projects / dragonfly.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | history | HEAD
igb: Change polling(4) support to ifpoll support
[dragonfly.git] / sys / dev / netif / igb / if_igb.c
CommitLineData
1f7e3916
SZ		 1/*
2 * Copyright (c) 2001-2011, Intel Corporation
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are met:
7 *
8 * 1. Redistributions of source code must retain the above copyright notice,
9 * this list of conditions and the following disclaimer.
10 *
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * 3. Neither the name of the Intel Corporation nor the names of its
16 * contributors may be used to endorse or promote products derived from
17 * this software without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
20 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
23 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
d0f59cad 32#include "opt_ifpoll.h"
8d6600da 33#include "opt_igb.h"
1f7e3916
SZ		 34
35#include <sys/param.h>
36#include <sys/bus.h>
37#include <sys/endian.h>
38#include <sys/interrupt.h>
39#include <sys/kernel.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/proc.h>
43#include <sys/rman.h>
44#include <sys/serialize.h>
45#include <sys/serialize2.h>
46#include <sys/socket.h>
47#include <sys/sockio.h>
48#include <sys/sysctl.h>
49#include <sys/systm.h>
50
51#include <net/bpf.h>
52#include <net/ethernet.h>
53#include <net/if.h>
54#include <net/if_arp.h>
55#include <net/if_dl.h>
56#include <net/if_media.h>
57#include <net/ifq_var.h>
58#include <net/toeplitz.h>
59#include <net/toeplitz2.h>
60#include <net/vlan/if_vlan_var.h>
61#include <net/vlan/if_vlan_ether.h>
62#include <net/if_poll.h>
63
64#include <netinet/in_systm.h>
65#include <netinet/in.h>
66#include <netinet/ip.h>
67#include <netinet/tcp.h>
68#include <netinet/udp.h>
69
70#include <bus/pci/pcivar.h>
71#include <bus/pci/pcireg.h>
72
73#include <dev/netif/ig_hal/e1000_api.h>
74#include <dev/netif/ig_hal/e1000_82575.h>
75#include <dev/netif/igb/if_igb.h>
76
8d6600da
SZ		 77#ifdef IGB_RSS_DEBUG
78#define IGB_RSS_DPRINTF(sc, lvl, fmt, ...) \
79do { \
80 if (sc->rss_debug >= lvl) \
81 if_printf(&sc->arpcom.ac_if, fmt, __VA_ARGS__); \
82} while (0)
83#else /* !IGB_RSS_DEBUG */
84#define IGB_RSS_DPRINTF(sc, lvl, fmt, ...) ((void)0)
85#endif /* IGB_RSS_DEBUG */
86
1f7e3916
SZ		 87#define IGB_NAME "Intel(R) PRO/1000 "
88#define IGB_DEVICE(id) \
89 { IGB_VENDOR_ID, E1000_DEV_ID_##id, IGB_NAME #id }
90#define IGB_DEVICE_NULL { 0, 0, NULL }
91
92static struct igb_device {
93 uint16_t vid;
94 uint16_t did;
95 const char *desc;
96} igb_devices[] = {
97 IGB_DEVICE(82575EB_COPPER),
98 IGB_DEVICE(82575EB_FIBER_SERDES),
99 IGB_DEVICE(82575GB_QUAD_COPPER),
100 IGB_DEVICE(82576),
101 IGB_DEVICE(82576_NS),
102 IGB_DEVICE(82576_NS_SERDES),
103 IGB_DEVICE(82576_FIBER),
104 IGB_DEVICE(82576_SERDES),
105 IGB_DEVICE(82576_SERDES_QUAD),
106 IGB_DEVICE(82576_QUAD_COPPER),
107 IGB_DEVICE(82576_QUAD_COPPER_ET2),
108 IGB_DEVICE(82576_VF),
109 IGB_DEVICE(82580_COPPER),
110 IGB_DEVICE(82580_FIBER),
111 IGB_DEVICE(82580_SERDES),
112 IGB_DEVICE(82580_SGMII),
113 IGB_DEVICE(82580_COPPER_DUAL),
114 IGB_DEVICE(82580_QUAD_FIBER),
115 IGB_DEVICE(DH89XXCC_SERDES),
116 IGB_DEVICE(DH89XXCC_SGMII),
117 IGB_DEVICE(DH89XXCC_SFP),
118 IGB_DEVICE(DH89XXCC_BACKPLANE),
119 IGB_DEVICE(I350_COPPER),
120 IGB_DEVICE(I350_FIBER),
121 IGB_DEVICE(I350_SERDES),
122 IGB_DEVICE(I350_SGMII),
123 IGB_DEVICE(I350_VF),
124
125 /* required last entry */
126 IGB_DEVICE_NULL
127};
128
129static int igb_probe(device_t);
130static int igb_attach(device_t);
131static int igb_detach(device_t);
132static int igb_shutdown(device_t);
133static int igb_suspend(device_t);
134static int igb_resume(device_t);
135
136static boolean_t igb_is_valid_ether_addr(const uint8_t *);
137static void igb_setup_ifp(struct igb_softc *);
48faa653 138static boolean_t igb_txcsum_ctx(struct igb_tx_ring *, struct mbuf *);
23f6ffe4
SZ		 139static int igb_tso_pullup(struct igb_tx_ring *, struct mbuf **);
140static void igb_tso_ctx(struct igb_tx_ring *, struct mbuf *, uint32_t *);
1f7e3916
SZ		 141static void igb_add_sysctl(struct igb_softc *);
142static int igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS);
9c0ecdcc 143static int igb_sysctl_msix_rate(SYSCTL_HANDLER_ARGS);
b6220144 144static int igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS);
9c0ecdcc 145static void igb_set_ring_inuse(struct igb_softc *, boolean_t);
d0f59cad
SZ		 146#ifdef IFPOLL_ENABLE
147static int igb_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS);
148static int igb_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS);
149#endif
1f7e3916
SZ		 150
151static void igb_vf_init_stats(struct igb_softc *);
152static void igb_reset(struct igb_softc *);
153static void igb_update_stats_counters(struct igb_softc *);
154static void igb_update_vf_stats_counters(struct igb_softc *);
155static void igb_update_link_status(struct igb_softc *);
156static void igb_init_tx_unit(struct igb_softc *);
157static void igb_init_rx_unit(struct igb_softc *);
158
159static void igb_set_vlan(struct igb_softc *);
160static void igb_set_multi(struct igb_softc *);
161static void igb_set_promisc(struct igb_softc *);
162static void igb_disable_promisc(struct igb_softc *);
163
a619b256
SZ		 164static int igb_alloc_rings(struct igb_softc *);
165static void igb_free_rings(struct igb_softc *);
1f7e3916
SZ		 166static int igb_create_tx_ring(struct igb_tx_ring *);
167static int igb_create_rx_ring(struct igb_rx_ring *);
168static void igb_free_tx_ring(struct igb_tx_ring *);
169static void igb_free_rx_ring(struct igb_rx_ring *);
170static void igb_destroy_tx_ring(struct igb_tx_ring *, int);
171static void igb_destroy_rx_ring(struct igb_rx_ring *, int);
172static void igb_init_tx_ring(struct igb_tx_ring *);
173static int igb_init_rx_ring(struct igb_rx_ring *);
174static int igb_newbuf(struct igb_rx_ring *, int, boolean_t);
175static int igb_encap(struct igb_tx_ring *, struct mbuf **);
176
177static void igb_stop(struct igb_softc *);
178static void igb_init(void *);
179static int igb_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
180static void igb_media_status(struct ifnet *, struct ifmediareq *);
181static int igb_media_change(struct ifnet *);
182static void igb_timer(void *);
183static void igb_watchdog(struct ifnet *);
184static void igb_start(struct ifnet *);
d0f59cad
SZ		 185#ifdef IFPOLL_ENABLE
186static void igb_npoll(struct ifnet *, struct ifpoll_info *);
187static void igb_npoll_rx(struct ifnet *, void *, int);
188static void igb_npoll_tx(struct ifnet *, void *, int);
189static void igb_npoll_status(struct ifnet *, int);
1f7e3916 190#endif
7d235eb5
SZ		 191static void igb_serialize(struct ifnet *, enum ifnet_serialize);
192static void igb_deserialize(struct ifnet *, enum ifnet_serialize);
193static int igb_tryserialize(struct ifnet *, enum ifnet_serialize);
194#ifdef INVARIANTS
195static void igb_serialize_assert(struct ifnet *, enum ifnet_serialize,
196 boolean_t);
197#endif
1f7e3916
SZ		 198
199static void igb_intr(void *);
9c0ecdcc 200static void igb_intr_shared(void *);
1f7e3916
SZ		 201static void igb_rxeof(struct igb_rx_ring *, int);
202static void igb_txeof(struct igb_tx_ring *);
9c0ecdcc 203static void igb_set_eitr(struct igb_softc *, int, int);
1f7e3916
SZ		 204static void igb_enable_intr(struct igb_softc *);
205static void igb_disable_intr(struct igb_softc *);
f6167a56
SZ		 206static void igb_init_unshared_intr(struct igb_softc *);
207static void igb_init_intr(struct igb_softc *);
208static int igb_setup_intr(struct igb_softc *);
9c0ecdcc
SZ		 209static void igb_set_txintr_mask(struct igb_tx_ring *, int *, int);
210static void igb_set_rxintr_mask(struct igb_rx_ring *, int *, int);
be922da6 211static void igb_set_intr_mask(struct igb_softc *);
3c7cc5e2
SZ		 212static int igb_alloc_intr(struct igb_softc *);
213static void igb_free_intr(struct igb_softc *);
9c0ecdcc
SZ		 214static void igb_teardown_intr(struct igb_softc *);
215static void igb_msix_try_alloc(struct igb_softc *);
216static void igb_msix_free(struct igb_softc *, boolean_t);
217static int igb_msix_setup(struct igb_softc *);
218static void igb_msix_teardown(struct igb_softc *, int);
219static void igb_msix_rx(void *);
220static void igb_msix_tx(void *);
221static void igb_msix_status(void *);
1f7e3916
SZ		 222
223/* Management and WOL Support */
224static void igb_get_mgmt(struct igb_softc *);
225static void igb_rel_mgmt(struct igb_softc *);
226static void igb_get_hw_control(struct igb_softc *);
73a4e8ed
SZ		 227static void igb_rel_hw_control(struct igb_softc *);
228static void igb_enable_wol(device_t);
1f7e3916
SZ		 229
230static device_method_t igb_methods[] = {
231 /* Device interface */
232 DEVMETHOD(device_probe, igb_probe),
233 DEVMETHOD(device_attach, igb_attach),
234 DEVMETHOD(device_detach, igb_detach),
235 DEVMETHOD(device_shutdown, igb_shutdown),
236 DEVMETHOD(device_suspend, igb_suspend),
237 DEVMETHOD(device_resume, igb_resume),
238 { 0, 0 }
239};
240
241static driver_t igb_driver = {
242 "igb",
243 igb_methods,
244 sizeof(struct igb_softc),
245};
246
247static devclass_t igb_devclass;
248
249DECLARE_DUMMY_MODULE(if_igb);
250MODULE_DEPEND(igb, ig_hal, 1, 1, 1);
251DRIVER_MODULE(if_igb, pci, igb_driver, igb_devclass, NULL, NULL);
252
253static int igb_rxd = IGB_DEFAULT_RXD;
254static int igb_txd = IGB_DEFAULT_TXD;
8d6600da 255static int igb_rxr = 0;
1f7e3916
SZ		 256static int igb_msi_enable = 1;
257static int igb_msix_enable = 1;
258static int igb_eee_disabled = 1; /* Energy Efficient Ethernet */
259static int igb_fc_setting = e1000_fc_full;
260
261/*
262 * DMA Coalescing, only for i350 - default to off,
263 * this feature is for power savings
264 */
265static int igb_dma_coalesce = 0;
266
267TUNABLE_INT("hw.igb.rxd", &igb_rxd);
268TUNABLE_INT("hw.igb.txd", &igb_txd);
8d6600da 269TUNABLE_INT("hw.igb.rxr", &igb_rxr);
1f7e3916
SZ		 270TUNABLE_INT("hw.igb.msi.enable", &igb_msi_enable);
271TUNABLE_INT("hw.igb.msix.enable", &igb_msix_enable);
272TUNABLE_INT("hw.igb.fc_setting", &igb_fc_setting);
273
274/* i350 specific */
275TUNABLE_INT("hw.igb.eee_disabled", &igb_eee_disabled);
276TUNABLE_INT("hw.igb.dma_coalesce", &igb_dma_coalesce);
277
278static __inline void
279igb_rxcsum(uint32_t staterr, struct mbuf *mp)
280{
281 /* Ignore Checksum bit is set */
282 if (staterr & E1000_RXD_STAT_IXSM)
283 return;
284
285 if ((staterr & (E1000_RXD_STAT_IPCS | E1000_RXDEXT_STATERR_IPE)) ==
286 E1000_RXD_STAT_IPCS)
287 mp->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
288
289 if (staterr & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)) {
290 if ((staterr & E1000_RXDEXT_STATERR_TCPE) == 0) {
291 mp->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
292 CSUM_PSEUDO_HDR | CSUM_FRAG_NOT_CHECKED;
293 mp->m_pkthdr.csum_data = htons(0xffff);
294 }
295 }
296}
297
8d6600da
SZ		 298static __inline struct pktinfo *
299igb_rssinfo(struct mbuf *m, struct pktinfo *pi,
300 uint32_t hash, uint32_t hashtype, uint32_t staterr)
301{
302 switch (hashtype) {
303 case E1000_RXDADV_RSSTYPE_IPV4_TCP:
304 pi->pi_netisr = NETISR_IP;
305 pi->pi_flags = 0;
306 pi->pi_l3proto = IPPROTO_TCP;
307 break;
308
309 case E1000_RXDADV_RSSTYPE_IPV4:
310 if (staterr & E1000_RXD_STAT_IXSM)
311 return NULL;
312
313 if ((staterr &
314 (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) ==
315 E1000_RXD_STAT_TCPCS) {
316 pi->pi_netisr = NETISR_IP;
317 pi->pi_flags = 0;
318 pi->pi_l3proto = IPPROTO_UDP;
319 break;
320 }
321 /* FALL THROUGH */
322 default:
323 return NULL;
324 }
325
326 m->m_flags |= M_HASH;
327 m->m_pkthdr.hash = toeplitz_hash(hash);
328 return pi;
329}
330
1f7e3916
SZ		 331static int
332igb_probe(device_t dev)
333{
334 const struct igb_device *d;
335 uint16_t vid, did;
336
337 vid = pci_get_vendor(dev);
338 did = pci_get_device(dev);
339
340 for (d = igb_devices; d->desc != NULL; ++d) {
341 if (vid == d->vid && did == d->did) {
342 device_set_desc(dev, d->desc);
343 return 0;
344 }
345 }
346 return ENXIO;
347}
348
349static int
350igb_attach(device_t dev)
351{
352 struct igb_softc *sc = device_get_softc(dev);
1f7e3916 353 uint16_t eeprom_data;
8d6600da 354 int error = 0, i, j, ring_max;
d0f59cad
SZ		 355#ifdef IFPOLL_ENABLE
356 int offset, offset_def;
357#endif
1f7e3916
SZ		 358
359#ifdef notyet
360 /* SYSCTL stuff */
361 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
362 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
363 OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
364 igb_sysctl_nvm_info, "I", "NVM Information");
365
366 SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
367 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
368 OID_AUTO, "enable_aim", CTLTYPE_INT|CTLFLAG_RW,
369 &igb_enable_aim, 1, "Interrupt Moderation");
370
371 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
372 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
373 OID_AUTO, "flow_control", CTLTYPE_INT|CTLFLAG_RW,
374 adapter, 0, igb_set_flowcntl, "I", "Flow Control");
375#endif
376
377 callout_init_mp(&sc->timer);
9c0ecdcc 378 lwkt_serialize_init(&sc->main_serialize);
1f7e3916 379
62be5890
SZ		 380 if_initname(&sc->arpcom.ac_if, device_get_name(dev),
381 device_get_unit(dev));
1f7e3916
SZ		 382 sc->dev = sc->osdep.dev = dev;
383
384 /*
385 * Determine hardware and mac type
386 */
387 sc->hw.vendor_id = pci_get_vendor(dev);
388 sc->hw.device_id = pci_get_device(dev);
389 sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
390 sc->hw.subsystem_vendor_id = pci_read_config(dev, PCIR_SUBVEND_0, 2);
391 sc->hw.subsystem_device_id = pci_read_config(dev, PCIR_SUBDEV_0, 2);
392
393 if (e1000_set_mac_type(&sc->hw))
394 return ENXIO;
395
396 /* Are we a VF device? */
397 if (sc->hw.mac.type == e1000_vfadapt ||
398 sc->hw.mac.type == e1000_vfadapt_i350)
399 sc->vf_ifp = 1;
400 else
401 sc->vf_ifp = 0;
402
9b7aa975
SZ		 403 /*
404 * Configure total supported RX/TX ring count
405 */
406 switch (sc->hw.mac.type) {
407 case e1000_82575:
408 ring_max = IGB_MAX_RING_82575;
409 break;
410 case e1000_82580:
411 ring_max = IGB_MAX_RING_82580;
412 break;
413 case e1000_i350:
414 ring_max = IGB_MAX_RING_I350;
415 break;
416 case e1000_82576:
417 ring_max = IGB_MAX_RING_82576;
418 break;
419 default:
420 ring_max = IGB_MIN_RING;
421 break;
422 }
423 sc->rx_ring_cnt = device_getenv_int(dev, "rxr", igb_rxr);
424 sc->rx_ring_cnt = if_ring_count2(sc->rx_ring_cnt, ring_max);
425#ifdef IGB_RSS_DEBUG
426 sc->rx_ring_cnt = device_getenv_int(dev, "rxr_debug", sc->rx_ring_cnt);
427#endif
428 sc->rx_ring_inuse = sc->rx_ring_cnt;
429 sc->tx_ring_cnt = 1; /* XXX */
430
23f6ffe4
SZ		 431 if (sc->hw.mac.type == e1000_82575)
432 sc->flags |= IGB_FLAG_TSO_IPLEN0;
433
1f7e3916
SZ		 434 /* Enable bus mastering */
435 pci_enable_busmaster(dev);
436
437 /*
438 * Allocate IO memory
439 */
440 sc->mem_rid = PCIR_BAR(0);
441 sc->mem_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->mem_rid,
442 RF_ACTIVE);
443 if (sc->mem_res == NULL) {
444 device_printf(dev, "Unable to allocate bus resource: memory\n");
445 error = ENXIO;
446 goto failed;
447 }
448 sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->mem_res);
449 sc->osdep.mem_bus_space_handle = rman_get_bushandle(sc->mem_res);
450
451 sc->hw.hw_addr = (uint8_t *)&sc->osdep.mem_bus_space_handle;
452
1f7e3916
SZ		 453 /* Save PCI command register for Shared Code */
454 sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
455 sc->hw.back = &sc->osdep;
456
1f7e3916
SZ		 457 /* Do Shared Code initialization */
458 if (e1000_setup_init_funcs(&sc->hw, TRUE)) {
459 device_printf(dev, "Setup of Shared code failed\n");
460 error = ENXIO;
461 goto failed;
462 }
463
464 e1000_get_bus_info(&sc->hw);
465
466 sc->hw.mac.autoneg = DO_AUTO_NEG;
467 sc->hw.phy.autoneg_wait_to_complete = FALSE;
468 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
469
470 /* Copper options */
471 if (sc->hw.phy.media_type == e1000_media_type_copper) {
472 sc->hw.phy.mdix = AUTO_ALL_MODES;
473 sc->hw.phy.disable_polarity_correction = FALSE;
474 sc->hw.phy.ms_type = IGB_MASTER_SLAVE;
475 }
476
477 /* Set the frame limits assuming standard ethernet sized frames. */
478 sc->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN;
1f7e3916 479
a619b256
SZ		 480 /* Allocate RX/TX rings */
481 error = igb_alloc_rings(sc);
1f7e3916
SZ		 482 if (error)
483 goto failed;
484
d0f59cad
SZ		 485#ifdef IFPOLL_ENABLE
486 /*
487 * NPOLLING RX CPU offset
488 */
489 if (sc->rx_ring_cnt == ncpus2) {
490 offset = 0;
491 } else {
492 offset_def = (sc->rx_ring_cnt * device_get_unit(dev)) % ncpus2;
493 offset = device_getenv_int(dev, "npoll.rxoff", offset_def);
494 if (offset >= ncpus2 ||
495 offset % sc->rx_ring_cnt != 0) {
496 device_printf(dev, "invalid npoll.rxoff %d, use %d\n",
497 offset, offset_def);
498 offset = offset_def;
499 }
500 }
501 sc->rx_npoll_off = offset;
502
503 /*
504 * NPOLLING TX CPU offset
505 */
506 offset_def = sc->rx_npoll_off;
507 offset = device_getenv_int(dev, "npoll.txoff", offset_def);
508 if (offset >= ncpus2) {
509 device_printf(dev, "invalid npoll.txoff %d, use %d\n",
510 offset, offset_def);
511 offset = offset_def;
512 }
513 sc->tx_npoll_off = offset;
514#endif
515
3c7cc5e2
SZ		 516 /* Allocate interrupt */
517 error = igb_alloc_intr(sc);
518 if (error)
a1647e40 519 goto failed;
a1647e40
SZ		 520
521 /*
7d235eb5
SZ		 522 * Setup serializers
523 */
7d235eb5
SZ		 524 i = 0;
525 sc->serializes[i++] = &sc->main_serialize;
526
527 sc->tx_serialize = i;
528 for (j = 0; j < sc->tx_ring_cnt; ++j)
529 sc->serializes[i++] = &sc->tx_rings[j].tx_serialize;
530
531 sc->rx_serialize = i;
532 for (j = 0; j < sc->rx_ring_cnt; ++j)
533 sc->serializes[i++] = &sc->rx_rings[j].rx_serialize;
534
535 sc->serialize_cnt = i;
536 KKASSERT(sc->serialize_cnt <= IGB_NSERIALIZE);
537
1f7e3916
SZ		 538 /* Allocate the appropriate stats memory */
539 if (sc->vf_ifp) {
540 sc->stats = kmalloc(sizeof(struct e1000_vf_stats), M_DEVBUF,
541 M_WAITOK | M_ZERO);
542 igb_vf_init_stats(sc);
543 } else {
544 sc->stats = kmalloc(sizeof(struct e1000_hw_stats), M_DEVBUF,
545 M_WAITOK | M_ZERO);
546 }
547
548 /* Allocate multicast array memory. */
549 sc->mta = kmalloc(ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES,
550 M_DEVBUF, M_WAITOK);
551
552 /* Some adapter-specific advanced features */
553 if (sc->hw.mac.type >= e1000_i350) {
554#ifdef notyet
555 igb_set_sysctl_value(adapter, "dma_coalesce",
556 "configure dma coalesce",
557 &adapter->dma_coalesce, igb_dma_coalesce);
558 igb_set_sysctl_value(adapter, "eee_disabled",
559 "enable Energy Efficient Ethernet",
560 &adapter->hw.dev_spec._82575.eee_disable,
561 igb_eee_disabled);
562#else
563 sc->dma_coalesce = igb_dma_coalesce;
564 sc->hw.dev_spec._82575.eee_disable = igb_eee_disabled;
565#endif
566 e1000_set_eee_i350(&sc->hw);
567 }
568
569 /*
570 * Start from a known state, this is important in reading the nvm and
571 * mac from that.
