rdev, false); + if (eg_pi->ls_clock_gating) + ni_ls_clockgating_enable(rdev, false); + ni_stop_dpm(rdev); + btc_reset_to_default(rdev); + ni_stop_smc(rdev); + ni_force_switch_to_arb_f0(rdev); + + ni_update_current_ps(rdev, boot_ps); +} + +static int ni_power_control_set_level(struct radeon_device *rdev) +{ + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + int ret; + + ret = ni_restrict_performance_levels_before_switch(rdev); + if (ret) + return ret; + ret = rv770_halt_smc(rdev); + if (ret) + return ret; + ret = ni_populate_smc_tdp_limits(rdev, new_ps); + if (ret) + return ret; + ret = rv770_resume_smc(rdev); + if (ret) + return ret; + ret = rv770_set_sw_state(rdev); + if (ret) + return ret; + + return 0; +} + +int ni_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps; + struct radeon_ps *new_ps = &requested_ps; + + ni_update_requested_ps(rdev, new_ps); + + ni_apply_state_adjust_rules(rdev, &eg_pi->requested_rps); + + return 0; +} + +int ni_dpm_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + struct radeon_ps *old_ps = &eg_pi->current_rps; + int ret; + + ret = ni_restrict_performance_levels_before_switch(rdev); + if (ret) { + DRM_ERROR("ni_restrict_performance_levels_before_switch failed\n"); + return ret; + } + ni_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + ret = ni_enable_power_containment(rdev, new_ps, false); + if (ret) { + DRM_ERROR("ni_enable_power_containment failed\n"); + return ret; + } + ret = ni_enable_smc_cac(rdev, new_ps, false); + if (ret) { + DRM_ERROR("ni_enable_smc_cac failed\n"); + return ret; + } + ret = rv770_halt_smc(rdev); + if (ret) { + DRM_ERROR("rv770_halt_smc failed\n"); + return ret; + } + if (eg_pi->smu_uvd_hs) + btc_notify_uvd_to_smc(rdev, new_ps); + ret = ni_upload_sw_state(rdev, new_ps); + if (ret) { + DRM_ERROR("ni_upload_sw_state failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = ni_upload_mc_reg_table(rdev, new_ps); + if (ret) { + DRM_ERROR("ni_upload_mc_reg_table failed\n"); + return ret; + } + } + ret = ni_program_memory_timing_parameters(rdev, new_ps); + if (ret) { + DRM_ERROR("ni_program_memory_timing_parameters failed\n"); + return ret; + } + ret = rv770_resume_smc(rdev); + if (ret) { + DRM_ERROR("rv770_resume_smc failed\n"); + return ret; + } + ret = rv770_set_sw_state(rdev); + if (ret) { + DRM_ERROR("rv770_set_sw_state failed\n"); + return ret; + } + ni_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + ret = ni_enable_smc_cac(rdev, new_ps, true); + if (ret) { + DRM_ERROR("ni_enable_smc_cac failed\n"); + return ret; + } + ret = ni_enable_power_containment(rdev, new_ps, true); + if (ret) { + DRM_ERROR("ni_enable_power_containment failed\n"); + return ret; + } + + /* update tdp */ + ret = ni_power_control_set_level(rdev); + if (ret) { + DRM_ERROR("ni_power_control_set_level failed\n"); + return ret; + } + + ret = ni_dpm_force_performance_level(rdev, RADEON_DPM_FORCED_LEVEL_AUTO); + if (ret) { + DRM_ERROR("ni_dpm_force_performance_level failed\n"); + return ret; + } + + return 0; +} + +void ni_dpm_post_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + + ni_update_current_ps(rdev, new_ps); +} + +void ni_dpm_reset_asic(struct radeon_device *rdev) +{ + ni_restrict_performance_levels_before_switch(rdev); + rv770_set_boot_state(rdev); +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void ni_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else if (r600_is_uvd_state(rps->class, rps->class2)) { + rps->vclk = RV770_DEFAULT_VCLK_FREQ; + rps->dclk = RV770_DEFAULT_DCLK_FREQ; + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) + rdev->pm.dpm.boot_ps = rps; + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void ni_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_ps *ps = ni_get_ps(rps); + u16 vddc; + struct rv7xx_pl *pl = &ps->performance_levels[index]; + + ps->performance_level_count = index + 1; + + pl->sclk = le16_to_cpu(clock_info->evergreen.usEngineClockLow); + pl->sclk |= clock_info->evergreen.ucEngineClockHigh << 16; + pl->mclk = le16_to_cpu(clock_info->evergreen.usMemoryClockLow); + pl->mclk |= clock_info->evergreen.ucMemoryClockHigh << 16; + + pl->vddc = le16_to_cpu(clock_info->evergreen.usVDDC); + pl->vddci = le16_to_cpu(clock_info->evergreen.usVDDCI); + pl->flags = le32_to_cpu(clock_info->evergreen.ulFlags); + + /* patch up vddc if necessary */ + if (pl->vddc == 0xff01) { + if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc) == 0) + pl->vddc = vddc; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) { + pi->acpi_vddc = pl->vddc; + eg_pi->acpi_vddci = pl->vddci; + if (ps->performance_levels[0].flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + pi->acpi_pcie_gen2 = true; + else + pi->acpi_pcie_gen2 = false; + } + + if (rps->class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) { + eg_pi->ulv.supported = true; + eg_pi->ulv.pl = pl; + } + + if (pi->min_vddc_in_table > pl->vddc) + pi->min_vddc_in_table = pl->vddc; + + if (pi->max_vddc_in_table < pl->vddc) + pi->max_vddc_in_table = pl->vddc; + + /* patch up boot state */ + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + u16 vddc, vddci, mvdd; + radeon_atombios_get_default_voltages(rdev, &vddc, &vddci, &mvdd); + pl->mclk = rdev->clock.default_mclk; + pl->sclk = rdev->clock.default_sclk; + pl->vddc = vddc; + pl->vddci = vddci; + } + + if ((rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == + ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE) { + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk = pl->sclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.mclk = pl->mclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddc = pl->vddc; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddci = pl->vddci; + } +} + +static int ni_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j; + union pplib_clock_info *clock_info; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + struct ni_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + power_info->pplib.ucNumStates, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + + for (i = 0; i < power_info->pplib.ucNumStates; i++) { + power_state = (union pplib_power_state *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset) + + i * power_info->pplib.ucStateEntrySize); + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset) + + (power_state->v1.ucNonClockStateIndex * + power_info->pplib.ucNonClockSize)); + if (power_info->pplib.ucStateEntrySize - 1) { + ps = kzalloc(sizeof(struct ni_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + ni_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + power_info->pplib.ucNonClockSize); + for (j = 0; j < (power_info->pplib.ucStateEntrySize - 1); j++) { + clock_info = (union pplib_clock_info *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset) + + (power_state->v1.ucClockStateIndices[j] * + power_info->pplib.ucClockInfoSize)); + ni_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], j, + clock_info); + } + } + } + rdev->pm.dpm.num_ps = power_info->pplib.ucNumStates; + return 0; +} + +int ni_dpm_init(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi; + struct evergreen_power_info *eg_pi; + struct ni_power_info *ni_pi; + struct atom_clock_dividers dividers; + int ret; + + ni_pi = kzalloc(sizeof(struct ni_power_info), GFP_KERNEL); + if (ni_pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = ni_pi; + eg_pi = &ni_pi->eg; + pi = &eg_pi->rv7xx; + + rv770_get_max_vddc(rdev); + + eg_pi->ulv.supported = false; + pi->acpi_vddc = 0; + eg_pi->acpi_vddci = 0; + pi->min_vddc_in_table = 0; + pi->max_vddc_in_table = 0; + + ret = ni_parse_power_table(rdev); + if (ret) + return ret; + ret = r600_parse_extended_power_table(rdev); + if (ret) + return ret; + + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries = + kzalloc(4 * sizeof(struct radeon_clock_voltage_dependency_entry), GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries) { + r600_free_extended_power_table(rdev); + return -ENOMEM; + } + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.count = 4; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].clk = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].v = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].clk = 36000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].v = 720; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].clk = 54000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].v = 810; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].clk = 72000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].v = 900; + + ni_patch_dependency_tables_based_on_leakage(rdev); + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->ref_div = dividers.ref_div + 1; + else + pi->ref_div = R600_REFERENCEDIVIDER_DFLT; + + pi->rlp = RV770_RLP_DFLT; + pi->rmp = RV770_RMP_DFLT; + pi->lhp = RV770_LHP_DFLT; + pi->lmp = RV770_LMP_DFLT; + + eg_pi->ats[0].rlp = RV770_RLP_DFLT; + eg_pi->ats[0].rmp = RV770_RMP_DFLT; + eg_pi->ats[0].lhp = RV770_LHP_DFLT; + eg_pi->ats[0].lmp = RV770_LMP_DFLT; + + eg_pi->ats[1].rlp = BTC_RLP_UVD_DFLT; + eg_pi->ats[1].rmp = BTC_RMP_UVD_DFLT; + eg_pi->ats[1].lhp = BTC_LHP_UVD_DFLT; + eg_pi->ats[1].lmp = BTC_LMP_UVD_DFLT; + + eg_pi->smu_uvd_hs = true; + + if (rdev->ddev->pci_device == 0x6707) { + pi->mclk_strobe_mode_threshold = 55000; + pi->mclk_edc_enable_threshold = 55000; + eg_pi->mclk_edc_wr_enable_threshold = 55000; + } else { + pi->mclk_strobe_mode_threshold = 40000; + pi->mclk_edc_enable_threshold = 40000; + eg_pi->mclk_edc_wr_enable_threshold = 40000; + } + ni_pi->mclk_rtt_mode_threshold = eg_pi->mclk_edc_wr_enable_threshold; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0); + + pi->mvdd_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_MVDDC, 0); + + eg_pi->vddci_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDCI, 0); + + rv770_get_engine_memory_ss(rdev); + + pi->asi = RV770_ASI_DFLT; + pi->pasi = CYPRESS_HASI_DFLT; + pi->vrc = CYPRESS_VRC_DFLT; + + pi->power_gating = false; + + pi->gfx_clock_gating = true; + + pi->mg_clock_gating = true; + pi->mgcgtssm = true; + eg_pi->ls_clock_gating = false; + eg_pi->sclk_deep_sleep = false; + + pi->dynamic_pcie_gen2 = true; + + if (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + pi->display_gap = true; + + pi->dcodt = true; + + pi->ulps = true; + + eg_pi->dynamic_ac_timing = true; + eg_pi->abm = true; + eg_pi->mcls = true; + eg_pi->light_sleep = true; + eg_pi->memory_transition = true; +#if defined(CONFIG_ACPI) + eg_pi->pcie_performance_request = + radeon_acpi_is_pcie_performance_request_supported(rdev); +#else + eg_pi->pcie_performance_request = false; +#endif + + eg_pi->dll_default_on = false; + + eg_pi->sclk_deep_sleep = false; + + pi->mclk_stutter_mode_threshold = 0; + + pi->sram_end = SMC_RAM_END; + + rdev->pm.dpm.dyn_state.mclk_sclk_ratio = 3; + rdev->pm.dpm.dyn_state.vddc_vddci_delta = 200; + rdev->pm.dpm.dyn_state.min_vddc_for_pcie_gen2 = 900; + rdev->pm.dpm.dyn_state.valid_sclk_values.count = ARRAY_SIZE(btc_valid_sclk); + rdev->pm.dpm.dyn_state.valid_sclk_values.values = btc_valid_sclk; + rdev->pm.dpm.dyn_state.valid_mclk_values.count = 0; + rdev->pm.dpm.dyn_state.valid_mclk_values.values = NULL; + rdev->pm.dpm.dyn_state.sclk_mclk_delta = 12500; + + ni_pi->cac_data.leakage_coefficients.at = 516; + ni_pi->cac_data.leakage_coefficients.bt = 18; + ni_pi->cac_data.leakage_coefficients.av = 51; + ni_pi->cac_data.leakage_coefficients.bv = 2957; + + switch (rdev->ddev->pci_device) { + case 0x6700: + case 0x6701: + case 0x6702: + case 0x6703: + case 0x6718: + ni_pi->cac_weights = &cac_weights_cayman_xt; + break; + case 0x6705: + case 0x6719: + case 0x671D: + case 0x671C: + default: + ni_pi->cac_weights = &cac_weights_cayman_pro; + break; + case 0x6704: + case 0x6706: + case 0x6707: + case 0x6708: + case 0x6709: + ni_pi->cac_weights = &cac_weights_cayman_le; + break; + } + + if (ni_pi->cac_weights->enable_power_containment_by_default) { + ni_pi->enable_power_containment = true; + ni_pi->enable_cac = true; + ni_pi->enable_sq_ramping = true; + } else { + ni_pi->enable_power_containment = false; + ni_pi->enable_cac = false; + ni_pi->enable_sq_ramping = false; + } + + ni_pi->driver_calculate_cac_leakage = false; + ni_pi->cac_configuration_required = true; + + if (ni_pi->cac_configuration_required) { + ni_pi->support_cac_long_term_average = true; + ni_pi->lta_window_size = ni_pi->cac_weights->l2_lta_window_size; + ni_pi->lts_truncate = ni_pi->cac_weights->lts_truncate; + } else { + ni_pi->support_cac_long_term_average = false; + ni_pi->lta_window_size = 0; + ni_pi->lts_truncate = 0; + } + + ni_pi->use_power_boost_limit = true; + + return 0; +} + +void ni_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries); + r600_free_extended_power_table(rdev); +} + +void ni_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct ni_ps *ps = ni_get_ps(rps); + struct rv7xx_pl *pl; + int i; + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + for (i = 0; i < ps->performance_level_count; i++) { + pl = &ps->performance_levels[i]; + if (rdev->family >= CHIP_TAHITI) + printk("\t\tpower level %d sclk: %u mclk: %u vddc: %u vddci: %u pcie gen: %u\n", + i, pl->sclk, pl->mclk, pl->vddc, pl->vddci, pl->pcie_gen + 1); + else + printk("\t\tpower level %d sclk: %u mclk: %u vddc: %u vddci: %u\n", + i, pl->sclk, pl->mclk, pl->vddc, pl->vddci); + } + r600_dpm_print_ps_status(rdev, rps); +} + +void ni_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct ni_ps *ps = ni_get_ps(rps); + struct rv7xx_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_STATE_INDEX_MASK) >> + CURRENT_STATE_INDEX_SHIFT; + + if (current_index >= ps->performance_level_count) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + pl = &ps->performance_levels[current_index]; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u mclk: %u vddc: %u vddci: %u\n", + current_index, pl->sclk, pl->mclk, pl->vddc, pl->vddci); + } +} + +u32 ni_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_ps *requested_state = ni_get_ps(&eg_pi->requested_rps); + + if (low) + return requested_state->performance_levels[0].sclk; + else + return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; +} + +u32 ni_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_ps *requested_state = ni_get_ps(&eg_pi->requested_rps); + + if (low) + return requested_state->performance_levels[0].mclk; + else + return requested_state->performance_levels[requested_state->performance_level_count - 1].mclk; +} diff --git a/sys/dev/drm/radeon/ni_dpm.h b/sys/dev/drm/radeon/ni_dpm.h new file mode 100644 index 0000000000..6bbee91809 --- /dev/null +++ b/sys/dev/drm/radeon/ni_dpm.h @@ -0,0 +1,250 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __NI_DPM_H__ +#define __NI_DPM_H__ + +#include "cypress_dpm.h" +#include "btc_dpm.h" +#include "nislands_smc.h" + +struct ni_clock_registers { + u32 cg_spll_func_cntl; + u32 cg_spll_func_cntl_2; + u32 cg_spll_func_cntl_3; + u32 cg_spll_func_cntl_4; + u32 cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2; + u32 mclk_pwrmgt_cntl; + u32 dll_cntl; + u32 mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2; + u32 mpll_ss1; + u32 mpll_ss2; +}; + +struct ni_mc_reg_entry { + u32 mclk_max; + u32 mc_data[SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +struct ni_mc_reg_table { + u8 last; + u8 num_entries; + u16 valid_flag; + struct ni_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES]; + SMC_NIslands_MCRegisterAddress mc_reg_address[SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +#define NISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT 2 + +enum ni_dc_cac_level +{ + NISLANDS_DCCAC_LEVEL_0 = 0, + NISLANDS_DCCAC_LEVEL_1, + NISLANDS_DCCAC_LEVEL_2, + NISLANDS_DCCAC_LEVEL_3, + NISLANDS_DCCAC_LEVEL_4, + NISLANDS_DCCAC_LEVEL_5, + NISLANDS_DCCAC_LEVEL_6, + NISLANDS_DCCAC_LEVEL_7, + NISLANDS_DCCAC_MAX_LEVELS +}; + +struct ni_leakage_coeffients +{ + u32 at; + u32 bt; + u32 av; + u32 bv; + s32 t_slope; + s32 t_intercept; + u32 t_ref; +}; + +struct ni_cac_data +{ + struct ni_leakage_coeffients leakage_coefficients; + u32 i_leakage; + s32 leakage_minimum_temperature; + u32 pwr_const; + u32 dc_cac_value; + u32 bif_cac_value; + u32 lkge_pwr; + u8 mc_wr_weight; + u8 mc_rd_weight; + u8 allow_ovrflw; + u8 num_win_tdp; + u8 l2num_win_tdp; + u8 lts_truncate_n; +}; + +struct ni_cac_weights +{ + u32 weight_tcp_sig0; + u32 weight_tcp_sig1; + u32 weight_ta_sig; + u32 weight_tcc_en0; + u32 weight_tcc_en1; + u32 weight_tcc_en2; + u32 weight_cb_en0; + u32 weight_cb_en1; + u32 weight_cb_en2; + u32 weight_cb_en3; + u32 weight_db_sig0; + u32 weight_db_sig1; + u32 weight_db_sig2; + u32 weight_db_sig3; + u32 weight_sxm_sig0; + u32 weight_sxm_sig1; + u32 weight_sxm_sig2; + u32 weight_sxs_sig0; + u32 weight_sxs_sig1; + u32 weight_xbr_0; + u32 weight_xbr_1; + u32 weight_xbr_2; + u32 weight_spi_sig0; + u32 weight_spi_sig1; + u32 weight_spi_sig2; + u32 weight_spi_sig3; + u32 weight_spi_sig4; + u32 weight_spi_sig5; + u32 weight_lds_sig0; + u32 weight_lds_sig1; + u32 weight_sc; + u32 weight_bif; + u32 weight_cp; + u32 weight_pa_sig0; + u32 weight_pa_sig1; + u32 weight_vgt_sig0; + u32 weight_vgt_sig1; + u32 weight_vgt_sig2; + u32 weight_dc_sig0; + u32 weight_dc_sig1; + u32 weight_dc_sig2; + u32 weight_dc_sig3; + u32 weight_uvd_sig0; + u32 weight_uvd_sig1; + u32 weight_spare0; + u32 weight_spare1; + u32 weight_sq_vsp; + u32 weight_sq_vsp0; + u32 weight_sq_gpr; + u32 ovr_mode_spare_0; + u32 ovr_val_spare_0; + u32 ovr_mode_spare_1; + u32 ovr_val_spare_1; + u32 vsp; + u32 vsp0; + u32 gpr; + u8 mc_read_weight; + u8 mc_write_weight; + u32 tid_cnt; + u32 tid_unit; + u32 l2_lta_window_size; + u32 lts_truncate; + u32 dc_cac[NISLANDS_DCCAC_MAX_LEVELS]; + u32 pcie_cac[SMC_NISLANDS_BIF_LUT_NUM_OF_ENTRIES]; + bool enable_power_containment_by_default; +}; + +struct ni_ps { + u16 performance_level_count; + bool dc_compatible; + struct rv7xx_pl performance_levels[NISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE]; +}; + +struct ni_power_info { + /* must be first! */ + struct evergreen_power_info eg; + struct ni_clock_registers clock_registers; + struct ni_mc_reg_table mc_reg_table; + u32 mclk_rtt_mode_threshold; + /* flags */ + bool use_power_boost_limit; + bool support_cac_long_term_average; + bool cac_enabled; + bool cac_configuration_required; + bool driver_calculate_cac_leakage; + bool pc_enabled; + bool enable_power_containment; + bool enable_cac; + bool enable_sq_ramping; + /* smc offsets */ + u16 arb_table_start; + u16 fan_table_start; + u16 cac_table_start; + u16 spll_table_start; + /* CAC stuff */ + struct ni_cac_data cac_data; + u32 dc_cac_table[NISLANDS_DCCAC_MAX_LEVELS]; + const struct ni_cac_weights *cac_weights; + u8 lta_window_size; + u8 lts_truncate; + struct ni_ps current_ps; + struct ni_ps requested_ps; + /* scratch structs */ + SMC_NIslands_MCRegisters smc_mc_reg_table; + NISLANDS_SMC_STATETABLE smc_statetable; +}; + +#define NISLANDS_INITIAL_STATE_ARB_INDEX 0 +#define NISLANDS_ACPI_STATE_ARB_INDEX 1 +#define NISLANDS_ULV_STATE_ARB_INDEX 2 +#define NISLANDS_DRIVER_STATE_ARB_INDEX 3 + +#define NISLANDS_DPM2_MAX_PULSE_SKIP 256 + +#define NISLANDS_DPM2_NEAR_TDP_DEC 10 +#define NISLANDS_DPM2_ABOVE_SAFE_INC 5 +#define NISLANDS_DPM2_BELOW_SAFE_INC 20 + +#define NISLANDS_DPM2_TDP_SAFE_LIMIT_PERCENT 80 + +#define NISLANDS_DPM2_MAXPS_PERCENT_H 90 +#define NISLANDS_DPM2_MAXPS_PERCENT_M 0 + +#define NISLANDS_DPM2_SQ_RAMP_MAX_POWER 0x3FFF +#define NISLANDS_DPM2_SQ_RAMP_MIN_POWER 0x12 +#define NISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA 0x15 +#define NISLANDS_DPM2_SQ_RAMP_STI_SIZE 0x1E +#define NISLANDS_DPM2_SQ_RAMP_LTI_RATIO 0xF + +int ni_copy_and_switch_arb_sets(struct radeon_device *rdev, + u32 arb_freq_src, u32 arb_freq_dest); +void ni_update_current_ps(struct radeon_device *rdev, + struct radeon_ps *rps); +void ni_update_requested_ps(struct radeon_device *rdev, + struct radeon_ps *rps); + +void ni_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps); +void ni_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps); + +bool ni_dpm_vblank_too_short(struct radeon_device *rdev); + +#endif diff --git a/sys/dev/drm/radeon/ni_reg.h b/sys/dev/drm/radeon/ni_reg.h index 5ed538c0a0..5db7b7d6fe 100644 --- a/sys/dev/drm/radeon/ni_reg.h +++ b/sys/dev/drm/radeon/ni_reg.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/ni_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __NI_REG_H__ #define __NI_REG_H__ diff --git a/sys/dev/drm/radeon/nid.h b/sys/dev/drm/radeon/nid.h index dbf85468f6..22421bc80c 100644 --- a/sys/dev/drm/radeon/nid.h +++ b/sys/dev/drm/radeon/nid.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/nid.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef NI_H #define NI_H @@ -135,6 +133,22 @@ #define VM_CONTEXT1_CNTL2 0x1434 #define VM_INVALIDATE_REQUEST 0x1478 #define VM_INVALIDATE_RESPONSE 0x147c +#define VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x14FC +#define VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x14DC +#define PROTECTIONS_MASK (0xf << 0) +#define PROTECTIONS_SHIFT 0 + /* bit 0: range + * bit 2: pde0 + * bit 3: valid + * bit 4: read + * bit 5: write + */ +#define MEMORY_CLIENT_ID_MASK (0xff << 12) +#define MEMORY_CLIENT_ID_SHIFT 12 +#define MEMORY_CLIENT_RW_MASK (1 << 24) +#define MEMORY_CLIENT_RW_SHIFT 24 +#define FAULT_VMID_MASK (0x7 << 25) +#define FAULT_VMID_SHIFT 25 #define VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR 0x1518 #define VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR 0x151c #define VM_CONTEXT0_PAGE_TABLE_BASE_ADDR 0x153C @@ -491,6 +505,571 @@ # define CACHE_FLUSH_AND_INV_EVENT_TS (0x14 << 0) # define CACHE_FLUSH_AND_INV_EVENT (0x16 << 0) +/* TN SMU registers */ +#define TN_CURRENT_GNB_TEMP 0x1F390 + +/* pm registers */ +#define SMC_MSG 0x20c +#define HOST_SMC_MSG(x) ((x) << 0) +#define HOST_SMC_MSG_MASK (0xff << 0) +#define HOST_SMC_MSG_SHIFT 0 +#define HOST_SMC_RESP(x) ((x) << 8) +#define HOST_SMC_RESP_MASK (0xff << 8) +#define HOST_SMC_RESP_SHIFT 8 +#define SMC_HOST_MSG(x) ((x) << 16) +#define SMC_HOST_MSG_MASK (0xff << 16) +#define SMC_HOST_MSG_SHIFT 16 +#define SMC_HOST_RESP(x) ((x) << 24) +#define SMC_HOST_RESP_MASK (0xff << 24) +#define SMC_HOST_RESP_SHIFT 24 + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_PDIV_A(x) ((x) << 20) +#define SPLL_PDIV_A_MASK (0x7f << 20) +#define SPLL_PDIV_A_SHIFT 20 +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_FB_DIV_SHIFT 0 +#define SPLL_DITHEN (1 << 28) + +#define MPLL_CNTL_MODE 0x61c +# define SS_SSEN (1 << 24) +# define SS_DSMODE_EN (1 << 25) + +#define MPLL_AD_FUNC_CNTL 0x624 +#define CLKF(x) ((x) << 0) +#define CLKF_MASK (0x7f << 0) +#define CLKR(x) ((x) << 7) +#define CLKR_MASK (0x1f << 7) +#define CLKFRAC(x) ((x) << 12) +#define CLKFRAC_MASK (0x1f << 12) +#define YCLK_POST_DIV(x) ((x) << 17) +#define YCLK_POST_DIV_MASK (3 << 17) +#define IBIAS(x) ((x) << 20) +#define IBIAS_MASK (0x3ff << 20) +#define RESET (1 << 30) +#define PDNB (1 << 31) +#define MPLL_AD_FUNC_CNTL_2 0x628 +#define BYPASS (1 << 19) +#define BIAS_GEN_PDNB (1 << 24) +#define RESET_EN (1 << 25) +#define VCO_MODE (1 << 29) +#define MPLL_DQ_FUNC_CNTL 0x62c +#define MPLL_DQ_FUNC_CNTL_2 0x630 + +#define GENERAL_PWRMGT 0x63c +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define ENABLE_GEN2PCIE (1 << 4) +# define ENABLE_GEN2XSP (1 << 5) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (3 << 6) +# define SW_SMIO_INDEX_SHIFT 6 +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 18) +# define BACKBIAS_VALUE (1 << 19) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +# define AC_DC_SW (1 << 24) + +#define SCLK_PWRMGT_CNTL 0x644 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) +# define DYN_LIGHT_SLEEP_EN (1 << 14) +#define MCLK_PWRMGT_CNTL 0x648 +# define DLL_SPEED(x) ((x) << 0) +# define DLL_SPEED_MASK (0x1f << 0) +# define MPLL_PWRMGT_OFF (1 << 5) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA0_PDNB (1 << 8) +# define MRDCKA1_PDNB (1 << 9) +# define MRDCKB0_PDNB (1 << 10) +# define MRDCKB1_PDNB (1 << 11) +# define MRDCKC0_PDNB (1 << 12) +# define MRDCKC1_PDNB (1 << 13) +# define MRDCKD0_PDNB (1 << 14) +# define MRDCKD1_PDNB (1 << 15) +# define MRDCKA0_RESET (1 << 16) +# define MRDCKA1_RESET (1 << 17) +# define MRDCKB0_RESET (1 << 18) +# define MRDCKB1_RESET (1 << 19) +# define MRDCKC0_RESET (1 << 20) +# define MRDCKC1_RESET (1 << 21) +# define MRDCKD0_RESET (1 << 22) +# define MRDCKD1_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define MPLL_TURNOFF_D2 (1 << 28) +#define DLL_CNTL 0x64c +# define MRDCKA0_BYPASS (1 << 24) +# define MRDCKA1_BYPASS (1 << 25) +# define MRDCKB0_BYPASS (1 << 26) +# define MRDCKB1_BYPASS (1 << 27) +# define MRDCKC0_BYPASS (1 << 28) +# define MRDCKC1_BYPASS (1 << 29) +# define MRDCKD0_BYPASS (1 << 30) +# define MRDCKD1_BYPASS (1 << 31) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x66c +# define CURRENT_STATE_INDEX_MASK (0xf << 4) +# define CURRENT_STATE_INDEX_SHIFT 4 + +#define CG_AT 0x6d4 +# define CG_R(x) ((x) << 0) +# define CG_R_MASK (0xffff << 0) +# define CG_L(x) ((x) << 16) +# define CG_L_MASK (0xffff << 16) + +#define CG_BIF_REQ_AND_RSP 0x7f4 +#define CG_CLIENT_REQ(x) ((x) << 0) +#define CG_CLIENT_REQ_MASK (0xff << 0) +#define CG_CLIENT_REQ_SHIFT 0 +#define CG_CLIENT_RESP(x) ((x) << 8) +#define CG_CLIENT_RESP_MASK (0xff << 8) +#define CG_CLIENT_RESP_SHIFT 8 +#define CLIENT_CG_REQ(x) ((x) << 16) +#define CLIENT_CG_REQ_MASK (0xff << 16) +#define CLIENT_CG_REQ_SHIFT 16 +#define CLIENT_CG_RESP(x) ((x) << 24) +#define CLIENT_CG_RESP_MASK (0xff << 24) +#define CLIENT_CG_RESP_SHIFT 24 + +#define CG_SPLL_SPREAD_SPECTRUM 0x790 +#define SSEN (1 << 0) +#define CLK_S(x) ((x) << 4) +#define CLK_S_MASK (0xfff << 4) +#define CLK_S_SHIFT 4 +#define CG_SPLL_SPREAD_SPECTRUM_2 0x794 +#define CLK_V(x) ((x) << 0) +#define CLK_V_MASK (0x3ffffff << 0) +#define CLK_V_SHIFT 0 + +#define SMC_SCRATCH0 0x81c + +#define CG_SPLL_FUNC_CNTL_4 0x850 + +#define MPLL_SS1 0x85c +#define CLKV(x) ((x) << 0) +#define CLKV_MASK (0x3ffffff << 0) +#define MPLL_SS2 0x860 +#define CLKS(x) ((x) << 0) +#define CLKS_MASK (0xfff << 0) + +#define CG_CAC_CTRL 0x88c +#define TID_CNT(x) ((x) << 0) +#define TID_CNT_MASK (0x3fff << 0) +#define TID_UNIT(x) ((x) << 14) +#define TID_UNIT_MASK (0xf << 14) + +#define CG_IND_ADDR 0x8f8 +#define CG_IND_DATA 0x8fc +/* CGIND regs */ +#define CG_CGTT_LOCAL_0 0x00 +#define CG_CGTT_LOCAL_1 0x01 + +#define MC_CG_CONFIG 0x25bc +#define MCDW_WR_ENABLE (1 << 0) +#define MCDX_WR_ENABLE (1 << 1) +#define MCDY_WR_ENABLE (1 << 2) +#define MCDZ_WR_ENABLE (1 << 3) +#define MC_RD_ENABLE(x) ((x) << 4) +#define MC_RD_ENABLE_MASK (3 << 4) +#define INDEX(x) ((x) << 6) +#define INDEX_MASK (0xfff << 6) +#define INDEX_SHIFT 6 + +#define MC_ARB_CAC_CNTL 0x2750 +#define ENABLE (1 << 0) +#define READ_WEIGHT(x) ((x) << 1) +#define READ_WEIGHT_MASK (0x3f << 1) +#define READ_WEIGHT_SHIFT 1 +#define WRITE_WEIGHT(x) ((x) << 7) +#define WRITE_WEIGHT_MASK (0x3f << 7) +#define WRITE_WEIGHT_SHIFT 7 +#define ALLOW_OVERFLOW (1 << 13) + +#define MC_ARB_DRAM_TIMING 0x2774 +#define MC_ARB_DRAM_TIMING2 0x2778 + +#define MC_ARB_RFSH_RATE 0x27b0 +#define POWERMODE0(x) ((x) << 0) +#define POWERMODE0_MASK (0xff << 0) +#define POWERMODE0_SHIFT 0 +#define POWERMODE1(x) ((x) << 8) +#define POWERMODE1_MASK (0xff << 8) +#define POWERMODE1_SHIFT 8 +#define POWERMODE2(x) ((x) << 16) +#define POWERMODE2_MASK (0xff << 16) +#define POWERMODE2_SHIFT 16 +#define POWERMODE3(x) ((x) << 24) +#define POWERMODE3_MASK (0xff << 24) +#define POWERMODE3_SHIFT 24 + +#define MC_ARB_CG 0x27e8 +#define CG_ARB_REQ(x) ((x) << 0) +#define CG_ARB_REQ_MASK (0xff << 0) +#define CG_ARB_REQ_SHIFT 0 +#define CG_ARB_RESP(x) ((x) << 8) +#define CG_ARB_RESP_MASK (0xff << 8) +#define CG_ARB_RESP_SHIFT 8 +#define ARB_CG_REQ(x) ((x) << 16) +#define ARB_CG_REQ_MASK (0xff << 16) +#define ARB_CG_REQ_SHIFT 16 +#define ARB_CG_RESP(x) ((x) << 24) +#define ARB_CG_RESP_MASK (0xff << 24) +#define ARB_CG_RESP_SHIFT 24 + +#define MC_ARB_DRAM_TIMING_1 0x27f0 +#define MC_ARB_DRAM_TIMING_2 0x27f4 +#define MC_ARB_DRAM_TIMING_3 0x27f8 +#define MC_ARB_DRAM_TIMING2_1 0x27fc +#define MC_ARB_DRAM_TIMING2_2 0x2800 +#define MC_ARB_DRAM_TIMING2_3 0x2804 +#define MC_ARB_BURST_TIME 0x2808 +#define STATE0(x) ((x) << 0) +#define STATE0_MASK (0x1f << 0) +#define STATE0_SHIFT 0 +#define STATE1(x) ((x) << 5) +#define STATE1_MASK (0x1f << 5) +#define STATE1_SHIFT 5 +#define STATE2(x) ((x) << 10) +#define STATE2_MASK (0x1f << 10) +#define STATE2_SHIFT 10 +#define STATE3(x) ((x) << 15) +#define STATE3_MASK (0x1f << 15) +#define STATE3_SHIFT 15 + +#define MC_CG_DATAPORT 0x2884 + +#define MC_SEQ_RAS_TIMING 0x28a0 +#define MC_SEQ_CAS_TIMING 0x28a4 +#define MC_SEQ_MISC_TIMING 0x28a8 +#define MC_SEQ_MISC_TIMING2 0x28ac +#define MC_SEQ_PMG_TIMING 0x28b0 +#define MC_SEQ_RD_CTL_D0 0x28b4 +#define MC_SEQ_RD_CTL_D1 0x28b8 +#define MC_SEQ_WR_CTL_D0 0x28bc +#define MC_SEQ_WR_CTL_D1 0x28c0 + +#define MC_SEQ_MISC0 0x2a00 +#define MC_SEQ_MISC0_GDDR5_SHIFT 28 +#define MC_SEQ_MISC0_GDDR5_MASK 0xf0000000 +#define MC_SEQ_MISC0_GDDR5_VALUE 5 +#define MC_SEQ_MISC1 0x2a04 +#define MC_SEQ_RESERVE_M 0x2a08 +#define MC_PMG_CMD_EMRS 0x2a0c + +#define MC_SEQ_MISC3 0x2a2c + +#define MC_SEQ_MISC5 0x2a54 +#define MC_SEQ_MISC6 0x2a58 + +#define MC_SEQ_MISC7 0x2a64 + +#define MC_SEQ_RAS_TIMING_LP 0x2a6c +#define MC_SEQ_CAS_TIMING_LP 0x2a70 +#define MC_SEQ_MISC_TIMING_LP 0x2a74 +#define MC_SEQ_MISC_TIMING2_LP 0x2a78 +#define MC_SEQ_WR_CTL_D0_LP 0x2a7c +#define MC_SEQ_WR_CTL_D1_LP 0x2a80 +#define MC_SEQ_PMG_CMD_EMRS_LP 0x2a84 +#define MC_SEQ_PMG_CMD_MRS_LP 0x2a88 + +#define MC_PMG_CMD_MRS 0x2aac + +#define MC_SEQ_RD_CTL_D0_LP 0x2b1c +#define MC_SEQ_RD_CTL_D1_LP 0x2b20 + +#define MC_PMG_CMD_MRS1 0x2b44 +#define MC_SEQ_PMG_CMD_MRS1_LP 0x2b48 +#define MC_SEQ_PMG_TIMING_LP 0x2b4c + +#define MC_PMG_CMD_MRS2 0x2b5c +#define MC_SEQ_PMG_CMD_MRS2_LP 0x2b60 + +#define LB_SYNC_RESET_SEL 0x6b28 +#define LB_SYNC_RESET_SEL_MASK (3 << 0) +#define LB_SYNC_RESET_SEL_SHIFT 0 + +#define DC_STUTTER_CNTL 0x6b30 +#define DC_STUTTER_ENABLE_A (1 << 0) +#define DC_STUTTER_ENABLE_B (1 << 1) + +#define SQ_CAC_THRESHOLD 0x8e4c +#define VSP(x) ((x) << 0) +#define VSP_MASK (0xff << 0) +#define VSP_SHIFT 0 +#define VSP0(x) ((x) << 8) +#define VSP0_MASK (0xff << 8) +#define VSP0_SHIFT 8 +#define GPR(x) ((x) << 16) +#define GPR_MASK (0xff << 16) +#define GPR_SHIFT 16 + +#define SQ_POWER_THROTTLE 0x8e58 +#define MIN_POWER(x) ((x) << 0) +#define MIN_POWER_MASK (0x3fff << 0) +#define MIN_POWER_SHIFT 0 +#define MAX_POWER(x) ((x) << 16) +#define MAX_POWER_MASK (0x3fff << 16) +#define MAX_POWER_SHIFT 0 +#define SQ_POWER_THROTTLE2 0x8e5c +#define MAX_POWER_DELTA(x) ((x) << 0) +#define MAX_POWER_DELTA_MASK (0x3fff << 0) +#define MAX_POWER_DELTA_SHIFT 0 +#define STI_SIZE(x) ((x) << 16) +#define STI_SIZE_MASK (0x3ff << 16) +#define STI_SIZE_SHIFT 16 +#define LTI_RATIO(x) ((x) << 27) +#define LTI_RATIO_MASK (0xf << 27) +#define LTI_RATIO_SHIFT 27 + +/* CG indirect registers */ +#define CG_CAC_REGION_1_WEIGHT_0 0x83 +#define WEIGHT_TCP_SIG0(x) ((x) << 0) +#define WEIGHT_TCP_SIG0_MASK (0x3f << 0) +#define WEIGHT_TCP_SIG0_SHIFT 0 +#define WEIGHT_TCP_SIG1(x) ((x) << 6) +#define WEIGHT_TCP_SIG1_MASK (0x3f << 6) +#define WEIGHT_TCP_SIG1_SHIFT 6 +#define WEIGHT_TA_SIG(x) ((x) << 12) +#define WEIGHT_TA_SIG_MASK (0x3f << 12) +#define WEIGHT_TA_SIG_SHIFT 12 +#define CG_CAC_REGION_1_WEIGHT_1 0x84 +#define WEIGHT_TCC_EN0(x) ((x) << 0) +#define WEIGHT_TCC_EN0_MASK (0x3f << 0) +#define WEIGHT_TCC_EN0_SHIFT 0 +#define WEIGHT_TCC_EN1(x) ((x) << 6) +#define WEIGHT_TCC_EN1_MASK (0x3f << 6) +#define WEIGHT_TCC_EN1_SHIFT 6 +#define WEIGHT_TCC_EN2(x) ((x) << 12) +#define WEIGHT_TCC_EN2_MASK (0x3f << 12) +#define WEIGHT_TCC_EN2_SHIFT 12 +#define WEIGHT_TCC_EN3(x) ((x) << 18) +#define WEIGHT_TCC_EN3_MASK (0x3f << 18) +#define WEIGHT_TCC_EN3_SHIFT 18 +#define CG_CAC_REGION_2_WEIGHT_0 0x85 +#define WEIGHT_CB_EN0(x) ((x) << 0) +#define WEIGHT_CB_EN0_MASK (0x3f << 0) +#define WEIGHT_CB_EN0_SHIFT 0 +#define WEIGHT_CB_EN1(x) ((x) << 6) +#define WEIGHT_CB_EN1_MASK (0x3f << 6) +#define WEIGHT_CB_EN1_SHIFT 6 +#define WEIGHT_CB_EN2(x) ((x) << 12) +#define WEIGHT_CB_EN2_MASK (0x3f << 12) +#define WEIGHT_CB_EN2_SHIFT 12 +#define WEIGHT_CB_EN3(x) ((x) << 18) +#define WEIGHT_CB_EN3_MASK (0x3f << 18) +#define WEIGHT_CB_EN3_SHIFT 18 +#define CG_CAC_REGION_2_WEIGHT_1 0x86 +#define WEIGHT_DB_SIG0(x) ((x) << 0) +#define WEIGHT_DB_SIG0_MASK (0x3f << 0) +#define WEIGHT_DB_SIG0_SHIFT 0 +#define WEIGHT_DB_SIG1(x) ((x) << 6) +#define WEIGHT_DB_SIG1_MASK (0x3f << 6) +#define WEIGHT_DB_SIG1_SHIFT 6 +#define WEIGHT_DB_SIG2(x) ((x) << 12) +#define WEIGHT_DB_SIG2_MASK (0x3f << 12) +#define WEIGHT_DB_SIG2_SHIFT 12 +#define WEIGHT_DB_SIG3(x) ((x) << 18) +#define WEIGHT_DB_SIG3_MASK (0x3f << 18) +#define WEIGHT_DB_SIG3_SHIFT 18 +#define CG_CAC_REGION_2_WEIGHT_2 0x87 +#define WEIGHT_SXM_SIG0(x) ((x) << 0) +#define WEIGHT_SXM_SIG0_MASK (0x3f << 0) +#define WEIGHT_SXM_SIG0_SHIFT 0 +#define WEIGHT_SXM_SIG1(x) ((x) << 6) +#define WEIGHT_SXM_SIG1_MASK (0x3f << 6) +#define WEIGHT_SXM_SIG1_SHIFT 6 +#define WEIGHT_SXM_SIG2(x) ((x) << 12) +#define WEIGHT_SXM_SIG2_MASK (0x3f << 12) +#define WEIGHT_SXM_SIG2_SHIFT 12 +#define WEIGHT_SXS_SIG0(x) ((x) << 18) +#define WEIGHT_SXS_SIG0_MASK (0x3f << 18) +#define WEIGHT_SXS_SIG0_SHIFT 18 +#define WEIGHT_SXS_SIG1(x) ((x) << 24) +#define WEIGHT_SXS_SIG1_MASK (0x3f << 24) +#define WEIGHT_SXS_SIG1_SHIFT 24 +#define CG_CAC_REGION_3_WEIGHT_0 0x88 +#define WEIGHT_XBR_0(x) ((x) << 0) +#define WEIGHT_XBR_0_MASK (0x3f << 0) +#define WEIGHT_XBR_0_SHIFT 0 +#define WEIGHT_XBR_1(x) ((x) << 6) +#define WEIGHT_XBR_1_MASK (0x3f << 6) +#define WEIGHT_XBR_1_SHIFT 6 +#define WEIGHT_XBR_2(x) ((x) << 12) +#define WEIGHT_XBR_2_MASK (0x3f << 12) +#define WEIGHT_XBR_2_SHIFT 12 +#define WEIGHT_SPI_SIG0(x) ((x) << 18) +#define WEIGHT_SPI_SIG0_MASK (0x3f << 18) +#define WEIGHT_SPI_SIG0_SHIFT 18 +#define CG_CAC_REGION_3_WEIGHT_1 0x89 +#define WEIGHT_SPI_SIG1(x) ((x) << 0) +#define WEIGHT_SPI_SIG1_MASK (0x3f << 0) +#define WEIGHT_SPI_SIG1_SHIFT 0 +#define WEIGHT_SPI_SIG2(x) ((x) << 6) +#define WEIGHT_SPI_SIG2_MASK (0x3f << 6) +#define WEIGHT_SPI_SIG2_SHIFT 6 +#define WEIGHT_SPI_SIG3(x) ((x) << 12) +#define WEIGHT_SPI_SIG3_MASK (0x3f << 12) +#define WEIGHT_SPI_SIG3_SHIFT 12 +#define WEIGHT_SPI_SIG4(x) ((x) << 18) +#define WEIGHT_SPI_SIG4_MASK (0x3f << 18) +#define WEIGHT_SPI_SIG4_SHIFT 18 +#define WEIGHT_SPI_SIG5(x) ((x) << 24) +#define WEIGHT_SPI_SIG5_MASK (0x3f << 24) +#define WEIGHT_SPI_SIG5_SHIFT 24 +#define CG_CAC_REGION_4_WEIGHT_0 0x8a +#define WEIGHT_LDS_SIG0(x) ((x) << 0) +#define WEIGHT_LDS_SIG0_MASK (0x3f << 0) +#define WEIGHT_LDS_SIG0_SHIFT 0 +#define WEIGHT_LDS_SIG1(x) ((x) << 6) +#define WEIGHT_LDS_SIG1_MASK (0x3f << 6) +#define WEIGHT_LDS_SIG1_SHIFT 6 +#define WEIGHT_SC(x) ((x) << 24) +#define WEIGHT_SC_MASK (0x3f << 24) +#define WEIGHT_SC_SHIFT 24 +#define CG_CAC_REGION_4_WEIGHT_1 0x8b +#define WEIGHT_BIF(x) ((x) << 0) +#define WEIGHT_BIF_MASK (0x3f << 0) +#define WEIGHT_BIF_SHIFT 0 +#define WEIGHT_CP(x) ((x) << 6) +#define WEIGHT_CP_MASK (0x3f << 6) +#define WEIGHT_CP_SHIFT 6 +#define WEIGHT_PA_SIG0(x) ((x) << 12) +#define WEIGHT_PA_SIG0_MASK (0x3f << 12) +#define WEIGHT_PA_SIG0_SHIFT 12 +#define WEIGHT_PA_SIG1(x) ((x) << 18) +#define WEIGHT_PA_SIG1_MASK (0x3f << 18) +#define WEIGHT_PA_SIG1_SHIFT 18 +#define WEIGHT_VGT_SIG0(x) ((x) << 24) +#define WEIGHT_VGT_SIG0_MASK (0x3f << 24) +#define WEIGHT_VGT_SIG0_SHIFT 24 +#define CG_CAC_REGION_4_WEIGHT_2 0x8c +#define WEIGHT_VGT_SIG1(x) ((x) << 0) +#define WEIGHT_VGT_SIG1_MASK (0x3f << 0) +#define WEIGHT_VGT_SIG1_SHIFT 0 +#define WEIGHT_VGT_SIG2(x) ((x) << 6) +#define WEIGHT_VGT_SIG2_MASK (0x3f << 6) +#define WEIGHT_VGT_SIG2_SHIFT 6 +#define WEIGHT_DC_SIG0(x) ((x) << 12) +#define WEIGHT_DC_SIG0_MASK (0x3f << 12) +#define WEIGHT_DC_SIG0_SHIFT 12 +#define WEIGHT_DC_SIG1(x) ((x) << 18) +#define WEIGHT_DC_SIG1_MASK (0x3f << 18) +#define WEIGHT_DC_SIG1_SHIFT 18 +#define WEIGHT_DC_SIG2(x) ((x) << 24) +#define WEIGHT_DC_SIG2_MASK (0x3f << 24) +#define WEIGHT_DC_SIG2_SHIFT 24 +#define CG_CAC_REGION_4_WEIGHT_3 0x8d +#define WEIGHT_DC_SIG3(x) ((x) << 0) +#define WEIGHT_DC_SIG3_MASK (0x3f << 0) +#define WEIGHT_DC_SIG3_SHIFT 0 +#define WEIGHT_UVD_SIG0(x) ((x) << 6) +#define WEIGHT_UVD_SIG0_MASK (0x3f << 6) +#define WEIGHT_UVD_SIG0_SHIFT 6 +#define WEIGHT_UVD_SIG1(x) ((x) << 12) +#define WEIGHT_UVD_SIG1_MASK (0x3f << 12) +#define WEIGHT_UVD_SIG1_SHIFT 12 +#define WEIGHT_SPARE0(x) ((x) << 18) +#define WEIGHT_SPARE0_MASK (0x3f << 18) +#define WEIGHT_SPARE0_SHIFT 18 +#define WEIGHT_SPARE1(x) ((x) << 24) +#define WEIGHT_SPARE1_MASK (0x3f << 24) +#define WEIGHT_SPARE1_SHIFT 24 +#define CG_CAC_REGION_5_WEIGHT_0 0x8e +#define WEIGHT_SQ_VSP(x) ((x) << 0) +#define WEIGHT_SQ_VSP_MASK (0x3fff << 0) +#define WEIGHT_SQ_VSP_SHIFT 0 +#define WEIGHT_SQ_VSP0(x) ((x) << 14) +#define WEIGHT_SQ_VSP0_MASK (0x3fff << 14) +#define WEIGHT_SQ_VSP0_SHIFT 14 +#define CG_CAC_REGION_4_OVERRIDE_4 0xab +#define OVR_MODE_SPARE_0(x) ((x) << 16) +#define OVR_MODE_SPARE_0_MASK (0x1 << 16) +#define OVR_MODE_SPARE_0_SHIFT 16 +#define OVR_VAL_SPARE_0(x) ((x) << 17) +#define OVR_VAL_SPARE_0_MASK (0x1 << 17) +#define OVR_VAL_SPARE_0_SHIFT 17 +#define OVR_MODE_SPARE_1(x) ((x) << 18) +#define OVR_MODE_SPARE_1_MASK (0x3f << 18) +#define OVR_MODE_SPARE_1_SHIFT 18 +#define OVR_VAL_SPARE_1(x) ((x) << 19) +#define OVR_VAL_SPARE_1_MASK (0x3f << 19) +#define OVR_VAL_SPARE_1_SHIFT 19 +#define CG_CAC_REGION_5_WEIGHT_1 0xb7 +#define WEIGHT_SQ_GPR(x) ((x) << 0) +#define WEIGHT_SQ_GPR_MASK (0x3fff << 0) +#define WEIGHT_SQ_GPR_SHIFT 0 +#define WEIGHT_SQ_LDS(x) ((x) << 14) +#define WEIGHT_SQ_LDS_MASK (0x3fff << 14) +#define WEIGHT_SQ_LDS_SHIFT 14 + +/* PCIE link stuff */ +#define PCIE_LC_TRAINING_CNTL 0xa1 /* PCIE_P */ +#define PCIE_LC_LINK_WIDTH_CNTL 0xa2 /* PCIE_P */ +# define LC_LINK_WIDTH_SHIFT 0 +# define LC_LINK_WIDTH_MASK 0x7 +# define LC_LINK_WIDTH_X0 0 +# define LC_LINK_WIDTH_X1 1 +# define LC_LINK_WIDTH_X2 2 +# define LC_LINK_WIDTH_X4 3 +# define LC_LINK_WIDTH_X8 4 +# define LC_LINK_WIDTH_X16 6 +# define LC_LINK_WIDTH_RD_SHIFT 4 +# define LC_LINK_WIDTH_RD_MASK 0x70 +# define LC_RECONFIG_ARC_MISSING_ESCAPE (1 << 7) +# define LC_RECONFIG_NOW (1 << 8) +# define LC_RENEGOTIATION_SUPPORT (1 << 9) +# define LC_RENEGOTIATE_EN (1 << 10) +# define LC_SHORT_RECONFIG_EN (1 << 11) +# define LC_UPCONFIGURE_SUPPORT (1 << 12) +# define LC_UPCONFIGURE_DIS (1 << 13) +#define PCIE_LC_SPEED_CNTL 0xa4 /* PCIE_P */ +# define LC_GEN2_EN_STRAP (1 << 0) +# define LC_TARGET_LINK_SPEED_OVERRIDE_EN (1 << 1) +# define LC_FORCE_EN_HW_SPEED_CHANGE (1 << 5) +# define LC_FORCE_DIS_HW_SPEED_CHANGE (1 << 6) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK (0x3 << 8) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT 3 +# define LC_CURRENT_DATA_RATE (1 << 11) +# define LC_HW_VOLTAGE_IF_CONTROL(x) ((x) << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_MASK (3 << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_SHIFT 12 +# define LC_VOLTAGE_TIMER_SEL_MASK (0xf << 14) +# define LC_CLR_FAILED_SPD_CHANGE_CNT (1 << 21) +# define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 23) +# define LC_OTHER_SIDE_SUPPORTS_GEN2 (1 << 24) +#define MM_CFGREGS_CNTL 0x544c +# define MM_WR_TO_CFG_EN (1 << 3) +#define LINK_CNTL2 0x88 /* F0 */ +# define TARGET_LINK_SPEED_MASK (0xf << 0) +# define SELECTABLE_DEEMPHASIS (1 << 6) + /* * UVD */ diff --git a/sys/dev/drm/radeon/nislands_smc.h b/sys/dev/drm/radeon/nislands_smc.h new file mode 100644 index 0000000000..78d4b6a681 --- /dev/null +++ b/sys/dev/drm/radeon/nislands_smc.h @@ -0,0 +1,328 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __NISLANDS_SMC_H__ +#define __NISLANDS_SMC_H__ + +#pragma pack(push, 1) + +#define NISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE 16 + +struct PP_NIslands_Dpm2PerfLevel +{ + uint8_t MaxPS; + uint8_t TgtAct; + uint8_t MaxPS_StepInc; + uint8_t MaxPS_StepDec; + uint8_t PSST; + uint8_t NearTDPDec; + uint8_t AboveSafeInc; + uint8_t BelowSafeInc; + uint8_t PSDeltaLimit; + uint8_t PSDeltaWin; + uint8_t Reserved[6]; +}; + +typedef struct PP_NIslands_Dpm2PerfLevel PP_NIslands_Dpm2PerfLevel; + +struct PP_NIslands_DPM2Parameters +{ + uint32_t TDPLimit; + uint32_t NearTDPLimit; + uint32_t SafePowerLimit; + uint32_t PowerBoostLimit; +}; +typedef struct PP_NIslands_DPM2Parameters PP_NIslands_DPM2Parameters; + +struct NISLANDS_SMC_SCLK_VALUE +{ + uint32_t vCG_SPLL_FUNC_CNTL; + uint32_t vCG_SPLL_FUNC_CNTL_2; + uint32_t vCG_SPLL_FUNC_CNTL_3; + uint32_t vCG_SPLL_FUNC_CNTL_4; + uint32_t vCG_SPLL_SPREAD_SPECTRUM; + uint32_t vCG_SPLL_SPREAD_SPECTRUM_2; + uint32_t sclk_value; +}; + +typedef struct NISLANDS_SMC_SCLK_VALUE NISLANDS_SMC_SCLK_VALUE; + +struct NISLANDS_SMC_MCLK_VALUE +{ + uint32_t vMPLL_FUNC_CNTL; + uint32_t vMPLL_FUNC_CNTL_1; + uint32_t vMPLL_FUNC_CNTL_2; + uint32_t vMPLL_AD_FUNC_CNTL; + uint32_t vMPLL_AD_FUNC_CNTL_2; + uint32_t vMPLL_DQ_FUNC_CNTL; + uint32_t vMPLL_DQ_FUNC_CNTL_2; + uint32_t vMCLK_PWRMGT_CNTL; + uint32_t vDLL_CNTL; + uint32_t vMPLL_SS; + uint32_t vMPLL_SS2; + uint32_t mclk_value; +}; + +typedef struct NISLANDS_SMC_MCLK_VALUE NISLANDS_SMC_MCLK_VALUE; + +struct NISLANDS_SMC_VOLTAGE_VALUE +{ + uint16_t value; + uint8_t index; + uint8_t padding; +}; + +typedef struct NISLANDS_SMC_VOLTAGE_VALUE NISLANDS_SMC_VOLTAGE_VALUE; + +struct NISLANDS_SMC_HW_PERFORMANCE_LEVEL +{ + uint8_t arbValue; + uint8_t ACIndex; + uint8_t displayWatermark; + uint8_t gen2PCIE; + uint8_t reserved1; + uint8_t reserved2; + uint8_t strobeMode; + uint8_t mcFlags; + uint32_t aT; + uint32_t bSP; + NISLANDS_SMC_SCLK_VALUE sclk; + NISLANDS_SMC_MCLK_VALUE mclk; + NISLANDS_SMC_VOLTAGE_VALUE vddc; + NISLANDS_SMC_VOLTAGE_VALUE mvdd; + NISLANDS_SMC_VOLTAGE_VALUE vddci; + NISLANDS_SMC_VOLTAGE_VALUE std_vddc; + uint32_t powergate_en; + uint8_t hUp; + uint8_t hDown; + uint8_t stateFlags; + uint8_t arbRefreshState; + uint32_t SQPowerThrottle; + uint32_t SQPowerThrottle_2; + uint32_t reserved[2]; + PP_NIslands_Dpm2PerfLevel dpm2; +}; + +#define NISLANDS_SMC_STROBE_RATIO 0x0F +#define NISLANDS_SMC_STROBE_ENABLE 0x10 + +#define NISLANDS_SMC_MC_EDC_RD_FLAG 0x01 +#define NISLANDS_SMC_MC_EDC_WR_FLAG 0x02 +#define NISLANDS_SMC_MC_RTT_ENABLE 0x04 +#define NISLANDS_SMC_MC_STUTTER_EN 0x08 + +typedef struct NISLANDS_SMC_HW_PERFORMANCE_LEVEL NISLANDS_SMC_HW_PERFORMANCE_LEVEL; + +struct NISLANDS_SMC_SWSTATE +{ + uint8_t flags; + uint8_t levelCount; + uint8_t padding2; + uint8_t padding3; + NISLANDS_SMC_HW_PERFORMANCE_LEVEL levels[1]; +}; + +typedef struct NISLANDS_SMC_SWSTATE NISLANDS_SMC_SWSTATE; + +#define NISLANDS_SMC_VOLTAGEMASK_VDDC 0 +#define NISLANDS_SMC_VOLTAGEMASK_MVDD 1 +#define NISLANDS_SMC_VOLTAGEMASK_VDDCI 2 +#define NISLANDS_SMC_VOLTAGEMASK_MAX 4 + +struct NISLANDS_SMC_VOLTAGEMASKTABLE +{ + uint8_t highMask[NISLANDS_SMC_VOLTAGEMASK_MAX]; + uint32_t lowMask[NISLANDS_SMC_VOLTAGEMASK_MAX]; +}; + +typedef struct NISLANDS_SMC_VOLTAGEMASKTABLE NISLANDS_SMC_VOLTAGEMASKTABLE; + +#define NISLANDS_MAX_NO_VREG_STEPS 32 + +struct NISLANDS_SMC_STATETABLE +{ + uint8_t thermalProtectType; + uint8_t systemFlags; + uint8_t maxVDDCIndexInPPTable; + uint8_t extraFlags; + uint8_t highSMIO[NISLANDS_MAX_NO_VREG_STEPS]; + uint32_t lowSMIO[NISLANDS_MAX_NO_VREG_STEPS]; + NISLANDS_SMC_VOLTAGEMASKTABLE voltageMaskTable; + PP_NIslands_DPM2Parameters dpm2Params; + NISLANDS_SMC_SWSTATE initialState; + NISLANDS_SMC_SWSTATE ACPIState; + NISLANDS_SMC_SWSTATE ULVState; + NISLANDS_SMC_SWSTATE driverState; + NISLANDS_SMC_HW_PERFORMANCE_LEVEL dpmLevels[NISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1]; +}; + +typedef struct NISLANDS_SMC_STATETABLE NISLANDS_SMC_STATETABLE; + +#define NI_SMC_SOFT_REGISTERS_START 0x108 + +#define NI_SMC_SOFT_REGISTER_mclk_chg_timeout 0x0 +#define NI_SMC_SOFT_REGISTER_delay_bbias 0xC +#define NI_SMC_SOFT_REGISTER_delay_vreg 0x10 +#define NI_SMC_SOFT_REGISTER_delay_acpi 0x2C +#define NI_SMC_SOFT_REGISTER_seq_index 0x64 +#define NI_SMC_SOFT_REGISTER_mvdd_chg_time 0x68 +#define NI_SMC_SOFT_REGISTER_mclk_switch_lim 0x78 +#define NI_SMC_SOFT_REGISTER_watermark_threshold 0x80 +#define NI_SMC_SOFT_REGISTER_mc_block_delay 0x84 +#define NI_SMC_SOFT_REGISTER_uvd_enabled 0x98 + +#define SMC_NISLANDS_MC_TPP_CAC_NUM_OF_ENTRIES 16 +#define SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES 16 +#define SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES 16 +#define SMC_NISLANDS_BIF_LUT_NUM_OF_ENTRIES 4 + +struct SMC_NISLANDS_MC_TPP_CAC_TABLE +{ + uint32_t tpp[SMC_NISLANDS_MC_TPP_CAC_NUM_OF_ENTRIES]; + uint32_t cacValue[SMC_NISLANDS_MC_TPP_CAC_NUM_OF_ENTRIES]; +}; + +typedef struct SMC_NISLANDS_MC_TPP_CAC_TABLE SMC_NISLANDS_MC_TPP_CAC_TABLE; + + +struct PP_NIslands_CACTABLES +{ + uint32_t cac_bif_lut[SMC_NISLANDS_BIF_LUT_NUM_OF_ENTRIES]; + uint32_t cac_lkge_lut[SMC_NISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES][SMC_NISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES]; + + uint32_t pwr_const; + + uint32_t dc_cacValue; + uint32_t bif_cacValue; + uint32_t lkge_pwr; + + uint8_t cac_width; + uint8_t window_size_p2; + + uint8_t num_drop_lsb; + uint8_t padding_0; + + uint32_t last_power; + + uint8_t AllowOvrflw; + uint8_t MCWrWeight; + uint8_t MCRdWeight; + uint8_t padding_1[9]; + + uint8_t enableWinAvg; + uint8_t numWin_TDP; + uint8_t l2numWin_TDP; + uint8_t WinIndex; + + uint32_t dynPwr_TDP[4]; + uint32_t lkgePwr_TDP[4]; + uint32_t power_TDP[4]; + uint32_t avg_dynPwr_TDP; + uint32_t avg_lkgePwr_TDP; + uint32_t avg_power_TDP; + uint32_t lts_power_TDP; + uint8_t lts_truncate_n; + uint8_t padding_2[7]; +}; + +typedef struct PP_NIslands_CACTABLES PP_NIslands_CACTABLES; + +#define SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE 32 +#define SMC_NISLANDS_MC_REGISTER_ARRAY_SET_COUNT 20 + +struct SMC_NIslands_MCRegisterAddress +{ + uint16_t s0; + uint16_t s1; +}; + +typedef struct SMC_NIslands_MCRegisterAddress SMC_NIslands_MCRegisterAddress; + + +struct SMC_NIslands_MCRegisterSet +{ + uint32_t value[SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +typedef struct SMC_NIslands_MCRegisterSet SMC_NIslands_MCRegisterSet; + +struct SMC_NIslands_MCRegisters +{ + uint8_t last; + uint8_t reserved[3]; + SMC_NIslands_MCRegisterAddress address[SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE]; + SMC_NIslands_MCRegisterSet data[SMC_NISLANDS_MC_REGISTER_ARRAY_SET_COUNT]; +}; + +typedef struct SMC_NIslands_MCRegisters SMC_NIslands_MCRegisters; + +struct SMC_NIslands_MCArbDramTimingRegisterSet +{ + uint32_t mc_arb_dram_timing; + uint32_t mc_arb_dram_timing2; + uint8_t mc_arb_rfsh_rate; + uint8_t padding[3]; +}; + +typedef struct SMC_NIslands_MCArbDramTimingRegisterSet SMC_NIslands_MCArbDramTimingRegisterSet; + +struct SMC_NIslands_MCArbDramTimingRegisters +{ + uint8_t arb_current; + uint8_t reserved[3]; + SMC_NIslands_MCArbDramTimingRegisterSet data[20]; +}; + +typedef struct SMC_NIslands_MCArbDramTimingRegisters SMC_NIslands_MCArbDramTimingRegisters; + +struct SMC_NISLANDS_SPLL_DIV_TABLE +{ + uint32_t freq[256]; + uint32_t ss[256]; +}; + +#define SMC_NISLANDS_SPLL_DIV_TABLE_FBDIV_MASK 0x01ffffff +#define SMC_NISLANDS_SPLL_DIV_TABLE_FBDIV_SHIFT 0 +#define SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_MASK 0xfe000000 +#define SMC_NISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT 25 +#define SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_MASK 0x000fffff +#define SMC_NISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT 0 +#define SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_MASK 0xfff00000 +#define SMC_NISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT 20 + +typedef struct SMC_NISLANDS_SPLL_DIV_TABLE SMC_NISLANDS_SPLL_DIV_TABLE; + +#define NISLANDS_SMC_FIRMWARE_HEADER_LOCATION 0x100 + +#define NISLANDS_SMC_FIRMWARE_HEADER_version 0x0 +#define NISLANDS_SMC_FIRMWARE_HEADER_flags 0x4 +#define NISLANDS_SMC_FIRMWARE_HEADER_softRegisters 0x8 +#define NISLANDS_SMC_FIRMWARE_HEADER_stateTable 0xC +#define NISLANDS_SMC_FIRMWARE_HEADER_fanTable 0x10 +#define NISLANDS_SMC_FIRMWARE_HEADER_cacTable 0x14 +#define NISLANDS_SMC_FIRMWARE_HEADER_mcRegisterTable 0x20 +#define NISLANDS_SMC_FIRMWARE_HEADER_mcArbDramAutoRefreshTable 0x2C +#define NISLANDS_SMC_FIRMWARE_HEADER_spllTable 0x30 + +#pragma pack(pop) + +#endif diff --git a/sys/dev/drm/radeon/ppsmc.h b/sys/dev/drm/radeon/ppsmc.h new file mode 100644 index 0000000000..b5564a3645 --- /dev/null +++ b/sys/dev/drm/radeon/ppsmc.h @@ -0,0 +1,116 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef PP_SMC_H +#define PP_SMC_H + +#pragma pack(push, 1) + +#define PPSMC_SWSTATE_FLAG_DC 0x01 +#define PPSMC_SWSTATE_FLAG_UVD 0x02 +#define PPSMC_SWSTATE_FLAG_VCE 0x04 +#define PPSMC_SWSTATE_FLAG_PCIE_X1 0x08 + +#define PPSMC_THERMAL_PROTECT_TYPE_INTERNAL 0x00 +#define PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL 0x01 +#define PPSMC_THERMAL_PROTECT_TYPE_NONE 0xff + +#define PPSMC_SYSTEMFLAG_GPIO_DC 0x01 +#define PPSMC_SYSTEMFLAG_STEPVDDC 0x02 +#define PPSMC_SYSTEMFLAG_GDDR5 0x04 +#define PPSMC_SYSTEMFLAG_DISABLE_BABYSTEP 0x08 +#define PPSMC_SYSTEMFLAG_REGULATOR_HOT 0x10 +#define PPSMC_SYSTEMFLAG_REGULATOR_HOT_ANALOG 0x20 +#define PPSMC_SYSTEMFLAG_REGULATOR_HOT_PROG_GPIO 0x40 + +#define PPSMC_EXTRAFLAGS_AC2DC_ACTION_MASK 0x07 +#define PPSMC_EXTRAFLAGS_AC2DC_DONT_WAIT_FOR_VBLANK 0x08 +#define PPSMC_EXTRAFLAGS_AC2DC_ACTION_GOTODPMLOWSTATE 0x00 +#define PPSMC_EXTRAFLAGS_AC2DC_ACTION_GOTOINITIALSTATE 0x01 +#define PPSMC_EXTRAFLAGS_AC2DC_GPIO5_POLARITY_HIGH 0x02 + +#define PPSMC_DISPLAY_WATERMARK_LOW 0 +#define PPSMC_DISPLAY_WATERMARK_HIGH 1 + +#define PPSMC_STATEFLAG_AUTO_PULSE_SKIP 0x01 +#define PPSMC_STATEFLAG_POWERBOOST 0x02 +#define PPSMC_STATEFLAG_DEEPSLEEP_THROTTLE 0x20 +#define PPSMC_STATEFLAG_DEEPSLEEP_BYPASS 0x40 + +#define PPSMC_Result_OK ((uint8_t)0x01) +#define PPSMC_Result_Failed ((uint8_t)0xFF) + +typedef uint8_t PPSMC_Result; + +#define PPSMC_MSG_Halt ((uint8_t)0x10) +#define PPSMC_MSG_Resume ((uint8_t)0x11) +#define PPSMC_MSG_ZeroLevelsDisabled ((uint8_t)0x13) +#define PPSMC_MSG_OneLevelsDisabled ((uint8_t)0x14) +#define PPSMC_MSG_TwoLevelsDisabled ((uint8_t)0x15) +#define PPSMC_MSG_EnableThermalInterrupt ((uint8_t)0x16) +#define PPSMC_MSG_RunningOnAC ((uint8_t)0x17) +#define PPSMC_MSG_SwitchToSwState ((uint8_t)0x20) +#define PPSMC_MSG_SwitchToInitialState ((uint8_t)0x40) +#define PPSMC_MSG_NoForcedLevel ((uint8_t)0x41) +#define PPSMC_MSG_ForceHigh ((uint8_t)0x42) +#define PPSMC_MSG_ForceMediumOrHigh ((uint8_t)0x43) +#define PPSMC_MSG_SwitchToMinimumPower ((uint8_t)0x51) +#define PPSMC_MSG_ResumeFromMinimumPower ((uint8_t)0x52) +#define PPSMC_MSG_EnableCac ((uint8_t)0x53) +#define PPSMC_MSG_DisableCac ((uint8_t)0x54) +#define PPSMC_TDPClampingActive ((uint8_t)0x59) +#define PPSMC_TDPClampingInactive ((uint8_t)0x5A) +#define PPSMC_MSG_NoDisplay ((uint8_t)0x5D) +#define PPSMC_MSG_HasDisplay ((uint8_t)0x5E) +#define PPSMC_MSG_UVDPowerOFF ((uint8_t)0x60) +#define PPSMC_MSG_UVDPowerON ((uint8_t)0x61) +#define PPSMC_MSG_EnableULV ((uint8_t)0x62) +#define PPSMC_MSG_DisableULV ((uint8_t)0x63) +#define PPSMC_MSG_EnterULV ((uint8_t)0x64) +#define PPSMC_MSG_ExitULV ((uint8_t)0x65) +#define PPSMC_CACLongTermAvgEnable ((uint8_t)0x6E) +#define PPSMC_CACLongTermAvgDisable ((uint8_t)0x6F) +#define PPSMC_MSG_CollectCAC_PowerCorreln ((uint8_t)0x7A) +#define PPSMC_FlushDataCache ((uint8_t)0x80) +#define PPSMC_MSG_SetEnabledLevels ((uint8_t)0x82) +#define PPSMC_MSG_SetForcedLevels ((uint8_t)0x83) +#define PPSMC_MSG_ResetToDefaults ((uint8_t)0x84) +#define PPSMC_MSG_EnableDTE ((uint8_t)0x87) +#define PPSMC_MSG_DisableDTE ((uint8_t)0x88) +#define PPSMC_MSG_ThrottleOVRDSCLKDS ((uint8_t)0x96) +#define PPSMC_MSG_CancelThrottleOVRDSCLKDS ((uint8_t)0x97) + +/* TN */ +#define PPSMC_MSG_DPM_Config ((uint32_t) 0x102) +#define PPSMC_MSG_DPM_ForceState ((uint32_t) 0x104) +#define PPSMC_MSG_PG_SIMD_Config ((uint32_t) 0x108) +#define PPSMC_MSG_DPM_N_LevelsDisabled ((uint32_t) 0x112) +#define PPSMC_MSG_DCE_RemoveVoltageAdjustment ((uint32_t) 0x11d) +#define PPSMC_MSG_DCE_AllowVoltageAdjustment ((uint32_t) 0x11e) +#define PPSMC_MSG_UVD_DPM_Config ((uint32_t) 0x124) + + +typedef uint16_t PPSMC_Msg; + +#pragma pack(pop) + +#endif diff --git a/sys/dev/drm/radeon/r100.c b/sys/dev/drm/radeon/r100.c index 138f19d410..50865b037b 100644 --- a/sys/dev/drm/radeon/r100.c +++ b/sys/dev/drm/radeon/r100.c @@ -3086,6 +3086,10 @@ int r100_set_surface_reg(struct radeon_device *rdev, int reg, flags |= RADEON_SURF_TILE_COLOR_BOTH; if (tiling_flags & RADEON_TILING_MACRO) flags |= RADEON_SURF_TILE_COLOR_MACRO; + /* setting pitch to 0 disables tiling */ + if ((tiling_flags & (RADEON_TILING_MACRO|RADEON_TILING_MICRO)) + == 0) + pitch = 0; } else if (rdev->family <= CHIP_RV280) { if (tiling_flags & (RADEON_TILING_MACRO)) flags |= R200_SURF_TILE_COLOR_MACRO; @@ -3103,13 +3107,6 @@ int r100_set_surface_reg(struct radeon_device *rdev, int reg, if (tiling_flags & RADEON_TILING_SWAP_32BIT) flags |= RADEON_SURF_AP0_SWP_32BPP | RADEON_SURF_AP1_SWP_32BPP; - /* when we aren't tiling the pitch seems to needs to be furtherdivided down. - tested on power5 + rn50 server */ - if (tiling_flags & (RADEON_TILING_SWAP_16BIT | RADEON_TILING_SWAP_32BIT)) { - if (!(tiling_flags & (RADEON_TILING_MACRO | RADEON_TILING_MICRO))) - if (ASIC_IS_RN50(rdev)) - pitch /= 16; - } - /* r100/r200 divide by 16 */ if (rdev->family < CHIP_R300) flags |= pitch / 16; diff --git a/sys/dev/drm/radeon/r100_track.h b/sys/dev/drm/radeon/r100_track.h index 3547d1b88f..eb40888bdf 100644 --- a/sys/dev/drm/radeon/r100_track.h +++ b/sys/dev/drm/radeon/r100_track.h @@ -1,4 +1,3 @@ -/* $FreeBSD: head/sys/dev/drm2/radeon/r100_track.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #define R100_TRACK_MAX_TEXTURE 3 #define R200_TRACK_MAX_TEXTURE 6 diff --git a/sys/dev/drm/radeon/r100d.h b/sys/dev/drm/radeon/r100d.h index 873676ff47..f0f8ee69f4 100644 --- a/sys/dev/drm/radeon/r100d.h +++ b/sys/dev/drm/radeon/r100d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r100d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __R100D_H__ #define __R100D_H__ diff --git a/sys/dev/drm/radeon/r200.c b/sys/dev/drm/radeon/r200.c index a052d5f40c..c3c3579785 100644 --- a/sys/dev/drm/radeon/r200.c +++ b/sys/dev/drm/radeon/r200.c @@ -24,10 +24,7 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r200.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include "radeon_reg.h" diff --git a/sys/dev/drm/radeon/r300_reg.h b/sys/dev/drm/radeon/r300_reg.h index 9dae676a0f..00c0d2ba22 100644 --- a/sys/dev/drm/radeon/r300_reg.h +++ b/sys/dev/drm/radeon/r300_reg.h @@ -23,8 +23,6 @@ * * Authors: Nicolai Haehnle * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r300_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef _R300_REG_H_ #define _R300_REG_H_ diff --git a/sys/dev/drm/radeon/r300d.h b/sys/dev/drm/radeon/r300d.h index c7debad358..ff229a00d2 100644 --- a/sys/dev/drm/radeon/r300d.h +++ b/sys/dev/drm/radeon/r300d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r300d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __R300D_H__ #define __R300D_H__ diff --git a/sys/dev/drm/radeon/r420d.h b/sys/dev/drm/radeon/r420d.h index 6e9f7b18f3..fc78d31a0b 100644 --- a/sys/dev/drm/radeon/r420d.h +++ b/sys/dev/drm/radeon/r420d.h @@ -24,10 +24,7 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r420d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ - #ifndef R420D_H #define R420D_H diff --git a/sys/dev/drm/radeon/r500_reg.h b/sys/dev/drm/radeon/r500_reg.h index 06a1f80f57..69b4f1efb9 100644 --- a/sys/dev/drm/radeon/r500_reg.h +++ b/sys/dev/drm/radeon/r500_reg.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r500_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __R500_REG_H__ #define __R500_REG_H__ diff --git a/sys/dev/drm/radeon/r520.c b/sys/dev/drm/radeon/r520.c index 587921b5b7..73722a74a0 100644 --- a/sys/dev/drm/radeon/r520.c +++ b/sys/dev/drm/radeon/r520.c @@ -24,10 +24,7 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r520.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include "radeon.h" #include "radeon_asic.h" diff --git a/sys/dev/drm/radeon/r520d.h b/sys/dev/drm/radeon/r520d.h index 0892fc2430..61af61f644 100644 --- a/sys/dev/drm/radeon/r520d.h +++ b/sys/dev/drm/radeon/r520d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r520d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __R520D_H__ #define __R520D_H__ diff --git a/sys/dev/drm/radeon/r600.c b/sys/dev/drm/radeon/r600.c index 8b04605655..bba86ece84 100644 --- a/sys/dev/drm/radeon/r600.c +++ b/sys/dev/drm/radeon/r600.c @@ -36,18 +36,7 @@ #include "r600d.h" #include "atom.h" #include "avivod.h" - -#define PFP_UCODE_SIZE 576 -#define PM4_UCODE_SIZE 1792 -#define RLC_UCODE_SIZE 768 -#define R700_PFP_UCODE_SIZE 848 -#define R700_PM4_UCODE_SIZE 1360 -#define R700_RLC_UCODE_SIZE 1024 -#define EVERGREEN_PFP_UCODE_SIZE 1120 -#define EVERGREEN_PM4_UCODE_SIZE 1376 -#define EVERGREEN_RLC_UCODE_SIZE 768 -#define CAYMAN_RLC_UCODE_SIZE 1024 -#define ARUBA_RLC_UCODE_SIZE 1536 +#include "radeon_ucode.h" #ifdef DUMBBELL_WIP /* Firmware Names */ @@ -67,24 +56,32 @@ MODULE_FIRMWARE("radeon/RS780_pfp.bin"); MODULE_FIRMWARE("radeon/RS780_me.bin"); MODULE_FIRMWARE("radeon/RV770_pfp.bin"); MODULE_FIRMWARE("radeon/RV770_me.bin"); +MODULE_FIRMWARE("radeon/RV770_smc.bin"); MODULE_FIRMWARE("radeon/RV730_pfp.bin"); MODULE_FIRMWARE("radeon/RV730_me.bin"); +MODULE_FIRMWARE("radeon/RV730_smc.bin"); +MODULE_FIRMWARE("radeon/RV740_smc.bin"); MODULE_FIRMWARE("radeon/RV710_pfp.bin"); MODULE_FIRMWARE("radeon/RV710_me.bin"); +MODULE_FIRMWARE("radeon/RV710_smc.bin"); MODULE_FIRMWARE("radeon/R600_rlc.bin"); MODULE_FIRMWARE("radeon/R700_rlc.bin"); MODULE_FIRMWARE("radeon/CEDAR_pfp.bin"); MODULE_FIRMWARE("radeon/CEDAR_me.bin"); MODULE_FIRMWARE("radeon/CEDAR_rlc.bin"); +MODULE_FIRMWARE("radeon/CEDAR_smc.bin"); MODULE_FIRMWARE("radeon/REDWOOD_pfp.bin"); MODULE_FIRMWARE("radeon/REDWOOD_me.bin"); MODULE_FIRMWARE("radeon/REDWOOD_rlc.bin"); +MODULE_FIRMWARE("radeon/REDWOOD_smc.bin"); MODULE_FIRMWARE("radeon/JUNIPER_pfp.bin"); MODULE_FIRMWARE("radeon/JUNIPER_me.bin"); MODULE_FIRMWARE("radeon/JUNIPER_rlc.bin"); +MODULE_FIRMWARE("radeon/JUNIPER_smc.bin"); MODULE_FIRMWARE("radeon/CYPRESS_pfp.bin"); MODULE_FIRMWARE("radeon/CYPRESS_me.bin"); MODULE_FIRMWARE("radeon/CYPRESS_rlc.bin"); +MODULE_FIRMWARE("radeon/CYPRESS_smc.bin"); MODULE_FIRMWARE("radeon/PALM_pfp.bin"); MODULE_FIRMWARE("radeon/PALM_me.bin"); MODULE_FIRMWARE("radeon/SUMO_rlc.bin"); @@ -2158,7 +2155,8 @@ int r600_init_microcode(struct radeon_device *rdev) { const char *chip_name; const char *rlc_chip_name; - size_t pfp_req_size, me_req_size, rlc_req_size; + const char *smc_chip_name = "RV770"; + size_t pfp_req_size, me_req_size, rlc_req_size, smc_req_size = 0; char fw_name[30]; int err; @@ -2197,32 +2195,51 @@ int r600_init_microcode(struct radeon_device *rdev) case CHIP_RV770: chip_name = "RV770"; rlc_chip_name = "R700"; + smc_chip_name = "RV770"; + smc_req_size = ALIGN(RV770_SMC_UCODE_SIZE, 4); break; case CHIP_RV730: - case CHIP_RV740: chip_name = "RV730"; rlc_chip_name = "R700"; + smc_chip_name = "RV730"; + smc_req_size = ALIGN(RV730_SMC_UCODE_SIZE, 4); break; case CHIP_RV710: chip_name = "RV710"; rlc_chip_name = "R700"; + smc_chip_name = "RV710"; + smc_req_size = ALIGN(RV710_SMC_UCODE_SIZE, 4); + break; + case CHIP_RV740: + chip_name = "RV730"; + rlc_chip_name = "R700"; + smc_chip_name = "RV740"; + smc_req_size = ALIGN(RV740_SMC_UCODE_SIZE, 4); break; case CHIP_CEDAR: chip_name = "CEDAR"; rlc_chip_name = "CEDAR"; + smc_chip_name = "CEDAR"; + smc_req_size = ALIGN(CEDAR_SMC_UCODE_SIZE, 4); break; case CHIP_REDWOOD: chip_name = "REDWOOD"; rlc_chip_name = "REDWOOD"; + smc_chip_name = "REDWOOD"; + smc_req_size = ALIGN(REDWOOD_SMC_UCODE_SIZE, 4); break; case CHIP_JUNIPER: chip_name = "JUNIPER"; rlc_chip_name = "JUNIPER"; + smc_chip_name = "JUNIPER"; + smc_req_size = ALIGN(JUNIPER_SMC_UCODE_SIZE, 4); break; case CHIP_CYPRESS: case CHIP_HEMLOCK: chip_name = "CYPRESS"; rlc_chip_name = "CYPRESS"; + smc_chip_name = "CYPRESS"; + smc_req_size = ALIGN(CYPRESS_SMC_UCODE_SIZE, 4); break; case CHIP_PALM: chip_name = "PALM"; @@ -2248,9 +2265,9 @@ int r600_init_microcode(struct radeon_device *rdev) me_req_size = R700_PM4_UCODE_SIZE * 4; rlc_req_size = R700_RLC_UCODE_SIZE * 4; } else { - pfp_req_size = PFP_UCODE_SIZE * 4; - me_req_size = PM4_UCODE_SIZE * 12; - rlc_req_size = RLC_UCODE_SIZE * 4; + pfp_req_size = R600_PFP_UCODE_SIZE * 4; + me_req_size = R600_PM4_UCODE_SIZE * 12; + rlc_req_size = R600_RLC_UCODE_SIZE * 4; } DRM_INFO("Loading %s Microcode\n", chip_name); @@ -2297,6 +2314,23 @@ int r600_init_microcode(struct radeon_device *rdev) err = -EINVAL; } + if ((rdev->family >= CHIP_RV770) && (rdev->family <= CHIP_HEMLOCK)) { + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_smc", smc_chip_name); + err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev); + if (err) { + printk(KERN_ERR + "smc: error loading firmware \"%s\"\n", + fw_name); + release_firmware(rdev->smc_fw); + rdev->smc_fw = NULL; + } else if (rdev->smc_fw->datasize != smc_req_size) { + printk(KERN_ERR + "smc: Bogus length %zu in firmware \"%s\"\n", + rdev->smc_fw->datasize, fw_name); + err = -EINVAL; + } + } + out: if (err) { if (err != -EINVAL) @@ -2343,6 +2377,8 @@ void r600_fini_microcode(struct radeon_device *rdev) if (rdev->rlc_fw != NULL) { firmware_put(rdev->rlc_fw, FIRMWARE_UNLOAD); rdev->rlc_fw = NULL; + firmware_put(rdev->smc_fw, FIRMWARE_UNLOAD); + rdev->smc_fw = NULL; } } @@ -2372,13 +2408,13 @@ static int r600_cp_load_microcode(struct radeon_device *rdev) fw_data = (const __be32 *)rdev->me_fw->data; WREG32(CP_ME_RAM_WADDR, 0); - for (i = 0; i < PM4_UCODE_SIZE * 3; i++) + for (i = 0; i < R600_PM4_UCODE_SIZE * 3; i++) WREG32(CP_ME_RAM_DATA, be32_to_cpup(fw_data++)); fw_data = (const __be32 *)rdev->pfp_fw->data; WREG32(CP_PFP_UCODE_ADDR, 0); - for (i = 0; i < PFP_UCODE_SIZE; i++) + for (i = 0; i < R600_PFP_UCODE_SIZE; i++) WREG32(CP_PFP_UCODE_DATA, be32_to_cpup(fw_data++)); @@ -2716,12 +2752,29 @@ int r600_uvd_rbc_start(struct radeon_device *rdev) return 0; } -void r600_uvd_rbc_stop(struct radeon_device *rdev) +void r600_uvd_stop(struct radeon_device *rdev) { struct radeon_ring *ring = &rdev->ring[R600_RING_TYPE_UVD_INDEX]; /* force RBC into idle state */ WREG32(UVD_RBC_RB_CNTL, 0x11010101); + + /* Stall UMC and register bus before resetting VCPU */ + WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); + DRM_MDELAY(1); + + /* put VCPU into reset */ + WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); + DRM_MDELAY(5); + + /* disable VCPU clock */ + WREG32(UVD_VCPU_CNTL, 0x0); + + /* Unstall UMC and register bus */ + WREG32_P(UVD_LMI_CTRL2, 0, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 0, ~(1 << 3)); + ring->ready = false; } @@ -2741,6 +2794,11 @@ int r600_uvd_init(struct radeon_device *rdev) /* disable interupt */ WREG32_P(UVD_MASTINT_EN, 0, ~(1 << 1)); + /* Stall UMC and register bus before resetting VCPU */ + WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); + WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); + DRM_MDELAY(1); + /* put LMI, VCPU, RBC etc... into reset */ WREG32(UVD_SOFT_RESET, LMI_SOFT_RESET | VCPU_SOFT_RESET | LBSI_SOFT_RESET | RBC_SOFT_RESET | CSM_SOFT_RESET | @@ -2770,10 +2828,6 @@ int r600_uvd_init(struct radeon_device *rdev) WREG32(UVD_MPC_SET_ALU, 0); WREG32(UVD_MPC_SET_MUX, 0x88); - /* Stall UMC */ - WREG32_P(UVD_LMI_CTRL2, 1 << 8, ~(1 << 8)); - WREG32_P(UVD_RB_ARB_CTRL, 1 << 3, ~(1 << 3)); - /* take all subblocks out of reset, except VCPU */ WREG32(UVD_SOFT_RESET, VCPU_SOFT_RESET); DRM_MDELAY(5); @@ -3027,7 +3081,7 @@ void r600_uvd_fence_emit(struct radeon_device *rdev, struct radeon_fence *fence) { struct radeon_ring *ring = &rdev->ring[fence->ring]; - uint32_t addr = rdev->fence_drv[fence->ring].gpu_addr; + uint64_t addr = rdev->fence_drv[fence->ring].gpu_addr; radeon_ring_write(ring, PACKET0(UVD_CONTEXT_ID, 0)); radeon_ring_write(ring, fence->seq); @@ -3152,6 +3206,90 @@ int r600_copy_blit(struct radeon_device *rdev, return 0; } +/** + * r600_copy_cpdma - copy pages using the CP DMA engine + * + * @rdev: radeon_device pointer + * @src_offset: src GPU address + * @dst_offset: dst GPU address + * @num_gpu_pages: number of GPU pages to xfer + * @fence: radeon fence object + * + * Copy GPU paging using the CP DMA engine (r6xx+). + * Used by the radeon ttm implementation to move pages if + * registered as the asic copy callback. + */ +int r600_copy_cpdma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence) +{ + struct radeon_semaphore *sem = NULL; + int ring_index = rdev->asic->copy.blit_ring_index; + struct radeon_ring *ring = &rdev->ring[ring_index]; + u32 size_in_bytes, cur_size_in_bytes, tmp; + int i, num_loops; + int r = 0; + + r = radeon_semaphore_create(rdev, &sem); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + return r; + } + + size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT); + num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff); + r = radeon_ring_lock(rdev, ring, num_loops * 6 + 24); + if (r) { + DRM_ERROR("radeon: moving bo (%d).\n", r); + radeon_semaphore_free(rdev, &sem, NULL); + return r; + } + + if (radeon_fence_need_sync(*fence, ring->idx)) { + radeon_semaphore_sync_rings(rdev, sem, (*fence)->ring, + ring->idx); + radeon_fence_note_sync(*fence, ring->idx); + } else { + radeon_semaphore_free(rdev, &sem, NULL); + } + + radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); + radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); + radeon_ring_write(ring, WAIT_3D_IDLE_bit); + for (i = 0; i < num_loops; i++) { + cur_size_in_bytes = size_in_bytes; + if (cur_size_in_bytes > 0x1fffff) + cur_size_in_bytes = 0x1fffff; + size_in_bytes -= cur_size_in_bytes; + tmp = upper_32_bits(src_offset) & 0xff; + if (size_in_bytes == 0) + tmp |= PACKET3_CP_DMA_CP_SYNC; + radeon_ring_write(ring, PACKET3(PACKET3_CP_DMA, 4)); + radeon_ring_write(ring, src_offset & 0xffffffff); + radeon_ring_write(ring, tmp); + radeon_ring_write(ring, dst_offset & 0xffffffff); + radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff); + radeon_ring_write(ring, cur_size_in_bytes); + src_offset += cur_size_in_bytes; + dst_offset += cur_size_in_bytes; + } + radeon_ring_write(ring, PACKET3(PACKET3_SET_CONFIG_REG, 1)); + radeon_ring_write(ring, (WAIT_UNTIL - PACKET3_SET_CONFIG_REG_OFFSET) >> 2); + radeon_ring_write(ring, WAIT_CP_DMA_IDLE_bit); + + r = radeon_fence_emit(rdev, fence, ring->idx); + if (r) { + radeon_ring_unlock_undo(rdev, ring); + return r; + } + + radeon_ring_unlock_commit(rdev, ring); + radeon_semaphore_free(rdev, &sem, *fence); + + return r; +} + /** * r600_copy_dma - copy pages using the DMA engine * @@ -3247,6 +3385,8 @@ static int r600_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ r600_pcie_gen2_enable(rdev); + r600_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { r = r600_init_microcode(rdev); if (r) { @@ -3259,7 +3399,6 @@ static int r600_startup(struct radeon_device *rdev) if (r) return r; - r600_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { r600_agp_enable(rdev); } else { @@ -3840,7 +3979,7 @@ static void r600_rlc_start(struct radeon_device *rdev) WREG32(RLC_CNTL, RLC_ENABLE); } -static int r600_rlc_init(struct radeon_device *rdev) +static int r600_rlc_resume(struct radeon_device *rdev) { u32 i; const __be32 *fw_data; @@ -3852,45 +3991,22 @@ static int r600_rlc_init(struct radeon_device *rdev) WREG32(RLC_HB_CNTL, 0); - if (rdev->family == CHIP_ARUBA) { - WREG32(TN_RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); - WREG32(TN_RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8); - } - if (rdev->family <= CHIP_CAYMAN) { - WREG32(RLC_HB_BASE, 0); - WREG32(RLC_HB_RPTR, 0); - WREG32(RLC_HB_WPTR, 0); - } - if (rdev->family <= CHIP_CAICOS) { - WREG32(RLC_HB_WPTR_LSB_ADDR, 0); - WREG32(RLC_HB_WPTR_MSB_ADDR, 0); - } + WREG32(RLC_HB_BASE, 0); + WREG32(RLC_HB_RPTR, 0); + WREG32(RLC_HB_WPTR, 0); + WREG32(RLC_HB_WPTR_LSB_ADDR, 0); + WREG32(RLC_HB_WPTR_MSB_ADDR, 0); WREG32(RLC_MC_CNTL, 0); WREG32(RLC_UCODE_CNTL, 0); fw_data = (const __be32 *)rdev->rlc_fw->data; - if (rdev->family >= CHIP_ARUBA) { - for (i = 0; i < ARUBA_RLC_UCODE_SIZE; i++) { - WREG32(RLC_UCODE_ADDR, i); - WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); - } - } else if (rdev->family >= CHIP_CAYMAN) { - for (i = 0; i < CAYMAN_RLC_UCODE_SIZE; i++) { - WREG32(RLC_UCODE_ADDR, i); - WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); - } - } else if (rdev->family >= CHIP_CEDAR) { - for (i = 0; i < EVERGREEN_RLC_UCODE_SIZE; i++) { - WREG32(RLC_UCODE_ADDR, i); - WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); - } - } else if (rdev->family >= CHIP_RV770) { + if (rdev->family >= CHIP_RV770) { for (i = 0; i < R700_RLC_UCODE_SIZE; i++) { WREG32(RLC_UCODE_ADDR, i); WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); } } else { - for (i = 0; i < RLC_UCODE_SIZE; i++) { + for (i = 0; i < R600_RLC_UCODE_SIZE; i++) { WREG32(RLC_UCODE_ADDR, i); WREG32(RLC_UCODE_DATA, be32_to_cpup(fw_data++)); } @@ -3998,7 +4114,10 @@ int r600_irq_init(struct radeon_device *rdev) r600_disable_interrupts(rdev); /* init rlc */ - ret = r600_rlc_init(rdev); + if (rdev->family >= CHIP_CEDAR) + ret = evergreen_rlc_resume(rdev); + else + ret = r600_rlc_resume(rdev); if (ret) { r600_ih_ring_fini(rdev); return ret; @@ -4079,6 +4198,7 @@ int r600_irq_set(struct radeon_device *rdev) u32 hdmi0, hdmi1; u32 d1grph = 0, d2grph = 0; u32 dma_cntl; + u32 thermal_int = 0; if (!rdev->irq.installed) { DRM_ERROR("Can't enable IRQ/MSI because no handler is installed\n"); @@ -4113,8 +4233,21 @@ int r600_irq_set(struct radeon_device *rdev) hdmi0 = RREG32(HDMI0_AUDIO_PACKET_CONTROL) & ~HDMI0_AZ_FORMAT_WTRIG_MASK; hdmi1 = RREG32(HDMI1_AUDIO_PACKET_CONTROL) & ~HDMI0_AZ_FORMAT_WTRIG_MASK; } + dma_cntl = RREG32(DMA_CNTL) & ~TRAP_ENABLE; + if ((rdev->family > CHIP_R600) && (rdev->family < CHIP_RV770)) { + thermal_int = RREG32(CG_THERMAL_INT) & + ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + } else if (rdev->family >= CHIP_RV770) { + thermal_int = RREG32(RV770_CG_THERMAL_INT) & + ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + } + if (rdev->irq.dpm_thermal) { + DRM_DEBUG("dpm thermal\n"); + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + } + if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("r600_irq_set: sw int\n"); cp_int_cntl |= RB_INT_ENABLE; @@ -4196,6 +4329,11 @@ int r600_irq_set(struct radeon_device *rdev) WREG32(HDMI0_AUDIO_PACKET_CONTROL, hdmi0); WREG32(HDMI1_AUDIO_PACKET_CONTROL, hdmi1); } + if ((rdev->family > CHIP_R600) && (rdev->family < CHIP_RV770)) { + WREG32(CG_THERMAL_INT, thermal_int); + } else if (rdev->family >= CHIP_RV770) { + WREG32(RV770_CG_THERMAL_INT, thermal_int); + } return 0; } @@ -4387,6 +4525,7 @@ irqreturn_t r600_irq_process(struct radeon_device *rdev) u32 ring_index; bool queue_hotplug = false; bool queue_hdmi = false; + bool queue_thermal = false; if (!rdev->ih.enabled || rdev->shutdown) return IRQ_NONE; @@ -4554,6 +4693,16 @@ restart_ih: DRM_DEBUG("IH: DMA trap\n"); radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX); break; + case 230: /* thermal low to high */ + DRM_DEBUG("IH: thermal low to high\n"); + rdev->pm.dpm.thermal.high_to_low = false; + queue_thermal = true; + break; + case 231: /* thermal high to low */ + DRM_DEBUG("IH: thermal high to low\n"); + rdev->pm.dpm.thermal.high_to_low = true; + queue_thermal = true; + break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); break; @@ -4570,6 +4719,9 @@ restart_ih: taskqueue_enqueue(rdev->tq, &rdev->hotplug_work); if (queue_hdmi) taskqueue_enqueue(rdev->tq, &rdev->audio_work); + if (queue_thermal && rdev->pm.dpm_enabled) + taskqueue_enqueue(rdev->tq, &rdev->pm.dpm.thermal.work); + rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); atomic_set(&rdev->ih.lock, 0); diff --git a/sys/dev/drm/radeon/r600_blit_shaders.c b/sys/dev/drm/radeon/r600_blit_shaders.c index db962c06bb..64599cd400 100644 --- a/sys/dev/drm/radeon/r600_blit_shaders.c +++ b/sys/dev/drm/radeon/r600_blit_shaders.c @@ -22,8 +22,6 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/r600_blit_shaders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/r600_blit_shaders.h b/sys/dev/drm/radeon/r600_blit_shaders.h index b2c080b7b5..2f3ce7a759 100644 --- a/sys/dev/drm/radeon/r600_blit_shaders.h +++ b/sys/dev/drm/radeon/r600_blit_shaders.h @@ -21,7 +21,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/r600_blit_shaders.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef R600_BLIT_SHADERS_H diff --git a/sys/dev/drm/radeon/r600_dpm.c b/sys/dev/drm/radeon/r600_dpm.c new file mode 100644 index 0000000000..444fc2e0f9 --- /dev/null +++ b/sys/dev/drm/radeon/r600_dpm.c @@ -0,0 +1,1049 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "r600d.h" +#include "r600_dpm.h" +#include "atom.h" + +const u32 r600_utc[R600_PM_NUMBER_OF_TC] = +{ + R600_UTC_DFLT_00, + R600_UTC_DFLT_01, + R600_UTC_DFLT_02, + R600_UTC_DFLT_03, + R600_UTC_DFLT_04, + R600_UTC_DFLT_05, + R600_UTC_DFLT_06, + R600_UTC_DFLT_07, + R600_UTC_DFLT_08, + R600_UTC_DFLT_09, + R600_UTC_DFLT_10, + R600_UTC_DFLT_11, + R600_UTC_DFLT_12, + R600_UTC_DFLT_13, + R600_UTC_DFLT_14, +}; + +const u32 r600_dtc[R600_PM_NUMBER_OF_TC] = +{ + R600_DTC_DFLT_00, + R600_DTC_DFLT_01, + R600_DTC_DFLT_02, + R600_DTC_DFLT_03, + R600_DTC_DFLT_04, + R600_DTC_DFLT_05, + R600_DTC_DFLT_06, + R600_DTC_DFLT_07, + R600_DTC_DFLT_08, + R600_DTC_DFLT_09, + R600_DTC_DFLT_10, + R600_DTC_DFLT_11, + R600_DTC_DFLT_12, + R600_DTC_DFLT_13, + R600_DTC_DFLT_14, +}; + +void r600_dpm_print_class_info(u32 class, u32 class2) +{ + printk("\tui class: "); + switch (class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) { + case ATOM_PPLIB_CLASSIFICATION_UI_NONE: + default: + printk("none\n"); + break; + case ATOM_PPLIB_CLASSIFICATION_UI_BATTERY: + printk("battery\n"); + break; + case ATOM_PPLIB_CLASSIFICATION_UI_BALANCED: + printk("balanced\n"); + break; + case ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE: + printk("performance\n"); + break; + } + printk("\tinternal class: "); + if (((class & ~ATOM_PPLIB_CLASSIFICATION_UI_MASK) == 0) && + (class2 == 0)) + printk("none"); + else { + if (class & ATOM_PPLIB_CLASSIFICATION_BOOT) + printk("boot "); + if (class & ATOM_PPLIB_CLASSIFICATION_THERMAL) + printk("thermal "); + if (class & ATOM_PPLIB_CLASSIFICATION_LIMITEDPOWERSOURCE) + printk("limited_pwr "); + if (class & ATOM_PPLIB_CLASSIFICATION_REST) + printk("rest "); + if (class & ATOM_PPLIB_CLASSIFICATION_FORCED) + printk("forced "); + if (class & ATOM_PPLIB_CLASSIFICATION_3DPERFORMANCE) + printk("3d_perf "); + if (class & ATOM_PPLIB_CLASSIFICATION_OVERDRIVETEMPLATE) + printk("ovrdrv "); + if (class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + printk("uvd "); + if (class & ATOM_PPLIB_CLASSIFICATION_3DLOW) + printk("3d_low "); + if (class & ATOM_PPLIB_CLASSIFICATION_ACPI) + printk("acpi "); + if (class & ATOM_PPLIB_CLASSIFICATION_HD2STATE) + printk("uvd_hd2 "); + if (class & ATOM_PPLIB_CLASSIFICATION_HDSTATE) + printk("uvd_hd "); + if (class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) + printk("uvd_sd "); + if (class2 & ATOM_PPLIB_CLASSIFICATION2_LIMITEDPOWERSOURCE_2) + printk("limited_pwr2 "); + if (class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) + printk("ulv "); + if (class2 & ATOM_PPLIB_CLASSIFICATION2_MVC) + printk("uvd_mvc "); + } + printk("\n"); +} + +void r600_dpm_print_cap_info(u32 caps) +{ + printk("\tcaps: "); + if (caps & ATOM_PPLIB_SINGLE_DISPLAY_ONLY) + printk("single_disp "); + if (caps & ATOM_PPLIB_SUPPORTS_VIDEO_PLAYBACK) + printk("video "); + if (caps & ATOM_PPLIB_DISALLOW_ON_DC) + printk("no_dc "); + printk("\n"); +} + +void r600_dpm_print_ps_status(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + printk("\tstatus: "); + if (rps == rdev->pm.dpm.current_ps) + printk("c "); + if (rps == rdev->pm.dpm.requested_ps) + printk("r "); + if (rps == rdev->pm.dpm.boot_ps) + printk("b "); + printk("\n"); +} + +u32 r600_dpm_get_vblank_time(struct radeon_device *rdev) +{ + struct drm_device *dev = rdev->ddev; + struct drm_crtc *crtc; + struct radeon_crtc *radeon_crtc; + u32 line_time_us, vblank_lines; + u32 vblank_time_us = 0xffffffff; /* if the displays are off, vblank time is max */ + + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { + radeon_crtc = to_radeon_crtc(crtc); + if (crtc->enabled && radeon_crtc->enabled && radeon_crtc->hw_mode.clock) { + line_time_us = (radeon_crtc->hw_mode.crtc_htotal * 1000) / + radeon_crtc->hw_mode.clock; + vblank_lines = radeon_crtc->hw_mode.crtc_vblank_end - + radeon_crtc->hw_mode.crtc_vdisplay + + (radeon_crtc->v_border * 2); + vblank_time_us = vblank_lines * line_time_us; + break; + } + } + + return vblank_time_us; +} + +void r600_calculate_u_and_p(u32 i, u32 r_c, u32 p_b, + u32 *p, u32 *u) +{ + u32 b_c = 0; + u32 i_c; + u32 tmp; + + i_c = (i * r_c) / 100; + tmp = i_c >> p_b; + + while (tmp) { + b_c++; + tmp >>= 1; + } + + *u = (b_c + 1) / 2; + *p = i_c / (1 << (2 * (*u))); +} + +int r600_calculate_at(u32 t, u32 h, u32 fh, u32 fl, u32 *tl, u32 *th) +{ + u32 k, a, ah, al; + u32 t1; + + if ((fl == 0) || (fh == 0) || (fl > fh)) + return -EINVAL; + + k = (100 * fh) / fl; + t1 = (t * (k - 100)); + a = (1000 * (100 * h + t1)) / (10000 + (t1 / 100)); + a = (a + 5) / 10; + ah = ((a * t) + 5000) / 10000; + al = a - ah; + + *th = t - ah; + *tl = t + al; + + return 0; +} + +void r600_gfx_clockgating_enable(struct radeon_device *rdev, bool enable) +{ + int i; + + if (enable) { + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + } else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + + WREG32(CG_RLC_REQ_AND_RSP, 0x2); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (((RREG32(CG_RLC_REQ_AND_RSP) & CG_RLC_RSP_TYPE_MASK) >> CG_RLC_RSP_TYPE_SHIFT) == 1) + break; + DRM_UDELAY(1); + } + + WREG32(CG_RLC_REQ_AND_RSP, 0x0); + + WREG32(GRBM_PWR_CNTL, 0x1); + RREG32(GRBM_PWR_CNTL); + } +} + +void r600_dynamicpm_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~GLOBAL_PWRMGT_EN); +} + +void r600_enable_thermal_protection(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + else + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); +} + +void r600_enable_acpi_pm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, STATIC_PM_EN, ~STATIC_PM_EN); +} + +void r600_enable_dynamic_pcie_gen2(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, ENABLE_GEN2PCIE, ~ENABLE_GEN2PCIE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~ENABLE_GEN2PCIE); +} + +bool r600_dynamicpm_enabled(struct radeon_device *rdev) +{ + if (RREG32(GENERAL_PWRMGT) & GLOBAL_PWRMGT_EN) + return true; + else + return false; +} + +void r600_enable_sclk_control(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~SCLK_PWRMGT_OFF); + else + WREG32_P(SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, ~SCLK_PWRMGT_OFF); +} + +void r600_enable_mclk_control(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(MCLK_PWRMGT_CNTL, 0, ~MPLL_PWRMGT_OFF); + else + WREG32_P(MCLK_PWRMGT_CNTL, MPLL_PWRMGT_OFF, ~MPLL_PWRMGT_OFF); +} + +void r600_enable_spll_bypass(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(CG_SPLL_FUNC_CNTL, SPLL_BYPASS_EN, ~SPLL_BYPASS_EN); + else + WREG32_P(CG_SPLL_FUNC_CNTL, 0, ~SPLL_BYPASS_EN); +} + +void r600_wait_for_spll_change(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(CG_SPLL_FUNC_CNTL) & SPLL_CHG_STATUS) + break; + DRM_UDELAY(1); + } +} + +void r600_set_bsp(struct radeon_device *rdev, u32 u, u32 p) +{ + WREG32(CG_BSP, BSP(p) | BSU(u)); +} + +void r600_set_at(struct radeon_device *rdev, + u32 l_to_m, u32 m_to_h, + u32 h_to_m, u32 m_to_l) +{ + WREG32(CG_RT, FLS(l_to_m) | FMS(m_to_h)); + WREG32(CG_LT, FHS(h_to_m) | FMS(m_to_l)); +} + +void r600_set_tc(struct radeon_device *rdev, + u32 index, u32 u_t, u32 d_t) +{ + WREG32(CG_FFCT_0 + (index * 4), UTC_0(u_t) | DTC_0(d_t)); +} + +void r600_select_td(struct radeon_device *rdev, + enum r600_td td) +{ + if (td == R600_TD_AUTO) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_FORCE_TREND_SEL); + else + WREG32_P(SCLK_PWRMGT_CNTL, FIR_FORCE_TREND_SEL, ~FIR_FORCE_TREND_SEL); + if (td == R600_TD_UP) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_TREND_MODE); + if (td == R600_TD_DOWN) + WREG32_P(SCLK_PWRMGT_CNTL, FIR_TREND_MODE, ~FIR_TREND_MODE); +} + +void r600_set_vrc(struct radeon_device *rdev, u32 vrv) +{ + WREG32(CG_FTV, vrv); +} + +void r600_set_tpu(struct radeon_device *rdev, u32 u) +{ + WREG32_P(CG_TPC, TPU(u), ~TPU_MASK); +} + +void r600_set_tpc(struct radeon_device *rdev, u32 c) +{ + WREG32_P(CG_TPC, TPCC(c), ~TPCC_MASK); +} + +void r600_set_sstu(struct radeon_device *rdev, u32 u) +{ + WREG32_P(CG_SSP, CG_SSTU(u), ~CG_SSTU_MASK); +} + +void r600_set_sst(struct radeon_device *rdev, u32 t) +{ + WREG32_P(CG_SSP, CG_SST(t), ~CG_SST_MASK); +} + +void r600_set_git(struct radeon_device *rdev, u32 t) +{ + WREG32_P(CG_GIT, CG_GICST(t), ~CG_GICST_MASK); +} + +void r600_set_fctu(struct radeon_device *rdev, u32 u) +{ + WREG32_P(CG_FC_T, FC_TU(u), ~FC_TU_MASK); +} + +void r600_set_fct(struct radeon_device *rdev, u32 t) +{ + WREG32_P(CG_FC_T, FC_T(t), ~FC_T_MASK); +} + +void r600_set_ctxcgtt3d_rphc(struct radeon_device *rdev, u32 p) +{ + WREG32_P(CG_CTX_CGTT3D_R, PHC(p), ~PHC_MASK); +} + +void r600_set_ctxcgtt3d_rsdc(struct radeon_device *rdev, u32 s) +{ + WREG32_P(CG_CTX_CGTT3D_R, SDC(s), ~SDC_MASK); +} + +void r600_set_vddc3d_oorsu(struct radeon_device *rdev, u32 u) +{ + WREG32_P(CG_VDDC3D_OOR, SU(u), ~SU_MASK); +} + +void r600_set_vddc3d_oorphc(struct radeon_device *rdev, u32 p) +{ + WREG32_P(CG_VDDC3D_OOR, PHC(p), ~PHC_MASK); +} + +void r600_set_vddc3d_oorsdc(struct radeon_device *rdev, u32 s) +{ + WREG32_P(CG_VDDC3D_OOR, SDC(s), ~SDC_MASK); +} + +void r600_set_mpll_lock_time(struct radeon_device *rdev, u32 lock_time) +{ + WREG32_P(MPLL_TIME, MPLL_LOCK_TIME(lock_time), ~MPLL_LOCK_TIME_MASK); +} + +void r600_set_mpll_reset_time(struct radeon_device *rdev, u32 reset_time) +{ + WREG32_P(MPLL_TIME, MPLL_RESET_TIME(reset_time), ~MPLL_RESET_TIME_MASK); +} + +void r600_engine_clock_entry_enable(struct radeon_device *rdev, + u32 index, bool enable) +{ + if (enable) + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + STEP_0_SPLL_ENTRY_VALID, ~STEP_0_SPLL_ENTRY_VALID); + else + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + 0, ~STEP_0_SPLL_ENTRY_VALID); +} + +void r600_engine_clock_entry_enable_pulse_skipping(struct radeon_device *rdev, + u32 index, bool enable) +{ + if (enable) + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + STEP_0_SPLL_STEP_ENABLE, ~STEP_0_SPLL_STEP_ENABLE); + else + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + 0, ~STEP_0_SPLL_STEP_ENABLE); +} + +void r600_engine_clock_entry_enable_post_divider(struct radeon_device *rdev, + u32 index, bool enable) +{ + if (enable) + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + STEP_0_POST_DIV_EN, ~STEP_0_POST_DIV_EN); + else + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART2 + (index * 4 * 2), + 0, ~STEP_0_POST_DIV_EN); +} + +void r600_engine_clock_entry_set_post_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART1 + (index * 4 * 2), + STEP_0_SPLL_POST_DIV(divider), ~STEP_0_SPLL_POST_DIV_MASK); +} + +void r600_engine_clock_entry_set_reference_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART1 + (index * 4 * 2), + STEP_0_SPLL_REF_DIV(divider), ~STEP_0_SPLL_REF_DIV_MASK); +} + +void r600_engine_clock_entry_set_feedback_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART1 + (index * 4 * 2), + STEP_0_SPLL_FB_DIV(divider), ~STEP_0_SPLL_FB_DIV_MASK); +} + +void r600_engine_clock_entry_set_step_time(struct radeon_device *rdev, + u32 index, u32 step_time) +{ + WREG32_P(SCLK_FREQ_SETTING_STEP_0_PART1 + (index * 4 * 2), + STEP_0_SPLL_STEP_TIME(step_time), ~STEP_0_SPLL_STEP_TIME_MASK); +} + +void r600_vid_rt_set_ssu(struct radeon_device *rdev, u32 u) +{ + WREG32_P(VID_RT, SSTU(u), ~SSTU_MASK); +} + +void r600_vid_rt_set_vru(struct radeon_device *rdev, u32 u) +{ + WREG32_P(VID_RT, VID_CRTU(u), ~VID_CRTU_MASK); +} + +void r600_vid_rt_set_vrt(struct radeon_device *rdev, u32 rt) +{ + WREG32_P(VID_RT, VID_CRT(rt), ~VID_CRT_MASK); +} + +void r600_voltage_control_enable_pins(struct radeon_device *rdev, + u64 mask) +{ + WREG32(LOWER_GPIO_ENABLE, mask & 0xffffffff); + WREG32(UPPER_GPIO_ENABLE, upper_32_bits(mask)); +} + + +void r600_voltage_control_program_voltages(struct radeon_device *rdev, + enum r600_power_level index, u64 pins) +{ + u32 tmp, mask; + u32 ix = 3 - (3 & index); + + WREG32(CTXSW_VID_LOWER_GPIO_CNTL + (ix * 4), pins & 0xffffffff); + + mask = 7 << (3 * ix); + tmp = RREG32(VID_UPPER_GPIO_CNTL); + tmp = (tmp & ~mask) | ((pins >> (32 - (3 * ix))) & mask); + WREG32(VID_UPPER_GPIO_CNTL, tmp); +} + +void r600_voltage_control_deactivate_static_control(struct radeon_device *rdev, + u64 mask) +{ + u32 gpio; + + gpio = RREG32(GPIOPAD_MASK); + gpio &= ~mask; + WREG32(GPIOPAD_MASK, gpio); + + gpio = RREG32(GPIOPAD_EN); + gpio &= ~mask; + WREG32(GPIOPAD_EN, gpio); + + gpio = RREG32(GPIOPAD_A); + gpio &= ~mask; + WREG32(GPIOPAD_A, gpio); +} + +void r600_power_level_enable(struct radeon_device *rdev, + enum r600_power_level index, bool enable) +{ + u32 ix = 3 - (3 & index); + + if (enable) + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), CTXSW_FREQ_STATE_ENABLE, + ~CTXSW_FREQ_STATE_ENABLE); + else + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), 0, + ~CTXSW_FREQ_STATE_ENABLE); +} + +void r600_power_level_set_voltage_index(struct radeon_device *rdev, + enum r600_power_level index, u32 voltage_index) +{ + u32 ix = 3 - (3 & index); + + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), + CTXSW_FREQ_VIDS_CFG_INDEX(voltage_index), ~CTXSW_FREQ_VIDS_CFG_INDEX_MASK); +} + +void r600_power_level_set_mem_clock_index(struct radeon_device *rdev, + enum r600_power_level index, u32 mem_clock_index) +{ + u32 ix = 3 - (3 & index); + + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), + CTXSW_FREQ_MCLK_CFG_INDEX(mem_clock_index), ~CTXSW_FREQ_MCLK_CFG_INDEX_MASK); +} + +void r600_power_level_set_eng_clock_index(struct radeon_device *rdev, + enum r600_power_level index, u32 eng_clock_index) +{ + u32 ix = 3 - (3 & index); + + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), + CTXSW_FREQ_SCLK_CFG_INDEX(eng_clock_index), ~CTXSW_FREQ_SCLK_CFG_INDEX_MASK); +} + +void r600_power_level_set_watermark_id(struct radeon_device *rdev, + enum r600_power_level index, + enum r600_display_watermark watermark_id) +{ + u32 ix = 3 - (3 & index); + u32 tmp = 0; + + if (watermark_id == R600_DISPLAY_WATERMARK_HIGH) + tmp = CTXSW_FREQ_DISPLAY_WATERMARK; + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), tmp, ~CTXSW_FREQ_DISPLAY_WATERMARK); +} + +void r600_power_level_set_pcie_gen2(struct radeon_device *rdev, + enum r600_power_level index, bool compatible) +{ + u32 ix = 3 - (3 & index); + u32 tmp = 0; + + if (compatible) + tmp = CTXSW_FREQ_GEN2PCIE_VOLT; + WREG32_P(CTXSW_PROFILE_INDEX + (ix * 4), tmp, ~CTXSW_FREQ_GEN2PCIE_VOLT); +} + +enum r600_power_level r600_power_level_get_current_index(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_PROFILE_INDEX_MASK; + tmp >>= CURRENT_PROFILE_INDEX_SHIFT; + return tmp; +} + +enum r600_power_level r600_power_level_get_target_index(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & TARGET_PROFILE_INDEX_MASK; + tmp >>= TARGET_PROFILE_INDEX_SHIFT; + return tmp; +} + +void r600_power_level_set_enter_index(struct radeon_device *rdev, + enum r600_power_level index) +{ + WREG32_P(TARGET_AND_CURRENT_PROFILE_INDEX, DYN_PWR_ENTER_INDEX(index), + ~DYN_PWR_ENTER_INDEX_MASK); +} + +void r600_wait_for_power_level_unequal(struct radeon_device *rdev, + enum r600_power_level index) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (r600_power_level_get_target_index(rdev) != index) + break; + DRM_UDELAY(1); + } + + for (i = 0; i < rdev->usec_timeout; i++) { + if (r600_power_level_get_current_index(rdev) != index) + break; + DRM_UDELAY(1); + } +} + +void r600_wait_for_power_level(struct radeon_device *rdev, + enum r600_power_level index) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (r600_power_level_get_target_index(rdev) == index) + break; + DRM_UDELAY(1); + } + + for (i = 0; i < rdev->usec_timeout; i++) { + if (r600_power_level_get_current_index(rdev) == index) + break; + DRM_UDELAY(1); + } +} + +void r600_start_dpm(struct radeon_device *rdev) +{ + r600_enable_sclk_control(rdev, false); + r600_enable_mclk_control(rdev, false); + + r600_dynamicpm_enable(rdev, true); + + radeon_wait_for_vblank(rdev, 0); + radeon_wait_for_vblank(rdev, 1); + + r600_enable_spll_bypass(rdev, true); + r600_wait_for_spll_change(rdev); + r600_enable_spll_bypass(rdev, false); + r600_wait_for_spll_change(rdev); + + r600_enable_spll_bypass(rdev, true); + r600_wait_for_spll_change(rdev); + r600_enable_spll_bypass(rdev, false); + r600_wait_for_spll_change(rdev); + + r600_enable_sclk_control(rdev, true); + r600_enable_mclk_control(rdev, true); +} + +void r600_stop_dpm(struct radeon_device *rdev) +{ + r600_dynamicpm_enable(rdev, false); +} + +int r600_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + return 0; +} + +void r600_dpm_post_set_power_state(struct radeon_device *rdev) +{ + +} + +bool r600_is_uvd_state(u32 class, u32 class2) +{ + if (class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + return true; + if (class & ATOM_PPLIB_CLASSIFICATION_HD2STATE) + return true; + if (class & ATOM_PPLIB_CLASSIFICATION_HDSTATE) + return true; + if (class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) + return true; + if (class2 & ATOM_PPLIB_CLASSIFICATION2_MVC) + return true; + return false; +} + +int r600_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) +{ + int low_temp = 0 * 1000; + int high_temp = 255 * 1000; + + if (low_temp < min_temp) + low_temp = min_temp; + if (high_temp > max_temp) + high_temp = max_temp; + if (high_temp < low_temp) { + DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); + return -EINVAL; + } + + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTH(high_temp / 1000), ~DIG_THERM_INTH_MASK); + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTL(low_temp / 1000), ~DIG_THERM_INTL_MASK); + WREG32_P(CG_THERMAL_CTRL, DIG_THERM_DPM(high_temp / 1000), ~DIG_THERM_DPM_MASK); + + rdev->pm.dpm.thermal.min_temp = low_temp; + rdev->pm.dpm.thermal.max_temp = high_temp; + + return 0; +} + +bool r600_is_internal_thermal_sensor(enum radeon_int_thermal_type sensor) +{ + switch (sensor) { + case THERMAL_TYPE_RV6XX: + case THERMAL_TYPE_RV770: + case THERMAL_TYPE_EVERGREEN: + case THERMAL_TYPE_SUMO: + case THERMAL_TYPE_NI: + case THERMAL_TYPE_SI: + return true; + case THERMAL_TYPE_ADT7473_WITH_INTERNAL: + case THERMAL_TYPE_EMC2103_WITH_INTERNAL: + return false; /* need special handling */ + case THERMAL_TYPE_NONE: + case THERMAL_TYPE_EXTERNAL: + case THERMAL_TYPE_EXTERNAL_GPIO: + default: + return false; + } +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; + struct _ATOM_PPLIB_POWERPLAYTABLE4 pplib4; + struct _ATOM_PPLIB_POWERPLAYTABLE5 pplib5; +}; + +union fan_info { + struct _ATOM_PPLIB_FANTABLE fan; + struct _ATOM_PPLIB_FANTABLE2 fan2; +}; + +static int r600_parse_clk_voltage_dep_table(struct radeon_clock_voltage_dependency_table *radeon_table, + ATOM_PPLIB_Clock_Voltage_Dependency_Table *atom_table) +{ + u32 size = atom_table->ucNumEntries * + sizeof(struct radeon_clock_voltage_dependency_entry); + int i; + + radeon_table->entries = kzalloc(size, GFP_KERNEL); + if (!radeon_table->entries) + return -ENOMEM; + + for (i = 0; i < atom_table->ucNumEntries; i++) { + radeon_table->entries[i].clk = le16_to_cpu(atom_table->entries[i].usClockLow) | + (atom_table->entries[i].ucClockHigh << 16); + radeon_table->entries[i].v = le16_to_cpu(atom_table->entries[i].usVoltage); + } + radeon_table->count = atom_table->ucNumEntries; + + return 0; +} + +/* sizeof(ATOM_PPLIB_EXTENDEDHEADER) */ +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V2 12 +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V3 14 +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V4 16 +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V5 18 +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V6 20 +#define SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V7 22 + +int r600_parse_extended_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + union power_info *power_info; + union fan_info *fan_info; + ATOM_PPLIB_Clock_Voltage_Dependency_Table *dep_table; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + int ret, i; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + /* fan table */ + if (le16_to_cpu(power_info->pplib.usTableSize) >= + sizeof(struct _ATOM_PPLIB_POWERPLAYTABLE3)) { + if (power_info->pplib3.usFanTableOffset) { + fan_info = (union fan_info *)((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib3.usFanTableOffset)); + rdev->pm.dpm.fan.t_hyst = fan_info->fan.ucTHyst; + rdev->pm.dpm.fan.t_min = le16_to_cpu(fan_info->fan.usTMin); + rdev->pm.dpm.fan.t_med = le16_to_cpu(fan_info->fan.usTMed); + rdev->pm.dpm.fan.t_high = le16_to_cpu(fan_info->fan.usTHigh); + rdev->pm.dpm.fan.pwm_min = le16_to_cpu(fan_info->fan.usPWMMin); + rdev->pm.dpm.fan.pwm_med = le16_to_cpu(fan_info->fan.usPWMMed); + rdev->pm.dpm.fan.pwm_high = le16_to_cpu(fan_info->fan.usPWMHigh); + if (fan_info->fan.ucFanTableFormat >= 2) + rdev->pm.dpm.fan.t_max = le16_to_cpu(fan_info->fan2.usTMax); + else + rdev->pm.dpm.fan.t_max = 10900; + rdev->pm.dpm.fan.cycle_delay = 100000; + rdev->pm.dpm.fan.ucode_fan_control = true; + } + } + + /* clock dependancy tables, shedding tables */ + if (le16_to_cpu(power_info->pplib.usTableSize) >= + sizeof(struct _ATOM_PPLIB_POWERPLAYTABLE4)) { + if (power_info->pplib4.usVddcDependencyOnSCLKOffset) { + dep_table = (ATOM_PPLIB_Clock_Voltage_Dependency_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib4.usVddcDependencyOnSCLKOffset)); + ret = r600_parse_clk_voltage_dep_table(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + dep_table); + if (ret) + return ret; + } + if (power_info->pplib4.usVddciDependencyOnMCLKOffset) { + dep_table = (ATOM_PPLIB_Clock_Voltage_Dependency_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib4.usVddciDependencyOnMCLKOffset)); + ret = r600_parse_clk_voltage_dep_table(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + dep_table); + if (ret) { + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + return ret; + } + } + if (power_info->pplib4.usVddcDependencyOnMCLKOffset) { + dep_table = (ATOM_PPLIB_Clock_Voltage_Dependency_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib4.usVddcDependencyOnMCLKOffset)); + ret = r600_parse_clk_voltage_dep_table(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + dep_table); + if (ret) { + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries); + return ret; + } + } + if (power_info->pplib4.usMaxClockVoltageOnDCOffset) { + ATOM_PPLIB_Clock_Voltage_Limit_Table *clk_v = + (ATOM_PPLIB_Clock_Voltage_Limit_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib4.usMaxClockVoltageOnDCOffset)); + if (clk_v->ucNumEntries) { + rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.sclk = + le16_to_cpu(clk_v->entries[0].usSclkLow) | + (clk_v->entries[0].ucSclkHigh << 16); + rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.mclk = + le16_to_cpu(clk_v->entries[0].usMclkLow) | + (clk_v->entries[0].ucMclkHigh << 16); + rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc = + le16_to_cpu(clk_v->entries[0].usVddc); + rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddci = + le16_to_cpu(clk_v->entries[0].usVddci); + } + } + if (power_info->pplib4.usVddcPhaseShedLimitsTableOffset) { + ATOM_PPLIB_PhaseSheddingLimits_Table *psl = + (ATOM_PPLIB_PhaseSheddingLimits_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib4.usVddcPhaseShedLimitsTableOffset)); + + rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries = + kzalloc(psl->ucNumEntries * + sizeof(struct radeon_phase_shedding_limits_entry), + GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries) { + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries); + return -ENOMEM; + } + + for (i = 0; i < psl->ucNumEntries; i++) { + rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries[i].sclk = + le16_to_cpu(psl->entries[i].usSclkLow) | + (psl->entries[i].ucSclkHigh << 16); + rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries[i].mclk = + le16_to_cpu(psl->entries[i].usMclkLow) | + (psl->entries[i].ucMclkHigh << 16); + rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries[i].voltage = + le16_to_cpu(psl->entries[i].usVoltage); + } + rdev->pm.dpm.dyn_state.phase_shedding_limits_table.count = + psl->ucNumEntries; + } + } + + /* cac data */ + if (le16_to_cpu(power_info->pplib.usTableSize) >= + sizeof(struct _ATOM_PPLIB_POWERPLAYTABLE5)) { + rdev->pm.dpm.tdp_limit = le32_to_cpu(power_info->pplib5.ulTDPLimit); + rdev->pm.dpm.near_tdp_limit = le32_to_cpu(power_info->pplib5.ulNearTDPLimit); + rdev->pm.dpm.near_tdp_limit_adjusted = rdev->pm.dpm.near_tdp_limit; + rdev->pm.dpm.tdp_od_limit = le16_to_cpu(power_info->pplib5.usTDPODLimit); + if (rdev->pm.dpm.tdp_od_limit) + rdev->pm.dpm.power_control = true; + else + rdev->pm.dpm.power_control = false; + rdev->pm.dpm.tdp_adjustment = 0; + rdev->pm.dpm.sq_ramping_threshold = le32_to_cpu(power_info->pplib5.ulSQRampingThreshold); + rdev->pm.dpm.cac_leakage = le32_to_cpu(power_info->pplib5.ulCACLeakage); + rdev->pm.dpm.load_line_slope = le16_to_cpu(power_info->pplib5.usLoadLineSlope); + if (power_info->pplib5.usCACLeakageTableOffset) { + ATOM_PPLIB_CAC_Leakage_Table *cac_table = + (ATOM_PPLIB_CAC_Leakage_Table *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib5.usCACLeakageTableOffset)); + u32 size = cac_table->ucNumEntries * sizeof(struct radeon_cac_leakage_table); + rdev->pm.dpm.dyn_state.cac_leakage_table.entries = kzalloc(size, GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.cac_leakage_table.entries) { + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries); + return -ENOMEM; + } + for (i = 0; i < cac_table->ucNumEntries; i++) { + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[i].vddc = + le16_to_cpu(cac_table->entries[i].usVddc); + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[i].leakage = + le32_to_cpu(cac_table->entries[i].ulLeakageValue); + } + rdev->pm.dpm.dyn_state.cac_leakage_table.count = cac_table->ucNumEntries; + } + } + + /* ppm table */ + if (le16_to_cpu(power_info->pplib.usTableSize) >= + sizeof(struct _ATOM_PPLIB_POWERPLAYTABLE3)) { + ATOM_PPLIB_EXTENDEDHEADER *ext_hdr = (ATOM_PPLIB_EXTENDEDHEADER *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib3.usExtendendedHeaderOffset)); + if ((le16_to_cpu(ext_hdr->usSize) >= SIZE_OF_ATOM_PPLIB_EXTENDEDHEADER_V5) && + ext_hdr->usPPMTableOffset) { + ATOM_PPLIB_PPM_Table *ppm = (ATOM_PPLIB_PPM_Table *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(ext_hdr->usPPMTableOffset)); + rdev->pm.dpm.dyn_state.ppm_table = + kzalloc(sizeof(struct radeon_ppm_table), GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.ppm_table) { + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries); + kfree(rdev->pm.dpm.dyn_state.cac_leakage_table.entries); + return -ENOMEM; + } + rdev->pm.dpm.dyn_state.ppm_table->ppm_design = ppm->ucPpmDesign; + rdev->pm.dpm.dyn_state.ppm_table->cpu_core_number = + le16_to_cpu(ppm->usCpuCoreNumber); + rdev->pm.dpm.dyn_state.ppm_table->platform_tdp = + le32_to_cpu(ppm->ulPlatformTDP); + rdev->pm.dpm.dyn_state.ppm_table->small_ac_platform_tdp = + le32_to_cpu(ppm->ulSmallACPlatformTDP); + rdev->pm.dpm.dyn_state.ppm_table->platform_tdc = + le32_to_cpu(ppm->ulPlatformTDC); + rdev->pm.dpm.dyn_state.ppm_table->small_ac_platform_tdc = + le32_to_cpu(ppm->ulSmallACPlatformTDC); + rdev->pm.dpm.dyn_state.ppm_table->apu_tdp = + le32_to_cpu(ppm->ulApuTDP); + rdev->pm.dpm.dyn_state.ppm_table->dgpu_tdp = + le32_to_cpu(ppm->ulDGpuTDP); + rdev->pm.dpm.dyn_state.ppm_table->dgpu_ulv_power = + le32_to_cpu(ppm->ulDGpuUlvPower); + rdev->pm.dpm.dyn_state.ppm_table->tj_max = + le32_to_cpu(ppm->ulTjmax); + } + } + + return 0; +} + +void r600_free_extended_power_table(struct radeon_device *rdev) +{ + if (rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries) + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries); + if (rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries) + kfree(rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk.entries); + if (rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries) + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk.entries); + if (rdev->pm.dpm.dyn_state.cac_leakage_table.entries) + kfree(rdev->pm.dpm.dyn_state.cac_leakage_table.entries); + if (rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries) + kfree(rdev->pm.dpm.dyn_state.phase_shedding_limits_table.entries); + if (rdev->pm.dpm.dyn_state.ppm_table) + kfree(rdev->pm.dpm.dyn_state.ppm_table); +} + +enum radeon_pcie_gen r600_get_pcie_gen_support(struct radeon_device *rdev, + u32 sys_mask, + enum radeon_pcie_gen asic_gen, + enum radeon_pcie_gen default_gen) +{ + switch (asic_gen) { + case RADEON_PCIE_GEN1: + return RADEON_PCIE_GEN1; + case RADEON_PCIE_GEN2: + return RADEON_PCIE_GEN2; + case RADEON_PCIE_GEN3: + return RADEON_PCIE_GEN3; + default: + if ((sys_mask & DRM_PCIE_SPEED_80) && (default_gen == RADEON_PCIE_GEN3)) + return RADEON_PCIE_GEN3; + else if ((sys_mask & DRM_PCIE_SPEED_50) && (default_gen == RADEON_PCIE_GEN2)) + return RADEON_PCIE_GEN2; + else + return RADEON_PCIE_GEN1; + } + return RADEON_PCIE_GEN1; +} diff --git a/sys/dev/drm/radeon/r600_dpm.h b/sys/dev/drm/radeon/r600_dpm.h new file mode 100644 index 0000000000..7c822d9ae5 --- /dev/null +++ b/sys/dev/drm/radeon/r600_dpm.h @@ -0,0 +1,227 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __R600_DPM_H__ +#define __R600_DPM_H__ + +#define R600_ASI_DFLT 10000 +#define R600_BSP_DFLT 0x41EB +#define R600_BSU_DFLT 0x2 +#define R600_AH_DFLT 5 +#define R600_RLP_DFLT 25 +#define R600_RMP_DFLT 65 +#define R600_LHP_DFLT 40 +#define R600_LMP_DFLT 15 +#define R600_TD_DFLT 0 +#define R600_UTC_DFLT_00 0x24 +#define R600_UTC_DFLT_01 0x22 +#define R600_UTC_DFLT_02 0x22 +#define R600_UTC_DFLT_03 0x22 +#define R600_UTC_DFLT_04 0x22 +#define R600_UTC_DFLT_05 0x22 +#define R600_UTC_DFLT_06 0x22 +#define R600_UTC_DFLT_07 0x22 +#define R600_UTC_DFLT_08 0x22 +#define R600_UTC_DFLT_09 0x22 +#define R600_UTC_DFLT_10 0x22 +#define R600_UTC_DFLT_11 0x22 +#define R600_UTC_DFLT_12 0x22 +#define R600_UTC_DFLT_13 0x22 +#define R600_UTC_DFLT_14 0x22 +#define R600_DTC_DFLT_00 0x24 +#define R600_DTC_DFLT_01 0x22 +#define R600_DTC_DFLT_02 0x22 +#define R600_DTC_DFLT_03 0x22 +#define R600_DTC_DFLT_04 0x22 +#define R600_DTC_DFLT_05 0x22 +#define R600_DTC_DFLT_06 0x22 +#define R600_DTC_DFLT_07 0x22 +#define R600_DTC_DFLT_08 0x22 +#define R600_DTC_DFLT_09 0x22 +#define R600_DTC_DFLT_10 0x22 +#define R600_DTC_DFLT_11 0x22 +#define R600_DTC_DFLT_12 0x22 +#define R600_DTC_DFLT_13 0x22 +#define R600_DTC_DFLT_14 0x22 +#define R600_VRC_DFLT 0x0000C003 +#define R600_VOLTAGERESPONSETIME_DFLT 1000 +#define R600_BACKBIASRESPONSETIME_DFLT 1000 +#define R600_VRU_DFLT 0x3 +#define R600_SPLLSTEPTIME_DFLT 0x1000 +#define R600_SPLLSTEPUNIT_DFLT 0x3 +#define R600_TPU_DFLT 0 +#define R600_TPC_DFLT 0x200 +#define R600_SSTU_DFLT 0 +#define R600_SST_DFLT 0x00C8 +#define R600_GICST_DFLT 0x200 +#define R600_FCT_DFLT 0x0400 +#define R600_FCTU_DFLT 0 +#define R600_CTXCGTT3DRPHC_DFLT 0x20 +#define R600_CTXCGTT3DRSDC_DFLT 0x40 +#define R600_VDDC3DOORPHC_DFLT 0x100 +#define R600_VDDC3DOORSDC_DFLT 0x7 +#define R600_VDDC3DOORSU_DFLT 0 +#define R600_MPLLLOCKTIME_DFLT 100 +#define R600_MPLLRESETTIME_DFLT 150 +#define R600_VCOSTEPPCT_DFLT 20 +#define R600_ENDINGVCOSTEPPCT_DFLT 5 +#define R600_REFERENCEDIVIDER_DFLT 4 + +#define R600_PM_NUMBER_OF_TC 15 +#define R600_PM_NUMBER_OF_SCLKS 20 +#define R600_PM_NUMBER_OF_MCLKS 4 +#define R600_PM_NUMBER_OF_VOLTAGE_LEVELS 4 +#define R600_PM_NUMBER_OF_ACTIVITY_LEVELS 3 + +/* XXX are these ok? */ +#define R600_TEMP_RANGE_MIN (90 * 1000) +#define R600_TEMP_RANGE_MAX (120 * 1000) + +enum r600_power_level { + R600_POWER_LEVEL_LOW = 0, + R600_POWER_LEVEL_MEDIUM = 1, + R600_POWER_LEVEL_HIGH = 2, + R600_POWER_LEVEL_CTXSW = 3, +}; + +enum r600_td { + R600_TD_AUTO, + R600_TD_UP, + R600_TD_DOWN, +}; + +enum r600_display_watermark { + R600_DISPLAY_WATERMARK_LOW = 0, + R600_DISPLAY_WATERMARK_HIGH = 1, +}; + +enum r600_display_gap +{ + R600_PM_DISPLAY_GAP_VBLANK_OR_WM = 0, + R600_PM_DISPLAY_GAP_VBLANK = 1, + R600_PM_DISPLAY_GAP_WATERMARK = 2, + R600_PM_DISPLAY_GAP_IGNORE = 3, +}; + +extern const u32 r600_utc[R600_PM_NUMBER_OF_TC]; +extern const u32 r600_dtc[R600_PM_NUMBER_OF_TC]; + +void r600_dpm_print_class_info(u32 class, u32 class2); +void r600_dpm_print_cap_info(u32 caps); +void r600_dpm_print_ps_status(struct radeon_device *rdev, + struct radeon_ps *rps); +u32 r600_dpm_get_vblank_time(struct radeon_device *rdev); +bool r600_is_uvd_state(u32 class, u32 class2); +void r600_calculate_u_and_p(u32 i, u32 r_c, u32 p_b, + u32 *p, u32 *u); +int r600_calculate_at(u32 t, u32 h, u32 fh, u32 fl, u32 *tl, u32 *th); +void r600_gfx_clockgating_enable(struct radeon_device *rdev, bool enable); +void r600_dynamicpm_enable(struct radeon_device *rdev, bool enable); +void r600_enable_thermal_protection(struct radeon_device *rdev, bool enable); +void r600_enable_acpi_pm(struct radeon_device *rdev); +void r600_enable_dynamic_pcie_gen2(struct radeon_device *rdev, bool enable); +bool r600_dynamicpm_enabled(struct radeon_device *rdev); +void r600_enable_sclk_control(struct radeon_device *rdev, bool enable); +void r600_enable_mclk_control(struct radeon_device *rdev, bool enable); +void r600_enable_spll_bypass(struct radeon_device *rdev, bool enable); +void r600_wait_for_spll_change(struct radeon_device *rdev); +void r600_set_bsp(struct radeon_device *rdev, u32 u, u32 p); +void r600_set_at(struct radeon_device *rdev, + u32 l_to_m, u32 m_to_h, + u32 h_to_m, u32 m_to_l); +void r600_set_tc(struct radeon_device *rdev, u32 index, u32 u_t, u32 d_t); +void r600_select_td(struct radeon_device *rdev, enum r600_td td); +void r600_set_vrc(struct radeon_device *rdev, u32 vrv); +void r600_set_tpu(struct radeon_device *rdev, u32 u); +void r600_set_tpc(struct radeon_device *rdev, u32 c); +void r600_set_sstu(struct radeon_device *rdev, u32 u); +void r600_set_sst(struct radeon_device *rdev, u32 t); +void r600_set_git(struct radeon_device *rdev, u32 t); +void r600_set_fctu(struct radeon_device *rdev, u32 u); +void r600_set_fct(struct radeon_device *rdev, u32 t); +void r600_set_ctxcgtt3d_rphc(struct radeon_device *rdev, u32 p); +void r600_set_ctxcgtt3d_rsdc(struct radeon_device *rdev, u32 s); +void r600_set_vddc3d_oorsu(struct radeon_device *rdev, u32 u); +void r600_set_vddc3d_oorphc(struct radeon_device *rdev, u32 p); +void r600_set_vddc3d_oorsdc(struct radeon_device *rdev, u32 s); +void r600_set_mpll_lock_time(struct radeon_device *rdev, u32 lock_time); +void r600_set_mpll_reset_time(struct radeon_device *rdev, u32 reset_time); +void r600_engine_clock_entry_enable(struct radeon_device *rdev, + u32 index, bool enable); +void r600_engine_clock_entry_enable_pulse_skipping(struct radeon_device *rdev, + u32 index, bool enable); +void r600_engine_clock_entry_enable_post_divider(struct radeon_device *rdev, + u32 index, bool enable); +void r600_engine_clock_entry_set_post_divider(struct radeon_device *rdev, + u32 index, u32 divider); +void r600_engine_clock_entry_set_reference_divider(struct radeon_device *rdev, + u32 index, u32 divider); +void r600_engine_clock_entry_set_feedback_divider(struct radeon_device *rdev, + u32 index, u32 divider); +void r600_engine_clock_entry_set_step_time(struct radeon_device *rdev, + u32 index, u32 step_time); +void r600_vid_rt_set_ssu(struct radeon_device *rdev, u32 u); +void r600_vid_rt_set_vru(struct radeon_device *rdev, u32 u); +void r600_vid_rt_set_vrt(struct radeon_device *rdev, u32 rt); +void r600_voltage_control_enable_pins(struct radeon_device *rdev, + u64 mask); +void r600_voltage_control_program_voltages(struct radeon_device *rdev, + enum r600_power_level index, u64 pins); +void r600_voltage_control_deactivate_static_control(struct radeon_device *rdev, + u64 mask); +void r600_power_level_enable(struct radeon_device *rdev, + enum r600_power_level index, bool enable); +void r600_power_level_set_voltage_index(struct radeon_device *rdev, + enum r600_power_level index, u32 voltage_index); +void r600_power_level_set_mem_clock_index(struct radeon_device *rdev, + enum r600_power_level index, u32 mem_clock_index); +void r600_power_level_set_eng_clock_index(struct radeon_device *rdev, + enum r600_power_level index, u32 eng_clock_index); +void r600_power_level_set_watermark_id(struct radeon_device *rdev, + enum r600_power_level index, + enum r600_display_watermark watermark_id); +void r600_power_level_set_pcie_gen2(struct radeon_device *rdev, + enum r600_power_level index, bool compatible); +enum r600_power_level r600_power_level_get_current_index(struct radeon_device *rdev); +enum r600_power_level r600_power_level_get_target_index(struct radeon_device *rdev); +void r600_power_level_set_enter_index(struct radeon_device *rdev, + enum r600_power_level index); +void r600_wait_for_power_level_unequal(struct radeon_device *rdev, + enum r600_power_level index); +void r600_wait_for_power_level(struct radeon_device *rdev, + enum r600_power_level index); +void r600_start_dpm(struct radeon_device *rdev); +void r600_stop_dpm(struct radeon_device *rdev); + +int r600_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp); +bool r600_is_internal_thermal_sensor(enum radeon_int_thermal_type sensor); + +int r600_parse_extended_power_table(struct radeon_device *rdev); +void r600_free_extended_power_table(struct radeon_device *rdev); + +enum radeon_pcie_gen r600_get_pcie_gen_support(struct radeon_device *rdev, + u32 sys_mask, + enum radeon_pcie_gen asic_gen, + enum radeon_pcie_gen default_gen); + +#endif diff --git a/sys/dev/drm/radeon/r600_hdmi.c b/sys/dev/drm/radeon/r600_hdmi.c index ccecbce9ce..553cd8344b 100644 --- a/sys/dev/drm/radeon/r600_hdmi.c +++ b/sys/dev/drm/radeon/r600_hdmi.c @@ -22,10 +22,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Christian König - * - * $FreeBSD: head/sys/dev/drm2/radeon/r600_hdmi.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include @@ -137,13 +134,7 @@ static void r600_hdmi_update_avi_infoframe(struct drm_encoder *encoder, uint32_t offset = dig->afmt->offset; uint8_t *frame = (uint8_t*)buffer + 3; - /* Our header values (type, version, length) should be alright, Intel - * is using the same. Checksum function also seems to be OK, it works - * fine for audio infoframe. However calculated value is always lower - * by 2 in comparison to fglrx. It breaks displaying anything in case - * of TVs that strictly check the checksum. Hack it manually here to - * workaround this issue. */ - frame[0x0] += 2; + uint8_t *header = buffer; WREG32(HDMI0_AVI_INFO0 + offset, frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24)); @@ -152,7 +143,7 @@ static void r600_hdmi_update_avi_infoframe(struct drm_encoder *encoder, WREG32(HDMI0_AVI_INFO2 + offset, frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24)); WREG32(HDMI0_AVI_INFO3 + offset, - frame[0xC] | (frame[0xD] << 8)); + frame[0xC] | (frame[0xD] << 8) | (header[1] << 24)); } /* @@ -236,10 +227,29 @@ void r600_audio_set_dto(struct drm_encoder *encoder, u32 clock) struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv; u32 base_rate = 24000; + u32 max_ratio = clock / base_rate; + u32 dto_phase; + u32 dto_modulo = clock; + u32 wallclock_ratio; + u32 dto_cntl; if (!dig || !dig->afmt) return; + if (max_ratio >= 8) { + dto_phase = 192 * 1000; + wallclock_ratio = 3; + } else if (max_ratio >= 4) { + dto_phase = 96 * 1000; + wallclock_ratio = 2; + } else if (max_ratio >= 2) { + dto_phase = 48 * 1000; + wallclock_ratio = 1; + } else { + dto_phase = 24 * 1000; + wallclock_ratio = 0; + } + /* there are two DTOs selected by DCCG_AUDIO_DTO_SELECT. * doesn't matter which one you use. Just use the first one. */ @@ -252,9 +262,21 @@ void r600_audio_set_dto(struct drm_encoder *encoder, u32 clock) /* according to the reg specs, this should DCE3.2 only, but in * practice it seems to cover DCE3.0 as well. */ - WREG32(DCCG_AUDIO_DTO0_PHASE, base_rate * 100); - WREG32(DCCG_AUDIO_DTO0_MODULE, clock * 100); - WREG32(DCCG_AUDIO_DTO_SELECT, 0); /* select DTO0 */ + if (dig->dig_encoder == 0) { + dto_cntl = RREG32(DCCG_AUDIO_DTO0_CNTL) & ~DCCG_AUDIO_DTO_WALLCLOCK_RATIO_MASK; + dto_cntl |= DCCG_AUDIO_DTO_WALLCLOCK_RATIO(wallclock_ratio); + WREG32(DCCG_AUDIO_DTO0_CNTL, dto_cntl); + WREG32(DCCG_AUDIO_DTO0_PHASE, dto_phase); + WREG32(DCCG_AUDIO_DTO0_MODULE, dto_modulo); + WREG32(DCCG_AUDIO_DTO_SELECT, 0); /* select DTO0 */ + } else { + dto_cntl = RREG32(DCCG_AUDIO_DTO1_CNTL) & ~DCCG_AUDIO_DTO_WALLCLOCK_RATIO_MASK; + dto_cntl |= DCCG_AUDIO_DTO_WALLCLOCK_RATIO(wallclock_ratio); + WREG32(DCCG_AUDIO_DTO1_CNTL, dto_cntl); + WREG32(DCCG_AUDIO_DTO1_PHASE, dto_phase); + WREG32(DCCG_AUDIO_DTO1_MODULE, dto_modulo); + WREG32(DCCG_AUDIO_DTO_SELECT, 1); /* select DTO1 */ + } } else { /* according to the reg specs, this should be DCE2.0 and DCE3.0 */ WREG32(AUDIO_DTO, AUDIO_DTO_PHASE(base_rate / 10) | @@ -276,6 +298,9 @@ void r600_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mod uint32_t offset; ssize_t err; + if (!dig || !dig->afmt) + return; + /* Silent, r600_hdmi_enable will raise WARN for us */ if (!dig->afmt->enabled) return; @@ -458,6 +483,9 @@ void r600_hdmi_enable(struct drm_encoder *encoder, bool enable) struct radeon_encoder_atom_dig *dig = radeon_encoder->enc_priv; u32 hdmi = HDMI0_ERROR_ACK; + if (!dig || !dig->afmt) + return; + /* Silent, r600_hdmi_enable will raise WARN for us */ if (enable && dig->afmt->enabled) return; diff --git a/sys/dev/drm/radeon/r600_reg.h b/sys/dev/drm/radeon/r600_reg.h index eca6abbda3..3ef202629e 100644 --- a/sys/dev/drm/radeon/r600_reg.h +++ b/sys/dev/drm/radeon/r600_reg.