sys/dev/netif/igb/if_igb.c

   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
  32 #include "opt_polling.h"
  33 #include "opt_igb.h"
  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
  77 #ifdef IGB_RSS_DEBUG
  78 #define IGB_RSS_DPRINTF(sc, lvl, fmt, ...) \
  79 do { \
  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
  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
  92 static 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
 129 static int      igb_probe(device_t);
 130 static int      igb_attach(device_t);
 131 static int      igb_detach(device_t);
 132 static int      igb_shutdown(device_t);
 133 static int      igb_suspend(device_t);
 134 static int      igb_resume(device_t);
 135
 136 static boolean_t igb_is_valid_ether_addr(const uint8_t *);
 137 static void     igb_setup_ifp(struct igb_softc *);
 138 static boolean_t igb_txcsum_ctx(struct igb_tx_ring *, struct mbuf *);
 139 static int      igb_tso_pullup(struct igb_tx_ring *, struct mbuf **);
 140 static void     igb_tso_ctx(struct igb_tx_ring *, struct mbuf *, uint32_t *);
 141 static void     igb_add_sysctl(struct igb_softc *);
 142 static int      igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS);
 143 static int      igb_sysctl_msix_rate(SYSCTL_HANDLER_ARGS);
 144 static int      igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS);
 145 static void     igb_set_ring_inuse(struct igb_softc *, boolean_t);
 146
 147 static void     igb_vf_init_stats(struct igb_softc *);
 148 static void     igb_reset(struct igb_softc *);
 149 static void     igb_update_stats_counters(struct igb_softc *);
 150 static void     igb_update_vf_stats_counters(struct igb_softc *);
 151 static void     igb_update_link_status(struct igb_softc *);
 152 static void     igb_init_tx_unit(struct igb_softc *);
 153 static void     igb_init_rx_unit(struct igb_softc *);
 154
 155 static void     igb_set_vlan(struct igb_softc *);
 156 static void     igb_set_multi(struct igb_softc *);
 157 static void     igb_set_promisc(struct igb_softc *);
 158 static void     igb_disable_promisc(struct igb_softc *);
 159
 160 static int      igb_alloc_rings(struct igb_softc *);
 161 static void     igb_free_rings(struct igb_softc *);
 162 static int      igb_create_tx_ring(struct igb_tx_ring *);
 163 static int      igb_create_rx_ring(struct igb_rx_ring *);
 164 static void     igb_free_tx_ring(struct igb_tx_ring *);
 165 static void     igb_free_rx_ring(struct igb_rx_ring *);
 166 static void     igb_destroy_tx_ring(struct igb_tx_ring *, int);
 167 static void     igb_destroy_rx_ring(struct igb_rx_ring *, int);
 168 static void     igb_init_tx_ring(struct igb_tx_ring *);
 169 static int      igb_init_rx_ring(struct igb_rx_ring *);
 170 static int      igb_newbuf(struct igb_rx_ring *, int, boolean_t);
 171 static int      igb_encap(struct igb_tx_ring *, struct mbuf **);
 172
 173 static void     igb_stop(struct igb_softc *);
 174 static void     igb_init(void *);
 175 static int      igb_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
 176 static void     igb_media_status(struct ifnet *, struct ifmediareq *);
 177 static int      igb_media_change(struct ifnet *);
 178 static void     igb_timer(void *);
 179 static void     igb_watchdog(struct ifnet *);
 180 static void     igb_start(struct ifnet *);
 181 #ifdef DEVICE_POLLING
 182 static void     igb_poll(struct ifnet *, enum poll_cmd, int);
 183 #endif
 184 static void     igb_serialize(struct ifnet *, enum ifnet_serialize);
 185 static void     igb_deserialize(struct ifnet *, enum ifnet_serialize);
 186 static int      igb_tryserialize(struct ifnet *, enum ifnet_serialize);
 187 #ifdef INVARIANTS
 188 static void     igb_serialize_assert(struct ifnet *, enum ifnet_serialize,
 189                     boolean_t);
 190 #endif
 191
 192 static void     igb_intr(void *);
 193 static void     igb_intr_shared(void *);
 194 static void     igb_rxeof(struct igb_rx_ring *, int);
 195 static void     igb_txeof(struct igb_tx_ring *);
 196 static void     igb_set_eitr(struct igb_softc *, int, int);
 197 static void     igb_enable_intr(struct igb_softc *);
 198 static void     igb_disable_intr(struct igb_softc *);
 199 static void     igb_init_unshared_intr(struct igb_softc *);
 200 static void     igb_init_intr(struct igb_softc *);
 201 static int      igb_setup_intr(struct igb_softc *);
 202 static void     igb_set_txintr_mask(struct igb_tx_ring *, int *, int);
 203 static void     igb_set_rxintr_mask(struct igb_rx_ring *, int *, int);
 204 static void     igb_set_intr_mask(struct igb_softc *);
 205 static int      igb_alloc_intr(struct igb_softc *);
 206 static void     igb_free_intr(struct igb_softc *);
 207 static void     igb_teardown_intr(struct igb_softc *);
 208 static void     igb_msix_try_alloc(struct igb_softc *);
 209 static void     igb_msix_free(struct igb_softc *, boolean_t);
 210 static int      igb_msix_setup(struct igb_softc *);
 211 static void     igb_msix_teardown(struct igb_softc *, int);
 212 static void     igb_msix_rx(void *);
 213 static void     igb_msix_tx(void *);
 214 static void     igb_msix_status(void *);
 215
 216 /* Management and WOL Support */
 217 static void     igb_get_mgmt(struct igb_softc *);
 218 static void     igb_rel_mgmt(struct igb_softc *);
 219 static void     igb_get_hw_control(struct igb_softc *);
 220 static void     igb_rel_hw_control(struct igb_softc *);
 221 static void     igb_enable_wol(device_t);
 222
 223 static device_method_t igb_methods[] = {
 224         /* Device interface */
 225         DEVMETHOD(device_probe,         igb_probe),
 226         DEVMETHOD(device_attach,        igb_attach),
 227         DEVMETHOD(device_detach,        igb_detach),
 228         DEVMETHOD(device_shutdown,      igb_shutdown),
 229         DEVMETHOD(device_suspend,       igb_suspend),
 230         DEVMETHOD(device_resume,        igb_resume),
 231         { 0, 0 }
 232 };
 233
 234 static driver_t igb_driver = {
 235         "igb",
 236         igb_methods,
 237         sizeof(struct igb_softc),
 238 };
 239
 240 static devclass_t igb_devclass;
 241
 242 DECLARE_DUMMY_MODULE(if_igb);
 243 MODULE_DEPEND(igb, ig_hal, 1, 1, 1);
 244 DRIVER_MODULE(if_igb, pci, igb_driver, igb_devclass, NULL, NULL);
 245
 246 static int      igb_rxd = IGB_DEFAULT_RXD;
 247 static int      igb_txd = IGB_DEFAULT_TXD;
 248 static int      igb_rxr = 0;
 249 static int      igb_msi_enable = 1;
 250 static int      igb_msix_enable = 1;
 251 static int      igb_eee_disabled = 1;   /* Energy Efficient Ethernet */
 252 static int      igb_fc_setting = e1000_fc_full;
 253
 254 /*
 255  * DMA Coalescing, only for i350 - default to off,
 256  * this feature is for power savings
 257  */
 258 static int      igb_dma_coalesce = 0;
 259
 260 TUNABLE_INT("hw.igb.rxd", &igb_rxd);
 261 TUNABLE_INT("hw.igb.txd", &igb_txd);
 262 TUNABLE_INT("hw.igb.rxr", &igb_rxr);
 263 TUNABLE_INT("hw.igb.msi.enable", &igb_msi_enable);
 264 TUNABLE_INT("hw.igb.msix.enable", &igb_msix_enable);
 265 TUNABLE_INT("hw.igb.fc_setting", &igb_fc_setting);
 266
 267 /* i350 specific */
 268 TUNABLE_INT("hw.igb.eee_disabled", &igb_eee_disabled);
 269 TUNABLE_INT("hw.igb.dma_coalesce", &igb_dma_coalesce);
 270
 271 static __inline void
 272 igb_rxcsum(uint32_t staterr, struct mbuf *mp)
 273 {
 274         /* Ignore Checksum bit is set */
 275         if (staterr & E1000_RXD_STAT_IXSM)
 276                 return;
 277
 278         if ((staterr & (E1000_RXD_STAT_IPCS | E1000_RXDEXT_STATERR_IPE)) ==
 279             E1000_RXD_STAT_IPCS)
 280                 mp->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
 281
 282         if (staterr & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)) {
 283                 if ((staterr & E1000_RXDEXT_STATERR_TCPE) == 0) {
 284                         mp->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
 285                             CSUM_PSEUDO_HDR | CSUM_FRAG_NOT_CHECKED;
 286                         mp->m_pkthdr.csum_data = htons(0xffff);
 287                 }
 288         }
 289 }
 290
 291 static __inline struct pktinfo *
 292 igb_rssinfo(struct mbuf *m, struct pktinfo *pi,
 293     uint32_t hash, uint32_t hashtype, uint32_t staterr)
 294 {
 295         switch (hashtype) {
 296         case E1000_RXDADV_RSSTYPE_IPV4_TCP:
 297                 pi->pi_netisr = NETISR_IP;
 298                 pi->pi_flags = 0;
 299                 pi->pi_l3proto = IPPROTO_TCP;
 300                 break;
 301
 302         case E1000_RXDADV_RSSTYPE_IPV4:
 303                 if (staterr & E1000_RXD_STAT_IXSM)
 304                         return NULL;
 305
 306                 if ((staterr &
 307                      (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) ==
 308                     E1000_RXD_STAT_TCPCS) {
 309                         pi->pi_netisr = NETISR_IP;
 310                         pi->pi_flags = 0;
 311                         pi->pi_l3proto = IPPROTO_UDP;
 312                         break;
 313                 }
 314                 /* FALL THROUGH */
 315         default:
 316                 return NULL;
 317         }
 318
 319         m->m_flags |= M_HASH;
 320         m->m_pkthdr.hash = toeplitz_hash(hash);
 321         return pi;
 322 }
 323
 324 static int
 325 igb_probe(device_t dev)
 326 {
 327         const struct igb_device *d;
 328         uint16_t vid, did;
 329
 330         vid = pci_get_vendor(dev);
 331         did = pci_get_device(dev);
 332
 333         for (d = igb_devices; d->desc != NULL; ++d) {
 334                 if (vid == d->vid && did == d->did) {
 335                         device_set_desc(dev, d->desc);
 336                         return 0;
 337                 }
 338         }
 339         return ENXIO;
 340 }
 341
 342 static int
 343 igb_attach(device_t dev)
 344 {
 345         struct igb_softc *sc = device_get_softc(dev);
 346         uint16_t eeprom_data;
 347         int error = 0, i, j, ring_max;
 348
 349 #ifdef notyet
 350         /* SYSCTL stuff */
 351         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 352             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 353             OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
 354             igb_sysctl_nvm_info, "I", "NVM Information");
 355
 356         SYSCTL_ADD_INT(device_get_sysctl_ctx(dev),
 357             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 358             OID_AUTO, "enable_aim", CTLTYPE_INT|CTLFLAG_RW,
 359             &igb_enable_aim, 1, "Interrupt Moderation");
 360
 361         SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev),
 362             SYSCTL_CHILDREN(device_get_sysctl_tree(dev)),
 363             OID_AUTO, "flow_control", CTLTYPE_INT|CTLFLAG_RW,
 364             adapter, 0, igb_set_flowcntl, "I", "Flow Control");
 365 #endif
 366
 367         callout_init_mp(&sc->timer);
 368         lwkt_serialize_init(&sc->main_serialize);
 369
 370         sc->dev = sc->osdep.dev = dev;
 371
 372         /*
 373          * Determine hardware and mac type
 374          */
 375         sc->hw.vendor_id = pci_get_vendor(dev);
 376         sc->hw.device_id = pci_get_device(dev);
 377         sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1);
 378         sc->hw.subsystem_vendor_id = pci_read_config(dev, PCIR_SUBVEND_0, 2);
 379         sc->hw.subsystem_device_id = pci_read_config(dev, PCIR_SUBDEV_0, 2);
 380
 381         if (e1000_set_mac_type(&sc->hw))
 382                 return ENXIO;
 383
 384         /* Are we a VF device? */
 385         if (sc->hw.mac.type == e1000_vfadapt ||
 386             sc->hw.mac.type == e1000_vfadapt_i350)
 387                 sc->vf_ifp = 1;
 388         else
 389                 sc->vf_ifp = 0;
 390
 391         /*
 392          * Configure total supported RX/TX ring count
 393          */
 394         switch (sc->hw.mac.type) {
 395         case e1000_82575:
 396                 ring_max = IGB_MAX_RING_82575;
 397                 break;
 398         case e1000_82580:
 399                 ring_max = IGB_MAX_RING_82580;
 400                 break;
 401         case e1000_i350:
 402                 ring_max = IGB_MAX_RING_I350;
 403                 break;
 404         case e1000_82576:
 405                 ring_max = IGB_MAX_RING_82576;
 406                 break;
 407         default:
 408                 ring_max = IGB_MIN_RING;
 409                 break;
 410         }
 411         sc->rx_ring_cnt = device_getenv_int(dev, "rxr", igb_rxr);
 412         sc->rx_ring_cnt = if_ring_count2(sc->rx_ring_cnt, ring_max);
 413 #ifdef IGB_RSS_DEBUG
 414         sc->rx_ring_cnt = device_getenv_int(dev, "rxr_debug", sc->rx_ring_cnt);
 415 #endif
 416         sc->rx_ring_inuse = sc->rx_ring_cnt;
 417         sc->tx_ring_cnt = 1; /* XXX */
 418
 419         if (sc->hw.mac.type == e1000_82575)
 420                 sc->flags |= IGB_FLAG_TSO_IPLEN0;
 421
 422         /* Enable bus mastering */
 423         pci_enable_busmaster(dev);
 424
 425         /*
 426          * Allocate IO memory
 427          */
 428         sc->mem_rid = PCIR_BAR(0);