572 */
573 e1000_reset_hw(&sc->hw);
574
575 /* Make sure we have a good EEPROM before we read from it */
576 if (e1000_validate_nvm_checksum(&sc->hw) < 0) {
577 /*
578 * Some PCI-E parts fail the first check due to
579 * the link being in sleep state, call it again,
580 * if it fails a second time its a real issue.
581 */
582 if (e1000_validate_nvm_checksum(&sc->hw) < 0) {
583 device_printf(dev,
584 "The EEPROM Checksum Is Not Valid\n");
585 error = EIO;
586 goto failed;
587 }
588 }
589
590 /* Copy the permanent MAC address out of the EEPROM */
591 if (e1000_read_mac_addr(&sc->hw) < 0) {
592 device_printf(dev, "EEPROM read error while reading MAC"
593 " address\n");
594 error = EIO;
595 goto failed;
596 }
597 if (!igb_is_valid_ether_addr(sc->hw.mac.addr)) {
598 device_printf(dev, "Invalid MAC address\n");
599 error = EIO;
600 goto failed;
601 }
602
603#ifdef notyet
604 /*
605 ** Configure Interrupts
606 */
607 if ((adapter->msix > 1) && (igb_enable_msix))
608 error = igb_allocate_msix(adapter);
609 else /* MSI or Legacy */
610 error = igb_allocate_legacy(adapter);
611 if (error)
612 goto err_late;
613#endif
614
615 /* Setup OS specific network interface */
616 igb_setup_ifp(sc);
617
618 /* Add sysctl tree, must after igb_setup_ifp() */
619 igb_add_sysctl(sc);
620
621 /* Now get a good starting state */
622 igb_reset(sc);
623
624 /* Initialize statistics */
625 igb_update_stats_counters(sc);
626
627 sc->hw.mac.get_link_status = 1;
628 igb_update_link_status(sc);
629
630 /* Indicate SOL/IDER usage */
631 if (e1000_check_reset_block(&sc->hw)) {
632 device_printf(dev,
633 "PHY reset is blocked due to SOL/IDER session.\n");
634 }
635
636 /* Determine if we have to control management hardware */
396b7048
SZ		 637 if (e1000_enable_mng_pass_thru(&sc->hw))
638 sc->flags |= IGB_FLAG_HAS_MGMT;
1f7e3916
SZ		 639
640 /*
641 * Setup Wake-on-Lan
642 */
643 /* APME bit in EEPROM is mapped to WUC.APME */
644 eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC) & E1000_WUC_APME;
645 if (eeprom_data)
646 sc->wol = E1000_WUFC_MAG;
647 /* XXX disable WOL */
648 sc->wol = 0;
649
650#ifdef notyet
651 /* Register for VLAN events */
652 adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
653 igb_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
654 adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
655 igb_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
656#endif
657
658#ifdef notyet
659 igb_add_hw_stats(adapter);
660#endif
661
f6167a56 662 error = igb_setup_intr(sc);
1f7e3916 663 if (error) {
1f7e3916
SZ		 664 ether_ifdetach(&sc->arpcom.ac_if);
665 goto failed;
666 }
1f7e3916
SZ		 667 return 0;
668
669failed:
670 igb_detach(dev);
671 return error;
672}
673
674static int
675igb_detach(device_t dev)
676{
677 struct igb_softc *sc = device_get_softc(dev);
678
679 if (device_is_attached(dev)) {
680 struct ifnet *ifp = &sc->arpcom.ac_if;
681
682 ifnet_serialize_all(ifp);
683
684 igb_stop(sc);
685
686 e1000_phy_hw_reset(&sc->hw);
687
688 /* Give control back to firmware */
689 igb_rel_mgmt(sc);
690 igb_rel_hw_control(sc);
691
692 if (sc->wol) {
693 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
694 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
695 igb_enable_wol(dev);
696 }
697
9c0ecdcc 698 igb_teardown_intr(sc);
1f7e3916
SZ		 699
700 ifnet_deserialize_all(ifp);
701
702 ether_ifdetach(ifp);
703 } else if (sc->mem_res != NULL) {
704 igb_rel_hw_control(sc);
705 }
706 bus_generic_detach(dev);
707
9c0ecdcc
SZ		 708 if (sc->sysctl_tree != NULL)
709 sysctl_ctx_free(&sc->sysctl_ctx);
710
3c7cc5e2 711 igb_free_intr(sc);
1f7e3916 712
9c0ecdcc
SZ		 713 if (sc->msix_mem_res != NULL) {
714 bus_release_resource(dev, SYS_RES_MEMORY, sc->msix_mem_rid,
715 sc->msix_mem_res);
716 }
1f7e3916
SZ		 717 if (sc->mem_res != NULL) {
718 bus_release_resource(dev, SYS_RES_MEMORY, sc->mem_rid,
719 sc->mem_res);
720 }
721
a619b256 722 igb_free_rings(sc);
1f7e3916
SZ		 723
724 if (sc->mta != NULL)
725 kfree(sc->mta, M_DEVBUF);
726 if (sc->stats != NULL)
727 kfree(sc->stats, M_DEVBUF);
728
1f7e3916
SZ		 729 return 0;
730}
731
732static int
733igb_shutdown(device_t dev)
734{
735 return igb_suspend(dev);
736}
737
738static int
739igb_suspend(device_t dev)
740{
741 struct igb_softc *sc = device_get_softc(dev);
742 struct ifnet *ifp = &sc->arpcom.ac_if;
743
744 ifnet_serialize_all(ifp);
745
746 igb_stop(sc);
747
748 igb_rel_mgmt(sc);
749 igb_rel_hw_control(sc);
750
751 if (sc->wol) {
752 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
753 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
754 igb_enable_wol(dev);
755 }
756
757 ifnet_deserialize_all(ifp);
758
759 return bus_generic_suspend(dev);
760}
761
762static int
763igb_resume(device_t dev)
764{
765 struct igb_softc *sc = device_get_softc(dev);
766 struct ifnet *ifp = &sc->arpcom.ac_if;
767
768 ifnet_serialize_all(ifp);
769
770 igb_init(sc);
771 igb_get_mgmt(sc);
772
773 if_devstart(ifp);
774
775 ifnet_deserialize_all(ifp);
776
777 return bus_generic_resume(dev);
778}
779
780static int
781igb_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr)
782{
783 struct igb_softc *sc = ifp->if_softc;
784 struct ifreq *ifr = (struct ifreq *)data;
785 int max_frame_size, mask, reinit;
786 int error = 0;
787
788 ASSERT_IFNET_SERIALIZED_ALL(ifp);
789
790 switch (command) {
791 case SIOCSIFMTU:
792 max_frame_size = 9234;
793 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
794 ETHER_CRC_LEN) {
795 error = EINVAL;
796 break;
797 }
798
799 ifp->if_mtu = ifr->ifr_mtu;
800 sc->max_frame_size = ifp->if_mtu + ETHER_HDR_LEN +
801 ETHER_CRC_LEN;
802
803 if (ifp->if_flags & IFF_RUNNING)
804 igb_init(sc);
805 break;
806
807 case SIOCSIFFLAGS:
808 if (ifp->if_flags & IFF_UP) {
809 if (ifp->if_flags & IFF_RUNNING) {
810 if ((ifp->if_flags ^ sc->if_flags) &
811 (IFF_PROMISC | IFF_ALLMULTI)) {
812 igb_disable_promisc(sc);
813 igb_set_promisc(sc);
814 }
815 } else {
816 igb_init(sc);
817 }
818 } else if (ifp->if_flags & IFF_RUNNING) {
819 igb_stop(sc);
820 }
821 sc->if_flags = ifp->if_flags;
822 break;
823
824 case SIOCADDMULTI:
825 case SIOCDELMULTI:
826 if (ifp->if_flags & IFF_RUNNING) {
827 igb_disable_intr(sc);
828 igb_set_multi(sc);
d0f59cad
SZ		 829#ifdef IFPOLL_ENABLE
830 if (!(ifp->if_flags & IFF_NPOLLING))
1f7e3916
SZ		 831#endif
832 igb_enable_intr(sc);
833 }
834 break;
835
836 case SIOCSIFMEDIA:
837 /*
838 * As the speed/duplex settings are being
839 * changed, we need toreset the PHY.
840 */
841 sc->hw.phy.reset_disable = FALSE;
842
843 /* Check SOL/IDER usage */
844 if (e1000_check_reset_block(&sc->hw)) {
845 if_printf(ifp, "Media change is "
846 "blocked due to SOL/IDER session.\n");
847 break;
848 }
849 /* FALL THROUGH */
850
851 case SIOCGIFMEDIA:
852 error = ifmedia_ioctl(ifp, ifr, &sc->media, command);
853 break;
854
855 case SIOCSIFCAP:
856 reinit = 0;
857 mask = ifr->ifr_reqcap ^ ifp->if_capenable;
77d8cab9
SZ		 858 if (mask & IFCAP_RXCSUM) {
859 ifp->if_capenable ^= IFCAP_RXCSUM;
1f7e3916
SZ		 860 reinit = 1;
861 }
862 if (mask & IFCAP_VLAN_HWTAGGING) {
863 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
864 reinit = 1;
865 }
77d8cab9
SZ		 866 if (mask & IFCAP_TXCSUM) {
867 ifp->if_capenable ^= IFCAP_TXCSUM;
868 if (ifp->if_capenable & IFCAP_TXCSUM)
869 ifp->if_hwassist |= IGB_CSUM_FEATURES;
870 else
871 ifp->if_hwassist &= ~IGB_CSUM_FEATURES;
872 }
23f6ffe4
SZ		 873 if (mask & IFCAP_TSO) {
874 ifp->if_capenable ^= IFCAP_TSO;
875 if (ifp->if_capenable & IFCAP_TSO)
876 ifp->if_hwassist |= CSUM_TSO;
877 else
878 ifp->if_hwassist &= ~CSUM_TSO;
879 }
8d6600da
SZ		 880 if (mask & IFCAP_RSS)
881 ifp->if_capenable ^= IFCAP_RSS;
1f7e3916
SZ		 882 if (reinit && (ifp->if_flags & IFF_RUNNING))
883 igb_init(sc);
884 break;
885
886 default:
887 error = ether_ioctl(ifp, command, data);
888 break;
889 }
890 return error;
891}
892
893static void
894igb_init(void *xsc)
895{
896 struct igb_softc *sc = xsc;
897 struct ifnet *ifp = &sc->arpcom.ac_if;
9c0ecdcc 898 boolean_t polling;
1f7e3916
SZ		 899 int i;
900
901 ASSERT_IFNET_SERIALIZED_ALL(ifp);
902
903 igb_stop(sc);
904
905 /* Get the latest mac address, User can use a LAA */
906 bcopy(IF_LLADDR(ifp), sc->hw.mac.addr, ETHER_ADDR_LEN);
907
908 /* Put the address into the Receive Address Array */
909 e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
910
911 igb_reset(sc);
912 igb_update_link_status(sc);
913
914 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
915
1f7e3916
SZ		 916 /* Configure for OS presence */
917 igb_get_mgmt(sc);
918
9c0ecdcc 919 polling = FALSE;
d0f59cad
SZ		 920#ifdef IFPOLL_ENABLE
921 if (ifp->if_flags & IFF_NPOLLING)
9c0ecdcc 922 polling = TRUE;
be922da6 923#endif
9c0ecdcc
SZ		 924
925 /* Configured used RX/TX rings */
926 igb_set_ring_inuse(sc, polling);
927
928 /* Initialize interrupt */
929 igb_init_intr(sc);
be922da6 930
1f7e3916 931 /* Prepare transmit descriptors and buffers */
27866bf1 932 for (i = 0; i < sc->tx_ring_cnt; ++i)
1f7e3916
SZ		 933 igb_init_tx_ring(&sc->tx_rings[i]);
934 igb_init_tx_unit(sc);
935
936 /* Setup Multicast table */
937 igb_set_multi(sc);
938
939#if 0
940 /*
941 * Figure out the desired mbuf pool
942 * for doing jumbo/packetsplit
943 */
944 if (adapter->max_frame_size <= 2048)
945 adapter->rx_mbuf_sz = MCLBYTES;
946 else if (adapter->max_frame_size <= 4096)
947 adapter->rx_mbuf_sz = MJUMPAGESIZE;
948 else
949 adapter->rx_mbuf_sz = MJUM9BYTES;
1f7e3916
SZ		 950#endif
951
952 /* Prepare receive descriptors and buffers */
be922da6 953 for (i = 0; i < sc->rx_ring_inuse; ++i) {
1f7e3916
SZ		 954 int error;
955
956 error = igb_init_rx_ring(&sc->rx_rings[i]);
957 if (error) {
958 if_printf(ifp, "Could not setup receive structures\n");
959 igb_stop(sc);
960 return;
961 }
962 }
963 igb_init_rx_unit(sc);
964
965 /* Enable VLAN support */
966 if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
967 igb_set_vlan(sc);
968
969 /* Don't lose promiscuous settings */
970 igb_set_promisc(sc);
971
1f7e3916
SZ		 972 ifp->if_flags |= IFF_RUNNING;
973 ifp->if_flags &= ~IFF_OACTIVE;
974
975 callout_reset(&sc->timer, hz, igb_timer, sc);
976 e1000_clear_hw_cntrs_base_generic(&sc->hw);
977
9c0ecdcc 978 /* This clears any pending interrupts */
1f7e3916 979 E1000_READ_REG(&sc->hw, E1000_ICR);
9c0ecdcc 980
1f7e3916
SZ		 981 /*
982 * Only enable interrupts if we are not polling, make sure
983 * they are off otherwise.