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r600_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __R600_REG_H__ #define __R600_REG_H__ @@ -33,6 +31,12 @@ #define R600_PCIE_PORT_INDEX 0x0038 #define R600_PCIE_PORT_DATA 0x003c +#define R600_RCU_INDEX 0x0100 +#define R600_RCU_DATA 0x0104 + +#define R600_UVD_CTX_INDEX 0xf4a0 +#define R600_UVD_CTX_DATA 0xf4a4 + #define R600_MC_VM_FB_LOCATION 0x2180 #define R600_MC_FB_BASE_MASK 0x0000FFFF #define R600_MC_FB_BASE_SHIFT 0 diff --git a/sys/dev/drm/radeon/r600d.h b/sys/dev/drm/radeon/r600d.h index 8b0cf6b1f2..7c780839a7 100644 --- a/sys/dev/drm/radeon/r600d.h +++ b/sys/dev/drm/radeon/r600d.h @@ -23,8 +23,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/r600d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef R600D_H #define R600D_H @@ -304,10 +302,25 @@ #define GRBM_SOFT_RESET 0x8020 #define SOFT_RESET_CP (1<<0) +#define CG_THERMAL_CTRL 0x7F0 +#define DIG_THERM_DPM(x) ((x) << 12) +#define DIG_THERM_DPM_MASK 0x000FF000 +#define DIG_THERM_DPM_SHIFT 12 #define CG_THERMAL_STATUS 0x7F4 #define ASIC_T(x) ((x) << 0) #define ASIC_T_MASK 0x1FF #define ASIC_T_SHIFT 0 +#define CG_THERMAL_INT 0x7F8 +#define DIG_THERM_INTH(x) ((x) << 8) +#define DIG_THERM_INTH_MASK 0x0000FF00 +#define DIG_THERM_INTH_SHIFT 8 +#define DIG_THERM_INTL(x) ((x) << 16) +#define DIG_THERM_INTL_MASK 0x00FF0000 +#define DIG_THERM_INTL_SHIFT 16 +#define THERM_INT_MASK_HIGH (1 << 24) +#define THERM_INT_MASK_LOW (1 << 25) + +#define RV770_CG_THERMAL_INT 0x734 #define HDP_HOST_PATH_CNTL 0x2C00 #define HDP_NONSURFACE_BASE 0x2C04 @@ -589,6 +602,7 @@ #define L2_BUSY (1 << 0) #define WAIT_UNTIL 0x8040 +#define WAIT_CP_DMA_IDLE_bit (1 << 8) #define WAIT_2D_IDLE_bit (1 << 14) #define WAIT_3D_IDLE_bit (1 << 15) #define WAIT_2D_IDLECLEAN_bit (1 << 16) @@ -686,10 +700,6 @@ #define RLC_UCODE_ADDR 0x3f2c #define RLC_UCODE_DATA 0x3f30 -/* new for TN */ -#define TN_RLC_SAVE_AND_RESTORE_BASE 0x3f10 -#define TN_RLC_CLEAR_STATE_RESTORE_BASE 0x3f20 - #define SRBM_SOFT_RESET 0xe60 # define SOFT_RESET_DMA (1 << 12) # define SOFT_RESET_RLC (1 << 13) @@ -923,6 +933,9 @@ #define DCCG_AUDIO_DTO0_LOAD 0x051c # define DTO_LOAD (1 << 31) #define DCCG_AUDIO_DTO0_CNTL 0x0520 +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO(x) (((x) & 7) << 0) +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO_MASK 7 +# define DCCG_AUDIO_DTO_WALLCLOCK_RATIO_SHIFT 0 #define DCCG_AUDIO_DTO1_PHASE 0x0524 #define DCCG_AUDIO_DTO1_MODULE 0x0528 @@ -1150,6 +1163,219 @@ # define AFMT_AZ_FORMAT_WTRIG_ACK (1 << 29) # define AFMT_AZ_AUDIO_ENABLE_CHG_ACK (1 << 30) +/* Power management */ +#define CG_SPLL_FUNC_CNTL 0x600 +# define SPLL_RESET (1 << 0) +# define SPLL_SLEEP (1 << 1) +# define SPLL_REF_DIV(x) ((x) << 2) +# define SPLL_REF_DIV_MASK (7 << 2) +# define SPLL_FB_DIV(x) ((x) << 5) +# define SPLL_FB_DIV_MASK (0xff << 5) +# define SPLL_PULSEEN (1 << 13) +# define SPLL_PULSENUM(x) ((x) << 14) +# define SPLL_PULSENUM_MASK (3 << 14) +# define SPLL_SW_HILEN(x) ((x) << 16) +# define SPLL_SW_HILEN_MASK (0xf << 16) +# define SPLL_SW_LOLEN(x) ((x) << 20) +# define SPLL_SW_LOLEN_MASK (0xf << 20) +# define SPLL_DIVEN (1 << 24) +# define SPLL_BYPASS_EN (1 << 25) +# define SPLL_CHG_STATUS (1 << 29) +# define SPLL_CTLREQ (1 << 30) +# define SPLL_CTLACK (1 << 31) + +#define GENERAL_PWRMGT 0x618 +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define MOBILE_SU (1 << 2) +# define THERMAL_PROTECTION_DIS (1 << 3) +# define THERMAL_PROTECTION_TYPE (1 << 4) +# define ENABLE_GEN2PCIE (1 << 5) +# define SW_GPIO_INDEX(x) ((x) << 6) +# define SW_GPIO_INDEX_MASK (3 << 6) +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +#define CG_TPC 0x61c +# define TPCC(x) ((x) << 0) +# define TPCC_MASK (0x7fffff << 0) +# define TPU(x) ((x) << 23) +# define TPU_MASK (0x1f << 23) +#define SCLK_PWRMGT_CNTL 0x620 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_TURNOFF (1 << 1) +# define SPLL_TURNOFF (1 << 2) +# define SU_SCLK_USE_BCLK (1 << 3) +# define DYNAMIC_GFX_ISLAND_PWR_DOWN (1 << 4) +# define DYNAMIC_GFX_ISLAND_PWR_LP (1 << 5) +# define CLK_TURN_ON_STAGGER (1 << 6) +# define CLK_TURN_OFF_STAGGER (1 << 7) +# define FIR_FORCE_TREND_SEL (1 << 8) +# define FIR_TREND_MODE (1 << 9) +# define DYN_GFX_CLK_OFF_EN (1 << 10) +# define VDDC3D_TURNOFF_D1 (1 << 11) +# define VDDC3D_TURNOFF_D2 (1 << 12) +# define VDDC3D_TURNOFF_D3 (1 << 13) +# define SPLL_TURNOFF_D2 (1 << 14) +# define SCLK_LOW_D1 (1 << 15) +# define DYN_GFX_CLK_OFF_MC_EN (1 << 16) +#define MCLK_PWRMGT_CNTL 0x624 +# define MPLL_PWRMGT_OFF (1 << 0) +# define YCLK_TURNOFF (1 << 1) +# define MPLL_TURNOFF (1 << 2) +# define SU_MCLK_USE_BCLK (1 << 3) +# define DLL_READY (1 << 4) +# define MC_BUSY (1 << 5) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA_SLEEP (1 << 8) +# define MRDCKB_SLEEP (1 << 9) +# define MRDCKC_SLEEP (1 << 10) +# define MRDCKD_SLEEP (1 << 11) +# define MRDCKE_SLEEP (1 << 12) +# define MRDCKF_SLEEP (1 << 13) +# define MRDCKG_SLEEP (1 << 14) +# define MRDCKH_SLEEP (1 << 15) +# define MRDCKA_RESET (1 << 16) +# define MRDCKB_RESET (1 << 17) +# define MRDCKC_RESET (1 << 18) +# define MRDCKD_RESET (1 << 19) +# define MRDCKE_RESET (1 << 20) +# define MRDCKF_RESET (1 << 21) +# define MRDCKG_RESET (1 << 22) +# define MRDCKH_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define USE_DISPLAY_GAP_CTXSW (1 << 27) +# define MPLL_TURNOFF_D2 (1 << 28) +# define USE_DISPLAY_URGENT_CTXSW (1 << 29) + +#define MPLL_TIME 0x634 +# define MPLL_LOCK_TIME(x) ((x) << 0) +# define MPLL_LOCK_TIME_MASK (0xffff << 0) +# define MPLL_RESET_TIME(x) ((x) << 16) +# define MPLL_RESET_TIME_MASK (0xffff << 16) + +#define SCLK_FREQ_SETTING_STEP_0_PART1 0x648 +# define STEP_0_SPLL_POST_DIV(x) ((x) << 0) +# define STEP_0_SPLL_POST_DIV_MASK (0xff << 0) +# define STEP_0_SPLL_FB_DIV(x) ((x) << 8) +# define STEP_0_SPLL_FB_DIV_MASK (0xff << 8) +# define STEP_0_SPLL_REF_DIV(x) ((x) << 16) +# define STEP_0_SPLL_REF_DIV_MASK (7 << 16) +# define STEP_0_SPLL_STEP_TIME(x) ((x) << 19) +# define STEP_0_SPLL_STEP_TIME_MASK (0x1fff << 19) +#define SCLK_FREQ_SETTING_STEP_0_PART2 0x64c +# define STEP_0_PULSE_HIGH_CNT(x) ((x) << 0) +# define STEP_0_PULSE_HIGH_CNT_MASK (0x1ff << 0) +# define STEP_0_POST_DIV_EN (1 << 9) +# define STEP_0_SPLL_STEP_ENABLE (1 << 30) +# define STEP_0_SPLL_ENTRY_VALID (1 << 31) + +#define VID_RT 0x6f8 +# define VID_CRT(x) ((x) << 0) +# define VID_CRT_MASK (0x1fff << 0) +# define VID_CRTU(x) ((x) << 13) +# define VID_CRTU_MASK (7 << 13) +# define SSTU(x) ((x) << 16) +# define SSTU_MASK (7 << 16) +#define CTXSW_PROFILE_INDEX 0x6fc +# define CTXSW_FREQ_VIDS_CFG_INDEX(x) ((x) << 0) +# define CTXSW_FREQ_VIDS_CFG_INDEX_MASK (3 << 0) +# define CTXSW_FREQ_VIDS_CFG_INDEX_SHIFT 0 +# define CTXSW_FREQ_MCLK_CFG_INDEX(x) ((x) << 2) +# define CTXSW_FREQ_MCLK_CFG_INDEX_MASK (3 << 2) +# define CTXSW_FREQ_MCLK_CFG_INDEX_SHIFT 2 +# define CTXSW_FREQ_SCLK_CFG_INDEX(x) ((x) << 4) +# define CTXSW_FREQ_SCLK_CFG_INDEX_MASK (0x1f << 4) +# define CTXSW_FREQ_SCLK_CFG_INDEX_SHIFT 4 +# define CTXSW_FREQ_STATE_SPLL_RESET_EN (1 << 9) +# define CTXSW_FREQ_STATE_ENABLE (1 << 10) +# define CTXSW_FREQ_DISPLAY_WATERMARK (1 << 11) +# define CTXSW_FREQ_GEN2PCIE_VOLT (1 << 12) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x70c +# define TARGET_PROFILE_INDEX_MASK (3 << 0) +# define TARGET_PROFILE_INDEX_SHIFT 0 +# define CURRENT_PROFILE_INDEX_MASK (3 << 2) +# define CURRENT_PROFILE_INDEX_SHIFT 2 +# define DYN_PWR_ENTER_INDEX(x) ((x) << 4) +# define DYN_PWR_ENTER_INDEX_MASK (3 << 4) +# define DYN_PWR_ENTER_INDEX_SHIFT 4 +# define CURR_MCLK_INDEX_MASK (3 << 6) +# define CURR_MCLK_INDEX_SHIFT 6 +# define CURR_SCLK_INDEX_MASK (0x1f << 8) +# define CURR_SCLK_INDEX_SHIFT 8 +# define CURR_VID_INDEX_MASK (3 << 13) +# define CURR_VID_INDEX_SHIFT 13 + +#define LOWER_GPIO_ENABLE 0x710 +#define UPPER_GPIO_ENABLE 0x714 +#define CTXSW_VID_LOWER_GPIO_CNTL 0x718 + +#define VID_UPPER_GPIO_CNTL 0x740 +#define CG_CTX_CGTT3D_R 0x744 +# define PHC(x) ((x) << 0) +# define PHC_MASK (0x1ff << 0) +# define SDC(x) ((x) << 9) +# define SDC_MASK (0x3fff << 9) +#define CG_VDDC3D_OOR 0x748 +# define SU(x) ((x) << 23) +# define SU_MASK (0xf << 23) +#define CG_FTV 0x74c +#define CG_FFCT_0 0x750 +# define UTC_0(x) ((x) << 0) +# define UTC_0_MASK (0x3ff << 0) +# define DTC_0(x) ((x) << 10) +# define DTC_0_MASK (0x3ff << 10) + +#define CG_BSP 0x78c +# define BSP(x) ((x) << 0) +# define BSP_MASK (0xffff << 0) +# define BSU(x) ((x) << 16) +# define BSU_MASK (0xf << 16) +#define CG_RT 0x790 +# define FLS(x) ((x) << 0) +# define FLS_MASK (0xffff << 0) +# define FMS(x) ((x) << 16) +# define FMS_MASK (0xffff << 16) +#define CG_LT 0x794 +# define FHS(x) ((x) << 0) +# define FHS_MASK (0xffff << 0) +#define CG_GIT 0x798 +# define CG_GICST(x) ((x) << 0) +# define CG_GICST_MASK (0xffff << 0) +# define CG_GIPOT(x) ((x) << 16) +# define CG_GIPOT_MASK (0xffff << 16) + +#define CG_SSP 0x7a8 +# define CG_SST(x) ((x) << 0) +# define CG_SST_MASK (0xffff << 0) +# define CG_SSTU(x) ((x) << 16) +# define CG_SSTU_MASK (0xf << 16) + +#define CG_RLC_REQ_AND_RSP 0x7c4 +# define RLC_CG_REQ_TYPE_MASK 0xf +# define RLC_CG_REQ_TYPE_SHIFT 0 +# define CG_RLC_RSP_TYPE_MASK 0xf0 +# define CG_RLC_RSP_TYPE_SHIFT 4 + +#define CG_FC_T 0x7cc +# define FC_T(x) ((x) << 0) +# define FC_T_MASK (0xffff << 0) +# define FC_TU(x) ((x) << 16) +# define FC_TU_MASK (0x1f << 16) + +#define GPIOPAD_MASK 0x1798 +#define GPIOPAD_A 0x179c +#define GPIOPAD_EN 0x17a0 + +#define GRBM_PWR_CNTL 0x800c +# define REQ_TYPE_MASK 0xf +# define REQ_TYPE_SHIFT 0 +# define RSP_TYPE_MASK 0xf0 +# define RSP_TYPE_SHIFT 4 + /* * UVD */ diff --git a/sys/dev/drm/radeon/radeon.h b/sys/dev/drm/radeon/radeon.h index 909be9f958..2001c7427c 100644 --- a/sys/dev/drm/radeon/radeon.h +++ b/sys/dev/drm/radeon/radeon.h @@ -67,7 +67,8 @@ #include #include #include -#include +#include +#include #include #include @@ -78,6 +79,8 @@ #include #include +#define CONFIG_ACPI 1 + #include "radeon_family.h" #include "radeon_mode.h" #include "radeon_reg.h" @@ -103,6 +106,8 @@ extern int radeon_pcie_gen2; extern int radeon_msi; extern int radeon_lockup_timeout; extern int radeon_fastfb; +extern int radeon_dpm; +extern int radeon_aspm; /* * Copy from radeon_drv.h so we don't have to include both and have conflicting @@ -157,6 +162,13 @@ extern int radeon_fastfb; #define RADEON_RESET_MC (1 << 10) #define RADEON_RESET_DISPLAY (1 << 11) +/* max cursor sizes (in pixels) */ +#define CURSOR_WIDTH 64 +#define CURSOR_HEIGHT 64 + +#define CIK_CURSOR_WIDTH 128 +#define CIK_CURSOR_HEIGHT 128 + /* * Errata workarounds. */ @@ -199,6 +211,7 @@ struct radeon_clock { uint32_t default_mclk; uint32_t default_sclk; uint32_t default_dispclk; + uint32_t current_dispclk; uint32_t dp_extclk; uint32_t max_pixel_clock; }; @@ -218,13 +231,49 @@ int radeon_atom_get_clock_dividers(struct radeon_device *rdev, u32 clock, bool strobe_mode, struct atom_clock_dividers *dividers); +int radeon_atom_get_memory_pll_dividers(struct radeon_device *rdev, + u32 clock, + bool strobe_mode, + struct atom_mpll_param *mpll_param); void radeon_atom_set_voltage(struct radeon_device *rdev, u16 voltage_level, u8 voltage_type); +int radeon_atom_get_voltage_gpio_settings(struct radeon_device *rdev, + u16 voltage_level, u8 voltage_type, + u32 *gpio_value, u32 *gpio_mask); +void radeon_atom_set_engine_dram_timings(struct radeon_device *rdev, + u32 eng_clock, u32 mem_clock); +int radeon_atom_get_voltage_step(struct radeon_device *rdev, + u8 voltage_type, u16 *voltage_step); +int radeon_atom_get_max_vddc(struct radeon_device *rdev, u8 voltage_type, + u16 voltage_id, u16 *voltage); +int radeon_atom_get_leakage_vddc_based_on_leakage_idx(struct radeon_device *rdev, + u16 *voltage, + u16 leakage_idx); +int radeon_atom_round_to_true_voltage(struct radeon_device *rdev, + u8 voltage_type, + u16 nominal_voltage, + u16 *true_voltage); +int radeon_atom_get_min_voltage(struct radeon_device *rdev, + u8 voltage_type, u16 *min_voltage); +int radeon_atom_get_max_voltage(struct radeon_device *rdev, + u8 voltage_type, u16 *max_voltage); +int radeon_atom_get_voltage_table(struct radeon_device *rdev, + u8 voltage_type, u8 voltage_mode, + struct atom_voltage_table *voltage_table); +bool radeon_atom_is_voltage_gpio(struct radeon_device *rdev, + u8 voltage_type, u8 voltage_mode); +void radeon_atom_update_memory_dll(struct radeon_device *rdev, + u32 mem_clock); +void radeon_atom_set_ac_timing(struct radeon_device *rdev, + u32 mem_clock); +int radeon_atom_init_mc_reg_table(struct radeon_device *rdev, + u8 module_index, + struct atom_mc_reg_table *reg_table); +int radeon_atom_get_memory_info(struct radeon_device *rdev, + u8 module_index, struct atom_memory_info *mem_info); +int radeon_atom_get_mclk_range_table(struct radeon_device *rdev, + bool gddr5, u8 module_index, + struct atom_memory_clock_range_table *mclk_range_table); void rs690_pm_info(struct radeon_device *rdev); -extern int rv6xx_get_temp(struct radeon_device *rdev); -extern int rv770_get_temp(struct radeon_device *rdev); -extern int evergreen_get_temp(struct radeon_device *rdev); -extern int sumo_get_temp(struct radeon_device *rdev); -extern int si_get_temp(struct radeon_device *rdev); extern void evergreen_tiling_fields(unsigned tiling_flags, unsigned *bankw, unsigned *bankh, unsigned *mtaspect, unsigned *tile_split); @@ -416,6 +465,7 @@ struct radeon_sa_manager { uint64_t gpu_addr; void *cpu_ptr; uint32_t domain; + uint32_t align; }; struct radeon_sa_bo; @@ -558,6 +608,22 @@ struct radeon_scratch { int radeon_scratch_get(struct radeon_device *rdev, uint32_t *reg); void radeon_scratch_free(struct radeon_device *rdev, uint32_t reg); +/* + * GPU doorbell structures, functions & helpers + */ +struct radeon_doorbell { + u32 num_pages; + bool free[1024]; + /* doorbell mmio */ + resource_size_t base; + resource_size_t size; + void __iomem *ptr; +}; + +int radeon_doorbell_init(struct radeon_device *rdev); +void radeon_doorbell_fini(struct radeon_device *rdev); +int radeon_doorbell_get(struct radeon_device *rdev, u32 *page); +void radeon_doorbell_free(struct radeon_device *rdev, u32 doorbell); /* * IRQS. @@ -609,10 +675,21 @@ struct evergreen_irq_stat_regs { u32 afmt_status6; }; +struct cik_irq_stat_regs { + u32 disp_int; + u32 disp_int_cont; + u32 disp_int_cont2; + u32 disp_int_cont3; + u32 disp_int_cont4; + u32 disp_int_cont5; + u32 disp_int_cont6; +}; + union radeon_irq_stat_regs { struct r500_irq_stat_regs r500; struct r600_irq_stat_regs r600; struct evergreen_irq_stat_regs evergreen; + struct cik_irq_stat_regs cik; }; #define RADEON_MAX_HPD_PINS 6 @@ -629,6 +706,7 @@ struct radeon_irq { bool hpd[RADEON_MAX_HPD_PINS]; bool afmt[RADEON_MAX_AFMT_BLOCKS]; union radeon_irq_stat_regs stat_regs; + bool dpm_thermal; }; int radeon_irq_kms_init(struct radeon_device *rdev); @@ -686,6 +764,22 @@ struct radeon_ring { u32 idx; u64 last_semaphore_signal_addr; u64 last_semaphore_wait_addr; + /* for CIK queues */ + u32 me; + u32 pipe; + u32 queue; + struct radeon_bo *mqd_obj; + u32 doorbell_page_num; + u32 doorbell_offset; + unsigned wptr_offs; +}; + +struct radeon_mec { + struct radeon_bo *hpd_eop_obj; + u64 hpd_eop_gpu_addr; + u32 num_pipe; + u32 num_mec; + u32 num_queue; }; /* @@ -703,6 +797,11 @@ struct radeon_ring { /* number of entries in page table */ #define RADEON_VM_PTE_COUNT (1 << RADEON_VM_BLOCK_SIZE) +/* PTBs (Page Table Blocks) need to be aligned to 32K */ +#define RADEON_VM_PTB_ALIGN_SIZE 32768 +#define RADEON_VM_PTB_ALIGN_MASK (RADEON_VM_PTB_ALIGN_SIZE - 1) +#define RADEON_VM_PTB_ALIGN(a) (((a) + RADEON_VM_PTB_ALIGN_MASK) & ~RADEON_VM_PTB_ALIGN_MASK) + struct radeon_vm { struct list_head list; struct list_head va; @@ -787,15 +886,22 @@ struct r600_blit { }; /* - * SI RLC stuff + * RLC stuff */ -struct si_rlc { +#include "clearstate_defs.h" + +struct radeon_rlc { /* for power gating */ struct radeon_bo *save_restore_obj; uint64_t save_restore_gpu_addr; + volatile uint32_t *sr_ptr; + u32 *reg_list; + u32 reg_list_size; /* for clear state */ struct radeon_bo *clear_state_obj; uint64_t clear_state_gpu_addr; + volatile uint32_t *cs_ptr; + struct cs_section_def *cs_data; }; int radeon_ib_get(struct radeon_device *rdev, int ring, @@ -811,6 +917,12 @@ int radeon_ib_ring_tests(struct radeon_device *rdev); /* Ring access between begin & end cannot sleep */ bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev, struct radeon_ring *ring); +u32 radeon_ring_generic_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +u32 radeon_ring_generic_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void radeon_ring_generic_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *cp); int radeon_ring_alloc(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw); int radeon_ring_lock(struct radeon_device *rdev, struct radeon_ring *cp, unsigned ndw); @@ -943,6 +1055,8 @@ struct radeon_wb { #define CAYMAN_WB_DMA1_RPTR_OFFSET 2304 #define R600_WB_UVD_RPTR_OFFSET 2560 #define R600_WB_EVENT_OFFSET 3072 +#define CIK_WB_CP1_WPTR_OFFSET 3328 +#define CIK_WB_CP2_WPTR_OFFSET 3584 /** * struct radeon_pm - power management datas @@ -967,6 +1081,7 @@ struct radeon_wb { enum radeon_pm_method { PM_METHOD_PROFILE, PM_METHOD_DYNPM, + PM_METHOD_DPM, }; enum radeon_dynpm_state { @@ -992,11 +1107,24 @@ enum radeon_voltage_type { }; enum radeon_pm_state_type { + /* not used for dpm */ POWER_STATE_TYPE_DEFAULT, POWER_STATE_TYPE_POWERSAVE, + /* user selectable states */ POWER_STATE_TYPE_BATTERY, POWER_STATE_TYPE_BALANCED, POWER_STATE_TYPE_PERFORMANCE, + /* internal states */ + POWER_STATE_TYPE_INTERNAL_UVD, + POWER_STATE_TYPE_INTERNAL_UVD_SD, + POWER_STATE_TYPE_INTERNAL_UVD_HD, + POWER_STATE_TYPE_INTERNAL_UVD_HD2, + POWER_STATE_TYPE_INTERNAL_UVD_MVC, + POWER_STATE_TYPE_INTERNAL_BOOT, + POWER_STATE_TYPE_INTERNAL_THERMAL, + POWER_STATE_TYPE_INTERNAL_ACPI, + POWER_STATE_TYPE_INTERNAL_ULV, + POWER_STATE_TYPE_INTERNAL_3DPERF, }; enum radeon_pm_profile_type { @@ -1025,12 +1153,17 @@ struct radeon_pm_profile { enum radeon_int_thermal_type { THERMAL_TYPE_NONE, + THERMAL_TYPE_EXTERNAL, + THERMAL_TYPE_EXTERNAL_GPIO, THERMAL_TYPE_RV6XX, THERMAL_TYPE_RV770, + THERMAL_TYPE_ADT7473_WITH_INTERNAL, THERMAL_TYPE_EVERGREEN, THERMAL_TYPE_SUMO, THERMAL_TYPE_NI, THERMAL_TYPE_SI, + THERMAL_TYPE_EMC2103_WITH_INTERNAL, + THERMAL_TYPE_CI, }; struct radeon_voltage { @@ -1084,6 +1217,201 @@ struct radeon_power_state { */ #define RADEON_MODE_OVERCLOCK_MARGIN 500 /* 5 MHz */ +enum radeon_dpm_auto_throttle_src { + RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, + RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL +}; + +enum radeon_dpm_event_src { + RADEON_DPM_EVENT_SRC_ANALOG = 0, + RADEON_DPM_EVENT_SRC_EXTERNAL = 1, + RADEON_DPM_EVENT_SRC_DIGITAL = 2, + RADEON_DPM_EVENT_SRC_ANALOG_OR_EXTERNAL = 3, + RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL = 4 +}; + +struct radeon_ps { + u32 caps; /* vbios flags */ + u32 class; /* vbios flags */ + u32 class2; /* vbios flags */ + /* UVD clocks */ + u32 vclk; + u32 dclk; + /* asic priv */ + void *ps_priv; +}; + +struct radeon_dpm_thermal { + /* thermal interrupt work */ + struct task work; + /* low temperature threshold */ + int min_temp; + /* high temperature threshold */ + int max_temp; + /* was interrupt low to high or high to low */ + bool high_to_low; +}; + +enum radeon_clk_action +{ + RADEON_SCLK_UP = 1, + RADEON_SCLK_DOWN +}; + +struct radeon_blacklist_clocks +{ + u32 sclk; + u32 mclk; + enum radeon_clk_action action; +}; + +struct radeon_clock_and_voltage_limits { + u32 sclk; + u32 mclk; + u32 vddc; + u32 vddci; +}; + +struct radeon_clock_array { + u32 count; + u32 *values; +}; + +struct radeon_clock_voltage_dependency_entry { + u32 clk; + u16 v; +}; + +struct radeon_clock_voltage_dependency_table { + u32 count; + struct radeon_clock_voltage_dependency_entry *entries; +}; + +struct radeon_cac_leakage_entry { + u16 vddc; + u32 leakage; +}; + +struct radeon_cac_leakage_table { + u32 count; + struct radeon_cac_leakage_entry *entries; +}; + +struct radeon_phase_shedding_limits_entry { + u16 voltage; + u32 sclk; + u32 mclk; +}; + +struct radeon_phase_shedding_limits_table { + u32 count; + struct radeon_phase_shedding_limits_entry *entries; +}; + +struct radeon_ppm_table { + u8 ppm_design; + u16 cpu_core_number; + u32 platform_tdp; + u32 small_ac_platform_tdp; + u32 platform_tdc; + u32 small_ac_platform_tdc; + u32 apu_tdp; + u32 dgpu_tdp; + u32 dgpu_ulv_power; + u32 tj_max; +}; + +struct radeon_dpm_dynamic_state { + struct radeon_clock_voltage_dependency_table vddc_dependency_on_sclk; + struct radeon_clock_voltage_dependency_table vddci_dependency_on_mclk; + struct radeon_clock_voltage_dependency_table vddc_dependency_on_mclk; + struct radeon_clock_voltage_dependency_table vddc_dependency_on_dispclk; + struct radeon_clock_array valid_sclk_values; + struct radeon_clock_array valid_mclk_values; + struct radeon_clock_and_voltage_limits max_clock_voltage_on_dc; + struct radeon_clock_and_voltage_limits max_clock_voltage_on_ac; + u32 mclk_sclk_ratio; + u32 sclk_mclk_delta; + u16 vddc_vddci_delta; + u16 min_vddc_for_pcie_gen2; + struct radeon_cac_leakage_table cac_leakage_table; + struct radeon_phase_shedding_limits_table phase_shedding_limits_table; + struct radeon_ppm_table *ppm_table; +}; + +struct radeon_dpm_fan { + u16 t_min; + u16 t_med; + u16 t_high; + u16 pwm_min; + u16 pwm_med; + u16 pwm_high; + u8 t_hyst; + u32 cycle_delay; + u16 t_max; + bool ucode_fan_control; +}; + +enum radeon_pcie_gen { + RADEON_PCIE_GEN1 = 0, + RADEON_PCIE_GEN2 = 1, + RADEON_PCIE_GEN3 = 2, + RADEON_PCIE_GEN_INVALID = 0xffff +}; + +enum radeon_dpm_forced_level { + RADEON_DPM_FORCED_LEVEL_AUTO = 0, + RADEON_DPM_FORCED_LEVEL_LOW = 1, + RADEON_DPM_FORCED_LEVEL_HIGH = 2, +}; + +struct radeon_dpm { + struct radeon_ps *ps; + /* number of valid power states */ + int num_ps; + /* current power state that is active */ + struct radeon_ps *current_ps; + /* requested power state */ + struct radeon_ps *requested_ps; + /* boot up power state */ + struct radeon_ps *boot_ps; + /* default uvd power state */ + struct radeon_ps *uvd_ps; + enum radeon_pm_state_type state; + enum radeon_pm_state_type user_state; + u32 platform_caps; + u32 voltage_response_time; + u32 backbias_response_time; + void *priv; + u32 new_active_crtcs; + int new_active_crtc_count; + u32 current_active_crtcs; + int current_active_crtc_count; + struct radeon_dpm_dynamic_state dyn_state; + struct radeon_dpm_fan fan; + u32 tdp_limit; + u32 near_tdp_limit; + u32 near_tdp_limit_adjusted; + u32 sq_ramping_threshold; + u32 cac_leakage; + u16 tdp_od_limit; + u32 tdp_adjustment; + u16 load_line_slope; + bool power_control; + bool ac_power; + /* special states active */ + bool thermal_active; + bool uvd_active; + /* thermal handling */ + struct radeon_dpm_thermal thermal; + /* forced levels */ + enum radeon_dpm_forced_level forced_level; +}; + +void radeon_dpm_enable_power_state(struct radeon_device *rdev, + enum radeon_pm_state_type dpm_state); + + struct radeon_pm { struct lock mutex; /* write locked while reprogramming mclk */ @@ -1141,6 +1469,9 @@ struct radeon_pm { #ifdef DUMBBELL_WIP struct device *int_hwmon_dev; #endif /* DUMBBELL_WIP */ + /* dpm */ + bool dpm_enabled; + struct radeon_dpm dpm; }; int radeon_pm_get_type_index(struct radeon_device *rdev, @@ -1157,6 +1488,7 @@ struct radeon_uvd { struct radeon_bo *vcpu_bo; void *cpu_addr; uint64_t gpu_addr; + void *saved_bo; atomic_t handles[RADEON_MAX_UVD_HANDLES]; struct drm_file *filp[RADEON_MAX_UVD_HANDLES]; struct delayed_work idle_work; @@ -1279,6 +1611,10 @@ struct radeon_asic { int (*ib_test)(struct radeon_device *rdev, struct radeon_ring *cp); bool (*is_lockup)(struct radeon_device *rdev, struct radeon_ring *cp); void (*vm_flush)(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); + + u32 (*get_rptr)(struct radeon_device *rdev, struct radeon_ring *ring); + u32 (*get_wptr)(struct radeon_device *rdev, struct radeon_ring *ring); + void (*set_wptr)(struct radeon_device *rdev, struct radeon_ring *ring); } ring[RADEON_NUM_RINGS]; /* irqs */ struct { @@ -1338,7 +1674,7 @@ struct radeon_asic { bool (*sense)(struct radeon_device *rdev, enum radeon_hpd_id hpd); void (*set_polarity)(struct radeon_device *rdev, enum radeon_hpd_id hpd); } hpd; - /* power management */ + /* static power management */ struct { void (*misc)(struct radeon_device *rdev); void (*prepare)(struct radeon_device *rdev); @@ -1353,7 +1689,26 @@ struct radeon_asic { void (*set_pcie_lanes)(struct radeon_device *rdev, int lanes); void (*set_clock_gating)(struct radeon_device *rdev, int enable); int (*set_uvd_clocks)(struct radeon_device *rdev, u32 vclk, u32 dclk); + int (*get_temperature)(struct radeon_device *rdev); } pm; + /* dynamic power management */ + struct { + int (*init)(struct radeon_device *rdev); + void (*setup_asic)(struct radeon_device *rdev); + int (*enable)(struct radeon_device *rdev); + void (*disable)(struct radeon_device *rdev); + int (*pre_set_power_state)(struct radeon_device *rdev); + int (*set_power_state)(struct radeon_device *rdev); + void (*post_set_power_state)(struct radeon_device *rdev); + void (*display_configuration_changed)(struct radeon_device *rdev); + void (*fini)(struct radeon_device *rdev); + u32 (*get_sclk)(struct radeon_device *rdev, bool low); + u32 (*get_mclk)(struct radeon_device *rdev, bool low); + void (*print_power_state)(struct radeon_device *rdev, struct radeon_ps *ps); + void (*debugfs_print_current_performance_level)(struct radeon_device *rdev, struct seq_file *m); + int (*force_performance_level)(struct radeon_device *rdev, enum radeon_dpm_forced_level level); + bool (*vblank_too_short)(struct radeon_device *rdev); + } dpm; /* pageflipping */ struct { void (*pre_page_flip)(struct radeon_device *rdev, int crtc); @@ -1518,6 +1873,36 @@ struct si_asic { uint32_t tile_mode_array[32]; }; +struct cik_asic { + unsigned max_shader_engines; + unsigned max_tile_pipes; + unsigned max_cu_per_sh; + unsigned max_sh_per_se; + unsigned max_backends_per_se; + unsigned max_texture_channel_caches; + unsigned max_gprs; + unsigned max_gs_threads; + unsigned max_hw_contexts; + unsigned sc_prim_fifo_size_frontend; + unsigned sc_prim_fifo_size_backend; + unsigned sc_hiz_tile_fifo_size; + unsigned sc_earlyz_tile_fifo_size; + + unsigned num_tile_pipes; + unsigned num_backends_per_se; + unsigned backend_disable_mask_per_asic; + unsigned backend_map; + unsigned num_texture_channel_caches; + unsigned mem_max_burst_length_bytes; + unsigned mem_row_size_in_kb; + unsigned shader_engine_tile_size; + unsigned num_gpus; + unsigned multi_gpu_tile_size; + + unsigned tile_config; + uint32_t tile_mode_array[32]; +}; + union radeon_asic_config { struct r300_asic r300; struct r100_asic r100; @@ -1526,6 +1911,7 @@ union radeon_asic_config { struct evergreen_asic evergreen; struct cayman_asic cayman; struct si_asic si; + struct cik_asic cik; }; /* @@ -1670,6 +2056,7 @@ struct radeon_device { struct radeon_gart gart; struct radeon_mode_info mode_info; struct radeon_scratch scratch; + struct radeon_doorbell doorbell; struct radeon_mman mman; struct radeon_fence_driver fence_drv[RADEON_NUM_RINGS]; wait_queue_head_t fence_queue; @@ -1697,15 +2084,20 @@ struct radeon_device { const struct firmware *rlc_fw; /* r6/700 RLC firmware */ const struct firmware *mc_fw; /* NI MC firmware */ const struct firmware *ce_fw; /* SI CE firmware */ + const struct firmware *mec_fw; /* CIK MEC firmware */ + const struct firmware *sdma_fw; /* CIK SDMA firmware */ + const struct firmware *smc_fw; /* SMC firmware */ const struct firmware *uvd_fw; /* UVD firmware */ struct r600_blit r600_blit; struct r600_vram_scratch vram_scratch; int msi_enabled; /* msi enabled */ struct r600_ih ih; /* r6/700 interrupt ring */ - struct si_rlc rlc; + struct radeon_rlc rlc; + struct radeon_mec mec; struct taskqueue *tq; struct task hotplug_work; struct task audio_work; + struct task reset_work; int num_crtc; /* number of crtcs */ struct lock dc_hw_i2c_mutex; /* display controller hw i2c mutex */ bool audio_enabled; @@ -1729,6 +2121,8 @@ struct radeon_device { /* ACPI interface */ struct radeon_atif atif; struct radeon_atcs atcs; + /* srbm instance registers */ + struct spinlock srbm_mutex; }; int radeon_device_init(struct radeon_device *rdev, @@ -1744,6 +2138,9 @@ void r100_mm_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v, u32 r100_io_rreg(struct radeon_device *rdev, u32 reg); void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v); +u32 cik_mm_rdoorbell(struct radeon_device *rdev, u32 offset); +void cik_mm_wdoorbell(struct radeon_device *rdev, u32 offset, u32 v); + /* * Cast helper */ @@ -1771,6 +2168,18 @@ void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v); #define WREG32_PCIE(reg, v) rv370_pcie_wreg(rdev, (reg), (v)) #define RREG32_PCIE_PORT(reg) rdev->pciep_rreg(rdev, (reg)) #define WREG32_PCIE_PORT(reg, v) rdev->pciep_wreg(rdev, (reg), (v)) +#define RREG32_SMC(reg) tn_smc_rreg(rdev, (reg)) +#define WREG32_SMC(reg, v) tn_smc_wreg(rdev, (reg), (v)) +#define RREG32_RCU(reg) r600_rcu_rreg(rdev, (reg)) +#define WREG32_RCU(reg, v) r600_rcu_wreg(rdev, (reg), (v)) +#define RREG32_CG(reg) eg_cg_rreg(rdev, (reg)) +#define WREG32_CG(reg, v) eg_cg_wreg(rdev, (reg), (v)) +#define RREG32_PIF_PHY0(reg) eg_pif_phy0_rreg(rdev, (reg)) +#define WREG32_PIF_PHY0(reg, v) eg_pif_phy0_wreg(rdev, (reg), (v)) +#define RREG32_PIF_PHY1(reg) eg_pif_phy1_rreg(rdev, (reg)) +#define WREG32_PIF_PHY1(reg, v) eg_pif_phy1_wreg(rdev, (reg), (v)) +#define RREG32_UVD_CTX(reg) r600_uvd_ctx_rreg(rdev, (reg)) +#define WREG32_UVD_CTX(reg, v) r600_uvd_ctx_wreg(rdev, (reg), (v)) #define WREG32_P(reg, val, mask) \ do { \ uint32_t tmp_ = RREG32(reg); \ @@ -1779,7 +2188,7 @@ void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v); WREG32(reg, tmp_); \ } while (0) #define WREG32_AND(reg, and) WREG32_P(reg, 0, and) -#define WREG32_OR(reg, or) WREG32_P(reg, or, ~or) +#define WREG32_OR(reg, or) WREG32_P(reg, or, ~(or)) #define WREG32_PLL_P(reg, val, mask) \ do { \ uint32_t tmp_ = RREG32_PLL(reg); \ @@ -1791,6 +2200,9 @@ void r100_io_wreg(struct radeon_device *rdev, u32 reg, u32 v); #define RREG32_IO(reg) r100_io_rreg(rdev, (reg)) #define WREG32_IO(reg, v) r100_io_wreg(rdev, (reg), (v)) +#define RDOORBELL32(offset) cik_mm_rdoorbell(rdev, (offset)) +#define WDOORBELL32(offset, v) cik_mm_wdoorbell(rdev, (offset), (v)) + /* * Indirect registers accessor */ @@ -1809,6 +2221,96 @@ static inline void rv370_pcie_wreg(struct radeon_device *rdev, uint32_t reg, uin WREG32(RADEON_PCIE_DATA, (v)); } +static inline u32 tn_smc_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(TN_SMC_IND_INDEX_0, (reg)); + r = RREG32(TN_SMC_IND_DATA_0); + return r; +} + +static inline void tn_smc_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(TN_SMC_IND_INDEX_0, (reg)); + WREG32(TN_SMC_IND_DATA_0, (v)); +} + +static inline u32 r600_rcu_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(R600_RCU_INDEX, ((reg) & 0x1fff)); + r = RREG32(R600_RCU_DATA); + return r; +} + +static inline void r600_rcu_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(R600_RCU_INDEX, ((reg) & 0x1fff)); + WREG32(R600_RCU_DATA, (v)); +} + +static inline u32 eg_cg_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(EVERGREEN_CG_IND_ADDR, ((reg) & 0xffff)); + r = RREG32(EVERGREEN_CG_IND_DATA); + return r; +} + +static inline void eg_cg_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(EVERGREEN_CG_IND_ADDR, ((reg) & 0xffff)); + WREG32(EVERGREEN_CG_IND_DATA, (v)); +} + +static inline u32 eg_pif_phy0_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(EVERGREEN_PIF_PHY0_INDEX, ((reg) & 0xffff)); + r = RREG32(EVERGREEN_PIF_PHY0_DATA); + return r; +} + +static inline void eg_pif_phy0_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(EVERGREEN_PIF_PHY0_INDEX, ((reg) & 0xffff)); + WREG32(EVERGREEN_PIF_PHY0_DATA, (v)); +} + +static inline u32 eg_pif_phy1_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(EVERGREEN_PIF_PHY1_INDEX, ((reg) & 0xffff)); + r = RREG32(EVERGREEN_PIF_PHY1_DATA); + return r; +} + +static inline void eg_pif_phy1_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(EVERGREEN_PIF_PHY1_INDEX, ((reg) & 0xffff)); + WREG32(EVERGREEN_PIF_PHY1_DATA, (v)); +} + +static inline u32 r600_uvd_ctx_rreg(struct radeon_device *rdev, u32 reg) +{ + u32 r; + + WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff)); + r = RREG32(R600_UVD_CTX_DATA); + return r; +} + +static inline void r600_uvd_ctx_wreg(struct radeon_device *rdev, u32 reg, u32 v) +{ + WREG32(R600_UVD_CTX_INDEX, ((reg) & 0x1ff)); + WREG32(R600_UVD_CTX_DATA, (v)); +} + void r100_pll_errata_after_index(struct radeon_device *rdev); @@ -1857,6 +2359,16 @@ void r100_pll_errata_after_index(struct radeon_device *rdev); (rdev->flags & RADEON_IS_IGP)) #define ASIC_IS_DCE64(rdev) ((rdev->family == CHIP_OLAND)) #define ASIC_IS_NODCE(rdev) ((rdev->family == CHIP_HAINAN)) +#define ASIC_IS_DCE8(rdev) ((rdev->family >= CHIP_BONAIRE)) + +#define ASIC_IS_LOMBOK(rdev) ((rdev->ddev->pci_device == 0x6849) || \ + (rdev->ddev->pci_device == 0x6850) || \ + (rdev->ddev->pci_device == 0x6858) || \ + (rdev->ddev->pci_device == 0x6859) || \ + (rdev->ddev->pci_device == 0x6840) || \ + (rdev->ddev->pci_device == 0x6841) || \ + (rdev->ddev->pci_device == 0x6842) || \ + (rdev->ddev->pci_device == 0x6843)) /* * BIOS helpers. @@ -1909,6 +2421,9 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_ring_ib_parse(rdev, r, ib) (rdev)->asic->ring[(r)].ib_parse((rdev), (ib)) #define radeon_ring_is_lockup(rdev, r, cp) (rdev)->asic->ring[(r)].is_lockup((rdev), (cp)) #define radeon_ring_vm_flush(rdev, r, vm) (rdev)->asic->ring[(r)].vm_flush((rdev), (r), (vm)) +#define radeon_ring_get_rptr(rdev, r) (rdev)->asic->ring[(r)->idx].get_rptr((rdev), (r)) +#define radeon_ring_get_wptr(rdev, r) (rdev)->asic->ring[(r)->idx].get_wptr((rdev), (r)) +#define radeon_ring_set_wptr(rdev, r) (rdev)->asic->ring[(r)->idx].set_wptr((rdev), (r)) #define radeon_irq_set(rdev) (rdev)->asic->irq.set((rdev)) #define radeon_irq_process(rdev) (rdev)->asic->irq.process((rdev)) #define radeon_get_vblank_counter(rdev, crtc) (rdev)->asic->display.get_vblank_counter((rdev), (crtc)) @@ -1932,6 +2447,7 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_set_pcie_lanes(rdev, l) (rdev)->asic->pm.set_pcie_lanes((rdev), (l)) #define radeon_set_clock_gating(rdev, e) (rdev)->asic->pm.set_clock_gating((rdev), (e)) #define radeon_set_uvd_clocks(rdev, v, d) (rdev)->asic->pm.set_uvd_clocks((rdev), (v), (d)) +#define radeon_get_temperature(rdev) (rdev)->asic->pm.get_temperature((rdev)) #define radeon_set_surface_reg(rdev, r, f, p, o, s) ((rdev)->asic->surface.set_reg((rdev), (r), (f), (p), (o), (s))) #define radeon_clear_surface_reg(rdev, r) ((rdev)->asic->surface.clear_reg((rdev), (r))) #define radeon_bandwidth_update(rdev) (rdev)->asic->display.bandwidth_update((rdev)) @@ -1952,6 +2468,21 @@ void radeon_ring_write(struct radeon_ring *ring, uint32_t v); #define radeon_mc_wait_for_idle(rdev) (rdev)->asic->mc_wait_for_idle((rdev)) #define radeon_get_xclk(rdev) (rdev)->asic->get_xclk((rdev)) #define radeon_get_gpu_clock_counter(rdev) (rdev)->asic->get_gpu_clock_counter((rdev)) +#define radeon_dpm_init(rdev) rdev->asic->dpm.init((rdev)) +#define radeon_dpm_setup_asic(rdev) rdev->asic->dpm.setup_asic((rdev)) +#define radeon_dpm_enable(rdev) rdev->asic->dpm.enable((rdev)) +#define radeon_dpm_disable(rdev) rdev->asic->dpm.disable((rdev)) +#define radeon_dpm_pre_set_power_state(rdev) rdev->asic->dpm.pre_set_power_state((rdev)) +#define radeon_dpm_set_power_state(rdev) rdev->asic->dpm.set_power_state((rdev)) +#define radeon_dpm_post_set_power_state(rdev) rdev->asic->dpm.post_set_power_state((rdev)) +#define radeon_dpm_display_configuration_changed(rdev) rdev->asic->dpm.display_configuration_changed((rdev)) +#define radeon_dpm_fini(rdev) rdev->asic->dpm.fini((rdev)) +#define radeon_dpm_get_sclk(rdev, l) rdev->asic->dpm.get_sclk((rdev), (l)) +#define radeon_dpm_get_mclk(rdev, l) rdev->asic->dpm.get_mclk((rdev), (l)) +#define radeon_dpm_print_power_state(rdev, ps) rdev->asic->dpm.print_power_state((rdev), (ps)) +#define radeon_dpm_debugfs_print_current_performance_level(rdev, m) rdev->asic->dpm.debugfs_print_current_performance_level((rdev), (m)) +#define radeon_dpm_force_performance_level(rdev, l) rdev->asic->dpm.force_performance_level((rdev), (l)) +#define radeon_dpm_vblank_too_short(rdev) rdev->asic->dpm.vblank_too_short((rdev)) /* Common functions */ /* AGP */ @@ -2069,6 +2600,10 @@ extern void ni_fini_microcode(struct radeon_device *rdev); /* radeon_acpi.c */ extern int radeon_acpi_init(struct radeon_device *rdev); extern void radeon_acpi_fini(struct radeon_device *rdev); +extern bool radeon_acpi_is_pcie_performance_request_supported(struct radeon_device *rdev); +extern int radeon_acpi_pcie_performance_request(struct radeon_device *rdev, + u8 perf_req, bool advertise); +extern int radeon_acpi_pcie_notify_device_ready(struct radeon_device *rdev); /* Prototypes added by @dumbbell. */ diff --git a/sys/dev/drm/radeon/radeon_acpi.c b/sys/dev/drm/radeon/radeon_acpi.c index 86e561fefd..04ef85d8ee 100644 --- a/sys/dev/drm/radeon/radeon_acpi.c +++ b/sys/dev/drm/radeon/radeon_acpi.c @@ -67,6 +67,22 @@ struct atcs_verify_interface { u32 function_bits; /* supported functions bit vector */ } __packed; +#define ATCS_VALID_FLAGS_MASK 0x3 + +struct atcs_pref_req_input { + u16 size; /* structure size in bytes (includes size field) */ + u16 client_id; /* client id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */ + u16 valid_flags_mask; /* valid flags mask */ + u16 flags; /* flags */ + u8 req_type; /* request type */ + u8 perf_req; /* performance request */ +} __packed; + +struct atcs_pref_req_output { + u16 size; /* structure size in bytes (includes size field) */ + u8 ret_val; /* return value */ +} __packed; + /* Call the ATIF method */ /** @@ -490,6 +506,142 @@ out: return err; } +/** + * radeon_acpi_is_pcie_performance_request_supported + * + * @rdev: radeon_device pointer + * + * Check if the ATCS pcie_perf_req and pcie_dev_rdy methods + * are supported (all asics). + * returns true if supported, false if not. + */ +bool radeon_acpi_is_pcie_performance_request_supported(struct radeon_device *rdev) +{ + struct radeon_atcs *atcs = &rdev->atcs; + + if (atcs->functions.pcie_perf_req && atcs->functions.pcie_dev_rdy) + return true; + + return false; +} + +/** + * radeon_acpi_pcie_notify_device_ready + * + * @rdev: radeon_device pointer + * + * Executes the PCIE_DEVICE_READY_NOTIFICATION method + * (all asics). + * returns 0 on success, error on failure. + */ +int radeon_acpi_pcie_notify_device_ready(struct radeon_device *rdev) +{ +#ifdef MN_TODO + acpi_handle handle; + union acpi_object *info; + struct radeon_atcs *atcs = &rdev->atcs; + + /* Get the device handle */ + handle = DEVICE_ACPI_HANDLE(&rdev->pdev->dev); + if (!handle) + return -EINVAL; + + if (!atcs->functions.pcie_dev_rdy) + return -EINVAL; + + info = radeon_atcs_call(handle, ATCS_FUNCTION_PCIE_DEVICE_READY_NOTIFICATION, NULL); + if (!info) + return -EIO; + + kfree(info); + return 0; +#else + return -EINVAL; +#endif +} + +/** + * radeon_acpi_pcie_performance_request + * + * @rdev: radeon_device pointer + * @perf_req: requested perf level (pcie gen speed) + * @advertise: set advertise caps flag if set + * + * Executes the PCIE_PERFORMANCE_REQUEST method to + * change the pcie gen speed (all asics). + * returns 0 on success, error on failure. + */ +int radeon_acpi_pcie_performance_request(struct radeon_device *rdev, + u8 perf_req, bool advertise) +{ +#ifdef MN_TODO + acpi_handle handle; + union acpi_object *info; + struct radeon_atcs *atcs = &rdev->atcs; + struct atcs_pref_req_input atcs_input; + struct atcs_pref_req_output atcs_output; + struct acpi_buffer params; + size_t size; + u32 retry = 3; + + /* Get the device handle */ + handle = DEVICE_ACPI_HANDLE(&rdev->pdev->dev); + if (!handle) + return -EINVAL; + + if (!atcs->functions.pcie_perf_req) + return -EINVAL; + + atcs_input.size = sizeof(struct atcs_pref_req_input); + /* client id (bit 2-0: func num, 7-3: dev num, 15-8: bus num) */ + atcs_input.client_id = rdev->pdev->devfn | (rdev->pdev->bus->number << 8); + atcs_input.valid_flags_mask = ATCS_VALID_FLAGS_MASK; + atcs_input.flags = ATCS_WAIT_FOR_COMPLETION; + if (advertise) + atcs_input.flags |= ATCS_ADVERTISE_CAPS; + atcs_input.req_type = ATCS_PCIE_LINK_SPEED; + atcs_input.perf_req = perf_req; + + params.length = sizeof(struct atcs_pref_req_input); + params.pointer = &atcs_input; + + while (retry--) { + info = radeon_atcs_call(handle, ATCS_FUNCTION_PCIE_PERFORMANCE_REQUEST, ¶ms); + if (!info) + return -EIO; + + memset(&atcs_output, 0, sizeof(atcs_output)); + + size = *(u16 *) info->buffer.pointer; + if (size < 3) { + DRM_INFO("ATCS buffer is too small: %zu\n", size); + kfree(info); + return -EINVAL; + } + size = min(sizeof(atcs_output), size); + + memcpy(&atcs_output, info->buffer.pointer, size); + + kfree(info); + + switch (atcs_output.ret_val) { + case ATCS_REQUEST_REFUSED: + default: + return -EINVAL; + case ATCS_REQUEST_COMPLETE: + return 0; + case ATCS_REQUEST_IN_PROGRESS: + DRM_UDELAY(10); + break; + } + } + + return 0; +#else + return -EINVAL; +#endif +} + /** * radeon_acpi_event - handle notify events * diff --git a/sys/dev/drm/radeon/radeon_asic.c b/sys/dev/drm/radeon/radeon_asic.c index 361785b93f..160108be14 100644 --- a/sys/dev/drm/radeon/radeon_asic.c +++ b/sys/dev/drm/radeon/radeon_asic.c @@ -35,6 +35,8 @@ #include "radeon.h" #include "radeon_asic.h" #include "atom.h" +#include "rv770_dpm.h" +#include "ni_dpm.h" /* * Registers accessors functions. @@ -123,7 +125,11 @@ static void radeon_register_accessor_init(struct radeon_device *rdev) rdev->mc_rreg = &rs780_mc_rreg; rdev->mc_wreg = &rs780_mc_wreg; } - if (rdev->family >= CHIP_R600) { + + if (rdev->family >= CHIP_BONAIRE) { + rdev->pciep_rreg = &cik_pciep_rreg; + rdev->pciep_wreg = &cik_pciep_wreg; + } else if (rdev->family >= CHIP_R600) { rdev->pciep_rreg = &r600_pciep_rreg; rdev->pciep_wreg = &r600_pciep_wreg; } @@ -189,6 +195,9 @@ static struct radeon_asic r100_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -265,6 +274,9 @@ static struct radeon_asic r200_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -341,6 +353,9 @@ static struct radeon_asic r300_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -417,6 +432,9 @@ static struct radeon_asic r300_asic_pcie = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -493,6 +511,9 @@ static struct radeon_asic r420_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -569,6 +590,9 @@ static struct radeon_asic rs400_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -645,6 +669,9 @@ static struct radeon_asic rs600_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -723,6 +750,9 @@ static struct radeon_asic rs690_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -801,6 +831,9 @@ static struct radeon_asic rv515_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -877,6 +910,9 @@ static struct radeon_asic r520_asic = { .ring_test = &r100_ring_test, .ib_test = &r100_ib_test, .is_lockup = &r100_gpu_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -954,6 +990,9 @@ static struct radeon_asic r600_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &r600_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &r600_dma_ring_ib_execute, @@ -963,6 +1002,9 @@ static struct radeon_asic r600_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &r600_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -983,8 +1025,8 @@ static struct radeon_asic r600_asic = { .blit_ring_index = RADEON_RING_TYPE_GFX_INDEX, .dma = &r600_copy_dma, .dma_ring_index = R600_RING_TYPE_DMA_INDEX, - .copy = &r600_copy_dma, - .copy_ring_index = R600_RING_TYPE_DMA_INDEX, + .copy = &r600_copy_cpdma, + .copy_ring_index = RADEON_RING_TYPE_GFX_INDEX, }, .surface = { .set_reg = r600_set_surface_reg, @@ -1009,6 +1051,116 @@ static struct radeon_asic r600_asic = { .get_pcie_lanes = &r600_get_pcie_lanes, .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = NULL, + .get_temperature = &rv6xx_get_temp, + }, + .pflip = { + .pre_page_flip = &rs600_pre_page_flip, + .page_flip = &rs600_page_flip, + .post_page_flip = &rs600_post_page_flip, + }, +}; + +static struct radeon_asic rv6xx_asic = { + .init = &r600_init, + .fini = &r600_fini, + .suspend = &r600_suspend, + .resume = &r600_resume, + .vga_set_state = &r600_vga_set_state, + .asic_reset = &r600_asic_reset, + .ioctl_wait_idle = r600_ioctl_wait_idle, + .gui_idle = &r600_gui_idle, + .mc_wait_for_idle = &r600_mc_wait_for_idle, + .get_xclk = &r600_get_xclk, + .get_gpu_clock_counter = &r600_get_gpu_clock_counter, + .gart = { + .tlb_flush = &r600_pcie_gart_tlb_flush, + .set_page = &rs600_gart_set_page, + }, + .ring = { + [RADEON_RING_TYPE_GFX_INDEX] = { + .ib_execute = &r600_ring_ib_execute, + .emit_fence = &r600_fence_ring_emit, + .emit_semaphore = &r600_semaphore_ring_emit, + .cs_parse = &r600_cs_parse, + .ring_test = &r600_ring_test, + .ib_test = &r600_ib_test, + .is_lockup = &r600_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [R600_RING_TYPE_DMA_INDEX] = { + .ib_execute = &r600_dma_ring_ib_execute, + .emit_fence = &r600_dma_fence_ring_emit, + .emit_semaphore = &r600_dma_semaphore_ring_emit, + .cs_parse = &r600_dma_cs_parse, + .ring_test = &r600_dma_ring_test, + .ib_test = &r600_dma_ib_test, + .is_lockup = &r600_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + } + }, + .irq = { + .set = &r600_irq_set, + .process = &r600_irq_process, + }, + .display = { + .bandwidth_update = &rv515_bandwidth_update, + .get_vblank_counter = &rs600_get_vblank_counter, + .wait_for_vblank = &avivo_wait_for_vblank, + .set_backlight_level = &atombios_set_backlight_level, + .get_backlight_level = &atombios_get_backlight_level, + }, + .copy = { + .blit = &r600_copy_blit, + .blit_ring_index = RADEON_RING_TYPE_GFX_INDEX, + .dma = &r600_copy_dma, + .dma_ring_index = R600_RING_TYPE_DMA_INDEX, + .copy = &r600_copy_cpdma, + .copy_ring_index = RADEON_RING_TYPE_GFX_INDEX, + }, + .surface = { + .set_reg = r600_set_surface_reg, + .clear_reg = r600_clear_surface_reg, + }, + .hpd = { + .init = &r600_hpd_init, + .fini = &r600_hpd_fini, + .sense = &r600_hpd_sense, + .set_polarity = &r600_hpd_set_polarity, + }, + .pm = { + .misc = &r600_pm_misc, + .prepare = &rs600_pm_prepare, + .finish = &rs600_pm_finish, + .init_profile = &r600_pm_init_profile, + .get_dynpm_state = &r600_pm_get_dynpm_state, + .get_engine_clock = &radeon_atom_get_engine_clock, + .set_engine_clock = &radeon_atom_set_engine_clock, + .get_memory_clock = &radeon_atom_get_memory_clock, + .set_memory_clock = &radeon_atom_set_memory_clock, + .get_pcie_lanes = &r600_get_pcie_lanes, + .set_pcie_lanes = &r600_set_pcie_lanes, + .set_clock_gating = NULL, + .get_temperature = &rv6xx_get_temp, + }, + .dpm = { + .init = &rv6xx_dpm_init, + .setup_asic = &rv6xx_setup_asic, + .enable = &rv6xx_dpm_enable, + .disable = &rv6xx_dpm_disable, + .pre_set_power_state = &r600_dpm_pre_set_power_state, + .set_power_state = &rv6xx_dpm_set_power_state, + .post_set_power_state = &r600_dpm_post_set_power_state, + .display_configuration_changed = &rv6xx_dpm_display_configuration_changed, + .fini = &rv6xx_dpm_fini, + .get_sclk = &rv6xx_dpm_get_sclk, + .get_mclk = &rv6xx_dpm_get_mclk, + .print_power_state = &rv6xx_dpm_print_power_state, + .debugfs_print_current_performance_level = &rv6xx_dpm_debugfs_print_current_performance_level, + .force_performance_level = &rv6xx_dpm_force_performance_level, }, .pflip = { .pre_page_flip = &rs600_pre_page_flip, @@ -1042,6 +1194,9 @@ static struct radeon_asic rs780_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &r600_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &r600_dma_ring_ib_execute, @@ -1051,6 +1206,9 @@ static struct radeon_asic rs780_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &r600_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1071,8 +1229,8 @@ static struct radeon_asic rs780_asic = { .blit_ring_index = RADEON_RING_TYPE_GFX_INDEX, .dma = &r600_copy_dma, .dma_ring_index = R600_RING_TYPE_DMA_INDEX, - .copy = &r600_copy_dma, - .copy_ring_index = R600_RING_TYPE_DMA_INDEX, + .copy = &r600_copy_cpdma, + .copy_ring_index = RADEON_RING_TYPE_GFX_INDEX, }, .surface = { .set_reg = r600_set_surface_reg, @@ -1097,6 +1255,22 @@ static struct radeon_asic rs780_asic = { .get_pcie_lanes = NULL, .set_pcie_lanes = NULL, .set_clock_gating = NULL, + .get_temperature = &rv6xx_get_temp, + }, + .dpm = { + .init = &rs780_dpm_init, + .setup_asic = &rs780_dpm_setup_asic, + .enable = &rs780_dpm_enable, + .disable = &rs780_dpm_disable, + .pre_set_power_state = &r600_dpm_pre_set_power_state, + .set_power_state = &rs780_dpm_set_power_state, + .post_set_power_state = &r600_dpm_post_set_power_state, + .display_configuration_changed = &rs780_dpm_display_configuration_changed, + .fini = &rs780_dpm_fini, + .get_sclk = &rs780_dpm_get_sclk, + .get_mclk = &rs780_dpm_get_mclk, + .print_power_state = &rs780_dpm_print_power_state, + .debugfs_print_current_performance_level = &rs780_dpm_debugfs_print_current_performance_level, }, .pflip = { .pre_page_flip = &rs600_pre_page_flip, @@ -1130,6 +1304,9 @@ static struct radeon_asic rv770_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &r600_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &r600_dma_ring_ib_execute, @@ -1139,6 +1316,9 @@ static struct radeon_asic rv770_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &r600_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1148,6 +1328,9 @@ static struct radeon_asic rv770_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1195,6 +1378,24 @@ static struct radeon_asic rv770_asic = { .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = &radeon_atom_set_clock_gating, .set_uvd_clocks = &rv770_set_uvd_clocks, + .get_temperature = &rv770_get_temp, + }, + .dpm = { + .init = &rv770_dpm_init, + .setup_asic = &rv770_dpm_setup_asic, + .enable = &rv770_dpm_enable, + .disable = &rv770_dpm_disable, + .pre_set_power_state = &r600_dpm_pre_set_power_state, + .set_power_state = &rv770_dpm_set_power_state, + .post_set_power_state = &r600_dpm_post_set_power_state, + .display_configuration_changed = &rv770_dpm_display_configuration_changed, + .fini = &rv770_dpm_fini, + .get_sclk = &rv770_dpm_get_sclk, + .get_mclk = &rv770_dpm_get_mclk, + .print_power_state = &rv770_dpm_print_power_state, + .debugfs_print_current_performance_level = &rv770_dpm_debugfs_print_current_performance_level, + .force_performance_level = &rv770_dpm_force_performance_level, + .vblank_too_short = &rv770_dpm_vblank_too_short, }, .pflip = { .pre_page_flip = &rs600_pre_page_flip, @@ -1228,6 +1429,9 @@ static struct radeon_asic evergreen_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &evergreen_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &evergreen_dma_ring_ib_execute, @@ -1237,6 +1441,9 @@ static struct radeon_asic evergreen_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &evergreen_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1246,6 +1453,9 @@ static struct radeon_asic evergreen_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1293,6 +1503,24 @@ static struct radeon_asic evergreen_asic = { .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = NULL, .set_uvd_clocks = &evergreen_set_uvd_clocks, + .get_temperature = &evergreen_get_temp, + }, + .dpm = { + .init = &cypress_dpm_init, + .setup_asic = &cypress_dpm_setup_asic, + .enable = &cypress_dpm_enable, + .disable = &cypress_dpm_disable, + .pre_set_power_state = &r600_dpm_pre_set_power_state, + .set_power_state = &cypress_dpm_set_power_state, + .post_set_power_state = &r600_dpm_post_set_power_state, + .display_configuration_changed = &cypress_dpm_display_configuration_changed, + .fini = &cypress_dpm_fini, + .get_sclk = &rv770_dpm_get_sclk, + .get_mclk = &rv770_dpm_get_mclk, + .print_power_state = &rv770_dpm_print_power_state, + .debugfs_print_current_performance_level = &rv770_dpm_debugfs_print_current_performance_level, + .force_performance_level = &rv770_dpm_force_performance_level, + .vblank_too_short = &cypress_dpm_vblank_too_short, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1326,6 +1554,9 @@ static struct radeon_asic sumo_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &evergreen_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &evergreen_dma_ring_ib_execute, @@ -1335,6 +1566,9 @@ static struct radeon_asic sumo_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &evergreen_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1344,6 +1578,9 @@ static struct radeon_asic sumo_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1391,6 +1628,23 @@ static struct radeon_asic sumo_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &sumo_set_uvd_clocks, + .get_temperature = &sumo_get_temp, + }, + .dpm = { + .init = &sumo_dpm_init, + .setup_asic = &sumo_dpm_setup_asic, + .enable = &sumo_dpm_enable, + .disable = &sumo_dpm_disable, + .pre_set_power_state = &sumo_dpm_pre_set_power_state, + .set_power_state = &sumo_dpm_set_power_state, + .post_set_power_state = &sumo_dpm_post_set_power_state, + .display_configuration_changed = &sumo_dpm_display_configuration_changed, + .fini = &sumo_dpm_fini, + .get_sclk = &sumo_dpm_get_sclk, + .get_mclk = &sumo_dpm_get_mclk, + .print_power_state = &sumo_dpm_print_power_state, + .debugfs_print_current_performance_level = &sumo_dpm_debugfs_print_current_performance_level, + .force_performance_level = &sumo_dpm_force_performance_level, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1424,6 +1678,9 @@ static struct radeon_asic btc_asic = { .ring_test = &r600_ring_test, .ib_test = &r600_ib_test, .is_lockup = &evergreen_gfx_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &evergreen_dma_ring_ib_execute, @@ -1433,6 +1690,9 @@ static struct radeon_asic btc_asic = { .ring_test = &r600_dma_ring_test, .ib_test = &r600_dma_ib_test, .is_lockup = &evergreen_dma_is_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1442,6 +1702,9 @@ static struct radeon_asic btc_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1489,6 +1752,24 @@ static struct radeon_asic btc_asic = { .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = NULL, .set_uvd_clocks = &evergreen_set_uvd_clocks, + .get_temperature = &evergreen_get_temp, + }, + .dpm = { + .init = &btc_dpm_init, + .setup_asic = &btc_dpm_setup_asic, + .enable = &btc_dpm_enable, + .disable = &btc_dpm_disable, + .pre_set_power_state = &btc_dpm_pre_set_power_state, + .set_power_state = &btc_dpm_set_power_state, + .post_set_power_state = &btc_dpm_post_set_power_state, + .display_configuration_changed = &cypress_dpm_display_configuration_changed, + .fini = &btc_dpm_fini, + .get_sclk = &btc_dpm_get_sclk, + .get_mclk = &btc_dpm_get_mclk, + .print_power_state = &rv770_dpm_print_power_state, + .debugfs_print_current_performance_level = &rv770_dpm_debugfs_print_current_performance_level, + .force_performance_level = &rv770_dpm_force_performance_level, + .vblank_too_short = &btc_dpm_vblank_too_short, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1530,6 +1811,9 @@ static struct radeon_asic cayman_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1541,6 +1825,9 @@ static struct radeon_asic cayman_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1552,6 +1839,9 @@ static struct radeon_asic cayman_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1563,6 +1853,9 @@ static struct radeon_asic cayman_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &cayman_dma_is_lockup, .vm_flush = &cayman_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_DMA1_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1574,6 +1867,9 @@ static struct radeon_asic cayman_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &cayman_dma_is_lockup, .vm_flush = &cayman_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1583,6 +1879,9 @@ static struct radeon_asic cayman_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1630,6 +1929,24 @@ static struct radeon_asic cayman_asic = { .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = NULL, .set_uvd_clocks = &evergreen_set_uvd_clocks, + .get_temperature = &evergreen_get_temp, + }, + .dpm = { + .init = &ni_dpm_init, + .setup_asic = &ni_dpm_setup_asic, + .enable = &ni_dpm_enable, + .disable = &ni_dpm_disable, + .pre_set_power_state = &ni_dpm_pre_set_power_state, + .set_power_state = &ni_dpm_set_power_state, + .post_set_power_state = &ni_dpm_post_set_power_state, + .display_configuration_changed = &cypress_dpm_display_configuration_changed, + .fini = &ni_dpm_fini, + .get_sclk = &ni_dpm_get_sclk, + .get_mclk = &ni_dpm_get_mclk, + .print_power_state = &ni_dpm_print_power_state, + .debugfs_print_current_performance_level = &ni_dpm_debugfs_print_current_performance_level, + .force_performance_level = &ni_dpm_force_performance_level, + .vblank_too_short = &ni_dpm_vblank_too_short, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1671,6 +1988,9 @@ static struct radeon_asic trinity_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1682,6 +2002,9 @@ static struct radeon_asic trinity_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &cayman_ring_ib_execute, @@ -1693,6 +2016,9 @@ static struct radeon_asic trinity_asic = { .ib_test = &r600_ib_test, .is_lockup = &cayman_gfx_is_lockup, .vm_flush = &cayman_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1704,6 +2030,9 @@ static struct radeon_asic trinity_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &cayman_dma_is_lockup, .vm_flush = &cayman_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_DMA1_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1715,6 +2044,9 @@ static struct radeon_asic trinity_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &cayman_dma_is_lockup, .vm_flush = &cayman_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1724,6 +2056,9 @@ static struct radeon_asic trinity_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1769,6 +2104,23 @@ static struct radeon_asic trinity_asic = { .set_pcie_lanes = NULL, .set_clock_gating = NULL, .set_uvd_clocks = &sumo_set_uvd_clocks, + .get_temperature = &tn_get_temp, + }, + .dpm = { + .init = &trinity_dpm_init, + .setup_asic = &trinity_dpm_setup_asic, + .enable = &trinity_dpm_enable, + .disable = &trinity_dpm_disable, + .pre_set_power_state = &trinity_dpm_pre_set_power_state, + .set_power_state = &trinity_dpm_set_power_state, + .post_set_power_state = &trinity_dpm_post_set_power_state, + .display_configuration_changed = &trinity_dpm_display_configuration_changed, + .fini = &trinity_dpm_fini, + .get_sclk = &trinity_dpm_get_sclk, + .get_mclk = &trinity_dpm_get_mclk, + .print_power_state = &trinity_dpm_print_power_state, + .debugfs_print_current_performance_level = &trinity_dpm_debugfs_print_current_performance_level, + .force_performance_level = &trinity_dpm_force_performance_level, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1810,6 +2162,9 @@ static struct radeon_asic si_asic = { .ib_test = &r600_ib_test, .is_lockup = &si_gfx_is_lockup, .vm_flush = &si_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP1_INDEX] = { .ib_execute = &si_ring_ib_execute, @@ -1821,6 +2176,9 @@ static struct radeon_asic si_asic = { .ib_test = &r600_ib_test, .is_lockup = &si_gfx_is_lockup, .vm_flush = &si_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_CP2_INDEX] = { .ib_execute = &si_ring_ib_execute, @@ -1832,6 +2190,9 @@ static struct radeon_asic si_asic = { .ib_test = &r600_ib_test, .is_lockup = &si_gfx_is_lockup, .vm_flush = &si_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_DMA_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1843,6 +2204,9 @@ static struct radeon_asic si_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &si_dma_is_lockup, .vm_flush = &si_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [CAYMAN_RING_TYPE_DMA1_INDEX] = { .ib_execute = &cayman_dma_ring_ib_execute, @@ -1854,6 +2218,9 @@ static struct radeon_asic si_asic = { .ib_test = &r600_dma_ib_test, .is_lockup = &si_dma_is_lockup, .vm_flush = &si_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, }, [R600_RING_TYPE_UVD_INDEX] = { .ib_execute = &r600_uvd_ib_execute, @@ -1863,6 +2230,9 @@ static struct radeon_asic si_asic = { .ring_test = &r600_uvd_ring_test, .ib_test = &r600_uvd_ib_test, .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, } }, .irq = { @@ -1908,6 +2278,334 @@ static struct radeon_asic si_asic = { .set_pcie_lanes = &r600_set_pcie_lanes, .set_clock_gating = NULL, .set_uvd_clocks = &si_set_uvd_clocks, + .get_temperature = &si_get_temp, + }, + .dpm = { + .init = &si_dpm_init, + .setup_asic = &si_dpm_setup_asic, + .enable = &si_dpm_enable, + .disable = &si_dpm_disable, + .pre_set_power_state = &si_dpm_pre_set_power_state, + .set_power_state = &si_dpm_set_power_state, + .post_set_power_state = &si_dpm_post_set_power_state, + .display_configuration_changed = &si_dpm_display_configuration_changed, + .fini = &si_dpm_fini, + .get_sclk = &ni_dpm_get_sclk, + .get_mclk = &ni_dpm_get_mclk, + .print_power_state = &ni_dpm_print_power_state, + .debugfs_print_current_performance_level = &si_dpm_debugfs_print_current_performance_level, + .force_performance_level = &si_dpm_force_performance_level, + .vblank_too_short = &ni_dpm_vblank_too_short, + }, + .pflip = { + .pre_page_flip = &evergreen_pre_page_flip, + .page_flip = &evergreen_page_flip, + .post_page_flip = &evergreen_post_page_flip, + }, +}; + +static struct radeon_asic ci_asic = { + .init = &cik_init, + .fini = &cik_fini, + .suspend = &cik_suspend, + .resume = &cik_resume, + .asic_reset = &cik_asic_reset, + .vga_set_state = &r600_vga_set_state, + .ioctl_wait_idle = NULL, + .gui_idle = &r600_gui_idle, + .mc_wait_for_idle = &evergreen_mc_wait_for_idle, + .get_xclk = &cik_get_xclk, + .get_gpu_clock_counter = &cik_get_gpu_clock_counter, + .gart = { + .tlb_flush = &cik_pcie_gart_tlb_flush, + .set_page = &rs600_gart_set_page, + }, + .vm = { + .init = &cik_vm_init, + .fini = &cik_vm_fini, + .pt_ring_index = R600_RING_TYPE_DMA_INDEX, + .set_page = &cik_vm_set_page, + }, + .ring = { + [RADEON_RING_TYPE_GFX_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_gfx_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [CAYMAN_RING_TYPE_CP1_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_compute_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &cik_compute_ring_get_rptr, + .get_wptr = &cik_compute_ring_get_wptr, + .set_wptr = &cik_compute_ring_set_wptr, + }, + [CAYMAN_RING_TYPE_CP2_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_compute_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &cik_compute_ring_get_rptr, + .get_wptr = &cik_compute_ring_get_wptr, + .set_wptr = &cik_compute_ring_set_wptr, + }, + [R600_RING_TYPE_DMA_INDEX] = { + .ib_execute = &cik_sdma_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_sdma_fence_ring_emit, + .emit_semaphore = &cik_sdma_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_sdma_ring_test, + .ib_test = &cik_sdma_ib_test, + .is_lockup = &cik_sdma_is_lockup, + .vm_flush = &cik_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [CAYMAN_RING_TYPE_DMA1_INDEX] = { + .ib_execute = &cik_sdma_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_sdma_fence_ring_emit, + .emit_semaphore = &cik_sdma_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_sdma_ring_test, + .ib_test = &cik_sdma_ib_test, + .is_lockup = &cik_sdma_is_lockup, + .vm_flush = &cik_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [R600_RING_TYPE_UVD_INDEX] = { + .ib_execute = &r600_uvd_ib_execute, + .emit_fence = &r600_uvd_fence_emit, + .emit_semaphore = &cayman_uvd_semaphore_emit, + .cs_parse = &radeon_uvd_cs_parse, + .ring_test = &r600_uvd_ring_test, + .ib_test = &r600_uvd_ib_test, + .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + } + }, + .irq = { + .set = &cik_irq_set, + .process = &cik_irq_process, + }, + .display = { + .bandwidth_update = &dce8_bandwidth_update, + .get_vblank_counter = &evergreen_get_vblank_counter, + .wait_for_vblank = &dce4_wait_for_vblank, + }, + .copy = { + .blit = NULL, + .blit_ring_index = RADEON_RING_TYPE_GFX_INDEX, + .dma = &cik_copy_dma, + .dma_ring_index = R600_RING_TYPE_DMA_INDEX, + .copy = &cik_copy_dma, + .copy_ring_index = R600_RING_TYPE_DMA_INDEX, + }, + .surface = { + .set_reg = r600_set_surface_reg, + .clear_reg = r600_clear_surface_reg, + }, + .hpd = { + .init = &evergreen_hpd_init, + .fini = &evergreen_hpd_fini, + .sense = &evergreen_hpd_sense, + .set_polarity = &evergreen_hpd_set_polarity, + }, + .pm = { + .misc = &evergreen_pm_misc, + .prepare = &evergreen_pm_prepare, + .finish = &evergreen_pm_finish, + .init_profile = &sumo_pm_init_profile, + .get_dynpm_state = &r600_pm_get_dynpm_state, + .get_engine_clock = &radeon_atom_get_engine_clock, + .set_engine_clock = &radeon_atom_set_engine_clock, + .get_memory_clock = &radeon_atom_get_memory_clock, + .set_memory_clock = &radeon_atom_set_memory_clock, + .get_pcie_lanes = NULL, + .set_pcie_lanes = NULL, + .set_clock_gating = NULL, + .set_uvd_clocks = &cik_set_uvd_clocks, + }, + .pflip = { + .pre_page_flip = &evergreen_pre_page_flip, + .page_flip = &evergreen_page_flip, + .post_page_flip = &evergreen_post_page_flip, + }, +}; + +static struct radeon_asic kv_asic = { + .init = &cik_init, + .fini = &cik_fini, + .suspend = &cik_suspend, + .resume = &cik_resume, + .asic_reset = &cik_asic_reset, + .vga_set_state = &r600_vga_set_state, + .ioctl_wait_idle = NULL, + .gui_idle = &r600_gui_idle, + .mc_wait_for_idle = &evergreen_mc_wait_for_idle, + .get_xclk = &cik_get_xclk, + .get_gpu_clock_counter = &cik_get_gpu_clock_counter, + .gart = { + .tlb_flush = &cik_pcie_gart_tlb_flush, + .set_page = &rs600_gart_set_page, + }, + .vm = { + .init = &cik_vm_init, + .fini = &cik_vm_fini, + .pt_ring_index = R600_RING_TYPE_DMA_INDEX, + .set_page = &cik_vm_set_page, + }, + .ring = { + [RADEON_RING_TYPE_GFX_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_gfx_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [CAYMAN_RING_TYPE_CP1_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_compute_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &cik_compute_ring_get_rptr, + .get_wptr = &cik_compute_ring_get_wptr, + .set_wptr = &cik_compute_ring_set_wptr, + }, + [CAYMAN_RING_TYPE_CP2_INDEX] = { + .ib_execute = &cik_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_fence_compute_ring_emit, + .emit_semaphore = &cik_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_ring_test, + .ib_test = &cik_ib_test, + .is_lockup = &cik_gfx_is_lockup, + .vm_flush = &cik_vm_flush, + .get_rptr = &cik_compute_ring_get_rptr, + .get_wptr = &cik_compute_ring_get_wptr, + .set_wptr = &cik_compute_ring_set_wptr, + }, + [R600_RING_TYPE_DMA_INDEX] = { + .ib_execute = &cik_sdma_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_sdma_fence_ring_emit, + .emit_semaphore = &cik_sdma_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_sdma_ring_test, + .ib_test = &cik_sdma_ib_test, + .is_lockup = &cik_sdma_is_lockup, + .vm_flush = &cik_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [CAYMAN_RING_TYPE_DMA1_INDEX] = { + .ib_execute = &cik_sdma_ring_ib_execute, + .ib_parse = &cik_ib_parse, + .emit_fence = &cik_sdma_fence_ring_emit, + .emit_semaphore = &cik_sdma_semaphore_ring_emit, + .cs_parse = NULL, + .ring_test = &cik_sdma_ring_test, + .ib_test = &cik_sdma_ib_test, + .is_lockup = &cik_sdma_is_lockup, + .vm_flush = &cik_dma_vm_flush, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + }, + [R600_RING_TYPE_UVD_INDEX] = { + .ib_execute = &r600_uvd_ib_execute, + .emit_fence = &r600_uvd_fence_emit, + .emit_semaphore = &cayman_uvd_semaphore_emit, + .cs_parse = &radeon_uvd_cs_parse, + .ring_test = &r600_uvd_ring_test, + .ib_test = &r600_uvd_ib_test, + .is_lockup = &radeon_ring_test_lockup, + .get_rptr = &radeon_ring_generic_get_rptr, + .get_wptr = &radeon_ring_generic_get_wptr, + .set_wptr = &radeon_ring_generic_set_wptr, + } + }, + .irq = { + .set = &cik_irq_set, + .process = &cik_irq_process, + }, + .display = { + .bandwidth_update = &dce8_bandwidth_update, + .get_vblank_counter = &evergreen_get_vblank_counter, + .wait_for_vblank = &dce4_wait_for_vblank, + }, + .copy = { + .blit = NULL, + .blit_ring_index = RADEON_RING_TYPE_GFX_INDEX, + .dma = &cik_copy_dma, + .dma_ring_index = R600_RING_TYPE_DMA_INDEX, + .copy = &cik_copy_dma, + .copy_ring_index = R600_RING_TYPE_DMA_INDEX, + }, + .surface = { + .set_reg = r600_set_surface_reg, + .clear_reg = r600_clear_surface_reg, + }, + .hpd = { + .init = &evergreen_hpd_init, + .fini = &evergreen_hpd_fini, + .sense = &evergreen_hpd_sense, + .set_polarity = &evergreen_hpd_set_polarity, + }, + .pm = { + .misc = &evergreen_pm_misc, + .prepare = &evergreen_pm_prepare, + .finish = &evergreen_pm_finish, + .init_profile = &sumo_pm_init_profile, + .get_dynpm_state = &r600_pm_get_dynpm_state, + .get_engine_clock = &radeon_atom_get_engine_clock, + .set_engine_clock = &radeon_atom_set_engine_clock, + .get_memory_clock = &radeon_atom_get_memory_clock, + .set_memory_clock = &radeon_atom_set_memory_clock, + .get_pcie_lanes = NULL, + .set_pcie_lanes = NULL, + .set_clock_gating = NULL, + .set_uvd_clocks = &cik_set_uvd_clocks, }, .pflip = { .pre_page_flip = &evergreen_pre_page_flip, @@ -1996,16 +2694,15 @@ int radeon_asic_init(struct radeon_device *rdev) rdev->asic = &r520_asic; break; case CHIP_R600: + rdev->asic = &r600_asic; + break; case CHIP_RV610: case CHIP_RV630: case CHIP_RV620: case CHIP_RV635: case CHIP_RV670: - rdev->asic = &r600_asic; - if (rdev->family == CHIP_R600) - rdev->has_uvd = false; - else - rdev->has_uvd = true; + rdev->asic = &rv6xx_asic; + rdev->has_uvd = true; break; case CHIP_RS780: case CHIP_RS880: @@ -2079,6 +2776,19 @@ int radeon_asic_init(struct radeon_device *rdev) else rdev->has_uvd = true; break; + case CHIP_BONAIRE: + rdev->asic = &ci_asic; + rdev->num_crtc = 6; + break; + case CHIP_KAVERI: + case CHIP_KABINI: + rdev->asic = &kv_asic; + /* set num crtcs */ + if (rdev->family == CHIP_KAVERI) + rdev->num_crtc = 4; + else + rdev->num_crtc = 2; + break; default: /* FIXME: not supported yet */ return -EINVAL; diff --git a/sys/dev/drm/radeon/radeon_asic.h b/sys/dev/drm/radeon/radeon_asic.h index 7595218012..63ac489b46 100644 --- a/sys/dev/drm/radeon/radeon_asic.h +++ b/sys/dev/drm/radeon/radeon_asic.h @@ -50,7 +50,6 @@ u8 atombios_get_backlight_level(struct radeon_encoder *radeon_encoder); void radeon_legacy_set_backlight_level(struct radeon_encoder *radeon_encoder, u8 level); u8 radeon_legacy_get_backlight_level(struct radeon_encoder *radeon_encoder); - /* * r100,rv100,rs100,rv200,rs200 */ @@ -336,6 +335,9 @@ int r600_uvd_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); int r600_copy_blit(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, struct radeon_fence **fence); +int r600_copy_cpdma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, struct radeon_fence **fence); int r600_copy_dma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, struct radeon_fence **fence); @@ -398,11 +400,44 @@ void r600_kms_blit_copy(struct radeon_device *rdev, struct radeon_sa_bo *vb); u32 r600_get_xclk(struct radeon_device *rdev); uint64_t r600_get_gpu_clock_counter(struct radeon_device *rdev); +int rv6xx_get_temp(struct radeon_device *rdev); +int r600_dpm_pre_set_power_state(struct radeon_device *rdev); +void r600_dpm_post_set_power_state(struct radeon_device *rdev); +/* rv6xx dpm */ +int rv6xx_dpm_init(struct radeon_device *rdev); +int rv6xx_dpm_enable(struct radeon_device *rdev); +void rv6xx_dpm_disable(struct radeon_device *rdev); +int rv6xx_dpm_set_power_state(struct radeon_device *rdev); +void rv6xx_setup_asic(struct radeon_device *rdev); +void rv6xx_dpm_display_configuration_changed(struct radeon_device *rdev); +void rv6xx_dpm_fini(struct radeon_device *rdev); +u32 rv6xx_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 rv6xx_dpm_get_mclk(struct radeon_device *rdev, bool low); +void rv6xx_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void rv6xx_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +int rv6xx_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); +/* rs780 dpm */ +int rs780_dpm_init(struct radeon_device *rdev); +int rs780_dpm_enable(struct radeon_device *rdev); +void rs780_dpm_disable(struct radeon_device *rdev); +int rs780_dpm_set_power_state(struct radeon_device *rdev); +void rs780_dpm_setup_asic(struct radeon_device *rdev); +void rs780_dpm_display_configuration_changed(struct radeon_device *rdev); +void rs780_dpm_fini(struct radeon_device *rdev); +u32 rs780_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 rs780_dpm_get_mclk(struct radeon_device *rdev, bool low); +void rs780_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); /* uvd */ int r600_uvd_init(struct radeon_device *rdev); int r600_uvd_rbc_start(struct radeon_device *rdev); -void r600_uvd_rbc_stop(struct radeon_device *rdev); +void r600_uvd_stop(struct radeon_device *rdev); int r600_uvd_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); void r600_uvd_fence_emit(struct radeon_device *rdev, struct radeon_fence *fence); @@ -431,6 +466,22 @@ int rv770_copy_dma(struct radeon_device *rdev, u32 rv770_get_xclk(struct radeon_device *rdev); int rv770_uvd_resume(struct radeon_device *rdev); int rv770_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); +int rv770_get_temp(struct radeon_device *rdev); +/* rv7xx pm */ +int rv770_dpm_init(struct radeon_device *rdev); +int rv770_dpm_enable(struct radeon_device *rdev); +void rv770_dpm_disable(struct radeon_device *rdev); +int rv770_dpm_set_power_state(struct radeon_device *rdev); +void rv770_dpm_setup_asic(struct radeon_device *rdev); +void rv770_dpm_display_configuration_changed(struct radeon_device *rdev); +void rv770_dpm_fini(struct radeon_device *rdev); +u32 rv770_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 rv770_dpm_get_mclk(struct radeon_device *rdev, bool low); +void rv770_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void rv770_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +bool rv770_dpm_vblank_too_short(struct radeon_device *rdev); /* * evergreen @@ -483,15 +534,58 @@ int evergreen_copy_dma(struct radeon_device *rdev, uint64_t src_offset, uint64_t dst_offset, unsigned num_gpu_pages, struct radeon_fence **fence); -void evergreen_hdmi_enable(struct drm_encoder *encoder, bool enable); -void evergreen_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mode); void evergreen_fix_pci_max_read_req_size(struct radeon_device *rdev); u32 evergreen_get_number_of_dram_channels(struct radeon_device *rdev); +void sumo_rlc_fini(struct radeon_device *rdev); +int sumo_rlc_init(struct radeon_device *rdev); +int evergreen_rlc_resume(struct radeon_device *rdev); void evergreen_mc_stop(struct radeon_device *rdev, struct evergreen_mc_save *save); void evergreen_mc_resume(struct radeon_device *rdev, struct evergreen_mc_save *save); void evergreen_mc_program(struct radeon_device *rdev); int evergreen_mc_init(struct radeon_device *rdev); void evergreen_irq_suspend(struct radeon_device *rdev); +bool evergreen_is_display_hung(struct radeon_device *rdev); +void evergreen_hdmi_enable(struct drm_encoder *encoder, bool enable); +void evergreen_hdmi_setmode(struct drm_encoder *encoder, struct drm_display_mode *mode); +int evergreen_get_temp(struct radeon_device *rdev); +int sumo_get_temp(struct radeon_device *rdev); +int tn_get_temp(struct radeon_device *rdev); +int cypress_dpm_init(struct radeon_device *rdev); +void cypress_dpm_setup_asic(struct radeon_device *rdev); +int cypress_dpm_enable(struct radeon_device *rdev); +void cypress_dpm_disable(struct radeon_device *rdev); +int cypress_dpm_set_power_state(struct radeon_device *rdev); +void cypress_dpm_display_configuration_changed(struct radeon_device *rdev); +void cypress_dpm_fini(struct radeon_device *rdev); +bool cypress_dpm_vblank_too_short(struct radeon_device *rdev); +int btc_dpm_init(struct radeon_device *rdev); +void btc_dpm_setup_asic(struct radeon_device *rdev); +int btc_dpm_enable(struct radeon_device *rdev); +void btc_dpm_disable(struct radeon_device *rdev); +int btc_dpm_pre_set_power_state(struct radeon_device *rdev); +int btc_dpm_set_power_state(struct radeon_device *rdev); +void btc_dpm_post_set_power_state(struct radeon_device *rdev); +void btc_dpm_fini(struct radeon_device *rdev); +u32 btc_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 btc_dpm_get_mclk(struct radeon_device *rdev, bool low); +bool btc_dpm_vblank_too_short(struct radeon_device *rdev); +int sumo_dpm_init(struct radeon_device *rdev); +int sumo_dpm_enable(struct radeon_device *rdev); +void sumo_dpm_disable(struct radeon_device *rdev); +int sumo_dpm_pre_set_power_state(struct radeon_device *rdev); +int sumo_dpm_set_power_state(struct radeon_device *rdev); +void sumo_dpm_post_set_power_state(struct radeon_device *rdev); +void sumo_dpm_setup_asic(struct radeon_device *rdev); +void sumo_dpm_display_configuration_changed(struct radeon_device *rdev); +void sumo_dpm_fini(struct radeon_device *rdev); +u32 sumo_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 sumo_dpm_get_mclk(struct radeon_device *rdev, bool low); +void sumo_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void sumo_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +int sumo_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); /* * cayman @@ -518,6 +612,8 @@ void cayman_vm_set_page(struct radeon_device *rdev, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint32_t flags); +void cayman_vm_decode_fault(struct radeon_device *rdev, + u32 status, u32 addr); int evergreen_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); int evergreen_dma_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); void cayman_dma_ring_ib_execute(struct radeon_device *rdev, @@ -528,6 +624,41 @@ void cayman_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm extern void cayman_cp_int_cntl_setup(struct radeon_device *rdev, int ring, u32 cp_int_cntl); +int ni_dpm_init(struct radeon_device *rdev); +void ni_dpm_setup_asic(struct radeon_device *rdev); +int ni_dpm_enable(struct radeon_device *rdev); +void ni_dpm_disable(struct radeon_device *rdev); +int ni_dpm_pre_set_power_state(struct radeon_device *rdev); +int ni_dpm_set_power_state(struct radeon_device *rdev); +void ni_dpm_post_set_power_state(struct radeon_device *rdev); +void ni_dpm_fini(struct radeon_device *rdev); +u32 ni_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 ni_dpm_get_mclk(struct radeon_device *rdev, bool low); +void ni_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void ni_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +int ni_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); +//bool ni_dpm_vblank_too_short(struct radeon_device *rdev); +int trinity_dpm_init(struct radeon_device *rdev); +int trinity_dpm_enable(struct radeon_device *rdev); +void trinity_dpm_disable(struct radeon_device *rdev); +int trinity_dpm_pre_set_power_state(struct radeon_device *rdev); +int trinity_dpm_set_power_state(struct radeon_device *rdev); +void trinity_dpm_post_set_power_state(struct radeon_device *rdev); +void trinity_dpm_setup_asic(struct radeon_device *rdev); +void trinity_dpm_display_configuration_changed(struct radeon_device *rdev); +void trinity_dpm_fini(struct radeon_device *rdev); +u32 trinity_dpm_get_sclk(struct radeon_device *rdev, bool low); +u32 trinity_dpm_get_mclk(struct radeon_device *rdev, bool low); +void trinity_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *ps); +void trinity_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +int trinity_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); + /* DCE6 - SI */ void dce6_bandwidth_update(struct radeon_device *rdev); @@ -564,7 +695,88 @@ void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) u32 si_get_xclk(struct radeon_device *rdev); uint64_t si_get_gpu_clock_counter(struct radeon_device *rdev); int si_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); +int si_get_temp(struct radeon_device *rdev); void si_rlc_fini(struct radeon_device *rdev); int si_rlc_init(struct radeon_device *rdev); +void si_vram_gtt_location(struct radeon_device *rdev, + struct radeon_mc *mc); +int si_dpm_init(struct radeon_device *rdev); +void si_dpm_setup_asic(struct radeon_device *rdev); +int si_dpm_enable(struct radeon_device *rdev); +void si_dpm_disable(struct radeon_device *rdev); +int si_dpm_pre_set_power_state(struct radeon_device *rdev); +int si_dpm_set_power_state(struct radeon_device *rdev); +void si_dpm_post_set_power_state(struct radeon_device *rdev); +void si_dpm_fini(struct radeon_device *rdev); +void si_dpm_display_configuration_changed(struct radeon_device *rdev); +void si_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m); +int si_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); +/* DCE8 - CIK */ +void dce8_bandwidth_update(struct radeon_device *rdev); + +/* + * cik + */ +uint64_t cik_get_gpu_clock_counter(struct radeon_device *rdev); +u32 cik_get_xclk(struct radeon_device *rdev); +uint32_t cik_pciep_rreg(struct radeon_device *rdev, uint32_t reg); +void cik_pciep_wreg(struct radeon_device *rdev, uint32_t reg, uint32_t v); +int cik_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk); +int cik_uvd_resume(struct radeon_device *rdev); +void cik_sdma_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence); +void cik_sdma_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *ring, + struct radeon_semaphore *semaphore, + bool emit_wait); +void cik_sdma_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); +int cik_copy_dma(struct radeon_device *rdev, + uint64_t src_offset, uint64_t dst_offset, + unsigned num_gpu_pages, + struct radeon_fence **fence); +int cik_sdma_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); +int cik_sdma_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); +bool cik_sdma_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring); +void cik_fence_gfx_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence); +void cik_fence_compute_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence); +void cik_semaphore_ring_emit(struct radeon_device *rdev, + struct radeon_ring *cp, + struct radeon_semaphore *semaphore, + bool emit_wait); +void cik_pcie_gart_tlb_flush(struct radeon_device *rdev); +int cik_init(struct radeon_device *rdev); +void cik_fini(struct radeon_device *rdev); +int cik_suspend(struct radeon_device *rdev); +int cik_resume(struct radeon_device *rdev); +bool cik_gfx_is_lockup(struct radeon_device *rdev, struct radeon_ring *cp); +int cik_asic_reset(struct radeon_device *rdev); +void cik_ring_ib_execute(struct radeon_device *rdev, struct radeon_ib *ib); +int cik_ring_test(struct radeon_device *rdev, struct radeon_ring *ring); +int cik_ib_test(struct radeon_device *rdev, struct radeon_ring *ring); +int cik_irq_set(struct radeon_device *rdev); +irqreturn_t cik_irq_process(struct radeon_device *rdev); +int cik_vm_init(struct radeon_device *rdev); +void cik_vm_fini(struct radeon_device *rdev); +void cik_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +void cik_vm_set_page(struct radeon_device *rdev, + struct radeon_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint32_t flags); +void cik_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm); +int cik_ib_parse(struct radeon_device *rdev, struct radeon_ib *ib); +u32 cik_compute_ring_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring); +u32 cik_compute_ring_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +void cik_compute_ring_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring); +bool cik_gpu_is_lockup(struct radeon_device *rdev, struct radeon_ring *ring); +void cik_fence_ring_emit(struct radeon_device *rdev, + struct radeon_fence *fence); #endif diff --git a/sys/dev/drm/radeon/radeon_atombios.c b/sys/dev/drm/radeon/radeon_atombios.c index 3f4aa37628..2a74a0cc19 100644 --- a/sys/dev/drm/radeon/radeon_atombios.c +++ b/sys/dev/drm/radeon/radeon_atombios.c @@ -34,10 +34,6 @@ #include "atom.h" #include "atom-bits.h" -/* local */ -static int radeon_atom_get_max_vddc(struct radeon_device *rdev, u8 voltage_type, - u16 voltage_id, u16 *voltage); - union atom_supported_devices { struct _ATOM_SUPPORTED_DEVICES_INFO info; struct _ATOM_SUPPORTED_DEVICES_INFO_2 info_2; @@ -1225,6 +1221,7 @@ bool radeon_atom_get_clock_info(struct drm_device *dev) } rdev->clock.dp_extclk = le16_to_cpu(firmware_info->info_21.usUniphyDPModeExtClkFreq); + rdev->clock.current_dispclk = rdev->clock.default_dispclk; } *dcpll = *p1pll; @@ -1247,6 +1244,7 @@ union igp_info { struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2; struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 info_6; struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 info_7; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_8 info_8; }; bool radeon_atombios_sideport_present(struct radeon_device *rdev) @@ -1416,6 +1414,22 @@ static void radeon_atombios_get_igp_ss_overrides(struct radeon_device *rdev, break; } break; + case 8: + switch (id) { + case ASIC_INTERNAL_SS_ON_TMDS: + percentage = le16_to_cpu(igp_info->info_8.usDVISSPercentage); + rate = le16_to_cpu(igp_info->info_8.usDVISSpreadRateIn10Hz); + break; + case ASIC_INTERNAL_SS_ON_HDMI: + percentage = le16_to_cpu(igp_info->info_8.usHDMISSPercentage); + rate = le16_to_cpu(igp_info->info_8.usHDMISSpreadRateIn10Hz); + break; + case ASIC_INTERNAL_SS_ON_LVDS: + percentage = le16_to_cpu(igp_info->info_8.usLvdsSSPercentage); + rate = le16_to_cpu(igp_info->info_8.usLvdsSSpreadRateIn10Hz); + break; + } + break; default: DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev); break; @@ -1477,6 +1491,10 @@ bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev, le16_to_cpu(ss_info->info_2.asSpreadSpectrum[i].usSpreadSpectrumPercentage); ss->type = ss_info->info_2.asSpreadSpectrum[i].ucSpreadSpectrumMode; ss->rate = le16_to_cpu(ss_info->info_2.asSpreadSpectrum[i].usSpreadRateIn10Hz); + if ((crev == 2) && + ((id == ASIC_INTERNAL_ENGINE_SS) || + (id == ASIC_INTERNAL_MEMORY_SS))) + ss->rate /= 100; return true; } } @@ -1491,6 +1509,9 @@ bool radeon_atombios_get_asic_ss_info(struct radeon_device *rdev, le16_to_cpu(ss_info->info_3.asSpreadSpectrum[i].usSpreadSpectrumPercentage); ss->type = ss_info->info_3.asSpreadSpectrum[i].ucSpreadSpectrumMode; ss->rate = le16_to_cpu(ss_info->info_3.asSpreadSpectrum[i].usSpreadRateIn10Hz); + if ((id == ASIC_INTERNAL_ENGINE_SS) || + (id == ASIC_INTERNAL_MEMORY_SS)) + ss->rate /= 100; if (rdev->flags & RADEON_IS_IGP) radeon_atombios_get_igp_ss_overrides(rdev, ss, id); return true; @@ -1909,6 +1930,7 @@ static const char *pp_lib_thermal_controller_names[] = { "Northern Islands", "Southern Islands", "lm96163", + "Sea Islands", }; union power_info { @@ -1926,6 +1948,7 @@ union pplib_clock_info { struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; struct _ATOM_PPLIB_SI_CLOCK_INFO si; + struct _ATOM_PPLIB_CI_CLOCK_INFO ci; }; union pplib_power_state { @@ -2195,6 +2218,11 @@ static void radeon_atombios_add_pplib_thermal_controller(struct radeon_device *r (controller->ucFanParameters & ATOM_PP_FANPARAMETERS_NOFAN) ? "without" : "with"); rdev->pm.int_thermal_type = THERMAL_TYPE_SI; + } else if (controller->ucType == ATOM_PP_THERMALCONTROLLER_CISLANDS) { + DRM_INFO("Internal thermal controller %s fan control\n", + (controller->ucFanParameters & + ATOM_PP_FANPARAMETERS_NOFAN) ? "without" : "with"); + rdev->pm.int_thermal_type = THERMAL_TYPE_CI; } else if ((controller->ucType == ATOM_PP_THERMALCONTROLLER_EXTERNAL_GPIO) || (controller->ucType == @@ -2229,8 +2257,8 @@ static void radeon_atombios_add_pplib_thermal_controller(struct radeon_device *r } } -static void radeon_atombios_get_default_voltages(struct radeon_device *rdev, - u16 *vddc, u16 *vddci) +void radeon_atombios_get_default_voltages(struct radeon_device *rdev, + u16 *vddc, u16 *vddci, u16 *mvdd) { struct radeon_mode_info *mode_info = &rdev->mode_info; int index = GetIndexIntoMasterTable(DATA, FirmwareInfo); @@ -2240,6 +2268,7 @@ static void radeon_atombios_get_default_voltages(struct radeon_device *rdev, *vddc = 0; *vddci = 0; + *mvdd = 0; if (atom_parse_data_header(mode_info->atom_context, index, NULL, &frev, &crev, &data_offset)) { @@ -2247,8 +2276,10 @@ static void radeon_atombios_get_default_voltages(struct radeon_device *rdev, (union firmware_info *)((char *)mode_info->atom_context->bios + data_offset); *vddc = le16_to_cpu(firmware_info->info_14.usBootUpVDDCVoltage); - if ((frev == 2) && (crev >= 2)) + if ((frev == 2) && (crev >= 2)) { *vddci = le16_to_cpu(firmware_info->info_22.usBootUpVDDCIVoltage); + *mvdd = le16_to_cpu(firmware_info->info_22.usBootUpMVDDCVoltage); + } } } @@ -2259,9 +2290,9 @@ static void radeon_atombios_parse_pplib_non_clock_info(struct radeon_device *rde int j; u32 misc = le32_to_cpu(non_clock_info->ulCapsAndSettings); u32 misc2 = le16_to_cpu(non_clock_info->usClassification); - u16 vddc, vddci; + u16 vddc, vddci, mvdd; - radeon_atombios_get_default_voltages(rdev, &vddc, &vddci); + radeon_atombios_get_default_voltages(rdev, &vddc, &vddci, &mvdd); rdev->pm.power_state[state_index].misc = misc; rdev->pm.power_state[state_index].misc2 = misc2; @@ -2304,7 +2335,13 @@ static void radeon_atombios_parse_pplib_non_clock_info(struct radeon_device *rde rdev->pm.default_vddc = rdev->pm.power_state[state_index].clock_info[0].voltage.voltage; rdev->pm.default_vddci = rdev->pm.power_state[state_index].clock_info[0].voltage.vddci; } else { - /* patch the table values with the default slck/mclk from firmware info */ + u16 max_vddci = 0; + + if (ASIC_IS_DCE4(rdev)) + radeon_atom_get_max_voltage(rdev, + SET_VOLTAGE_TYPE_ASIC_VDDCI, + &max_vddci); + /* patch the table values with the default sclk/mclk from firmware info */ for (j = 0; j < mode_index; j++) { rdev->pm.power_state[state_index].clock_info[j].mclk = rdev->clock.default_mclk; @@ -2313,6 +2350,9 @@ static void radeon_atombios_parse_pplib_non_clock_info(struct radeon_device *rde if (vddc) rdev->pm.power_state[state_index].clock_info[j].voltage.voltage = vddc; + if (max_vddci) + rdev->pm.power_state[state_index].clock_info[j].voltage.vddci = + max_vddci; } } } @@ -2335,6 +2375,15 @@ static bool radeon_atombios_parse_pplib_clock_info(struct radeon_device *rdev, sclk |= clock_info->rs780.ucLowEngineClockHigh << 16; rdev->pm.power_state[state_index].clock_info[mode_index].sclk = sclk; } + } else if (rdev->family >= CHIP_BONAIRE) { + sclk = le16_to_cpu(clock_info->ci.usEngineClockLow); + sclk |= clock_info->ci.ucEngineClockHigh << 16; + mclk = le16_to_cpu(clock_info->ci.usMemoryClockLow); + mclk |= clock_info->ci.ucMemoryClockHigh << 16; + rdev->pm.power_state[state_index].clock_info[mode_index].mclk = mclk; + rdev->pm.power_state[state_index].clock_info[mode_index].sclk = sclk; + rdev->pm.power_state[state_index].clock_info[mode_index].voltage.type = + VOLTAGE_NONE; } else if (rdev->family >= CHIP_TAHITI) { sclk = le16_to_cpu(clock_info->si.usEngineClockLow); sclk |= clock_info->si.ucEngineClockHigh << 16; @@ -2380,6 +2429,10 @@ static bool radeon_atombios_parse_pplib_clock_info(struct radeon_device *rdev, case ATOM_VIRTUAL_VOLTAGE_ID1: case ATOM_VIRTUAL_VOLTAGE_ID2: case ATOM_VIRTUAL_VOLTAGE_ID3: + case ATOM_VIRTUAL_VOLTAGE_ID4: + case ATOM_VIRTUAL_VOLTAGE_ID5: + case ATOM_VIRTUAL_VOLTAGE_ID6: + case ATOM_VIRTUAL_VOLTAGE_ID7: if (radeon_atom_get_max_vddc(rdev, VOLTAGE_TYPE_VDDC, rdev->pm.power_state[state_index].clock_info[mode_index].voltage.voltage, &vddc) == 0) @@ -2659,6 +2712,8 @@ union get_clock_dividers { struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V3 v3; struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4 v4; struct _COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V5 v5; + struct _COMPUTE_GPU_CLOCK_INPUT_PARAMETERS_V1_6 v6_in; + struct _COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_6 v6_out; }; int radeon_atom_get_clock_dividers(struct radeon_device *rdev, @@ -2691,7 +2746,8 @@ int radeon_atom_get_clock_dividers(struct radeon_device *rdev, break; case 2: case 3: - /* r6xx, r7xx, evergreen, ni */ + case 5: + /* r6xx, r7xx, evergreen, ni, si */ if (rdev->family <= CHIP_RV770) { args.v2.ucAction = clock_type; args.v2.ulClock = cpu_to_le32(clock); /* 10 khz */ @@ -2718,12 +2774,15 @@ int radeon_atom_get_clock_dividers(struct radeon_device *rdev, ATOM_PLL_CNTL_FLAG_PLL_POST_DIV_EN) ? true : false; dividers->enable_dithen = (args.v3.ucCntlFlag & ATOM_PLL_CNTL_FLAG_FRACTION_DISABLE) ? false : true; - dividers->fb_div = le16_to_cpu(args.v3.ulFbDiv.usFbDiv); + dividers->whole_fb_div = le16_to_cpu(args.v3.ulFbDiv.usFbDiv); dividers->frac_fb_div = le16_to_cpu(args.v3.ulFbDiv.usFbDivFrac); dividers->ref_div = args.v3.ucRefDiv; dividers->vco_mode = (args.v3.ucCntlFlag & ATOM_PLL_CNTL_FLAG_MPLL_VCO_MODE) ? 1 : 0; } else { + /* for SI we use ComputeMemoryClockParam for memory plls */ + if (rdev->family >= CHIP_TAHITI) + return -EINVAL; args.v5.ulClockParams = cpu_to_le32((clock_type << 24) | clock); if (strobe_mode) args.v5.ucInputFlag = ATOM_PLL_INPUT_FLAG_PLL_STROBE_MODE_EN; @@ -2749,9 +2808,76 @@ int radeon_atom_get_clock_dividers(struct radeon_device *rdev, atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); - dividers->post_div = args.v4.ucPostDiv; + dividers->post_divider = dividers->post_div = args.v4.ucPostDiv; dividers->real_clock = le32_to_cpu(args.v4.ulClock); break; + case 6: + /* CI */ + /* COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, COMPUTE_GPUCLK_INPUT_FLAG_SCLK */ + args.v6_in.ulClock.ulComputeClockFlag = clock_type; + args.v6_in.ulClock.ulClockFreq = cpu_to_le32(clock); /* 10 khz */ + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + + dividers->whole_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDiv); + dividers->frac_fb_div = le16_to_cpu(args.v6_out.ulFbDiv.usFbDivFrac); + dividers->ref_div = args.v6_out.ucPllRefDiv; + dividers->post_div = args.v6_out.ucPllPostDiv; + dividers->flags = args.v6_out.ucPllCntlFlag; + dividers->real_clock = le32_to_cpu(args.v6_out.ulClock.ulClock); + dividers->post_divider = args.v6_out.ulClock.ucPostDiv; + break; + default: + return -EINVAL; + } + return 0; +} + +int radeon_atom_get_memory_pll_dividers(struct radeon_device *rdev, + u32 clock, + bool strobe_mode, + struct atom_mpll_param *mpll_param) +{ + COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_1 args; + int index = GetIndexIntoMasterTable(COMMAND, ComputeMemoryClockParam); + u8 frev, crev; + + memset(&args, 0, sizeof(args)); + memset(mpll_param, 0, sizeof(struct atom_mpll_param)); + + if (!atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, &crev)) + return -EINVAL; + + switch (frev) { + case 2: + switch (crev) { + case 1: + /* SI */ + args.ulClock = cpu_to_le32(clock); /* 10 khz */ + args.ucInputFlag = 0; + if (strobe_mode) + args.ucInputFlag |= MPLL_INPUT_FLAG_STROBE_MODE_EN; + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + + mpll_param->clkfrac = le16_to_cpu(args.ulFbDiv.usFbDivFrac); + mpll_param->clkf = le16_to_cpu(args.ulFbDiv.usFbDiv); + mpll_param->post_div = args.ucPostDiv; + mpll_param->dll_speed = args.ucDllSpeed; + mpll_param->bwcntl = args.ucBWCntl; + mpll_param->vco_mode = + (args.ucPllCntlFlag & MPLL_CNTL_FLAG_VCO_MODE_MASK) ? 1 : 0; + mpll_param->yclk_sel = + (args.ucPllCntlFlag & MPLL_CNTL_FLAG_BYPASS_DQ_PLL) ? 1 : 0; + mpll_param->qdr = + (args.ucPllCntlFlag & MPLL_CNTL_FLAG_QDR_ENABLE) ? 1 : 0; + mpll_param->half_rate = + (args.ucPllCntlFlag & MPLL_CNTL_FLAG_AD_HALF_RATE) ? 1 : 0; + break; + default: + return -EINVAL; + } + break; default: return -EINVAL; } @@ -2811,6 +2937,48 @@ void radeon_atom_set_memory_clock(struct radeon_device *rdev, atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); } +void radeon_atom_set_engine_dram_timings(struct radeon_device *rdev, + u32 eng_clock, u32 mem_clock) +{ + SET_ENGINE_CLOCK_PS_ALLOCATION args; + int index = GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings); + u32 tmp; + + memset(&args, 0, sizeof(args)); + + tmp = eng_clock & SET_CLOCK_FREQ_MASK; + tmp |= (COMPUTE_ENGINE_PLL_PARAM << 24); + + args.ulTargetEngineClock = cpu_to_le32(tmp); + if (mem_clock) + args.sReserved.ulClock = cpu_to_le32(mem_clock & SET_CLOCK_FREQ_MASK); + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); +} + +void radeon_atom_update_memory_dll(struct radeon_device *rdev, + u32 mem_clock) +{ + u32 args; + int index = GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings); + + args = cpu_to_le32(mem_clock); /* 10 khz */ + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); +} + +void radeon_atom_set_ac_timing(struct radeon_device *rdev, + u32 mem_clock) +{ + SET_MEMORY_CLOCK_PS_ALLOCATION args; + int index = GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings); + u32 tmp = mem_clock | (COMPUTE_MEMORY_PLL_PARAM << 24); + + args.ulTargetMemoryClock = cpu_to_le32(tmp); /* 10 khz */ + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); +} + union set_voltage { struct _SET_VOLTAGE_PS_ALLOCATION alloc; struct _SET_VOLTAGE_PARAMETERS v1; @@ -2855,8 +3023,8 @@ void radeon_atom_set_voltage(struct radeon_device *rdev, u16 voltage_level, u8 v atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); } -static int radeon_atom_get_max_vddc(struct radeon_device *rdev, u8 voltage_type, - u16 voltage_id, u16 *voltage) +int radeon_atom_get_max_vddc(struct radeon_device *rdev, u8 voltage_type, + u16 voltage_id, u16 *voltage) { union set_voltage args; int index = GetIndexIntoMasterTable(COMMAND, SetVoltage); @@ -2894,6 +3062,693 @@ static int radeon_atom_get_max_vddc(struct radeon_device *rdev, u8 voltage_type, return 0; } +int radeon_atom_get_leakage_vddc_based_on_leakage_idx(struct radeon_device *rdev, + u16 *voltage, + u16 leakage_idx) +{ + return radeon_atom_get_max_vddc(rdev, VOLTAGE_TYPE_VDDC, leakage_idx, voltage); +} + +int radeon_atom_get_voltage_gpio_settings(struct radeon_device *rdev, + u16 voltage_level, u8 voltage_type, + u32 *gpio_value, u32 *gpio_mask) +{ + union set_voltage args; + int index = GetIndexIntoMasterTable(COMMAND, SetVoltage); + u8 frev, crev; + + if (!atom_parse_cmd_header(rdev->mode_info.atom_context, index, &frev, &crev)) + return -EINVAL; + + switch (crev) { + case 1: + return -EINVAL; + case 2: + args.v2.ucVoltageType = voltage_type; + args.v2.ucVoltageMode = SET_ASIC_VOLTAGE_MODE_GET_GPIOMASK; + args.v2.usVoltageLevel = cpu_to_le16(voltage_level); + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + + *gpio_mask = le32_to_cpu(*(u32 *)&args.v2); + + args.v2.ucVoltageType = voltage_type; + args.v2.ucVoltageMode = SET_ASIC_VOLTAGE_MODE_GET_GPIOVAL; + args.v2.usVoltageLevel = cpu_to_le16(voltage_level); + + atom_execute_table(rdev->mode_info.atom_context, index, (uint32_t *)&args); + + *gpio_value = le32_to_cpu(*(u32 *)&args.v2); + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + + return 0; +} + +union voltage_object_info { + struct _ATOM_VOLTAGE_OBJECT_INFO v1; + struct _ATOM_VOLTAGE_OBJECT_INFO_V2 v2; + struct _ATOM_VOLTAGE_OBJECT_INFO_V3_1 v3; +}; + +union voltage_object { + struct _ATOM_VOLTAGE_OBJECT v1; + struct _ATOM_VOLTAGE_OBJECT_V2 v2; + union _ATOM_VOLTAGE_OBJECT_V3 v3; +}; + +static ATOM_VOLTAGE_OBJECT *atom_lookup_voltage_object_v1(ATOM_VOLTAGE_OBJECT_INFO *v1, + u8 voltage_type) +{ + u32 size = le16_to_cpu(v1->sHeader.usStructureSize); + u32 offset = offsetof(ATOM_VOLTAGE_OBJECT_INFO, asVoltageObj[0]); + u8 *start = (u8 *)v1; + + while (offset < size) { + ATOM_VOLTAGE_OBJECT *vo = (ATOM_VOLTAGE_OBJECT *)(start + offset); + if (vo->ucVoltageType == voltage_type) + return vo; + offset += offsetof(ATOM_VOLTAGE_OBJECT, asFormula.ucVIDAdjustEntries) + + vo->asFormula.ucNumOfVoltageEntries; + } + return NULL; +} + +static ATOM_VOLTAGE_OBJECT_V2 *atom_lookup_voltage_object_v2(ATOM_VOLTAGE_OBJECT_INFO_V2 *v2, + u8 voltage_type) +{ + u32 size = le16_to_cpu(v2->sHeader.usStructureSize); + u32 offset = offsetof(ATOM_VOLTAGE_OBJECT_INFO_V2, asVoltageObj[0]); + u8 *start = (u8*)v2; + + while (offset < size) { + ATOM_VOLTAGE_OBJECT_V2 *vo = (ATOM_VOLTAGE_OBJECT_V2 *)(start + offset); + if (vo->ucVoltageType == voltage_type) + return vo; + offset += offsetof(ATOM_VOLTAGE_OBJECT_V2, asFormula.asVIDAdjustEntries) + + (vo->asFormula.ucNumOfVoltageEntries * sizeof(VOLTAGE_LUT_ENTRY)); + } + return NULL; +} + +static ATOM_VOLTAGE_OBJECT_V3 *atom_lookup_voltage_object_v3(ATOM_VOLTAGE_OBJECT_INFO_V3_1 *v3, + u8 voltage_type, u8 voltage_mode) +{ + u32 size = le16_to_cpu(v3->sHeader.usStructureSize); + u32 offset = offsetof(ATOM_VOLTAGE_OBJECT_INFO_V3_1, asVoltageObj[0]); + u8 *start = (u8*)v3; + + while (offset < size) { + ATOM_VOLTAGE_OBJECT_V3 *vo = (ATOM_VOLTAGE_OBJECT_V3 *)(start + offset); + if ((vo->asGpioVoltageObj.sHeader.ucVoltageType == voltage_type) && + (vo->asGpioVoltageObj.sHeader.ucVoltageMode == voltage_mode)) + return vo; + offset += le16_to_cpu(vo->asGpioVoltageObj.sHeader.usSize); + } + return NULL; +} + +bool +radeon_atom_is_voltage_gpio(struct radeon_device *rdev, + u8 voltage_type, u8 voltage_mode) +{ + int index = GetIndexIntoMasterTable(DATA, VoltageObjectInfo); + u8 frev, crev; + u16 data_offset, size; + union voltage_object_info *voltage_info; + union voltage_object *voltage_object = NULL; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + voltage_info = (union voltage_object_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + switch (frev) { + case 1: + case 2: + switch (crev) { + case 1: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v1(&voltage_info->v1, voltage_type); + if (voltage_object && + (voltage_object->v1.asControl.ucVoltageControlId == VOLTAGE_CONTROLLED_BY_GPIO)) + return true; + break; + case 2: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v2(&voltage_info->v2, voltage_type); + if (voltage_object && + (voltage_object->v2.asControl.ucVoltageControlId == VOLTAGE_CONTROLLED_BY_GPIO)) + return true; + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return false; + } + break; + case 3: + switch (crev) { + case 1: + if (atom_lookup_voltage_object_v3(&voltage_info->v3, + voltage_type, voltage_mode)) + return true; + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return false; + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return false; + } + + } + return false; +} + +int radeon_atom_get_max_voltage(struct radeon_device *rdev, + u8 voltage_type, u16 *max_voltage) +{ + int index = GetIndexIntoMasterTable(DATA, VoltageObjectInfo); + u8 frev, crev; + u16 data_offset, size; + union voltage_object_info *voltage_info; + union voltage_object *voltage_object = NULL; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + voltage_info = (union voltage_object_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + switch (crev) { + case 1: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v1(&voltage_info->v1, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA *formula = + &voltage_object->v1.asFormula; + if (formula->ucFlag & 1) + *max_voltage = + le16_to_cpu(formula->usVoltageBaseLevel) + + formula->ucNumOfVoltageEntries / 2 * + le16_to_cpu(formula->usVoltageStep); + else + *max_voltage = + le16_to_cpu(formula->usVoltageBaseLevel) + + (formula->ucNumOfVoltageEntries - 1) * + le16_to_cpu(formula->usVoltageStep); + return 0; + } + break; + case 2: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v2(&voltage_info->v2, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA_V2 *formula = + &voltage_object->v2.asFormula; + if (formula->ucNumOfVoltageEntries) { + *max_voltage = + le16_to_cpu(formula->asVIDAdjustEntries[ + formula->ucNumOfVoltageEntries - 1 + ].usVoltageValue); + return 0; + } + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + + } + return -EINVAL; +} + +int radeon_atom_get_min_voltage(struct radeon_device *rdev, + u8 voltage_type, u16 *min_voltage) +{ + int index = GetIndexIntoMasterTable(DATA, VoltageObjectInfo); + u8 frev, crev; + u16 data_offset, size; + union voltage_object_info *voltage_info; + union voltage_object *voltage_object = NULL; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + voltage_info = (union voltage_object_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + switch (crev) { + case 1: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v1(&voltage_info->v1, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA *formula = + &voltage_object->v1.asFormula; + *min_voltage = + le16_to_cpu(formula->usVoltageBaseLevel); + return 0; + } + break; + case 2: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v2(&voltage_info->v2, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA_V2 *formula = + &voltage_object->v2.asFormula; + if (formula->ucNumOfVoltageEntries) { + *min_voltage = + le16_to_cpu(formula->asVIDAdjustEntries[ + 0 + ].usVoltageValue); + return 0; + } + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + + } + return -EINVAL; +} + +int radeon_atom_get_voltage_step(struct radeon_device *rdev, + u8 voltage_type, u16 *voltage_step) +{ + int index = GetIndexIntoMasterTable(DATA, VoltageObjectInfo); + u8 frev, crev; + u16 data_offset, size; + union voltage_object_info *voltage_info; + union voltage_object *voltage_object = NULL; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + voltage_info = (union voltage_object_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + switch (crev) { + case 1: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v1(&voltage_info->v1, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA *formula = + &voltage_object->v1.asFormula; + if (formula->ucFlag & 1) + *voltage_step = + (le16_to_cpu(formula->usVoltageStep) + 1) / 2; + else + *voltage_step = + le16_to_cpu(formula->usVoltageStep); + return 0; + } + break; + case 2: + return -EINVAL; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + + } + return -EINVAL; +} + +int radeon_atom_round_to_true_voltage(struct radeon_device *rdev, + u8 voltage_type, + u16 nominal_voltage, + u16 *true_voltage) +{ + u16 min_voltage, max_voltage, voltage_step; + + if (radeon_atom_get_max_voltage(rdev, voltage_type, &max_voltage)) + return -EINVAL; + if (radeon_atom_get_min_voltage(rdev, voltage_type, &min_voltage)) + return -EINVAL; + if (radeon_atom_get_voltage_step(rdev, voltage_type, &voltage_step)) + return -EINVAL; + + if (nominal_voltage <= min_voltage) + *true_voltage = min_voltage; + else if (nominal_voltage >= max_voltage) + *true_voltage = max_voltage; + else + *true_voltage = min_voltage + + ((nominal_voltage - min_voltage) / voltage_step) * + voltage_step; + + return 0; +} + +int radeon_atom_get_voltage_table(struct radeon_device *rdev, + u8 voltage_type, u8 voltage_mode, + struct atom_voltage_table *voltage_table) +{ + int index = GetIndexIntoMasterTable(DATA, VoltageObjectInfo); + u8 frev, crev; + u16 data_offset, size; + int i, ret; + union voltage_object_info *voltage_info; + union voltage_object *voltage_object = NULL; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + voltage_info = (union voltage_object_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + switch (frev) { + case 1: + case 2: + switch (crev) { + case 1: + DRM_ERROR("old table version %d, %d\n", frev, crev); + return -EINVAL; + case 2: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v2(&voltage_info->v2, voltage_type); + if (voltage_object) { + ATOM_VOLTAGE_FORMULA_V2 *formula = + &voltage_object->v2.asFormula; + if (formula->ucNumOfVoltageEntries > MAX_VOLTAGE_ENTRIES) + return -EINVAL; + for (i = 0; i < formula->ucNumOfVoltageEntries; i++) { + voltage_table->entries[i].value = + le16_to_cpu(formula->asVIDAdjustEntries[i].usVoltageValue); + ret = radeon_atom_get_voltage_gpio_settings(rdev, + voltage_table->entries[i].value, + voltage_type, + &voltage_table->entries[i].smio_low, + &voltage_table->mask_low); + if (ret) + return ret; + } + voltage_table->count = formula->ucNumOfVoltageEntries; + return 0; + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + break; + case 3: + switch (crev) { + case 1: + voltage_object = (union voltage_object *) + atom_lookup_voltage_object_v3(&voltage_info->v3, + voltage_type, voltage_mode); + if (voltage_object) { + ATOM_GPIO_VOLTAGE_OBJECT_V3 *gpio = + &voltage_object->v3.asGpioVoltageObj; + if (gpio->ucGpioEntryNum > MAX_VOLTAGE_ENTRIES) + return -EINVAL; + for (i = 0; i < gpio->ucGpioEntryNum; i++) { + voltage_table->entries[i].value = + le16_to_cpu(gpio->asVolGpioLut[i].usVoltageValue); + voltage_table->entries[i].smio_low = + le32_to_cpu(gpio->asVolGpioLut[i].ulVoltageId); + } + voltage_table->mask_low = le32_to_cpu(gpio->ulGpioMaskVal); + voltage_table->count = gpio->ucGpioEntryNum; + voltage_table->phase_delay = gpio->ucPhaseDelay; + return 0; + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + break; + default: + DRM_ERROR("unknown voltage object table\n"); + return -EINVAL; + } + } + return -EINVAL; +} + +union vram_info { + struct _ATOM_VRAM_INFO_V3 v1_3; + struct _ATOM_VRAM_INFO_V4 v1_4; + struct _ATOM_VRAM_INFO_HEADER_V2_1 v2_1; +}; + +int radeon_atom_get_memory_info(struct radeon_device *rdev, + u8 module_index, struct atom_memory_info *mem_info) +{ + int index = GetIndexIntoMasterTable(DATA, VRAM_Info); + u8 frev, crev, i; + u16 data_offset, size; + union vram_info *vram_info; + + memset(mem_info, 0, sizeof(struct atom_memory_info)); + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + vram_info = (union vram_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + switch (frev) { + case 1: + switch (crev) { + case 3: + /* r6xx */ + if (module_index < vram_info->v1_3.ucNumOfVRAMModule) { + ATOM_VRAM_MODULE_V3 *vram_module = + (ATOM_VRAM_MODULE_V3 *)vram_info->v1_3.aVramInfo; + + for (i = 0; i < module_index; i++) { + if (le16_to_cpu(vram_module->usSize) == 0) + return -EINVAL; + vram_module = (ATOM_VRAM_MODULE_V3 *) + ((u8 *)vram_module + le16_to_cpu(vram_module->usSize)); + } + mem_info->mem_vendor = vram_module->asMemory.ucMemoryVenderID & 0xf; + mem_info->mem_type = vram_module->asMemory.ucMemoryType & 0xf0; + } else + return -EINVAL; + break; + case 4: + /* r7xx, evergreen */ + if (module_index < vram_info->v1_4.ucNumOfVRAMModule) { + ATOM_VRAM_MODULE_V4 *vram_module = + (ATOM_VRAM_MODULE_V4 *)vram_info->v1_4.aVramInfo; + + for (i = 0; i < module_index; i++) { + if (le16_to_cpu(vram_module->usModuleSize) == 0) + return -EINVAL; + vram_module = (ATOM_VRAM_MODULE_V4 *) + ((u8 *)vram_module + le16_to_cpu(vram_module->usModuleSize)); + } + mem_info->mem_vendor = vram_module->ucMemoryVenderID & 0xf; + mem_info->mem_type = vram_module->ucMemoryType & 0xf0; + } else + return -EINVAL; + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + break; + case 2: + switch (crev) { + case 1: + /* ni */ + if (module_index < vram_info->v2_1.ucNumOfVRAMModule) { + ATOM_VRAM_MODULE_V7 *vram_module = + (ATOM_VRAM_MODULE_V7 *)vram_info->v2_1.aVramInfo; + + for (i = 0; i < module_index; i++) { + if (le16_to_cpu(vram_module->usModuleSize) == 0) + return -EINVAL; + vram_module = (ATOM_VRAM_MODULE_V7 *) + ((u8 *)vram_module + le16_to_cpu(vram_module->usModuleSize)); + } + mem_info->mem_vendor = vram_module->ucMemoryVenderID & 0xf; + mem_info->mem_type = vram_module->ucMemoryType & 0xf0; + } else + return -EINVAL; + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + return 0; + } + return -EINVAL; +} + +int radeon_atom_get_mclk_range_table(struct radeon_device *rdev, + bool gddr5, u8 module_index, + struct atom_memory_clock_range_table *mclk_range_table) +{ + int index = GetIndexIntoMasterTable(DATA, VRAM_Info); + u8 frev, crev, i; + u16 data_offset, size; + union vram_info *vram_info; + u32 mem_timing_size = gddr5 ? + sizeof(ATOM_MEMORY_TIMING_FORMAT_V2) : sizeof(ATOM_MEMORY_TIMING_FORMAT); + u8 *p; + + memset(mclk_range_table, 0, sizeof(struct atom_memory_clock_range_table)); + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + vram_info = (union vram_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + switch (frev) { + case 1: + switch (crev) { + case 3: + DRM_ERROR("old table version %d, %d\n", frev, crev); + return -EINVAL; + case 4: + /* r7xx, evergreen */ + if (module_index < vram_info->v1_4.ucNumOfVRAMModule) { + ATOM_VRAM_MODULE_V4 *vram_module = + (ATOM_VRAM_MODULE_V4 *)vram_info->v1_4.aVramInfo; + + for (i = 0; i < module_index; i++) { + if (le16_to_cpu(vram_module->usModuleSize) == 0) + return -EINVAL; + vram_module = (ATOM_VRAM_MODULE_V4 *) + ((u8 *)vram_module + le16_to_cpu(vram_module->usModuleSize)); + } + mclk_range_table->num_entries = (u8) + ((le16_to_cpu(vram_module->usModuleSize) - offsetof(ATOM_VRAM_MODULE_V4, asMemTiming)) / + mem_timing_size); + p = (u8 *)&vram_module->asMemTiming[0]; + for (i = 0; i < mclk_range_table->num_entries; i++) { + ATOM_MEMORY_TIMING_FORMAT *format = (ATOM_MEMORY_TIMING_FORMAT *)p; + mclk_range_table->mclk[i] = le32_to_cpu(format->ulClkRange); + p += mem_timing_size; + } + } else + return -EINVAL; + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + break; + case 2: + DRM_ERROR("new table version %d, %d\n", frev, crev); + return -EINVAL; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + return 0; + } + return -EINVAL; +} + +#define MEM_ID_MASK 0xff000000 +#define MEM_ID_SHIFT 24 +#define CLOCK_RANGE_MASK 0x00ffffff +#define CLOCK_RANGE_SHIFT 0 +#define LOW_NIBBLE_MASK 0xf +#define DATA_EQU_PREV 0 +#define DATA_FROM_TABLE 4 + +int radeon_atom_init_mc_reg_table(struct radeon_device *rdev, + u8 module_index, + struct atom_mc_reg_table *reg_table) +{ + int index = GetIndexIntoMasterTable(DATA, VRAM_Info); + u8 frev, crev, num_entries, t_mem_id, num_ranges = 0; + u32 i = 0, j; + u16 data_offset, size; + union vram_info *vram_info; + + memset(reg_table, 0, sizeof(struct atom_mc_reg_table)); + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, &size, + &frev, &crev, &data_offset)) { + vram_info = (union vram_info *) + ((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + switch (frev) { + case 1: + DRM_ERROR("old table version %d, %d\n", frev, crev); + return -EINVAL; + case 2: + switch (crev) { + case 1: + if (module_index < vram_info->v2_1.ucNumOfVRAMModule) { + ATOM_INIT_REG_BLOCK *reg_block = + (ATOM_INIT_REG_BLOCK *) + ((u8 *)vram_info + le16_to_cpu(vram_info->v2_1.usMemClkPatchTblOffset)); + ATOM_MEMORY_SETTING_DATA_BLOCK *reg_data = + (ATOM_MEMORY_SETTING_DATA_BLOCK *) + ((u8 *)reg_block + (2 * sizeof(u16)) + + le16_to_cpu(reg_block->usRegIndexTblSize)); + ATOM_INIT_REG_INDEX_FORMAT *format = ®_block->asRegIndexBuf[0]; + num_entries = (u8)((le16_to_cpu(reg_block->usRegIndexTblSize)) / + sizeof(ATOM_INIT_REG_INDEX_FORMAT)) - 1; + if (num_entries > VBIOS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + while (i < num_entries) { + if (format->ucPreRegDataLength & ACCESS_PLACEHOLDER) + break; + reg_table->mc_reg_address[i].s1 = + (u16)(le16_to_cpu(format->usRegIndex)); + reg_table->mc_reg_address[i].pre_reg_data = + (u8)(format->ucPreRegDataLength); + i++; + format = (ATOM_INIT_REG_INDEX_FORMAT *) + ((u8 *)format + sizeof(ATOM_INIT_REG_INDEX_FORMAT)); + } + reg_table->last = i; + while ((*(u32 *)reg_data != END_OF_REG_DATA_BLOCK) && + (num_ranges < VBIOS_MAX_AC_TIMING_ENTRIES)) { + t_mem_id = (u8)((*(u32 *)reg_data & MEM_ID_MASK) >> MEM_ID_SHIFT); + if (module_index == t_mem_id) { + reg_table->mc_reg_table_entry[num_ranges].mclk_max = + (u32)((*(u32 *)reg_data & CLOCK_RANGE_MASK) >> CLOCK_RANGE_SHIFT); + for (i = 0, j = 1; i < reg_table->last; i++) { + if ((reg_table->mc_reg_address[i].pre_reg_data & LOW_NIBBLE_MASK) == DATA_FROM_TABLE) { + reg_table->mc_reg_table_entry[num_ranges].mc_data[i] = + (u32)*((u32 *)reg_data + j); + j++; + } else if ((reg_table->mc_reg_address[i].pre_reg_data & LOW_NIBBLE_MASK) == DATA_EQU_PREV) { + reg_table->mc_reg_table_entry[num_ranges].mc_data[i] = + reg_table->mc_reg_table_entry[num_ranges].mc_data[i - 1]; + } + } + num_ranges++; + } + reg_data = (ATOM_MEMORY_SETTING_DATA_BLOCK *) + ((u8 *)reg_data + le16_to_cpu(reg_block->usRegDataBlkSize)); + } + if (*(u32 *)reg_data != END_OF_REG_DATA_BLOCK) + return -EINVAL; + reg_table->num_entries = num_ranges; + } else + return -EINVAL; + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + break; + default: + DRM_ERROR("Unknown table version %d, %d\n", frev, crev); + return -EINVAL; + } + return 0; + } + return -EINVAL; +} + void radeon_atom_initialize_bios_scratch_regs(struct drm_device *dev) { struct radeon_device *rdev = dev->dev_private; diff --git a/sys/dev/drm/radeon/radeon_benchmark.c b/sys/dev/drm/radeon/radeon_benchmark.c index 9fa7125523..c1377bbbf0 100644 --- a/sys/dev/drm/radeon/radeon_benchmark.c +++ b/sys/dev/drm/radeon/radeon_benchmark.c @@ -20,10 +20,7 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_benchmark.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include "radeon_reg.h" diff --git a/sys/dev/drm/radeon/radeon_blit_common.h b/sys/dev/drm/radeon/radeon_blit_common.h index 42c900bdbb..4ecbe72c9d 100644 --- a/sys/dev/drm/radeon/radeon_blit_common.h +++ b/sys/dev/drm/radeon/radeon_blit_common.h @@ -22,7 +22,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_blit_common.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RADEON_BLIT_COMMON_H__ diff --git a/sys/dev/drm/radeon/radeon_combios.c b/sys/dev/drm/radeon/radeon_combios.c index 0301dbdc91..29898f5686 100644 --- a/sys/dev/drm/radeon/radeon_combios.c +++ b/sys/dev/drm/radeon/radeon_combios.c @@ -129,7 +129,7 @@ static uint16_t combios_get_table_offset(struct drm_device *dev, enum radeon_combios_table_offset table) { struct radeon_device *rdev = dev->dev_private; - int rev; + int rev, size; uint16_t offset = 0, check_offset; if (!rdev->bios) @@ -138,174 +138,106 @@ static uint16_t combios_get_table_offset(struct drm_device *dev, switch (table) { /* absolute offset tables */ case COMBIOS_ASIC_INIT_1_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0xc); - if (check_offset) - offset = check_offset; + check_offset = 0xc; break; case COMBIOS_BIOS_SUPPORT_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x14); - if (check_offset) - offset = check_offset; + check_offset = 0x14; break; case COMBIOS_DAC_PROGRAMMING_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x2a); - if (check_offset) - offset = check_offset; + check_offset = 0x2a; break; case COMBIOS_MAX_COLOR_DEPTH_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x2c); - if (check_offset) - offset = check_offset; + check_offset = 0x2c; break; case COMBIOS_CRTC_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x2e); - if (check_offset) - offset = check_offset; + check_offset = 0x2e; break; case COMBIOS_PLL_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x30); - if (check_offset) - offset = check_offset; + check_offset = 0x30; break; case COMBIOS_TV_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x32); - if (check_offset) - offset = check_offset; + check_offset = 0x32; break; case COMBIOS_DFP_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x34); - if (check_offset) - offset = check_offset; + check_offset = 0x34; break; case COMBIOS_HW_CONFIG_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x36); - if (check_offset) - offset = check_offset; + check_offset = 0x36; break; case COMBIOS_MULTIMEDIA_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x38); - if (check_offset) - offset = check_offset; + check_offset = 0x38; break; case COMBIOS_TV_STD_PATCH_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x3e); - if (check_offset) - offset = check_offset; + check_offset = 0x3e; break; case COMBIOS_LCD_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x40); - if (check_offset) - offset = check_offset; + check_offset = 0x40; break; case COMBIOS_MOBILE_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x42); - if (check_offset) - offset = check_offset; + check_offset = 0x42; break; case COMBIOS_PLL_INIT_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x46); - if (check_offset) - offset = check_offset; + check_offset = 0x46; break; case COMBIOS_MEM_CONFIG_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x48); - if (check_offset) - offset = check_offset; + check_offset = 0x48; break; case COMBIOS_SAVE_MASK_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x4a); - if (check_offset) - offset = check_offset; + check_offset = 0x4a; break; case COMBIOS_HARDCODED_EDID_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x4c); - if (check_offset) - offset = check_offset; + check_offset = 0x4c; break; case COMBIOS_ASIC_INIT_2_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x4e); - if (check_offset) - offset = check_offset; + check_offset = 0x4e; break; case COMBIOS_CONNECTOR_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x50); - if (check_offset) - offset = check_offset; + check_offset = 0x50; break; case COMBIOS_DYN_CLK_1_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x52); - if (check_offset) - offset = check_offset; + check_offset = 0x52; break; case COMBIOS_RESERVED_MEM_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x54); - if (check_offset) - offset = check_offset; + check_offset = 0x54; break; case COMBIOS_EXT_TMDS_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x58); - if (check_offset) - offset = check_offset; + check_offset = 0x58; break; case COMBIOS_MEM_CLK_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x5a); - if (check_offset) - offset = check_offset; + check_offset = 0x5a; break; case COMBIOS_EXT_DAC_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x5c); - if (check_offset) - offset = check_offset; + check_offset = 0x5c; break; case COMBIOS_MISC_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x5e); - if (check_offset) - offset = check_offset; + check_offset = 0x5e; break; case COMBIOS_CRT_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x60); - if (check_offset) - offset = check_offset; + check_offset = 0x60; break; case COMBIOS_INTEGRATED_SYSTEM_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x62); - if (check_offset) - offset = check_offset; + check_offset = 0x62; break; case COMBIOS_COMPONENT_VIDEO_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x64); - if (check_offset) - offset = check_offset; + check_offset = 0x64; break; case COMBIOS_FAN_SPEED_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x66); - if (check_offset) - offset = check_offset; + check_offset = 0x66; break; case COMBIOS_OVERDRIVE_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x68); - if (check_offset) - offset = check_offset; + check_offset = 0x68; break; case COMBIOS_OEM_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x6a); - if (check_offset) - offset = check_offset; + check_offset = 0x6a; break; case COMBIOS_DYN_CLK_2_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x6c); - if (check_offset) - offset = check_offset; + check_offset = 0x6c; break; case COMBIOS_POWER_CONNECTOR_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x6e); - if (check_offset) - offset = check_offset; + check_offset = 0x6e; break; case COMBIOS_I2C_INFO_TABLE: - check_offset = RBIOS16(rdev->bios_header_start + 0x70); - if (check_offset) - offset = check_offset; + check_offset = 0x70; break; /* relative offset tables */ case COMBIOS_ASIC_INIT_3_TABLE: /* offset from misc info */ @@ -421,11 +353,16 @@ static uint16_t combios_get_table_offset(struct drm_device *dev, } break; default: + check_offset = 0; break; } - return offset; + size = RBIOS8(rdev->bios_header_start + 0x6); + /* check absolute offset tables */ + if (table < COMBIOS_ASIC_INIT_3_TABLE && check_offset && check_offset < size) + offset = RBIOS16(rdev->bios_header_start + check_offset); + return offset; } bool radeon_combios_check_hardcoded_edid(struct radeon_device *rdev) @@ -948,16 +885,22 @@ struct radeon_encoder_primary_dac *radeon_combios_get_primary_dac_info(struct dac = RBIOS8(dac_info + 0x3) & 0xf; p_dac->ps2_pdac_adj = (bg << 8) | (dac); } - /* if the values are all zeros, use the table */ - if (p_dac->ps2_pdac_adj) + /* if the values are zeros, use the table */ + if ((dac == 0) || (bg == 0)) + found = 0; + else found = 1; } /* quirks */ + /* Radeon 7000 (RV100) */ + if (((rdev->ddev->pci_device == 0x5159) && ++ (rdev->ddev->pci_subvendor == 0x174B) && ++ (rdev->ddev->pci_subdevice == 0x7c28)) || /* Radeon 9100 (R200) */ - if ((rdev->ddev->pci_device == 0x514D) && + ((rdev->ddev->pci_device == 0x514D) && (rdev->ddev->pci_subvendor == 0x174B) && - (rdev->ddev->pci_subdevice == 0x7149)) { + (rdev->ddev->pci_subdevice == 0x7149))) { /* vbios value is bad, use the default */ found = 0; } diff --git a/sys/dev/drm/radeon/radeon_cs.c b/sys/dev/drm/radeon/radeon_cs.c index 087862d144..36ceacec8e 100644 --- a/sys/dev/drm/radeon/radeon_cs.c +++ b/sys/dev/drm/radeon/radeon_cs.c @@ -564,6 +564,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) return r; } + /* XXX pick SD/HD/MVC */ if (parser.ring == R600_RING_TYPE_UVD_INDEX) radeon_uvd_note_usage(rdev); diff --git a/sys/dev/drm/radeon/radeon_cursor.c b/sys/dev/drm/radeon/radeon_cursor.c index 461dbd4382..883f674e9d 100644 --- a/sys/dev/drm/radeon/radeon_cursor.c +++ b/sys/dev/drm/radeon/radeon_cursor.c @@ -30,9 +30,6 @@ #include #include "radeon.h" -#define CURSOR_WIDTH 64 -#define CURSOR_HEIGHT 64 - static void radeon_lock_cursor(struct drm_crtc *crtc, bool lock) { struct radeon_device *rdev = crtc->dev->dev_private; @@ -170,7 +167,8 @@ int radeon_crtc_cursor_set(struct drm_crtc *crtc, goto unpin; } - if ((width > CURSOR_WIDTH) || (height > CURSOR_HEIGHT)) { + if ((width > radeon_crtc->max_cursor_width) || + (height > radeon_crtc->max_cursor_height)) { DRM_ERROR("bad cursor width or height %d x %d\n", width, height); return -EINVAL; } @@ -236,11 +234,11 @@ int radeon_crtc_cursor_move(struct drm_crtc *crtc, DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y); if (x < 0) { - xorigin = min(-x, CURSOR_WIDTH - 1); + xorigin = min(-x, radeon_crtc->max_cursor_width - 1); x = 0; } if (y < 0) { - yorigin = min(-y, CURSOR_HEIGHT - 1); + yorigin = min(-y, radeon_crtc->max_cursor_height - 1); y = 0; } diff --git a/sys/dev/drm/radeon/radeon_device.c b/sys/dev/drm/radeon/radeon_device.c index 697de3ac52..6e026074e8 100644 --- a/sys/dev/drm/radeon/radeon_device.c +++ b/sys/dev/drm/radeon/radeon_device.c @@ -31,6 +31,7 @@ #include #include #include +#include #include "radeon_reg.h" #include "radeon.h" #include "atom.h" @@ -93,6 +94,9 @@ static const char radeon_family_name[][16] = { "VERDE", "OLAND", "HAINAN", + "BONAIRE", + "KAVERI", + "KABINI", "LAST", }; @@ -226,6 +230,94 @@ void radeon_scratch_free(struct radeon_device *rdev, uint32_t reg) } } +/* + * GPU doorbell aperture helpers function. + */ +/** + * radeon_doorbell_init - Init doorbell driver information. + * + * @rdev: radeon_device pointer + * + * Init doorbell driver information (CIK) + * Returns 0 on success, error on failure. + */ +int radeon_doorbell_init(struct radeon_device *rdev) +{ + int i; + + /* doorbell bar mapping */ + rdev->doorbell.base = drm_get_resource_start(rdev->ddev, 2); + rdev->doorbell.size = drm_get_resource_len(rdev->ddev, 2); + + /* limit to 4 MB for now */ + if (rdev->doorbell.size > (4 * 1024 * 1024)) + rdev->doorbell.size = 4 * 1024 * 1024; + + rdev->doorbell.ptr = ioremap(rdev->doorbell.base, rdev->doorbell.size); + if (rdev->doorbell.ptr == NULL) { + return -ENOMEM; + } + DRM_INFO("doorbell mmio base: 0x%08X\n", (uint32_t)rdev->doorbell.base); + DRM_INFO("doorbell mmio size: %u\n", (unsigned)rdev->doorbell.size); + + rdev->doorbell.num_pages = rdev->doorbell.size / PAGE_SIZE; + + for (i = 0; i < rdev->doorbell.num_pages; i++) { + rdev->doorbell.free[i] = true; + } + return 0; +} + +/** + * radeon_doorbell_fini - Tear down doorbell driver information. + * + * @rdev: radeon_device pointer + * + * Tear down doorbell driver information (CIK) + */ +void radeon_doorbell_fini(struct radeon_device *rdev) +{ + iounmap(rdev->doorbell.ptr, rdev->doorbell.size); + rdev->doorbell.ptr = NULL; +} + +/** + * radeon_doorbell_get - Allocate a doorbell page + * + * @rdev: radeon_device pointer + * @doorbell: doorbell page number + * + * Allocate a doorbell page for use by the driver (all asics). + * Returns 0 on success or -EINVAL on failure. + */ +int radeon_doorbell_get(struct radeon_device *rdev, u32 *doorbell) +{ + int i; + + for (i = 0; i < rdev->doorbell.num_pages; i++) { + if (rdev->doorbell.free[i]) { + rdev->doorbell.free[i] = false; + *doorbell = i; + return 0; + } + } + return -EINVAL; +} + +/** + * radeon_doorbell_free - Free a doorbell page + * + * @rdev: radeon_device pointer + * @doorbell: doorbell page number + * + * Free a doorbell page allocated for use by the driver (all asics) + */ +void radeon_doorbell_free(struct radeon_device *rdev, u32 doorbell) +{ + if (doorbell < rdev->doorbell.num_pages) + rdev->doorbell.free[doorbell] = true; +} + /* * radeon_wb_*() * Writeback is the the method by which the the GPU updates special pages @@ -1079,6 +1171,7 @@ int radeon_device_init(struct radeon_device *rdev, lockinit(&rdev->pm.mutex, "drm__radeon_device__pm__mutex", 0, LK_CANRECURSE); spin_init(&rdev->gpu_clock_mutex, "radeon_clockmtx"); + spin_init(&rdev->srbm_mutex, "radeon_srbm_mutex"); lockinit(&rdev->pm.mclk_lock, "drm__radeon_device__pm__mclk_lock", 0, LK_CANRECURSE); lockinit(&rdev->exclusive_lock, "drm__radeon_device__exclusive_lock", @@ -1157,7 +1250,11 @@ int radeon_device_init(struct radeon_device *rdev, /* Registers mapping */ /* TODO: block userspace mapping of io register */ spin_init(&rdev->mmio_idx_lock, "radeon_mpio"); - rdev->rmmio_rid = PCIR_BAR(2); + if (rdev->family >= CHIP_BONAIRE) { + rdev->rmmio_rid = PCIR_BAR(5); + } else { + rdev->rmmio_rid = PCIR_BAR(2); + } rdev->rmmio = bus_alloc_resource_any(rdev->dev, SYS_RES_MEMORY, &rdev->rmmio_rid, RF_ACTIVE | RF_SHAREABLE); if (rdev->rmmio == NULL) { @@ -1168,6 +1265,10 @@ int radeon_device_init(struct radeon_device *rdev, DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)rdev->rmmio_base); DRM_INFO("register mmio size: %u\n", (unsigned)rdev->rmmio_size); + /* doorbell bar mapping */ + if (rdev->family >= CHIP_BONAIRE) + radeon_doorbell_init(rdev); + /* io port mapping */ for (i = 0; i < DRM_MAX_PCI_RESOURCE; i++) { uint32_t data; @@ -1291,6 +1392,8 @@ void radeon_device_fini(struct radeon_device *rdev) bus_release_resource(rdev->dev, SYS_RES_MEMORY, rdev->rmmio_rid, rdev->rmmio); rdev->rmmio = NULL; + if (rdev->family >= CHIP_BONAIRE) + radeon_doorbell_fini(rdev); #ifdef DUMBBELL_WIP radeon_debugfs_remove_files(rdev); #endif /* DUMBBELL_WIP */ @@ -1493,6 +1596,7 @@ int radeon_gpu_reset(struct radeon_device *rdev) radeon_save_bios_scratch_regs(rdev); /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&rdev->mman.bdev); + radeon_pm_suspend(rdev); radeon_suspend(rdev); for (i = 0; i < RADEON_NUM_RINGS; ++i) { @@ -1538,6 +1642,7 @@ retry: } } + radeon_pm_resume(rdev); drm_helper_resume_force_mode(rdev->ddev); ttm_bo_unlock_delayed_workqueue(&rdev->mman.bdev, resched); diff --git a/sys/dev/drm/radeon/radeon_display.c b/sys/dev/drm/radeon/radeon_display.c index 2d90e6492d..4a35581b6c 100644 --- a/sys/dev/drm/radeon/radeon_display.c +++ b/sys/dev/drm/radeon/radeon_display.c @@ -156,7 +156,13 @@ static void dce5_crtc_load_lut(struct drm_crtc *crtc) NI_OUTPUT_CSC_OVL_MODE(NI_OUTPUT_CSC_BYPASS))); /* XXX match this to the depth of the crtc fmt block, move to modeset? */ WREG32(0x6940 + radeon_crtc->crtc_offset, 0); - + if (ASIC_IS_DCE8(rdev)) { + /* XXX this only needs to be programmed once per crtc at startup, + * not sure where the best place for it is + */ + WREG32(CIK_ALPHA_CONTROL + radeon_crtc->crtc_offset, + CIK_CURSOR_ALPHA_BLND_ENA); + } } static void legacy_crtc_load_lut(struct drm_crtc *crtc) @@ -514,6 +520,14 @@ static void radeon_crtc_init(struct drm_device *dev, int index) radeon_crtc->crtc_id = index; rdev->mode_info.crtcs[index] = radeon_crtc; + if (rdev->family >= CHIP_BONAIRE) { + radeon_crtc->max_cursor_width = CIK_CURSOR_WIDTH; + radeon_crtc->max_cursor_height = CIK_CURSOR_HEIGHT; + } else { + radeon_crtc->max_cursor_width = CURSOR_WIDTH; + radeon_crtc->max_cursor_height = CURSOR_HEIGHT; + } + #if 0 radeon_crtc->mode_set.crtc = &radeon_crtc->base; radeon_crtc->mode_set.connectors = (struct drm_connector **)(radeon_crtc + 1); @@ -532,7 +546,7 @@ static void radeon_crtc_init(struct drm_device *dev, int index) radeon_legacy_init_crtc(dev, radeon_crtc); } -static const char *encoder_names[37] = { +static const char *encoder_names[38] = { "NONE", "INTERNAL_LVDS", "INTERNAL_TMDS1", @@ -569,7 +583,8 @@ static const char *encoder_names[37] = { "INTERNAL_UNIPHY2", "NUTMEG", "TRAVIS", - "INTERNAL_VCE" + "INTERNAL_VCE", + "INTERNAL_UNIPHY3", }; static const char *hpd_names[6] = { diff --git a/sys/dev/drm/radeon/radeon_drv.c b/sys/dev/drm/radeon/radeon_drv.c index fda457fa9a..7313154773 100644 --- a/sys/dev/drm/radeon/radeon_drv.c +++ b/sys/dev/drm/radeon/radeon_drv.c @@ -76,9 +76,10 @@ * 2.31.0 - Add fastfb support for rs690 * 2.32.0 - new info request for rings working * 2.33.0 - Add SI tiling mode array query + * 2.34.0 - Add CIK tiling mode array query */ #define KMS_DRIVER_MAJOR 2 -#define KMS_DRIVER_MINOR 33 +#define KMS_DRIVER_MINOR 34 #define KMS_DRIVER_PATCHLEVEL 0 int radeon_suspend_kms(struct drm_device *dev); int radeon_resume_kms(struct drm_device *dev); @@ -103,6 +104,7 @@ struct drm_gem_object *radeon_gem_prime_import_sg_table(struct drm_device *dev, size_t size, struct sg_table *sg); int radeon_gem_prime_pin(struct drm_gem_object *obj); +void radeon_gem_prime_unpin(struct drm_gem_object *obj); void *radeon_gem_prime_vmap(struct drm_gem_object *obj); void radeon_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); @@ -138,6 +140,8 @@ int radeon_pcie_gen2 = -1; int radeon_msi = -1; int radeon_lockup_timeout = 10000; int radeon_fastfb = 0; +int radeon_dpm = -1; +int radeon_aspm = -1; static drm_pci_id_list_t pciidlist[] = { radeon_PCI_IDS @@ -200,6 +204,12 @@ module_param_named(lockup_timeout, radeon_lockup_timeout, int, 0444); MODULE_PARM_DESC(fastfb, "Direct FB access for IGP chips (0 = disable, 1 = enable)"); module_param_named(fastfb, radeon_fastfb, int, 0444); +MODULE_PARM_DESC(dpm, "DPM support (1 = enable, 0 = disable, -1 = auto)"); +module_param_named(dpm, radeon_dpm, int, 0444); + +MODULE_PARM_DESC(aspm, "ASPM support (1 = enable, 0 = disable, -1 = auto)"); +module_param_named(aspm, radeon_aspm, int, 0444); + static int radeon_suspend(struct drm_device *dev, pm_message_t state) { drm_radeon_private_t *dev_priv = dev->dev_private; @@ -405,6 +415,7 @@ static struct drm_driver kms_driver = { .gem_prime_export = drm_gem_prime_export, .gem_prime_import = drm_gem_prime_import, .gem_prime_pin = radeon_gem_prime_pin, + .gem_prime_unpin = radeon_gem_prime_unpin, .gem_prime_get_sg_table = radeon_gem_prime_get_sg_table, .gem_prime_import_sg_table = radeon_gem_prime_import_sg_table, .gem_prime_vmap = radeon_gem_prime_vmap, diff --git a/sys/dev/drm/radeon/radeon_drv.h b/sys/dev/drm/radeon/radeon_drv.h index 31dc5713a0..ab3b728ef7 100644 --- a/sys/dev/drm/radeon/radeon_drv.h +++ b/sys/dev/drm/radeon/radeon_drv.h @@ -26,8 +26,6 @@ * Authors: * Kevin E. Martin * Gareth Hughes - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_drv.