 429         sc->mem_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->mem_rid,
 430             RF_ACTIVE);
 431         if (sc->mem_res == NULL) {
 432                 device_printf(dev, "Unable to allocate bus resource: memory\n");
 433                 error = ENXIO;
 434                 goto failed;
 435         }
 436         sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->mem_res);
 437         sc->osdep.mem_bus_space_handle = rman_get_bushandle(sc->mem_res);
 438
 439         sc->hw.hw_addr = (uint8_t *)&sc->osdep.mem_bus_space_handle;
 440
 441         /* Save PCI command register for Shared Code */
 442         sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
 443         sc->hw.back = &sc->osdep;
 444
 445         /* Do Shared Code initialization */
 446         if (e1000_setup_init_funcs(&sc->hw, TRUE)) {
 447                 device_printf(dev, "Setup of Shared code failed\n");
 448                 error = ENXIO;
 449                 goto failed;
 450         }
 451
 452         e1000_get_bus_info(&sc->hw);
 453
 454         sc->hw.mac.autoneg = DO_AUTO_NEG;
 455         sc->hw.phy.autoneg_wait_to_complete = FALSE;
 456         sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
 457
 458         /* Copper options */
 459         if (sc->hw.phy.media_type == e1000_media_type_copper) {
 460                 sc->hw.phy.mdix = AUTO_ALL_MODES;
 461                 sc->hw.phy.disable_polarity_correction = FALSE;
 462                 sc->hw.phy.ms_type = IGB_MASTER_SLAVE;
 463         }
 464
 465         /* Set the frame limits assuming  standard ethernet sized frames. */
 466         sc->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN;
 467
 468         /* Allocate RX/TX rings */
 469         error = igb_alloc_rings(sc);
 470         if (error)
 471                 goto failed;
 472
 473         /* Allocate interrupt */
 474         error = igb_alloc_intr(sc);
 475         if (error)
 476                 goto failed;
 477
 478         /*
 479          * Setup serializers
 480          */
 481         i = 0;
 482         sc->serializes[i++] = &sc->main_serialize;
 483
 484         sc->tx_serialize = i;
 485         for (j = 0; j < sc->tx_ring_cnt; ++j)
 486                 sc->serializes[i++] = &sc->tx_rings[j].tx_serialize;
 487
 488         sc->rx_serialize = i;
 489         for (j = 0; j < sc->rx_ring_cnt; ++j)
 490                 sc->serializes[i++] = &sc->rx_rings[j].rx_serialize;
 491
 492         sc->serialize_cnt = i;
 493         KKASSERT(sc->serialize_cnt <= IGB_NSERIALIZE);
 494
 495         /* Allocate the appropriate stats memory */
 496         if (sc->vf_ifp) {
 497                 sc->stats = kmalloc(sizeof(struct e1000_vf_stats), M_DEVBUF,
 498                     M_WAITOK | M_ZERO);
 499                 igb_vf_init_stats(sc);
 500         } else {
 501                 sc->stats = kmalloc(sizeof(struct e1000_hw_stats), M_DEVBUF,
 502                     M_WAITOK | M_ZERO);
 503         }
 504
 505         /* Allocate multicast array memory. */
 506         sc->mta = kmalloc(ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES,
 507             M_DEVBUF, M_WAITOK);
 508
 509         /* Some adapter-specific advanced features */
 510         if (sc->hw.mac.type >= e1000_i350) {
 511 #ifdef notyet
 512                 igb_set_sysctl_value(adapter, "dma_coalesce",
 513                     "configure dma coalesce",
 514                     &adapter->dma_coalesce, igb_dma_coalesce);
 515                 igb_set_sysctl_value(adapter, "eee_disabled",
 516                     "enable Energy Efficient Ethernet",
 517                     &adapter->hw.dev_spec._82575.eee_disable,
 518                     igb_eee_disabled);
 519 #else
 520                 sc->dma_coalesce = igb_dma_coalesce;
 521                 sc->hw.dev_spec._82575.eee_disable = igb_eee_disabled;
 522 #endif
 523                 e1000_set_eee_i350(&sc->hw);
 524         }
 525
 526         /*
 527          * Start from a known state, this is important in reading the nvm and
 528          * mac from that.
 529          */
 530         e1000_reset_hw(&sc->hw);
 531
 532         /* Make sure we have a good EEPROM before we read from it */
 533         if (e1000_validate_nvm_checksum(&sc->hw) < 0) {
 534                 /*
 535                  * Some PCI-E parts fail the first check due to
 536                  * the link being in sleep state, call it again,
 537                  * if it fails a second time its a real issue.
 538                  */
 539                 if (e1000_validate_nvm_checksum(&sc->hw) < 0) {
 540                         device_printf(dev,
 541                             "The EEPROM Checksum Is Not Valid\n");
 542                         error = EIO;
 543                         goto failed;
 544                 }
 545         }
 546
 547         /* Copy the permanent MAC address out of the EEPROM */
 548         if (e1000_read_mac_addr(&sc->hw) < 0) {
 549                 device_printf(dev, "EEPROM read error while reading MAC"
 550                     " address\n");
 551                 error = EIO;
 552                 goto failed;
 553         }
 554         if (!igb_is_valid_ether_addr(sc->hw.mac.addr)) {
 555                 device_printf(dev, "Invalid MAC address\n");
 556                 error = EIO;
 557                 goto failed;
 558         }
 559
 560 #ifdef notyet
 561         /*
 562         ** Configure Interrupts
 563         */
 564         if ((adapter->msix > 1) && (igb_enable_msix))
 565                 error = igb_allocate_msix(adapter);
 566         else /* MSI or Legacy */
 567                 error = igb_allocate_legacy(adapter);
 568         if (error)
 569                 goto err_late;
 570 #endif
 571
 572         /* Setup OS specific network interface */
 573         igb_setup_ifp(sc);
 574
 575         /* Add sysctl tree, must after igb_setup_ifp() */
 576         igb_add_sysctl(sc);
 577
 578         /* Now get a good starting state */
 579         igb_reset(sc);
 580
 581         /* Initialize statistics */
 582         igb_update_stats_counters(sc);
 583
 584         sc->hw.mac.get_link_status = 1;
 585         igb_update_link_status(sc);
 586
 587         /* Indicate SOL/IDER usage */
 588         if (e1000_check_reset_block(&sc->hw)) {
 589                 device_printf(dev,
 590                     "PHY reset is blocked due to SOL/IDER session.\n");
 591         }
 592
 593         /* Determine if we have to control management hardware */
 594         if (e1000_enable_mng_pass_thru(&sc->hw))
 595                 sc->flags |= IGB_FLAG_HAS_MGMT;
 596
 597         /*
 598          * Setup Wake-on-Lan
 599          */
 600         /* APME bit in EEPROM is mapped to WUC.APME */
 601         eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC) & E1000_WUC_APME;
 602         if (eeprom_data)
 603                 sc->wol = E1000_WUFC_MAG;
 604         /* XXX disable WOL */
 605         sc->wol = 0;
 606
 607 #ifdef notyet
 608         /* Register for VLAN events */
 609         adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config,
 610              igb_register_vlan, adapter, EVENTHANDLER_PRI_FIRST);
 611         adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig,
 612              igb_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST);
 613 #endif
 614
 615 #ifdef notyet
 616         igb_add_hw_stats(adapter);
 617 #endif
 618
 619         error = igb_setup_intr(sc);
 620         if (error) {
 621                 ether_ifdetach(&sc->arpcom.ac_if);
 622                 goto failed;
 623         }
 624         return 0;
 625
 626 failed:
 627         igb_detach(dev);
 628         return error;
 629 }
 630
 631 static int
 632 igb_detach(device_t dev)
 633 {
 634         struct igb_softc *sc = device_get_softc(dev);
 635
 636         if (device_is_attached(dev)) {
 637                 struct ifnet *ifp = &sc->arpcom.ac_if;
 638
 639                 ifnet_serialize_all(ifp);
 640
 641                 igb_stop(sc);
 642
 643                 e1000_phy_hw_reset(&sc->hw);
 644
 645                 /* Give control back to firmware */
 646                 igb_rel_mgmt(sc);
 647                 igb_rel_hw_control(sc);
 648
 649                 if (sc->wol) {
 650                         E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
 651                         E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
 652                         igb_enable_wol(dev);
 653                 }
 654
 655                 igb_teardown_intr(sc);
 656
 657                 ifnet_deserialize_all(ifp);
 658
 659                 ether_ifdetach(ifp);
 660         } else if (sc->mem_res != NULL) {
 661                 igb_rel_hw_control(sc);
 662         }
 663         bus_generic_detach(dev);
 664
 665         if (sc->sysctl_tree != NULL)
 666                 sysctl_ctx_free(&sc->sysctl_ctx);
 667
 668         igb_free_intr(sc);
 669
 670         if (sc->msix_mem_res != NULL) {
 671                 bus_release_resource(dev, SYS_RES_MEMORY, sc->msix_mem_rid,
 672                     sc->msix_mem_res);
 673         }
 674         if (sc->mem_res != NULL) {
 675                 bus_release_resource(dev, SYS_RES_MEMORY, sc->mem_rid,
 676                     sc->mem_res);
 677         }
 678
 679         igb_free_rings(sc);
 680
 681         if (sc->mta != NULL)
 682                 kfree(sc->mta, M_DEVBUF);
 683         if (sc->stats != NULL)
 684                 kfree(sc->stats, M_DEVBUF);
 685
 686         return 0;
 687 }
 688
 689 static int
 690 igb_shutdown(device_t dev)
 691 {
 692         return igb_suspend(dev);
 693 }
 694
 695 static int
 696 igb_suspend(device_t dev)
 697 {
 698         struct igb_softc *sc = device_get_softc(dev);
 699         struct ifnet *ifp = &sc->arpcom.ac_if;
 700
 701         ifnet_serialize_all(ifp);
 702
 703         igb_stop(sc);
 704
 705         igb_rel_mgmt(sc);
 706         igb_rel_hw_control(sc);
 707
 708         if (sc->wol) {
 709                 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN);
 710                 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol);
 711                 igb_enable_wol(dev);
 712         }
 713
 714         ifnet_deserialize_all(ifp);
 715
 716         return bus_generic_suspend(dev);
 717 }
 718
 719 static int
 720 igb_resume(device_t dev)
 721 {
 722         struct igb_softc *sc = device_get_softc(dev);
 723         struct ifnet *ifp = &sc->arpcom.ac_if;
 724
 725         ifnet_serialize_all(ifp);
 726
 727         igb_init(sc);
 728         igb_get_mgmt(sc);
 729
 730         if_devstart(ifp);
 731
 732         ifnet_deserialize_all(ifp);
 733
 734         return bus_generic_resume(dev);
 735 }
 736
 737 static int
 738 igb_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr)
 739 {
 740         struct igb_softc *sc = ifp->if_softc;
 741         struct ifreq *ifr = (struct ifreq *)data;
 742         int max_frame_size, mask, reinit;
 743         int error = 0;
 744
 745         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 746
 747         switch (command) {
 748         case SIOCSIFMTU:
 749                 max_frame_size = 9234;
 750                 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
 751                     ETHER_CRC_LEN) {
 752                         error = EINVAL;
 753                         break;
 754                 }
 755
 756                 ifp->if_mtu = ifr->ifr_mtu;
 757                 sc->max_frame_size = ifp->if_mtu + ETHER_HDR_LEN +
 758                     ETHER_CRC_LEN;
 759
 760                 if (ifp->if_flags & IFF_RUNNING)
 761                         igb_init(sc);
 762                 break;
 763
 764         case SIOCSIFFLAGS:
 765                 if (ifp->if_flags & IFF_UP) {
 766                         if (ifp->if_flags & IFF_RUNNING) {
 767                                 if ((ifp->if_flags ^ sc->if_flags) &
 768                                     (IFF_PROMISC | IFF_ALLMULTI)) {
 769                                         igb_disable_promisc(sc);
 770                                         igb_set_promisc(sc);
 771                                 }
 772                         } else {
 773                                 igb_init(sc);
 774                         }
 775                 } else if (ifp->if_flags & IFF_RUNNING) {
 776                         igb_stop(sc);
 777                 }
 778                 sc->if_flags = ifp->if_flags;
 779                 break;
 780
 781         case SIOCADDMULTI:
 782         case SIOCDELMULTI:
 783                 if (ifp->if_flags & IFF_RUNNING) {
 784                         igb_disable_intr(sc);
 785                         igb_set_multi(sc);
 786 #ifdef DEVICE_POLLING
 787                         if (!(ifp->if_flags & IFF_POLLING))
 788 #endif
 789                                 igb_enable_intr(sc);
 790                 }
 791                 break;
 792
 793         case SIOCSIFMEDIA:
 794                 /*
 795                  * As the speed/duplex settings are being
 796                  * changed, we need toreset the PHY.