984 */
9c0ecdcc 985 if (polling) {
1f7e3916 986 igb_disable_intr(sc);
9c0ecdcc 987 } else {
1f7e3916
SZ		 988 igb_enable_intr(sc);
989 E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
990 }
991
992 /* Set Energy Efficient Ethernet */
993 e1000_set_eee_i350(&sc->hw);
994
995 /* Don't reset the phy next time init gets called */
996 sc->hw.phy.reset_disable = TRUE;
997}
998
999static void
1000igb_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
1001{
1002 struct igb_softc *sc = ifp->if_softc;
1003 u_char fiber_type = IFM_1000_SX;
1004
1005 ASSERT_IFNET_SERIALIZED_ALL(ifp);
1006
1007 igb_update_link_status(sc);
1008
1009 ifmr->ifm_status = IFM_AVALID;
1010 ifmr->ifm_active = IFM_ETHER;
1011
1012 if (!sc->link_active)
1013 return;
1014
1015 ifmr->ifm_status |= IFM_ACTIVE;
1016
1017 if (sc->hw.phy.media_type == e1000_media_type_fiber ||
1018 sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
1019 ifmr->ifm_active |= fiber_type | IFM_FDX;
1020 } else {
1021 switch (sc->link_speed) {
1022 case 10:
1023 ifmr->ifm_active |= IFM_10_T;
1024 break;
1025
1026 case 100:
1027 ifmr->ifm_active |= IFM_100_TX;
1028 break;
1029
1030 case 1000:
1031 ifmr->ifm_active |= IFM_1000_T;
1032 break;
1033 }
1034 if (sc->link_duplex == FULL_DUPLEX)
1035 ifmr->ifm_active |= IFM_FDX;
1036 else
1037 ifmr->ifm_active |= IFM_HDX;
1038 }
1039}
1040
1041static int
1042igb_media_change(struct ifnet *ifp)
1043{
1044 struct igb_softc *sc = ifp->if_softc;
1045 struct ifmedia *ifm = &sc->media;
1046
1047 ASSERT_IFNET_SERIALIZED_ALL(ifp);
1048
1049 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1050 return EINVAL;
1051
1052 switch (IFM_SUBTYPE(ifm->ifm_media)) {
1053 case IFM_AUTO:
1054 sc->hw.mac.autoneg = DO_AUTO_NEG;
1055 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1056 break;
1057
1058 case IFM_1000_LX:
1059 case IFM_1000_SX:
1060 case IFM_1000_T:
1061 sc->hw.mac.autoneg = DO_AUTO_NEG;
1062 sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1063 break;
1064
1065 case IFM_100_TX:
1066 sc->hw.mac.autoneg = FALSE;
1067 sc->hw.phy.autoneg_advertised = 0;
1068 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1069 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1070 else
1071 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1072 break;
1073
1074 case IFM_10_T:
1075 sc->hw.mac.autoneg = FALSE;
1076 sc->hw.phy.autoneg_advertised = 0;
1077 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1078 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1079 else
1080 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1081 break;
1082
1083 default:
1084 if_printf(ifp, "Unsupported media type\n");
1085 break;
1086 }
1087
1088 igb_init(sc);
1089
1090 return 0;
1091}
1092
1093static void
1094igb_set_promisc(struct igb_softc *sc)
1095{
1096 struct ifnet *ifp = &sc->arpcom.ac_if;
1097 struct e1000_hw *hw = &sc->hw;
1098 uint32_t reg;
1099
1100 if (sc->vf_ifp) {
1101 e1000_promisc_set_vf(hw, e1000_promisc_enabled);
1102 return;
1103 }
1104
1105 reg = E1000_READ_REG(hw, E1000_RCTL);
1106 if (ifp->if_flags & IFF_PROMISC) {
1107 reg |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1108 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1109 } else if (ifp->if_flags & IFF_ALLMULTI) {
1110 reg |= E1000_RCTL_MPE;
1111 reg &= ~E1000_RCTL_UPE;
1112 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1113 }
1114}
1115
1116static void
1117igb_disable_promisc(struct igb_softc *sc)
1118{
1119 struct e1000_hw *hw = &sc->hw;
1120 uint32_t reg;
1121
1122 if (sc->vf_ifp) {
1123 e1000_promisc_set_vf(hw, e1000_promisc_disabled);
1124 return;
1125 }
1126 reg = E1000_READ_REG(hw, E1000_RCTL);
1127 reg &= ~E1000_RCTL_UPE;
1128 reg &= ~E1000_RCTL_MPE;
1129 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1130}
1131
1132static void
1133igb_set_multi(struct igb_softc *sc)
1134{
1135 struct ifnet *ifp = &sc->arpcom.ac_if;
1136 struct ifmultiaddr *ifma;
1137 uint32_t reg_rctl = 0;
1138 uint8_t *mta;
1139 int mcnt = 0;
1140
1141 mta = sc->mta;
1142 bzero(mta, ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1143
1144 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1145 if (ifma->ifma_addr->sa_family != AF_LINK)
1146 continue;
1147
1148 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
1149 break;
1150
1151 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
1152 &mta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
1153 mcnt++;
1154 }
1155
1156 if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1157 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1158 reg_rctl |= E1000_RCTL_MPE;
1159 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1160 } else {
1161 e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1162 }
1163}
1164
1165static void
1166igb_timer(void *xsc)
1167{
1168 struct igb_softc *sc = xsc;
1f7e3916 1169
27dd00d6 1170 lwkt_serialize_enter(&sc->main_serialize);
1f7e3916
SZ		 1171
1172 igb_update_link_status(sc);
1173 igb_update_stats_counters(sc);
1174
1175 callout_reset(&sc->timer, hz, igb_timer, sc);
1176
27dd00d6 1177 lwkt_serialize_exit(&sc->main_serialize);
1f7e3916
SZ		 1178}
1179
1180static void
1181igb_update_link_status(struct igb_softc *sc)
1182{
1183 struct ifnet *ifp = &sc->arpcom.ac_if;
1184 struct e1000_hw *hw = &sc->hw;
1185 uint32_t link_check, thstat, ctrl;
1186
1187 link_check = thstat = ctrl = 0;
1188
1189 /* Get the cached link value or read for real */
1190 switch (hw->phy.media_type) {
1191 case e1000_media_type_copper:
1192 if (hw->mac.get_link_status) {
1193 /* Do the work to read phy */
1194 e1000_check_for_link(hw);
1195 link_check = !hw->mac.get_link_status;
1196 } else {
1197 link_check = TRUE;
1198 }
1199 break;
1200
1201 case e1000_media_type_fiber:
1202 e1000_check_for_link(hw);
1203 link_check = E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU;
1204 break;
1205
1206 case e1000_media_type_internal_serdes:
1207 e1000_check_for_link(hw);
1208 link_check = hw->mac.serdes_has_link;
1209 break;
1210
1211 /* VF device is type_unknown */
1212 case e1000_media_type_unknown:
1213 e1000_check_for_link(hw);
1214 link_check = !hw->mac.get_link_status;
1215 /* Fall thru */
1216 default:
1217 break;
1218 }
1219
1220 /* Check for thermal downshift or shutdown */
1221 if (hw->mac.type == e1000_i350) {
1222 thstat = E1000_READ_REG(hw, E1000_THSTAT);
1223 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1224 }
1225
1226 /* Now we check if a transition has happened */
1227 if (link_check && sc->link_active == 0) {
1228 e1000_get_speed_and_duplex(hw,
1229 &sc->link_speed, &sc->link_duplex);
1230 if (bootverbose) {
1231 if_printf(ifp, "Link is up %d Mbps %s\n",
1232 sc->link_speed,
1233 sc->link_duplex == FULL_DUPLEX ?
1234 "Full Duplex" : "Half Duplex");
1235 }
1236 sc->link_active = 1;
1237
1238 ifp->if_baudrate = sc->link_speed * 1000000;
1239 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1240 (thstat & E1000_THSTAT_LINK_THROTTLE))
1241 if_printf(ifp, "Link: thermal downshift\n");
1242 /* This can sleep */
1243 ifp->if_link_state = LINK_STATE_UP;
1244 if_link_state_change(ifp);
1245 } else if (!link_check && sc->link_active == 1) {
1246 ifp->if_baudrate = sc->link_speed = 0;
1247 sc->link_duplex = 0;
1248 if (bootverbose)
1249 if_printf(ifp, "Link is Down\n");
1250 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1251 (thstat & E1000_THSTAT_PWR_DOWN))
1252 if_printf(ifp, "Link: thermal shutdown\n");
1253 sc->link_active = 0;
1254 /* This can sleep */
1255 ifp->if_link_state = LINK_STATE_DOWN;
1256 if_link_state_change(ifp);
1257 }
1258}
1259
1260static void
1261igb_stop(struct igb_softc *sc)
1262{
1263 struct ifnet *ifp = &sc->arpcom.ac_if;
1264 int i;
1265
1266 ASSERT_IFNET_SERIALIZED_ALL(ifp);
1267
1268 igb_disable_intr(sc);
1269
1270 callout_stop(&sc->timer);
1271
1272 ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
1273 ifp->if_timer = 0;
1274
1275 e1000_reset_hw(&sc->hw);
1276 E1000_WRITE_REG(&sc->hw, E1000_WUC, 0);
1277
1278 e1000_led_off(&sc->hw);
1279 e1000_cleanup_led(&sc->hw);
1280
27866bf1 1281 for (i = 0; i < sc->tx_ring_cnt; ++i)
1f7e3916 1282 igb_free_tx_ring(&sc->tx_rings[i]);
27866bf1 1283 for (i = 0; i < sc->rx_ring_cnt; ++i)
1f7e3916
SZ		 1284 igb_free_rx_ring(&sc->rx_rings[i]);
1285}
1286
1287static void
1288igb_reset(struct igb_softc *sc)
1289{
1290 struct ifnet *ifp = &sc->arpcom.ac_if;
1291 struct e1000_hw *hw = &sc->hw;
1292 struct e1000_fc_info *fc = &hw->fc;
1293 uint32_t pba = 0;
1294 uint16_t hwm;
1295
1296 /* Let the firmware know the OS is in control */
1297 igb_get_hw_control(sc);
1298
1299 /*
1300 * Packet Buffer Allocation (PBA)
1301 * Writing PBA sets the receive portion of the buffer
1302 * the remainder is used for the transmit buffer.
1303 */
1304 switch (hw->mac.type) {
1305 case e1000_82575:
1306 pba = E1000_PBA_32K;
1307 break;
1308
1309 case e1000_82576:
1310 case e1000_vfadapt:
1311 pba = E1000_READ_REG(hw, E1000_RXPBS);
1312 pba &= E1000_RXPBS_SIZE_MASK_82576;
1313 break;
1314
1315 case e1000_82580:
1316 case e1000_i350:
1317 case e1000_vfadapt_i350:
1318 pba = E1000_READ_REG(hw, E1000_RXPBS);
1319 pba = e1000_rxpbs_adjust_82580(pba);
1320 break;
1321 /* XXX pba = E1000_PBA_35K; */
1322
1323 default:
1324 break;
1325 }
1326
1327 /* Special needs in case of Jumbo frames */
1328 if (hw->mac.type == e1000_82575 && ifp->if_mtu > ETHERMTU) {
1329 uint32_t tx_space, min_tx, min_rx;
1330
1331 pba = E1000_READ_REG(hw, E1000_PBA);
1332 tx_space = pba >> 16;
1333 pba &= 0xffff;
1334
1335 min_tx = (sc->max_frame_size +
1336 sizeof(struct e1000_tx_desc) - ETHER_CRC_LEN) * 2;
1337 min_tx = roundup2(min_tx, 1024);
1338 min_tx >>= 10;
1339 min_rx = sc->max_frame_size;
1340 min_rx = roundup2(min_rx, 1024);
1341 min_rx >>= 10;
1342 if (tx_space < min_tx && (min_tx - tx_space) < pba) {
1343 pba = pba - (min_tx - tx_space);
1344 /*
1345 * if short on rx space, rx wins
1346 * and must trump tx adjustment
1347 */
1348 if (pba < min_rx)
1349 pba = min_rx;
1350 }
1351 E1000_WRITE_REG(hw, E1000_PBA, pba);
1352 }
1353
1354 /*
1355 * These parameters control the automatic generation (Tx) and
1356 * response (Rx) to Ethernet PAUSE frames.
1357 * - High water mark should allow for at least two frames to be
1358 * received after sending an XOFF.
1359 * - Low water mark works best when it is very near the high water mark.
1360 * This allows the receiver to restart by sending XON when it has
1361 * drained a bit.
1362 */
1363 hwm = min(((pba << 10) * 9 / 10),
1364 ((pba << 10) - 2 * sc->max_frame_size));
1365
1366 if (hw->mac.type < e1000_82576) {
1367 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */
1368 fc->low_water = fc->high_water - 8;
1369 } else {
1370 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */
1371 fc->low_water = fc->high_water - 16;
1372 }
1373 fc->pause_time = IGB_FC_PAUSE_TIME;
1374 fc->send_xon = TRUE;
1375
1376 /* Issue a global reset */
1377 e1000_reset_hw(hw);
1378 E1000_WRITE_REG(hw, E1000_WUC, 0);
1379
1380 if (e1000_init_hw(hw) < 0)
1381 if_printf(ifp, "Hardware Initialization Failed\n");
1382
1383 /* Setup DMA Coalescing */
1384 if (hw->mac.type == e1000_i350 && sc->dma_coalesce) {
1385 uint32_t reg;
1386
1387 hwm = (pba - 4) << 10;
1388 reg = ((pba - 6) << E1000_DMACR_DMACTHR_SHIFT)
1389 & E1000_DMACR_DMACTHR_MASK;
1390
1391 /* transition to L0x or L1 if available..*/
1392 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
1393
1394 /* timer = +-1000 usec in 32usec intervals */
1395 reg |= (1000 >> 5);
1396 E1000_WRITE_REG(hw, E1000_DMACR, reg);
1397
1398 /* No lower threshold */
1399 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
1400
1401 /* set hwm to PBA - 2 * max frame size */
1402 E1000_WRITE_REG(hw, E1000_FCRTC, hwm);
1403
1404 /* Set the interval before transition */
1405 reg = E1000_READ_REG(hw, E1000_DMCTLX);
1406 reg |= 0x800000FF; /* 255 usec */
1407 E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
1408
1409 /* free space in tx packet buffer to wake from DMA coal */
1410 E1000_WRITE_REG(hw, E1000_DMCTXTH,
1411 (20480 - (2 * sc->max_frame_size)) >> 6);
1412
1413 /* make low power state decision controlled by DMA coal */
1414 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
1415 E1000_WRITE_REG(hw, E1000_PCIEMISC,
1416 reg | E1000_PCIEMISC_LX_DECISION);
1417 if_printf(ifp, "DMA Coalescing enabled\n");
1418 }
1419
1420 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1421 e1000_get_phy_info(hw);
1422 e1000_check_for_link(hw);
1423}
1424
1425static void
1426igb_setup_ifp(struct igb_softc *sc)
1427{
1428 struct ifnet *ifp = &sc->arpcom.ac_if;
1429
1f7e3916
SZ		 1430 ifp->if_softc = sc;
1431 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
9c0ecdcc 1432 ifp->if_init = igb_init;
1f7e3916
SZ		 1433 ifp->if_ioctl = igb_ioctl;
1434 ifp->if_start = igb_start;
7d235eb5
SZ		 1435 ifp->if_serialize = igb_serialize;
1436 ifp->if_deserialize = igb_deserialize;
1437 ifp->if_tryserialize = igb_tryserialize;
1438#ifdef INVARIANTS
1439 ifp->if_serialize_assert = igb_serialize_assert;
1440#endif
d0f59cad
SZ		 1441#ifdef IFPOLL_ENABLE
1442 ifp->if_npoll = igb_npoll;
1f7e3916
SZ		 1443#endif
1444 ifp->if_watchdog = igb_watchdog;
1445
91b8700a 1446 ifq_set_maxlen(&ifp->if_snd, sc->tx_rings[0].num_tx_desc - 1);
1f7e3916
SZ		 1447 ifq_set_ready(&ifp->if_snd);
1448
1449 ether_ifattach(ifp, sc->hw.mac.addr, NULL);
1450
1451 ifp->if_capabilities =
23f6ffe4 1452 IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_TSO;
8d6600da
SZ		 1453 if (IGB_ENABLE_HWRSS(sc))
1454 ifp->if_capabilities |= IFCAP_RSS;
1f7e3916 1455 ifp->if_capenable = ifp->if_capabilities;
23f6ffe4 1456 ifp->if_hwassist = IGB_CSUM_FEATURES | CSUM_TSO;
1f7e3916
SZ		 1457
1458 /*
1459 * Tell the upper layer(s) we support long frames
1460 */
1461 ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
1462
1463 /*
1464 * Specify the media types supported by this adapter and register
1465 * callbacks to update media and link information
1466 */
1467 ifmedia_init(&sc->media, IFM_IMASK, igb_media_change, igb_media_status);
1468 if (sc->hw.phy.media_type == e1000_media_type_fiber ||
1469 sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
1470 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX | IFM_FDX,
1471 0, NULL);
1472 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
1473 } else {
1474 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