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RADEON_DRV_H__ diff --git a/sys/dev/drm/radeon/radeon_encoders.c b/sys/dev/drm/radeon/radeon_encoders.c index 57af5408a3..79575c5e18 100644 --- a/sys/dev/drm/radeon/radeon_encoders.c +++ b/sys/dev/drm/radeon/radeon_encoders.c @@ -22,10 +22,7 @@ * * Authors: Dave Airlie * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_encoders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include diff --git a/sys/dev/drm/radeon/radeon_family.h b/sys/dev/drm/radeon/radeon_family.h index 0158e51155..3c8289083f 100644 --- a/sys/dev/drm/radeon/radeon_family.h +++ b/sys/dev/drm/radeon/radeon_family.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_family.h 254885 2013-08-25 19:37:15Z dumbbell $ */ /* this file defines the CHIP_ and family flags used in the pciids, @@ -95,6 +93,9 @@ enum radeon_family { CHIP_VERDE, CHIP_OLAND, CHIP_HAINAN, + CHIP_BONAIRE, + CHIP_KAVERI, + CHIP_KABINI, CHIP_LAST, }; diff --git a/sys/dev/drm/radeon/radeon_fb.c b/sys/dev/drm/radeon/radeon_fb.c index 012589bc86..1fab672c8c 100644 --- a/sys/dev/drm/radeon/radeon_fb.c +++ b/sys/dev/drm/radeon/radeon_fb.c @@ -237,7 +237,7 @@ static int radeonfb_create(struct drm_fb_helper *helper, ret = radeon_framebuffer_init(rdev->ddev, &rfbdev->rfb, &mode_cmd, gobj); if (ret) { - DRM_ERROR("failed to initalise framebuffer %d\n", ret); + DRM_ERROR("failed to initialize framebuffer %d\n", ret); goto out_unref; } diff --git a/sys/dev/drm/radeon/radeon_gart.c b/sys/dev/drm/radeon/radeon_gart.c index a0395f3705..99a894bdf3 100644 --- a/sys/dev/drm/radeon/radeon_gart.c +++ b/sys/dev/drm/radeon/radeon_gart.c @@ -211,7 +211,6 @@ void radeon_gart_table_vram_free(struct radeon_device *rdev) if (rdev->gart.robj == NULL) { return; } - radeon_gart_table_vram_unpin(rdev); radeon_bo_unref(&rdev->gart.robj); } @@ -472,6 +471,7 @@ int radeon_vm_manager_init(struct radeon_device *rdev) size *= 2; r = radeon_sa_bo_manager_init(rdev, &rdev->vm_manager.sa_manager, RADEON_GPU_PAGE_ALIGN(size), + RADEON_VM_PTB_ALIGN_SIZE, RADEON_GEM_DOMAIN_VRAM); if (r) { dev_err(rdev->dev, "failed to allocate vm bo (%dKB)\n", @@ -625,10 +625,10 @@ int radeon_vm_alloc_pt(struct radeon_device *rdev, struct radeon_vm *vm) } retry: - pd_size = RADEON_GPU_PAGE_ALIGN(radeon_vm_directory_size(rdev)); + pd_size = radeon_vm_directory_size(rdev); r = radeon_sa_bo_new(rdev, &rdev->vm_manager.sa_manager, &vm->page_directory, pd_size, - RADEON_GPU_PAGE_SIZE, false); + RADEON_VM_PTB_ALIGN_SIZE, false); if (r == -ENOMEM) { r = radeon_vm_evict(rdev, vm); if (r) diff --git a/sys/dev/drm/radeon/radeon_irq.c b/sys/dev/drm/radeon/radeon_irq.c index c83be0c36f..e1ecc7852d 100644 --- a/sys/dev/drm/radeon/radeon_irq.c +++ b/sys/dev/drm/radeon/radeon_irq.c @@ -30,8 +30,6 @@ * Michel D�zer * * ------------------------ This file is DEPRECATED! ------------------------- - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_irq.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/radeon_irq_kms.c b/sys/dev/drm/radeon/radeon_irq_kms.c index cc39709de9..d65bf4af8f 100644 --- a/sys/dev/drm/radeon/radeon_irq_kms.c +++ b/sys/dev/drm/radeon/radeon_irq_kms.c @@ -84,6 +84,22 @@ static void radeon_hotplug_work_func(void *arg, int pending) drm_helper_hpd_irq_event(dev); } +/** + * radeon_irq_reset_work_func - execute gpu reset + * + * @work: work struct + * + * Execute scheduled gpu reset (cayman+). + * This function is called when the irq handler + * thinks we need a gpu reset. + */ +static void radeon_irq_reset_work_func(void *arg, int pending) +{ + struct radeon_device *rdev = arg; + + radeon_gpu_reset(rdev); +} + /** * radeon_driver_irq_preinstall_kms - drm irq preinstall callback * @@ -101,6 +117,7 @@ void radeon_driver_irq_preinstall_kms(struct drm_device *dev) /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) atomic_set(&rdev->irq.ring_int[i], 0); + rdev->irq.dpm_thermal = false; for (i = 0; i < RADEON_MAX_HPD_PINS; i++) rdev->irq.hpd[i] = false; for (i = 0; i < RADEON_MAX_CRTCS; i++) { @@ -147,6 +164,7 @@ void radeon_driver_irq_uninstall_kms(struct drm_device *dev) /* Disable *all* interrupts */ for (i = 0; i < RADEON_NUM_RINGS; i++) atomic_set(&rdev->irq.ring_int[i], 0); + rdev->irq.dpm_thermal = false; for (i = 0; i < RADEON_MAX_HPD_PINS; i++) rdev->irq.hpd[i] = false; for (i = 0; i < RADEON_MAX_CRTCS; i++) { @@ -246,8 +264,6 @@ int radeon_irq_kms_init(struct radeon_device *rdev) { int r = 0; - TASK_INIT(&rdev->hotplug_work, 0, radeon_hotplug_work_func, rdev); - TASK_INIT(&rdev->audio_work, 0, r600_audio_update_hdmi, rdev); lockinit(&rdev->irq.lock, "drm__radeon_device__irq__lock", 0, LK_CANRECURSE); r = drm_vblank_init(rdev->ddev, rdev->num_crtc); @@ -265,6 +281,11 @@ int radeon_irq_kms_init(struct radeon_device *rdev) rdev->irq.installed = false; return r; } + + TASK_INIT(&rdev->hotplug_work, 0, radeon_hotplug_work_func, rdev); + TASK_INIT(&rdev->audio_work, 0, r600_audio_update_hdmi, rdev); + TASK_INIT(&rdev->reset_work, 0, radeon_irq_reset_work_func, rdev); + DRM_INFO("radeon: irq initialized.\n"); return 0; } diff --git a/sys/dev/drm/radeon/radeon_kms.c b/sys/dev/drm/radeon/radeon_kms.c index 8f7a409c68..7d3f385520 100644 --- a/sys/dev/drm/radeon/radeon_kms.c +++ b/sys/dev/drm/radeon/radeon_kms.c @@ -234,7 +234,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *value = rdev->accel_working; break; case RADEON_INFO_TILING_CONFIG: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.tile_config; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.tile_config; else if (rdev->family >= CHIP_CAYMAN) *value = rdev->config.cayman.tile_config; @@ -286,7 +288,10 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *value = rdev->clock.spll.reference_freq * 10; break; case RADEON_INFO_NUM_BACKENDS: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.max_backends_per_se * + rdev->config.cik.max_shader_engines; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.max_backends_per_se * rdev->config.si.max_shader_engines; else if (rdev->family >= CHIP_CAYMAN) @@ -303,7 +308,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file } break; case RADEON_INFO_NUM_TILE_PIPES: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.max_tile_pipes; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.max_tile_pipes; else if (rdev->family >= CHIP_CAYMAN) *value = rdev->config.cayman.max_tile_pipes; @@ -321,7 +328,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *value = 1; break; case RADEON_INFO_BACKEND_MAP: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + return -EINVAL; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.backend_map; else if (rdev->family >= CHIP_CAYMAN) *value = rdev->config.cayman.backend_map; @@ -348,7 +357,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *value = RADEON_IB_VM_MAX_SIZE; break; case RADEON_INFO_MAX_PIPES: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.max_cu_per_sh; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.max_cu_per_sh; else if (rdev->family >= CHIP_CAYMAN) *value = rdev->config.cayman.max_pipes_per_simd; @@ -372,7 +383,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file value64 = radeon_get_gpu_clock_counter(rdev); break; case RADEON_INFO_MAX_SE: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.max_shader_engines; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.max_shader_engines; else if (rdev->family >= CHIP_CAYMAN) *value = rdev->config.cayman.max_shader_engines; @@ -382,7 +395,9 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file *value = 1; break; case RADEON_INFO_MAX_SH_PER_SE: - if (rdev->family >= CHIP_TAHITI) + if (rdev->family >= CHIP_BONAIRE) + *value = rdev->config.cik.max_sh_per_se; + else if (rdev->family >= CHIP_TAHITI) *value = rdev->config.si.max_sh_per_se; else return -EINVAL; @@ -412,12 +427,16 @@ static int radeon_info_ioctl(struct drm_device *dev, void *data, struct drm_file } break; case RADEON_INFO_SI_TILE_MODE_ARRAY: - if (rdev->family < CHIP_TAHITI) { - DRM_DEBUG_KMS("tile mode array is si only!\n"); + if (rdev->family >= CHIP_BONAIRE) { + value = rdev->config.cik.tile_mode_array; + value_size = sizeof(uint32_t)*32; + } else if (rdev->family >= CHIP_TAHITI) { + value = rdev->config.si.tile_mode_array; + value_size = sizeof(uint32_t)*32; + } else { + DRM_DEBUG_KMS("tile mode array is si+ only!\n"); return -EINVAL; } - value = rdev->config.si.tile_mode_array; - value_size = sizeof(uint32_t)*32; break; default: DRM_DEBUG_KMS("Invalid request %d\n", info->request); diff --git a/sys/dev/drm/radeon/radeon_legacy_crtc.c b/sys/dev/drm/radeon/radeon_legacy_crtc.c index d9c01ae8ed..0616f9d4da 100644 --- a/sys/dev/drm/radeon/radeon_legacy_crtc.c +++ b/sys/dev/drm/radeon/radeon_legacy_crtc.c @@ -22,10 +22,7 @@ * * Authors: Dave Airlie * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_legacy_crtc.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include diff --git a/sys/dev/drm/radeon/radeon_legacy_tv.c b/sys/dev/drm/radeon/radeon_legacy_tv.c index 9f08b55599..ad5af4284d 100644 --- a/sys/dev/drm/radeon/radeon_legacy_tv.c +++ b/sys/dev/drm/radeon/radeon_legacy_tv.c @@ -1,5 +1,3 @@ -/* $FreeBSD: head/sys/dev/drm2/radeon/radeon_legacy_tv.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include "radeon.h" diff --git a/sys/dev/drm/radeon/radeon_mode.h b/sys/dev/drm/radeon/radeon_mode.h index b30cd8def8..a69b36cc6c 100644 --- a/sys/dev/drm/radeon/radeon_mode.h +++ b/sys/dev/drm/radeon/radeon_mode.h @@ -305,6 +305,8 @@ struct radeon_crtc { uint64_t cursor_addr; int cursor_width; int cursor_height; + int max_cursor_width; + int max_cursor_height; uint32_t legacy_display_base_addr; uint32_t legacy_cursor_offset; enum radeon_rmx_type rmx_type; @@ -327,6 +329,11 @@ struct radeon_crtc { u32 pll_flags; struct drm_encoder *encoder; struct drm_connector *connector; + /* for dpm */ + u32 line_time; + u32 wm_low; + u32 wm_high; + struct drm_display_mode hw_mode; }; struct radeon_encoder_primary_dac { @@ -510,12 +517,99 @@ struct atom_clock_dividers { bool enable_dithen; u32 vco_mode; u32 real_clock; + /* added for CI */ + u32 post_divider; + u32 flags; +}; + +struct atom_mpll_param { + union { + struct { +#ifdef __BIG_ENDIAN + u32 reserved : 8; + u32 clkfrac : 12; + u32 clkf : 12; +#else + u32 clkf : 12; + u32 clkfrac : 12; + u32 reserved : 8; +#endif + }; + u32 fb_div; + }; + u32 post_div; + u32 bwcntl; + u32 dll_speed; + u32 vco_mode; + u32 yclk_sel; + u32 qdr; + u32 half_rate; +}; + +#define MEM_TYPE_GDDR5 0x50 +#define MEM_TYPE_GDDR4 0x40 +#define MEM_TYPE_GDDR3 0x30 +#define MEM_TYPE_DDR2 0x20 +#define MEM_TYPE_GDDR1 0x10 +#define MEM_TYPE_DDR3 0xb0 +#define MEM_TYPE_MASK 0xf0 + +struct atom_memory_info { + u8 mem_vendor; + u8 mem_type; +}; + +#define MAX_AC_TIMING_ENTRIES 16 + +struct atom_memory_clock_range_table +{ + u8 num_entries; + u8 rsv[3]; + u32 mclk[MAX_AC_TIMING_ENTRIES]; +}; + +#define VBIOS_MC_REGISTER_ARRAY_SIZE 32 +#define VBIOS_MAX_AC_TIMING_ENTRIES 20 + +struct atom_mc_reg_entry { + u32 mclk_max; + u32 mc_data[VBIOS_MC_REGISTER_ARRAY_SIZE]; +}; + +struct atom_mc_register_address { + u16 s1; + u8 pre_reg_data; +}; + +struct atom_mc_reg_table { + u8 last; + u8 num_entries; + struct atom_mc_reg_entry mc_reg_table_entry[VBIOS_MAX_AC_TIMING_ENTRIES]; + struct atom_mc_register_address mc_reg_address[VBIOS_MC_REGISTER_ARRAY_SIZE]; +}; + +#define MAX_VOLTAGE_ENTRIES 32 + +struct atom_voltage_table_entry +{ + u16 value; + u32 smio_low; +}; + +struct atom_voltage_table +{ + u32 count; + u32 mask_low; + u32 phase_delay; + struct atom_voltage_table_entry entries[MAX_VOLTAGE_ENTRIES]; }; extern enum radeon_tv_std radeon_combios_get_tv_info(struct radeon_device *rdev); extern enum radeon_tv_std radeon_atombios_get_tv_info(struct radeon_device *rdev); +extern void radeon_atombios_get_default_voltages(struct radeon_device *rdev, + u16 *vddc, u16 *vddci, u16 *mvdd); extern struct drm_connector * radeon_get_connector_for_encoder(struct drm_encoder *encoder); diff --git a/sys/dev/drm/radeon/radeon_object.c b/sys/dev/drm/radeon/radeon_object.c index c9b8afa143..2df578408a 100644 --- a/sys/dev/drm/radeon/radeon_object.c +++ b/sys/dev/drm/radeon/radeon_object.c @@ -38,6 +38,7 @@ #ifdef DUMBBELL_WIP #include "radeon_trace.h" #endif /* DUMBBELL_WIP */ +#include static void radeon_bo_clear_surface_reg(struct radeon_bo *bo); @@ -328,8 +329,7 @@ int radeon_bo_init(struct radeon_device *rdev) { /* Add an MTRR for the VRAM */ if (!rdev->fastfb_working) { - rdev->mc.vram_mtrr = drm_mtrr_add(rdev->mc.aper_base, rdev->mc.aper_size, - DRM_MTRR_WC); + rdev->mc.vram_mtrr = arch_phys_wc_add(rdev->mc.aper_base, rdev->mc.aper_size); } DRM_INFO("Detected VRAM RAM=%juM, BAR=%juM\n", (uintmax_t)rdev->mc.mc_vram_size >> 20, @@ -342,6 +342,7 @@ int radeon_bo_init(struct radeon_device *rdev) void radeon_bo_fini(struct radeon_device *rdev) { radeon_ttm_fini(rdev); + arch_phys_wc_del(rdev->mc.vram_mtrr); } void radeon_bo_list_add_object(struct radeon_bo_list *lobj, diff --git a/sys/dev/drm/radeon/radeon_object.h b/sys/dev/drm/radeon/radeon_object.h index 823545c938..cb04135c83 100644 --- a/sys/dev/drm/radeon/radeon_object.h +++ b/sys/dev/drm/radeon/radeon_object.h @@ -169,7 +169,7 @@ static inline void * radeon_sa_bo_cpu_addr(struct radeon_sa_bo *sa_bo) extern int radeon_sa_bo_manager_init(struct radeon_device *rdev, struct radeon_sa_manager *sa_manager, - unsigned size, u32 domain); + unsigned size, u32 align, u32 domain); extern void radeon_sa_bo_manager_fini(struct radeon_device *rdev, struct radeon_sa_manager *sa_manager); extern int radeon_sa_bo_manager_start(struct radeon_device *rdev, diff --git a/sys/dev/drm/radeon/radeon_pm.c b/sys/dev/drm/radeon/radeon_pm.c index 979c253a9e..d15716e689 100644 --- a/sys/dev/drm/radeon/radeon_pm.c +++ b/sys/dev/drm/radeon/radeon_pm.c @@ -23,6 +23,7 @@ * $FreeBSD: head/sys/dev/drm2/radeon/radeon_pm.c 254885 2013-08-25 19:37:15Z dumbbell $ */ +#include #include #include "radeon.h" #include "avivod.h" @@ -86,8 +87,7 @@ static void radeon_pm_update_profile(struct radeon_device *rdev) rdev->pm.profile_index = PM_PROFILE_DEFAULT_IDX; break; case PM_PROFILE_AUTO: -#ifdef DUMBBELL_WIP - if (power_supply_is_system_supplied() > 0) { + if (power_profile_get_state() == POWER_PROFILE_PERFORMANCE) { if (rdev->pm.active_crtc_count > 1) rdev->pm.profile_index = PM_PROFILE_HIGH_MH_IDX; else @@ -98,7 +98,6 @@ static void radeon_pm_update_profile(struct radeon_device *rdev) else rdev->pm.profile_index = PM_PROFILE_MID_SH_IDX; } -#endif /* DUMBBELL_WIP */ break; case PM_PROFILE_LOW: if (rdev->pm.active_crtc_count > 1) @@ -248,7 +247,7 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev) return; DRM_LOCK(rdev->ddev); - lockmgr(&rdev->pm.mclk_lock, LK_EXCLUSIVE); + lockmgr(&rdev->pm.mclk_lock, LK_EXCLUSIVE); // down_write lockmgr(&rdev->ring_lock, LK_EXCLUSIVE); /* wait for the rings to drain */ @@ -261,7 +260,7 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev) if (r) { /* needs a GPU reset dont reset here */ lockmgr(&rdev->ring_lock, LK_RELEASE); - lockmgr(&rdev->pm.mclk_lock, LK_RELEASE); + lockmgr(&rdev->pm.mclk_lock, LK_RELEASE); // up_write DRM_UNLOCK(rdev->ddev); return; } @@ -297,7 +296,7 @@ static void radeon_pm_set_clocks(struct radeon_device *rdev) rdev->pm.dynpm_planned_action = DYNPM_ACTION_NONE; lockmgr(&rdev->ring_lock, LK_RELEASE); - lockmgr(&rdev->pm.mclk_lock, LK_RELEASE); + lockmgr(&rdev->pm.mclk_lock, LK_RELEASE); // up_write DRM_UNLOCK(rdev->ddev); } @@ -395,7 +394,8 @@ static ssize_t radeon_get_pm_method(struct device *dev, int pm = rdev->pm.pm_method; return ksnprintf(buf, PAGE_SIZE, "%s\n", - (pm == PM_METHOD_DYNPM) ? "dynpm" : "profile"); + (pm == PM_METHOD_DYNPM) ? "dynpm" : + (pm == PM_METHOD_PROFILE) ? "profile" : "dpm"); } static ssize_t radeon_set_pm_method(struct device *dev, @@ -406,6 +406,11 @@ static ssize_t radeon_set_pm_method(struct device *dev, struct drm_device *ddev = pci_get_drvdata(to_pci_dev(dev)); struct radeon_device *rdev = ddev->dev_private; + /* we don't support the legacy modes with dpm */ + if (rdev->pm.pm_method == PM_METHOD_DPM) { + count = -EINVAL; + goto fail; + } if (strncmp("dynpm", buf, strlen("dynpm")) == 0) { lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); @@ -432,8 +437,96 @@ fail: return count; } +static ssize_t radeon_get_dpm_state(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = pci_get_drvdata(to_pci_dev(dev)); + struct radeon_device *rdev = ddev->dev_private; + enum radeon_pm_state_type pm = rdev->pm.dpm.user_state; + + return snprintf(buf, PAGE_SIZE, "%s\n", + (pm == POWER_STATE_TYPE_BATTERY) ? "battery" : + (pm == POWER_STATE_TYPE_BALANCED) ? "balanced" : "performance"); +} + +static ssize_t radeon_set_dpm_state(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = pci_get_drvdata(to_pci_dev(dev)); + struct radeon_device *rdev = ddev->dev_private; + + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + if (strncmp("battery", buf, strlen("battery")) == 0) + rdev->pm.dpm.user_state = POWER_STATE_TYPE_BATTERY; + else if (strncmp("balanced", buf, strlen("balanced")) == 0) + rdev->pm.dpm.user_state = POWER_STATE_TYPE_BALANCED; + else if (strncmp("performance", buf, strlen("performance")) == 0) + rdev->pm.dpm.user_state = POWER_STATE_TYPE_PERFORMANCE; + else { + lockmgr(&rdev->pm.mutex, LK_RELEASE); + count = -EINVAL; + goto fail; + } + lockmgr(&rdev->pm.mutex, LK_RELEASE); + radeon_pm_compute_clocks(rdev); +fail: + return count; +} + +static ssize_t radeon_get_dpm_forced_performance_level(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = pci_get_drvdata(to_pci_dev(dev)); + struct radeon_device *rdev = ddev->dev_private; + enum radeon_dpm_forced_level level = rdev->pm.dpm.forced_level; + + return snprintf(buf, PAGE_SIZE, "%s\n", + (level == RADEON_DPM_FORCED_LEVEL_AUTO) ? "auto" : + (level == RADEON_DPM_FORCED_LEVEL_LOW) ? "low" : "high"); +} + +static ssize_t radeon_set_dpm_forced_performance_level(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = pci_get_drvdata(to_pci_dev(dev)); + struct radeon_device *rdev = ddev->dev_private; + enum radeon_dpm_forced_level level; + int ret = 0; + + spin_lock(&rdev->pm.mutex); + if (strncmp("low", buf, strlen("low")) == 0) { + level = RADEON_DPM_FORCED_LEVEL_LOW; + } else if (strncmp("high", buf, strlen("high")) == 0) { + level = RADEON_DPM_FORCED_LEVEL_HIGH; + } else if (strncmp("auto", buf, strlen("auto")) == 0) { + level = RADEON_DPM_FORCED_LEVEL_AUTO; + } else { + spin_unlock(&rdev->pm.mutex); + count = -EINVAL; + goto fail; + } + if (rdev->asic->dpm.force_performance_level) { + ret = radeon_dpm_force_performance_level(rdev, level); + if (ret) + count = -EINVAL; + } + spin_unlock(&rdev->pm.mutex); +fail: + return count; +} + static DEVICE_ATTR(power_profile, S_IRUGO | S_IWUSR, radeon_get_pm_profile, radeon_set_pm_profile); static DEVICE_ATTR(power_method, S_IRUGO | S_IWUSR, radeon_get_pm_method, radeon_set_pm_method); +static DEVICE_ATTR(power_dpm_state, S_IRUGO | S_IWUSR, radeon_get_dpm_state, radeon_set_dpm_state); +static DEVICE_ATTR(power_dpm_force_performance_level, S_IRUGO | S_IWUSR, + radeon_get_dpm_forced_performance_level, + radeon_set_dpm_forced_performance_level); static ssize_t radeon_hwmon_show_temp(struct device *dev, struct device_attribute *attr, @@ -443,27 +536,10 @@ static ssize_t radeon_hwmon_show_temp(struct device *dev, struct radeon_device *rdev = ddev->dev_private; int temp; - switch (rdev->pm.int_thermal_type) { - case THERMAL_TYPE_RV6XX: - temp = rv6xx_get_temp(rdev); - break; - case THERMAL_TYPE_RV770: - temp = rv770_get_temp(rdev); - break; - case THERMAL_TYPE_EVERGREEN: - case THERMAL_TYPE_NI: - temp = evergreen_get_temp(rdev); - break; - case THERMAL_TYPE_SUMO: - temp = sumo_get_temp(rdev); - break; - case THERMAL_TYPE_SI: - temp = si_get_temp(rdev); - break; - default: + if (rdev->asic->pm.get_temperature) + temp = radeon_get_temperature(rdev); + else temp = 0; - break; - } return ksnprintf(buf, PAGE_SIZE, "%d\n", temp); } @@ -504,8 +580,7 @@ static int radeon_hwmon_init(struct radeon_device *rdev) case THERMAL_TYPE_NI: case THERMAL_TYPE_SUMO: case THERMAL_TYPE_SI: - /* No support for TN yet */ - if (rdev->family == CHIP_ARUBA) + if (rdev->asic->pm.get_temperature == NULL) return err; #ifdef DUMBBELL_WIP rdev->pm.int_hwmon_dev = hwmon_device_register(rdev->dev); @@ -542,7 +617,287 @@ static void radeon_hwmon_fini(struct radeon_device *rdev) #endif /* DUMBBELL_WIP */ } -void radeon_pm_suspend(struct radeon_device *rdev) +static void radeon_dpm_thermal_work_handler(void *arg, int pending) +{ + struct radeon_device *rdev = arg; + /* switch to the thermal state */ + enum radeon_pm_state_type dpm_state = POWER_STATE_TYPE_INTERNAL_THERMAL; + + if (!rdev->pm.dpm_enabled) + return; + + if (rdev->asic->pm.get_temperature) { + int temp = radeon_get_temperature(rdev); + + if (temp < rdev->pm.dpm.thermal.min_temp) + /* switch back the user state */ + dpm_state = rdev->pm.dpm.user_state; + } else { + if (rdev->pm.dpm.thermal.high_to_low) + /* switch back the user state */ + dpm_state = rdev->pm.dpm.user_state; + } + radeon_dpm_enable_power_state(rdev, dpm_state); +} + +static struct radeon_ps *radeon_dpm_pick_power_state(struct radeon_device *rdev, + enum radeon_pm_state_type dpm_state) +{ + int i; + struct radeon_ps *ps; + u32 ui_class; + bool single_display = (rdev->pm.dpm.new_active_crtc_count < 2) ? + true : false; + + /* check if the vblank period is too short to adjust the mclk */ + if (single_display && rdev->asic->dpm.vblank_too_short) { + if (radeon_dpm_vblank_too_short(rdev)) + single_display = false; + } + + /* certain older asics have a separare 3D performance state, + * so try that first if the user selected performance + */ + if (dpm_state == POWER_STATE_TYPE_PERFORMANCE) + dpm_state = POWER_STATE_TYPE_INTERNAL_3DPERF; + /* balanced states don't exist at the moment */ + if (dpm_state == POWER_STATE_TYPE_BALANCED) + dpm_state = POWER_STATE_TYPE_PERFORMANCE; + +restart_search: + /* Pick the best power state based on current conditions */ + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + ps = &rdev->pm.dpm.ps[i]; + ui_class = ps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK; + switch (dpm_state) { + /* user states */ + case POWER_STATE_TYPE_BATTERY: + if (ui_class == ATOM_PPLIB_CLASSIFICATION_UI_BATTERY) { + if (ps->caps & ATOM_PPLIB_SINGLE_DISPLAY_ONLY) { + if (single_display) + return ps; + } else + return ps; + } + break; + case POWER_STATE_TYPE_BALANCED: + if (ui_class == ATOM_PPLIB_CLASSIFICATION_UI_BALANCED) { + if (ps->caps & ATOM_PPLIB_SINGLE_DISPLAY_ONLY) { + if (single_display) + return ps; + } else + return ps; + } + break; + case POWER_STATE_TYPE_PERFORMANCE: + if (ui_class == ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE) { + if (ps->caps & ATOM_PPLIB_SINGLE_DISPLAY_ONLY) { + if (single_display) + return ps; + } else + return ps; + } + break; + /* internal states */ + case POWER_STATE_TYPE_INTERNAL_UVD: + return rdev->pm.dpm.uvd_ps; + case POWER_STATE_TYPE_INTERNAL_UVD_SD: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_UVD_HD: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_HDSTATE) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_UVD_HD2: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_HD2STATE) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_UVD_MVC: + if (ps->class2 & ATOM_PPLIB_CLASSIFICATION2_MVC) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_BOOT: + return rdev->pm.dpm.boot_ps; + case POWER_STATE_TYPE_INTERNAL_THERMAL: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_THERMAL) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_ACPI: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_ULV: + if (ps->class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) + return ps; + break; + case POWER_STATE_TYPE_INTERNAL_3DPERF: + if (ps->class & ATOM_PPLIB_CLASSIFICATION_3DPERFORMANCE) + return ps; + break; + default: + break; + } + } + /* use a fallback state if we didn't match */ + switch (dpm_state) { + case POWER_STATE_TYPE_INTERNAL_UVD_SD: + case POWER_STATE_TYPE_INTERNAL_UVD_HD: + case POWER_STATE_TYPE_INTERNAL_UVD_HD2: + case POWER_STATE_TYPE_INTERNAL_UVD_MVC: + return rdev->pm.dpm.uvd_ps; + case POWER_STATE_TYPE_INTERNAL_THERMAL: + dpm_state = POWER_STATE_TYPE_INTERNAL_ACPI; + goto restart_search; + case POWER_STATE_TYPE_INTERNAL_ACPI: + dpm_state = POWER_STATE_TYPE_BATTERY; + goto restart_search; + case POWER_STATE_TYPE_BATTERY: + case POWER_STATE_TYPE_BALANCED: + case POWER_STATE_TYPE_INTERNAL_3DPERF: + dpm_state = POWER_STATE_TYPE_PERFORMANCE; + goto restart_search; + default: + break; + } + + return NULL; +} + +static void radeon_dpm_change_power_state_locked(struct radeon_device *rdev) +{ + int i; + struct radeon_ps *ps; + enum radeon_pm_state_type dpm_state; + int ret; + + /* if dpm init failed */ + if (!rdev->pm.dpm_enabled) + return; + + if (rdev->pm.dpm.user_state != rdev->pm.dpm.state) { + /* add other state override checks here */ + if ((!rdev->pm.dpm.thermal_active) && + (!rdev->pm.dpm.uvd_active)) + rdev->pm.dpm.state = rdev->pm.dpm.user_state; + } + dpm_state = rdev->pm.dpm.state; + + ps = radeon_dpm_pick_power_state(rdev, dpm_state); + if (ps) + rdev->pm.dpm.requested_ps = ps; + else + return; + + /* no need to reprogram if nothing changed unless we are on BTC+ */ + if (rdev->pm.dpm.current_ps == rdev->pm.dpm.requested_ps) { + if ((rdev->family < CHIP_BARTS) || (rdev->flags & RADEON_IS_IGP)) { + /* for pre-BTC and APUs if the num crtcs changed but state is the same, + * all we need to do is update the display configuration. + */ + if (rdev->pm.dpm.new_active_crtcs != rdev->pm.dpm.current_active_crtcs) { + /* update display watermarks based on new power state */ + radeon_bandwidth_update(rdev); + /* update displays */ + radeon_dpm_display_configuration_changed(rdev); + rdev->pm.dpm.current_active_crtcs = rdev->pm.dpm.new_active_crtcs; + rdev->pm.dpm.current_active_crtc_count = rdev->pm.dpm.new_active_crtc_count; + } + return; + } else { + /* for BTC+ if the num crtcs hasn't changed and state is the same, + * nothing to do, if the num crtcs is > 1 and state is the same, + * update display configuration. + */ + if (rdev->pm.dpm.new_active_crtcs == + rdev->pm.dpm.current_active_crtcs) { + return; + } else { + if ((rdev->pm.dpm.current_active_crtc_count > 1) && + (rdev->pm.dpm.new_active_crtc_count > 1)) { + /* update display watermarks based on new power state */ + radeon_bandwidth_update(rdev); + /* update displays */ + radeon_dpm_display_configuration_changed(rdev); + rdev->pm.dpm.current_active_crtcs = rdev->pm.dpm.new_active_crtcs; + rdev->pm.dpm.current_active_crtc_count = rdev->pm.dpm.new_active_crtc_count; + return; + } + } + } + } + + printk("switching from power state:\n"); + radeon_dpm_print_power_state(rdev, rdev->pm.dpm.current_ps); + printk("switching to power state:\n"); + radeon_dpm_print_power_state(rdev, rdev->pm.dpm.requested_ps); + + lockmgr(&rdev->ddev->struct_mutex, LK_EXCLUSIVE); + lockmgr(&rdev->pm.mclk_lock, LK_EXCLUSIVE); // down_write + lockmgr(&rdev->ring_lock, LK_EXCLUSIVE); + + ret = radeon_dpm_pre_set_power_state(rdev); + if (ret) + goto done; + + /* update display watermarks based on new power state */ + radeon_bandwidth_update(rdev); + /* update displays */ + radeon_dpm_display_configuration_changed(rdev); + + rdev->pm.dpm.current_active_crtcs = rdev->pm.dpm.new_active_crtcs; + rdev->pm.dpm.current_active_crtc_count = rdev->pm.dpm.new_active_crtc_count; + + /* wait for the rings to drain */ + for (i = 0; i < RADEON_NUM_RINGS; i++) { + struct radeon_ring *ring = &rdev->ring[i]; + if (ring->ready) + radeon_fence_wait_empty_locked(rdev, i); + } + + /* program the new power state */ + radeon_dpm_set_power_state(rdev); + + /* update current power state */ + rdev->pm.dpm.current_ps = rdev->pm.dpm.requested_ps; + + radeon_dpm_post_set_power_state(rdev); + +done: + lockmgr(&rdev->ring_lock, LK_RELEASE); + lockmgr(&rdev->pm.mclk_lock, LK_RELEASE); // up_write + lockmgr(&rdev->ddev->struct_mutex, LK_RELEASE); +} + +void radeon_dpm_enable_power_state(struct radeon_device *rdev, + enum radeon_pm_state_type dpm_state) +{ + if (!rdev->pm.dpm_enabled) + return; + + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + switch (dpm_state) { + case POWER_STATE_TYPE_INTERNAL_THERMAL: + rdev->pm.dpm.thermal_active = true; + break; + case POWER_STATE_TYPE_INTERNAL_UVD: + case POWER_STATE_TYPE_INTERNAL_UVD_SD: + case POWER_STATE_TYPE_INTERNAL_UVD_HD: + case POWER_STATE_TYPE_INTERNAL_UVD_HD2: + case POWER_STATE_TYPE_INTERNAL_UVD_MVC: + rdev->pm.dpm.uvd_active = true; + break; + default: + rdev->pm.dpm.thermal_active = false; + rdev->pm.dpm.uvd_active = false; + break; + } + rdev->pm.dpm.state = dpm_state; + lockmgr(&rdev->pm.mutex, LK_RELEASE); + radeon_pm_compute_clocks(rdev); +} + +static void radeon_pm_suspend_old(struct radeon_device *rdev) { lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); if (rdev->pm.pm_method == PM_METHOD_DYNPM) { @@ -556,11 +911,30 @@ void radeon_pm_suspend(struct radeon_device *rdev) #endif /* DUMBBELL_WIP */ } -void radeon_pm_resume(struct radeon_device *rdev) +static void radeon_pm_suspend_dpm(struct radeon_device *rdev) +{ + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + /* disable dpm */ + radeon_dpm_disable(rdev); + /* reset the power state */ + rdev->pm.dpm.current_ps = rdev->pm.dpm.requested_ps = rdev->pm.dpm.boot_ps; + rdev->pm.dpm_enabled = false; + lockmgr(&rdev->pm.mutex, LK_RELEASE); +} + +void radeon_pm_suspend(struct radeon_device *rdev) +{ + if (rdev->pm.pm_method == PM_METHOD_DPM) + radeon_pm_suspend_dpm(rdev); + else + radeon_pm_suspend_old(rdev); +} + +static void radeon_pm_resume_old(struct radeon_device *rdev) { /* set up the default clocks if the MC ucode is loaded */ if ((rdev->family >= CHIP_BARTS) && - (rdev->family <= CHIP_CAYMAN) && + (rdev->family <= CHIP_HAINAN) && rdev->mc_fw) { if (rdev->pm.default_vddc) radeon_atom_set_voltage(rdev, rdev->pm.default_vddc, @@ -593,12 +967,50 @@ void radeon_pm_resume(struct radeon_device *rdev) radeon_pm_compute_clocks(rdev); } -int radeon_pm_init(struct radeon_device *rdev) +static void radeon_pm_resume_dpm(struct radeon_device *rdev) +{ + int ret; + + /* asic init will reset to the boot state */ + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + rdev->pm.dpm.current_ps = rdev->pm.dpm.requested_ps = rdev->pm.dpm.boot_ps; + radeon_dpm_setup_asic(rdev); + ret = radeon_dpm_enable(rdev); + lockmgr(&rdev->pm.mutex, LK_RELEASE); + if (ret) { + DRM_ERROR("radeon: dpm resume failed\n"); + if ((rdev->family >= CHIP_BARTS) && + (rdev->family <= CHIP_HAINAN) && + rdev->mc_fw) { + if (rdev->pm.default_vddc) + radeon_atom_set_voltage(rdev, rdev->pm.default_vddc, + SET_VOLTAGE_TYPE_ASIC_VDDC); + if (rdev->pm.default_vddci) + radeon_atom_set_voltage(rdev, rdev->pm.default_vddci, + SET_VOLTAGE_TYPE_ASIC_VDDCI); + if (rdev->pm.default_sclk) + radeon_set_engine_clock(rdev, rdev->pm.default_sclk); + if (rdev->pm.default_mclk) + radeon_set_memory_clock(rdev, rdev->pm.default_mclk); + } + } else { + rdev->pm.dpm_enabled = true; + radeon_pm_compute_clocks(rdev); + } +} + +void radeon_pm_resume(struct radeon_device *rdev) +{ + if (rdev->pm.pm_method == PM_METHOD_DPM) + radeon_pm_resume_dpm(rdev); + else + radeon_pm_resume_old(rdev); +} + +static int radeon_pm_init_old(struct radeon_device *rdev) { int ret; - /* default to profile method */ - rdev->pm.pm_method = PM_METHOD_PROFILE; rdev->pm.profile = PM_PROFILE_DEFAULT; rdev->pm.dynpm_state = DYNPM_STATE_DISABLED; rdev->pm.dynpm_planned_action = DYNPM_ACTION_NONE; @@ -619,7 +1031,7 @@ int radeon_pm_init(struct radeon_device *rdev) radeon_pm_init_profile(rdev); /* set up the default clocks if the MC ucode is loaded */ if ((rdev->family >= CHIP_BARTS) && - (rdev->family <= CHIP_CAYMAN) && + (rdev->family <= CHIP_HAINAN) && rdev->mc_fw) { if (rdev->pm.default_vddc) radeon_atom_set_voltage(rdev, rdev->pm.default_vddc, @@ -666,7 +1078,154 @@ int radeon_pm_init(struct radeon_device *rdev) return 0; } -void radeon_pm_fini(struct radeon_device *rdev) +static void radeon_dpm_print_power_states(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + printk("== power state %d ==\n", i); + radeon_dpm_print_power_state(rdev, &rdev->pm.dpm.ps[i]); + } +} + +static int radeon_pm_init_dpm(struct radeon_device *rdev) +{ + int ret; + + /* default to performance state */ + rdev->pm.dpm.state = POWER_STATE_TYPE_BALANCED; + rdev->pm.dpm.user_state = POWER_STATE_TYPE_BALANCED; + rdev->pm.default_sclk = rdev->clock.default_sclk; + rdev->pm.default_mclk = rdev->clock.default_mclk; + rdev->pm.current_sclk = rdev->clock.default_sclk; + rdev->pm.current_mclk = rdev->clock.default_mclk; + rdev->pm.int_thermal_type = THERMAL_TYPE_NONE; + + if (rdev->bios && rdev->is_atom_bios) + radeon_atombios_get_power_modes(rdev); + else + return -EINVAL; + + /* set up the internal thermal sensor if applicable */ + ret = radeon_hwmon_init(rdev); + if (ret) + return ret; + + TASK_INIT(&rdev->pm.dpm.thermal.work, 0, radeon_dpm_thermal_work_handler, rdev); + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + radeon_dpm_init(rdev); + rdev->pm.dpm.current_ps = rdev->pm.dpm.requested_ps = rdev->pm.dpm.boot_ps; + radeon_dpm_print_power_states(rdev); + radeon_dpm_setup_asic(rdev); + ret = radeon_dpm_enable(rdev); + lockmgr(&rdev->pm.mutex, LK_RELEASE); + if (ret) { + rdev->pm.dpm_enabled = false; + if ((rdev->family >= CHIP_BARTS) && + (rdev->family <= CHIP_HAINAN) && + rdev->mc_fw) { + if (rdev->pm.default_vddc) + radeon_atom_set_voltage(rdev, rdev->pm.default_vddc, + SET_VOLTAGE_TYPE_ASIC_VDDC); + if (rdev->pm.default_vddci) + radeon_atom_set_voltage(rdev, rdev->pm.default_vddci, + SET_VOLTAGE_TYPE_ASIC_VDDCI); + if (rdev->pm.default_sclk) + radeon_set_engine_clock(rdev, rdev->pm.default_sclk); + if (rdev->pm.default_mclk) + radeon_set_memory_clock(rdev, rdev->pm.default_mclk); + } + DRM_ERROR("radeon: dpm initialization failed\n"); + return ret; + } + rdev->pm.dpm_enabled = true; + radeon_pm_compute_clocks(rdev); + +#ifdef TODO_DEVICE_FILE + if (rdev->pm.num_power_states > 1) { + ret = device_create_file(rdev->dev, &dev_attr_power_dpm_state); + if (ret) + DRM_ERROR("failed to create device file for dpm state\n"); + ret = device_create_file(rdev->dev, &dev_attr_power_dpm_force_performance_level); + if (ret) + DRM_ERROR("failed to create device file for dpm state\n"); + /* XXX: these are noops for dpm but are here for backwards compat */ + ret = device_create_file(rdev->dev, &dev_attr_power_profile); + if (ret) + DRM_ERROR("failed to create device file for power profile\n"); + ret = device_create_file(rdev->dev, &dev_attr_power_method); + if (ret) + DRM_ERROR("failed to create device file for power method\n"); + + if (radeon_debugfs_pm_init(rdev)) { + DRM_ERROR("Failed to register debugfs file for dpm!\n"); + } + + DRM_INFO("radeon: dpm initialized\n"); + } +#endif + + return 0; +} + +int radeon_pm_init(struct radeon_device *rdev) +{ + /* enable dpm on rv6xx+ */ + switch (rdev->family) { + case CHIP_RV610: + case CHIP_RV630: + case CHIP_RV620: + case CHIP_RV635: + case CHIP_RV670: + case CHIP_RS780: + case CHIP_RS880: + case CHIP_RV770: + case CHIP_RV730: + case CHIP_RV710: + case CHIP_RV740: + case CHIP_CEDAR: + case CHIP_REDWOOD: + case CHIP_JUNIPER: + case CHIP_CYPRESS: + case CHIP_HEMLOCK: + case CHIP_PALM: + case CHIP_SUMO: + case CHIP_SUMO2: + case CHIP_BARTS: + case CHIP_TURKS: + case CHIP_CAICOS: + case CHIP_CAYMAN: + case CHIP_ARUBA: + case CHIP_TAHITI: + case CHIP_PITCAIRN: + case CHIP_VERDE: + case CHIP_OLAND: + case CHIP_HAINAN: + /* DPM requires the RLC, RV770+ dGPU requires SMC */ + if (!rdev->rlc_fw) + rdev->pm.pm_method = PM_METHOD_PROFILE; + else if ((rdev->family >= CHIP_RV770) && + (!(rdev->flags & RADEON_IS_IGP)) && + (!rdev->smc_fw)) + rdev->pm.pm_method = PM_METHOD_PROFILE; + else if (radeon_dpm == 1) + rdev->pm.pm_method = PM_METHOD_DPM; + else + rdev->pm.pm_method = PM_METHOD_PROFILE; + break; + default: + /* default to profile method */ + rdev->pm.pm_method = PM_METHOD_PROFILE; + break; + } + + if (rdev->pm.pm_method == PM_METHOD_DPM) + return radeon_pm_init_dpm(rdev); + else + return radeon_pm_init_old(rdev); +} + +static void radeon_pm_fini_old(struct radeon_device *rdev) { if (rdev->pm.num_power_states > 1) { DRM_UNLOCK(rdev->ddev); /* Work around LOR. */ @@ -706,7 +1265,38 @@ void radeon_pm_fini(struct radeon_device *rdev) radeon_hwmon_fini(rdev); } -void radeon_pm_compute_clocks(struct radeon_device *rdev) +static void radeon_pm_fini_dpm(struct radeon_device *rdev) +{ + if (rdev->pm.num_power_states > 1) { + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + radeon_dpm_disable(rdev); + lockmgr(&rdev->pm.mutex, LK_RELEASE); + +#ifdef TODO_DEVICE_FILE + device_remove_file(rdev->dev, &dev_attr_power_dpm_state); + device_remove_file(rdev->dev, &dev_attr_power_dpm_force_performance_level); + /* XXX backwards compat */ + device_remove_file(rdev->dev, &dev_attr_power_profile); + device_remove_file(rdev->dev, &dev_attr_power_method); +#endif + } + radeon_dpm_fini(rdev); + + if (rdev->pm.power_state) + kfree(rdev->pm.power_state); + + radeon_hwmon_fini(rdev); +} + +void radeon_pm_fini(struct radeon_device *rdev) +{ + if (rdev->pm.pm_method == PM_METHOD_DPM) + radeon_pm_fini_dpm(rdev); + else + radeon_pm_fini_old(rdev); +} + +static void radeon_pm_compute_clocks_old(struct radeon_device *rdev) { struct drm_device *ddev = rdev->ddev; struct drm_crtc *crtc; @@ -785,6 +1375,45 @@ void radeon_pm_compute_clocks(struct radeon_device *rdev) lockmgr(&rdev->pm.mutex, LK_RELEASE); } +static void radeon_pm_compute_clocks_dpm(struct radeon_device *rdev) +{ + struct drm_device *ddev = rdev->ddev; + struct drm_crtc *crtc; + struct radeon_crtc *radeon_crtc; + + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + + /* update active crtc counts */ + rdev->pm.dpm.new_active_crtcs = 0; + rdev->pm.dpm.new_active_crtc_count = 0; + list_for_each_entry(crtc, + &ddev->mode_config.crtc_list, head) { + radeon_crtc = to_radeon_crtc(crtc); + if (crtc->enabled) { + rdev->pm.dpm.new_active_crtcs |= (1 << radeon_crtc->crtc_id); + rdev->pm.dpm.new_active_crtc_count++; + } + } + + /* update battery/ac status */ + if (power_profile_get_state() == POWER_PROFILE_PERFORMANCE) + rdev->pm.dpm.ac_power = true; + else + rdev->pm.dpm.ac_power = false; + + radeon_dpm_change_power_state_locked(rdev); + + lockmgr(&rdev->pm.mutex, LK_RELEASE); +} + +void radeon_pm_compute_clocks(struct radeon_device *rdev) +{ + if (rdev->pm.pm_method == PM_METHOD_DPM) + radeon_pm_compute_clocks_dpm(rdev); + else + radeon_pm_compute_clocks_old(rdev); +} + static bool radeon_pm_in_vbl(struct radeon_device *rdev) { int crtc, vpos, hpos, vbl_status; @@ -890,19 +1519,28 @@ static int radeon_debugfs_pm_info(struct seq_file *m, void *data) struct drm_device *dev = node->minor->dev; struct radeon_device *rdev = dev->dev_private; - seq_printf(m, "default engine clock: %u0 kHz\n", rdev->pm.default_sclk); - /* radeon_get_engine_clock is not reliable on APUs so just print the current clock */ - if ((rdev->family >= CHIP_PALM) && (rdev->flags & RADEON_IS_IGP)) - seq_printf(m, "current engine clock: %u0 kHz\n", rdev->pm.current_sclk); - else - seq_printf(m, "current engine clock: %u0 kHz\n", radeon_get_engine_clock(rdev)); - seq_printf(m, "default memory clock: %u0 kHz\n", rdev->pm.default_mclk); - if (rdev->asic->pm.get_memory_clock) - seq_printf(m, "current memory clock: %u0 kHz\n", radeon_get_memory_clock(rdev)); - if (rdev->pm.current_vddc) - seq_printf(m, "voltage: %u mV\n", rdev->pm.current_vddc); - if (rdev->asic->pm.get_pcie_lanes) - seq_printf(m, "PCIE lanes: %d\n", radeon_get_pcie_lanes(rdev)); + if (rdev->pm.dpm_enabled) { + spin_lock(&rdev->pm.mutex); + if (rdev->asic->dpm.debugfs_print_current_performance_level) + radeon_dpm_debugfs_print_current_performance_level(rdev, m); + else + seq_printf(m, "Debugfs support not implemented for this asic\n"); + spin_unlock(&rdev->pm.mutex); + } else { + seq_printf(m, "default engine clock: %u0 kHz\n", rdev->pm.default_sclk); + /* radeon_get_engine_clock is not reliable on APUs so just print the current clock */ + if ((rdev->family >= CHIP_PALM) && (rdev->flags & RADEON_IS_IGP)) + seq_printf(m, "current engine clock: %u0 kHz\n", rdev->pm.current_sclk); + else + seq_printf(m, "current engine clock: %u0 kHz\n", radeon_get_engine_clock(rdev)); + seq_printf(m, "default memory clock: %u0 kHz\n", rdev->pm.default_mclk); + if (rdev->asic->pm.get_memory_clock) + seq_printf(m, "current memory clock: %u0 kHz\n", radeon_get_memory_clock(rdev)); + if (rdev->pm.current_vddc) + seq_printf(m, "voltage: %u mV\n", rdev->pm.current_vddc); + if (rdev->asic->pm.get_pcie_lanes) + seq_printf(m, "PCIE lanes: %d\n", radeon_get_pcie_lanes(rdev)); + } return 0; } diff --git a/sys/dev/drm/radeon/radeon_prime.c b/sys/dev/drm/radeon/radeon_prime.c index e70dcba74d..65b9eabd5a 100644 --- a/sys/dev/drm/radeon/radeon_prime.c +++ b/sys/dev/drm/radeon/radeon_prime.c @@ -22,10 +22,7 @@ * based on nouveau_prime.c * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_prime.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include "radeon.h" @@ -91,11 +88,19 @@ int radeon_gem_prime_pin(struct drm_gem_object *obj) /* pin buffer into GTT */ ret = radeon_bo_pin(bo, RADEON_GEM_DOMAIN_GTT, NULL); - if (ret) { - radeon_bo_unreserve(bo); - return ret; - } radeon_bo_unreserve(bo); + return ret; +} + +void radeon_gem_prime_unpin(struct drm_gem_object *obj) +{ + struct radeon_bo *bo = gem_to_radeon_bo(obj); + int ret = 0; + + ret = radeon_bo_reserve(bo, false); + if (unlikely(ret != 0)) + return; - return 0; + radeon_bo_unpin(bo); + radeon_bo_unreserve(bo); } diff --git a/sys/dev/drm/radeon/radeon_reg.h b/sys/dev/drm/radeon/radeon_reg.h index 2eea531581..62d54976d2 100644 --- a/sys/dev/drm/radeon/radeon_reg.h +++ b/sys/dev/drm/radeon/radeon_reg.h @@ -24,8 +24,6 @@ * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. - * - * $FreeBSD: head/sys/dev/drm2/radeon/radeon_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ /* @@ -50,7 +48,6 @@ /* !!!! FIXME !!!! NOTE: THIS FILE HAS BEEN CONVERTED FROM r128_reg.h * AND CONTAINS REGISTERS AND REGISTER DEFINITIONS THAT ARE NOT CORRECT * ON THE RADEON. A FULL AUDIT OF THIS CODE IS NEEDED! */ - #ifndef _RADEON_REG_H_ #define _RADEON_REG_H_ @@ -60,6 +57,7 @@ #include "evergreen_reg.h" #include "ni_reg.h" #include "si_reg.h" +#include "cik_reg.h" #define RADEON_MC_AGP_LOCATION 0x014c #define RADEON_MC_AGP_START_MASK 0x0000FFFF diff --git a/sys/dev/drm/radeon/radeon_ring.c b/sys/dev/drm/radeon/radeon_ring.c index 3b88d0797d..39da3ef173 100644 --- a/sys/dev/drm/radeon/radeon_ring.c +++ b/sys/dev/drm/radeon/radeon_ring.c @@ -226,6 +226,7 @@ int radeon_ib_pool_init(struct radeon_device *rdev) } r = radeon_sa_bo_manager_init(rdev, &rdev->ring_tmp_bo, RADEON_IB_POOL_SIZE*64*1024, + RADEON_GPU_PAGE_SIZE, RADEON_GEM_DOMAIN_GTT); if (r) { return r; @@ -365,6 +366,38 @@ bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev, } } +u32 radeon_ring_generic_get_rptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + u32 rptr; + + if (rdev->wb.enabled && ring != &rdev->ring[R600_RING_TYPE_UVD_INDEX]) + rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]); + else + rptr = RREG32(ring->rptr_reg); + rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + + return rptr; +} + +u32 radeon_ring_generic_get_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + u32 wptr; + + wptr = RREG32(ring->wptr_reg); + wptr = (wptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + + return wptr; +} + +void radeon_ring_generic_set_wptr(struct radeon_device *rdev, + struct radeon_ring *ring) +{ + WREG32(ring->wptr_reg, (ring->wptr << ring->ptr_reg_shift) & ring->ptr_reg_mask); + (void)RREG32(ring->wptr_reg); +} + /** * radeon_ring_free_size - update the free size * @@ -375,13 +408,7 @@ bool radeon_ring_supports_scratch_reg(struct radeon_device *rdev, */ void radeon_ring_free_size(struct radeon_device *rdev, struct radeon_ring *ring) { - u32 rptr; - - if (rdev->wb.enabled && ring != &rdev->ring[R600_RING_TYPE_UVD_INDEX]) - rptr = le32_to_cpu(rdev->wb.wb[ring->rptr_offs/4]); - else - rptr = RREG32(ring->rptr_reg); - ring->rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + ring->rptr = radeon_ring_get_rptr(rdev, ring); /* This works because ring_size is a power of 2 */ ring->ring_free_dw = (ring->rptr + (ring->ring_size / 4)); ring->ring_free_dw -= ring->wptr; @@ -473,8 +500,7 @@ void radeon_ring_commit(struct radeon_device *rdev, struct radeon_ring *ring) radeon_ring_write(ring, ring->nop); } DRM_MEMORYBARRIER(); - WREG32(ring->wptr_reg, (ring->wptr << ring->ptr_reg_shift) & ring->ptr_reg_mask); - (void)RREG32(ring->wptr_reg); + radeon_ring_set_wptr(rdev, ring); } /** @@ -576,7 +602,6 @@ void radeon_ring_lockup_update(struct radeon_ring *ring) bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *ring) { unsigned long cjiffies, elapsed; - uint32_t rptr; cjiffies = jiffies; if (!time_after(cjiffies, ring->last_activity)) { @@ -584,8 +609,7 @@ bool radeon_ring_test_lockup(struct radeon_device *rdev, struct radeon_ring *rin radeon_ring_lockup_update(ring); return false; } - rptr = RREG32(ring->rptr_reg); - ring->rptr = (rptr & ring->ptr_reg_mask) >> ring->ptr_reg_shift; + ring->rptr = radeon_ring_get_rptr(rdev, ring); if (ring->rptr != ring->last_rptr) { /* CP is still working no lockup */ radeon_ring_lockup_update(ring); @@ -820,9 +844,9 @@ static int radeon_debugfs_ring_info(struct seq_file *m, void *data) radeon_ring_free_size(rdev, ring); count = (ring->ring_size / 4) - ring->ring_free_dw; - tmp = RREG32(ring->wptr_reg) >> ring->ptr_reg_shift; + tmp = radeon_ring_get_wptr(rdev, ring); seq_printf(m, "wptr(0x%04x): 0x%08x [%5d]\n", ring->wptr_reg, tmp, tmp); - tmp = RREG32(ring->rptr_reg) >> ring->ptr_reg_shift; + tmp = radeon_ring_get_rptr(rdev, ring); seq_printf(m, "rptr(0x%04x): 0x%08x [%5d]\n", ring->rptr_reg, tmp, tmp); if (ring->rptr_save_reg) { seq_printf(m, "rptr next(0x%04x): 0x%08x\n", ring->rptr_save_reg, diff --git a/sys/dev/drm/radeon/radeon_sa.c b/sys/dev/drm/radeon/radeon_sa.c index da9e92cfcf..19ca568480 100644 --- a/sys/dev/drm/radeon/radeon_sa.c +++ b/sys/dev/drm/radeon/radeon_sa.c @@ -52,7 +52,7 @@ static void radeon_sa_bo_try_free(struct radeon_sa_manager *sa_manager); int radeon_sa_bo_manager_init(struct radeon_device *rdev, struct radeon_sa_manager *sa_manager, - unsigned size, u32 domain) + unsigned size, u32 align, u32 domain) { int i, r; @@ -62,13 +62,14 @@ int radeon_sa_bo_manager_init(struct radeon_device *rdev, sa_manager->bo = NULL; sa_manager->size = size; sa_manager->domain = domain; + sa_manager->align = align; sa_manager->hole = &sa_manager->olist; INIT_LIST_HEAD(&sa_manager->olist); for (i = 0; i < RADEON_NUM_RINGS; ++i) { INIT_LIST_HEAD(&sa_manager->flist[i]); } - r = radeon_bo_create(rdev, size, RADEON_GPU_PAGE_SIZE, true, + r = radeon_bo_create(rdev, size, align, true, domain, NULL, &sa_manager->bo); if (r) { dev_err(rdev->dev, "(%d) failed to allocate bo for manager\n", r); @@ -324,8 +325,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev, unsigned tries[RADEON_NUM_RINGS]; int i, r; - KASSERT(align <= RADEON_GPU_PAGE_SIZE, ("align > RADEON_GPU_PAGE_SIZE")); - KASSERT(size <= sa_manager->size, ("size > sa_manager->size")); + BUG_ON(align > sa_manager->align); + BUG_ON(size > sa_manager->size); *sa_bo = kmalloc(sizeof(struct radeon_sa_bo), M_DRM, M_WAITOK | M_ZERO); diff --git a/sys/dev/drm/radeon/radeon_trace.h b/sys/dev/drm/radeon/radeon_trace.h index 1fde2905e9..eafd8160a1 100644 --- a/sys/dev/drm/radeon/radeon_trace.h +++ b/sys/dev/drm/radeon/radeon_trace.h @@ -1,5 +1,3 @@ -/* $FreeBSD: head/sys/dev/drm2/radeon/radeon_trace.h 254885 2013-08-25 19:37:15Z dumbbell $ */ - #if !defined(_RADEON_TRACE_H) || defined(TRACE_HEADER_MULTI_READ) #define _RADEON_TRACE_H_ diff --git a/sys/dev/drm/radeon/radeon_trace_points.c b/sys/dev/drm/radeon/radeon_trace_points.c index 9a06a320c4..e51d357597 100644 --- a/sys/dev/drm/radeon/radeon_trace_points.c +++ b/sys/dev/drm/radeon/radeon_trace_points.c @@ -1,9 +1,6 @@ /* Copyright Red Hat Inc 2010. * Author : Dave Airlie - * - $FreeBSD: head/sys/dev/drm2/radeon/radeon_trace_points.c 254885 2013-08-25 19:37:15Z dumbbell $ */ - #include #include #include "radeon.h" diff --git a/sys/dev/drm/radeon/radeon_ucode.h b/sys/dev/drm/radeon/radeon_ucode.h new file mode 100644 index 0000000000..d8b05f7bcf --- /dev/null +++ b/sys/dev/drm/radeon/radeon_ucode.h @@ -0,0 +1,129 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __RADEON_UCODE_H__ +#define __RADEON_UCODE_H__ + +/* CP */ +#define R600_PFP_UCODE_SIZE 576 +#define R600_PM4_UCODE_SIZE 1792 +#define R700_PFP_UCODE_SIZE 848 +#define R700_PM4_UCODE_SIZE 1360 +#define EVERGREEN_PFP_UCODE_SIZE 1120 +#define EVERGREEN_PM4_UCODE_SIZE 1376 +#define CAYMAN_PFP_UCODE_SIZE 2176 +#define CAYMAN_PM4_UCODE_SIZE 2176 +#define SI_PFP_UCODE_SIZE 2144 +#define SI_PM4_UCODE_SIZE 2144 +#define SI_CE_UCODE_SIZE 2144 + +/* RLC */ +#define R600_RLC_UCODE_SIZE 768 +#define R700_RLC_UCODE_SIZE 1024 +#define EVERGREEN_RLC_UCODE_SIZE 768 +#define CAYMAN_RLC_UCODE_SIZE 1024 +#define ARUBA_RLC_UCODE_SIZE 1536 +#define SI_RLC_UCODE_SIZE 2048 + +/* MC */ +#define BTC_MC_UCODE_SIZE 6024 +#define CAYMAN_MC_UCODE_SIZE 6037 +#define SI_MC_UCODE_SIZE 7769 +#define OLAND_MC_UCODE_SIZE 7863 + +/* SMC */ +#define RV770_SMC_UCODE_START 0x0100 +#define RV770_SMC_UCODE_SIZE 0x410d +#define RV770_SMC_INT_VECTOR_START 0xffc0 +#define RV770_SMC_INT_VECTOR_SIZE 0x0040 + +#define RV730_SMC_UCODE_START 0x0100 +#define RV730_SMC_UCODE_SIZE 0x412c +#define RV730_SMC_INT_VECTOR_START 0xffc0 +#define RV730_SMC_INT_VECTOR_SIZE 0x0040 + +#define RV710_SMC_UCODE_START 0x0100 +#define RV710_SMC_UCODE_SIZE 0x3f1f +#define RV710_SMC_INT_VECTOR_START 0xffc0 +#define RV710_SMC_INT_VECTOR_SIZE 0x0040 + +#define RV740_SMC_UCODE_START 0x0100 +#define RV740_SMC_UCODE_SIZE 0x41c5 +#define RV740_SMC_INT_VECTOR_START 0xffc0 +#define RV740_SMC_INT_VECTOR_SIZE 0x0040 + +#define CEDAR_SMC_UCODE_START 0x0100 +#define CEDAR_SMC_UCODE_SIZE 0x5d50 +#define CEDAR_SMC_INT_VECTOR_START 0xffc0 +#define CEDAR_SMC_INT_VECTOR_SIZE 0x0040 + +#define REDWOOD_SMC_UCODE_START 0x0100 +#define REDWOOD_SMC_UCODE_SIZE 0x5f0a +#define REDWOOD_SMC_INT_VECTOR_START 0xffc0 +#define REDWOOD_SMC_INT_VECTOR_SIZE 0x0040 + +#define JUNIPER_SMC_UCODE_START 0x0100 +#define JUNIPER_SMC_UCODE_SIZE 0x5f1f +#define JUNIPER_SMC_INT_VECTOR_START 0xffc0 +#define JUNIPER_SMC_INT_VECTOR_SIZE 0x0040 + +#define CYPRESS_SMC_UCODE_START 0x0100 +#define CYPRESS_SMC_UCODE_SIZE 0x61f7 +#define CYPRESS_SMC_INT_VECTOR_START 0xffc0 +#define CYPRESS_SMC_INT_VECTOR_SIZE 0x0040 + +#define BARTS_SMC_UCODE_START 0x0100 +#define BARTS_SMC_UCODE_SIZE 0x6107 +#define BARTS_SMC_INT_VECTOR_START 0xffc0 +#define BARTS_SMC_INT_VECTOR_SIZE 0x0040 + +#define TURKS_SMC_UCODE_START 0x0100 +#define TURKS_SMC_UCODE_SIZE 0x605b +#define TURKS_SMC_INT_VECTOR_START 0xffc0 +#define TURKS_SMC_INT_VECTOR_SIZE 0x0040 + +#define CAICOS_SMC_UCODE_START 0x0100 +#define CAICOS_SMC_UCODE_SIZE 0x5fbd +#define CAICOS_SMC_INT_VECTOR_START 0xffc0 +#define CAICOS_SMC_INT_VECTOR_SIZE 0x0040 + +#define CAYMAN_SMC_UCODE_START 0x0100 +#define CAYMAN_SMC_UCODE_SIZE 0x79ec +#define CAYMAN_SMC_INT_VECTOR_START 0xffc0 +#define CAYMAN_SMC_INT_VECTOR_SIZE 0x0040 + +#define TAHITI_SMC_UCODE_START 0x10000 +#define TAHITI_SMC_UCODE_SIZE 0xf458 + +#define PITCAIRN_SMC_UCODE_START 0x10000 +#define PITCAIRN_SMC_UCODE_SIZE 0xe9f4 + +#define VERDE_SMC_UCODE_START 0x10000 +#define VERDE_SMC_UCODE_SIZE 0xebe4 + +#define OLAND_SMC_UCODE_START 0x10000 +#define OLAND_SMC_UCODE_SIZE 0xe7b4 + +#define HAINAN_SMC_UCODE_START 0x10000 +#define HAINAN_SMC_UCODE_SIZE 0xe67C + +#endif diff --git a/sys/dev/drm/radeon/radeon_uvd.c b/sys/dev/drm/radeon/radeon_uvd.c index 397120820e..61b69b09c0 100644 --- a/sys/dev/drm/radeon/radeon_uvd.c +++ b/sys/dev/drm/radeon/radeon_uvd.c @@ -29,6 +29,7 @@ */ #include +#include #include #include "radeon.h" @@ -42,13 +43,13 @@ #define FIRMWARE_CYPRESS "radeonkmsfw_CYPRESS_uvd" #define FIRMWARE_SUMO "radeonkmsfw_SUMO_uvd" #define FIRMWARE_TAHITI "radeonkmsfw_TAHITI_uvd" +#define FIRMWARE_BONAIRE "radeonkmsfw_BONAIRE_uvd" -#if 0 MODULE_FIRMWARE(FIRMWARE_RV710); MODULE_FIRMWARE(FIRMWARE_CYPRESS); MODULE_FIRMWARE(FIRMWARE_SUMO); MODULE_FIRMWARE(FIRMWARE_TAHITI); -#endif +MODULE_FIRMWARE(FIRMWARE_BONAIRE); static void radeon_uvd_idle_work_handler(struct work_struct *work); @@ -92,12 +93,18 @@ int radeon_uvd_init(struct radeon_device *rdev) fw_name = FIRMWARE_TAHITI; break; + case CHIP_BONAIRE: + case CHIP_KABINI: + case CHIP_KAVERI: + fw_name = FIRMWARE_BONAIRE; + break; + default: return -EINVAL; } - rdev->uvd_fw = firmware_get(fw_name); - if (!rdev->uvd_fw) { + r = request_firmware(&rdev->uvd_fw, fw_name, rdev->dev); + if (r) { dev_err(rdev->dev, "radeon_uvd: Can't load firmware \"%s\"\n", fw_name); return -EINVAL; @@ -112,16 +119,29 @@ int radeon_uvd_init(struct radeon_device *rdev) return r; } - r = radeon_uvd_resume(rdev); - if (r) + r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); + if (r) { + radeon_bo_unref(&rdev->uvd.vcpu_bo); + dev_err(rdev->dev, "(%d) failed to reserve UVD bo\n", r); return r; + } - memset(rdev->uvd.cpu_addr, 0, bo_size); - memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->datasize); + r = radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, + &rdev->uvd.gpu_addr); + if (r) { + radeon_bo_unreserve(rdev->uvd.vcpu_bo); + radeon_bo_unref(&rdev->uvd.vcpu_bo); + dev_err(rdev->dev, "(%d) UVD bo pin failed\n", r); + return r; + } - r = radeon_uvd_suspend(rdev); - if (r) + r = radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr); + if (r) { + dev_err(rdev->dev, "(%d) UVD map failed\n", r); return r; + } + + radeon_bo_unreserve(rdev->uvd.vcpu_bo); for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { atomic_set(&rdev->uvd.handles[i], 0); @@ -132,71 +152,74 @@ int radeon_uvd_init(struct radeon_device *rdev) } void radeon_uvd_fini(struct radeon_device *rdev) -{ - radeon_uvd_suspend(rdev); - radeon_bo_unref(&rdev->uvd.vcpu_bo); -} - -int radeon_uvd_suspend(struct radeon_device *rdev) { int r; if (rdev->uvd.vcpu_bo == NULL) - return 0; + return; r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); if (!r) { radeon_bo_kunmap(rdev->uvd.vcpu_bo); radeon_bo_unpin(rdev->uvd.vcpu_bo); - rdev->uvd.cpu_addr = NULL; - if (!radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_CPU, NULL)) { - radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr); - } radeon_bo_unreserve(rdev->uvd.vcpu_bo); - - if (rdev->uvd.cpu_addr) { - radeon_fence_driver_start_ring(rdev, R600_RING_TYPE_UVD_INDEX); - } else { - rdev->fence_drv[R600_RING_TYPE_UVD_INDEX].cpu_addr = NULL; - } } - return r; + + radeon_bo_unref(&rdev->uvd.vcpu_bo); + + release_firmware(rdev->uvd_fw); +} + +int radeon_uvd_suspend(struct radeon_device *rdev) +{ + unsigned size; + char *ptr; + int i; + + if (rdev->uvd.vcpu_bo == NULL) + return 0; + + for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) + if (atomic_read(&rdev->uvd.handles[i])) + break; + + if (i == RADEON_MAX_UVD_HANDLES) + return 0; + + size = radeon_bo_size(rdev->uvd.vcpu_bo); + size -= rdev->uvd_fw->datasize; + + ptr = rdev->uvd.cpu_addr; + ptr += rdev->uvd_fw->datasize; + + rdev->uvd.saved_bo = kmalloc(size, M_DRM, M_WAITOK); + memcpy(rdev->uvd.saved_bo, ptr, size); + + return 0; } int radeon_uvd_resume(struct radeon_device *rdev) { - int r; + unsigned size; + char *ptr; if (rdev->uvd.vcpu_bo == NULL) return -EINVAL; - r = radeon_bo_reserve(rdev->uvd.vcpu_bo, false); - if (r) { - radeon_bo_unref(&rdev->uvd.vcpu_bo); - dev_err(rdev->dev, "(%d) failed to reserve UVD bo\n", r); - return r; - } + memcpy(rdev->uvd.cpu_addr, rdev->uvd_fw->data, rdev->uvd_fw->datasize); - /* Have been pin in cpu unmap unpin */ - radeon_bo_kunmap(rdev->uvd.vcpu_bo); - radeon_bo_unpin(rdev->uvd.vcpu_bo); + size = radeon_bo_size(rdev->uvd.vcpu_bo); + size -= rdev->uvd_fw->datasize; - r = radeon_bo_pin(rdev->uvd.vcpu_bo, RADEON_GEM_DOMAIN_VRAM, - &rdev->uvd.gpu_addr); - if (r) { - radeon_bo_unreserve(rdev->uvd.vcpu_bo); - radeon_bo_unref(&rdev->uvd.vcpu_bo); - dev_err(rdev->dev, "(%d) UVD bo pin failed\n", r); - return r; - } + ptr = rdev->uvd.cpu_addr; + ptr += rdev->uvd_fw->datasize; - r = radeon_bo_kmap(rdev->uvd.vcpu_bo, &rdev->uvd.cpu_addr); - if (r) { - dev_err(rdev->dev, "(%d) UVD map failed\n", r); - return r; - } - - radeon_bo_unreserve(rdev->uvd.vcpu_bo); + if (rdev->uvd.saved_bo != NULL) { + memcpy(ptr, rdev->uvd.saved_bo, size); + kfree(rdev->uvd.saved_bo); + rdev->uvd.saved_bo = NULL; + } else + memset(ptr, 0, size); return 0; } @@ -211,8 +234,8 @@ void radeon_uvd_free_handles(struct radeon_device *rdev, struct drm_file *filp) { int i, r; for (i = 0; i < RADEON_MAX_UVD_HANDLES; ++i) { - if (rdev->uvd.filp[i] == filp) { - uint32_t handle = atomic_read(&rdev->uvd.handles[i]); + uint32_t handle = atomic_read(&rdev->uvd.handles[i]); + if (handle != 0 && rdev->uvd.filp[i] == filp) { struct radeon_fence *fence; r = radeon_uvd_get_destroy_msg(rdev, @@ -332,9 +355,19 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, return -EINVAL; } + if (bo->tbo.sync_obj) { + r = radeon_fence_wait(bo->tbo.sync_obj, false); + if (r) { + DRM_ERROR("Failed waiting for UVD message (%d)!\n", r); + return r; + } + } + r = radeon_bo_kmap(bo, &ptr); - if (r) + if (r) { + DRM_ERROR("Failed mapping the UVD message (%d)!\n", r); return r; + } msg = (uint32_t*)((uint8_t*)ptr + offset); @@ -360,8 +393,14 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, radeon_bo_kunmap(bo); return 0; } else { - /* it's a create msg, no special handling needed */ radeon_bo_kunmap(bo); + + if (msg_type != 0) { + DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type); + return -EINVAL; + } + + /* it's a create msg, no special handling needed */ } /* create or decode, validate the handle */ @@ -387,7 +426,7 @@ static int radeon_uvd_cs_msg(struct radeon_cs_parser *p, struct radeon_bo *bo, static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, int data0, int data1, - unsigned buf_sizes[]) + unsigned buf_sizes[], bool *has_msg_cmd) { struct radeon_cs_chunk *relocs_chunk; struct radeon_cs_reloc *reloc; @@ -416,7 +455,7 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, if (cmd < 0x4) { if ((end - start) < buf_sizes[cmd]) { - DRM_ERROR("buffer to small (%d / %d)!\n", + DRM_ERROR("buffer (%d) to small (%d / %d)!\n", cmd, (unsigned)(end - start), buf_sizes[cmd]); return -EINVAL; } @@ -441,9 +480,17 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, } if (cmd == 0) { + if (*has_msg_cmd) { + DRM_ERROR("More than one message in a UVD-IB!\n"); + return -EINVAL; + } + *has_msg_cmd = true; r = radeon_uvd_cs_msg(p, reloc->robj, offset, buf_sizes); if (r) return r; + } else if (!*has_msg_cmd) { + DRM_ERROR("Message needed before other commands are send!\n"); + return -EINVAL; } return 0; @@ -452,7 +499,8 @@ static int radeon_uvd_cs_reloc(struct radeon_cs_parser *p, static int radeon_uvd_cs_reg(struct radeon_cs_parser *p, struct radeon_cs_packet *pkt, int *data0, int *data1, - unsigned buf_sizes[]) + unsigned buf_sizes[], + bool *has_msg_cmd) { int i, r; @@ -466,7 +514,8 @@ static int radeon_uvd_cs_reg(struct radeon_cs_parser *p, *data1 = p->idx; break; case UVD_GPCOM_VCPU_CMD: - r = radeon_uvd_cs_reloc(p, *data0, *data1, buf_sizes); + r = radeon_uvd_cs_reloc(p, *data0, *data1, + buf_sizes, has_msg_cmd); if (r) return r; break; @@ -487,6 +536,9 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) struct radeon_cs_packet pkt; int r, data0 = 0, data1 = 0; + /* does the IB has a msg command */ + bool has_msg_cmd = false; + /* minimum buffer sizes */ unsigned buf_sizes[] = { [0x00000000] = 2048, @@ -513,8 +565,8 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) return r; switch (pkt.type) { case RADEON_PACKET_TYPE0: - r = radeon_uvd_cs_reg(p, &pkt, &data0, - &data1, buf_sizes); + r = radeon_uvd_cs_reg(p, &pkt, &data0, &data1, + buf_sizes, &has_msg_cmd); if (r) return r; break; @@ -526,6 +578,12 @@ int radeon_uvd_cs_parse(struct radeon_cs_parser *p) return -EINVAL; } } while (p->idx < p->chunks[p->chunk_ib_idx].length_dw); + + if (!has_msg_cmd) { + DRM_ERROR("UVD-IBs need a msg command!\n"); + return -EINVAL; + } + return 0; } @@ -683,11 +741,19 @@ static void radeon_uvd_idle_work_handler(struct work_struct *work) struct radeon_device *rdev = container_of(work, struct radeon_device, uvd.idle_work.work); - if (radeon_fence_count_emitted(rdev, R600_RING_TYPE_UVD_INDEX) == 0) - radeon_set_uvd_clocks(rdev, 0, 0); - else + if (radeon_fence_count_emitted(rdev, R600_RING_TYPE_UVD_INDEX) == 0) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + lockmgr(&rdev->pm.mutex, LK_EXCLUSIVE); + rdev->pm.dpm.uvd_active = false; + lockmgr(&rdev->pm.mutex, LK_RELEASE); + radeon_pm_compute_clocks(rdev); + } else { + radeon_set_uvd_clocks(rdev, 0, 0); + } + } else { schedule_delayed_work(&rdev->uvd.idle_work, msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); + } } void radeon_uvd_note_usage(struct radeon_device *rdev) @@ -695,8 +761,14 @@ void radeon_uvd_note_usage(struct radeon_device *rdev) bool set_clocks = !cancel_delayed_work_sync(&rdev->uvd.idle_work); set_clocks &= schedule_delayed_work(&rdev->uvd.idle_work, msecs_to_jiffies(UVD_IDLE_TIMEOUT_MS)); - if (set_clocks) - radeon_set_uvd_clocks(rdev, 53300, 40000); + if (set_clocks) { + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + /* XXX pick SD/HD/MVC */ + radeon_dpm_enable_power_state(rdev, POWER_STATE_TYPE_INTERNAL_UVD); + } else { + radeon_set_uvd_clocks(rdev, 53300, 40000); + } + } } static unsigned radeon_uvd_calc_upll_post_div(unsigned vco_freq, diff --git a/sys/dev/drm/radeon/rs100d.h b/sys/dev/drm/radeon/rs100d.h index e45ec1aab1..48a913a06c 100644 --- a/sys/dev/drm/radeon/rs100d.h +++ b/sys/dev/drm/radeon/rs100d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rs100d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RS100D_H__ #define __RS100D_H__ diff --git a/sys/dev/drm/radeon/rs400d.h b/sys/dev/drm/radeon/rs400d.h index 31328ca621..6d8bac58ce 100644 --- a/sys/dev/drm/radeon/rs400d.h +++ b/sys/dev/drm/radeon/rs400d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rs400d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RS400D_H__ #define __RS400D_H__ diff --git a/sys/dev/drm/radeon/rs600d.h b/sys/dev/drm/radeon/rs600d.h index 9d03cebf04..f1f89414dc 100644 --- a/sys/dev/drm/radeon/rs600d.h +++ b/sys/dev/drm/radeon/rs600d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rs600d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RS600D_H__ #define __RS600D_H__ diff --git a/sys/dev/drm/radeon/rs690.c b/sys/dev/drm/radeon/rs690.c index 17b95f4e5f..1cbb9a478c 100644 --- a/sys/dev/drm/radeon/rs690.c +++ b/sys/dev/drm/radeon/rs690.c @@ -251,13 +251,16 @@ struct rs690_watermark { }; static void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, - struct radeon_crtc *crtc, - struct rs690_watermark *wm) + struct radeon_crtc *crtc, + struct rs690_watermark *wm, + bool low) { struct drm_display_mode *mode = &crtc->base.mode; fixed20_12 a, b, c; fixed20_12 pclk, request_fifo_depth, tolerable_latency, estimated_width; fixed20_12 consumption_time, line_time, chunk_time, read_delay_latency; + fixed20_12 sclk, core_bandwidth, max_bandwidth; + u32 selected_sclk; if (!crtc->base.enabled) { /* FIXME: wouldn't it better to set priority mark to maximum */ @@ -265,6 +268,21 @@ static void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, return; } + if (((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880)) && + (rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) + selected_sclk = radeon_dpm_get_sclk(rdev, low); + else + selected_sclk = rdev->pm.current_sclk; + + /* sclk in Mhz */ + a.full = dfixed_const(100); + sclk.full = dfixed_const(selected_sclk); + sclk.full = dfixed_div(sclk, a); + + /* core_bandwidth = sclk(Mhz) * 16 */ + a.full = dfixed_const(16); + core_bandwidth.full = dfixed_div(rdev->pm.sclk, a); + if (crtc->vsc.full > dfixed_const(2)) wm->num_line_pair.full = dfixed_const(2); else @@ -325,11 +343,11 @@ static void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, wm->active_time.full = dfixed_div(wm->active_time, a); /* Maximun bandwidth is the minimun bandwidth of all component */ - rdev->pm.max_bandwidth = rdev->pm.core_bandwidth; + max_bandwidth = core_bandwidth; if (rdev->mc.igp_sideport_enabled) { - if (rdev->pm.max_bandwidth.full > rdev->pm.sideport_bandwidth.full && + if (max_bandwidth.full > rdev->pm.sideport_bandwidth.full && rdev->pm.sideport_bandwidth.full) - rdev->pm.max_bandwidth = rdev->pm.sideport_bandwidth; + max_bandwidth = rdev->pm.sideport_bandwidth; #ifdef DUMBBELL_WIP read_delay_latency.full = dfixed_const(370 * 800 * 1000); #endif /* DUMBBELL_WIP */ @@ -337,27 +355,27 @@ static void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, read_delay_latency.full = dfixed_div(read_delay_latency, rdev->pm.igp_sideport_mclk); } else { - if (rdev->pm.max_bandwidth.full > rdev->pm.k8_bandwidth.full && + if (max_bandwidth.full > rdev->pm.k8_bandwidth.full && rdev->pm.k8_bandwidth.full) - rdev->pm.max_bandwidth = rdev->pm.k8_bandwidth; - if (rdev->pm.max_bandwidth.full > rdev->pm.ht_bandwidth.full && + max_bandwidth = rdev->pm.k8_bandwidth; + if (max_bandwidth.full > rdev->pm.ht_bandwidth.full && rdev->pm.ht_bandwidth.full) - rdev->pm.max_bandwidth = rdev->pm.ht_bandwidth; + max_bandwidth = rdev->pm.ht_bandwidth; read_delay_latency.full = dfixed_const(5000); } /* sclk = system clocks(ns) = 1000 / max_bandwidth / 16 */ a.full = dfixed_const(16); - rdev->pm.sclk.full = dfixed_mul(rdev->pm.max_bandwidth, a); + sclk.full = dfixed_mul(max_bandwidth, a); a.full = dfixed_const(1000); - rdev->pm.sclk.full = dfixed_div(a, rdev->pm.sclk); + sclk.full = dfixed_div(a, sclk); /* Determine chunk time * ChunkTime = the time it takes the DCP to send one chunk of data * to the LB which consists of pipeline delay and inter chunk gap * sclk = system clock(ns) */ a.full = dfixed_const(256 * 13); - chunk_time.full = dfixed_mul(rdev->pm.sclk, a); + chunk_time.full = dfixed_mul(sclk, a); a.full = dfixed_const(10); chunk_time.full = dfixed_div(chunk_time, a); @@ -421,175 +439,200 @@ static void rs690_crtc_bandwidth_compute(struct radeon_device *rdev, } } -void rs690_bandwidth_update(struct radeon_device *rdev) +static void rs690_compute_mode_priority(struct radeon_device *rdev, + struct rs690_watermark *wm0, + struct rs690_watermark *wm1, + struct drm_display_mode *mode0, + struct drm_display_mode *mode1, + u32 *d1mode_priority_a_cnt, + u32 *d2mode_priority_a_cnt) { - struct drm_display_mode *mode0 = NULL; - struct drm_display_mode *mode1 = NULL; - struct rs690_watermark wm0; - struct rs690_watermark wm1; - u32 tmp; - u32 d1mode_priority_a_cnt = S_006548_D1MODE_PRIORITY_A_OFF(1); - u32 d2mode_priority_a_cnt = S_006548_D1MODE_PRIORITY_A_OFF(1); fixed20_12 priority_mark02, priority_mark12, fill_rate; fixed20_12 a, b; - radeon_update_display_priority(rdev); - - if (rdev->mode_info.crtcs[0]->base.enabled) - mode0 = &rdev->mode_info.crtcs[0]->base.mode; - if (rdev->mode_info.crtcs[1]->base.enabled) - mode1 = &rdev->mode_info.crtcs[1]->base.mode; - /* - * Set display0/1 priority up in the memory controller for - * modes if the user specifies HIGH for displaypriority - * option. - */ - if ((rdev->disp_priority == 2) && - ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740))) { - tmp = RREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER); - tmp &= C_000104_MC_DISP0R_INIT_LAT; - tmp &= C_000104_MC_DISP1R_INIT_LAT; - if (mode0) - tmp |= S_000104_MC_DISP0R_INIT_LAT(1); - if (mode1) - tmp |= S_000104_MC_DISP1R_INIT_LAT(1); - WREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER, tmp); - } - rs690_line_buffer_adjust(rdev, mode0, mode1); - - if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740)) - WREG32(R_006C9C_DCP_CONTROL, 0); - if ((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880)) - WREG32(R_006C9C_DCP_CONTROL, 2); - - rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0); - rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1); - - tmp = (wm0.lb_request_fifo_depth - 1); - tmp |= (wm1.lb_request_fifo_depth - 1) << 16; - WREG32(R_006D58_LB_MAX_REQ_OUTSTANDING, tmp); + *d1mode_priority_a_cnt = S_006548_D1MODE_PRIORITY_A_OFF(1); + *d2mode_priority_a_cnt = S_006548_D1MODE_PRIORITY_A_OFF(1); if (mode0 && mode1) { - if (dfixed_trunc(wm0.dbpp) > 64) - a.full = dfixed_mul(wm0.dbpp, wm0.num_line_pair); + if (dfixed_trunc(wm0->dbpp) > 64) + a.full = dfixed_mul(wm0->dbpp, wm0->num_line_pair); else - a.full = wm0.num_line_pair.full; - if (dfixed_trunc(wm1.dbpp) > 64) - b.full = dfixed_mul(wm1.dbpp, wm1.num_line_pair); + a.full = wm0->num_line_pair.full; + if (dfixed_trunc(wm1->dbpp) > 64) + b.full = dfixed_mul(wm1->dbpp, wm1->num_line_pair); else - b.full = wm1.num_line_pair.full; + b.full = wm1->num_line_pair.full; a.full += b.full; - fill_rate.full = dfixed_div(wm0.sclk, a); - if (wm0.consumption_rate.full > fill_rate.full) { - b.full = wm0.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm0.active_time); - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + fill_rate.full = dfixed_div(wm0->sclk, a); + if (wm0->consumption_rate.full > fill_rate.full) { + b.full = wm0->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm0->active_time); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); a.full = a.full + b.full; b.full = dfixed_const(16 * 1000); priority_mark02.full = dfixed_div(a, b); } else { - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark02.full = dfixed_div(a, b); } - if (wm1.consumption_rate.full > fill_rate.full) { - b.full = wm1.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm1.active_time); - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + if (wm1->consumption_rate.full > fill_rate.full) { + b.full = wm1->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm1->active_time); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); a.full = a.full + b.full; b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } else { - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } - if (wm0.priority_mark.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark.full; + if (wm0->priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark.full; if (dfixed_trunc(priority_mark02) < 0) priority_mark02.full = 0; - if (wm0.priority_mark_max.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark_max.full; - if (wm1.priority_mark.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark.full; + if (wm0->priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark_max.full; + if (wm1->priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark.full; if (dfixed_trunc(priority_mark12) < 0) priority_mark12.full = 0; - if (wm1.priority_mark_max.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark_max.full; - d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); - d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); + if (wm1->priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark_max.full; + *d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); + *d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); if (rdev->disp_priority == 2) { - d1mode_priority_a_cnt |= S_006548_D1MODE_PRIORITY_A_ALWAYS_ON(1); - d2mode_priority_a_cnt |= S_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(1); + *d1mode_priority_a_cnt |= S_006548_D1MODE_PRIORITY_A_ALWAYS_ON(1); + *d2mode_priority_a_cnt |= S_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(1); } } else if (mode0) { - if (dfixed_trunc(wm0.dbpp) > 64) - a.full = dfixed_mul(wm0.dbpp, wm0.num_line_pair); + if (dfixed_trunc(wm0->dbpp) > 64) + a.full = dfixed_mul(wm0->dbpp, wm0->num_line_pair); else - a.full = wm0.num_line_pair.full; - fill_rate.full = dfixed_div(wm0.sclk, a); - if (wm0.consumption_rate.full > fill_rate.full) { - b.full = wm0.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm0.active_time); - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = wm0->num_line_pair.full; + fill_rate.full = dfixed_div(wm0->sclk, a); + if (wm0->consumption_rate.full > fill_rate.full) { + b.full = wm0->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm0->active_time); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); a.full = a.full + b.full; b.full = dfixed_const(16 * 1000); priority_mark02.full = dfixed_div(a, b); } else { - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark02.full = dfixed_div(a, b); } - if (wm0.priority_mark.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark.full; + if (wm0->priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark.full; if (dfixed_trunc(priority_mark02) < 0) priority_mark02.full = 0; - if (wm0.priority_mark_max.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark_max.full; - d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); + if (wm0->priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark_max.full; + *d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); if (rdev->disp_priority == 2) - d1mode_priority_a_cnt |= S_006548_D1MODE_PRIORITY_A_ALWAYS_ON(1); + *d1mode_priority_a_cnt |= S_006548_D1MODE_PRIORITY_A_ALWAYS_ON(1); } else if (mode1) { - if (dfixed_trunc(wm1.dbpp) > 64) - a.full = dfixed_mul(wm1.dbpp, wm1.num_line_pair); + if (dfixed_trunc(wm1->dbpp) > 64) + a.full = dfixed_mul(wm1->dbpp, wm1->num_line_pair); else - a.full = wm1.num_line_pair.full; - fill_rate.full = dfixed_div(wm1.sclk, a); - if (wm1.consumption_rate.full > fill_rate.full) { - b.full = wm1.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm1.active_time); - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = wm1->num_line_pair.full; + fill_rate.full = dfixed_div(wm1->sclk, a); + if (wm1->consumption_rate.full > fill_rate.full) { + b.full = wm1->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm1->active_time); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); a.full = a.full + b.full; b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } else { - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } - if (wm1.priority_mark.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark.full; + if (wm1->priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark.full; if (dfixed_trunc(priority_mark12) < 0) priority_mark12.full = 0; - if (wm1.priority_mark_max.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark_max.full; - d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); + if (wm1->priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark_max.full; + *d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); if (rdev->disp_priority == 2) - d2mode_priority_a_cnt |= S_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(1); + *d2mode_priority_a_cnt |= S_006D48_D2MODE_PRIORITY_A_ALWAYS_ON(1); } +} + +void rs690_bandwidth_update(struct radeon_device *rdev) +{ + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + struct rs690_watermark wm0_high, wm0_low; + struct rs690_watermark wm1_high, wm1_low; + u32 tmp; + u32 d1mode_priority_a_cnt, d1mode_priority_b_cnt; + u32 d2mode_priority_a_cnt, d2mode_priority_b_cnt; + + radeon_update_display_priority(rdev); + + if (rdev->mode_info.crtcs[0]->base.enabled) + mode0 = &rdev->mode_info.crtcs[0]->base.mode; + if (rdev->mode_info.crtcs[1]->base.enabled) + mode1 = &rdev->mode_info.crtcs[1]->base.mode; + /* + * Set display0/1 priority up in the memory controller for + * modes if the user specifies HIGH for displaypriority + * option. + */ + if ((rdev->disp_priority == 2) && + ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740))) { + tmp = RREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER); + tmp &= C_000104_MC_DISP0R_INIT_LAT; + tmp &= C_000104_MC_DISP1R_INIT_LAT; + if (mode0) + tmp |= S_000104_MC_DISP0R_INIT_LAT(1); + if (mode1) + tmp |= S_000104_MC_DISP1R_INIT_LAT(1); + WREG32_MC(R_000104_MC_INIT_MISC_LAT_TIMER, tmp); + } + rs690_line_buffer_adjust(rdev, mode0, mode1); + + if ((rdev->family == CHIP_RS690) || (rdev->family == CHIP_RS740)) + WREG32(R_006C9C_DCP_CONTROL, 0); + if ((rdev->family == CHIP_RS780) || (rdev->family == CHIP_RS880)) + WREG32(R_006C9C_DCP_CONTROL, 2); + + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0_high, false); + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1_high, false); + + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0_low, true); + rs690_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1_low, true); + + tmp = (wm0_high.lb_request_fifo_depth - 1); + tmp |= (wm1_high.lb_request_fifo_depth - 1) << 16; + WREG32(R_006D58_LB_MAX_REQ_OUTSTANDING, tmp); + + rs690_compute_mode_priority(rdev, + &wm0_high, &wm1_high, + mode0, mode1, + &d1mode_priority_a_cnt, &d2mode_priority_a_cnt); + rs690_compute_mode_priority(rdev, + &wm0_low, &wm1_low, + mode0, mode1, + &d1mode_priority_b_cnt, &d2mode_priority_b_cnt); WREG32(R_006548_D1MODE_PRIORITY_A_CNT, d1mode_priority_a_cnt); - WREG32(R_00654C_D1MODE_PRIORITY_B_CNT, d1mode_priority_a_cnt); + WREG32(R_00654C_D1MODE_PRIORITY_B_CNT, d1mode_priority_b_cnt); WREG32(R_006D48_D2MODE_PRIORITY_A_CNT, d2mode_priority_a_cnt); - WREG32(R_006D4C_D2MODE_PRIORITY_B_CNT, d2mode_priority_a_cnt); + WREG32(R_006D4C_D2MODE_PRIORITY_B_CNT, d2mode_priority_b_cnt); } uint32_t rs690_mc_rreg(struct radeon_device *rdev, uint32_t reg) diff --git a/sys/dev/drm/radeon/rs690d.h b/sys/dev/drm/radeon/rs690d.h index a6df62e20b..8af3ccf20c 100644 --- a/sys/dev/drm/radeon/rs690d.h +++ b/sys/dev/drm/radeon/rs690d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rs690d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RS690D_H__ #define __RS690D_H__ diff --git a/sys/dev/drm/radeon/rs780_dpm.c b/sys/dev/drm/radeon/rs780_dpm.c new file mode 100644 index 0000000000..9775d0cb8e --- /dev/null +++ b/sys/dev/drm/radeon/rs780_dpm.c @@ -0,0 +1,989 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "rs780d.h" +#include "r600_dpm.h" +#include "rs780_dpm.h" +#include "atom.h" +#include + +static struct igp_ps *rs780_get_ps(struct radeon_ps *rps) +{ + struct igp_ps *ps = rps->ps_priv; + + return ps; +} + +static struct igp_power_info *rs780_get_pi(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void rs780_get_pm_mode_parameters(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + struct radeon_mode_info *minfo = &rdev->mode_info; + struct drm_crtc *crtc; + struct radeon_crtc *radeon_crtc; + int i; + + /* defaults */ + pi->crtc_id = 0; + pi->refresh_rate = 60; + + for (i = 0; i < rdev->num_crtc; i++) { + crtc = (struct drm_crtc *)minfo->crtcs[i]; + if (crtc && crtc->enabled) { + radeon_crtc = to_radeon_crtc(crtc); + pi->crtc_id = radeon_crtc->crtc_id; + if (crtc->mode.htotal && crtc->mode.vtotal) + pi->refresh_rate = + (crtc->mode.clock * 1000) / + (crtc->mode.htotal * crtc->mode.vtotal); + break; + } + } +} + +static void rs780_voltage_scaling_enable(struct radeon_device *rdev, bool enable); + +static int rs780_initialize_dpm_power_state(struct radeon_device *rdev, + struct radeon_ps *boot_ps) +{ + struct atom_clock_dividers dividers; + struct igp_ps *default_state = rs780_get_ps(boot_ps); + int i, ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + default_state->sclk_low, false, ÷rs); + if (ret) + return ret; + + r600_engine_clock_entry_set_reference_divider(rdev, 0, dividers.ref_div); + r600_engine_clock_entry_set_feedback_divider(rdev, 0, dividers.fb_div); + r600_engine_clock_entry_set_post_divider(rdev, 0, dividers.post_div); + + if (dividers.enable_post_div) + r600_engine_clock_entry_enable_post_divider(rdev, 0, true); + else + r600_engine_clock_entry_enable_post_divider(rdev, 0, false); + + r600_engine_clock_entry_set_step_time(rdev, 0, R600_SST_DFLT); + r600_engine_clock_entry_enable_pulse_skipping(rdev, 0, false); + + r600_engine_clock_entry_enable(rdev, 0, true); + for (i = 1; i < R600_PM_NUMBER_OF_SCLKS; i++) + r600_engine_clock_entry_enable(rdev, i, false); + + r600_enable_mclk_control(rdev, false); + r600_voltage_control_enable_pins(rdev, 0); + + return 0; +} + +static int rs780_initialize_dpm_parameters(struct radeon_device *rdev, + struct radeon_ps *boot_ps) +{ + int ret = 0; + int i; + + r600_set_bsp(rdev, R600_BSU_DFLT, R600_BSP_DFLT); + + r600_set_at(rdev, 0, 0, 0, 0); + + r600_set_git(rdev, R600_GICST_DFLT); + + for (i = 0; i < R600_PM_NUMBER_OF_TC; i++) + r600_set_tc(rdev, i, 0, 0); + + r600_select_td(rdev, R600_TD_DFLT); + r600_set_vrc(rdev, 0); + + r600_set_tpu(rdev, R600_TPU_DFLT); + r600_set_tpc(rdev, R600_TPC_DFLT); + + r600_set_sstu(rdev, R600_SSTU_DFLT); + r600_set_sst(rdev, R600_SST_DFLT); + + r600_set_fctu(rdev, R600_FCTU_DFLT); + r600_set_fct(rdev, R600_FCT_DFLT); + + r600_set_vddc3d_oorsu(rdev, R600_VDDC3DOORSU_DFLT); + r600_set_vddc3d_oorphc(rdev, R600_VDDC3DOORPHC_DFLT); + r600_set_vddc3d_oorsdc(rdev, R600_VDDC3DOORSDC_DFLT); + r600_set_ctxcgtt3d_rphc(rdev, R600_CTXCGTT3DRPHC_DFLT); + r600_set_ctxcgtt3d_rsdc(rdev, R600_CTXCGTT3DRSDC_DFLT); + + r600_vid_rt_set_vru(rdev, R600_VRU_DFLT); + r600_vid_rt_set_vrt(rdev, R600_VOLTAGERESPONSETIME_DFLT); + r600_vid_rt_set_ssu(rdev, R600_SPLLSTEPUNIT_DFLT); + + ret = rs780_initialize_dpm_power_state(rdev, boot_ps); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_LOW, 0); + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_MEDIUM, 0); + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_HIGH, 0); + + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_LOW, 0); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, 0); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_HIGH, 0); + + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_LOW, 0); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, 0); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_HIGH, 0); + + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_LOW, R600_DISPLAY_WATERMARK_HIGH); + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_MEDIUM, R600_DISPLAY_WATERMARK_HIGH); + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_HIGH, R600_DISPLAY_WATERMARK_HIGH); + + r600_power_level_enable(rdev, R600_POWER_LEVEL_CTXSW, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + + r600_power_level_set_enter_index(rdev, R600_POWER_LEVEL_LOW); + + r600_set_vrc(rdev, RS780_CGFTV_DFLT); + + return ret; +} + +static void rs780_start_dpm(struct radeon_device *rdev) +{ + r600_enable_sclk_control(rdev, false); + r600_enable_mclk_control(rdev, false); + + r600_dynamicpm_enable(rdev, true); + + radeon_wait_for_vblank(rdev, 0); + radeon_wait_for_vblank(rdev, 1); + + r600_enable_spll_bypass(rdev, true); + r600_wait_for_spll_change(rdev); + r600_enable_spll_bypass(rdev, false); + r600_wait_for_spll_change(rdev); + + r600_enable_spll_bypass(rdev, true); + r600_wait_for_spll_change(rdev); + r600_enable_spll_bypass(rdev, false); + r600_wait_for_spll_change(rdev); + + r600_enable_sclk_control(rdev, true); +} + + +static void rs780_preset_ranges_slow_clk_fbdiv_en(struct radeon_device *rdev) +{ + WREG32_P(FVTHROT_SLOW_CLK_FEEDBACK_DIV_REG1, RANGE_SLOW_CLK_FEEDBACK_DIV_EN, + ~RANGE_SLOW_CLK_FEEDBACK_DIV_EN); + + WREG32_P(FVTHROT_SLOW_CLK_FEEDBACK_DIV_REG1, + RANGE0_SLOW_CLK_FEEDBACK_DIV(RS780_SLOWCLKFEEDBACKDIV_DFLT), + ~RANGE0_SLOW_CLK_FEEDBACK_DIV_MASK); +} + +static void rs780_preset_starting_fbdiv(struct radeon_device *rdev) +{ + u32 fbdiv = (RREG32(CG_SPLL_FUNC_CNTL) & SPLL_FB_DIV_MASK) >> SPLL_FB_DIV_SHIFT; + + WREG32_P(FVTHROT_FBDIV_REG1, STARTING_FEEDBACK_DIV(fbdiv), + ~STARTING_FEEDBACK_DIV_MASK); + + WREG32_P(FVTHROT_FBDIV_REG2, FORCED_FEEDBACK_DIV(fbdiv), + ~FORCED_FEEDBACK_DIV_MASK); + + WREG32_P(FVTHROT_FBDIV_REG1, FORCE_FEEDBACK_DIV, ~FORCE_FEEDBACK_DIV); +} + +static void rs780_voltage_scaling_init(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + struct drm_device *dev = rdev->ddev; + u32 fv_throt_pwm_fb_div_range[3]; + u32 fv_throt_pwm_range[4]; + + if (dev->pci_device == 0x9614) { + fv_throt_pwm_fb_div_range[0] = RS780D_FVTHROTPWMFBDIVRANGEREG0_DFLT; + fv_throt_pwm_fb_div_range[1] = RS780D_FVTHROTPWMFBDIVRANGEREG1_DFLT; + fv_throt_pwm_fb_div_range[2] = RS780D_FVTHROTPWMFBDIVRANGEREG2_DFLT; + } else if ((dev->pci_device == 0x9714) || + (dev->pci_device == 0x9715)) { + fv_throt_pwm_fb_div_range[0] = RS880D_FVTHROTPWMFBDIVRANGEREG0_DFLT; + fv_throt_pwm_fb_div_range[1] = RS880D_FVTHROTPWMFBDIVRANGEREG1_DFLT; + fv_throt_pwm_fb_div_range[2] = RS880D_FVTHROTPWMFBDIVRANGEREG2_DFLT; + } else { + fv_throt_pwm_fb_div_range[0] = RS780_FVTHROTPWMFBDIVRANGEREG0_DFLT; + fv_throt_pwm_fb_div_range[1] = RS780_FVTHROTPWMFBDIVRANGEREG1_DFLT; + fv_throt_pwm_fb_div_range[2] = RS780_FVTHROTPWMFBDIVRANGEREG2_DFLT; + } + + if (pi->pwm_voltage_control) { + fv_throt_pwm_range[0] = pi->min_voltage; + fv_throt_pwm_range[1] = pi->min_voltage; + fv_throt_pwm_range[2] = pi->max_voltage; + fv_throt_pwm_range[3] = pi->max_voltage; + } else { + fv_throt_pwm_range[0] = pi->invert_pwm_required ? + RS780_FVTHROTPWMRANGE3_GPIO_DFLT : RS780_FVTHROTPWMRANGE0_GPIO_DFLT; + fv_throt_pwm_range[1] = pi->invert_pwm_required ? + RS780_FVTHROTPWMRANGE2_GPIO_DFLT : RS780_FVTHROTPWMRANGE1_GPIO_DFLT; + fv_throt_pwm_range[2] = pi->invert_pwm_required ? + RS780_FVTHROTPWMRANGE1_GPIO_DFLT : RS780_FVTHROTPWMRANGE2_GPIO_DFLT; + fv_throt_pwm_range[3] = pi->invert_pwm_required ? + RS780_FVTHROTPWMRANGE0_GPIO_DFLT : RS780_FVTHROTPWMRANGE3_GPIO_DFLT; + } + + WREG32_P(FVTHROT_PWM_CTRL_REG0, + STARTING_PWM_HIGHTIME(pi->max_voltage), + ~STARTING_PWM_HIGHTIME_MASK); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, + NUMBER_OF_CYCLES_IN_PERIOD(pi->num_of_cycles_in_period), + ~NUMBER_OF_CYCLES_IN_PERIOD_MASK); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, FORCE_STARTING_PWM_HIGHTIME, + ~FORCE_STARTING_PWM_HIGHTIME); + + if (pi->invert_pwm_required) + WREG32_P(FVTHROT_PWM_CTRL_REG0, INVERT_PWM_WAVEFORM, ~INVERT_PWM_WAVEFORM); + else + WREG32_P(FVTHROT_PWM_CTRL_REG0, 0, ~INVERT_PWM_WAVEFORM); + + rs780_voltage_scaling_enable(rdev, true); + + WREG32(FVTHROT_PWM_CTRL_REG1, + (MIN_PWM_HIGHTIME(pi->min_voltage) | + MAX_PWM_HIGHTIME(pi->max_voltage))); + + WREG32(FVTHROT_PWM_US_REG0, RS780_FVTHROTPWMUSREG0_DFLT); + WREG32(FVTHROT_PWM_US_REG1, RS780_FVTHROTPWMUSREG1_DFLT); + WREG32(FVTHROT_PWM_DS_REG0, RS780_FVTHROTPWMDSREG0_DFLT); + WREG32(FVTHROT_PWM_DS_REG1, RS780_FVTHROTPWMDSREG1_DFLT); + + WREG32_P(FVTHROT_PWM_FEEDBACK_DIV_REG1, + RANGE0_PWM_FEEDBACK_DIV(fv_throt_pwm_fb_div_range[0]), + ~RANGE0_PWM_FEEDBACK_DIV_MASK); + + WREG32(FVTHROT_PWM_FEEDBACK_DIV_REG2, + (RANGE1_PWM_FEEDBACK_DIV(fv_throt_pwm_fb_div_range[1]) | + RANGE2_PWM_FEEDBACK_DIV(fv_throt_pwm_fb_div_range[2]))); + + WREG32(FVTHROT_PWM_FEEDBACK_DIV_REG3, + (RANGE0_PWM(fv_throt_pwm_range[1]) | + RANGE1_PWM(fv_throt_pwm_range[2]))); + WREG32(FVTHROT_PWM_FEEDBACK_DIV_REG4, + (RANGE2_PWM(fv_throt_pwm_range[1]) | + RANGE3_PWM(fv_throt_pwm_range[2]))); +} + +static void rs780_clk_scaling_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(FVTHROT_CNTRL_REG, ENABLE_FV_THROT | ENABLE_FV_UPDATE, + ~(ENABLE_FV_THROT | ENABLE_FV_UPDATE)); + else + WREG32_P(FVTHROT_CNTRL_REG, 0, + ~(ENABLE_FV_THROT | ENABLE_FV_UPDATE)); +} + +static void rs780_voltage_scaling_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(FVTHROT_CNTRL_REG, ENABLE_FV_THROT_IO, ~ENABLE_FV_THROT_IO); + else + WREG32_P(FVTHROT_CNTRL_REG, 0, ~ENABLE_FV_THROT_IO); +} + +static void rs780_set_engine_clock_wfc(struct radeon_device *rdev) +{ + WREG32(FVTHROT_UTC0, RS780_FVTHROTUTC0_DFLT); + WREG32(FVTHROT_UTC1, RS780_FVTHROTUTC1_DFLT); + WREG32(FVTHROT_UTC2, RS780_FVTHROTUTC2_DFLT); + WREG32(FVTHROT_UTC3, RS780_FVTHROTUTC3_DFLT); + WREG32(FVTHROT_UTC4, RS780_FVTHROTUTC4_DFLT); + + WREG32(FVTHROT_DTC0, RS780_FVTHROTDTC0_DFLT); + WREG32(FVTHROT_DTC1, RS780_FVTHROTDTC1_DFLT); + WREG32(FVTHROT_DTC2, RS780_FVTHROTDTC2_DFLT); + WREG32(FVTHROT_DTC3, RS780_FVTHROTDTC3_DFLT); + WREG32(FVTHROT_DTC4, RS780_FVTHROTDTC4_DFLT); +} + +static void rs780_set_engine_clock_sc(struct radeon_device *rdev) +{ + WREG32_P(FVTHROT_FBDIV_REG2, + FB_DIV_TIMER_VAL(RS780_FBDIVTIMERVAL_DFLT), + ~FB_DIV_TIMER_VAL_MASK); + + WREG32_P(FVTHROT_CNTRL_REG, + REFRESH_RATE_DIVISOR(0) | MINIMUM_CIP(0xf), + ~(REFRESH_RATE_DIVISOR_MASK | MINIMUM_CIP_MASK)); +} + +static void rs780_set_engine_clock_tdc(struct radeon_device *rdev) +{ + WREG32_P(FVTHROT_CNTRL_REG, 0, ~(FORCE_TREND_SEL | TREND_SEL_MODE)); +} + +static void rs780_set_engine_clock_ssc(struct radeon_device *rdev) +{ + WREG32(FVTHROT_FB_US_REG0, RS780_FVTHROTFBUSREG0_DFLT); + WREG32(FVTHROT_FB_US_REG1, RS780_FVTHROTFBUSREG1_DFLT); + WREG32(FVTHROT_FB_DS_REG0, RS780_FVTHROTFBDSREG0_DFLT); + WREG32(FVTHROT_FB_DS_REG1, RS780_FVTHROTFBDSREG1_DFLT); + + WREG32_P(FVTHROT_FBDIV_REG1, MAX_FEEDBACK_STEP(1), ~MAX_FEEDBACK_STEP_MASK); +} + +static void rs780_program_at(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + + WREG32(FVTHROT_TARGET_REG, 30000000 / pi->refresh_rate); + WREG32(FVTHROT_CB1, 1000000 * 5 / pi->refresh_rate); + WREG32(FVTHROT_CB2, 1000000 * 10 / pi->refresh_rate); + WREG32(FVTHROT_CB3, 1000000 * 30 / pi->refresh_rate); + WREG32(FVTHROT_CB4, 1000000 * 50 / pi->refresh_rate); +} + +static void rs780_disable_vbios_powersaving(struct radeon_device *rdev) +{ + WREG32_P(CG_INTGFX_MISC, 0, ~0xFFF00000); +} + +static void rs780_force_voltage_to_high(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + struct igp_ps *current_state = rs780_get_ps(rdev->pm.dpm.current_ps); + + if ((current_state->max_voltage == RS780_VDDC_LEVEL_HIGH) && + (current_state->min_voltage == RS780_VDDC_LEVEL_HIGH)) + return; + + WREG32_P(GFX_MACRO_BYPASS_CNTL, SPLL_BYPASS_CNTL, ~SPLL_BYPASS_CNTL); + + DRM_UDELAY(1); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, + STARTING_PWM_HIGHTIME(pi->max_voltage), + ~STARTING_PWM_HIGHTIME_MASK); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, + FORCE_STARTING_PWM_HIGHTIME, ~FORCE_STARTING_PWM_HIGHTIME); + + WREG32_P(FVTHROT_PWM_FEEDBACK_DIV_REG1, 0, + ~RANGE_PWM_FEEDBACK_DIV_EN); + + DRM_UDELAY(1); + + WREG32_P(GFX_MACRO_BYPASS_CNTL, 0, ~SPLL_BYPASS_CNTL); +} + +static int rs780_set_engine_clock_scaling(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct atom_clock_dividers min_dividers, max_dividers, current_max_dividers; + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_ps *old_state = rs780_get_ps(old_ps); + int ret; + + if ((new_state->sclk_high == old_state->sclk_high) && + (new_state->sclk_low == old_state->sclk_low)) + return 0; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + new_state->sclk_low, false, &min_dividers); + if (ret) + return ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + new_state->sclk_high, false, &max_dividers); + if (ret) + return ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + old_state->sclk_high, false, ¤t_max_dividers); + if (ret) + return ret; + + WREG32_P(GFX_MACRO_BYPASS_CNTL, SPLL_BYPASS_CNTL, ~SPLL_BYPASS_CNTL); + + WREG32_P(FVTHROT_FBDIV_REG2, FORCED_FEEDBACK_DIV(max_dividers.fb_div), + ~FORCED_FEEDBACK_DIV_MASK); + WREG32_P(FVTHROT_FBDIV_REG1, STARTING_FEEDBACK_DIV(max_dividers.fb_div), + ~STARTING_FEEDBACK_DIV_MASK); + WREG32_P(FVTHROT_FBDIV_REG1, FORCE_FEEDBACK_DIV, ~FORCE_FEEDBACK_DIV); + + DRM_UDELAY(100); + + WREG32_P(GFX_MACRO_BYPASS_CNTL, 0, ~SPLL_BYPASS_CNTL); + + if (max_dividers.fb_div > min_dividers.fb_div) { + WREG32_P(FVTHROT_FBDIV_REG0, + MIN_FEEDBACK_DIV(min_dividers.fb_div) | + MAX_FEEDBACK_DIV(max_dividers.fb_div), + ~(MIN_FEEDBACK_DIV_MASK | MAX_FEEDBACK_DIV_MASK)); + + WREG32_P(FVTHROT_FBDIV_REG1, 0, ~FORCE_FEEDBACK_DIV); + } + + return 0; +} + +static void rs780_set_engine_clock_spc(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_ps *old_state = rs780_get_ps(old_ps); + struct igp_power_info *pi = rs780_get_pi(rdev); + + if ((new_state->sclk_high == old_state->sclk_high) && + (new_state->sclk_low == old_state->sclk_low)) + return; + + if (pi->crtc_id == 0) + WREG32_P(CG_INTGFX_MISC, 0, ~FVTHROT_VBLANK_SEL); + else + WREG32_P(CG_INTGFX_MISC, FVTHROT_VBLANK_SEL, ~FVTHROT_VBLANK_SEL); + +} + +static void rs780_activate_engine_clk_scaling(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_ps *old_state = rs780_get_ps(old_ps); + + if ((new_state->sclk_high == old_state->sclk_high) && + (new_state->sclk_low == old_state->sclk_low)) + return; + + rs780_clk_scaling_enable(rdev, true); +} + +static u32 rs780_get_voltage_for_vddc_level(struct radeon_device *rdev, + enum rs780_vddc_level vddc) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + + if (vddc == RS780_VDDC_LEVEL_HIGH) + return pi->max_voltage; + else if (vddc == RS780_VDDC_LEVEL_LOW) + return pi->min_voltage; + else + return pi->max_voltage; +} + +static void rs780_enable_voltage_scaling(struct radeon_device *rdev, + struct radeon_ps *new_ps) +{ + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_power_info *pi = rs780_get_pi(rdev); + enum rs780_vddc_level vddc_high, vddc_low; + + DRM_UDELAY(100); + + if ((new_state->max_voltage == RS780_VDDC_LEVEL_HIGH) && + (new_state->min_voltage == RS780_VDDC_LEVEL_HIGH)) + return; + + vddc_high = rs780_get_voltage_for_vddc_level(rdev, + new_state->max_voltage); + vddc_low = rs780_get_voltage_for_vddc_level(rdev, + new_state->min_voltage); + + WREG32_P(GFX_MACRO_BYPASS_CNTL, SPLL_BYPASS_CNTL, ~SPLL_BYPASS_CNTL); + + DRM_UDELAY(1); + if (vddc_high > vddc_low) { + WREG32_P(FVTHROT_PWM_FEEDBACK_DIV_REG1, + RANGE_PWM_FEEDBACK_DIV_EN, ~RANGE_PWM_FEEDBACK_DIV_EN); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, 0, ~FORCE_STARTING_PWM_HIGHTIME); + } else if (vddc_high == vddc_low) { + if (pi->max_voltage != vddc_high) { + WREG32_P(FVTHROT_PWM_CTRL_REG0, + STARTING_PWM_HIGHTIME(vddc_high), + ~STARTING_PWM_HIGHTIME_MASK); + + WREG32_P(FVTHROT_PWM_CTRL_REG0, + FORCE_STARTING_PWM_HIGHTIME, + ~FORCE_STARTING_PWM_HIGHTIME); + } + } + + WREG32_P(GFX_MACRO_BYPASS_CNTL, 0, ~SPLL_BYPASS_CNTL); +} + +static void rs780_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_ps *current_state = rs780_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->sclk_high >= current_state->sclk_high) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +static void rs780_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct igp_ps *new_state = rs780_get_ps(new_ps); + struct igp_ps *current_state = rs780_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->sclk_high < current_state->sclk_high) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +int rs780_dpm_enable(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + rs780_get_pm_mode_parameters(rdev); + rs780_disable_vbios_powersaving(rdev); + + if (r600_dynamicpm_enabled(rdev)) + return -EINVAL; + ret = rs780_initialize_dpm_parameters(rdev, boot_ps); + if (ret) + return ret; + rs780_start_dpm(rdev); + + rs780_preset_ranges_slow_clk_fbdiv_en(rdev); + rs780_preset_starting_fbdiv(rdev); + if (pi->voltage_control) + rs780_voltage_scaling_init(rdev); + rs780_clk_scaling_enable(rdev, true); + rs780_set_engine_clock_sc(rdev); + rs780_set_engine_clock_wfc(rdev); + rs780_program_at(rdev); + rs780_set_engine_clock_tdc(rdev); + rs780_set_engine_clock_ssc(rdev); + + if (pi->gfx_clock_gating) + r600_gfx_clockgating_enable(rdev, true); + + if (rdev->irq.installed && (rdev->pm.int_thermal_type == THERMAL_TYPE_RV6XX)) { + ret = r600_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + } + + return 0; +} + +void rs780_dpm_disable(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + + r600_dynamicpm_enable(rdev, false); + + rs780_clk_scaling_enable(rdev, false); + rs780_voltage_scaling_enable(rdev, false); + + if (pi->gfx_clock_gating) + r600_gfx_clockgating_enable(rdev, false); + + if (rdev->irq.installed && + (rdev->pm.int_thermal_type == THERMAL_TYPE_RV6XX)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } +} + +int rs780_dpm_set_power_state(struct radeon_device *rdev) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + struct radeon_ps *old_ps = rdev->pm.dpm.current_ps; + int ret; + + rs780_get_pm_mode_parameters(rdev); + + rs780_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + + if (pi->voltage_control) { + rs780_force_voltage_to_high(rdev); + DRM_MDELAY(5); + } + + ret = rs780_set_engine_clock_scaling(rdev, new_ps, old_ps); + if (ret) + return ret; + rs780_set_engine_clock_spc(rdev, new_ps, old_ps); + + rs780_activate_engine_clk_scaling(rdev, new_ps, old_ps); + + if (pi->voltage_control) + rs780_enable_voltage_scaling(rdev, new_ps); + + rs780_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + return 0; +} + +void rs780_dpm_setup_asic(struct radeon_device *rdev) +{ + +} + +void rs780_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + rs780_get_pm_mode_parameters(rdev); + rs780_program_at(rdev); +} + +union igp_info { + struct _ATOM_INTEGRATED_SYSTEM_INFO info; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2; +}; + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void rs780_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else if (r600_is_uvd_state(rps->class, rps->class2)) { + rps->vclk = RS780_DEFAULT_VCLK_FREQ; + rps->dclk = RS780_DEFAULT_DCLK_FREQ; + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) + rdev->pm.dpm.boot_ps = rps; + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void rs780_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + union pplib_clock_info *clock_info) +{ + struct igp_ps *ps = rs780_get_ps(rps); + u32 sclk; + + sclk = le16_to_cpu(clock_info->rs780.usLowEngineClockLow); + sclk |= clock_info->rs780.ucLowEngineClockHigh << 16; + ps->sclk_low = sclk; + sclk = le16_to_cpu(clock_info->rs780.usHighEngineClockLow); + sclk |= clock_info->rs780.ucHighEngineClockHigh << 16; + ps->sclk_high = sclk; + switch (le16_to_cpu(clock_info->rs780.usVDDC)) { + case ATOM_PPLIB_RS780_VOLTAGE_NONE: + default: + ps->min_voltage = RS780_VDDC_LEVEL_UNKNOWN; + ps->max_voltage = RS780_VDDC_LEVEL_UNKNOWN; + break; + case ATOM_PPLIB_RS780_VOLTAGE_LOW: + ps->min_voltage = RS780_VDDC_LEVEL_LOW; + ps->max_voltage = RS780_VDDC_LEVEL_LOW; + break; + case ATOM_PPLIB_RS780_VOLTAGE_HIGH: + ps->min_voltage = RS780_VDDC_LEVEL_HIGH; + ps->max_voltage = RS780_VDDC_LEVEL_HIGH; + break; + case ATOM_PPLIB_RS780_VOLTAGE_VARIABLE: + ps->min_voltage = RS780_VDDC_LEVEL_LOW; + ps->max_voltage = RS780_VDDC_LEVEL_HIGH; + break; + } + ps->flags = le32_to_cpu(clock_info->rs780.ulFlags); + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + ps->sclk_low = rdev->clock.default_sclk; + ps->sclk_high = rdev->clock.default_sclk; + ps->min_voltage = RS780_VDDC_LEVEL_HIGH; + ps->max_voltage = RS780_VDDC_LEVEL_HIGH; + } +} + +static int rs780_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i; + union pplib_clock_info *clock_info; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + struct igp_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + power_info->pplib.ucNumStates, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + + for (i = 0; i < power_info->pplib.ucNumStates; i++) { + power_state = (union pplib_power_state *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset) + + i * power_info->pplib.ucStateEntrySize); + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset) + + (power_state->v1.ucNonClockStateIndex * + power_info->pplib.ucNonClockSize)); + if (power_info->pplib.ucStateEntrySize - 1) { + clock_info = (union pplib_clock_info *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset) + + (power_state->v1.ucClockStateIndices[0] * + power_info->pplib.ucClockInfoSize)); + ps = kzalloc(sizeof(struct igp_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + rs780_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + power_info->pplib.ucNonClockSize); + rs780_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], + clock_info); + } + } + rdev->pm.dpm.num_ps = power_info->pplib.ucNumStates; + return 0; +} + +int rs780_dpm_init(struct radeon_device *rdev) +{ + struct igp_power_info *pi; + int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo); + union igp_info *info; + u16 data_offset; + u8 frev, crev; + int ret; + + pi = kzalloc(sizeof(struct igp_power_info), GFP_KERNEL); + if (pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = pi; + + ret = rs780_parse_power_table(rdev); + if (ret) + return ret; + + pi->voltage_control = false; + pi->gfx_clock_gating = true; + + if (atom_parse_data_header(rdev->mode_info.atom_context, index, NULL, + &frev, &crev, &data_offset)) { + info = (union igp_info *)((uint8_t*)rdev->mode_info.atom_context->bios + data_offset); + + /* Get various system informations from bios */ + switch (crev) { + case 1: + pi->num_of_cycles_in_period = + info->info.ucNumberOfCyclesInPeriod; + pi->num_of_cycles_in_period |= + info->info.ucNumberOfCyclesInPeriodHi << 8; + pi->invert_pwm_required = + (pi->num_of_cycles_in_period & 0x8000) ? true : false; + pi->boot_voltage = info->info.ucStartingPWM_HighTime; + pi->max_voltage = info->info.ucMaxNBVoltage; + pi->max_voltage |= info->info.ucMaxNBVoltageHigh << 8; + pi->min_voltage = info->info.ucMinNBVoltage; + pi->min_voltage |= info->info.ucMinNBVoltageHigh << 8; + pi->inter_voltage_low = + le16_to_cpu(info->info.usInterNBVoltageLow); + pi->inter_voltage_high = + le16_to_cpu(info->info.usInterNBVoltageHigh); + pi->voltage_control = true; + pi->bootup_uma_clk = info->info.usK8MemoryClock * 100; + break; + case 2: + pi->num_of_cycles_in_period = + le16_to_cpu(info->info_2.usNumberOfCyclesInPeriod); + pi->invert_pwm_required = + (pi->num_of_cycles_in_period & 0x8000) ? true : false; + pi->boot_voltage = + le16_to_cpu(info->info_2.usBootUpNBVoltage); + pi->max_voltage = + le16_to_cpu(info->info_2.usMaxNBVoltage); + pi->min_voltage = + le16_to_cpu(info->info_2.usMinNBVoltage); + pi->system_config = + le32_to_cpu(info->info_2.ulSystemConfig); + pi->pwm_voltage_control = + (pi->system_config & 0x4) ? true : false; + pi->voltage_control = true; + pi->bootup_uma_clk = le32_to_cpu(info->info_2.ulBootUpUMAClock); + break; + default: + DRM_ERROR("No integrated system info for your GPU\n"); + return -EINVAL; + } + if (pi->min_voltage > pi->max_voltage) + pi->voltage_control = false; + if (pi->pwm_voltage_control) { + if ((pi->num_of_cycles_in_period == 0) || + (pi->max_voltage == 0) || + (pi->min_voltage == 0)) + pi->voltage_control = false; + } else { + if ((pi->num_of_cycles_in_period == 0) || + (pi->max_voltage == 0)) + pi->voltage_control = false; + } + + return 0; + } + radeon_dpm_fini(rdev); + return -EINVAL; +} + +void rs780_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct igp_ps *ps = rs780_get_ps(rps); + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + printk("\t\tpower level 0 sclk: %u vddc_index: %d\n", + ps->sclk_low, ps->min_voltage); + printk("\t\tpower level 1 sclk: %u vddc_index: %d\n", + ps->sclk_high, ps->max_voltage); + r600_dpm_print_ps_status(rdev, rps); +} + +void rs780_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +u32 rs780_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct igp_ps *requested_state = rs780_get_ps(rdev->pm.dpm.requested_ps); + + if (low) + return requested_state->sclk_low; + else + return requested_state->sclk_high; +} + +u32 rs780_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct igp_power_info *pi = rs780_get_pi(rdev); + + return pi->bootup_uma_clk; +} + +void rs780_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct igp_ps *ps = rs780_get_ps(rps); + u32 current_fb_div = RREG32(FVTHROT_STATUS_REG0) & CURRENT_FEEDBACK_DIV_MASK; + u32 func_cntl = RREG32(CG_SPLL_FUNC_CNTL); + u32 ref_div = ((func_cntl & SPLL_REF_DIV_MASK) >> SPLL_REF_DIV_SHIFT) + 1; + u32 post_div = ((func_cntl & SPLL_SW_HILEN_MASK) >> SPLL_SW_HILEN_SHIFT) + 1 + + ((func_cntl & SPLL_SW_LOLEN_MASK) >> SPLL_SW_LOLEN_SHIFT) + 1; + u32 sclk = (rdev->clock.spll.reference_freq * current_fb_div) / + (post_div * ref_div); + + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + + /* guess based on the current sclk */ + if (sclk < (ps->sclk_low + 500)) + seq_printf(m, "power level 0 sclk: %u vddc_index: %d\n", + ps->sclk_low, ps->min_voltage); + else + seq_printf(m, "power level 1 sclk: %u vddc_index: %d\n", + ps->sclk_high, ps->max_voltage); +} diff --git a/sys/dev/drm/radeon/rs780_dpm.h b/sys/dev/drm/radeon/rs780_dpm.h new file mode 100644 index 0000000000..47a40b14fa --- /dev/null +++ b/sys/dev/drm/radeon/rs780_dpm.h @@ -0,0 +1,109 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __RS780_DPM_H__ +#define __RS780_DPM_H__ + +enum rs780_vddc_level { + RS780_VDDC_LEVEL_UNKNOWN = 0, + RS780_VDDC_LEVEL_LOW = 1, + RS780_VDDC_LEVEL_HIGH = 2, +}; + +struct igp_power_info { + /* flags */ + bool invert_pwm_required; + bool pwm_voltage_control; + bool voltage_control; + bool gfx_clock_gating; + /* stored values */ + u32 system_config; + u32 bootup_uma_clk; + u16 max_voltage; + u16 min_voltage; + u16 boot_voltage; + u16 inter_voltage_low; + u16 inter_voltage_high; + u16 num_of_cycles_in_period; + /* variable */ + int crtc_id; + int refresh_rate; +}; + +struct igp_ps { + enum rs780_vddc_level min_voltage; + enum rs780_vddc_level max_voltage; + u32 sclk_low; + u32 sclk_high; + u32 flags; +}; + +#define RS780_CGFTV_DFLT 0x0303000f +#define RS780_FBDIVTIMERVAL_DFLT 0x2710 + +#define RS780_FVTHROTUTC0_DFLT 0x04010040 +#define RS780_FVTHROTUTC1_DFLT 0x04010040 +#define RS780_FVTHROTUTC2_DFLT 0x04010040 +#define RS780_FVTHROTUTC3_DFLT 0x04010040 +#define RS780_FVTHROTUTC4_DFLT 0x04010040 + +#define RS780_FVTHROTDTC0_DFLT 0x04010040 +#define RS780_FVTHROTDTC1_DFLT 0x04010040 +#define RS780_FVTHROTDTC2_DFLT 0x04010040 +#define RS780_FVTHROTDTC3_DFLT 0x04010040 +#define RS780_FVTHROTDTC4_DFLT 0x04010040 + +#define RS780_FVTHROTFBUSREG0_DFLT 0x00001001 +#define RS780_FVTHROTFBUSREG1_DFLT 0x00002002 +#define RS780_FVTHROTFBDSREG0_DFLT 0x00004001 +#define RS780_FVTHROTFBDSREG1_DFLT 0x00020010 + +#define RS780_FVTHROTPWMUSREG0_DFLT 0x00002001 +#define RS780_FVTHROTPWMUSREG1_DFLT 0x00004003 +#define RS780_FVTHROTPWMDSREG0_DFLT 0x00002001 +#define RS780_FVTHROTPWMDSREG1_DFLT 0x00004003 + +#define RS780_FVTHROTPWMFBDIVRANGEREG0_DFLT 0x37 +#define RS780_FVTHROTPWMFBDIVRANGEREG1_DFLT 0x4b +#define RS780_FVTHROTPWMFBDIVRANGEREG2_DFLT 0x8b + +#define RS780D_FVTHROTPWMFBDIVRANGEREG0_DFLT 0x8b +#define RS780D_FVTHROTPWMFBDIVRANGEREG1_DFLT 0x8c +#define RS780D_FVTHROTPWMFBDIVRANGEREG2_DFLT 0xb5 + +#define RS880D_FVTHROTPWMFBDIVRANGEREG0_DFLT 0x8d +#define RS880D_FVTHROTPWMFBDIVRANGEREG1_DFLT 0x8e +#define RS880D_FVTHROTPWMFBDIVRANGEREG2_DFLT 0xBa + +#define RS780_FVTHROTPWMRANGE0_GPIO_DFLT 0x1a +#define RS780_FVTHROTPWMRANGE1_GPIO_DFLT 0x1a +#define RS780_FVTHROTPWMRANGE2_GPIO_DFLT 0x0 +#define RS780_FVTHROTPWMRANGE3_GPIO_DFLT 0x0 + +#define RS780_SLOWCLKFEEDBACKDIV_DFLT 110 + +#define RS780_CGCLKGATING_DFLT 0x0000E204 + +#define RS780_DEFAULT_VCLK_FREQ 53300 /* 10 khz */ +#define RS780_DEFAULT_DCLK_FREQ 40000 /* 10 khz */ + +#endif diff --git a/sys/dev/drm/radeon/rs780d.h b/sys/dev/drm/radeon/rs780d.h new file mode 100644 index 0000000000..cfbe9a43d9 --- /dev/null +++ b/sys/dev/drm/radeon/rs780d.h @@ -0,0 +1,171 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __RS780D_H__ +#define __RS780D_H__ + +#define CG_SPLL_FUNC_CNTL 0x600 +# define SPLL_RESET (1 << 0) +# define SPLL_SLEEP (1 << 1) +# define SPLL_REF_DIV(x) ((x) << 2) +# define SPLL_REF_DIV_MASK (7 << 2) +# define SPLL_REF_DIV_SHIFT 2 +# define SPLL_FB_DIV(x) ((x) << 5) +# define SPLL_FB_DIV_MASK (0xff << 2) +# define SPLL_FB_DIV_SHIFT 2 +# define SPLL_PULSEEN (1 << 13) +# define SPLL_PULSENUM(x) ((x) << 14) +# define SPLL_PULSENUM_MASK (3 << 14) +# define SPLL_SW_HILEN(x) ((x) << 16) +# define SPLL_SW_HILEN_MASK (0xf << 16) +# define SPLL_SW_HILEN_SHIFT 16 +# define SPLL_SW_LOLEN(x) ((x) << 20) +# define SPLL_SW_LOLEN_MASK (0xf << 20) +# define SPLL_SW_LOLEN_SHIFT 20 +# define SPLL_DIVEN (1 << 24) +# define SPLL_BYPASS_EN (1 << 25) +# define SPLL_CHG_STATUS (1 << 29) +# define SPLL_CTLREQ (1 << 30) +# define SPLL_CTLACK (1 << 31) + +/* RS780/RS880 PM */ +#define FVTHROT_CNTRL_REG 0x3000 +#define DONT_WAIT_FOR_FBDIV_WRAP (1 << 0) +#define MINIMUM_CIP(x) ((x) << 1) +#define MINIMUM_CIP_SHIFT 1 +#define MINIMUM_CIP_MASK 0x1fffffe +#define REFRESH_RATE_DIVISOR(x) ((x) << 25) +#define REFRESH_RATE_DIVISOR_SHIFT 25 +#define REFRESH_RATE_DIVISOR_MASK (0x3 << 25) +#define ENABLE_FV_THROT (1 << 27) +#define ENABLE_FV_UPDATE (1 << 28) +#define TREND_SEL_MODE (1 << 29) +#define FORCE_TREND_SEL (1 << 30) +#define ENABLE_FV_THROT_IO (1 << 31) +#define FVTHROT_TARGET_REG 0x3004 +#define TARGET_IDLE_COUNT(x) ((x) << 0) +#define TARGET_IDLE_COUNT_MASK 0xffffff +#define TARGET_IDLE_COUNT_SHIFT 0 +#define FVTHROT_CB1 0x3008 +#define FVTHROT_CB2 0x300c +#define FVTHROT_CB3 0x3010 +#define FVTHROT_CB4 0x3014 +#define FVTHROT_UTC0 0x3018 +#define FVTHROT_UTC1 0x301c +#define FVTHROT_UTC2 0x3020 +#define FVTHROT_UTC3 0x3024 +#define FVTHROT_UTC4 0x3028 +#define FVTHROT_DTC0 0x302c +#define FVTHROT_DTC1 0x3030 +#define FVTHROT_DTC2 0x3034 +#define FVTHROT_DTC3 0x3038 +#define FVTHROT_DTC4 0x303c +#define FVTHROT_FBDIV_REG0 0x3040 +#define MIN_FEEDBACK_DIV(x) ((x) << 0) +#define MIN_FEEDBACK_DIV_MASK 0xfff +#define MIN_FEEDBACK_DIV_SHIFT 0 +#define MAX_FEEDBACK_DIV(x) ((x) << 12) +#define MAX_FEEDBACK_DIV_MASK (0xfff << 12) +#define MAX_FEEDBACK_DIV_SHIFT 12 +#define FVTHROT_FBDIV_REG1 0x3044 +#define MAX_FEEDBACK_STEP(x) ((x) << 0) +#define MAX_FEEDBACK_STEP_MASK 0xfff +#define MAX_FEEDBACK_STEP_SHIFT 0 +#define STARTING_FEEDBACK_DIV(x) ((x) << 12) +#define STARTING_FEEDBACK_DIV_MASK (0xfff << 12) +#define STARTING_FEEDBACK_DIV_SHIFT 12 +#define FORCE_FEEDBACK_DIV (1 << 24) +#define FVTHROT_FBDIV_REG2 0x3048 +#define FORCED_FEEDBACK_DIV(x) ((x) << 0) +#define FORCED_FEEDBACK_DIV_MASK 0xfff +#define FORCED_FEEDBACK_DIV_SHIFT 0 +#define FB_DIV_TIMER_VAL(x) ((x) << 12) +#define FB_DIV_TIMER_VAL_MASK (0xffff << 12) +#define FB_DIV_TIMER_VAL_SHIFT 12 +#define FVTHROT_FB_US_REG0 0x304c +#define FVTHROT_FB_US_REG1 0x3050 +#define FVTHROT_FB_DS_REG0 0x3054 +#define FVTHROT_FB_DS_REG1 0x3058 +#define FVTHROT_PWM_CTRL_REG0 0x305c +#define STARTING_PWM_HIGHTIME(x) ((x) << 0) +#define STARTING_PWM_HIGHTIME_MASK 0xfff +#define STARTING_PWM_HIGHTIME_SHIFT 0 +#define NUMBER_OF_CYCLES_IN_PERIOD(x) ((x) << 12) +#define NUMBER_OF_CYCLES_IN_PERIOD_MASK (0xfff << 12) +#define NUMBER_OF_CYCLES_IN_PERIOD_SHIFT 12 +#define FORCE_STARTING_PWM_HIGHTIME (1 << 24) +#define INVERT_PWM_WAVEFORM (1 << 25) +#define FVTHROT_PWM_CTRL_REG1 0x3060 +#define MIN_PWM_HIGHTIME(x) ((x) << 0) +#define MIN_PWM_HIGHTIME_MASK 0xfff +#define MIN_PWM_HIGHTIME_SHIFT 0 +#define MAX_PWM_HIGHTIME(x) ((x) << 12) +#define MAX_PWM_HIGHTIME_MASK (0xfff << 12) +#define MAX_PWM_HIGHTIME_SHIFT 12 +#define FVTHROT_PWM_US_REG0 0x3064 +#define FVTHROT_PWM_US_REG1 0x3068 +#define FVTHROT_PWM_DS_REG0 0x306c +#define FVTHROT_PWM_DS_REG1 0x3070 +#define FVTHROT_STATUS_REG0 0x3074 +#define CURRENT_FEEDBACK_DIV_MASK 0xfff +#define CURRENT_FEEDBACK_DIV_SHIFT 0 +#define FVTHROT_STATUS_REG1 0x3078 +#define FVTHROT_STATUS_REG2 0x307c +#define CG_INTGFX_MISC 0x3080 +#define FVTHROT_VBLANK_SEL (1 << 9) +#define FVTHROT_PWM_FEEDBACK_DIV_REG1 0x308c +#define RANGE0_PWM_FEEDBACK_DIV(x) ((x) << 0) +#define RANGE0_PWM_FEEDBACK_DIV_MASK 0xfff +#define RANGE0_PWM_FEEDBACK_DIV_SHIFT 0 +#define RANGE_PWM_FEEDBACK_DIV_EN (1 << 12) +#define FVTHROT_PWM_FEEDBACK_DIV_REG2 0x3090 +#define RANGE1_PWM_FEEDBACK_DIV(x) ((x) << 0) +#define RANGE1_PWM_FEEDBACK_DIV_MASK 0xfff +#define RANGE1_PWM_FEEDBACK_DIV_SHIFT 0 +#define RANGE2_PWM_FEEDBACK_DIV(x) ((x) << 12) +#define RANGE2_PWM_FEEDBACK_DIV_MASK (0xfff << 12) +#define RANGE2_PWM_FEEDBACK_DIV_SHIFT 12 +#define FVTHROT_PWM_FEEDBACK_DIV_REG3 0x3094 +#define RANGE0_PWM(x) ((x) << 0) +#define RANGE0_PWM_MASK 0xfff +#define RANGE0_PWM_SHIFT 0 +#define RANGE1_PWM(x) ((x) << 12) +#define RANGE1_PWM_MASK (0xfff << 12) +#define RANGE1_PWM_SHIFT 12 +#define FVTHROT_PWM_FEEDBACK_DIV_REG4 0x3098 +#define RANGE2_PWM(x) ((x) << 0) +#define RANGE2_PWM_MASK 0xfff +#define RANGE2_PWM_SHIFT 0 +#define RANGE3_PWM(x) ((x) << 12) +#define RANGE3_PWM_MASK (0xfff << 12) +#define RANGE3_PWM_SHIFT 12 +#define FVTHROT_SLOW_CLK_FEEDBACK_DIV_REG1 0x30ac +#define RANGE0_SLOW_CLK_FEEDBACK_DIV(x) ((x) << 0) +#define RANGE0_SLOW_CLK_FEEDBACK_DIV_MASK 0xfff +#define RANGE0_SLOW_CLK_FEEDBACK_DIV_SHIFT 0 +#define RANGE_SLOW_CLK_FEEDBACK_DIV_EN (1 << 12) + +#define GFX_MACRO_BYPASS_CNTL 0x30c0 +#define SPLL_BYPASS_CNTL (1 << 0) +#define UPLL_BYPASS_CNTL (1 << 1) + +#endif diff --git a/sys/dev/drm/radeon/rv200d.h b/sys/dev/drm/radeon/rv200d.h index 8b6c5f6046..c5b398330c 100644 --- a/sys/dev/drm/radeon/rv200d.h +++ b/sys/dev/drm/radeon/rv200d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rv200d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RV200D_H__ #define __RV200D_H__ diff --git a/sys/dev/drm/radeon/rv250d.h b/sys/dev/drm/radeon/rv250d.h index f9e14fc9f9..e5a70b06fe 100644 --- a/sys/dev/drm/radeon/rv250d.h +++ b/sys/dev/drm/radeon/rv250d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rv250d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RV250D_H__ #define __RV250D_H__ diff --git a/sys/dev/drm/radeon/rv350d.h b/sys/dev/drm/radeon/rv350d.h index 2c9de0ec78..c75c5ed9e6 100644 --- a/sys/dev/drm/radeon/rv350d.h +++ b/sys/dev/drm/radeon/rv350d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rv350d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RV350D_H__ #define __RV350D_H__ diff --git a/sys/dev/drm/radeon/rv515.c b/sys/dev/drm/radeon/rv515.c index 88ef31beb8..57e2a05584 100644 --- a/sys/dev/drm/radeon/rv515.c +++ b/sys/dev/drm/radeon/rv515.c @@ -937,13 +937,16 @@ struct rv515_watermark { }; static void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, - struct radeon_crtc *crtc, - struct rv515_watermark *wm) + struct radeon_crtc *crtc, + struct rv515_watermark *wm, + bool low) { struct drm_display_mode *mode = &crtc->base.mode; fixed20_12 a, b, c; fixed20_12 pclk, request_fifo_depth, tolerable_latency, estimated_width; fixed20_12 consumption_time, line_time, chunk_time, read_delay_latency; + fixed20_12 sclk; + u32 selected_sclk; if (!crtc->base.enabled) { /* FIXME: wouldn't it better to set priority mark to maximum */ @@ -951,6 +954,18 @@ static void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, return; } + /* rv6xx, rv7xx */ + if ((rdev->family >= CHIP_RV610) && + (rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) + selected_sclk = radeon_dpm_get_sclk(rdev, low); + else + selected_sclk = rdev->pm.current_sclk; + + /* sclk in Mhz */ + a.full = dfixed_const(100); + sclk.full = dfixed_const(selected_sclk); + sclk.full = dfixed_div(sclk, a); + if (crtc->vsc.full > dfixed_const(2)) wm->num_line_pair.full = dfixed_const(2); else @@ -1016,7 +1031,7 @@ static void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, * sclk = system clock(Mhz) */ a.full = dfixed_const(600 * 1000); - chunk_time.full = dfixed_div(a, rdev->pm.sclk); + chunk_time.full = dfixed_div(a, sclk); read_delay_latency.full = dfixed_const(1000); /* Determine the worst case latency @@ -1077,152 +1092,177 @@ static void rv515_crtc_bandwidth_compute(struct radeon_device *rdev, } } -void rv515_bandwidth_avivo_update(struct radeon_device *rdev) +static void rv515_compute_mode_priority(struct radeon_device *rdev, + struct rv515_watermark *wm0, + struct rv515_watermark *wm1, + struct drm_display_mode *mode0, + struct drm_display_mode *mode1, + u32 *d1mode_priority_a_cnt, + u32 *d2mode_priority_a_cnt) { - struct drm_display_mode *mode0 = NULL; - struct drm_display_mode *mode1 = NULL; - struct rv515_watermark wm0; - struct rv515_watermark wm1; - u32 tmp; - u32 d1mode_priority_a_cnt = MODE_PRIORITY_OFF; - u32 d2mode_priority_a_cnt = MODE_PRIORITY_OFF; fixed20_12 priority_mark02, priority_mark12, fill_rate; fixed20_12 a, b; - if (rdev->mode_info.crtcs[0]->base.enabled) - mode0 = &rdev->mode_info.crtcs[0]->base.mode; - if (rdev->mode_info.crtcs[1]->base.enabled) - mode1 = &rdev->mode_info.crtcs[1]->base.mode; - rs690_line_buffer_adjust(rdev, mode0, mode1); - - rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0); - rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1); - - tmp = wm0.lb_request_fifo_depth; - tmp |= wm1.lb_request_fifo_depth << 16; - WREG32(LB_MAX_REQ_OUTSTANDING, tmp); + *d1mode_priority_a_cnt = MODE_PRIORITY_OFF; + *d2mode_priority_a_cnt = MODE_PRIORITY_OFF; if (mode0 && mode1) { - if (dfixed_trunc(wm0.dbpp) > 64) - a.full = dfixed_div(wm0.dbpp, wm0.num_line_pair); + if (dfixed_trunc(wm0->dbpp) > 64) + a.full = dfixed_div(wm0->dbpp, wm0->num_line_pair); else - a.full = wm0.num_line_pair.full; - if (dfixed_trunc(wm1.dbpp) > 64) - b.full = dfixed_div(wm1.dbpp, wm1.num_line_pair); + a.full = wm0->num_line_pair.full; + if (dfixed_trunc(wm1->dbpp) > 64) + b.full = dfixed_div(wm1->dbpp, wm1->num_line_pair); else - b.full = wm1.num_line_pair.full; + b.full = wm1->num_line_pair.full; a.full += b.full; - fill_rate.full = dfixed_div(wm0.sclk, a); - if (wm0.consumption_rate.full > fill_rate.full) { - b.full = wm0.