 797                  */
 798                 sc->hw.phy.reset_disable = FALSE;
 799
 800                 /* Check SOL/IDER usage */
 801                 if (e1000_check_reset_block(&sc->hw)) {
 802                         if_printf(ifp, "Media change is "
 803                             "blocked due to SOL/IDER session.\n");
 804                         break;
 805                 }
 806                 /* FALL THROUGH */
 807
 808         case SIOCGIFMEDIA:
 809                 error = ifmedia_ioctl(ifp, ifr, &sc->media, command);
 810                 break;
 811
 812         case SIOCSIFCAP:
 813                 reinit = 0;
 814                 mask = ifr->ifr_reqcap ^ ifp->if_capenable;
 815                 if (mask & IFCAP_RXCSUM) {
 816                         ifp->if_capenable ^= IFCAP_RXCSUM;
 817                         reinit = 1;
 818                 }
 819                 if (mask & IFCAP_VLAN_HWTAGGING) {
 820                         ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
 821                         reinit = 1;
 822                 }
 823                 if (mask & IFCAP_TXCSUM) {
 824                         ifp->if_capenable ^= IFCAP_TXCSUM;
 825                         if (ifp->if_capenable & IFCAP_TXCSUM)
 826                                 ifp->if_hwassist |= IGB_CSUM_FEATURES;
 827                         else
 828                                 ifp->if_hwassist &= ~IGB_CSUM_FEATURES;
 829                 }
 830                 if (mask & IFCAP_TSO) {
 831                         ifp->if_capenable ^= IFCAP_TSO;
 832                         if (ifp->if_capenable & IFCAP_TSO)
 833                                 ifp->if_hwassist |= CSUM_TSO;
 834                         else
 835                                 ifp->if_hwassist &= ~CSUM_TSO;
 836                 }
 837                 if (mask & IFCAP_RSS)
 838                         ifp->if_capenable ^= IFCAP_RSS;
 839                 if (reinit && (ifp->if_flags & IFF_RUNNING))
 840                         igb_init(sc);
 841                 break;
 842
 843         default:
 844                 error = ether_ioctl(ifp, command, data);
 845                 break;
 846         }
 847         return error;
 848 }
 849
 850 static void
 851 igb_init(void *xsc)
 852 {
 853         struct igb_softc *sc = xsc;
 854         struct ifnet *ifp = &sc->arpcom.ac_if;
 855         boolean_t polling;
 856         int i;
 857
 858         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 859
 860         igb_stop(sc);
 861
 862         /* Get the latest mac address, User can use a LAA */
 863         bcopy(IF_LLADDR(ifp), sc->hw.mac.addr, ETHER_ADDR_LEN);
 864
 865         /* Put the address into the Receive Address Array */
 866         e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0);
 867
 868         igb_reset(sc);
 869         igb_update_link_status(sc);
 870
 871         E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
 872
 873         /* Configure for OS presence */
 874         igb_get_mgmt(sc);
 875
 876         polling = FALSE;
 877 #ifdef DEVICE_POLLING
 878         if (ifp->if_flags & IFF_POLLING)
 879                 polling = TRUE;
 880 #endif
 881
 882         /* Configured used RX/TX rings */
 883         igb_set_ring_inuse(sc, polling);
 884
 885         /* Initialize interrupt */
 886         igb_init_intr(sc);
 887
 888         /* Prepare transmit descriptors and buffers */
 889         for (i = 0; i < sc->tx_ring_cnt; ++i)
 890                 igb_init_tx_ring(&sc->tx_rings[i]);
 891         igb_init_tx_unit(sc);
 892
 893         /* Setup Multicast table */
 894         igb_set_multi(sc);
 895
 896 #if 0
 897         /*
 898          * Figure out the desired mbuf pool
 899          * for doing jumbo/packetsplit
 900          */
 901         if (adapter->max_frame_size <= 2048)
 902                 adapter->rx_mbuf_sz = MCLBYTES;
 903         else if (adapter->max_frame_size <= 4096)
 904                 adapter->rx_mbuf_sz = MJUMPAGESIZE;
 905         else
 906                 adapter->rx_mbuf_sz = MJUM9BYTES;
 907 #endif
 908
 909         /* Prepare receive descriptors and buffers */
 910         for (i = 0; i < sc->rx_ring_inuse; ++i) {
 911                 int error;
 912
 913                 error = igb_init_rx_ring(&sc->rx_rings[i]);
 914                 if (error) {
 915                         if_printf(ifp, "Could not setup receive structures\n");
 916                         igb_stop(sc);
 917                         return;
 918                 }
 919         }
 920         igb_init_rx_unit(sc);
 921
 922         /* Enable VLAN support */
 923         if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING)
 924                 igb_set_vlan(sc);
 925
 926         /* Don't lose promiscuous settings */
 927         igb_set_promisc(sc);
 928
 929         ifp->if_flags |= IFF_RUNNING;
 930         ifp->if_flags &= ~IFF_OACTIVE;
 931
 932         callout_reset(&sc->timer, hz, igb_timer, sc);
 933         e1000_clear_hw_cntrs_base_generic(&sc->hw);
 934
 935 #if 0
 936         if (adapter->msix > 1) /* Set up queue routing */
 937                 igb_configure_queues(adapter);
 938 #endif
 939
 940         /* This clears any pending interrupts */
 941         E1000_READ_REG(&sc->hw, E1000_ICR);
 942
 943         /*
 944          * Only enable interrupts if we are not polling, make sure
 945          * they are off otherwise.
 946          */
 947         if (polling) {
 948                 igb_disable_intr(sc);
 949         } else {
 950                 igb_enable_intr(sc);
 951                 E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC);
 952         }
 953
 954         /* Set Energy Efficient Ethernet */
 955         e1000_set_eee_i350(&sc->hw);
 956
 957         /* Don't reset the phy next time init gets called */
 958         sc->hw.phy.reset_disable = TRUE;
 959 }
 960
 961 static void
 962 igb_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
 963 {
 964         struct igb_softc *sc = ifp->if_softc;
 965         u_char fiber_type = IFM_1000_SX;
 966
 967         ASSERT_IFNET_SERIALIZED_ALL(ifp);
 968
 969         igb_update_link_status(sc);
 970
 971         ifmr->ifm_status = IFM_AVALID;
 972         ifmr->ifm_active = IFM_ETHER;
 973
 974         if (!sc->link_active)
 975                 return;
 976
 977         ifmr->ifm_status |= IFM_ACTIVE;
 978
 979         if (sc->hw.phy.media_type == e1000_media_type_fiber ||
 980             sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
 981                 ifmr->ifm_active |= fiber_type | IFM_FDX;
 982         } else {
 983                 switch (sc->link_speed) {
 984                 case 10:
 985                         ifmr->ifm_active |= IFM_10_T;
 986                         break;
 987
 988                 case 100:
 989                         ifmr->ifm_active |= IFM_100_TX;
 990                         break;
 991
 992                 case 1000:
 993                         ifmr->ifm_active |= IFM_1000_T;
 994                         break;
 995                 }
 996                 if (sc->link_duplex == FULL_DUPLEX)
 997                         ifmr->ifm_active |= IFM_FDX;
 998                 else
 999                         ifmr->ifm_active |= IFM_HDX;
1000         }
1001 }
1002
1003 static int
1004 igb_media_change(struct ifnet *ifp)
1005 {
1006         struct igb_softc *sc = ifp->if_softc;
1007         struct ifmedia *ifm = &sc->media;
1008
1009         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1010
1011         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1012                 return EINVAL;
1013
1014         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1015         case IFM_AUTO:
1016                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1017                 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1018                 break;
1019
1020         case IFM_1000_LX:
1021         case IFM_1000_SX:
1022         case IFM_1000_T:
1023                 sc->hw.mac.autoneg = DO_AUTO_NEG;
1024                 sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1025                 break;
1026
1027         case IFM_100_TX:
1028                 sc->hw.mac.autoneg = FALSE;
1029                 sc->hw.phy.autoneg_advertised = 0;
1030                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1031                         sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1032                 else
1033                         sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1034                 break;
1035
1036         case IFM_10_T:
1037                 sc->hw.mac.autoneg = FALSE;
1038                 sc->hw.phy.autoneg_advertised = 0;
1039                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1040                         sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1041                 else
1042                         sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1043                 break;
1044
1045         default:
1046                 if_printf(ifp, "Unsupported media type\n");
1047                 break;
1048         }
1049
1050         igb_init(sc);
1051
1052         return 0;
1053 }
1054
1055 static void
1056 igb_set_promisc(struct igb_softc *sc)
1057 {
1058         struct ifnet *ifp = &sc->arpcom.ac_if;
1059         struct e1000_hw *hw = &sc->hw;
1060         uint32_t reg;
1061
1062         if (sc->vf_ifp) {
1063                 e1000_promisc_set_vf(hw, e1000_promisc_enabled);
1064                 return;
1065         }
1066
1067         reg = E1000_READ_REG(hw, E1000_RCTL);
1068         if (ifp->if_flags & IFF_PROMISC) {
1069                 reg |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1070                 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1071         } else if (ifp->if_flags & IFF_ALLMULTI) {
1072                 reg |= E1000_RCTL_MPE;
1073                 reg &= ~E1000_RCTL_UPE;
1074                 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1075         }
1076 }
1077
1078 static void
1079 igb_disable_promisc(struct igb_softc *sc)
1080 {
1081         struct e1000_hw *hw = &sc->hw;
1082         uint32_t reg;
1083
1084         if (sc->vf_ifp) {
1085                 e1000_promisc_set_vf(hw, e1000_promisc_disabled);
1086                 return;
1087         }
1088         reg = E1000_READ_REG(hw, E1000_RCTL);
1089         reg &= ~E1000_RCTL_UPE;
1090         reg &= ~E1000_RCTL_MPE;
1091         E1000_WRITE_REG(hw, E1000_RCTL, reg);
1092 }
1093
1094 static void
1095 igb_set_multi(struct igb_softc *sc)
1096 {
1097         struct ifnet *ifp = &sc->arpcom.ac_if;
1098         struct ifmultiaddr *ifma;
1099         uint32_t reg_rctl = 0;
1100         uint8_t *mta;
1101         int mcnt = 0;
1102
1103         mta = sc->mta;
1104         bzero(mta, ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES);
1105
1106         TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1107                 if (ifma->ifma_addr->sa_family != AF_LINK)
1108                         continue;
1109
1110                 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
1111                         break;
1112
1113                 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
1114                     &mta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN);
1115                 mcnt++;
1116         }
1117
1118         if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1119                 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL);
1120                 reg_rctl |= E1000_RCTL_MPE;
1121                 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl);
1122         } else {
1123                 e1000_update_mc_addr_list(&sc->hw, mta, mcnt);
1124         }
1125 }
1126
1127 static void
1128 igb_timer(void *xsc)
1129 {
1130         struct igb_softc *sc = xsc;
1131
1132         lwkt_serialize_enter(&sc->main_serialize);
1133
1134         igb_update_link_status(sc);
1135         igb_update_stats_counters(sc);
1136
1137         callout_reset(&sc->timer, hz, igb_timer, sc);
1138
1139         lwkt_serialize_exit(&sc->main_serialize);
1140 }
1141
1142 static void
1143 igb_update_link_status(struct igb_softc *sc)
1144 {
1145         struct ifnet *ifp = &sc->arpcom.ac_if;
1146         struct e1000_hw *hw = &sc->hw;
1147         uint32_t link_check, thstat, ctrl;
1148
1149         link_check = thstat = ctrl = 0;
1150
1151         /* Get the cached link value or read for real */
1152         switch (hw->phy.media_type) {
1153         case e1000_media_type_copper:
1154                 if (hw->mac.get_link_status) {
1155                         /* Do the work to read phy */
1156                         e1000_check_for_link(hw);
1157                         link_check = !hw->mac.get_link_status;
1158                 } else {
1159                         link_check = TRUE;
1160                 }
1161                 break;
1162
1163         case e1000_media_type_fiber:
1164                 e1000_check_for_link(hw);
1165                 link_check = E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU;
1166                 break;
1167
1168         case e1000_media_type_internal_serdes:
1169                 e1000_check_for_link(hw);
1170                 link_check = hw->mac.serdes_has_link;
1171                 break;
1172
1173         /* VF device is type_unknown */
1174         case e1000_media_type_unknown:
1175                 e1000_check_for_link(hw);
1176                 link_check = !hw->mac.get_link_status;
1177                 /* Fall thru */
1178         default:
1179                 break;
1180         }
1181
1182         /* Check for thermal downshift or shutdown */
1183         if (hw->mac.type == e1000_i350) {
1184                 thstat = E1000_READ_REG(hw, E1000_THSTAT);
1185                 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT);
1186         }
1187
1188         /* Now we check if a transition has happened */
1189         if (link_check && sc->link_active == 0) {
1190                 e1000_get_speed_and_duplex(hw,
1191                     &sc->link_speed, &sc->link_duplex);
1192                 if (bootverbose) {
1193                         if_printf(ifp, "Link is up %d Mbps %s\n",
1194                             sc->link_speed,
1195                             sc->link_duplex == FULL_DUPLEX ?