1475 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T | IFM_FDX,
1476 0, NULL);
1477 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1478 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
1479 0, NULL);
1480 if (sc->hw.phy.type != e1000_phy_ife) {
1481 ifmedia_add(&sc->media,
1482 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
1483 ifmedia_add(&sc->media,
1484 IFM_ETHER | IFM_1000_T, 0, NULL);
1485 }
1486 }
1487 ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1488 ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO);
1489}
1490
1491static void
1492igb_add_sysctl(struct igb_softc *sc)
1493{
9c0ecdcc 1494 char node[32];
8d6600da 1495 int i;
8d6600da 1496
1f7e3916
SZ		 1497 sysctl_ctx_init(&sc->sysctl_ctx);
1498 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
1499 SYSCTL_STATIC_CHILDREN(_hw), OID_AUTO,
1500 device_get_nameunit(sc->dev), CTLFLAG_RD, 0, "");
1501 if (sc->sysctl_tree == NULL) {
1502 device_printf(sc->dev, "can't add sysctl node\n");
1503 return;
1504 }
1505
1506 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
8d6600da
SZ		 1507 OID_AUTO, "rxr", CTLFLAG_RD, &sc->rx_ring_cnt, 0, "# of RX rings");
1508 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
be922da6
SZ		 1509 OID_AUTO, "rxr_inuse", CTLFLAG_RD, &sc->rx_ring_inuse, 0,
1510 "# of RX rings used");
1511 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
8d6600da
SZ		 1512 OID_AUTO, "rxd", CTLFLAG_RD, &sc->rx_rings[0].num_rx_desc, 0,
1513 "# of RX descs");
1f7e3916 1514 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
8d6600da
SZ		 1515 OID_AUTO, "txd", CTLFLAG_RD, &sc->tx_rings[0].num_tx_desc, 0,
1516 "# of TX descs");
1f7e3916 1517
9c0ecdcc
SZ		 1518 if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
1519 SYSCTL_ADD_PROC(&sc->sysctl_ctx,
1520 SYSCTL_CHILDREN(sc->sysctl_tree),
1521 OID_AUTO, "intr_rate", CTLTYPE_INT | CTLFLAG_RW,
1522 sc, 0, igb_sysctl_intr_rate, "I", "interrupt rate");
1523 } else {
1524 for (i = 0; i < sc->msix_cnt; ++i) {
1525 struct igb_msix_data *msix = &sc->msix_data[i];
1526
1527 ksnprintf(node, sizeof(node), "msix%d_rate", i);
1528 SYSCTL_ADD_PROC(&sc->sysctl_ctx,
1529 SYSCTL_CHILDREN(sc->sysctl_tree),
1530 OID_AUTO, node, CTLTYPE_INT | CTLFLAG_RW,
1531 msix, 0, igb_sysctl_msix_rate, "I",
1532 msix->msix_rate_desc);
1533 }
1534 }
b6220144
SZ		 1535
1536 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1537 OID_AUTO, "tx_intr_nsegs", CTLTYPE_INT | CTLFLAG_RW,
1538 sc, 0, igb_sysctl_tx_intr_nsegs, "I",
8d6600da
SZ		 1539 "# of segments per TX interrupt");
1540
d0f59cad
SZ		 1541#ifdef IFPOLL_ENABLE
1542 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1543 OID_AUTO, "npoll_rxoff", CTLTYPE_INT|CTLFLAG_RW,
1544 sc, 0, igb_sysctl_npoll_rxoff, "I", "NPOLLING RX cpu offset");
1545 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1546 OID_AUTO, "npoll_txoff", CTLTYPE_INT|CTLFLAG_RW,
1547 sc, 0, igb_sysctl_npoll_txoff, "I", "NPOLLING TX cpu offset");
1548#endif
1549
8d6600da
SZ		 1550#ifdef IGB_RSS_DEBUG
1551 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1552 OID_AUTO, "rss_debug", CTLFLAG_RW, &sc->rss_debug, 0,
1553 "RSS debug level");
1554 for (i = 0; i < sc->rx_ring_cnt; ++i) {
9c0ecdcc 1555 ksnprintf(node, sizeof(node), "rx%d_pkt", i);
8d6600da 1556 SYSCTL_ADD_ULONG(&sc->sysctl_ctx,
9c0ecdcc 1557 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, node,
8d6600da
SZ		 1558 CTLFLAG_RW, &sc->rx_rings[i].rx_packets, "RXed packets");
1559 }
1560#endif
1f7e3916
SZ		 1561}
1562
1563static int
a619b256 1564igb_alloc_rings(struct igb_softc *sc)
1f7e3916
SZ		 1565{
1566 int error, i;
1567
1f7e3916
SZ		 1568 /*
1569 * Create top level busdma tag
1570 */
1571 error = bus_dma_tag_create(NULL, 1, 0,
1572 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
1573 BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0,
1574 &sc->parent_tag);
1575 if (error) {
1576 device_printf(sc->dev, "could not create top level DMA tag\n");
1577 return error;
1578 }
1579
1580 /*
1581 * Allocate TX descriptor rings and buffers
1582 */
27866bf1 1583 sc->tx_rings = kmalloc(sizeof(struct igb_tx_ring) * sc->tx_ring_cnt,
1f7e3916 1584 M_DEVBUF, M_WAITOK | M_ZERO);
27866bf1 1585 for (i = 0; i < sc->tx_ring_cnt; ++i) {
1f7e3916
SZ		 1586 struct igb_tx_ring *txr = &sc->tx_rings[i];
1587
1588 /* Set up some basics */
1589 txr->sc = sc;
1590 txr->me = i;
7d235eb5 1591 lwkt_serialize_init(&txr->tx_serialize);
1f7e3916
SZ		 1592
1593 error = igb_create_tx_ring(txr);
1594 if (error)
1595 return error;
1596 }
1597
1598 /*
1599 * Allocate RX descriptor rings and buffers
1600 */
27866bf1 1601 sc->rx_rings = kmalloc(sizeof(struct igb_rx_ring) * sc->rx_ring_cnt,
1f7e3916 1602 M_DEVBUF, M_WAITOK | M_ZERO);
27866bf1 1603 for (i = 0; i < sc->rx_ring_cnt; ++i) {
1f7e3916
SZ		 1604 struct igb_rx_ring *rxr = &sc->rx_rings[i];
1605
1606 /* Set up some basics */
1607 rxr->sc = sc;
1608 rxr->me = i;
7d235eb5 1609 lwkt_serialize_init(&rxr->rx_serialize);
1f7e3916
SZ		 1610
1611 error = igb_create_rx_ring(rxr);
1612 if (error)
1613 return error;
1614 }
1615
1f7e3916
SZ		 1616 return 0;
1617}
1618
1619static void
a619b256 1620igb_free_rings(struct igb_softc *sc)
1f7e3916
SZ		 1621{
1622 int i;
1623
1f7e3916 1624 if (sc->tx_rings != NULL) {
91b8700a
SZ		 1625 for (i = 0; i < sc->tx_ring_cnt; ++i) {
1626 struct igb_tx_ring *txr = &sc->tx_rings[i];
1627
1628 igb_destroy_tx_ring(txr, txr->num_tx_desc);
1629 }
1f7e3916
SZ		 1630 kfree(sc->tx_rings, M_DEVBUF);
1631 }
1632
1633 if (sc->rx_rings != NULL) {
91b8700a
SZ		 1634 for (i = 0; i < sc->rx_ring_cnt; ++i) {
1635 struct igb_rx_ring *rxr = &sc->rx_rings[i];
1636
1637 igb_destroy_rx_ring(rxr, rxr->num_rx_desc);
1638 }
1f7e3916
SZ		 1639 kfree(sc->rx_rings, M_DEVBUF);
1640 }
1641}
1642
1643static int
1644igb_create_tx_ring(struct igb_tx_ring *txr)
1645{
1646 int tsize, error, i;
1647
1648 /*
1649 * Validate number of transmit descriptors. It must not exceed
1650 * hardware maximum, and must be multiple of IGB_DBA_ALIGN.
1651 */
1652 if (((igb_txd * sizeof(struct e1000_tx_desc)) % IGB_DBA_ALIGN) != 0 ||
1653 (igb_txd > IGB_MAX_TXD) || (igb_txd < IGB_MIN_TXD)) {
1654 device_printf(txr->sc->dev,
1655 "Using %d TX descriptors instead of %d!\n",
1656 IGB_DEFAULT_TXD, igb_txd);
91b8700a 1657 txr->num_tx_desc = IGB_DEFAULT_TXD;
1f7e3916 1658 } else {
91b8700a 1659 txr->num_tx_desc = igb_txd;
1f7e3916
SZ		 1660 }
1661
1662 /*
1663 * Allocate TX descriptor ring
1664 */
91b8700a 1665 tsize = roundup2(txr->num_tx_desc * sizeof(union e1000_adv_tx_desc),
1f7e3916
SZ		 1666 IGB_DBA_ALIGN);
1667 txr->txdma.dma_vaddr = bus_dmamem_coherent_any(txr->sc->parent_tag,
1668 IGB_DBA_ALIGN, tsize, BUS_DMA_WAITOK,
1669 &txr->txdma.dma_tag, &txr->txdma.dma_map, &txr->txdma.dma_paddr);
1670 if (txr->txdma.dma_vaddr == NULL) {
1671 device_printf(txr->sc->dev,
1672 "Unable to allocate TX Descriptor memory\n");
1673 return ENOMEM;
1674 }
1675 txr->tx_base = txr->txdma.dma_vaddr;
1676 bzero(txr->tx_base, tsize);
1677
91b8700a 1678 txr->tx_buf = kmalloc(sizeof(struct igb_tx_buf) * txr->num_tx_desc,
1f7e3916
SZ		 1679 M_DEVBUF, M_WAITOK | M_ZERO);
1680
1681 /*
b6220144
SZ		 1682 * Allocate TX head write-back buffer
1683 */
1684 txr->tx_hdr = bus_dmamem_coherent_any(txr->sc->parent_tag,
1685 __VM_CACHELINE_SIZE, __VM_CACHELINE_SIZE, BUS_DMA_WAITOK,
1686 &txr->tx_hdr_dtag, &txr->tx_hdr_dmap, &txr->tx_hdr_paddr);
1687 if (txr->tx_hdr == NULL) {
1688 device_printf(txr->sc->dev,
1689 "Unable to allocate TX head write-back buffer\n");
1690 return ENOMEM;
1691 }
1692
1693 /*
1f7e3916
SZ		 1694 * Create DMA tag for TX buffers
1695 */
1696 error = bus_dma_tag_create(txr->sc->parent_tag,
1697 1, 0, /* alignment, bounds */
1698 BUS_SPACE_MAXADDR, /* lowaddr */
1699 BUS_SPACE_MAXADDR, /* highaddr */
1700 NULL, NULL, /* filter, filterarg */
1701 IGB_TSO_SIZE, /* maxsize */
1702 IGB_MAX_SCATTER, /* nsegments */
1703 PAGE_SIZE, /* maxsegsize */
1704 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW |
1705 BUS_DMA_ONEBPAGE, /* flags */
1706 &txr->tx_tag);
1707 if (error) {
1708 device_printf(txr->sc->dev, "Unable to allocate TX DMA tag\n");
1709 kfree(txr->tx_buf, M_DEVBUF);
1710 txr->tx_buf = NULL;
1711 return error;
1712 }
1713
1714 /*
1715 * Create DMA maps for TX buffers
1716 */
91b8700a 1717 for (i = 0; i < txr->num_tx_desc; ++i) {
1f7e3916
SZ		 1718 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1719
1720 error = bus_dmamap_create(txr->tx_tag,
1721 BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE, &txbuf->map);
1722 if (error) {
1723 device_printf(txr->sc->dev,
1724 "Unable to create TX DMA map\n");
1725 igb_destroy_tx_ring(txr, i);
1726 return error;
1727 }
1728 }
b6220144
SZ		 1729
1730 /*
1731 * Initialize various watermark
1732 */
1733 txr->spare_desc = IGB_TX_SPARE;
91b8700a
SZ		 1734 txr->intr_nsegs = txr->num_tx_desc / 16;
1735 txr->oact_hi_desc = txr->num_tx_desc / 2;
1736 txr->oact_lo_desc = txr->num_tx_desc / 8;
b6220144
SZ		 1737 if (txr->oact_lo_desc > IGB_TX_OACTIVE_MAX)
1738 txr->oact_lo_desc = IGB_TX_OACTIVE_MAX;
1739 if (txr->oact_lo_desc < txr->spare_desc + IGB_TX_RESERVED)
1740 txr->oact_lo_desc = txr->spare_desc + IGB_TX_RESERVED;
1741
1f7e3916
SZ		 1742 return 0;
1743}
1744
1745static void
1746igb_free_tx_ring(struct igb_tx_ring *txr)
1747{
1748 int i;
1749
91b8700a 1750 for (i = 0; i < txr->num_tx_desc; ++i) {
1f7e3916
SZ		 1751 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1752
1753 if (txbuf->m_head != NULL) {
1754 bus_dmamap_unload(txr->tx_tag, txbuf->map);
1755 m_freem(txbuf->m_head);
1756 txbuf->m_head = NULL;
1757 }
1758 }
1759}
1760
1761static void
1762igb_destroy_tx_ring(struct igb_tx_ring *txr, int ndesc)
1763{
1764 int i;
1765
1766 if (txr->txdma.dma_vaddr != NULL) {
1767 bus_dmamap_unload(txr->txdma.dma_tag, txr->txdma.dma_map);
1768 bus_dmamem_free(txr->txdma.dma_tag, txr->txdma.dma_vaddr,
1769 txr->txdma.dma_map);
1770 bus_dma_tag_destroy(txr->txdma.dma_tag);
1771 txr->txdma.dma_vaddr = NULL;
1772 }
1773
b6220144
SZ		 1774 if (txr->tx_hdr != NULL) {
1775 bus_dmamap_unload(txr->tx_hdr_dtag, txr->tx_hdr_dmap);
1776 bus_dmamem_free(txr->tx_hdr_dtag, txr->tx_hdr,
1777 txr->tx_hdr_dmap);
1778 bus_dma_tag_destroy(txr->tx_hdr_dtag);
1779 txr->tx_hdr = NULL;
1780 }
1781
1f7e3916
SZ		 1782 if (txr->tx_buf == NULL)
1783 return;
1784
1785 for (i = 0; i < ndesc; ++i) {
1786 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1787
1788 KKASSERT(txbuf->m_head == NULL);
1789 bus_dmamap_destroy(txr->tx_tag, txbuf->map);
1790 }
1791 bus_dma_tag_destroy(txr->tx_tag);
1792
1793 kfree(txr->tx_buf, M_DEVBUF);
1794 txr->tx_buf = NULL;
1795}
1796
1797static void
1798igb_init_tx_ring(struct igb_tx_ring *txr)
1799{
1f7e3916
SZ		 1800 /* Clear the old descriptor contents */
1801 bzero(txr->tx_base,
91b8700a 1802 sizeof(union e1000_adv_tx_desc) * txr->num_tx_desc);
1f7e3916 1803
b6220144
SZ		 1804 /* Clear TX head write-back buffer */
1805 *(txr->tx_hdr) = 0;
1806
1f7e3916
SZ		 1807 /* Reset indices */
1808 txr->next_avail_desc = 0;
1809 txr->next_to_clean = 0;
b6220144 1810 txr->tx_nsegs = 0;
1f7e3916
SZ		 1811
1812 /* Set number of descriptors available */
91b8700a 1813 txr->tx_avail = txr->num_tx_desc;
1f7e3916
SZ		 1814}
1815
1816static void
1817igb_init_tx_unit(struct igb_softc *sc)
1818{
1819 struct e1000_hw *hw = &sc->hw;
1820 uint32_t tctl;
1821 int i;
1822
1823 /* Setup the Tx Descriptor Rings */
27866bf1 1824 for (i = 0; i < sc->tx_ring_cnt; ++i) {
1f7e3916
SZ		 1825 struct igb_tx_ring *txr = &sc->tx_rings[i];
1826 uint64_t bus_addr = txr->txdma.dma_paddr;
c3162c4e 1827 uint64_t hdr_paddr = txr->tx_hdr_paddr;
1f7e3916 1828 uint32_t txdctl = 0;
b6220144 1829 uint32_t dca_txctrl;
1f7e3916
SZ		 1830
1831 E1000_WRITE_REG(hw, E1000_TDLEN(i),
91b8700a 1832 txr->num_tx_desc * sizeof(struct e1000_tx_desc));
1f7e3916
SZ		 1833 E1000_WRITE_REG(hw, E1000_TDBAH(i),
1834 (uint32_t)(bus_addr >> 32));
1835 E1000_WRITE_REG(hw, E1000_TDBAL(i),
1836 (uint32_t)bus_addr);
1837
1838 /* Setup the HW Tx Head and Tail descriptor pointers */
1839 E1000_WRITE_REG(hw, E1000_TDT(i), 0);
1840 E1000_WRITE_REG(hw, E1000_TDH(i), 0);
1841
54691ff1
SZ		 1842 /*
1843 * WTHRESH is ignored by the hardware, since header
1844 * write back mode is used.
1845 */
1f7e3916
SZ		 1846 txdctl |= IGB_TX_PTHRESH;
1847 txdctl |= IGB_TX_HTHRESH << 8;
1848 txdctl |= IGB_TX_WTHRESH << 16;
1849 txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
1850 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
b6220144
SZ		 1851
1852 dca_txctrl = E1000_READ_REG(hw, E1000_DCA_TXCTRL(i));
1853 dca_txctrl &= ~E1000_DCA_TXCTRL_TX_WB_RO_EN;
1854 E1000_WRITE_REG(hw, E1000_DCA_TXCTRL(i), dca_txctrl);
1855
54691ff1
SZ		 1856 /*
1857 * Don't set WB_on_EITR:
1858 * - 82575 does not have it
1859 * - It almost has no effect on 82576, see:
1860 * 82576 specification update errata #26
1861 * - It causes unnecessary bus traffic
1862 */
b6220144 1863 E1000_WRITE_REG(hw, E1000_TDWBAH(i),
c3162c4e 1864 (uint32_t)(hdr_paddr >> 32));
b6220144 1865 E1000_WRITE_REG(hw, E1000_TDWBAL(i),
c3162c4e 1866 ((uint32_t)hdr_paddr) | E1000_TX_HEAD_WB_ENABLE);
1f7e3916
SZ		 1867 }
1868
1869 if (sc->vf_ifp)
1870 return;
1871
1872 e1000_config_collision_dist(hw);
1873
1874 /* Program the Transmit Control Register */
1875 tctl = E1000_READ_REG(hw, E1000_TCTL);
1876 tctl &= ~E1000_TCTL_CT;
1877 tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
1878 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
1879
1880 /* This write will effectively turn on the transmit unit. */
1881 E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1882}
1883
1884static boolean_t
48faa653 1885igb_txcsum_ctx(struct igb_tx_ring *txr, struct mbuf *mp)
1f7e3916
SZ		 1886{
1887 struct e1000_adv_tx_context_desc *TXD;
1f7e3916 1888 uint32_t vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx;
1f7e3916 1889 int ehdrlen, ctxd, ip_hlen = 0;
1f7e3916
SZ		 1890 boolean_t offload = TRUE;
1891
1892 if ((mp->m_pkthdr.csum_flags & IGB_CSUM_FEATURES) == 0)
1893 offload = FALSE;
1894
1895 vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0;
48faa653 1896
1f7e3916 1897 ctxd = txr->next_avail_desc;
1f7e3916
SZ		 1898 TXD = (struct e1000_adv_tx_context_desc *)&txr->tx_base[ctxd];
1899
1900 /*
1901 * In advanced descriptors the vlan tag must
1902 * be placed into the context descriptor, thus
1903 * we need to be here just for that setup.