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm0.active_time); + fill_rate.full = dfixed_div(wm0->sclk, a); + if (wm0->consumption_rate.full > fill_rate.full) { + b.full = wm0->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm0->active_time); a.full = dfixed_const(16); b.full = dfixed_div(b, a); - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); priority_mark02.full = a.full + b.full; } else { - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark02.full = dfixed_div(a, b); } - if (wm1.consumption_rate.full > fill_rate.full) { - b.full = wm1.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm1.active_time); + if (wm1->consumption_rate.full > fill_rate.full) { + b.full = wm1->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm1->active_time); a.full = dfixed_const(16); b.full = dfixed_div(b, a); - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); priority_mark12.full = a.full + b.full; } else { - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } - if (wm0.priority_mark.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark.full; + if (wm0->priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark.full; if (dfixed_trunc(priority_mark02) < 0) priority_mark02.full = 0; - if (wm0.priority_mark_max.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark_max.full; - if (wm1.priority_mark.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark.full; + if (wm0->priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark_max.full; + if (wm1->priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark.full; if (dfixed_trunc(priority_mark12) < 0) priority_mark12.full = 0; - if (wm1.priority_mark_max.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark_max.full; - d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); - d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); + if (wm1->priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark_max.full; + *d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); + *d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); if (rdev->disp_priority == 2) { - d1mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; - d2mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; + *d1mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; + *d2mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; } } else if (mode0) { - if (dfixed_trunc(wm0.dbpp) > 64) - a.full = dfixed_div(wm0.dbpp, wm0.num_line_pair); + if (dfixed_trunc(wm0->dbpp) > 64) + a.full = dfixed_div(wm0->dbpp, wm0->num_line_pair); else - a.full = wm0.num_line_pair.full; - fill_rate.full = dfixed_div(wm0.sclk, a); - if (wm0.consumption_rate.full > fill_rate.full) { - b.full = wm0.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm0.active_time); + a.full = wm0->num_line_pair.full; + fill_rate.full = dfixed_div(wm0->sclk, a); + if (wm0->consumption_rate.full > fill_rate.full) { + b.full = wm0->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm0->active_time); a.full = dfixed_const(16); b.full = dfixed_div(b, a); - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); priority_mark02.full = a.full + b.full; } else { - a.full = dfixed_mul(wm0.worst_case_latency, - wm0.consumption_rate); + a.full = dfixed_mul(wm0->worst_case_latency, + wm0->consumption_rate); b.full = dfixed_const(16); priority_mark02.full = dfixed_div(a, b); } - if (wm0.priority_mark.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark.full; + if (wm0->priority_mark.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark.full; if (dfixed_trunc(priority_mark02) < 0) priority_mark02.full = 0; - if (wm0.priority_mark_max.full > priority_mark02.full) - priority_mark02.full = wm0.priority_mark_max.full; - d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); + if (wm0->priority_mark_max.full > priority_mark02.full) + priority_mark02.full = wm0->priority_mark_max.full; + *d1mode_priority_a_cnt = dfixed_trunc(priority_mark02); if (rdev->disp_priority == 2) - d1mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; + *d1mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; } else if (mode1) { - if (dfixed_trunc(wm1.dbpp) > 64) - a.full = dfixed_div(wm1.dbpp, wm1.num_line_pair); + if (dfixed_trunc(wm1->dbpp) > 64) + a.full = dfixed_div(wm1->dbpp, wm1->num_line_pair); else - a.full = wm1.num_line_pair.full; - fill_rate.full = dfixed_div(wm1.sclk, a); - if (wm1.consumption_rate.full > fill_rate.full) { - b.full = wm1.consumption_rate.full - fill_rate.full; - b.full = dfixed_mul(b, wm1.active_time); + a.full = wm1->num_line_pair.full; + fill_rate.full = dfixed_div(wm1->sclk, a); + if (wm1->consumption_rate.full > fill_rate.full) { + b.full = wm1->consumption_rate.full - fill_rate.full; + b.full = dfixed_mul(b, wm1->active_time); a.full = dfixed_const(16); b.full = dfixed_div(b, a); - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); priority_mark12.full = a.full + b.full; } else { - a.full = dfixed_mul(wm1.worst_case_latency, - wm1.consumption_rate); + a.full = dfixed_mul(wm1->worst_case_latency, + wm1->consumption_rate); b.full = dfixed_const(16 * 1000); priority_mark12.full = dfixed_div(a, b); } - if (wm1.priority_mark.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark.full; + if (wm1->priority_mark.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark.full; if (dfixed_trunc(priority_mark12) < 0) priority_mark12.full = 0; - if (wm1.priority_mark_max.full > priority_mark12.full) - priority_mark12.full = wm1.priority_mark_max.full; - d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); + if (wm1->priority_mark_max.full > priority_mark12.full) + priority_mark12.full = wm1->priority_mark_max.full; + *d2mode_priority_a_cnt = dfixed_trunc(priority_mark12); if (rdev->disp_priority == 2) - d2mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; + *d2mode_priority_a_cnt |= MODE_PRIORITY_ALWAYS_ON; } +} + +void rv515_bandwidth_avivo_update(struct radeon_device *rdev) +{ + struct drm_display_mode *mode0 = NULL; + struct drm_display_mode *mode1 = NULL; + struct rv515_watermark wm0_high, wm0_low; + struct rv515_watermark wm1_high, wm1_low; + u32 tmp; + u32 d1mode_priority_a_cnt, d1mode_priority_b_cnt; + u32 d2mode_priority_a_cnt, d2mode_priority_b_cnt; + + if (rdev->mode_info.crtcs[0]->base.enabled) + mode0 = &rdev->mode_info.crtcs[0]->base.mode; + if (rdev->mode_info.crtcs[1]->base.enabled) + mode1 = &rdev->mode_info.crtcs[1]->base.mode; + rs690_line_buffer_adjust(rdev, mode0, mode1); + + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0_high, false); + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1_high, false); + + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[0], &wm0_low, false); + rv515_crtc_bandwidth_compute(rdev, rdev->mode_info.crtcs[1], &wm1_low, false); + + tmp = wm0_high.lb_request_fifo_depth; + tmp |= wm1_high.lb_request_fifo_depth << 16; + WREG32(LB_MAX_REQ_OUTSTANDING, tmp); + + rv515_compute_mode_priority(rdev, + &wm0_high, &wm1_high, + mode0, mode1, + &d1mode_priority_a_cnt, &d2mode_priority_a_cnt); + rv515_compute_mode_priority(rdev, + &wm0_low, &wm1_low, + mode0, mode1, + &d1mode_priority_b_cnt, &d2mode_priority_b_cnt); WREG32(D1MODE_PRIORITY_A_CNT, d1mode_priority_a_cnt); - WREG32(D1MODE_PRIORITY_B_CNT, d1mode_priority_a_cnt); + WREG32(D1MODE_PRIORITY_B_CNT, d1mode_priority_b_cnt); WREG32(D2MODE_PRIORITY_A_CNT, d2mode_priority_a_cnt); - WREG32(D2MODE_PRIORITY_B_CNT, d2mode_priority_a_cnt); + WREG32(D2MODE_PRIORITY_B_CNT, d2mode_priority_b_cnt); } void rv515_bandwidth_update(struct radeon_device *rdev) diff --git a/sys/dev/drm/radeon/rv515d.h b/sys/dev/drm/radeon/rv515d.h index 9f2d573fd5..6927a200da 100644 --- a/sys/dev/drm/radeon/rv515d.h +++ b/sys/dev/drm/radeon/rv515d.h @@ -24,8 +24,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rv515d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __RV515D_H__ #define __RV515D_H__ diff --git a/sys/dev/drm/radeon/rv6xx_dpm.c b/sys/dev/drm/radeon/rv6xx_dpm.c new file mode 100644 index 0000000000..3da1ccf51b --- /dev/null +++ b/sys/dev/drm/radeon/rv6xx_dpm.c @@ -0,0 +1,2123 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "rv6xxd.h" +#include "r600_dpm.h" +#include "rv6xx_dpm.h" +#include "atom.h" +#include + +static u32 rv6xx_scale_count_given_unit(struct radeon_device *rdev, + u32 unscaled_count, u32 unit); + +static struct rv6xx_ps *rv6xx_get_ps(struct radeon_ps *rps) +{ + struct rv6xx_ps *ps = rps->ps_priv; + + return ps; +} + +static struct rv6xx_power_info *rv6xx_get_pi(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void rv6xx_force_pcie_gen1(struct radeon_device *rdev) +{ + u32 tmp; + int i; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + tmp &= LC_GEN2_EN; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + tmp |= LC_INITIATE_LINK_SPEED_CHANGE; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (!(RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL) & LC_CURRENT_DATA_RATE)) + break; + DRM_UDELAY(1); + } + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + tmp &= ~LC_INITIATE_LINK_SPEED_CHANGE; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); +} + +static void rv6xx_enable_pcie_gen2_support(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) { + tmp |= LC_GEN2_EN; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + } +} + +static void rv6xx_enable_bif_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL) & ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + if (enable) + tmp |= LC_HW_VOLTAGE_IF_CONTROL(1); + else + tmp |= LC_HW_VOLTAGE_IF_CONTROL(0); + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); +} + +static void rv6xx_enable_l0s(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL) & ~LC_L0S_INACTIVITY_MASK; + tmp |= LC_L0S_INACTIVITY(3); + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); +} + +static void rv6xx_enable_l1(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL); + tmp &= ~LC_L1_INACTIVITY_MASK; + tmp |= LC_L1_INACTIVITY(4); + tmp &= ~LC_PMI_TO_L1_DIS; + tmp &= ~LC_ASPM_TO_L1_DIS; + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); +} + +static void rv6xx_enable_pll_sleep_in_l1(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL) & ~LC_L1_INACTIVITY_MASK; + tmp |= LC_L1_INACTIVITY(8); + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); + + /* NOTE, this is a PCIE indirect reg, not PCIE PORT */ + tmp = RREG32_PCIE(PCIE_P_CNTL); + tmp |= P_PLL_PWRDN_IN_L1L23; + tmp &= ~P_PLL_BUF_PDNB; + tmp &= ~P_PLL_PDNB; + tmp |= P_ALLOW_PRX_FRONTEND_SHUTOFF; + WREG32_PCIE(PCIE_P_CNTL, tmp); +} + +static int rv6xx_convert_clock_to_stepping(struct radeon_device *rdev, + u32 clock, struct rv6xx_sclk_stepping *step) +{ + int ret; + struct atom_clock_dividers dividers; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + clock, false, ÷rs); + if (ret) + return ret; + + if (dividers.enable_post_div) + step->post_divider = 2 + (dividers.post_div & 0xF) + (dividers.post_div >> 4); + else + step->post_divider = 1; + + step->vco_frequency = clock * step->post_divider; + + return 0; +} + +static void rv6xx_output_stepping(struct radeon_device *rdev, + u32 step_index, struct rv6xx_sclk_stepping *step) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + u32 ref_clk = rdev->clock.spll.reference_freq; + u32 fb_divider; + u32 spll_step_count = rv6xx_scale_count_given_unit(rdev, + R600_SPLLSTEPTIME_DFLT * + pi->spll_ref_div, + R600_SPLLSTEPUNIT_DFLT); + + r600_engine_clock_entry_enable(rdev, step_index, true); + r600_engine_clock_entry_enable_pulse_skipping(rdev, step_index, false); + + if (step->post_divider == 1) + r600_engine_clock_entry_enable_post_divider(rdev, step_index, false); + else { + u32 lo_len = (step->post_divider - 2) / 2; + u32 hi_len = step->post_divider - 2 - lo_len; + + r600_engine_clock_entry_enable_post_divider(rdev, step_index, true); + r600_engine_clock_entry_set_post_divider(rdev, step_index, (hi_len << 4) | lo_len); + } + + fb_divider = ((step->vco_frequency * pi->spll_ref_div) / ref_clk) >> + pi->fb_div_scale; + + r600_engine_clock_entry_set_reference_divider(rdev, step_index, + pi->spll_ref_div - 1); + r600_engine_clock_entry_set_feedback_divider(rdev, step_index, fb_divider); + r600_engine_clock_entry_set_step_time(rdev, step_index, spll_step_count); + +} + +static struct rv6xx_sclk_stepping rv6xx_next_vco_step(struct radeon_device *rdev, + struct rv6xx_sclk_stepping *cur, + bool increasing_vco, u32 step_size) +{ + struct rv6xx_sclk_stepping next; + + next.post_divider = cur->post_divider; + + if (increasing_vco) + next.vco_frequency = (cur->vco_frequency * (100 + step_size)) / 100; + else + next.vco_frequency = (cur->vco_frequency * 100 + 99 + step_size) / (100 + step_size); + + return next; +} + +static bool rv6xx_can_step_post_div(struct radeon_device *rdev, + struct rv6xx_sclk_stepping *cur, + struct rv6xx_sclk_stepping *target) +{ + return (cur->post_divider > target->post_divider) && + ((cur->vco_frequency * target->post_divider) <= + (target->vco_frequency * (cur->post_divider - 1))); +} + +static struct rv6xx_sclk_stepping rv6xx_next_post_div_step(struct radeon_device *rdev, + struct rv6xx_sclk_stepping *cur, + struct rv6xx_sclk_stepping *target) +{ + struct rv6xx_sclk_stepping next = *cur; + + while (rv6xx_can_step_post_div(rdev, &next, target)) + next.post_divider--; + + return next; +} + +static bool rv6xx_reached_stepping_target(struct radeon_device *rdev, + struct rv6xx_sclk_stepping *cur, + struct rv6xx_sclk_stepping *target, + bool increasing_vco) +{ + return (increasing_vco && (cur->vco_frequency >= target->vco_frequency)) || + (!increasing_vco && (cur->vco_frequency <= target->vco_frequency)); +} + +static void rv6xx_generate_steps(struct radeon_device *rdev, + u32 low, u32 high, + u32 start_index, u8 *end_index) +{ + struct rv6xx_sclk_stepping cur; + struct rv6xx_sclk_stepping target; + bool increasing_vco; + u32 step_index = start_index; + + rv6xx_convert_clock_to_stepping(rdev, low, &cur); + rv6xx_convert_clock_to_stepping(rdev, high, &target); + + rv6xx_output_stepping(rdev, step_index++, &cur); + + increasing_vco = (target.vco_frequency >= cur.vco_frequency); + + if (target.post_divider > cur.post_divider) + cur.post_divider = target.post_divider; + + while (1) { + struct rv6xx_sclk_stepping next; + + if (rv6xx_can_step_post_div(rdev, &cur, &target)) + next = rv6xx_next_post_div_step(rdev, &cur, &target); + else + next = rv6xx_next_vco_step(rdev, &cur, increasing_vco, R600_VCOSTEPPCT_DFLT); + + if (rv6xx_reached_stepping_target(rdev, &next, &target, increasing_vco)) { + struct rv6xx_sclk_stepping tiny = + rv6xx_next_vco_step(rdev, &target, !increasing_vco, R600_ENDINGVCOSTEPPCT_DFLT); + tiny.post_divider = next.post_divider; + + if (!rv6xx_reached_stepping_target(rdev, &tiny, &cur, !increasing_vco)) + rv6xx_output_stepping(rdev, step_index++, &tiny); + + if ((next.post_divider != target.post_divider) && + (next.vco_frequency != target.vco_frequency)) { + struct rv6xx_sclk_stepping final_vco; + + final_vco.vco_frequency = target.vco_frequency; + final_vco.post_divider = next.post_divider; + + rv6xx_output_stepping(rdev, step_index++, &final_vco); + } + + rv6xx_output_stepping(rdev, step_index++, &target); + break; + } else + rv6xx_output_stepping(rdev, step_index++, &next); + + cur = next; + } + + *end_index = (u8)step_index - 1; + +} + +static void rv6xx_generate_single_step(struct radeon_device *rdev, + u32 clock, u32 index) +{ + struct rv6xx_sclk_stepping step; + + rv6xx_convert_clock_to_stepping(rdev, clock, &step); + rv6xx_output_stepping(rdev, index, &step); +} + +static void rv6xx_invalidate_intermediate_steps_range(struct radeon_device *rdev, + u32 start_index, u32 end_index) +{ + u32 step_index; + + for (step_index = start_index + 1; step_index < end_index; step_index++) + r600_engine_clock_entry_enable(rdev, step_index, false); +} + +static void rv6xx_set_engine_spread_spectrum_clk_s(struct radeon_device *rdev, + u32 index, u32 clk_s) +{ + WREG32_P(CG_SPLL_SPREAD_SPECTRUM_LOW + (index * 4), + CLKS(clk_s), ~CLKS_MASK); +} + +static void rv6xx_set_engine_spread_spectrum_clk_v(struct radeon_device *rdev, + u32 index, u32 clk_v) +{ + WREG32_P(CG_SPLL_SPREAD_SPECTRUM_LOW + (index * 4), + CLKV(clk_v), ~CLKV_MASK); +} + +static void rv6xx_enable_engine_spread_spectrum(struct radeon_device *rdev, + u32 index, bool enable) +{ + if (enable) + WREG32_P(CG_SPLL_SPREAD_SPECTRUM_LOW + (index * 4), + SSEN, ~SSEN); + else + WREG32_P(CG_SPLL_SPREAD_SPECTRUM_LOW + (index * 4), + 0, ~SSEN); +} + +static void rv6xx_set_memory_spread_spectrum_clk_s(struct radeon_device *rdev, + u32 clk_s) +{ + WREG32_P(CG_MPLL_SPREAD_SPECTRUM, CLKS(clk_s), ~CLKS_MASK); +} + +static void rv6xx_set_memory_spread_spectrum_clk_v(struct radeon_device *rdev, + u32 clk_v) +{ + WREG32_P(CG_MPLL_SPREAD_SPECTRUM, CLKV(clk_v), ~CLKV_MASK); +} + +static void rv6xx_enable_memory_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(CG_MPLL_SPREAD_SPECTRUM, SSEN, ~SSEN); + else + WREG32_P(CG_MPLL_SPREAD_SPECTRUM, 0, ~SSEN); +} + +static void rv6xx_enable_dynamic_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, DYN_SPREAD_SPECTRUM_EN, ~DYN_SPREAD_SPECTRUM_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~DYN_SPREAD_SPECTRUM_EN); +} + +static void rv6xx_memory_clock_entry_enable_post_divider(struct radeon_device *rdev, + u32 index, bool enable) +{ + if (enable) + WREG32_P(MPLL_FREQ_LEVEL_0 + (index * 4), + LEVEL0_MPLL_DIV_EN, ~LEVEL0_MPLL_DIV_EN); + else + WREG32_P(MPLL_FREQ_LEVEL_0 + (index * 4), 0, ~LEVEL0_MPLL_DIV_EN); +} + +static void rv6xx_memory_clock_entry_set_post_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(MPLL_FREQ_LEVEL_0 + (index * 4), + LEVEL0_MPLL_POST_DIV(divider), ~LEVEL0_MPLL_POST_DIV_MASK); +} + +static void rv6xx_memory_clock_entry_set_feedback_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(MPLL_FREQ_LEVEL_0 + (index * 4), LEVEL0_MPLL_FB_DIV(divider), + ~LEVEL0_MPLL_FB_DIV_MASK); +} + +static void rv6xx_memory_clock_entry_set_reference_divider(struct radeon_device *rdev, + u32 index, u32 divider) +{ + WREG32_P(MPLL_FREQ_LEVEL_0 + (index * 4), + LEVEL0_MPLL_REF_DIV(divider), ~LEVEL0_MPLL_REF_DIV_MASK); +} + +static void rv6xx_vid_response_set_brt(struct radeon_device *rdev, u32 rt) +{ + WREG32_P(VID_RT, BRT(rt), ~BRT_MASK); +} + +static void rv6xx_enable_engine_feedback_and_reference_sync(struct radeon_device *rdev) +{ + WREG32_P(SPLL_CNTL_MODE, SPLL_DIV_SYNC, ~SPLL_DIV_SYNC); +} + +static u64 rv6xx_clocks_per_unit(u32 unit) +{ + u64 tmp = 1 << (2 * unit); + + return tmp; +} + +static u32 rv6xx_scale_count_given_unit(struct radeon_device *rdev, + u32 unscaled_count, u32 unit) +{ + u32 count_per_unit = (u32)rv6xx_clocks_per_unit(unit); + + return (unscaled_count + count_per_unit - 1) / count_per_unit; +} + +static u32 rv6xx_compute_count_for_delay(struct radeon_device *rdev, + u32 delay_us, u32 unit) +{ + u32 ref_clk = rdev->clock.spll.reference_freq; + + return rv6xx_scale_count_given_unit(rdev, delay_us * (ref_clk / 100), unit); +} + +static void rv6xx_calculate_engine_speed_stepping_parameters(struct radeon_device *rdev, + struct rv6xx_ps *state) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.sclks[R600_POWER_LEVEL_LOW] = + state->low.sclk; + pi->hw.sclks[R600_POWER_LEVEL_MEDIUM] = + state->medium.sclk; + pi->hw.sclks[R600_POWER_LEVEL_HIGH] = + state->high.sclk; + + pi->hw.low_sclk_index = R600_POWER_LEVEL_LOW; + pi->hw.medium_sclk_index = R600_POWER_LEVEL_MEDIUM; + pi->hw.high_sclk_index = R600_POWER_LEVEL_HIGH; +} + +static void rv6xx_calculate_memory_clock_stepping_parameters(struct radeon_device *rdev, + struct rv6xx_ps *state) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.mclks[R600_POWER_LEVEL_CTXSW] = + state->high.mclk; + pi->hw.mclks[R600_POWER_LEVEL_HIGH] = + state->high.mclk; + pi->hw.mclks[R600_POWER_LEVEL_MEDIUM] = + state->medium.mclk; + pi->hw.mclks[R600_POWER_LEVEL_LOW] = + state->low.mclk; + + pi->hw.high_mclk_index = R600_POWER_LEVEL_HIGH; + + if (state->high.mclk == state->medium.mclk) + pi->hw.medium_mclk_index = + pi->hw.high_mclk_index; + else + pi->hw.medium_mclk_index = R600_POWER_LEVEL_MEDIUM; + + + if (state->medium.mclk == state->low.mclk) + pi->hw.low_mclk_index = + pi->hw.medium_mclk_index; + else + pi->hw.low_mclk_index = R600_POWER_LEVEL_LOW; +} + +static void rv6xx_calculate_voltage_stepping_parameters(struct radeon_device *rdev, + struct rv6xx_ps *state) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.vddc[R600_POWER_LEVEL_CTXSW] = state->high.vddc; + pi->hw.vddc[R600_POWER_LEVEL_HIGH] = state->high.vddc; + pi->hw.vddc[R600_POWER_LEVEL_MEDIUM] = state->medium.vddc; + pi->hw.vddc[R600_POWER_LEVEL_LOW] = state->low.vddc; + + pi->hw.backbias[R600_POWER_LEVEL_CTXSW] = + (state->high.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? true : false; + pi->hw.backbias[R600_POWER_LEVEL_HIGH] = + (state->high.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? true : false; + pi->hw.backbias[R600_POWER_LEVEL_MEDIUM] = + (state->medium.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? true : false; + pi->hw.backbias[R600_POWER_LEVEL_LOW] = + (state->low.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? true : false; + + pi->hw.pcie_gen2[R600_POWER_LEVEL_HIGH] = + (state->high.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? true : false; + pi->hw.pcie_gen2[R600_POWER_LEVEL_MEDIUM] = + (state->medium.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? true : false; + pi->hw.pcie_gen2[R600_POWER_LEVEL_LOW] = + (state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? true : false; + + pi->hw.high_vddc_index = R600_POWER_LEVEL_HIGH; + + if ((state->high.vddc == state->medium.vddc) && + ((state->high.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) == + (state->medium.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE))) + pi->hw.medium_vddc_index = + pi->hw.high_vddc_index; + else + pi->hw.medium_vddc_index = R600_POWER_LEVEL_MEDIUM; + + if ((state->medium.vddc == state->low.vddc) && + ((state->medium.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) == + (state->low.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE))) + pi->hw.low_vddc_index = + pi->hw.medium_vddc_index; + else + pi->hw.medium_vddc_index = R600_POWER_LEVEL_LOW; +} + +static inline u32 rv6xx_calculate_vco_frequency(u32 ref_clock, + struct atom_clock_dividers *dividers, + u32 fb_divider_scale) +{ + return ref_clock * ((dividers->fb_div & ~1) << fb_divider_scale) / + (dividers->ref_div + 1); +} + +static inline u32 rv6xx_calculate_spread_spectrum_clk_v(u32 vco_freq, u32 ref_freq, + u32 ss_rate, u32 ss_percent, + u32 fb_divider_scale) +{ + u32 fb_divider = vco_freq / ref_freq; + + return (ss_percent * ss_rate * 4 * (fb_divider * fb_divider) / + (5375 * ((vco_freq * 10) / (4096 >> fb_divider_scale)))); +} + +static inline u32 rv6xx_calculate_spread_spectrum_clk_s(u32 ss_rate, u32 ref_freq) +{ + return (((ref_freq * 10) / (ss_rate * 2)) - 1) / 4; +} + +static void rv6xx_program_engine_spread_spectrum(struct radeon_device *rdev, + u32 clock, enum r600_power_level level) +{ + u32 ref_clk = rdev->clock.spll.reference_freq; + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + struct atom_clock_dividers dividers; + struct radeon_atom_ss ss; + u32 vco_freq, clk_v, clk_s; + + rv6xx_enable_engine_spread_spectrum(rdev, level, false); + + if (clock && pi->sclk_ss) { + if (radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, clock, false, ÷rs) == 0) { + vco_freq = rv6xx_calculate_vco_frequency(ref_clk, ÷rs, + pi->fb_div_scale); + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + clk_v = rv6xx_calculate_spread_spectrum_clk_v(vco_freq, + (ref_clk / (dividers.ref_div + 1)), + ss.rate, + ss.percentage, + pi->fb_div_scale); + + clk_s = rv6xx_calculate_spread_spectrum_clk_s(ss.rate, + (ref_clk / (dividers.ref_div + 1))); + + rv6xx_set_engine_spread_spectrum_clk_v(rdev, level, clk_v); + rv6xx_set_engine_spread_spectrum_clk_s(rdev, level, clk_s); + rv6xx_enable_engine_spread_spectrum(rdev, level, true); + } + } + } +} + +static void rv6xx_program_sclk_spread_spectrum_parameters_except_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_engine_spread_spectrum(rdev, + pi->hw.sclks[R600_POWER_LEVEL_HIGH], + R600_POWER_LEVEL_HIGH); + + rv6xx_program_engine_spread_spectrum(rdev, + pi->hw.sclks[R600_POWER_LEVEL_MEDIUM], + R600_POWER_LEVEL_MEDIUM); + +} + +static int rv6xx_program_mclk_stepping_entry(struct radeon_device *rdev, + u32 entry, u32 clock) +{ + struct atom_clock_dividers dividers; + + if (radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, clock, false, ÷rs)) + return -EINVAL; + + + rv6xx_memory_clock_entry_set_reference_divider(rdev, entry, dividers.ref_div); + rv6xx_memory_clock_entry_set_feedback_divider(rdev, entry, dividers.fb_div); + rv6xx_memory_clock_entry_set_post_divider(rdev, entry, dividers.post_div); + + if (dividers.enable_post_div) + rv6xx_memory_clock_entry_enable_post_divider(rdev, entry, true); + else + rv6xx_memory_clock_entry_enable_post_divider(rdev, entry, false); + + return 0; +} + +static void rv6xx_program_mclk_stepping_parameters_except_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + int i; + + for (i = 1; i < R600_PM_NUMBER_OF_MCLKS; i++) { + if (pi->hw.mclks[i]) + rv6xx_program_mclk_stepping_entry(rdev, i, + pi->hw.mclks[i]); + } +} + +static void rv6xx_find_memory_clock_with_highest_vco(struct radeon_device *rdev, + u32 requested_memory_clock, + u32 ref_clk, + struct atom_clock_dividers *dividers, + u32 *vco_freq) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + struct atom_clock_dividers req_dividers; + u32 vco_freq_temp; + + if (radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + requested_memory_clock, false, &req_dividers) == 0) { + vco_freq_temp = rv6xx_calculate_vco_frequency(ref_clk, &req_dividers, + pi->fb_div_scale); + + if (vco_freq_temp > *vco_freq) { + *dividers = req_dividers; + *vco_freq = vco_freq_temp; + } + } +} + +static void rv6xx_program_mclk_spread_spectrum_parameters(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + u32 ref_clk = rdev->clock.mpll.reference_freq; + struct atom_clock_dividers dividers; + struct radeon_atom_ss ss; + u32 vco_freq = 0, clk_v, clk_s; + + rv6xx_enable_memory_spread_spectrum(rdev, false); + + if (pi->mclk_ss) { + rv6xx_find_memory_clock_with_highest_vco(rdev, + pi->hw.mclks[pi->hw.high_mclk_index], + ref_clk, + ÷rs, + &vco_freq); + + rv6xx_find_memory_clock_with_highest_vco(rdev, + pi->hw.mclks[pi->hw.medium_mclk_index], + ref_clk, + ÷rs, + &vco_freq); + + rv6xx_find_memory_clock_with_highest_vco(rdev, + pi->hw.mclks[pi->hw.low_mclk_index], + ref_clk, + ÷rs, + &vco_freq); + + if (vco_freq) { + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, vco_freq)) { + clk_v = rv6xx_calculate_spread_spectrum_clk_v(vco_freq, + (ref_clk / (dividers.ref_div + 1)), + ss.rate, + ss.percentage, + pi->fb_div_scale); + + clk_s = rv6xx_calculate_spread_spectrum_clk_s(ss.rate, + (ref_clk / (dividers.ref_div + 1))); + + rv6xx_set_memory_spread_spectrum_clk_v(rdev, clk_v); + rv6xx_set_memory_spread_spectrum_clk_s(rdev, clk_s); + rv6xx_enable_memory_spread_spectrum(rdev, true); + } + } + } +} + +static int rv6xx_program_voltage_stepping_entry(struct radeon_device *rdev, + u32 entry, u16 voltage) +{ + u32 mask, set_pins; + int ret; + + ret = radeon_atom_get_voltage_gpio_settings(rdev, voltage, + SET_VOLTAGE_TYPE_ASIC_VDDC, + &set_pins, &mask); + if (ret) + return ret; + + r600_voltage_control_program_voltages(rdev, entry, set_pins); + + return 0; +} + +static void rv6xx_program_voltage_stepping_parameters_except_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + int i; + + for (i = 1; i < R600_PM_NUMBER_OF_VOLTAGE_LEVELS; i++) + rv6xx_program_voltage_stepping_entry(rdev, i, + pi->hw.vddc[i]); + +} + +static void rv6xx_program_backbias_stepping_parameters_except_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (pi->hw.backbias[1]) + WREG32_P(VID_UPPER_GPIO_CNTL, MEDIUM_BACKBIAS_VALUE, ~MEDIUM_BACKBIAS_VALUE); + else + WREG32_P(VID_UPPER_GPIO_CNTL, 0, ~MEDIUM_BACKBIAS_VALUE); + + if (pi->hw.backbias[2]) + WREG32_P(VID_UPPER_GPIO_CNTL, HIGH_BACKBIAS_VALUE, ~HIGH_BACKBIAS_VALUE); + else + WREG32_P(VID_UPPER_GPIO_CNTL, 0, ~HIGH_BACKBIAS_VALUE); +} + +static void rv6xx_program_sclk_spread_spectrum_parameters_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_engine_spread_spectrum(rdev, + pi->hw.sclks[R600_POWER_LEVEL_LOW], + R600_POWER_LEVEL_LOW); +} + +static void rv6xx_program_mclk_stepping_parameters_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (pi->hw.mclks[0]) + rv6xx_program_mclk_stepping_entry(rdev, 0, + pi->hw.mclks[0]); +} + +static void rv6xx_program_voltage_stepping_parameters_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_voltage_stepping_entry(rdev, 0, + pi->hw.vddc[0]); + +} + +static void rv6xx_program_backbias_stepping_parameters_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (pi->hw.backbias[0]) + WREG32_P(VID_UPPER_GPIO_CNTL, LOW_BACKBIAS_VALUE, ~LOW_BACKBIAS_VALUE); + else + WREG32_P(VID_UPPER_GPIO_CNTL, 0, ~LOW_BACKBIAS_VALUE); +} + +static u32 calculate_memory_refresh_rate(struct radeon_device *rdev, + u32 engine_clock) +{ + u32 dram_rows, dram_refresh_rate; + u32 tmp; + + tmp = (RREG32(RAMCFG) & NOOFROWS_MASK) >> NOOFROWS_SHIFT; + dram_rows = 1 << (tmp + 10); + dram_refresh_rate = 1 << ((RREG32(MC_SEQ_RESERVE_M) & 0x3) + 3); + + return ((engine_clock * 10) * dram_refresh_rate / dram_rows - 32) / 64; +} + +static void rv6xx_program_memory_timing_parameters(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + u32 sqm_ratio; + u32 arb_refresh_rate; + u32 high_clock; + + if (pi->hw.sclks[R600_POWER_LEVEL_HIGH] < + (pi->hw.sclks[R600_POWER_LEVEL_LOW] * 0xFF / 0x40)) + high_clock = pi->hw.sclks[R600_POWER_LEVEL_HIGH]; + else + high_clock = + pi->hw.sclks[R600_POWER_LEVEL_LOW] * 0xFF / 0x40; + + radeon_atom_set_engine_dram_timings(rdev, high_clock, 0); + + sqm_ratio = (STATE0(64 * high_clock / pi->hw.sclks[R600_POWER_LEVEL_LOW]) | + STATE1(64 * high_clock / pi->hw.sclks[R600_POWER_LEVEL_MEDIUM]) | + STATE2(64 * high_clock / pi->hw.sclks[R600_POWER_LEVEL_HIGH]) | + STATE3(64 * high_clock / pi->hw.sclks[R600_POWER_LEVEL_HIGH])); + WREG32(SQM_RATIO, sqm_ratio); + + arb_refresh_rate = + (POWERMODE0(calculate_memory_refresh_rate(rdev, + pi->hw.sclks[R600_POWER_LEVEL_LOW])) | + POWERMODE1(calculate_memory_refresh_rate(rdev, + pi->hw.sclks[R600_POWER_LEVEL_MEDIUM])) | + POWERMODE2(calculate_memory_refresh_rate(rdev, + pi->hw.sclks[R600_POWER_LEVEL_HIGH])) | + POWERMODE3(calculate_memory_refresh_rate(rdev, + pi->hw.sclks[R600_POWER_LEVEL_HIGH]))); + WREG32(ARB_RFSH_RATE, arb_refresh_rate); +} + +static void rv6xx_program_mpll_timing_parameters(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_set_mpll_lock_time(rdev, R600_MPLLLOCKTIME_DFLT * + pi->mpll_ref_div); + r600_set_mpll_reset_time(rdev, R600_MPLLRESETTIME_DFLT); +} + +static void rv6xx_program_bsp(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + u32 ref_clk = rdev->clock.spll.reference_freq; + + r600_calculate_u_and_p(R600_ASI_DFLT, + ref_clk, 16, + &pi->bsp, + &pi->bsu); + + r600_set_bsp(rdev, pi->bsu, pi->bsp); +} + +static void rv6xx_program_at(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_set_at(rdev, + (pi->hw.rp[0] * pi->bsp) / 200, + (pi->hw.rp[1] * pi->bsp) / 200, + (pi->hw.lp[2] * pi->bsp) / 200, + (pi->hw.lp[1] * pi->bsp) / 200); +} + +static void rv6xx_program_git(struct radeon_device *rdev) +{ + r600_set_git(rdev, R600_GICST_DFLT); +} + +static void rv6xx_program_tp(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < R600_PM_NUMBER_OF_TC; i++) + r600_set_tc(rdev, i, r600_utc[i], r600_dtc[i]); + + r600_select_td(rdev, R600_TD_DFLT); +} + +static void rv6xx_program_vc(struct radeon_device *rdev) +{ + r600_set_vrc(rdev, R600_VRC_DFLT); +} + +static void rv6xx_clear_vc(struct radeon_device *rdev) +{ + r600_set_vrc(rdev, 0); +} + +static void rv6xx_program_tpp(struct radeon_device *rdev) +{ + r600_set_tpu(rdev, R600_TPU_DFLT); + r600_set_tpc(rdev, R600_TPC_DFLT); +} + +static void rv6xx_program_sstp(struct radeon_device *rdev) +{ + r600_set_sstu(rdev, R600_SSTU_DFLT); + r600_set_sst(rdev, R600_SST_DFLT); +} + +static void rv6xx_program_fcp(struct radeon_device *rdev) +{ + r600_set_fctu(rdev, R600_FCTU_DFLT); + r600_set_fct(rdev, R600_FCT_DFLT); +} + +static void rv6xx_program_vddc3d_parameters(struct radeon_device *rdev) +{ + r600_set_vddc3d_oorsu(rdev, R600_VDDC3DOORSU_DFLT); + r600_set_vddc3d_oorphc(rdev, R600_VDDC3DOORPHC_DFLT); + r600_set_vddc3d_oorsdc(rdev, R600_VDDC3DOORSDC_DFLT); + r600_set_ctxcgtt3d_rphc(rdev, R600_CTXCGTT3DRPHC_DFLT); + r600_set_ctxcgtt3d_rsdc(rdev, R600_CTXCGTT3DRSDC_DFLT); +} + +static void rv6xx_program_voltage_timing_parameters(struct radeon_device *rdev) +{ + u32 rt; + + r600_vid_rt_set_vru(rdev, R600_VRU_DFLT); + + r600_vid_rt_set_vrt(rdev, + rv6xx_compute_count_for_delay(rdev, + rdev->pm.dpm.voltage_response_time, + R600_VRU_DFLT)); + + rt = rv6xx_compute_count_for_delay(rdev, + rdev->pm.dpm.backbias_response_time, + R600_VRU_DFLT); + + rv6xx_vid_response_set_brt(rdev, (rt + 0x1F) >> 5); +} + +static void rv6xx_program_engine_speed_parameters(struct radeon_device *rdev) +{ + r600_vid_rt_set_ssu(rdev, R600_SPLLSTEPUNIT_DFLT); + rv6xx_enable_engine_feedback_and_reference_sync(rdev); +} + +static u64 rv6xx_get_master_voltage_mask(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + u64 master_mask = 0; + int i; + + for (i = 0; i < R600_PM_NUMBER_OF_VOLTAGE_LEVELS; i++) { + u32 tmp_mask, tmp_set_pins; + int ret; + + ret = radeon_atom_get_voltage_gpio_settings(rdev, + pi->hw.vddc[i], + SET_VOLTAGE_TYPE_ASIC_VDDC, + &tmp_set_pins, &tmp_mask); + + if (ret == 0) + master_mask |= tmp_mask; + } + + return master_mask; +} + +static void rv6xx_program_voltage_gpio_pins(struct radeon_device *rdev) +{ + r600_voltage_control_enable_pins(rdev, + rv6xx_get_master_voltage_mask(rdev)); +} + +static void rv6xx_enable_static_voltage_control(struct radeon_device *rdev, + struct radeon_ps *new_ps, + bool enable) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + + if (enable) + radeon_atom_set_voltage(rdev, + new_state->low.vddc, + SET_VOLTAGE_TYPE_ASIC_VDDC); + else + r600_voltage_control_deactivate_static_control(rdev, + rv6xx_get_master_voltage_mask(rdev)); +} + +static void rv6xx_enable_display_gap(struct radeon_device *rdev, bool enable) +{ + if (enable) { + u32 tmp = (DISP1_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM) | + DISP2_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM) | + DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE) | + DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE) | + VBI_TIMER_COUNT(0x3FFF) | + VBI_TIMER_UNIT(7)); + WREG32(CG_DISPLAY_GAP_CNTL, tmp); + + WREG32_P(MCLK_PWRMGT_CNTL, USE_DISPLAY_GAP, ~USE_DISPLAY_GAP); + } else + WREG32_P(MCLK_PWRMGT_CNTL, 0, ~USE_DISPLAY_GAP); +} + +static void rv6xx_program_power_level_enter_state(struct radeon_device *rdev) +{ + r600_power_level_set_enter_index(rdev, R600_POWER_LEVEL_MEDIUM); +} + +static void rv6xx_calculate_t(u32 l_f, u32 h_f, int h, + int d_l, int d_r, u8 *l, u8 *r) +{ + int a_n, a_d, h_r, l_r; + + h_r = d_l; + l_r = 100 - d_r; + + a_n = (int)h_f * d_l + (int)l_f * (h - d_r); + a_d = (int)l_f * l_r + (int)h_f * h_r; + + if (a_d != 0) { + *l = d_l - h_r * a_n / a_d; + *r = d_r + l_r * a_n / a_d; + } +} + +static void rv6xx_calculate_ap(struct radeon_device *rdev, + struct rv6xx_ps *state) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.lp[0] = 0; + pi->hw.rp[R600_PM_NUMBER_OF_ACTIVITY_LEVELS - 1] + = 100; + + rv6xx_calculate_t(state->low.sclk, + state->medium.sclk, + R600_AH_DFLT, + R600_LMP_DFLT, + R600_RLP_DFLT, + &pi->hw.lp[1], + &pi->hw.rp[0]); + + rv6xx_calculate_t(state->medium.sclk, + state->high.sclk, + R600_AH_DFLT, + R600_LHP_DFLT, + R600_RMP_DFLT, + &pi->hw.lp[2], + &pi->hw.rp[1]); + +} + +static void rv6xx_calculate_stepping_parameters(struct radeon_device *rdev, + struct radeon_ps *new_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + + rv6xx_calculate_engine_speed_stepping_parameters(rdev, new_state); + rv6xx_calculate_memory_clock_stepping_parameters(rdev, new_state); + rv6xx_calculate_voltage_stepping_parameters(rdev, new_state); + rv6xx_calculate_ap(rdev, new_state); +} + +static void rv6xx_program_stepping_parameters_except_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_mclk_stepping_parameters_except_lowest_entry(rdev); + if (pi->voltage_control) + rv6xx_program_voltage_stepping_parameters_except_lowest_entry(rdev); + rv6xx_program_backbias_stepping_parameters_except_lowest_entry(rdev); + rv6xx_program_sclk_spread_spectrum_parameters_except_lowest_entry(rdev); + rv6xx_program_mclk_spread_spectrum_parameters(rdev); + rv6xx_program_memory_timing_parameters(rdev); +} + +static void rv6xx_program_stepping_parameters_lowest_entry(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_mclk_stepping_parameters_lowest_entry(rdev); + if (pi->voltage_control) + rv6xx_program_voltage_stepping_parameters_lowest_entry(rdev); + rv6xx_program_backbias_stepping_parameters_lowest_entry(rdev); + rv6xx_program_sclk_spread_spectrum_parameters_lowest_entry(rdev); +} + +static void rv6xx_program_power_level_low(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_LOW, + pi->hw.low_vddc_index); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_LOW, + pi->hw.low_mclk_index); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_LOW, + pi->hw.low_sclk_index); + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_LOW, + R600_DISPLAY_WATERMARK_LOW); + r600_power_level_set_pcie_gen2(rdev, R600_POWER_LEVEL_LOW, + pi->hw.pcie_gen2[R600_POWER_LEVEL_LOW]); +} + +static void rv6xx_program_power_level_low_to_lowest_state(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_LOW, 0); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_LOW, 0); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_LOW, 0); + + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_LOW, + R600_DISPLAY_WATERMARK_LOW); + + r600_power_level_set_pcie_gen2(rdev, R600_POWER_LEVEL_LOW, + pi->hw.pcie_gen2[R600_POWER_LEVEL_LOW]); + +} + +static void rv6xx_program_power_level_medium(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.medium_vddc_index); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.medium_mclk_index); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.medium_sclk_index); + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_MEDIUM, + R600_DISPLAY_WATERMARK_LOW); + r600_power_level_set_pcie_gen2(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.pcie_gen2[R600_POWER_LEVEL_MEDIUM]); +} + +static void rv6xx_program_power_level_medium_for_transition(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_program_mclk_stepping_entry(rdev, + R600_POWER_LEVEL_CTXSW, + pi->hw.mclks[pi->hw.low_mclk_index]); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_MEDIUM, 1); + + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, + R600_POWER_LEVEL_CTXSW); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.medium_sclk_index); + + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_MEDIUM, + R600_DISPLAY_WATERMARK_LOW); + + rv6xx_enable_engine_spread_spectrum(rdev, R600_POWER_LEVEL_MEDIUM, false); + + r600_power_level_set_pcie_gen2(rdev, R600_POWER_LEVEL_MEDIUM, + pi->hw.pcie_gen2[R600_POWER_LEVEL_LOW]); +} + +static void rv6xx_program_power_level_high(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + r600_power_level_set_voltage_index(rdev, R600_POWER_LEVEL_HIGH, + pi->hw.high_vddc_index); + r600_power_level_set_mem_clock_index(rdev, R600_POWER_LEVEL_HIGH, + pi->hw.high_mclk_index); + r600_power_level_set_eng_clock_index(rdev, R600_POWER_LEVEL_HIGH, + pi->hw.high_sclk_index); + + r600_power_level_set_watermark_id(rdev, R600_POWER_LEVEL_HIGH, + R600_DISPLAY_WATERMARK_HIGH); + + r600_power_level_set_pcie_gen2(rdev, R600_POWER_LEVEL_HIGH, + pi->hw.pcie_gen2[R600_POWER_LEVEL_HIGH]); +} + +static void rv6xx_enable_backbias(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, BACKBIAS_PAD_EN | BACKBIAS_DPM_CNTL, + ~(BACKBIAS_PAD_EN | BACKBIAS_DPM_CNTL)); + else + WREG32_P(GENERAL_PWRMGT, 0, + ~(BACKBIAS_VALUE | BACKBIAS_PAD_EN | BACKBIAS_DPM_CNTL)); +} + +static void rv6xx_program_display_gap(struct radeon_device *rdev) +{ + u32 tmp = RREG32(CG_DISPLAY_GAP_CNTL); + + tmp &= ~(DISP1_GAP_MCHG_MASK | DISP2_GAP_MCHG_MASK); + if (rdev->pm.dpm.new_active_crtcs & 1) { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + } else if (rdev->pm.dpm.new_active_crtcs & 2) { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK); + } else { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + } + WREG32(CG_DISPLAY_GAP_CNTL, tmp); +} + +static void rv6xx_set_sw_voltage_to_safe(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + u16 safe_voltage; + + safe_voltage = (new_state->low.vddc >= old_state->low.vddc) ? + new_state->low.vddc : old_state->low.vddc; + + rv6xx_program_voltage_stepping_entry(rdev, R600_POWER_LEVEL_CTXSW, + safe_voltage); + + WREG32_P(GENERAL_PWRMGT, SW_GPIO_INDEX(R600_POWER_LEVEL_CTXSW), + ~SW_GPIO_INDEX_MASK); +} + +static void rv6xx_set_sw_voltage_to_low(struct radeon_device *rdev, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + + rv6xx_program_voltage_stepping_entry(rdev, R600_POWER_LEVEL_CTXSW, + old_state->low.vddc); + + WREG32_P(GENERAL_PWRMGT, SW_GPIO_INDEX(R600_POWER_LEVEL_CTXSW), + ~SW_GPIO_INDEX_MASK); +} + +static void rv6xx_set_safe_backbias(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + + if ((new_state->low.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) && + (old_state->low.flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE)) + WREG32_P(GENERAL_PWRMGT, BACKBIAS_VALUE, ~BACKBIAS_VALUE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~BACKBIAS_VALUE); +} + +static void rv6xx_set_safe_pcie_gen2(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + + if ((new_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) != + (old_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2)) + rv6xx_force_pcie_gen1(rdev); +} + +static void rv6xx_enable_dynamic_voltage_control(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, VOLT_PWRMGT_EN, ~VOLT_PWRMGT_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~VOLT_PWRMGT_EN); +} + +static void rv6xx_enable_dynamic_backbias_control(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, BACKBIAS_DPM_CNTL, ~BACKBIAS_DPM_CNTL); + else + WREG32_P(GENERAL_PWRMGT, 0, ~BACKBIAS_DPM_CNTL); +} + +static int rv6xx_step_sw_voltage(struct radeon_device *rdev, + u16 initial_voltage, + u16 target_voltage) +{ + u16 current_voltage; + u16 true_target_voltage; + u16 voltage_step; + int signed_voltage_step; + + if ((radeon_atom_get_voltage_step(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, + &voltage_step)) || + (radeon_atom_round_to_true_voltage(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, + initial_voltage, ¤t_voltage)) || + (radeon_atom_round_to_true_voltage(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, + target_voltage, &true_target_voltage))) + return -EINVAL; + + if (true_target_voltage < current_voltage) + signed_voltage_step = -(int)voltage_step; + else + signed_voltage_step = voltage_step; + + while (current_voltage != true_target_voltage) { + current_voltage += signed_voltage_step; + rv6xx_program_voltage_stepping_entry(rdev, R600_POWER_LEVEL_CTXSW, + current_voltage); + msleep((rdev->pm.dpm.voltage_response_time + 999) / 1000); + } + + return 0; +} + +static int rv6xx_step_voltage_if_increasing(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + + if (new_state->low.vddc > old_state->low.vddc) + return rv6xx_step_sw_voltage(rdev, + old_state->low.vddc, + new_state->low.vddc); + + return 0; +} + +static int rv6xx_step_voltage_if_decreasing(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + + if (new_state->low.vddc < old_state->low.vddc) + return rv6xx_step_sw_voltage(rdev, + old_state->low.vddc, + new_state->low.vddc); + else + return 0; +} + +static void rv6xx_enable_high(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if ((pi->restricted_levels < 1) || + (pi->restricted_levels == 3)) + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, true); +} + +static void rv6xx_enable_medium(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (pi->restricted_levels < 2) + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, true); +} + +static void rv6xx_set_dpm_event_sources(struct radeon_device *rdev, u32 sources) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + bool want_thermal_protection; + enum radeon_dpm_event_src dpm_event_src; + + switch (sources) { + case 0: + default: + want_thermal_protection = false; + break; + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGITAL; + break; + + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_EXTERNAL; + break; + + case ((1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL) | + (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL)): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL; + break; + } + + if (want_thermal_protection) { + WREG32_P(CG_THERMAL_CTRL, DPM_EVENT_SRC(dpm_event_src), ~DPM_EVENT_SRC_MASK); + if (pi->thermal_protection) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + } else { + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); + } +} + +static void rv6xx_enable_auto_throttle_source(struct radeon_device *rdev, + enum radeon_dpm_auto_throttle_src source, + bool enable) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (enable) { + if (!(pi->active_auto_throttle_sources & (1 << source))) { + pi->active_auto_throttle_sources |= 1 << source; + rv6xx_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } else { + if (pi->active_auto_throttle_sources & (1 << source)) { + pi->active_auto_throttle_sources &= ~(1 << source); + rv6xx_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } +} + + +static void rv6xx_enable_thermal_protection(struct radeon_device *rdev, + bool enable) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (pi->active_auto_throttle_sources) + r600_enable_thermal_protection(rdev, enable); +} + +static void rv6xx_generate_transition_stepping(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *old_state = rv6xx_get_ps(old_ps); + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_generate_steps(rdev, + old_state->low.sclk, + new_state->low.sclk, + 0, &pi->hw.medium_sclk_index); +} + +static void rv6xx_generate_low_step(struct radeon_device *rdev, + struct radeon_ps *new_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.low_sclk_index = 0; + rv6xx_generate_single_step(rdev, + new_state->low.sclk, + 0); +} + +static void rv6xx_invalidate_intermediate_steps(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + rv6xx_invalidate_intermediate_steps_range(rdev, 0, + pi->hw.medium_sclk_index); +} + +static void rv6xx_generate_stepping_table(struct radeon_device *rdev, + struct radeon_ps *new_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + pi->hw.low_sclk_index = 0; + + rv6xx_generate_steps(rdev, + new_state->low.sclk, + new_state->medium.sclk, + 0, + &pi->hw.medium_sclk_index); + rv6xx_generate_steps(rdev, + new_state->medium.sclk, + new_state->high.sclk, + pi->hw.medium_sclk_index, + &pi->hw.high_sclk_index); +} + +static void rv6xx_enable_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + if (enable) + rv6xx_enable_dynamic_spread_spectrum(rdev, true); + else { + rv6xx_enable_engine_spread_spectrum(rdev, R600_POWER_LEVEL_LOW, false); + rv6xx_enable_engine_spread_spectrum(rdev, R600_POWER_LEVEL_MEDIUM, false); + rv6xx_enable_engine_spread_spectrum(rdev, R600_POWER_LEVEL_HIGH, false); + rv6xx_enable_dynamic_spread_spectrum(rdev, false); + rv6xx_enable_memory_spread_spectrum(rdev, false); + } +} + +static void rv6xx_reset_lvtm_data_sync(struct radeon_device *rdev) +{ + if (ASIC_IS_DCE3(rdev)) + WREG32_P(DCE3_LVTMA_DATA_SYNCHRONIZATION, LVTMA_PFREQCHG, ~LVTMA_PFREQCHG); + else + WREG32_P(LVTMA_DATA_SYNCHRONIZATION, LVTMA_PFREQCHG, ~LVTMA_PFREQCHG); +} + +static void rv6xx_enable_dynamic_pcie_gen2(struct radeon_device *rdev, + struct radeon_ps *new_ps, + bool enable) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + + if (enable) { + rv6xx_enable_bif_dynamic_pcie_gen2(rdev, true); + rv6xx_enable_pcie_gen2_support(rdev); + r600_enable_dynamic_pcie_gen2(rdev, true); + } else { + if (!(new_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2)) + rv6xx_force_pcie_gen1(rdev); + rv6xx_enable_bif_dynamic_pcie_gen2(rdev, false); + r600_enable_dynamic_pcie_gen2(rdev, false); + } +} + +static void rv6xx_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *current_state = rv6xx_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->high.sclk >= current_state->high.sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +static void rv6xx_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv6xx_ps *new_state = rv6xx_get_ps(new_ps); + struct rv6xx_ps *current_state = rv6xx_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->high.sclk < current_state->high.sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +int rv6xx_dpm_enable(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (r600_dynamicpm_enabled(rdev)) + return -EINVAL; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv6xx_enable_backbias(rdev, true); + + if (pi->dynamic_ss) + rv6xx_enable_spread_spectrum(rdev, true); + + rv6xx_program_mpll_timing_parameters(rdev); + rv6xx_program_bsp(rdev); + rv6xx_program_git(rdev); + rv6xx_program_tp(rdev); + rv6xx_program_tpp(rdev); + rv6xx_program_sstp(rdev); + rv6xx_program_fcp(rdev); + rv6xx_program_vddc3d_parameters(rdev); + rv6xx_program_voltage_timing_parameters(rdev); + rv6xx_program_engine_speed_parameters(rdev); + + rv6xx_enable_display_gap(rdev, true); + if (pi->display_gap == false) + rv6xx_enable_display_gap(rdev, false); + + rv6xx_program_power_level_enter_state(rdev); + + rv6xx_calculate_stepping_parameters(rdev, boot_ps); + + if (pi->voltage_control) + rv6xx_program_voltage_gpio_pins(rdev); + + rv6xx_generate_stepping_table(rdev, boot_ps); + + rv6xx_program_stepping_parameters_except_lowest_entry(rdev); + rv6xx_program_stepping_parameters_lowest_entry(rdev); + + rv6xx_program_power_level_low(rdev); + rv6xx_program_power_level_medium(rdev); + rv6xx_program_power_level_high(rdev); + rv6xx_program_vc(rdev); + rv6xx_program_at(rdev); + + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, true); + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, true); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + ret = r600_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + } + + rv6xx_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + r600_start_dpm(rdev); + + if (pi->voltage_control) + rv6xx_enable_static_voltage_control(rdev, boot_ps, false); + + if (pi->dynamic_pcie_gen2) + rv6xx_enable_dynamic_pcie_gen2(rdev, boot_ps, true); + + if (pi->gfx_clock_gating) + r600_gfx_clockgating_enable(rdev, true); + + return 0; +} + +void rv6xx_dpm_disable(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + + if (!r600_dynamicpm_enabled(rdev)) + return; + + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, true); + rv6xx_enable_display_gap(rdev, false); + rv6xx_clear_vc(rdev); + r600_set_at(rdev, 0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF); + + if (pi->thermal_protection) + r600_enable_thermal_protection(rdev, false); + + r600_wait_for_power_level(rdev, R600_POWER_LEVEL_LOW); + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, false); + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv6xx_enable_backbias(rdev, false); + + rv6xx_enable_spread_spectrum(rdev, false); + + if (pi->voltage_control) + rv6xx_enable_static_voltage_control(rdev, boot_ps, true); + + if (pi->dynamic_pcie_gen2) + rv6xx_enable_dynamic_pcie_gen2(rdev, boot_ps, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + if (pi->gfx_clock_gating) + r600_gfx_clockgating_enable(rdev, false); + + r600_stop_dpm(rdev); +} + +int rv6xx_dpm_set_power_state(struct radeon_device *rdev) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + struct radeon_ps *old_ps = rdev->pm.dpm.current_ps; + int ret; + + pi->restricted_levels = 0; + + rv6xx_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + + rv6xx_clear_vc(rdev); + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + r600_set_at(rdev, 0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF); + + if (pi->thermal_protection) + r600_enable_thermal_protection(rdev, false); + + r600_wait_for_power_level(rdev, R600_POWER_LEVEL_LOW); + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, false); + + rv6xx_generate_transition_stepping(rdev, new_ps, old_ps); + rv6xx_program_power_level_medium_for_transition(rdev); + + if (pi->voltage_control) { + rv6xx_set_sw_voltage_to_safe(rdev, new_ps, old_ps); + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + rv6xx_set_sw_voltage_to_low(rdev, old_ps); + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv6xx_set_safe_backbias(rdev, new_ps, old_ps); + + if (pi->dynamic_pcie_gen2) + rv6xx_set_safe_pcie_gen2(rdev, new_ps, old_ps); + + if (pi->voltage_control) + rv6xx_enable_dynamic_voltage_control(rdev, false); + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv6xx_enable_dynamic_backbias_control(rdev, false); + + if (pi->voltage_control) { + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + rv6xx_step_voltage_if_increasing(rdev, new_ps, old_ps); + msleep((rdev->pm.dpm.voltage_response_time + 999) / 1000); + } + + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, true); + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, false); + r600_wait_for_power_level_unequal(rdev, R600_POWER_LEVEL_LOW); + + rv6xx_generate_low_step(rdev, new_ps); + rv6xx_invalidate_intermediate_steps(rdev); + rv6xx_calculate_stepping_parameters(rdev, new_ps); + rv6xx_program_stepping_parameters_lowest_entry(rdev); + rv6xx_program_power_level_low_to_lowest_state(rdev); + + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + r600_wait_for_power_level(rdev, R600_POWER_LEVEL_LOW); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, false); + + if (pi->voltage_control) { + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) { + ret = rv6xx_step_voltage_if_decreasing(rdev, new_ps, old_ps); + if (ret) + return ret; + } + rv6xx_enable_dynamic_voltage_control(rdev, true); + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv6xx_enable_dynamic_backbias_control(rdev, true); + + if (pi->dynamic_pcie_gen2) + rv6xx_enable_dynamic_pcie_gen2(rdev, new_ps, true); + + rv6xx_reset_lvtm_data_sync(rdev); + + rv6xx_generate_stepping_table(rdev, new_ps); + rv6xx_program_stepping_parameters_except_lowest_entry(rdev); + rv6xx_program_power_level_low(rdev); + rv6xx_program_power_level_medium(rdev); + rv6xx_program_power_level_high(rdev); + rv6xx_enable_medium(rdev); + rv6xx_enable_high(rdev); + + if (pi->thermal_protection) + rv6xx_enable_thermal_protection(rdev, true); + rv6xx_program_vc(rdev); + rv6xx_program_at(rdev); + + rv6xx_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + rdev->pm.dpm.forced_level = RADEON_DPM_FORCED_LEVEL_AUTO; + + return 0; +} + +void rv6xx_setup_asic(struct radeon_device *rdev) +{ + r600_enable_acpi_pm(rdev); + + if (radeon_aspm != 0) { + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_ASPM_L0s) + rv6xx_enable_l0s(rdev); + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_ASPM_L1) + rv6xx_enable_l1(rdev); + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_TURNOFFPLL_ASPML1) + rv6xx_enable_pll_sleep_in_l1(rdev); + } +} + +void rv6xx_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + rv6xx_program_display_gap(rdev); +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void rv6xx_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info) +{ + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (r600_is_uvd_state(rps->class, rps->class2)) { + rps->vclk = RV6XX_DEFAULT_VCLK_FREQ; + rps->dclk = RV6XX_DEFAULT_DCLK_FREQ; + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) + rdev->pm.dpm.boot_ps = rps; + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void rv6xx_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct rv6xx_ps *ps = rv6xx_get_ps(rps); + u32 sclk, mclk; + u16 vddc; + struct rv6xx_pl *pl; + + switch (index) { + case 0: + pl = &ps->low; + break; + case 1: + pl = &ps->medium; + break; + case 2: + default: + pl = &ps->high; + break; + } + + sclk = le16_to_cpu(clock_info->r600.usEngineClockLow); + sclk |= clock_info->r600.ucEngineClockHigh << 16; + mclk = le16_to_cpu(clock_info->r600.usMemoryClockLow); + mclk |= clock_info->r600.ucMemoryClockHigh << 16; + + pl->mclk = mclk; + pl->sclk = sclk; + pl->vddc = le16_to_cpu(clock_info->r600.usVDDC); + pl->flags = le32_to_cpu(clock_info->r600.ulFlags); + + /* patch up vddc if necessary */ + if (pl->vddc == 0xff01) { + if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc) == 0) + pl->vddc = vddc; + } + + /* fix up pcie gen2 */ + if (pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) { + if ((rdev->family == CHIP_RV610) || (rdev->family == CHIP_RV630)) { + if (pl->vddc < 1100) + pl->flags &= ~ATOM_PPLIB_R600_FLAGS_PCIEGEN2; + } + } + + /* patch up boot state */ + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + u16 vddc, vddci, mvdd; + radeon_atombios_get_default_voltages(rdev, &vddc, &vddci, &mvdd); + pl->mclk = rdev->clock.default_mclk; + pl->sclk = rdev->clock.default_sclk; + pl->vddc = vddc; + } +} + +static int rv6xx_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j; + union pplib_clock_info *clock_info; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + struct rv6xx_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + power_info->pplib.ucNumStates, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + + for (i = 0; i < power_info->pplib.ucNumStates; i++) { + power_state = (union pplib_power_state *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset) + + i * power_info->pplib.ucStateEntrySize); + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset) + + (power_state->v1.ucNonClockStateIndex * + power_info->pplib.ucNonClockSize)); + if (power_info->pplib.ucStateEntrySize - 1) { + ps = kzalloc(sizeof(struct rv6xx_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + rv6xx_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info); + for (j = 0; j < (power_info->pplib.ucStateEntrySize - 1); j++) { + clock_info = (union pplib_clock_info *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset) + + (power_state->v1.ucClockStateIndices[j] * + power_info->pplib.ucClockInfoSize)); + rv6xx_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], j, + clock_info); + } + } + } + rdev->pm.dpm.num_ps = power_info->pplib.ucNumStates; + return 0; +} + +int rv6xx_dpm_init(struct radeon_device *rdev) +{ + struct radeon_atom_ss ss; + struct atom_clock_dividers dividers; + struct rv6xx_power_info *pi; + int ret; + + pi = kzalloc(sizeof(struct rv6xx_power_info), GFP_KERNEL); + if (pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = pi; + + ret = rv6xx_parse_power_table(rdev); + if (ret) + return ret; + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->spll_ref_div = dividers.ref_div + 1; + else + pi->spll_ref_div = R600_REFERENCEDIVIDER_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->mpll_ref_div = dividers.ref_div + 1; + else + pi->mpll_ref_div = R600_REFERENCEDIVIDER_DFLT; + + if (rdev->family >= CHIP_RV670) + pi->fb_div_scale = 1; + else + pi->fb_div_scale = 0; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0); + + pi->gfx_clock_gating = true; + + pi->sclk_ss = radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, 0); + pi->mclk_ss = radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, 0); + + /* Disable sclk ss, causes hangs on a lot of systems */ + pi->sclk_ss = false; + + if (pi->sclk_ss || pi->mclk_ss) + pi->dynamic_ss = true; + else + pi->dynamic_ss = false; + + pi->dynamic_pcie_gen2 = true; + + if (pi->gfx_clock_gating && + (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE)) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + pi->display_gap = true; + + return 0; +} + +void rv6xx_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct rv6xx_ps *ps = rv6xx_get_ps(rps); + struct rv6xx_pl *pl; + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + pl = &ps->low; + printk("\t\tpower level 0 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + pl = &ps->medium; + printk("\t\tpower level 1 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + pl = &ps->high; + printk("\t\tpower level 2 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + r600_dpm_print_ps_status(rdev, rps); +} + +void rv6xx_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct rv6xx_ps *ps = rv6xx_get_ps(rps); + struct rv6xx_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_PROFILE_INDEX_MASK) >> + CURRENT_PROFILE_INDEX_SHIFT; + + if (current_index > 2) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + if (current_index == 0) + pl = &ps->low; + else if (current_index == 1) + pl = &ps->medium; + else /* current_index == 2 */ + pl = &ps->high; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u mclk: %u vddc: %u\n", + current_index, pl->sclk, pl->mclk, pl->vddc); + } +} + +void rv6xx_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +u32 rv6xx_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct rv6xx_ps *requested_state = rv6xx_get_ps(rdev->pm.dpm.requested_ps); + + if (low) + return requested_state->low.sclk; + else + return requested_state->high.sclk; +} + +u32 rv6xx_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct rv6xx_ps *requested_state = rv6xx_get_ps(rdev->pm.dpm.requested_ps); + + if (low) + return requested_state->low.mclk; + else + return requested_state->high.mclk; +} + +int rv6xx_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + struct rv6xx_power_info *pi = rv6xx_get_pi(rdev); + + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + pi->restricted_levels = 3; + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + pi->restricted_levels = 2; + } else { + pi->restricted_levels = 0; + } + + rv6xx_clear_vc(rdev); + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, true); + r600_set_at(rdev, 0xFFFF, 0xFFFF, 0xFFFF, 0xFFFF); + r600_wait_for_power_level(rdev, R600_POWER_LEVEL_LOW); + r600_power_level_enable(rdev, R600_POWER_LEVEL_HIGH, false); + r600_power_level_enable(rdev, R600_POWER_LEVEL_MEDIUM, false); + rv6xx_enable_medium(rdev); + rv6xx_enable_high(rdev); + if (pi->restricted_levels == 3) + r600_power_level_enable(rdev, R600_POWER_LEVEL_LOW, false); + rv6xx_program_vc(rdev); + rv6xx_program_at(rdev); + + rdev->pm.dpm.forced_level = level; + + return 0; +} diff --git a/sys/dev/drm/radeon/rv6xx_dpm.h b/sys/dev/drm/radeon/rv6xx_dpm.h new file mode 100644 index 0000000000..8035d53ebe --- /dev/null +++ b/sys/dev/drm/radeon/rv6xx_dpm.h @@ -0,0 +1,95 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#ifndef __RV6XX_DPM_H__ +#define __RV6XX_DPM_H__ + +#include "r600_dpm.h" + +/* Represents a single SCLK step. */ +struct rv6xx_sclk_stepping +{ + u32 vco_frequency; + u32 post_divider; +}; + +struct rv6xx_pm_hw_state { + u32 sclks[R600_PM_NUMBER_OF_ACTIVITY_LEVELS]; + u32 mclks[R600_PM_NUMBER_OF_MCLKS]; + u16 vddc[R600_PM_NUMBER_OF_VOLTAGE_LEVELS]; + bool backbias[R600_PM_NUMBER_OF_VOLTAGE_LEVELS]; + bool pcie_gen2[R600_PM_NUMBER_OF_ACTIVITY_LEVELS]; + u8 high_sclk_index; + u8 medium_sclk_index; + u8 low_sclk_index; + u8 high_mclk_index; + u8 medium_mclk_index; + u8 low_mclk_index; + u8 high_vddc_index; + u8 medium_vddc_index; + u8 low_vddc_index; + u8 rp[R600_PM_NUMBER_OF_ACTIVITY_LEVELS]; + u8 lp[R600_PM_NUMBER_OF_ACTIVITY_LEVELS]; +}; + +struct rv6xx_power_info { + /* flags */ + bool voltage_control; + bool sclk_ss; + bool mclk_ss; + bool dynamic_ss; + bool dynamic_pcie_gen2; + bool thermal_protection; + bool display_gap; + bool gfx_clock_gating; + /* clk values */ + u32 fb_div_scale; + u32 spll_ref_div; + u32 mpll_ref_div; + u32 bsu; + u32 bsp; + /* */ + u32 active_auto_throttle_sources; + /* current power state */ + u32 restricted_levels; + struct rv6xx_pm_hw_state hw; +}; + +struct rv6xx_pl { + u32 sclk; + u32 mclk; + u16 vddc; + u32 flags; +}; + +struct rv6xx_ps { + struct rv6xx_pl high; + struct rv6xx_pl medium; + struct rv6xx_pl low; +}; + +#define RV6XX_DEFAULT_VCLK_FREQ 40000 /* 10 khz */ +#define RV6XX_DEFAULT_DCLK_FREQ 30000 /* 10 khz */ + +#endif diff --git a/sys/dev/drm/radeon/rv6xxd.h b/sys/dev/drm/radeon/rv6xxd.h new file mode 100644 index 0000000000..34e86f90b4 --- /dev/null +++ b/sys/dev/drm/radeon/rv6xxd.h @@ -0,0 +1,246 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef RV6XXD_H +#define RV6XXD_H + +/* RV6xx power management */ +#define SPLL_CNTL_MODE 0x60c +# define SPLL_DIV_SYNC (1 << 5) + +#define GENERAL_PWRMGT 0x618 +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define MOBILE_SU (1 << 2) +# define THERMAL_PROTECTION_DIS (1 << 3) +# define THERMAL_PROTECTION_TYPE (1 << 4) +# define ENABLE_GEN2PCIE (1 << 5) +# define SW_GPIO_INDEX(x) ((x) << 6) +# define SW_GPIO_INDEX_MASK (3 << 6) +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 16) +# define BACKBIAS_VALUE (1 << 17) +# define BACKBIAS_DPM_CNTL (1 << 18) +# define DYN_SPREAD_SPECTRUM_EN (1 << 21) + +#define MCLK_PWRMGT_CNTL 0x624 +# define MPLL_PWRMGT_OFF (1 << 0) +# define YCLK_TURNOFF (1 << 1) +# define MPLL_TURNOFF (1 << 2) +# define SU_MCLK_USE_BCLK (1 << 3) +# define DLL_READY (1 << 4) +# define MC_BUSY (1 << 5) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA_SLEEP (1 << 8) +# define MRDCKB_SLEEP (1 << 9) +# define MRDCKC_SLEEP (1 << 10) +# define MRDCKD_SLEEP (1 << 11) +# define MRDCKE_SLEEP (1 << 12) +# define MRDCKF_SLEEP (1 << 13) +# define MRDCKG_SLEEP (1 << 14) +# define MRDCKH_SLEEP (1 << 15) +# define MRDCKA_RESET (1 << 16) +# define MRDCKB_RESET (1 << 17) +# define MRDCKC_RESET (1 << 18) +# define MRDCKD_RESET (1 << 19) +# define MRDCKE_RESET (1 << 20) +# define MRDCKF_RESET (1 << 21) +# define MRDCKG_RESET (1 << 22) +# define MRDCKH_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define USE_DISPLAY_GAP_CTXSW (1 << 27) +# define MPLL_TURNOFF_D2 (1 << 28) +# define USE_DISPLAY_URGENT_CTXSW (1 << 29) + +#define MPLL_FREQ_LEVEL_0 0x6e8 +# define LEVEL0_MPLL_POST_DIV(x) ((x) << 0) +# define LEVEL0_MPLL_POST_DIV_MASK (0xff << 0) +# define LEVEL0_MPLL_FB_DIV(x) ((x) << 8) +# define LEVEL0_MPLL_FB_DIV_MASK (0xfff << 8) +# define LEVEL0_MPLL_REF_DIV(x) ((x) << 20) +# define LEVEL0_MPLL_REF_DIV_MASK (0x3f << 20) +# define LEVEL0_MPLL_DIV_EN (1 << 28) +# define LEVEL0_DLL_BYPASS (1 << 29) +# define LEVEL0_DLL_RESET (1 << 30) + +#define VID_RT 0x6f8 +# define VID_CRT(x) ((x) << 0) +# define VID_CRT_MASK (0x1fff << 0) +# define VID_CRTU(x) ((x) << 13) +# define VID_CRTU_MASK (7 << 13) +# define SSTU(x) ((x) << 16) +# define SSTU_MASK (7 << 16) +# define VID_SWT(x) ((x) << 19) +# define VID_SWT_MASK (0x1f << 19) +# define BRT(x) ((x) << 24) +# define BRT_MASK (0xff << 24) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x70c +# define TARGET_PROFILE_INDEX_MASK (3 << 0) +# define TARGET_PROFILE_INDEX_SHIFT 0 +# define CURRENT_PROFILE_INDEX_MASK (3 << 2) +# define CURRENT_PROFILE_INDEX_SHIFT 2 +# define DYN_PWR_ENTER_INDEX(x) ((x) << 4) +# define DYN_PWR_ENTER_INDEX_MASK (3 << 4) +# define DYN_PWR_ENTER_INDEX_SHIFT 4 +# define CURR_MCLK_INDEX_MASK (3 << 6) +# define CURR_MCLK_INDEX_SHIFT 6 +# define CURR_SCLK_INDEX_MASK (0x1f << 8) +# define CURR_SCLK_INDEX_SHIFT 8 +# define CURR_VID_INDEX_MASK (3 << 13) +# define CURR_VID_INDEX_SHIFT 13 + +#define VID_UPPER_GPIO_CNTL 0x740 +# define CTXSW_UPPER_GPIO_VALUES(x) ((x) << 0) +# define CTXSW_UPPER_GPIO_VALUES_MASK (7 << 0) +# define HIGH_UPPER_GPIO_VALUES(x) ((x) << 3) +# define HIGH_UPPER_GPIO_VALUES_MASK (7 << 3) +# define MEDIUM_UPPER_GPIO_VALUES(x) ((x) << 6) +# define MEDIUM_UPPER_GPIO_VALUES_MASK (7 << 6) +# define LOW_UPPER_GPIO_VALUES(x) ((x) << 9) +# define LOW_UPPER_GPIO_VALUES_MASK (7 << 9) +# define CTXSW_BACKBIAS_VALUE (1 << 12) +# define HIGH_BACKBIAS_VALUE (1 << 13) +# define MEDIUM_BACKBIAS_VALUE (1 << 14) +# define LOW_BACKBIAS_VALUE (1 << 15) + +#define CG_DISPLAY_GAP_CNTL 0x7dc +# define DISP1_GAP(x) ((x) << 0) +# define DISP1_GAP_MASK (3 << 0) +# define DISP2_GAP(x) ((x) << 2) +# define DISP2_GAP_MASK (3 << 2) +# define VBI_TIMER_COUNT(x) ((x) << 4) +# define VBI_TIMER_COUNT_MASK (0x3fff << 4) +# define VBI_TIMER_UNIT(x) ((x) << 20) +# define VBI_TIMER_UNIT_MASK (7 << 20) +# define DISP1_GAP_MCHG(x) ((x) << 24) +# define DISP1_GAP_MCHG_MASK (3 << 24) +# define DISP2_GAP_MCHG(x) ((x) << 26) +# define DISP2_GAP_MCHG_MASK (3 << 26) + +#define CG_THERMAL_CTRL 0x7f0 +# define DPM_EVENT_SRC(x) ((x) << 0) +# define DPM_EVENT_SRC_MASK (7 << 0) +# define THERM_INC_CLK (1 << 3) +# define TOFFSET(x) ((x) << 4) +# define TOFFSET_MASK (0xff << 4) +# define DIG_THERM_DPM(x) ((x) << 12) +# define DIG_THERM_DPM_MASK (0xff << 12) +# define CTF_SEL(x) ((x) << 20) +# define CTF_SEL_MASK (7 << 20) +# define CTF_PAD_POLARITY (1 << 23) +# define CTF_PAD_EN (1 << 24) + +#define CG_SPLL_SPREAD_SPECTRUM_LOW 0x820 +# define SSEN (1 << 0) +# define CLKS(x) ((x) << 3) +# define CLKS_MASK (0xff << 3) +# define CLKS_SHIFT 3 +# define CLKV(x) ((x) << 11) +# define CLKV_MASK (0x7ff << 11) +# define CLKV_SHIFT 11 +#define CG_MPLL_SPREAD_SPECTRUM 0x830 + +#define CITF_CNTL 0x200c +# define BLACKOUT_RD (1 << 0) +# define BLACKOUT_WR (1 << 1) + +#define RAMCFG 0x2408 +#define NOOFBANK_SHIFT 0 +#define NOOFBANK_MASK 0x00000001 +#define NOOFRANK_SHIFT 1 +#define NOOFRANK_MASK 0x00000002 +#define NOOFROWS_SHIFT 2 +#define NOOFROWS_MASK 0x0000001C +#define NOOFCOLS_SHIFT 5 +#define NOOFCOLS_MASK 0x00000060 +#define CHANSIZE_SHIFT 7 +#define CHANSIZE_MASK 0x00000080 +#define BURSTLENGTH_SHIFT 8 +#define BURSTLENGTH_MASK 0x00000100 +#define CHANSIZE_OVERRIDE (1 << 10) + +#define SQM_RATIO 0x2424 +# define STATE0(x) ((x) << 0) +# define STATE0_MASK (0xff << 0) +# define STATE1(x) ((x) << 8) +# define STATE1_MASK (0xff << 8) +# define STATE2(x) ((x) << 16) +# define STATE2_MASK (0xff << 16) +# define STATE3(x) ((x) << 24) +# define STATE3_MASK (0xff << 24) + +#define ARB_RFSH_CNTL 0x2460 +# define ENABLE (1 << 0) +#define ARB_RFSH_RATE 0x2464 +# define POWERMODE0(x) ((x) << 0) +# define POWERMODE0_MASK (0xff << 0) +# define POWERMODE1(x) ((x) << 8) +# define POWERMODE1_MASK (0xff << 8) +# define POWERMODE2(x) ((x) << 16) +# define POWERMODE2_MASK (0xff << 16) +# define POWERMODE3(x) ((x) << 24) +# define POWERMODE3_MASK (0xff << 24) + +#define MC_SEQ_DRAM 0x2608 +# define CKE_DYN (1 << 12) + +#define MC_SEQ_CMD 0x26c4 + +#define MC_SEQ_RESERVE_S 0x2890 +#define MC_SEQ_RESERVE_M 0x2894 + +#define LVTMA_DATA_SYNCHRONIZATION 0x7adc +# define LVTMA_PFREQCHG (1 << 8) +#define DCE3_LVTMA_DATA_SYNCHRONIZATION 0x7f98 + +/* PCIE indirect regs */ +#define PCIE_P_CNTL 0x40 +# define P_PLL_PWRDN_IN_L1L23 (1 << 3) +# define P_PLL_BUF_PDNB (1 << 4) +# define P_PLL_PDNB (1 << 9) +# define P_ALLOW_PRX_FRONTEND_SHUTOFF (1 << 12) +/* PCIE PORT indirect regs */ +#define PCIE_LC_CNTL 0xa0 +# define LC_L0S_INACTIVITY(x) ((x) << 8) +# define LC_L0S_INACTIVITY_MASK (0xf << 8) +# define LC_L0S_INACTIVITY_SHIFT 8 +# define LC_L1_INACTIVITY(x) ((x) << 12) +# define LC_L1_INACTIVITY_MASK (0xf << 12) +# define LC_L1_INACTIVITY_SHIFT 12 +# define LC_PMI_TO_L1_DIS (1 << 16) +# define LC_ASPM_TO_L1_DIS (1 << 24) +#define PCIE_LC_SPEED_CNTL 0xa4 +# define LC_GEN2_EN (1 << 0) +# define LC_INITIATE_LINK_SPEED_CHANGE (1 << 7) +# define LC_CURRENT_DATA_RATE (1 << 11) +# define LC_HW_VOLTAGE_IF_CONTROL(x) ((x) << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_MASK (3 << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_SHIFT 12 +# define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 23) +# define LC_OTHER_SIDE_SUPPORTS_GEN2 (1 << 24) + +#endif diff --git a/sys/dev/drm/radeon/rv730_dpm.c b/sys/dev/drm/radeon/rv730_dpm.c new file mode 100644 index 0000000000..678571e044 --- /dev/null +++ b/sys/dev/drm/radeon/rv730_dpm.c @@ -0,0 +1,508 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "rv730d.h" +#include "r600_dpm.h" +#include "rv770_dpm.h" +#include "atom.h" + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +struct rv7xx_ps *rv770_get_ps(struct radeon_ps *rps); +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); + +int rv730_populate_sclk_value(struct radeon_device *rdev, + u32 engine_clock, + RV770_SMC_SCLK_VALUE *sclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct atom_clock_dividers dividers; + u32 spll_func_cntl = pi->clk_regs.rv730.cg_spll_func_cntl; + u32 spll_func_cntl_2 = pi->clk_regs.rv730.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = pi->clk_regs.rv730.cg_spll_func_cntl_3; + u32 cg_spll_spread_spectrum = pi->clk_regs.rv730.cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2 = pi->clk_regs.rv730.cg_spll_spread_spectrum_2; + u64 tmp; + u32 reference_clock = rdev->clock.spll.reference_freq; + u32 reference_divider, post_divider; + u32 fbdiv; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + engine_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = 1 + dividers.ref_div; + + if (dividers.enable_post_div) + post_divider = ((dividers.post_div >> 4) & 0xf) + + (dividers.post_div & 0xf) + 2; + else + post_divider = 1; + + tmp = (u64) engine_clock * reference_divider * post_divider * 16384; + do_div(tmp, reference_clock); + fbdiv = (u32) tmp; + + /* set up registers */ + if (dividers.enable_post_div) + spll_func_cntl |= SPLL_DIVEN; + else + spll_func_cntl &= ~SPLL_DIVEN; + spll_func_cntl &= ~(SPLL_HILEN_MASK | SPLL_LOLEN_MASK | SPLL_REF_DIV_MASK); + spll_func_cntl |= SPLL_REF_DIV(dividers.ref_div); + spll_func_cntl |= SPLL_HILEN((dividers.post_div >> 4) & 0xf); + spll_func_cntl |= SPLL_LOLEN(dividers.post_div & 0xf); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(2); + + spll_func_cntl_3 &= ~SPLL_FB_DIV_MASK; + spll_func_cntl_3 |= SPLL_FB_DIV(fbdiv); + spll_func_cntl_3 |= SPLL_DITHEN; + + if (pi->sclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = engine_clock * post_divider; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = ss.percentage * fbdiv / (clk_s * 10000); + + cg_spll_spread_spectrum &= ~CLK_S_MASK; + cg_spll_spread_spectrum |= CLK_S(clk_s); + cg_spll_spread_spectrum |= SSEN; + + cg_spll_spread_spectrum_2 &= ~CLK_V_MASK; + cg_spll_spread_spectrum_2 |= CLK_V(clk_v); + } + } + + sclk->sclk_value = cpu_to_be32(engine_clock); + sclk->vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + sclk->vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + sclk->vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + sclk->vCG_SPLL_SPREAD_SPECTRUM = cpu_to_be32(cg_spll_spread_spectrum); + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cpu_to_be32(cg_spll_spread_spectrum_2); + + return 0; +} + +int rv730_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + LPRV7XX_SMC_MCLK_VALUE mclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 mclk_pwrmgt_cntl = pi->clk_regs.rv730.mclk_pwrmgt_cntl; + u32 dll_cntl = pi->clk_regs.rv730.dll_cntl; + u32 mpll_func_cntl = pi->clk_regs.rv730.mpll_func_cntl; + u32 mpll_func_cntl_2 = pi->clk_regs.rv730.mpll_func_cntl2; + u32 mpll_func_cntl_3 = pi->clk_regs.rv730.mpll_func_cntl3; + u32 mpll_ss = pi->clk_regs.rv730.mpll_ss; + u32 mpll_ss2 = pi->clk_regs.rv730.mpll_ss2; + struct atom_clock_dividers dividers; + u32 post_divider, reference_divider; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + memory_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = dividers.ref_div + 1; + + if (dividers.enable_post_div) + post_divider = ((dividers.post_div >> 4) & 0xf) + + (dividers.post_div & 0xf) + 2; + else + post_divider = 1; + + /* setup the registers */ + if (dividers.enable_post_div) + mpll_func_cntl |= MPLL_DIVEN; + else + mpll_func_cntl &= ~MPLL_DIVEN; + + mpll_func_cntl &= ~(MPLL_REF_DIV_MASK | MPLL_HILEN_MASK | MPLL_LOLEN_MASK); + mpll_func_cntl |= MPLL_REF_DIV(dividers.ref_div); + mpll_func_cntl |= MPLL_HILEN((dividers.post_div >> 4) & 0xf); + mpll_func_cntl |= MPLL_LOLEN(dividers.post_div & 0xf); + + mpll_func_cntl_3 &= ~MPLL_FB_DIV_MASK; + mpll_func_cntl_3 |= MPLL_FB_DIV(dividers.fb_div); + if (dividers.enable_dithen) + mpll_func_cntl_3 |= MPLL_DITHEN; + else + mpll_func_cntl_3 &= ~MPLL_DITHEN; + + if (pi->mclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = memory_clock * post_divider; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, vco_freq)) { + u32 reference_clock = rdev->clock.mpll.reference_freq; + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = ss.percentage * dividers.fb_div / (clk_s * 10000); + + mpll_ss &= ~CLK_S_MASK; + mpll_ss |= CLK_S(clk_s); + mpll_ss |= SSEN; + + mpll_ss2 &= ~CLK_V_MASK; + mpll_ss |= CLK_V(clk_v); + } + } + + + mclk->mclk730.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->mclk730.vDLL_CNTL = cpu_to_be32(dll_cntl); + mclk->mclk730.mclk_value = cpu_to_be32(memory_clock); + mclk->mclk730.vMPLL_FUNC_CNTL = cpu_to_be32(mpll_func_cntl); + mclk->mclk730.vMPLL_FUNC_CNTL2 = cpu_to_be32(mpll_func_cntl_2); + mclk->mclk730.vMPLL_FUNC_CNTL3 = cpu_to_be32(mpll_func_cntl_3); + mclk->mclk730.vMPLL_SS = cpu_to_be32(mpll_ss); + mclk->mclk730.vMPLL_SS2 = cpu_to_be32(mpll_ss2); + + return 0; +} + +void rv730_read_clock_registers(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + pi->clk_regs.rv730.cg_spll_func_cntl = + RREG32(CG_SPLL_FUNC_CNTL); + pi->clk_regs.rv730.cg_spll_func_cntl_2 = + RREG32(CG_SPLL_FUNC_CNTL_2); + pi->clk_regs.rv730.cg_spll_func_cntl_3 = + RREG32(CG_SPLL_FUNC_CNTL_3); + pi->clk_regs.rv730.cg_spll_spread_spectrum = + RREG32(CG_SPLL_SPREAD_SPECTRUM); + pi->clk_regs.rv730.cg_spll_spread_spectrum_2 = + RREG32(CG_SPLL_SPREAD_SPECTRUM_2); + + pi->clk_regs.rv730.mclk_pwrmgt_cntl = + RREG32(TCI_MCLK_PWRMGT_CNTL); + pi->clk_regs.rv730.dll_cntl = + RREG32(TCI_DLL_CNTL); + pi->clk_regs.rv730.mpll_func_cntl = + RREG32(CG_MPLL_FUNC_CNTL); + pi->clk_regs.rv730.mpll_func_cntl2 = + RREG32(CG_MPLL_FUNC_CNTL_2); + pi->clk_regs.rv730.mpll_func_cntl3 = + RREG32(CG_MPLL_FUNC_CNTL_3); + pi->clk_regs.rv730.mpll_ss = + RREG32(CG_TCI_MPLL_SPREAD_SPECTRUM); + pi->clk_regs.rv730.mpll_ss2 = + RREG32(CG_TCI_MPLL_SPREAD_SPECTRUM_2); +} + +int rv730_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 mpll_func_cntl = 0; + u32 mpll_func_cntl_2 = 0 ; + u32 mpll_func_cntl_3 = 0; + u32 mclk_pwrmgt_cntl; + u32 dll_cntl; + u32 spll_func_cntl; + u32 spll_func_cntl_2; + u32 spll_func_cntl_3; + + table->ACPIState = table->initialState; + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + rv770_populate_vddc_value(rdev, pi->acpi_vddc, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = pi->pcie_gen2 ? + pi->acpi_pcie_gen2 : 0; + table->ACPIState.levels[0].gen2XSP = + pi->acpi_pcie_gen2; + } else { + rv770_populate_vddc_value(rdev, pi->min_vddc_in_table, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = 0; + } + + mpll_func_cntl = pi->clk_regs.rv730.mpll_func_cntl; + mpll_func_cntl_2 = pi->clk_regs.rv730.mpll_func_cntl2; + mpll_func_cntl_3 = pi->clk_regs.rv730.mpll_func_cntl3; + + mpll_func_cntl |= MPLL_RESET | MPLL_BYPASS_EN; + mpll_func_cntl &= ~MPLL_SLEEP; + + mpll_func_cntl_2 &= ~MCLK_MUX_SEL_MASK; + mpll_func_cntl_2 |= MCLK_MUX_SEL(1); + + mclk_pwrmgt_cntl = (MRDCKA_RESET | + MRDCKB_RESET | + MRDCKC_RESET | + MRDCKD_RESET | + MRDCKE_RESET | + MRDCKF_RESET | + MRDCKG_RESET | + MRDCKH_RESET | + MRDCKA_SLEEP | + MRDCKB_SLEEP | + MRDCKC_SLEEP | + MRDCKD_SLEEP | + MRDCKE_SLEEP | + MRDCKF_SLEEP | + MRDCKG_SLEEP | + MRDCKH_SLEEP); + + dll_cntl = 0xff000000; + + spll_func_cntl = pi->clk_regs.rv730.cg_spll_func_cntl; + spll_func_cntl_2 = pi->clk_regs.rv730.cg_spll_func_cntl_2; + spll_func_cntl_3 = pi->clk_regs.rv730.cg_spll_func_cntl_3; + + spll_func_cntl |= SPLL_RESET | SPLL_BYPASS_EN; + spll_func_cntl &= ~SPLL_SLEEP; + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL = cpu_to_be32(mpll_func_cntl); + table->ACPIState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL2 = cpu_to_be32(mpll_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL3 = cpu_to_be32(mpll_func_cntl_3); + table->ACPIState.levels[0].mclk.mclk730.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.mclk730.vDLL_CNTL = cpu_to_be32(dll_cntl); + + table->ACPIState.levels[0].mclk.mclk730.mclk_value = 0; + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + + table->ACPIState.levels[0].sclk.sclk_value = 0; + + rv770_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + table->ACPIState.levels[1] = table->ACPIState.levels[0]; + table->ACPIState.levels[2] = table->ACPIState.levels[0]; + + return 0; +} + +int rv730_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_ps *initial_state = rv770_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 a_t; + + table->initialState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv730.mpll_func_cntl); + table->initialState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL2 = + cpu_to_be32(pi->clk_regs.rv730.mpll_func_cntl2); + table->initialState.levels[0].mclk.mclk730.vMPLL_FUNC_CNTL3 = + cpu_to_be32(pi->clk_regs.rv730.mpll_func_cntl3); + table->initialState.levels[0].mclk.mclk730.vMCLK_PWRMGT_CNTL = + cpu_to_be32(pi->clk_regs.rv730.mclk_pwrmgt_cntl); + table->initialState.levels[0].mclk.mclk730.vDLL_CNTL = + cpu_to_be32(pi->clk_regs.rv730.dll_cntl); + table->initialState.levels[0].mclk.mclk730.vMPLL_SS = + cpu_to_be32(pi->clk_regs.rv730.mpll_ss); + table->initialState.levels[0].mclk.mclk730.vMPLL_SS2 = + cpu_to_be32(pi->clk_regs.rv730.mpll_ss2); + + table->initialState.levels[0].mclk.mclk730.mclk_value = + cpu_to_be32(initial_state->low.mclk); + + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv730.cg_spll_func_cntl); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv730.cg_spll_func_cntl_2); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(pi->clk_regs.rv730.cg_spll_func_cntl_3); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM = + cpu_to_be32(pi->clk_regs.rv730.cg_spll_spread_spectrum); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM_2 = + cpu_to_be32(pi->clk_regs.rv730.cg_spll_spread_spectrum_2); + + table->initialState.levels[0].sclk.sclk_value = + cpu_to_be32(initial_state->low.sclk); + + table->initialState.levels[0].arbValue = MC_CG_ARB_FREQ_F0; + + table->initialState.levels[0].seqValue = + rv770_get_seq_value(rdev, &initial_state->low); + + rv770_populate_vddc_value(rdev, + initial_state->low.vddc, + &table->initialState.levels[0].vddc); + rv770_populate_initial_mvdd_value(rdev, + &table->initialState.levels[0].mvdd); + + a_t = CG_R(0xffff) | CG_L(0); + + table->initialState.levels[0].aT = cpu_to_be32(a_t); + + table->initialState.levels[0].bSP = cpu_to_be32(pi->dsp); + + if (pi->boot_in_gen2) + table->initialState.levels[0].gen2PCIE = 1; + else + table->initialState.levels[0].gen2PCIE = 0; + if (initial_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + table->initialState.levels[0].gen2XSP = 1; + else + table->initialState.levels[0].gen2XSP = 0; + + table->initialState.levels[1] = table->initialState.levels[0]; + table->initialState.levels[2] = table->initialState.levels[0]; + + table->initialState.flags |= PPSMC_SWSTATE_FLAG_DC; + + return 0; +} + +void rv730_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + u32 arb_refresh_rate = 0; + u32 dram_timing = 0; + u32 dram_timing2 = 0; + u32 old_dram_timing = 0; + u32 old_dram_timing2 = 0; + + arb_refresh_rate = RREG32(MC_ARB_RFSH_RATE) & + ~(POWERMODE1_MASK | POWERMODE2_MASK | POWERMODE3_MASK); + arb_refresh_rate |= + (POWERMODE1(rv770_calculate_memory_refresh_rate(rdev, state->low.sclk)) | + POWERMODE2(rv770_calculate_memory_refresh_rate(rdev, state->medium.sclk)) | + POWERMODE3(rv770_calculate_memory_refresh_rate(rdev, state->high.sclk))); + WREG32(MC_ARB_RFSH_RATE, arb_refresh_rate); + + /* save the boot dram timings */ + old_dram_timing = RREG32(MC_ARB_DRAM_TIMING); + old_dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + + radeon_atom_set_engine_dram_timings(rdev, + state->high.sclk, + state->high.mclk); + + dram_timing = RREG32(MC_ARB_DRAM_TIMING); + dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + + WREG32(MC_ARB_DRAM_TIMING_3, dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_3, dram_timing2); + + radeon_atom_set_engine_dram_timings(rdev, + state->medium.sclk, + state->medium.mclk); + + dram_timing = RREG32(MC_ARB_DRAM_TIMING); + dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + + WREG32(MC_ARB_DRAM_TIMING_2, dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_2, dram_timing2); + + radeon_atom_set_engine_dram_timings(rdev, + state->low.sclk, + state->low.mclk); + + dram_timing = RREG32(MC_ARB_DRAM_TIMING); + dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + + WREG32(MC_ARB_DRAM_TIMING_1, dram_timing); + WREG32(MC_ARB_DRAM_TIMING2_1, dram_timing2); + + /* restore the boot dram timings */ + WREG32(MC_ARB_DRAM_TIMING, old_dram_timing); + WREG32(MC_ARB_DRAM_TIMING2, old_dram_timing2); + +} + +void rv730_start_dpm(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~SCLK_PWRMGT_OFF); + + WREG32_P(TCI_MCLK_PWRMGT_CNTL, 0, ~MPLL_PWRMGT_OFF); + + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); +} + +void rv730_stop_dpm(struct radeon_device *rdev) +{ + PPSMC_Result result; + + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_TwoLevelsDisabled); + + if (result != PPSMC_Result_OK) + DRM_ERROR("Could not force DPM to low\n"); + + WREG32_P(GENERAL_PWRMGT, 0, ~GLOBAL_PWRMGT_EN); + + WREG32_P(SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, ~SCLK_PWRMGT_OFF); + + WREG32_P(TCI_MCLK_PWRMGT_CNTL, MPLL_PWRMGT_OFF, ~MPLL_PWRMGT_OFF); +} + +void rv730_program_dcodt(struct radeon_device *rdev, bool use_dcodt) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 i = use_dcodt ? 0 : 1; + u32 mc4_io_pad_cntl; + + mc4_io_pad_cntl = RREG32(MC4_IO_DQ_PAD_CNTL_D0_I0); + mc4_io_pad_cntl &= 0xFFFFFF00; + mc4_io_pad_cntl |= pi->odt_value_0[i]; + WREG32(MC4_IO_DQ_PAD_CNTL_D0_I0, mc4_io_pad_cntl); + WREG32(MC4_IO_DQ_PAD_CNTL_D0_I1, mc4_io_pad_cntl); + + mc4_io_pad_cntl = RREG32(MC4_IO_QS_PAD_CNTL_D0_I0); + mc4_io_pad_cntl &= 0xFFFFFF00; + mc4_io_pad_cntl |= pi->odt_value_1[i]; + WREG32(MC4_IO_QS_PAD_CNTL_D0_I0, mc4_io_pad_cntl); + WREG32(MC4_IO_QS_PAD_CNTL_D0_I1, mc4_io_pad_cntl); +} + +void rv730_get_odt_values(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 mc4_io_pad_cntl; + + pi->odt_value_0[0] = (u8)0; + pi->odt_value_1[0] = (u8)0x80; + + mc4_io_pad_cntl = RREG32(MC4_IO_DQ_PAD_CNTL_D0_I0); + pi->odt_value_0[1] = (u8)(mc4_io_pad_cntl & 0xff); + + mc4_io_pad_cntl = RREG32(MC4_IO_QS_PAD_CNTL_D0_I0); + pi->odt_value_1[1] = (u8)(mc4_io_pad_cntl & 0xff); +} diff --git a/sys/dev/drm/radeon/rv730d.h b/sys/dev/drm/radeon/rv730d.h new file mode 100644 index 0000000000..f0a7954fb1 --- /dev/null +++ b/sys/dev/drm/radeon/rv730d.h @@ -0,0 +1,165 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef RV730_H +#define RV730_H + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_DIVEN (1 << 2) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_HILEN(x) ((x) << 12) +#define SPLL_HILEN_MASK (0xf << 12) +#define SPLL_LOLEN(x) ((x) << 16) +#define SPLL_LOLEN_MASK (0xf << 16) +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_DITHEN (1 << 28) + +#define CG_MPLL_FUNC_CNTL 0x624 +#define MPLL_RESET (1 << 0) +#define MPLL_SLEEP (1 << 1) +#define MPLL_DIVEN (1 << 2) +#define MPLL_BYPASS_EN (1 << 3) +#define MPLL_REF_DIV(x) ((x) << 4) +#define MPLL_REF_DIV_MASK (0x3f << 4) +#define MPLL_HILEN(x) ((x) << 12) +#define MPLL_HILEN_MASK (0xf << 12) +#define MPLL_LOLEN(x) ((x) << 16) +#define MPLL_LOLEN_MASK (0xf << 16) +#define CG_MPLL_FUNC_CNTL_2 0x628 +#define MCLK_MUX_SEL(x) ((x) << 0) +#define MCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_MPLL_FUNC_CNTL_3 0x62c +#define MPLL_FB_DIV(x) ((x) << 0) +#define MPLL_FB_DIV_MASK (0x3ffffff << 0) +#define MPLL_DITHEN (1 << 28) + +#define CG_TCI_MPLL_SPREAD_SPECTRUM 0x634 +#define CG_TCI_MPLL_SPREAD_SPECTRUM_2 0x638 +#define GENERAL_PWRMGT 0x63c +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define ENABLE_GEN2PCIE (1 << 4) +# define ENABLE_GEN2XSP (1 << 5) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (3 << 6) +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 18) +# define BACKBIAS_VALUE (1 << 19) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +# define AC_DC_SW (1 << 24) + +#define SCLK_PWRMGT_CNTL 0x644 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) + +#define TCI_MCLK_PWRMGT_CNTL 0x648 +# define MPLL_PWRMGT_OFF (1 << 5) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA_SLEEP (1 << 8) +# define MRDCKB_SLEEP (1 << 9) +# define MRDCKC_SLEEP (1 << 10) +# define MRDCKD_SLEEP (1 << 11) +# define MRDCKE_SLEEP (1 << 12) +# define MRDCKF_SLEEP (1 << 13) +# define MRDCKG_SLEEP (1 << 14) +# define MRDCKH_SLEEP (1 << 15) +# define MRDCKA_RESET (1 << 16) +# define MRDCKB_RESET (1 << 17) +# define MRDCKC_RESET (1 << 18) +# define MRDCKD_RESET (1 << 19) +# define MRDCKE_RESET (1 << 20) +# define MRDCKF_RESET (1 << 21) +# define MRDCKG_RESET (1 << 22) +# define MRDCKH_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define MPLL_TURNOFF_D2 (1 << 28) +#define TCI_DLL_CNTL 0x64c + +#define CG_PG_CNTL 0x858 +# define PWRGATE_ENABLE (1 << 0) + +#define CG_AT 0x6d4 +#define CG_R(x) ((x) << 0) +#define CG_R_MASK (0xffff << 0) +#define CG_L(x) ((x) << 16) +#define CG_L_MASK (0xffff << 16) + +#define CG_SPLL_SPREAD_SPECTRUM 0x790 +#define SSEN (1 << 0) +#define CLK_S(x) ((x) << 4) +#define CLK_S_MASK (0xfff << 4) +#define CG_SPLL_SPREAD_SPECTRUM_2 0x794 +#define CLK_V(x) ((x) << 0) +#define CLK_V_MASK (0x3ffffff << 0) + +#define MC_ARB_DRAM_TIMING 0x2774 +#define MC_ARB_DRAM_TIMING2 0x2778 + +#define MC_ARB_RFSH_RATE 0x27b0 +#define POWERMODE0(x) ((x) << 0) +#define POWERMODE0_MASK (0xff << 0) +#define POWERMODE1(x) ((x) << 8) +#define POWERMODE1_MASK (0xff << 8) +#define POWERMODE2(x) ((x) << 16) +#define POWERMODE2_MASK (0xff << 16) +#define POWERMODE3(x) ((x) << 24) +#define POWERMODE3_MASK (0xff << 24) + +#define MC_ARB_DRAM_TIMING_1 0x27f0 +#define MC_ARB_DRAM_TIMING_2 0x27f4 +#define MC_ARB_DRAM_TIMING_3 0x27f8 +#define MC_ARB_DRAM_TIMING2_1 0x27fc +#define MC_ARB_DRAM_TIMING2_2 0x2800 +#define MC_ARB_DRAM_TIMING2_3 0x2804 + +#define MC4_IO_DQ_PAD_CNTL_D0_I0 0x2978 +#define MC4_IO_DQ_PAD_CNTL_D0_I1 0x297c +#define MC4_IO_QS_PAD_CNTL_D0_I0 0x2980 +#define MC4_IO_QS_PAD_CNTL_D0_I1 0x2984 + +#endif diff --git a/sys/dev/drm/radeon/rv740_dpm.c b/sys/dev/drm/radeon/rv740_dpm.c new file mode 100644 index 0000000000..d20adf5663 --- /dev/null +++ b/sys/dev/drm/radeon/rv740_dpm.c @@ -0,0 +1,416 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "rv740d.h" +#include "r600_dpm.h" +#include "rv770_dpm.h" +#include "atom.h" + +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); + +u32 rv740_get_decoded_reference_divider(u32 encoded_ref) +{ + u32 ref = 0; + + switch (encoded_ref) { + case 0: + ref = 1; + break; + case 16: + ref = 2; + break; + case 17: + ref = 3; + break; + case 18: + ref = 2; + break; + case 19: + ref = 3; + break; + case 20: + ref = 4; + break; + case 21: + ref = 5; + break; + default: + DRM_ERROR("Invalid encoded Reference Divider\n"); + ref = 0; + break; + } + + return ref; +} + +struct dll_speed_setting { + u16 min; + u16 max; + u32 dll_speed; +}; + +static struct dll_speed_setting dll_speed_table[16] = +{ + { 270, 320, 0x0f }, + { 240, 270, 0x0e }, + { 200, 240, 0x0d }, + { 180, 200, 0x0c }, + { 160, 180, 0x0b }, + { 140, 160, 0x0a }, + { 120, 140, 0x09 }, + { 110, 120, 0x08 }, + { 95, 110, 0x07 }, + { 85, 95, 0x06 }, + { 78, 85, 0x05 }, + { 70, 78, 0x04 }, + { 65, 70, 0x03 }, + { 60, 65, 0x02 }, + { 42, 60, 0x01 }, + { 00, 42, 0x00 } +}; + +u32 rv740_get_dll_speed(bool is_gddr5, u32 memory_clock) +{ + int i; + u32 factor; + u16 data_rate; + + if (is_gddr5) + factor = 4; + else + factor = 2; + + data_rate = (u16)(memory_clock * factor / 1000); + + if (data_rate < dll_speed_table[0].max) { + for (i = 0; i < 16; i++) { + if (data_rate > dll_speed_table[i].min && + data_rate <= dll_speed_table[i].max) + return dll_speed_table[i].dll_speed; + } + } + + DRM_DEBUG_KMS("Target MCLK greater than largest MCLK in DLL speed table\n"); + + return 0x0f; +} + +int rv740_populate_sclk_value(struct radeon_device *rdev, u32 engine_clock, + RV770_SMC_SCLK_VALUE *sclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct atom_clock_dividers dividers; + u32 spll_func_cntl = pi->clk_regs.rv770.cg_spll_func_cntl; + u32 spll_func_cntl_2 = pi->clk_regs.rv770.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = pi->clk_regs.rv770.cg_spll_func_cntl_3; + u32 cg_spll_spread_spectrum = pi->clk_regs.rv770.cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2 = pi->clk_regs.rv770.cg_spll_spread_spectrum_2; + u64 tmp; + u32 reference_clock = rdev->clock.spll.reference_freq; + u32 reference_divider; + u32 fbdiv; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + engine_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = 1 + dividers.ref_div; + + tmp = (u64) engine_clock * reference_divider * dividers.post_div * 16384; + do_div(tmp, reference_clock); + fbdiv = (u32) tmp; + + spll_func_cntl &= ~(SPLL_PDIV_A_MASK | SPLL_REF_DIV_MASK); + spll_func_cntl |= SPLL_REF_DIV(dividers.ref_div); + spll_func_cntl |= SPLL_PDIV_A(dividers.post_div); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(2); + + spll_func_cntl_3 &= ~SPLL_FB_DIV_MASK; + spll_func_cntl_3 |= SPLL_FB_DIV(fbdiv); + spll_func_cntl_3 |= SPLL_DITHEN; + + if (pi->sclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = engine_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = 4 * ss.percentage * fbdiv / (clk_s * 10000); + + cg_spll_spread_spectrum &= ~CLK_S_MASK; + cg_spll_spread_spectrum |= CLK_S(clk_s); + cg_spll_spread_spectrum |= SSEN; + + cg_spll_spread_spectrum_2 &= ~CLK_V_MASK; + cg_spll_spread_spectrum_2 |= CLK_V(clk_v); + } + } + + sclk->sclk_value = cpu_to_be32(engine_clock); + sclk->vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + sclk->vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + sclk->vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + sclk->vCG_SPLL_SPREAD_SPECTRUM = cpu_to_be32(cg_spll_spread_spectrum); + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cpu_to_be32(cg_spll_spread_spectrum_2); + + return 0; +} + +int rv740_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + RV7XX_SMC_MCLK_VALUE *mclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 mpll_ad_func_cntl = pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 mclk_pwrmgt_cntl = pi->clk_regs.rv770.mclk_pwrmgt_cntl; + u32 dll_cntl = pi->clk_regs.rv770.dll_cntl; + u32 mpll_ss1 = pi->clk_regs.rv770.mpll_ss1; + u32 mpll_ss2 = pi->clk_regs.rv770.mpll_ss2; + struct atom_clock_dividers dividers; + u32 ibias; + u32 dll_speed; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + memory_clock, false, ÷rs); + if (ret) + return ret; + + ibias = rv770_map_clkf_to_ibias(rdev, dividers.whole_fb_div); + + mpll_ad_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_ad_func_cntl |= CLKR(dividers.ref_div); + mpll_ad_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_ad_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_ad_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_ad_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_ad_func_cntl_2 |= VCO_MODE; + else + mpll_ad_func_cntl_2 &= ~VCO_MODE; + + if (pi->mem_gddr5) { + mpll_dq_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_dq_func_cntl |= CLKR(dividers.ref_div); + mpll_dq_func_cntl |= YCLK_POST_DIV(dividers.post_div); + mpll_dq_func_cntl |= CLKF(dividers.whole_fb_div); + mpll_dq_func_cntl |= CLKFRAC(dividers.frac_fb_div); + mpll_dq_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_dq_func_cntl_2 |= VCO_MODE; + else + mpll_dq_func_cntl_2 &= ~VCO_MODE; + } + + if (pi->mclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = memory_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, vco_freq)) { + u32 reference_clock = rdev->clock.mpll.reference_freq; + u32 decoded_ref = rv740_get_decoded_reference_divider(dividers.ref_div); + u32 clk_s = reference_clock * 5 / (decoded_ref * ss.rate); + u32 clk_v = 0x40000 * ss.percentage * + (dividers.whole_fb_div + (dividers.frac_fb_div / 8)) / (clk_s * 10000); + + mpll_ss1 &= ~CLKV_MASK; + mpll_ss1 |= CLKV(clk_v); + + mpll_ss2 &= ~CLKS_MASK; + mpll_ss2 |= CLKS(clk_s); + } + } + + dll_speed = rv740_get_dll_speed(pi->mem_gddr5, + memory_clock); + + mclk_pwrmgt_cntl &= ~DLL_SPEED_MASK; + mclk_pwrmgt_cntl |= DLL_SPEED(dll_speed); + + mclk->mclk770.mclk_value = cpu_to_be32(memory_clock); + mclk->mclk770.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + mclk->mclk770.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + mclk->mclk770.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + mclk->mclk770.vMPLL_SS = cpu_to_be32(mpll_ss1); + mclk->mclk770.vMPLL_SS2 = cpu_to_be32(mpll_ss2); + + return 0; +} + +void rv740_read_clock_registers(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + pi->clk_regs.rv770.cg_spll_func_cntl = + RREG32(CG_SPLL_FUNC_CNTL); + pi->clk_regs.rv770.cg_spll_func_cntl_2 = + RREG32(CG_SPLL_FUNC_CNTL_2); + pi->clk_regs.rv770.cg_spll_func_cntl_3 = + RREG32(CG_SPLL_FUNC_CNTL_3); + pi->clk_regs.rv770.cg_spll_spread_spectrum = + RREG32(CG_SPLL_SPREAD_SPECTRUM); + pi->clk_regs.rv770.cg_spll_spread_spectrum_2 = + RREG32(CG_SPLL_SPREAD_SPECTRUM_2); + + pi->clk_regs.rv770.mpll_ad_func_cntl = + RREG32(MPLL_AD_FUNC_CNTL); + pi->clk_regs.rv770.mpll_ad_func_cntl_2 = + RREG32(MPLL_AD_FUNC_CNTL_2); + pi->clk_regs.rv770.mpll_dq_func_cntl = + RREG32(MPLL_DQ_FUNC_CNTL); + pi->clk_regs.rv770.mpll_dq_func_cntl_2 = + RREG32(MPLL_DQ_FUNC_CNTL_2); + pi->clk_regs.rv770.mclk_pwrmgt_cntl = + RREG32(MCLK_PWRMGT_CNTL); + pi->clk_regs.rv770.dll_cntl = RREG32(DLL_CNTL); + pi->clk_regs.rv770.mpll_ss1 = RREG32(MPLL_SS1); + pi->clk_regs.rv770.mpll_ss2 = RREG32(MPLL_SS2); +} + +int rv740_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 mpll_ad_func_cntl = pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 spll_func_cntl = pi->clk_regs.rv770.cg_spll_func_cntl; + u32 spll_func_cntl_2 = pi->clk_regs.rv770.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = pi->clk_regs.rv770.cg_spll_func_cntl_3; + u32 mclk_pwrmgt_cntl = pi->clk_regs.rv770.mclk_pwrmgt_cntl; + u32 dll_cntl = pi->clk_regs.rv770.dll_cntl; + + table->ACPIState = table->initialState; + + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + rv770_populate_vddc_value(rdev, pi->acpi_vddc, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = + pi->pcie_gen2 ? + pi->acpi_pcie_gen2 : 0; + table->ACPIState.levels[0].gen2XSP = + pi->acpi_pcie_gen2; + } else { + rv770_populate_vddc_value(rdev, pi->min_vddc_in_table, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = 0; + } + + mpll_ad_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN; + + mpll_dq_func_cntl_2 |= BYPASS | BIAS_GEN_PDNB | RESET_EN; + + mclk_pwrmgt_cntl |= (MRDCKA0_RESET | + MRDCKA1_RESET | + MRDCKB0_RESET | + MRDCKB1_RESET | + MRDCKC0_RESET | + MRDCKC1_RESET | + MRDCKD0_RESET | + MRDCKD1_RESET); + + dll_cntl |= (MRDCKA0_BYPASS | + MRDCKA1_BYPASS | + MRDCKB0_BYPASS | + MRDCKB1_BYPASS | + MRDCKC0_BYPASS | + MRDCKC1_BYPASS | + MRDCKD0_BYPASS | + MRDCKD1_BYPASS); + + spll_func_cntl |= SPLL_RESET | SPLL_SLEEP | SPLL_BYPASS_EN; + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk770.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + + table->ACPIState.levels[0].mclk.mclk770.mclk_value = 0; + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + + table->ACPIState.levels[0].sclk.sclk_value = 0; + + table->ACPIState.levels[1] = table->ACPIState.levels[0]; + table->ACPIState.levels[2] = table->ACPIState.levels[0]; + + rv770_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + return 0; +} + +void rv740_enable_mclk_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(MPLL_CNTL_MODE, SS_SSEN, ~SS_SSEN); + else + WREG32_P(MPLL_CNTL_MODE, 0, ~SS_SSEN); +} + +u8 rv740_get_mclk_frequency_ratio(u32 memory_clock) +{ + u8 mc_para_index; + + if ((memory_clock < 10000) || (memory_clock > 47500)) + mc_para_index = 0x00; + else + mc_para_index = (u8)((memory_clock - 10000) / 2500); + + return mc_para_index; +} diff --git a/sys/dev/drm/radeon/rv740d.h b/sys/dev/drm/radeon/rv740d.h new file mode 100644 index 0000000000..fe5ab075dc --- /dev/null +++ b/sys/dev/drm/radeon/rv740d.h @@ -0,0 +1,117 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef RV740_H +#define RV740_H + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_PDIV_A(x) ((x) << 20) +#define SPLL_PDIV_A_MASK (0x7f << 20) +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_DITHEN (1 << 28) + +#define MPLL_CNTL_MODE 0x61c +#define SS_SSEN (1 << 24) + +#define MPLL_AD_FUNC_CNTL 0x624 +#define CLKF(x) ((x) << 0) +#define CLKF_MASK (0x7f << 0) +#define CLKR(x) ((x) << 7) +#define CLKR_MASK (0x1f << 7) +#define CLKFRAC(x) ((x) << 12) +#define CLKFRAC_MASK (0x1f << 12) +#define YCLK_POST_DIV(x) ((x) << 17) +#define YCLK_POST_DIV_MASK (3 << 17) +#define IBIAS(x) ((x) << 20) +#define IBIAS_MASK (0x3ff << 20) +#define RESET (1 << 30) +#define PDNB (1 << 31) +#define MPLL_AD_FUNC_CNTL_2 0x628 +#define BYPASS (1 << 19) +#define BIAS_GEN_PDNB (1 << 24) +#define RESET_EN (1 << 25) +#define VCO_MODE (1 << 29) +#define MPLL_DQ_FUNC_CNTL 0x62c +#define MPLL_DQ_FUNC_CNTL_2 0x630 + +#define MCLK_PWRMGT_CNTL 0x648 +#define DLL_SPEED(x) ((x) << 0) +#define DLL_SPEED_MASK (0x1f << 0) +# define MPLL_PWRMGT_OFF (1 << 5) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA0_SLEEP (1 << 8) +# define MRDCKA1_SLEEP (1 << 9) +# define MRDCKB0_SLEEP (1 << 10) +# define MRDCKB1_SLEEP (1 << 11) +# define MRDCKC0_SLEEP (1 << 12) +# define MRDCKC1_SLEEP (1 << 13) +# define MRDCKD0_SLEEP (1 << 14) +# define MRDCKD1_SLEEP (1 << 15) +# define MRDCKA0_RESET (1 << 16) +# define MRDCKA1_RESET (1 << 17) +# define MRDCKB0_RESET (1 << 18) +# define MRDCKB1_RESET (1 << 19) +# define MRDCKC0_RESET (1 << 20) +# define MRDCKC1_RESET (1 << 21) +# define MRDCKD0_RESET (1 << 22) +# define MRDCKD1_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define MPLL_TURNOFF_D2 (1 << 28) +#define DLL_CNTL 0x64c +# define MRDCKA0_BYPASS (1 << 24) +# define MRDCKA1_BYPASS (1 << 25) +# define MRDCKB0_BYPASS (1 << 26) +# define MRDCKB1_BYPASS (1 << 27) +# define MRDCKC0_BYPASS (1 << 28) +# define MRDCKC1_BYPASS (1 << 29) +# define MRDCKD0_BYPASS (1 << 30) +# define MRDCKD1_BYPASS (1 << 31) + +#define CG_SPLL_SPREAD_SPECTRUM 0x790 +#define SSEN (1 << 0) +#define CLK_S(x) ((x) << 4) +#define CLK_S_MASK (0xfff << 4) +#define CG_SPLL_SPREAD_SPECTRUM_2 0x794 +#define CLK_V(x) ((x) << 0) +#define CLK_V_MASK (0x3ffffff << 0) + +#define MPLL_SS1 0x85c +#define CLKV(x) ((x) << 0) +#define CLKV_MASK (0x3ffffff << 0) +#define MPLL_SS2 0x860 +#define CLKS(x) ((x) << 0) +#define CLKS_MASK (0xfff << 0) + +#endif diff --git a/sys/dev/drm/radeon/rv770.c b/sys/dev/drm/radeon/rv770.c index 2b915b47ea..e405ba0ccf 100644 --- a/sys/dev/drm/radeon/rv770.c +++ b/sys/dev/drm/radeon/rv770.c @@ -742,10 +742,10 @@ static void rv770_init_golden_registers(struct radeon_device *rdev) (const u32)ARRAY_SIZE(r7xx_golden_dyn_gpr_registers)); radeon_program_register_sequence(rdev, rv730_golden_registers, - (const u32)ARRAY_SIZE(rv770_golden_registers)); + (const u32)ARRAY_SIZE(rv730_golden_registers)); radeon_program_register_sequence(rdev, rv730_mgcg_init, - (const u32)ARRAY_SIZE(rv770_mgcg_init)); + (const u32)ARRAY_SIZE(rv730_mgcg_init)); break; case CHIP_RV710: radeon_program_register_sequence(rdev, @@ -756,18 +756,18 @@ static void rv770_init_golden_registers(struct radeon_device *rdev) (const u32)ARRAY_SIZE(r7xx_golden_dyn_gpr_registers)); radeon_program_register_sequence(rdev, rv710_golden_registers, - (const u32)ARRAY_SIZE(rv770_golden_registers)); + (const u32)ARRAY_SIZE(rv710_golden_registers)); radeon_program_register_sequence(rdev, rv710_mgcg_init, - (const u32)ARRAY_SIZE(rv770_mgcg_init)); + (const u32)ARRAY_SIZE(rv710_mgcg_init)); break; case CHIP_RV740: radeon_program_register_sequence(rdev, rv740_golden_registers, - (const u32)ARRAY_SIZE(rv770_golden_registers)); + (const u32)ARRAY_SIZE(rv740_golden_registers)); radeon_program_register_sequence(rdev, rv740_mgcg_init, - (const u32)ARRAY_SIZE(rv770_mgcg_init)); + (const u32)ARRAY_SIZE(rv740_mgcg_init)); break; default: break; @@ -1827,6 +1827,8 @@ static int rv770_startup(struct radeon_device *rdev) /* enable pcie gen2 link */ rv770_pcie_gen2_enable(rdev); + rv770_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->rlc_fw) { r = r600_init_microcode(rdev); if (r) { @@ -1839,7 +1841,6 @@ static int rv770_startup(struct radeon_device *rdev) if (r) return r; - rv770_mc_program(rdev); if (rdev->flags & RADEON_IS_AGP) { rv770_agp_enable(rdev); } else { @@ -1981,6 +1982,7 @@ int rv770_resume(struct radeon_device *rdev) int rv770_suspend(struct radeon_device *rdev) { r600_audio_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_suspend(rdev); r700_cp_stop(rdev); r600_dma_stop(rdev); @@ -2096,6 +2098,7 @@ void rv770_fini(struct radeon_device *rdev) radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); rv770_pcie_gart_fini(rdev); + r600_uvd_stop(rdev); radeon_uvd_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); diff --git a/sys/dev/drm/radeon/rv770_dpm.c b/sys/dev/drm/radeon/rv770_dpm.c new file mode 100644 index 0000000000..0f3b25fd1c --- /dev/null +++ b/sys/dev/drm/radeon/rv770_dpm.c @@ -0,0 +1,2532 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "rv770d.h" +#include "r600_dpm.h" +#include "rv770_dpm.h" +#include "cypress_dpm.h" +#include "atom.h" +#include + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define MC_CG_SEQ_DRAMCONF_S0 0x05 +#define MC_CG_SEQ_DRAMCONF_S1 0x06 + +#define PCIE_BUS_CLK 10000 +#define TCLK (PCIE_BUS_CLK / 10) + +#define SMC_RAM_END 0xC000 + +struct rv7xx_ps *rv770_get_ps(struct radeon_ps *rps); +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); +void rv770_dpm_reset_asic(struct radeon_device *rdev); +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev); + +struct rv7xx_ps *rv770_get_ps(struct radeon_ps *rps) +{ + struct rv7xx_ps *ps = rps->ps_priv; + + return ps; +} + +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev) +{ + struct evergreen_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void rv770_enable_bif_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + if (enable) { + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp |= LC_HW_VOLTAGE_IF_CONTROL(1); + tmp |= LC_GEN2_EN_STRAP; + } else { + if (!pi->boot_in_gen2) { + tmp &= ~LC_HW_VOLTAGE_IF_CONTROL_MASK; + tmp &= ~LC_GEN2_EN_STRAP; + } + } + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) || + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, tmp); + +} + +static void rv770_enable_l0s(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL) & ~LC_L0S_INACTIVITY_MASK; + tmp |= LC_L0S_INACTIVITY(3); + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); +} + +static void rv770_enable_l1(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL); + tmp &= ~LC_L1_INACTIVITY_MASK; + tmp |= LC_L1_INACTIVITY(4); + tmp &= ~LC_PMI_TO_L1_DIS; + tmp &= ~LC_ASPM_TO_L1_DIS; + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); +} + +static void rv770_enable_pll_sleep_in_l1(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL) & ~LC_L1_INACTIVITY_MASK; + tmp |= LC_L1_INACTIVITY(8); + WREG32_PCIE_PORT(PCIE_LC_CNTL, tmp); + + /* NOTE, this is a PCIE indirect reg, not PCIE PORT */ + tmp = RREG32_PCIE(PCIE_P_CNTL); + tmp |= P_PLL_PWRDN_IN_L1L23; + tmp &= ~P_PLL_BUF_PDNB; + tmp &= ~P_PLL_PDNB; + tmp |= P_ALLOW_PRX_FRONTEND_SHUTOFF; + WREG32_PCIE(PCIE_P_CNTL, tmp); +} + +static void rv770_gfx_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_TILING_CONFIG); + } +} + +static void rv770_mg_clock_gating_enable(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + u32 mgcg_cgtt_local0; + + if (rdev->family == CHIP_RV770) + mgcg_cgtt_local0 = RV770_MGCGTTLOCAL0_DFLT; + else + mgcg_cgtt_local0 = RV7XX_MGCGTTLOCAL0_DFLT; + + WREG32(CG_CGTT_LOCAL_0, mgcg_cgtt_local0); + WREG32(CG_CGTT_LOCAL_1, (RV770_MGCGTTLOCAL1_DFLT & 0xFFFFCFFF)); + + if (pi->mgcgtssm) + WREG32(CGTS_SM_CTRL_REG, RV770_MGCGCGTSSMCTRL_DFLT); + } else { + WREG32(CG_CGTT_LOCAL_0, 0xFFFFFFFF); + WREG32(CG_CGTT_LOCAL_1, 0xFFFFCFFF); + } +} + +void rv770_restore_cgcg(struct radeon_device *rdev) +{ + bool dpm_en = false, cg_en = false; + + if (RREG32(GENERAL_PWRMGT) & GLOBAL_PWRMGT_EN) + dpm_en = true; + if (RREG32(SCLK_PWRMGT_CNTL) & DYN_GFX_CLK_OFF_EN) + cg_en = true; + + if (dpm_en && !cg_en) + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); +} + +static void rv770_start_dpm(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~SCLK_PWRMGT_OFF); + + WREG32_P(MCLK_PWRMGT_CNTL, 0, ~MPLL_PWRMGT_OFF); + + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); +} + +void rv770_stop_dpm(struct radeon_device *rdev) +{ + PPSMC_Result result; + + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_TwoLevelsDisabled); + + if (result != PPSMC_Result_OK) + DRM_ERROR("Could not force DPM to low.\n"); + + WREG32_P(GENERAL_PWRMGT, 0, ~GLOBAL_PWRMGT_EN); + + WREG32_P(SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, ~SCLK_PWRMGT_OFF); + + WREG32_P(MCLK_PWRMGT_CNTL, MPLL_PWRMGT_OFF, ~MPLL_PWRMGT_OFF); +} + +bool rv770_dpm_enabled(struct radeon_device *rdev) +{ + if (RREG32(GENERAL_PWRMGT) & GLOBAL_PWRMGT_EN) + return true; + else + return false; +} + +void rv770_enable_thermal_protection(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + else + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); +} + +void rv770_enable_acpi_pm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, STATIC_PM_EN, ~STATIC_PM_EN); +} + +u8 rv770_get_seq_value(struct radeon_device *rdev, + struct rv7xx_pl *pl) +{ + return (pl->flags & ATOM_PPLIB_R600_FLAGS_LOWPOWER) ? + MC_CG_SEQ_DRAMCONF_S0 : MC_CG_SEQ_DRAMCONF_S1; +} + +int rv770_read_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 *value) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + return rv770_read_smc_sram_dword(rdev, + pi->soft_regs_start + reg_offset, + value, pi->sram_end); +} + +int rv770_write_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 value) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + return rv770_write_smc_sram_dword(rdev, + pi->soft_regs_start + reg_offset, + value, pi->sram_end); +} + +int rv770_populate_smc_t(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + int a_n; + int a_d; + u8 l[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE]; + u8 r[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE]; + u32 a_t; + + l[0] = 0; + r[2] = 100; + + a_n = (int)state->medium.sclk * pi->lmp + + (int)state->low.sclk * (R600_AH_DFLT - pi->rlp); + a_d = (int)state->low.sclk * (100 - (int)pi->rlp) + + (int)state->medium.sclk * pi->lmp; + + l[1] = (u8)(pi->lmp - (int)pi->lmp * a_n / a_d); + r[0] = (u8)(pi->rlp + (100 - (int)pi->rlp) * a_n / a_d); + + a_n = (int)state->high.sclk * pi->lhp + (int)state->medium.sclk * + (R600_AH_DFLT - pi->rmp); + a_d = (int)state->medium.sclk * (100 - (int)pi->rmp) + + (int)state->high.sclk * pi->lhp; + + l[2] = (u8)(pi->lhp - (int)pi->lhp * a_n / a_d); + r[1] = (u8)(pi->rmp + (100 - (int)pi->rmp) * a_n / a_d); + + for (i = 0; i < (RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1); i++) { + a_t = CG_R(r[i] * pi->bsp / 200) | CG_L(l[i] * pi->bsp / 200); + smc_state->levels[i].aT = cpu_to_be32(a_t); + } + + a_t = CG_R(r[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1] * pi->pbsp / 200) | + CG_L(l[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1] * pi->pbsp / 200); + + smc_state->levels[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1].aT = + cpu_to_be32(a_t); + + return 0; +} + +int rv770_populate_smc_sp(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + for (i = 0; i < (RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1); i++) + smc_state->levels[i].bSP = cpu_to_be32(pi->dsp); + + smc_state->levels[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1].bSP = + cpu_to_be32(pi->psp); + + return 0; +} + +static void rv770_calculate_fractional_mpll_feedback_divider(u32 memory_clock, + u32 reference_clock, + bool gddr5, + struct atom_clock_dividers *dividers, + u32 *clkf, + u32 *clkfrac) +{ + u32 post_divider, reference_divider, feedback_divider8; + u32 fyclk; + + if (gddr5) + fyclk = (memory_clock * 8) / 2; + else + fyclk = (memory_clock * 4) / 2; + + post_divider = dividers->post_div; + reference_divider = dividers->ref_div; + + feedback_divider8 = + (8 * fyclk * reference_divider * post_divider) / reference_clock; + + *clkf = feedback_divider8 / 8; + *clkfrac = feedback_divider8 % 8; +} + +static int rv770_encode_yclk_post_div(u32 postdiv, u32 *encoded_postdiv) +{ + int ret = 0; + + switch (postdiv) { + case 1: + *encoded_postdiv = 0; + break; + case 2: + *encoded_postdiv = 1; + break; + case 4: + *encoded_postdiv = 2; + break; + case 8: + *encoded_postdiv = 3; + break; + case 16: + *encoded_postdiv = 4; + break; + default: + ret = -EINVAL; + break; + } + + return ret; +} + +u32 rv770_map_clkf_to_ibias(struct radeon_device *rdev, u32 clkf) +{ + if (clkf <= 0x10) + return 0x4B; + if (clkf <= 0x19) + return 0x5B; + if (clkf <= 0x21) + return 0x2B; + if (clkf <= 0x27) + return 0x6C; + if (clkf <= 0x31) + return 0x9D; + return 0xC6; +} + +static int rv770_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + RV7XX_SMC_MCLK_VALUE *mclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 encoded_reference_dividers[] = { 0, 16, 17, 20, 21 }; + u32 mpll_ad_func_cntl = + pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = + pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = + pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = + pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 mclk_pwrmgt_cntl = + pi->clk_regs.rv770.mclk_pwrmgt_cntl; + u32 dll_cntl = pi->clk_regs.rv770.dll_cntl; + struct atom_clock_dividers dividers; + u32 reference_clock = rdev->clock.mpll.reference_freq; + u32 clkf, clkfrac; + u32 postdiv_yclk; + u32 ibias; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_MEMORY_PLL_PARAM, + memory_clock, false, ÷rs); + if (ret) + return ret; + + if ((dividers.ref_div < 1) || (dividers.ref_div > 5)) + return -EINVAL; + + rv770_calculate_fractional_mpll_feedback_divider(memory_clock, reference_clock, + pi->mem_gddr5, + ÷rs, &clkf, &clkfrac); + + ret = rv770_encode_yclk_post_div(dividers.post_div, &postdiv_yclk); + if (ret) + return ret; + + ibias = rv770_map_clkf_to_ibias(rdev, clkf); + + mpll_ad_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_ad_func_cntl |= CLKR(encoded_reference_dividers[dividers.ref_div - 1]); + mpll_ad_func_cntl |= YCLK_POST_DIV(postdiv_yclk); + mpll_ad_func_cntl |= CLKF(clkf); + mpll_ad_func_cntl |= CLKFRAC(clkfrac); + mpll_ad_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_ad_func_cntl_2 |= VCO_MODE; + else + mpll_ad_func_cntl_2 &= ~VCO_MODE; + + if (pi->mem_gddr5) { + rv770_calculate_fractional_mpll_feedback_divider(memory_clock, + reference_clock, + pi->mem_gddr5, + ÷rs, &clkf, &clkfrac); + + ibias = rv770_map_clkf_to_ibias(rdev, clkf); + + ret = rv770_encode_yclk_post_div(dividers.post_div, &postdiv_yclk); + if (ret) + return ret; + + mpll_dq_func_cntl &= ~(CLKR_MASK | + YCLK_POST_DIV_MASK | + CLKF_MASK | + CLKFRAC_MASK | + IBIAS_MASK); + mpll_dq_func_cntl |= CLKR(encoded_reference_dividers[dividers.ref_div - 1]); + mpll_dq_func_cntl |= YCLK_POST_DIV(postdiv_yclk); + mpll_dq_func_cntl |= CLKF(clkf); + mpll_dq_func_cntl |= CLKFRAC(clkfrac); + mpll_dq_func_cntl |= IBIAS(ibias); + + if (dividers.vco_mode) + mpll_dq_func_cntl_2 |= VCO_MODE; + else + mpll_dq_func_cntl_2 &= ~VCO_MODE; + } + + mclk->mclk770.mclk_value = cpu_to_be32(memory_clock); + mclk->mclk770.