1196                             "Full Duplex" : "Half Duplex");
1197                 }
1198                 sc->link_active = 1;
1199
1200                 ifp->if_baudrate = sc->link_speed * 1000000;
1201                 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1202                     (thstat & E1000_THSTAT_LINK_THROTTLE))
1203                         if_printf(ifp, "Link: thermal downshift\n");
1204                 /* This can sleep */
1205                 ifp->if_link_state = LINK_STATE_UP;
1206                 if_link_state_change(ifp);
1207         } else if (!link_check && sc->link_active == 1) {
1208                 ifp->if_baudrate = sc->link_speed = 0;
1209                 sc->link_duplex = 0;
1210                 if (bootverbose)
1211                         if_printf(ifp, "Link is Down\n");
1212                 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) &&
1213                     (thstat & E1000_THSTAT_PWR_DOWN))
1214                         if_printf(ifp, "Link: thermal shutdown\n");
1215                 sc->link_active = 0;
1216                 /* This can sleep */
1217                 ifp->if_link_state = LINK_STATE_DOWN;
1218                 if_link_state_change(ifp);
1219         }
1220 }
1221
1222 static void
1223 igb_stop(struct igb_softc *sc)
1224 {
1225         struct ifnet *ifp = &sc->arpcom.ac_if;
1226         int i;
1227
1228         ASSERT_IFNET_SERIALIZED_ALL(ifp);
1229
1230         igb_disable_intr(sc);
1231
1232         callout_stop(&sc->timer);
1233
1234         ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
1235         ifp->if_timer = 0;
1236
1237         e1000_reset_hw(&sc->hw);
1238         E1000_WRITE_REG(&sc->hw, E1000_WUC, 0);
1239
1240         e1000_led_off(&sc->hw);
1241         e1000_cleanup_led(&sc->hw);
1242
1243         for (i = 0; i < sc->tx_ring_cnt; ++i)
1244                 igb_free_tx_ring(&sc->tx_rings[i]);
1245         for (i = 0; i < sc->rx_ring_cnt; ++i)
1246                 igb_free_rx_ring(&sc->rx_rings[i]);
1247 }
1248
1249 static void
1250 igb_reset(struct igb_softc *sc)
1251 {
1252         struct ifnet *ifp = &sc->arpcom.ac_if;
1253         struct e1000_hw *hw = &sc->hw;
1254         struct e1000_fc_info *fc = &hw->fc;
1255         uint32_t pba = 0;
1256         uint16_t hwm;
1257
1258         /* Let the firmware know the OS is in control */
1259         igb_get_hw_control(sc);
1260
1261         /*
1262          * Packet Buffer Allocation (PBA)
1263          * Writing PBA sets the receive portion of the buffer
1264          * the remainder is used for the transmit buffer.
1265          */
1266         switch (hw->mac.type) {
1267         case e1000_82575:
1268                 pba = E1000_PBA_32K;
1269                 break;
1270
1271         case e1000_82576:
1272         case e1000_vfadapt:
1273                 pba = E1000_READ_REG(hw, E1000_RXPBS);
1274                 pba &= E1000_RXPBS_SIZE_MASK_82576;
1275                 break;
1276
1277         case e1000_82580:
1278         case e1000_i350:
1279         case e1000_vfadapt_i350:
1280                 pba = E1000_READ_REG(hw, E1000_RXPBS);
1281                 pba = e1000_rxpbs_adjust_82580(pba);
1282                 break;
1283                 /* XXX pba = E1000_PBA_35K; */
1284
1285         default:
1286                 break;
1287         }
1288
1289         /* Special needs in case of Jumbo frames */
1290         if (hw->mac.type == e1000_82575 && ifp->if_mtu > ETHERMTU) {
1291                 uint32_t tx_space, min_tx, min_rx;
1292
1293                 pba = E1000_READ_REG(hw, E1000_PBA);
1294                 tx_space = pba >> 16;
1295                 pba &= 0xffff;
1296
1297                 min_tx = (sc->max_frame_size +
1298                     sizeof(struct e1000_tx_desc) - ETHER_CRC_LEN) * 2;
1299                 min_tx = roundup2(min_tx, 1024);
1300                 min_tx >>= 10;
1301                 min_rx = sc->max_frame_size;
1302                 min_rx = roundup2(min_rx, 1024);
1303                 min_rx >>= 10;
1304                 if (tx_space < min_tx && (min_tx - tx_space) < pba) {
1305                         pba = pba - (min_tx - tx_space);
1306                         /*
1307                          * if short on rx space, rx wins
1308                          * and must trump tx adjustment
1309                          */
1310                         if (pba < min_rx)
1311                                 pba = min_rx;
1312                 }
1313                 E1000_WRITE_REG(hw, E1000_PBA, pba);
1314         }
1315
1316         /*
1317          * These parameters control the automatic generation (Tx) and
1318          * response (Rx) to Ethernet PAUSE frames.
1319          * - High water mark should allow for at least two frames to be
1320          *   received after sending an XOFF.
1321          * - Low water mark works best when it is very near the high water mark.
1322          *   This allows the receiver to restart by sending XON when it has
1323          *   drained a bit.
1324          */
1325         hwm = min(((pba << 10) * 9 / 10),
1326             ((pba << 10) - 2 * sc->max_frame_size));
1327
1328         if (hw->mac.type < e1000_82576) {
1329                 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */
1330                 fc->low_water = fc->high_water - 8;
1331         } else {
1332                 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */
1333                 fc->low_water = fc->high_water - 16;
1334         }
1335         fc->pause_time = IGB_FC_PAUSE_TIME;
1336         fc->send_xon = TRUE;
1337
1338         /* Issue a global reset */
1339         e1000_reset_hw(hw);
1340         E1000_WRITE_REG(hw, E1000_WUC, 0);
1341
1342         if (e1000_init_hw(hw) < 0)
1343                 if_printf(ifp, "Hardware Initialization Failed\n");
1344
1345         /* Setup DMA Coalescing */
1346         if (hw->mac.type == e1000_i350 && sc->dma_coalesce) {
1347                 uint32_t reg;
1348
1349                 hwm = (pba - 4) << 10;
1350                 reg = ((pba - 6) << E1000_DMACR_DMACTHR_SHIFT)
1351                     & E1000_DMACR_DMACTHR_MASK;
1352
1353                 /* transition to L0x or L1 if available..*/
1354                 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
1355
1356                 /* timer = +-1000 usec in 32usec intervals */
1357                 reg |= (1000 >> 5);
1358                 E1000_WRITE_REG(hw, E1000_DMACR, reg);
1359
1360                 /* No lower threshold */
1361                 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0);
1362
1363                 /* set hwm to PBA -  2 * max frame size */
1364                 E1000_WRITE_REG(hw, E1000_FCRTC, hwm);
1365
1366                 /* Set the interval before transition */
1367                 reg = E1000_READ_REG(hw, E1000_DMCTLX);
1368                 reg |= 0x800000FF; /* 255 usec */
1369                 E1000_WRITE_REG(hw, E1000_DMCTLX, reg);
1370
1371                 /* free space in tx packet buffer to wake from DMA coal */
1372                 E1000_WRITE_REG(hw, E1000_DMCTXTH,
1373                     (20480 - (2 * sc->max_frame_size)) >> 6);
1374
1375                 /* make low power state decision controlled by DMA coal */
1376                 reg = E1000_READ_REG(hw, E1000_PCIEMISC);
1377                 E1000_WRITE_REG(hw, E1000_PCIEMISC,
1378                     reg | E1000_PCIEMISC_LX_DECISION);
1379                 if_printf(ifp, "DMA Coalescing enabled\n");
1380         }
1381
1382         E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN);
1383         e1000_get_phy_info(hw);
1384         e1000_check_for_link(hw);
1385 }
1386
1387 static void
1388 igb_setup_ifp(struct igb_softc *sc)
1389 {
1390         struct ifnet *ifp = &sc->arpcom.ac_if;
1391
1392         if_initname(ifp, device_get_name(sc->dev), device_get_unit(sc->dev));
1393         ifp->if_softc = sc;
1394         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
1395         ifp->if_init = igb_init;
1396         ifp->if_ioctl = igb_ioctl;
1397         ifp->if_start = igb_start;
1398         ifp->if_serialize = igb_serialize;
1399         ifp->if_deserialize = igb_deserialize;
1400         ifp->if_tryserialize = igb_tryserialize;
1401 #ifdef INVARIANTS
1402         ifp->if_serialize_assert = igb_serialize_assert;
1403 #endif
1404 #ifdef DEVICE_POLLING
1405         ifp->if_poll = igb_poll;
1406 #endif
1407         ifp->if_watchdog = igb_watchdog;
1408
1409         ifq_set_maxlen(&ifp->if_snd, sc->tx_rings[0].num_tx_desc - 1);
1410         ifq_set_ready(&ifp->if_snd);
1411
1412         ether_ifattach(ifp, sc->hw.mac.addr, NULL);
1413
1414         ifp->if_capabilities =
1415             IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_TSO;
1416         if (IGB_ENABLE_HWRSS(sc))
1417                 ifp->if_capabilities |= IFCAP_RSS;
1418         ifp->if_capenable = ifp->if_capabilities;
1419         ifp->if_hwassist = IGB_CSUM_FEATURES | CSUM_TSO;
1420
1421         /*
1422          * Tell the upper layer(s) we support long frames
1423          */
1424         ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
1425
1426         /*
1427          * Specify the media types supported by this adapter and register
1428          * callbacks to update media and link information
1429          */
1430         ifmedia_init(&sc->media, IFM_IMASK, igb_media_change, igb_media_status);
1431         if (sc->hw.phy.media_type == e1000_media_type_fiber ||
1432             sc->hw.phy.media_type == e1000_media_type_internal_serdes) {
1433                 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX | IFM_FDX,
1434                     0, NULL);
1435                 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL);
1436         } else {
1437                 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T, 0, NULL);
1438                 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T | IFM_FDX,
1439                     0, NULL);
1440                 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX, 0, NULL);
1441                 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
1442                     0, NULL);
1443                 if (sc->hw.phy.type != e1000_phy_ife) {
1444                         ifmedia_add(&sc->media,
1445                             IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
1446                         ifmedia_add(&sc->media,
1447                             IFM_ETHER | IFM_1000_T, 0, NULL);
1448                 }
1449         }
1450         ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL);
1451         ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO);
1452 }
1453
1454 static void
1455 igb_add_sysctl(struct igb_softc *sc)
1456 {
1457         char node[32];
1458         int i;
1459
1460         sysctl_ctx_init(&sc->sysctl_ctx);
1461         sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
1462             SYSCTL_STATIC_CHILDREN(_hw), OID_AUTO,
1463             device_get_nameunit(sc->dev), CTLFLAG_RD, 0, "");
1464         if (sc->sysctl_tree == NULL) {
1465                 device_printf(sc->dev, "can't add sysctl node\n");
1466                 return;
1467         }
1468
1469         SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1470             OID_AUTO, "rxr", CTLFLAG_RD, &sc->rx_ring_cnt, 0, "# of RX rings");
1471         SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1472             OID_AUTO, "rxr_inuse", CTLFLAG_RD, &sc->rx_ring_inuse, 0,
1473             "# of RX rings used");
1474         SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1475             OID_AUTO, "rxd", CTLFLAG_RD, &sc->rx_rings[0].num_rx_desc, 0,
1476             "# of RX descs");
1477         SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1478             OID_AUTO, "txd", CTLFLAG_RD, &sc->tx_rings[0].num_tx_desc, 0,
1479             "# of TX descs");
1480
1481         if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
1482                 SYSCTL_ADD_PROC(&sc->sysctl_ctx,
1483                     SYSCTL_CHILDREN(sc->sysctl_tree),
1484                     OID_AUTO, "intr_rate", CTLTYPE_INT | CTLFLAG_RW,
1485                     sc, 0, igb_sysctl_intr_rate, "I", "interrupt rate");
1486         } else {
1487                 for (i = 0; i < sc->msix_cnt; ++i) {
1488                         struct igb_msix_data *msix = &sc->msix_data[i];
1489
1490                         ksnprintf(node, sizeof(node), "msix%d_rate", i);
1491                         SYSCTL_ADD_PROC(&sc->sysctl_ctx,
1492                             SYSCTL_CHILDREN(sc->sysctl_tree),
1493                             OID_AUTO, node, CTLTYPE_INT | CTLFLAG_RW,
1494                             msix, 0, igb_sysctl_msix_rate, "I",
1495                             msix->msix_rate_desc);
1496                 }
1497         }
1498
1499         SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1500             OID_AUTO, "tx_intr_nsegs", CTLTYPE_INT | CTLFLAG_RW,
1501             sc, 0, igb_sysctl_tx_intr_nsegs, "I",
1502             "# of segments per TX interrupt");
1503
1504 #ifdef IGB_RSS_DEBUG
1505         SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree),
1506             OID_AUTO, "rss_debug", CTLFLAG_RW, &sc->rss_debug, 0,
1507             "RSS debug level");
1508         for (i = 0; i < sc->rx_ring_cnt; ++i) {
1509                 ksnprintf(node, sizeof(node), "rx%d_pkt", i);
1510                 SYSCTL_ADD_ULONG(&sc->sysctl_ctx,
1511                     SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, node,
1512                     CTLFLAG_RW, &sc->rx_rings[i].rx_packets, "RXed packets");
1513         }
1514 #endif
1515 }
1516
1517 static int
1518 igb_alloc_rings(struct igb_softc *sc)
1519 {
1520         int error, i;
1521
1522         /*
1523          * Create top level busdma tag
1524          */
1525         error = bus_dma_tag_create(NULL, 1, 0,
1526             BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL,
1527             BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0,
1528             &sc->parent_tag);
1529         if (error) {
1530                 device_printf(sc->dev, "could not create top level DMA tag\n");
1531                 return error;
1532         }
1533
1534         /*
1535          * Allocate TX descriptor rings and buffers
1536          */
1537         sc->tx_rings = kmalloc(sizeof(struct igb_tx_ring) * sc->tx_ring_cnt,
1538             M_DEVBUF, M_WAITOK | M_ZERO);
1539         for (i = 0; i < sc->tx_ring_cnt; ++i) {
1540                 struct igb_tx_ring *txr = &sc->tx_rings[i];
1541
1542                 /* Set up some basics */
1543                 txr->sc = sc;
1544                 txr->me = i;
1545                 lwkt_serialize_init(&txr->tx_serialize);
1546
1547                 error = igb_create_tx_ring(txr);
1548                 if (error)
1549                         return error;
1550         }
1551
1552         /*
1553          * Allocate RX descriptor rings and buffers
1554          */
1555         sc->rx_rings = kmalloc(sizeof(struct igb_rx_ring) * sc->rx_ring_cnt,
1556             M_DEVBUF, M_WAITOK | M_ZERO);
1557         for (i = 0; i < sc->rx_ring_cnt; ++i) {
1558                 struct igb_rx_ring *rxr = &sc->rx_rings[i];
1559
1560                 /* Set up some basics */
1561                 rxr->sc = sc;
1562                 rxr->me = i;
1563                 lwkt_serialize_init(&rxr->rx_serialize);
1564
1565                 error = igb_create_rx_ring(rxr);
1566                 if (error)
1567                         return error;
1568         }
1569
1570         return 0;
1571 }
1572
1573 static void
1574 igb_free_rings(struct igb_softc *sc)
1575 {
1576         int i;
1577
1578         if (sc->tx_rings != NULL) {
1579                 for (i = 0; i < sc->tx_ring_cnt; ++i) {
1580                         struct igb_tx_ring *txr = &sc->tx_rings[i];
1581
1582                         igb_destroy_tx_ring(txr, txr->num_tx_desc);
1583                 }
1584                 kfree(sc->tx_rings, M_DEVBUF);
1585         }
1586
1587         if (sc->rx_rings != NULL) {
1588                 for (i = 0; i < sc->rx_ring_cnt; ++i) {
1589                         struct igb_rx_ring *rxr = &sc->rx_rings[i];
1590
1591                         igb_destroy_rx_ring(rxr, rxr->num_rx_desc);
1592                 }
1593                 kfree(sc->rx_rings, M_DEVBUF);
1594         }
1595 }
1596
1597 static int
1598 igb_create_tx_ring(struct igb_tx_ring *txr)
1599 {
1600         int tsize, error, i;
1601
1602         /*
1603          * Validate number of transmit descriptors. It must not exceed
1604          * hardware maximum, and must be multiple of IGB_DBA_ALIGN.