1904 */
1905 if (mp->m_flags & M_VLANTAG) {
23f6ffe4
SZ		 1906 uint16_t vlantag;
1907
1f7e3916
SZ		 1908 vlantag = htole16(mp->m_pkthdr.ether_vlantag);
1909 vlan_macip_lens |= (vlantag << E1000_ADVTXD_VLAN_SHIFT);
1910 } else if (!offload) {
1911 return FALSE;
1912 }
1913
48faa653
SZ		 1914 ehdrlen = mp->m_pkthdr.csum_lhlen;
1915 KASSERT(ehdrlen > 0, ("invalid ether hlen"));
1f7e3916
SZ		 1916
1917 /* Set the ether header length */
1918 vlan_macip_lens |= ehdrlen << E1000_ADVTXD_MACLEN_SHIFT;
48faa653
SZ		 1919 if (mp->m_pkthdr.csum_flags & CSUM_IP) {
1920 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
1921 ip_hlen = mp->m_pkthdr.csum_iphlen;
1922 KASSERT(ip_hlen > 0, ("invalid ip hlen"));
1f7e3916 1923 }
1f7e3916 1924 vlan_macip_lens |= ip_hlen;
1f7e3916 1925
23f6ffe4 1926 type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
1f7e3916
SZ		 1927 if (mp->m_pkthdr.csum_flags & CSUM_TCP)
1928 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
1929 else if (mp->m_pkthdr.csum_flags & CSUM_UDP)
1930 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP;
1931
1932 /* 82575 needs the queue index added */
1933 if (txr->sc->hw.mac.type == e1000_82575)
1934 mss_l4len_idx = txr->me << 4;
1935
1936 /* Now copy bits into descriptor */
1937 TXD->vlan_macip_lens = htole32(vlan_macip_lens);
1938 TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
1939 TXD->seqnum_seed = htole32(0);
1940 TXD->mss_l4len_idx = htole32(mss_l4len_idx);
1941
1f7e3916 1942 /* We've consumed the first desc, adjust counters */
91b8700a 1943 if (++ctxd == txr->num_tx_desc)
1f7e3916
SZ		 1944 ctxd = 0;
1945 txr->next_avail_desc = ctxd;
1946 --txr->tx_avail;
1947
1948 return offload;
1949}
1950
1951static void
1952igb_txeof(struct igb_tx_ring *txr)
1953{
1954 struct ifnet *ifp = &txr->sc->arpcom.ac_if;
b6220144 1955 int first, hdr, avail;
1f7e3916 1956
91b8700a 1957 if (txr->tx_avail == txr->num_tx_desc)
1f7e3916
SZ		 1958 return;
1959
1960 first = txr->next_to_clean;
b6220144 1961 hdr = *(txr->tx_hdr);
1f7e3916 1962
b6220144
SZ		 1963 if (first == hdr)
1964 return;
1f7e3916 1965
b6220144
SZ		 1966 avail = txr->tx_avail;
1967 while (first != hdr) {
1968 struct igb_tx_buf *txbuf = &txr->tx_buf[first];
1f7e3916 1969
b6220144
SZ		 1970 ++avail;
1971 if (txbuf->m_head) {
1972 bus_dmamap_unload(txr->tx_tag, txbuf->map);
1973 m_freem(txbuf->m_head);
1974 txbuf->m_head = NULL;
1975 ++ifp->if_opackets;
1f7e3916 1976 }
91b8700a 1977 if (++first == txr->num_tx_desc)
b6220144 1978 first = 0;
1f7e3916
SZ		 1979 }
1980 txr->next_to_clean = first;
b6220144 1981 txr->tx_avail = avail;
1f7e3916
SZ		 1982
1983 /*
1984 * If we have a minimum free, clear IFF_OACTIVE
1985 * to tell the stack that it is OK to send packets.
1986 */
b6220144 1987 if (IGB_IS_NOT_OACTIVE(txr)) {
1f7e3916
SZ		 1988 ifp->if_flags &= ~IFF_OACTIVE;
1989
1f7e3916
SZ		 1990 /*
1991 * We have enough TX descriptors, turn off
b6220144
SZ		 1992 * the watchdog. We allow small amount of
1993 * packets (roughly intr_nsegs) pending on
1994 * the transmit ring.
1f7e3916
SZ		 1995 */
1996 ifp->if_timer = 0;
1f7e3916
SZ		 1997 }
1998}
1999
2000static int
2001igb_create_rx_ring(struct igb_rx_ring *rxr)
2002{
2003 int rsize, i, error;
2004
2005 /*
2006 * Validate number of receive descriptors. It must not exceed
2007 * hardware maximum, and must be multiple of IGB_DBA_ALIGN.
2008 */
2009 if (((igb_rxd * sizeof(struct e1000_rx_desc)) % IGB_DBA_ALIGN) != 0 ||
2010 (igb_rxd > IGB_MAX_RXD) || (igb_rxd < IGB_MIN_RXD)) {
2011 device_printf(rxr->sc->dev,
2012 "Using %d RX descriptors instead of %d!\n",
2013 IGB_DEFAULT_RXD, igb_rxd);
91b8700a 2014 rxr->num_rx_desc = IGB_DEFAULT_RXD;
1f7e3916 2015 } else {
91b8700a 2016 rxr->num_rx_desc = igb_rxd;
1f7e3916
SZ		 2017 }
2018
2019 /*
2020 * Allocate RX descriptor ring
2021 */
91b8700a 2022 rsize = roundup2(rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc),
1f7e3916
SZ		 2023 IGB_DBA_ALIGN);
2024 rxr->rxdma.dma_vaddr = bus_dmamem_coherent_any(rxr->sc->parent_tag,
2025 IGB_DBA_ALIGN, rsize, BUS_DMA_WAITOK,
2026 &rxr->rxdma.dma_tag, &rxr->rxdma.dma_map,
2027 &rxr->rxdma.dma_paddr);
2028 if (rxr->rxdma.dma_vaddr == NULL) {
2029 device_printf(rxr->sc->dev,
2030 "Unable to allocate RxDescriptor memory\n");
2031 return ENOMEM;
2032 }
2033 rxr->rx_base = rxr->rxdma.dma_vaddr;
2034 bzero(rxr->rx_base, rsize);
2035
91b8700a 2036 rxr->rx_buf = kmalloc(sizeof(struct igb_rx_buf) * rxr->num_rx_desc,
1f7e3916
SZ		 2037 M_DEVBUF, M_WAITOK | M_ZERO);
2038
2039 /*
2040 * Create DMA tag for RX buffers
2041 */
2042 error = bus_dma_tag_create(rxr->sc->parent_tag,
2043 1, 0, /* alignment, bounds */
2044 BUS_SPACE_MAXADDR, /* lowaddr */
2045 BUS_SPACE_MAXADDR, /* highaddr */
2046 NULL, NULL, /* filter, filterarg */
2047 MCLBYTES, /* maxsize */
2048 1, /* nsegments */
2049 MCLBYTES, /* maxsegsize */
2050 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW, /* flags */
2051 &rxr->rx_tag);
2052 if (error) {
2053 device_printf(rxr->sc->dev,
2054 "Unable to create RX payload DMA tag\n");
2055 kfree(rxr->rx_buf, M_DEVBUF);
2056 rxr->rx_buf = NULL;
2057 return error;
2058 }
2059
2060 /*
2061 * Create spare DMA map for RX buffers
2062 */
2063 error = bus_dmamap_create(rxr->rx_tag, BUS_DMA_WAITOK,
2064 &rxr->rx_sparemap);
2065 if (error) {
2066 device_printf(rxr->sc->dev,
2067 "Unable to create spare RX DMA maps\n");
2068 bus_dma_tag_destroy(rxr->rx_tag);
2069 kfree(rxr->rx_buf, M_DEVBUF);
2070 rxr->rx_buf = NULL;
2071 return error;
2072 }
2073
2074 /*
2075 * Create DMA maps for RX buffers
2076 */
91b8700a 2077 for (i = 0; i < rxr->num_rx_desc; i++) {
1f7e3916
SZ		 2078 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2079
2080 error = bus_dmamap_create(rxr->rx_tag,
2081 BUS_DMA_WAITOK, &rxbuf->map);
2082 if (error) {
2083 device_printf(rxr->sc->dev,
2084 "Unable to create RX DMA maps\n");
2085 igb_destroy_rx_ring(rxr, i);
2086 return error;
2087 }
2088 }
2089 return 0;
2090}
2091
2092static void
2093igb_free_rx_ring(struct igb_rx_ring *rxr)
2094{
2095 int i;
2096
91b8700a 2097 for (i = 0; i < rxr->num_rx_desc; ++i) {
1f7e3916
SZ		 2098 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2099
2100 if (rxbuf->m_head != NULL) {
2101 bus_dmamap_unload(rxr->rx_tag, rxbuf->map);
2102 m_freem(rxbuf->m_head);
2103 rxbuf->m_head = NULL;
2104 }
2105 }
2106
2107 if (rxr->fmp != NULL)
2108 m_freem(rxr->fmp);
2109 rxr->fmp = NULL;
2110 rxr->lmp = NULL;
2111}
2112
2113static void
2114igb_destroy_rx_ring(struct igb_rx_ring *rxr, int ndesc)
2115{
2116 int i;
2117
2118 if (rxr->rxdma.dma_vaddr != NULL) {
2119 bus_dmamap_unload(rxr->rxdma.dma_tag, rxr->rxdma.dma_map);
2120 bus_dmamem_free(rxr->rxdma.dma_tag, rxr->rxdma.dma_vaddr,
2121 rxr->rxdma.dma_map);
2122 bus_dma_tag_destroy(rxr->rxdma.dma_tag);
2123 rxr->rxdma.dma_vaddr = NULL;
2124 }
2125
2126 if (rxr->rx_buf == NULL)
2127 return;
2128
2129 for (i = 0; i < ndesc; ++i) {
2130 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2131
2132 KKASSERT(rxbuf->m_head == NULL);
2133 bus_dmamap_destroy(rxr->rx_tag, rxbuf->map);
2134 }
2135 bus_dmamap_destroy(rxr->rx_tag, rxr->rx_sparemap);
2136 bus_dma_tag_destroy(rxr->rx_tag);
2137
2138 kfree(rxr->rx_buf, M_DEVBUF);
2139 rxr->rx_buf = NULL;
2140}
2141
2142static void
2143igb_setup_rxdesc(union e1000_adv_rx_desc *rxd, const struct igb_rx_buf *rxbuf)
2144{
2145 rxd->read.pkt_addr = htole64(rxbuf->paddr);
2146 rxd->wb.upper.status_error = 0;
2147}
2148
2149static int
2150igb_newbuf(struct igb_rx_ring *rxr, int i, boolean_t wait)
2151{
2152 struct mbuf *m;
2153 bus_dma_segment_t seg;
2154 bus_dmamap_t map;
2155 struct igb_rx_buf *rxbuf;
2156 int error, nseg;
2157
2158 m = m_getcl(wait ? MB_WAIT : MB_DONTWAIT, MT_DATA, M_PKTHDR);
2159 if (m == NULL) {
2160 if (wait) {
2161 if_printf(&rxr->sc->arpcom.ac_if,
2162 "Unable to allocate RX mbuf\n");
2163 }
2164 return ENOBUFS;
2165 }
2166 m->m_len = m->m_pkthdr.len = MCLBYTES;
2167
2168 if (rxr->sc->max_frame_size <= MCLBYTES - ETHER_ALIGN)
2169 m_adj(m, ETHER_ALIGN);
2170
2171 error = bus_dmamap_load_mbuf_segment(rxr->rx_tag,
2172 rxr->rx_sparemap, m, &seg, 1, &nseg, BUS_DMA_NOWAIT);
2173 if (error) {
2174 m_freem(m);
2175 if (wait) {
2176 if_printf(&rxr->sc->arpcom.ac_if,
2177 "Unable to load RX mbuf\n");
2178 }
2179 return error;
2180 }
2181
2182 rxbuf = &rxr->rx_buf[i];
2183 if (rxbuf->m_head != NULL)
2184 bus_dmamap_unload(rxr->rx_tag, rxbuf->map);
2185
2186 map = rxbuf->map;
2187 rxbuf->map = rxr->rx_sparemap;
2188 rxr->rx_sparemap = map;
2189
2190 rxbuf->m_head = m;
2191 rxbuf->paddr = seg.ds_addr;
2192
2193 igb_setup_rxdesc(&rxr->rx_base[i], rxbuf);
2194 return 0;
2195}
2196
2197static int
2198igb_init_rx_ring(struct igb_rx_ring *rxr)
2199{
2200 int i;
2201
2202 /* Clear the ring contents */
2203 bzero(rxr->rx_base,
91b8700a 2204 rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc));
1f7e3916
SZ		 2205
2206 /* Now replenish the ring mbufs */
91b8700a 2207 for (i = 0; i < rxr->num_rx_desc; ++i) {
1f7e3916
SZ		 2208 int error;
2209
2210 error = igb_newbuf(rxr, i, TRUE);
2211 if (error)
2212 return error;
2213 }
2214
2215 /* Setup our descriptor indices */
2216 rxr->next_to_check = 0;
2217
2218 rxr->fmp = NULL;
2219 rxr->lmp = NULL;
2220 rxr->discard = FALSE;
2221
2222 return 0;
2223}
2224
2225static void
2226igb_init_rx_unit(struct igb_softc *sc)
2227{
2228 struct ifnet *ifp = &sc->arpcom.ac_if;
2229 struct e1000_hw *hw = &sc->hw;
2230 uint32_t rctl, rxcsum, srrctl = 0;
2231 int i;
2232
2233 /*
2234 * Make sure receives are disabled while setting
2235 * up the descriptor ring
2236 */
2237 rctl = E1000_READ_REG(hw, E1000_RCTL);
2238 E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2239
2240#if 0
2241 /*
2242 ** Set up for header split
2243 */
2244 if (igb_header_split) {
2245 /* Use a standard mbuf for the header */
2246 srrctl |= IGB_HDR_BUF << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
2247 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
2248 } else
2249#endif
2250 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
2251
2252 /*
2253 ** Set up for jumbo frames
2254 */
2255 if (ifp->if_mtu > ETHERMTU) {
2256 rctl |= E1000_RCTL_LPE;
2257#if 0
2258 if (adapter->rx_mbuf_sz == MJUMPAGESIZE) {
2259 srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2260 rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
2261 } else if (adapter->rx_mbuf_sz > MJUMPAGESIZE) {
2262 srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2263 rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
2264 }
2265 /* Set maximum packet len */
2266 psize = adapter->max_frame_size;
2267 /* are we on a vlan? */
2268 if (adapter->ifp->if_vlantrunk != NULL)
2269 psize += VLAN_TAG_SIZE;
2270 E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
2271#else
2272 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2273 rctl |= E1000_RCTL_SZ_2048;
2274#endif
2275 } else {
2276 rctl &= ~E1000_RCTL_LPE;
2277 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2278 rctl |= E1000_RCTL_SZ_2048;
2279 }
2280
2281 /* Setup the Base and Length of the Rx Descriptor Rings */
be922da6 2282 for (i = 0; i < sc->rx_ring_inuse; ++i) {
1f7e3916
SZ		 2283 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2284 uint64_t bus_addr = rxr->rxdma.dma_paddr;
2285 uint32_t rxdctl;
2286
2287 E1000_WRITE_REG(hw, E1000_RDLEN(i),
91b8700a 2288 rxr->num_rx_desc * sizeof(struct e1000_rx_desc));
1f7e3916
SZ		 2289 E1000_WRITE_REG(hw, E1000_RDBAH(i),
2290 (uint32_t)(bus_addr >> 32));
2291 E1000_WRITE_REG(hw, E1000_RDBAL(i),
2292 (uint32_t)bus_addr);
2293 E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
2294 /* Enable this Queue */
2295 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
2296 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
2297 rxdctl &= 0xFFF00000;
2298 rxdctl |= IGB_RX_PTHRESH;
2299 rxdctl |= IGB_RX_HTHRESH << 8;
54691ff1
SZ		 2300 /*
2301 * Don't set WTHRESH to a value above 1 on 82576, see:
2302 * 82576 specification update errata #26
2303 */
1f7e3916
SZ		 2304 rxdctl |= IGB_RX_WTHRESH << 16;
2305 E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
2306 }
2307
8d6600da
SZ		 2308 rxcsum = E1000_READ_REG(&sc->hw, E1000_RXCSUM);
2309 rxcsum &= ~(E1000_RXCSUM_PCSS_MASK | E1000_RXCSUM_IPPCSE);
2310
1f7e3916 2311 /*
8d6600da
SZ		 2312 * Receive Checksum Offload for TCP and UDP
2313 *
2314 * Checksum offloading is also enabled if multiple receive
2315 * queue is to be supported, since we need it to figure out
2316 * fragments.
1f7e3916 2317 */
8d6600da
SZ		 2318 if ((ifp->if_capenable & IFCAP_RXCSUM) || IGB_ENABLE_HWRSS(sc)) {
2319 /*
2320 * NOTE:
2321 * PCSD must be enabled to enable multiple
2322 * receive queues.
2323 */
2324 rxcsum |= E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL |
2325 E1000_RXCSUM_PCSD;
2326 } else {
2327 rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL |
2328 E1000_RXCSUM_PCSD);
2329 }
2330 E1000_WRITE_REG(&sc->hw, E1000_RXCSUM, rxcsum);
2331
2332 if (IGB_ENABLE_HWRSS(sc)) {
2333 uint8_t key[IGB_NRSSRK * IGB_RSSRK_SIZE];
d1218435
SZ		 2334 uint32_t reta_shift;
2335 int j, r;
8d6600da
SZ		 2336
2337 /*
2338 * NOTE:
2339 * When we reach here, RSS has already been disabled
2340 * in igb_stop(), so we could safely configure RSS key
2341 * and redirect table.
2342 */
2343
2344 /*
2345 * Configure RSS key
2346 */
2347 toeplitz_get_key(key, sizeof(key));
2348 for (i = 0; i < IGB_NRSSRK; ++i) {
2349 uint32_t rssrk;
2350
2351 rssrk = IGB_RSSRK_VAL(key, i);
2352 IGB_RSS_DPRINTF(sc, 1, "rssrk%d 0x%08x\n", i, rssrk);
2353
2354 E1000_WRITE_REG(hw, E1000_RSSRK(i), rssrk);
1f7e3916 2355 }
1f7e3916
SZ		 2356
2357 /*
8d6600da
SZ		 2358 * Configure RSS redirect table in following fashion:
2359 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2360 */
2361 reta_shift = IGB_RETA_SHIFT;
2362 if (hw->mac.type == e1000_82575)
2363 reta_shift = IGB_RETA_SHIFT_82575;
8d6600da 2364
d1218435
SZ		 2365 r = 0;
2366 for (j = 0; j < IGB_NRETA; ++j) {
2367 uint32_t reta = 0;
8d6600da 2368
d1218435
SZ		 2369 for (i = 0; i < IGB_RETA_SIZE; ++i) {
2370 uint32_t q;
2371
be922da6 2372 q = (r % sc->rx_ring_inuse) << reta_shift;
d1218435
SZ		 2373 reta |= q << (8 * i);
2374 ++r;
2375 }
2376 IGB_RSS_DPRINTF(sc, 1, "reta 0x%08x\n", reta);
2377 E1000_WRITE_REG(hw, E1000_RETA(j), reta);
2378 }
8d6600da
SZ		 2379
2380 /*
2381 * Enable multiple receive queues.
2382 * Enable IPv4 RSS standard hash functions.