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + mclk->mclk770.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + mclk->mclk770.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + mclk->mclk770.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + + return 0; +} + +static int rv770_populate_sclk_value(struct radeon_device *rdev, + u32 engine_clock, + RV770_SMC_SCLK_VALUE *sclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct atom_clock_dividers dividers; + u32 spll_func_cntl = + pi->clk_regs.rv770.cg_spll_func_cntl; + u32 spll_func_cntl_2 = + pi->clk_regs.rv770.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = + pi->clk_regs.rv770.cg_spll_func_cntl_3; + u32 cg_spll_spread_spectrum = + pi->clk_regs.rv770.cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2 = + pi->clk_regs.rv770.cg_spll_spread_spectrum_2; + u64 tmp; + u32 reference_clock = rdev->clock.spll.reference_freq; + u32 reference_divider, post_divider; + u32 fbdiv; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + engine_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = 1 + dividers.ref_div; + + if (dividers.enable_post_div) + post_divider = (0x0f & (dividers.post_div >> 4)) + (0x0f & dividers.post_div) + 2; + else + post_divider = 1; + + tmp = (u64) engine_clock * reference_divider * post_divider * 16384; + do_div(tmp, reference_clock); + fbdiv = (u32) tmp; + + if (dividers.enable_post_div) + spll_func_cntl |= SPLL_DIVEN; + else + spll_func_cntl &= ~SPLL_DIVEN; + spll_func_cntl &= ~(SPLL_HILEN_MASK | SPLL_LOLEN_MASK | SPLL_REF_DIV_MASK); + spll_func_cntl |= SPLL_REF_DIV(dividers.ref_div); + spll_func_cntl |= SPLL_HILEN((dividers.post_div >> 4) & 0xf); + spll_func_cntl |= SPLL_LOLEN(dividers.post_div & 0xf); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(2); + + spll_func_cntl_3 &= ~SPLL_FB_DIV_MASK; + spll_func_cntl_3 |= SPLL_FB_DIV(fbdiv); + spll_func_cntl_3 |= SPLL_DITHEN; + + if (pi->sclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = engine_clock * post_divider; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = ss.percentage * fbdiv / (clk_s * 10000); + + cg_spll_spread_spectrum &= ~CLKS_MASK; + cg_spll_spread_spectrum |= CLKS(clk_s); + cg_spll_spread_spectrum |= SSEN; + + cg_spll_spread_spectrum_2 &= ~CLKV_MASK; + cg_spll_spread_spectrum_2 |= CLKV(clk_v); + } + } + + sclk->sclk_value = cpu_to_be32(engine_clock); + sclk->vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + sclk->vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + sclk->vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + sclk->vCG_SPLL_SPREAD_SPECTRUM = cpu_to_be32(cg_spll_spread_spectrum); + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cpu_to_be32(cg_spll_spread_spectrum_2); + + return 0; +} + +int rv770_populate_vddc_value(struct radeon_device *rdev, u16 vddc, + RV770_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + if (!pi->voltage_control) { + voltage->index = 0; + voltage->value = 0; + return 0; + } + + for (i = 0; i < pi->valid_vddc_entries; i++) { + if (vddc <= pi->vddc_table[i].vddc) { + voltage->index = pi->vddc_table[i].vddc_index; + voltage->value = cpu_to_be16(vddc); + break; + } + } + + if (i == pi->valid_vddc_entries) + return -EINVAL; + + return 0; +} + +int rv770_populate_mvdd_value(struct radeon_device *rdev, u32 mclk, + RV770_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (!pi->mvdd_control) { + voltage->index = MVDD_HIGH_INDEX; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + return 0; + } + + if (mclk <= pi->mvdd_split_frequency) { + voltage->index = MVDD_LOW_INDEX; + voltage->value = cpu_to_be16(MVDD_LOW_VALUE); + } else { + voltage->index = MVDD_HIGH_INDEX; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + } + + return 0; +} + +static int rv770_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + RV770_SMC_HW_PERFORMANCE_LEVEL *level, + u8 watermark_level) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int ret; + + level->gen2PCIE = pi->pcie_gen2 ? + ((pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? 1 : 0) : 0; + level->gen2XSP = (pl->flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) ? 1 : 0; + level->backbias = (pl->flags & ATOM_PPLIB_R600_FLAGS_BACKBIASENABLE) ? 1 : 0; + level->displayWatermark = watermark_level; + + if (rdev->family == CHIP_RV740) + ret = rv740_populate_sclk_value(rdev, pl->sclk, + &level->sclk); + else if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + ret = rv730_populate_sclk_value(rdev, pl->sclk, + &level->sclk); + else + ret = rv770_populate_sclk_value(rdev, pl->sclk, + &level->sclk); + if (ret) + return ret; + + if (rdev->family == CHIP_RV740) { + if (pi->mem_gddr5) { + if (pl->mclk <= pi->mclk_strobe_mode_threshold) + level->strobeMode = + rv740_get_mclk_frequency_ratio(pl->mclk) | 0x10; + else + level->strobeMode = 0; + + if (pl->mclk > pi->mclk_edc_enable_threshold) + level->mcFlags = SMC_MC_EDC_RD_FLAG | SMC_MC_EDC_WR_FLAG; + else + level->mcFlags = 0; + } + ret = rv740_populate_mclk_value(rdev, pl->sclk, + pl->mclk, &level->mclk); + } else if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + ret = rv730_populate_mclk_value(rdev, pl->sclk, + pl->mclk, &level->mclk); + else + ret = rv770_populate_mclk_value(rdev, pl->sclk, + pl->mclk, &level->mclk); + if (ret) + return ret; + + ret = rv770_populate_vddc_value(rdev, pl->vddc, + &level->vddc); + if (ret) + return ret; + + ret = rv770_populate_mvdd_value(rdev, pl->mclk, &level->mvdd); + + return ret; +} + +static int rv770_convert_power_state_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + int ret; + + if (!(radeon_state->caps & ATOM_PPLIB_DISALLOW_ON_DC)) + smc_state->flags |= PPSMC_SWSTATE_FLAG_DC; + + ret = rv770_convert_power_level_to_smc(rdev, + &state->low, + &smc_state->levels[0], + PPSMC_DISPLAY_WATERMARK_LOW); + if (ret) + return ret; + + ret = rv770_convert_power_level_to_smc(rdev, + &state->medium, + &smc_state->levels[1], + PPSMC_DISPLAY_WATERMARK_LOW); + if (ret) + return ret; + + ret = rv770_convert_power_level_to_smc(rdev, + &state->high, + &smc_state->levels[2], + PPSMC_DISPLAY_WATERMARK_HIGH); + if (ret) + return ret; + + smc_state->levels[0].arbValue = MC_CG_ARB_FREQ_F1; + smc_state->levels[1].arbValue = MC_CG_ARB_FREQ_F2; + smc_state->levels[2].arbValue = MC_CG_ARB_FREQ_F3; + + smc_state->levels[0].seqValue = rv770_get_seq_value(rdev, + &state->low); + smc_state->levels[1].seqValue = rv770_get_seq_value(rdev, + &state->medium); + smc_state->levels[2].seqValue = rv770_get_seq_value(rdev, + &state->high); + + rv770_populate_smc_sp(rdev, radeon_state, smc_state); + + return rv770_populate_smc_t(rdev, radeon_state, smc_state); + +} + +u32 rv770_calculate_memory_refresh_rate(struct radeon_device *rdev, + u32 engine_clock) +{ + u32 dram_rows; + u32 dram_refresh_rate; + u32 mc_arb_rfsh_rate; + u32 tmp; + + tmp = (RREG32(MC_ARB_RAMCFG) & NOOFROWS_MASK) >> NOOFROWS_SHIFT; + dram_rows = 1 << (tmp + 10); + tmp = RREG32(MC_SEQ_MISC0) & 3; + dram_refresh_rate = 1 << (tmp + 3); + mc_arb_rfsh_rate = ((engine_clock * 10) * dram_refresh_rate / dram_rows - 32) / 64; + + return mc_arb_rfsh_rate; +} + +static void rv770_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct rv7xx_ps *state = rv770_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 sqm_ratio; + u32 arb_refresh_rate; + u32 high_clock; + + if (state->high.sclk < (state->low.sclk * 0xFF / 0x40)) + high_clock = state->high.sclk; + else + high_clock = (state->low.sclk * 0xFF / 0x40); + + radeon_atom_set_engine_dram_timings(rdev, high_clock, + state->high.mclk); + + sqm_ratio = + STATE0(64 * high_clock / pi->boot_sclk) | + STATE1(64 * high_clock / state->low.sclk) | + STATE2(64 * high_clock / state->medium.sclk) | + STATE3(64 * high_clock / state->high.sclk); + WREG32(MC_ARB_SQM_RATIO, sqm_ratio); + + arb_refresh_rate = + POWERMODE0(rv770_calculate_memory_refresh_rate(rdev, pi->boot_sclk)) | + POWERMODE1(rv770_calculate_memory_refresh_rate(rdev, state->low.sclk)) | + POWERMODE2(rv770_calculate_memory_refresh_rate(rdev, state->medium.sclk)) | + POWERMODE3(rv770_calculate_memory_refresh_rate(rdev, state->high.sclk)); + WREG32(MC_ARB_RFSH_RATE, arb_refresh_rate); +} + +void rv770_enable_backbias(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, BACKBIAS_PAD_EN, ~BACKBIAS_PAD_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~(BACKBIAS_VALUE | BACKBIAS_PAD_EN)); +} + +static void rv770_enable_spread_spectrum(struct radeon_device *rdev, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + if (pi->sclk_ss) + WREG32_P(GENERAL_PWRMGT, DYN_SPREAD_SPECTRUM_EN, ~DYN_SPREAD_SPECTRUM_EN); + + if (pi->mclk_ss) { + if (rdev->family == CHIP_RV740) + rv740_enable_mclk_spread_spectrum(rdev, true); + } + } else { + WREG32_P(CG_SPLL_SPREAD_SPECTRUM, 0, ~SSEN); + + WREG32_P(GENERAL_PWRMGT, 0, ~DYN_SPREAD_SPECTRUM_EN); + + WREG32_P(CG_MPLL_SPREAD_SPECTRUM, 0, ~SSEN); + + if (rdev->family == CHIP_RV740) + rv740_enable_mclk_spread_spectrum(rdev, false); + } +} + +static void rv770_program_mpll_timing_parameters(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if ((rdev->family == CHIP_RV770) && !pi->mem_gddr5) { + WREG32(MPLL_TIME, + (MPLL_LOCK_TIME(R600_MPLLLOCKTIME_DFLT * pi->ref_div) | + MPLL_RESET_TIME(R600_MPLLRESETTIME_DFLT))); + } +} + +void rv770_setup_bsp(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(pi->asi, + xclk, + 16, + &pi->bsp, + &pi->bsu); + + r600_calculate_u_and_p(pi->pasi, + xclk, + 16, + &pi->pbsp, + &pi->pbsu); + + pi->dsp = BSP(pi->bsp) | BSU(pi->bsu); + pi->psp = BSP(pi->pbsp) | BSU(pi->pbsu); + + WREG32(CG_BSP, pi->dsp); + +} + +void rv770_program_git(struct radeon_device *rdev) +{ + WREG32_P(CG_GIT, CG_GICST(R600_GICST_DFLT), ~CG_GICST_MASK); +} + +void rv770_program_tp(struct radeon_device *rdev) +{ + int i; + enum r600_td td = R600_TD_DFLT; + + for (i = 0; i < R600_PM_NUMBER_OF_TC; i++) + WREG32(CG_FFCT_0 + (i * 4), (UTC_0(r600_utc[i]) | DTC_0(r600_dtc[i]))); + + if (td == R600_TD_AUTO) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_FORCE_TREND_SEL); + else + WREG32_P(SCLK_PWRMGT_CNTL, FIR_FORCE_TREND_SEL, ~FIR_FORCE_TREND_SEL); + if (td == R600_TD_UP) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_TREND_MODE); + if (td == R600_TD_DOWN) + WREG32_P(SCLK_PWRMGT_CNTL, FIR_TREND_MODE, ~FIR_TREND_MODE); +} + +void rv770_program_tpp(struct radeon_device *rdev) +{ + WREG32(CG_TPC, R600_TPC_DFLT); +} + +void rv770_program_sstp(struct radeon_device *rdev) +{ + WREG32(CG_SSP, (SSTU(R600_SSTU_DFLT) | SST(R600_SST_DFLT))); +} + +void rv770_program_engine_speed_parameters(struct radeon_device *rdev) +{ + WREG32_P(SPLL_CNTL_MODE, SPLL_DIV_SYNC, ~SPLL_DIV_SYNC); +} + +static void rv770_enable_display_gap(struct radeon_device *rdev) +{ + u32 tmp = RREG32(CG_DISPLAY_GAP_CNTL); + + tmp &= ~(DISP1_GAP_MCHG_MASK | DISP2_GAP_MCHG_MASK); + tmp |= (DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE) | + DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE)); + WREG32(CG_DISPLAY_GAP_CNTL, tmp); +} + +void rv770_program_vc(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + WREG32(CG_FTV, pi->vrc); +} + +void rv770_clear_vc(struct radeon_device *rdev) +{ + WREG32(CG_FTV, 0); +} + +int rv770_upload_firmware(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int ret; + + rv770_reset_smc(rdev); + rv770_stop_smc_clock(rdev); + + ret = rv770_load_smc_ucode(rdev, pi->sram_end); + if (ret) + return ret; + + return 0; +} + +static int rv770_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + u32 mpll_ad_func_cntl = + pi->clk_regs.rv770.mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2 = + pi->clk_regs.rv770.mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl = + pi->clk_regs.rv770.mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2 = + pi->clk_regs.rv770.mpll_dq_func_cntl_2; + u32 spll_func_cntl = + pi->clk_regs.rv770.cg_spll_func_cntl; + u32 spll_func_cntl_2 = + pi->clk_regs.rv770.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = + pi->clk_regs.rv770.cg_spll_func_cntl_3; + u32 mclk_pwrmgt_cntl; + u32 dll_cntl; + + table->ACPIState = table->initialState; + + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + rv770_populate_vddc_value(rdev, pi->acpi_vddc, + &table->ACPIState.levels[0].vddc); + if (pi->pcie_gen2) { + if (pi->acpi_pcie_gen2) + table->ACPIState.levels[0].gen2PCIE = 1; + else + table->ACPIState.levels[0].gen2PCIE = 0; + } else + table->ACPIState.levels[0].gen2PCIE = 0; + if (pi->acpi_pcie_gen2) + table->ACPIState.levels[0].gen2XSP = 1; + else + table->ACPIState.levels[0].gen2XSP = 0; + } else { + rv770_populate_vddc_value(rdev, pi->min_vddc_in_table, + &table->ACPIState.levels[0].vddc); + table->ACPIState.levels[0].gen2PCIE = 0; + } + + + mpll_ad_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN; + + mpll_dq_func_cntl_2 |= BIAS_GEN_PDNB | RESET_EN; + + mclk_pwrmgt_cntl = (MRDCKA0_RESET | + MRDCKA1_RESET | + MRDCKB0_RESET | + MRDCKB1_RESET | + MRDCKC0_RESET | + MRDCKC1_RESET | + MRDCKD0_RESET | + MRDCKD1_RESET); + + dll_cntl = 0xff000000; + + spll_func_cntl |= SPLL_RESET | SPLL_SLEEP | SPLL_BYPASS_EN; + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL_2 = cpu_to_be32(mpll_ad_func_cntl_2); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + table->ACPIState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL_2 = cpu_to_be32(mpll_dq_func_cntl_2); + + table->ACPIState.levels[0].mclk.mclk770.vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.mclk770.vDLL_CNTL = cpu_to_be32(dll_cntl); + + table->ACPIState.levels[0].mclk.mclk770.mclk_value = 0; + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(spll_func_cntl_3); + + table->ACPIState.levels[0].sclk.sclk_value = 0; + + rv770_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + table->ACPIState.levels[1] = table->ACPIState.levels[0]; + table->ACPIState.levels[2] = table->ACPIState.levels[0]; + + return 0; +} + +int rv770_populate_initial_mvdd_value(struct radeon_device *rdev, + RV770_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if ((pi->s0_vid_lower_smio_cntl & pi->mvdd_mask_low) == + (pi->mvdd_low_smio[MVDD_LOW_INDEX] & pi->mvdd_mask_low) ) { + voltage->index = MVDD_LOW_INDEX; + voltage->value = cpu_to_be16(MVDD_LOW_VALUE); + } else { + voltage->index = MVDD_HIGH_INDEX; + voltage->value = cpu_to_be16(MVDD_HIGH_VALUE); + } + + return 0; +} + +static int rv770_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_ps *initial_state = rv770_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 a_t; + + table->initialState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mpll_ad_func_cntl); + table->initialState.levels[0].mclk.mclk770.vMPLL_AD_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_ad_func_cntl_2); + table->initialState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mpll_dq_func_cntl); + table->initialState.levels[0].mclk.mclk770.vMPLL_DQ_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_dq_func_cntl_2); + table->initialState.levels[0].mclk.mclk770.vMCLK_PWRMGT_CNTL = + cpu_to_be32(pi->clk_regs.rv770.mclk_pwrmgt_cntl); + table->initialState.levels[0].mclk.mclk770.vDLL_CNTL = + cpu_to_be32(pi->clk_regs.rv770.dll_cntl); + + table->initialState.levels[0].mclk.mclk770.vMPLL_SS = + cpu_to_be32(pi->clk_regs.rv770.mpll_ss1); + table->initialState.levels[0].mclk.mclk770.vMPLL_SS2 = + cpu_to_be32(pi->clk_regs.rv770.mpll_ss2); + + table->initialState.levels[0].mclk.mclk770.mclk_value = + cpu_to_be32(initial_state->low.mclk); + + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl_2); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_func_cntl_3); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_spread_spectrum); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM_2 = + cpu_to_be32(pi->clk_regs.rv770.cg_spll_spread_spectrum_2); + + table->initialState.levels[0].sclk.sclk_value = + cpu_to_be32(initial_state->low.sclk); + + table->initialState.levels[0].arbValue = MC_CG_ARB_FREQ_F0; + + table->initialState.levels[0].seqValue = + rv770_get_seq_value(rdev, &initial_state->low); + + rv770_populate_vddc_value(rdev, + initial_state->low.vddc, + &table->initialState.levels[0].vddc); + rv770_populate_initial_mvdd_value(rdev, + &table->initialState.levels[0].mvdd); + + a_t = CG_R(0xffff) | CG_L(0); + table->initialState.levels[0].aT = cpu_to_be32(a_t); + + table->initialState.levels[0].bSP = cpu_to_be32(pi->dsp); + + if (pi->boot_in_gen2) + table->initialState.levels[0].gen2PCIE = 1; + else + table->initialState.levels[0].gen2PCIE = 0; + if (initial_state->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + table->initialState.levels[0].gen2XSP = 1; + else + table->initialState.levels[0].gen2XSP = 0; + + if (rdev->family == CHIP_RV740) { + if (pi->mem_gddr5) { + if (initial_state->low.mclk <= pi->mclk_strobe_mode_threshold) + table->initialState.levels[0].strobeMode = + rv740_get_mclk_frequency_ratio(initial_state->low.mclk) | 0x10; + else + table->initialState.levels[0].strobeMode = 0; + + if (initial_state->low.mclk >= pi->mclk_edc_enable_threshold) + table->initialState.levels[0].mcFlags = SMC_MC_EDC_RD_FLAG | SMC_MC_EDC_WR_FLAG; + else + table->initialState.levels[0].mcFlags = 0; + } + } + + table->initialState.levels[1] = table->initialState.levels[0]; + table->initialState.levels[2] = table->initialState.levels[0]; + + table->initialState.flags |= PPSMC_SWSTATE_FLAG_DC; + + return 0; +} + +static int rv770_populate_smc_vddc_table(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + for (i = 0; i < pi->valid_vddc_entries; i++) { + table->highSMIO[pi->vddc_table[i].vddc_index] = + pi->vddc_table[i].high_smio; + table->lowSMIO[pi->vddc_table[i].vddc_index] = + cpu_to_be32(pi->vddc_table[i].low_smio); + } + + table->voltageMaskTable.highMask[RV770_SMC_VOLTAGEMASK_VDDC] = 0; + table->voltageMaskTable.lowMask[RV770_SMC_VOLTAGEMASK_VDDC] = + cpu_to_be32(pi->vddc_mask_low); + + for (i = 0; + ((i < pi->valid_vddc_entries) && + (pi->max_vddc_in_table > + pi->vddc_table[i].vddc)); + i++); + + table->maxVDDCIndexInPPTable = + pi->vddc_table[i].vddc_index; + + return 0; +} + +static int rv770_populate_smc_mvdd_table(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (pi->mvdd_control) { + table->lowSMIO[MVDD_HIGH_INDEX] |= + cpu_to_be32(pi->mvdd_low_smio[MVDD_HIGH_INDEX]); + table->lowSMIO[MVDD_LOW_INDEX] |= + cpu_to_be32(pi->mvdd_low_smio[MVDD_LOW_INDEX]); + + table->voltageMaskTable.highMask[RV770_SMC_VOLTAGEMASK_MVDD] = 0; + table->voltageMaskTable.lowMask[RV770_SMC_VOLTAGEMASK_MVDD] = + cpu_to_be32(pi->mvdd_mask_low); + } + + return 0; +} + +static int rv770_init_smc_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct rv7xx_ps *boot_state = rv770_get_ps(radeon_boot_state); + RV770_SMC_STATETABLE *table = &pi->smc_statetable; + int ret; + + memset(table, 0, sizeof(RV770_SMC_STATETABLE)); + + pi->boot_sclk = boot_state->low.sclk; + + rv770_populate_smc_vddc_table(rdev, table); + rv770_populate_smc_mvdd_table(rdev, table); + + switch (rdev->pm.int_thermal_type) { + case THERMAL_TYPE_RV770: + case THERMAL_TYPE_ADT7473_WITH_INTERNAL: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_INTERNAL; + break; + case THERMAL_TYPE_NONE: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_NONE; + break; + case THERMAL_TYPE_EXTERNAL_GPIO: + default: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL; + break; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) { + table->systemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_DONT_WAIT_FOR_VBLANK_ON_ALERT) + table->extraFlags |= PPSMC_EXTRAFLAGS_AC2DC_DONT_WAIT_FOR_VBLANK; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_GOTO_BOOT_ON_ALERT) + table->extraFlags |= PPSMC_EXTRAFLAGS_AC2DC_ACTION_GOTOINITIALSTATE; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (pi->mem_gddr5) + table->systemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + ret = rv730_populate_smc_initial_state(rdev, radeon_boot_state, table); + else + ret = rv770_populate_smc_initial_state(rdev, radeon_boot_state, table); + if (ret) + return ret; + + if (rdev->family == CHIP_RV740) + ret = rv740_populate_smc_acpi_state(rdev, table); + else if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + ret = rv730_populate_smc_acpi_state(rdev, table); + else + ret = rv770_populate_smc_acpi_state(rdev, table); + if (ret) + return ret; + + table->driverState = table->initialState; + + return rv770_copy_bytes_to_smc(rdev, + pi->state_table_start, + (const u8 *)table, + sizeof(RV770_SMC_STATETABLE), + pi->sram_end); +} + +static int rv770_construct_vddc_table(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u16 min, max, step; + u32 steps = 0; + u8 vddc_index = 0; + u32 i; + + radeon_atom_get_min_voltage(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, &min); + radeon_atom_get_max_voltage(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, &max); + radeon_atom_get_voltage_step(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, &step); + + steps = (max - min) / step + 1; + + if (steps > MAX_NO_VREG_STEPS) + return -EINVAL; + + for (i = 0; i < steps; i++) { + u32 gpio_pins, gpio_mask; + + pi->vddc_table[i].vddc = (u16)(min + i * step); + radeon_atom_get_voltage_gpio_settings(rdev, + pi->vddc_table[i].vddc, + SET_VOLTAGE_TYPE_ASIC_VDDC, + &gpio_pins, &gpio_mask); + pi->vddc_table[i].low_smio = gpio_pins & gpio_mask; + pi->vddc_table[i].high_smio = 0; + pi->vddc_mask_low = gpio_mask; + if (i > 0) { + if ((pi->vddc_table[i].low_smio != + pi->vddc_table[i - 1].low_smio ) || + (pi->vddc_table[i].high_smio != + pi->vddc_table[i - 1].high_smio)) + vddc_index++; + } + pi->vddc_table[i].vddc_index = vddc_index; + } + + pi->valid_vddc_entries = (u8)steps; + + return 0; +} + +static u32 rv770_get_mclk_split_point(struct atom_memory_info *memory_info) +{ + if (memory_info->mem_type == MEM_TYPE_GDDR3) + return 30000; + + return 0; +} + +static int rv770_get_mvdd_pin_configuration(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 gpio_pins, gpio_mask; + + radeon_atom_get_voltage_gpio_settings(rdev, + MVDD_HIGH_VALUE, SET_VOLTAGE_TYPE_ASIC_MVDDC, + &gpio_pins, &gpio_mask); + pi->mvdd_mask_low = gpio_mask; + pi->mvdd_low_smio[MVDD_HIGH_INDEX] = + gpio_pins & gpio_mask; + + radeon_atom_get_voltage_gpio_settings(rdev, + MVDD_LOW_VALUE, SET_VOLTAGE_TYPE_ASIC_MVDDC, + &gpio_pins, &gpio_mask); + pi->mvdd_low_smio[MVDD_LOW_INDEX] = + gpio_pins & gpio_mask; + + return 0; +} + +u8 rv770_get_memory_module_index(struct radeon_device *rdev) +{ + return (u8) ((RREG32(BIOS_SCRATCH_4) >> 16) & 0xff); +} + +static int rv770_get_mvdd_configuration(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 memory_module_index; + struct atom_memory_info memory_info; + + memory_module_index = rv770_get_memory_module_index(rdev); + + if (radeon_atom_get_memory_info(rdev, memory_module_index, &memory_info)) { + pi->mvdd_control = false; + return 0; + } + + pi->mvdd_split_frequency = + rv770_get_mclk_split_point(&memory_info); + + if (pi->mvdd_split_frequency == 0) { + pi->mvdd_control = false; + return 0; + } + + return rv770_get_mvdd_pin_configuration(rdev); +} + +void rv770_enable_voltage_control(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, VOLT_PWRMGT_EN, ~VOLT_PWRMGT_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~VOLT_PWRMGT_EN); +} + +static void rv770_program_display_gap(struct radeon_device *rdev) +{ + u32 tmp = RREG32(CG_DISPLAY_GAP_CNTL); + + tmp &= ~(DISP1_GAP_MCHG_MASK | DISP2_GAP_MCHG_MASK); + if (rdev->pm.dpm.new_active_crtcs & 1) { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + } else if (rdev->pm.dpm.new_active_crtcs & 2) { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK); + } else { + tmp |= DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + tmp |= DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE); + } + WREG32(CG_DISPLAY_GAP_CNTL, tmp); +} + +static void rv770_enable_dynamic_pcie_gen2(struct radeon_device *rdev, + bool enable) +{ + rv770_enable_bif_dynamic_pcie_gen2(rdev, enable); + + if (enable) + WREG32_P(GENERAL_PWRMGT, ENABLE_GEN2PCIE, ~ENABLE_GEN2PCIE); + else + WREG32_P(GENERAL_PWRMGT, 0, ~ENABLE_GEN2PCIE); +} + +static void r7xx_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + if ((rdev->family == CHIP_RV730) || + (rdev->family == CHIP_RV710) || + (rdev->family == CHIP_RV740)) + rv730_program_memory_timing_parameters(rdev, radeon_new_state); + else + rv770_program_memory_timing_parameters(rdev, radeon_new_state); +} + +static int rv770_upload_sw_state(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u16 address = pi->state_table_start + + offsetof(RV770_SMC_STATETABLE, driverState); + RV770_SMC_SWSTATE state = { 0 }; + int ret; + + ret = rv770_convert_power_state_to_smc(rdev, radeon_new_state, &state); + if (ret) + return ret; + + return rv770_copy_bytes_to_smc(rdev, address, (const u8 *)&state, + sizeof(RV770_SMC_SWSTATE), + pi->sram_end); +} + +int rv770_halt_smc(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_Halt) != PPSMC_Result_OK) + return -EINVAL; + + if (rv770_wait_for_smc_inactive(rdev) != PPSMC_Result_OK) + return -EINVAL; + + return 0; +} + +int rv770_resume_smc(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_Resume) != PPSMC_Result_OK) + return -EINVAL; + return 0; +} + +int rv770_set_sw_state(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_SwitchToSwState) != PPSMC_Result_OK) + return -EINVAL; + return 0; +} + +int rv770_set_boot_state(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_SwitchToInitialState) != PPSMC_Result_OK) + return -EINVAL; + return 0; +} + +void rv770_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv7xx_ps *new_state = rv770_get_ps(new_ps); + struct rv7xx_ps *current_state = rv770_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->high.sclk >= current_state->high.sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +void rv770_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps) +{ + struct rv7xx_ps *new_state = rv770_get_ps(new_ps); + struct rv7xx_ps *current_state = rv770_get_ps(old_ps); + + if ((new_ps->vclk == old_ps->vclk) && + (new_ps->dclk == old_ps->dclk)) + return; + + if (new_state->high.sclk < current_state->high.sclk) + return; + + radeon_set_uvd_clocks(rdev, new_ps->vclk, new_ps->dclk); +} + +int rv770_restrict_performance_levels_before_switch(struct radeon_device *rdev) +{ + if (rv770_send_msg_to_smc(rdev, (PPSMC_Msg)(PPSMC_MSG_NoForcedLevel)) != PPSMC_Result_OK) + return -EINVAL; + + if (rv770_send_msg_to_smc(rdev, (PPSMC_Msg)(PPSMC_MSG_TwoLevelsDisabled)) != PPSMC_Result_OK) + return -EINVAL; + + return 0; +} + +int rv770_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + PPSMC_Msg msg; + + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_ZeroLevelsDisabled) != PPSMC_Result_OK) + return -EINVAL; + msg = PPSMC_MSG_ForceHigh; + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_NoForcedLevel) != PPSMC_Result_OK) + return -EINVAL; + msg = (PPSMC_Msg)(PPSMC_MSG_TwoLevelsDisabled); + } else { + if (rv770_send_msg_to_smc(rdev, PPSMC_MSG_NoForcedLevel) != PPSMC_Result_OK) + return -EINVAL; + msg = (PPSMC_Msg)(PPSMC_MSG_ZeroLevelsDisabled); + } + + if (rv770_send_msg_to_smc(rdev, msg) != PPSMC_Result_OK) + return -EINVAL; + + rdev->pm.dpm.forced_level = level; + + return 0; +} + +void r7xx_start_smc(struct radeon_device *rdev) +{ + rv770_start_smc(rdev); + rv770_start_smc_clock(rdev); +} + + +void r7xx_stop_smc(struct radeon_device *rdev) +{ + rv770_reset_smc(rdev); + rv770_stop_smc_clock(rdev); +} + +static void rv770_read_clock_registers(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + pi->clk_regs.rv770.cg_spll_func_cntl = + RREG32(CG_SPLL_FUNC_CNTL); + pi->clk_regs.rv770.cg_spll_func_cntl_2 = + RREG32(CG_SPLL_FUNC_CNTL_2); + pi->clk_regs.rv770.cg_spll_func_cntl_3 = + RREG32(CG_SPLL_FUNC_CNTL_3); + pi->clk_regs.rv770.cg_spll_spread_spectrum = + RREG32(CG_SPLL_SPREAD_SPECTRUM); + pi->clk_regs.rv770.cg_spll_spread_spectrum_2 = + RREG32(CG_SPLL_SPREAD_SPECTRUM_2); + pi->clk_regs.rv770.mpll_ad_func_cntl = + RREG32(MPLL_AD_FUNC_CNTL); + pi->clk_regs.rv770.mpll_ad_func_cntl_2 = + RREG32(MPLL_AD_FUNC_CNTL_2); + pi->clk_regs.rv770.mpll_dq_func_cntl = + RREG32(MPLL_DQ_FUNC_CNTL); + pi->clk_regs.rv770.mpll_dq_func_cntl_2 = + RREG32(MPLL_DQ_FUNC_CNTL_2); + pi->clk_regs.rv770.mclk_pwrmgt_cntl = + RREG32(MCLK_PWRMGT_CNTL); + pi->clk_regs.rv770.dll_cntl = RREG32(DLL_CNTL); +} + +static void r7xx_read_clock_registers(struct radeon_device *rdev) +{ + if (rdev->family == CHIP_RV740) + rv740_read_clock_registers(rdev); + else if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_read_clock_registers(rdev); + else + rv770_read_clock_registers(rdev); +} + +void rv770_read_voltage_smio_registers(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + pi->s0_vid_lower_smio_cntl = + RREG32(S0_VID_LOWER_SMIO_CNTL); +} + +void rv770_reset_smio_status(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 sw_smio_index, vid_smio_cntl; + + sw_smio_index = + (RREG32(GENERAL_PWRMGT) & SW_SMIO_INDEX_MASK) >> SW_SMIO_INDEX_SHIFT; + switch (sw_smio_index) { + case 3: + vid_smio_cntl = RREG32(S3_VID_LOWER_SMIO_CNTL); + break; + case 2: + vid_smio_cntl = RREG32(S2_VID_LOWER_SMIO_CNTL); + break; + case 1: + vid_smio_cntl = RREG32(S1_VID_LOWER_SMIO_CNTL); + break; + case 0: + return; + default: + vid_smio_cntl = pi->s0_vid_lower_smio_cntl; + break; + } + + WREG32(S0_VID_LOWER_SMIO_CNTL, vid_smio_cntl); + WREG32_P(GENERAL_PWRMGT, SW_SMIO_INDEX(0), ~SW_SMIO_INDEX_MASK); +} + +void rv770_get_memory_type(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + + tmp = RREG32(MC_SEQ_MISC0); + + if (((tmp & MC_SEQ_MISC0_GDDR5_MASK) >> MC_SEQ_MISC0_GDDR5_SHIFT) == + MC_SEQ_MISC0_GDDR5_VALUE) + pi->mem_gddr5 = true; + else + pi->mem_gddr5 = false; + +} + +void rv770_get_pcie_gen2_status(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 tmp; + + tmp = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + + if ((tmp & LC_OTHER_SIDE_EVER_SENT_GEN2) && + (tmp & LC_OTHER_SIDE_SUPPORTS_GEN2)) + pi->pcie_gen2 = true; + else + pi->pcie_gen2 = false; + + if (pi->pcie_gen2) { + if (tmp & LC_CURRENT_DATA_RATE) + pi->boot_in_gen2 = true; + else + pi->boot_in_gen2 = false; + } else + pi->boot_in_gen2 = false; +} + +#if 0 +static int rv770_enter_ulp_state(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (pi->gfx_clock_gating) { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_TILING_CONFIG); + } + + WREG32_P(SMC_MSG, HOST_SMC_MSG(PPSMC_MSG_SwitchToMinimumPower), + ~HOST_SMC_MSG_MASK); + + DRM_UDELAY(7000); + + return 0; +} + +static int rv770_exit_ulp_state(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + WREG32_P(SMC_MSG, HOST_SMC_MSG(PPSMC_MSG_ResumeFromMinimumPower), + ~HOST_SMC_MSG_MASK); + + DRM_UDELAY(7000); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (((RREG32(SMC_MSG) & HOST_SMC_RESP_MASK) >> HOST_SMC_RESP_SHIFT) == 1) + break; + DRM_UDELAY(1000); + } + + if (pi->gfx_clock_gating) + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + + return 0; +} +#endif + +static void rv770_get_mclk_odt_threshold(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 memory_module_index; + struct atom_memory_info memory_info; + + pi->mclk_odt_threshold = 0; + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) { + memory_module_index = rv770_get_memory_module_index(rdev); + + if (radeon_atom_get_memory_info(rdev, memory_module_index, &memory_info)) + return; + + if (memory_info.mem_type == MEM_TYPE_DDR2 || + memory_info.mem_type == MEM_TYPE_DDR3) + pi->mclk_odt_threshold = 30000; + } +} + +void rv770_get_max_vddc(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u16 vddc; + + if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc)) + pi->max_vddc = 0; + else + pi->max_vddc = vddc; +} + +void rv770_program_response_times(struct radeon_device *rdev) +{ + u32 voltage_response_time, backbias_response_time; + u32 acpi_delay_time, vbi_time_out; + u32 vddc_dly, bb_dly, acpi_dly, vbi_dly; + u32 reference_clock; + + voltage_response_time = (u32)rdev->pm.dpm.voltage_response_time; + backbias_response_time = (u32)rdev->pm.dpm.backbias_response_time; + + if (voltage_response_time == 0) + voltage_response_time = 1000; + + if (backbias_response_time == 0) + backbias_response_time = 1000; + + acpi_delay_time = 15000; + vbi_time_out = 100000; + + reference_clock = radeon_get_xclk(rdev); + + vddc_dly = (voltage_response_time * reference_clock) / 1600; + bb_dly = (backbias_response_time * reference_clock) / 1600; + acpi_dly = (acpi_delay_time * reference_clock) / 1600; + vbi_dly = (vbi_time_out * reference_clock) / 1600; + + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_delay_vreg, vddc_dly); + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_delay_bbias, bb_dly); + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_delay_acpi, acpi_dly); + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_mclk_chg_timeout, vbi_dly); +#if 0 + /* XXX look up hw revision */ + if (WEKIVA_A21) + rv770_write_smc_soft_register(rdev, + RV770_SMC_SOFT_REGISTER_baby_step_timer, + 0x10); +#endif +} + +static void rv770_program_dcodt_before_state_switch(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct rv7xx_ps *new_state = rv770_get_ps(radeon_new_state); + struct rv7xx_ps *current_state = rv770_get_ps(radeon_current_state); + bool current_use_dc = false; + bool new_use_dc = false; + + if (pi->mclk_odt_threshold == 0) + return; + + if (current_state->high.mclk <= pi->mclk_odt_threshold) + current_use_dc = true; + + if (new_state->high.mclk <= pi->mclk_odt_threshold) + new_use_dc = true; + + if (current_use_dc == new_use_dc) + return; + + if (!current_use_dc && new_use_dc) + return; + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_program_dcodt(rdev, new_use_dc); +} + +static void rv770_program_dcodt_after_state_switch(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct rv7xx_ps *new_state = rv770_get_ps(radeon_new_state); + struct rv7xx_ps *current_state = rv770_get_ps(radeon_current_state); + bool current_use_dc = false; + bool new_use_dc = false; + + if (pi->mclk_odt_threshold == 0) + return; + + if (current_state->high.mclk <= pi->mclk_odt_threshold) + current_use_dc = true; + + if (new_state->high.mclk <= pi->mclk_odt_threshold) + new_use_dc = true; + + if (current_use_dc == new_use_dc) + return; + + if (current_use_dc && !new_use_dc) + return; + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_program_dcodt(rdev, new_use_dc); +} + +static void rv770_retrieve_odt_values(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (pi->mclk_odt_threshold == 0) + return; + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_get_odt_values(rdev); +} + +static void rv770_set_dpm_event_sources(struct radeon_device *rdev, u32 sources) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + bool want_thermal_protection; + enum radeon_dpm_event_src dpm_event_src; + + switch (sources) { + case 0: + default: + want_thermal_protection = false; + break; + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGITAL; + break; + + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_EXTERNAL; + break; + + case ((1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL) | + (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL)): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL; + break; + } + + if (want_thermal_protection) { + WREG32_P(CG_THERMAL_CTRL, DPM_EVENT_SRC(dpm_event_src), ~DPM_EVENT_SRC_MASK); + if (pi->thermal_protection) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + } else { + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); + } +} + +void rv770_enable_auto_throttle_source(struct radeon_device *rdev, + enum radeon_dpm_auto_throttle_src source, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + if (!(pi->active_auto_throttle_sources & (1 << source))) { + pi->active_auto_throttle_sources |= 1 << source; + rv770_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } else { + if (pi->active_auto_throttle_sources & (1 << source)) { + pi->active_auto_throttle_sources &= ~(1 << source); + rv770_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } +} + +int rv770_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) +{ + int low_temp = 0 * 1000; + int high_temp = 255 * 1000; + + if (low_temp < min_temp) + low_temp = min_temp; + if (high_temp > max_temp) + high_temp = max_temp; + if (high_temp < low_temp) { + DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); + return -EINVAL; + } + + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTH(high_temp / 1000), ~DIG_THERM_INTH_MASK); + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTL(low_temp / 1000), ~DIG_THERM_INTL_MASK); + WREG32_P(CG_THERMAL_CTRL, DIG_THERM_DPM(high_temp / 1000), ~DIG_THERM_DPM_MASK); + + rdev->pm.dpm.thermal.min_temp = low_temp; + rdev->pm.dpm.thermal.max_temp = high_temp; + + return 0; +} + +int rv770_dpm_enable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (pi->gfx_clock_gating) + rv770_restore_cgcg(rdev); + + if (rv770_dpm_enabled(rdev)) + return -EINVAL; + + if (pi->voltage_control) { + rv770_enable_voltage_control(rdev, true); + ret = rv770_construct_vddc_table(rdev); + if (ret) { + DRM_ERROR("rv770_construct_vddc_table failed\n"); + return ret; + } + } + + if (pi->dcodt) + rv770_retrieve_odt_values(rdev); + + if (pi->mvdd_control) { + ret = rv770_get_mvdd_configuration(rdev); + if (ret) { + DRM_ERROR("rv770_get_mvdd_configuration failed\n"); + return ret; + } + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_BACKBIAS) + rv770_enable_backbias(rdev, true); + + rv770_enable_spread_spectrum(rdev, true); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, true); + + rv770_program_mpll_timing_parameters(rdev); + rv770_setup_bsp(rdev); + rv770_program_git(rdev); + rv770_program_tp(rdev); + rv770_program_tpp(rdev); + rv770_program_sstp(rdev); + rv770_program_engine_speed_parameters(rdev); + rv770_enable_display_gap(rdev); + rv770_program_vc(rdev); + + if (pi->dynamic_pcie_gen2) + rv770_enable_dynamic_pcie_gen2(rdev, true); + + ret = rv770_upload_firmware(rdev); + if (ret) { + DRM_ERROR("rv770_upload_firmware failed\n"); + return ret; + } + ret = rv770_init_smc_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("rv770_init_smc_table failed\n"); + return ret; + } + + rv770_program_response_times(rdev); + r7xx_start_smc(rdev); + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_start_dpm(rdev); + else + rv770_start_dpm(rdev); + + if (pi->gfx_clock_gating) + rv770_gfx_clock_gating_enable(rdev, true); + + if (pi->mg_clock_gating) + rv770_mg_clock_gating_enable(rdev, true); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + PPSMC_Result result; + + ret = rv770_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + result = rv770_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + + if (result != PPSMC_Result_OK) + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + } + + rv770_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + return 0; +} + +void rv770_dpm_disable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (!rv770_dpm_enabled(rdev)) + return; + + rv770_clear_vc(rdev); + + if (pi->thermal_protection) + rv770_enable_thermal_protection(rdev, false); + + rv770_enable_spread_spectrum(rdev, false); + + if (pi->dynamic_pcie_gen2) + rv770_enable_dynamic_pcie_gen2(rdev, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + if (pi->gfx_clock_gating) + rv770_gfx_clock_gating_enable(rdev, false); + + if (pi->mg_clock_gating) + rv770_mg_clock_gating_enable(rdev, false); + + if ((rdev->family == CHIP_RV730) || (rdev->family == CHIP_RV710)) + rv730_stop_dpm(rdev); + else + rv770_stop_dpm(rdev); + + r7xx_stop_smc(rdev); + rv770_reset_smio_status(rdev); +} + +int rv770_dpm_set_power_state(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + struct radeon_ps *old_ps = rdev->pm.dpm.current_ps; + int ret; + + ret = rv770_restrict_performance_levels_before_switch(rdev); + if (ret) { + DRM_ERROR("rv770_restrict_performance_levels_before_switch failed\n"); + return ret; + } + rv770_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + ret = rv770_halt_smc(rdev); + if (ret) { + DRM_ERROR("rv770_halt_smc failed\n"); + return ret; + } + ret = rv770_upload_sw_state(rdev, new_ps); + if (ret) { + DRM_ERROR("rv770_upload_sw_state failed\n"); + return ret; + } + r7xx_program_memory_timing_parameters(rdev, new_ps); + if (pi->dcodt) + rv770_program_dcodt_before_state_switch(rdev, new_ps, old_ps); + ret = rv770_resume_smc(rdev); + if (ret) { + DRM_ERROR("rv770_resume_smc failed\n"); + return ret; + } + ret = rv770_set_sw_state(rdev); + if (ret) { + DRM_ERROR("rv770_set_sw_state failed\n"); + return ret; + } + if (pi->dcodt) + rv770_program_dcodt_after_state_switch(rdev, new_ps, old_ps); + rv770_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + ret = rv770_dpm_force_performance_level(rdev, RADEON_DPM_FORCED_LEVEL_AUTO); + if (ret) { + DRM_ERROR("rv770_dpm_force_performance_level failed\n"); + return ret; + } + + return 0; +} + +void rv770_dpm_reset_asic(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + + rv770_restrict_performance_levels_before_switch(rdev); + if (pi->dcodt) + rv770_program_dcodt_before_state_switch(rdev, boot_ps, boot_ps); + rv770_set_boot_state(rdev); + if (pi->dcodt) + rv770_program_dcodt_after_state_switch(rdev, boot_ps, boot_ps); +} + +void rv770_dpm_setup_asic(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + r7xx_read_clock_registers(rdev); + rv770_read_voltage_smio_registers(rdev); + rv770_get_memory_type(rdev); + if (pi->dcodt) + rv770_get_mclk_odt_threshold(rdev); + rv770_get_pcie_gen2_status(rdev); + + rv770_enable_acpi_pm(rdev); + + if (radeon_aspm != 0) { + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_ASPM_L0s) + rv770_enable_l0s(rdev); + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_ASPM_L1) + rv770_enable_l1(rdev); + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_TURNOFFPLL_ASPML1) + rv770_enable_pll_sleep_in_l1(rdev); + } +} + +void rv770_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + rv770_program_display_gap(rdev); +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void rv7xx_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else if (r600_is_uvd_state(rps->class, rps->class2)) { + rps->vclk = RV770_DEFAULT_VCLK_FREQ; + rps->dclk = RV770_DEFAULT_DCLK_FREQ; + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) + rdev->pm.dpm.boot_ps = rps; + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void rv7xx_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct rv7xx_ps *ps = rv770_get_ps(rps); + u32 sclk, mclk; + u16 vddc; + struct rv7xx_pl *pl; + + switch (index) { + case 0: + pl = &ps->low; + break; + case 1: + pl = &ps->medium; + break; + case 2: + default: + pl = &ps->high; + break; + } + + if (rdev->family >= CHIP_CEDAR) { + sclk = le16_to_cpu(clock_info->evergreen.usEngineClockLow); + sclk |= clock_info->evergreen.ucEngineClockHigh << 16; + mclk = le16_to_cpu(clock_info->evergreen.usMemoryClockLow); + mclk |= clock_info->evergreen.ucMemoryClockHigh << 16; + + pl->vddc = le16_to_cpu(clock_info->evergreen.usVDDC); + pl->vddci = le16_to_cpu(clock_info->evergreen.usVDDCI); + pl->flags = le32_to_cpu(clock_info->evergreen.ulFlags); + } else { + sclk = le16_to_cpu(clock_info->r600.usEngineClockLow); + sclk |= clock_info->r600.ucEngineClockHigh << 16; + mclk = le16_to_cpu(clock_info->r600.usMemoryClockLow); + mclk |= clock_info->r600.ucMemoryClockHigh << 16; + + pl->vddc = le16_to_cpu(clock_info->r600.usVDDC); + pl->flags = le32_to_cpu(clock_info->r600.ulFlags); + } + + pl->mclk = mclk; + pl->sclk = sclk; + + /* patch up vddc if necessary */ + if (pl->vddc == 0xff01) { + if (radeon_atom_get_max_vddc(rdev, 0, 0, &vddc) == 0) + pl->vddc = vddc; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) { + pi->acpi_vddc = pl->vddc; + if (rdev->family >= CHIP_CEDAR) + eg_pi->acpi_vddci = pl->vddci; + if (ps->low.flags & ATOM_PPLIB_R600_FLAGS_PCIEGEN2) + pi->acpi_pcie_gen2 = true; + else + pi->acpi_pcie_gen2 = false; + } + + if (rps->class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) { + if (rdev->family >= CHIP_BARTS) { + eg_pi->ulv.supported = true; + eg_pi->ulv.pl = pl; + } + } + + if (pi->min_vddc_in_table > pl->vddc) + pi->min_vddc_in_table = pl->vddc; + + if (pi->max_vddc_in_table < pl->vddc) + pi->max_vddc_in_table = pl->vddc; + + /* patch up boot state */ + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + u16 vddc, vddci, mvdd; + radeon_atombios_get_default_voltages(rdev, &vddc, &vddci, &mvdd); + pl->mclk = rdev->clock.default_mclk; + pl->sclk = rdev->clock.default_sclk; + pl->vddc = vddc; + pl->vddci = vddci; + } + + if (rdev->family >= CHIP_BARTS) { + if ((rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == + ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE) { + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk = pl->sclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.mclk = pl->mclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddc = pl->vddc; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddci = pl->vddci; + } + } +} + +int rv7xx_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j; + union pplib_clock_info *clock_info; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + struct rv7xx_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + power_info->pplib.ucNumStates, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + + for (i = 0; i < power_info->pplib.ucNumStates; i++) { + power_state = (union pplib_power_state *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset) + + i * power_info->pplib.ucStateEntrySize); + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset) + + (power_state->v1.ucNonClockStateIndex * + power_info->pplib.ucNonClockSize)); + if (power_info->pplib.ucStateEntrySize - 1) { + ps = kzalloc(sizeof(struct rv7xx_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + rv7xx_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + power_info->pplib.ucNonClockSize); + for (j = 0; j < (power_info->pplib.ucStateEntrySize - 1); j++) { + clock_info = (union pplib_clock_info *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset) + + (power_state->v1.ucClockStateIndices[j] * + power_info->pplib.ucClockInfoSize)); + rv7xx_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], j, + clock_info); + } + } + } + rdev->pm.dpm.num_ps = power_info->pplib.ucNumStates; + return 0; +} + +void rv770_get_engine_memory_ss(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct radeon_atom_ss ss; + + pi->sclk_ss = radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, 0); + pi->mclk_ss = radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, 0); + + if (pi->sclk_ss || pi->mclk_ss) + pi->dynamic_ss = true; + else + pi->dynamic_ss = false; +} + +int rv770_dpm_init(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi; + struct atom_clock_dividers dividers; + int ret; + + pi = kzalloc(sizeof(struct rv7xx_power_info), GFP_KERNEL); + if (pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = pi; + + rv770_get_max_vddc(rdev); + + pi->acpi_vddc = 0; + pi->min_vddc_in_table = 0; + pi->max_vddc_in_table = 0; + + ret = rv7xx_parse_power_table(rdev); + if (ret) + return ret; + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->ref_div = dividers.ref_div + 1; + else + pi->ref_div = R600_REFERENCEDIVIDER_DFLT; + + pi->mclk_strobe_mode_threshold = 30000; + pi->mclk_edc_enable_threshold = 30000; + + pi->rlp = RV770_RLP_DFLT; + pi->rmp = RV770_RMP_DFLT; + pi->lhp = RV770_LHP_DFLT; + pi->lmp = RV770_LMP_DFLT; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, 0); + + pi->mvdd_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_MVDDC, 0); + + rv770_get_engine_memory_ss(rdev); + + pi->asi = RV770_ASI_DFLT; + pi->pasi = RV770_HASI_DFLT; + pi->vrc = RV770_VRC_DFLT; + + pi->power_gating = false; + + pi->gfx_clock_gating = true; + + pi->mg_clock_gating = true; + pi->mgcgtssm = true; + + pi->dynamic_pcie_gen2 = true; + + if (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + pi->display_gap = true; + + if (rdev->flags & RADEON_IS_MOBILITY) + pi->dcodt = true; + else + pi->dcodt = false; + + pi->ulps = true; + + pi->mclk_stutter_mode_threshold = 0; + + pi->sram_end = SMC_RAM_END; + pi->state_table_start = RV770_SMC_TABLE_ADDRESS; + pi->soft_regs_start = RV770_SMC_SOFT_REGISTERS_START; + + return 0; +} + +void rv770_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct rv7xx_ps *ps = rv770_get_ps(rps); + struct rv7xx_pl *pl; + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + if (rdev->family >= CHIP_CEDAR) { + pl = &ps->low; + printk("\t\tpower level 0 sclk: %u mclk: %u vddc: %u vddci: %u\n", + pl->sclk, pl->mclk, pl->vddc, pl->vddci); + pl = &ps->medium; + printk("\t\tpower level 1 sclk: %u mclk: %u vddc: %u vddci: %u\n", + pl->sclk, pl->mclk, pl->vddc, pl->vddci); + pl = &ps->high; + printk("\t\tpower level 2 sclk: %u mclk: %u vddc: %u vddci: %u\n", + pl->sclk, pl->mclk, pl->vddc, pl->vddci); + } else { + pl = &ps->low; + printk("\t\tpower level 0 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + pl = &ps->medium; + printk("\t\tpower level 1 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + pl = &ps->high; + printk("\t\tpower level 2 sclk: %u mclk: %u vddc: %u\n", + pl->sclk, pl->mclk, pl->vddc); + } + r600_dpm_print_ps_status(rdev, rps); +} + +void rv770_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct rv7xx_ps *ps = rv770_get_ps(rps); + struct rv7xx_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_PROFILE_INDEX_MASK) >> + CURRENT_PROFILE_INDEX_SHIFT; + + if (current_index > 2) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + if (current_index == 0) + pl = &ps->low; + else if (current_index == 1) + pl = &ps->medium; + else /* current_index == 2 */ + pl = &ps->high; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + if (rdev->family >= CHIP_CEDAR) { + seq_printf(m, "power level %d sclk: %u mclk: %u vddc: %u vddci: %u\n", + current_index, pl->sclk, pl->mclk, pl->vddc, pl->vddci); + } else { + seq_printf(m, "power level %d sclk: %u mclk: %u vddc: %u\n", + current_index, pl->sclk, pl->mclk, pl->vddc); + } + } +} + +void rv770_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +u32 rv770_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct rv7xx_ps *requested_state = rv770_get_ps(rdev->pm.dpm.requested_ps); + + if (low) + return requested_state->low.sclk; + else + return requested_state->high.sclk; +} + +u32 rv770_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct rv7xx_ps *requested_state = rv770_get_ps(rdev->pm.dpm.requested_ps); + + if (low) + return requested_state->low.mclk; + else + return requested_state->high.mclk; +} + +bool rv770_dpm_vblank_too_short(struct radeon_device *rdev) +{ + u32 vblank_time = r600_dpm_get_vblank_time(rdev); + + if (vblank_time < 300) + return true; + else + return false; + +} diff --git a/sys/dev/drm/radeon/rv770_dpm.h b/sys/dev/drm/radeon/rv770_dpm.h new file mode 100644 index 0000000000..9244effc6b --- /dev/null +++ b/sys/dev/drm/radeon/rv770_dpm.h @@ -0,0 +1,290 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __RV770_DPM_H__ +#define __RV770_DPM_H__ + +#include "rv770_smc.h" + +struct rv770_clock_registers { + u32 cg_spll_func_cntl; + u32 cg_spll_func_cntl_2; + u32 cg_spll_func_cntl_3; + u32 cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2; + u32 mpll_ad_func_cntl; + u32 mpll_ad_func_cntl_2; + u32 mpll_dq_func_cntl; + u32 mpll_dq_func_cntl_2; + u32 mclk_pwrmgt_cntl; + u32 dll_cntl; + u32 mpll_ss1; + u32 mpll_ss2; +}; + +struct rv730_clock_registers { + u32 cg_spll_func_cntl; + u32 cg_spll_func_cntl_2; + u32 cg_spll_func_cntl_3; + u32 cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2; + u32 mclk_pwrmgt_cntl; + u32 dll_cntl; + u32 mpll_func_cntl; + u32 mpll_func_cntl2; + u32 mpll_func_cntl3; + u32 mpll_ss; + u32 mpll_ss2; +}; + +union r7xx_clock_registers { + struct rv770_clock_registers rv770; + struct rv730_clock_registers rv730; +}; + +struct vddc_table_entry { + u16 vddc; + u8 vddc_index; + u8 high_smio; + u32 low_smio; +}; + +#define MAX_NO_OF_MVDD_VALUES 2 +#define MAX_NO_VREG_STEPS 32 + +struct rv7xx_power_info { + /* flags */ + bool mem_gddr5; + bool pcie_gen2; + bool dynamic_pcie_gen2; + bool acpi_pcie_gen2; + bool boot_in_gen2; + bool voltage_control; /* vddc */ + bool mvdd_control; + bool sclk_ss; + bool mclk_ss; + bool dynamic_ss; + bool gfx_clock_gating; + bool mg_clock_gating; + bool mgcgtssm; + bool power_gating; + bool thermal_protection; + bool display_gap; + bool dcodt; + bool ulps; + /* registers */ + union r7xx_clock_registers clk_regs; + u32 s0_vid_lower_smio_cntl; + /* voltage */ + u32 vddc_mask_low; + u32 mvdd_mask_low; + u32 mvdd_split_frequency; + u32 mvdd_low_smio[MAX_NO_OF_MVDD_VALUES]; + u16 max_vddc; + u16 max_vddc_in_table; + u16 min_vddc_in_table; + struct vddc_table_entry vddc_table[MAX_NO_VREG_STEPS]; + u8 valid_vddc_entries; + /* dc odt */ + u32 mclk_odt_threshold; + u8 odt_value_0[2]; + u8 odt_value_1[2]; + /* stored values */ + u32 boot_sclk; + u16 acpi_vddc; + u32 ref_div; + u32 active_auto_throttle_sources; + u32 mclk_stutter_mode_threshold; + u32 mclk_strobe_mode_threshold; + u32 mclk_edc_enable_threshold; + u32 bsp; + u32 bsu; + u32 pbsp; + u32 pbsu; + u32 dsp; + u32 psp; + u32 asi; + u32 pasi; + u32 vrc; + u32 restricted_levels; + u32 rlp; + u32 rmp; + u32 lhp; + u32 lmp; + /* smc offsets */ + u16 state_table_start; + u16 soft_regs_start; + u16 sram_end; + /* scratch structs */ + RV770_SMC_STATETABLE smc_statetable; +}; + +struct rv7xx_pl { + u32 sclk; + u32 mclk; + u16 vddc; + u16 vddci; /* eg+ only */ + u32 flags; + enum radeon_pcie_gen pcie_gen; /* si+ only */ +}; + +struct rv7xx_ps { + struct rv7xx_pl high; + struct rv7xx_pl medium; + struct rv7xx_pl low; + bool dc_compatible; +}; + +#define RV770_RLP_DFLT 10 +#define RV770_RMP_DFLT 25 +#define RV770_LHP_DFLT 25 +#define RV770_LMP_DFLT 10 +#define RV770_VRC_DFLT 0x003f +#define RV770_ASI_DFLT 1000 +#define RV770_HASI_DFLT 200000 +#define RV770_MGCGTTLOCAL0_DFLT 0x00100000 +#define RV7XX_MGCGTTLOCAL0_DFLT 0 +#define RV770_MGCGTTLOCAL1_DFLT 0xFFFF0000 +#define RV770_MGCGCGTSSMCTRL_DFLT 0x55940000 + +#define MVDD_LOW_INDEX 0 +#define MVDD_HIGH_INDEX 1 + +#define MVDD_LOW_VALUE 0 +#define MVDD_HIGH_VALUE 0xffff + +#define RV770_DEFAULT_VCLK_FREQ 53300 /* 10 khz */ +#define RV770_DEFAULT_DCLK_FREQ 40000 /* 10 khz */ + +/* rv730/rv710 */ +int rv730_populate_sclk_value(struct radeon_device *rdev, + u32 engine_clock, + RV770_SMC_SCLK_VALUE *sclk); +int rv730_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + LPRV7XX_SMC_MCLK_VALUE mclk); +void rv730_read_clock_registers(struct radeon_device *rdev); +int rv730_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table); +int rv730_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_initial_state, + RV770_SMC_STATETABLE *table); +void rv730_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_state); +void rv730_power_gating_enable(struct radeon_device *rdev, + bool enable); +void rv730_start_dpm(struct radeon_device *rdev); +void rv730_stop_dpm(struct radeon_device *rdev); +void rv730_program_dcodt(struct radeon_device *rdev, bool use_dcodt); +void rv730_get_odt_values(struct radeon_device *rdev); + +/* rv740 */ +int rv740_populate_sclk_value(struct radeon_device *rdev, u32 engine_clock, + RV770_SMC_SCLK_VALUE *sclk); +int rv740_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, u32 memory_clock, + RV7XX_SMC_MCLK_VALUE *mclk); +void rv740_read_clock_registers(struct radeon_device *rdev); +int rv740_populate_smc_acpi_state(struct radeon_device *rdev, + RV770_SMC_STATETABLE *table); +void rv740_enable_mclk_spread_spectrum(struct radeon_device *rdev, + bool enable); +u8 rv740_get_mclk_frequency_ratio(u32 memory_clock); +u32 rv740_get_dll_speed(bool is_gddr5, u32 memory_clock); +u32 rv740_get_decoded_reference_divider(u32 encoded_ref); + +/* rv770 */ +u32 rv770_map_clkf_to_ibias(struct radeon_device *rdev, u32 clkf); +int rv770_populate_vddc_value(struct radeon_device *rdev, u16 vddc, + RV770_SMC_VOLTAGE_VALUE *voltage); +int rv770_populate_mvdd_value(struct radeon_device *rdev, u32 mclk, + RV770_SMC_VOLTAGE_VALUE *voltage); +u8 rv770_get_seq_value(struct radeon_device *rdev, + struct rv7xx_pl *pl); +int rv770_populate_initial_mvdd_value(struct radeon_device *rdev, + RV770_SMC_VOLTAGE_VALUE *voltage); +u32 rv770_calculate_memory_refresh_rate(struct radeon_device *rdev, + u32 engine_clock); +void rv770_program_response_times(struct radeon_device *rdev); +int rv770_populate_smc_sp(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state); +int rv770_populate_smc_t(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + RV770_SMC_SWSTATE *smc_state); +void rv770_read_voltage_smio_registers(struct radeon_device *rdev); +void rv770_get_memory_type(struct radeon_device *rdev); +void r7xx_start_smc(struct radeon_device *rdev); +u8 rv770_get_memory_module_index(struct radeon_device *rdev); +void rv770_get_max_vddc(struct radeon_device *rdev); +void rv770_get_pcie_gen2_status(struct radeon_device *rdev); +void rv770_enable_acpi_pm(struct radeon_device *rdev); +void rv770_restore_cgcg(struct radeon_device *rdev); +bool rv770_dpm_enabled(struct radeon_device *rdev); +void rv770_enable_voltage_control(struct radeon_device *rdev, + bool enable); +void rv770_enable_backbias(struct radeon_device *rdev, + bool enable); +void rv770_enable_thermal_protection(struct radeon_device *rdev, + bool enable); +void rv770_enable_auto_throttle_source(struct radeon_device *rdev, + enum radeon_dpm_auto_throttle_src source, + bool enable); +void rv770_setup_bsp(struct radeon_device *rdev); +void rv770_program_git(struct radeon_device *rdev); +void rv770_program_tp(struct radeon_device *rdev); +void rv770_program_tpp(struct radeon_device *rdev); +void rv770_program_sstp(struct radeon_device *rdev); +void rv770_program_engine_speed_parameters(struct radeon_device *rdev); +void rv770_program_vc(struct radeon_device *rdev); +void rv770_clear_vc(struct radeon_device *rdev); +int rv770_upload_firmware(struct radeon_device *rdev); +void rv770_stop_dpm(struct radeon_device *rdev); +void r7xx_stop_smc(struct radeon_device *rdev); +void rv770_reset_smio_status(struct radeon_device *rdev); +int rv770_restrict_performance_levels_before_switch(struct radeon_device *rdev); +int rv770_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level); +int rv770_halt_smc(struct radeon_device *rdev); +int rv770_resume_smc(struct radeon_device *rdev); +int rv770_set_sw_state(struct radeon_device *rdev); +int rv770_set_boot_state(struct radeon_device *rdev); +int rv7xx_parse_power_table(struct radeon_device *rdev); +void rv770_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps); +void rv770_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_ps, + struct radeon_ps *old_ps); +void rv770_get_engine_memory_ss(struct radeon_device *rdev); + +/* smc */ +int rv770_read_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 *value); +int rv770_write_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 value); + +/* thermal */ +int rv770_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp); + +#endif diff --git a/sys/dev/drm/radeon/rv770_smc.c b/sys/dev/drm/radeon/rv770_smc.c new file mode 100644 index 0000000000..1bd2415cf2 --- /dev/null +++ b/sys/dev/drm/radeon/rv770_smc.c @@ -0,0 +1,621 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include +#include "radeon.h" +#include "rv770d.h" +#include "rv770_dpm.h" +#include "rv770_smc.h" +#include "atom.h" +#include "radeon_ucode.h" + +#define FIRST_SMC_INT_VECT_REG 0xFFD8 +#define FIRST_INT_VECT_S19 0xFFC0 + +static const u8 rv770_smc_int_vectors[] = +{ + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x0C, 0xD7, + 0x08, 0x2B, 0x08, 0x10, + 0x03, 0x51, 0x03, 0x51, + 0x03, 0x51, 0x03, 0x51 +}; + +static const u8 rv730_smc_int_vectors[] = +{ + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x08, 0x15, + 0x08, 0x15, 0x0C, 0xBB, + 0x08, 0x30, 0x08, 0x15, + 0x03, 0x56, 0x03, 0x56, + 0x03, 0x56, 0x03, 0x56 +}; + +static const u8 rv710_smc_int_vectors[] = +{ + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x08, 0x04, + 0x08, 0x04, 0x0C, 0xCB, + 0x08, 0x1F, 0x08, 0x04, + 0x03, 0x51, 0x03, 0x51, + 0x03, 0x51, 0x03, 0x51 +}; + +static const u8 rv740_smc_int_vectors[] = +{ + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x08, 0x10, + 0x08, 0x10, 0x0C, 0xD7, + 0x08, 0x2B, 0x08, 0x10, + 0x03, 0x51, 0x03, 0x51, + 0x03, 0x51, 0x03, 0x51 +}; + +static const u8 cedar_smc_int_vectors[] = +{ + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x11, 0x8B, + 0x0B, 0x20, 0x0B, 0x05, + 0x04, 0xF6, 0x04, 0xF6, + 0x04, 0xF6, 0x04, 0xF6 +}; + +static const u8 redwood_smc_int_vectors[] = +{ + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x11, 0x8B, + 0x0B, 0x20, 0x0B, 0x05, + 0x04, 0xF6, 0x04, 0xF6, + 0x04, 0xF6, 0x04, 0xF6 +}; + +static const u8 juniper_smc_int_vectors[] = +{ + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x11, 0x8B, + 0x0B, 0x20, 0x0B, 0x05, + 0x04, 0xF6, 0x04, 0xF6, + 0x04, 0xF6, 0x04, 0xF6 +}; + +static const u8 cypress_smc_int_vectors[] = +{ + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x0B, 0x05, + 0x0B, 0x05, 0x11, 0x8B, + 0x0B, 0x20, 0x0B, 0x05, + 0x04, 0xF6, 0x04, 0xF6, + 0x04, 0xF6, 0x04, 0xF6 +}; + +static const u8 barts_smc_int_vectors[] = +{ + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x12, 0xAA, + 0x0C, 0x2F, 0x15, 0xF6, + 0x15, 0xF6, 0x05, 0x0A, + 0x05, 0x0A, 0x05, 0x0A +}; + +static const u8 turks_smc_int_vectors[] = +{ + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x12, 0xAA, + 0x0C, 0x2F, 0x15, 0xF6, + 0x15, 0xF6, 0x05, 0x0A, + 0x05, 0x0A, 0x05, 0x0A +}; + +static const u8 caicos_smc_int_vectors[] = +{ + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x0C, 0x14, + 0x0C, 0x14, 0x12, 0xAA, + 0x0C, 0x2F, 0x15, 0xF6, + 0x15, 0xF6, 0x05, 0x0A, + 0x05, 0x0A, 0x05, 0x0A +}; + +static const u8 cayman_smc_int_vectors[] = +{ + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x12, 0x05, + 0x12, 0x05, 0x18, 0xEA, + 0x12, 0x20, 0x1C, 0x34, + 0x1C, 0x34, 0x08, 0x72, + 0x08, 0x72, 0x08, 0x72 +}; + +int rv770_set_smc_sram_address(struct radeon_device *rdev, + u16 smc_address, u16 limit) +{ + u32 addr; + + if (smc_address & 3) + return -EINVAL; + if ((smc_address + 3) > limit) + return -EINVAL; + + addr = smc_address; + addr |= SMC_SRAM_AUTO_INC_DIS; + + WREG32(SMC_SRAM_ADDR, addr); + + return 0; +} + +int rv770_copy_bytes_to_smc(struct radeon_device *rdev, + u16 smc_start_address, const u8 *src, + u16 byte_count, u16 limit) +{ + u32 data, original_data, extra_shift; + u16 addr; + int ret; + + if (smc_start_address & 3) + return -EINVAL; + if ((smc_start_address + byte_count) > limit) + return -EINVAL; + + addr = smc_start_address; + + while (byte_count >= 4) { + /* SMC address space is BE */ + data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3]; + + ret = rv770_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + WREG32(SMC_SRAM_DATA, data); + + src += 4; + byte_count -= 4; + addr += 4; + } + + /* RMW for final bytes */ + if (byte_count > 0) { + data = 0; + + ret = rv770_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + original_data = RREG32(SMC_SRAM_DATA); + + extra_shift = 8 * (4 - byte_count); + + while (byte_count > 0) { + /* SMC address space is BE */ + data = (data << 8) + *src++; + byte_count--; + } + + data <<= extra_shift; + + data |= (original_data & ~((~0UL) << extra_shift)); + + ret = rv770_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + WREG32(SMC_SRAM_DATA, data); + } + + return 0; +} + +static int rv770_program_interrupt_vectors(struct radeon_device *rdev, + u32 smc_first_vector, const u8 *src, + u32 byte_count) +{ + u32 tmp, i; + + if (byte_count % 4) + return -EINVAL; + + if (smc_first_vector < FIRST_SMC_INT_VECT_REG) { + tmp = FIRST_SMC_INT_VECT_REG - smc_first_vector; + + if (tmp > byte_count) + return 0; + + byte_count -= tmp; + src += tmp; + smc_first_vector = FIRST_SMC_INT_VECT_REG; + } + + for (i = 0; i < byte_count; i += 4) { + /* SMC address space is BE */ + tmp = (src[i] << 24) | (src[i + 1] << 16) | (src[i + 2] << 8) | src[i + 3]; + + WREG32(SMC_ISR_FFD8_FFDB + i, tmp); + } + + return 0; +} + +void rv770_start_smc(struct radeon_device *rdev) +{ + WREG32_P(SMC_IO, SMC_RST_N, ~SMC_RST_N); +} + +void rv770_reset_smc(struct radeon_device *rdev) +{ + WREG32_P(SMC_IO, 0, ~SMC_RST_N); +} + +void rv770_stop_smc_clock(struct radeon_device *rdev) +{ + WREG32_P(SMC_IO, 0, ~SMC_CLK_EN); +} + +void rv770_start_smc_clock(struct radeon_device *rdev) +{ + WREG32_P(SMC_IO, SMC_CLK_EN, ~SMC_CLK_EN); +} + +bool rv770_is_smc_running(struct radeon_device *rdev) +{ + u32 tmp; + + tmp = RREG32(SMC_IO); + + if ((tmp & SMC_RST_N) && (tmp & SMC_CLK_EN)) + return true; + else + return false; +} + +PPSMC_Result rv770_send_msg_to_smc(struct radeon_device *rdev, PPSMC_Msg msg) +{ + u32 tmp; + int i; + PPSMC_Result result; + + if (!rv770_is_smc_running(rdev)) + return PPSMC_Result_Failed; + + WREG32_P(SMC_MSG, HOST_SMC_MSG(msg), ~HOST_SMC_MSG_MASK); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(SMC_MSG) & HOST_SMC_RESP_MASK; + tmp >>= HOST_SMC_RESP_SHIFT; + if (tmp != 0) + break; + DRM_UDELAY(1); + } + + tmp = RREG32(SMC_MSG) & HOST_SMC_RESP_MASK; + tmp >>= HOST_SMC_RESP_SHIFT; + + result = (PPSMC_Result)tmp; + return result; +} + +PPSMC_Result rv770_wait_for_smc_inactive(struct radeon_device *rdev) +{ + int i; + PPSMC_Result result = PPSMC_Result_OK; + + if (!rv770_is_smc_running(rdev)) + return result; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(SMC_IO) & SMC_STOP_MODE) + break; + DRM_UDELAY(1); + } + + return result; +} + +static void rv770_clear_smc_sram(struct radeon_device *rdev, u16 limit) +{ + u16 i; + + for (i = 0; i < limit; i += 4) { + rv770_set_smc_sram_address(rdev, i, limit); + WREG32(SMC_SRAM_DATA, 0); + } +} + +int rv770_load_smc_ucode(struct radeon_device *rdev, + u16 limit) +{ + int ret; + const u8 *int_vect; + u16 int_vect_start_address; + u16 int_vect_size; + const u8 *ucode_data; + u16 ucode_start_address; + u16 ucode_size; + + if (!rdev->smc_fw) + return -EINVAL; + + rv770_clear_smc_sram(rdev, limit); + + switch (rdev->family) { + case CHIP_RV770: + ucode_start_address = RV770_SMC_UCODE_START; + ucode_size = RV770_SMC_UCODE_SIZE; + int_vect = (const u8 *)&rv770_smc_int_vectors; + int_vect_start_address = RV770_SMC_INT_VECTOR_START; + int_vect_size = RV770_SMC_INT_VECTOR_SIZE; + break; + case CHIP_RV730: + ucode_start_address = RV730_SMC_UCODE_START; + ucode_size = RV730_SMC_UCODE_SIZE; + int_vect = (const u8 *)&rv730_smc_int_vectors; + int_vect_start_address = RV730_SMC_INT_VECTOR_START; + int_vect_size = RV730_SMC_INT_VECTOR_SIZE; + break; + case CHIP_RV710: + ucode_start_address = RV710_SMC_UCODE_START; + ucode_size = RV710_SMC_UCODE_SIZE; + int_vect = (const u8 *)&rv710_smc_int_vectors; + int_vect_start_address = RV710_SMC_INT_VECTOR_START; + int_vect_size = RV710_SMC_INT_VECTOR_SIZE; + break; + case CHIP_RV740: + ucode_start_address = RV740_SMC_UCODE_START; + ucode_size = RV740_SMC_UCODE_SIZE; + int_vect = (const u8 *)&rv740_smc_int_vectors; + int_vect_start_address = RV740_SMC_INT_VECTOR_START; + int_vect_size = RV740_SMC_INT_VECTOR_SIZE; + break; + case CHIP_CEDAR: + ucode_start_address = CEDAR_SMC_UCODE_START; + ucode_size = CEDAR_SMC_UCODE_SIZE; + int_vect = (const u8 *)&cedar_smc_int_vectors; + int_vect_start_address = CEDAR_SMC_INT_VECTOR_START; + int_vect_size = CEDAR_SMC_INT_VECTOR_SIZE; + break; + case CHIP_REDWOOD: + ucode_start_address = REDWOOD_SMC_UCODE_START; + ucode_size = REDWOOD_SMC_UCODE_SIZE; + int_vect = (const u8 *)&redwood_smc_int_vectors; + int_vect_start_address = REDWOOD_SMC_INT_VECTOR_START; + int_vect_size = REDWOOD_SMC_INT_VECTOR_SIZE; + break; + case CHIP_JUNIPER: + ucode_start_address = JUNIPER_SMC_UCODE_START; + ucode_size = JUNIPER_SMC_UCODE_SIZE; + int_vect = (const u8 *)&juniper_smc_int_vectors; + int_vect_start_address = JUNIPER_SMC_INT_VECTOR_START; + int_vect_size = JUNIPER_SMC_INT_VECTOR_SIZE; + break; + case CHIP_CYPRESS: + case CHIP_HEMLOCK: + ucode_start_address = CYPRESS_SMC_UCODE_START; + ucode_size = CYPRESS_SMC_UCODE_SIZE; + int_vect = (const u8 *)&cypress_smc_int_vectors; + int_vect_start_address = CYPRESS_SMC_INT_VECTOR_START; + int_vect_size = CYPRESS_SMC_INT_VECTOR_SIZE; + break; + case CHIP_BARTS: + ucode_start_address = BARTS_SMC_UCODE_START; + ucode_size = BARTS_SMC_UCODE_SIZE; + int_vect = (const u8 *)&barts_smc_int_vectors; + int_vect_start_address = BARTS_SMC_INT_VECTOR_START; + int_vect_size = BARTS_SMC_INT_VECTOR_SIZE; + break; + case CHIP_TURKS: + ucode_start_address = TURKS_SMC_UCODE_START; + ucode_size = TURKS_SMC_UCODE_SIZE; + int_vect = (const u8 *)&turks_smc_int_vectors; + int_vect_start_address = TURKS_SMC_INT_VECTOR_START; + int_vect_size = TURKS_SMC_INT_VECTOR_SIZE; + break; + case CHIP_CAICOS: + ucode_start_address = CAICOS_SMC_UCODE_START; + ucode_size = CAICOS_SMC_UCODE_SIZE; + int_vect = (const u8 *)&caicos_smc_int_vectors; + int_vect_start_address = CAICOS_SMC_INT_VECTOR_START; + int_vect_size = CAICOS_SMC_INT_VECTOR_SIZE; + break; + case CHIP_CAYMAN: + ucode_start_address = CAYMAN_SMC_UCODE_START; + ucode_size = CAYMAN_SMC_UCODE_SIZE; + int_vect = (const u8 *)&cayman_smc_int_vectors; + int_vect_start_address = CAYMAN_SMC_INT_VECTOR_START; + int_vect_size = CAYMAN_SMC_INT_VECTOR_SIZE; + break; + default: + DRM_ERROR("unknown asic in smc ucode loader\n"); + BUG(); + } + + /* load the ucode */ + ucode_data = (const u8 *)rdev->smc_fw->data; + ret = rv770_copy_bytes_to_smc(rdev, ucode_start_address, + ucode_data, ucode_size, limit); + if (ret) + return ret; + + /* set up the int vectors */ + ret = rv770_program_interrupt_vectors(rdev, int_vect_start_address, + int_vect, int_vect_size); + if (ret) + return ret; + + return 0; +} + +int rv770_read_smc_sram_dword(struct radeon_device *rdev, + u16 smc_address, u32 *value, u16 limit) +{ + int ret; + + ret = rv770_set_smc_sram_address(rdev, smc_address, limit); + if (ret) + return ret; + + *value = RREG32(SMC_SRAM_DATA); + + return 0; +} + +int rv770_write_smc_sram_dword(struct radeon_device *rdev, + u16 smc_address, u32 value, u16 limit) +{ + int ret; + + ret = rv770_set_smc_sram_address(rdev, smc_address, limit); + if (ret) + return ret; + + WREG32(SMC_SRAM_DATA, value); + + return 0; +} diff --git a/sys/dev/drm/radeon/rv770_smc.h b/sys/dev/drm/radeon/rv770_smc.h new file mode 100644 index 0000000000..f78d92a4b3 --- /dev/null +++ b/sys/dev/drm/radeon/rv770_smc.h @@ -0,0 +1,209 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __RV770_SMC_H__ +#define __RV770_SMC_H__ + +#include "ppsmc.h" + +#pragma pack(push, 1) + +#define RV770_SMC_TABLE_ADDRESS 0xB000 + +#define RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE 3 + +struct RV770_SMC_SCLK_VALUE +{ + uint32_t vCG_SPLL_FUNC_CNTL; + uint32_t vCG_SPLL_FUNC_CNTL_2; + uint32_t vCG_SPLL_FUNC_CNTL_3; + uint32_t vCG_SPLL_SPREAD_SPECTRUM; + uint32_t vCG_SPLL_SPREAD_SPECTRUM_2; + uint32_t sclk_value; +}; + +typedef struct RV770_SMC_SCLK_VALUE RV770_SMC_SCLK_VALUE; + +struct RV770_SMC_MCLK_VALUE +{ + uint32_t vMPLL_AD_FUNC_CNTL; + uint32_t vMPLL_AD_FUNC_CNTL_2; + uint32_t vMPLL_DQ_FUNC_CNTL; + uint32_t vMPLL_DQ_FUNC_CNTL_2; + uint32_t vMCLK_PWRMGT_CNTL; + uint32_t vDLL_CNTL; + uint32_t vMPLL_SS; + uint32_t vMPLL_SS2; + uint32_t mclk_value; +}; + +typedef struct RV770_SMC_MCLK_VALUE RV770_SMC_MCLK_VALUE; + + +struct RV730_SMC_MCLK_VALUE +{ + uint32_t vMCLK_PWRMGT_CNTL; + uint32_t vDLL_CNTL; + uint32_t vMPLL_FUNC_CNTL; + uint32_t vMPLL_FUNC_CNTL2; + uint32_t vMPLL_FUNC_CNTL3; + uint32_t vMPLL_SS; + uint32_t vMPLL_SS2; + uint32_t mclk_value; +}; + +typedef struct RV730_SMC_MCLK_VALUE RV730_SMC_MCLK_VALUE; + +struct RV770_SMC_VOLTAGE_VALUE +{ + uint16_t value; + uint8_t index; + uint8_t padding; +}; + +typedef struct RV770_SMC_VOLTAGE_VALUE RV770_SMC_VOLTAGE_VALUE; + +union RV7XX_SMC_MCLK_VALUE +{ + RV770_SMC_MCLK_VALUE mclk770; + RV730_SMC_MCLK_VALUE mclk730; +}; + +typedef union RV7XX_SMC_MCLK_VALUE RV7XX_SMC_MCLK_VALUE, *LPRV7XX_SMC_MCLK_VALUE; + +struct RV770_SMC_HW_PERFORMANCE_LEVEL +{ + uint8_t arbValue; + union{ + uint8_t seqValue; + uint8_t ACIndex; + }; + uint8_t displayWatermark; + uint8_t gen2PCIE; + uint8_t gen2XSP; + uint8_t backbias; + uint8_t strobeMode; + uint8_t mcFlags; + uint32_t aT; + uint32_t bSP; + RV770_SMC_SCLK_VALUE sclk; + RV7XX_SMC_MCLK_VALUE mclk; + RV770_SMC_VOLTAGE_VALUE vddc; + RV770_SMC_VOLTAGE_VALUE mvdd; + RV770_SMC_VOLTAGE_VALUE vddci; + uint8_t reserved1; + uint8_t reserved2; + uint8_t stateFlags; + uint8_t padding; +}; + +#define SMC_STROBE_RATIO 0x0F +#define SMC_STROBE_ENABLE 0x10 + +#define SMC_MC_EDC_RD_FLAG 0x01 +#define SMC_MC_EDC_WR_FLAG 0x02 +#define SMC_MC_RTT_ENABLE 0x04 +#define SMC_MC_STUTTER_EN 0x08 + +typedef struct RV770_SMC_HW_PERFORMANCE_LEVEL RV770_SMC_HW_PERFORMANCE_LEVEL; + +struct RV770_SMC_SWSTATE +{ + uint8_t flags; + uint8_t padding1; + uint8_t padding2; + uint8_t padding3; + RV770_SMC_HW_PERFORMANCE_LEVEL levels[RV770_SMC_PERFORMANCE_LEVELS_PER_SWSTATE]; +}; + +typedef struct RV770_SMC_SWSTATE RV770_SMC_SWSTATE; + +#define RV770_SMC_VOLTAGEMASK_VDDC 0 +#define RV770_SMC_VOLTAGEMASK_MVDD 1 +#define RV770_SMC_VOLTAGEMASK_VDDCI 2 +#define RV770_SMC_VOLTAGEMASK_MAX 4 + +struct RV770_SMC_VOLTAGEMASKTABLE +{ + uint8_t highMask[RV770_SMC_VOLTAGEMASK_MAX]; + uint32_t lowMask[RV770_SMC_VOLTAGEMASK_MAX]; +}; + +typedef struct RV770_SMC_VOLTAGEMASKTABLE RV770_SMC_VOLTAGEMASKTABLE; + +#define MAX_NO_VREG_STEPS 32 + +struct RV770_SMC_STATETABLE +{ + uint8_t thermalProtectType; + uint8_t systemFlags; + uint8_t maxVDDCIndexInPPTable; + uint8_t extraFlags; + uint8_t highSMIO[MAX_NO_VREG_STEPS]; + uint32_t lowSMIO[MAX_NO_VREG_STEPS]; + RV770_SMC_VOLTAGEMASKTABLE voltageMaskTable; + RV770_SMC_SWSTATE initialState; + RV770_SMC_SWSTATE ACPIState; + RV770_SMC_SWSTATE driverState; + RV770_SMC_SWSTATE ULVState; +}; + +typedef struct RV770_SMC_STATETABLE RV770_SMC_STATETABLE; + +#define PPSMC_STATEFLAG_AUTO_PULSE_SKIP 0x01 + +#pragma pack(pop) + +#define RV770_SMC_SOFT_REGISTERS_START 0x104 + +#define RV770_SMC_SOFT_REGISTER_mclk_chg_timeout 0x0 +#define RV770_SMC_SOFT_REGISTER_baby_step_timer 0x8 +#define RV770_SMC_SOFT_REGISTER_delay_bbias 0xC +#define RV770_SMC_SOFT_REGISTER_delay_vreg 0x10 +#define RV770_SMC_SOFT_REGISTER_delay_acpi 0x2C +#define RV770_SMC_SOFT_REGISTER_seq_index 0x64 +#define RV770_SMC_SOFT_REGISTER_mvdd_chg_time 0x68 +#define RV770_SMC_SOFT_REGISTER_mclk_switch_lim 0x78 +#define RV770_SMC_SOFT_REGISTER_mc_block_delay 0x90 +#define RV770_SMC_SOFT_REGISTER_uvd_enabled 0x9C +#define RV770_SMC_SOFT_REGISTER_is_asic_lombok 0xA0 + +int rv770_set_smc_sram_address(struct radeon_device *rdev, + u16 smc_address, u16 limit); +int rv770_copy_bytes_to_smc(struct radeon_device *rdev, + u16 smc_start_address, const u8 *src, + u16 byte_count, u16 limit); +void rv770_start_smc(struct radeon_device *rdev); +void rv770_reset_smc(struct radeon_device *rdev); +void rv770_stop_smc_clock(struct radeon_device *rdev); +void rv770_start_smc_clock(struct radeon_device *rdev); +bool rv770_is_smc_running(struct radeon_device *rdev); +PPSMC_Result rv770_send_msg_to_smc(struct radeon_device *rdev, PPSMC_Msg msg); +PPSMC_Result rv770_wait_for_smc_inactive(struct radeon_device *rdev); +int rv770_read_smc_sram_dword(struct radeon_device *rdev, + u16 smc_address, u32 *value, u16 limit); +int rv770_write_smc_sram_dword(struct radeon_device *rdev, + u16 smc_address, u32 value, u16 limit); +int rv770_load_smc_ucode(struct radeon_device *rdev, + u16 limit); + +#endif diff --git a/sys/dev/drm/radeon/rv770d.h b/sys/dev/drm/radeon/rv770d.h index e71adba805..6bef2b7d60 100644 --- a/sys/dev/drm/radeon/rv770d.h +++ b/sys/dev/drm/radeon/rv770d.h @@ -23,8 +23,6 @@ * Authors: Dave Airlie * Alex Deucher * Jerome Glisse - * - * $FreeBSD: head/sys/dev/drm2/radeon/rv770d.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef RV770_H #define RV770_H @@ -64,6 +62,246 @@ # define UPLL_FB_DIV(x) ((x) << 0) # define UPLL_FB_DIV_MASK 0x01FFFFFF +/* pm registers */ +#define SMC_SRAM_ADDR 0x200 +#define SMC_SRAM_AUTO_INC_DIS (1 << 16) +#define SMC_SRAM_DATA 0x204 +#define SMC_IO 0x208 +#define SMC_RST_N (1 << 0) +#define SMC_STOP_MODE (1 << 2) +#define SMC_CLK_EN (1 << 11) +#define SMC_MSG 0x20c +#define HOST_SMC_MSG(x) ((x) << 0) +#define HOST_SMC_MSG_MASK (0xff << 0) +#define HOST_SMC_MSG_SHIFT 0 +#define HOST_SMC_RESP(x) ((x) << 8) +#define HOST_SMC_RESP_MASK (0xff << 8) +#define HOST_SMC_RESP_SHIFT 8 +#define SMC_HOST_MSG(x) ((x) << 16) +#define SMC_HOST_MSG_MASK (0xff << 16) +#define SMC_HOST_MSG_SHIFT 16 +#define SMC_HOST_RESP(x) ((x) << 24) +#define SMC_HOST_RESP_MASK (0xff << 24) +#define SMC_HOST_RESP_SHIFT 24 + +#define SMC_ISR_FFD8_FFDB 0x218 + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_DIVEN (1 << 2) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_HILEN(x) ((x) << 12) +#define SPLL_HILEN_MASK (0xf << 12) +#define SPLL_LOLEN(x) ((x) << 16) +#define SPLL_LOLEN_MASK (0xf << 16) +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_DITHEN (1 << 28) + +#define SPLL_CNTL_MODE 0x610 +#define SPLL_DIV_SYNC (1 << 5) + +#define MPLL_AD_FUNC_CNTL 0x624 +#define CLKF(x) ((x) << 0) +#define CLKF_MASK (0x7f << 0) +#define CLKR(x) ((x) << 7) +#define CLKR_MASK (0x1f << 7) +#define CLKFRAC(x) ((x) << 12) +#define CLKFRAC_MASK (0x1f << 12) +#define YCLK_POST_DIV(x) ((x) << 17) +#define YCLK_POST_DIV_MASK (3 << 17) +#define IBIAS(x) ((x) << 20) +#define IBIAS_MASK (0x3ff << 20) +#define RESET (1 << 30) +#define PDNB (1 << 31) +#define MPLL_AD_FUNC_CNTL_2 0x628 +#define BYPASS (1 << 19) +#define BIAS_GEN_PDNB (1 << 24) +#define RESET_EN (1 << 25) +#define VCO_MODE (1 << 29) +#define MPLL_DQ_FUNC_CNTL 0x62c +#define MPLL_DQ_FUNC_CNTL_2 0x630 + +#define GENERAL_PWRMGT 0x63c +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define ENABLE_GEN2PCIE (1 << 4) +# define ENABLE_GEN2XSP (1 << 5) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (3 << 6) +# define SW_SMIO_INDEX_SHIFT 6 +# define LOW_VOLT_D2_ACPI (1 << 8) +# define LOW_VOLT_D3_ACPI (1 << 9) +# define VOLT_PWRMGT_EN (1 << 10) +# define BACKBIAS_PAD_EN (1 << 18) +# define BACKBIAS_VALUE (1 << 19) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +# define AC_DC_SW (1 << 24) + +#define CG_TPC 0x640 +#define SCLK_PWRMGT_CNTL 0x644 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) +#define MCLK_PWRMGT_CNTL 0x648 +# define DLL_SPEED(x) ((x) << 0) +# define DLL_SPEED_MASK (0x1f << 0) +# define MPLL_PWRMGT_OFF (1 << 5) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCKA0_SLEEP (1 << 8) +# define MRDCKA1_SLEEP (1 << 9) +# define MRDCKB0_SLEEP (1 << 10) +# define MRDCKB1_SLEEP (1 << 11) +# define MRDCKC0_SLEEP (1 << 12) +# define MRDCKC1_SLEEP (1 << 13) +# define MRDCKD0_SLEEP (1 << 14) +# define MRDCKD1_SLEEP (1 << 15) +# define MRDCKA0_RESET (1 << 16) +# define MRDCKA1_RESET (1 << 17) +# define MRDCKB0_RESET (1 << 18) +# define MRDCKB1_RESET (1 << 19) +# define MRDCKC0_RESET (1 << 20) +# define MRDCKC1_RESET (1 << 21) +# define MRDCKD0_RESET (1 << 22) +# define MRDCKD1_RESET (1 << 23) +# define DLL_READY_READ (1 << 24) +# define USE_DISPLAY_GAP (1 << 25) +# define USE_DISPLAY_URGENT_NORMAL (1 << 26) +# define MPLL_TURNOFF_D2 (1 << 28) +#define DLL_CNTL 0x64c +# define MRDCKA0_BYPASS (1 << 24) +# define MRDCKA1_BYPASS (1 << 25) +# define MRDCKB0_BYPASS (1 << 26) +# define MRDCKB1_BYPASS (1 << 27) +# define MRDCKC0_BYPASS (1 << 28) +# define MRDCKC1_BYPASS (1 << 29) +# define MRDCKD0_BYPASS (1 << 30) +# define MRDCKD1_BYPASS (1 << 31) + +#define MPLL_TIME 0x654 +# define MPLL_LOCK_TIME(x) ((x) << 0) +# define MPLL_LOCK_TIME_MASK (0xffff << 0) +# define MPLL_RESET_TIME(x) ((x) << 16) +# define MPLL_RESET_TIME_MASK (0xffff << 16) + +#define CG_CLKPIN_CNTL 0x660 +# define MUX_TCLK_TO_XCLK (1 << 8) +# define XTALIN_DIVIDE (1 << 9) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x66c +# define CURRENT_PROFILE_INDEX_MASK (0xf << 4) +# define CURRENT_PROFILE_INDEX_SHIFT 4 + +#define S0_VID_LOWER_SMIO_CNTL 0x678 +#define S1_VID_LOWER_SMIO_CNTL 0x67c +#define S2_VID_LOWER_SMIO_CNTL 0x680 +#define S3_VID_LOWER_SMIO_CNTL 0x684 + +#define CG_FTV 0x690 +#define CG_FFCT_0 0x694 +# define UTC_0(x) ((x) << 0) +# define UTC_0_MASK (0x3ff << 0) +# define DTC_0(x) ((x) << 10) +# define DTC_0_MASK (0x3ff << 10) + +#define CG_BSP 0x6d0 +# define BSP(x) ((x) << 0) +# define BSP_MASK (0xffff << 0) +# define BSU(x) ((x) << 16) +# define BSU_MASK (0xf << 16) +#define CG_AT 0x6d4 +# define CG_R(x) ((x) << 0) +# define CG_R_MASK (0xffff << 0) +# define CG_L(x) ((x) << 16) +# define CG_L_MASK (0xffff << 16) +#define CG_GIT 0x6d8 +# define CG_GICST(x) ((x) << 0) +# define CG_GICST_MASK (0xffff << 0) +# define CG_GIPOT(x) ((x) << 16) +# define CG_GIPOT_MASK (0xffff << 16) + +#define CG_SSP 0x6e8 +# define SST(x) ((x) << 0) +# define SST_MASK (0xffff << 0) +# define SSTU(x) ((x) << 16) +# define SSTU_MASK (0xf << 16) + +#define CG_DISPLAY_GAP_CNTL 0x714 +# define DISP1_GAP(x) ((x) << 0) +# define DISP1_GAP_MASK (3 << 0) +# define DISP2_GAP(x) ((x) << 2) +# define DISP2_GAP_MASK (3 << 2) +# define VBI_TIMER_COUNT(x) ((x) << 4) +# define VBI_TIMER_COUNT_MASK (0x3fff << 4) +# define VBI_TIMER_UNIT(x) ((x) << 20) +# define VBI_TIMER_UNIT_MASK (7 << 20) +# define DISP1_GAP_MCHG(x) ((x) << 24) +# define DISP1_GAP_MCHG_MASK (3 << 24) +# define DISP2_GAP_MCHG(x) ((x) << 26) +# define DISP2_GAP_MCHG_MASK (3 << 26) + +#define CG_SPLL_SPREAD_SPECTRUM 0x790 +#define SSEN (1 << 0) +#define CLKS(x) ((x) << 4) +#define CLKS_MASK (0xfff << 4) +#define CG_SPLL_SPREAD_SPECTRUM_2 0x794 +#define CLKV(x) ((x) << 0) +#define CLKV_MASK (0x3ffffff << 0) +#define CG_MPLL_SPREAD_SPECTRUM 0x798 +#define CG_UPLL_SPREAD_SPECTRUM 0x79c +# define SSEN_MASK 0x00000001 + +#define CG_CGTT_LOCAL_0 0x7d0 +#define CG_CGTT_LOCAL_1 0x7d4 + +#define BIOS_SCRATCH_4 0x1734 + +#define MC_SEQ_MISC0 0x2a00 +#define MC_SEQ_MISC0_GDDR5_SHIFT 28 +#define MC_SEQ_MISC0_GDDR5_MASK 0xf0000000 +#define MC_SEQ_MISC0_GDDR5_VALUE 5 + +#define MC_ARB_SQM_RATIO 0x2770 +#define STATE0(x) ((x) << 0) +#define STATE0_MASK (0xff << 0) +#define STATE1(x) ((x) << 8) +#define STATE1_MASK (0xff << 8) +#define STATE2(x) ((x) << 16) +#define STATE2_MASK (0xff << 16) +#define STATE3(x) ((x) << 24) +#define STATE3_MASK (0xff << 24) + +#define MC_ARB_RFSH_RATE 0x27b0 +#define POWERMODE0(x) ((x) << 0) +#define POWERMODE0_MASK (0xff << 0) +#define POWERMODE1(x) ((x) << 8) +#define POWERMODE1_MASK (0xff << 8) +#define POWERMODE2(x) ((x) << 16) +#define POWERMODE2_MASK (0xff << 16) +#define POWERMODE3(x) ((x) << 24) +#define POWERMODE3_MASK (0xff << 24) + +#define CGTS_SM_CTRL_REG 0x9150 + /* Registers */ #define CB_COLOR0_BASE 0x28040 #define CB_COLOR1_BASE 0x28044 @@ -88,8 +326,8 @@ #define CONFIG_MEMSIZE 0x5428 #define CP_ME_CNTL 0x86D8 -#define CP_ME_HALT (1<<28) -#define CP_PFP_HALT (1<<26) +#define CP_ME_HALT (1 << 28) +#define CP_PFP_HALT (1 << 26) #define CP_ME_RAM_DATA 0xC160 #define CP_ME_RAM_RADDR 0xC158 #define CP_ME_RAM_WADDR 0xC15C @@ -159,9 +397,22 @@ #define GUI_ACTIVE (1<<31) #define GRBM_STATUS2 0x8014 -#define CG_CLKPIN_CNTL 0x660 -# define MUX_TCLK_TO_XCLK (1 << 8) -# define XTALIN_DIVIDE (1 << 9) +#define CG_THERMAL_CTRL 0x72C +#define DPM_EVENT_SRC(x) ((x) << 0) +#define DPM_EVENT_SRC_MASK (7 << 0) +#define DIG_THERM_DPM(x) ((x) << 14) +#define DIG_THERM_DPM_MASK 0x003FC000 +#define DIG_THERM_DPM_SHIFT 14 + +#define CG_THERMAL_INT 0x734 +#define DIG_THERM_INTH(x) ((x) << 8) +#define DIG_THERM_INTH_MASK 0x0000FF00 +#define DIG_THERM_INTH_SHIFT 8 +#define DIG_THERM_INTL(x) ((x) << 16) +#define DIG_THERM_INTL_MASK 0x00FF0000 +#define DIG_THERM_INTL_SHIFT 16 +#define THERM_INT_MASK_HIGH (1 << 24) +#define THERM_INT_MASK_LOW (1 << 25) #define CG_MULT_THERMAL_STATUS 0x740 #define ASIC_T(x) ((x) << 16) @@ -664,7 +915,22 @@ #define D1GRPH_SECONDARY_SURFACE_ADDRESS_HIGH 0x691c #define D2GRPH_SECONDARY_SURFACE_ADDRESS_HIGH 0x611c -/* PCIE link stuff */ +/* PCIE indirect regs */ +#define PCIE_P_CNTL 0x40 +# define P_PLL_PWRDN_IN_L1L23 (1 << 3) +# define P_PLL_BUF_PDNB (1 << 4) +# define P_PLL_PDNB (1 << 9) +# define P_ALLOW_PRX_FRONTEND_SHUTOFF (1 << 12) +/* PCIE PORT regs */ +#define PCIE_LC_CNTL 0xa0 +# define LC_L0S_INACTIVITY(x) ((x) << 8) +# define LC_L0S_INACTIVITY_MASK (0xf << 8) +# define LC_L0S_INACTIVITY_SHIFT 8 +# define LC_L1_INACTIVITY(x) ((x) << 12) +# define LC_L1_INACTIVITY_MASK (0xf << 12) +# define LC_L1_INACTIVITY_SHIFT 12 +# define LC_PMI_TO_L1_DIS (1 << 16) +# define LC_ASPM_TO_L1_DIS (1 << 24) #define PCIE_LC_TRAINING_CNTL 0xa1 /* PCIE_P */ #define PCIE_LC_LINK_WIDTH_CNTL 0xa2 /* PCIE_P */ # define LC_LINK_WIDTH_SHIFT 0 @@ -692,6 +958,9 @@ # define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK (0x3 << 8) # define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT 3 # define LC_CURRENT_DATA_RATE (1 << 11) +# define LC_HW_VOLTAGE_IF_CONTROL(x) ((x) << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_MASK (3 << 12) +# define LC_HW_VOLTAGE_IF_CONTROL_SHIFT 12 # define LC_VOLTAGE_TIMER_SEL_MASK (0xf << 14) # define LC_CLR_FAILED_SPD_CHANGE_CNT (1 << 21) # define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 23) diff --git a/sys/dev/drm/radeon/si.c b/sys/dev/drm/radeon/si.c index df9196b407..7545ab9dee 100644 --- a/sys/dev/drm/radeon/si.c +++ b/sys/dev/drm/radeon/si.c @@ -25,22 +25,278 @@ */ #include +#include #include "radeon.h" #include "radeon_asic.h" #include #include "sid.h" #include "atom.h" #include "si_blit_shaders.h" - -#define SI_PFP_UCODE_SIZE 2144 -#define SI_PM4_UCODE_SIZE 2144 -#define SI_CE_UCODE_SIZE 2144 -#define SI_RLC_UCODE_SIZE 2048 -#define SI_MC_UCODE_SIZE 7769 -#define OLAND_MC_UCODE_SIZE 7863 - +#include "clearstate_si.h" +#include "radeon_ucode.h" + + +#define PCI_EXP_LNKCTL PCIER_LINKCTRL /* 16 */ +#define PCI_EXP_LNKCTL2 48 +#define PCI_EXP_LNKCTL_HAWD PCIEM_LNKCTL_HAWD /* 0x0200 */ +#define PCI_EXP_DEVSTA PCIER_DEVSTS /* 10 */ +#define PCI_EXP_DEVSTA_TRPND 0x0020 +#define PCI_EXP_LNKCAP_CLKPM 0x00040000 + +MODULE_FIRMWARE("radeon/TAHITI_ce.bin"); +MODULE_FIRMWARE("radeon/TAHITI_mc.bin"); +MODULE_FIRMWARE("radeon/TAHITI_rlc.bin"); +MODULE_FIRMWARE("radeon/TAHITI_smc.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_pfp.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_me.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_ce.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_mc.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_rlc.bin"); +MODULE_FIRMWARE("radeon/PITCAIRN_smc.bin"); +MODULE_FIRMWARE("radeon/VERDE_pfp.bin"); +MODULE_FIRMWARE("radeon/VERDE_me.bin"); +MODULE_FIRMWARE("radeon/VERDE_ce.bin"); +MODULE_FIRMWARE("radeon/VERDE_mc.bin"); +MODULE_FIRMWARE("radeon/VERDE_rlc.bin"); +MODULE_FIRMWARE("radeon/VERDE_smc.bin"); +MODULE_FIRMWARE("radeon/OLAND_pfp.bin"); +MODULE_FIRMWARE("radeon/OLAND_me.bin"); +MODULE_FIRMWARE("radeon/OLAND_ce.bin"); +MODULE_FIRMWARE("radeon/OLAND_mc.bin"); +MODULE_FIRMWARE("radeon/OLAND_rlc.bin"); +MODULE_FIRMWARE("radeon/OLAND_smc.bin"); +MODULE_FIRMWARE("radeon/HAINAN_pfp.bin"); +MODULE_FIRMWARE("radeon/HAINAN_me.bin"); +MODULE_FIRMWARE("radeon/HAINAN_ce.bin"); +MODULE_FIRMWARE("radeon/HAINAN_mc.bin"); +MODULE_FIRMWARE("radeon/HAINAN_rlc.bin"); +MODULE_FIRMWARE("radeon/HAINAN_smc.bin"); + +static void si_pcie_gen3_enable(struct radeon_device *rdev); +static void si_program_aspm(struct radeon_device *rdev); extern void evergreen_print_gpu_status_regs(struct radeon_device *rdev); -extern bool evergreen_is_display_hung(struct radeon_device *rdev); + +static const u32 verde_rlc_save_restore_register_list[] = +{ + (0x8000 << 16) | (0x98f4 >> 2), + 0x00000000, + (0x8040 << 16) | (0x98f4 >> 2), + 0x00000000, + (0x8000 << 16) | (0xe80 >> 2), + 0x00000000, + (0x8040 << 16) | (0xe80 >> 2), + 0x00000000, + (0x8000 << 16) | (0x89bc >> 2), + 0x00000000, + (0x8040 << 16) | (0x89bc >> 2), + 0x00000000, + (0x8000 << 16) | (0x8c1c >> 2), + 0x00000000, + (0x8040 << 16) | (0x8c1c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x98f0 >> 2), + 0x00000000, + (0x9c00 << 16) | (0xe7c >> 2), + 0x00000000, + (0x8000 << 16) | (0x9148 >> 2), + 0x00000000, + (0x8040 << 16) | (0x9148 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9150 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x897c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8d8c >> 2), + 0x00000000, + (0x9c00 << 16) | (0xac54 >> 2), + 0X00000000, + 0x3, + (0x9c00 << 16) | (0x98f8 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9910 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9914 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9918 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x991c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9920 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9924 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9928 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x992c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9930 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9934 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9938 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x993c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9940 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9944 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9948 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x994c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9950 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9954 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9958 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x995c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9960 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9964 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9968 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x996c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9970 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9974 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9978 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x997c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9980 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9984 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9988 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x998c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8c00 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8c14 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8c04 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8c08 >> 2), + 0x00000000, + (0x8000 << 16) | (0x9b7c >> 2), + 0x00000000, + (0x8040 << 16) | (0x9b7c >> 2), + 0x00000000, + (0x8000 << 16) | (0xe84 >> 2), + 0x00000000, + (0x8040 << 16) | (0xe84 >> 2), + 0x00000000, + (0x8000 << 16) | (0x89c0 >> 2), + 0x00000000, + (0x8040 << 16) | (0x89c0 >> 2), + 0x00000000, + (0x8000 << 16) | (0x914c >> 2), + 0x00000000, + (0x8040 << 16) | (0x914c >> 2), + 0x00000000, + (0x8000 << 16) | (0x8c20 >> 2), + 0x00000000, + (0x8040 << 16) | (0x8c20 >> 2), + 0x00000000, + (0x8000 << 16) | (0x9354 >> 2), + 0x00000000, + (0x8040 << 16) | (0x9354 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9060 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9364 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9100 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x913c >> 2), + 0x00000000, + (0x8000 << 16) | (0x90e0 >> 2), + 0x00000000, + (0x8000 << 16) | (0x90e4 >> 2), + 0x00000000, + (0x8000 << 16) | (0x90e8 >> 2), + 0x00000000, + (0x8040 << 16) | (0x90e0 >> 2), + 0x00000000, + (0x8040 << 16) | (0x90e4 >> 2), + 0x00000000, + (0x8040 << 16) | (0x90e8 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8bcc >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8b24 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x88c4 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8e50 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8c0c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8e58 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8e5c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9508 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x950c >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9494 >> 2), + 0x00000000, + (0x9c00 << 16) | (0xac0c >> 2), + 0x00000000, + (0x9c00 << 16) | (0xac10 >> 2), + 0x00000000, + (0x9c00 << 16) | (0xac14 >> 2), + 0x00000000, + (0x9c00 << 16) | (0xae00 >> 2), + 0x00000000, + (0x9c00 << 16) | (0xac08 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x88d4 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x88c8 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x88cc >> 2), + 0x00000000, + (0x9c00 << 16) | (0x89b0 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8b10 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x8a14 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9830 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9834 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9838 >> 2), + 0x00000000, + (0x9c00 << 16) | (0x9a10 >> 2), + 0x00000000, + (0x8000 << 16) | (0x9870 >> 2), + 0x00000000, + (0x8000 << 16) | (0x9874 >> 2), + 0x00000000, + (0x8001 << 16) | (0x9870 >> 2), + 0x00000000, + (0x8001 << 16) | (0x9874 >> 2), + 0x00000000, + (0x8040 << 16) | (0x9870 >> 2), + 0x00000000, + (0x8040 << 16) | (0x9874 >> 2), + 0x00000000, + (0x8041 << 16) | (0x9870 >> 2), + 0x00000000, + (0x8041 << 16) | (0x9874 >> 2), + 0x00000000, + 0x00000000 +}; static const u32 tahiti_golden_rlc_registers[] = { @@ -1286,6 +1542,7 @@ static int si_init_microcode(struct radeon_device *rdev) const char *chip_name; const char *rlc_chip_name; size_t pfp_req_size, me_req_size, ce_req_size, rlc_req_size, mc_req_size; + size_t smc_req_size; char fw_name[30]; int err; @@ -1300,6 +1557,7 @@ static int si_init_microcode(struct radeon_device *rdev) ce_req_size = SI_CE_UCODE_SIZE * 4; rlc_req_size = SI_RLC_UCODE_SIZE * 4; mc_req_size = SI_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(TAHITI_SMC_UCODE_SIZE, 4); break; case CHIP_PITCAIRN: chip_name = "PITCAIRN"; @@ -1309,6 +1567,7 @@ static int si_init_microcode(struct radeon_device *rdev) ce_req_size = SI_CE_UCODE_SIZE * 4; rlc_req_size = SI_RLC_UCODE_SIZE * 4; mc_req_size = SI_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(PITCAIRN_SMC_UCODE_SIZE, 4); break; case CHIP_VERDE: chip_name = "VERDE"; @@ -1318,6 +1577,7 @@ static int si_init_microcode(struct radeon_device *rdev) ce_req_size = SI_CE_UCODE_SIZE * 4; rlc_req_size = SI_RLC_UCODE_SIZE * 4; mc_req_size = SI_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(VERDE_SMC_UCODE_SIZE, 4); break; case CHIP_OLAND: chip_name = "OLAND"; @@ -1327,6 +1587,7 @@ static int si_init_microcode(struct radeon_device *rdev) ce_req_size = SI_CE_UCODE_SIZE * 4; rlc_req_size = SI_RLC_UCODE_SIZE * 4; mc_req_size = OLAND_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(OLAND_SMC_UCODE_SIZE, 4); break; case CHIP_HAINAN: chip_name = "HAINAN"; @@ -1336,6 +1597,7 @@ static int si_init_microcode(struct radeon_device *rdev) ce_req_size = SI_CE_UCODE_SIZE * 4; rlc_req_size = SI_RLC_UCODE_SIZE * 4; mc_req_size = OLAND_MC_UCODE_SIZE * 4; + smc_req_size = ALIGN(HAINAN_SMC_UCODE_SIZE, 4); break; default: panic("%s: Unsupported family %d", __func__, rdev->family); } @@ -1410,6 +1672,21 @@ static int si_init_microcode(struct radeon_device *rdev) err = -EINVAL; } + ksnprintf(fw_name, sizeof(fw_name), "radeonkmsfw_%s_smc", chip_name); + err = request_firmware(&rdev->smc_fw, fw_name, rdev->dev); + if (err) { + printk(KERN_ERR + "smc: error loading firmware \"%s\"\n", + fw_name); + release_firmware(rdev->smc_fw); + rdev->smc_fw = NULL; + } else if (rdev->smc_fw->datasize != smc_req_size) { + DRM_ERROR( + "si_smc: Bogus length %zu in firmware \"%s\"\n", + rdev->smc_fw->datasize, fw_name); + err = -EINVAL; + } + out: if (err) { if (err != -EINVAL) @@ -1436,6 +1713,10 @@ out: firmware_put(rdev->mc_fw, FIRMWARE_UNLOAD); rdev->mc_fw = NULL; } + if (rdev->smc_fw != NULL) { + firmware_put(rdev->smc_fw, FIRMWARE_UNLOAD); + rdev->smc_fw = NULL; + } } return err; } @@ -1471,6 +1752,11 @@ static void si_fini_microcode(struct radeon_device *rdev) rdev->mc_fw = NULL; } + if (rdev->smc_fw != NULL) { + firmware_put(rdev->smc_fw, FIRMWARE_UNLOAD); + rdev->smc_fw = NULL; + } + if (rdev->ce_fw != NULL) { firmware_put(rdev->ce_fw, FIRMWARE_UNLOAD); rdev->ce_fw = NULL; @@ -1808,7 +2094,8 @@ static void dce6_program_watermarks(struct radeon_device *rdev, u32 lb_size, u32 num_heads) { struct drm_display_mode *mode = &radeon_crtc->base.mode; - struct dce6_wm_params wm; + struct dce6_wm_params wm_low, wm_high; + u32 dram_channels; u32 pixel_period; u32 line_time = 0; u32 latency_watermark_a = 0, latency_watermark_b = 0; @@ -1824,38 +2111,83 @@ static void dce6_program_watermarks(struct radeon_device *rdev, priority_a_cnt = 0; priority_b_cnt = 0; - wm.yclk = rdev->pm.current_mclk * 10; - wm.sclk = rdev->pm.current_sclk * 10; - wm.disp_clk = mode->clock; - wm.src_width = mode->crtc_hdisplay; - wm.active_time = mode->crtc_hdisplay * pixel_period; - wm.blank_time = line_time - wm.active_time; - wm.interlaced = false; - if (mode->flags & DRM_MODE_FLAG_INTERLACE) - wm.interlaced = true; - wm.vsc = radeon_crtc->vsc; - wm.vtaps = 1; - if (radeon_crtc->rmx_type != RMX_OFF) - wm.vtaps = 2; - wm.bytes_per_pixel = 4; /* XXX: get this from fb config */ - wm.lb_size = lb_size; if (rdev->family == CHIP_ARUBA) - wm.dram_channels = evergreen_get_number_of_dram_channels(rdev); + dram_channels = evergreen_get_number_of_dram_channels(rdev); else - wm.dram_channels = si_get_number_of_dram_channels(rdev); - wm.num_heads = num_heads; + dram_channels = si_get_number_of_dram_channels(rdev); + + /* watermark for high clocks */ + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + wm_high.yclk = + radeon_dpm_get_mclk(rdev, false) * 10; + wm_high.sclk = + radeon_dpm_get_sclk(rdev, false) * 10; + } else { + wm_high.yclk = rdev->pm.current_mclk * 10; + wm_high.sclk = rdev->pm.current_sclk * 10; + } + + wm_high.disp_clk = mode->clock; + wm_high.src_width = mode->crtc_hdisplay; + wm_high.active_time = mode->crtc_hdisplay * pixel_period; + wm_high.blank_time = line_time - wm_high.active_time; + wm_high.interlaced = false; + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + wm_high.interlaced = true; + wm_high.vsc = radeon_crtc->vsc; + wm_high.vtaps = 1; + if (radeon_crtc->rmx_type != RMX_OFF) + wm_high.vtaps = 2; + wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm_high.lb_size = lb_size; + wm_high.dram_channels = dram_channels; + wm_high.num_heads = num_heads; + + /* watermark for low clocks */ + if ((rdev->pm.pm_method == PM_METHOD_DPM) && rdev->pm.dpm_enabled) { + wm_low.yclk = + radeon_dpm_get_mclk(rdev, true) * 10; + wm_low.sclk = + radeon_dpm_get_sclk(rdev, true) * 10; + } else { + wm_low.yclk = rdev->pm.current_mclk * 10; + wm_low.sclk = rdev->pm.current_sclk * 10; + } + + wm_low.disp_clk = mode->clock; + wm_low.src_width = mode->crtc_hdisplay; + wm_low.active_time = mode->crtc_hdisplay * pixel_period; + wm_low.blank_time = line_time - wm_low.active_time; + wm_low.interlaced = false; + if (mode->flags & DRM_MODE_FLAG_INTERLACE) + wm_low.interlaced = true; + wm_low.vsc = radeon_crtc->vsc; + wm_low.vtaps = 1; + if (radeon_crtc->rmx_type != RMX_OFF) + wm_low.vtaps = 2; + wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */ + wm_low.lb_size = lb_size; + wm_low.dram_channels = dram_channels; + wm_low.num_heads = num_heads; /* set for high clocks */ - latency_watermark_a = min(dce6_latency_watermark(&wm), (u32)65535); + latency_watermark_a = min(dce6_latency_watermark(&wm_high), (u32)65535); /* set for low clocks */ - /* wm.yclk = low clk; wm.sclk = low clk */ - latency_watermark_b = min(dce6_latency_watermark(&wm), (u32)65535); + latency_watermark_b = min(dce6_latency_watermark(&wm_low), (u32)65535); /* possibly force display priority to high */ /* should really do this at mode validation time... */ - if (!dce6_average_bandwidth_vs_dram_bandwidth_for_display(&wm) || - !dce6_average_bandwidth_vs_available_bandwidth(&wm) || - !dce6_check_latency_hiding(&wm) || + if (!dce6_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) || + !dce6_average_bandwidth_vs_available_bandwidth(&wm_high) || + !dce6_check_latency_hiding(&wm_high) || + (rdev->disp_priority == 2)) { + DRM_DEBUG_KMS("force priority to high\n"); + priority_a_cnt |= PRIORITY_ALWAYS_ON; + priority_b_cnt |= PRIORITY_ALWAYS_ON; + } + if (!dce6_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) || + !dce6_average_bandwidth_vs_available_bandwidth(&wm_low) || + !dce6_check_latency_hiding(&wm_low) || (rdev->disp_priority == 2)) { DRM_DEBUG_KMS("force priority to high\n"); priority_a_cnt |= PRIORITY_ALWAYS_ON; @@ -1911,6 +2243,10 @@ static void dce6_program_watermarks(struct radeon_device *rdev, WREG32(PRIORITY_A_CNT + radeon_crtc->crtc_offset, priority_a_cnt); WREG32(PRIORITY_B_CNT + radeon_crtc->crtc_offset, priority_b_cnt); + /* save values for DPM */ + radeon_crtc->line_time = line_time; + radeon_crtc->wm_high = latency_watermark_a; + radeon_crtc->wm_low = latency_watermark_b; } void dce6_bandwidth_update(struct radeon_device *rdev) @@ -3551,8 +3887,8 @@ static void si_mc_program(struct radeon_device *rdev) } } -static void si_vram_gtt_location(struct radeon_device *rdev, - struct radeon_mc *mc) +void si_vram_gtt_location(struct radeon_device *rdev, + struct radeon_mc *mc) { if (mc->mc_vram_size > 0xFFC0000000ULL) { /* leave room for at least 1024M GTT */ @@ -4122,6 +4458,270 @@ void si_vm_fini(struct radeon_device *rdev) { } +/** + * si_vm_decode_fault - print human readable fault info + * + * @rdev: radeon_device pointer + * @status: VM_CONTEXT1_PROTECTION_FAULT_STATUS register value + * @addr: VM_CONTEXT1_PROTECTION_FAULT_ADDR register value + * + * Print human readable fault information (SI). + */ +static void si_vm_decode_fault(struct radeon_device *rdev, + u32 status, u32 addr) +{ + u32 mc_id = (status & MEMORY_CLIENT_ID_MASK) >> MEMORY_CLIENT_ID_SHIFT; + u32 vmid = (status & FAULT_VMID_MASK) >> FAULT_VMID_SHIFT; + u32 protections = (status & PROTECTIONS_MASK) >> PROTECTIONS_SHIFT; + char *block; + + if (rdev->family == CHIP_TAHITI) { + switch (mc_id) { + case 160: + case 144: + case 96: + case 80: + case 224: + case 208: + case 32: + case 16: + block = "CB"; + break; + case 161: + case 145: + case 97: + case 81: + case 225: + case 209: + case 33: + case 17: + block = "CB_FMASK"; + break; + case 162: + case 146: + case 98: + case 82: + case 226: + case 210: + case 34: + case 18: + block = "CB_CMASK"; + break; + case 163: + case 147: + case 99: + case 83: + case 227: + case 211: + case 35: + case 19: + block = "CB_IMMED"; + break; + case 164: + case 148: + case 100: + case 84: + case 228: + case 212: + case 36: + case 20: + block = "DB"; + break; + case 165: + case 149: + case 101: + case 85: + case 229: + case 213: + case 37: + case 21: + block = "DB_HTILE"; + break; + case 167: + case 151: + case 103: + case 87: + case 231: + case 215: + case 39: + case 23: + block = "DB_STEN"; + break; + case 72: + case 68: + case 64: + case 8: + case 4: + case 0: + case 136: + case 132: + case 128: + case 200: + case 196: + case 192: + block = "TC"; + break; + case 112: + case 48: + block = "CP"; + break; + case 49: + case 177: + case 50: + case 178: + block = "SH"; + break; + case 53: + case 190: + block = "VGT"; + break; + case 117: + block = "IH"; + break; + case 51: + case 115: + block = "RLC"; + break; + case 119: + case 183: + block = "DMA0"; + break; + case 61: + block = "DMA1"; + break; + case 248: + case 120: + block = "HDP"; + break; + default: + block = "unknown"; + break; + } + } else { + switch (mc_id) { + case 32: + case 16: + case 96: + case 80: + case 160: + case 144: + case 224: + case 208: + block = "CB"; + break; + case 33: + case 17: + case 97: + case 81: + case 161: + case 145: + case 225: + case 209: + block = "CB_FMASK"; + break; + case 34: + case 18: + case 98: + case 82: + case 162: + case 146: + case 226: + case 210: + block = "CB_CMASK"; + break; + case 35: + case 19: + case 99: + case 83: + case 163: + case 147: + case 227: + case 211: + block = "CB_IMMED"; + break; + case 36: + case 20: + case 100: + case 84: + case 164: + case 148: + case 228: + case 212: + block = "DB"; + break; + case 37: + case 21: + case 101: + case 85: + case 165: + case 149: + case 229: + case 213: + block = "DB_HTILE"; + break; + case 39: + case 23: + case 103: + case 87: + case 167: + case 151: + case 231: + case 215: + block = "DB_STEN"; + break; + case 72: + case 68: + case 8: + case 4: + case 136: + case 132: + case 200: + case 196: + block = "TC"; + break; + case 112: + case 48: + block = "CP"; + break; + case 49: + case 177: + case 50: + case 178: + block = "SH"; + break; + case 53: + block = "VGT"; + break; + case 117: + block = "IH"; + break; + case 51: + case 115: + block = "RLC"; + break; + case 119: + case 183: + block = "DMA0"; + break; + case 61: + block = "DMA1"; + break; + case 248: + case 120: + block = "HDP"; + break; + default: + block = "unknown"; + break; + } + } + + printk("VM fault (0x%02x, vmid %d) at page %u, %s from %s (%d)\n", + protections, vmid, addr, + (status & MEMORY_CLIENT_RW_MASK) ? "write" : "read", + block, mc_id); +} + /** * si_vm_set_page - update the page tables using the CP * @@ -4297,6 +4897,446 @@ void si_dma_vm_flush(struct radeon_device *rdev, int ridx, struct radeon_vm *vm) radeon_ring_write(ring, 1 << vm->id); } +/* + * Power and clock gating + */ +static void si_wait_for_rlc_serdes(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(RLC_SERDES_MASTER_BUSY_0) == 0) + break; + DRM_UDELAY(1); + } + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(RLC_SERDES_MASTER_BUSY_1) == 0) + break; + DRM_UDELAY(1); + } +} + +static void si_enable_gui_idle_interrupt(struct radeon_device *rdev, + bool enable) +{ + u32 tmp = RREG32(CP_INT_CNTL_RING0); + u32 mask; + int i; + + if (enable) + tmp |= (CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); + else + tmp &= ~(CNTX_BUSY_INT_ENABLE | CNTX_EMPTY_INT_ENABLE); + WREG32(CP_INT_CNTL_RING0, tmp); + + if (!enable) { + /* read a gfx register */ + tmp = RREG32(DB_DEPTH_INFO); + + mask = RLC_BUSY_STATUS | GFX_POWER_STATUS | GFX_CLOCK_STATUS | GFX_LS_STATUS; + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(RLC_STAT) & mask) == (GFX_CLOCK_STATUS | GFX_POWER_STATUS)) + break; + DRM_UDELAY(1); + } + } +} + +static void si_set_uvd_dcm(struct radeon_device *rdev, + bool sw_mode) +{ + u32 tmp, tmp2; + + tmp = RREG32(UVD_CGC_CTRL); + tmp &= ~(CLK_OD_MASK | CG_DT_MASK); + tmp |= DCM | CG_DT(1) | CLK_OD(4); + + if (sw_mode) { + tmp &= ~0x7ffff800; + tmp2 = DYN_OR_EN | DYN_RR_EN | G_DIV_ID(7); + } else { + tmp |= 0x7ffff800; + tmp2 = 0; + } + + WREG32(UVD_CGC_CTRL, tmp); + WREG32_UVD_CTX(UVD_CGC_CTRL2, tmp2); +} + +static void si_init_uvd_internal_cg(struct radeon_device *rdev) +{ + bool hw_mode = true; + + if (hw_mode) { + si_set_uvd_dcm(rdev, false); + } else { + u32 tmp = RREG32(UVD_CGC_CTRL); + tmp &= ~DCM; + WREG32(UVD_CGC_CTRL, tmp); + } +} + +static u32 si_halt_rlc(struct radeon_device *rdev) +{ + u32 data, orig; + + orig = data = RREG32(RLC_CNTL); + + if (data & RLC_ENABLE) { + data &= ~RLC_ENABLE; + WREG32(RLC_CNTL, data); + + si_wait_for_rlc_serdes(rdev); + } + + return orig; +} + +static void si_update_rlc(struct radeon_device *rdev, u32 rlc) +{ + u32 tmp; + + tmp = RREG32(RLC_CNTL); + if (tmp != rlc) + WREG32(RLC_CNTL, rlc); +} + +static void si_enable_dma_pg(struct radeon_device *rdev, bool enable) +{ + u32 data, orig; + + orig = data = RREG32(DMA_PG); + if (enable) + data |= PG_CNTL_ENABLE; + else + data &= ~PG_CNTL_ENABLE; + if (orig != data) + WREG32(DMA_PG, data); +} + +static void si_init_dma_pg(struct radeon_device *rdev) +{ + u32 tmp; + + WREG32(DMA_PGFSM_WRITE, 0x00002000); + WREG32(DMA_PGFSM_CONFIG, 0x100010ff); + + for (tmp = 0; tmp < 5; tmp++) + WREG32(DMA_PGFSM_WRITE, 0); +} + +static void si_enable_gfx_cgpg(struct radeon_device *rdev, + bool enable) +{ + u32 tmp; + + if (enable) { + tmp = RLC_PUD(0x10) | RLC_PDD(0x10) | RLC_TTPD(0x10) | RLC_MSD(0x10); + WREG32(RLC_TTOP_D, tmp); + + tmp = RREG32(RLC_PG_CNTL); + tmp |= GFX_PG_ENABLE; + WREG32(RLC_PG_CNTL, tmp); + + tmp = RREG32(RLC_AUTO_PG_CTRL); + tmp |= AUTO_PG_EN; + WREG32(RLC_AUTO_PG_CTRL, tmp); + } else { + tmp = RREG32(RLC_AUTO_PG_CTRL); + tmp &= ~AUTO_PG_EN; + WREG32(RLC_AUTO_PG_CTRL, tmp); + + tmp = RREG32(DB_RENDER_CONTROL); + } +} + +static void si_init_gfx_cgpg(struct radeon_device *rdev) +{ + u32 tmp; + + WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); + + tmp = RREG32(RLC_PG_CNTL); + tmp |= GFX_PG_SRC; + WREG32(RLC_PG_CNTL, tmp); + + WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8); + + tmp = RREG32(RLC_AUTO_PG_CTRL); + + tmp &= ~GRBM_REG_SGIT_MASK; + tmp |= GRBM_REG_SGIT(0x700); + tmp &= ~PG_AFTER_GRBM_REG_ST_MASK; + WREG32(RLC_AUTO_PG_CTRL, tmp); +} + +static u32 si_get_cu_active_bitmap(struct radeon_device *rdev, u32 se, u32 sh) +{ + u32 mask = 0, tmp, tmp1; + int i; + + si_select_se_sh(rdev, se, sh); + tmp = RREG32(CC_GC_SHADER_ARRAY_CONFIG); + tmp1 = RREG32(GC_USER_SHADER_ARRAY_CONFIG); + si_select_se_sh(rdev, 0xffffffff, 0xffffffff); + + tmp &= 0xffff0000; + + tmp |= tmp1; + tmp >>= 16; + + for (i = 0; i < rdev->config.si.max_cu_per_sh; i ++) { + mask <<= 1; + mask |= 1; + } + + return (~tmp) & mask; +} + +static void si_init_ao_cu_mask(struct radeon_device *rdev) +{ + u32 i, j, k, active_cu_number = 0; + u32 mask, counter, cu_bitmap; + u32 tmp = 0; + + for (i = 0; i < rdev->config.si.max_shader_engines; i++) { + for (j = 0; j < rdev->config.si.max_sh_per_se; j++) { + mask = 1; + cu_bitmap = 0; + counter = 0; + for (k = 0; k < rdev->config.si.max_cu_per_sh; k++) { + if (si_get_cu_active_bitmap(rdev, i, j) & mask) { + if (counter < 2) + cu_bitmap |= mask; + counter++; + } + mask <<= 1; + } + + active_cu_number += counter; + tmp |= (cu_bitmap << (i * 16 + j * 8)); + } + } + + WREG32(RLC_PG_AO_CU_MASK, tmp); + + tmp = RREG32(RLC_MAX_PG_CU); + tmp &= ~MAX_PU_CU_MASK; + tmp |= MAX_PU_CU(active_cu_number); + WREG32(RLC_MAX_PG_CU, tmp); +} + +static void si_enable_cgcg(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig, tmp; + + orig = data = RREG32(RLC_CGCG_CGLS_CTRL); + + si_enable_gui_idle_interrupt(rdev, enable); + + if (enable) { + WREG32(RLC_GCPM_GENERAL_3, 0x00000080); + + tmp = si_halt_rlc(rdev); + + WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff); + WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff); + WREG32(RLC_SERDES_WR_CTRL, 0x00b000ff); + + si_wait_for_rlc_serdes(rdev); + + si_update_rlc(rdev, tmp); + + WREG32(RLC_SERDES_WR_CTRL, 0x007000ff); + + data |= CGCG_EN | CGLS_EN; + } else { + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + + data &= ~(CGCG_EN | CGLS_EN); + } + + if (orig != data) + WREG32(RLC_CGCG_CGLS_CTRL, data); +} + +static void si_enable_mgcg(struct radeon_device *rdev, + bool enable) +{ + u32 data, orig, tmp = 0; + + if (enable) { + orig = data = RREG32(CGTS_SM_CTRL_REG); + data = 0x96940200; + if (orig != data) + WREG32(CGTS_SM_CTRL_REG, data); + + orig = data = RREG32(CP_MEM_SLP_CNTL); + data |= CP_MEM_LS_EN; + if (orig != data) + WREG32(CP_MEM_SLP_CNTL, data); + + orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); + data &= 0xffffffc0; + if (orig != data) + WREG32(RLC_CGTT_MGCG_OVERRIDE, data); + + tmp = si_halt_rlc(rdev); + + WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff); + WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff); + WREG32(RLC_SERDES_WR_CTRL, 0x00d000ff); + + si_update_rlc(rdev, tmp); + } else { + orig = data = RREG32(RLC_CGTT_MGCG_OVERRIDE); + data |= 0x00000003; + if (orig != data) + WREG32(RLC_CGTT_MGCG_OVERRIDE, data); + + data = RREG32(CP_MEM_SLP_CNTL); + if (data & CP_MEM_LS_EN) { + data &= ~CP_MEM_LS_EN; + WREG32(CP_MEM_SLP_CNTL, data); + } + orig = data = RREG32(CGTS_SM_CTRL_REG); + data |= LS_OVERRIDE | OVERRIDE; + if (orig != data) + WREG32(CGTS_SM_CTRL_REG, data); + + tmp = si_halt_rlc(rdev); + + WREG32(RLC_SERDES_WR_MASTER_MASK_0, 0xffffffff); + WREG32(RLC_SERDES_WR_MASTER_MASK_1, 0xffffffff); + WREG32(RLC_SERDES_WR_CTRL, 0x00e000ff); + + si_update_rlc(rdev, tmp); + } +} + +static void si_enable_uvd_mgcg(struct radeon_device *rdev, + bool enable) +{ + u32 orig, data, tmp; + + if (enable) { + tmp = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL); + tmp |= 0x3fff; + WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, tmp); + + orig = data = RREG32(UVD_CGC_CTRL); + data |= DCM; + if (orig != data) + WREG32(UVD_CGC_CTRL, data); + + WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_0, 0); + WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_1, 0); + } else { + tmp = RREG32_UVD_CTX(UVD_CGC_MEM_CTRL); + tmp &= ~0x3fff; + WREG32_UVD_CTX(UVD_CGC_MEM_CTRL, tmp); + + orig = data = RREG32(UVD_CGC_CTRL); + data &= ~DCM; + if (orig != data) + WREG32(UVD_CGC_CTRL, data); + + WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_0, 0xffffffff); + WREG32_SMC(SMC_CG_IND_START + CG_CGTT_LOCAL_1, 0xffffffff); + } +} + +static const u32 mc_cg_registers[] = +{ + MC_HUB_MISC_HUB_CG, + MC_HUB_MISC_SIP_CG, + MC_HUB_MISC_VM_CG, + MC_XPB_CLK_GAT, + ATC_MISC_CG, + MC_CITF_MISC_WR_CG, + MC_CITF_MISC_RD_CG, + MC_CITF_MISC_VM_CG, + VM_L2_CG, +}; + +static void si_enable_mc_ls(struct radeon_device *rdev, + bool enable) +{ + int i; + u32 orig, data; + + for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) { + orig = data = RREG32(mc_cg_registers[i]); + if (enable) + data |= MC_LS_ENABLE; + else + data &= ~MC_LS_ENABLE; + if (data != orig) + WREG32(mc_cg_registers[i], data); + } +} + + +static void si_init_cg(struct radeon_device *rdev) +{ + si_enable_mgcg(rdev, true); + si_enable_cgcg(rdev, false); + /* disable MC LS on Tahiti */ + if (rdev->family == CHIP_TAHITI) + si_enable_mc_ls(rdev, false); + if (rdev->has_uvd) { + si_enable_uvd_mgcg(rdev, true); + si_init_uvd_internal_cg(rdev); + } +} + +static void si_fini_cg(struct radeon_device *rdev) +{ + if (rdev->has_uvd) + si_enable_uvd_mgcg(rdev, false); + si_enable_cgcg(rdev, false); + si_enable_mgcg(rdev, false); +} + +static void si_init_pg(struct radeon_device *rdev) +{ + bool has_pg = false; +#if 0 + /* only cape verde supports PG */ + if (rdev->family == CHIP_VERDE) + has_pg = true; +#endif + if (has_pg) { + si_init_ao_cu_mask(rdev); + si_init_dma_pg(rdev); + si_enable_dma_pg(rdev, true); + si_init_gfx_cgpg(rdev); + si_enable_gfx_cgpg(rdev, true); + } else { + WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); + WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8); + } +} + +static void si_fini_pg(struct radeon_device *rdev) +{ + bool has_pg = false; + + /* only cape verde supports PG */ + if (rdev->family == CHIP_VERDE) + has_pg = true; + + if (has_pg) { + si_enable_dma_pg(rdev, false); + si_enable_gfx_cgpg(rdev, false); + } +} + /* * RLC */ @@ -4329,8 +5369,16 @@ void si_rlc_fini(struct radeon_device *rdev) } } +#define RLC_CLEAR_STATE_END_MARKER 0x00000001 + int si_rlc_init(struct radeon_device *rdev) { + volatile u32 *dst_ptr; + void *ptr; + u32 dws, data, i, j, k, reg_num; + u32 reg_list_num, reg_list_hdr_blk_index, reg_list_blk_index; + u64 reg_list_mc_addr; + const struct cs_section_def *cs_data = si_cs_data; int r; /* save restore block */ @@ -4351,18 +5399,45 @@ int si_rlc_init(struct radeon_device *rdev) } r = radeon_bo_pin(rdev->rlc.save_restore_obj, RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.save_restore_gpu_addr); - radeon_bo_unreserve(rdev->rlc.save_restore_obj); if (r) { + radeon_bo_unreserve(rdev->rlc.save_restore_obj); dev_warn(rdev->dev, "(%d) pin RLC sr bo failed\n", r); si_rlc_fini(rdev); return r; } + if (rdev->family == CHIP_VERDE) { + r = radeon_bo_kmap(rdev->rlc.save_restore_obj, (void **)&ptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC sr bo failed\n", r); + si_rlc_fini(rdev); + return r; + } + rdev->rlc.sr_ptr = ptr; + /* write the sr buffer */ + dst_ptr = rdev->rlc.sr_ptr; + for (i = 0; i < ARRAY_SIZE(verde_rlc_save_restore_register_list); i++) { + dst_ptr[i] = verde_rlc_save_restore_register_list[i]; + } + radeon_bo_kunmap(rdev->rlc.save_restore_obj); + } + radeon_bo_unreserve(rdev->rlc.save_restore_obj); + /* clear state block */ + reg_list_num = 0; + dws = 0; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_list_num++; + dws += cs_data[i].section[j].reg_count; + } + } + reg_list_blk_index = (3 * reg_list_num + 2); + dws += reg_list_blk_index; + if (rdev->rlc.clear_state_obj == NULL) { - r = radeon_bo_create(rdev, RADEON_GPU_PAGE_SIZE, PAGE_SIZE, true, - RADEON_GEM_DOMAIN_VRAM, NULL, - &rdev->rlc.clear_state_obj); + r = radeon_bo_create(rdev, dws * 4, PAGE_SIZE, true, + RADEON_GEM_DOMAIN_VRAM, NULL, &rdev->rlc.clear_state_obj); if (r) { dev_warn(rdev->dev, "(%d) create RLC c bo failed\n", r); si_rlc_fini(rdev); @@ -4376,24 +5451,114 @@ int si_rlc_init(struct radeon_device *rdev) } r = radeon_bo_pin(rdev->rlc.clear_state_obj, RADEON_GEM_DOMAIN_VRAM, &rdev->rlc.clear_state_gpu_addr); - radeon_bo_unreserve(rdev->rlc.clear_state_obj); if (r) { + + radeon_bo_unreserve(rdev->rlc.clear_state_obj); dev_warn(rdev->dev, "(%d) pin RLC c bo failed\n", r); si_rlc_fini(rdev); return r; } + r = radeon_bo_kmap(rdev->rlc.clear_state_obj, (void **)&ptr); + if (r) { + dev_warn(rdev->dev, "(%d) map RLC c bo failed\n", r); + si_rlc_fini(rdev); + return r; + } + rdev->rlc.cs_ptr = ptr; + /* set up the cs buffer */ + dst_ptr = rdev->rlc.cs_ptr; + reg_list_hdr_blk_index = 0; + reg_list_mc_addr = rdev->rlc.clear_state_gpu_addr + (reg_list_blk_index * 4); + data = upper_32_bits(reg_list_mc_addr); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + for (i = 0; cs_data[i].section != NULL; i++) { + for (j = 0; cs_data[i].section[j].extent != NULL; j++) { + reg_num = cs_data[i].section[j].reg_count; + data = reg_list_mc_addr & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = (cs_data[i].section[j].reg_index * 4) & 0xffffffff; + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + data = 0x08000000 | (reg_num * 4); + dst_ptr[reg_list_hdr_blk_index] = data; + reg_list_hdr_blk_index++; + + for (k = 0; k < reg_num; k++) { + data = cs_data[i].section[j].extent[k]; + dst_ptr[reg_list_blk_index + k] = data; + } + reg_list_mc_addr += reg_num * 4; + reg_list_blk_index += reg_num; + } + } + dst_ptr[reg_list_hdr_blk_index] = RLC_CLEAR_STATE_END_MARKER; + + radeon_bo_kunmap(rdev->rlc.clear_state_obj); + radeon_bo_unreserve(rdev->rlc.clear_state_obj); return 0; } +static void si_rlc_reset(struct radeon_device *rdev) +{ + u32 tmp = RREG32(GRBM_SOFT_RESET); + + tmp |= SOFT_RESET_RLC; + WREG32(GRBM_SOFT_RESET, tmp); + DRM_UDELAY(50); + tmp &= ~SOFT_RESET_RLC; + WREG32(GRBM_SOFT_RESET, tmp); + DRM_UDELAY(50); +} + static void si_rlc_stop(struct radeon_device *rdev) { WREG32(RLC_CNTL, 0); + + si_enable_gui_idle_interrupt(rdev, false); + + si_wait_for_rlc_serdes(rdev); } static void si_rlc_start(struct radeon_device *rdev) { WREG32(RLC_CNTL, RLC_ENABLE); + + si_enable_gui_idle_interrupt(rdev, true); + + DRM_UDELAY(50); +} + +static bool si_lbpw_supported(struct radeon_device *rdev) +{ + u32 tmp; + + /* Enable LBPW only for DDR3 */ + tmp = RREG32(MC_SEQ_MISC0); + if ((tmp & 0xF0000000) == 0xB0000000) + return true; + return false; +} + +static void si_enable_lbpw(struct radeon_device *rdev, bool enable) +{ + u32 tmp; + + tmp = RREG32(RLC_LB_CNTL); + if (enable) + tmp |= LOAD_BALANCE_ENABLE; + else + tmp &= ~LOAD_BALANCE_ENABLE; + WREG32(RLC_LB_CNTL, tmp); + + if (!enable) { + si_select_se_sh(rdev, 0xffffffff, 0xffffffff); + WREG32(SPI_LB_CU_MASK, 0x00ff); + } } static int si_rlc_resume(struct radeon_device *rdev) @@ -4406,14 +5571,18 @@ static int si_rlc_resume(struct radeon_device *rdev) si_rlc_stop(rdev); + si_rlc_reset(rdev); + + si_init_pg(rdev); + + si_init_cg(rdev); + WREG32(RLC_RL_BASE, 0); WREG32(RLC_RL_SIZE, 0); WREG32(RLC_LB_CNTL, 0); WREG32(RLC_LB_CNTR_MAX, 0xffffffff); WREG32(RLC_LB_CNTR_INIT, 0); - - WREG32(RLC_SAVE_AND_RESTORE_BASE, rdev->rlc.save_restore_gpu_addr >> 8); - WREG32(RLC_CLEAR_STATE_RESTORE_BASE, rdev->rlc.clear_state_gpu_addr >> 8); + WREG32(RLC_LB_INIT_CU_MASK, 0xffffffff); WREG32(RLC_MC_CNTL, 0); WREG32(RLC_UCODE_CNTL, 0); @@ -4425,6 +5594,8 @@ static int si_rlc_resume(struct radeon_device *rdev) } WREG32(RLC_UCODE_ADDR, 0); + si_enable_lbpw(rdev, si_lbpw_supported(rdev)); + si_rlc_start(rdev); return 0; @@ -4594,6 +5765,7 @@ int si_irq_set(struct radeon_device *rdev) u32 grbm_int_cntl = 0; u32 grph1 = 0, grph2 = 0, grph3 = 0, grph4 = 0, grph5 = 0, grph6 = 0; u32 dma_cntl, dma_cntl1; + u32 thermal_int = 0; if (!rdev->irq.installed) { DRM_ERROR("Can't enable IRQ/MSI because no handler is installed\n"); @@ -4619,6 +5791,9 @@ int si_irq_set(struct radeon_device *rdev) dma_cntl = RREG32(DMA_CNTL + DMA0_REGISTER_OFFSET) & ~TRAP_ENABLE; dma_cntl1 = RREG32(DMA_CNTL + DMA1_REGISTER_OFFSET) & ~TRAP_ENABLE; + thermal_int = RREG32(CG_THERMAL_INT) & + ~(THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW); + /* enable CP interrupts on all rings */ if (atomic_read(&rdev->irq.ring_int[RADEON_RING_TYPE_GFX_INDEX])) { DRM_DEBUG("si_irq_set: sw int gfx\n"); @@ -4705,6 +5880,11 @@ int si_irq_set(struct radeon_device *rdev) WREG32(GRBM_INT_CNTL, grbm_int_cntl); + if (rdev->irq.dpm_thermal) { + DRM_DEBUG("dpm thermal\n"); + thermal_int |= THERM_INT_MASK_HIGH | THERM_INT_MASK_LOW; + } + if (rdev->num_crtc >= 2) { WREG32(INT_MASK + EVERGREEN_CRTC0_REGISTER_OFFSET, crtc1); WREG32(INT_MASK + EVERGREEN_CRTC1_REGISTER_OFFSET, crtc2); @@ -4740,6 +5920,8 @@ int si_irq_set(struct radeon_device *rdev) WREG32(DC_HPD6_INT_CONTROL, hpd6); } + WREG32(CG_THERMAL_INT, thermal_int); + return 0; } @@ -4904,6 +6086,8 @@ irqreturn_t si_irq_process(struct radeon_device *rdev) u32 src_id, src_data, ring_id; u32 ring_index; bool queue_hotplug = false; + bool queue_thermal = false; + u32 status, addr; if (!rdev->ih.enabled || rdev->shutdown) return IRQ_NONE; @@ -5139,11 +6323,14 @@ restart_ih: break; case 146: case 147: + addr = RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR); + status = RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS); dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data); dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n", - RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR)); + addr); dev_err(rdev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n", - RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS)); + status); + si_vm_decode_fault(rdev, status, addr); /* reset addr and status */ WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1); break; @@ -5174,6 +6361,16 @@ restart_ih: DRM_DEBUG("IH: DMA trap\n"); radeon_fence_process(rdev, R600_RING_TYPE_DMA_INDEX); break; + case 230: /* thermal low to high */ + DRM_DEBUG("IH: thermal low to high\n"); + rdev->pm.dpm.thermal.high_to_low = false; + queue_thermal = true; + break; + case 231: /* thermal high to low */ + DRM_DEBUG("IH: thermal high to low\n"); + rdev->pm.dpm.thermal.high_to_low = true; + queue_thermal = true; + break; case 233: /* GUI IDLE */ DRM_DEBUG("IH: GUI idle\n"); break; @@ -5192,6 +6389,9 @@ restart_ih: } if (queue_hotplug) taskqueue_enqueue(rdev->tq, &rdev->hotplug_work); + if (queue_thermal && rdev->pm.dpm_enabled) + taskqueue_enqueue(rdev->tq, &rdev->pm.dpm.thermal.work); + rdev->ih.rptr = rptr; WREG32(IH_RB_RPTR, rdev->ih.rptr); atomic_set(&rdev->ih.lock, 0); @@ -5286,6 +6486,13 @@ static int si_startup(struct radeon_device *rdev) struct radeon_ring *ring; int r; + /* enable pcie gen2/3 link */ + si_pcie_gen3_enable(rdev); + /* enable aspm */ + si_program_aspm(rdev); + + si_mc_program(rdev); + if (!rdev->me_fw || !rdev->pfp_fw || !rdev->ce_fw || !rdev->rlc_fw || !rdev->mc_fw) { r = si_init_microcode(rdev); @@ -5305,7 +6512,6 @@ static int si_startup(struct radeon_device *rdev) if (r) return r; - si_mc_program(rdev); r = si_pcie_gart_enable(rdev); if (r) return r; @@ -5489,7 +6695,7 @@ int si_suspend(struct radeon_device *rdev) si_cp_enable(rdev, false); cayman_dma_stop(rdev); if (rdev->has_uvd) { - r600_uvd_rbc_stop(rdev); + r600_uvd_stop(rdev); radeon_uvd_suspend(rdev); } si_irq_suspend(rdev); @@ -5625,12 +6831,16 @@ void si_fini(struct radeon_device *rdev) cayman_dma_fini(rdev); si_irq_fini(rdev); si_rlc_fini(rdev); + si_fini_cg(rdev); + si_fini_pg(rdev); radeon_wb_fini(rdev); radeon_vm_manager_fini(rdev); radeon_ib_pool_fini(rdev); radeon_irq_kms_fini(rdev); - if (rdev->has_uvd) + if (rdev->has_uvd) { + r600_uvd_stop(rdev); radeon_uvd_fini(rdev); + } si_pcie_gart_fini(rdev); r600_vram_scratch_fini(rdev); radeon_gem_fini(rdev); @@ -5752,3 +6962,381 @@ int si_set_uvd_clocks(struct radeon_device *rdev, u32 vclk, u32 dclk) return 0; } + +static struct pci_dev dev_to_pcidev(device_t dev) +{ + struct pci_dev pdev; + pdev.dev = dev; + return pdev; +} + +static void si_pcie_gen3_enable(struct radeon_device *rdev) +{ +#if 0 + struct pci_dev *root = rdev->dev->bus->self; +#else + device_t root = device_get_parent(rdev->dev); +#endif + int bridge_pos, gpu_pos; + u32 speed_cntl, mask, current_data_rate; + int ret, i; + u16 tmp16; + struct pci_dev root_pdev = dev_to_pcidev(root); + struct pci_dev pdev = dev_to_pcidev(rdev->dev); + + if (radeon_pcie_gen2 == 0) + return; + + if (rdev->flags & RADEON_IS_IGP) + return; + + if (!