1605          */
1606         if (((igb_txd * sizeof(struct e1000_tx_desc)) % IGB_DBA_ALIGN) != 0 ||
1607             (igb_txd > IGB_MAX_TXD) || (igb_txd < IGB_MIN_TXD)) {
1608                 device_printf(txr->sc->dev,
1609                     "Using %d TX descriptors instead of %d!\n",
1610                     IGB_DEFAULT_TXD, igb_txd);
1611                 txr->num_tx_desc = IGB_DEFAULT_TXD;
1612         } else {
1613                 txr->num_tx_desc = igb_txd;
1614         }
1615
1616         /*
1617          * Allocate TX descriptor ring
1618          */
1619         tsize = roundup2(txr->num_tx_desc * sizeof(union e1000_adv_tx_desc),
1620             IGB_DBA_ALIGN);
1621         txr->txdma.dma_vaddr = bus_dmamem_coherent_any(txr->sc->parent_tag,
1622             IGB_DBA_ALIGN, tsize, BUS_DMA_WAITOK,
1623             &txr->txdma.dma_tag, &txr->txdma.dma_map, &txr->txdma.dma_paddr);
1624         if (txr->txdma.dma_vaddr == NULL) {
1625                 device_printf(txr->sc->dev,
1626                     "Unable to allocate TX Descriptor memory\n");
1627                 return ENOMEM;
1628         }
1629         txr->tx_base = txr->txdma.dma_vaddr;
1630         bzero(txr->tx_base, tsize);
1631
1632         txr->tx_buf = kmalloc(sizeof(struct igb_tx_buf) * txr->num_tx_desc,
1633             M_DEVBUF, M_WAITOK | M_ZERO);
1634
1635         /*
1636          * Allocate TX head write-back buffer
1637          */
1638         txr->tx_hdr = bus_dmamem_coherent_any(txr->sc->parent_tag,
1639             __VM_CACHELINE_SIZE, __VM_CACHELINE_SIZE, BUS_DMA_WAITOK,
1640             &txr->tx_hdr_dtag, &txr->tx_hdr_dmap, &txr->tx_hdr_paddr);
1641         if (txr->tx_hdr == NULL) {
1642                 device_printf(txr->sc->dev,
1643                     "Unable to allocate TX head write-back buffer\n");
1644                 return ENOMEM;
1645         }
1646
1647         /*
1648          * Create DMA tag for TX buffers
1649          */
1650         error = bus_dma_tag_create(txr->sc->parent_tag,
1651             1, 0,               /* alignment, bounds */
1652             BUS_SPACE_MAXADDR,  /* lowaddr */
1653             BUS_SPACE_MAXADDR,  /* highaddr */
1654             NULL, NULL,         /* filter, filterarg */
1655             IGB_TSO_SIZE,       /* maxsize */
1656             IGB_MAX_SCATTER,    /* nsegments */
1657             PAGE_SIZE,          /* maxsegsize */
1658             BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW |
1659             BUS_DMA_ONEBPAGE,   /* flags */
1660             &txr->tx_tag);
1661         if (error) {
1662                 device_printf(txr->sc->dev, "Unable to allocate TX DMA tag\n");
1663                 kfree(txr->tx_buf, M_DEVBUF);
1664                 txr->tx_buf = NULL;
1665                 return error;
1666         }
1667
1668         /*
1669          * Create DMA maps for TX buffers
1670          */
1671         for (i = 0; i < txr->num_tx_desc; ++i) {
1672                 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1673
1674                 error = bus_dmamap_create(txr->tx_tag,
1675                     BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE, &txbuf->map);
1676                 if (error) {
1677                         device_printf(txr->sc->dev,
1678                             "Unable to create TX DMA map\n");
1679                         igb_destroy_tx_ring(txr, i);
1680                         return error;
1681                 }
1682         }
1683
1684         /*
1685          * Initialize various watermark
1686          */
1687         txr->spare_desc = IGB_TX_SPARE;
1688         txr->intr_nsegs = txr->num_tx_desc / 16;
1689         txr->oact_hi_desc = txr->num_tx_desc / 2;
1690         txr->oact_lo_desc = txr->num_tx_desc / 8;
1691         if (txr->oact_lo_desc > IGB_TX_OACTIVE_MAX)
1692                 txr->oact_lo_desc = IGB_TX_OACTIVE_MAX;
1693         if (txr->oact_lo_desc < txr->spare_desc + IGB_TX_RESERVED)
1694                 txr->oact_lo_desc = txr->spare_desc + IGB_TX_RESERVED;
1695
1696         return 0;
1697 }
1698
1699 static void
1700 igb_free_tx_ring(struct igb_tx_ring *txr)
1701 {
1702         int i;
1703
1704         for (i = 0; i < txr->num_tx_desc; ++i) {
1705                 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1706
1707                 if (txbuf->m_head != NULL) {
1708                         bus_dmamap_unload(txr->tx_tag, txbuf->map);
1709                         m_freem(txbuf->m_head);
1710                         txbuf->m_head = NULL;
1711                 }
1712         }
1713 }
1714
1715 static void
1716 igb_destroy_tx_ring(struct igb_tx_ring *txr, int ndesc)
1717 {
1718         int i;
1719
1720         if (txr->txdma.dma_vaddr != NULL) {
1721                 bus_dmamap_unload(txr->txdma.dma_tag, txr->txdma.dma_map);
1722                 bus_dmamem_free(txr->txdma.dma_tag, txr->txdma.dma_vaddr,
1723                     txr->txdma.dma_map);
1724                 bus_dma_tag_destroy(txr->txdma.dma_tag);
1725                 txr->txdma.dma_vaddr = NULL;
1726         }
1727
1728         if (txr->tx_hdr != NULL) {
1729                 bus_dmamap_unload(txr->tx_hdr_dtag, txr->tx_hdr_dmap);
1730                 bus_dmamem_free(txr->tx_hdr_dtag, txr->tx_hdr,
1731                     txr->tx_hdr_dmap);
1732                 bus_dma_tag_destroy(txr->tx_hdr_dtag);
1733                 txr->tx_hdr = NULL;
1734         }
1735
1736         if (txr->tx_buf == NULL)
1737                 return;
1738
1739         for (i = 0; i < ndesc; ++i) {
1740                 struct igb_tx_buf *txbuf = &txr->tx_buf[i];
1741
1742                 KKASSERT(txbuf->m_head == NULL);
1743                 bus_dmamap_destroy(txr->tx_tag, txbuf->map);
1744         }
1745         bus_dma_tag_destroy(txr->tx_tag);
1746
1747         kfree(txr->tx_buf, M_DEVBUF);
1748         txr->tx_buf = NULL;
1749 }
1750
1751 static void
1752 igb_init_tx_ring(struct igb_tx_ring *txr)
1753 {
1754         /* Clear the old descriptor contents */
1755         bzero(txr->tx_base,
1756             sizeof(union e1000_adv_tx_desc) * txr->num_tx_desc);
1757
1758         /* Clear TX head write-back buffer */
1759         *(txr->tx_hdr) = 0;
1760
1761         /* Reset indices */
1762         txr->next_avail_desc = 0;
1763         txr->next_to_clean = 0;
1764         txr->tx_nsegs = 0;
1765
1766         /* Set number of descriptors available */
1767         txr->tx_avail = txr->num_tx_desc;
1768 }
1769
1770 static void
1771 igb_init_tx_unit(struct igb_softc *sc)
1772 {
1773         struct e1000_hw *hw = &sc->hw;
1774         uint32_t tctl;
1775         int i;
1776
1777         /* Setup the Tx Descriptor Rings */
1778         for (i = 0; i < sc->tx_ring_cnt; ++i) {
1779                 struct igb_tx_ring *txr = &sc->tx_rings[i];
1780                 uint64_t bus_addr = txr->txdma.dma_paddr;
1781                 uint64_t hdr_paddr = txr->tx_hdr_paddr;
1782                 uint32_t txdctl = 0;
1783                 uint32_t dca_txctrl;
1784
1785                 E1000_WRITE_REG(hw, E1000_TDLEN(i),
1786                     txr->num_tx_desc * sizeof(struct e1000_tx_desc));
1787                 E1000_WRITE_REG(hw, E1000_TDBAH(i),
1788                     (uint32_t)(bus_addr >> 32));
1789                 E1000_WRITE_REG(hw, E1000_TDBAL(i),
1790                     (uint32_t)bus_addr);
1791
1792                 /* Setup the HW Tx Head and Tail descriptor pointers */
1793                 E1000_WRITE_REG(hw, E1000_TDT(i), 0);
1794                 E1000_WRITE_REG(hw, E1000_TDH(i), 0);
1795
1796                 /*
1797                  * WTHRESH is ignored by the hardware, since header
1798                  * write back mode is used.
1799                  */
1800                 txdctl |= IGB_TX_PTHRESH;
1801                 txdctl |= IGB_TX_HTHRESH << 8;
1802                 txdctl |= IGB_TX_WTHRESH << 16;
1803                 txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
1804                 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl);
1805
1806                 dca_txctrl = E1000_READ_REG(hw, E1000_DCA_TXCTRL(i));
1807                 dca_txctrl &= ~E1000_DCA_TXCTRL_TX_WB_RO_EN;
1808                 E1000_WRITE_REG(hw, E1000_DCA_TXCTRL(i), dca_txctrl);
1809
1810                 /*
1811                  * Don't set WB_on_EITR:
1812                  * - 82575 does not have it
1813                  * - It almost has no effect on 82576, see:
1814                  *   82576 specification update errata #26
1815                  * - It causes unnecessary bus traffic
1816                  */
1817                 E1000_WRITE_REG(hw, E1000_TDWBAH(i),
1818                     (uint32_t)(hdr_paddr >> 32));
1819                 E1000_WRITE_REG(hw, E1000_TDWBAL(i),
1820                     ((uint32_t)hdr_paddr) | E1000_TX_HEAD_WB_ENABLE);
1821         }
1822
1823         if (sc->vf_ifp)
1824                 return;
1825
1826         e1000_config_collision_dist(hw);
1827
1828         /* Program the Transmit Control Register */
1829         tctl = E1000_READ_REG(hw, E1000_TCTL);
1830         tctl &= ~E1000_TCTL_CT;
1831         tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
1832             (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT));
1833
1834         /* This write will effectively turn on the transmit unit. */
1835         E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1836 }
1837
1838 static boolean_t
1839 igb_txcsum_ctx(struct igb_tx_ring *txr, struct mbuf *mp)
1840 {
1841         struct e1000_adv_tx_context_desc *TXD;
1842         uint32_t vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx;
1843         int ehdrlen, ctxd, ip_hlen = 0;
1844         boolean_t offload = TRUE;
1845
1846         if ((mp->m_pkthdr.csum_flags & IGB_CSUM_FEATURES) == 0)
1847                 offload = FALSE;
1848
1849         vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0;
1850
1851         ctxd = txr->next_avail_desc;
1852         TXD = (struct e1000_adv_tx_context_desc *)&txr->tx_base[ctxd];
1853
1854         /*
1855          * In advanced descriptors the vlan tag must
1856          * be placed into the context descriptor, thus
1857          * we need to be here just for that setup.