2383 * Disable RSS interrupt on 82575
2384 */
2385 E1000_WRITE_REG(&sc->hw, E1000_MRQC,
2386 E1000_MRQC_ENABLE_RSS_4Q |
2387 E1000_MRQC_RSS_FIELD_IPV4_TCP |
2388 E1000_MRQC_RSS_FIELD_IPV4);
1f7e3916 2389 }
1f7e3916
SZ		 2390
2391 /* Setup the Receive Control Register */
2392 rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
2393 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
2394 E1000_RCTL_RDMTS_HALF |
2395 (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
2396 /* Strip CRC bytes. */
2397 rctl |= E1000_RCTL_SECRC;
2398 /* Make sure VLAN Filters are off */
2399 rctl &= ~E1000_RCTL_VFE;
2400 /* Don't store bad packets */
2401 rctl &= ~E1000_RCTL_SBP;
2402
2403 /* Enable Receives */
2404 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2405
2406 /*
2407 * Setup the HW Rx Head and Tail Descriptor Pointers
2408 * - needs to be after enable
2409 */
be922da6 2410 for (i = 0; i < sc->rx_ring_inuse; ++i) {
1f7e3916
SZ		 2411 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2412
2413 E1000_WRITE_REG(hw, E1000_RDH(i), rxr->next_to_check);
91b8700a 2414 E1000_WRITE_REG(hw, E1000_RDT(i), rxr->num_rx_desc - 1);
1f7e3916
SZ		 2415 }
2416}
2417
2418static void
2419igb_rxeof(struct igb_rx_ring *rxr, int count)
2420{
2421 struct ifnet *ifp = &rxr->sc->arpcom.ac_if;
2422 union e1000_adv_rx_desc *cur;
2423 uint32_t staterr;
2424 int i;
2425
2426 i = rxr->next_to_check;
2427 cur = &rxr->rx_base[i];
2428 staterr = le32toh(cur->wb.upper.status_error);
2429
2430 if ((staterr & E1000_RXD_STAT_DD) == 0)
2431 return;
2432
2433 while ((staterr & E1000_RXD_STAT_DD) && count != 0) {
8d6600da 2434 struct pktinfo *pi = NULL, pi0;
1f7e3916
SZ		 2435 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2436 struct mbuf *m = NULL;
2437 boolean_t eop;
2438
2439 eop = (staterr & E1000_RXD_STAT_EOP) ? TRUE : FALSE;
2440 if (eop)
2441 --count;
2442
2443 if ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) == 0 &&
2444 !rxr->discard) {
2445 struct mbuf *mp = rxbuf->m_head;
8d6600da 2446 uint32_t hash, hashtype;
1f7e3916
SZ		 2447 uint16_t vlan;
2448 int len;
2449
2450 len = le16toh(cur->wb.upper.length);
2451 if (rxr->sc->hw.mac.type == e1000_i350 &&
2452 (staterr & E1000_RXDEXT_STATERR_LB))
2453 vlan = be16toh(cur->wb.upper.vlan);
2454 else
2455 vlan = le16toh(cur->wb.upper.vlan);
2456
8d6600da
SZ		 2457 hash = le32toh(cur->wb.lower.hi_dword.rss);
2458 hashtype = le32toh(cur->wb.lower.lo_dword.data) &
2459 E1000_RXDADV_RSSTYPE_MASK;
2460
2461 IGB_RSS_DPRINTF(rxr->sc, 10,
2462 "ring%d, hash 0x%08x, hashtype %u\n",
2463 rxr->me, hash, hashtype);
2464
1f7e3916
SZ		 2465 bus_dmamap_sync(rxr->rx_tag, rxbuf->map,
2466 BUS_DMASYNC_POSTREAD);
2467
2468 if (igb_newbuf(rxr, i, FALSE) != 0) {
2469 ifp->if_iqdrops++;
2470 goto discard;
2471 }
2472
2473 mp->m_len = len;
2474 if (rxr->fmp == NULL) {
2475 mp->m_pkthdr.len = len;
2476 rxr->fmp = mp;
2477 rxr->lmp = mp;
2478 } else {
2479 rxr->lmp->m_next = mp;
2480 rxr->lmp = rxr->lmp->m_next;
2481 rxr->fmp->m_pkthdr.len += len;
2482 }
2483
2484 if (eop) {
2485 m = rxr->fmp;
2486 rxr->fmp = NULL;
2487 rxr->lmp = NULL;
2488
2489 m->m_pkthdr.rcvif = ifp;
2490 ifp->if_ipackets++;
2491
2492 if (ifp->if_capenable & IFCAP_RXCSUM)
2493 igb_rxcsum(staterr, m);
2494
2495 if (staterr & E1000_RXD_STAT_VP) {
2496 m->m_pkthdr.ether_vlantag = vlan;
2497 m->m_flags |= M_VLANTAG;
2498 }
2499
1f7e3916 2500 if (ifp->if_capenable & IFCAP_RSS) {
8d6600da
SZ		 2501 pi = igb_rssinfo(m, &pi0,
2502 hash, hashtype, staterr);
1f7e3916 2503 }
8d6600da
SZ		 2504#ifdef IGB_RSS_DEBUG
2505 rxr->rx_packets++;
1f7e3916
SZ		 2506#endif
2507 }
2508 } else {
2509 ifp->if_ierrors++;
2510discard:
2511 igb_setup_rxdesc(cur, rxbuf);
2512 if (!eop)
2513 rxr->discard = TRUE;
2514 else
2515 rxr->discard = FALSE;
2516 if (rxr->fmp != NULL) {
2517 m_freem(rxr->fmp);
2518 rxr->fmp = NULL;
2519 rxr->lmp = NULL;
2520 }
2521 m = NULL;
2522 }
2523
2524 if (m != NULL)
8d6600da 2525 ether_input_pkt(ifp, m, pi);
1f7e3916
SZ		 2526
2527 /* Advance our pointers to the next descriptor. */
91b8700a 2528 if (++i == rxr->num_rx_desc)
1f7e3916
SZ		 2529 i = 0;
2530
2531 cur = &rxr->rx_base[i];
2532 staterr = le32toh(cur->wb.upper.status_error);
2533 }
2534 rxr->next_to_check = i;
2535
2536 if (--i < 0)
91b8700a 2537 i = rxr->num_rx_desc - 1;
1f7e3916
SZ		 2538 E1000_WRITE_REG(&rxr->sc->hw, E1000_RDT(rxr->me), i);
2539}
2540
2541
2542static void
2543igb_set_vlan(struct igb_softc *sc)
2544{
2545 struct e1000_hw *hw = &sc->hw;
2546 uint32_t reg;
2547#if 0
2548 struct ifnet *ifp = sc->arpcom.ac_if;
2549#endif
2550
2551 if (sc->vf_ifp) {
2552 e1000_rlpml_set_vf(hw, sc->max_frame_size + VLAN_TAG_SIZE);
2553 return;
2554 }
2555
2556 reg = E1000_READ_REG(hw, E1000_CTRL);
2557 reg |= E1000_CTRL_VME;
2558 E1000_WRITE_REG(hw, E1000_CTRL, reg);
2559
2560#if 0
2561 /* Enable the Filter Table */
2562 if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) {
2563 reg = E1000_READ_REG(hw, E1000_RCTL);
2564 reg &= ~E1000_RCTL_CFIEN;
2565 reg |= E1000_RCTL_VFE;
2566 E1000_WRITE_REG(hw, E1000_RCTL, reg);
2567 }
2568#endif
2569
2570 /* Update the frame size */
2571 E1000_WRITE_REG(&sc->hw, E1000_RLPML,
2572 sc->max_frame_size + VLAN_TAG_SIZE);
2573
2574#if 0
2575 /* Don't bother with table if no vlans */
2576 if ((adapter->num_vlans == 0) ||
2577 ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0))
2578 return;
2579 /*
2580 ** A soft reset zero's out the VFTA, so
2581 ** we need to repopulate it now.
2582 */
2583 for (int i = 0; i < IGB_VFTA_SIZE; i++)
2584 if (adapter->shadow_vfta[i] != 0) {
2585 if (adapter->vf_ifp)
2586 e1000_vfta_set_vf(hw,
2587 adapter->shadow_vfta[i], TRUE);
2588 else
2589 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
2590 i, adapter->shadow_vfta[i]);
2591 }
2592#endif
2593}
2594
2595static void
2596igb_enable_intr(struct igb_softc *sc)
2597{
9c0ecdcc
SZ		 2598 if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
2599 lwkt_serialize_handler_enable(&sc->main_serialize);
2600 } else {
2601 int i;
2602
2603 for (i = 0; i < sc->msix_cnt; ++i) {
2604 lwkt_serialize_handler_enable(
2605 sc->msix_data[i].msix_serialize);
2606 }
2607 }
1f7e3916 2608
f6167a56 2609 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) {
9c0ecdcc
SZ		 2610 if (sc->intr_type == PCI_INTR_TYPE_MSIX)
2611 E1000_WRITE_REG(&sc->hw, E1000_EIAC, sc->intr_mask);
2612 else
2613 E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0);
f6167a56
SZ		 2614 E1000_WRITE_REG(&sc->hw, E1000_EIAM, sc->intr_mask);
2615 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask);
1f7e3916
SZ		 2616 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
2617 } else {
2618 E1000_WRITE_REG(&sc->hw, E1000_IMS, IMS_ENABLE_MASK);
2619 }
2620 E1000_WRITE_FLUSH(&sc->hw);
2621}
2622
2623static void
2624igb_disable_intr(struct igb_softc *sc)
2625{
f6167a56 2626 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) {
1f7e3916
SZ		 2627 E1000_WRITE_REG(&sc->hw, E1000_EIMC, 0xffffffff);
2628 E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0);
f6167a56 2629 }
1f7e3916
SZ		 2630 E1000_WRITE_REG(&sc->hw, E1000_IMC, 0xffffffff);
2631 E1000_WRITE_FLUSH(&sc->hw);
2632
9c0ecdcc
SZ		 2633 if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
2634 lwkt_serialize_handler_disable(&sc->main_serialize);
2635 } else {
2636 int i;
2637
2638 for (i = 0; i < sc->msix_cnt; ++i) {
2639 lwkt_serialize_handler_disable(
2640 sc->msix_data[i].msix_serialize);
2641 }
2642 }
1f7e3916
SZ		 2643}
2644
2645/*
2646 * Bit of a misnomer, what this really means is
2647 * to enable OS management of the system... aka
2648 * to disable special hardware management features
2649 */
2650static void
2651igb_get_mgmt(struct igb_softc *sc)
2652{
396b7048 2653 if (sc->flags & IGB_FLAG_HAS_MGMT) {
1f7e3916
SZ		 2654 int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
2655 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
2656
2657 /* disable hardware interception of ARP */
2658 manc &= ~E1000_MANC_ARP_EN;
2659
2660 /* enable receiving management packets to the host */
2661 manc |= E1000_MANC_EN_MNG2HOST;
2662 manc2h |= 1 << 5; /* Mng Port 623 */
2663 manc2h |= 1 << 6; /* Mng Port 664 */
2664 E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
2665 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
2666 }
2667}
2668
2669/*
2670 * Give control back to hardware management controller
2671 * if there is one.
2672 */
2673static void
2674igb_rel_mgmt(struct igb_softc *sc)
2675{
396b7048 2676 if (sc->flags & IGB_FLAG_HAS_MGMT) {
1f7e3916
SZ		 2677 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
2678
2679 /* Re-enable hardware interception of ARP */
2680 manc |= E1000_MANC_ARP_EN;
2681 manc &= ~E1000_MANC_EN_MNG2HOST;
2682
2683 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
2684 }
2685}
2686
2687/*
2688 * Sets CTRL_EXT:DRV_LOAD bit.
2689 *
2690 * For ASF and Pass Through versions of f/w this means that
2691 * the driver is loaded.
2692 */
2693static void
2694igb_get_hw_control(struct igb_softc *sc)
2695{
2696 uint32_t ctrl_ext;
2697
2698 if (sc->vf_ifp)
2699 return;
2700
2701 /* Let firmware know the driver has taken over */
2702 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
2703 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
2704 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2705}
2706
2707/*
2708 * Resets CTRL_EXT:DRV_LOAD bit.
2709 *
2710 * For ASF and Pass Through versions of f/w this means that the
2711 * driver is no longer loaded.
2712 */
2713static void
2714igb_rel_hw_control(struct igb_softc *sc)
2715{
2716 uint32_t ctrl_ext;
2717
2718 if (sc->vf_ifp)
2719 return;
2720
2721 /* Let firmware taken over control of h/w */
2722 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
2723 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
2724 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2725}
2726
2727static int
2728igb_is_valid_ether_addr(const uint8_t *addr)
2729{
2730 uint8_t zero_addr[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 };
2731
2732 if ((addr[0] & 1) || !bcmp(addr, zero_addr, ETHER_ADDR_LEN))
2733 return FALSE;
2734 return TRUE;
2735}
2736
2737/*
2738 * Enable PCI Wake On Lan capability
2739 */
2740static void
2741igb_enable_wol(device_t dev)
2742{
2743 uint16_t cap, status;
2744 uint8_t id;
2745
2746 /* First find the capabilities pointer*/
2747 cap = pci_read_config(dev, PCIR_CAP_PTR, 2);
2748
2749 /* Read the PM Capabilities */
2750 id = pci_read_config(dev, cap, 1);
2751 if (id != PCIY_PMG) /* Something wrong */
2752 return;
2753
2754 /*
2755 * OK, we have the power capabilities,
2756 * so now get the status register
2757 */
2758 cap += PCIR_POWER_STATUS;
2759 status = pci_read_config(dev, cap, 2);
2760 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
2761 pci_write_config(dev, cap, status, 2);
2762}
2763
2764static void
2765igb_update_stats_counters(struct igb_softc *sc)
2766{
2767 struct e1000_hw *hw = &sc->hw;
2768 struct e1000_hw_stats *stats;
2769 struct ifnet *ifp = &sc->arpcom.ac_if;
2770
2771 /*
2772 * The virtual function adapter has only a
2773 * small controlled set of stats, do only
2774 * those and return.
2775 */
2776 if (sc->vf_ifp) {
2777 igb_update_vf_stats_counters(sc);
2778 return;
2779 }
2780 stats = sc->stats;
2781
2782 if (sc->hw.phy.media_type == e1000_media_type_copper ||
2783 (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
2784 stats->symerrs +=
2785 E1000_READ_REG(hw,E1000_SYMERRS);
2786 stats->sec += E1000_READ_REG(hw, E1000_SEC);
2787 }
2788
2789 stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
2790 stats->mpc += E1000_READ_REG(hw, E1000_MPC);
2791 stats->scc += E1000_READ_REG(hw, E1000_SCC);
2792 stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
2793
2794 stats->mcc += E1000_READ_REG(hw, E1000_MCC);
2795 stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
2796 stats->colc += E1000_READ_REG(hw, E1000_COLC);
2797 stats->dc += E1000_READ_REG(hw, E1000_DC);
2798 stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
2799 stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
2800 stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
2801
2802 /*
2803 * For watchdog management we need to know if we have been
2804 * paused during the last interval, so capture that here.