(rdev->flags & RADEON_IS_PCIE)) + return; + + ret = drm_pcie_get_speed_cap_mask(rdev->ddev, &mask); + if (ret != 0) + return; + + if (!(mask & (DRM_PCIE_SPEED_50 | DRM_PCIE_SPEED_80))) + return; + + speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + current_data_rate = (speed_cntl & LC_CURRENT_DATA_RATE_MASK) >> + LC_CURRENT_DATA_RATE_SHIFT; + if (mask & DRM_PCIE_SPEED_80) { + if (current_data_rate == 2) { + DRM_INFO("PCIE gen 3 link speeds already enabled\n"); + return; + } + DRM_INFO("enabling PCIE gen 3 link speeds, disable with radeon.pcie_gen2=0\n"); + } else if (mask & DRM_PCIE_SPEED_50) { + if (current_data_rate == 1) { + DRM_INFO("PCIE gen 2 link speeds already enabled\n"); + return; + } + DRM_INFO("enabling PCIE gen 2 link speeds, disable with radeon.pcie_gen2=0\n"); + } + + bridge_pos = pci_get_pciecap_ptr(root); + if (!bridge_pos) + return; + + gpu_pos = pci_get_pciecap_ptr(rdev->dev); + if (!gpu_pos) + return; + + if (mask & DRM_PCIE_SPEED_80) { + /* re-try equalization if gen3 is not already enabled */ + if (current_data_rate != 2) { + u16 bridge_cfg, gpu_cfg; + u16 bridge_cfg2, gpu_cfg2; + u32 max_lw, current_lw, tmp; + + pci_read_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg); + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg); + + tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD; + pci_write_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL, tmp16); + + tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD; + pci_write_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16); + + tmp = RREG32_PCIE(PCIE_LC_STATUS1); + max_lw = (tmp & LC_DETECTED_LINK_WIDTH_MASK) >> LC_DETECTED_LINK_WIDTH_SHIFT; + current_lw = (tmp & LC_OPERATING_LINK_WIDTH_MASK) >> LC_OPERATING_LINK_WIDTH_SHIFT; + + if (current_lw < max_lw) { + tmp = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL); + if (tmp & LC_RENEGOTIATION_SUPPORT) { + tmp &= ~(LC_LINK_WIDTH_MASK | LC_UPCONFIGURE_DIS); + tmp |= (max_lw << LC_LINK_WIDTH_SHIFT); + tmp |= LC_UPCONFIGURE_SUPPORT | LC_RENEGOTIATE_EN | LC_RECONFIG_NOW; + WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, tmp); + } + } + + for (i = 0; i < 10; i++) { + /* check status */ + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_DEVSTA, &tmp16); + if (tmp16 & PCI_EXP_DEVSTA_TRPND) + break; + + pci_read_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg); + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg); + + pci_read_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL2, &bridge_cfg2); + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL2, &gpu_cfg2); + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4); + tmp |= LC_SET_QUIESCE; + WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp); + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4); + tmp |= LC_REDO_EQ; + WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp); + + DRM_MDELAY(100); + + /* linkctl */ + pci_read_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL, &tmp16); + tmp16 &= ~PCI_EXP_LNKCTL_HAWD; + tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD); + pci_write_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL, tmp16); + + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL, &tmp16); + tmp16 &= ~PCI_EXP_LNKCTL_HAWD; + tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD); + pci_write_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16); + + /* linkctl2 */ + pci_read_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL2, &tmp16); + tmp16 &= ~((1 << 4) | (7 << 9)); + tmp16 |= (bridge_cfg2 & ((1 << 4) | (7 << 9))); + pci_write_config_word(&root_pdev, bridge_pos + PCI_EXP_LNKCTL2, tmp16); + + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16); + tmp16 &= ~((1 << 4) | (7 << 9)); + tmp16 |= (gpu_cfg2 & ((1 << 4) | (7 << 9))); + pci_write_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16); + + tmp = RREG32_PCIE_PORT(PCIE_LC_CNTL4); + tmp &= ~LC_SET_QUIESCE; + WREG32_PCIE_PORT(PCIE_LC_CNTL4, tmp); + } + } + } + + /* set the link speed */ + speed_cntl |= LC_FORCE_EN_SW_SPEED_CHANGE | LC_FORCE_DIS_HW_SPEED_CHANGE; + speed_cntl &= ~LC_FORCE_DIS_SW_SPEED_CHANGE; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl); + + pci_read_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16); + tmp16 &= ~0xf; + if (mask & DRM_PCIE_SPEED_80) + tmp16 |= 3; /* gen3 */ + else if (mask & DRM_PCIE_SPEED_50) + tmp16 |= 2; /* gen2 */ + else + tmp16 |= 1; /* gen1 */ + pci_write_config_word(&pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16); + + speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + speed_cntl |= LC_INITIATE_LINK_SPEED_CHANGE; + WREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL, speed_cntl); + + for (i = 0; i < rdev->usec_timeout; i++) { + speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL); + if ((speed_cntl & LC_INITIATE_LINK_SPEED_CHANGE) == 0) + break; + DRM_UDELAY(1); + } +} + +static void si_program_aspm(struct radeon_device *rdev) +{ + u32 data, orig; + bool disable_l0s = false, disable_l1 = false, disable_plloff_in_l1 = false; + bool disable_clkreq = false; + + if (radeon_aspm == 0) + return; + + if (!(rdev->flags & RADEON_IS_PCIE)) + return; + + orig = data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL); + data &= ~LC_XMIT_N_FTS_MASK; + data |= LC_XMIT_N_FTS(0x24) | LC_XMIT_N_FTS_OVERRIDE_EN; + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL, data); + + orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL3); + data |= LC_GO_TO_RECOVERY; + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_CNTL3, data); + + orig = data = RREG32_PCIE(PCIE_P_CNTL); + data |= P_IGNORE_EDB_ERR; + if (orig != data) + WREG32_PCIE(PCIE_P_CNTL, data); + + orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL); + data &= ~(LC_L0S_INACTIVITY_MASK | LC_L1_INACTIVITY_MASK); + data |= LC_PMI_TO_L1_DIS; + if (!disable_l0s) + data |= LC_L0S_INACTIVITY(7); + + if (!disable_l1) { + data |= LC_L1_INACTIVITY(7); + data &= ~LC_PMI_TO_L1_DIS; + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_CNTL, data); + + if (!disable_plloff_in_l1) { + bool clk_req_support; + + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0); + data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK); + data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7); + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data); + + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1); + data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK); + data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7); + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0); + data &= ~(PLL_POWER_STATE_IN_OFF_0_MASK | PLL_POWER_STATE_IN_TXS2_0_MASK); + data |= PLL_POWER_STATE_IN_OFF_0(7) | PLL_POWER_STATE_IN_TXS2_0(7); + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1); + data &= ~(PLL_POWER_STATE_IN_OFF_1_MASK | PLL_POWER_STATE_IN_TXS2_1_MASK); + data |= PLL_POWER_STATE_IN_OFF_1(7) | PLL_POWER_STATE_IN_TXS2_1(7); + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data); + + if ((rdev->family != CHIP_OLAND) && (rdev->family != CHIP_HAINAN)) { + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0); + data &= ~PLL_RAMP_UP_TIME_0_MASK; + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_0, data); + + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1); + data &= ~PLL_RAMP_UP_TIME_1_MASK; + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_1, data); + + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_2); + data &= ~PLL_RAMP_UP_TIME_2_MASK; + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_2, data); + + orig = data = RREG32_PIF_PHY0(PB0_PIF_PWRDOWN_3); + data &= ~PLL_RAMP_UP_TIME_3_MASK; + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_PWRDOWN_3, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0); + data &= ~PLL_RAMP_UP_TIME_0_MASK; + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_0, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1); + data &= ~PLL_RAMP_UP_TIME_1_MASK; + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_1, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_2); + data &= ~PLL_RAMP_UP_TIME_2_MASK; + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_2, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_PWRDOWN_3); + data &= ~PLL_RAMP_UP_TIME_3_MASK; + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_PWRDOWN_3, data); + } + orig = data = RREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL); + data &= ~LC_DYN_LANES_PWR_STATE_MASK; + data |= LC_DYN_LANES_PWR_STATE(3); + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_LINK_WIDTH_CNTL, data); + + orig = data = RREG32_PIF_PHY0(PB0_PIF_CNTL); + data &= ~LS2_EXIT_TIME_MASK; + if ((rdev->family == CHIP_OLAND) || (rdev->family == CHIP_HAINAN)) + data |= LS2_EXIT_TIME(5); + if (orig != data) + WREG32_PIF_PHY0(PB0_PIF_CNTL, data); + + orig = data = RREG32_PIF_PHY1(PB1_PIF_CNTL); + data &= ~LS2_EXIT_TIME_MASK; + if ((rdev->family == CHIP_OLAND) || (rdev->family == CHIP_HAINAN)) + data |= LS2_EXIT_TIME(5); + if (orig != data) + WREG32_PIF_PHY1(PB1_PIF_CNTL, data); + + if (!disable_clkreq) { +#ifdef MN_TODO + struct pci_dev *root = rdev->pdev->bus->self; + u32 lnkcap; + + clk_req_support = false; + pcie_capability_read_dword(root, PCI_EXP_LNKCAP, &lnkcap); + if (lnkcap & PCI_EXP_LNKCAP_CLKPM) + clk_req_support = true; +#else + clk_req_support = false; +#endif + } else { + clk_req_support = false; + } + + if (clk_req_support) { + orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL2); + data |= LC_ALLOW_PDWN_IN_L1 | LC_ALLOW_PDWN_IN_L23; + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_CNTL2, data); + + orig = data = RREG32(THM_CLK_CNTL); + data &= ~(CMON_CLK_SEL_MASK | TMON_CLK_SEL_MASK); + data |= CMON_CLK_SEL(1) | TMON_CLK_SEL(1); + if (orig != data) + WREG32(THM_CLK_CNTL, data); + + orig = data = RREG32(MISC_CLK_CNTL); + data &= ~(DEEP_SLEEP_CLK_SEL_MASK | ZCLK_SEL_MASK); + data |= DEEP_SLEEP_CLK_SEL(1) | ZCLK_SEL(1); + if (orig != data) + WREG32(MISC_CLK_CNTL, data); + + orig = data = RREG32(CG_CLKPIN_CNTL); + data &= ~BCLK_AS_XCLK; + if (orig != data) + WREG32(CG_CLKPIN_CNTL, data); + + orig = data = RREG32(CG_CLKPIN_CNTL_2); + data &= ~FORCE_BIF_REFCLK_EN; + if (orig != data) + WREG32(CG_CLKPIN_CNTL_2, data); + + orig = data = RREG32(MPLL_BYPASSCLK_SEL); + data &= ~MPLL_CLKOUT_SEL_MASK; + data |= MPLL_CLKOUT_SEL(4); + if (orig != data) + WREG32(MPLL_BYPASSCLK_SEL, data); + + orig = data = RREG32(SPLL_CNTL_MODE); + data &= ~SPLL_REFCLK_SEL_MASK; + if (orig != data) + WREG32(SPLL_CNTL_MODE, data); + } + } + } else { + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_CNTL, data); + } + + orig = data = RREG32_PCIE(PCIE_CNTL2); + data |= SLV_MEM_LS_EN | MST_MEM_LS_EN | REPLAY_MEM_LS_EN; + if (orig != data) + WREG32_PCIE(PCIE_CNTL2, data); + + if (!disable_l0s) { + data = RREG32_PCIE_PORT(PCIE_LC_N_FTS_CNTL); + if((data & LC_N_FTS_MASK) == LC_N_FTS_MASK) { + data = RREG32_PCIE(PCIE_LC_STATUS1); + if ((data & LC_REVERSE_XMIT) && (data & LC_REVERSE_RCVR)) { + orig = data = RREG32_PCIE_PORT(PCIE_LC_CNTL); + data &= ~LC_L0S_INACTIVITY_MASK; + if (orig != data) + WREG32_PCIE_PORT(PCIE_LC_CNTL, data); + } + } + } +} diff --git a/sys/dev/drm/radeon/si_blit_shaders.c b/sys/dev/drm/radeon/si_blit_shaders.c index 6798cbdde4..8e7017b6a6 100644 --- a/sys/dev/drm/radeon/si_blit_shaders.c +++ b/sys/dev/drm/radeon/si_blit_shaders.c @@ -22,8 +22,6 @@ * * Authors: * Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/si_blit_shaders.c 254885 2013-08-25 19:37:15Z dumbbell $ */ #include diff --git a/sys/dev/drm/radeon/si_blit_shaders.h b/sys/dev/drm/radeon/si_blit_shaders.h index 24de9c413f..c739e51e39 100644 --- a/sys/dev/drm/radeon/si_blit_shaders.h +++ b/sys/dev/drm/radeon/si_blit_shaders.h @@ -20,7 +20,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $FreeBSD: head/sys/dev/drm2/radeon/si_blit_shaders.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef SI_BLIT_SHADERS_H diff --git a/sys/dev/drm/radeon/si_dpm.c b/sys/dev/drm/radeon/si_dpm.c new file mode 100644 index 0000000000..671a9ba1aa --- /dev/null +++ b/sys/dev/drm/radeon/si_dpm.c @@ -0,0 +1,6441 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "sid.h" +#include "r600_dpm.h" +#include "si_dpm.h" +#include "atom.h" +#include +#include + +#define MC_CG_ARB_FREQ_F0 0x0a +#define MC_CG_ARB_FREQ_F1 0x0b +#define MC_CG_ARB_FREQ_F2 0x0c +#define MC_CG_ARB_FREQ_F3 0x0d + +#define SMC_RAM_END 0x20000 + +#define SCLK_MIN_DEEPSLEEP_FREQ 1350 + +void si_dpm_reset_asic(struct radeon_device *rdev); + +static const struct si_cac_config_reg cac_weights_tahiti[] = +{ + { 0x0, 0x0000ffff, 0, 0xc, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x101, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0xc, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x8fc, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x95, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x34e, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x1a1, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0xda, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x46, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x208, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0xe7, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x948, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x167, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x31, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x18e, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg lcac_tahiti[] = +{ + { 0x143, 0x0001fffe, 1, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x0001fffe, 1, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x149, 0x0001fffe, 1, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0x149, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14c, 0x0001fffe, 1, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0x14c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x9e, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x9e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x101, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x101, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10a, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10d, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x92, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x92, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x95, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x95, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x155, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x155, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x158, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x158, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x116, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x116, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x119, 0x0001fffe, 1, 0x8, SISLANDS_CACCONFIG_CGIND }, + { 0x119, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x122, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x122, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x125, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x125, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x128, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x128, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x12b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x12b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } + +}; + +static const struct si_cac_config_reg cac_override_tahiti[] = +{ + { 0xFFFFFFFF } +}; + +static const struct si_powertune_data powertune_data_tahiti = +{ + ((1 << 16) | 27027), + 6, + 0, + 4, + 95, + { + 0UL, + 0UL, + 4521550UL, + 309631529UL, + -1270850L, + 4513710L, + 40 + }, + 595000000UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +static const struct si_dte_data dte_data_tahiti = +{ + { 1159409, 0, 0, 0, 0 }, + { 777, 0, 0, 0, 0 }, + 2, + 54000, + 127000, + 25, + 2, + 10, + 13, + { 27, 31, 35, 39, 43, 47, 54, 61, 67, 74, 81, 88, 95, 0, 0, 0 }, + { 240888759, 221057860, 235370597, 162287531, 158510299, 131423027, 116673180, 103067515, 87941937, 76209048, 68209175, 64090048, 58301890, 0, 0, 0 }, + { 12024, 11189, 11451, 8411, 7939, 6666, 5681, 4905, 4241, 3720, 3354, 3122, 2890, 0, 0, 0 }, + 85, + false +}; + +static const struct si_dte_data dte_data_tahiti_le = +{ + { 0x1E8480, 0x7A1200, 0x2160EC0, 0x3938700, 0 }, + { 0x7D, 0x7D, 0x4E4, 0xB00, 0 }, + 0x5, + 0xAFC8, + 0x64, + 0x32, + 1, + 0, + 0x10, + { 0x78, 0x7C, 0x82, 0x88, 0x8E, 0x94, 0x9A, 0xA0, 0xA6, 0xAC, 0xB0, 0xB4, 0xB8, 0xBC, 0xC0, 0xC4 }, + { 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700, 0x3938700 }, + { 0x2AF8, 0x2AF8, 0x29BB, 0x27F9, 0x2637, 0x2475, 0x22B3, 0x20F1, 0x1F2F, 0x1D6D, 0x1734, 0x1414, 0x10F4, 0xDD4, 0xAB4, 0x794 }, + 85, + true +}; + +static const struct si_dte_data dte_data_tahiti_pro = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x7D0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_new_zealand = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0 }, + { 0x29B, 0x3E9, 0x537, 0x7D2, 0 }, + 0x5, + 0xAFC8, + 0x69, + 0x32, + 1, + 0, + 0x10, + { 0x82, 0xA0, 0xB4, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE, 0xFE }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0xDAC, 0x1388, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685, 0x685 }, + 85, + true +}; + +static const struct si_dte_data dte_data_aruba_pro = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x1000, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_malta = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +struct si_cac_config_reg cac_weights_pitcairn[] = +{ + { 0x0, 0x0000ffff, 0, 0x8a, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x24d, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x19, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x76, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0xc11, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x7f3, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x403, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x367, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x4c9, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x45d, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x36d, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x534, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x5da, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x880, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0x201, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x9f, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x1f, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5de, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x7b, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x13, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0xf9, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x66, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x13, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x186, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg lcac_pitcairn[] = +{ + { 0x98, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x9e, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x9e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10a, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x116, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x116, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x155, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x155, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x92, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x92, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x149, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x149, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x101, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x101, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10d, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x119, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x119, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x158, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x158, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x95, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x95, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x14c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x122, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x122, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x125, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x125, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x128, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x128, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x12b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x12b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_override_pitcairn[] = +{ + { 0xFFFFFFFF } +}; + +static const struct si_powertune_data powertune_data_pitcairn = +{ + ((1 << 16) | 27027), + 5, + 0, + 6, + 100, + { + 51600000UL, + 1800000UL, + 7194395UL, + 309631529UL, + -1270850L, + 4513710L, + 100 + }, + 117830498UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +static const struct si_dte_data dte_data_pitcairn = +{ + { 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0 }, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + 0, + false +}; + +static const struct si_dte_data dte_data_curacao_xt = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x1D17, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_curacao_pro = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x1D17, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_neptune_xt = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 45000, + 100, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0x3A2F, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_cac_config_reg cac_weights_chelsea_pro[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x2BD, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_chelsea_xt[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x30A, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_heathrow[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x362, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_cape_verde_pro[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x315, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_cape_verde[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x3BA, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg lcac_cape_verde[] = +{ + { 0x98, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x9b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x107, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x113, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x0001fffe, 1, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x152, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x146, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_override_cape_verde[] = +{ + { 0xFFFFFFFF } +}; + +static const struct si_powertune_data powertune_data_cape_verde = +{ + ((1 << 16) | 0x6993), + 5, + 0, + 7, + 105, + { + 0UL, + 0UL, + 7194395UL, + 309631529UL, + -1270850L, + 4513710L, + 100 + }, + 117830498UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +static const struct si_dte_data dte_data_cape_verde = +{ + { 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0 }, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + 0, + false +}; + +static const struct si_dte_data dte_data_venus_xtx = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x71C, 0xAAB, 0xE39, 0x11C7, 0x0 }, + 5, + 55000, + 0x69, + 0xA, + 1, + 0, + 0x3, + { 0x96, 0xB4, 0xFF, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0x895440, 0x3D0900, 0x989680, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0xD6D8, 0x88B8, 0x1555, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_venus_xt = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0xBDA, 0x11C7, 0x17B4, 0x1DA1, 0x0 }, + 5, + 55000, + 0x69, + 0xA, + 1, + 0, + 0x3, + { 0x96, 0xB4, 0xFF, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0x895440, 0x3D0900, 0x989680, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0xAFC8, 0x88B8, 0x238E, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_venus_pro = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x11C7, 0x1AAB, 0x238E, 0x2C72, 0x0 }, + 5, + 55000, + 0x69, + 0xA, + 1, + 0, + 0x3, + { 0x96, 0xB4, 0xFF, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0x895440, 0x3D0900, 0x989680, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + { 0x88B8, 0x88B8, 0x3555, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +struct si_cac_config_reg cac_weights_oland[] = +{ + { 0x0, 0x0000ffff, 0, 0x82, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x153, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x52, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x4F, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x135, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0xAC, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x118, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0xBE, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x110, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x4CD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x37, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x27, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0xC3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x35, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0x28, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x26C, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3B2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x99D, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA3F, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0xA, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x34, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x3BA, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x30, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7A, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x100, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_mars_pro[] = +{ + { 0x0, 0x0000ffff, 0, 0x43, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0xAF, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x2A, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x59, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x1A5, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x1D6, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x2A3, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x8FD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x76, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x8A, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0xA3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x71, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0xA6, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x81, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3D2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x27C, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA96, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0xB, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x32, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7E, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x280, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x3C, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x203, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0xB4, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_mars_xt[] = +{ + { 0x0, 0x0000ffff, 0, 0x43, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0xAF, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x2A, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x59, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x1A5, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x1D6, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x2A3, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x8FD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x76, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x8A, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0xA3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x71, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0xA6, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x81, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3D2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x27C, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA96, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0xB, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x60, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x32, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7E, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x280, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x3C, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x203, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0xB4, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_oland_pro[] = +{ + { 0x0, 0x0000ffff, 0, 0x43, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0xAF, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x2A, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x59, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x1A5, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x1D6, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x2A3, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x8FD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x76, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x8A, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0xA3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x71, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0xA6, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x81, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3D2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x27C, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA96, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0xB, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x90, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x32, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7E, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x280, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x3C, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x203, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0xB4, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_weights_oland_xt[] = +{ + { 0x0, 0x0000ffff, 0, 0x43, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0xAF, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x2A, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x29, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0xA0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x59, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x1A5, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x1D6, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0x2A3, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x8FD, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x76, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x8A, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0xA3, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x71, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0xA6, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x81, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0x3D2, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0x27C, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xA96, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x5, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0xB, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x15, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x36, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x10, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x120, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x32, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x7E, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0x280, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0x7, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0x3C, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0x203, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0xB4, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg lcac_oland[] = +{ + { 0x98, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x0001fffe, 1, 0x4, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg lcac_mars_pro[] = +{ + { 0x98, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x98, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x104, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x110, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x14f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x0001fffe, 1, 0x6, SISLANDS_CACCONFIG_CGIND }, + { 0x8c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x143, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x11f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x164, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x167, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16a, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15e, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x161, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x15b, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x0001fffe, 1, 0x2, SISLANDS_CACCONFIG_CGIND }, + { 0x16d, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x170, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x173, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x176, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x179, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17c, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x0001fffe, 1, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0x17f, 0x00000001, 0, 0x1, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_cac_config_reg cac_override_oland[] = +{ + { 0xFFFFFFFF } +}; + +static const struct si_powertune_data powertune_data_oland = +{ + ((1 << 16) | 0x6993), + 5, + 0, + 7, + 105, + { + 0UL, + 0UL, + 7194395UL, + 309631529UL, + -1270850L, + 4513710L, + 100 + }, + 117830498UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +static const struct si_powertune_data powertune_data_mars_pro = +{ + ((1 << 16) | 0x6993), + 5, + 0, + 7, + 105, + { + 0UL, + 0UL, + 7194395UL, + 309631529UL, + -1270850L, + 4513710L, + 100 + }, + 117830498UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +static const struct si_dte_data dte_data_oland = +{ + { 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0 }, + 0, + 0, + 0, + 0, + 0, + 0, + 0, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, + 0, + false +}; + +static const struct si_dte_data dte_data_mars_pro = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 55000, + 105, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0xF627, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + +static const struct si_dte_data dte_data_sun_xt = +{ + { 0x1E8480, 0x3D0900, 0x989680, 0x2625A00, 0x0 }, + { 0x0, 0x0, 0x0, 0x0, 0x0 }, + 5, + 55000, + 105, + 0xA, + 1, + 0, + 0x10, + { 0x96, 0xB4, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF }, + { 0x895440, 0x3D0900, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680, 0x989680 }, + { 0xD555, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }, + 90, + true +}; + + +static const struct si_cac_config_reg cac_weights_hainan[] = +{ + { 0x0, 0x0000ffff, 0, 0x2d9, SISLANDS_CACCONFIG_CGIND }, + { 0x0, 0xffff0000, 16, 0x22b, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0x0000ffff, 0, 0x21c, SISLANDS_CACCONFIG_CGIND }, + { 0x1, 0xffff0000, 16, 0x1dc, SISLANDS_CACCONFIG_CGIND }, + { 0x2, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0x0000ffff, 0, 0x24e, SISLANDS_CACCONFIG_CGIND }, + { 0x3, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0x0000ffff, 0, 0x35e, SISLANDS_CACCONFIG_CGIND }, + { 0x5, 0xffff0000, 16, 0x1143, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0x0000ffff, 0, 0xe17, SISLANDS_CACCONFIG_CGIND }, + { 0x6, 0xffff0000, 16, 0x441, SISLANDS_CACCONFIG_CGIND }, + { 0x18f, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0x0000ffff, 0, 0x28b, SISLANDS_CACCONFIG_CGIND }, + { 0x7, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x8, 0xffff0000, 16, 0xabe, SISLANDS_CACCONFIG_CGIND }, + { 0x9, 0x0000ffff, 0, 0xf11, SISLANDS_CACCONFIG_CGIND }, + { 0xa, 0x0000ffff, 0, 0x907, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0x0000ffff, 0, 0xb45, SISLANDS_CACCONFIG_CGIND }, + { 0xb, 0xffff0000, 16, 0xd1e, SISLANDS_CACCONFIG_CGIND }, + { 0xc, 0x0000ffff, 0, 0xa2c, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0x0000ffff, 0, 0x62, SISLANDS_CACCONFIG_CGIND }, + { 0xd, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0xe, 0x0000ffff, 0, 0x1f3, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0x0000ffff, 0, 0x42, SISLANDS_CACCONFIG_CGIND }, + { 0xf, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x10, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0x0000ffff, 0, 0x709, SISLANDS_CACCONFIG_CGIND }, + { 0x11, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x12, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x13, 0xffff0000, 16, 0x3a, SISLANDS_CACCONFIG_CGIND }, + { 0x14, 0x0000ffff, 0, 0x357, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0x0000ffff, 0, 0x9f, SISLANDS_CACCONFIG_CGIND }, + { 0x15, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x4e, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0x0000ffff, 0, 0x314, SISLANDS_CACCONFIG_CGIND }, + { 0x16, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x17, 0x0000ffff, 0, 0x6d, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x18, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0x0000ffff, 0, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x19, 0xffff0000, 16, 0x0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1a, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1b, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1c, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1d, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1e, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x1f, 0xffff0000, 16, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x20, 0x0000ffff, 0, 0, SISLANDS_CACCONFIG_CGIND }, + { 0x6d, 0x0000ffff, 0, 0x1b9, SISLANDS_CACCONFIG_CGIND }, + { 0xFFFFFFFF } +}; + +static const struct si_powertune_data powertune_data_hainan = +{ + ((1 << 16) | 0x6993), + 5, + 0, + 9, + 105, + { + 0UL, + 0UL, + 7194395UL, + 309631529UL, + -1270850L, + 4513710L, + 100 + }, + 117830498UL, + 12, + { + 0, + 0, + 0, + 0, + 0, + 0, + 0, + 0 + }, + true +}; + +struct rv7xx_power_info *rv770_get_pi(struct radeon_device *rdev); +struct evergreen_power_info *evergreen_get_pi(struct radeon_device *rdev); +struct ni_power_info *ni_get_pi(struct radeon_device *rdev); +struct ni_ps *ni_get_ps(struct radeon_ps *rps); + +static int si_populate_voltage_value(struct radeon_device *rdev, + const struct atom_voltage_table *table, + u16 value, SISLANDS_SMC_VOLTAGE_VALUE *voltage); +static int si_get_std_voltage_value(struct radeon_device *rdev, + SISLANDS_SMC_VOLTAGE_VALUE *voltage, + u16 *std_voltage); +static int si_write_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 value); +static int si_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SISLANDS_SMC_HW_PERFORMANCE_LEVEL *level); +static int si_calculate_sclk_params(struct radeon_device *rdev, + u32 engine_clock, + SISLANDS_SMC_SCLK_VALUE *sclk); + +static struct si_power_info *si_get_pi(struct radeon_device *rdev) +{ + struct si_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void si_calculate_leakage_for_v_and_t_formula(const struct ni_leakage_coeffients *coeff, + u16 v, s32 t, u32 ileakage, u32 *leakage) +{ + s64 kt, kv, leakage_w, i_leakage, vddc; + s64 temperature, t_slope, t_intercept, av, bv, t_ref; + s64 tmp; + + i_leakage = div64_s64(drm_int2fixp(ileakage), 100); + vddc = div64_s64(drm_int2fixp(v), 1000); + temperature = div64_s64(drm_int2fixp(t), 1000); + + t_slope = div64_s64(drm_int2fixp(coeff->t_slope), 100000000); + t_intercept = div64_s64(drm_int2fixp(coeff->t_intercept), 100000000); + av = div64_s64(drm_int2fixp(coeff->av), 100000000); + bv = div64_s64(drm_int2fixp(coeff->bv), 100000000); + t_ref = drm_int2fixp(coeff->t_ref); + + tmp = drm_fixp_mul(t_slope, vddc) + t_intercept; + kt = drm_fixp_exp(drm_fixp_mul(tmp, temperature)); + kt = drm_fixp_div(kt, drm_fixp_exp(drm_fixp_mul(tmp, t_ref))); + kv = drm_fixp_mul(av, drm_fixp_exp(drm_fixp_mul(bv, vddc))); + + leakage_w = drm_fixp_mul(drm_fixp_mul(drm_fixp_mul(i_leakage, kt), kv), vddc); + + *leakage = drm_fixp2int(leakage_w * 1000); +} + +static void si_calculate_leakage_for_v_and_t(struct radeon_device *rdev, + const struct ni_leakage_coeffients *coeff, + u16 v, + s32 t, + u32 i_leakage, + u32 *leakage) +{ + si_calculate_leakage_for_v_and_t_formula(coeff, v, t, i_leakage, leakage); +} + +static void si_calculate_leakage_for_v_formula(const struct ni_leakage_coeffients *coeff, + const u32 fixed_kt, u16 v, + u32 ileakage, u32 *leakage) +{ + s64 kt, kv, leakage_w, i_leakage, vddc; + + i_leakage = div64_s64(drm_int2fixp(ileakage), 100); + vddc = div64_s64(drm_int2fixp(v), 1000); + + kt = div64_s64(drm_int2fixp(fixed_kt), 100000000); + kv = drm_fixp_mul(div64_s64(drm_int2fixp(coeff->av), 100000000), + drm_fixp_exp(drm_fixp_mul(div64_s64(drm_int2fixp(coeff->bv), 100000000), vddc))); + + leakage_w = drm_fixp_mul(drm_fixp_mul(drm_fixp_mul(i_leakage, kt), kv), vddc); + + *leakage = drm_fixp2int(leakage_w * 1000); +} + +static void si_calculate_leakage_for_v(struct radeon_device *rdev, + const struct ni_leakage_coeffients *coeff, + const u32 fixed_kt, + u16 v, + u32 i_leakage, + u32 *leakage) +{ + si_calculate_leakage_for_v_formula(coeff, fixed_kt, v, i_leakage, leakage); +} + + +static void si_update_dte_from_pl2(struct radeon_device *rdev, + struct si_dte_data *dte_data) +{ + u32 p_limit1 = rdev->pm.dpm.tdp_limit; + u32 p_limit2 = rdev->pm.dpm.near_tdp_limit; + u32 k = dte_data->k; + u32 t_max = dte_data->max_t; + u32 t_split[5] = { 10, 15, 20, 25, 30 }; + u32 t_0 = dte_data->t0; + u32 i; + + if (p_limit2 != 0 && p_limit2 <= p_limit1) { + dte_data->tdep_count = 3; + + for (i = 0; i < k; i++) { + dte_data->r[i] = + (t_split[i] * (t_max - t_0/(u32)1000) * (1 << 14)) / + (p_limit2 * (u32)100); + } + + dte_data->tdep_r[1] = dte_data->r[4] * 2; + + for (i = 2; i < SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE; i++) { + dte_data->tdep_r[i] = dte_data->r[4]; + } + } else { + DRM_ERROR("Invalid PL2! DTE will not be updated.\n"); + } +} + +static void si_initialize_powertune_defaults(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + bool update_dte_from_pl2 = false; + + if (rdev->family == CHIP_TAHITI) { + si_pi->cac_weights = cac_weights_tahiti; + si_pi->lcac_config = lcac_tahiti; + si_pi->cac_override = cac_override_tahiti; + si_pi->powertune_data = &powertune_data_tahiti; + si_pi->dte_data = dte_data_tahiti; + + switch (rdev->ddev->pci_device) { + case 0x6798: + si_pi->dte_data.enable_dte_by_default = true; + break; + case 0x6799: + si_pi->dte_data = dte_data_new_zealand; + break; + case 0x6790: + case 0x6791: + case 0x6792: + case 0x679E: + si_pi->dte_data = dte_data_aruba_pro; + update_dte_from_pl2 = true; + break; + case 0x679B: + si_pi->dte_data = dte_data_malta; + update_dte_from_pl2 = true; + break; + case 0x679A: + si_pi->dte_data = dte_data_tahiti_pro; + update_dte_from_pl2 = true; + break; + default: + if (si_pi->dte_data.enable_dte_by_default == true) + DRM_ERROR("DTE is not enabled!\n"); + break; + } + } else if (rdev->family == CHIP_PITCAIRN) { + switch (rdev->ddev->pci_device) { + case 0x6810: + case 0x6818: + si_pi->cac_weights = cac_weights_pitcairn; + si_pi->lcac_config = lcac_pitcairn; + si_pi->cac_override = cac_override_pitcairn; + si_pi->powertune_data = &powertune_data_pitcairn; + si_pi->dte_data = dte_data_curacao_xt; + update_dte_from_pl2 = true; + break; + case 0x6819: + case 0x6811: + si_pi->cac_weights = cac_weights_pitcairn; + si_pi->lcac_config = lcac_pitcairn; + si_pi->cac_override = cac_override_pitcairn; + si_pi->powertune_data = &powertune_data_pitcairn; + si_pi->dte_data = dte_data_curacao_pro; + update_dte_from_pl2 = true; + break; + case 0x6800: + case 0x6806: + si_pi->cac_weights = cac_weights_pitcairn; + si_pi->lcac_config = lcac_pitcairn; + si_pi->cac_override = cac_override_pitcairn; + si_pi->powertune_data = &powertune_data_pitcairn; + si_pi->dte_data = dte_data_neptune_xt; + update_dte_from_pl2 = true; + break; + default: + si_pi->cac_weights = cac_weights_pitcairn; + si_pi->lcac_config = lcac_pitcairn; + si_pi->cac_override = cac_override_pitcairn; + si_pi->powertune_data = &powertune_data_pitcairn; + si_pi->dte_data = dte_data_pitcairn; + break; + } + } else if (rdev->family == CHIP_VERDE) { + si_pi->lcac_config = lcac_cape_verde; + si_pi->cac_override = cac_override_cape_verde; + si_pi->powertune_data = &powertune_data_cape_verde; + + switch (rdev->ddev->pci_device) { + case 0x683B: + case 0x683F: + case 0x6829: + case 0x6835: + si_pi->cac_weights = cac_weights_cape_verde_pro; + si_pi->dte_data = dte_data_cape_verde; + break; + case 0x6825: + case 0x6827: + si_pi->cac_weights = cac_weights_heathrow; + si_pi->dte_data = dte_data_cape_verde; + break; + case 0x6824: + case 0x682D: + si_pi->cac_weights = cac_weights_chelsea_xt; + si_pi->dte_data = dte_data_cape_verde; + break; + case 0x682F: + si_pi->cac_weights = cac_weights_chelsea_pro; + si_pi->dte_data = dte_data_cape_verde; + break; + case 0x6820: + si_pi->cac_weights = cac_weights_heathrow; + si_pi->dte_data = dte_data_venus_xtx; + break; + case 0x6821: + si_pi->cac_weights = cac_weights_heathrow; + si_pi->dte_data = dte_data_venus_xt; + break; + case 0x6823: + si_pi->cac_weights = cac_weights_chelsea_pro; + si_pi->dte_data = dte_data_venus_pro; + break; + case 0x682B: + si_pi->cac_weights = cac_weights_chelsea_pro; + si_pi->dte_data = dte_data_venus_pro; + break; + default: + si_pi->cac_weights = cac_weights_cape_verde; + si_pi->dte_data = dte_data_cape_verde; + break; + } + } else if (rdev->family == CHIP_OLAND) { + switch (rdev->ddev->pci_device) { + case 0x6601: + case 0x6621: + case 0x6603: + si_pi->cac_weights = cac_weights_mars_pro; + si_pi->lcac_config = lcac_mars_pro; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_mars_pro; + si_pi->dte_data = dte_data_mars_pro; + update_dte_from_pl2 = true; + break; + case 0x6600: + case 0x6606: + case 0x6620: + si_pi->cac_weights = cac_weights_mars_xt; + si_pi->lcac_config = lcac_mars_pro; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_mars_pro; + si_pi->dte_data = dte_data_mars_pro; + update_dte_from_pl2 = true; + break; + case 0x6611: + si_pi->cac_weights = cac_weights_oland_pro; + si_pi->lcac_config = lcac_mars_pro; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_mars_pro; + si_pi->dte_data = dte_data_mars_pro; + update_dte_from_pl2 = true; + break; + case 0x6610: + si_pi->cac_weights = cac_weights_oland_xt; + si_pi->lcac_config = lcac_mars_pro; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_mars_pro; + si_pi->dte_data = dte_data_mars_pro; + update_dte_from_pl2 = true; + break; + default: + si_pi->cac_weights = cac_weights_oland; + si_pi->lcac_config = lcac_oland; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_oland; + si_pi->dte_data = dte_data_oland; + break; + } + } else if (rdev->family == CHIP_HAINAN) { + si_pi->cac_weights = cac_weights_hainan; + si_pi->lcac_config = lcac_oland; + si_pi->cac_override = cac_override_oland; + si_pi->powertune_data = &powertune_data_hainan; + si_pi->dte_data = dte_data_sun_xt; + update_dte_from_pl2 = true; + } else { + DRM_ERROR("Unknown SI asic revision, failed to initialize PowerTune!\n"); + return; + } + + ni_pi->enable_power_containment = false; + ni_pi->enable_cac = false; + ni_pi->enable_sq_ramping = false; + si_pi->enable_dte = false; + + if (si_pi->powertune_data->enable_powertune_by_default) { + ni_pi->enable_power_containment= true; + ni_pi->enable_cac = true; + if (si_pi->dte_data.enable_dte_by_default) { + si_pi->enable_dte = true; + if (update_dte_from_pl2) + si_update_dte_from_pl2(rdev, &si_pi->dte_data); + + } + ni_pi->enable_sq_ramping = true; + } + + ni_pi->driver_calculate_cac_leakage = true; + ni_pi->cac_configuration_required = true; + + if (ni_pi->cac_configuration_required) { + ni_pi->support_cac_long_term_average = true; + si_pi->dyn_powertune_data.l2_lta_window_size = + si_pi->powertune_data->l2_lta_window_size_default; + si_pi->dyn_powertune_data.lts_truncate = + si_pi->powertune_data->lts_truncate_default; + } else { + ni_pi->support_cac_long_term_average = false; + si_pi->dyn_powertune_data.l2_lta_window_size = 0; + si_pi->dyn_powertune_data.lts_truncate = 0; + } + + si_pi->dyn_powertune_data.disable_uvd_powertune = false; +} + +static u32 si_get_smc_power_scaling_factor(struct radeon_device *rdev) +{ + return 1; +} + +static u32 si_calculate_cac_wintime(struct radeon_device *rdev) +{ + u32 xclk; + u32 wintime; + u32 cac_window; + u32 cac_window_size; + + xclk = radeon_get_xclk(rdev); + + if (xclk == 0) + return 0; + + cac_window = RREG32(CG_CAC_CTRL) & CAC_WINDOW_MASK; + cac_window_size = ((cac_window & 0xFFFF0000) >> 16) * (cac_window & 0x0000FFFF); + + wintime = (cac_window_size * 100) / xclk; + + return wintime; +} + +static u32 si_scale_power_for_smc(u32 power_in_watts, u32 scaling_factor) +{ + return power_in_watts; +} + +static int si_calculate_adjusted_tdp_limits(struct radeon_device *rdev, + bool adjust_polarity, + u32 tdp_adjustment, + u32 *tdp_limit, + u32 *near_tdp_limit) +{ + u32 adjustment_delta, max_tdp_limit; + + if (tdp_adjustment > (u32)rdev->pm.dpm.tdp_od_limit) + return -EINVAL; + + max_tdp_limit = ((100 + 100) * rdev->pm.dpm.tdp_limit) / 100; + + if (adjust_polarity) { + *tdp_limit = ((100 + tdp_adjustment) * rdev->pm.dpm.tdp_limit) / 100; + *near_tdp_limit = rdev->pm.dpm.near_tdp_limit_adjusted + (*tdp_limit - rdev->pm.dpm.tdp_limit); + } else { + *tdp_limit = ((100 - tdp_adjustment) * rdev->pm.dpm.tdp_limit) / 100; + adjustment_delta = rdev->pm.dpm.tdp_limit - *tdp_limit; + if (adjustment_delta < rdev->pm.dpm.near_tdp_limit_adjusted) + *near_tdp_limit = rdev->pm.dpm.near_tdp_limit_adjusted - adjustment_delta; + else + *near_tdp_limit = 0; + } + + if ((*tdp_limit <= 0) || (*tdp_limit > max_tdp_limit)) + return -EINVAL; + if ((*near_tdp_limit <= 0) || (*near_tdp_limit > *tdp_limit)) + return -EINVAL; + + return 0; +} + +static int si_populate_smc_tdp_limits(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + + if (ni_pi->enable_power_containment) { + SISLANDS_SMC_STATETABLE *smc_table = &si_pi->smc_statetable; + PP_SIslands_PAPMParameters *papm_parm; + struct radeon_ppm_table *ppm = rdev->pm.dpm.dyn_state.ppm_table; + u32 scaling_factor = si_get_smc_power_scaling_factor(rdev); + u32 tdp_limit; + u32 near_tdp_limit; + int ret; + + if (scaling_factor == 0) + return -EINVAL; + + memset(smc_table, 0, sizeof(SISLANDS_SMC_STATETABLE)); + + ret = si_calculate_adjusted_tdp_limits(rdev, + false, /* ??? */ + rdev->pm.dpm.tdp_adjustment, + &tdp_limit, + &near_tdp_limit); + if (ret) + return ret; + + smc_table->dpm2Params.TDPLimit = + cpu_to_be32(si_scale_power_for_smc(tdp_limit, scaling_factor) * 1000); + smc_table->dpm2Params.NearTDPLimit = + cpu_to_be32(si_scale_power_for_smc(near_tdp_limit, scaling_factor) * 1000); + smc_table->dpm2Params.SafePowerLimit = + cpu_to_be32(si_scale_power_for_smc((near_tdp_limit * SISLANDS_DPM2_TDP_SAFE_LIMIT_PERCENT) / 100, scaling_factor) * 1000); + + ret = si_copy_bytes_to_smc(rdev, + (si_pi->state_table_start + offsetof(SISLANDS_SMC_STATETABLE, dpm2Params) + + offsetof(PP_SIslands_DPM2Parameters, TDPLimit)), + (u8 *)(&(smc_table->dpm2Params.TDPLimit)), + sizeof(u32) * 3, + si_pi->sram_end); + if (ret) + return ret; + + if (si_pi->enable_ppm) { + papm_parm = &si_pi->papm_parm; + memset(papm_parm, 0, sizeof(PP_SIslands_PAPMParameters)); + papm_parm->NearTDPLimitTherm = cpu_to_be32(ppm->dgpu_tdp); + papm_parm->dGPU_T_Limit = cpu_to_be32(ppm->tj_max); + papm_parm->dGPU_T_Warning = cpu_to_be32(95); + papm_parm->dGPU_T_Hysteresis = cpu_to_be32(5); + papm_parm->PlatformPowerLimit = 0xffffffff; + papm_parm->NearTDPLimitPAPM = 0xffffffff; + + ret = si_copy_bytes_to_smc(rdev, si_pi->papm_cfg_table_start, + (u8 *)papm_parm, + sizeof(PP_SIslands_PAPMParameters), + si_pi->sram_end); + if (ret) + return ret; + } + } + return 0; +} + +static int si_populate_smc_tdp_limits_2(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + + if (ni_pi->enable_power_containment) { + SISLANDS_SMC_STATETABLE *smc_table = &si_pi->smc_statetable; + u32 scaling_factor = si_get_smc_power_scaling_factor(rdev); + int ret; + + memset(smc_table, 0, sizeof(SISLANDS_SMC_STATETABLE)); + + smc_table->dpm2Params.NearTDPLimit = + cpu_to_be32(si_scale_power_for_smc(rdev->pm.dpm.near_tdp_limit_adjusted, scaling_factor) * 1000); + smc_table->dpm2Params.SafePowerLimit = + cpu_to_be32(si_scale_power_for_smc((rdev->pm.dpm.near_tdp_limit_adjusted * SISLANDS_DPM2_TDP_SAFE_LIMIT_PERCENT) / 100, scaling_factor) * 1000); + + ret = si_copy_bytes_to_smc(rdev, + (si_pi->state_table_start + + offsetof(SISLANDS_SMC_STATETABLE, dpm2Params) + + offsetof(PP_SIslands_DPM2Parameters, NearTDPLimit)), + (u8 *)(&(smc_table->dpm2Params.NearTDPLimit)), + sizeof(u32) * 2, + si_pi->sram_end); + if (ret) + return ret; + } + + return 0; +} + +static u16 si_calculate_power_efficiency_ratio(struct radeon_device *rdev, + const u16 prev_std_vddc, + const u16 curr_std_vddc) +{ + u64 margin = (u64)SISLANDS_DPM2_PWREFFICIENCYRATIO_MARGIN; + u64 prev_vddc = (u64)prev_std_vddc; + u64 curr_vddc = (u64)curr_std_vddc; + u64 pwr_efficiency_ratio, n, d; + + if ((prev_vddc == 0) || (curr_vddc == 0)) + return 0; + + n = div64_u64((u64)1024 * curr_vddc * curr_vddc * ((u64)1000 + margin), (u64)1000); + d = prev_vddc * prev_vddc; + pwr_efficiency_ratio = div64_u64(n, d); + + if (pwr_efficiency_ratio > (u64)0xFFFF) + return 0; + + return (u16)pwr_efficiency_ratio; +} + +static bool si_should_disable_uvd_powertune(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + + if (si_pi->dyn_powertune_data.disable_uvd_powertune && + radeon_state->vclk && radeon_state->dclk) + return true; + + return false; +} + +static int si_populate_power_containment_values(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SISLANDS_SMC_SWSTATE *smc_state) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + SISLANDS_SMC_VOLTAGE_VALUE vddc; + u32 prev_sclk; + u32 max_sclk; + u32 min_sclk; + u16 prev_std_vddc; + u16 curr_std_vddc; + int i; + u16 pwr_efficiency_ratio; + u8 max_ps_percent; + bool disable_uvd_power_tune; + int ret; + + if (ni_pi->enable_power_containment == false) + return 0; + + if (state->performance_level_count == 0) + return -EINVAL; + + if (smc_state->levelCount != state->performance_level_count) + return -EINVAL; + + disable_uvd_power_tune = si_should_disable_uvd_powertune(rdev, radeon_state); + + smc_state->levels[0].dpm2.MaxPS = 0; + smc_state->levels[0].dpm2.NearTDPDec = 0; + smc_state->levels[0].dpm2.AboveSafeInc = 0; + smc_state->levels[0].dpm2.BelowSafeInc = 0; + smc_state->levels[0].dpm2.PwrEfficiencyRatio = 0; + + for (i = 1; i < state->performance_level_count; i++) { + prev_sclk = state->performance_levels[i-1].sclk; + max_sclk = state->performance_levels[i].sclk; + if (i == 1) + max_ps_percent = SISLANDS_DPM2_MAXPS_PERCENT_M; + else + max_ps_percent = SISLANDS_DPM2_MAXPS_PERCENT_H; + + if (prev_sclk > max_sclk) + return -EINVAL; + + if ((max_ps_percent == 0) || + (prev_sclk == max_sclk) || + disable_uvd_power_tune) { + min_sclk = max_sclk; + } else if (i == 1) { + min_sclk = prev_sclk; + } else { + min_sclk = (prev_sclk * (u32)max_ps_percent) / 100; + } + + if (min_sclk < state->performance_levels[0].sclk) + min_sclk = state->performance_levels[0].sclk; + + if (min_sclk == 0) + return -EINVAL; + + ret = si_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + state->performance_levels[i-1].vddc, &vddc); + if (ret) + return ret; + + ret = si_get_std_voltage_value(rdev, &vddc, &prev_std_vddc); + if (ret) + return ret; + + ret = si_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + state->performance_levels[i].vddc, &vddc); + if (ret) + return ret; + + ret = si_get_std_voltage_value(rdev, &vddc, &curr_std_vddc); + if (ret) + return ret; + + pwr_efficiency_ratio = si_calculate_power_efficiency_ratio(rdev, + prev_std_vddc, curr_std_vddc); + + smc_state->levels[i].dpm2.MaxPS = (u8)((SISLANDS_DPM2_MAX_PULSE_SKIP * (max_sclk - min_sclk)) / max_sclk); + smc_state->levels[i].dpm2.NearTDPDec = SISLANDS_DPM2_NEAR_TDP_DEC; + smc_state->levels[i].dpm2.AboveSafeInc = SISLANDS_DPM2_ABOVE_SAFE_INC; + smc_state->levels[i].dpm2.BelowSafeInc = SISLANDS_DPM2_BELOW_SAFE_INC; + smc_state->levels[i].dpm2.PwrEfficiencyRatio = cpu_to_be16(pwr_efficiency_ratio); + } + + return 0; +} + +static int si_populate_sq_ramping_values(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SISLANDS_SMC_SWSTATE *smc_state) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + u32 sq_power_throttle, sq_power_throttle2; + bool enable_sq_ramping = ni_pi->enable_sq_ramping; + int i; + + if (state->performance_level_count == 0) + return -EINVAL; + + if (smc_state->levelCount != state->performance_level_count) + return -EINVAL; + + if (rdev->pm.dpm.sq_ramping_threshold == 0) + return -EINVAL; + + if (SISLANDS_DPM2_SQ_RAMP_MAX_POWER > (MAX_POWER_MASK >> MAX_POWER_SHIFT)) + enable_sq_ramping = false; + + if (SISLANDS_DPM2_SQ_RAMP_MIN_POWER > (MIN_POWER_MASK >> MIN_POWER_SHIFT)) + enable_sq_ramping = false; + + if (SISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA > (MAX_POWER_DELTA_MASK >> MAX_POWER_DELTA_SHIFT)) + enable_sq_ramping = false; + + if (SISLANDS_DPM2_SQ_RAMP_STI_SIZE > (STI_SIZE_MASK >> STI_SIZE_SHIFT)) + enable_sq_ramping = false; + + if (NISLANDS_DPM2_SQ_RAMP_LTI_RATIO <= (LTI_RATIO_MASK >> LTI_RATIO_SHIFT)) + enable_sq_ramping = false; + + for (i = 0; i < state->performance_level_count; i++) { + sq_power_throttle = 0; + sq_power_throttle2 = 0; + + if ((state->performance_levels[i].sclk >= rdev->pm.dpm.sq_ramping_threshold) && + enable_sq_ramping) { + sq_power_throttle |= MAX_POWER(SISLANDS_DPM2_SQ_RAMP_MAX_POWER); + sq_power_throttle |= MIN_POWER(SISLANDS_DPM2_SQ_RAMP_MIN_POWER); + sq_power_throttle2 |= MAX_POWER_DELTA(SISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA); + sq_power_throttle2 |= STI_SIZE(SISLANDS_DPM2_SQ_RAMP_STI_SIZE); + sq_power_throttle2 |= LTI_RATIO(SISLANDS_DPM2_SQ_RAMP_LTI_RATIO); + } else { + sq_power_throttle |= MAX_POWER_MASK | MIN_POWER_MASK; + sq_power_throttle2 |= MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + } + + smc_state->levels[i].SQPowerThrottle = cpu_to_be32(sq_power_throttle); + smc_state->levels[i].SQPowerThrottle_2 = cpu_to_be32(sq_power_throttle2); + } + + return 0; +} + +static int si_enable_power_containment(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + bool enable) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + PPSMC_Result smc_result; + int ret = 0; + + if (ni_pi->enable_power_containment) { + if (enable) { + if (!si_should_disable_uvd_powertune(rdev, radeon_new_state)) { + smc_result = si_send_msg_to_smc(rdev, PPSMC_TDPClampingActive); + if (smc_result != PPSMC_Result_OK) { + ret = -EINVAL; + ni_pi->pc_enabled = false; + } else { + ni_pi->pc_enabled = true; + } + } + } else { + smc_result = si_send_msg_to_smc(rdev, PPSMC_TDPClampingInactive); + if (smc_result != PPSMC_Result_OK) + ret = -EINVAL; + ni_pi->pc_enabled = false; + } + } + + return ret; +} + +static int si_initialize_smc_dte_tables(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + int ret = 0; + struct si_dte_data *dte_data = &si_pi->dte_data; + Smc_SIslands_DTE_Configuration *dte_tables = NULL; + u32 table_size; + u8 tdep_count; + u32 i; + + if (dte_data == NULL) + si_pi->enable_dte = false; + + if (si_pi->enable_dte == false) + return 0; + + if (dte_data->k <= 0) + return -EINVAL; + + dte_tables = kzalloc(sizeof(Smc_SIslands_DTE_Configuration), GFP_KERNEL); + if (dte_tables == NULL) { + si_pi->enable_dte = false; + return -ENOMEM; + } + + table_size = dte_data->k; + + if (table_size > SMC_SISLANDS_DTE_MAX_FILTER_STAGES) + table_size = SMC_SISLANDS_DTE_MAX_FILTER_STAGES; + + tdep_count = dte_data->tdep_count; + if (tdep_count > SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE) + tdep_count = SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE; + + dte_tables->K = cpu_to_be32(table_size); + dte_tables->T0 = cpu_to_be32(dte_data->t0); + dte_tables->MaxT = cpu_to_be32(dte_data->max_t); + dte_tables->WindowSize = dte_data->window_size; + dte_tables->temp_select = dte_data->temp_select; + dte_tables->DTE_mode = dte_data->dte_mode; + dte_tables->Tthreshold = cpu_to_be32(dte_data->t_threshold); + + if (tdep_count > 0) + table_size--; + + for (i = 0; i < table_size; i++) { + dte_tables->tau[i] = cpu_to_be32(dte_data->tau[i]); + dte_tables->R[i] = cpu_to_be32(dte_data->r[i]); + } + + dte_tables->Tdep_count = tdep_count; + + for (i = 0; i < (u32)tdep_count; i++) { + dte_tables->T_limits[i] = dte_data->t_limits[i]; + dte_tables->Tdep_tau[i] = cpu_to_be32(dte_data->tdep_tau[i]); + dte_tables->Tdep_R[i] = cpu_to_be32(dte_data->tdep_r[i]); + } + + ret = si_copy_bytes_to_smc(rdev, si_pi->dte_table_start, (u8 *)dte_tables, + sizeof(Smc_SIslands_DTE_Configuration), si_pi->sram_end); + kfree(dte_tables); + + return ret; +} + +static int si_get_cac_std_voltage_max_min(struct radeon_device *rdev, + u16 *max, u16 *min) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct radeon_cac_leakage_table *table = + &rdev->pm.dpm.dyn_state.cac_leakage_table; + u32 i; + u32 v0_loadline; + + + if (table == NULL) + return -EINVAL; + + *max = 0; + *min = 0xFFFF; + + for (i = 0; i < table->count; i++) { + if (table->entries[i].vddc > *max) + *max = table->entries[i].vddc; + if (table->entries[i].vddc < *min) + *min = table->entries[i].vddc; + } + + if (si_pi->powertune_data->lkge_lut_v0_percent > 100) + return -EINVAL; + + v0_loadline = (*min) * (100 - si_pi->powertune_data->lkge_lut_v0_percent) / 100; + + if (v0_loadline > 0xFFFFUL) + return -EINVAL; + + *min = (u16)v0_loadline; + + if ((*min > *max) || (*max == 0) || (*min == 0)) + return -EINVAL; + + return 0; +} + +static u16 si_get_cac_std_voltage_step(u16 max, u16 min) +{ + return ((max - min) + (SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES - 1)) / + SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; +} + +static int si_init_dte_leakage_table(struct radeon_device *rdev, + PP_SIslands_CacConfig *cac_tables, + u16 vddc_max, u16 vddc_min, u16 vddc_step, + u16 t0, u16 t_step) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 leakage; + unsigned int i, j; + s32 t; + u32 smc_leakage; + u32 scaling_factor; + u16 voltage; + + scaling_factor = si_get_smc_power_scaling_factor(rdev); + + for (i = 0; i < SMC_SISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES ; i++) { + t = (1000 * (i * t_step + t0)); + + for (j = 0; j < SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; j++) { + voltage = vddc_max - (vddc_step * j); + + si_calculate_leakage_for_v_and_t(rdev, + &si_pi->powertune_data->leakage_coefficients, + voltage, + t, + si_pi->dyn_powertune_data.cac_leakage, + &leakage); + + smc_leakage = si_scale_power_for_smc(leakage, scaling_factor) / 4; + + if (smc_leakage > 0xFFFF) + smc_leakage = 0xFFFF; + + cac_tables->cac_lkge_lut[i][SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES-1-j] = + cpu_to_be16((u16)smc_leakage); + } + } + return 0; +} + +static int si_init_simplified_leakage_table(struct radeon_device *rdev, + PP_SIslands_CacConfig *cac_tables, + u16 vddc_max, u16 vddc_min, u16 vddc_step) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 leakage; + unsigned int i, j; + u32 smc_leakage; + u32 scaling_factor; + u16 voltage; + + scaling_factor = si_get_smc_power_scaling_factor(rdev); + + for (j = 0; j < SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES; j++) { + voltage = vddc_max - (vddc_step * j); + + si_calculate_leakage_for_v(rdev, + &si_pi->powertune_data->leakage_coefficients, + si_pi->powertune_data->fixed_kt, + voltage, + si_pi->dyn_powertune_data.cac_leakage, + &leakage); + + smc_leakage = si_scale_power_for_smc(leakage, scaling_factor) / 4; + + if (smc_leakage > 0xFFFF) + smc_leakage = 0xFFFF; + + for (i = 0; i < SMC_SISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES ; i++) + cac_tables->cac_lkge_lut[i][SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES-1-j] = + cpu_to_be16((u16)smc_leakage); + } + return 0; +} + +static int si_initialize_smc_cac_tables(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + PP_SIslands_CacConfig *cac_tables = NULL; + u16 vddc_max, vddc_min, vddc_step; + u16 t0, t_step; + u32 load_line_slope, reg; + int ret = 0; + u32 ticks_per_us = radeon_get_xclk(rdev) / 100; + + if (ni_pi->enable_cac == false) + return 0; + + cac_tables = kzalloc(sizeof(PP_SIslands_CacConfig), GFP_KERNEL); + if (!cac_tables) + return -ENOMEM; + + reg = RREG32(CG_CAC_CTRL) & ~CAC_WINDOW_MASK; + reg |= CAC_WINDOW(si_pi->powertune_data->cac_window); + WREG32(CG_CAC_CTRL, reg); + + si_pi->dyn_powertune_data.cac_leakage = rdev->pm.dpm.cac_leakage; + si_pi->dyn_powertune_data.dc_pwr_value = + si_pi->powertune_data->dc_cac[NISLANDS_DCCAC_LEVEL_0]; + si_pi->dyn_powertune_data.wintime = si_calculate_cac_wintime(rdev); + si_pi->dyn_powertune_data.shift_n = si_pi->powertune_data->shift_n_default; + + si_pi->dyn_powertune_data.leakage_minimum_temperature = 80 * 1000; + + ret = si_get_cac_std_voltage_max_min(rdev, &vddc_max, &vddc_min); + if (ret) + goto done_free; + + vddc_step = si_get_cac_std_voltage_step(vddc_max, vddc_min); + vddc_min = vddc_max - (vddc_step * (SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES - 1)); + t_step = 4; + t0 = 60; + + if (si_pi->enable_dte || ni_pi->driver_calculate_cac_leakage) + ret = si_init_dte_leakage_table(rdev, cac_tables, + vddc_max, vddc_min, vddc_step, + t0, t_step); + else + ret = si_init_simplified_leakage_table(rdev, cac_tables, + vddc_max, vddc_min, vddc_step); + if (ret) + goto done_free; + + load_line_slope = ((u32)rdev->pm.dpm.load_line_slope << SMC_SISLANDS_SCALE_R) / 100; + + cac_tables->l2numWin_TDP = cpu_to_be32(si_pi->dyn_powertune_data.l2_lta_window_size); + cac_tables->lts_truncate_n = si_pi->dyn_powertune_data.lts_truncate; + cac_tables->SHIFT_N = si_pi->dyn_powertune_data.shift_n; + cac_tables->lkge_lut_V0 = cpu_to_be32((u32)vddc_min); + cac_tables->lkge_lut_Vstep = cpu_to_be32((u32)vddc_step); + cac_tables->R_LL = cpu_to_be32(load_line_slope); + cac_tables->WinTime = cpu_to_be32(si_pi->dyn_powertune_data.wintime); + cac_tables->calculation_repeats = cpu_to_be32(2); + cac_tables->dc_cac = cpu_to_be32(0); + cac_tables->log2_PG_LKG_SCALE = 12; + cac_tables->cac_temp = si_pi->powertune_data->operating_temp; + cac_tables->lkge_lut_T0 = cpu_to_be32((u32)t0); + cac_tables->lkge_lut_Tstep = cpu_to_be32((u32)t_step); + + ret = si_copy_bytes_to_smc(rdev, si_pi->cac_table_start, (u8 *)cac_tables, + sizeof(PP_SIslands_CacConfig), si_pi->sram_end); + + if (ret) + goto done_free; + + ret = si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_ticks_per_us, ticks_per_us); + +done_free: + if (ret) { + ni_pi->enable_cac = false; + ni_pi->enable_power_containment = false; + } + + kfree(cac_tables); + + return 0; +} + +static int si_program_cac_config_registers(struct radeon_device *rdev, + const struct si_cac_config_reg *cac_config_regs) +{ + const struct si_cac_config_reg *config_regs = cac_config_regs; + u32 data = 0, offset; + + if (!config_regs) + return -EINVAL; + + while (config_regs->offset != 0xFFFFFFFF) { + switch (config_regs->type) { + case SISLANDS_CACCONFIG_CGIND: + offset = SMC_CG_IND_START + config_regs->offset; + if (offset < SMC_CG_IND_END) + data = RREG32_SMC(offset); + break; + default: + data = RREG32(config_regs->offset << 2); + break; + } + + data &= ~config_regs->mask; + data |= ((config_regs->value << config_regs->shift) & config_regs->mask); + + switch (config_regs->type) { + case SISLANDS_CACCONFIG_CGIND: + offset = SMC_CG_IND_START + config_regs->offset; + if (offset < SMC_CG_IND_END) + WREG32_SMC(offset, data); + break; + default: + WREG32(config_regs->offset << 2, data); + break; + } + config_regs++; + } + return 0; +} + +static int si_initialize_hardware_cac_manager(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + int ret; + + if ((ni_pi->enable_cac == false) || + (ni_pi->cac_configuration_required == false)) + return 0; + + ret = si_program_cac_config_registers(rdev, si_pi->lcac_config); + if (ret) + return ret; + ret = si_program_cac_config_registers(rdev, si_pi->cac_override); + if (ret) + return ret; + ret = si_program_cac_config_registers(rdev, si_pi->cac_weights); + if (ret) + return ret; + + return 0; +} + +static int si_enable_smc_cac(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + bool enable) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + PPSMC_Result smc_result; + int ret = 0; + + if (ni_pi->enable_cac) { + if (enable) { + if (!si_should_disable_uvd_powertune(rdev, radeon_new_state)) { + if (ni_pi->support_cac_long_term_average) { + smc_result = si_send_msg_to_smc(rdev, PPSMC_CACLongTermAvgEnable); + if (smc_result != PPSMC_Result_OK) + ni_pi->support_cac_long_term_average = false; + } + + smc_result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableCac); + if (smc_result != PPSMC_Result_OK) { + ret = -EINVAL; + ni_pi->cac_enabled = false; + } else { + ni_pi->cac_enabled = true; + } + + if (si_pi->enable_dte) { + smc_result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableDTE); + if (smc_result != PPSMC_Result_OK) + ret = -EINVAL; + } + } + } else if (ni_pi->cac_enabled) { + if (si_pi->enable_dte) + smc_result = si_send_msg_to_smc(rdev, PPSMC_MSG_DisableDTE); + + smc_result = si_send_msg_to_smc(rdev, PPSMC_MSG_DisableCac); + + ni_pi->cac_enabled = false; + + if (ni_pi->support_cac_long_term_average) + smc_result = si_send_msg_to_smc(rdev, PPSMC_CACLongTermAvgDisable); + } + } + return ret; +} + +static int si_init_smc_spll_table(struct radeon_device *rdev) +{ + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + SMC_SISLANDS_SPLL_DIV_TABLE *spll_table; + SISLANDS_SMC_SCLK_VALUE sclk_params; + u32 fb_div, p_div; + u32 clk_s, clk_v; + u32 sclk = 0; + int ret = 0; + u32 tmp; + int i; + + if (si_pi->spll_table_start == 0) + return -EINVAL; + + spll_table = kzalloc(sizeof(SMC_SISLANDS_SPLL_DIV_TABLE), GFP_KERNEL); + if (spll_table == NULL) + return -ENOMEM; + + for (i = 0; i < 256; i++) { + ret = si_calculate_sclk_params(rdev, sclk, &sclk_params); + if (ret) + break; + + p_div = (sclk_params.vCG_SPLL_FUNC_CNTL & SPLL_PDIV_A_MASK) >> SPLL_PDIV_A_SHIFT; + fb_div = (sclk_params.vCG_SPLL_FUNC_CNTL_3 & SPLL_FB_DIV_MASK) >> SPLL_FB_DIV_SHIFT; + clk_s = (sclk_params.vCG_SPLL_SPREAD_SPECTRUM & CLK_S_MASK) >> CLK_S_SHIFT; + clk_v = (sclk_params.vCG_SPLL_SPREAD_SPECTRUM_2 & CLK_V_MASK) >> CLK_V_SHIFT; + + fb_div &= ~0x00001FFF; + fb_div >>= 1; + clk_v >>= 6; + + if (p_div & ~(SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_MASK >> SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT)) + ret = -EINVAL; + if (fb_div & ~(SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_MASK >> SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_SHIFT)) + ret = -EINVAL; + if (clk_s & ~(SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_MASK >> SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT)) + ret = -EINVAL; + if (clk_v & ~(SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_MASK >> SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT)) + ret = -EINVAL; + + if (ret) + break; + + tmp = ((fb_div << SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_SHIFT) & SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_MASK) | + ((p_div << SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT) & SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_MASK); + spll_table->freq[i] = cpu_to_be32(tmp); + + tmp = ((clk_v << SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT) & SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_MASK) | + ((clk_s << SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT) & SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_MASK); + spll_table->ss[i] = cpu_to_be32(tmp); + + sclk += 512; + } + + + if (!ret) + ret = si_copy_bytes_to_smc(rdev, si_pi->spll_table_start, + (u8 *)spll_table, sizeof(SMC_SISLANDS_SPLL_DIV_TABLE), + si_pi->sram_end); + + if (ret) + ni_pi->enable_power_containment = false; + + kfree(spll_table); + + return ret; +} + +static void si_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct ni_ps *ps = ni_get_ps(rps); + struct radeon_clock_and_voltage_limits *max_limits; + bool disable_mclk_switching = false; + bool disable_sclk_switching = false; + u32 mclk, sclk; + u16 vddc, vddci; + int i; + + if ((rdev->pm.dpm.new_active_crtc_count > 1) || + ni_dpm_vblank_too_short(rdev)) + disable_mclk_switching = true; + + if (rps->vclk || rps->dclk) { + disable_mclk_switching = true; + disable_sclk_switching = true; + } + + if (rdev->pm.dpm.ac_power) + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac; + else + max_limits = &rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc; + + for (i = ps->performance_level_count - 2; i >= 0; i--) { + if (ps->performance_levels[i].vddc > ps->performance_levels[i+1].vddc) + ps->performance_levels[i].vddc = ps->performance_levels[i+1].vddc; + } + if (rdev->pm.dpm.ac_power == false) { + for (i = 0; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].mclk > max_limits->mclk) + ps->performance_levels[i].mclk = max_limits->mclk; + if (ps->performance_levels[i].sclk > max_limits->sclk) + ps->performance_levels[i].sclk = max_limits->sclk; + if (ps->performance_levels[i].vddc > max_limits->vddc) + ps->performance_levels[i].vddc = max_limits->vddc; + if (ps->performance_levels[i].vddci > max_limits->vddci) + ps->performance_levels[i].vddci = max_limits->vddci; + } + } + + /* XXX validate the min clocks required for display */ + + if (disable_mclk_switching) { + mclk = ps->performance_levels[ps->performance_level_count - 1].mclk; + vddci = ps->performance_levels[ps->performance_level_count - 1].vddci; + } else { + mclk = ps->performance_levels[0].mclk; + vddci = ps->performance_levels[0].vddci; + } + + if (disable_sclk_switching) { + sclk = ps->performance_levels[ps->performance_level_count - 1].sclk; + vddc = ps->performance_levels[ps->performance_level_count - 1].vddc; + } else { + sclk = ps->performance_levels[0].sclk; + vddc = ps->performance_levels[0].vddc; + } + + /* adjusted low state */ + ps->performance_levels[0].sclk = sclk; + ps->performance_levels[0].mclk = mclk; + ps->performance_levels[0].vddc = vddc; + ps->performance_levels[0].vddci = vddci; + + if (disable_sclk_switching) { + sclk = ps->performance_levels[0].sclk; + for (i = 1; i < ps->performance_level_count; i++) { + if (sclk < ps->performance_levels[i].sclk) + sclk = ps->performance_levels[i].sclk; + } + for (i = 0; i < ps->performance_level_count; i++) { + ps->performance_levels[i].sclk = sclk; + ps->performance_levels[i].vddc = vddc; + } + } else { + for (i = 1; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].sclk < ps->performance_levels[i - 1].sclk) + ps->performance_levels[i].sclk = ps->performance_levels[i - 1].sclk; + if (ps->performance_levels[i].vddc < ps->performance_levels[i - 1].vddc) + ps->performance_levels[i].vddc = ps->performance_levels[i - 1].vddc; + } + } + + if (disable_mclk_switching) { + mclk = ps->performance_levels[0].mclk; + for (i = 1; i < ps->performance_level_count; i++) { + if (mclk < ps->performance_levels[i].mclk) + mclk = ps->performance_levels[i].mclk; + } + for (i = 0; i < ps->performance_level_count; i++) { + ps->performance_levels[i].mclk = mclk; + ps->performance_levels[i].vddci = vddci; + } + } else { + for (i = 1; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].mclk < ps->performance_levels[i - 1].mclk) + ps->performance_levels[i].mclk = ps->performance_levels[i - 1].mclk; + if (ps->performance_levels[i].vddci < ps->performance_levels[i - 1].vddci) + ps->performance_levels[i].vddci = ps->performance_levels[i - 1].vddci; + } + } + + for (i = 0; i < ps->performance_level_count; i++) + btc_adjust_clock_combinations(rdev, max_limits, + &ps->performance_levels[i]); + + for (i = 0; i < ps->performance_level_count; i++) { + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk, + ps->performance_levels[i].sclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk, + ps->performance_levels[i].mclk, + max_limits->vddci, &ps->performance_levels[i].vddci); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk, + ps->performance_levels[i].mclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + btc_apply_voltage_dependency_rules(&rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk, + rdev->clock.current_dispclk, + max_limits->vddc, &ps->performance_levels[i].vddc); + } + + for (i = 0; i < ps->performance_level_count; i++) { + btc_apply_voltage_delta_rules(rdev, + max_limits->vddc, max_limits->vddci, + &ps->performance_levels[i].vddc, + &ps->performance_levels[i].vddci); + } + + ps->dc_compatible = true; + for (i = 0; i < ps->performance_level_count; i++) { + if (ps->performance_levels[i].vddc > rdev->pm.dpm.dyn_state.max_clock_voltage_on_dc.vddc) + ps->dc_compatible = false; + } + +} + +#if 0 +static int si_read_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 *value) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + + return si_read_smc_sram_dword(rdev, + si_pi->soft_regs_start + reg_offset, value, + si_pi->sram_end); +} +#endif + +static int si_write_smc_soft_register(struct radeon_device *rdev, + u16 reg_offset, u32 value) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + + return si_write_smc_sram_dword(rdev, + si_pi->soft_regs_start + reg_offset, + value, si_pi->sram_end); +} + +static bool si_is_special_1gb_platform(struct radeon_device *rdev) +{ + bool ret = false; + u32 tmp, width, row, column, bank, density; + bool is_memory_gddr5, is_special; + + tmp = RREG32(MC_SEQ_MISC0); + is_memory_gddr5 = (MC_SEQ_MISC0_GDDR5_VALUE == ((tmp & MC_SEQ_MISC0_GDDR5_MASK) >> MC_SEQ_MISC0_GDDR5_SHIFT)); + is_special = (MC_SEQ_MISC0_REV_ID_VALUE == ((tmp & MC_SEQ_MISC0_REV_ID_MASK) >> MC_SEQ_MISC0_REV_ID_SHIFT)) + & (MC_SEQ_MISC0_VEN_ID_VALUE == ((tmp & MC_SEQ_MISC0_VEN_ID_MASK) >> MC_SEQ_MISC0_VEN_ID_SHIFT)); + + WREG32(MC_SEQ_IO_DEBUG_INDEX, 0xb); + width = ((RREG32(MC_SEQ_IO_DEBUG_DATA) >> 1) & 1) ? 16 : 32; + + tmp = RREG32(MC_ARB_RAMCFG); + row = ((tmp & NOOFROWS_MASK) >> NOOFROWS_SHIFT) + 10; + column = ((tmp & NOOFCOLS_MASK) >> NOOFCOLS_SHIFT) + 8; + bank = ((tmp & NOOFBANK_MASK) >> NOOFBANK_SHIFT) + 2; + + density = (1 << (row + column - 20 + bank)) * width; + + if ((rdev->ddev->pci_device == 0x6819) && + is_memory_gddr5 && is_special && (density == 0x400)) + ret = true; + + return ret; +} + +static void si_get_leakage_vddc(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u16 vddc, count = 0; + int i, ret; + + for (i = 0; i < SISLANDS_MAX_LEAKAGE_COUNT; i++) { + ret = radeon_atom_get_leakage_vddc_based_on_leakage_idx(rdev, &vddc, SISLANDS_LEAKAGE_INDEX0 + i); + + if (!ret && (vddc > 0) && (vddc != (SISLANDS_LEAKAGE_INDEX0 + i))) { + si_pi->leakage_voltage.entries[count].voltage = vddc; + si_pi->leakage_voltage.entries[count].leakage_index = + SISLANDS_LEAKAGE_INDEX0 + i; + count++; + } + } + si_pi->leakage_voltage.count = count; +} + +static int si_get_leakage_voltage_from_leakage_index(struct radeon_device *rdev, + u32 index, u16 *leakage_voltage) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + int i; + + if (leakage_voltage == NULL) + return -EINVAL; + + if ((index & 0xff00) != 0xff00) + return -EINVAL; + + if ((index & 0xff) > SISLANDS_MAX_LEAKAGE_COUNT + 1) + return -EINVAL; + + if (index < SISLANDS_LEAKAGE_INDEX0) + return -EINVAL; + + for (i = 0; i < si_pi->leakage_voltage.count; i++) { + if (si_pi->leakage_voltage.entries[i].leakage_index == index) { + *leakage_voltage = si_pi->leakage_voltage.entries[i].voltage; + return 0; + } + } + return -EAGAIN; +} + +static void si_set_dpm_event_sources(struct radeon_device *rdev, u32 sources) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + bool want_thermal_protection; + enum radeon_dpm_event_src dpm_event_src; + + switch (sources) { + case 0: + default: + want_thermal_protection = false; + break; + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGITAL; + break; + case (1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_EXTERNAL; + break; + case ((1 << RADEON_DPM_AUTO_THROTTLE_SRC_EXTERNAL) | + (1 << RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL)): + want_thermal_protection = true; + dpm_event_src = RADEON_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL; + break; + } + + if (want_thermal_protection) { + WREG32_P(CG_THERMAL_CTRL, DPM_EVENT_SRC(dpm_event_src), ~DPM_EVENT_SRC_MASK); + if (pi->thermal_protection) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + } else { + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); + } +} + +static void si_enable_auto_throttle_source(struct radeon_device *rdev, + enum radeon_dpm_auto_throttle_src source, + bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + if (!(pi->active_auto_throttle_sources & (1 << source))) { + pi->active_auto_throttle_sources |= 1 << source; + si_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } else { + if (pi->active_auto_throttle_sources & (1 << source)) { + pi->active_auto_throttle_sources &= ~(1 << source); + si_set_dpm_event_sources(rdev, pi->active_auto_throttle_sources); + } + } +} + +static void si_start_dpm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); +} + +static void si_stop_dpm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, 0, ~GLOBAL_PWRMGT_EN); +} + +static void si_enable_sclk_control(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~SCLK_PWRMGT_OFF); + else + WREG32_P(SCLK_PWRMGT_CNTL, SCLK_PWRMGT_OFF, ~SCLK_PWRMGT_OFF); + +} + +#if 0 +static int si_notify_hardware_of_thermal_state(struct radeon_device *rdev, + u32 thermal_level) +{ + PPSMC_Result ret; + + if (thermal_level == 0) { + ret = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + if (ret == PPSMC_Result_OK) + return 0; + else + return -EINVAL; + } + return 0; +} + +static void si_notify_hardware_vpu_recovery_event(struct radeon_device *rdev) +{ + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_tdr_is_about_to_happen, true); +} +#endif + +#if 0 +static int si_notify_hw_of_powersource(struct radeon_device *rdev, bool ac_power) +{ + if (ac_power) + return (si_send_msg_to_smc(rdev, PPSMC_MSG_RunningOnAC) == PPSMC_Result_OK) ? + 0 : -EINVAL; + + return 0; +} +#endif + +static PPSMC_Result si_send_msg_to_smc_with_parameter(struct radeon_device *rdev, + PPSMC_Msg msg, u32 parameter) +{ + WREG32(SMC_SCRATCH0, parameter); + return si_send_msg_to_smc(rdev, msg); +} + +static int si_restrict_performance_levels_before_switch(struct radeon_device *rdev) +{ + if (si_send_msg_to_smc(rdev, PPSMC_MSG_NoForcedLevel) != PPSMC_Result_OK) + return -EINVAL; + + return (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 1) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +int si_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct ni_ps *ps = ni_get_ps(rps); + u32 levels = ps->performance_level_count; + + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, levels) != PPSMC_Result_OK) + return -EINVAL; + + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 1) != PPSMC_Result_OK) + return -EINVAL; + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, 1) != PPSMC_Result_OK) + return -EINVAL; + } else if (level == RADEON_DPM_FORCED_LEVEL_AUTO) { + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetForcedLevels, 0) != PPSMC_Result_OK) + return -EINVAL; + + if (si_send_msg_to_smc_with_parameter(rdev, PPSMC_MSG_SetEnabledLevels, levels) != PPSMC_Result_OK) + return -EINVAL; + } + + rdev->pm.dpm.forced_level = level; + + return 0; +} + +static int si_set_boot_state(struct radeon_device *rdev) +{ + return (si_send_msg_to_smc(rdev, PPSMC_MSG_SwitchToInitialState) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static int si_set_sw_state(struct radeon_device *rdev) +{ + return (si_send_msg_to_smc(rdev, PPSMC_MSG_SwitchToSwState) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static int si_halt_smc(struct radeon_device *rdev) +{ + if (si_send_msg_to_smc(rdev, PPSMC_MSG_Halt) != PPSMC_Result_OK) + return -EINVAL; + + return (si_wait_for_smc_inactive(rdev) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static int si_resume_smc(struct radeon_device *rdev) +{ + if (si_send_msg_to_smc(rdev, PPSMC_FlushDataCache) != PPSMC_Result_OK) + return -EINVAL; + + return (si_send_msg_to_smc(rdev, PPSMC_MSG_Resume) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static void si_dpm_start_smc(struct radeon_device *rdev) +{ + si_program_jump_on_start(rdev); + si_start_smc(rdev); + si_start_smc_clock(rdev); +} + +static void si_dpm_stop_smc(struct radeon_device *rdev) +{ + si_reset_smc(rdev); + si_stop_smc_clock(rdev); +} + +static int si_process_firmware_header(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 tmp; + int ret; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_stateTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->state_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_softRegisters, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->soft_regs_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_mcRegisterTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->mc_reg_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_mcArbDramAutoRefreshTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->arb_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_CacConfigTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->cac_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_DteConfiguration, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->dte_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_spllTable, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->spll_table_start = tmp; + + ret = si_read_smc_sram_dword(rdev, + SISLANDS_SMC_FIRMWARE_HEADER_LOCATION + + SISLANDS_SMC_FIRMWARE_HEADER_PAPMParameters, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + si_pi->papm_cfg_table_start = tmp; + + return ret; +} + +static void si_read_clock_registers(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + + si_pi->clock_registers.cg_spll_func_cntl = RREG32(CG_SPLL_FUNC_CNTL); + si_pi->clock_registers.cg_spll_func_cntl_2 = RREG32(CG_SPLL_FUNC_CNTL_2); + si_pi->clock_registers.cg_spll_func_cntl_3 = RREG32(CG_SPLL_FUNC_CNTL_3); + si_pi->clock_registers.cg_spll_func_cntl_4 = RREG32(CG_SPLL_FUNC_CNTL_4); + si_pi->clock_registers.cg_spll_spread_spectrum = RREG32(CG_SPLL_SPREAD_SPECTRUM); + si_pi->clock_registers.cg_spll_spread_spectrum_2 = RREG32(CG_SPLL_SPREAD_SPECTRUM_2); + si_pi->clock_registers.dll_cntl = RREG32(DLL_CNTL); + si_pi->clock_registers.mclk_pwrmgt_cntl = RREG32(MCLK_PWRMGT_CNTL); + si_pi->clock_registers.mpll_ad_func_cntl = RREG32(MPLL_AD_FUNC_CNTL); + si_pi->clock_registers.mpll_dq_func_cntl = RREG32(MPLL_DQ_FUNC_CNTL); + si_pi->clock_registers.mpll_func_cntl = RREG32(MPLL_FUNC_CNTL); + si_pi->clock_registers.mpll_func_cntl_1 = RREG32(MPLL_FUNC_CNTL_1); + si_pi->clock_registers.mpll_func_cntl_2 = RREG32(MPLL_FUNC_CNTL_2); + si_pi->clock_registers.mpll_ss1 = RREG32(MPLL_SS1); + si_pi->clock_registers.mpll_ss2 = RREG32(MPLL_SS2); +} + +static void si_enable_thermal_protection(struct radeon_device *rdev, + bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, 0, ~THERMAL_PROTECTION_DIS); + else + WREG32_P(GENERAL_PWRMGT, THERMAL_PROTECTION_DIS, ~THERMAL_PROTECTION_DIS); +} + +static void si_enable_acpi_power_management(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, STATIC_PM_EN, ~STATIC_PM_EN); +} + +#if 0 +static int si_enter_ulp_state(struct radeon_device *rdev) +{ + WREG32(SMC_MESSAGE_0, PPSMC_MSG_SwitchToMinimumPower); + + DRM_UDELAY(25000); + + return 0; +} + +static int si_exit_ulp_state(struct radeon_device *rdev) +{ + int i; + + WREG32(SMC_MESSAGE_0, PPSMC_MSG_ResumeFromMinimumPower); + + DRM_UDELAY(7000); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(SMC_RESP_0) == 1) + break; + DRM_UDELAY(1000); + } + + return 0; +} +#endif + +static int si_notify_smc_display_change(struct radeon_device *rdev, + bool has_display) +{ + PPSMC_Msg msg = has_display ? + PPSMC_MSG_HasDisplay : PPSMC_MSG_NoDisplay; + + return (si_send_msg_to_smc(rdev, msg) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static void si_program_response_times(struct radeon_device *rdev) +{ + u32 voltage_response_time, backbias_response_time, acpi_delay_time, vbi_time_out; + u32 vddc_dly, acpi_dly, vbi_dly; + u32 reference_clock; + + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_mvdd_chg_time, 1); + + voltage_response_time = (u32)rdev->pm.dpm.voltage_response_time; + backbias_response_time = (u32)rdev->pm.dpm.backbias_response_time; + + if (voltage_response_time == 0) + voltage_response_time = 1000; + + acpi_delay_time = 15000; + vbi_time_out = 100000; + + reference_clock = radeon_get_xclk(rdev); + + vddc_dly = (voltage_response_time * reference_clock) / 100; + acpi_dly = (acpi_delay_time * reference_clock) / 100; + vbi_dly = (vbi_time_out * reference_clock) / 100; + + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_delay_vreg, vddc_dly); + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_delay_acpi, acpi_dly); + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_mclk_chg_timeout, vbi_dly); + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_mc_block_delay, 0xAA); +} + +static void si_program_ds_registers(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + u32 tmp = 1; /* XXX: 0x10 on tahiti A0 */ + + if (eg_pi->sclk_deep_sleep) { + WREG32_P(MISC_CLK_CNTL, DEEP_SLEEP_CLK_SEL(tmp), ~DEEP_SLEEP_CLK_SEL_MASK); + WREG32_P(CG_SPLL_AUTOSCALE_CNTL, AUTOSCALE_ON_SS_CLEAR, + ~AUTOSCALE_ON_SS_CLEAR); + } +} + +static void si_program_display_gap(struct radeon_device *rdev) +{ + u32 tmp, pipe; + int i; + + tmp = RREG32(CG_DISPLAY_GAP_CNTL) & ~(DISP1_GAP_MASK | DISP2_GAP_MASK); + if (rdev->pm.dpm.new_active_crtc_count > 0) + tmp |= DISP1_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM); + else + tmp |= DISP1_GAP(R600_PM_DISPLAY_GAP_IGNORE); + + if (rdev->pm.dpm.new_active_crtc_count > 1) + tmp |= DISP2_GAP(R600_PM_DISPLAY_GAP_VBLANK_OR_WM); + else + tmp |= DISP2_GAP(R600_PM_DISPLAY_GAP_IGNORE); + + WREG32(CG_DISPLAY_GAP_CNTL, tmp); + + tmp = RREG32(DCCG_DISP_SLOW_SELECT_REG); + pipe = (tmp & DCCG_DISP1_SLOW_SELECT_MASK) >> DCCG_DISP1_SLOW_SELECT_SHIFT; + + if ((rdev->pm.dpm.new_active_crtc_count > 0) && + (!(rdev->pm.dpm.new_active_crtcs & (1 << pipe)))) { + /* find the first active crtc */ + for (i = 0; i < rdev->num_crtc; i++) { + if (rdev->pm.dpm.new_active_crtcs & (1 << i)) + break; + } + if (i == rdev->num_crtc) + pipe = 0; + else + pipe = i; + + tmp &= ~DCCG_DISP1_SLOW_SELECT_MASK; + tmp |= DCCG_DISP1_SLOW_SELECT(pipe); + WREG32(DCCG_DISP_SLOW_SELECT_REG, tmp); + } + + si_notify_smc_display_change(rdev, rdev->pm.dpm.new_active_crtc_count > 0); +} + +static void si_enable_spread_spectrum(struct radeon_device *rdev, bool enable) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + if (enable) { + if (pi->sclk_ss) + WREG32_P(GENERAL_PWRMGT, DYN_SPREAD_SPECTRUM_EN, ~DYN_SPREAD_SPECTRUM_EN); + } else { + WREG32_P(CG_SPLL_SPREAD_SPECTRUM, 0, ~SSEN); + WREG32_P(GENERAL_PWRMGT, 0, ~DYN_SPREAD_SPECTRUM_EN); + } +} + +static void si_setup_bsp(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(pi->asi, + xclk, + 16, + &pi->bsp, + &pi->bsu); + + r600_calculate_u_and_p(pi->pasi, + xclk, + 16, + &pi->pbsp, + &pi->pbsu); + + + pi->dsp = BSP(pi->bsp) | BSU(pi->bsu); + pi->psp = BSP(pi->pbsp) | BSU(pi->pbsu); + + WREG32(CG_BSP, pi->dsp); +} + +static void si_program_git(struct radeon_device *rdev) +{ + WREG32_P(CG_GIT, CG_GICST(R600_GICST_DFLT), ~CG_GICST_MASK); +} + +static void si_program_tp(struct radeon_device *rdev) +{ + int i; + enum r600_td td = R600_TD_DFLT; + + for (i = 0; i < R600_PM_NUMBER_OF_TC; i++) + WREG32(CG_FFCT_0 + (i * 4), (UTC_0(r600_utc[i]) | DTC_0(r600_dtc[i]))); + + if (td == R600_TD_AUTO) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_FORCE_TREND_SEL); + else + WREG32_P(SCLK_PWRMGT_CNTL, FIR_FORCE_TREND_SEL, ~FIR_FORCE_TREND_SEL); + + if (td == R600_TD_UP) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_TREND_MODE); + + if (td == R600_TD_DOWN) + WREG32_P(SCLK_PWRMGT_CNTL, FIR_TREND_MODE, ~FIR_TREND_MODE); +} + +static void si_program_tpp(struct radeon_device *rdev) +{ + WREG32(CG_TPC, R600_TPC_DFLT); +} + +static void si_program_sstp(struct radeon_device *rdev) +{ + WREG32(CG_SSP, (SSTU(R600_SSTU_DFLT) | SST(R600_SST_DFLT))); +} + +static void si_enable_display_gap(struct radeon_device *rdev) +{ + u32 tmp = RREG32(CG_DISPLAY_GAP_CNTL); + + tmp &= ~(DISP1_GAP_MASK | DISP2_GAP_MASK); + tmp |= (DISP1_GAP(R600_PM_DISPLAY_GAP_IGNORE) | + DISP2_GAP(R600_PM_DISPLAY_GAP_IGNORE)); + + tmp &= ~(DISP1_GAP_MCHG_MASK | DISP2_GAP_MCHG_MASK); + tmp |= (DISP1_GAP_MCHG(R600_PM_DISPLAY_GAP_VBLANK) | + DISP2_GAP_MCHG(R600_PM_DISPLAY_GAP_IGNORE)); + WREG32(CG_DISPLAY_GAP_CNTL, tmp); +} + +static void si_program_vc(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + WREG32(CG_FTV, pi->vrc); +} + +static void si_clear_vc(struct radeon_device *rdev) +{ + WREG32(CG_FTV, 0); +} + +static u8 si_get_ddr3_mclk_frequency_ratio(u32 memory_clock) +{ + u8 mc_para_index; + + if (memory_clock < 10000) + mc_para_index = 0; + else if (memory_clock >= 80000) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 10000) / 5000 + 1); + return mc_para_index; +} + +static u8 si_get_mclk_frequency_ratio(u32 memory_clock, bool strobe_mode) +{ + u8 mc_para_index; + + if (strobe_mode) { + if (memory_clock < 12500) + mc_para_index = 0x00; + else if (memory_clock > 47500) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 10000) / 2500); + } else { + if (memory_clock < 65000) + mc_para_index = 0x00; + else if (memory_clock > 135000) + mc_para_index = 0x0f; + else + mc_para_index = (u8)((memory_clock - 60000) / 5000); + } + return mc_para_index; +} + +static u8 si_get_strobe_mode_settings(struct radeon_device *rdev, u32 mclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + bool strobe_mode = false; + u8 result = 0; + + if (mclk <= pi->mclk_strobe_mode_threshold) + strobe_mode = true; + + if (pi->mem_gddr5) + result = si_get_mclk_frequency_ratio(mclk, strobe_mode); + else + result = si_get_ddr3_mclk_frequency_ratio(mclk); + + if (strobe_mode) + result |= SISLANDS_SMC_STROBE_ENABLE; + + return result; +} + +static int si_upload_firmware(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + int ret; + + si_reset_smc(rdev); + si_stop_smc_clock(rdev); + + ret = si_load_smc_ucode(rdev, si_pi->sram_end); + + return ret; +} + +static bool si_validate_phase_shedding_tables(struct radeon_device *rdev, + const struct atom_voltage_table *table, + const struct radeon_phase_shedding_limits_table *limits) +{ + u32 data, num_bits, num_levels; + + if ((table == NULL) || (limits == NULL)) + return false; + + data = table->mask_low; + + num_bits = hweight32(data); + + if (num_bits == 0) + return false; + + num_levels = (1 << num_bits); + + if (table->count != num_levels) + return false; + + if (limits->count != (num_levels - 1)) + return false; + + return true; +} + +static void si_trim_voltage_table_to_fit_state_table(struct radeon_device *rdev, + struct atom_voltage_table *voltage_table) +{ + unsigned int i, diff; + + if (voltage_table->count <= SISLANDS_MAX_NO_VREG_STEPS) + return; + + diff = voltage_table->count - SISLANDS_MAX_NO_VREG_STEPS; + + for (i= 0; i < SISLANDS_MAX_NO_VREG_STEPS; i++) + voltage_table->entries[i] = voltage_table->entries[i + diff]; + + voltage_table->count = SISLANDS_MAX_NO_VREG_STEPS; +} + +static int si_construct_voltage_tables(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + int ret; + + ret = radeon_atom_get_voltage_table(rdev, VOLTAGE_TYPE_VDDC, + VOLTAGE_OBJ_GPIO_LUT, &eg_pi->vddc_voltage_table); + if (ret) + return ret; + + if (eg_pi->vddc_voltage_table.count > SISLANDS_MAX_NO_VREG_STEPS) + si_trim_voltage_table_to_fit_state_table(rdev, &eg_pi->vddc_voltage_table); + + if (eg_pi->vddci_control) { + ret = radeon_atom_get_voltage_table(rdev, VOLTAGE_TYPE_VDDCI, + VOLTAGE_OBJ_GPIO_LUT, &eg_pi->vddci_voltage_table); + if (ret) + return ret; + + if (eg_pi->vddci_voltage_table.count > SISLANDS_MAX_NO_VREG_STEPS) + si_trim_voltage_table_to_fit_state_table(rdev, &eg_pi->vddci_voltage_table); + } + + if (pi->mvdd_control) { + ret = radeon_atom_get_voltage_table(rdev, VOLTAGE_TYPE_MVDDC, + VOLTAGE_OBJ_GPIO_LUT, &si_pi->mvdd_voltage_table); + + if (ret) { + pi->mvdd_control = false; + return ret; + } + + if (si_pi->mvdd_voltage_table.count == 0) { + pi->mvdd_control = false; + return -EINVAL; + } + + if (si_pi->mvdd_voltage_table.count > SISLANDS_MAX_NO_VREG_STEPS) + si_trim_voltage_table_to_fit_state_table(rdev, &si_pi->mvdd_voltage_table); + } + + if (si_pi->vddc_phase_shed_control) { + ret = radeon_atom_get_voltage_table(rdev, VOLTAGE_TYPE_VDDC, + VOLTAGE_OBJ_PHASE_LUT, &si_pi->vddc_phase_shed_table); + if (ret) + si_pi->vddc_phase_shed_control = false; + + if ((si_pi->vddc_phase_shed_table.count == 0) || + (si_pi->vddc_phase_shed_table.count > SISLANDS_MAX_NO_VREG_STEPS)) + si_pi->vddc_phase_shed_control = false; + } + + return 0; +} + +static void si_populate_smc_voltage_table(struct radeon_device *rdev, + const struct atom_voltage_table *voltage_table, + SISLANDS_SMC_STATETABLE *table) +{ + unsigned int i; + + for (i = 0; i < voltage_table->count; i++) + table->lowSMIO[i] |= cpu_to_be32(voltage_table->entries[i].smio_low); +} + +static int si_populate_smc_voltage_tables(struct radeon_device *rdev, + SISLANDS_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + u8 i; + + if (eg_pi->vddc_voltage_table.count) { + si_populate_smc_voltage_table(rdev, &eg_pi->vddc_voltage_table, table); + table->voltageMaskTable.lowMask[SISLANDS_SMC_VOLTAGEMASK_VDDC] = + cpu_to_be32(eg_pi->vddc_voltage_table.mask_low); + + for (i = 0; i < eg_pi->vddc_voltage_table.count; i++) { + if (pi->max_vddc_in_table <= eg_pi->vddc_voltage_table.entries[i].value) { + table->maxVDDCIndexInPPTable = i; + break; + } + } + } + + if (eg_pi->vddci_voltage_table.count) { + si_populate_smc_voltage_table(rdev, &eg_pi->vddci_voltage_table, table); + + table->voltageMaskTable.lowMask[SISLANDS_SMC_VOLTAGEMASK_VDDCI] = + cpu_to_be32(eg_pi->vddci_voltage_table.mask_low); + } + + + if (si_pi->mvdd_voltage_table.count) { + si_populate_smc_voltage_table(rdev, &si_pi->mvdd_voltage_table, table); + + table->voltageMaskTable.lowMask[SISLANDS_SMC_VOLTAGEMASK_MVDD] = + cpu_to_be32(si_pi->mvdd_voltage_table.mask_low); + } + + if (si_pi->vddc_phase_shed_control) { + if (si_validate_phase_shedding_tables(rdev, &si_pi->vddc_phase_shed_table, + &rdev->pm.dpm.dyn_state.phase_shedding_limits_table)) { + si_populate_smc_voltage_table(rdev, &si_pi->vddc_phase_shed_table, table); + + table->phaseMaskTable.lowMask[SISLANDS_SMC_VOLTAGEMASK_VDDC] = + cpu_to_be32(si_pi->vddc_phase_shed_table.mask_low); + + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_phase_shedding_delay, + (u32)si_pi->vddc_phase_shed_table.phase_delay); + } else { + si_pi->vddc_phase_shed_control = false; + } + } + + return 0; +} + +static int si_populate_voltage_value(struct radeon_device *rdev, + const struct atom_voltage_table *table, + u16 value, SISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + unsigned int i; + + for (i = 0; i < table->count; i++) { + if (value <= table->entries[i].value) { + voltage->index = (u8)i; + voltage->value = cpu_to_be16(table->entries[i].value); + break; + } + } + + if (i >= table->count) + return -EINVAL; + + return 0; +} + +static int si_populate_mvdd_value(struct radeon_device *rdev, u32 mclk, + SISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + + if (pi->mvdd_control) { + if (mclk <= pi->mvdd_split_frequency) + voltage->index = 0; + else + voltage->index = (u8)(si_pi->mvdd_voltage_table.count) - 1; + + voltage->value = cpu_to_be16(si_pi->mvdd_voltage_table.entries[voltage->index].value); + } + return 0; +} + +static int si_get_std_voltage_value(struct radeon_device *rdev, + SISLANDS_SMC_VOLTAGE_VALUE *voltage, + u16 *std_voltage) +{ + u16 v_index; + bool voltage_found = false; + *std_voltage = be16_to_cpu(voltage->value); + + if (rdev->pm.dpm.dyn_state.cac_leakage_table.entries) { + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_NEW_CAC_VOLTAGE) { + if (rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries == NULL) + return -EINVAL; + + for (v_index = 0; (u32)v_index < rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count; v_index++) { + if (be16_to_cpu(voltage->value) == + (u16)rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[v_index].v) { + voltage_found = true; + if ((u32)v_index < rdev->pm.dpm.dyn_state.cac_leakage_table.count) + *std_voltage = + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[v_index].vddc; + else + *std_voltage = + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[rdev->pm.dpm.dyn_state.cac_leakage_table.count-1].vddc; + break; + } + } + + if (!voltage_found) { + for (v_index = 0; (u32)v_index < rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.count; v_index++) { + if (be16_to_cpu(voltage->value) <= + (u16)rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk.entries[v_index].v) { + voltage_found = true; + if ((u32)v_index < rdev->pm.dpm.dyn_state.cac_leakage_table.count) + *std_voltage = + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[v_index].vddc; + else + *std_voltage = + rdev->pm.dpm.dyn_state.cac_leakage_table.entries[rdev->pm.dpm.dyn_state.cac_leakage_table.count-1].vddc; + break; + } + } + } + } else { + if ((u32)voltage->index < rdev->pm.dpm.dyn_state.cac_leakage_table.count) + *std_voltage = rdev->pm.dpm.dyn_state.cac_leakage_table.entries[voltage->index].vddc; + } + } + + return 0; +} + +static int si_populate_std_voltage_value(struct radeon_device *rdev, + u16 value, u8 index, + SISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + voltage->index = index; + voltage->value = cpu_to_be16(value); + + return 0; +} + +static int si_populate_phase_shedding_value(struct radeon_device *rdev, + const struct radeon_phase_shedding_limits_table *limits, + u16 voltage, u32 sclk, u32 mclk, + SISLANDS_SMC_VOLTAGE_VALUE *smc_voltage) +{ + unsigned int i; + + for (i = 0; i < limits->count; i++) { + if ((voltage <= limits->entries[i].voltage) && + (sclk <= limits->entries[i].sclk) && + (mclk <= limits->entries[i].mclk)) + break; + } + + smc_voltage->phase_settings = (u8)i; + + return 0; +} + +static int si_init_arb_table_index(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 tmp; + int ret; + + ret = si_read_smc_sram_dword(rdev, si_pi->arb_table_start, &tmp, si_pi->sram_end); + if (ret) + return ret; + + tmp &= 0x00FFFFFF; + tmp |= MC_CG_ARB_FREQ_F1 << 24; + + return si_write_smc_sram_dword(rdev, si_pi->arb_table_start, tmp, si_pi->sram_end); +} + +static int si_initial_switch_from_arb_f0_to_f1(struct radeon_device *rdev) +{ + return ni_copy_and_switch_arb_sets(rdev, MC_CG_ARB_FREQ_F0, MC_CG_ARB_FREQ_F1); +} + +static int si_reset_to_default(struct radeon_device *rdev) +{ + return (si_send_msg_to_smc(rdev, PPSMC_MSG_ResetToDefaults) == PPSMC_Result_OK) ? + 0 : -EINVAL; +} + +static int si_force_switch_to_arb_f0(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 tmp; + int ret; + + ret = si_read_smc_sram_dword(rdev, si_pi->arb_table_start, + &tmp, si_pi->sram_end); + if (ret) + return ret; + + tmp = (tmp >> 24) & 0xff; + + if (tmp == MC_CG_ARB_FREQ_F0) + return 0; + + return ni_copy_and_switch_arb_sets(rdev, tmp, MC_CG_ARB_FREQ_F0); +} + +static u32 si_calculate_memory_refresh_rate(struct radeon_device *rdev, + u32 engine_clock) +{ + u32 dram_rows; + u32 dram_refresh_rate; + u32 mc_arb_rfsh_rate; + u32 tmp = (RREG32(MC_ARB_RAMCFG) & NOOFROWS_MASK) >> NOOFROWS_SHIFT; + + if (tmp >= 4) + dram_rows = 16384; + else + dram_rows = 1 << (tmp + 10); + + dram_refresh_rate = 1 << ((RREG32(MC_SEQ_MISC0) & 0x3) + 3); + mc_arb_rfsh_rate = ((engine_clock * 10) * dram_refresh_rate / dram_rows - 32) / 64; + + return mc_arb_rfsh_rate; +} + +static int si_populate_memory_timing_parameters(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SMC_SIslands_MCArbDramTimingRegisterSet *arb_regs) +{ + u32 dram_timing; + u32 dram_timing2; + u32 burst_time; + + arb_regs->mc_arb_rfsh_rate = + (u8)si_calculate_memory_refresh_rate(rdev, pl->sclk); + + radeon_atom_set_engine_dram_timings(rdev, + pl->sclk, + pl->mclk); + + dram_timing = RREG32(MC_ARB_DRAM_TIMING); + dram_timing2 = RREG32(MC_ARB_DRAM_TIMING2); + burst_time = RREG32(MC_ARB_BURST_TIME) & STATE0_MASK; + + arb_regs->mc_arb_dram_timing = cpu_to_be32(dram_timing); + arb_regs->mc_arb_dram_timing2 = cpu_to_be32(dram_timing2); + arb_regs->mc_arb_burst_time = (u8)burst_time; + + return 0; +} + +static int si_do_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + unsigned int first_arb_set) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + SMC_SIslands_MCArbDramTimingRegisterSet arb_regs = { 0 }; + int i, ret = 0; + + for (i = 0; i < state->performance_level_count; i++) { + ret = si_populate_memory_timing_parameters(rdev, &state->performance_levels[i], &arb_regs); + if (ret) + break; + ret = si_copy_bytes_to_smc(rdev, + si_pi->arb_table_start + + offsetof(SMC_SIslands_MCArbDramTimingRegisters, data) + + sizeof(SMC_SIslands_MCArbDramTimingRegisterSet) * (first_arb_set + i), + (u8 *)&arb_regs, + sizeof(SMC_SIslands_MCArbDramTimingRegisterSet), + si_pi->sram_end); + if (ret) + break; + } + + return ret; +} + +static int si_program_memory_timing_parameters(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + return si_do_program_memory_timing_parameters(rdev, radeon_new_state, + SISLANDS_DRIVER_STATE_ARB_INDEX); +} + +static int si_populate_initial_mvdd_value(struct radeon_device *rdev, + struct SISLANDS_SMC_VOLTAGE_VALUE *voltage) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + + if (pi->mvdd_control) + return si_populate_voltage_value(rdev, &si_pi->mvdd_voltage_table, + si_pi->mvdd_bootup_value, voltage); + + return 0; +} + +static int si_populate_smc_initial_state(struct radeon_device *rdev, + struct radeon_ps *radeon_initial_state, + SISLANDS_SMC_STATETABLE *table) +{ + struct ni_ps *initial_state = ni_get_ps(radeon_initial_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + u32 reg; + int ret; + + table->initialState.levels[0].mclk.vDLL_CNTL = + cpu_to_be32(si_pi->clock_registers.dll_cntl); + table->initialState.levels[0].mclk.vMCLK_PWRMGT_CNTL = + cpu_to_be32(si_pi->clock_registers.mclk_pwrmgt_cntl); + table->initialState.levels[0].mclk.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(si_pi->clock_registers.mpll_ad_func_cntl); + table->initialState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(si_pi->clock_registers.mpll_dq_func_cntl); + table->initialState.levels[0].mclk.vMPLL_FUNC_CNTL = + cpu_to_be32(si_pi->clock_registers.mpll_func_cntl); + table->initialState.levels[0].mclk.vMPLL_FUNC_CNTL_1 = + cpu_to_be32(si_pi->clock_registers.mpll_func_cntl_1); + table->initialState.levels[0].mclk.vMPLL_FUNC_CNTL_2 = + cpu_to_be32(si_pi->clock_registers.mpll_func_cntl_2); + table->initialState.levels[0].mclk.vMPLL_SS = + cpu_to_be32(si_pi->clock_registers.mpll_ss1); + table->initialState.levels[0].mclk.vMPLL_SS2 = + cpu_to_be32(si_pi->clock_registers.mpll_ss2); + + table->initialState.levels[0].mclk.mclk_value = + cpu_to_be32(initial_state->performance_levels[0].mclk); + + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(si_pi->clock_registers.cg_spll_func_cntl); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(si_pi->clock_registers.cg_spll_func_cntl_2); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(si_pi->clock_registers.cg_spll_func_cntl_3); + table->initialState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_4 = + cpu_to_be32(si_pi->clock_registers.cg_spll_func_cntl_4); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM = + cpu_to_be32(si_pi->clock_registers.cg_spll_spread_spectrum); + table->initialState.levels[0].sclk.vCG_SPLL_SPREAD_SPECTRUM_2 = + cpu_to_be32(si_pi->clock_registers.cg_spll_spread_spectrum_2); + + table->initialState.levels[0].sclk.sclk_value = + cpu_to_be32(initial_state->performance_levels[0].sclk); + + table->initialState.levels[0].arbRefreshState = + SISLANDS_INITIAL_STATE_ARB_INDEX; + + table->initialState.levels[0].ACIndex = 0; + + ret = si_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + initial_state->performance_levels[0].vddc, + &table->initialState.levels[0].vddc); + + if (!ret) { + u16 std_vddc; + + ret = si_get_std_voltage_value(rdev, + &table->initialState.levels[0].vddc, + &std_vddc); + if (!ret) + si_populate_std_voltage_value(rdev, std_vddc, + table->initialState.levels[0].vddc.index, + &table->initialState.levels[0].std_vddc); + } + + if (eg_pi->vddci_control) + si_populate_voltage_value(rdev, + &eg_pi->vddci_voltage_table, + initial_state->performance_levels[0].vddci, + &table->initialState.levels[0].vddci); + + if (si_pi->vddc_phase_shed_control) + si_populate_phase_shedding_value(rdev, + &rdev->pm.dpm.dyn_state.phase_shedding_limits_table, + initial_state->performance_levels[0].vddc, + initial_state->performance_levels[0].sclk, + initial_state->performance_levels[0].mclk, + &table->initialState.levels[0].vddc); + + si_populate_initial_mvdd_value(rdev, &table->initialState.levels[0].mvdd); + + reg = CG_R(0xffff) | CG_L(0); + table->initialState.levels[0].aT = cpu_to_be32(reg); + + table->initialState.levels[0].bSP = cpu_to_be32(pi->dsp); + + table->initialState.levels[0].gen2PCIE = (u8)si_pi->boot_pcie_gen; + + if (pi->mem_gddr5) { + table->initialState.levels[0].strobeMode = + si_get_strobe_mode_settings(rdev, + initial_state->performance_levels[0].mclk); + + if (initial_state->performance_levels[0].mclk > pi->mclk_edc_enable_threshold) + table->initialState.levels[0].mcFlags = SISLANDS_SMC_MC_EDC_RD_FLAG | SISLANDS_SMC_MC_EDC_WR_FLAG; + else + table->initialState.levels[0].mcFlags = 0; + } + + table->initialState.levelCount = 1; + + table->initialState.flags |= PPSMC_SWSTATE_FLAG_DC; + + table->initialState.levels[0].dpm2.MaxPS = 0; + table->initialState.levels[0].dpm2.NearTDPDec = 0; + table->initialState.levels[0].dpm2.AboveSafeInc = 0; + table->initialState.levels[0].dpm2.BelowSafeInc = 0; + table->initialState.levels[0].dpm2.PwrEfficiencyRatio = 0; + + reg = MIN_POWER_MASK | MAX_POWER_MASK; + table->initialState.levels[0].SQPowerThrottle = cpu_to_be32(reg); + + reg = MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + table->initialState.levels[0].SQPowerThrottle_2 = cpu_to_be32(reg); + + return 0; +} + +static int si_populate_smc_acpi_state(struct radeon_device *rdev, + SISLANDS_SMC_STATETABLE *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + u32 spll_func_cntl = si_pi->clock_registers.cg_spll_func_cntl; + u32 spll_func_cntl_2 = si_pi->clock_registers.