1858          */
1859         if (mp->m_flags & M_VLANTAG) {
1860                 uint16_t vlantag;
1861
1862                 vlantag = htole16(mp->m_pkthdr.ether_vlantag);
1863                 vlan_macip_lens |= (vlantag << E1000_ADVTXD_VLAN_SHIFT);
1864         } else if (!offload) {
1865                 return FALSE;
1866         }
1867
1868         ehdrlen = mp->m_pkthdr.csum_lhlen;
1869         KASSERT(ehdrlen > 0, ("invalid ether hlen"));
1870
1871         /* Set the ether header length */
1872         vlan_macip_lens |= ehdrlen << E1000_ADVTXD_MACLEN_SHIFT;
1873         if (mp->m_pkthdr.csum_flags & CSUM_IP) {
1874                 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
1875                 ip_hlen = mp->m_pkthdr.csum_iphlen;
1876                 KASSERT(ip_hlen > 0, ("invalid ip hlen"));
1877         }
1878         vlan_macip_lens |= ip_hlen;
1879
1880         type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
1881         if (mp->m_pkthdr.csum_flags & CSUM_TCP)
1882                 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
1883         else if (mp->m_pkthdr.csum_flags & CSUM_UDP)
1884                 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP;
1885
1886         /* 82575 needs the queue index added */
1887         if (txr->sc->hw.mac.type == e1000_82575)
1888                 mss_l4len_idx = txr->me << 4;
1889
1890         /* Now copy bits into descriptor */
1891         TXD->vlan_macip_lens = htole32(vlan_macip_lens);
1892         TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
1893         TXD->seqnum_seed = htole32(0);
1894         TXD->mss_l4len_idx = htole32(mss_l4len_idx);
1895
1896         /* We've consumed the first desc, adjust counters */
1897         if (++ctxd == txr->num_tx_desc)
1898                 ctxd = 0;
1899         txr->next_avail_desc = ctxd;
1900         --txr->tx_avail;
1901
1902         return offload;
1903 }
1904
1905 static void
1906 igb_txeof(struct igb_tx_ring *txr)
1907 {
1908         struct ifnet *ifp = &txr->sc->arpcom.ac_if;
1909         int first, hdr, avail;
1910
1911         if (txr->tx_avail == txr->num_tx_desc)
1912                 return;
1913
1914         first = txr->next_to_clean;
1915         hdr = *(txr->tx_hdr);
1916
1917         if (first == hdr)
1918                 return;
1919
1920         avail = txr->tx_avail;
1921         while (first != hdr) {
1922                 struct igb_tx_buf *txbuf = &txr->tx_buf[first];
1923
1924                 ++avail;
1925                 if (txbuf->m_head) {
1926                         bus_dmamap_unload(txr->tx_tag, txbuf->map);
1927                         m_freem(txbuf->m_head);
1928                         txbuf->m_head = NULL;
1929                         ++ifp->if_opackets;
1930                 }
1931                 if (++first == txr->num_tx_desc)
1932                         first = 0;
1933         }
1934         txr->next_to_clean = first;
1935         txr->tx_avail = avail;
1936
1937         /*
1938          * If we have a minimum free, clear IFF_OACTIVE
1939          * to tell the stack that it is OK to send packets.
1940          */
1941         if (IGB_IS_NOT_OACTIVE(txr)) {
1942                 ifp->if_flags &= ~IFF_OACTIVE;
1943
1944                 /*
1945                  * We have enough TX descriptors, turn off
1946                  * the watchdog.  We allow small amount of
1947                  * packets (roughly intr_nsegs) pending on
1948                  * the transmit ring.
1949                  */
1950                 ifp->if_timer = 0;
1951         }
1952 }
1953
1954 static int
1955 igb_create_rx_ring(struct igb_rx_ring *rxr)
1956 {
1957         int rsize, i, error;
1958
1959         /*
1960          * Validate number of receive descriptors. It must not exceed
1961          * hardware maximum, and must be multiple of IGB_DBA_ALIGN.
1962          */
1963         if (((igb_rxd * sizeof(struct e1000_rx_desc)) % IGB_DBA_ALIGN) != 0 ||
1964             (igb_rxd > IGB_MAX_RXD) || (igb_rxd < IGB_MIN_RXD)) {
1965                 device_printf(rxr->sc->dev,
1966                     "Using %d RX descriptors instead of %d!\n",
1967                     IGB_DEFAULT_RXD, igb_rxd);
1968                 rxr->num_rx_desc = IGB_DEFAULT_RXD;
1969         } else {
1970                 rxr->num_rx_desc = igb_rxd;
1971         }
1972
1973         /*
1974          * Allocate RX descriptor ring
1975          */
1976         rsize = roundup2(rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc),
1977             IGB_DBA_ALIGN);
1978         rxr->rxdma.dma_vaddr = bus_dmamem_coherent_any(rxr->sc->parent_tag,
1979             IGB_DBA_ALIGN, rsize, BUS_DMA_WAITOK,
1980             &rxr->rxdma.dma_tag, &rxr->rxdma.dma_map,
1981             &rxr->rxdma.dma_paddr);
1982         if (rxr->rxdma.dma_vaddr == NULL) {
1983                 device_printf(rxr->sc->dev,
1984                     "Unable to allocate RxDescriptor memory\n");
1985                 return ENOMEM;
1986         }
1987         rxr->rx_base = rxr->rxdma.dma_vaddr;
1988         bzero(rxr->rx_base, rsize);
1989
1990         rxr->rx_buf = kmalloc(sizeof(struct igb_rx_buf) * rxr->num_rx_desc,
1991             M_DEVBUF, M_WAITOK | M_ZERO);
1992
1993         /*
1994          * Create DMA tag for RX buffers
1995          */
1996         error = bus_dma_tag_create(rxr->sc->parent_tag,
1997             1, 0,               /* alignment, bounds */
1998             BUS_SPACE_MAXADDR,  /* lowaddr */
1999             BUS_SPACE_MAXADDR,  /* highaddr */
2000             NULL, NULL,         /* filter, filterarg */
2001             MCLBYTES,           /* maxsize */
2002             1,                  /* nsegments */
2003             MCLBYTES,           /* maxsegsize */
2004             BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW, /* flags */
2005             &rxr->rx_tag);
2006         if (error) {
2007                 device_printf(rxr->sc->dev,
2008                     "Unable to create RX payload DMA tag\n");
2009                 kfree(rxr->rx_buf, M_DEVBUF);
2010                 rxr->rx_buf = NULL;
2011                 return error;
2012         }
2013
2014         /*
2015          * Create spare DMA map for RX buffers
2016          */
2017         error = bus_dmamap_create(rxr->rx_tag, BUS_DMA_WAITOK,
2018             &rxr->rx_sparemap);
2019         if (error) {
2020                 device_printf(rxr->sc->dev,
2021                     "Unable to create spare RX DMA maps\n");
2022                 bus_dma_tag_destroy(rxr->rx_tag);
2023                 kfree(rxr->rx_buf, M_DEVBUF);
2024                 rxr->rx_buf = NULL;
2025                 return error;
2026         }
2027
2028         /*
2029          * Create DMA maps for RX buffers
2030          */
2031         for (i = 0; i < rxr->num_rx_desc; i++) {
2032                 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2033
2034                 error = bus_dmamap_create(rxr->rx_tag,
2035                     BUS_DMA_WAITOK, &rxbuf->map);
2036                 if (error) {
2037                         device_printf(rxr->sc->dev,
2038                             "Unable to create RX DMA maps\n");
2039                         igb_destroy_rx_ring(rxr, i);
2040                         return error;
2041                 }
2042         }
2043         return 0;
2044 }
2045
2046 static void
2047 igb_free_rx_ring(struct igb_rx_ring *rxr)
2048 {
2049         int i;
2050
2051         for (i = 0; i < rxr->num_rx_desc; ++i) {
2052                 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2053
2054                 if (rxbuf->m_head != NULL) {
2055                         bus_dmamap_unload(rxr->rx_tag, rxbuf->map);
2056                         m_freem(rxbuf->m_head);
2057                         rxbuf->m_head = NULL;
2058                 }
2059         }
2060
2061         if (rxr->fmp != NULL)
2062                 m_freem(rxr->fmp);
2063         rxr->fmp = NULL;
2064         rxr->lmp = NULL;
2065 }
2066
2067 static void
2068 igb_destroy_rx_ring(struct igb_rx_ring *rxr, int ndesc)
2069 {
2070         int i;
2071
2072         if (rxr->rxdma.dma_vaddr != NULL) {
2073                 bus_dmamap_unload(rxr->rxdma.dma_tag, rxr->rxdma.dma_map);
2074                 bus_dmamem_free(rxr->rxdma.dma_tag, rxr->rxdma.dma_vaddr,
2075                     rxr->rxdma.dma_map);
2076                 bus_dma_tag_destroy(rxr->rxdma.dma_tag);
2077                 rxr->rxdma.dma_vaddr = NULL;
2078         }
2079
2080         if (rxr->rx_buf == NULL)
2081                 return;
2082
2083         for (i = 0; i < ndesc; ++i) {
2084                 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2085
2086                 KKASSERT(rxbuf->m_head == NULL);
2087                 bus_dmamap_destroy(rxr->rx_tag, rxbuf->map);
2088         }
2089         bus_dmamap_destroy(rxr->rx_tag, rxr->rx_sparemap);
2090         bus_dma_tag_destroy(rxr->rx_tag);
2091
2092         kfree(rxr->rx_buf, M_DEVBUF);
2093         rxr->rx_buf = NULL;
2094 }
2095
2096 static void
2097 igb_setup_rxdesc(union e1000_adv_rx_desc *rxd, const struct igb_rx_buf *rxbuf)
2098 {
2099         rxd->read.pkt_addr = htole64(rxbuf->paddr);
2100         rxd->wb.upper.status_error = 0;
2101 }
2102
2103 static int
2104 igb_newbuf(struct igb_rx_ring *rxr, int i, boolean_t wait)
2105 {
2106         struct mbuf *m;
2107         bus_dma_segment_t seg;
2108         bus_dmamap_t map;
2109         struct igb_rx_buf *rxbuf;
2110         int error, nseg;
2111
2112         m = m_getcl(wait ? MB_WAIT : MB_DONTWAIT, MT_DATA, M_PKTHDR);
2113         if (m == NULL) {
2114                 if (wait) {
2115                         if_printf(&rxr->sc->arpcom.ac_if,
2116                             "Unable to allocate RX mbuf\n");
2117                 }
2118                 return ENOBUFS;
2119         }
2120         m->m_len = m->m_pkthdr.len = MCLBYTES;
2121
2122         if (rxr->sc->max_frame_size <= MCLBYTES - ETHER_ALIGN)
2123                 m_adj(m, ETHER_ALIGN);
2124
2125         error = bus_dmamap_load_mbuf_segment(rxr->rx_tag,
2126             rxr->rx_sparemap, m, &seg, 1, &nseg, BUS_DMA_NOWAIT);
2127         if (error) {
2128                 m_freem(m);
2129                 if (wait) {
2130                         if_printf(&rxr->sc->arpcom.ac_if,
2131                             "Unable to load RX mbuf\n");
2132                 }
2133                 return error;
2134         }
2135
2136         rxbuf = &rxr->rx_buf[i];
2137         if (rxbuf->m_head != NULL)
2138                 bus_dmamap_unload(rxr->rx_tag, rxbuf->map);
2139
2140         map = rxbuf->map;
2141         rxbuf->map = rxr->rx_sparemap;
2142         rxr->rx_sparemap = map;
2143
2144         rxbuf->m_head = m;
2145         rxbuf->paddr = seg.ds_addr;
2146
2147         igb_setup_rxdesc(&rxr->rx_base[i], rxbuf);
2148         return 0;
2149 }
2150
2151 static int
2152 igb_init_rx_ring(struct igb_rx_ring *rxr)
2153 {
2154         int i;
2155
2156         /* Clear the ring contents */
2157         bzero(rxr->rx_base,
2158             rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc));
2159
2160         /* Now replenish the ring mbufs */
2161         for (i = 0; i < rxr->num_rx_desc; ++i) {
2162                 int error;
2163
2164                 error = igb_newbuf(rxr, i, TRUE);
2165                 if (error)
2166                         return error;
2167         }
2168
2169         /* Setup our descriptor indices */
2170         rxr->next_to_check = 0;
2171
2172         rxr->fmp = NULL;
2173         rxr->lmp = NULL;
2174         rxr->discard = FALSE;
2175
2176         return 0;
2177 }
2178
2179 static void
2180 igb_init_rx_unit(struct igb_softc *sc)
2181 {
2182         struct ifnet *ifp = &sc->arpcom.ac_if;
2183         struct e1000_hw *hw = &sc->hw;
2184         uint32_t rctl, rxcsum, srrctl = 0;
2185         int i;
2186
2187         /*
2188          * Make sure receives are disabled while setting
2189          * up the descriptor ring
2190          */
2191         rctl = E1000_READ_REG(hw, E1000_RCTL);
2192         E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
2193
2194 #if 0
2195         /*
2196         ** Set up for header split
2197         */
2198         if (igb_header_split) {
2199                 /* Use a standard mbuf for the header */
2200                 srrctl |= IGB_HDR_BUF << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
2201                 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS;
2202         } else
2203 #endif
2204                 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
2205
2206         /*
2207         ** Set up for jumbo frames
2208         */
2209         if (ifp->if_mtu > ETHERMTU) {
2210                 rctl |= E1000_RCTL_LPE;
2211 #if 0
2212                 if (adapter->rx_mbuf_sz == MJUMPAGESIZE) {
2213                         srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2214                         rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX;
2215                 } else if (adapter->rx_mbuf_sz > MJUMPAGESIZE) {
2216                         srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2217                         rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX;
2218                 }
2219                 /* Set maximum packet len */
2220                 psize = adapter->max_frame_size;
2221                 /* are we on a vlan? */
2222                 if (adapter->ifp->if_vlantrunk != NULL)
2223                         psize += VLAN_TAG_SIZE;
2224                 E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize);
2225 #else
2226                 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2227                 rctl |= E1000_RCTL_SZ_2048;
2228 #endif
2229         } else {
2230                 rctl &= ~E1000_RCTL_LPE;
2231                 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
2232                 rctl |= E1000_RCTL_SZ_2048;
2233         }
2234
2235         /* Setup the Base and Length of the Rx Descriptor Rings */
2236         for (i = 0; i < sc->rx_ring_inuse; ++i) {
2237                 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2238                 uint64_t bus_addr = rxr->rxdma.dma_paddr;
2239                 uint32_t rxdctl;
2240
2241                 E1000_WRITE_REG(hw, E1000_RDLEN(i),
2242                     rxr->num_rx_desc * sizeof(struct e1000_rx_desc));
2243                 E1000_WRITE_REG(hw, E1000_RDBAH(i),
2244                     (uint32_t)(bus_addr >> 32));
2245                 E1000_WRITE_REG(hw, E1000_RDBAL(i),
2246                     (uint32_t)bus_addr);
2247                 E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl);
2248                 /* Enable this Queue */
2249                 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i));
2250                 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
2251                 rxdctl &= 0xFFF00000;
2252                 rxdctl |= IGB_RX_PTHRESH;
2253                 rxdctl |= IGB_RX_HTHRESH << 8;
2254                 /*
2255                  * Don't set WTHRESH to a value above 1 on 82576, see:
2256                  * 82576 specification update errata #26
2257                  */
2258                 rxdctl |= IGB_RX_WTHRESH << 16;
2259                 E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl);
2260         }
2261
2262         rxcsum = E1000_READ_REG(&sc->hw, E1000_RXCSUM);
2263         rxcsum &= ~(E1000_RXCSUM_PCSS_MASK | E1000_RXCSUM_IPPCSE);
2264
2265         /*
2266          * Receive Checksum Offload for TCP and UDP
2267          *
2268          * Checksum offloading is also enabled if multiple receive
2269          * queue is to be supported, since we need it to figure out
2270          * fragments.