2805 */
2806 sc->pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
2807 stats->xoffrxc += sc->pause_frames;
2808 stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
2809 stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
2810 stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
2811 stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
2812 stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
2813 stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
2814 stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
2815 stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
2816 stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
2817 stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
2818 stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
2819 stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
2820
2821 /* For the 64-bit byte counters the low dword must be read first. */
2822 /* Both registers clear on the read of the high dword */
2823
2824 stats->gorc += E1000_READ_REG(hw, E1000_GORCL) +
2825 ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
2826 stats->gotc += E1000_READ_REG(hw, E1000_GOTCL) +
2827 ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
2828
2829 stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
2830 stats->ruc += E1000_READ_REG(hw, E1000_RUC);
2831 stats->rfc += E1000_READ_REG(hw, E1000_RFC);
2832 stats->roc += E1000_READ_REG(hw, E1000_ROC);
2833 stats->rjc += E1000_READ_REG(hw, E1000_RJC);
2834
2835 stats->tor += E1000_READ_REG(hw, E1000_TORH);
2836 stats->tot += E1000_READ_REG(hw, E1000_TOTH);
2837
2838 stats->tpr += E1000_READ_REG(hw, E1000_TPR);
2839 stats->tpt += E1000_READ_REG(hw, E1000_TPT);
2840 stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
2841 stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
2842 stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
2843 stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
2844 stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
2845 stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
2846 stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
2847 stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
2848
2849 /* Interrupt Counts */
2850
2851 stats->iac += E1000_READ_REG(hw, E1000_IAC);
2852 stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
2853 stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
2854 stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
2855 stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
2856 stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
2857 stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
2858 stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
2859 stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
2860
2861 /* Host to Card Statistics */
2862
2863 stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
2864 stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
2865 stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
2866 stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
2867 stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
2868 stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
2869 stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
2870 stats->hgorc += (E1000_READ_REG(hw, E1000_HGORCL) +
2871 ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32));
2872 stats->hgotc += (E1000_READ_REG(hw, E1000_HGOTCL) +
2873 ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32));
2874 stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
2875 stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
2876 stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
2877
2878 stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
2879 stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
2880 stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
2881 stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
2882 stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
2883 stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
2884
2885 ifp->if_collisions = stats->colc;
2886
2887 /* Rx Errors */
2888 ifp->if_ierrors = stats->rxerrc + stats->crcerrs + stats->algnerrc +
2889 stats->ruc + stats->roc + stats->mpc + stats->cexterr;
2890
2891 /* Tx Errors */
2892 ifp->if_oerrors = stats->ecol + stats->latecol + sc->watchdog_events;
2893
2894 /* Driver specific counters */
2895 sc->device_control = E1000_READ_REG(hw, E1000_CTRL);
2896 sc->rx_control = E1000_READ_REG(hw, E1000_RCTL);
2897 sc->int_mask = E1000_READ_REG(hw, E1000_IMS);
2898 sc->eint_mask = E1000_READ_REG(hw, E1000_EIMS);
2899 sc->packet_buf_alloc_tx =
2900 ((E1000_READ_REG(hw, E1000_PBA) & 0xffff0000) >> 16);
2901 sc->packet_buf_alloc_rx =
2902 (E1000_READ_REG(hw, E1000_PBA) & 0xffff);
2903}
2904
2905static void
2906igb_vf_init_stats(struct igb_softc *sc)
2907{
2908 struct e1000_hw *hw = &sc->hw;
2909 struct e1000_vf_stats *stats;
2910
2911 stats = sc->stats;
2912 stats->last_gprc = E1000_READ_REG(hw, E1000_VFGPRC);
2913 stats->last_gorc = E1000_READ_REG(hw, E1000_VFGORC);
2914 stats->last_gptc = E1000_READ_REG(hw, E1000_VFGPTC);
2915 stats->last_gotc = E1000_READ_REG(hw, E1000_VFGOTC);
2916 stats->last_mprc = E1000_READ_REG(hw, E1000_VFMPRC);
2917}
2918
2919static void
2920igb_update_vf_stats_counters(struct igb_softc *sc)
2921{
2922 struct e1000_hw *hw = &sc->hw;
2923 struct e1000_vf_stats *stats;
2924
2925 if (sc->link_speed == 0)
2926 return;
2927
2928 stats = sc->stats;
2929 UPDATE_VF_REG(E1000_VFGPRC, stats->last_gprc, stats->gprc);
2930 UPDATE_VF_REG(E1000_VFGORC, stats->last_gorc, stats->gorc);
2931 UPDATE_VF_REG(E1000_VFGPTC, stats->last_gptc, stats->gptc);
2932 UPDATE_VF_REG(E1000_VFGOTC, stats->last_gotc, stats->gotc);
2933 UPDATE_VF_REG(E1000_VFMPRC, stats->last_mprc, stats->mprc);
2934}
2935
d0f59cad 2936#ifdef IFPOLL_ENABLE
1f7e3916
SZ		 2937
2938static void
d0f59cad 2939igb_npoll_status(struct ifnet *ifp, int pollhz __unused)
1f7e3916
SZ		 2940{
2941 struct igb_softc *sc = ifp->if_softc;
2942 uint32_t reg_icr;
2943
d0f59cad 2944 ASSERT_SERIALIZED(&sc->main_serialize);
1f7e3916 2945
d0f59cad
SZ		 2946 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
2947 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
2948 sc->hw.mac.get_link_status = 1;
2949 igb_update_link_status(sc);
2950 }
2951}
7d235eb5 2952
d0f59cad
SZ		 2953static void
2954igb_npoll_tx(struct ifnet *ifp, void *arg, int cycle __unused)
2955{
2956 struct igb_tx_ring *txr = arg;
7d235eb5 2957
d0f59cad 2958 ASSERT_SERIALIZED(&txr->tx_serialize);
1f7e3916 2959
d0f59cad
SZ		 2960 igb_txeof(txr);
2961 if (!ifq_is_empty(&ifp->if_snd))
2962 if_devstart(ifp);
2963}
2964
2965static void
2966igb_npoll_rx(struct ifnet *ifp __unused, void *arg, int cycle)
2967{
2968 struct igb_rx_ring *rxr = arg;
2969
2970 ASSERT_SERIALIZED(&rxr->rx_serialize);
2971
2972 igb_rxeof(rxr, cycle);
2973}
2974
2975static void
2976igb_npoll(struct ifnet *ifp, struct ifpoll_info *info)
2977{
2978 struct igb_softc *sc = ifp->if_softc;
2979
2980 ASSERT_IFNET_SERIALIZED_ALL(ifp);
2981
2982 if (info) {
2983 struct igb_tx_ring *txr;
2984 int i, off;
2985
2986 info->ifpi_status.status_func = igb_npoll_status;
2987 info->ifpi_status.serializer = &sc->main_serialize;
2988
2989 off = sc->tx_npoll_off;
2990 KKASSERT(off < ncpus2);
2991 txr = &sc->tx_rings[0];
2992 info->ifpi_tx[off].poll_func = igb_npoll_tx;
2993 info->ifpi_tx[off].arg = txr;
2994 info->ifpi_tx[off].serializer = &txr->tx_serialize;
2995
2996 off = sc->rx_npoll_off;
2997 for (i = 0; i < sc->rx_ring_cnt; ++i) {
2998 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2999 int idx = i + off;
3000
3001 KKASSERT(idx < ncpus2);
3002 info->ifpi_rx[idx].poll_func = igb_npoll_rx;
3003 info->ifpi_rx[idx].arg = rxr;
3004 info->ifpi_rx[idx].serializer = &rxr->rx_serialize;
1f7e3916 3005 }
d0f59cad
SZ		 3006
3007 if (ifp->if_flags & IFF_RUNNING) {
3008 if (sc->rx_ring_inuse == sc->rx_ring_cnt)
3009 igb_disable_intr(sc);
3010 else
3011 igb_init(sc);
3012 }
3013 ifp->if_npoll_cpuid = sc->tx_npoll_off;
3014 } else {
3015 if (ifp->if_flags & IFF_RUNNING) {
3016 if (sc->rx_ring_inuse == sc->rx_ring_cnt)
3017 igb_enable_intr(sc);
3018 else
3019 igb_init(sc);
3020 }
3021 ifp->if_npoll_cpuid = -1;
1f7e3916
SZ		 3022 }
3023}
3024
d0f59cad 3025#endif /* IFPOLL_ENABLE */
1f7e3916
SZ		 3026
3027static void
3028igb_intr(void *xsc)
3029{
3030 struct igb_softc *sc = xsc;
3031 struct ifnet *ifp = &sc->arpcom.ac_if;
f6167a56
SZ		 3032 uint32_t eicr;
3033
7d235eb5 3034 ASSERT_SERIALIZED(&sc->main_serialize);
f6167a56
SZ		 3035
3036 eicr = E1000_READ_REG(&sc->hw, E1000_EICR);
3037
3038 if (eicr == 0)
3039 return;
3040
3041 if (ifp->if_flags & IFF_RUNNING) {
7d235eb5
SZ		 3042 struct igb_tx_ring *txr;
3043 int i;
f6167a56 3044
be922da6 3045 for (i = 0; i < sc->rx_ring_inuse; ++i) {
7d235eb5
SZ		 3046 struct igb_rx_ring *rxr = &sc->rx_rings[i];
3047
3048 if (eicr & rxr->rx_intr_mask) {
3049 lwkt_serialize_enter(&rxr->rx_serialize);
3050 igb_rxeof(rxr, -1);
3051 lwkt_serialize_exit(&rxr->rx_serialize);
3052 }
3053 }
3054
3055 txr = &sc->tx_rings[0];
3056 if (eicr & txr->tx_intr_mask) {
3057 lwkt_serialize_enter(&txr->tx_serialize);
3058 igb_txeof(txr);
f6167a56
SZ		 3059 if (!ifq_is_empty(&ifp->if_snd))
3060 if_devstart(ifp);
7d235eb5 3061 lwkt_serialize_exit(&txr->tx_serialize);
f6167a56
SZ		 3062 }
3063 }
3064
3065 if (eicr & E1000_EICR_OTHER) {
3066 uint32_t icr = E1000_READ_REG(&sc->hw, E1000_ICR);
3067
3068 /* Link status change */
3069 if (icr & E1000_ICR_LSC) {
3070 sc->hw.mac.get_link_status = 1;
3071 igb_update_link_status(sc);
3072 }
3073 }
3074
3075 /*
3076 * Reading EICR has the side effect to clear interrupt mask,
3077 * so all interrupts need to be enabled here.
3078 */
3079 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask);
3080}
3081
3082static void
9c0ecdcc 3083igb_intr_shared(void *xsc)
f6167a56
SZ		 3084{
3085 struct igb_softc *sc = xsc;
3086 struct ifnet *ifp = &sc->arpcom.ac_if;
1f7e3916
SZ		 3087 uint32_t reg_icr;
3088
7d235eb5 3089 ASSERT_SERIALIZED(&sc->main_serialize);
1f7e3916
SZ		 3090
3091 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
3092
3093 /* Hot eject? */
3094 if (reg_icr == 0xffffffff)
3095 return;
3096
3097 /* Definitely not our interrupt. */
3098 if (reg_icr == 0x0)
3099 return;
3100
3101 if ((reg_icr & E1000_ICR_INT_ASSERTED) == 0)
3102 return;
3103
3104 if (ifp->if_flags & IFF_RUNNING) {
71b8b086
SZ		 3105 if (reg_icr &
3106 (E1000_ICR_RXT0 | E1000_ICR_RXDMT0 | E1000_ICR_RXO)) {
3107 int i;
7d235eb5 3108
71b8b086
SZ		 3109 for (i = 0; i < sc->rx_ring_inuse; ++i) {
3110 struct igb_rx_ring *rxr = &sc->rx_rings[i];
1f7e3916 3111
71b8b086
SZ		 3112 lwkt_serialize_enter(&rxr->rx_serialize);
3113 igb_rxeof(rxr, -1);
3114 lwkt_serialize_exit(&rxr->rx_serialize);
3115 }
7d235eb5
SZ		 3116 }
3117
71b8b086
SZ		 3118 if (reg_icr & E1000_ICR_TXDW) {
3119 struct igb_tx_ring *txr = &sc->tx_rings[0];
3120
3121 lwkt_serialize_enter(&txr->tx_serialize);
3122 igb_txeof(txr);
3123 if (!ifq_is_empty(&ifp->if_snd))
3124 if_devstart(ifp);
3125 lwkt_serialize_exit(&txr->tx_serialize);
3126 }
1f7e3916
SZ		 3127 }
3128
3129 /* Link status change */
3130 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
3131 sc->hw.mac.get_link_status = 1;
3132 igb_update_link_status(sc);
3133 }
3134
3135 if (reg_icr & E1000_ICR_RXO)
3136 sc->rx_overruns++;
3137}
3138
3139static int
1f7e3916
SZ		 3140igb_encap(struct igb_tx_ring *txr, struct mbuf **m_headp)
3141{
3142 bus_dma_segment_t segs[IGB_MAX_SCATTER];
3143 bus_dmamap_t map;
3144 struct igb_tx_buf *tx_buf, *tx_buf_mapped;
3145 union e1000_adv_tx_desc *txd = NULL;
3146 struct mbuf *m_head = *m_headp;
b6220144
SZ		 3147 uint32_t olinfo_status = 0, cmd_type_len = 0, cmd_rs = 0;
3148 int maxsegs, nsegs, i, j, error, last = 0;
1f7e3916
SZ		 3149 uint32_t hdrlen = 0;
3150
23f6ffe4
SZ		 3151 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3152 error = igb_tso_pullup(txr, m_headp);
3153 if (error)
3154 return error;
3155 m_head = *m_headp;
3156 }
3157
1f7e3916
SZ		 3158 /* Set basic descriptor constants */
3159 cmd_type_len |= E1000_ADVTXD_DTYP_DATA;
3160 cmd_type_len |= E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT;
3161 if (m_head->m_flags & M_VLANTAG)
3162 cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
3163
3164 /*
3165 * Map the packet for DMA.
1f7e3916 3166 */
b6220144 3167 tx_buf = &txr->tx_buf[txr->next_avail_desc];
1f7e3916
SZ		 3168 tx_buf_mapped = tx_buf;
3169 map = tx_buf->map;
3170
b6220144
SZ		 3171 maxsegs = txr->tx_avail - IGB_TX_RESERVED;
3172 KASSERT(maxsegs >= txr->spare_desc, ("not enough spare TX desc\n"));
1f7e3916
SZ		 3173 if (maxsegs > IGB_MAX_SCATTER)
3174 maxsegs = IGB_MAX_SCATTER;
3175
3176 error = bus_dmamap_load_mbuf_defrag(txr->tx_tag, map, m_headp,
3177 segs, maxsegs, &nsegs, BUS_DMA_NOWAIT);
3178 if (error) {
3179 if (error == ENOBUFS)
3180 txr->sc->mbuf_defrag_failed++;
3181 else
3182 txr->sc->no_tx_dma_setup++;
3183
3184 m_freem(*m_headp);
3185 *m_headp = NULL;
3186 return error;
3187 }
3188 bus_dmamap_sync(txr->tx_tag, map, BUS_DMASYNC_PREWRITE);
3189
3190 m_head = *m_headp;
3191
1f7e3916 3192 /*
66c68b4b
SZ		 3193 * Set up the TX context descriptor, if any hardware offloading is
3194 * needed. This includes CSUM, VLAN, and TSO. It will consume one
3195 * TX descriptor.
3196 *
3197 * Unlike these chips' predecessors (em/emx), TX context descriptor
3198 * will _not_ interfere TX data fetching pipelining.
1f7e3916
SZ		 3199 */
3200 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
23f6ffe4
SZ		 3201 igb_tso_ctx(txr, m_head, &hdrlen);
3202 cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
3203 olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
3204 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3205 txr->tx_nsegs++;
3206 } else if (igb_txcsum_ctx(txr, m_head)) {
48faa653
SZ		 3207 if (m_head->m_pkthdr.csum_flags & CSUM_IP)
3208 olinfo_status |= (E1000_TXD_POPTS_IXSM << 8);
1f7e3916
SZ		 3209 if (m_head->m_pkthdr.csum_flags & (CSUM_UDP | CSUM_TCP))
3210 olinfo_status |= (E1000_TXD_POPTS_TXSM << 8);
b6220144 3211 txr->tx_nsegs++;
1f7e3916 3212 }
1f7e3916 3213
b6220144
SZ		 3214 txr->tx_nsegs += nsegs;
3215 if (txr->tx_nsegs >= txr->intr_nsegs) {
3216 /*
3217 * Report Status (RS) is turned on every intr_nsegs
3218 * descriptors (roughly).
3219 */
3220 txr->tx_nsegs = 0;
3221 cmd_rs = E1000_ADVTXD_DCMD_RS;
3222 }
3223
1f7e3916
SZ		 3224 /* Calculate payload length */
3225 olinfo_status |= ((m_head->m_pkthdr.len - hdrlen)
3226 << E1000_ADVTXD_PAYLEN_SHIFT);
3227
3228 /* 82575 needs the queue index added */
3229 if (txr->sc->hw.mac.type == e1000_82575)
3230 olinfo_status |= txr->me << 4;
3231
3232 /* Set up our transmit descriptors */
3233 i = txr->next_avail_desc;
3234 for (j = 0; j < nsegs; j++) {
3235 bus_size_t seg_len;
3236 bus_addr_t seg_addr;
3237
3238 tx_buf = &txr->tx_buf[i];
3239 txd = (union e1000_adv_tx_desc *)&txr->tx_base[i];
3240 seg_addr = segs[j].ds_addr;
3241 seg_len = segs[j].ds_len;
3242
3243 txd->read.buffer_addr = htole64(seg_addr);
3244 txd->read.cmd_type_len = htole32(cmd_type_len | seg_len);
3245 txd->read.olinfo_status = htole32(olinfo_status);
3246 last = i;
91b8700a 3247 if (++i == txr->num_tx_desc)
1f7e3916
SZ		 3248 i = 0;
3249 tx_buf->m_head = NULL;
1f7e3916
SZ		 3250 }
3251
3252 KASSERT(txr->tx_avail > nsegs, ("invalid avail TX desc\n"));
3253 txr->next_avail_desc = i;
3254 txr->tx_avail -= nsegs;
3255
3256 tx_buf->m_head = m_head;
3257 tx_buf_mapped->map = tx_buf->map;
3258 tx_buf->map = map;
3259
3260 /*
b6220144 3261 * Last Descriptor of Packet needs End Of Packet (EOP)
1f7e3916 3262 */
b6220144 3263 txd->read.cmd_type_len |= htole32(E1000_ADVTXD_DCMD_EOP | cmd_rs);
1f7e3916
SZ		 3264
3265 /*
3266 * Advance the Transmit Descriptor Tail (TDT), this tells the E1000
3267 * that this frame is available to transmit.
3268 */
3269 E1000_WRITE_REG(&txr->sc->hw, E1000_TDT(txr->me), i);
3270 ++txr->tx_packets;
3271
3272 return 0;
3273}
3274
3275static void
3276igb_start(struct ifnet *ifp)
3277{
3278 struct igb_softc *sc = ifp->if_softc;
27866bf1 3279 struct igb_tx_ring *txr = &sc->tx_rings[0];
1f7e3916
SZ		 3280 struct mbuf *m_head;
3281
7d235eb5 3282 ASSERT_SERIALIZED(&txr->tx_serialize);
1f7e3916 3283
9c0ecdcc 3284 if ((ifp->if_flags & (IFF_RUNNING | IFF_OACTIVE)) != IFF_RUNNING)
1f7e3916
SZ		 3285 return;
3286
3287 if (!sc->link_active) {
3288 ifq_purge(&ifp->if_snd);
3289 return;
3290 }
3291
b6220144 3292 if (!IGB_IS_NOT_OACTIVE(txr))
1f7e3916
SZ		 3293 igb_txeof(txr);
3294
3295 while (!ifq_is_empty(&ifp->if_snd)) {
b6220144 3296 if (IGB_IS_OACTIVE(txr)) {
1f7e3916
SZ		 3297 ifp->if_flags |= IFF_OACTIVE;
3298 /* Set watchdog on */
3299 ifp->if_timer = 5;
3300 break;
3301 }
3302
3303 m_head = ifq_dequeue(&ifp->if_snd, NULL);
3304 if (m_head == NULL)
3305 break;
3306
3307 if (igb_encap(txr, &m_head)) {
3308 ifp->if_oerrors++;
3309 continue;
3310 }
3311
3312 /* Send a copy of the frame to the BPF listener */
3313 ETHER_BPF_MTAP(ifp, m_head);
3314 }
3315}
3316
3317static void
3318igb_watchdog(struct ifnet *ifp)
3319{
3320 struct igb_softc *sc = ifp->if_softc;
27866bf1 3321 struct igb_tx_ring *txr = &sc->tx_rings[0];
1f7e3916
SZ		 3322
3323 ASSERT_IFNET_SERIALIZED_ALL(ifp);
3324
3325 /*
3326 * If flow control has paused us since last checking
3327 * it invalidates the watchdog timing, so dont run it.