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = si_pi->clock_registers.cg_spll_func_cntl_3; + u32 spll_func_cntl_4 = si_pi->clock_registers.cg_spll_func_cntl_4; + u32 dll_cntl = si_pi->clock_registers.dll_cntl; + u32 mclk_pwrmgt_cntl = si_pi->clock_registers.mclk_pwrmgt_cntl; + u32 mpll_ad_func_cntl = si_pi->clock_registers.mpll_ad_func_cntl; + u32 mpll_dq_func_cntl = si_pi->clock_registers.mpll_dq_func_cntl; + u32 mpll_func_cntl = si_pi->clock_registers.mpll_func_cntl; + u32 mpll_func_cntl_1 = si_pi->clock_registers.mpll_func_cntl_1; + u32 mpll_func_cntl_2 = si_pi->clock_registers.mpll_func_cntl_2; + u32 reg; + int ret; + + table->ACPIState = table->initialState; + + table->ACPIState.flags &= ~PPSMC_SWSTATE_FLAG_DC; + + if (pi->acpi_vddc) { + ret = si_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + pi->acpi_vddc, &table->ACPIState.levels[0].vddc); + if (!ret) { + u16 std_vddc; + + ret = si_get_std_voltage_value(rdev, + &table->ACPIState.levels[0].vddc, &std_vddc); + if (!ret) + si_populate_std_voltage_value(rdev, std_vddc, + table->ACPIState.levels[0].vddc.index, + &table->ACPIState.levels[0].std_vddc); + } + table->ACPIState.levels[0].gen2PCIE = si_pi->acpi_pcie_gen; + + if (si_pi->vddc_phase_shed_control) { + si_populate_phase_shedding_value(rdev, + &rdev->pm.dpm.dyn_state.phase_shedding_limits_table, + pi->acpi_vddc, + 0, + 0, + &table->ACPIState.levels[0].vddc); + } + } else { + ret = si_populate_voltage_value(rdev, &eg_pi->vddc_voltage_table, + pi->min_vddc_in_table, &table->ACPIState.levels[0].vddc); + if (!ret) { + u16 std_vddc; + + ret = si_get_std_voltage_value(rdev, + &table->ACPIState.levels[0].vddc, &std_vddc); + + if (!ret) + si_populate_std_voltage_value(rdev, std_vddc, + table->ACPIState.levels[0].vddc.index, + &table->ACPIState.levels[0].std_vddc); + } + table->ACPIState.levels[0].gen2PCIE = (u8)r600_get_pcie_gen_support(rdev, + si_pi->sys_pcie_mask, + si_pi->boot_pcie_gen, + RADEON_PCIE_GEN1); + + if (si_pi->vddc_phase_shed_control) + si_populate_phase_shedding_value(rdev, + &rdev->pm.dpm.dyn_state.phase_shedding_limits_table, + pi->min_vddc_in_table, + 0, + 0, + &table->ACPIState.levels[0].vddc); + } + + if (pi->acpi_vddc) { + if (eg_pi->acpi_vddci) + si_populate_voltage_value(rdev, &eg_pi->vddci_voltage_table, + eg_pi->acpi_vddci, + &table->ACPIState.levels[0].vddci); + } + + mclk_pwrmgt_cntl |= MRDCK0_RESET | MRDCK1_RESET; + mclk_pwrmgt_cntl &= ~(MRDCK0_PDNB | MRDCK1_PDNB); + + dll_cntl &= ~(MRDCK0_BYPASS | MRDCK1_BYPASS); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(4); + + table->ACPIState.levels[0].mclk.vDLL_CNTL = + cpu_to_be32(dll_cntl); + table->ACPIState.levels[0].mclk.vMCLK_PWRMGT_CNTL = + cpu_to_be32(mclk_pwrmgt_cntl); + table->ACPIState.levels[0].mclk.vMPLL_AD_FUNC_CNTL = + cpu_to_be32(mpll_ad_func_cntl); + table->ACPIState.levels[0].mclk.vMPLL_DQ_FUNC_CNTL = + cpu_to_be32(mpll_dq_func_cntl); + table->ACPIState.levels[0].mclk.vMPLL_FUNC_CNTL = + cpu_to_be32(mpll_func_cntl); + table->ACPIState.levels[0].mclk.vMPLL_FUNC_CNTL_1 = + cpu_to_be32(mpll_func_cntl_1); + table->ACPIState.levels[0].mclk.vMPLL_FUNC_CNTL_2 = + cpu_to_be32(mpll_func_cntl_2); + table->ACPIState.levels[0].mclk.vMPLL_SS = + cpu_to_be32(si_pi->clock_registers.mpll_ss1); + table->ACPIState.levels[0].mclk.vMPLL_SS2 = + cpu_to_be32(si_pi->clock_registers.mpll_ss2); + + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL = + cpu_to_be32(spll_func_cntl); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_2 = + cpu_to_be32(spll_func_cntl_2); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_3 = + cpu_to_be32(spll_func_cntl_3); + table->ACPIState.levels[0].sclk.vCG_SPLL_FUNC_CNTL_4 = + cpu_to_be32(spll_func_cntl_4); + + table->ACPIState.levels[0].mclk.mclk_value = 0; + table->ACPIState.levels[0].sclk.sclk_value = 0; + + si_populate_mvdd_value(rdev, 0, &table->ACPIState.levels[0].mvdd); + + if (eg_pi->dynamic_ac_timing) + table->ACPIState.levels[0].ACIndex = 0; + + table->ACPIState.levels[0].dpm2.MaxPS = 0; + table->ACPIState.levels[0].dpm2.NearTDPDec = 0; + table->ACPIState.levels[0].dpm2.AboveSafeInc = 0; + table->ACPIState.levels[0].dpm2.BelowSafeInc = 0; + table->ACPIState.levels[0].dpm2.PwrEfficiencyRatio = 0; + + reg = MIN_POWER_MASK | MAX_POWER_MASK; + table->ACPIState.levels[0].SQPowerThrottle = cpu_to_be32(reg); + + reg = MAX_POWER_DELTA_MASK | STI_SIZE_MASK | LTI_RATIO_MASK; + table->ACPIState.levels[0].SQPowerThrottle_2 = cpu_to_be32(reg); + + return 0; +} + +static int si_populate_ulv_state(struct radeon_device *rdev, + SISLANDS_SMC_SWSTATE *state) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + struct si_ulv_param *ulv = &si_pi->ulv; + u32 sclk_in_sr = 1350; /* ??? */ + int ret; + + ret = si_convert_power_level_to_smc(rdev, &ulv->pl, + &state->levels[0]); + if (!ret) { + if (eg_pi->sclk_deep_sleep) { + if (sclk_in_sr <= SCLK_MIN_DEEPSLEEP_FREQ) + state->levels[0].stateFlags |= PPSMC_STATEFLAG_DEEPSLEEP_BYPASS; + else + state->levels[0].stateFlags |= PPSMC_STATEFLAG_DEEPSLEEP_THROTTLE; + } + if (ulv->one_pcie_lane_in_ulv) + state->flags |= PPSMC_SWSTATE_FLAG_PCIE_X1; + state->levels[0].arbRefreshState = (u8)(SISLANDS_ULV_STATE_ARB_INDEX); + state->levels[0].ACIndex = 1; + state->levels[0].std_vddc = state->levels[0].vddc; + state->levelCount = 1; + + state->flags |= PPSMC_SWSTATE_FLAG_DC; + } + + return ret; +} + +static int si_program_ulv_memory_timing_parameters(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct si_ulv_param *ulv = &si_pi->ulv; + SMC_SIslands_MCArbDramTimingRegisterSet arb_regs = { 0 }; + int ret; + + ret = si_populate_memory_timing_parameters(rdev, &ulv->pl, + &arb_regs); + if (ret) + return ret; + + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_ulv_volt_change_delay, + ulv->volt_change_delay); + + ret = si_copy_bytes_to_smc(rdev, + si_pi->arb_table_start + + offsetof(SMC_SIslands_MCArbDramTimingRegisters, data) + + sizeof(SMC_SIslands_MCArbDramTimingRegisterSet) * SISLANDS_ULV_STATE_ARB_INDEX, + (u8 *)&arb_regs, + sizeof(SMC_SIslands_MCArbDramTimingRegisterSet), + si_pi->sram_end); + + return ret; +} + +static void si_get_mvdd_configuration(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + + pi->mvdd_split_frequency = 30000; +} + +static int si_init_smc_table(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + struct radeon_ps *radeon_boot_state = rdev->pm.dpm.boot_ps; + const struct si_ulv_param *ulv = &si_pi->ulv; + SISLANDS_SMC_STATETABLE *table = &si_pi->smc_statetable; + int ret; + u32 lane_width; + u32 vr_hot_gpio; + + si_populate_smc_voltage_tables(rdev, table); + + switch (rdev->pm.int_thermal_type) { + case THERMAL_TYPE_SI: + case THERMAL_TYPE_EMC2103_WITH_INTERNAL: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_INTERNAL; + break; + case THERMAL_TYPE_NONE: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_NONE; + break; + default: + table->thermalProtectType = PPSMC_THERMAL_PROTECT_TYPE_EXTERNAL; + break; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_HARDWAREDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REGULATOR_HOT) { + if ((rdev->ddev->pci_device != 0x6818) && (rdev->ddev->pci_device != 0x6819)) + table->systemFlags |= PPSMC_SYSTEMFLAG_REGULATOR_HOT; + } + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_STEPVDDC) + table->systemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (pi->mem_gddr5) + table->systemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_REVERT_GPIO5_POLARITY) + table->systemFlags |= PPSMC_EXTRAFLAGS_AC2DC_GPIO5_POLARITY_HIGH; + + if (rdev->pm.dpm.platform_caps & ATOM_PP_PLATFORM_CAP_VRHOT_GPIO_CONFIGURABLE) { + table->systemFlags |= PPSMC_SYSTEMFLAG_REGULATOR_HOT_PROG_GPIO; + vr_hot_gpio = rdev->pm.dpm.backbias_response_time; + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_vr_hot_gpio, + vr_hot_gpio); + } + + ret = si_populate_smc_initial_state(rdev, radeon_boot_state, table); + if (ret) + return ret; + + ret = si_populate_smc_acpi_state(rdev, table); + if (ret) + return ret; + + table->driverState = table->initialState; + + ret = si_do_program_memory_timing_parameters(rdev, radeon_boot_state, + SISLANDS_INITIAL_STATE_ARB_INDEX); + if (ret) + return ret; + + if (ulv->supported && ulv->pl.vddc) { + ret = si_populate_ulv_state(rdev, &table->ULVState); + if (ret) + return ret; + + ret = si_program_ulv_memory_timing_parameters(rdev); + if (ret) + return ret; + + WREG32(CG_ULV_CONTROL, ulv->cg_ulv_control); + WREG32(CG_ULV_PARAMETER, ulv->cg_ulv_parameter); + + lane_width = radeon_get_pcie_lanes(rdev); + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_non_ulv_pcie_link_width, lane_width); + } else { + table->ULVState = table->initialState; + } + + return si_copy_bytes_to_smc(rdev, si_pi->state_table_start, + (u8 *)table, sizeof(SISLANDS_SMC_STATETABLE), + si_pi->sram_end); +} + +static int si_calculate_sclk_params(struct radeon_device *rdev, + u32 engine_clock, + SISLANDS_SMC_SCLK_VALUE *sclk) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + struct atom_clock_dividers dividers; + u32 spll_func_cntl = si_pi->clock_registers.cg_spll_func_cntl; + u32 spll_func_cntl_2 = si_pi->clock_registers.cg_spll_func_cntl_2; + u32 spll_func_cntl_3 = si_pi->clock_registers.cg_spll_func_cntl_3; + u32 spll_func_cntl_4 = si_pi->clock_registers.cg_spll_func_cntl_4; + u32 cg_spll_spread_spectrum = si_pi->clock_registers.cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2 = si_pi->clock_registers.cg_spll_spread_spectrum_2; + u64 tmp; + u32 reference_clock = rdev->clock.spll.reference_freq; + u32 reference_divider; + u32 fbdiv; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + engine_clock, false, ÷rs); + if (ret) + return ret; + + reference_divider = 1 + dividers.ref_div; + + tmp = (u64) engine_clock * reference_divider * dividers.post_div * 16384; + do_div(tmp, reference_clock); + fbdiv = (u32) tmp; + + spll_func_cntl &= ~(SPLL_PDIV_A_MASK | SPLL_REF_DIV_MASK); + spll_func_cntl |= SPLL_REF_DIV(dividers.ref_div); + spll_func_cntl |= SPLL_PDIV_A(dividers.post_div); + + spll_func_cntl_2 &= ~SCLK_MUX_SEL_MASK; + spll_func_cntl_2 |= SCLK_MUX_SEL(2); + + spll_func_cntl_3 &= ~SPLL_FB_DIV_MASK; + spll_func_cntl_3 |= SPLL_FB_DIV(fbdiv); + spll_func_cntl_3 |= SPLL_DITHEN; + + if (pi->sclk_ss) { + struct radeon_atom_ss ss; + u32 vco_freq = engine_clock * dividers.post_div; + + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_ENGINE_SS, vco_freq)) { + u32 clk_s = reference_clock * 5 / (reference_divider * ss.rate); + u32 clk_v = 4 * ss.percentage * fbdiv / (clk_s * 10000); + + cg_spll_spread_spectrum &= ~CLK_S_MASK; + cg_spll_spread_spectrum |= CLK_S(clk_s); + cg_spll_spread_spectrum |= SSEN; + + cg_spll_spread_spectrum_2 &= ~CLK_V_MASK; + cg_spll_spread_spectrum_2 |= CLK_V(clk_v); + } + } + + sclk->sclk_value = engine_clock; + sclk->vCG_SPLL_FUNC_CNTL = spll_func_cntl; + sclk->vCG_SPLL_FUNC_CNTL_2 = spll_func_cntl_2; + sclk->vCG_SPLL_FUNC_CNTL_3 = spll_func_cntl_3; + sclk->vCG_SPLL_FUNC_CNTL_4 = spll_func_cntl_4; + sclk->vCG_SPLL_SPREAD_SPECTRUM = cg_spll_spread_spectrum; + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cg_spll_spread_spectrum_2; + + return 0; +} + +static int si_populate_sclk_value(struct radeon_device *rdev, + u32 engine_clock, + SISLANDS_SMC_SCLK_VALUE *sclk) +{ + SISLANDS_SMC_SCLK_VALUE sclk_tmp; + int ret; + + ret = si_calculate_sclk_params(rdev, engine_clock, &sclk_tmp); + if (!ret) { + sclk->sclk_value = cpu_to_be32(sclk_tmp.sclk_value); + sclk->vCG_SPLL_FUNC_CNTL = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL); + sclk->vCG_SPLL_FUNC_CNTL_2 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_2); + sclk->vCG_SPLL_FUNC_CNTL_3 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_3); + sclk->vCG_SPLL_FUNC_CNTL_4 = cpu_to_be32(sclk_tmp.vCG_SPLL_FUNC_CNTL_4); + sclk->vCG_SPLL_SPREAD_SPECTRUM = cpu_to_be32(sclk_tmp.vCG_SPLL_SPREAD_SPECTRUM); + sclk->vCG_SPLL_SPREAD_SPECTRUM_2 = cpu_to_be32(sclk_tmp.vCG_SPLL_SPREAD_SPECTRUM_2); + } + + return ret; +} + +static int si_populate_mclk_value(struct radeon_device *rdev, + u32 engine_clock, + u32 memory_clock, + SISLANDS_SMC_MCLK_VALUE *mclk, + bool strobe_mode, + bool dll_state_on) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + u32 dll_cntl = si_pi->clock_registers.dll_cntl; + u32 mclk_pwrmgt_cntl = si_pi->clock_registers.mclk_pwrmgt_cntl; + u32 mpll_ad_func_cntl = si_pi->clock_registers.mpll_ad_func_cntl; + u32 mpll_dq_func_cntl = si_pi->clock_registers.mpll_dq_func_cntl; + u32 mpll_func_cntl = si_pi->clock_registers.mpll_func_cntl; + u32 mpll_func_cntl_1 = si_pi->clock_registers.mpll_func_cntl_1; + u32 mpll_func_cntl_2 = si_pi->clock_registers.mpll_func_cntl_2; + u32 mpll_ss1 = si_pi->clock_registers.mpll_ss1; + u32 mpll_ss2 = si_pi->clock_registers.mpll_ss2; + struct atom_mpll_param mpll_param; + int ret; + + ret = radeon_atom_get_memory_pll_dividers(rdev, memory_clock, strobe_mode, &mpll_param); + if (ret) + return ret; + + mpll_func_cntl &= ~BWCTRL_MASK; + mpll_func_cntl |= BWCTRL(mpll_param.bwcntl); + + mpll_func_cntl_1 &= ~(CLKF_MASK | CLKFRAC_MASK | VCO_MODE_MASK); + mpll_func_cntl_1 |= CLKF(mpll_param.clkf) | + CLKFRAC(mpll_param.clkfrac) | VCO_MODE(mpll_param.vco_mode); + + mpll_ad_func_cntl &= ~YCLK_POST_DIV_MASK; + mpll_ad_func_cntl |= YCLK_POST_DIV(mpll_param.post_div); + + if (pi->mem_gddr5) { + mpll_dq_func_cntl &= ~(YCLK_SEL_MASK | YCLK_POST_DIV_MASK); + mpll_dq_func_cntl |= YCLK_SEL(mpll_param.yclk_sel) | + YCLK_POST_DIV(mpll_param.post_div); + } + + if (pi->mclk_ss) { + struct radeon_atom_ss ss; + u32 freq_nom; + u32 tmp; + u32 reference_clock = rdev->clock.mpll.reference_freq; + + if (pi->mem_gddr5) + freq_nom = memory_clock * 4; + else + freq_nom = memory_clock * 2; + + tmp = freq_nom / reference_clock; + tmp = tmp * tmp; + if (radeon_atombios_get_asic_ss_info(rdev, &ss, + ASIC_INTERNAL_MEMORY_SS, freq_nom)) { + u32 clks = reference_clock * 5 / ss.rate; + u32 clkv = (u32)((((131 * ss.percentage * ss.rate) / 100) * tmp) / freq_nom); + + mpll_ss1 &= ~CLKV_MASK; + mpll_ss1 |= CLKV(clkv); + + mpll_ss2 &= ~CLKS_MASK; + mpll_ss2 |= CLKS(clks); + } + } + + mclk_pwrmgt_cntl &= ~DLL_SPEED_MASK; + mclk_pwrmgt_cntl |= DLL_SPEED(mpll_param.dll_speed); + + if (dll_state_on) + mclk_pwrmgt_cntl |= MRDCK0_PDNB | MRDCK1_PDNB; + else + mclk_pwrmgt_cntl &= ~(MRDCK0_PDNB | MRDCK1_PDNB); + + mclk->mclk_value = cpu_to_be32(memory_clock); + mclk->vMPLL_FUNC_CNTL = cpu_to_be32(mpll_func_cntl); + mclk->vMPLL_FUNC_CNTL_1 = cpu_to_be32(mpll_func_cntl_1); + mclk->vMPLL_FUNC_CNTL_2 = cpu_to_be32(mpll_func_cntl_2); + mclk->vMPLL_AD_FUNC_CNTL = cpu_to_be32(mpll_ad_func_cntl); + mclk->vMPLL_DQ_FUNC_CNTL = cpu_to_be32(mpll_dq_func_cntl); + mclk->vMCLK_PWRMGT_CNTL = cpu_to_be32(mclk_pwrmgt_cntl); + mclk->vDLL_CNTL = cpu_to_be32(dll_cntl); + mclk->vMPLL_SS = cpu_to_be32(mpll_ss1); + mclk->vMPLL_SS2 = cpu_to_be32(mpll_ss2); + + return 0; +} + +static void si_populate_smc_sp(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SISLANDS_SMC_SWSTATE *smc_state) +{ + struct ni_ps *ps = ni_get_ps(radeon_state); + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + int i; + + for (i = 0; i < ps->performance_level_count - 1; i++) + smc_state->levels[i].bSP = cpu_to_be32(pi->dsp); + + smc_state->levels[ps->performance_level_count - 1].bSP = + cpu_to_be32(pi->psp); +} + +static int si_convert_power_level_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SISLANDS_SMC_HW_PERFORMANCE_LEVEL *level) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + int ret; + bool dll_state_on; + u16 std_vddc; + bool gmc_pg = false; + + if (eg_pi->pcie_performance_request && + (si_pi->force_pcie_gen != RADEON_PCIE_GEN_INVALID)) + level->gen2PCIE = (u8)si_pi->force_pcie_gen; + else + level->gen2PCIE = (u8)pl->pcie_gen; + + ret = si_populate_sclk_value(rdev, pl->sclk, &level->sclk); + if (ret) + return ret; + + level->mcFlags = 0; + + if (pi->mclk_stutter_mode_threshold && + (pl->mclk <= pi->mclk_stutter_mode_threshold) && + !eg_pi->uvd_enabled && + (RREG32(DPG_PIPE_STUTTER_CONTROL) & STUTTER_ENABLE) && + (rdev->pm.dpm.new_active_crtc_count <= 2)) { + level->mcFlags |= SISLANDS_SMC_MC_STUTTER_EN; + + if (gmc_pg) + level->mcFlags |= SISLANDS_SMC_MC_PG_EN; + } + + if (pi->mem_gddr5) { + if (pl->mclk > pi->mclk_edc_enable_threshold) + level->mcFlags |= SISLANDS_SMC_MC_EDC_RD_FLAG; + + if (pl->mclk > eg_pi->mclk_edc_wr_enable_threshold) + level->mcFlags |= SISLANDS_SMC_MC_EDC_WR_FLAG; + + level->strobeMode = si_get_strobe_mode_settings(rdev, pl->mclk); + + if (level->strobeMode & SISLANDS_SMC_STROBE_ENABLE) { + if (si_get_mclk_frequency_ratio(pl->mclk, true) >= + ((RREG32(MC_SEQ_MISC7) >> 16) & 0xf)) + dll_state_on = ((RREG32(MC_SEQ_MISC5) >> 1) & 0x1) ? true : false; + else + dll_state_on = ((RREG32(MC_SEQ_MISC6) >> 1) & 0x1) ? true : false; + } else { + dll_state_on = false; + } + } else { + level->strobeMode = si_get_strobe_mode_settings(rdev, + pl->mclk); + + dll_state_on = ((RREG32(MC_SEQ_MISC5) >> 1) & 0x1) ? true : false; + } + + ret = si_populate_mclk_value(rdev, + pl->sclk, + pl->mclk, + &level->mclk, + (level->strobeMode & SISLANDS_SMC_STROBE_ENABLE) != 0, dll_state_on); + if (ret) + return ret; + + ret = si_populate_voltage_value(rdev, + &eg_pi->vddc_voltage_table, + pl->vddc, &level->vddc); + if (ret) + return ret; + + + ret = si_get_std_voltage_value(rdev, &level->vddc, &std_vddc); + if (ret) + return ret; + + ret = si_populate_std_voltage_value(rdev, std_vddc, + level->vddc.index, &level->std_vddc); + if (ret) + return ret; + + if (eg_pi->vddci_control) { + ret = si_populate_voltage_value(rdev, &eg_pi->vddci_voltage_table, + pl->vddci, &level->vddci); + if (ret) + return ret; + } + + if (si_pi->vddc_phase_shed_control) { + ret = si_populate_phase_shedding_value(rdev, + &rdev->pm.dpm.dyn_state.phase_shedding_limits_table, + pl->vddc, + pl->sclk, + pl->mclk, + &level->vddc); + if (ret) + return ret; + } + + level->MaxPoweredUpCU = si_pi->max_cu; + + ret = si_populate_mvdd_value(rdev, pl->mclk, &level->mvdd); + + return ret; +} + +static int si_populate_smc_t(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SISLANDS_SMC_SWSTATE *smc_state) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + u32 a_t; + u32 t_l, t_h; + u32 high_bsp; + int i, ret; + + if (state->performance_level_count >= 9) + return -EINVAL; + + if (state->performance_level_count < 2) { + a_t = CG_R(0xffff) | CG_L(0); + smc_state->levels[0].aT = cpu_to_be32(a_t); + return 0; + } + + smc_state->levels[0].aT = cpu_to_be32(0); + + for (i = 0; i <= state->performance_level_count - 2; i++) { + ret = r600_calculate_at( + (50 / SISLANDS_MAX_HARDWARE_POWERLEVELS) * 100 * (i + 1), + 100 * R600_AH_DFLT, + state->performance_levels[i + 1].sclk, + state->performance_levels[i].sclk, + &t_l, + &t_h); + + if (ret) { + t_h = (i + 1) * 1000 - 50 * R600_AH_DFLT; + t_l = (i + 1) * 1000 + 50 * R600_AH_DFLT; + } + + a_t = be32_to_cpu(smc_state->levels[i].aT) & ~CG_R_MASK; + a_t |= CG_R(t_l * pi->bsp / 20000); + smc_state->levels[i].aT = cpu_to_be32(a_t); + + high_bsp = (i == state->performance_level_count - 2) ? + pi->pbsp : pi->bsp; + a_t = CG_R(0xffff) | CG_L(t_h * high_bsp / 20000); + smc_state->levels[i + 1].aT = cpu_to_be32(a_t); + } + + return 0; +} + +static int si_disable_ulv(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct si_ulv_param *ulv = &si_pi->ulv; + + if (ulv->supported) + return (si_send_msg_to_smc(rdev, PPSMC_MSG_DisableULV) == PPSMC_Result_OK) ? + 0 : -EINVAL; + + return 0; +} + +static bool si_is_state_ulv_compatible(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + const struct si_power_info *si_pi = si_get_pi(rdev); + const struct si_ulv_param *ulv = &si_pi->ulv; + const struct ni_ps *state = ni_get_ps(radeon_state); + int i; + + if (state->performance_levels[0].mclk != ulv->pl.mclk) + return false; + + /* XXX validate against display requirements! */ + + for (i = 0; i < rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.count; i++) { + if (rdev->clock.current_dispclk <= + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[i].clk) { + if (ulv->pl.vddc < + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[i].v) + return false; + } + } + + if ((radeon_state->vclk != 0) || (radeon_state->dclk != 0)) + return false; + + return true; +} + +static int si_set_power_state_conditionally_enable_ulv(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + const struct si_power_info *si_pi = si_get_pi(rdev); + const struct si_ulv_param *ulv = &si_pi->ulv; + + if (ulv->supported) { + if (si_is_state_ulv_compatible(rdev, radeon_new_state)) + return (si_send_msg_to_smc(rdev, PPSMC_MSG_EnableULV) == PPSMC_Result_OK) ? + 0 : -EINVAL; + } + return 0; +} + +static int si_convert_power_state_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SISLANDS_SMC_SWSTATE *smc_state) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct ni_power_info *ni_pi = ni_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + struct ni_ps *state = ni_get_ps(radeon_state); + int i, ret; + u32 threshold; + u32 sclk_in_sr = 1350; /* ??? */ + + if (state->performance_level_count > SISLANDS_MAX_HARDWARE_POWERLEVELS) + return -EINVAL; + + threshold = state->performance_levels[state->performance_level_count-1].sclk * 100 / 100; + + if (radeon_state->vclk && radeon_state->dclk) { + eg_pi->uvd_enabled = true; + if (eg_pi->smu_uvd_hs) + smc_state->flags |= PPSMC_SWSTATE_FLAG_UVD; + } else { + eg_pi->uvd_enabled = false; + } + + if (state->dc_compatible) + smc_state->flags |= PPSMC_SWSTATE_FLAG_DC; + + smc_state->levelCount = 0; + for (i = 0; i < state->performance_level_count; i++) { + if (eg_pi->sclk_deep_sleep) { + if ((i == 0) || si_pi->sclk_deep_sleep_above_low) { + if (sclk_in_sr <= SCLK_MIN_DEEPSLEEP_FREQ) + smc_state->levels[i].stateFlags |= PPSMC_STATEFLAG_DEEPSLEEP_BYPASS; + else + smc_state->levels[i].stateFlags |= PPSMC_STATEFLAG_DEEPSLEEP_THROTTLE; + } + } + + ret = si_convert_power_level_to_smc(rdev, &state->performance_levels[i], + &smc_state->levels[i]); + smc_state->levels[i].arbRefreshState = + (u8)(SISLANDS_DRIVER_STATE_ARB_INDEX + i); + + if (ret) + return ret; + + if (ni_pi->enable_power_containment) + smc_state->levels[i].displayWatermark = + (state->performance_levels[i].sclk < threshold) ? + PPSMC_DISPLAY_WATERMARK_LOW : PPSMC_DISPLAY_WATERMARK_HIGH; + else + smc_state->levels[i].displayWatermark = (i < 2) ? + PPSMC_DISPLAY_WATERMARK_LOW : PPSMC_DISPLAY_WATERMARK_HIGH; + + if (eg_pi->dynamic_ac_timing) + smc_state->levels[i].ACIndex = SISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT + i; + else + smc_state->levels[i].ACIndex = 0; + + smc_state->levelCount++; + } + + si_write_smc_soft_register(rdev, + SI_SMC_SOFT_REGISTER_watermark_threshold, + threshold / 512); + + si_populate_smc_sp(rdev, radeon_state, smc_state); + + ret = si_populate_power_containment_values(rdev, radeon_state, smc_state); + if (ret) + ni_pi->enable_power_containment = false; + + ret = si_populate_sq_ramping_values(rdev, radeon_state, smc_state); + if (ret) + ni_pi->enable_sq_ramping = false; + + return si_populate_smc_t(rdev, radeon_state, smc_state); +} + +static int si_upload_sw_state(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct ni_ps *new_state = ni_get_ps(radeon_new_state); + int ret; + u32 address = si_pi->state_table_start + + offsetof(SISLANDS_SMC_STATETABLE, driverState); + u32 state_size = sizeof(SISLANDS_SMC_SWSTATE) + + ((new_state->performance_level_count - 1) * + sizeof(SISLANDS_SMC_HW_PERFORMANCE_LEVEL)); + SISLANDS_SMC_SWSTATE *smc_state = &si_pi->smc_statetable.driverState; + + memset(smc_state, 0, state_size); + + ret = si_convert_power_state_to_smc(rdev, radeon_new_state, smc_state); + if (ret) + return ret; + + ret = si_copy_bytes_to_smc(rdev, address, (u8 *)smc_state, + state_size, si_pi->sram_end); + + return ret; +} + +static int si_upload_ulv_state(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct si_ulv_param *ulv = &si_pi->ulv; + int ret = 0; + + if (ulv->supported && ulv->pl.vddc) { + u32 address = si_pi->state_table_start + + offsetof(SISLANDS_SMC_STATETABLE, ULVState); + SISLANDS_SMC_SWSTATE *smc_state = &si_pi->smc_statetable.ULVState; + u32 state_size = sizeof(SISLANDS_SMC_SWSTATE); + + memset(smc_state, 0, state_size); + + ret = si_populate_ulv_state(rdev, smc_state); + if (!ret) + ret = si_copy_bytes_to_smc(rdev, address, (u8 *)smc_state, + state_size, si_pi->sram_end); + } + + return ret; +} + +static int si_upload_smc_data(struct radeon_device *rdev) +{ + struct radeon_crtc *radeon_crtc = NULL; + int i; + + if (rdev->pm.dpm.new_active_crtc_count == 0) + return 0; + + for (i = 0; i < rdev->num_crtc; i++) { + if (rdev->pm.dpm.new_active_crtcs & (1 << i)) { + radeon_crtc = rdev->mode_info.crtcs[i]; + break; + } + } + + if (radeon_crtc == NULL) + return 0; + + if (radeon_crtc->line_time <= 0) + return 0; + + if (si_write_smc_soft_register(rdev, + SI_SMC_SOFT_REGISTER_crtc_index, + radeon_crtc->crtc_id) != PPSMC_Result_OK) + return 0; + + if (si_write_smc_soft_register(rdev, + SI_SMC_SOFT_REGISTER_mclk_change_block_cp_min, + radeon_crtc->wm_high / radeon_crtc->line_time) != PPSMC_Result_OK) + return 0; + + if (si_write_smc_soft_register(rdev, + SI_SMC_SOFT_REGISTER_mclk_change_block_cp_max, + radeon_crtc->wm_low / radeon_crtc->line_time) != PPSMC_Result_OK) + return 0; + + return 0; +} + +static int si_set_mc_special_registers(struct radeon_device *rdev, + struct si_mc_reg_table *table) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + u8 i, j, k; + u32 temp_reg; + + for (i = 0, j = table->last; i < table->last; i++) { + if (j >= SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + switch (table->mc_reg_address[i].s1 << 2) { + case MC_SEQ_MISC1: + temp_reg = RREG32(MC_PMG_CMD_EMRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_EMRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + for (k = 0; k < table->num_entries; k++) + table->mc_reg_table_entry[k].mc_data[j] = + ((temp_reg & 0xffff0000)) | + ((table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16); + j++; + if (j >= SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + + temp_reg = RREG32(MC_PMG_CMD_MRS); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS_LP >> 2; + for (k = 0; k < table->num_entries; k++) { + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + if (!pi->mem_gddr5) + table->mc_reg_table_entry[k].mc_data[j] |= 0x100; + } + j++; + if (j > SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + + if (!pi->mem_gddr5) { + table->mc_reg_address[j].s1 = MC_PMG_AUTO_CMD >> 2; + table->mc_reg_address[j].s0 = MC_PMG_AUTO_CMD >> 2; + for (k = 0; k < table->num_entries; k++) + table->mc_reg_table_entry[k].mc_data[j] = + (table->mc_reg_table_entry[k].mc_data[i] & 0xffff0000) >> 16; + j++; + if (j > SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + } + break; + case MC_SEQ_RESERVE_M: + temp_reg = RREG32(MC_PMG_CMD_MRS1); + table->mc_reg_address[j].s1 = MC_PMG_CMD_MRS1 >> 2; + table->mc_reg_address[j].s0 = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + for(k = 0; k < table->num_entries; k++) + table->mc_reg_table_entry[k].mc_data[j] = + (temp_reg & 0xffff0000) | + (table->mc_reg_table_entry[k].mc_data[i] & 0x0000ffff); + j++; + if (j > SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + break; + default: + break; + } + } + + table->last = j; + + return 0; +} + +static bool si_check_s0_mc_reg_index(u16 in_reg, u16 *out_reg) +{ + bool result = true; + + switch (in_reg) { + case MC_SEQ_RAS_TIMING >> 2: + *out_reg = MC_SEQ_RAS_TIMING_LP >> 2; + break; + case MC_SEQ_CAS_TIMING >> 2: + *out_reg = MC_SEQ_CAS_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING >> 2: + *out_reg = MC_SEQ_MISC_TIMING_LP >> 2; + break; + case MC_SEQ_MISC_TIMING2 >> 2: + *out_reg = MC_SEQ_MISC_TIMING2_LP >> 2; + break; + case MC_SEQ_RD_CTL_D0 >> 2: + *out_reg = MC_SEQ_RD_CTL_D0_LP >> 2; + break; + case MC_SEQ_RD_CTL_D1 >> 2: + *out_reg = MC_SEQ_RD_CTL_D1_LP >> 2; + break; + case MC_SEQ_WR_CTL_D0 >> 2: + *out_reg = MC_SEQ_WR_CTL_D0_LP >> 2; + break; + case MC_SEQ_WR_CTL_D1 >> 2: + *out_reg = MC_SEQ_WR_CTL_D1_LP >> 2; + break; + case MC_PMG_CMD_EMRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_EMRS_LP >> 2; + break; + case MC_PMG_CMD_MRS >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS_LP >> 2; + break; + case MC_PMG_CMD_MRS1 >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS1_LP >> 2; + break; + case MC_SEQ_PMG_TIMING >> 2: + *out_reg = MC_SEQ_PMG_TIMING_LP >> 2; + break; + case MC_PMG_CMD_MRS2 >> 2: + *out_reg = MC_SEQ_PMG_CMD_MRS2_LP >> 2; + break; + case MC_SEQ_WR_CTL_2 >> 2: + *out_reg = MC_SEQ_WR_CTL_2_LP >> 2; + break; + default: + result = false; + break; + } + + return result; +} + +static void si_set_valid_flag(struct si_mc_reg_table *table) +{ + u8 i, j; + + for (i = 0; i < table->last; i++) { + for (j = 1; j < table->num_entries; j++) { + if (table->mc_reg_table_entry[j-1].mc_data[i] != table->mc_reg_table_entry[j].mc_data[i]) { + table->valid_flag |= 1 << i; + break; + } + } + } +} + +static void si_set_s0_mc_reg_index(struct si_mc_reg_table *table) +{ + u32 i; + u16 address; + + for (i = 0; i < table->last; i++) + table->mc_reg_address[i].s0 = si_check_s0_mc_reg_index(table->mc_reg_address[i].s1, &address) ? + address : table->mc_reg_address[i].s1; + +} + +static int si_copy_vbios_mc_reg_table(struct atom_mc_reg_table *table, + struct si_mc_reg_table *si_table) +{ + u8 i, j; + + if (table->last > SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE) + return -EINVAL; + if (table->num_entries > MAX_AC_TIMING_ENTRIES) + return -EINVAL; + + for (i = 0; i < table->last; i++) + si_table->mc_reg_address[i].s1 = table->mc_reg_address[i].s1; + si_table->last = table->last; + + for (i = 0; i < table->num_entries; i++) { + si_table->mc_reg_table_entry[i].mclk_max = + table->mc_reg_table_entry[i].mclk_max; + for (j = 0; j < table->last; j++) { + si_table->mc_reg_table_entry[i].mc_data[j] = + table->mc_reg_table_entry[i].mc_data[j]; + } + } + si_table->num_entries = table->num_entries; + + return 0; +} + +static int si_initialize_mc_reg_table(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + struct atom_mc_reg_table *table; + struct si_mc_reg_table *si_table = &si_pi->mc_reg_table; + u8 module_index = rv770_get_memory_module_index(rdev); + int ret; + + table = kzalloc(sizeof(struct atom_mc_reg_table), GFP_KERNEL); + if (!table) + return -ENOMEM; + + WREG32(MC_SEQ_RAS_TIMING_LP, RREG32(MC_SEQ_RAS_TIMING)); + WREG32(MC_SEQ_CAS_TIMING_LP, RREG32(MC_SEQ_CAS_TIMING)); + WREG32(MC_SEQ_MISC_TIMING_LP, RREG32(MC_SEQ_MISC_TIMING)); + WREG32(MC_SEQ_MISC_TIMING2_LP, RREG32(MC_SEQ_MISC_TIMING2)); + WREG32(MC_SEQ_PMG_CMD_EMRS_LP, RREG32(MC_PMG_CMD_EMRS)); + WREG32(MC_SEQ_PMG_CMD_MRS_LP, RREG32(MC_PMG_CMD_MRS)); + WREG32(MC_SEQ_PMG_CMD_MRS1_LP, RREG32(MC_PMG_CMD_MRS1)); + WREG32(MC_SEQ_WR_CTL_D0_LP, RREG32(MC_SEQ_WR_CTL_D0)); + WREG32(MC_SEQ_WR_CTL_D1_LP, RREG32(MC_SEQ_WR_CTL_D1)); + WREG32(MC_SEQ_RD_CTL_D0_LP, RREG32(MC_SEQ_RD_CTL_D0)); + WREG32(MC_SEQ_RD_CTL_D1_LP, RREG32(MC_SEQ_RD_CTL_D1)); + WREG32(MC_SEQ_PMG_TIMING_LP, RREG32(MC_SEQ_PMG_TIMING)); + WREG32(MC_SEQ_PMG_CMD_MRS2_LP, RREG32(MC_PMG_CMD_MRS2)); + WREG32(MC_SEQ_WR_CTL_2_LP, RREG32(MC_SEQ_WR_CTL_2)); + + ret = radeon_atom_init_mc_reg_table(rdev, module_index, table); + if (ret) + goto init_mc_done; + + ret = si_copy_vbios_mc_reg_table(table, si_table); + if (ret) + goto init_mc_done; + + si_set_s0_mc_reg_index(si_table); + + ret = si_set_mc_special_registers(rdev, si_table); + if (ret) + goto init_mc_done; + + si_set_valid_flag(si_table); + +init_mc_done: + kfree(table); + + return ret; + +} + +static void si_populate_mc_reg_addresses(struct radeon_device *rdev, + SMC_SIslands_MCRegisters *mc_reg_table) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 i, j; + + for (i = 0, j = 0; j < si_pi->mc_reg_table.last; j++) { + if (si_pi->mc_reg_table.valid_flag & (1 << j)) { + if (i >= SMC_NISLANDS_MC_REGISTER_ARRAY_SIZE) + break; + mc_reg_table->address[i].s0 = + cpu_to_be16(si_pi->mc_reg_table.mc_reg_address[j].s0); + mc_reg_table->address[i].s1 = + cpu_to_be16(si_pi->mc_reg_table.mc_reg_address[j].s1); + i++; + } + } + mc_reg_table->last = (u8)i; +} + +static void si_convert_mc_registers(const struct si_mc_reg_entry *entry, + SMC_SIslands_MCRegisterSet *data, + u32 num_entries, u32 valid_flag) +{ + u32 i, j; + + for(i = 0, j = 0; j < num_entries; j++) { + if (valid_flag & (1 << j)) { + data->value[i] = cpu_to_be32(entry->mc_data[j]); + i++; + } + } +} + +static void si_convert_mc_reg_table_entry_to_smc(struct radeon_device *rdev, + struct rv7xx_pl *pl, + SMC_SIslands_MCRegisterSet *mc_reg_table_data) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + u32 i = 0; + + for (i = 0; i < si_pi->mc_reg_table.num_entries; i++) { + if (pl->mclk <= si_pi->mc_reg_table.mc_reg_table_entry[i].mclk_max) + break; + } + + if ((i == si_pi->mc_reg_table.num_entries) && (i > 0)) + --i; + + si_convert_mc_registers(&si_pi->mc_reg_table.mc_reg_table_entry[i], + mc_reg_table_data, si_pi->mc_reg_table.last, + si_pi->mc_reg_table.valid_flag); +} + +static void si_convert_mc_reg_table_to_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_state, + SMC_SIslands_MCRegisters *mc_reg_table) +{ + struct ni_ps *state = ni_get_ps(radeon_state); + int i; + + for (i = 0; i < state->performance_level_count; i++) { + si_convert_mc_reg_table_entry_to_smc(rdev, + &state->performance_levels[i], + &mc_reg_table->data[SISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT + i]); + } +} + +static int si_populate_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_boot_state) +{ + struct ni_ps *boot_state = ni_get_ps(radeon_boot_state); + struct si_power_info *si_pi = si_get_pi(rdev); + struct si_ulv_param *ulv = &si_pi->ulv; + SMC_SIslands_MCRegisters *smc_mc_reg_table = &si_pi->smc_mc_reg_table; + + memset(smc_mc_reg_table, 0, sizeof(SMC_SIslands_MCRegisters)); + + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_seq_index, 1); + + si_populate_mc_reg_addresses(rdev, smc_mc_reg_table); + + si_convert_mc_reg_table_entry_to_smc(rdev, &boot_state->performance_levels[0], + &smc_mc_reg_table->data[SISLANDS_MCREGISTERTABLE_INITIAL_SLOT]); + + si_convert_mc_registers(&si_pi->mc_reg_table.mc_reg_table_entry[0], + &smc_mc_reg_table->data[SISLANDS_MCREGISTERTABLE_ACPI_SLOT], + si_pi->mc_reg_table.last, + si_pi->mc_reg_table.valid_flag); + + if (ulv->supported && ulv->pl.vddc != 0) + si_convert_mc_reg_table_entry_to_smc(rdev, &ulv->pl, + &smc_mc_reg_table->data[SISLANDS_MCREGISTERTABLE_ULV_SLOT]); + else + si_convert_mc_registers(&si_pi->mc_reg_table.mc_reg_table_entry[0], + &smc_mc_reg_table->data[SISLANDS_MCREGISTERTABLE_ULV_SLOT], + si_pi->mc_reg_table.last, + si_pi->mc_reg_table.valid_flag); + + si_convert_mc_reg_table_to_smc(rdev, radeon_boot_state, smc_mc_reg_table); + + return si_copy_bytes_to_smc(rdev, si_pi->mc_reg_table_start, + (u8 *)smc_mc_reg_table, + sizeof(SMC_SIslands_MCRegisters), si_pi->sram_end); +} + +static int si_upload_mc_reg_table(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state) +{ + struct ni_ps *new_state = ni_get_ps(radeon_new_state); + struct si_power_info *si_pi = si_get_pi(rdev); + u32 address = si_pi->mc_reg_table_start + + offsetof(SMC_SIslands_MCRegisters, + data[SISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT]); + SMC_SIslands_MCRegisters *smc_mc_reg_table = &si_pi->smc_mc_reg_table; + + memset(smc_mc_reg_table, 0, sizeof(SMC_SIslands_MCRegisters)); + + si_convert_mc_reg_table_to_smc(rdev, radeon_new_state, smc_mc_reg_table); + + + return si_copy_bytes_to_smc(rdev, address, + (u8 *)&smc_mc_reg_table->data[SISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT], + sizeof(SMC_SIslands_MCRegisterSet) * new_state->performance_level_count, + si_pi->sram_end); + +} + +static void si_enable_voltage_control(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(GENERAL_PWRMGT, VOLT_PWRMGT_EN, ~VOLT_PWRMGT_EN); + else + WREG32_P(GENERAL_PWRMGT, 0, ~VOLT_PWRMGT_EN); +} + +static enum radeon_pcie_gen si_get_maximum_link_speed(struct radeon_device *rdev, + struct radeon_ps *radeon_state) +{ + struct ni_ps *state = ni_get_ps(radeon_state); + int i; + u16 pcie_speed, max_speed = 0; + + for (i = 0; i < state->performance_level_count; i++) { + pcie_speed = state->performance_levels[i].pcie_gen; + if (max_speed < pcie_speed) + max_speed = pcie_speed; + } + return max_speed; +} + +static u16 si_get_current_pcie_speed(struct radeon_device *rdev) +{ + u32 speed_cntl; + + speed_cntl = RREG32_PCIE_PORT(PCIE_LC_SPEED_CNTL) & LC_CURRENT_DATA_RATE_MASK; + speed_cntl >>= LC_CURRENT_DATA_RATE_SHIFT; + + return (u16)speed_cntl; +} + +static void si_request_link_speed_change_before_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + enum radeon_pcie_gen target_link_speed = si_get_maximum_link_speed(rdev, radeon_new_state); + enum radeon_pcie_gen current_link_speed; + + if (si_pi->force_pcie_gen == RADEON_PCIE_GEN_INVALID) + current_link_speed = si_get_maximum_link_speed(rdev, radeon_current_state); + else + current_link_speed = si_pi->force_pcie_gen; + + si_pi->force_pcie_gen = RADEON_PCIE_GEN_INVALID; + si_pi->pspp_notify_required = false; + if (target_link_speed > current_link_speed) { + switch (target_link_speed) { +#if defined(CONFIG_ACPI) + case RADEON_PCIE_GEN3: + if (radeon_acpi_pcie_performance_request(rdev, PCIE_PERF_REQ_PECI_GEN3, false) == 0) + break; + si_pi->force_pcie_gen = RADEON_PCIE_GEN2; + if (current_link_speed == RADEON_PCIE_GEN2) + break; + case RADEON_PCIE_GEN2: + if (radeon_acpi_pcie_performance_request(rdev, PCIE_PERF_REQ_PECI_GEN2, false) == 0) + break; +#endif + default: + si_pi->force_pcie_gen = si_get_current_pcie_speed(rdev); + break; + } + } else { + if (target_link_speed < current_link_speed) + si_pi->pspp_notify_required = true; + } +} + +static void si_notify_link_speed_change_after_state_change(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + enum radeon_pcie_gen target_link_speed = si_get_maximum_link_speed(rdev, radeon_new_state); + u8 request; + + if (si_pi->pspp_notify_required) { + if (target_link_speed == RADEON_PCIE_GEN3) + request = PCIE_PERF_REQ_PECI_GEN3; + else if (target_link_speed == RADEON_PCIE_GEN2) + request = PCIE_PERF_REQ_PECI_GEN2; + else + request = PCIE_PERF_REQ_PECI_GEN1; + + if ((request == PCIE_PERF_REQ_PECI_GEN1) && + (si_get_current_pcie_speed(rdev) > 0)) + return; + +#if defined(CONFIG_ACPI) + radeon_acpi_pcie_performance_request(rdev, request, false); +#endif + } +} + +#if 0 +static int si_ds_request(struct radeon_device *rdev, + bool ds_status_on, u32 count_write) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + + if (eg_pi->sclk_deep_sleep) { + if (ds_status_on) + return (si_send_msg_to_smc(rdev, PPSMC_MSG_CancelThrottleOVRDSCLKDS) == + PPSMC_Result_OK) ? + 0 : -EINVAL; + else + return (si_send_msg_to_smc(rdev, PPSMC_MSG_ThrottleOVRDSCLKDS) == + PPSMC_Result_OK) ? 0 : -EINVAL; + } + return 0; +} +#endif + +static void si_set_max_cu_value(struct radeon_device *rdev) +{ + struct si_power_info *si_pi = si_get_pi(rdev); + + if (rdev->family == CHIP_VERDE) { + switch (rdev->ddev->pci_device) { + case 0x6820: + case 0x6825: + case 0x6821: + case 0x6823: + case 0x6827: + si_pi->max_cu = 10; + break; + case 0x682D: + case 0x6824: + case 0x682F: + case 0x6826: + si_pi->max_cu = 8; + break; + case 0x6828: + case 0x6830: + case 0x6831: + case 0x6838: + case 0x6839: + case 0x683D: + si_pi->max_cu = 10; + break; + case 0x683B: + case 0x683F: + case 0x6829: + si_pi->max_cu = 8; + break; + default: + si_pi->max_cu = 0; + break; + } + } else { + si_pi->max_cu = 0; + } +} + +static int si_patch_single_dependency_table_based_on_leakage(struct radeon_device *rdev, + struct radeon_clock_voltage_dependency_table *table) +{ + u32 i; + int j; + u16 leakage_voltage; + + if (table) { + for (i = 0; i < table->count; i++) { + switch (si_get_leakage_voltage_from_leakage_index(rdev, + table->entries[i].v, + &leakage_voltage)) { + case 0: + table->entries[i].v = leakage_voltage; + break; + case -EAGAIN: + return -EINVAL; + case -EINVAL: + default: + break; + } + } + + for (j = (table->count - 2); j >= 0; j--) { + table->entries[j].v = (table->entries[j].v <= table->entries[j + 1].v) ? + table->entries[j].v : table->entries[j + 1].v; + } + } + return 0; +} + +static int si_patch_dependency_tables_based_on_leakage(struct radeon_device *rdev) +{ + int ret = 0; + + ret = si_patch_single_dependency_table_based_on_leakage(rdev, + &rdev->pm.dpm.dyn_state.vddc_dependency_on_sclk); + ret = si_patch_single_dependency_table_based_on_leakage(rdev, + &rdev->pm.dpm.dyn_state.vddc_dependency_on_mclk); + ret = si_patch_single_dependency_table_based_on_leakage(rdev, + &rdev->pm.dpm.dyn_state.vddci_dependency_on_mclk); + return ret; +} + +static void si_set_pcie_lane_width_in_smc(struct radeon_device *rdev, + struct radeon_ps *radeon_new_state, + struct radeon_ps *radeon_current_state) +{ + u32 lane_width; + u32 new_lane_width = + (radeon_new_state->caps & ATOM_PPLIB_PCIE_LINK_WIDTH_MASK) >> ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT; + u32 current_lane_width = + (radeon_current_state->caps & ATOM_PPLIB_PCIE_LINK_WIDTH_MASK) >> ATOM_PPLIB_PCIE_LINK_WIDTH_SHIFT; + + if (new_lane_width != current_lane_width) { + radeon_set_pcie_lanes(rdev, new_lane_width); + lane_width = radeon_get_pcie_lanes(rdev); + si_write_smc_soft_register(rdev, SI_SMC_SOFT_REGISTER_non_ulv_pcie_link_width, lane_width); + } +} + +void si_dpm_setup_asic(struct radeon_device *rdev) +{ + rv770_get_memory_type(rdev); + si_read_clock_registers(rdev); + si_enable_acpi_power_management(rdev); +} + +static int si_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) +{ + int low_temp = 0 * 1000; + int high_temp = 255 * 1000; + + if (low_temp < min_temp) + low_temp = min_temp; + if (high_temp > max_temp) + high_temp = max_temp; + if (high_temp < low_temp) { + DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); + return -EINVAL; + } + + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTH(high_temp / 1000), ~DIG_THERM_INTH_MASK); + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTL(low_temp / 1000), ~DIG_THERM_INTL_MASK); + WREG32_P(CG_THERMAL_CTRL, DIG_THERM_DPM(high_temp / 1000), ~DIG_THERM_DPM_MASK); + + rdev->pm.dpm.thermal.min_temp = low_temp; + rdev->pm.dpm.thermal.max_temp = high_temp; + + return 0; +} + +int si_dpm_enable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + int ret; + + if (si_is_smc_running(rdev)) + return -EINVAL; + if (pi->voltage_control) + si_enable_voltage_control(rdev, true); + if (pi->mvdd_control) + si_get_mvdd_configuration(rdev); + if (pi->voltage_control) { + ret = si_construct_voltage_tables(rdev); + if (ret) { + DRM_ERROR("si_construct_voltage_tables failed\n"); + return ret; + } + } + if (eg_pi->dynamic_ac_timing) { + ret = si_initialize_mc_reg_table(rdev); + if (ret) + eg_pi->dynamic_ac_timing = false; + } + if (pi->dynamic_ss) + si_enable_spread_spectrum(rdev, true); + if (pi->thermal_protection) + si_enable_thermal_protection(rdev, true); + si_setup_bsp(rdev); + si_program_git(rdev); + si_program_tp(rdev); + si_program_tpp(rdev); + si_program_sstp(rdev); + si_enable_display_gap(rdev); + si_program_vc(rdev); + ret = si_upload_firmware(rdev); + if (ret) { + DRM_ERROR("si_upload_firmware failed\n"); + return ret; + } + ret = si_process_firmware_header(rdev); + if (ret) { + DRM_ERROR("si_process_firmware_header failed\n"); + return ret; + } + ret = si_initial_switch_from_arb_f0_to_f1(rdev); + if (ret) { + DRM_ERROR("si_initial_switch_from_arb_f0_to_f1 failed\n"); + return ret; + } + ret = si_init_smc_table(rdev); + if (ret) { + DRM_ERROR("si_init_smc_table failed\n"); + return ret; + } + ret = si_init_smc_spll_table(rdev); + if (ret) { + DRM_ERROR("si_init_smc_spll_table failed\n"); + return ret; + } + ret = si_init_arb_table_index(rdev); + if (ret) { + DRM_ERROR("si_init_arb_table_index failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = si_populate_mc_reg_table(rdev, boot_ps); + if (ret) { + DRM_ERROR("si_populate_mc_reg_table failed\n"); + return ret; + } + } + ret = si_initialize_smc_cac_tables(rdev); + if (ret) { + DRM_ERROR("si_initialize_smc_cac_tables failed\n"); + return ret; + } + ret = si_initialize_hardware_cac_manager(rdev); + if (ret) { + DRM_ERROR("si_initialize_hardware_cac_manager failed\n"); + return ret; + } + ret = si_initialize_smc_dte_tables(rdev); + if (ret) { + DRM_ERROR("si_initialize_smc_dte_tables failed\n"); + return ret; + } + ret = si_populate_smc_tdp_limits(rdev, boot_ps); + if (ret) { + DRM_ERROR("si_populate_smc_tdp_limits failed\n"); + return ret; + } + ret = si_populate_smc_tdp_limits_2(rdev, boot_ps); + if (ret) { + DRM_ERROR("si_populate_smc_tdp_limits_2 failed\n"); + return ret; + } + si_program_response_times(rdev); + si_program_ds_registers(rdev); + si_dpm_start_smc(rdev); + ret = si_notify_smc_display_change(rdev, false); + if (ret) { + DRM_ERROR("si_notify_smc_display_change failed\n"); + return ret; + } + si_enable_sclk_control(rdev, true); + si_start_dpm(rdev); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + PPSMC_Result result; + + ret = si_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + result = si_send_msg_to_smc(rdev, PPSMC_MSG_EnableThermalInterrupt); + + if (result != PPSMC_Result_OK) + DRM_DEBUG_KMS("Could not enable thermal interrupts.\n"); + } + + si_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, true); + + ni_update_current_ps(rdev, boot_ps); + + return 0; +} + +void si_dpm_disable(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct radeon_ps *boot_ps = rdev->pm.dpm.boot_ps; + + if (!si_is_smc_running(rdev)) + return; + si_disable_ulv(rdev); + si_clear_vc(rdev); + if (pi->thermal_protection) + si_enable_thermal_protection(rdev, false); + si_enable_power_containment(rdev, boot_ps, false); + si_enable_smc_cac(rdev, boot_ps, false); + si_enable_spread_spectrum(rdev, false); + si_enable_auto_throttle_source(rdev, RADEON_DPM_AUTO_THROTTLE_SRC_THERMAL, false); + si_stop_dpm(rdev); + si_reset_to_default(rdev); + si_dpm_stop_smc(rdev); + si_force_switch_to_arb_f0(rdev); + + ni_update_current_ps(rdev, boot_ps); +} + +int si_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps; + struct radeon_ps *new_ps = &requested_ps; + + ni_update_requested_ps(rdev, new_ps); + + si_apply_state_adjust_rules(rdev, &eg_pi->requested_rps); + + return 0; +} + +static int si_power_control_set_level(struct radeon_device *rdev) +{ + struct radeon_ps *new_ps = rdev->pm.dpm.requested_ps; + int ret; + + ret = si_restrict_performance_levels_before_switch(rdev); + if (ret) + return ret; + ret = si_halt_smc(rdev); + if (ret) + return ret; + ret = si_populate_smc_tdp_limits(rdev, new_ps); + if (ret) + return ret; + ret = si_populate_smc_tdp_limits_2(rdev, new_ps); + if (ret) + return ret; + ret = si_resume_smc(rdev); + if (ret) + return ret; + ret = si_set_sw_state(rdev); + if (ret) + return ret; + return 0; +} + +int si_dpm_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + struct radeon_ps *old_ps = &eg_pi->current_rps; + int ret; + + ret = si_disable_ulv(rdev); + if (ret) { + DRM_ERROR("si_disable_ulv failed\n"); + return ret; + } + ret = si_restrict_performance_levels_before_switch(rdev); + if (ret) { + DRM_ERROR("si_restrict_performance_levels_before_switch failed\n"); + return ret; + } + if (eg_pi->pcie_performance_request) + si_request_link_speed_change_before_state_change(rdev, new_ps, old_ps); + ni_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + ret = si_enable_power_containment(rdev, new_ps, false); + if (ret) { + DRM_ERROR("si_enable_power_containment failed\n"); + return ret; + } + ret = si_enable_smc_cac(rdev, new_ps, false); + if (ret) { + DRM_ERROR("si_enable_smc_cac failed\n"); + return ret; + } + ret = si_halt_smc(rdev); + if (ret) { + DRM_ERROR("si_halt_smc failed\n"); + return ret; + } + ret = si_upload_sw_state(rdev, new_ps); + if (ret) { + DRM_ERROR("si_upload_sw_state failed\n"); + return ret; + } + ret = si_upload_smc_data(rdev); + if (ret) { + DRM_ERROR("si_upload_smc_data failed\n"); + return ret; + } + ret = si_upload_ulv_state(rdev); + if (ret) { + DRM_ERROR("si_upload_ulv_state failed\n"); + return ret; + } + if (eg_pi->dynamic_ac_timing) { + ret = si_upload_mc_reg_table(rdev, new_ps); + if (ret) { + DRM_ERROR("si_upload_mc_reg_table failed\n"); + return ret; + } + } + ret = si_program_memory_timing_parameters(rdev, new_ps); + if (ret) { + DRM_ERROR("si_program_memory_timing_parameters failed\n"); + return ret; + } + si_set_pcie_lane_width_in_smc(rdev, new_ps, old_ps); + + ret = si_resume_smc(rdev); + if (ret) { + DRM_ERROR("si_resume_smc failed\n"); + return ret; + } + ret = si_set_sw_state(rdev); + if (ret) { + DRM_ERROR("si_set_sw_state failed\n"); + return ret; + } + ni_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + if (eg_pi->pcie_performance_request) + si_notify_link_speed_change_after_state_change(rdev, new_ps, old_ps); + ret = si_set_power_state_conditionally_enable_ulv(rdev, new_ps); + if (ret) { + DRM_ERROR("si_set_power_state_conditionally_enable_ulv failed\n"); + return ret; + } + ret = si_enable_smc_cac(rdev, new_ps, true); + if (ret) { + DRM_ERROR("si_enable_smc_cac failed\n"); + return ret; + } + ret = si_enable_power_containment(rdev, new_ps, true); + if (ret) { + DRM_ERROR("si_enable_power_containment failed\n"); + return ret; + } + + ret = si_power_control_set_level(rdev); + if (ret) { + DRM_ERROR("si_power_control_set_level failed\n"); + return ret; + } + + ret = si_dpm_force_performance_level(rdev, RADEON_DPM_FORCED_LEVEL_AUTO); + if (ret) { + DRM_ERROR("si_dpm_force_performance_level failed\n"); + return ret; + } + + return 0; +} + +void si_dpm_post_set_power_state(struct radeon_device *rdev) +{ + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct radeon_ps *new_ps = &eg_pi->requested_rps; + + ni_update_current_ps(rdev, new_ps); +} + + +void si_dpm_reset_asic(struct radeon_device *rdev) +{ + si_restrict_performance_levels_before_switch(rdev); + si_disable_ulv(rdev); + si_set_boot_state(rdev); +} + +void si_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + si_program_display_gap(rdev); +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; + struct _ATOM_PPLIB_SI_CLOCK_INFO si; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void si_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else if (r600_is_uvd_state(rps->class, rps->class2)) { + rps->vclk = RV770_DEFAULT_VCLK_FREQ; + rps->dclk = RV770_DEFAULT_DCLK_FREQ; + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) + rdev->pm.dpm.boot_ps = rps; + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void si_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct rv7xx_power_info *pi = rv770_get_pi(rdev); + struct evergreen_power_info *eg_pi = evergreen_get_pi(rdev); + struct si_power_info *si_pi = si_get_pi(rdev); + struct ni_ps *ps = ni_get_ps(rps); + u16 leakage_voltage; + struct rv7xx_pl *pl = &ps->performance_levels[index]; + int ret; + + ps->performance_level_count = index + 1; + + pl->sclk = le16_to_cpu(clock_info->si.usEngineClockLow); + pl->sclk |= clock_info->si.ucEngineClockHigh << 16; + pl->mclk = le16_to_cpu(clock_info->si.usMemoryClockLow); + pl->mclk |= clock_info->si.ucMemoryClockHigh << 16; + + pl->vddc = le16_to_cpu(clock_info->si.usVDDC); + pl->vddci = le16_to_cpu(clock_info->si.usVDDCI); + pl->flags = le32_to_cpu(clock_info->si.ulFlags); + pl->pcie_gen = r600_get_pcie_gen_support(rdev, + si_pi->sys_pcie_mask, + si_pi->boot_pcie_gen, + clock_info->si.ucPCIEGen); + + /* patch up vddc if necessary */ + ret = si_get_leakage_voltage_from_leakage_index(rdev, pl->vddc, + &leakage_voltage); + if (ret == 0) + pl->vddc = leakage_voltage; + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_ACPI) { + pi->acpi_vddc = pl->vddc; + eg_pi->acpi_vddci = pl->vddci; + si_pi->acpi_pcie_gen = pl->pcie_gen; + } + + if ((rps->class2 & ATOM_PPLIB_CLASSIFICATION2_ULV) && + index == 0) { + /* XXX disable for A0 tahiti */ + si_pi->ulv.supported = true; + si_pi->ulv.pl = *pl; + si_pi->ulv.one_pcie_lane_in_ulv = false; + si_pi->ulv.volt_change_delay = SISLANDS_ULVVOLTAGECHANGEDELAY_DFLT; + si_pi->ulv.cg_ulv_parameter = SISLANDS_CGULVPARAMETER_DFLT; + si_pi->ulv.cg_ulv_control = SISLANDS_CGULVCONTROL_DFLT; + } + + if (pi->min_vddc_in_table > pl->vddc) + pi->min_vddc_in_table = pl->vddc; + + if (pi->max_vddc_in_table < pl->vddc) + pi->max_vddc_in_table = pl->vddc; + + /* patch up boot state */ + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + u16 vddc, vddci, mvdd; + radeon_atombios_get_default_voltages(rdev, &vddc, &vddci, &mvdd); + pl->mclk = rdev->clock.default_mclk; + pl->sclk = rdev->clock.default_sclk; + pl->vddc = vddc; + pl->vddci = vddci; + si_pi->mvdd_bootup_value = mvdd; + } + + if ((rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == + ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE) { + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.sclk = pl->sclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.mclk = pl->mclk; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddc = pl->vddc; + rdev->pm.dpm.dyn_state.max_clock_voltage_on_ac.vddci = pl->vddci; + } +} + +static int si_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j, k, non_clock_array_index, clock_array_index; + union pplib_clock_info *clock_info; + struct _StateArray *state_array; + struct _ClockInfoArray *clock_info_array; + struct _NonClockInfoArray *non_clock_info_array; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + u8 *power_state_offset; + struct ni_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)(mode_info->atom_context->bios + data_offset); + + state_array = (struct _StateArray *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset)); + clock_info_array = (struct _ClockInfoArray *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset)); + non_clock_info_array = (struct _NonClockInfoArray *) + (mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset)); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + state_array->ucNumEntries, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + power_state_offset = (u8 *)state_array->states; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + for (i = 0; i < state_array->ucNumEntries; i++) { + power_state = (union pplib_power_state *)power_state_offset; + non_clock_array_index = power_state->v2.nonClockInfoIndex; + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + &non_clock_info_array->nonClockInfo[non_clock_array_index]; + if (!rdev->pm.power_state[i].clock_info) + return -EINVAL; + ps = kzalloc(sizeof(struct ni_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + si_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + non_clock_info_array->ucEntrySize); + k = 0; + for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) { + clock_array_index = power_state->v2.clockInfoIndex[j]; + if (clock_array_index >= clock_info_array->ucNumEntries) + continue; + if (k >= SISLANDS_MAX_HARDWARE_POWERLEVELS) + break; + clock_info = (union pplib_clock_info *) + &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; + si_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], k, + clock_info); + k++; + } + power_state_offset += 2 + power_state->v2.ucNumDPMLevels; + } + rdev->pm.dpm.num_ps = state_array->ucNumEntries; + return 0; +} + +int si_dpm_init(struct radeon_device *rdev) +{ + struct rv7xx_power_info *pi; + struct evergreen_power_info *eg_pi; + struct ni_power_info *ni_pi; + struct si_power_info *si_pi; + struct atom_clock_dividers dividers; + int ret; + u32 mask; + + si_pi = kzalloc(sizeof(struct si_power_info), GFP_KERNEL); + if (si_pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = si_pi; + ni_pi = &si_pi->ni; + eg_pi = &ni_pi->eg; + pi = &eg_pi->rv7xx; + + ret = drm_pcie_get_speed_cap_mask(rdev->ddev, &mask); + if (ret) + si_pi->sys_pcie_mask = 0; + else + si_pi->sys_pcie_mask = mask; + si_pi->force_pcie_gen = RADEON_PCIE_GEN_INVALID; + si_pi->boot_pcie_gen = si_get_current_pcie_speed(rdev); + + si_set_max_cu_value(rdev); + + rv770_get_max_vddc(rdev); + si_get_leakage_vddc(rdev); + si_patch_dependency_tables_based_on_leakage(rdev); + + pi->acpi_vddc = 0; + eg_pi->acpi_vddci = 0; + pi->min_vddc_in_table = 0; + pi->max_vddc_in_table = 0; + + ret = si_parse_power_table(rdev); + if (ret) + return ret; + ret = r600_parse_extended_power_table(rdev); + if (ret) + return ret; + + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries = + kzalloc(4 * sizeof(struct radeon_clock_voltage_dependency_entry), GFP_KERNEL); + if (!rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries) { + r600_free_extended_power_table(rdev); + return -ENOMEM; + } + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.count = 4; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].clk = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[0].v = 0; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].clk = 36000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[1].v = 720; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].clk = 54000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[2].v = 810; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].clk = 72000; + rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries[3].v = 900; + + if (rdev->pm.dpm.voltage_response_time == 0) + rdev->pm.dpm.voltage_response_time = R600_VOLTAGERESPONSETIME_DFLT; + if (rdev->pm.dpm.backbias_response_time == 0) + rdev->pm.dpm.backbias_response_time = R600_BACKBIASRESPONSETIME_DFLT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 0, false, ÷rs); + if (ret) + pi->ref_div = dividers.ref_div + 1; + else + pi->ref_div = R600_REFERENCEDIVIDER_DFLT; + + eg_pi->smu_uvd_hs = false; + + pi->mclk_strobe_mode_threshold = 40000; + if (si_is_special_1gb_platform(rdev)) + pi->mclk_stutter_mode_threshold = 0; + else + pi->mclk_stutter_mode_threshold = pi->mclk_strobe_mode_threshold; + pi->mclk_edc_enable_threshold = 40000; + eg_pi->mclk_edc_wr_enable_threshold = 40000; + + ni_pi->mclk_rtt_mode_threshold = eg_pi->mclk_edc_wr_enable_threshold; + + pi->voltage_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, VOLTAGE_OBJ_GPIO_LUT); + + pi->mvdd_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_MVDDC, VOLTAGE_OBJ_GPIO_LUT); + + eg_pi->vddci_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDCI, VOLTAGE_OBJ_GPIO_LUT); + + si_pi->vddc_phase_shed_control = + radeon_atom_is_voltage_gpio(rdev, SET_VOLTAGE_TYPE_ASIC_VDDC, VOLTAGE_OBJ_PHASE_LUT); + + rv770_get_engine_memory_ss(rdev); + + pi->asi = RV770_ASI_DFLT; + pi->pasi = CYPRESS_HASI_DFLT; + pi->vrc = SISLANDS_VRC_DFLT; + + pi->gfx_clock_gating = true; + + eg_pi->sclk_deep_sleep = true; + si_pi->sclk_deep_sleep_above_low = false; + + if (rdev->pm.int_thermal_type != THERMAL_TYPE_NONE) + pi->thermal_protection = true; + else + pi->thermal_protection = false; + + eg_pi->dynamic_ac_timing = true; + + eg_pi->light_sleep = true; +#if defined(CONFIG_ACPI) + eg_pi->pcie_performance_request = + radeon_acpi_is_pcie_performance_request_supported(rdev); +#else + eg_pi->pcie_performance_request = false; +#endif + + si_pi->sram_end = SMC_RAM_END; + + rdev->pm.dpm.dyn_state.mclk_sclk_ratio = 4; + rdev->pm.dpm.dyn_state.sclk_mclk_delta = 15000; + rdev->pm.dpm.dyn_state.vddc_vddci_delta = 200; + rdev->pm.dpm.dyn_state.valid_sclk_values.count = 0; + rdev->pm.dpm.dyn_state.valid_sclk_values.values = NULL; + rdev->pm.dpm.dyn_state.valid_mclk_values.count = 0; + rdev->pm.dpm.dyn_state.valid_mclk_values.values = NULL; + + si_initialize_powertune_defaults(rdev); + + return 0; +} + +void si_dpm_fini(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); + kfree(rdev->pm.dpm.dyn_state.vddc_dependency_on_dispclk.entries); + r600_free_extended_power_table(rdev); +} + +void si_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct ni_ps *ps = ni_get_ps(rps); + struct rv7xx_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_STATE_INDEX_MASK) >> + CURRENT_STATE_INDEX_SHIFT; + + if (current_index >= ps->performance_level_count) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + pl = &ps->performance_levels[current_index]; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u mclk: %u vddc: %u vddci: %u pcie gen: %u\n", + current_index, pl->sclk, pl->mclk, pl->vddc, pl->vddci, pl->pcie_gen + 1); + } +} diff --git a/sys/dev/drm/radeon/si_dpm.h b/sys/dev/drm/radeon/si_dpm.h new file mode 100644 index 0000000000..4ce5032cdf --- /dev/null +++ b/sys/dev/drm/radeon/si_dpm.h @@ -0,0 +1,227 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __SI_DPM_H__ +#define __SI_DPM_H__ + +#include "ni_dpm.h" +#include "sislands_smc.h" + +enum si_cac_config_reg_type +{ + SISLANDS_CACCONFIG_MMR = 0, + SISLANDS_CACCONFIG_CGIND, + SISLANDS_CACCONFIG_MAX +}; + +struct si_cac_config_reg +{ + u32 offset; + u32 mask; + u32 shift; + u32 value; + enum si_cac_config_reg_type type; +}; + +struct si_powertune_data +{ + u32 cac_window; + u32 l2_lta_window_size_default; + u8 lts_truncate_default; + u8 shift_n_default; + u8 operating_temp; + struct ni_leakage_coeffients leakage_coefficients; + u32 fixed_kt; + u32 lkge_lut_v0_percent; + u8 dc_cac[NISLANDS_DCCAC_MAX_LEVELS]; + bool enable_powertune_by_default; +}; + +struct si_dyn_powertune_data +{ + u32 cac_leakage; + s32 leakage_minimum_temperature; + u32 wintime; + u32 l2_lta_window_size; + u8 lts_truncate; + u8 shift_n; + u8 dc_pwr_value; + bool disable_uvd_powertune; +}; + +struct si_dte_data +{ + u32 tau[SMC_SISLANDS_DTE_MAX_FILTER_STAGES]; + u32 r[SMC_SISLANDS_DTE_MAX_FILTER_STAGES]; + u32 k; + u32 t0; + u32 max_t; + u8 window_size; + u8 temp_select; + u8 dte_mode; + u8 tdep_count; + u8 t_limits[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + u32 tdep_tau[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + u32 tdep_r[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + u32 t_threshold; + bool enable_dte_by_default; +}; + +struct si_clock_registers { + u32 cg_spll_func_cntl; + u32 cg_spll_func_cntl_2; + u32 cg_spll_func_cntl_3; + u32 cg_spll_func_cntl_4; + u32 cg_spll_spread_spectrum; + u32 cg_spll_spread_spectrum_2; + u32 dll_cntl; + u32 mclk_pwrmgt_cntl; + u32 mpll_ad_func_cntl; + u32 mpll_dq_func_cntl; + u32 mpll_func_cntl; + u32 mpll_func_cntl_1; + u32 mpll_func_cntl_2; + u32 mpll_ss1; + u32 mpll_ss2; +}; + +struct si_mc_reg_entry { + u32 mclk_max; + u32 mc_data[SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +struct si_mc_reg_table { + u8 last; + u8 num_entries; + u16 valid_flag; + struct si_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES]; + SMC_NIslands_MCRegisterAddress mc_reg_address[SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +#define SISLANDS_MCREGISTERTABLE_INITIAL_SLOT 0 +#define SISLANDS_MCREGISTERTABLE_ACPI_SLOT 1 +#define SISLANDS_MCREGISTERTABLE_ULV_SLOT 2 +#define SISLANDS_MCREGISTERTABLE_FIRST_DRIVERSTATE_SLOT 3 + +struct si_leakage_voltage_entry +{ + u16 voltage; + u16 leakage_index; +}; + +#define SISLANDS_LEAKAGE_INDEX0 0xff01 +#define SISLANDS_MAX_LEAKAGE_COUNT 4 + +struct si_leakage_voltage +{ + u16 count; + struct si_leakage_voltage_entry entries[SISLANDS_MAX_LEAKAGE_COUNT]; +}; + +#define SISLANDS_MAX_HARDWARE_POWERLEVELS 5 + +struct si_ulv_param { + bool supported; + u32 cg_ulv_control; + u32 cg_ulv_parameter; + u32 volt_change_delay; + struct rv7xx_pl pl; + bool one_pcie_lane_in_ulv; +}; + +struct si_power_info { + /* must be first! */ + struct ni_power_info ni; + struct si_clock_registers clock_registers; + struct si_mc_reg_table mc_reg_table; + struct atom_voltage_table mvdd_voltage_table; + struct atom_voltage_table vddc_phase_shed_table; + struct si_leakage_voltage leakage_voltage; + u16 mvdd_bootup_value; + struct si_ulv_param ulv; + u32 max_cu; + /* pcie gen */ + enum radeon_pcie_gen force_pcie_gen; + enum radeon_pcie_gen boot_pcie_gen; + enum radeon_pcie_gen acpi_pcie_gen; + u32 sys_pcie_mask; + /* flags */ + bool enable_dte; + bool enable_ppm; + bool vddc_phase_shed_control; + bool pspp_notify_required; + bool sclk_deep_sleep_above_low; + /* smc offsets */ + u32 sram_end; + u32 state_table_start; + u32 soft_regs_start; + u32 mc_reg_table_start; + u32 arb_table_start; + u32 cac_table_start; + u32 dte_table_start; + u32 spll_table_start; + u32 papm_cfg_table_start; + /* CAC stuff */ + const struct si_cac_config_reg *cac_weights; + const struct si_cac_config_reg *lcac_config; + const struct si_cac_config_reg *cac_override; + const struct si_powertune_data *powertune_data; + struct si_dyn_powertune_data dyn_powertune_data; + /* DTE stuff */ + struct si_dte_data dte_data; + /* scratch structs */ + SMC_SIslands_MCRegisters smc_mc_reg_table; + SISLANDS_SMC_STATETABLE smc_statetable; + PP_SIslands_PAPMParameters papm_parm; +}; + +#define SISLANDS_INITIAL_STATE_ARB_INDEX 0 +#define SISLANDS_ACPI_STATE_ARB_INDEX 1 +#define SISLANDS_ULV_STATE_ARB_INDEX 2 +#define SISLANDS_DRIVER_STATE_ARB_INDEX 3 + +#define SISLANDS_DPM2_MAX_PULSE_SKIP 256 + +#define SISLANDS_DPM2_NEAR_TDP_DEC 10 +#define SISLANDS_DPM2_ABOVE_SAFE_INC 5 +#define SISLANDS_DPM2_BELOW_SAFE_INC 20 + +#define SISLANDS_DPM2_TDP_SAFE_LIMIT_PERCENT 80 + +#define SISLANDS_DPM2_MAXPS_PERCENT_H 99 +#define SISLANDS_DPM2_MAXPS_PERCENT_M 99 + +#define SISLANDS_DPM2_SQ_RAMP_MAX_POWER 0x3FFF +#define SISLANDS_DPM2_SQ_RAMP_MIN_POWER 0x12 +#define SISLANDS_DPM2_SQ_RAMP_MAX_POWER_DELTA 0x15 +#define SISLANDS_DPM2_SQ_RAMP_STI_SIZE 0x1E +#define SISLANDS_DPM2_SQ_RAMP_LTI_RATIO 0xF + +#define SISLANDS_DPM2_PWREFFICIENCYRATIO_MARGIN 10 + +#define SISLANDS_VRC_DFLT 0xC000B3 +#define SISLANDS_ULVVOLTAGECHANGEDELAY_DFLT 1687 +#define SISLANDS_CGULVPARAMETER_DFLT 0x00040035 +#define SISLANDS_CGULVCONTROL_DFLT 0x1f007550 + + +#endif diff --git a/sys/dev/drm/radeon/si_reg.h b/sys/dev/drm/radeon/si_reg.h index 53f8a2e36f..501f9d431d 100644 --- a/sys/dev/drm/radeon/si_reg.h +++ b/sys/dev/drm/radeon/si_reg.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/si_reg.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef __SI_REG_H__ #define __SI_REG_H__ diff --git a/sys/dev/drm/radeon/si_smc.c b/sys/dev/drm/radeon/si_smc.c new file mode 100644 index 0000000000..8e90ff8e87 --- /dev/null +++ b/sys/dev/drm/radeon/si_smc.c @@ -0,0 +1,303 @@ +/* + * Copyright 2011 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ + +#include +#include +#include "radeon.h" +#include "sid.h" +#include "ppsmc.h" +#include "radeon_ucode.h" + +int si_set_smc_sram_address(struct radeon_device *rdev, + u32 smc_address, u32 limit); +int si_copy_bytes_to_smc(struct radeon_device *rdev, + u32 smc_start_address, + const u8 *src, u32 byte_count, u32 limit); +void si_start_smc(struct radeon_device *rdev); +void si_reset_smc(struct radeon_device *rdev); +int si_program_jump_on_start(struct radeon_device *rdev); +void si_stop_smc_clock(struct radeon_device *rdev); +void si_start_smc_clock(struct radeon_device *rdev); +bool si_is_smc_running(struct radeon_device *rdev); +PPSMC_Result si_send_msg_to_smc(struct radeon_device *rdev, PPSMC_Msg msg); +PPSMC_Result si_wait_for_smc_inactive(struct radeon_device *rdev); +int si_load_smc_ucode(struct radeon_device *rdev, u32 limit); +int si_read_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 *value, u32 limit); +int si_write_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 value, u32 limit); + +int si_set_smc_sram_address(struct radeon_device *rdev, + u32 smc_address, u32 limit) +{ + if (smc_address & 3) + return -EINVAL; + if ((smc_address + 3) > limit) + return -EINVAL; + + WREG32(SMC_IND_INDEX_0, smc_address); + WREG32_P(SMC_IND_ACCESS_CNTL, 0, ~AUTO_INCREMENT_IND_0); + + return 0; +} + +int si_copy_bytes_to_smc(struct radeon_device *rdev, + u32 smc_start_address, + const u8 *src, u32 byte_count, u32 limit) +{ + int ret; + u32 data, original_data, addr, extra_shift; + + if (smc_start_address & 3) + return -EINVAL; + if ((smc_start_address + byte_count) > limit) + return -EINVAL; + + addr = smc_start_address; + + while (byte_count >= 4) { + /* SMC address space is BE */ + data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3]; + + ret = si_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + WREG32(SMC_IND_DATA_0, data); + + src += 4; + byte_count -= 4; + addr += 4; + } + + /* RMW for the final bytes */ + if (byte_count > 0) { + data = 0; + + ret = si_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + original_data = RREG32(SMC_IND_DATA_0); + + extra_shift = 8 * (4 - byte_count); + + while (byte_count > 0) { + /* SMC address space is BE */ + data = (data << 8) + *src++; + byte_count--; + } + + data <<= extra_shift; + + data |= (original_data & ~((~0UL) << extra_shift)); + + ret = si_set_smc_sram_address(rdev, addr, limit); + if (ret) + return ret; + + WREG32(SMC_IND_DATA_0, data); + } + return 0; +} + +void si_start_smc(struct radeon_device *rdev) +{ + u32 tmp = RREG32_SMC(SMC_SYSCON_RESET_CNTL); + + tmp &= ~RST_REG; + + WREG32_SMC(SMC_SYSCON_RESET_CNTL, tmp); +} + +void si_reset_smc(struct radeon_device *rdev) +{ + u32 tmp; + + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + RREG32(CB_CGTT_SCLK_CTRL); + + tmp = RREG32_SMC(SMC_SYSCON_RESET_CNTL); + tmp |= RST_REG; + WREG32_SMC(SMC_SYSCON_RESET_CNTL, tmp); +} + +int si_program_jump_on_start(struct radeon_device *rdev) +{ + static u8 data[] = { 0x0E, 0x00, 0x40, 0x40 }; + + return si_copy_bytes_to_smc(rdev, 0x0, data, 4, sizeof(data)+1); +} + +void si_stop_smc_clock(struct radeon_device *rdev) +{ + u32 tmp = RREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0); + + tmp |= CK_DISABLE; + + WREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0, tmp); +} + +void si_start_smc_clock(struct radeon_device *rdev) +{ + u32 tmp = RREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0); + + tmp &= ~CK_DISABLE; + + WREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0, tmp); +} + +bool si_is_smc_running(struct radeon_device *rdev) +{ + u32 rst = RREG32_SMC(SMC_SYSCON_RESET_CNTL); + u32 clk = RREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0); + + if (!(rst & RST_REG) && !(clk & CK_DISABLE)) + return true; + + return false; +} + +PPSMC_Result si_send_msg_to_smc(struct radeon_device *rdev, PPSMC_Msg msg) +{ + u32 tmp; + int i; + + if (!si_is_smc_running(rdev)) + return PPSMC_Result_Failed; + + WREG32(SMC_MESSAGE_0, msg); + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32(SMC_RESP_0); + if (tmp != 0) + break; + DRM_UDELAY(1); + } + tmp = RREG32(SMC_RESP_0); + + return (PPSMC_Result)tmp; +} + +PPSMC_Result si_wait_for_smc_inactive(struct radeon_device *rdev) +{ + u32 tmp; + int i; + + if (!si_is_smc_running(rdev)) + return PPSMC_Result_OK; + + for (i = 0; i < rdev->usec_timeout; i++) { + tmp = RREG32_SMC(SMC_SYSCON_CLOCK_CNTL_0); + if ((tmp & CKEN) == 0) + break; + DRM_UDELAY(1); + } + + return PPSMC_Result_OK; +} + +int si_load_smc_ucode(struct radeon_device *rdev, u32 limit) +{ + u32 ucode_start_address; + u32 ucode_size; + const u8 *src; + u32 data; + + if (!rdev->smc_fw) + return -EINVAL; + + switch (rdev->family) { + case CHIP_TAHITI: + ucode_start_address = TAHITI_SMC_UCODE_START; + ucode_size = TAHITI_SMC_UCODE_SIZE; + break; + case CHIP_PITCAIRN: + ucode_start_address = PITCAIRN_SMC_UCODE_START; + ucode_size = PITCAIRN_SMC_UCODE_SIZE; + break; + case CHIP_VERDE: + ucode_start_address = VERDE_SMC_UCODE_START; + ucode_size = VERDE_SMC_UCODE_SIZE; + break; + case CHIP_OLAND: + ucode_start_address = OLAND_SMC_UCODE_START; + ucode_size = OLAND_SMC_UCODE_SIZE; + break; + case CHIP_HAINAN: + ucode_start_address = HAINAN_SMC_UCODE_START; + ucode_size = HAINAN_SMC_UCODE_SIZE; + break; + default: + DRM_ERROR("unknown asic in smc ucode loader\n"); + BUG(); + } + + if (ucode_size & 3) + return -EINVAL; + + src = (const u8 *)rdev->smc_fw->data; + WREG32(SMC_IND_INDEX_0, ucode_start_address); + WREG32_P(SMC_IND_ACCESS_CNTL, AUTO_INCREMENT_IND_0, ~AUTO_INCREMENT_IND_0); + while (ucode_size >= 4) { + /* SMC address space is BE */ + data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3]; + + WREG32(SMC_IND_DATA_0, data); + + src += 4; + ucode_size -= 4; + } + WREG32_P(SMC_IND_ACCESS_CNTL, 0, ~AUTO_INCREMENT_IND_0); + + return 0; +} + +int si_read_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 *value, u32 limit) +{ + int ret; + + ret = si_set_smc_sram_address(rdev, smc_address, limit); + if (ret) + return ret; + + *value = RREG32(SMC_IND_DATA_0); + return 0; +} + +int si_write_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 value, u32 limit) +{ + int ret; + + ret = si_set_smc_sram_address(rdev, smc_address, limit); + if (ret) + return ret; + + WREG32(SMC_IND_DATA_0, value); + return 0; +} diff --git a/sys/dev/drm/radeon/sid.h b/sys/dev/drm/radeon/sid.h index b0003c1640..2c8da27a92 100644 --- a/sys/dev/drm/radeon/sid.h +++ b/sys/dev/drm/radeon/sid.h @@ -20,8 +20,6 @@ * OTHER DEALINGS IN THE SOFTWARE. * * Authors: Alex Deucher - * - * $FreeBSD: head/sys/dev/drm2/radeon/sid.h 254885 2013-08-25 19:37:15Z dumbbell $ */ #ifndef SI_H #define SI_H @@ -32,6 +30,94 @@ #define VERDE_GB_ADDR_CONFIG_GOLDEN 0x12010002 #define HAINAN_GB_ADDR_CONFIG_GOLDEN 0x02010001 +#define SI_MAX_SH_GPRS 256 +#define SI_MAX_TEMP_GPRS 16 +#define SI_MAX_SH_THREADS 256 +#define SI_MAX_SH_STACK_ENTRIES 4096 +#define SI_MAX_FRC_EOV_CNT 16384 +#define SI_MAX_BACKENDS 8 +#define SI_MAX_BACKENDS_MASK 0xFF +#define SI_MAX_BACKENDS_PER_SE_MASK 0x0F +#define SI_MAX_SIMDS 12 +#define SI_MAX_SIMDS_MASK 0x0FFF +#define SI_MAX_SIMDS_PER_SE_MASK 0x00FF +#define SI_MAX_PIPES 8 +#define SI_MAX_PIPES_MASK 0xFF +#define SI_MAX_PIPES_PER_SIMD_MASK 0x3F +#define SI_MAX_LDS_NUM 0xFFFF +#define SI_MAX_TCC 16 +#define SI_MAX_TCC_MASK 0xFFFF + +/* SMC IND accessor regs */ +#define SMC_IND_INDEX_0 0x200 +#define SMC_IND_DATA_0 0x204 + +#define SMC_IND_ACCESS_CNTL 0x228 +# define AUTO_INCREMENT_IND_0 (1 << 0) +#define SMC_MESSAGE_0 0x22c +#define SMC_RESP_0 0x230 + +/* CG IND registers are accessed via SMC indirect space + SMC_CG_IND_START */ +#define SMC_CG_IND_START 0xc0030000 +#define SMC_CG_IND_END 0xc0040000 + +#define CG_CGTT_LOCAL_0 0x400 +#define CG_CGTT_LOCAL_1 0x401 + +/* SMC IND registers */ +#define SMC_SYSCON_RESET_CNTL 0x80000000 +# define RST_REG (1 << 0) +#define SMC_SYSCON_CLOCK_CNTL_0 0x80000004 +# define CK_DISABLE (1 << 0) +# define CKEN (1 << 24) + +#define VGA_HDP_CONTROL 0x328 +#define VGA_MEMORY_DISABLE (1 << 4) + +#define DCCG_DISP_SLOW_SELECT_REG 0x4fc +#define DCCG_DISP1_SLOW_SELECT(x) ((x) << 0) +#define DCCG_DISP1_SLOW_SELECT_MASK (7 << 0) +#define DCCG_DISP1_SLOW_SELECT_SHIFT 0 +#define DCCG_DISP2_SLOW_SELECT(x) ((x) << 4) +#define DCCG_DISP2_SLOW_SELECT_MASK (7 << 4) +#define DCCG_DISP2_SLOW_SELECT_SHIFT 4 + +#define CG_SPLL_FUNC_CNTL 0x600 +#define SPLL_RESET (1 << 0) +#define SPLL_SLEEP (1 << 1) +#define SPLL_BYPASS_EN (1 << 3) +#define SPLL_REF_DIV(x) ((x) << 4) +#define SPLL_REF_DIV_MASK (0x3f << 4) +#define SPLL_PDIV_A(x) ((x) << 20) +#define SPLL_PDIV_A_MASK (0x7f << 20) +#define SPLL_PDIV_A_SHIFT 20 +#define CG_SPLL_FUNC_CNTL_2 0x604 +#define SCLK_MUX_SEL(x) ((x) << 0) +#define SCLK_MUX_SEL_MASK (0x1ff << 0) +#define CG_SPLL_FUNC_CNTL_3 0x608 +#define SPLL_FB_DIV(x) ((x) << 0) +#define SPLL_FB_DIV_MASK (0x3ffffff << 0) +#define SPLL_FB_DIV_SHIFT 0 +#define SPLL_DITHEN (1 << 28) +#define CG_SPLL_FUNC_CNTL_4 0x60c + +#define SPLL_CNTL_MODE 0x618 +# define SPLL_REFCLK_SEL(x) ((x) << 8) +# define SPLL_REFCLK_SEL_MASK 0xFF00 + +#define CG_SPLL_SPREAD_SPECTRUM 0x620 +#define SSEN (1 << 0) +#define CLK_S(x) ((x) << 4) +#define CLK_S_MASK (0xfff << 4) +#define CLK_S_SHIFT 4 +#define CG_SPLL_SPREAD_SPECTRUM_2 0x624 +#define CLK_V(x) ((x) << 0) +#define CLK_V_MASK (0x3ffffff << 0) +#define CLK_V_SHIFT 0 + +#define CG_SPLL_AUTOSCALE_CNTL 0x62c +# define AUTOSCALE_ON_SS_CLEAR (1 << 9) + /* discrete uvd clocks */ #define CG_UPLL_FUNC_CNTL 0x634 # define UPLL_RESET_MASK 0x00000001 @@ -61,6 +147,45 @@ #define CG_UPLL_SPREAD_SPECTRUM 0x650 # define SSEN_MASK 0x00000001 +#define MPLL_BYPASSCLK_SEL 0x65c +# define MPLL_CLKOUT_SEL(x) ((x) << 8) +# define MPLL_CLKOUT_SEL_MASK 0xFF00 + +#define CG_CLKPIN_CNTL 0x660 +# define XTALIN_DIVIDE (1 << 1) +# define BCLK_AS_XCLK (1 << 2) +#define CG_CLKPIN_CNTL_2 0x664 +# define FORCE_BIF_REFCLK_EN (1 << 3) +# define MUX_TCLK_TO_XCLK (1 << 8) + +#define THM_CLK_CNTL 0x66c +# define CMON_CLK_SEL(x) ((x) << 0) +# define CMON_CLK_SEL_MASK 0xFF +# define TMON_CLK_SEL(x) ((x) << 8) +# define TMON_CLK_SEL_MASK 0xFF00 +#define MISC_CLK_CNTL 0x670 +# define DEEP_SLEEP_CLK_SEL(x) ((x) << 0) +# define DEEP_SLEEP_CLK_SEL_MASK 0xFF +# define ZCLK_SEL(x) ((x) << 8) +# define ZCLK_SEL_MASK 0xFF00 + +#define CG_THERMAL_CTRL 0x700 +#define DPM_EVENT_SRC(x) ((x) << 0) +#define DPM_EVENT_SRC_MASK (7 << 0) +#define DIG_THERM_DPM(x) ((x) << 14) +#define DIG_THERM_DPM_MASK 0x003FC000 +#define DIG_THERM_DPM_SHIFT 14 + +#define CG_THERMAL_INT 0x708 +#define DIG_THERM_INTH(x) ((x) << 8) +#define DIG_THERM_INTH_MASK 0x0000FF00 +#define DIG_THERM_INTH_SHIFT 8 +#define DIG_THERM_INTL(x) ((x) << 16) +#define DIG_THERM_INTL_MASK 0x00FF0000 +#define DIG_THERM_INTL_SHIFT 16 +#define THERM_INT_MASK_HIGH (1 << 24) +#define THERM_INT_MASK_LOW (1 << 25) + #define CG_MULT_THERMAL_STATUS 0x714 #define ASIC_MAX_TEMP(x) ((x) << 0) #define ASIC_MAX_TEMP_MASK 0x000001ff @@ -69,31 +194,89 @@ #define CTF_TEMP_MASK 0x0003fe00 #define CTF_TEMP_SHIFT 9 -#define SI_MAX_SH_GPRS 256 -#define SI_MAX_TEMP_GPRS 16 -#define SI_MAX_SH_THREADS 256 -#define SI_MAX_SH_STACK_ENTRIES 4096 -#define SI_MAX_FRC_EOV_CNT 16384 -#define SI_MAX_BACKENDS 8 -#define SI_MAX_BACKENDS_MASK 0xFF -#define SI_MAX_BACKENDS_PER_SE_MASK 0x0F -#define SI_MAX_SIMDS 12 -#define SI_MAX_SIMDS_MASK 0x0FFF -#define SI_MAX_SIMDS_PER_SE_MASK 0x00FF -#define SI_MAX_PIPES 8 -#define SI_MAX_PIPES_MASK 0xFF -#define SI_MAX_PIPES_PER_SIMD_MASK 0x3F -#define SI_MAX_LDS_NUM 0xFFFF -#define SI_MAX_TCC 16 -#define SI_MAX_TCC_MASK 0xFFFF - -#define VGA_HDP_CONTROL 0x328 -#define VGA_MEMORY_DISABLE (1 << 4) - -#define CG_CLKPIN_CNTL 0x660 -# define XTALIN_DIVIDE (1 << 1) -#define CG_CLKPIN_CNTL_2 0x664 -# define MUX_TCLK_TO_XCLK (1 << 8) +#define GENERAL_PWRMGT 0x780 +# define GLOBAL_PWRMGT_EN (1 << 0) +# define STATIC_PM_EN (1 << 1) +# define THERMAL_PROTECTION_DIS (1 << 2) +# define THERMAL_PROTECTION_TYPE (1 << 3) +# define SW_SMIO_INDEX(x) ((x) << 6) +# define SW_SMIO_INDEX_MASK (1 << 6) +# define SW_SMIO_INDEX_SHIFT 6 +# define VOLT_PWRMGT_EN (1 << 10) +# define DYN_SPREAD_SPECTRUM_EN (1 << 23) +#define CG_TPC 0x784 +#define SCLK_PWRMGT_CNTL 0x788 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) +# define DYN_LIGHT_SLEEP_EN (1 << 14) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x798 +# define CURRENT_STATE_INDEX_MASK (0xf << 4) +# define CURRENT_STATE_INDEX_SHIFT 4 + +#define CG_FTV 0x7bc + +#define CG_FFCT_0 0x7c0 +# define UTC_0(x) ((x) << 0) +# define UTC_0_MASK (0x3ff << 0) +# define DTC_0(x) ((x) << 10) +# define DTC_0_MASK (0x3ff << 10) + +#define CG_BSP 0x7fc +# define BSP(x) ((x) << 0) +# define BSP_MASK (0xffff << 0) +# define BSU(x) ((x) << 16) +# define BSU_MASK (0xf << 16) +#define CG_AT 0x800 +# define CG_R(x) ((x) << 0) +# define CG_R_MASK (0xffff << 0) +# define CG_L(x) ((x) << 16) +# define CG_L_MASK (0xffff << 16) + +#define CG_GIT 0x804 +# define CG_GICST(x) ((x) << 0) +# define CG_GICST_MASK (0xffff << 0) +# define CG_GIPOT(x) ((x) << 16) +# define CG_GIPOT_MASK (0xffff << 16) + +#define CG_SSP 0x80c +# define SST(x) ((x) << 0) +# define SST_MASK (0xffff << 0) +# define SSTU(x) ((x) << 16) +# define SSTU_MASK (0xf << 16) + +#define CG_DISPLAY_GAP_CNTL 0x828 +# define DISP1_GAP(x) ((x) << 0) +# define DISP1_GAP_MASK (3 << 0) +# define DISP2_GAP(x) ((x) << 2) +# define DISP2_GAP_MASK (3 << 2) +# define VBI_TIMER_COUNT(x) ((x) << 4) +# define VBI_TIMER_COUNT_MASK (0x3fff << 4) +# define VBI_TIMER_UNIT(x) ((x) << 20) +# define VBI_TIMER_UNIT_MASK (7 << 20) +# define DISP1_GAP_MCHG(x) ((x) << 24) +# define DISP1_GAP_MCHG_MASK (3 << 24) +# define DISP2_GAP_MCHG(x) ((x) << 26) +# define DISP2_GAP_MCHG_MASK (3 << 26) + +#define CG_ULV_CONTROL 0x878 +#define CG_ULV_PARAMETER 0x87c + +#define SMC_SCRATCH0 0x884 + +#define CG_CAC_CTRL 0x8b8 +# define CAC_WINDOW(x) ((x) << 0) +# define CAC_WINDOW_MASK 0x00ffffff #define DMIF_ADDR_CONFIG 0xBD4 @@ -184,6 +367,20 @@ #define VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x14FC #define VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x14DC +#define PROTECTIONS_MASK (0xf << 0) +#define PROTECTIONS_SHIFT 0 + /* bit 0: range + * bit 1: pde0 + * bit 2: valid + * bit 3: read + * bit 4: write + */ +#define MEMORY_CLIENT_ID_MASK (0xff << 12) +#define MEMORY_CLIENT_ID_SHIFT 12 +#define MEMORY_CLIENT_RW_MASK (1 << 24) +#define MEMORY_CLIENT_RW_SHIFT 24 +#define FAULT_VMID_MASK (0xf << 25) +#define FAULT_VMID_SHIFT 25 #define VM_INVALIDATE_REQUEST 0x1478 #define VM_INVALIDATE_RESPONSE 0x147c @@ -205,6 +402,10 @@ #define VM_CONTEXT0_PAGE_TABLE_END_ADDR 0x157C #define VM_CONTEXT1_PAGE_TABLE_END_ADDR 0x1580 +#define VM_L2_CG 0x15c0 +#define MC_CG_ENABLE (1 << 18) +#define MC_LS_ENABLE (1 << 19) + #define MC_SHARED_CHMAP 0x2004 #define NOOFCHAN_SHIFT 12 #define NOOFCHAN_MASK 0x0000f000 @@ -230,6 +431,17 @@ #define MC_SHARED_BLACKOUT_CNTL 0x20ac +#define MC_HUB_MISC_HUB_CG 0x20b8 +#define MC_HUB_MISC_VM_CG 0x20bc + +#define MC_HUB_MISC_SIP_CG 0x20c0 + +#define MC_XPB_CLK_GAT 0x2478 + +#define MC_CITF_MISC_RD_CG 0x2648 +#define MC_CITF_MISC_WR_CG 0x264c +#define MC_CITF_MISC_VM_CG 0x2650 + #define MC_ARB_RAMCFG 0x2760 #define NOOFBANK_SHIFT 0 #define NOOFBANK_MASK 0x00000003 @@ -245,6 +457,23 @@ #define NOOFGROUPS_SHIFT 12 #define NOOFGROUPS_MASK 0x00001000 +#define MC_ARB_DRAM_TIMING 0x2774 +#define MC_ARB_DRAM_TIMING2 0x2778 + +#define MC_ARB_BURST_TIME 0x2808 +#define STATE0(x) ((x) << 0) +#define STATE0_MASK (0x1f << 0) +#define STATE0_SHIFT 0 +#define STATE1(x) ((x) << 5) +#define STATE1_MASK (0x1f << 5) +#define STATE1_SHIFT 5 +#define STATE2(x) ((x) << 10) +#define STATE2_MASK (0x1f << 10) +#define STATE2_SHIFT 10 +#define STATE3(x) ((x) << 15) +#define STATE3_MASK (0x1f << 15) +#define STATE3_SHIFT 15 + #define MC_SEQ_TRAIN_WAKEUP_CNTL 0x2808 #define TRAIN_DONE_D0 (1 << 30) #define TRAIN_DONE_D1 (1 << 31) @@ -252,13 +481,105 @@ #define MC_SEQ_SUP_CNTL 0x28c8 #define RUN_MASK (1 << 0) #define MC_SEQ_SUP_PGM 0x28cc +#define MC_PMG_AUTO_CMD 0x28d0 #define MC_IO_PAD_CNTL_D0 0x29d0 #define MEM_FALL_OUT_CMD (1 << 8) +#define MC_SEQ_RAS_TIMING 0x28a0 +#define MC_SEQ_CAS_TIMING 0x28a4 +#define MC_SEQ_MISC_TIMING 0x28a8 +#define MC_SEQ_MISC_TIMING2 0x28ac +#define MC_SEQ_PMG_TIMING 0x28b0 +#define MC_SEQ_RD_CTL_D0 0x28b4 +#define MC_SEQ_RD_CTL_D1 0x28b8 +#define MC_SEQ_WR_CTL_D0 0x28bc +#define MC_SEQ_WR_CTL_D1 0x28c0 + +#define MC_SEQ_MISC0 0x2a00 +#define MC_SEQ_MISC0_VEN_ID_SHIFT 8 +#define MC_SEQ_MISC0_VEN_ID_MASK 0x00000f00 +#define MC_SEQ_MISC0_VEN_ID_VALUE 3 +#define MC_SEQ_MISC0_REV_ID_SHIFT 12 +#define MC_SEQ_MISC0_REV_ID_MASK 0x0000f000 +#define MC_SEQ_MISC0_REV_ID_VALUE 1 +#define MC_SEQ_MISC0_GDDR5_SHIFT 28 +#define MC_SEQ_MISC0_GDDR5_MASK 0xf0000000 +#define MC_SEQ_MISC0_GDDR5_VALUE 5 +#define MC_SEQ_MISC1 0x2a04 +#define MC_SEQ_RESERVE_M 0x2a08 +#define MC_PMG_CMD_EMRS 0x2a0c + #define MC_SEQ_IO_DEBUG_INDEX 0x2a44 #define MC_SEQ_IO_DEBUG_DATA 0x2a48 +#define MC_SEQ_MISC5 0x2a54 +#define MC_SEQ_MISC6 0x2a58 + +#define MC_SEQ_MISC7 0x2a64 + +#define MC_SEQ_RAS_TIMING_LP 0x2a6c +#define MC_SEQ_CAS_TIMING_LP 0x2a70 +#define MC_SEQ_MISC_TIMING_LP 0x2a74 +#define MC_SEQ_MISC_TIMING2_LP 0x2a78 +#define MC_SEQ_WR_CTL_D0_LP 0x2a7c +#define MC_SEQ_WR_CTL_D1_LP 0x2a80 +#define MC_SEQ_PMG_CMD_EMRS_LP 0x2a84 +#define MC_SEQ_PMG_CMD_MRS_LP 0x2a88 + +#define MC_PMG_CMD_MRS 0x2aac + +#define MC_SEQ_RD_CTL_D0_LP 0x2b1c +#define MC_SEQ_RD_CTL_D1_LP 0x2b20 + +#define MC_PMG_CMD_MRS1 0x2b44 +#define MC_SEQ_PMG_CMD_MRS1_LP 0x2b48 +#define MC_SEQ_PMG_TIMING_LP 0x2b4c + +#define MC_SEQ_WR_CTL_2 0x2b54 +#define MC_SEQ_WR_CTL_2_LP 0x2b58 +#define MC_PMG_CMD_MRS2 0x2b5c +#define MC_SEQ_PMG_CMD_MRS2_LP 0x2b60 + +#define MCLK_PWRMGT_CNTL 0x2ba0 +# define DLL_SPEED(x) ((x) << 0) +# define DLL_SPEED_MASK (0x1f << 0) +# define DLL_READY (1 << 6) +# define MC_INT_CNTL (1 << 7) +# define MRDCK0_PDNB (1 << 8) +# define MRDCK1_PDNB (1 << 9) +# define MRDCK0_RESET (1 << 16) +# define MRDCK1_RESET (1 << 17) +# define DLL_READY_READ (1 << 24) +#define DLL_CNTL 0x2ba4 +# define MRDCK0_BYPASS (1 << 24) +# define MRDCK1_BYPASS (1 << 25) + +#define MPLL_FUNC_CNTL 0x2bb4 +#define BWCTRL(x) ((x) << 20) +#define BWCTRL_MASK (0xff << 20) +#define MPLL_FUNC_CNTL_1 0x2bb8 +#define VCO_MODE(x) ((x) << 0) +#define VCO_MODE_MASK (3 << 0) +#define CLKFRAC(x) ((x) << 4) +#define CLKFRAC_MASK (0xfff << 4) +#define CLKF(x) ((x) << 16) +#define CLKF_MASK (0xfff << 16) +#define MPLL_FUNC_CNTL_2 0x2bbc +#define MPLL_AD_FUNC_CNTL 0x2bc0 +#define YCLK_POST_DIV(x) ((x) << 0) +#define YCLK_POST_DIV_MASK (7 << 0) +#define MPLL_DQ_FUNC_CNTL 0x2bc4 +#define YCLK_SEL(x) ((x) << 4) +#define YCLK_SEL_MASK (1 << 4) + +#define MPLL_SS1 0x2bcc +#define CLKV(x) ((x) << 0) +#define CLKV_MASK (0x3ffffff << 0) +#define MPLL_SS2 0x2bd0 +#define CLKS(x) ((x) << 0) +#define CLKS_MASK (0xfff << 0) + #define HDP_HOST_PATH_CNTL 0x2C00 #define HDP_NONSURFACE_BASE 0x2C04 #define HDP_NONSURFACE_INFO 0x2C08 @@ -268,6 +589,8 @@ #define HDP_MISC_CNTL 0x2F4C #define HDP_FLUSH_INVALIDATE_CACHE (1 << 0) +#define ATC_MISC_CG 0x3350 + #define IH_RB_CNTL 0x3e00 # define IH_RB_ENABLE (1 << 0) # define IH_IB_SIZE(x) ((x) << 1) /* log2 */ @@ -426,6 +749,9 @@ # define DC_HPDx_RX_INT_TIMER(x) ((x) << 16) # define DC_HPDx_EN (1 << 28) +#define DPG_PIPE_STUTTER_CONTROL 0x6cd4 +# define STUTTER_ENABLE (1 << 0) + /* 0x6e98, 0x7a98, 0x10698, 0x11298, 0x11e98, 0x12a98 */ #define CRTC_STATUS_FRAME_COUNT 0x6e98 @@ -601,6 +927,24 @@ #define SQC_CACHES 0x8C08 +#define SQ_POWER_THROTTLE 0x8e58 +#define MIN_POWER(x) ((x) << 0) +#define MIN_POWER_MASK (0x3fff << 0) +#define MIN_POWER_SHIFT 0 +#define MAX_POWER(x) ((x) << 16) +#define MAX_POWER_MASK (0x3fff << 16) +#define MAX_POWER_SHIFT 0 +#define SQ_POWER_THROTTLE2 0x8e5c +#define MAX_POWER_DELTA(x) ((x) << 0) +#define MAX_POWER_DELTA_MASK (0x3fff << 0) +#define MAX_POWER_DELTA_SHIFT 0 +#define STI_SIZE(x) ((x) << 16) +#define STI_SIZE_MASK (0x3ff << 16) +#define STI_SIZE_SHIFT 16 +#define LTI_RATIO(x) ((x) << 27) +#define LTI_RATIO_MASK (0xf << 27) +#define LTI_RATIO_SHIFT 27 + #define SX_DEBUG_1 0x9060 #define SPI_STATIC_THREAD_MGMT_1 0x90E0 @@ -618,6 +962,11 @@ #define CGTS_USER_TCC_DISABLE 0x914C #define TCC_DISABLE_MASK 0xFFFF0000 #define TCC_DISABLE_SHIFT 16 +#define CGTS_SM_CTRL_REG 0x9150 +#define OVERRIDE (1 << 21) +#define LS_OVERRIDE (1 << 22) + +#define SPI_LB_CU_MASK 0x9354 #define TA_CNTL_AUX 0x9508 @@ -707,6 +1056,8 @@ #define CB_PERFCOUNTER3_SELECT0 0x9a38 #define CB_PERFCOUNTER3_SELECT1 0x9a3c +#define CB_CGTT_SCLK_CTRL 0x9a60 + #define GC_USER_RB_BACKEND_DISABLE 0x9B7C #define BACKEND_DISABLE_MASK 0x00FF0000 #define BACKEND_DISABLE_SHIFT 16 @@ -764,6 +1115,9 @@ # define CP_RINGID1_INT_STAT (1 << 30) # define CP_RINGID0_INT_STAT (1 << 31) +#define CP_MEM_SLP_CNTL 0xC1E4 +# define CP_MEM_LS_EN (1 << 0) + #define CP_DEBUG 0xC1FC #define RLC_CNTL 0xC300 @@ -771,6 +1125,7 @@ #define RLC_RL_BASE 0xC304 #define RLC_RL_SIZE 0xC308 #define RLC_LB_CNTL 0xC30C +# define LOAD_BALANCE_ENABLE (1 << 0) #define RLC_SAVE_AND_RESTORE_BASE 0xC310 #define RLC_LB_CNTR_MAX 0xC314 #define RLC_LB_CNTR_INIT 0xC318 @@ -785,6 +1140,56 @@ #define RLC_CAPTURE_GPU_CLOCK_COUNT 0xC340 #define RLC_MC_CNTL 0xC344 #define RLC_UCODE_CNTL 0xC348 +#define RLC_STAT 0xC34C +# define RLC_BUSY_STATUS (1 << 0) +# define GFX_POWER_STATUS (1 << 1) +# define GFX_CLOCK_STATUS (1 << 2) +# define GFX_LS_STATUS (1 << 3) + +#define RLC_PG_CNTL 0xC35C +# define GFX_PG_ENABLE (1 << 0) +# define GFX_PG_SRC (1 << 1) + +#define RLC_CGTT_MGCG_OVERRIDE 0xC400 +#define RLC_CGCG_CGLS_CTRL 0xC404 +# define CGCG_EN (1 << 0) +# define CGLS_EN (1 << 1) + +#define RLC_TTOP_D 0xC414 +# define RLC_PUD(x) ((x) << 0) +# define RLC_PUD_MASK (0xff << 0) +# define RLC_PDD(x) ((x) << 8) +# define RLC_PDD_MASK (0xff << 8) +# define RLC_TTPD(x) ((x) << 16) +# define RLC_TTPD_MASK (0xff << 16) +# define RLC_MSD(x) ((x) << 24) +# define RLC_MSD_MASK (0xff << 24) + +#define RLC_LB_INIT_CU_MASK 0xC41C + +#define RLC_PG_AO_CU_MASK 0xC42C +#define RLC_MAX_PG_CU 0xC430 +# define MAX_PU_CU(x) ((x) << 0) +# define MAX_PU_CU_MASK (0xff << 0) +#define RLC_AUTO_PG_CTRL 0xC434 +# define AUTO_PG_EN (1 << 0) +# define GRBM_REG_SGIT(x) ((x) << 3) +# define GRBM_REG_SGIT_MASK (0xffff << 3) +# define PG_AFTER_GRBM_REG_ST(x) ((x) << 19) +# define PG_AFTER_GRBM_REG_ST_MASK (0x1fff << 19) + +#define RLC_SERDES_WR_MASTER_MASK_0 0xC454 +#define RLC_SERDES_WR_MASTER_MASK_1 0xC458 +#define RLC_SERDES_WR_CTRL 0xC45C + +#define RLC_SERDES_MASTER_BUSY_0 0xC464 +#define RLC_SERDES_MASTER_BUSY_1 0xC468 + +#define RLC_GCPM_GENERAL_3 0xC478 + +#define DB_RENDER_CONTROL 0x28000 + +#define DB_DEPTH_INFO 0x2803c #define PA_SC_RASTER_CONFIG 0x28350 # define RASTER_CONFIG_RB_MAP_0 0 @@ -831,6 +1236,146 @@ # define THREAD_TRACE_FLUSH (54 << 0) # define THREAD_TRACE_FINISH (55 << 0) +/* PIF PHY0 registers idx/data 0x8/0xc */ +#define PB0_PIF_CNTL 0x10 +# define LS2_EXIT_TIME(x) ((x) << 17) +# define LS2_EXIT_TIME_MASK (0x7 << 17) +# define LS2_EXIT_TIME_SHIFT 17 +#define PB0_PIF_PAIRING 0x11 +# define MULTI_PIF (1 << 25) +#define PB0_PIF_PWRDOWN_0 0x12 +# define PLL_POWER_STATE_IN_TXS2_0(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_0_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_0_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_0(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_0_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_0_SHIFT 10 +# define PLL_RAMP_UP_TIME_0(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_0_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_0_SHIFT 24 +#define PB0_PIF_PWRDOWN_1 0x13 +# define PLL_POWER_STATE_IN_TXS2_1(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_1_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_1_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_1(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_1_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_1_SHIFT 10 +# define PLL_RAMP_UP_TIME_1(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_1_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_1_SHIFT 24 + +#define PB0_PIF_PWRDOWN_2 0x17 +# define PLL_POWER_STATE_IN_TXS2_2(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_2_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_2_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_2(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_2_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_2_SHIFT 10 +# define PLL_RAMP_UP_TIME_2(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_2_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_2_SHIFT 24 +#define PB0_PIF_PWRDOWN_3 0x18 +# define PLL_POWER_STATE_IN_TXS2_3(x) ((x) << 7) +# define PLL_POWER_STATE_IN_TXS2_3_MASK (0x7 << 7) +# define PLL_POWER_STATE_IN_TXS2_3_SHIFT 7 +# define PLL_POWER_STATE_IN_OFF_3(x) ((x) << 10) +# define PLL_POWER_STATE_IN_OFF_3_MASK (0x7 << 10) +# define PLL_POWER_STATE_IN_OFF_3_SHIFT 10 +# define PLL_RAMP_UP_TIME_3(x) ((x) << 24) +# define PLL_RAMP_UP_TIME_3_MASK (0x7 << 24) +# define PLL_RAMP_UP_TIME_3_SHIFT 24 +/* PIF PHY1 registers idx/data 0x10/0x14 */ +#define PB1_PIF_CNTL 0x10 +#define PB1_PIF_PAIRING 0x11 +#define PB1_PIF_PWRDOWN_0 0x12 +#define PB1_PIF_PWRDOWN_1 0x13 + +#define PB1_PIF_PWRDOWN_2 0x17 +#define PB1_PIF_PWRDOWN_3 0x18 +/* PCIE registers idx/data 0x30/0x34 */ +#define PCIE_CNTL2 0x1c /* PCIE */ +# define SLV_MEM_LS_EN (1 << 16) +# define MST_MEM_LS_EN (1 << 18) +# define REPLAY_MEM_LS_EN (1 << 19) +#define PCIE_LC_STATUS1 0x28 /* PCIE */ +# define LC_REVERSE_RCVR (1 << 0) +# define LC_REVERSE_XMIT (1 << 1) +# define LC_OPERATING_LINK_WIDTH_MASK (0x7 << 2) +# define LC_OPERATING_LINK_WIDTH_SHIFT 2 +# define LC_DETECTED_LINK_WIDTH_MASK (0x7 << 5) +# define LC_DETECTED_LINK_WIDTH_SHIFT 5 + +#define PCIE_P_CNTL 0x40 /* PCIE */ +# define P_IGNORE_EDB_ERR (1 << 6) + +/* PCIE PORT registers idx/data 0x38/0x3c */ +#define PCIE_LC_CNTL 0xa0 +# define LC_L0S_INACTIVITY(x) ((x) << 8) +# define LC_L0S_INACTIVITY_MASK (0xf << 8) +# define LC_L0S_INACTIVITY_SHIFT 8 +# define LC_L1_INACTIVITY(x) ((x) << 12) +# define LC_L1_INACTIVITY_MASK (0xf << 12) +# define LC_L1_INACTIVITY_SHIFT 12 +# define LC_PMI_TO_L1_DIS (1 << 16) +# define LC_ASPM_TO_L1_DIS (1 << 24) +#define PCIE_LC_LINK_WIDTH_CNTL 0xa2 /* PCIE_P */ +# define LC_LINK_WIDTH_SHIFT 0 +# define LC_LINK_WIDTH_MASK 0x7 +# define LC_LINK_WIDTH_X0 0 +# define LC_LINK_WIDTH_X1 1 +# define LC_LINK_WIDTH_X2 2 +# define LC_LINK_WIDTH_X4 3 +# define LC_LINK_WIDTH_X8 4 +# define LC_LINK_WIDTH_X16 6 +# define LC_LINK_WIDTH_RD_SHIFT 4 +# define LC_LINK_WIDTH_RD_MASK 0x70 +# define LC_RECONFIG_ARC_MISSING_ESCAPE (1 << 7) +# define LC_RECONFIG_NOW (1 << 8) +# define LC_RENEGOTIATION_SUPPORT (1 << 9) +# define LC_RENEGOTIATE_EN (1 << 10) +# define LC_SHORT_RECONFIG_EN (1 << 11) +# define LC_UPCONFIGURE_SUPPORT (1 << 12) +# define LC_UPCONFIGURE_DIS (1 << 13) +# define LC_DYN_LANES_PWR_STATE(x) ((x) << 21) +# define LC_DYN_LANES_PWR_STATE_MASK (0x3 << 21) +# define LC_DYN_LANES_PWR_STATE_SHIFT 21 +#define PCIE_LC_N_FTS_CNTL 0xa3 /* PCIE_P */ +# define LC_XMIT_N_FTS(x) ((x) << 0) +# define LC_XMIT_N_FTS_MASK (0xff << 0) +# define LC_XMIT_N_FTS_SHIFT 0 +# define LC_XMIT_N_FTS_OVERRIDE_EN (1 << 8) +# define LC_N_FTS_MASK (0xff << 24) +#define PCIE_LC_SPEED_CNTL 0xa4 /* PCIE_P */ +# define LC_GEN2_EN_STRAP (1 << 0) +# define LC_GEN3_EN_STRAP (1 << 1) +# define LC_TARGET_LINK_SPEED_OVERRIDE_EN (1 << 2) +# define LC_TARGET_LINK_SPEED_OVERRIDE_MASK (0x3 << 3) +# define LC_TARGET_LINK_SPEED_OVERRIDE_SHIFT 3 +# define LC_FORCE_EN_SW_SPEED_CHANGE (1 << 5) +# define LC_FORCE_DIS_SW_SPEED_CHANGE (1 << 6) +# define LC_FORCE_EN_HW_SPEED_CHANGE (1 << 7) +# define LC_FORCE_DIS_HW_SPEED_CHANGE (1 << 8) +# define LC_INITIATE_LINK_SPEED_CHANGE (1 << 9) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_MASK (0x3 << 10) +# define LC_SPEED_CHANGE_ATTEMPTS_ALLOWED_SHIFT 10 +# define LC_CURRENT_DATA_RATE_MASK (0x3 << 13) /* 0/1/2 = gen1/2/3 */ +# define LC_CURRENT_DATA_RATE_SHIFT 13 +# define LC_CLR_FAILED_SPD_CHANGE_CNT (1 << 16) +# define LC_OTHER_SIDE_EVER_SENT_GEN2 (1 << 18) +# define LC_OTHER_SIDE_SUPPORTS_GEN2 (1 << 19) +# define LC_OTHER_SIDE_EVER_SENT_GEN3 (1 << 20) +# define LC_OTHER_SIDE_SUPPORTS_GEN3 (1 << 21) + +#define PCIE_LC_CNTL2 0xb1 +# define LC_ALLOW_PDWN_IN_L1 (1 << 17) +# define LC_ALLOW_PDWN_IN_L23 (1 << 18) + +#define PCIE_LC_CNTL3 0xb5 /* PCIE_P */ +# define LC_GO_TO_RECOVERY (1 << 30) +#define PCIE_LC_CNTL4 0xb6 /* PCIE_P */ +# define LC_REDO_EQ (1 << 5) +# define LC_SET_QUIESCE (1 << 13) + /* * UVD */ @@ -840,6 +1385,21 @@ #define UVD_RBC_RB_RPTR 0xF690 #define UVD_RBC_RB_WPTR 0xF694 +#define UVD_CGC_CTRL 0xF4B0 +# define DCM (1 << 0) +# define CG_DT(x) ((x) << 2) +# define CG_DT_MASK (0xf << 2) +# define CLK_OD(x) ((x) << 6) +# define CLK_OD_MASK (0x1f << 6) + + /* UVD CTX indirect */ +#define UVD_CGC_MEM_CTRL 0xC0 +#define UVD_CGC_CTRL2 0xC1 +# define DYN_OR_EN (1 << 0) +# define DYN_RR_EN (1 << 1) +# define G_DIV_ID(x) ((x) << 2) +# define G_DIV_ID_MASK (0x7 << 2) + /* * PM4 */ @@ -1084,6 +1644,11 @@ # define DMA_IDLE (1 << 0) #define DMA_TILING_CONFIG 0xd0b8 +#define DMA_PG 0xd0d4 +# define PG_CNTL_ENABLE (1 << 0) +#define DMA_PGFSM_CONFIG 0xd0d8 +#define DMA_PGFSM_WRITE 0xd0dc + #define DMA_PACKET(cmd, b, t, s, n) ((((cmd) & 0xF) << 28) | \ (((b) & 0x1) << 26) | \ (((t) & 0x1) << 23) | \ diff --git a/sys/dev/drm/radeon/sislands_smc.h b/sys/dev/drm/radeon/sislands_smc.h new file mode 100644 index 0000000000..16e7bf4712 --- /dev/null +++ b/sys/dev/drm/radeon/sislands_smc.h @@ -0,0 +1,396 @@ +/* + * Copyright 2013 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef PP_SISLANDS_SMC_H +#define PP_SISLANDS_SMC_H + +#include "ppsmc.h" + +#pragma pack(push, 1) + +#define SISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE 16 + +struct PP_SIslands_Dpm2PerfLevel +{ + uint8_t MaxPS; + uint8_t TgtAct; + uint8_t MaxPS_StepInc; + uint8_t MaxPS_StepDec; + uint8_t PSSamplingTime; + uint8_t NearTDPDec; + uint8_t AboveSafeInc; + uint8_t BelowSafeInc; + uint8_t PSDeltaLimit; + uint8_t PSDeltaWin; + uint16_t PwrEfficiencyRatio; + uint8_t Reserved[4]; +}; + +typedef struct PP_SIslands_Dpm2PerfLevel PP_SIslands_Dpm2PerfLevel; + +struct PP_SIslands_DPM2Status +{ + uint32_t dpm2Flags; + uint8_t CurrPSkip; + uint8_t CurrPSkipPowerShift; + uint8_t CurrPSkipTDP; + uint8_t CurrPSkipOCP; + uint8_t MaxSPLLIndex; + uint8_t MinSPLLIndex; + uint8_t CurrSPLLIndex; + uint8_t InfSweepMode; + uint8_t InfSweepDir; + uint8_t TDPexceeded; + uint8_t reserved; + uint8_t SwitchDownThreshold; + uint32_t SwitchDownCounter; + uint32_t SysScalingFactor; +}; + +typedef struct PP_SIslands_DPM2Status PP_SIslands_DPM2Status; + +struct PP_SIslands_DPM2Parameters +{ + uint32_t TDPLimit; + uint32_t NearTDPLimit; + uint32_t SafePowerLimit; + uint32_t PowerBoostLimit; + uint32_t MinLimitDelta; +}; +typedef struct PP_SIslands_DPM2Parameters PP_SIslands_DPM2Parameters; + +struct PP_SIslands_PAPMStatus +{ + uint32_t EstimatedDGPU_T; + uint32_t EstimatedDGPU_P; + uint32_t EstimatedAPU_T; + uint32_t EstimatedAPU_P; + uint8_t dGPU_T_Limit_Exceeded; + uint8_t reserved[3]; +}; +typedef struct PP_SIslands_PAPMStatus PP_SIslands_PAPMStatus; + +struct PP_SIslands_PAPMParameters +{ + uint32_t NearTDPLimitTherm; + uint32_t NearTDPLimitPAPM; + uint32_t PlatformPowerLimit; + uint32_t dGPU_T_Limit; + uint32_t dGPU_T_Warning; + uint32_t dGPU_T_Hysteresis; +}; +typedef struct PP_SIslands_PAPMParameters PP_SIslands_PAPMParameters; + +struct SISLANDS_SMC_SCLK_VALUE +{ + uint32_t vCG_SPLL_FUNC_CNTL; + uint32_t vCG_SPLL_FUNC_CNTL_2; + uint32_t vCG_SPLL_FUNC_CNTL_3; + uint32_t vCG_SPLL_FUNC_CNTL_4; + uint32_t vCG_SPLL_SPREAD_SPECTRUM; + uint32_t vCG_SPLL_SPREAD_SPECTRUM_2; + uint32_t sclk_value; +}; + +typedef struct SISLANDS_SMC_SCLK_VALUE SISLANDS_SMC_SCLK_VALUE; + +struct SISLANDS_SMC_MCLK_VALUE +{ + uint32_t vMPLL_FUNC_CNTL; + uint32_t vMPLL_FUNC_CNTL_1; + uint32_t vMPLL_FUNC_CNTL_2; + uint32_t vMPLL_AD_FUNC_CNTL; + uint32_t vMPLL_DQ_FUNC_CNTL; + uint32_t vMCLK_PWRMGT_CNTL; + uint32_t vDLL_CNTL; + uint32_t vMPLL_SS; + uint32_t vMPLL_SS2; + uint32_t mclk_value; +}; + +typedef struct SISLANDS_SMC_MCLK_VALUE SISLANDS_SMC_MCLK_VALUE; + +struct SISLANDS_SMC_VOLTAGE_VALUE +{ + uint16_t value; + uint8_t index; + uint8_t phase_settings; +}; + +typedef struct SISLANDS_SMC_VOLTAGE_VALUE SISLANDS_SMC_VOLTAGE_VALUE; + +struct SISLANDS_SMC_HW_PERFORMANCE_LEVEL +{ + uint8_t ACIndex; + uint8_t displayWatermark; + uint8_t gen2PCIE; + uint8_t UVDWatermark; + uint8_t VCEWatermark; + uint8_t strobeMode; + uint8_t mcFlags; + uint8_t padding; + uint32_t aT; + uint32_t bSP; + SISLANDS_SMC_SCLK_VALUE sclk; + SISLANDS_SMC_MCLK_VALUE mclk; + SISLANDS_SMC_VOLTAGE_VALUE vddc; + SISLANDS_SMC_VOLTAGE_VALUE mvdd; + SISLANDS_SMC_VOLTAGE_VALUE vddci; + SISLANDS_SMC_VOLTAGE_VALUE std_vddc; + uint8_t hysteresisUp; + uint8_t hysteresisDown; + uint8_t stateFlags; + uint8_t arbRefreshState; + uint32_t SQPowerThrottle; + uint32_t SQPowerThrottle_2; + uint32_t MaxPoweredUpCU; + SISLANDS_SMC_VOLTAGE_VALUE high_temp_vddc; + SISLANDS_SMC_VOLTAGE_VALUE low_temp_vddc; + uint32_t reserved[2]; + PP_SIslands_Dpm2PerfLevel dpm2; +}; + +#define SISLANDS_SMC_STROBE_RATIO 0x0F +#define SISLANDS_SMC_STROBE_ENABLE 0x10 + +#define SISLANDS_SMC_MC_EDC_RD_FLAG 0x01 +#define SISLANDS_SMC_MC_EDC_WR_FLAG 0x02 +#define SISLANDS_SMC_MC_RTT_ENABLE 0x04 +#define SISLANDS_SMC_MC_STUTTER_EN 0x08 +#define SISLANDS_SMC_MC_PG_EN 0x10 + +typedef struct SISLANDS_SMC_HW_PERFORMANCE_LEVEL SISLANDS_SMC_HW_PERFORMANCE_LEVEL; + +struct SISLANDS_SMC_SWSTATE +{ + uint8_t flags; + uint8_t levelCount; + uint8_t padding2; + uint8_t padding3; + SISLANDS_SMC_HW_PERFORMANCE_LEVEL levels[1]; +}; + +typedef struct SISLANDS_SMC_SWSTATE SISLANDS_SMC_SWSTATE; + +#define SISLANDS_SMC_VOLTAGEMASK_VDDC 0 +#define SISLANDS_SMC_VOLTAGEMASK_MVDD 1 +#define SISLANDS_SMC_VOLTAGEMASK_VDDCI 2 +#define SISLANDS_SMC_VOLTAGEMASK_MAX 4 + +struct SISLANDS_SMC_VOLTAGEMASKTABLE +{ + uint32_t lowMask[SISLANDS_SMC_VOLTAGEMASK_MAX]; +}; + +typedef struct SISLANDS_SMC_VOLTAGEMASKTABLE SISLANDS_SMC_VOLTAGEMASKTABLE; + +#define SISLANDS_MAX_NO_VREG_STEPS 32 + +struct SISLANDS_SMC_STATETABLE +{ + uint8_t thermalProtectType; + uint8_t systemFlags; + uint8_t maxVDDCIndexInPPTable; + uint8_t extraFlags; + uint32_t lowSMIO[SISLANDS_MAX_NO_VREG_STEPS]; + SISLANDS_SMC_VOLTAGEMASKTABLE voltageMaskTable; + SISLANDS_SMC_VOLTAGEMASKTABLE phaseMaskTable; + PP_SIslands_DPM2Parameters dpm2Params; + SISLANDS_SMC_SWSTATE initialState; + SISLANDS_SMC_SWSTATE ACPIState; + SISLANDS_SMC_SWSTATE ULVState; + SISLANDS_SMC_SWSTATE driverState; + SISLANDS_SMC_HW_PERFORMANCE_LEVEL dpmLevels[SISLANDS_MAX_SMC_PERFORMANCE_LEVELS_PER_SWSTATE - 1]; +}; + +typedef struct SISLANDS_SMC_STATETABLE SISLANDS_SMC_STATETABLE; + +#define SI_SMC_SOFT_REGISTER_mclk_chg_timeout 0x0 +#define SI_SMC_SOFT_REGISTER_delay_vreg 0xC +#define SI_SMC_SOFT_REGISTER_delay_acpi 0x28 +#define SI_SMC_SOFT_REGISTER_seq_index 0x5C +#define SI_SMC_SOFT_REGISTER_mvdd_chg_time 0x60 +#define SI_SMC_SOFT_REGISTER_mclk_switch_lim 0x70 +#define SI_SMC_SOFT_REGISTER_watermark_threshold 0x78 +#define SI_SMC_SOFT_REGISTER_phase_shedding_delay 0x88 +#define SI_SMC_SOFT_REGISTER_ulv_volt_change_delay 0x8C +#define SI_SMC_SOFT_REGISTER_mc_block_delay 0x98 +#define SI_SMC_SOFT_REGISTER_ticks_per_us 0xA8 +#define SI_SMC_SOFT_REGISTER_crtc_index 0xC4 +#define SI_SMC_SOFT_REGISTER_mclk_change_block_cp_min 0xC8 +#define SI_SMC_SOFT_REGISTER_mclk_change_block_cp_max 0xCC +#define SI_SMC_SOFT_REGISTER_non_ulv_pcie_link_width 0xF4 +#define SI_SMC_SOFT_REGISTER_tdr_is_about_to_happen 0xFC +#define SI_SMC_SOFT_REGISTER_vr_hot_gpio 0x100 + +#define SMC_SISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES 16 +#define SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES 32 + +#define SMC_SISLANDS_SCALE_I 7 +#define SMC_SISLANDS_SCALE_R 12 + +struct PP_SIslands_CacConfig +{ + uint16_t cac_lkge_lut[SMC_SISLANDS_LKGE_LUT_NUM_OF_TEMP_ENTRIES][SMC_SISLANDS_LKGE_LUT_NUM_OF_VOLT_ENTRIES]; + uint32_t lkge_lut_V0; + uint32_t lkge_lut_Vstep; + uint32_t WinTime; + uint32_t R_LL; + uint32_t calculation_repeats; + uint32_t l2numWin_TDP; + uint32_t dc_cac; + uint8_t lts_truncate_n; + uint8_t SHIFT_N; + uint8_t log2_PG_LKG_SCALE; + uint8_t cac_temp; + uint32_t lkge_lut_T0; + uint32_t lkge_lut_Tstep; +}; + +typedef struct PP_SIslands_CacConfig PP_SIslands_CacConfig; + +#define SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE 16 +#define SMC_SISLANDS_MC_REGISTER_ARRAY_SET_COUNT 20 + +struct SMC_SIslands_MCRegisterAddress +{ + uint16_t s0; + uint16_t s1; +}; + +typedef struct SMC_SIslands_MCRegisterAddress SMC_SIslands_MCRegisterAddress; + +struct SMC_SIslands_MCRegisterSet +{ + uint32_t value[SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE]; +}; + +typedef struct SMC_SIslands_MCRegisterSet SMC_SIslands_MCRegisterSet; + +struct SMC_SIslands_MCRegisters +{ + uint8_t last; + uint8_t reserved[3]; + SMC_SIslands_MCRegisterAddress address[SMC_SISLANDS_MC_REGISTER_ARRAY_SIZE]; + SMC_SIslands_MCRegisterSet data[SMC_SISLANDS_MC_REGISTER_ARRAY_SET_COUNT]; +}; + +typedef struct SMC_SIslands_MCRegisters SMC_SIslands_MCRegisters; + +struct SMC_SIslands_MCArbDramTimingRegisterSet +{ + uint32_t mc_arb_dram_timing; + uint32_t mc_arb_dram_timing2; + uint8_t mc_arb_rfsh_rate; + uint8_t mc_arb_burst_time; + uint8_t padding[2]; +}; + +typedef struct SMC_SIslands_MCArbDramTimingRegisterSet SMC_SIslands_MCArbDramTimingRegisterSet; + +struct SMC_SIslands_MCArbDramTimingRegisters +{ + uint8_t arb_current; + uint8_t reserved[3]; + SMC_SIslands_MCArbDramTimingRegisterSet data[16]; +}; + +typedef struct SMC_SIslands_MCArbDramTimingRegisters SMC_SIslands_MCArbDramTimingRegisters; + +struct SMC_SISLANDS_SPLL_DIV_TABLE +{ + uint32_t freq[256]; + uint32_t ss[256]; +}; + +#define SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_MASK 0x01ffffff +#define SMC_SISLANDS_SPLL_DIV_TABLE_FBDIV_SHIFT 0 +#define SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_MASK 0xfe000000 +#define SMC_SISLANDS_SPLL_DIV_TABLE_PDIV_SHIFT 25 +#define SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_MASK 0x000fffff +#define SMC_SISLANDS_SPLL_DIV_TABLE_CLKV_SHIFT 0 +#define SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_MASK 0xfff00000 +#define SMC_SISLANDS_SPLL_DIV_TABLE_CLKS_SHIFT 20 + +typedef struct SMC_SISLANDS_SPLL_DIV_TABLE SMC_SISLANDS_SPLL_DIV_TABLE; + +#define SMC_SISLANDS_DTE_MAX_FILTER_STAGES 5 + +#define SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE 16 + +struct Smc_SIslands_DTE_Configuration +{ + uint32_t tau[SMC_SISLANDS_DTE_MAX_FILTER_STAGES]; + uint32_t R[SMC_SISLANDS_DTE_MAX_FILTER_STAGES]; + uint32_t K; + uint32_t T0; + uint32_t MaxT; + uint8_t WindowSize; + uint8_t Tdep_count; + uint8_t temp_select; + uint8_t DTE_mode; + uint8_t T_limits[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + uint32_t Tdep_tau[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + uint32_t Tdep_R[SMC_SISLANDS_DTE_MAX_TEMPERATURE_DEPENDENT_ARRAY_SIZE]; + uint32_t Tthreshold; +}; + +typedef struct Smc_SIslands_DTE_Configuration Smc_SIslands_DTE_Configuration; + +#define SMC_SISLANDS_DTE_STATUS_FLAG_DTE_ON 1 + +#define SISLANDS_SMC_FIRMWARE_HEADER_LOCATION 0x10000 + +#define SISLANDS_SMC_FIRMWARE_HEADER_version 0x0 +#define SISLANDS_SMC_FIRMWARE_HEADER_flags 0x4 +#define SISLANDS_SMC_FIRMWARE_HEADER_softRegisters 0xC +#define SISLANDS_SMC_FIRMWARE_HEADER_stateTable 0x10 +#define SISLANDS_SMC_FIRMWARE_HEADER_fanTable 0x14 +#define SISLANDS_SMC_FIRMWARE_HEADER_CacConfigTable 0x18 +#define SISLANDS_SMC_FIRMWARE_HEADER_mcRegisterTable 0x24 +#define SISLANDS_SMC_FIRMWARE_HEADER_mcArbDramAutoRefreshTable 0x30 +#define SISLANDS_SMC_FIRMWARE_HEADER_spllTable 0x38 +#define SISLANDS_SMC_FIRMWARE_HEADER_DteConfiguration 0x40 +#define SISLANDS_SMC_FIRMWARE_HEADER_PAPMParameters 0x48 + +#pragma pack(pop) + +int si_set_smc_sram_address(struct radeon_device *rdev, + u32 smc_address, u32 limit); +int si_copy_bytes_to_smc(struct radeon_device *rdev, + u32 smc_start_address, + const u8 *src, u32 byte_count, u32 limit); +void si_start_smc(struct radeon_device *rdev); +void si_reset_smc(struct radeon_device *rdev); +int si_program_jump_on_start(struct radeon_device *rdev); +void si_stop_smc_clock(struct radeon_device *rdev); +void si_start_smc_clock(struct radeon_device *rdev); +bool si_is_smc_running(struct radeon_device *rdev); +PPSMC_Result si_send_msg_to_smc(struct radeon_device *rdev, PPSMC_Msg msg); +PPSMC_Result si_wait_for_smc_inactive(struct radeon_device *rdev); +int si_load_smc_ucode(struct radeon_device *rdev, u32 limit); +int si_read_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 *value, u32 limit); +int si_write_smc_sram_dword(struct radeon_device *rdev, u32 smc_address, + u32 value, u32 limit); + +#endif diff --git a/sys/dev/drm/radeon/sumo_dpm.c b/sys/dev/drm/radeon/sumo_dpm.c new file mode 100644 index 0000000000..eff5da209a --- /dev/null +++ b/sys/dev/drm/radeon/sumo_dpm.c @@ -0,0 +1,1893 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "sumod.h" +#include "r600_dpm.h" +#include "cypress_dpm.h" +#include "sumo_dpm.h" +#include + +#define SUMO_MAX_DEEPSLEEP_DIVIDER_ID 5 +#define SUMO_MINIMUM_ENGINE_CLOCK 800 +#define BOOST_DPM_LEVEL 7 + +static const u32 sumo_utc[SUMO_PM_NUMBER_OF_TC] = +{ + SUMO_UTC_DFLT_00, + SUMO_UTC_DFLT_01, + SUMO_UTC_DFLT_02, + SUMO_UTC_DFLT_03, + SUMO_UTC_DFLT_04, + SUMO_UTC_DFLT_05, + SUMO_UTC_DFLT_06, + SUMO_UTC_DFLT_07, + SUMO_UTC_DFLT_08, + SUMO_UTC_DFLT_09, + SUMO_UTC_DFLT_10, + SUMO_UTC_DFLT_11, + SUMO_UTC_DFLT_12, + SUMO_UTC_DFLT_13, + SUMO_UTC_DFLT_14, +}; + +static const u32 sumo_dtc[SUMO_PM_NUMBER_OF_TC] = +{ + SUMO_DTC_DFLT_00, + SUMO_DTC_DFLT_01, + SUMO_DTC_DFLT_02, + SUMO_DTC_DFLT_03, + SUMO_DTC_DFLT_04, + SUMO_DTC_DFLT_05, + SUMO_DTC_DFLT_06, + SUMO_DTC_DFLT_07, + SUMO_DTC_DFLT_08, + SUMO_DTC_DFLT_09, + SUMO_DTC_DFLT_10, + SUMO_DTC_DFLT_11, + SUMO_DTC_DFLT_12, + SUMO_DTC_DFLT_13, + SUMO_DTC_DFLT_14, +}; + +struct sumo_ps *sumo_get_ps(struct radeon_ps *rps); +struct sumo_power_info *sumo_get_pi(struct radeon_device *rdev); +void sumo_dpm_reset_asic(struct radeon_device *rdev); + +struct sumo_ps *sumo_get_ps(struct radeon_ps *rps) +{ + struct sumo_ps *ps = rps->ps_priv; + + return ps; +} + +struct sumo_power_info *sumo_get_pi(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void sumo_gfx_clockgating_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_ADDR_CONFIG); + } +} + +#define CGCG_CGTT_LOCAL0_MASK 0xE5BFFFFF +#define CGCG_CGTT_LOCAL1_MASK 0xEFFF07FF + +static void sumo_mg_clockgating_enable(struct radeon_device *rdev, bool enable) +{ + u32 local0; + u32 local1; + + local0 = RREG32(CG_CGTT_LOCAL_0); + local1 = RREG32(CG_CGTT_LOCAL_1); + + if (enable) { + WREG32(CG_CGTT_LOCAL_0, (0 & CGCG_CGTT_LOCAL0_MASK) | (local0 & ~CGCG_CGTT_LOCAL0_MASK) ); + WREG32(CG_CGTT_LOCAL_1, (0 & CGCG_CGTT_LOCAL1_MASK) | (local1 & ~CGCG_CGTT_LOCAL1_MASK) ); + } else { + WREG32(CG_CGTT_LOCAL_0, (0xFFFFFFFF & CGCG_CGTT_LOCAL0_MASK) | (local0 & ~CGCG_CGTT_LOCAL0_MASK) ); + WREG32(CG_CGTT_LOCAL_1, (0xFFFFCFFF & CGCG_CGTT_LOCAL1_MASK) | (local1 & ~CGCG_CGTT_LOCAL1_MASK) ); + } +} + +static void sumo_program_git(struct radeon_device *rdev) +{ + u32 p, u; + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(SUMO_GICST_DFLT, + xclk, 16, &p, &u); + + WREG32_P(CG_GIT, CG_GICST(p), ~CG_GICST_MASK); +} + +static void sumo_program_grsd(struct radeon_device *rdev) +{ + u32 p, u; + u32 xclk = radeon_get_xclk(rdev); + u32 grs = 256 * 25 / 100; + + r600_calculate_u_and_p(1, xclk, 14, &p, &u); + + WREG32(CG_GCOOR, PHC(grs) | SDC(p) | SU(u)); +} + +void sumo_gfx_clockgating_initialize(struct radeon_device *rdev) +{ + sumo_program_git(rdev); + sumo_program_grsd(rdev); +} + +static void sumo_gfx_powergating_initialize(struct radeon_device *rdev) +{ + u32 rcu_pwr_gating_cntl; + u32 p, u; + u32 p_c, p_p, d_p; + u32 r_t, i_t; + u32 xclk = radeon_get_xclk(rdev); + + if (rdev->family == CHIP_PALM) { + p_c = 4; + d_p = 10; + r_t = 10; + i_t = 4; + p_p = 50 + 1000/200 + 6 * 32; + } else { + p_c = 16; + d_p = 50; + r_t = 50; + i_t = 50; + p_p = 113; + } + + WREG32(CG_SCRATCH2, 0x01B60A17); + + r600_calculate_u_and_p(SUMO_GFXPOWERGATINGT_DFLT, + xclk, 16, &p, &u); + + WREG32_P(CG_PWR_GATING_CNTL, PGP(p) | PGU(u), + ~(PGP_MASK | PGU_MASK)); + + r600_calculate_u_and_p(SUMO_VOLTAGEDROPT_DFLT, + xclk, 16, &p, &u); + + WREG32_P(CG_CG_VOLTAGE_CNTL, PGP(p) | PGU(u), + ~(PGP_MASK | PGU_MASK)); + + if (rdev->family == CHIP_PALM) { + WREG32_RCU(RCU_PWR_GATING_SEQ0, 0x10103210); + WREG32_RCU(RCU_PWR_GATING_SEQ1, 0x10101010); + } else { + WREG32_RCU(RCU_PWR_GATING_SEQ0, 0x76543210); + WREG32_RCU(RCU_PWR_GATING_SEQ1, 0xFEDCBA98); + } + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL); + rcu_pwr_gating_cntl &= + ~(RSVD_MASK | PCV_MASK | PGS_MASK); + rcu_pwr_gating_cntl |= PCV(p_c) | PGS(1) | PWR_GATING_EN; + if (rdev->family == CHIP_PALM) { + rcu_pwr_gating_cntl &= ~PCP_MASK; + rcu_pwr_gating_cntl |= PCP(0x77); + } + WREG32_RCU(RCU_PWR_GATING_CNTL, rcu_pwr_gating_cntl); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_2); + rcu_pwr_gating_cntl &= ~(MPPU_MASK | MPPD_MASK); + rcu_pwr_gating_cntl |= MPPU(p_p) | MPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_2, rcu_pwr_gating_cntl); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_3); + rcu_pwr_gating_cntl &= ~(DPPU_MASK | DPPD_MASK); + rcu_pwr_gating_cntl |= DPPU(d_p) | DPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_3, rcu_pwr_gating_cntl); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_4); + rcu_pwr_gating_cntl &= ~(RT_MASK | IT_MASK); + rcu_pwr_gating_cntl |= RT(r_t) | IT(i_t); + WREG32_RCU(RCU_PWR_GATING_CNTL_4, rcu_pwr_gating_cntl); + + if (rdev->family == CHIP_PALM) + WREG32_RCU(RCU_PWR_GATING_CNTL_5, 0xA02); + + sumo_smu_pg_init(rdev); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL); + rcu_pwr_gating_cntl &= + ~(RSVD_MASK | PCV_MASK | PGS_MASK); + rcu_pwr_gating_cntl |= PCV(p_c) | PGS(4) | PWR_GATING_EN; + if (rdev->family == CHIP_PALM) { + rcu_pwr_gating_cntl &= ~PCP_MASK; + rcu_pwr_gating_cntl |= PCP(0x77); + } + WREG32_RCU(RCU_PWR_GATING_CNTL, rcu_pwr_gating_cntl); + + if (rdev->family == CHIP_PALM) { + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_2); + rcu_pwr_gating_cntl &= ~(MPPU_MASK | MPPD_MASK); + rcu_pwr_gating_cntl |= MPPU(113) | MPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_2, rcu_pwr_gating_cntl); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_3); + rcu_pwr_gating_cntl &= ~(DPPU_MASK | DPPD_MASK); + rcu_pwr_gating_cntl |= DPPU(16) | DPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_3, rcu_pwr_gating_cntl); + } + + sumo_smu_pg_init(rdev); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL); + rcu_pwr_gating_cntl &= + ~(RSVD_MASK | PCV_MASK | PGS_MASK); + rcu_pwr_gating_cntl |= PGS(5) | PWR_GATING_EN; + + if (rdev->family == CHIP_PALM) { + rcu_pwr_gating_cntl |= PCV(4); + rcu_pwr_gating_cntl &= ~PCP_MASK; + rcu_pwr_gating_cntl |= PCP(0x77); + } else + rcu_pwr_gating_cntl |= PCV(11); + WREG32_RCU(RCU_PWR_GATING_CNTL, rcu_pwr_gating_cntl); + + if (rdev->family == CHIP_PALM) { + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_2); + rcu_pwr_gating_cntl &= ~(MPPU_MASK | MPPD_MASK); + rcu_pwr_gating_cntl |= MPPU(113) | MPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_2, rcu_pwr_gating_cntl); + + rcu_pwr_gating_cntl = RREG32_RCU(RCU_PWR_GATING_CNTL_3); + rcu_pwr_gating_cntl &= ~(DPPU_MASK | DPPD_MASK); + rcu_pwr_gating_cntl |= DPPU(22) | DPPD(50); + WREG32_RCU(RCU_PWR_GATING_CNTL_3, rcu_pwr_gating_cntl); + } + + sumo_smu_pg_init(rdev); +} + +static void sumo_gfx_powergating_enable(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(CG_PWR_GATING_CNTL, DYN_PWR_DOWN_EN, ~DYN_PWR_DOWN_EN); + else { + WREG32_P(CG_PWR_GATING_CNTL, 0, ~DYN_PWR_DOWN_EN); + RREG32(GB_ADDR_CONFIG); + } +} + +static int sumo_enable_clock_power_gating(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (pi->enable_gfx_clock_gating) + sumo_gfx_clockgating_initialize(rdev); + if (pi->enable_gfx_power_gating) + sumo_gfx_powergating_initialize(rdev); + if (pi->enable_mg_clock_gating) + sumo_mg_clockgating_enable(rdev, true); + if (pi->enable_gfx_clock_gating) + sumo_gfx_clockgating_enable(rdev, true); + if (pi->enable_gfx_power_gating) + sumo_gfx_powergating_enable(rdev, true); + + return 0; +} + +static void sumo_disable_clock_power_gating(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (pi->enable_gfx_clock_gating) + sumo_gfx_clockgating_enable(rdev, false); + if (pi->enable_gfx_power_gating) + sumo_gfx_powergating_enable(rdev, false); + if (pi->enable_mg_clock_gating) + sumo_mg_clockgating_enable(rdev, false); +} + +static void sumo_calculate_bsp(struct radeon_device *rdev, + u32 high_clk) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 xclk = radeon_get_xclk(rdev); + + pi->pasi = 65535 * 100 / high_clk; + pi->asi = 65535 * 100 / high_clk; + + r600_calculate_u_and_p(pi->asi, + xclk, 16, &pi->bsp, &pi->bsu); + + r600_calculate_u_and_p(pi->pasi, + xclk, 16, &pi->pbsp, &pi->pbsu); + + pi->dsp = BSP(pi->bsp) | BSU(pi->bsu); + pi->psp = BSP(pi->pbsp) | BSU(pi->pbsu); +} + +static void sumo_init_bsp(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + WREG32(CG_BSP_0, pi->psp); +} + + +static void sumo_program_bsp(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *ps = sumo_get_ps(rps); + u32 i; + u32 highest_engine_clock = ps->levels[ps->num_levels - 1].sclk; + + if (ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) + highest_engine_clock = pi->boost_pl.sclk; + + sumo_calculate_bsp(rdev, highest_engine_clock); + + for (i = 0; i < ps->num_levels - 1; i++) + WREG32(CG_BSP_0 + (i * 4), pi->dsp); + + WREG32(CG_BSP_0 + (i * 4), pi->psp); + + if (ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) + WREG32(CG_BSP_0 + (BOOST_DPM_LEVEL * 4), pi->psp); +} + +static void sumo_write_at(struct radeon_device *rdev, + u32 index, u32 value) +{ + if (index == 0) + WREG32(CG_AT_0, value); + else if (index == 1) + WREG32(CG_AT_1, value); + else if (index == 2) + WREG32(CG_AT_2, value); + else if (index == 3) + WREG32(CG_AT_3, value); + else if (index == 4) + WREG32(CG_AT_4, value); + else if (index == 5) + WREG32(CG_AT_5, value); + else if (index == 6) + WREG32(CG_AT_6, value); + else if (index == 7) + WREG32(CG_AT_7, value); +} + +static void sumo_program_at(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *ps = sumo_get_ps(rps); + u32 asi; + u32 i; + u32 m_a; + u32 a_t; + u32 r[SUMO_MAX_HARDWARE_POWERLEVELS]; + u32 l[SUMO_MAX_HARDWARE_POWERLEVELS]; + + r[0] = SUMO_R_DFLT0; + r[1] = SUMO_R_DFLT1; + r[2] = SUMO_R_DFLT2; + r[3] = SUMO_R_DFLT3; + r[4] = SUMO_R_DFLT4; + + l[0] = SUMO_L_DFLT0; + l[1] = SUMO_L_DFLT1; + l[2] = SUMO_L_DFLT2; + l[3] = SUMO_L_DFLT3; + l[4] = SUMO_L_DFLT4; + + for (i = 0; i < ps->num_levels; i++) { + asi = (i == ps->num_levels - 1) ? pi->pasi : pi->asi; + + m_a = asi * ps->levels[i].sclk / 100; + + a_t = CG_R(m_a * r[i] / 100) | CG_L(m_a * l[i] / 100); + + sumo_write_at(rdev, i, a_t); + } + + if (ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) { + asi = pi->pasi; + + m_a = asi * pi->boost_pl.sclk / 100; + + a_t = CG_R(m_a * r[ps->num_levels - 1] / 100) | + CG_L(m_a * l[ps->num_levels - 1] / 100); + + sumo_write_at(rdev, BOOST_DPM_LEVEL, a_t); + } +} + +static void sumo_program_tp(struct radeon_device *rdev) +{ + int i; + enum r600_td td = R600_TD_DFLT; + + for (i = 0; i < SUMO_PM_NUMBER_OF_TC; i++) { + WREG32_P(CG_FFCT_0 + (i * 4), UTC_0(sumo_utc[i]), ~UTC_0_MASK); + WREG32_P(CG_FFCT_0 + (i * 4), DTC_0(sumo_dtc[i]), ~DTC_0_MASK); + } + + if (td == R600_TD_AUTO) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_FORCE_TREND_SEL); + else + WREG32_P(SCLK_PWRMGT_CNTL, FIR_FORCE_TREND_SEL, ~FIR_FORCE_TREND_SEL); + + if (td == R600_TD_UP) + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_TREND_MODE); + + if (td == R600_TD_DOWN) + WREG32_P(SCLK_PWRMGT_CNTL, FIR_TREND_MODE, ~FIR_TREND_MODE); +} + +void sumo_program_vc(struct radeon_device *rdev, u32 vrc) +{ + WREG32(CG_FTV, vrc); +} + +void sumo_clear_vc(struct radeon_device *rdev) +{ + WREG32(CG_FTV, 0); +} + +void sumo_program_sstp(struct radeon_device *rdev) +{ + u32 p, u; + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(SUMO_SST_DFLT, + xclk, 16, &p, &u); + + WREG32(CG_SSP, SSTU(u) | SST(p)); +} + +static void sumo_set_divider_value(struct radeon_device *rdev, + u32 index, u32 divider) +{ + u32 reg_index = index / 4; + u32 field_index = index % 4; + + if (field_index == 0) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + SCLK_FSTATE_0_DIV(divider), ~SCLK_FSTATE_0_DIV_MASK); + else if (field_index == 1) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + SCLK_FSTATE_1_DIV(divider), ~SCLK_FSTATE_1_DIV_MASK); + else if (field_index == 2) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + SCLK_FSTATE_2_DIV(divider), ~SCLK_FSTATE_2_DIV_MASK); + else if (field_index == 3) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + SCLK_FSTATE_3_DIV(divider), ~SCLK_FSTATE_3_DIV_MASK); +} + +static void sumo_set_ds_dividers(struct radeon_device *rdev, + u32 index, u32 divider) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (pi->enable_sclk_ds) { + u32 dpm_ctrl = RREG32(CG_SCLK_DPM_CTRL_6); + + dpm_ctrl &= ~(0x7 << (index * 3)); + dpm_ctrl |= (divider << (index * 3)); + WREG32(CG_SCLK_DPM_CTRL_6, dpm_ctrl); + } +} + +static void sumo_set_ss_dividers(struct radeon_device *rdev, + u32 index, u32 divider) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (pi->enable_sclk_ds) { + u32 dpm_ctrl = RREG32(CG_SCLK_DPM_CTRL_11); + + dpm_ctrl &= ~(0x7 << (index * 3)); + dpm_ctrl |= (divider << (index * 3)); + WREG32(CG_SCLK_DPM_CTRL_11, dpm_ctrl); + } +} + +static void sumo_set_vid(struct radeon_device *rdev, u32 index, u32 vid) +{ + u32 voltage_cntl = RREG32(CG_DPM_VOLTAGE_CNTL); + + voltage_cntl &= ~(DPM_STATE0_LEVEL_MASK << (index * 2)); + voltage_cntl |= (vid << (DPM_STATE0_LEVEL_SHIFT + index * 2)); + WREG32(CG_DPM_VOLTAGE_CNTL, voltage_cntl); +} + +static void sumo_set_allos_gnb_slow(struct radeon_device *rdev, u32 index, u32 gnb_slow) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 temp = gnb_slow; + u32 cg_sclk_dpm_ctrl_3; + + if (pi->driver_nbps_policy_disable) + temp = 1; + + cg_sclk_dpm_ctrl_3 = RREG32(CG_SCLK_DPM_CTRL_3); + cg_sclk_dpm_ctrl_3 &= ~(GNB_SLOW_FSTATE_0_MASK << index); + cg_sclk_dpm_ctrl_3 |= (temp << (GNB_SLOW_FSTATE_0_SHIFT + index)); + + WREG32(CG_SCLK_DPM_CTRL_3, cg_sclk_dpm_ctrl_3); +} + +static void sumo_program_power_level(struct radeon_device *rdev, + struct sumo_pl *pl, u32 index) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + int ret; + struct atom_clock_dividers dividers; + u32 ds_en = RREG32(DEEP_SLEEP_CNTL) & ENABLE_DS; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + pl->sclk, false, ÷rs); + if (ret) + return; + + sumo_set_divider_value(rdev, index, dividers.post_div); + + sumo_set_vid(rdev, index, pl->vddc_index); + + if (pl->ss_divider_index == 0 || pl->ds_divider_index == 0) { + if (ds_en) + WREG32_P(DEEP_SLEEP_CNTL, 0, ~ENABLE_DS); + } else { + sumo_set_ss_dividers(rdev, index, pl->ss_divider_index); + sumo_set_ds_dividers(rdev, index, pl->ds_divider_index); + + if (!ds_en) + WREG32_P(DEEP_SLEEP_CNTL, ENABLE_DS, ~ENABLE_DS); + } + + sumo_set_allos_gnb_slow(rdev, index, pl->allow_gnb_slow); + + if (pi->enable_boost) + sumo_set_tdp_limit(rdev, index, pl->sclk_dpm_tdp_limit); +} + +static void sumo_power_level_enable(struct radeon_device *rdev, u32 index, bool enable) +{ + u32 reg_index = index / 4; + u32 field_index = index % 4; + + if (field_index == 0) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + enable ? SCLK_FSTATE_0_VLD : 0, ~SCLK_FSTATE_0_VLD); + else if (field_index == 1) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + enable ? SCLK_FSTATE_1_VLD : 0, ~SCLK_FSTATE_1_VLD); + else if (field_index == 2) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + enable ? SCLK_FSTATE_2_VLD : 0, ~SCLK_FSTATE_2_VLD); + else if (field_index == 3) + WREG32_P(CG_SCLK_DPM_CTRL + (reg_index * 4), + enable ? SCLK_FSTATE_3_VLD : 0, ~SCLK_FSTATE_3_VLD); +} + +static bool sumo_dpm_enabled(struct radeon_device *rdev) +{ + if (RREG32(CG_SCLK_DPM_CTRL_3) & DPM_SCLK_ENABLE) + return true; + else + return false; +} + +static void sumo_start_dpm(struct radeon_device *rdev) +{ + WREG32_P(CG_SCLK_DPM_CTRL_3, DPM_SCLK_ENABLE, ~DPM_SCLK_ENABLE); +} + +static void sumo_stop_dpm(struct radeon_device *rdev) +{ + WREG32_P(CG_SCLK_DPM_CTRL_3, 0, ~DPM_SCLK_ENABLE); +} + +static void sumo_set_forced_mode(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_P(CG_SCLK_DPM_CTRL_3, FORCE_SCLK_STATE_EN, ~FORCE_SCLK_STATE_EN); + else + WREG32_P(CG_SCLK_DPM_CTRL_3, 0, ~FORCE_SCLK_STATE_EN); +} + +static void sumo_set_forced_mode_enabled(struct radeon_device *rdev) +{ + int i; + + sumo_set_forced_mode(rdev, true); + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(CG_SCLK_STATUS) & SCLK_OVERCLK_DETECT) + break; + DRM_UDELAY(1); + } +} + +static void sumo_wait_for_level_0(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURR_SCLK_INDEX_MASK) == 0) + break; + DRM_UDELAY(1); + } + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURR_INDEX_MASK) == 0) + break; + DRM_UDELAY(1); + } +} + +static void sumo_set_forced_mode_disabled(struct radeon_device *rdev) +{ + sumo_set_forced_mode(rdev, false); +} + +static void sumo_enable_power_level_0(struct radeon_device *rdev) +{ + sumo_power_level_enable(rdev, 0, true); +} + +static void sumo_patch_boost_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *new_ps = sumo_get_ps(rps); + + if (new_ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) { + pi->boost_pl = new_ps->levels[new_ps->num_levels - 1]; + pi->boost_pl.sclk = pi->sys_info.boost_sclk; + pi->boost_pl.vddc_index = pi->sys_info.boost_vid_2bit; + pi->boost_pl.sclk_dpm_tdp_limit = pi->sys_info.sclk_dpm_tdp_limit_boost; + } +} + +static void sumo_pre_notify_alt_vddnb_change(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(new_rps); + struct sumo_ps *old_ps = sumo_get_ps(old_rps); + u32 nbps1_old = 0; + u32 nbps1_new = 0; + + if (old_ps != NULL) + nbps1_old = (old_ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE) ? 