2271          */
2272         if ((ifp->if_capenable & IFCAP_RXCSUM) || IGB_ENABLE_HWRSS(sc)) {
2273                 /*
2274                  * NOTE:
2275                  * PCSD must be enabled to enable multiple
2276                  * receive queues.
2277                  */
2278                 rxcsum |= E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL |
2279                     E1000_RXCSUM_PCSD;
2280         } else {
2281                 rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL |
2282                     E1000_RXCSUM_PCSD);
2283         }
2284         E1000_WRITE_REG(&sc->hw, E1000_RXCSUM, rxcsum);
2285
2286         if (IGB_ENABLE_HWRSS(sc)) {
2287                 uint8_t key[IGB_NRSSRK * IGB_RSSRK_SIZE];
2288                 uint32_t reta_shift;
2289                 int j, r;
2290
2291                 /*
2292                  * NOTE:
2293                  * When we reach here, RSS has already been disabled
2294                  * in igb_stop(), so we could safely configure RSS key
2295                  * and redirect table.
2296                  */
2297
2298                 /*
2299                  * Configure RSS key
2300                  */
2301                 toeplitz_get_key(key, sizeof(key));
2302                 for (i = 0; i < IGB_NRSSRK; ++i) {
2303                         uint32_t rssrk;
2304
2305                         rssrk = IGB_RSSRK_VAL(key, i);
2306                         IGB_RSS_DPRINTF(sc, 1, "rssrk%d 0x%08x\n", i, rssrk);
2307
2308                         E1000_WRITE_REG(hw, E1000_RSSRK(i), rssrk);
2309                 }
2310
2311                 /*
2312                  * Configure RSS redirect table in following fashion:
2313                  * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)]
2314                  */
2315                 reta_shift = IGB_RETA_SHIFT;
2316                 if (hw->mac.type == e1000_82575)
2317                         reta_shift = IGB_RETA_SHIFT_82575;
2318
2319                 r = 0;
2320                 for (j = 0; j < IGB_NRETA; ++j) {
2321                         uint32_t reta = 0;
2322
2323                         for (i = 0; i < IGB_RETA_SIZE; ++i) {
2324                                 uint32_t q;
2325
2326                                 q = (r % sc->rx_ring_inuse) << reta_shift;
2327                                 reta |= q << (8 * i);
2328                                 ++r;
2329                         }
2330                         IGB_RSS_DPRINTF(sc, 1, "reta 0x%08x\n", reta);
2331                         E1000_WRITE_REG(hw, E1000_RETA(j), reta);
2332                 }
2333
2334                 /*
2335                  * Enable multiple receive queues.
2336                  * Enable IPv4 RSS standard hash functions.
2337                  * Disable RSS interrupt on 82575
2338                  */
2339                 E1000_WRITE_REG(&sc->hw, E1000_MRQC,
2340                                 E1000_MRQC_ENABLE_RSS_4Q |
2341                                 E1000_MRQC_RSS_FIELD_IPV4_TCP |
2342                                 E1000_MRQC_RSS_FIELD_IPV4);
2343         }
2344
2345         /* Setup the Receive Control Register */
2346         rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
2347         rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
2348             E1000_RCTL_RDMTS_HALF |
2349             (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
2350         /* Strip CRC bytes. */
2351         rctl |= E1000_RCTL_SECRC;
2352         /* Make sure VLAN Filters are off */
2353         rctl &= ~E1000_RCTL_VFE;
2354         /* Don't store bad packets */
2355         rctl &= ~E1000_RCTL_SBP;
2356
2357         /* Enable Receives */
2358         E1000_WRITE_REG(hw, E1000_RCTL, rctl);
2359
2360         /*
2361          * Setup the HW Rx Head and Tail Descriptor Pointers
2362          *   - needs to be after enable
2363          */
2364         for (i = 0; i < sc->rx_ring_inuse; ++i) {
2365                 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2366
2367                 E1000_WRITE_REG(hw, E1000_RDH(i), rxr->next_to_check);
2368                 E1000_WRITE_REG(hw, E1000_RDT(i), rxr->num_rx_desc - 1);
2369         }
2370 }
2371
2372 static void
2373 igb_rxeof(struct igb_rx_ring *rxr, int count)
2374 {
2375         struct ifnet *ifp = &rxr->sc->arpcom.ac_if;
2376         union e1000_adv_rx_desc *cur;
2377         uint32_t staterr;
2378         int i;
2379
2380         i = rxr->next_to_check;
2381         cur = &rxr->rx_base[i];
2382         staterr = le32toh(cur->wb.upper.status_error);
2383
2384         if ((staterr & E1000_RXD_STAT_DD) == 0)
2385                 return;
2386
2387         while ((staterr & E1000_RXD_STAT_DD) && count != 0) {
2388                 struct pktinfo *pi = NULL, pi0;
2389                 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i];
2390                 struct mbuf *m = NULL;
2391                 boolean_t eop;
2392
2393                 eop = (staterr & E1000_RXD_STAT_EOP) ? TRUE : FALSE;
2394                 if (eop)
2395                         --count;
2396
2397                 if ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) == 0 &&
2398                     !rxr->discard) {
2399                         struct mbuf *mp = rxbuf->m_head;
2400                         uint32_t hash, hashtype;
2401                         uint16_t vlan;
2402                         int len;
2403
2404                         len = le16toh(cur->wb.upper.length);
2405                         if (rxr->sc->hw.mac.type == e1000_i350 &&
2406                             (staterr & E1000_RXDEXT_STATERR_LB))
2407                                 vlan = be16toh(cur->wb.upper.vlan);
2408                         else
2409                                 vlan = le16toh(cur->wb.upper.vlan);
2410
2411                         hash = le32toh(cur->wb.lower.hi_dword.rss);
2412                         hashtype = le32toh(cur->wb.lower.lo_dword.data) &
2413                             E1000_RXDADV_RSSTYPE_MASK;
2414
2415                         IGB_RSS_DPRINTF(rxr->sc, 10,
2416                             "ring%d, hash 0x%08x, hashtype %u\n",
2417                             rxr->me, hash, hashtype);
2418
2419                         bus_dmamap_sync(rxr->rx_tag, rxbuf->map,
2420                             BUS_DMASYNC_POSTREAD);
2421
2422                         if (igb_newbuf(rxr, i, FALSE) != 0) {
2423                                 ifp->if_iqdrops++;
2424                                 goto discard;
2425                         }
2426
2427                         mp->m_len = len;
2428                         if (rxr->fmp == NULL) {
2429                                 mp->m_pkthdr.len = len;
2430                                 rxr->fmp = mp;
2431                                 rxr->lmp = mp;
2432                         } else {
2433                                 rxr->lmp->m_next = mp;
2434                                 rxr->lmp = rxr->lmp->m_next;
2435                                 rxr->fmp->m_pkthdr.len += len;
2436                         }
2437
2438                         if (eop) {
2439                                 m = rxr->fmp;
2440                                 rxr->fmp = NULL;
2441                                 rxr->lmp = NULL;
2442
2443                                 m->m_pkthdr.rcvif = ifp;
2444                                 ifp->if_ipackets++;
2445
2446                                 if (ifp->if_capenable & IFCAP_RXCSUM)
2447                                         igb_rxcsum(staterr, m);
2448
2449                                 if (staterr & E1000_RXD_STAT_VP) {
2450                                         m->m_pkthdr.ether_vlantag = vlan;
2451                                         m->m_flags |= M_VLANTAG;
2452                                 }
2453
2454                                 if (ifp->if_capenable & IFCAP_RSS) {
2455                                         pi = igb_rssinfo(m, &pi0,
2456                                             hash, hashtype, staterr);
2457                                 }
2458 #ifdef IGB_RSS_DEBUG
2459                                 rxr->rx_packets++;
2460 #endif
2461                         }
2462                 } else {
2463                         ifp->if_ierrors++;
2464 discard:
2465                         igb_setup_rxdesc(cur, rxbuf);
2466                         if (!eop)
2467                                 rxr->discard = TRUE;
2468                         else
2469                                 rxr->discard = FALSE;
2470                         if (rxr->fmp != NULL) {
2471                                 m_freem(rxr->fmp);
2472                                 rxr->fmp = NULL;
2473                                 rxr->lmp = NULL;
2474                         }
2475                         m = NULL;
2476                 }
2477
2478                 if (m != NULL)
2479                         ether_input_pkt(ifp, m, pi);
2480
2481                 /* Advance our pointers to the next descriptor. */
2482                 if (++i == rxr->num_rx_desc)
2483                         i = 0;
2484
2485                 cur = &rxr->rx_base[i];
2486                 staterr = le32toh(cur->wb.upper.status_error);
2487         }
2488         rxr->next_to_check = i;
2489
2490         if (--i < 0)
2491                 i = rxr->num_rx_desc - 1;
2492         E1000_WRITE_REG(&rxr->sc->hw, E1000_RDT(rxr->me), i);
2493 }
2494
2495
2496 static void
2497 igb_set_vlan(struct igb_softc *sc)
2498 {
2499         struct e1000_hw *hw = &sc->hw;
2500         uint32_t reg;
2501 #if 0
2502         struct ifnet *ifp = sc->arpcom.ac_if;
2503 #endif
2504
2505         if (sc->vf_ifp) {
2506                 e1000_rlpml_set_vf(hw, sc->max_frame_size + VLAN_TAG_SIZE);
2507                 return;
2508         }
2509
2510         reg = E1000_READ_REG(hw, E1000_CTRL);
2511         reg |= E1000_CTRL_VME;
2512         E1000_WRITE_REG(hw, E1000_CTRL, reg);
2513
2514 #if 0
2515         /* Enable the Filter Table */
2516         if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) {
2517                 reg = E1000_READ_REG(hw, E1000_RCTL);
2518                 reg &= ~E1000_RCTL_CFIEN;
2519                 reg |= E1000_RCTL_VFE;
2520                 E1000_WRITE_REG(hw, E1000_RCTL, reg);
2521         }
2522 #endif
2523
2524         /* Update the frame size */
2525         E1000_WRITE_REG(&sc->hw, E1000_RLPML,
2526             sc->max_frame_size + VLAN_TAG_SIZE);
2527
2528 #if 0
2529         /* Don't bother with table if no vlans */
2530         if ((adapter->num_vlans == 0) ||
2531             ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0))
2532                 return;
2533         /*
2534         ** A soft reset zero's out the VFTA, so
2535         ** we need to repopulate it now.