3328 */
3329 if (sc->pause_frames) {
3330 sc->pause_frames = 0;
3331 ifp->if_timer = 5;
3332 return;
3333 }
3334
3335 if_printf(ifp, "Watchdog timeout -- resetting\n");
3336 if_printf(ifp, "Queue(%d) tdh = %d, hw tdt = %d\n", txr->me,
3337 E1000_READ_REG(&sc->hw, E1000_TDH(txr->me)),
3338 E1000_READ_REG(&sc->hw, E1000_TDT(txr->me)));
3339 if_printf(ifp, "TX(%d) desc avail = %d, "
3340 "Next TX to Clean = %d\n",
3341 txr->me, txr->tx_avail, txr->next_to_clean);
3342
3343 ifp->if_oerrors++;
3344 sc->watchdog_events++;
3345
3346 igb_init(sc);
3347 if (!ifq_is_empty(&ifp->if_snd))
3348 if_devstart(ifp);
3349}
3350
3351static void
9c0ecdcc 3352igb_set_eitr(struct igb_softc *sc, int idx, int rate)
1f7e3916 3353{
9c0ecdcc 3354 uint32_t eitr = 0;
1f7e3916 3355
9c0ecdcc 3356 if (rate > 0) {
1f7e3916 3357 if (sc->hw.mac.type == e1000_82575) {
9c0ecdcc 3358 eitr = 1000000000 / 256 / rate;
1f7e3916
SZ		 3359 /*
3360 * NOTE:
3361 * Document is wrong on the 2 bits left shift
3362 */
3363 } else {
9c0ecdcc 3364 eitr = 1000000 / rate;
d4beffa9
SZ		 3365 eitr <<= IGB_EITR_INTVL_SHIFT;
3366 }
3367
3368 if (eitr == 0) {
3369 /* Don't disable it */
3370 eitr = 1 << IGB_EITR_INTVL_SHIFT;
3371 } else if (eitr > IGB_EITR_INTVL_MASK) {
3372 /* Don't allow it to be too large */
3373 eitr = IGB_EITR_INTVL_MASK;
1f7e3916 3374 }
1f7e3916
SZ		 3375 }
3376 if (sc->hw.mac.type == e1000_82575)
9c0ecdcc 3377 eitr |= eitr << 16;
1f7e3916 3378 else
9c0ecdcc
SZ		 3379 eitr |= E1000_EITR_CNT_IGNR;
3380 E1000_WRITE_REG(&sc->hw, E1000_EITR(idx), eitr);
1f7e3916
SZ		 3381}
3382
3383static int
3384igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS)
3385{
3386 struct igb_softc *sc = (void *)arg1;
3387 struct ifnet *ifp = &sc->arpcom.ac_if;
3388 int error, intr_rate;
3389
3390 intr_rate = sc->intr_rate;
3391 error = sysctl_handle_int(oidp, &intr_rate, 0, req);
3392 if (error || req->newptr == NULL)
3393 return error;
3394 if (intr_rate < 0)
3395 return EINVAL;
3396
3397 ifnet_serialize_all(ifp);
3398
3399 sc->intr_rate = intr_rate;
3400 if (ifp->if_flags & IFF_RUNNING)
9c0ecdcc
SZ		 3401 igb_set_eitr(sc, 0, sc->intr_rate);
3402
3403 if (bootverbose)
3404 if_printf(ifp, "interrupt rate set to %d/sec\n", sc->intr_rate);
1f7e3916
SZ		 3405
3406 ifnet_deserialize_all(ifp);
3407
9c0ecdcc
SZ		 3408 return 0;
3409}
3410
3411static int
3412igb_sysctl_msix_rate(SYSCTL_HANDLER_ARGS)
3413{
3414 struct igb_msix_data *msix = (void *)arg1;
3415 struct igb_softc *sc = msix->msix_sc;
3416 struct ifnet *ifp = &sc->arpcom.ac_if;
3417 int error, msix_rate;
3418
3419 msix_rate = msix->msix_rate;
3420 error = sysctl_handle_int(oidp, &msix_rate, 0, req);
3421 if (error || req->newptr == NULL)
3422 return error;
3423 if (msix_rate < 0)
3424 return EINVAL;
3425
3426 lwkt_serialize_enter(msix->msix_serialize);
3427
3428 msix->msix_rate = msix_rate;
3429 if (ifp->if_flags & IFF_RUNNING)
3430 igb_set_eitr(sc, msix->msix_vector, msix->msix_rate);
3431
3432 if (bootverbose) {
3433 if_printf(ifp, "%s set to %d/sec\n", msix->msix_rate_desc,
3434 msix->msix_rate);
3435 }
3436
3437 lwkt_serialize_exit(msix->msix_serialize);
3438
1f7e3916
SZ		 3439 return 0;
3440}
b6220144
SZ		 3441
3442static int
3443igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS)
3444{
3445 struct igb_softc *sc = (void *)arg1;
3446 struct ifnet *ifp = &sc->arpcom.ac_if;
27866bf1 3447 struct igb_tx_ring *txr = &sc->tx_rings[0];
b6220144
SZ		 3448 int error, nsegs;
3449
3450 nsegs = txr->intr_nsegs;
3451 error = sysctl_handle_int(oidp, &nsegs, 0, req);
3452 if (error || req->newptr == NULL)
3453 return error;
3454 if (nsegs <= 0)
3455 return EINVAL;
3456
3457 ifnet_serialize_all(ifp);
3458
91b8700a 3459 if (nsegs >= txr->num_tx_desc - txr->oact_lo_desc ||
b6220144
SZ		 3460 nsegs >= txr->oact_hi_desc - IGB_MAX_SCATTER) {
3461 error = EINVAL;
3462 } else {
3463 error = 0;
3464 txr->intr_nsegs = nsegs;
3465 }
3466
3467 ifnet_deserialize_all(ifp);
3468
3469 return error;
3470}
f6167a56 3471
d0f59cad
SZ		 3472#ifdef IFPOLL_ENABLE
3473
3474static int
3475igb_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS)
3476{
3477 struct igb_softc *sc = (void *)arg1;
3478 struct ifnet *ifp = &sc->arpcom.ac_if;
3479 int error, off;
3480
3481 off = sc->rx_npoll_off;
3482 error = sysctl_handle_int(oidp, &off, 0, req);
3483 if (error || req->newptr == NULL)
3484 return error;
3485 if (off < 0)
3486 return EINVAL;
3487
3488 ifnet_serialize_all(ifp);
3489 if (off >= ncpus2 || off % sc->rx_ring_cnt != 0) {
3490 error = EINVAL;
3491 } else {
3492 error = 0;
3493 sc->rx_npoll_off = off;
3494 }
3495 ifnet_deserialize_all(ifp);
3496
3497 return error;
3498}
3499
3500static int
3501igb_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS)
3502{
3503 struct igb_softc *sc = (void *)arg1;
3504 struct ifnet *ifp = &sc->arpcom.ac_if;
3505 int error, off;
3506
3507 off = sc->tx_npoll_off;
3508 error = sysctl_handle_int(oidp, &off, 0, req);
3509 if (error || req->newptr == NULL)
3510 return error;
3511 if (off < 0)
3512 return EINVAL;
3513
3514 ifnet_serialize_all(ifp);
3515 if (off >= ncpus2) {
3516 error = EINVAL;
3517 } else {
3518 error = 0;
3519 sc->tx_npoll_off = off;
3520 }
3521 ifnet_deserialize_all(ifp);
3522
3523 return error;
3524}
3525
3526#endif /* IFPOLL_ENABLE */
3527
f6167a56
SZ		 3528static void
3529igb_init_intr(struct igb_softc *sc)
3530{
be922da6 3531 igb_set_intr_mask(sc);
9c0ecdcc
SZ		 3532
3533 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0)
f6167a56 3534 igb_init_unshared_intr(sc);
9c0ecdcc
SZ		 3535
3536 if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
3537 igb_set_eitr(sc, 0, sc->intr_rate);
3538 } else {
3539 int i;
3540
3541 for (i = 0; i < sc->msix_cnt; ++i)
3542 igb_set_eitr(sc, i, sc->msix_data[i].msix_rate);
3543 }
f6167a56
SZ		 3544}
3545
3546static void
3547igb_init_unshared_intr(struct igb_softc *sc)
3548{
3549 struct e1000_hw *hw = &sc->hw;
3550 const struct igb_rx_ring *rxr;
3551 const struct igb_tx_ring *txr;
3552 uint32_t ivar, index;
3553 int i;
3554
3555 /*
3556 * Enable extended mode
3557 */
3558 if (sc->hw.mac.type != e1000_82575) {
9c0ecdcc
SZ		 3559 uint32_t gpie;
3560 int ivar_max;
3561
3562 gpie = E1000_GPIE_NSICR;
3563 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3564 gpie |= E1000_GPIE_MSIX_MODE |
3565 E1000_GPIE_EIAME |
3566 E1000_GPIE_PBA;
3567 }
3568 E1000_WRITE_REG(hw, E1000_GPIE, gpie);
3569
3570 /*
3571 * Clear IVARs
3572 */
3573 switch (sc->hw.mac.type) {
3574 case e1000_82580:
3575 ivar_max = IGB_MAX_IVAR_82580;
3576 break;
3577
3578 case e1000_i350:
3579 ivar_max = IGB_MAX_IVAR_I350;
3580 break;
3581
3582 case e1000_vfadapt:
3583 case e1000_vfadapt_i350:
3584 ivar_max = IGB_MAX_IVAR_VF;
3585 break;
3586
3587 case e1000_82576:
3588 ivar_max = IGB_MAX_IVAR_82576;
3589 break;
3590
3591 default:
3592 panic("unknown mac type %d\n", sc->hw.mac.type);
3593 }
3594 for (i = 0; i < ivar_max; ++i)
3595 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, i, 0);
3596 E1000_WRITE_REG(hw, E1000_IVAR_MISC, 0);
f6167a56
SZ		 3597 } else {
3598 uint32_t tmp;
3599
9c0ecdcc
SZ		 3600 KASSERT(sc->intr_type != PCI_INTR_TYPE_MSIX,
3601 ("82575 w/ MSI-X"));
f6167a56
SZ		 3602 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
3603 tmp |= E1000_CTRL_EXT_IRCA;
3604 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
3605 }
3606
3607 /*
3608 * Map TX/RX interrupts to EICR
3609 */
3610 switch (sc->hw.mac.type) {
3611 case e1000_82580:
3612 case e1000_i350:
3613 case e1000_vfadapt:
3614 case e1000_vfadapt_i350:
3615 /* RX entries */
be922da6 3616 for (i = 0; i < sc->rx_ring_inuse; ++i) {
f6167a56
SZ		 3617 rxr = &sc->rx_rings[i];
3618
3619 index = i >> 1;
3620 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3621
3622 if (i & 1) {
3623 ivar &= 0xff00ffff;
3624 ivar |=
3625 (rxr->rx_intr_bit | E1000_IVAR_VALID) << 16;
3626 } else {
3627 ivar &= 0xffffff00;
3628 ivar |=
3629 (rxr->rx_intr_bit | E1000_IVAR_VALID);
3630 }
3631 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3632 }
3633 /* TX entries */
27866bf1 3634 for (i = 0; i < sc->tx_ring_cnt; ++i) {
f6167a56
SZ		 3635 txr = &sc->tx_rings[i];
3636
3637 index = i >> 1;
3638 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3639
3640 if (i & 1) {
3641 ivar &= 0x00ffffff;
3642 ivar |=
3643 (txr->tx_intr_bit | E1000_IVAR_VALID) << 24;
3644 } else {
3645 ivar &= 0xffff00ff;
3646 ivar |=
3647 (txr->tx_intr_bit | E1000_IVAR_VALID) << 8;
3648 }
3649 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3650 }
9c0ecdcc
SZ		 3651 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3652 ivar = (sc->sts_intr_bit | E1000_IVAR_VALID) << 8;
3653 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
3654 }
f6167a56
SZ		 3655 break;
3656
3657 case e1000_82576:
3658 /* RX entries */
be922da6 3659 for (i = 0; i < sc->rx_ring_inuse; ++i) {
f6167a56
SZ		 3660 rxr = &sc->rx_rings[i];
3661
3662 index = i & 0x7; /* Each IVAR has two entries */
3663 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3664
3665 if (i < 8) {
3666 ivar &= 0xffffff00;
3667 ivar |=
3668 (rxr->rx_intr_bit | E1000_IVAR_VALID);
3669 } else {
3670 ivar &= 0xff00ffff;
3671 ivar |=
3672 (rxr->rx_intr_bit | E1000_IVAR_VALID) << 16;
3673 }
3674 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3675 }
3676 /* TX entries */
27866bf1 3677 for (i = 0; i < sc->tx_ring_cnt; ++i) {
f6167a56
SZ		 3678 txr = &sc->tx_rings[i];
3679
3680 index = i & 0x7; /* Each IVAR has two entries */
3681 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3682
3683 if (i < 8) {
3684 ivar &= 0xffff00ff;
3685 ivar |=
3686 (txr->tx_intr_bit | E1000_IVAR_VALID) << 8;
3687 } else {
3688 ivar &= 0x00ffffff;
3689 ivar |=
3690 (txr->tx_intr_bit | E1000_IVAR_VALID) << 24;
3691 }
3692 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3693 }
9c0ecdcc
SZ		 3694 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3695 ivar = (sc->sts_intr_bit | E1000_IVAR_VALID) << 8;
3696 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
3697 }
f6167a56
SZ		 3698 break;
3699
3700 case e1000_82575:
3701 /*
3702 * Enable necessary interrupt bits.
3703 *
3704 * The name of the register is confusing; in addition to
3705 * configuring the first vector of MSI-X, it also configures
3706 * which bits of EICR could be set by the hardware even when
3707 * MSI or line interrupt is used; it thus controls interrupt
3708 * generation. It MUST be configured explicitly; the default
3709 * value mentioned in the datasheet is wrong: RX queue0 and
3710 * TX queue0 are NOT enabled by default.
3711 */
3712 E1000_WRITE_REG(&sc->hw, E1000_MSIXBM(0), sc->intr_mask);
3713 break;
3714
3715 default:
9c0ecdcc 3716 panic("unknown mac type %d\n", sc->hw.mac.type);
f6167a56 3717 }
f6167a56
SZ		 3718}
3719
3720static int
3721igb_setup_intr(struct igb_softc *sc)
3722{
3723 struct ifnet *ifp = &sc->arpcom.ac_if;
9c0ecdcc 3724 int error;
f6167a56 3725
9c0ecdcc
SZ		 3726 if (sc->intr_type == PCI_INTR_TYPE_MSIX)
3727 return igb_msix_setup(sc);
f6167a56 3728
f6167a56 3729 error = bus_setup_intr(sc->dev, sc->intr_res, INTR_MPSAFE,
9c0ecdcc 3730 (sc->flags & IGB_FLAG_SHARED_INTR) ? igb_intr_shared : igb_intr,
7d235eb5 3731 sc, &sc->intr_tag, &sc->main_serialize);
f6167a56
SZ		 3732 if (error) {
3733 device_printf(sc->dev, "Failed to register interrupt handler");
3734 return error;
3735 }
3736
3737 ifp->if_cpuid = rman_get_cpuid(sc->intr_res);
3738 KKASSERT(ifp->if_cpuid >= 0 && ifp->if_cpuid < ncpus);
3739
3740 return 0;
3741}
3742
3743static void
9c0ecdcc 3744igb_set_txintr_mask(struct igb_tx_ring *txr, int *intr_bit0, int intr_bitmax)
f6167a56
SZ		 3745{
3746 if (txr->sc->hw.mac.type == e1000_82575) {
3747 txr->tx_intr_bit = 0; /* unused */
3748 switch (txr->me) {
3749 case 0:
3750 txr->tx_intr_mask = E1000_EICR_TX_QUEUE0;
3751 break;
3752 case 1:
3753 txr->tx_intr_mask = E1000_EICR_TX_QUEUE1;
3754 break;
3755 case 2:
3756 txr->tx_intr_mask = E1000_EICR_TX_QUEUE2;
3757 break;
3758 case 3:
3759 txr->tx_intr_mask = E1000_EICR_TX_QUEUE3;
3760 break;
3761 default:
3762 panic("unsupported # of TX ring, %d\n", txr->me);
3763 }
3764 } else {
8d6600da
SZ		 3765 int intr_bit = *intr_bit0;
3766
3767 txr->tx_intr_bit = intr_bit % intr_bitmax;
f6167a56 3768 txr->tx_intr_mask = 1 << txr->tx_intr_bit;
8d6600da
SZ		 3769
3770 *intr_bit0 = intr_bit + 1;
f6167a56
SZ		 3771 }
3772}
3773
3774static void
9c0ecdcc 3775igb_set_rxintr_mask(struct igb_rx_ring *rxr, int *intr_bit0, int intr_bitmax)
f6167a56
SZ		 3776{
3777 if (rxr->sc->hw.mac.type == e1000_82575) {
3778 rxr->rx_intr_bit = 0; /* unused */
3779 switch (rxr->me) {
3780 case 0:
3781 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE0;
3782 break;
3783 case 1:
3784 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE1;
3785 break;
3786 case 2:
3787 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE2;
3788 break;
3789 case 3:
3790 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE3;
3791 break;
3792 default:
3793 panic("unsupported # of RX ring, %d\n", rxr->me);
3794 }
3795 } else {
8d6600da
SZ		 3796 int intr_bit = *intr_bit0;
3797
3798 rxr->rx_intr_bit = intr_bit % intr_bitmax;
f6167a56 3799 rxr->rx_intr_mask = 1 << rxr->rx_intr_bit;
8d6600da
SZ		 3800
3801 *intr_bit0 = intr_bit + 1;
f6167a56
SZ		 3802 }
3803}
7d235eb5
SZ		 3804
3805static void
3806igb_serialize(struct ifnet *ifp, enum ifnet_serialize slz)
3807{
3808 struct igb_softc *sc = ifp->if_softc;
3809
3810 ifnet_serialize_array_enter(sc->serializes, sc->serialize_cnt,
3811 sc->tx_serialize, sc->rx_serialize, slz);
3812}
3813
3814static void
3815igb_deserialize(struct ifnet *ifp, enum ifnet_serialize slz)
3816{
3817 struct igb_softc *sc = ifp->if_softc;
3818
3819 ifnet_serialize_array_exit(sc->serializes, sc->serialize_cnt,
3820 sc->tx_serialize, sc->rx_serialize, slz);
3821}
3822
3823static int
3824igb_tryserialize(struct ifnet *ifp, enum ifnet_serialize slz)
3825{
3826 struct igb_softc *sc = ifp->if_softc;
3827
3828 return ifnet_serialize_array_try(sc->serializes, sc->serialize_cnt,
3829 sc->tx_serialize, sc->rx_serialize, slz);
3830}
3831
7d235eb5
SZ		 3832#ifdef INVARIANTS
3833
3834static void
3835igb_serialize_assert(struct ifnet *ifp, enum ifnet_serialize slz,
3836 boolean_t serialized)
3837{
3838 struct igb_softc *sc = ifp->if_softc;
3839
3840 ifnet_serialize_array_assert(sc->serializes, sc->serialize_cnt,
3841 sc->tx_serialize, sc->rx_serialize, slz, serialized);
3842}
3843
3844#endif /* INVARIANTS */
be922da6
SZ		 3845
3846static void
3847igb_set_intr_mask(struct igb_softc *sc)
3848{
3849 int i;
3850
9c0ecdcc 3851 sc->intr_mask = sc->sts_intr_mask;
be922da6
SZ		 3852 for (i = 0; i < sc->rx_ring_inuse; ++i)
3853 sc->intr_mask |= sc->rx_rings[i].rx_intr_mask;
3854 for (i = 0; i < sc->tx_ring_cnt; ++i)
3855 sc->intr_mask |= sc->tx_rings[i].tx_intr_mask;
62be5890
SZ		 3856 if (bootverbose) {
3857 if_printf(&sc->arpcom.ac_if, "intr mask 0x%08x\n",
3858 sc->intr_mask);
3859 }
be922da6 3860}
3c7cc5e2
SZ		 3861
3862static int
3863igb_alloc_intr(struct igb_softc *sc)