1 : 0; + + nbps1_new = (new_ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE) ? 1 : 0; + + if (nbps1_old == 1 && nbps1_new == 0) + sumo_smu_notify_alt_vddnb_change(rdev, 0, 0); +} + +static void sumo_post_notify_alt_vddnb_change(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(new_rps); + struct sumo_ps *old_ps = sumo_get_ps(old_rps); + u32 nbps1_old = 0; + u32 nbps1_new = 0; + + if (old_ps != NULL) + nbps1_old = (old_ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE)? 1 : 0; + + nbps1_new = (new_ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE)? 1 : 0; + + if (nbps1_old == 0 && nbps1_new == 1) + sumo_smu_notify_alt_vddnb_change(rdev, 1, 1); +} + +static void sumo_enable_boost(struct radeon_device *rdev, + struct radeon_ps *rps, + bool enable) +{ + struct sumo_ps *new_ps = sumo_get_ps(rps); + + if (enable) { + if (new_ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) + sumo_boost_state_enable(rdev, true); + } else + sumo_boost_state_enable(rdev, false); +} + +static void sumo_set_forced_level(struct radeon_device *rdev, u32 index) +{ + WREG32_P(CG_SCLK_DPM_CTRL_3, FORCE_SCLK_STATE(index), ~FORCE_SCLK_STATE_MASK); +} + +static void sumo_set_forced_level_0(struct radeon_device *rdev) +{ + sumo_set_forced_level(rdev, 0); +} + +static void sumo_program_wl(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(rps); + u32 dpm_ctrl4 = RREG32(CG_SCLK_DPM_CTRL_4); + + dpm_ctrl4 &= 0xFFFFFF00; + dpm_ctrl4 |= (1 << (new_ps->num_levels - 1)); + + if (new_ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) + dpm_ctrl4 |= (1 << BOOST_DPM_LEVEL); + + WREG32(CG_SCLK_DPM_CTRL_4, dpm_ctrl4); +} + +static void sumo_program_power_levels_0_to_n(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *new_ps = sumo_get_ps(new_rps); + struct sumo_ps *old_ps = sumo_get_ps(old_rps); + u32 i; + u32 n_current_state_levels = (old_ps == NULL) ? 1 : old_ps->num_levels; + + for (i = 0; i < new_ps->num_levels; i++) { + sumo_program_power_level(rdev, &new_ps->levels[i], i); + sumo_power_level_enable(rdev, i, true); + } + + for (i = new_ps->num_levels; i < n_current_state_levels; i++) + sumo_power_level_enable(rdev, i, false); + + if (new_ps->flags & SUMO_POWERSTATE_FLAGS_BOOST_STATE) + sumo_program_power_level(rdev, &pi->boost_pl, BOOST_DPM_LEVEL); +} + +static void sumo_enable_acpi_pm(struct radeon_device *rdev) +{ + WREG32_P(GENERAL_PWRMGT, STATIC_PM_EN, ~STATIC_PM_EN); +} + +static void sumo_program_power_level_enter_state(struct radeon_device *rdev) +{ + WREG32_P(CG_SCLK_DPM_CTRL_5, SCLK_FSTATE_BOOTUP(0), ~SCLK_FSTATE_BOOTUP_MASK); +} + +static void sumo_program_acpi_power_level(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct atom_clock_dividers dividers; + int ret; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + pi->acpi_pl.sclk, + false, ÷rs); + if (ret) + return; + + WREG32_P(CG_ACPI_CNTL, SCLK_ACPI_DIV(dividers.post_div), ~SCLK_ACPI_DIV_MASK); + WREG32_P(CG_ACPI_VOLTAGE_CNTL, 0, ~ACPI_VOLTAGE_EN); +} + +static void sumo_program_bootup_state(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 dpm_ctrl4 = RREG32(CG_SCLK_DPM_CTRL_4); + u32 i; + + sumo_program_power_level(rdev, &pi->boot_pl, 0); + + dpm_ctrl4 &= 0xFFFFFF00; + WREG32(CG_SCLK_DPM_CTRL_4, dpm_ctrl4); + + for (i = 1; i < 8; i++) + sumo_power_level_enable(rdev, i, false); +} + +static void sumo_setup_uvd_clocks(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (pi->enable_gfx_power_gating) { + sumo_gfx_powergating_enable(rdev, false); + } + + radeon_set_uvd_clocks(rdev, new_rps->vclk, new_rps->dclk); + + if (pi->enable_gfx_power_gating) { + if (!pi->disable_gfx_power_gating_in_uvd || + !r600_is_uvd_state(new_rps->class, new_rps->class2)) + sumo_gfx_powergating_enable(rdev, true); + } +} + +static void sumo_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(new_rps); + struct sumo_ps *current_ps = sumo_get_ps(old_rps); + + if ((new_rps->vclk == old_rps->vclk) && + (new_rps->dclk == old_rps->dclk)) + return; + + if (new_ps->levels[new_ps->num_levels - 1].sclk >= + current_ps->levels[current_ps->num_levels - 1].sclk) + return; + + sumo_setup_uvd_clocks(rdev, new_rps, old_rps); +} + +static void sumo_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(new_rps); + struct sumo_ps *current_ps = sumo_get_ps(old_rps); + + if ((new_rps->vclk == old_rps->vclk) && + (new_rps->dclk == old_rps->dclk)) + return; + + if (new_ps->levels[new_ps->num_levels - 1].sclk < + current_ps->levels[current_ps->num_levels - 1].sclk) + return; + + sumo_setup_uvd_clocks(rdev, new_rps, old_rps); +} + +void sumo_take_smu_control(struct radeon_device *rdev, bool enable) +{ +/* This bit selects who handles display phy powergating. + * Clear the bit to let atom handle it. + * Set it to let the driver handle it. + * For now we just let atom handle it. + */ +#if 0 + u32 v = RREG32(DOUT_SCRATCH3); + + if (enable) + v |= 0x4; + else + v &= 0xFFFFFFFB; + + WREG32(DOUT_SCRATCH3, v); +#endif +} + +static void sumo_enable_sclk_ds(struct radeon_device *rdev, bool enable) +{ + if (enable) { + u32 deep_sleep_cntl = RREG32(DEEP_SLEEP_CNTL); + u32 deep_sleep_cntl2 = RREG32(DEEP_SLEEP_CNTL2); + u32 t = 1; + + deep_sleep_cntl &= ~R_DIS; + deep_sleep_cntl &= ~HS_MASK; + deep_sleep_cntl |= HS(t > 4095 ? 4095 : t); + + deep_sleep_cntl2 |= LB_UFP_EN; + deep_sleep_cntl2 &= INOUT_C_MASK; + deep_sleep_cntl2 |= INOUT_C(0xf); + + WREG32(DEEP_SLEEP_CNTL2, deep_sleep_cntl2); + WREG32(DEEP_SLEEP_CNTL, deep_sleep_cntl); + } else + WREG32_P(DEEP_SLEEP_CNTL, 0, ~ENABLE_DS); +} + +static void sumo_program_bootup_at(struct radeon_device *rdev) +{ + WREG32_P(CG_AT_0, CG_R(0xffff), ~CG_R_MASK); + WREG32_P(CG_AT_0, CG_L(0), ~CG_L_MASK); +} + +static void sumo_reset_am(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, FIR_RESET, ~FIR_RESET); +} + +static void sumo_start_am(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~FIR_RESET); +} + +static void sumo_program_ttp(struct radeon_device *rdev) +{ + u32 xclk = radeon_get_xclk(rdev); + u32 p, u; + u32 cg_sclk_dpm_ctrl_5 = RREG32(CG_SCLK_DPM_CTRL_5); + + r600_calculate_u_and_p(1000, + xclk, 16, &p, &u); + + cg_sclk_dpm_ctrl_5 &= ~(TT_TP_MASK | TT_TU_MASK); + cg_sclk_dpm_ctrl_5 |= TT_TP(p) | TT_TU(u); + + WREG32(CG_SCLK_DPM_CTRL_5, cg_sclk_dpm_ctrl_5); +} + +static void sumo_program_ttt(struct radeon_device *rdev) +{ + u32 cg_sclk_dpm_ctrl_3 = RREG32(CG_SCLK_DPM_CTRL_3); + struct sumo_power_info *pi = sumo_get_pi(rdev); + + cg_sclk_dpm_ctrl_3 &= ~(GNB_TT_MASK | GNB_THERMTHRO_MASK); + cg_sclk_dpm_ctrl_3 |= GNB_TT(pi->thermal_auto_throttling + 49); + + WREG32(CG_SCLK_DPM_CTRL_3, cg_sclk_dpm_ctrl_3); +} + + +static void sumo_enable_voltage_scaling(struct radeon_device *rdev, bool enable) +{ + if (enable) { + WREG32_P(CG_DPM_VOLTAGE_CNTL, DPM_VOLTAGE_EN, ~DPM_VOLTAGE_EN); + WREG32_P(CG_CG_VOLTAGE_CNTL, 0, ~CG_VOLTAGE_EN); + } else { + WREG32_P(CG_CG_VOLTAGE_CNTL, CG_VOLTAGE_EN, ~CG_VOLTAGE_EN); + WREG32_P(CG_DPM_VOLTAGE_CNTL, 0, ~DPM_VOLTAGE_EN); + } +} + +static void sumo_override_cnb_thermal_events(struct radeon_device *rdev) +{ + WREG32_P(CG_SCLK_DPM_CTRL_3, CNB_THERMTHRO_MASK_SCLK, + ~CNB_THERMTHRO_MASK_SCLK); +} + +static void sumo_program_dc_hto(struct radeon_device *rdev) +{ + u32 cg_sclk_dpm_ctrl_4 = RREG32(CG_SCLK_DPM_CTRL_4); + u32 p, u; + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(100000, + xclk, 14, &p, &u); + + cg_sclk_dpm_ctrl_4 &= ~(DC_HDC_MASK | DC_HU_MASK); + cg_sclk_dpm_ctrl_4 |= DC_HDC(p) | DC_HU(u); + + WREG32(CG_SCLK_DPM_CTRL_4, cg_sclk_dpm_ctrl_4); +} + +static void sumo_force_nbp_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *new_ps = sumo_get_ps(rps); + + if (!pi->driver_nbps_policy_disable) { + if (new_ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE) + WREG32_P(CG_SCLK_DPM_CTRL_3, FORCE_NB_PSTATE_1, ~FORCE_NB_PSTATE_1); + else + WREG32_P(CG_SCLK_DPM_CTRL_3, 0, ~FORCE_NB_PSTATE_1); + } +} + +u32 sumo_get_sleep_divider_from_id(u32 id) +{ + return 1 << id; +} + +u32 sumo_get_sleep_divider_id_from_clock(struct radeon_device *rdev, + u32 sclk, + u32 min_sclk_in_sr) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 i; + u32 temp; + u32 min = (min_sclk_in_sr > SUMO_MINIMUM_ENGINE_CLOCK) ? + min_sclk_in_sr : SUMO_MINIMUM_ENGINE_CLOCK; + + if (sclk < min) + return 0; + + if (!pi->enable_sclk_ds) + return 0; + + for (i = SUMO_MAX_DEEPSLEEP_DIVIDER_ID; ; i--) { + temp = sclk / sumo_get_sleep_divider_from_id(i); + + if (temp >= min || i == 0) + break; + } + return i; +} + +static u32 sumo_get_valid_engine_clock(struct radeon_device *rdev, + u32 lower_limit) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 i; + + for (i = 0; i < pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries; i++) { + if (pi->sys_info.sclk_voltage_mapping_table.entries[i].sclk_frequency >= lower_limit) + return pi->sys_info.sclk_voltage_mapping_table.entries[i].sclk_frequency; + } + + return pi->sys_info.sclk_voltage_mapping_table.entries[pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries - 1].sclk_frequency; +} + +static void sumo_patch_thermal_state(struct radeon_device *rdev, + struct sumo_ps *ps, + struct sumo_ps *current_ps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 sclk_in_sr = pi->sys_info.min_sclk; /* ??? */ + u32 current_vddc; + u32 current_sclk; + u32 current_index = 0; + + if (current_ps) { + current_vddc = current_ps->levels[current_index].vddc_index; + current_sclk = current_ps->levels[current_index].sclk; + } else { + current_vddc = pi->boot_pl.vddc_index; + current_sclk = pi->boot_pl.sclk; + } + + ps->levels[0].vddc_index = current_vddc; + + if (ps->levels[0].sclk > current_sclk) + ps->levels[0].sclk = current_sclk; + + ps->levels[0].ss_divider_index = + sumo_get_sleep_divider_id_from_clock(rdev, ps->levels[0].sclk, sclk_in_sr); + + ps->levels[0].ds_divider_index = + sumo_get_sleep_divider_id_from_clock(rdev, ps->levels[0].sclk, SUMO_MINIMUM_ENGINE_CLOCK); + + if (ps->levels[0].ds_divider_index > ps->levels[0].ss_divider_index + 1) + ps->levels[0].ds_divider_index = ps->levels[0].ss_divider_index + 1; + + if (ps->levels[0].ss_divider_index == ps->levels[0].ds_divider_index) { + if (ps->levels[0].ss_divider_index > 1) + ps->levels[0].ss_divider_index = ps->levels[0].ss_divider_index - 1; + } + + if (ps->levels[0].ss_divider_index == 0) + ps->levels[0].ds_divider_index = 0; + + if (ps->levels[0].ds_divider_index == 0) + ps->levels[0].ss_divider_index = 0; +} + +static void sumo_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct sumo_ps *ps = sumo_get_ps(new_rps); + struct sumo_ps *current_ps = sumo_get_ps(old_rps); + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 min_voltage = 0; /* ??? */ + u32 min_sclk = pi->sys_info.min_sclk; /* XXX check against disp reqs */ + u32 sclk_in_sr = pi->sys_info.min_sclk; /* ??? */ + u32 i; + + if (new_rps->class & ATOM_PPLIB_CLASSIFICATION_THERMAL) + return sumo_patch_thermal_state(rdev, ps, current_ps); + + if (pi->enable_boost) { + if (new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_PERFORMANCE) + ps->flags |= SUMO_POWERSTATE_FLAGS_BOOST_STATE; + } + + if ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_BATTERY) || + (new_rps->class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) || + (new_rps->class & ATOM_PPLIB_CLASSIFICATION_HDSTATE)) + ps->flags |= SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE; + + for (i = 0; i < ps->num_levels; i++) { + if (ps->levels[i].vddc_index < min_voltage) + ps->levels[i].vddc_index = min_voltage; + + if (ps->levels[i].sclk < min_sclk) + ps->levels[i].sclk = + sumo_get_valid_engine_clock(rdev, min_sclk); + + ps->levels[i].ss_divider_index = + sumo_get_sleep_divider_id_from_clock(rdev, ps->levels[i].sclk, sclk_in_sr); + + ps->levels[i].ds_divider_index = + sumo_get_sleep_divider_id_from_clock(rdev, ps->levels[i].sclk, SUMO_MINIMUM_ENGINE_CLOCK); + + if (ps->levels[i].ds_divider_index > ps->levels[i].ss_divider_index + 1) + ps->levels[i].ds_divider_index = ps->levels[i].ss_divider_index + 1; + + if (ps->levels[i].ss_divider_index == ps->levels[i].ds_divider_index) { + if (ps->levels[i].ss_divider_index > 1) + ps->levels[i].ss_divider_index = ps->levels[i].ss_divider_index - 1; + } + + if (ps->levels[i].ss_divider_index == 0) + ps->levels[i].ds_divider_index = 0; + + if (ps->levels[i].ds_divider_index == 0) + ps->levels[i].ss_divider_index = 0; + + if (ps->flags & SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE) + ps->levels[i].allow_gnb_slow = 1; + else if ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) || + (new_rps->class2 & ATOM_PPLIB_CLASSIFICATION2_MVC)) + ps->levels[i].allow_gnb_slow = 0; + else if (i == ps->num_levels - 1) + ps->levels[i].allow_gnb_slow = 0; + else + ps->levels[i].allow_gnb_slow = 1; + } +} + +static void sumo_cleanup_asic(struct radeon_device *rdev) +{ + sumo_take_smu_control(rdev, false); +} + +static int sumo_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) +{ + int low_temp = 0 * 1000; + int high_temp = 255 * 1000; + + if (low_temp < min_temp) + low_temp = min_temp; + if (high_temp > max_temp) + high_temp = max_temp; + if (high_temp < low_temp) { + DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); + return -EINVAL; + } + + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTH(49 + (high_temp / 1000)), ~DIG_THERM_INTH_MASK); + WREG32_P(CG_THERMAL_INT, DIG_THERM_INTL(49 + (low_temp / 1000)), ~DIG_THERM_INTL_MASK); + + rdev->pm.dpm.thermal.min_temp = low_temp; + rdev->pm.dpm.thermal.max_temp = high_temp; + + return 0; +} + +static void sumo_update_current_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(rps); + struct sumo_power_info *pi = sumo_get_pi(rdev); + + pi->current_rps = *rps; + pi->current_ps = *new_ps; + pi->current_rps.ps_priv = &pi->current_ps; +} + +static void sumo_update_requested_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct sumo_ps *new_ps = sumo_get_ps(rps); + struct sumo_power_info *pi = sumo_get_pi(rdev); + + pi->requested_rps = *rps; + pi->requested_ps = *new_ps; + pi->requested_rps.ps_priv = &pi->requested_ps; +} + +int sumo_dpm_enable(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + int ret; + + if (sumo_dpm_enabled(rdev)) + return -EINVAL; + + ret = sumo_enable_clock_power_gating(rdev); + if (ret) + return ret; + sumo_program_bootup_state(rdev); + sumo_init_bsp(rdev); + sumo_reset_am(rdev); + sumo_program_tp(rdev); + sumo_program_bootup_at(rdev); + sumo_start_am(rdev); + if (pi->enable_auto_thermal_throttling) { + sumo_program_ttp(rdev); + sumo_program_ttt(rdev); + } + sumo_program_dc_hto(rdev); + sumo_program_power_level_enter_state(rdev); + sumo_enable_voltage_scaling(rdev, true); + sumo_program_sstp(rdev); + sumo_program_vc(rdev, SUMO_VRC_DFLT); + sumo_override_cnb_thermal_events(rdev); + sumo_start_dpm(rdev); + sumo_wait_for_level_0(rdev); + if (pi->enable_sclk_ds) + sumo_enable_sclk_ds(rdev, true); + if (pi->enable_boost) + sumo_enable_boost_timer(rdev); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + ret = sumo_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) + return ret; + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + } + + sumo_update_current_ps(rdev, rdev->pm.dpm.boot_ps); + + return 0; +} + +void sumo_dpm_disable(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + if (!sumo_dpm_enabled(rdev)) + return; + sumo_disable_clock_power_gating(rdev); + if (pi->enable_sclk_ds) + sumo_enable_sclk_ds(rdev, false); + sumo_clear_vc(rdev); + sumo_wait_for_level_0(rdev); + sumo_stop_dpm(rdev); + sumo_enable_voltage_scaling(rdev, false); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + sumo_update_current_ps(rdev, rdev->pm.dpm.boot_ps); +} + +int sumo_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps; + struct radeon_ps *new_ps = &requested_ps; + + sumo_update_requested_ps(rdev, new_ps); + + if (pi->enable_dynamic_patch_ps) + sumo_apply_state_adjust_rules(rdev, + &pi->requested_rps, + &pi->current_rps); + + return 0; +} + +int sumo_dpm_set_power_state(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_ps *new_ps = &pi->requested_rps; + struct radeon_ps *old_ps = &pi->current_rps; + + if (pi->enable_dpm) + sumo_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + if (pi->enable_boost) { + sumo_enable_boost(rdev, new_ps, false); + sumo_patch_boost_state(rdev, new_ps); + } + if (pi->enable_dpm) { + sumo_pre_notify_alt_vddnb_change(rdev, new_ps, old_ps); + sumo_enable_power_level_0(rdev); + sumo_set_forced_level_0(rdev); + sumo_set_forced_mode_enabled(rdev); + sumo_wait_for_level_0(rdev); + sumo_program_power_levels_0_to_n(rdev, new_ps, old_ps); + sumo_program_wl(rdev, new_ps); + sumo_program_bsp(rdev, new_ps); + sumo_program_at(rdev, new_ps); + sumo_force_nbp_state(rdev, new_ps); + sumo_set_forced_mode_disabled(rdev); + sumo_set_forced_mode_enabled(rdev); + sumo_set_forced_mode_disabled(rdev); + sumo_post_notify_alt_vddnb_change(rdev, new_ps, old_ps); + } + if (pi->enable_boost) + sumo_enable_boost(rdev, new_ps, true); + if (pi->enable_dpm) + sumo_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + + rdev->pm.dpm.forced_level = RADEON_DPM_FORCED_LEVEL_AUTO; + + return 0; +} + +void sumo_dpm_post_set_power_state(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_ps *new_ps = &pi->requested_rps; + + sumo_update_current_ps(rdev, new_ps); +} + +void sumo_dpm_reset_asic(struct radeon_device *rdev) +{ + sumo_program_bootup_state(rdev); + sumo_enable_power_level_0(rdev); + sumo_set_forced_level_0(rdev); + sumo_set_forced_mode_enabled(rdev); + sumo_wait_for_level_0(rdev); + sumo_set_forced_mode_disabled(rdev); + sumo_set_forced_mode_enabled(rdev); + sumo_set_forced_mode_disabled(rdev); +} + +void sumo_dpm_setup_asic(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + sumo_initialize_m3_arb(rdev); + pi->fw_version = sumo_get_running_fw_version(rdev); + DRM_INFO("Found smc ucode version: 0x%08x\n", pi->fw_version); + sumo_program_acpi_power_level(rdev); + sumo_enable_acpi_pm(rdev); + sumo_take_smu_control(rdev, true); +} + +void sumo_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void sumo_patch_boot_state(struct radeon_device *rdev, + struct sumo_ps *ps) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + ps->num_levels = 1; + ps->flags = 0; + ps->levels[0] = pi->boot_pl; +} + +static void sumo_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + struct sumo_ps *ps = sumo_get_ps(rps); + + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + rdev->pm.dpm.boot_ps = rps; + sumo_patch_boot_state(rdev, ps); + } + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void sumo_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *ps = sumo_get_ps(rps); + struct sumo_pl *pl = &ps->levels[index]; + u32 sclk; + + sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow); + sclk |= clock_info->sumo.ucEngineClockHigh << 16; + pl->sclk = sclk; + pl->vddc_index = clock_info->sumo.vddcIndex; + pl->sclk_dpm_tdp_limit = clock_info->sumo.tdpLimit; + + ps->num_levels = index + 1; + + if (pi->enable_sclk_ds) { + pl->ds_divider_index = 5; + pl->ss_divider_index = 4; + } +} + +static int sumo_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j, k, non_clock_array_index, clock_array_index; + union pplib_clock_info *clock_info; + struct _StateArray *state_array; + struct _ClockInfoArray *clock_info_array; + struct _NonClockInfoArray *non_clock_info_array; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + u8 *power_state_offset; + struct sumo_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + state_array = (struct _StateArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset)); + clock_info_array = (struct _ClockInfoArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset)); + non_clock_info_array = (struct _NonClockInfoArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset)); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + state_array->ucNumEntries, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + power_state_offset = (u8 *)state_array->states; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + for (i = 0; i < state_array->ucNumEntries; i++) { + power_state = (union pplib_power_state *)power_state_offset; + non_clock_array_index = power_state->v2.nonClockInfoIndex; + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + &non_clock_info_array->nonClockInfo[non_clock_array_index]; + if (!rdev->pm.power_state[i].clock_info) + return -EINVAL; + ps = kzalloc(sizeof(struct sumo_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + k = 0; + for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) { + clock_array_index = power_state->v2.clockInfoIndex[j]; + if (k >= SUMO_MAX_HARDWARE_POWERLEVELS) + break; + clock_info = (union pplib_clock_info *) + &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; + sumo_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], k, + clock_info); + k++; + } + sumo_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + non_clock_info_array->ucEntrySize); + power_state_offset += 2 + power_state->v2.ucNumDPMLevels; + } + rdev->pm.dpm.num_ps = state_array->ucNumEntries; + return 0; +} + +u32 sumo_convert_vid2_to_vid7(struct radeon_device *rdev, + struct sumo_vid_mapping_table *vid_mapping_table, + u32 vid_2bit) +{ + u32 i; + + for (i = 0; i < vid_mapping_table->num_entries; i++) { + if (vid_mapping_table->entries[i].vid_2bit == vid_2bit) + return vid_mapping_table->entries[i].vid_7bit; + } + + return vid_mapping_table->entries[vid_mapping_table->num_entries - 1].vid_7bit; +} + +static u16 sumo_convert_voltage_index_to_value(struct radeon_device *rdev, + u32 vid_2bit) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 vid_7bit = sumo_convert_vid2_to_vid7(rdev, &pi->sys_info.vid_mapping_table, vid_2bit); + + if (vid_7bit > 0x7C) + return 0; + + return (15500 - vid_7bit * 125 + 5) / 10; +} + +static void sumo_construct_display_voltage_mapping_table(struct radeon_device *rdev, + struct sumo_disp_clock_voltage_mapping_table *disp_clk_voltage_mapping_table, + ATOM_CLK_VOLT_CAPABILITY *table) +{ + u32 i; + + for (i = 0; i < SUMO_MAX_NUMBER_VOLTAGES; i++) { + if (table[i].ulMaximumSupportedCLK == 0) + break; + + disp_clk_voltage_mapping_table->display_clock_frequency[i] = + table[i].ulMaximumSupportedCLK; + } + + disp_clk_voltage_mapping_table->num_max_voltage_levels = i; + + if (disp_clk_voltage_mapping_table->num_max_voltage_levels == 0) { + disp_clk_voltage_mapping_table->display_clock_frequency[0] = 80000; + disp_clk_voltage_mapping_table->num_max_voltage_levels = 1; + } +} + +void sumo_construct_sclk_voltage_mapping_table(struct radeon_device *rdev, + struct sumo_sclk_voltage_mapping_table *sclk_voltage_mapping_table, + ATOM_AVAILABLE_SCLK_LIST *table) +{ + u32 i; + u32 n = 0; + u32 prev_sclk = 0; + + for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) { + if (table[i].ulSupportedSCLK > prev_sclk) { + sclk_voltage_mapping_table->entries[n].sclk_frequency = + table[i].ulSupportedSCLK; + sclk_voltage_mapping_table->entries[n].vid_2bit = + table[i].usVoltageIndex; + prev_sclk = table[i].ulSupportedSCLK; + n++; + } + } + + sclk_voltage_mapping_table->num_max_dpm_entries = n; +} + +void sumo_construct_vid_mapping_table(struct radeon_device *rdev, + struct sumo_vid_mapping_table *vid_mapping_table, + ATOM_AVAILABLE_SCLK_LIST *table) +{ + u32 i, j; + + for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) { + if (table[i].ulSupportedSCLK != 0) { + vid_mapping_table->entries[table[i].usVoltageIndex].vid_7bit = + table[i].usVoltageID; + vid_mapping_table->entries[table[i].usVoltageIndex].vid_2bit = + table[i].usVoltageIndex; + } + } + + for (i = 0; i < SUMO_MAX_NUMBER_VOLTAGES; i++) { + if (vid_mapping_table->entries[i].vid_7bit == 0) { + for (j = i + 1; j < SUMO_MAX_NUMBER_VOLTAGES; j++) { + if (vid_mapping_table->entries[j].vid_7bit != 0) { + vid_mapping_table->entries[i] = + vid_mapping_table->entries[j]; + vid_mapping_table->entries[j].vid_7bit = 0; + break; + } + } + + if (j == SUMO_MAX_NUMBER_VOLTAGES) + break; + } + } + + vid_mapping_table->num_entries = i; +} + +union igp_info { + struct _ATOM_INTEGRATED_SYSTEM_INFO info; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V5 info_5; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 info_6; +}; + +static int sumo_parse_sys_info_table(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_mode_info *mode_info = &rdev->mode_info; + int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo); + union igp_info *igp_info; + u8 frev, crev; + u16 data_offset; + int i; + + if (atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + igp_info = (union igp_info *)((uint8_t*)mode_info->atom_context->bios + + data_offset); + + if (crev != 6) { + DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev); + return -EINVAL; + } + pi->sys_info.bootup_sclk = le32_to_cpu(igp_info->info_6.ulBootUpEngineClock); + pi->sys_info.min_sclk = le32_to_cpu(igp_info->info_6.ulMinEngineClock); + pi->sys_info.bootup_uma_clk = le32_to_cpu(igp_info->info_6.ulBootUpUMAClock); + pi->sys_info.bootup_nb_voltage_index = + le16_to_cpu(igp_info->info_6.usBootUpNBVoltage); + if (igp_info->info_6.ucHtcTmpLmt == 0) + pi->sys_info.htc_tmp_lmt = 203; + else + pi->sys_info.htc_tmp_lmt = igp_info->info_6.ucHtcTmpLmt; + if (igp_info->info_6.ucHtcHystLmt == 0) + pi->sys_info.htc_hyst_lmt = 5; + else + pi->sys_info.htc_hyst_lmt = igp_info->info_6.ucHtcHystLmt; + if (pi->sys_info.htc_tmp_lmt <= pi->sys_info.htc_hyst_lmt) { + DRM_ERROR("The htcTmpLmt should be larger than htcHystLmt.\n"); + } + for (i = 0; i < NUMBER_OF_M3ARB_PARAM_SETS; i++) { + pi->sys_info.csr_m3_arb_cntl_default[i] = + le32_to_cpu(igp_info->info_6.ulCSR_M3_ARB_CNTL_DEFAULT[i]); + pi->sys_info.csr_m3_arb_cntl_uvd[i] = + le32_to_cpu(igp_info->info_6.ulCSR_M3_ARB_CNTL_UVD[i]); + pi->sys_info.csr_m3_arb_cntl_fs3d[i] = + le32_to_cpu(igp_info->info_6.ulCSR_M3_ARB_CNTL_FS3D[i]); + } + pi->sys_info.sclk_dpm_boost_margin = + le32_to_cpu(igp_info->info_6.SclkDpmBoostMargin); + pi->sys_info.sclk_dpm_throttle_margin = + le32_to_cpu(igp_info->info_6.SclkDpmThrottleMargin); + pi->sys_info.sclk_dpm_tdp_limit_pg = + le16_to_cpu(igp_info->info_6.SclkDpmTdpLimitPG); + pi->sys_info.gnb_tdp_limit = le16_to_cpu(igp_info->info_6.GnbTdpLimit); + pi->sys_info.sclk_dpm_tdp_limit_boost = + le16_to_cpu(igp_info->info_6.SclkDpmTdpLimitBoost); + pi->sys_info.boost_sclk = le32_to_cpu(igp_info->info_6.ulBoostEngineCLock); + pi->sys_info.boost_vid_2bit = igp_info->info_6.ulBoostVid_2bit; + if (igp_info->info_6.EnableBoost) + pi->sys_info.enable_boost = true; + else + pi->sys_info.enable_boost = false; + sumo_construct_display_voltage_mapping_table(rdev, + &pi->sys_info.disp_clk_voltage_mapping_table, + igp_info->info_6.sDISPCLK_Voltage); + sumo_construct_sclk_voltage_mapping_table(rdev, + &pi->sys_info.sclk_voltage_mapping_table, + igp_info->info_6.sAvail_SCLK); + sumo_construct_vid_mapping_table(rdev, &pi->sys_info.vid_mapping_table, + igp_info->info_6.sAvail_SCLK); + + } + return 0; +} + +static void sumo_construct_boot_and_acpi_state(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + pi->boot_pl.sclk = pi->sys_info.bootup_sclk; + pi->boot_pl.vddc_index = pi->sys_info.bootup_nb_voltage_index; + pi->boot_pl.ds_divider_index = 0; + pi->boot_pl.ss_divider_index = 0; + pi->boot_pl.allow_gnb_slow = 1; + pi->acpi_pl = pi->boot_pl; + pi->current_ps.num_levels = 1; + pi->current_ps.levels[0] = pi->boot_pl; +} + +int sumo_dpm_init(struct radeon_device *rdev) +{ + struct sumo_power_info *pi; + u32 hw_rev = (RREG32(HW_REV) & ATI_REV_ID_MASK) >> ATI_REV_ID_SHIFT; + int ret; + + pi = kzalloc(sizeof(struct sumo_power_info), GFP_KERNEL); + if (pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = pi; + + pi->driver_nbps_policy_disable = false; + if ((rdev->family == CHIP_PALM) && (hw_rev < 3)) + pi->disable_gfx_power_gating_in_uvd = true; + else + pi->disable_gfx_power_gating_in_uvd = false; + pi->enable_alt_vddnb = true; + pi->enable_sclk_ds = true; + pi->enable_dynamic_m3_arbiter = false; + pi->enable_dynamic_patch_ps = true; + /* Some PALM chips don't seem to properly ungate gfx when UVD is in use; + * for now just disable gfx PG. + */ + if (rdev->family == CHIP_PALM) + pi->enable_gfx_power_gating = false; + else + pi->enable_gfx_power_gating = true; + pi->enable_gfx_clock_gating = true; + pi->enable_mg_clock_gating = true; + pi->enable_auto_thermal_throttling = true; + + ret = sumo_parse_sys_info_table(rdev); + if (ret) + return ret; + + sumo_construct_boot_and_acpi_state(rdev); + + ret = sumo_parse_power_table(rdev); + if (ret) + return ret; + + pi->pasi = CYPRESS_HASI_DFLT; + pi->asi = RV770_ASI_DFLT; + pi->thermal_auto_throttling = pi->sys_info.htc_tmp_lmt; + pi->enable_boost = pi->sys_info.enable_boost; + pi->enable_dpm = true; + + return 0; +} + +void sumo_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + int i; + struct sumo_ps *ps = sumo_get_ps(rps); + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + for (i = 0; i < ps->num_levels; i++) { + struct sumo_pl *pl = &ps->levels[i]; + printk("\t\tpower level %d sclk: %u vddc: %u\n", + i, pl->sclk, + sumo_convert_voltage_index_to_value(rdev, pl->vddc_index)); + } + r600_dpm_print_ps_status(rdev, rps); +} + +void sumo_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct sumo_ps *ps = sumo_get_ps(rps); + struct sumo_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURR_INDEX_MASK) >> + CURR_INDEX_SHIFT; + + if (current_index == BOOST_DPM_LEVEL) { + pl = &pi->boost_pl; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u vddc: %u\n", + current_index, pl->sclk, + sumo_convert_voltage_index_to_value(rdev, pl->vddc_index)); + } else if (current_index >= ps->num_levels) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + pl = &ps->levels[current_index]; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u vddc: %u\n", + current_index, pl->sclk, + sumo_convert_voltage_index_to_value(rdev, pl->vddc_index)); + } +} + +void sumo_dpm_fini(struct radeon_device *rdev) +{ + int i; + + sumo_cleanup_asic(rdev); /* ??? */ + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +u32 sumo_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct sumo_ps *requested_state = sumo_get_ps(&pi->requested_rps); + + if (low) + return requested_state->levels[0].sclk; + else + return requested_state->levels[requested_state->num_levels - 1].sclk; +} + +u32 sumo_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + + return pi->sys_info.bootup_uma_clk; +} + +int sumo_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + struct radeon_ps *rps = &pi->current_rps; + struct sumo_ps *ps = sumo_get_ps(rps); + int i; + + if (ps->num_levels <= 1) + return 0; + + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + if (pi->enable_boost) + sumo_enable_boost(rdev, rps, false); + sumo_power_level_enable(rdev, ps->num_levels - 1, true); + sumo_set_forced_level(rdev, ps->num_levels - 1); + sumo_set_forced_mode_enabled(rdev); + for (i = 0; i < ps->num_levels - 1; i++) { + sumo_power_level_enable(rdev, i, false); + } + sumo_set_forced_mode(rdev, false); + sumo_set_forced_mode_enabled(rdev); + sumo_set_forced_mode(rdev, false); + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + if (pi->enable_boost) + sumo_enable_boost(rdev, rps, false); + sumo_power_level_enable(rdev, 0, true); + sumo_set_forced_level(rdev, 0); + sumo_set_forced_mode_enabled(rdev); + for (i = 1; i < ps->num_levels; i++) { + sumo_power_level_enable(rdev, i, false); + } + sumo_set_forced_mode(rdev, false); + sumo_set_forced_mode_enabled(rdev); + sumo_set_forced_mode(rdev, false); + } else { + for (i = 0; i < ps->num_levels; i++) { + sumo_power_level_enable(rdev, i, true); + } + if (pi->enable_boost) + sumo_enable_boost(rdev, rps, true); + } + + rdev->pm.dpm.forced_level = level; + + return 0; +} diff --git a/sys/dev/drm/radeon/sumo_dpm.h b/sys/dev/drm/radeon/sumo_dpm.h new file mode 100644 index 0000000000..07dda299c7 --- /dev/null +++ b/sys/dev/drm/radeon/sumo_dpm.h @@ -0,0 +1,220 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __SUMO_DPM_H__ +#define __SUMO_DPM_H__ + +#include "atom.h" + +#define SUMO_MAX_HARDWARE_POWERLEVELS 5 +#define SUMO_PM_NUMBER_OF_TC 15 + +struct sumo_pl { + u32 sclk; + u32 vddc_index; + u32 ds_divider_index; + u32 ss_divider_index; + u32 allow_gnb_slow; + u32 sclk_dpm_tdp_limit; +}; + +/* used for the flags field */ +#define SUMO_POWERSTATE_FLAGS_FORCE_NBPS1_STATE (1 << 0) +#define SUMO_POWERSTATE_FLAGS_BOOST_STATE (1 << 1) + +struct sumo_ps { + struct sumo_pl levels[SUMO_MAX_HARDWARE_POWERLEVELS]; + u32 num_levels; + /* flags */ + u32 flags; +}; + +#define NUMBER_OF_M3ARB_PARAM_SETS 10 +#define SUMO_MAX_NUMBER_VOLTAGES 4 + +struct sumo_disp_clock_voltage_mapping_table { + u32 num_max_voltage_levels; + u32 display_clock_frequency[SUMO_MAX_NUMBER_VOLTAGES]; +}; + +struct sumo_vid_mapping_entry { + u16 vid_2bit; + u16 vid_7bit; +}; + +struct sumo_vid_mapping_table { + u32 num_entries; + struct sumo_vid_mapping_entry entries[SUMO_MAX_NUMBER_VOLTAGES]; +}; + +struct sumo_sclk_voltage_mapping_entry { + u32 sclk_frequency; + u16 vid_2bit; + u16 rsv; +}; + +struct sumo_sclk_voltage_mapping_table { + u32 num_max_dpm_entries; + struct sumo_sclk_voltage_mapping_entry entries[SUMO_MAX_HARDWARE_POWERLEVELS]; +}; + +struct sumo_sys_info { + u32 bootup_sclk; + u32 min_sclk; + u32 bootup_uma_clk; + u16 bootup_nb_voltage_index; + u8 htc_tmp_lmt; + u8 htc_hyst_lmt; + struct sumo_sclk_voltage_mapping_table sclk_voltage_mapping_table; + struct sumo_disp_clock_voltage_mapping_table disp_clk_voltage_mapping_table; + struct sumo_vid_mapping_table vid_mapping_table; + u32 csr_m3_arb_cntl_default[NUMBER_OF_M3ARB_PARAM_SETS]; + u32 csr_m3_arb_cntl_uvd[NUMBER_OF_M3ARB_PARAM_SETS]; + u32 csr_m3_arb_cntl_fs3d[NUMBER_OF_M3ARB_PARAM_SETS]; + u32 sclk_dpm_boost_margin; + u32 sclk_dpm_throttle_margin; + u32 sclk_dpm_tdp_limit_pg; + u32 gnb_tdp_limit; + u32 sclk_dpm_tdp_limit_boost; + u32 boost_sclk; + u32 boost_vid_2bit; + bool enable_boost; +}; + +struct sumo_power_info { + u32 asi; + u32 pasi; + u32 bsp; + u32 bsu; + u32 pbsp; + u32 pbsu; + u32 dsp; + u32 psp; + u32 thermal_auto_throttling; + u32 uvd_m3_arbiter; + u32 fw_version; + struct sumo_sys_info sys_info; + struct sumo_pl acpi_pl; + struct sumo_pl boot_pl; + struct sumo_pl boost_pl; + bool disable_gfx_power_gating_in_uvd; + bool driver_nbps_policy_disable; + bool enable_alt_vddnb; + bool enable_dynamic_m3_arbiter; + bool enable_gfx_clock_gating; + bool enable_gfx_power_gating; + bool enable_mg_clock_gating; + bool enable_sclk_ds; + bool enable_auto_thermal_throttling; + bool enable_dynamic_patch_ps; + bool enable_dpm; + bool enable_boost; + struct radeon_ps current_rps; + struct sumo_ps current_ps; + struct radeon_ps requested_rps; + struct sumo_ps requested_ps; +}; + +#define SUMO_UTC_DFLT_00 0x48 +#define SUMO_UTC_DFLT_01 0x44 +#define SUMO_UTC_DFLT_02 0x44 +#define SUMO_UTC_DFLT_03 0x44 +#define SUMO_UTC_DFLT_04 0x44 +#define SUMO_UTC_DFLT_05 0x44 +#define SUMO_UTC_DFLT_06 0x44 +#define SUMO_UTC_DFLT_07 0x44 +#define SUMO_UTC_DFLT_08 0x44 +#define SUMO_UTC_DFLT_09 0x44 +#define SUMO_UTC_DFLT_10 0x44 +#define SUMO_UTC_DFLT_11 0x44 +#define SUMO_UTC_DFLT_12 0x44 +#define SUMO_UTC_DFLT_13 0x44 +#define SUMO_UTC_DFLT_14 0x44 + +#define SUMO_DTC_DFLT_00 0x48 +#define SUMO_DTC_DFLT_01 0x44 +#define SUMO_DTC_DFLT_02 0x44 +#define SUMO_DTC_DFLT_03 0x44 +#define SUMO_DTC_DFLT_04 0x44 +#define SUMO_DTC_DFLT_05 0x44 +#define SUMO_DTC_DFLT_06 0x44 +#define SUMO_DTC_DFLT_07 0x44 +#define SUMO_DTC_DFLT_08 0x44 +#define SUMO_DTC_DFLT_09 0x44 +#define SUMO_DTC_DFLT_10 0x44 +#define SUMO_DTC_DFLT_11 0x44 +#define SUMO_DTC_DFLT_12 0x44 +#define SUMO_DTC_DFLT_13 0x44 +#define SUMO_DTC_DFLT_14 0x44 + +#define SUMO_AH_DFLT 5 + +#define SUMO_R_DFLT0 70 +#define SUMO_R_DFLT1 70 +#define SUMO_R_DFLT2 70 +#define SUMO_R_DFLT3 70 +#define SUMO_R_DFLT4 100 + +#define SUMO_L_DFLT0 0 +#define SUMO_L_DFLT1 20 +#define SUMO_L_DFLT2 20 +#define SUMO_L_DFLT3 20 +#define SUMO_L_DFLT4 20 +#define SUMO_VRC_DFLT 0x30033 +#define SUMO_MGCGTTLOCAL0_DFLT 0 +#define SUMO_MGCGTTLOCAL1_DFLT 0 +#define SUMO_GICST_DFLT 19 +#define SUMO_SST_DFLT 8 +#define SUMO_VOLTAGEDROPT_DFLT 1 +#define SUMO_GFXPOWERGATINGT_DFLT 100 + +/* sumo_dpm.c */ +void sumo_gfx_clockgating_initialize(struct radeon_device *rdev); +void sumo_program_vc(struct radeon_device *rdev, u32 vrc); +void sumo_clear_vc(struct radeon_device *rdev); +void sumo_program_sstp(struct radeon_device *rdev); +void sumo_take_smu_control(struct radeon_device *rdev, bool enable); +void sumo_construct_sclk_voltage_mapping_table(struct radeon_device *rdev, + struct sumo_sclk_voltage_mapping_table *sclk_voltage_mapping_table, + ATOM_AVAILABLE_SCLK_LIST *table); +void sumo_construct_vid_mapping_table(struct radeon_device *rdev, + struct sumo_vid_mapping_table *vid_mapping_table, + ATOM_AVAILABLE_SCLK_LIST *table); +u32 sumo_convert_vid2_to_vid7(struct radeon_device *rdev, + struct sumo_vid_mapping_table *vid_mapping_table, + u32 vid_2bit); +u32 sumo_get_sleep_divider_from_id(u32 id); +u32 sumo_get_sleep_divider_id_from_clock(struct radeon_device *rdev, + u32 sclk, + u32 min_sclk_in_sr); + +/* sumo_smc.c */ +void sumo_initialize_m3_arb(struct radeon_device *rdev); +void sumo_smu_pg_init(struct radeon_device *rdev); +void sumo_set_tdp_limit(struct radeon_device *rdev, u32 index, u32 tdp_limit); +void sumo_smu_notify_alt_vddnb_change(struct radeon_device *rdev, + bool powersaving, bool force_nbps1); +void sumo_boost_state_enable(struct radeon_device *rdev, bool enable); +void sumo_enable_boost_timer(struct radeon_device *rdev); +u32 sumo_get_running_fw_version(struct radeon_device *rdev); + +#endif diff --git a/sys/dev/drm/radeon/sumo_smc.c b/sys/dev/drm/radeon/sumo_smc.c new file mode 100644 index 0000000000..fe2d6788b6 --- /dev/null +++ b/sys/dev/drm/radeon/sumo_smc.c @@ -0,0 +1,221 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "sumod.h" +#include "sumo_dpm.h" +#include "ppsmc.h" + +#define SUMO_SMU_SERVICE_ROUTINE_PG_INIT 1 +#define SUMO_SMU_SERVICE_ROUTINE_ALTVDDNB_NOTIFY 27 +#define SUMO_SMU_SERVICE_ROUTINE_GFX_SRV_ID_20 20 + +struct sumo_ps *sumo_get_ps(struct radeon_ps *rps); +struct sumo_power_info *sumo_get_pi(struct radeon_device *rdev); + +static void sumo_send_msg_to_smu(struct radeon_device *rdev, u32 id) +{ + u32 gfx_int_req; + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(GFX_INT_STATUS) & INT_DONE) + break; + DRM_UDELAY(1); + } + + gfx_int_req = SERV_INDEX(id) | INT_REQ; + WREG32(GFX_INT_REQ, gfx_int_req); + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(GFX_INT_REQ) & INT_REQ) + break; + DRM_UDELAY(1); + } + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(GFX_INT_STATUS) & INT_ACK) + break; + DRM_UDELAY(1); + } + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(GFX_INT_STATUS) & INT_DONE) + break; + DRM_UDELAY(1); + } + + gfx_int_req &= ~INT_REQ; + WREG32(GFX_INT_REQ, gfx_int_req); +} + +void sumo_initialize_m3_arb(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 i; + + if (!pi->enable_dynamic_m3_arbiter) + return; + + for (i = 0; i < NUMBER_OF_M3ARB_PARAM_SETS; i++) + WREG32_RCU(MCU_M3ARB_PARAMS + (i * 4), + pi->sys_info.csr_m3_arb_cntl_default[i]); + + for (; i < NUMBER_OF_M3ARB_PARAM_SETS * 2; i++) + WREG32_RCU(MCU_M3ARB_PARAMS + (i * 4), + pi->sys_info.csr_m3_arb_cntl_uvd[i % NUMBER_OF_M3ARB_PARAM_SETS]); + + for (; i < NUMBER_OF_M3ARB_PARAM_SETS * 3; i++) + WREG32_RCU(MCU_M3ARB_PARAMS + (i * 4), + pi->sys_info.csr_m3_arb_cntl_fs3d[i % NUMBER_OF_M3ARB_PARAM_SETS]); +} + +static bool sumo_is_alt_vddnb_supported(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + bool return_code = false; + + if (!pi->enable_alt_vddnb) + return return_code; + + if ((rdev->family == CHIP_SUMO) || (rdev->family == CHIP_SUMO2)) { + if (pi->fw_version >= 0x00010C00) + return_code = true; + } + + return return_code; +} + +void sumo_smu_notify_alt_vddnb_change(struct radeon_device *rdev, + bool powersaving, bool force_nbps1) +{ + u32 param = 0; + + if (!sumo_is_alt_vddnb_supported(rdev)) + return; + + if (powersaving) + param |= 1; + + if (force_nbps1) + param |= 2; + + WREG32_RCU(RCU_ALTVDDNB_NOTIFY, param); + + sumo_send_msg_to_smu(rdev, SUMO_SMU_SERVICE_ROUTINE_ALTVDDNB_NOTIFY); +} + +void sumo_smu_pg_init(struct radeon_device *rdev) +{ + sumo_send_msg_to_smu(rdev, SUMO_SMU_SERVICE_ROUTINE_PG_INIT); +} + +static u32 sumo_power_of_4(u32 unit) +{ + u32 ret = 1; + u32 i; + + for (i = 0; i < unit; i++) + ret *= 4; + + return ret; +} + +void sumo_enable_boost_timer(struct radeon_device *rdev) +{ + struct sumo_power_info *pi = sumo_get_pi(rdev); + u32 period, unit, timer_value; + u32 xclk = radeon_get_xclk(rdev); + + unit = (RREG32_RCU(RCU_LCLK_SCALING_CNTL) & LCLK_SCALING_TIMER_PRESCALER_MASK) + >> LCLK_SCALING_TIMER_PRESCALER_SHIFT; + + period = 100 * (xclk / 100 / sumo_power_of_4(unit)); + + timer_value = (period << 16) | (unit << 4); + + WREG32_RCU(RCU_GNB_PWR_REP_TIMER_CNTL, timer_value); + WREG32_RCU(RCU_BOOST_MARGIN, pi->sys_info.sclk_dpm_boost_margin); + WREG32_RCU(RCU_THROTTLE_MARGIN, pi->sys_info.sclk_dpm_throttle_margin); + WREG32_RCU(GNB_TDP_LIMIT, pi->sys_info.gnb_tdp_limit); + WREG32_RCU(RCU_SclkDpmTdpLimitPG, pi->sys_info.sclk_dpm_tdp_limit_pg); + + sumo_send_msg_to_smu(rdev, SUMO_SMU_SERVICE_ROUTINE_GFX_SRV_ID_20); +} + +void sumo_set_tdp_limit(struct radeon_device *rdev, u32 index, u32 tdp_limit) +{ + u32 regoffset = 0; + u32 shift = 0; + u32 mask = 0xFFF; + u32 sclk_dpm_tdp_limit; + + switch (index) { + case 0: + regoffset = RCU_SclkDpmTdpLimit01; + shift = 16; + break; + case 1: + regoffset = RCU_SclkDpmTdpLimit01; + shift = 0; + break; + case 2: + regoffset = RCU_SclkDpmTdpLimit23; + shift = 16; + break; + case 3: + regoffset = RCU_SclkDpmTdpLimit23; + shift = 0; + break; + case 4: + regoffset = RCU_SclkDpmTdpLimit47; + shift = 16; + break; + case 7: + regoffset = RCU_SclkDpmTdpLimit47; + shift = 0; + break; + default: + break; + } + + sclk_dpm_tdp_limit = RREG32_RCU(regoffset); + sclk_dpm_tdp_limit &= ~(mask << shift); + sclk_dpm_tdp_limit |= (tdp_limit << shift); + WREG32_RCU(regoffset, sclk_dpm_tdp_limit); +} + +void sumo_boost_state_enable(struct radeon_device *rdev, bool enable) +{ + u32 boost_disable = RREG32_RCU(RCU_GPU_BOOST_DISABLE); + + boost_disable &= 0xFFFFFFFE; + boost_disable |= (enable ? 0 : 1); + WREG32_RCU(RCU_GPU_BOOST_DISABLE, boost_disable); +} + +u32 sumo_get_running_fw_version(struct radeon_device *rdev) +{ + return RREG32_RCU(RCU_FW_VERSION); +} diff --git a/sys/dev/drm/radeon/sumod.h b/sys/dev/drm/radeon/sumod.h new file mode 100644 index 0000000000..7c9c2d4b86 --- /dev/null +++ b/sys/dev/drm/radeon/sumod.h @@ -0,0 +1,372 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#ifndef _SUMOD_H_ +#define _SUMOD_H_ + +/* pm registers */ + +/* rcu */ +#define RCU_FW_VERSION 0x30c + +#define RCU_PWR_GATING_SEQ0 0x408 +#define RCU_PWR_GATING_SEQ1 0x40c +#define RCU_PWR_GATING_CNTL 0x410 +# define PWR_GATING_EN (1 << 0) +# define RSVD_MASK (0x3 << 1) +# define PCV(x) ((x) << 3) +# define PCV_MASK (0x1f << 3) +# define PCV_SHIFT 3 +# define PCP(x) ((x) << 8) +# define PCP_MASK (0xf << 8) +# define PCP_SHIFT 8 +# define RPW(x) ((x) << 16) +# define RPW_MASK (0xf << 16) +# define RPW_SHIFT 16 +# define ID(x) ((x) << 24) +# define ID_MASK (0xf << 24) +# define ID_SHIFT 24 +# define PGS(x) ((x) << 28) +# define PGS_MASK (0xf << 28) +# define PGS_SHIFT 28 + +#define RCU_ALTVDDNB_NOTIFY 0x430 +#define RCU_LCLK_SCALING_CNTL 0x434 +# define LCLK_SCALING_EN (1 << 0) +# define LCLK_SCALING_TYPE (1 << 1) +# define LCLK_SCALING_TIMER_PRESCALER(x) ((x) << 4) +# define LCLK_SCALING_TIMER_PRESCALER_MASK (0xf << 4) +# define LCLK_SCALING_TIMER_PRESCALER_SHIFT 4 +# define LCLK_SCALING_TIMER_PERIOD(x) ((x) << 16) +# define LCLK_SCALING_TIMER_PERIOD_MASK (0xf << 16) +# define LCLK_SCALING_TIMER_PERIOD_SHIFT 16 + +#define RCU_PWR_GATING_CNTL_2 0x4a0 +# define MPPU(x) ((x) << 0) +# define MPPU_MASK (0xffff << 0) +# define MPPU_SHIFT 0 +# define MPPD(x) ((x) << 16) +# define MPPD_MASK (0xffff << 16) +# define MPPD_SHIFT 16 +#define RCU_PWR_GATING_CNTL_3 0x4a4 +# define DPPU(x) ((x) << 0) +# define DPPU_MASK (0xffff << 0) +# define DPPU_SHIFT 0 +# define DPPD(x) ((x) << 16) +# define DPPD_MASK (0xffff << 16) +# define DPPD_SHIFT 16 +#define RCU_PWR_GATING_CNTL_4 0x4a8 +# define RT(x) ((x) << 0) +# define RT_MASK (0xffff << 0) +# define RT_SHIFT 0 +# define IT(x) ((x) << 16) +# define IT_MASK (0xffff << 16) +# define IT_SHIFT 16 + +/* yes these two have the same address */ +#define RCU_PWR_GATING_CNTL_5 0x504 +#define RCU_GPU_BOOST_DISABLE 0x508 + +#define MCU_M3ARB_INDEX 0x504 +#define MCU_M3ARB_PARAMS 0x508 + +#define RCU_GNB_PWR_REP_TIMER_CNTL 0x50C + +#define RCU_SclkDpmTdpLimit01 0x514 +#define RCU_SclkDpmTdpLimit23 0x518 +#define RCU_SclkDpmTdpLimit47 0x51C +#define RCU_SclkDpmTdpLimitPG 0x520 + +#define GNB_TDP_LIMIT 0x540 +#define RCU_BOOST_MARGIN 0x544 +#define RCU_THROTTLE_MARGIN 0x548 + +#define SMU_PCIE_PG_ARGS 0x58C +#define SMU_PCIE_PG_ARGS_2 0x598 +#define SMU_PCIE_PG_ARGS_3 0x59C + +/* mmio */ +#define RCU_STATUS 0x11c +# define GMC_PWR_GATER_BUSY (1 << 8) +# define GFX_PWR_GATER_BUSY (1 << 9) +# define UVD_PWR_GATER_BUSY (1 << 10) +# define PCIE_PWR_GATER_BUSY (1 << 11) +# define GMC_PWR_GATER_STATE (1 << 12) +# define GFX_PWR_GATER_STATE (1 << 13) +# define UVD_PWR_GATER_STATE (1 << 14) +# define PCIE_PWR_GATER_STATE (1 << 15) +# define GFX1_PWR_GATER_BUSY (1 << 16) +# define GFX2_PWR_GATER_BUSY (1 << 17) +# define GFX1_PWR_GATER_STATE (1 << 18) +# define GFX2_PWR_GATER_STATE (1 << 19) + +#define GFX_INT_REQ 0x120 +# define INT_REQ (1 << 0) +# define SERV_INDEX(x) ((x) << 1) +# define SERV_INDEX_MASK (0xff << 1) +# define SERV_INDEX_SHIFT 1 +#define GFX_INT_STATUS 0x124 +# define INT_ACK (1 << 0) +# define INT_DONE (1 << 1) + +#define CG_SCLK_CNTL 0x600 +# define SCLK_DIVIDER(x) ((x) << 0) +# define SCLK_DIVIDER_MASK (0x7f << 0) +# define SCLK_DIVIDER_SHIFT 0 +#define CG_SCLK_STATUS 0x604 +# define SCLK_OVERCLK_DETECT (1 << 2) + +#define CG_DCLK_CNTL 0x610 +# define DCLK_DIVIDER_MASK 0x7f +# define DCLK_DIR_CNTL_EN (1 << 8) +#define CG_DCLK_STATUS 0x614 +# define DCLK_STATUS (1 << 0) +#define CG_VCLK_CNTL 0x618 +# define VCLK_DIVIDER_MASK 0x7f +# define VCLK_DIR_CNTL_EN (1 << 8) +#define CG_VCLK_STATUS 0x61c + +#define GENERAL_PWRMGT 0x63c +# define STATIC_PM_EN (1 << 1) + +#define SCLK_PWRMGT_CNTL 0x644 +# define SCLK_PWRMGT_OFF (1 << 0) +# define SCLK_LOW_D1 (1 << 1) +# define FIR_RESET (1 << 4) +# define FIR_FORCE_TREND_SEL (1 << 5) +# define FIR_TREND_MODE (1 << 6) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define GFX_CLK_REQUEST_OFF (1 << 9) +# define GFX_CLK_FORCE_OFF (1 << 10) +# define GFX_CLK_OFF_ACPI_D1 (1 << 11) +# define GFX_CLK_OFF_ACPI_D2 (1 << 12) +# define GFX_CLK_OFF_ACPI_D3 (1 << 13) +# define GFX_VOLTAGE_CHANGE_EN (1 << 16) +# define GFX_VOLTAGE_CHANGE_MODE (1 << 17) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x66c +# define TARG_SCLK_INDEX(x) ((x) << 6) +# define TARG_SCLK_INDEX_MASK (0x7 << 6) +# define TARG_SCLK_INDEX_SHIFT 6 +# define CURR_SCLK_INDEX(x) ((x) << 9) +# define CURR_SCLK_INDEX_MASK (0x7 << 9) +# define CURR_SCLK_INDEX_SHIFT 9 +# define TARG_INDEX(x) ((x) << 12) +# define TARG_INDEX_MASK (0x7 << 12) +# define TARG_INDEX_SHIFT 12 +# define CURR_INDEX(x) ((x) << 15) +# define CURR_INDEX_MASK (0x7 << 15) +# define CURR_INDEX_SHIFT 15 + +#define CG_SCLK_DPM_CTRL 0x684 +# define SCLK_FSTATE_0_DIV(x) ((x) << 0) +# define SCLK_FSTATE_0_DIV_MASK (0x7f << 0) +# define SCLK_FSTATE_0_DIV_SHIFT 0 +# define SCLK_FSTATE_0_VLD (1 << 7) +# define SCLK_FSTATE_1_DIV(x) ((x) << 8) +# define SCLK_FSTATE_1_DIV_MASK (0x7f << 8) +# define SCLK_FSTATE_1_DIV_SHIFT 8 +# define SCLK_FSTATE_1_VLD (1 << 15) +# define SCLK_FSTATE_2_DIV(x) ((x) << 16) +# define SCLK_FSTATE_2_DIV_MASK (0x7f << 16) +# define SCLK_FSTATE_2_DIV_SHIFT 16 +# define SCLK_FSTATE_2_VLD (1 << 23) +# define SCLK_FSTATE_3_DIV(x) ((x) << 24) +# define SCLK_FSTATE_3_DIV_MASK (0x7f << 24) +# define SCLK_FSTATE_3_DIV_SHIFT 24 +# define SCLK_FSTATE_3_VLD (1 << 31) +#define CG_SCLK_DPM_CTRL_2 0x688 +#define CG_GCOOR 0x68c +# define PHC(x) ((x) << 0) +# define PHC_MASK (0x1f << 0) +# define PHC_SHIFT 0 +# define SDC(x) ((x) << 9) +# define SDC_MASK (0x3ff << 9) +# define SDC_SHIFT 9 +# define SU(x) ((x) << 23) +# define SU_MASK (0xf << 23) +# define SU_SHIFT 23 +# define DIV_ID(x) ((x) << 28) +# define DIV_ID_MASK (0x7 << 28) +# define DIV_ID_SHIFT 28 + +#define CG_FTV 0x690 +#define CG_FFCT_0 0x694 +# define UTC_0(x) ((x) << 0) +# define UTC_0_MASK (0x3ff << 0) +# define UTC_0_SHIFT 0 +# define DTC_0(x) ((x) << 10) +# define DTC_0_MASK (0x3ff << 10) +# define DTC_0_SHIFT 10 + +#define CG_GIT 0x6d8 +# define CG_GICST(x) ((x) << 0) +# define CG_GICST_MASK (0xffff << 0) +# define CG_GICST_SHIFT 0 +# define CG_GIPOT(x) ((x) << 16) +# define CG_GIPOT_MASK (0xffff << 16) +# define CG_GIPOT_SHIFT 16 + +#define CG_SCLK_DPM_CTRL_3 0x6e0 +# define FORCE_SCLK_STATE(x) ((x) << 0) +# define FORCE_SCLK_STATE_MASK (0x7 << 0) +# define FORCE_SCLK_STATE_SHIFT 0 +# define FORCE_SCLK_STATE_EN (1 << 3) +# define GNB_TT(x) ((x) << 8) +# define GNB_TT_MASK (0xff << 8) +# define GNB_TT_SHIFT 8 +# define GNB_THERMTHRO_MASK (1 << 16) +# define CNB_THERMTHRO_MASK_SCLK (1 << 17) +# define DPM_SCLK_ENABLE (1 << 18) +# define GNB_SLOW_FSTATE_0_MASK (1 << 23) +# define GNB_SLOW_FSTATE_0_SHIFT 23 +# define FORCE_NB_PSTATE_1 (1 << 31) + +#define CG_SSP 0x6e8 +# define SST(x) ((x) << 0) +# define SST_MASK (0xffff << 0) +# define SST_SHIFT 0 +# define SSTU(x) ((x) << 16) +# define SSTU_MASK (0xffff << 16) +# define SSTU_SHIFT 16 + +#define CG_ACPI_CNTL 0x70c +# define SCLK_ACPI_DIV(x) ((x) << 0) +# define SCLK_ACPI_DIV_MASK (0x7f << 0) +# define SCLK_ACPI_DIV_SHIFT 0 + +#define CG_SCLK_DPM_CTRL_4 0x71c +# define DC_HDC(x) ((x) << 14) +# define DC_HDC_MASK (0x3fff << 14) +# define DC_HDC_SHIFT 14 +# define DC_HU(x) ((x) << 28) +# define DC_HU_MASK (0xf << 28) +# define DC_HU_SHIFT 28 +#define CG_SCLK_DPM_CTRL_5 0x720 +# define SCLK_FSTATE_BOOTUP(x) ((x) << 0) +# define SCLK_FSTATE_BOOTUP_MASK (0x7 << 0) +# define SCLK_FSTATE_BOOTUP_SHIFT 0 +# define TT_TP(x) ((x) << 3) +# define TT_TP_MASK (0xffff << 3) +# define TT_TP_SHIFT 3 +# define TT_TU(x) ((x) << 19) +# define TT_TU_MASK (0xff << 19) +# define TT_TU_SHIFT 19 +#define CG_SCLK_DPM_CTRL_6 0x724 +#define CG_AT_0 0x728 +# define CG_R(x) ((x) << 0) +# define CG_R_MASK (0xffff << 0) +# define CG_R_SHIFT 0 +# define CG_L(x) ((x) << 16) +# define CG_L_MASK (0xffff << 16) +# define CG_L_SHIFT 16 +#define CG_AT_1 0x72c +#define CG_AT_2 0x730 +#define CG_THERMAL_INT 0x734 +#define DIG_THERM_INTH(x) ((x) << 8) +#define DIG_THERM_INTH_MASK 0x0000FF00 +#define DIG_THERM_INTH_SHIFT 8 +#define DIG_THERM_INTL(x) ((x) << 16) +#define DIG_THERM_INTL_MASK 0x00FF0000 +#define DIG_THERM_INTL_SHIFT 16 +#define THERM_INT_MASK_HIGH (1 << 24) +#define THERM_INT_MASK_LOW (1 << 25) +#define CG_AT_3 0x738 +#define CG_AT_4 0x73c +#define CG_AT_5 0x740 +#define CG_AT_6 0x744 +#define CG_AT_7 0x748 + +#define CG_BSP_0 0x750 +# define BSP(x) ((x) << 0) +# define BSP_MASK (0xffff << 0) +# define BSP_SHIFT 0 +# define BSU(x) ((x) << 16) +# define BSU_MASK (0xf << 16) +# define BSU_SHIFT 16 + +#define CG_CG_VOLTAGE_CNTL 0x770 +# define REQ (1 << 0) +# define LEVEL(x) ((x) << 1) +# define LEVEL_MASK (0x3 << 1) +# define LEVEL_SHIFT 1 +# define CG_VOLTAGE_EN (1 << 3) +# define FORCE (1 << 4) +# define PERIOD(x) ((x) << 8) +# define PERIOD_MASK (0xffff << 8) +# define PERIOD_SHIFT 8 +# define UNIT(x) ((x) << 24) +# define UNIT_MASK (0xf << 24) +# define UNIT_SHIFT 24 + +#define CG_ACPI_VOLTAGE_CNTL 0x780 +# define ACPI_VOLTAGE_EN (1 << 8) + +#define CG_DPM_VOLTAGE_CNTL 0x788 +# define DPM_STATE0_LEVEL_MASK (0x3 << 0) +# define DPM_STATE0_LEVEL_SHIFT 0 +# define DPM_VOLTAGE_EN (1 << 16) + +#define CG_PWR_GATING_CNTL 0x7ac +# define DYN_PWR_DOWN_EN (1 << 0) +# define ACPI_PWR_DOWN_EN (1 << 1) +# define GFX_CLK_OFF_PWR_DOWN_EN (1 << 2) +# define IOC_DISGPU_PWR_DOWN_EN (1 << 3) +# define FORCE_POWR_ON (1 << 4) +# define PGP(x) ((x) << 8) +# define PGP_MASK (0xffff << 8) +# define PGP_SHIFT 8 +# define PGU(x) ((x) << 24) +# define PGU_MASK (0xf << 24) +# define PGU_SHIFT 24 + +#define CG_CGTT_LOCAL_0 0x7d0 +#define CG_CGTT_LOCAL_1 0x7d4 + +#define DEEP_SLEEP_CNTL 0x818 +# define R_DIS (1 << 3) +# define HS(x) ((x) << 4) +# define HS_MASK (0xfff << 4) +# define HS_SHIFT 4 +# define ENABLE_DS (1 << 31) +#define DEEP_SLEEP_CNTL2 0x81c +# define LB_UFP_EN (1 << 0) +# define INOUT_C(x) ((x) << 4) +# define INOUT_C_MASK (0xff << 4) +# define INOUT_C_SHIFT 4 + +#define CG_SCRATCH2 0x824 + +#define CG_SCLK_DPM_CTRL_11 0x830 + +#define HW_REV 0x5564 +# define ATI_REV_ID_MASK (0xf << 28) +# define ATI_REV_ID_SHIFT 28 +/* 0 = A0, 1 = A1, 2 = B0, 3 = C0, etc. */ + +#define DOUT_SCRATCH3 0x611c + +#define GB_ADDR_CONFIG 0x98f8 + +#endif diff --git a/sys/dev/drm/radeon/trinity_dpm.c b/sys/dev/drm/radeon/trinity_dpm.c new file mode 100644 index 0000000000..971a6ed0fa --- /dev/null +++ b/sys/dev/drm/radeon/trinity_dpm.c @@ -0,0 +1,1953 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "radeon_asic.h" +#include "trinityd.h" +#include "r600_dpm.h" +#include "trinity_dpm.h" +#include + +#define TRINITY_MAX_DEEPSLEEP_DIVIDER_ID 5 +#define TRINITY_MINIMUM_ENGINE_CLOCK 800 +#define SCLK_MIN_DIV_INTV_SHIFT 12 +#define TRINITY_DISPCLK_BYPASS_THRESHOLD 10000 + +#ifndef TRINITY_MGCG_SEQUENCE +#define TRINITY_MGCG_SEQUENCE 100 + +static const u32 trinity_mgcg_shls_default[] = +{ + /* Register, Value, Mask */ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00003fc4, 0xc0000000, 0xffffffff, + 0x00005448, 0x00000100, 0xffffffff, + 0x000055e4, 0x00000100, 0xffffffff, + 0x0000160c, 0x00000100, 0xffffffff, + 0x00008984, 0x06000100, 0xffffffff, + 0x0000c164, 0x00000100, 0xffffffff, + 0x00008a18, 0x00000100, 0xffffffff, + 0x0000897c, 0x06000100, 0xffffffff, + 0x00008b28, 0x00000100, 0xffffffff, + 0x00009144, 0x00800200, 0xffffffff, + 0x00009a60, 0x00000100, 0xffffffff, + 0x00009868, 0x00000100, 0xffffffff, + 0x00008d58, 0x00000100, 0xffffffff, + 0x00009510, 0x00000100, 0xffffffff, + 0x0000949c, 0x00000100, 0xffffffff, + 0x00009654, 0x00000100, 0xffffffff, + 0x00009030, 0x00000100, 0xffffffff, + 0x00009034, 0x00000100, 0xffffffff, + 0x00009038, 0x00000100, 0xffffffff, + 0x0000903c, 0x00000100, 0xffffffff, + 0x00009040, 0x00000100, 0xffffffff, + 0x0000a200, 0x00000100, 0xffffffff, + 0x0000a204, 0x00000100, 0xffffffff, + 0x0000a208, 0x00000100, 0xffffffff, + 0x0000a20c, 0x00000100, 0xffffffff, + 0x00009744, 0x00000100, 0xffffffff, + 0x00003f80, 0x00000100, 0xffffffff, + 0x0000a210, 0x00000100, 0xffffffff, + 0x0000a214, 0x00000100, 0xffffffff, + 0x000004d8, 0x00000100, 0xffffffff, + 0x00009664, 0x00000100, 0xffffffff, + 0x00009698, 0x00000100, 0xffffffff, + 0x000004d4, 0x00000200, 0xffffffff, + 0x000004d0, 0x00000000, 0xffffffff, + 0x000030cc, 0x00000104, 0xffffffff, + 0x0000d0c0, 0x00000100, 0xffffffff, + 0x0000d8c0, 0x00000100, 0xffffffff, + 0x0000951c, 0x00010000, 0xffffffff, + 0x00009160, 0x00030002, 0xffffffff, + 0x00009164, 0x00050004, 0xffffffff, + 0x00009168, 0x00070006, 0xffffffff, + 0x00009178, 0x00070000, 0xffffffff, + 0x0000917c, 0x00030002, 0xffffffff, + 0x00009180, 0x00050004, 0xffffffff, + 0x0000918c, 0x00010006, 0xffffffff, + 0x00009190, 0x00090008, 0xffffffff, + 0x00009194, 0x00070000, 0xffffffff, + 0x00009198, 0x00030002, 0xffffffff, + 0x0000919c, 0x00050004, 0xffffffff, + 0x000091a8, 0x00010006, 0xffffffff, + 0x000091ac, 0x00090008, 0xffffffff, + 0x000091b0, 0x00070000, 0xffffffff, + 0x000091b4, 0x00030002, 0xffffffff, + 0x000091b8, 0x00050004, 0xffffffff, + 0x000091c4, 0x00010006, 0xffffffff, + 0x000091c8, 0x00090008, 0xffffffff, + 0x000091cc, 0x00070000, 0xffffffff, + 0x000091d0, 0x00030002, 0xffffffff, + 0x000091d4, 0x00050004, 0xffffffff, + 0x000091e0, 0x00010006, 0xffffffff, + 0x000091e4, 0x00090008, 0xffffffff, + 0x000091e8, 0x00000000, 0xffffffff, + 0x000091ec, 0x00070000, 0xffffffff, + 0x000091f0, 0x00030002, 0xffffffff, + 0x000091f4, 0x00050004, 0xffffffff, + 0x00009200, 0x00010006, 0xffffffff, + 0x00009204, 0x00090008, 0xffffffff, + 0x00009208, 0x00070000, 0xffffffff, + 0x0000920c, 0x00030002, 0xffffffff, + 0x00009210, 0x00050004, 0xffffffff, + 0x0000921c, 0x00010006, 0xffffffff, + 0x00009220, 0x00090008, 0xffffffff, + 0x00009294, 0x00000000, 0xffffffff +}; + +static const u32 trinity_mgcg_shls_enable[] = +{ + /* Register, Value, Mask */ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0x00000000, 0x000133FF, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0x00000000, 0xE00B03FC, + 0x00009150, 0x96944200, 0xffffffff +}; + +static const u32 trinity_mgcg_shls_disable[] = +{ + /* Register, Value, Mask */ + 0x0000802c, 0xc0000000, 0xffffffff, + 0x00009150, 0x00600000, 0xffffffff, + 0x000008f8, 0x00000000, 0xffffffff, + 0x000008fc, 0xffffffff, 0x000133FF, + 0x000008f8, 0x00000001, 0xffffffff, + 0x000008fc, 0xffffffff, 0xE00B03FC +}; +#endif + +#ifndef TRINITY_SYSLS_SEQUENCE +#define TRINITY_SYSLS_SEQUENCE 100 + +static const u32 trinity_sysls_default[] = +{ + /* Register, Value, Mask */ + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x0000d8bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x00002f50, 0x00000404, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x0000641c, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; + +static const u32 trinity_sysls_disable[] = +{ + /* Register, Value, Mask */ + 0x0000d0c0, 0x00000000, 0xffffffff, + 0x0000d8c0, 0x00000000, 0xffffffff, + 0x000055e8, 0x00000000, 0xffffffff, + 0x0000d0bc, 0x00000000, 0xffffffff, + 0x0000d8bc, 0x00000000, 0xffffffff, + 0x000015c0, 0x00041401, 0xffffffff, + 0x0000264c, 0x00040400, 0xffffffff, + 0x00002648, 0x00040400, 0xffffffff, + 0x00002650, 0x00040400, 0xffffffff, + 0x000020b8, 0x00040400, 0xffffffff, + 0x000020bc, 0x00040400, 0xffffffff, + 0x000020c0, 0x00040c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680000, 0xffffffff, + 0x00002f50, 0x00000404, 0xffffffff, + 0x000004c8, 0x00000001, 0xffffffff, + 0x0000641c, 0x00007ffd, 0xffffffff, + 0x00000c7c, 0x0000ff00, 0xffffffff, + 0x00006dfc, 0x0000007f, 0xffffffff +}; + +static const u32 trinity_sysls_enable[] = +{ + /* Register, Value, Mask */ + 0x000055e8, 0x00000001, 0xffffffff, + 0x0000d0bc, 0x00000100, 0xffffffff, + 0x0000d8bc, 0x00000100, 0xffffffff, + 0x000015c0, 0x000c1401, 0xffffffff, + 0x0000264c, 0x000c0400, 0xffffffff, + 0x00002648, 0x000c0400, 0xffffffff, + 0x00002650, 0x000c0400, 0xffffffff, + 0x000020b8, 0x000c0400, 0xffffffff, + 0x000020bc, 0x000c0400, 0xffffffff, + 0x000020c0, 0x000c0c80, 0xffffffff, + 0x0000f4a0, 0x000000c0, 0xffffffff, + 0x0000f4a4, 0x00680fff, 0xffffffff, + 0x00002f50, 0x00000903, 0xffffffff, + 0x000004c8, 0x00000000, 0xffffffff, + 0x0000641c, 0x00000000, 0xffffffff, + 0x00000c7c, 0x00000000, 0xffffffff, + 0x00006dfc, 0x00000000, 0xffffffff +}; +#endif + +static const u32 trinity_override_mgpg_sequences[] = +{ + /* Register, Value */ + 0x00000200, 0xE030032C, + 0x00000204, 0x00000FFF, + 0x00000200, 0xE0300058, + 0x00000204, 0x00030301, + 0x00000200, 0xE0300054, + 0x00000204, 0x500010FF, + 0x00000200, 0xE0300074, + 0x00000204, 0x00030301, + 0x00000200, 0xE0300070, + 0x00000204, 0x500010FF, + 0x00000200, 0xE0300090, + 0x00000204, 0x00030301, + 0x00000200, 0xE030008C, + 0x00000204, 0x500010FF, + 0x00000200, 0xE03000AC, + 0x00000204, 0x00030301, + 0x00000200, 0xE03000A8, + 0x00000204, 0x500010FF, + 0x00000200, 0xE03000C8, + 0x00000204, 0x00030301, + 0x00000200, 0xE03000C4, + 0x00000204, 0x500010FF, + 0x00000200, 0xE03000E4, + 0x00000204, 0x00030301, + 0x00000200, 0xE03000E0, + 0x00000204, 0x500010FF, + 0x00000200, 0xE0300100, + 0x00000204, 0x00030301, + 0x00000200, 0xE03000FC, + 0x00000204, 0x500010FF, + 0x00000200, 0xE0300058, + 0x00000204, 0x00030303, + 0x00000200, 0xE0300054, + 0x00000204, 0x600010FF, + 0x00000200, 0xE0300074, + 0x00000204, 0x00030303, + 0x00000200, 0xE0300070, + 0x00000204, 0x600010FF, + 0x00000200, 0xE0300090, + 0x00000204, 0x00030303, + 0x00000200, 0xE030008C, + 0x00000204, 0x600010FF, + 0x00000200, 0xE03000AC, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000A8, + 0x00000204, 0x600010FF, + 0x00000200, 0xE03000C8, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000C4, + 0x00000204, 0x600010FF, + 0x00000200, 0xE03000E4, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000E0, + 0x00000204, 0x600010FF, + 0x00000200, 0xE0300100, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000FC, + 0x00000204, 0x600010FF, + 0x00000200, 0xE0300058, + 0x00000204, 0x00030303, + 0x00000200, 0xE0300054, + 0x00000204, 0x700010FF, + 0x00000200, 0xE0300074, + 0x00000204, 0x00030303, + 0x00000200, 0xE0300070, + 0x00000204, 0x700010FF, + 0x00000200, 0xE0300090, + 0x00000204, 0x00030303, + 0x00000200, 0xE030008C, + 0x00000204, 0x700010FF, + 0x00000200, 0xE03000AC, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000A8, + 0x00000204, 0x700010FF, + 0x00000200, 0xE03000C8, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000C4, + 0x00000204, 0x700010FF, + 0x00000200, 0xE03000E4, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000E0, + 0x00000204, 0x700010FF, + 0x00000200, 0xE0300100, + 0x00000204, 0x00030303, + 0x00000200, 0xE03000FC, + 0x00000204, 0x700010FF, + 0x00000200, 0xE0300058, + 0x00000204, 0x00010303, + 0x00000200, 0xE0300054, + 0x00000204, 0x800010FF, + 0x00000200, 0xE0300074, + 0x00000204, 0x00010303, + 0x00000200, 0xE0300070, + 0x00000204, 0x800010FF, + 0x00000200, 0xE0300090, + 0x00000204, 0x00010303, + 0x00000200, 0xE030008C, + 0x00000204, 0x800010FF, + 0x00000200, 0xE03000AC, + 0x00000204, 0x00010303, + 0x00000200, 0xE03000A8, + 0x00000204, 0x800010FF, + 0x00000200, 0xE03000C4, + 0x00000204, 0x800010FF, + 0x00000200, 0xE03000C8, + 0x00000204, 0x00010303, + 0x00000200, 0xE03000E4, + 0x00000204, 0x00010303, + 0x00000200, 0xE03000E0, + 0x00000204, 0x800010FF, + 0x00000200, 0xE0300100, + 0x00000204, 0x00010303, + 0x00000200, 0xE03000FC, + 0x00000204, 0x800010FF, + 0x00000200, 0x0001f198, + 0x00000204, 0x0003ffff, + 0x00000200, 0x0001f19C, + 0x00000204, 0x3fffffff, + 0x00000200, 0xE030032C, + 0x00000204, 0x00000000, +}; + +static void trinity_program_clk_gating_hw_sequence(struct radeon_device *rdev, + const u32 *seq, u32 count); +static void trinity_override_dynamic_mg_powergating(struct radeon_device *rdev); +static void trinity_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps); +struct trinity_ps *trinity_get_ps(struct radeon_ps *rps); +struct trinity_power_info *trinity_get_pi(struct radeon_device *rdev); +void trinity_dpm_reset_asic(struct radeon_device *rdev); + +struct trinity_ps *trinity_get_ps(struct radeon_ps *rps) +{ + struct trinity_ps *ps = rps->ps_priv; + + return ps; +} + +struct trinity_power_info *trinity_get_pi(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = rdev->pm.dpm.priv; + + return pi; +} + +static void trinity_gfx_powergating_initialize(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 p, u; + u32 value; + struct atom_clock_dividers dividers; + u32 xclk = radeon_get_xclk(rdev); + u32 sssd = 1; + int ret; + u32 hw_rev = (RREG32(HW_REV) & ATI_REV_ID_MASK) >> ATI_REV_ID_SHIFT; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + 25000, false, ÷rs); + if (ret) + return; + + value = RREG32_SMC(GFX_POWER_GATING_CNTL); + value &= ~(SSSD_MASK | PDS_DIV_MASK); + if (sssd) + value |= SSSD(1); + value |= PDS_DIV(dividers.post_div); + WREG32_SMC(GFX_POWER_GATING_CNTL, value); + + r600_calculate_u_and_p(500, xclk, 16, &p, &u); + + WREG32(CG_PG_CTRL, SP(p) | SU(u)); + + WREG32_P(CG_GIPOTS, CG_GIPOT(p), ~CG_GIPOT_MASK); + + /* XXX double check hw_rev */ + if (pi->override_dynamic_mgpg && (hw_rev == 0)) + trinity_override_dynamic_mg_powergating(rdev); + +} + +#define CGCG_CGTT_LOCAL0_MASK 0xFFFF33FF +#define CGCG_CGTT_LOCAL1_MASK 0xFFFB0FFE +#define CGTS_SM_CTRL_REG_DISABLE 0x00600000 +#define CGTS_SM_CTRL_REG_ENABLE 0x96944200 + +static void trinity_mg_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 local0; + u32 local1; + + if (enable) { + local0 = RREG32_CG(CG_CGTT_LOCAL_0); + local1 = RREG32_CG(CG_CGTT_LOCAL_1); + + WREG32_CG(CG_CGTT_LOCAL_0, + (0x00380000 & CGCG_CGTT_LOCAL0_MASK) | (local0 & ~CGCG_CGTT_LOCAL0_MASK) ); + WREG32_CG(CG_CGTT_LOCAL_1, + (0x0E000000 & CGCG_CGTT_LOCAL1_MASK) | (local1 & ~CGCG_CGTT_LOCAL1_MASK) ); + + WREG32(CGTS_SM_CTRL_REG, CGTS_SM_CTRL_REG_ENABLE); + } else { + WREG32(CGTS_SM_CTRL_REG, CGTS_SM_CTRL_REG_DISABLE); + + local0 = RREG32_CG(CG_CGTT_LOCAL_0); + local1 = RREG32_CG(CG_CGTT_LOCAL_1); + + WREG32_CG(CG_CGTT_LOCAL_0, + CGCG_CGTT_LOCAL0_MASK | (local0 & ~CGCG_CGTT_LOCAL0_MASK) ); + WREG32_CG(CG_CGTT_LOCAL_1, + CGCG_CGTT_LOCAL1_MASK | (local1 & ~CGCG_CGTT_LOCAL1_MASK) ); + } +} + +static void trinity_mg_clockgating_initialize(struct radeon_device *rdev) +{ + u32 count; + const u32 *seq = NULL; + + seq = &trinity_mgcg_shls_default[0]; + count = sizeof(trinity_mgcg_shls_default) / (3 * sizeof(u32)); + + trinity_program_clk_gating_hw_sequence(rdev, seq, count); +} + +static void trinity_gfx_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + if (enable) { + WREG32_P(SCLK_PWRMGT_CNTL, DYN_GFX_CLK_OFF_EN, ~DYN_GFX_CLK_OFF_EN); + } else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_GFX_CLK_OFF_EN); + WREG32_P(SCLK_PWRMGT_CNTL, GFX_CLK_FORCE_ON, ~GFX_CLK_FORCE_ON); + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~GFX_CLK_FORCE_ON); + RREG32(GB_ADDR_CONFIG); + } +} + +static void trinity_program_clk_gating_hw_sequence(struct radeon_device *rdev, + const u32 *seq, u32 count) +{ + u32 i, length = count * 3; + + for (i = 0; i < length; i += 3) + WREG32_P(seq[i], seq[i+1], ~seq[i+2]); +} + +static void trinity_program_override_mgpg_sequences(struct radeon_device *rdev, + const u32 *seq, u32 count) +{ + u32 i, length = count * 2; + + for (i = 0; i < length; i += 2) + WREG32(seq[i], seq[i+1]); + +} + +static void trinity_override_dynamic_mg_powergating(struct radeon_device *rdev) +{ + u32 count; + const u32 *seq = NULL; + + seq = &trinity_override_mgpg_sequences[0]; + count = sizeof(trinity_override_mgpg_sequences) / (2 * sizeof(u32)); + + trinity_program_override_mgpg_sequences(rdev, seq, count); +} + +static void trinity_ls_clockgating_enable(struct radeon_device *rdev, + bool enable) +{ + u32 count; + const u32 *seq = NULL; + + if (enable) { + seq = &trinity_sysls_enable[0]; + count = sizeof(trinity_sysls_enable) / (3 * sizeof(u32)); + } else { + seq = &trinity_sysls_disable[0]; + count = sizeof(trinity_sysls_disable) / (3 * sizeof(u32)); + } + + trinity_program_clk_gating_hw_sequence(rdev, seq, count); +} + +static void trinity_gfx_powergating_enable(struct radeon_device *rdev, + bool enable) +{ + if (enable) { + if (RREG32_SMC(CC_SMU_TST_EFUSE1_MISC) & RB_BACKEND_DISABLE_MASK) + WREG32_SMC(SMU_SCRATCH_A, (RREG32_SMC(SMU_SCRATCH_A) | 0x01)); + + WREG32_P(SCLK_PWRMGT_CNTL, DYN_PWR_DOWN_EN, ~DYN_PWR_DOWN_EN); + } else { + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~DYN_PWR_DOWN_EN); + RREG32(GB_ADDR_CONFIG); + } +} + +static void trinity_gfx_dynamic_mgpg_enable(struct radeon_device *rdev, + bool enable) +{ + u32 value; + + if (enable) { + value = RREG32_SMC(PM_I_CNTL_1); + value &= ~DS_PG_CNTL_MASK; + value |= DS_PG_CNTL(1); + WREG32_SMC(PM_I_CNTL_1, value); + + value = RREG32_SMC(SMU_S_PG_CNTL); + value &= ~DS_PG_EN_MASK; + value |= DS_PG_EN(1); + WREG32_SMC(SMU_S_PG_CNTL, value); + } else { + value = RREG32_SMC(SMU_S_PG_CNTL); + value &= ~DS_PG_EN_MASK; + WREG32_SMC(SMU_S_PG_CNTL, value); + + value = RREG32_SMC(PM_I_CNTL_1); + value &= ~DS_PG_CNTL_MASK; + WREG32_SMC(PM_I_CNTL_1, value); + } + + trinity_gfx_dynamic_mgpg_config(rdev); + +} + +static void trinity_enable_clock_power_gating(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (pi->enable_gfx_clock_gating) + sumo_gfx_clockgating_initialize(rdev); + if (pi->enable_mg_clock_gating) + trinity_mg_clockgating_initialize(rdev); + if (pi->enable_gfx_power_gating) + trinity_gfx_powergating_initialize(rdev); + if (pi->enable_mg_clock_gating) { + trinity_ls_clockgating_enable(rdev, true); + trinity_mg_clockgating_enable(rdev, true); + } + if (pi->enable_gfx_clock_gating) + trinity_gfx_clockgating_enable(rdev, true); + if (pi->enable_gfx_dynamic_mgpg) + trinity_gfx_dynamic_mgpg_enable(rdev, true); + if (pi->enable_gfx_power_gating) + trinity_gfx_powergating_enable(rdev, true); +} + +static void trinity_disable_clock_power_gating(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (pi->enable_gfx_power_gating) + trinity_gfx_powergating_enable(rdev, false); + if (pi->enable_gfx_dynamic_mgpg) + trinity_gfx_dynamic_mgpg_enable(rdev, false); + if (pi->enable_gfx_clock_gating) + trinity_gfx_clockgating_enable(rdev, false); + if (pi->enable_mg_clock_gating) { + trinity_mg_clockgating_enable(rdev, false); + trinity_ls_clockgating_enable(rdev, false); + } +} + +static void trinity_set_divider_value(struct radeon_device *rdev, + u32 index, u32 sclk) +{ + struct atom_clock_dividers dividers; + int ret; + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + sclk, false, ÷rs); + if (ret) + return; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix); + value &= ~CLK_DIVIDER_MASK; + value |= CLK_DIVIDER(dividers.post_div); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix, value); + + ret = radeon_atom_get_clock_dividers(rdev, COMPUTE_ENGINE_PLL_PARAM, + sclk/2, false, ÷rs); + if (ret) + return; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_PG_CNTL + ix); + value &= ~PD_SCLK_DIVIDER_MASK; + value |= PD_SCLK_DIVIDER(dividers.post_div); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_PG_CNTL + ix, value); +} + +static void trinity_set_ds_dividers(struct radeon_device *rdev, + u32 index, u32 divider) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix); + value &= ~DS_DIV_MASK; + value |= DS_DIV(divider); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix, value); +} + +static void trinity_set_ss_dividers(struct radeon_device *rdev, + u32 index, u32 divider) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix); + value &= ~DS_SH_DIV_MASK; + value |= DS_SH_DIV(divider); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix, value); +} + +static void trinity_set_vid(struct radeon_device *rdev, u32 index, u32 vid) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 vid_7bit = sumo_convert_vid2_to_vid7(rdev, &pi->sys_info.vid_mapping_table, vid); + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix); + value &= ~VID_MASK; + value |= VID(vid_7bit); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix, value); + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix); + value &= ~LVRT_MASK; + value |= LVRT(0); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix, value); +} + +static void trinity_set_allos_gnb_slow(struct radeon_device *rdev, + u32 index, u32 gnb_slow) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_3 + ix); + value &= ~GNB_SLOW_MASK; + value |= GNB_SLOW(gnb_slow); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_3 + ix, value); +} + +static void trinity_set_force_nbp_state(struct radeon_device *rdev, + u32 index, u32 force_nbp_state) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_3 + ix); + value &= ~FORCE_NBPS1_MASK; + value |= FORCE_NBPS1(force_nbp_state); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_3 + ix, value); +} + +static void trinity_set_display_wm(struct radeon_device *rdev, + u32 index, u32 wm) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix); + value &= ~DISPLAY_WM_MASK; + value |= DISPLAY_WM(wm); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix, value); +} + +static void trinity_set_vce_wm(struct radeon_device *rdev, + u32 index, u32 wm) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix); + value &= ~VCE_WM_MASK; + value |= VCE_WM(wm); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_1 + ix, value); +} + +static void trinity_set_at(struct radeon_device *rdev, + u32 index, u32 at) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_AT + ix); + value &= ~AT_MASK; + value |= AT(at); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_AT + ix, value); +} + +static void trinity_program_power_level(struct radeon_device *rdev, + struct trinity_pl *pl, u32 index) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (index >= SUMO_MAX_HARDWARE_POWERLEVELS) + return; + + trinity_set_divider_value(rdev, index, pl->sclk); + trinity_set_vid(rdev, index, pl->vddc_index); + trinity_set_ss_dividers(rdev, index, pl->ss_divider_index); + trinity_set_ds_dividers(rdev, index, pl->ds_divider_index); + trinity_set_allos_gnb_slow(rdev, index, pl->allow_gnb_slow); + trinity_set_force_nbp_state(rdev, index, pl->force_nbp_state); + trinity_set_display_wm(rdev, index, pl->display_wm); + trinity_set_vce_wm(rdev, index, pl->vce_wm); + trinity_set_at(rdev, index, pi->at[index]); +} + +static void trinity_power_level_enable_disable(struct radeon_device *rdev, + u32 index, bool enable) +{ + u32 value; + u32 ix = index * TRINITY_SIZEOF_DPM_STATE_TABLE; + + value = RREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix); + value &= ~STATE_VALID_MASK; + if (enable) + value |= STATE_VALID(1); + WREG32_SMC(SMU_SCLK_DPM_STATE_0_CNTL_0 + ix, value); +} + +static bool trinity_dpm_enabled(struct radeon_device *rdev) +{ + if (RREG32_SMC(SMU_SCLK_DPM_CNTL) & SCLK_DPM_EN(1)) + return true; + else + return false; +} + +static void trinity_start_dpm(struct radeon_device *rdev) +{ + u32 value = RREG32_SMC(SMU_SCLK_DPM_CNTL); + + value &= ~(SCLK_DPM_EN_MASK | SCLK_DPM_BOOT_STATE_MASK | VOLTAGE_CHG_EN_MASK); + value |= SCLK_DPM_EN(1) | SCLK_DPM_BOOT_STATE(0) | VOLTAGE_CHG_EN(1); + WREG32_SMC(SMU_SCLK_DPM_CNTL, value); + + WREG32_P(GENERAL_PWRMGT, GLOBAL_PWRMGT_EN, ~GLOBAL_PWRMGT_EN); + WREG32_P(CG_CG_VOLTAGE_CNTL, 0, ~EN); + + trinity_dpm_config(rdev, true); +} + +static void trinity_wait_for_dpm_enabled(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(SCLK_PWRMGT_CNTL) & DYNAMIC_PM_EN) + break; + DRM_UDELAY(1); + } + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & TARGET_STATE_MASK) == 0) + break; + DRM_UDELAY(1); + } + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_STATE_MASK) == 0) + break; + DRM_UDELAY(1); + } +} + +static void trinity_stop_dpm(struct radeon_device *rdev) +{ + u32 sclk_dpm_cntl; + + WREG32_P(CG_CG_VOLTAGE_CNTL, EN, ~EN); + + sclk_dpm_cntl = RREG32_SMC(SMU_SCLK_DPM_CNTL); + sclk_dpm_cntl &= ~(SCLK_DPM_EN_MASK | VOLTAGE_CHG_EN_MASK); + WREG32_SMC(SMU_SCLK_DPM_CNTL, sclk_dpm_cntl); + + trinity_dpm_config(rdev, false); +} + +static void trinity_start_am(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, 0, ~(RESET_SCLK_CNT | RESET_BUSY_CNT)); +} + +static void trinity_reset_am(struct radeon_device *rdev) +{ + WREG32_P(SCLK_PWRMGT_CNTL, RESET_SCLK_CNT | RESET_BUSY_CNT, + ~(RESET_SCLK_CNT | RESET_BUSY_CNT)); +} + +static void trinity_wait_for_level_0(struct radeon_device *rdev) +{ + int i; + + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_STATE_MASK) == 0) + break; + DRM_UDELAY(1); + } +} + +static void trinity_enable_power_level_0(struct radeon_device *rdev) +{ + trinity_power_level_enable_disable(rdev, 0, true); +} + +static void trinity_force_level_0(struct radeon_device *rdev) +{ + trinity_dpm_force_state(rdev, 0); +} + +static void trinity_unforce_levels(struct radeon_device *rdev) +{ + trinity_dpm_no_forced_level(rdev); +} + +static void trinity_program_power_levels_0_to_n(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct trinity_ps *new_ps = trinity_get_ps(new_rps); + struct trinity_ps *old_ps = trinity_get_ps(old_rps); + u32 i; + u32 n_current_state_levels = (old_ps == NULL) ? 1 : old_ps->num_levels; + + for (i = 0; i < new_ps->num_levels; i++) { + trinity_program_power_level(rdev, &new_ps->levels[i], i); + trinity_power_level_enable_disable(rdev, i, true); + } + + for (i = new_ps->num_levels; i < n_current_state_levels; i++) + trinity_power_level_enable_disable(rdev, i, false); +} + +static void trinity_program_bootup_state(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 i; + + trinity_program_power_level(rdev, &pi->boot_pl, 0); + trinity_power_level_enable_disable(rdev, 0, true); + + for (i = 1; i < 8; i++) + trinity_power_level_enable_disable(rdev, i, false); +} + +static void trinity_setup_uvd_clock_table(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_ps *ps = trinity_get_ps(rps); + u32 uvdstates = (ps->vclk_low_divider | + ps->vclk_high_divider << 8 | + ps->dclk_low_divider << 16 | + ps->dclk_high_divider << 24); + + WREG32_SMC(SMU_UVD_DPM_STATES, uvdstates); +} + +static void trinity_setup_uvd_dpm_interval(struct radeon_device *rdev, + u32 interval) +{ + u32 p, u; + u32 tp = RREG32_SMC(PM_TP); + u32 val; + u32 xclk = radeon_get_xclk(rdev); + + r600_calculate_u_and_p(interval, xclk, 16, &p, &u); + + val = (p + tp - 1) / tp; + + WREG32_SMC(SMU_UVD_DPM_CNTL, val); +} + +static bool trinity_uvd_clocks_zero(struct radeon_ps *rps) +{ + if ((rps->vclk == 0) && (rps->dclk == 0)) + return true; + else + return false; +} + +static bool trinity_uvd_clocks_equal(struct radeon_ps *rps1, + struct radeon_ps *rps2) +{ + struct trinity_ps *ps1 = trinity_get_ps(rps1); + struct trinity_ps *ps2 = trinity_get_ps(rps2); + + if ((rps1->vclk == rps2->vclk) && + (rps1->dclk == rps2->dclk) && + (ps1->vclk_low_divider == ps2->vclk_low_divider) && + (ps1->vclk_high_divider == ps2->vclk_high_divider) && + (ps1->dclk_low_divider == ps2->dclk_low_divider) && + (ps1->dclk_high_divider == ps2->dclk_high_divider)) + return true; + else + return false; +} + +static void trinity_setup_uvd_clocks(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (pi->enable_gfx_power_gating) { + trinity_gfx_powergating_enable(rdev, false); + } + + if (pi->uvd_dpm) { + if (trinity_uvd_clocks_zero(new_rps) && + !trinity_uvd_clocks_zero(old_rps)) { + trinity_setup_uvd_dpm_interval(rdev, 0); + } else if (!trinity_uvd_clocks_zero(new_rps)) { + trinity_setup_uvd_clock_table(rdev, new_rps); + + if (trinity_uvd_clocks_zero(old_rps)) { + u32 tmp = RREG32(CG_MISC_REG); + tmp &= 0xfffffffd; + WREG32(CG_MISC_REG, tmp); + + radeon_set_uvd_clocks(rdev, new_rps->vclk, new_rps->dclk); + + trinity_setup_uvd_dpm_interval(rdev, 3000); + } + } + trinity_uvd_dpm_config(rdev); + } else { + if (trinity_uvd_clocks_zero(new_rps) || + trinity_uvd_clocks_equal(new_rps, old_rps)) + return; + + radeon_set_uvd_clocks(rdev, new_rps->vclk, new_rps->dclk); + } + + if (pi->enable_gfx_power_gating) { + trinity_gfx_powergating_enable(rdev, true); + } +} + +static void trinity_set_uvd_clock_before_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct trinity_ps *new_ps = trinity_get_ps(new_rps); + struct trinity_ps *current_ps = trinity_get_ps(new_rps); + + if (new_ps->levels[new_ps->num_levels - 1].sclk >= + current_ps->levels[current_ps->num_levels - 1].sclk) + return; + + trinity_setup_uvd_clocks(rdev, new_rps, old_rps); +} + +static void trinity_set_uvd_clock_after_set_eng_clock(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct trinity_ps *new_ps = trinity_get_ps(new_rps); + struct trinity_ps *current_ps = trinity_get_ps(old_rps); + + if (new_ps->levels[new_ps->num_levels - 1].sclk < + current_ps->levels[current_ps->num_levels - 1].sclk) + return; + + trinity_setup_uvd_clocks(rdev, new_rps, old_rps); +} + +static void trinity_program_ttt(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 value = RREG32_SMC(SMU_SCLK_DPM_TTT); + + value &= ~(HT_MASK | LT_MASK); + value |= HT((pi->thermal_auto_throttling + 49) * 8); + value |= LT((pi->thermal_auto_throttling + 49 - pi->sys_info.htc_hyst_lmt) * 8); + WREG32_SMC(SMU_SCLK_DPM_TTT, value); +} + +static void trinity_enable_att(struct radeon_device *rdev) +{ + u32 value = RREG32_SMC(SMU_SCLK_DPM_TT_CNTL); + + value &= ~SCLK_TT_EN_MASK; + value |= SCLK_TT_EN(1); + WREG32_SMC(SMU_SCLK_DPM_TT_CNTL, value); +} + +static void trinity_program_sclk_dpm(struct radeon_device *rdev) +{ + u32 p, u; + u32 tp = RREG32_SMC(PM_TP); + u32 ni; + u32 xclk = radeon_get_xclk(rdev); + u32 value; + + r600_calculate_u_and_p(400, xclk, 16, &p, &u); + + ni = (p + tp - 1) / tp; + + value = RREG32_SMC(PM_I_CNTL_1); + value &= ~SCLK_DPM_MASK; + value |= SCLK_DPM(ni); + WREG32_SMC(PM_I_CNTL_1, value); +} + +static int trinity_set_thermal_temperature_range(struct radeon_device *rdev, + int min_temp, int max_temp) +{ + int low_temp = 0 * 1000; + int high_temp = 255 * 1000; + + if (low_temp < min_temp) + low_temp = min_temp; + if (high_temp > max_temp) + high_temp = max_temp; + if (high_temp < low_temp) { + DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp); + return -EINVAL; + } + + WREG32_P(CG_THERMAL_INT_CTRL, DIG_THERM_INTH(49 + (high_temp / 1000)), ~DIG_THERM_INTH_MASK); + WREG32_P(CG_THERMAL_INT_CTRL, DIG_THERM_INTL(49 + (low_temp / 1000)), ~DIG_THERM_INTL_MASK); + + rdev->pm.dpm.thermal.min_temp = low_temp; + rdev->pm.dpm.thermal.max_temp = high_temp; + + return 0; +} + +static void trinity_update_current_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_ps *new_ps = trinity_get_ps(rps); + struct trinity_power_info *pi = trinity_get_pi(rdev); + + pi->current_rps = *rps; + pi->current_ps = *new_ps; + pi->current_rps.ps_priv = &pi->current_ps; +} + +static void trinity_update_requested_ps(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_ps *new_ps = trinity_get_ps(rps); + struct trinity_power_info *pi = trinity_get_pi(rdev); + + pi->requested_rps = *rps; + pi->requested_ps = *new_ps; + pi->requested_rps.ps_priv = &pi->requested_ps; +} + +int trinity_dpm_enable(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + int ret; + + trinity_acquire_mutex(rdev); + + if (trinity_dpm_enabled(rdev)) { + trinity_release_mutex(rdev); + return -EINVAL; + } + + trinity_enable_clock_power_gating(rdev); + trinity_program_bootup_state(rdev); + sumo_program_vc(rdev, 0x00C00033); + trinity_start_am(rdev); + if (pi->enable_auto_thermal_throttling) { + trinity_program_ttt(rdev); + trinity_enable_att(rdev); + } + trinity_program_sclk_dpm(rdev); + trinity_start_dpm(rdev); + trinity_wait_for_dpm_enabled(rdev); + trinity_release_mutex(rdev); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + ret = trinity_set_thermal_temperature_range(rdev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); + if (ret) { + trinity_release_mutex(rdev); + return ret; + } + rdev->irq.dpm_thermal = true; + radeon_irq_set(rdev); + } + + trinity_update_current_ps(rdev, rdev->pm.dpm.boot_ps); + + return 0; +} + +void trinity_dpm_disable(struct radeon_device *rdev) +{ + trinity_acquire_mutex(rdev); + if (!trinity_dpm_enabled(rdev)) { + trinity_release_mutex(rdev); + return; + } + trinity_disable_clock_power_gating(rdev); + sumo_clear_vc(rdev); + trinity_wait_for_level_0(rdev); + trinity_stop_dpm(rdev); + trinity_reset_am(rdev); + trinity_release_mutex(rdev); + + if (rdev->irq.installed && + r600_is_internal_thermal_sensor(rdev->pm.int_thermal_type)) { + rdev->irq.dpm_thermal = false; + radeon_irq_set(rdev); + } + + trinity_update_current_ps(rdev, rdev->pm.dpm.boot_ps); +} + +static void trinity_get_min_sclk_divider(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + pi->min_sclk_did = + (RREG32_SMC(CC_SMU_MISC_FUSES) & MinSClkDid_MASK) >> MinSClkDid_SHIFT; +} + +static void trinity_setup_nbp_sim(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct trinity_ps *new_ps = trinity_get_ps(rps); + u32 nbpsconfig; + + if (pi->sys_info.nb_dpm_enable) { + nbpsconfig = RREG32_SMC(NB_PSTATE_CONFIG); + nbpsconfig &= ~(Dpm0PgNbPsLo_MASK | Dpm0PgNbPsHi_MASK | DpmXNbPsLo_MASK | DpmXNbPsHi_MASK); + nbpsconfig |= (Dpm0PgNbPsLo(new_ps->Dpm0PgNbPsLo) | + Dpm0PgNbPsHi(new_ps->Dpm0PgNbPsHi) | + DpmXNbPsLo(new_ps->DpmXNbPsLo) | + DpmXNbPsHi(new_ps->DpmXNbPsHi)); + WREG32_SMC(NB_PSTATE_CONFIG, nbpsconfig); + } +} + +int trinity_dpm_force_performance_level(struct radeon_device *rdev, + enum radeon_dpm_forced_level level) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct radeon_ps *rps = &pi->current_rps; + struct trinity_ps *ps = trinity_get_ps(rps); + int i, ret; + + if (ps->num_levels <= 1) + return 0; + + if (level == RADEON_DPM_FORCED_LEVEL_HIGH) { + /* not supported by the hw */ + return -EINVAL; + } else if (level == RADEON_DPM_FORCED_LEVEL_LOW) { + ret = trinity_dpm_n_levels_disabled(rdev, ps->num_levels - 1); + if (ret) + return ret; + } else { + for (i = 0; i < ps->num_levels; i++) { + ret = trinity_dpm_n_levels_disabled(rdev, 0); + if (ret) + return ret; + } + } + + rdev->pm.dpm.forced_level = level; + + return 0; +} + +int trinity_dpm_pre_set_power_state(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct radeon_ps requested_ps = *rdev->pm.dpm.requested_ps; + struct radeon_ps *new_ps = &requested_ps; + + trinity_update_requested_ps(rdev, new_ps); + + trinity_apply_state_adjust_rules(rdev, + &pi->requested_rps, + &pi->current_rps); + + return 0; +} + +int trinity_dpm_set_power_state(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct radeon_ps *new_ps = &pi->requested_rps; + struct radeon_ps *old_ps = &pi->current_rps; + + trinity_acquire_mutex(rdev); + if (pi->enable_dpm) { + trinity_set_uvd_clock_before_set_eng_clock(rdev, new_ps, old_ps); + trinity_enable_power_level_0(rdev); + trinity_force_level_0(rdev); + trinity_wait_for_level_0(rdev); + trinity_setup_nbp_sim(rdev, new_ps); + trinity_program_power_levels_0_to_n(rdev, new_ps, old_ps); + trinity_force_level_0(rdev); + trinity_unforce_levels(rdev); + trinity_set_uvd_clock_after_set_eng_clock(rdev, new_ps, old_ps); + rdev->pm.dpm.forced_level = RADEON_DPM_FORCED_LEVEL_AUTO; + } + trinity_release_mutex(rdev); + + return 0; +} + +void trinity_dpm_post_set_power_state(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct radeon_ps *new_ps = &pi->requested_rps; + + trinity_update_current_ps(rdev, new_ps); +} + +void trinity_dpm_setup_asic(struct radeon_device *rdev) +{ + trinity_acquire_mutex(rdev); + sumo_program_sstp(rdev); + sumo_take_smu_control(rdev, true); + trinity_get_min_sclk_divider(rdev); + trinity_release_mutex(rdev); +} + +void trinity_dpm_reset_asic(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + trinity_acquire_mutex(rdev); + if (pi->enable_dpm) { + trinity_enable_power_level_0(rdev); + trinity_force_level_0(rdev); + trinity_wait_for_level_0(rdev); + trinity_program_bootup_state(rdev); + trinity_force_level_0(rdev); + trinity_unforce_levels(rdev); + } + trinity_release_mutex(rdev); +} + +static u16 trinity_convert_voltage_index_to_value(struct radeon_device *rdev, + u32 vid_2bit) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 vid_7bit = sumo_convert_vid2_to_vid7(rdev, &pi->sys_info.vid_mapping_table, vid_2bit); + u32 svi_mode = (RREG32_SMC(PM_CONFIG) & SVI_Mode) ? 1 : 0; + u32 step = (svi_mode == 0) ? 1250 : 625; + u32 delta = vid_7bit * step + 50; + + if (delta > 155000) + return 0; + + return (155000 - delta) / 100; +} + +static void trinity_patch_boot_state(struct radeon_device *rdev, + struct trinity_ps *ps) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + ps->num_levels = 1; + ps->nbps_flags = 0; + ps->bapm_flags = 0; + ps->levels[0] = pi->boot_pl; +} + +static u8 trinity_calculate_vce_wm(struct radeon_device *rdev, u32 sclk) +{ + if (sclk < 20000) + return 1; + return 0; +} + +static void trinity_construct_boot_state(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + pi->boot_pl.sclk = pi->sys_info.bootup_sclk; + pi->boot_pl.vddc_index = pi->sys_info.bootup_nb_voltage_index; + pi->boot_pl.ds_divider_index = 0; + pi->boot_pl.ss_divider_index = 0; + pi->boot_pl.allow_gnb_slow = 1; + pi->boot_pl.force_nbp_state = 0; + pi->boot_pl.display_wm = 0; + pi->boot_pl.vce_wm = 0; + pi->current_ps.num_levels = 1; + pi->current_ps.levels[0] = pi->boot_pl; +} + +static u8 trinity_get_sleep_divider_id_from_clock(struct radeon_device *rdev, + u32 sclk, u32 min_sclk_in_sr) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 i; + u32 temp; + u32 min = (min_sclk_in_sr > TRINITY_MINIMUM_ENGINE_CLOCK) ? + min_sclk_in_sr : TRINITY_MINIMUM_ENGINE_CLOCK; + + if (sclk < min) + return 0; + + if (!pi->enable_sclk_ds) + return 0; + + for (i = TRINITY_MAX_DEEPSLEEP_DIVIDER_ID; ; i--) { + temp = sclk / sumo_get_sleep_divider_from_id(i); + if (temp >= min || i == 0) + break; + } + + return (u8)i; +} + +static u32 trinity_get_valid_engine_clock(struct radeon_device *rdev, + u32 lower_limit) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 i; + + for (i = 0; i < pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries; i++) { + if (pi->sys_info.sclk_voltage_mapping_table.entries[i].sclk_frequency >= lower_limit) + return pi->sys_info.sclk_voltage_mapping_table.entries[i].sclk_frequency; + } + + if (i == pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries) + DRM_ERROR("engine clock out of range!"); + + return 0; +} + +static void trinity_patch_thermal_state(struct radeon_device *rdev, + struct trinity_ps *ps, + struct trinity_ps *current_ps) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 sclk_in_sr = pi->sys_info.min_sclk; /* ??? */ + u32 current_vddc; + u32 current_sclk; + u32 current_index = 0; + + if (current_ps) { + current_vddc = current_ps->levels[current_index].vddc_index; + current_sclk = current_ps->levels[current_index].sclk; + } else { + current_vddc = pi->boot_pl.vddc_index; + current_sclk = pi->boot_pl.sclk; + } + + ps->levels[0].vddc_index = current_vddc; + + if (ps->levels[0].sclk > current_sclk) + ps->levels[0].sclk = current_sclk; + + ps->levels[0].ds_divider_index = + trinity_get_sleep_divider_id_from_clock(rdev, ps->levels[0].sclk, sclk_in_sr); + ps->levels[0].ss_divider_index = ps->levels[0].ds_divider_index; + ps->levels[0].allow_gnb_slow = 1; + ps->levels[0].force_nbp_state = 0; + ps->levels[0].display_wm = 0; + ps->levels[0].vce_wm = + trinity_calculate_vce_wm(rdev, ps->levels[0].sclk); +} + +static u8 trinity_calculate_display_wm(struct radeon_device *rdev, + struct trinity_ps *ps, u32 index) +{ + if (ps == NULL || ps->num_levels <= 1) + return 0; + else if (ps->num_levels == 2) { + if (index == 0) + return 0; + else + return 1; + } else { + if (index == 0) + return 0; + else if (ps->levels[index].sclk < 30000) + return 0; + else + return 1; + } +} + +static u32 trinity_get_uvd_clock_index(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 i = 0; + + for (i = 0; i < 4; i++) { + if ((rps->vclk == pi->sys_info.uvd_clock_table_entries[i].vclk) && + (rps->dclk == pi->sys_info.uvd_clock_table_entries[i].dclk)) + break; + } + + if (i >= 4) { + DRM_ERROR("UVD clock index not found!\n"); + i = 3; + } + return i; +} + +static void trinity_adjust_uvd_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + struct trinity_ps *ps = trinity_get_ps(rps); + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 high_index = 0; + u32 low_index = 0; + + if (pi->uvd_dpm && r600_is_uvd_state(rps->class, rps->class2)) { + high_index = trinity_get_uvd_clock_index(rdev, rps); + + switch(high_index) { + case 3: + case 2: + low_index = 1; + break; + case 1: + case 0: + default: + low_index = 0; + break; + } + + ps->vclk_low_divider = + pi->sys_info.uvd_clock_table_entries[high_index].vclk_did; + ps->dclk_low_divider = + pi->sys_info.uvd_clock_table_entries[high_index].dclk_did; + ps->vclk_high_divider = + pi->sys_info.uvd_clock_table_entries[low_index].vclk_did; + ps->dclk_high_divider = + pi->sys_info.uvd_clock_table_entries[low_index].dclk_did; + } +} + + + +static void trinity_apply_state_adjust_rules(struct radeon_device *rdev, + struct radeon_ps *new_rps, + struct radeon_ps *old_rps) +{ + struct trinity_ps *ps = trinity_get_ps(new_rps); + struct trinity_ps *current_ps = trinity_get_ps(old_rps); + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 min_voltage = 0; /* ??? */ + u32 min_sclk = pi->sys_info.min_sclk; /* XXX check against disp reqs */ + u32 sclk_in_sr = pi->sys_info.min_sclk; /* ??? */ + u32 i; + bool force_high; + u32 num_active_displays = rdev->pm.dpm.new_active_crtc_count; + + if (new_rps->class & ATOM_PPLIB_CLASSIFICATION_THERMAL) + return trinity_patch_thermal_state(rdev, ps, current_ps); + + trinity_adjust_uvd_state(rdev, new_rps); + + for (i = 0; i < ps->num_levels; i++) { + if (ps->levels[i].vddc_index < min_voltage) + ps->levels[i].vddc_index = min_voltage; + + if (ps->levels[i].sclk < min_sclk) + ps->levels[i].sclk = + trinity_get_valid_engine_clock(rdev, min_sclk); + + ps->levels[i].ds_divider_index = + sumo_get_sleep_divider_id_from_clock(rdev, ps->levels[i].sclk, sclk_in_sr); + + ps->levels[i].ss_divider_index = ps->levels[i].ds_divider_index; + + ps->levels[i].allow_gnb_slow = 1; + ps->levels[i].force_nbp_state = 0; + ps->levels[i].display_wm = + trinity_calculate_display_wm(rdev, ps, i); + ps->levels[i].vce_wm = + trinity_calculate_vce_wm(rdev, ps->levels[0].sclk); + } + + if ((new_rps->class & (ATOM_PPLIB_CLASSIFICATION_HDSTATE | ATOM_PPLIB_CLASSIFICATION_SDSTATE)) || + ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == ATOM_PPLIB_CLASSIFICATION_UI_BATTERY)) + ps->bapm_flags |= TRINITY_POWERSTATE_FLAGS_BAPM_DISABLE; + + if (pi->sys_info.nb_dpm_enable) { + ps->Dpm0PgNbPsLo = 0x1; + ps->Dpm0PgNbPsHi = 0x0; + ps->DpmXNbPsLo = 0x2; + ps->DpmXNbPsHi = 0x1; + + if ((new_rps->class & (ATOM_PPLIB_CLASSIFICATION_HDSTATE | ATOM_PPLIB_CLASSIFICATION_SDSTATE)) || + ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) == ATOM_PPLIB_CLASSIFICATION_UI_BATTERY)) { + force_high = ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_HDSTATE) || + ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_SDSTATE) && + (pi->sys_info.uma_channel_number == 1))); + force_high = (num_active_displays >= 3) || force_high; + ps->Dpm0PgNbPsLo = force_high ? 0x2 : 0x3; + ps->Dpm0PgNbPsHi = 0x1; + ps->DpmXNbPsLo = force_high ? 0x2 : 0x3; + ps->DpmXNbPsHi = 0x2; + ps->levels[ps->num_levels - 1].allow_gnb_slow = 0; + } + } +} + +static void trinity_cleanup_asic(struct radeon_device *rdev) +{ + sumo_take_smu_control(rdev, false); +} + +#if 0 +static void trinity_pre_display_configuration_change(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (pi->voltage_drop_in_dce) + trinity_dce_enable_voltage_adjustment(rdev, false); +} +#endif + +static void trinity_add_dccac_value(struct radeon_device *rdev) +{ + u32 gpu_cac_avrg_cntl_window_size; + u32 num_active_displays = rdev->pm.dpm.new_active_crtc_count; + u64 disp_clk = rdev->clock.default_dispclk / 100; + u32 dc_cac_value; + + gpu_cac_avrg_cntl_window_size = + (RREG32_SMC(GPU_CAC_AVRG_CNTL) & WINDOW_SIZE_MASK) >> WINDOW_SIZE_SHIFT; + + dc_cac_value = (u32)((14213 * disp_clk * disp_clk * (u64)num_active_displays) >> + (32 - gpu_cac_avrg_cntl_window_size)); + + WREG32_SMC(DC_CAC_VALUE, dc_cac_value); +} + +void trinity_dpm_display_configuration_changed(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + if (pi->voltage_drop_in_dce) + trinity_dce_enable_voltage_adjustment(rdev, true); + trinity_add_dccac_value(rdev); +} + +union power_info { + struct _ATOM_POWERPLAY_INFO info; + struct _ATOM_POWERPLAY_INFO_V2 info_2; + struct _ATOM_POWERPLAY_INFO_V3 info_3; + struct _ATOM_PPLIB_POWERPLAYTABLE pplib; + struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2; + struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3; +}; + +union pplib_clock_info { + struct _ATOM_PPLIB_R600_CLOCK_INFO r600; + struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780; + struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen; + struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo; +}; + +union pplib_power_state { + struct _ATOM_PPLIB_STATE v1; + struct _ATOM_PPLIB_STATE_V2 v2; +}; + +static void trinity_parse_pplib_non_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info, + u8 table_rev) +{ + struct trinity_ps *ps = trinity_get_ps(rps); + + rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings); + rps->class = le16_to_cpu(non_clock_info->usClassification); + rps->class2 = le16_to_cpu(non_clock_info->usClassification2); + + if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) { + rps->vclk = le32_to_cpu(non_clock_info->ulVCLK); + rps->dclk = le32_to_cpu(non_clock_info->ulDCLK); + } else { + rps->vclk = 0; + rps->dclk = 0; + } + + if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) { + rdev->pm.dpm.boot_ps = rps; + trinity_patch_boot_state(rdev, ps); + } + if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE) + rdev->pm.dpm.uvd_ps = rps; +} + +static void trinity_parse_pplib_clock_info(struct radeon_device *rdev, + struct radeon_ps *rps, int index, + union pplib_clock_info *clock_info) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct trinity_ps *ps = trinity_get_ps(rps); + struct trinity_pl *pl = &ps->levels[index]; + u32 sclk; + + sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow); + sclk |= clock_info->sumo.ucEngineClockHigh << 16; + pl->sclk = sclk; + pl->vddc_index = clock_info->sumo.vddcIndex; + + ps->num_levels = index + 1; + + if (pi->enable_sclk_ds) { + pl->ds_divider_index = 5; + pl->ss_divider_index = 5; + } +} + +static int trinity_parse_power_table(struct radeon_device *rdev) +{ + struct radeon_mode_info *mode_info = &rdev->mode_info; + struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info; + union pplib_power_state *power_state; + int i, j, k, non_clock_array_index, clock_array_index; + union pplib_clock_info *clock_info; + struct _StateArray *state_array; + struct _ClockInfoArray *clock_info_array; + struct _NonClockInfoArray *non_clock_info_array; + union power_info *power_info; + int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo); + u16 data_offset; + u8 frev, crev; + u8 *power_state_offset; + struct sumo_ps *ps; + + if (!atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) + return -EINVAL; + power_info = (union power_info *)((uint8_t*)mode_info->atom_context->bios + data_offset); + + state_array = (struct _StateArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usStateArrayOffset)); + clock_info_array = (struct _ClockInfoArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usClockInfoArrayOffset)); + non_clock_info_array = (struct _NonClockInfoArray *) + ((uint8_t*)mode_info->atom_context->bios + data_offset + + le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset)); + + rdev->pm.dpm.ps = kzalloc(sizeof(struct radeon_ps) * + state_array->ucNumEntries, GFP_KERNEL); + if (!rdev->pm.dpm.ps) + return -ENOMEM; + power_state_offset = (u8 *)state_array->states; + rdev->pm.dpm.platform_caps = le32_to_cpu(power_info->pplib.ulPlatformCaps); + rdev->pm.dpm.backbias_response_time = le16_to_cpu(power_info->pplib.usBackbiasTime); + rdev->pm.dpm.voltage_response_time = le16_to_cpu(power_info->pplib.usVoltageTime); + for (i = 0; i < state_array->ucNumEntries; i++) { + power_state = (union pplib_power_state *)power_state_offset; + non_clock_array_index = power_state->v2.nonClockInfoIndex; + non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *) + &non_clock_info_array->nonClockInfo[non_clock_array_index]; + if (!rdev->pm.power_state[i].clock_info) + return -EINVAL; + ps = kzalloc(sizeof(struct sumo_ps), GFP_KERNEL); + if (ps == NULL) { + kfree(rdev->pm.dpm.ps); + return -ENOMEM; + } + rdev->pm.dpm.ps[i].ps_priv = ps; + k = 0; + for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) { + clock_array_index = power_state->v2.clockInfoIndex[j]; + if (clock_array_index >= clock_info_array->ucNumEntries) + continue; + if (k >= SUMO_MAX_HARDWARE_POWERLEVELS) + break; + clock_info = (union pplib_clock_info *) + &clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize]; + trinity_parse_pplib_clock_info(rdev, + &rdev->pm.dpm.ps[i], k, + clock_info); + k++; + } + trinity_parse_pplib_non_clock_info(rdev, &rdev->pm.dpm.ps[i], + non_clock_info, + non_clock_info_array->ucEntrySize); + power_state_offset += 2 + power_state->v2.ucNumDPMLevels; + } + rdev->pm.dpm.num_ps = state_array->ucNumEntries; + return 0; +} + +union igp_info { + struct _ATOM_INTEGRATED_SYSTEM_INFO info; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V5 info_5; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 info_6; + struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 info_7; +}; + +static u32 trinity_convert_did_to_freq(struct radeon_device *rdev, u8 did) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + u32 divider; + + if (did >= 8 && did <= 0x3f) + divider = did * 25; + else if (did > 0x3f && did <= 0x5f) + divider = (did - 64) * 50 + 1600; + else if (did > 0x5f && did <= 0x7e) + divider = (did - 96) * 100 + 3200; + else if (did == 0x7f) + divider = 128 * 100; + else + return 10000; + + return ((pi->sys_info.dentist_vco_freq * 100) + (divider - 1)) / divider; +} + +static int trinity_parse_sys_info_table(struct radeon_device *rdev) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct radeon_mode_info *mode_info = &rdev->mode_info; + int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo); + union igp_info *igp_info; + u8 frev, crev; + u16 data_offset; + int i; + + if (atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + igp_info = (union igp_info *)((uint8_t*)mode_info->atom_context->bios + + data_offset); + + if (crev != 7) { + DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev); + return -EINVAL; + } + pi->sys_info.bootup_sclk = le32_to_cpu(igp_info->info_7.ulBootUpEngineClock); + pi->sys_info.min_sclk = le32_to_cpu(igp_info->info_7.ulMinEngineClock); + pi->sys_info.bootup_uma_clk = le32_to_cpu(igp_info->info_7.ulBootUpUMAClock); + pi->sys_info.dentist_vco_freq = le32_to_cpu(igp_info->info_7.ulDentistVCOFreq); + pi->sys_info.bootup_nb_voltage_index = + le16_to_cpu(igp_info->info_7.usBootUpNBVoltage); + if (igp_info->info_7.ucHtcTmpLmt == 0) + pi->sys_info.htc_tmp_lmt = 203; + else + pi->sys_info.htc_tmp_lmt = igp_info->info_7.ucHtcTmpLmt; + if (igp_info->info_7.ucHtcHystLmt == 0) + pi->sys_info.htc_hyst_lmt = 5; + else + pi->sys_info.htc_hyst_lmt = igp_info->info_7.ucHtcHystLmt; + if (pi->sys_info.htc_tmp_lmt <= pi->sys_info.htc_hyst_lmt) { + DRM_ERROR("The htcTmpLmt should be larger than htcHystLmt.\n"); + } + + if (pi->enable_nbps_policy) + pi->sys_info.nb_dpm_enable = igp_info->info_7.ucNBDPMEnable; + else + pi->sys_info.nb_dpm_enable = 0; + + for (i = 0; i < TRINITY_NUM_NBPSTATES; i++) { + pi->sys_info.nbp_mclk[i] = le32_to_cpu(igp_info->info_7.ulNbpStateMemclkFreq[i]); + pi->sys_info.nbp_nclk[i] = le32_to_cpu(igp_info->info_7.ulNbpStateNClkFreq[i]); + } + + pi->sys_info.nbp_voltage_index[0] = le16_to_cpu(igp_info->info_7.usNBP0Voltage); + pi->sys_info.nbp_voltage_index[1] = le16_to_cpu(igp_info->info_7.usNBP1Voltage); + pi->sys_info.nbp_voltage_index[2] = le16_to_cpu(igp_info->info_7.usNBP2Voltage); + pi->sys_info.nbp_voltage_index[3] = le16_to_cpu(igp_info->info_7.usNBP3Voltage); + + if (!pi->sys_info.nb_dpm_enable) { + for (i = 1; i < TRINITY_NUM_NBPSTATES; i++) { + pi->sys_info.nbp_mclk[i] = pi->sys_info.nbp_mclk[0]; + pi->sys_info.nbp_nclk[i] = pi->sys_info.nbp_nclk[0]; + pi->sys_info.nbp_voltage_index[i] = pi->sys_info.nbp_voltage_index[0]; + } + } + + pi->sys_info.uma_channel_number = igp_info->info_7.ucUMAChannelNumber; + + sumo_construct_sclk_voltage_mapping_table(rdev, + &pi->sys_info.sclk_voltage_mapping_table, + igp_info->info_7.sAvail_SCLK); + sumo_construct_vid_mapping_table(rdev, &pi->sys_info.vid_mapping_table, + igp_info->info_7.sAvail_SCLK); + + pi->sys_info.uvd_clock_table_entries[0].vclk_did = + igp_info->info_7.ucDPMState0VclkFid; + pi->sys_info.uvd_clock_table_entries[1].vclk_did = + igp_info->info_7.ucDPMState1VclkFid; + pi->sys_info.uvd_clock_table_entries[2].vclk_did = + igp_info->info_7.ucDPMState2VclkFid; + pi->sys_info.uvd_clock_table_entries[3].vclk_did = + igp_info->info_7.ucDPMState3VclkFid; + + pi->sys_info.uvd_clock_table_entries[0].dclk_did = + igp_info->info_7.ucDPMState0DclkFid; + pi->sys_info.uvd_clock_table_entries[1].dclk_did = + igp_info->info_7.ucDPMState1DclkFid; + pi->sys_info.uvd_clock_table_entries[2].dclk_did = + igp_info->info_7.ucDPMState2DclkFid; + pi->sys_info.uvd_clock_table_entries[3].dclk_did = + igp_info->info_7.ucDPMState3DclkFid; + + for (i = 0; i < 4; i++) { + pi->sys_info.uvd_clock_table_entries[i].vclk = + trinity_convert_did_to_freq(rdev, + pi->sys_info.uvd_clock_table_entries[i].vclk_did); + pi->sys_info.uvd_clock_table_entries[i].dclk = + trinity_convert_did_to_freq(rdev, + pi->sys_info.uvd_clock_table_entries[i].dclk_did); + } + + + + } + return 0; +} + +int trinity_dpm_init(struct radeon_device *rdev) +{ + struct trinity_power_info *pi; + int ret, i; + + pi = kzalloc(sizeof(struct trinity_power_info), GFP_KERNEL); + if (pi == NULL) + return -ENOMEM; + rdev->pm.dpm.priv = pi; + + for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) + pi->at[i] = TRINITY_AT_DFLT; + + pi->enable_nbps_policy = true; + pi->enable_sclk_ds = true; + pi->enable_gfx_power_gating = true; + pi->enable_gfx_clock_gating = true; + pi->enable_mg_clock_gating = true; + pi->enable_gfx_dynamic_mgpg = true; /* ??? */ + pi->override_dynamic_mgpg = true; + pi->enable_auto_thermal_throttling = true; + pi->voltage_drop_in_dce = false; /* need to restructure dpm/modeset interaction */ + pi->uvd_dpm = true; /* ??? */ + + ret = trinity_parse_sys_info_table(rdev); + if (ret) + return ret; + + trinity_construct_boot_state(rdev); + + ret = trinity_parse_power_table(rdev); + if (ret) + return ret; + + pi->thermal_auto_throttling = pi->sys_info.htc_tmp_lmt; + pi->enable_dpm = true; + + return 0; +} + +void trinity_dpm_print_power_state(struct radeon_device *rdev, + struct radeon_ps *rps) +{ + int i; + struct trinity_ps *ps = trinity_get_ps(rps); + + r600_dpm_print_class_info(rps->class, rps->class2); + r600_dpm_print_cap_info(rps->caps); + printk("\tuvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + for (i = 0; i < ps->num_levels; i++) { + struct trinity_pl *pl = &ps->levels[i]; + printk("\t\tpower level %d sclk: %u vddc: %u\n", + i, pl->sclk, + trinity_convert_voltage_index_to_value(rdev, pl->vddc_index)); + } + r600_dpm_print_ps_status(rdev, rps); +} + +void trinity_dpm_debugfs_print_current_performance_level(struct radeon_device *rdev, + struct seq_file *m) +{ + struct radeon_ps *rps = rdev->pm.dpm.current_ps; + struct trinity_ps *ps = trinity_get_ps(rps); + struct trinity_pl *pl; + u32 current_index = + (RREG32(TARGET_AND_CURRENT_PROFILE_INDEX) & CURRENT_STATE_MASK) >> + CURRENT_STATE_SHIFT; + + if (current_index >= ps->num_levels) { + seq_printf(m, "invalid dpm profile %d\n", current_index); + } else { + pl = &ps->levels[current_index]; + seq_printf(m, "uvd vclk: %d dclk: %d\n", rps->vclk, rps->dclk); + seq_printf(m, "power level %d sclk: %u vddc: %u\n", + current_index, pl->sclk, + trinity_convert_voltage_index_to_value(rdev, pl->vddc_index)); + } +} + +void trinity_dpm_fini(struct radeon_device *rdev) +{ + int i; + + trinity_cleanup_asic(rdev); /* ??? */ + + for (i = 0; i < rdev->pm.dpm.num_ps; i++) { + kfree(rdev->pm.dpm.ps[i].ps_priv); + } + kfree(rdev->pm.dpm.ps); + kfree(rdev->pm.dpm.priv); +} + +u32 trinity_dpm_get_sclk(struct radeon_device *rdev, bool low) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + struct trinity_ps *requested_state = trinity_get_ps(&pi->requested_rps); + + if (low) + return requested_state->levels[0].sclk; + else + return requested_state->levels[requested_state->num_levels - 1].sclk; +} + +u32 trinity_dpm_get_mclk(struct radeon_device *rdev, bool low) +{ + struct trinity_power_info *pi = trinity_get_pi(rdev); + + return pi->sys_info.bootup_uma_clk; +} diff --git a/sys/dev/drm/radeon/trinity_dpm.h b/sys/dev/drm/radeon/trinity_dpm.h new file mode 100644 index 0000000000..e82df071f8 --- /dev/null +++ b/sys/dev/drm/radeon/trinity_dpm.h @@ -0,0 +1,132 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __TRINITY_DPM_H__ +#define __TRINITY_DPM_H__ + +#include "sumo_dpm.h" + +#define TRINITY_SIZEOF_DPM_STATE_TABLE (SMU_SCLK_DPM_STATE_1_CNTL_0 - SMU_SCLK_DPM_STATE_0_CNTL_0) + +struct trinity_pl { + u32 sclk; + u8 vddc_index; + u8 ds_divider_index; + u8 ss_divider_index; + u8 allow_gnb_slow; + u8 force_nbp_state; + u8 display_wm; + u8 vce_wm; +}; + +#define TRINITY_POWERSTATE_FLAGS_NBPS_FORCEHIGH (1 << 0) +#define TRINITY_POWERSTATE_FLAGS_NBPS_LOCKTOHIGH (1 << 1) +#define TRINITY_POWERSTATE_FLAGS_NBPS_LOCKTOLOW (1 << 2) + +#define TRINITY_POWERSTATE_FLAGS_BAPM_DISABLE (1 << 0) + +struct trinity_ps { + u32 num_levels; + struct trinity_pl levels[SUMO_MAX_HARDWARE_POWERLEVELS]; + + u32 nbps_flags; + u32 bapm_flags; + + u8 Dpm0PgNbPsLo; + u8 Dpm0PgNbPsHi; + u8 DpmXNbPsLo; + u8 DpmXNbPsHi; + + u32 vclk_low_divider; + u32 vclk_high_divider; + u32 dclk_low_divider; + u32 dclk_high_divider; +}; + +#define TRINITY_NUM_NBPSTATES 4 + +struct trinity_uvd_clock_table_entry +{ + u32 vclk; + u32 dclk; + u8 vclk_did; + u8 dclk_did; + u8 rsv[2]; +}; + +struct trinity_sys_info { + u32 bootup_uma_clk; + u32 bootup_sclk; + u32 min_sclk; + u32 dentist_vco_freq; + u32 nb_dpm_enable; + u32 nbp_mclk[TRINITY_NUM_NBPSTATES]; + u32 nbp_nclk[TRINITY_NUM_NBPSTATES]; + u16 nbp_voltage_index[TRINITY_NUM_NBPSTATES]; + u16 bootup_nb_voltage_index; + u8 htc_tmp_lmt; + u8 htc_hyst_lmt; + struct sumo_sclk_voltage_mapping_table sclk_voltage_mapping_table; + struct sumo_vid_mapping_table vid_mapping_table; + u32 uma_channel_number; + struct trinity_uvd_clock_table_entry uvd_clock_table_entries[4]; +}; + +struct trinity_power_info { + u32 at[SUMO_MAX_HARDWARE_POWERLEVELS]; + u32 dpm_interval; + u32 thermal_auto_throttling; + struct trinity_sys_info sys_info; + struct trinity_pl boot_pl; + u32 min_sclk_did; + bool enable_nbps_policy; + bool voltage_drop_in_dce; + bool override_dynamic_mgpg; + bool enable_gfx_clock_gating; + bool enable_gfx_power_gating; + bool enable_mg_clock_gating; + bool enable_gfx_dynamic_mgpg; + bool enable_auto_thermal_throttling; + bool enable_dpm; + bool enable_sclk_ds; + bool uvd_dpm; + struct radeon_ps current_rps; + struct trinity_ps current_ps; + struct radeon_ps requested_rps; + struct trinity_ps requested_ps; +}; + +#define TRINITY_AT_DFLT 30 + +/* trinity_smc.c */ +int trinity_dpm_config(struct radeon_device *rdev, bool enable); +int trinity_uvd_dpm_config(struct radeon_device *rdev); +int trinity_dpm_force_state(struct radeon_device *rdev, u32 n); +int trinity_dpm_n_levels_disabled(struct radeon_device *rdev, u32 n); +int trinity_dpm_no_forced_level(struct radeon_device *rdev); +int trinity_dce_enable_voltage_adjustment(struct radeon_device *rdev, + bool enable); +int trinity_gfx_dynamic_mgpg_config(struct radeon_device *rdev); +void trinity_acquire_mutex(struct radeon_device *rdev); +void trinity_release_mutex(struct radeon_device *rdev); + +#endif diff --git a/sys/dev/drm/radeon/trinity_smc.c b/sys/dev/drm/radeon/trinity_smc.c new file mode 100644 index 0000000000..ea906d9ba4 --- /dev/null +++ b/sys/dev/drm/radeon/trinity_smc.c @@ -0,0 +1,122 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include +#include "radeon.h" +#include "trinityd.h" +#include "trinity_dpm.h" +#include "ppsmc.h" + +struct trinity_ps *trinity_get_ps(struct radeon_ps *rps); +struct trinity_power_info *trinity_get_pi(struct radeon_device *rdev); + +static int trinity_notify_message_to_smu(struct radeon_device *rdev, u32 id) +{ + int i; + u32 v = 0; + + WREG32(SMC_MESSAGE_0, id); + for (i = 0; i < rdev->usec_timeout; i++) { + if (RREG32(SMC_RESP_0) != 0) + break; + DRM_UDELAY(1); + } + v = RREG32(SMC_RESP_0); + + if (v != 1) { + if (v == 0xFF) { + DRM_ERROR("SMC failed to handle the message!\n"); + return -EINVAL; + } else if (v == 0xFE) { + DRM_ERROR("Unknown SMC message!\n"); + return -EINVAL; + } + } + + return 0; +} + +int trinity_dpm_config(struct radeon_device *rdev, bool enable) +{ + if (enable) + WREG32_SMC(SMU_SCRATCH0, 1); + else + WREG32_SMC(SMU_SCRATCH0, 0); + + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_DPM_Config); +} + +int trinity_dpm_force_state(struct radeon_device *rdev, u32 n) +{ + WREG32_SMC(SMU_SCRATCH0, n); + + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_DPM_ForceState); +} + +int trinity_dpm_n_levels_disabled(struct radeon_device *rdev, u32 n) +{ + WREG32_SMC(SMU_SCRATCH0, n); + + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_DPM_N_LevelsDisabled); +} + +int trinity_uvd_dpm_config(struct radeon_device *rdev) +{ + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_UVD_DPM_Config); +} + +int trinity_dpm_no_forced_level(struct radeon_device *rdev) +{ + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_NoForcedLevel); +} + +int trinity_dce_enable_voltage_adjustment(struct radeon_device *rdev, + bool enable) +{ + if (enable) + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_DCE_AllowVoltageAdjustment); + else + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_DCE_RemoveVoltageAdjustment); +} + +int trinity_gfx_dynamic_mgpg_config(struct radeon_device *rdev) +{ + return trinity_notify_message_to_smu(rdev, PPSMC_MSG_PG_SIMD_Config); +} + +void trinity_acquire_mutex(struct radeon_device *rdev) +{ + int i; + + WREG32(SMC_INT_REQ, 1); + for (i = 0; i < rdev->usec_timeout; i++) { + if ((RREG32(SMC_INT_REQ) & 0xffff) == 1) + break; + DRM_UDELAY(1); + } +} + +void trinity_release_mutex(struct radeon_device *rdev) +{ + WREG32(SMC_INT_REQ, 0); +} diff --git a/sys/dev/drm/radeon/trinityd.h b/sys/dev/drm/radeon/trinityd.h new file mode 100644 index 0000000000..fd32e27717 --- /dev/null +++ b/sys/dev/drm/radeon/trinityd.h @@ -0,0 +1,228 @@ +/* + * Copyright 2012 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Alex Deucher + */ +#ifndef _TRINITYD_H_ +#define _TRINITYD_H_ + +/* pm registers */ + +/* cg */ +#define CG_CGTT_LOCAL_0 0x0 +#define CG_CGTT_LOCAL_1 0x1 + +/* smc */ +#define SMU_SCLK_DPM_STATE_0_CNTL_0 0x1f000 +# define STATE_VALID(x) ((x) << 0) +# define STATE_VALID_MASK (0xff << 0) +# define STATE_VALID_SHIFT 0 +# define CLK_DIVIDER(x) ((x) << 8) +# define CLK_DIVIDER_MASK (0xff << 8) +# define CLK_DIVIDER_SHIFT 8 +# define VID(x) ((x) << 16) +# define VID_MASK (0xff << 16) +# define VID_SHIFT 16 +# define LVRT(x) ((x) << 24) +# define LVRT_MASK (0xff << 24) +# define LVRT_SHIFT 24 +#define SMU_SCLK_DPM_STATE_0_CNTL_1 0x1f004 +# define DS_DIV(x) ((x) << 0) +# define DS_DIV_MASK (0xff << 0) +# define DS_DIV_SHIFT 0 +# define DS_SH_DIV(x) ((x) << 8) +# define DS_SH_DIV_MASK (0xff << 8) +# define DS_SH_DIV_SHIFT 8 +# define DISPLAY_WM(x) ((x) << 16) +# define DISPLAY_WM_MASK (0xff << 16) +# define DISPLAY_WM_SHIFT 16 +# define VCE_WM(x) ((x) << 24) +# define VCE_WM_MASK (0xff << 24) +# define VCE_WM_SHIFT 24 + +#define SMU_SCLK_DPM_STATE_0_CNTL_3 0x1f00c +# define GNB_SLOW(x) ((x) << 0) +# define GNB_SLOW_MASK (0xff << 0) +# define GNB_SLOW_SHIFT 0 +# define FORCE_NBPS1(x) ((x) << 8) +# define FORCE_NBPS1_MASK (0xff << 8) +# define FORCE_NBPS1_SHIFT 8 +#define SMU_SCLK_DPM_STATE_0_AT 0x1f010 +# define AT(x) ((x) << 0) +# define AT_MASK (0xff << 0) +# define AT_SHIFT 0 + +#define SMU_SCLK_DPM_STATE_0_PG_CNTL 0x1f014 +# define PD_SCLK_DIVIDER(x) ((x) << 16) +# define PD_SCLK_DIVIDER_MASK (0xff << 16) +# define PD_SCLK_DIVIDER_SHIFT 16 + +#define SMU_SCLK_DPM_STATE_1_CNTL_0 0x1f020 + +#define SMU_SCLK_DPM_CNTL 0x1f100 +# define SCLK_DPM_EN(x) ((x) << 0) +# define SCLK_DPM_EN_MASK (0xff << 0) +# define SCLK_DPM_EN_SHIFT 0 +# define SCLK_DPM_BOOT_STATE(x) ((x) << 16) +# define SCLK_DPM_BOOT_STATE_MASK (0xff << 16) +# define SCLK_DPM_BOOT_STATE_SHIFT 16 +# define VOLTAGE_CHG_EN(x) ((x) << 24) +# define VOLTAGE_CHG_EN_MASK (0xff << 24) +# define VOLTAGE_CHG_EN_SHIFT 24 + +#define SMU_SCLK_DPM_TT_CNTL 0x1f108 +# define SCLK_TT_EN(x) ((x) << 0) +# define SCLK_TT_EN_MASK (0xff << 0) +# define SCLK_TT_EN_SHIFT 0 +#define SMU_SCLK_DPM_TTT 0x1f10c +# define LT(x) ((x) << 0) +# define LT_MASK (0xffff << 0) +# define LT_SHIFT 0 +# define HT(x) ((x) << 16) +# define HT_MASK (0xffff << 16) +# define HT_SHIFT 16 + +#define SMU_UVD_DPM_STATES 0x1f1a0 +#define SMU_UVD_DPM_CNTL 0x1f1a4 + +#define SMU_S_PG_CNTL 0x1f118 +# define DS_PG_EN(x) ((x) << 16) +# define DS_PG_EN_MASK (0xff << 16) +# define DS_PG_EN_SHIFT 16 + +#define GFX_POWER_GATING_CNTL 0x1f38c +# define PDS_DIV(x) ((x) << 0) +# define PDS_DIV_MASK (0xff << 0) +# define PDS_DIV_SHIFT 0 +# define SSSD(x) ((x) << 8) +# define SSSD_MASK (0xff << 8) +# define SSSD_SHIFT 8 + +#define PM_CONFIG 0x1f428 +# define SVI_Mode (1 << 29) + +#define PM_I_CNTL_1 0x1f464 +# define SCLK_DPM(x) ((x) << 0) +# define SCLK_DPM_MASK (0xff << 0) +# define SCLK_DPM_SHIFT 0 +# define DS_PG_CNTL(x) ((x) << 16) +# define DS_PG_CNTL_MASK (0xff << 16) +# define DS_PG_CNTL_SHIFT 16 +#define PM_TP 0x1f468 + +#define NB_PSTATE_CONFIG 0x1f5f8 +# define Dpm0PgNbPsLo(x) ((x) << 0) +# define Dpm0PgNbPsLo_MASK (3 << 0) +# define Dpm0PgNbPsLo_SHIFT 0 +# define Dpm0PgNbPsHi(x) ((x) << 2) +# define Dpm0PgNbPsHi_MASK (3 << 2) +# define Dpm0PgNbPsHi_SHIFT 2 +# define DpmXNbPsLo(x) ((x) << 4) +# define DpmXNbPsLo_MASK (3 << 4) +# define DpmXNbPsLo_SHIFT 4 +# define DpmXNbPsHi(x) ((x) << 6) +# define DpmXNbPsHi_MASK (3 << 6) +# define DpmXNbPsHi_SHIFT 6 + +#define DC_CAC_VALUE 0x1f908 + +#define GPU_CAC_AVRG_CNTL 0x1f920 +# define WINDOW_SIZE(x) ((x) << 0) +# define WINDOW_SIZE_MASK (0xff << 0) +# define WINDOW_SIZE_SHIFT 0 + +#define CC_SMU_MISC_FUSES 0xe0001004 +# define MinSClkDid(x) ((x) << 2) +# define MinSClkDid_MASK (0x7f << 2) +# define MinSClkDid_SHIFT 2 + +#define CC_SMU_TST_EFUSE1_MISC 0xe000101c +# define RB_BACKEND_DISABLE(x) ((x) << 16) +# define RB_BACKEND_DISABLE_MASK (3 << 16) +# define RB_BACKEND_DISABLE_SHIFT 16 + +#define SMU_SCRATCH_A 0xe0003024 + +#define SMU_SCRATCH0 0xe0003040 + +/* mmio */ +#define SMC_INT_REQ 0x220 + +#define SMC_MESSAGE_0 0x22c +#define SMC_RESP_0 0x230 + +#define GENERAL_PWRMGT 0x670 +# define GLOBAL_PWRMGT_EN (1 << 0) + +#define SCLK_PWRMGT_CNTL 0x678 +# define DYN_PWR_DOWN_EN (1 << 2) +# define RESET_BUSY_CNT (1 << 4) +# define RESET_SCLK_CNT (1 << 5) +# define DYN_GFX_CLK_OFF_EN (1 << 7) +# define GFX_CLK_FORCE_ON (1 << 8) +# define DYNAMIC_PM_EN (1 << 21) + +#define TARGET_AND_CURRENT_PROFILE_INDEX 0x684 +# define TARGET_STATE(x) ((x) << 0) +# define TARGET_STATE_MASK (0xf << 0) +# define TARGET_STATE_SHIFT 0 +# define CURRENT_STATE(x) ((x) << 4) +# define CURRENT_STATE_MASK (0xf << 4) +# define CURRENT_STATE_SHIFT 4 + +#define CG_GIPOTS 0x6d8 +# define CG_GIPOT(x) ((x) << 16) +# define CG_GIPOT_MASK (0xffff << 16) +# define CG_GIPOT_SHIFT 16 + +#define CG_PG_CTRL 0x6e0 +# define SP(x) ((x) << 0) +# define SP_MASK (0xffff << 0) +# define SP_SHIFT 0 +# define SU(x) ((x) << 16) +# define SU_MASK (0xffff << 16) +# define SU_SHIFT 16 + +#define CG_MISC_REG 0x708 + +#define CG_THERMAL_INT_CTRL 0x738 +# define DIG_THERM_INTH(x) ((x) << 0) +# define DIG_THERM_INTH_MASK (0xff << 0) +# define DIG_THERM_INTH_SHIFT 0 +# define DIG_THERM_INTL(x) ((x) << 8) +# define DIG_THERM_INTL_MASK (0xff << 8) +# define DIG_THERM_INTL_SHIFT 8 +# define THERM_INTH_MASK (1 << 24) +# define THERM_INTL_MASK (1 << 25) + +#define CG_CG_VOLTAGE_CNTL 0x770 +# define EN (1 << 9) + +#define HW_REV 0x5564 +# define ATI_REV_ID_MASK (0xf << 28) +# define ATI_REV_ID_SHIFT 28 +/* 0 = A0, 1 = A1, 2 = B0, 3 = C0, etc. */ + +#define CGTS_SM_CTRL_REG 0x9150 + +#define GB_ADDR_CONFIG 0x98f8 + +#endif -- 2.41.0