2536         */
2537         for (int i = 0; i < IGB_VFTA_SIZE; i++)
2538                 if (adapter->shadow_vfta[i] != 0) {
2539                         if (adapter->vf_ifp)
2540                                 e1000_vfta_set_vf(hw,
2541                                     adapter->shadow_vfta[i], TRUE);
2542                         else
2543                                 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA,
2544                                  i, adapter->shadow_vfta[i]);
2545                 }
2546 #endif
2547 }
2548
2549 static void
2550 igb_enable_intr(struct igb_softc *sc)
2551 {
2552         if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
2553                 lwkt_serialize_handler_enable(&sc->main_serialize);
2554         } else {
2555                 int i;
2556
2557                 for (i = 0; i < sc->msix_cnt; ++i) {
2558                         lwkt_serialize_handler_enable(
2559                             sc->msix_data[i].msix_serialize);
2560                 }
2561         }
2562
2563         if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) {
2564                 if (sc->intr_type == PCI_INTR_TYPE_MSIX)
2565                         E1000_WRITE_REG(&sc->hw, E1000_EIAC, sc->intr_mask);
2566                 else
2567                         E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0);
2568                 E1000_WRITE_REG(&sc->hw, E1000_EIAM, sc->intr_mask);
2569                 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask);
2570                 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC);
2571         } else {
2572                 E1000_WRITE_REG(&sc->hw, E1000_IMS, IMS_ENABLE_MASK);
2573         }
2574         E1000_WRITE_FLUSH(&sc->hw);
2575 }
2576
2577 static void
2578 igb_disable_intr(struct igb_softc *sc)
2579 {
2580         if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) {
2581                 E1000_WRITE_REG(&sc->hw, E1000_EIMC, 0xffffffff);
2582                 E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0);
2583         }
2584         E1000_WRITE_REG(&sc->hw, E1000_IMC, 0xffffffff);
2585         E1000_WRITE_FLUSH(&sc->hw);
2586
2587         if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
2588                 lwkt_serialize_handler_disable(&sc->main_serialize);
2589         } else {
2590                 int i;
2591
2592                 for (i = 0; i < sc->msix_cnt; ++i) {
2593                         lwkt_serialize_handler_disable(
2594                             sc->msix_data[i].msix_serialize);
2595                 }
2596         }
2597 }
2598
2599 /*
2600  * Bit of a misnomer, what this really means is
2601  * to enable OS management of the system... aka
2602  * to disable special hardware management features
2603  */
2604 static void
2605 igb_get_mgmt(struct igb_softc *sc)
2606 {
2607         if (sc->flags & IGB_FLAG_HAS_MGMT) {
2608                 int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H);
2609                 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
2610
2611                 /* disable hardware interception of ARP */
2612                 manc &= ~E1000_MANC_ARP_EN;
2613
2614                 /* enable receiving management packets to the host */
2615                 manc |= E1000_MANC_EN_MNG2HOST;
2616                 manc2h |= 1 << 5; /* Mng Port 623 */
2617                 manc2h |= 1 << 6; /* Mng Port 664 */
2618                 E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h);
2619                 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
2620         }
2621 }
2622
2623 /*
2624  * Give control back to hardware management controller
2625  * if there is one.
2626  */
2627 static void
2628 igb_rel_mgmt(struct igb_softc *sc)
2629 {
2630         if (sc->flags & IGB_FLAG_HAS_MGMT) {
2631                 int manc = E1000_READ_REG(&sc->hw, E1000_MANC);
2632
2633                 /* Re-enable hardware interception of ARP */
2634                 manc |= E1000_MANC_ARP_EN;
2635                 manc &= ~E1000_MANC_EN_MNG2HOST;
2636
2637                 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc);
2638         }
2639 }
2640
2641 /*
2642  * Sets CTRL_EXT:DRV_LOAD bit.
2643  *
2644  * For ASF and Pass Through versions of f/w this means that
2645  * the driver is loaded.
2646  */
2647 static void
2648 igb_get_hw_control(struct igb_softc *sc)
2649 {
2650         uint32_t ctrl_ext;
2651
2652         if (sc->vf_ifp)
2653                 return;
2654
2655         /* Let firmware know the driver has taken over */
2656         ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
2657         E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
2658             ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2659 }
2660
2661 /*
2662  * Resets CTRL_EXT:DRV_LOAD bit.
2663  *
2664  * For ASF and Pass Through versions of f/w this means that the
2665  * driver is no longer loaded.
2666  */
2667 static void
2668 igb_rel_hw_control(struct igb_softc *sc)
2669 {
2670         uint32_t ctrl_ext;
2671
2672         if (sc->vf_ifp)
2673                 return;
2674
2675         /* Let firmware taken over control of h/w */
2676         ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT);
2677         E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT,
2678             ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2679 }
2680
2681 static int
2682 igb_is_valid_ether_addr(const uint8_t *addr)
2683 {
2684         uint8_t zero_addr[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 };
2685
2686         if ((addr[0] & 1) || !bcmp(addr, zero_addr, ETHER_ADDR_LEN))
2687                 return FALSE;
2688         return TRUE;
2689 }
2690
2691 /*
2692  * Enable PCI Wake On Lan capability
2693  */
2694 static void
2695 igb_enable_wol(device_t dev)
2696 {
2697         uint16_t cap, status;
2698         uint8_t id;
2699
2700         /* First find the capabilities pointer*/
2701         cap = pci_read_config(dev, PCIR_CAP_PTR, 2);
2702
2703         /* Read the PM Capabilities */
2704         id = pci_read_config(dev, cap, 1);
2705         if (id != PCIY_PMG)     /* Something wrong */
2706                 return;
2707
2708         /*
2709          * OK, we have the power capabilities,
2710          * so now get the status register
2711          */
2712         cap += PCIR_POWER_STATUS;
2713         status = pci_read_config(dev, cap, 2);
2714         status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
2715         pci_write_config(dev, cap, status, 2);
2716 }
2717
2718 static void
2719 igb_update_stats_counters(struct igb_softc *sc)
2720 {
2721         struct e1000_hw *hw = &sc->hw;
2722         struct e1000_hw_stats *stats;
2723         struct ifnet *ifp = &sc->arpcom.ac_if;
2724
2725         /*
2726          * The virtual function adapter has only a
2727          * small controlled set of stats, do only
2728          * those and return.
2729          */
2730         if (sc->vf_ifp) {
2731                 igb_update_vf_stats_counters(sc);
2732                 return;
2733         }
2734         stats = sc->stats;
2735
2736         if (sc->hw.phy.media_type == e1000_media_type_copper ||
2737             (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
2738                 stats->symerrs +=
2739                     E1000_READ_REG(hw,E1000_SYMERRS);
2740                 stats->sec += E1000_READ_REG(hw, E1000_SEC);
2741         }
2742
2743         stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
2744         stats->mpc += E1000_READ_REG(hw, E1000_MPC);
2745         stats->scc += E1000_READ_REG(hw, E1000_SCC);
2746         stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
2747
2748         stats->mcc += E1000_READ_REG(hw, E1000_MCC);
2749         stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
2750         stats->colc += E1000_READ_REG(hw, E1000_COLC);
2751         stats->dc += E1000_READ_REG(hw, E1000_DC);
2752         stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
2753         stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
2754         stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
2755
2756         /*
2757          * For watchdog management we need to know if we have been
2758          * paused during the last interval, so capture that here.
2759          */
2760         sc->pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
2761         stats->xoffrxc += sc->pause_frames;
2762         stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
2763         stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
2764         stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
2765         stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
2766         stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
2767         stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
2768         stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
2769         stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
2770         stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
2771         stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
2772         stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
2773         stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
2774
2775         /* For the 64-bit byte counters the low dword must be read first. */
2776         /* Both registers clear on the read of the high dword */
2777
2778         stats->gorc += E1000_READ_REG(hw, E1000_GORCL) +
2779             ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
2780         stats->gotc += E1000_READ_REG(hw, E1000_GOTCL) +
2781             ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
2782
2783         stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
2784         stats->ruc += E1000_READ_REG(hw, E1000_RUC);
2785         stats->rfc += E1000_READ_REG(hw, E1000_RFC);
2786         stats->roc += E1000_READ_REG(hw, E1000_ROC);
2787         stats->rjc += E1000_READ_REG(hw, E1000_RJC);
2788
2789         stats->tor += E1000_READ_REG(hw, E1000_TORH);
2790         stats->tot += E1000_READ_REG(hw, E1000_TOTH);
2791
2792         stats->tpr += E1000_READ_REG(hw, E1000_TPR);
2793         stats->tpt += E1000_READ_REG(hw, E1000_TPT);
2794         stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
2795         stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
2796         stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
2797         stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
2798         stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
2799         stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
2800         stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
2801         stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
2802
2803         /* Interrupt Counts */
2804
2805         stats->iac += E1000_READ_REG(hw, E1000_IAC);
2806         stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
2807         stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
2808         stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
2809         stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
2810         stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
2811         stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
2812         stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
2813         stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
2814
2815         /* Host to Card Statistics */
2816
2817         stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC);
2818         stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC);
2819         stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC);
2820         stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC);
2821         stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC);
2822         stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC);
2823         stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC);
2824         stats->hgorc += (E1000_READ_REG(hw, E1000_HGORCL) +
2825             ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32));
2826         stats->hgotc += (E1000_READ_REG(hw, E1000_HGOTCL) +
2827             ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32));
2828         stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS);
2829         stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC);
2830         stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC);
2831
2832         stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
2833         stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
2834         stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
2835         stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
2836         stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
2837         stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
2838
2839         ifp->if_collisions = stats->colc;
2840
2841         /* Rx Errors */
2842         ifp->if_ierrors = stats->rxerrc + stats->crcerrs + stats->algnerrc +
2843             stats->ruc + stats->roc + stats->mpc + stats->cexterr;
2844
2845         /* Tx Errors */
2846         ifp->if_oerrors = stats->ecol + stats->latecol + sc->watchdog_events;
2847
2848         /* Driver specific counters */
2849         sc->device_control = E1000_READ_REG(hw, E1000_CTRL);
2850         sc->rx_control = E1000_READ_REG(hw, E1000_RCTL);
2851         sc->int_mask = E1000_READ_REG(hw, E1000_IMS);
2852         sc->eint_mask = E1000_READ_REG(hw, E1000_EIMS);
2853         sc->packet_buf_alloc_tx =
2854             ((E1000_READ_REG(hw, E1000_PBA) & 0xffff0000) >> 16);
2855         sc->packet_buf_alloc_rx =
2856             (E1000_READ_REG(hw, E1000_PBA) & 0xffff);
2857 }
2858
2859 static void
2860 igb_vf_init_stats(struct igb_softc *sc)
2861 {
2862         struct e1000_hw *hw = &sc->hw;
2863         struct e1000_vf_stats *stats;
2864
2865         stats = sc->stats;
2866         stats->last_gprc = E1000_READ_REG(hw, E1000_VFGPRC);
2867         stats->last_gorc = E1000_READ_REG(hw, E1000_VFGORC);
2868         stats->last_gptc = E1000_READ_REG(hw, E1000_VFGPTC);
2869         stats->last_gotc = E1000_READ_REG(hw, E1000_VFGOTC);
2870         stats->last_mprc = E1000_READ_REG(hw, E1000_VFMPRC);
2871 }
2872
2873 static void
2874 igb_update_vf_stats_counters(struct igb_softc *sc)
2875 {
2876         struct e1000_hw *hw = &sc->hw;
2877         struct e1000_vf_stats *stats;
2878
2879         if (sc->link_speed == 0)
2880                 return;
2881
2882         stats = sc->stats;
2883         UPDATE_VF_REG(E1000_VFGPRC, stats->last_gprc, stats->gprc);
2884         UPDATE_VF_REG(E1000_VFGORC, stats->last_gorc, stats->gorc);
2885         UPDATE_VF_REG(E1000_VFGPTC, stats->last_gptc, stats->gptc);
2886         UPDATE_VF_REG(E1000_VFGOTC, stats->last_gotc, stats->gotc);
2887         UPDATE_VF_REG(E1000_VFMPRC, stats->last_mprc, stats->mprc);
2888 }
2889
2890 #ifdef DEVICE_POLLING
2891
2892 static void
2893 igb_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
2894 {
2895         struct igb_softc *sc = ifp->if_softc;
2896         uint32_t reg_icr;
2897
2898         switch (cmd) {
2899         case POLL_REGISTER:
2900         case POLL_DEREGISTER:
2901                 ASSERT_IFNET_SERIALIZED_ALL(ifp);
2902                 igb_init(sc);
2903                 break;
2904
2905         case POLL_AND_CHECK_STATUS:
2906                 ASSERT_SERIALIZED(&sc->main_serialize);
2907                 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
2908                 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
2909                         sc->hw.mac.get_link_status = 1;
2910                         igb_update_link_status(sc);
2911                 }
2912                 /* FALL THROUGH */
2913         case POLL_ONLY:
2914                 ASSERT_SERIALIZED(&sc->main_serialize);
2915                 if (ifp->if_flags & IFF_RUNNING) {
2916                         struct igb_tx_ring *txr;
2917                         int i;
2918
2919                         for (i = 0; i < sc->rx_ring_inuse; ++i) {
2920                                 struct igb_rx_ring *rxr = &sc->rx_rings[i];
2921
2922                                 lwkt_serialize_enter(&rxr->rx_serialize);
2923                                 igb_rxeof(rxr, count);
2924                                 lwkt_serialize_exit(&rxr->rx_serialize);
2925                         }
2926
2927                         txr = &sc->tx_rings[0];
2928                         lwkt_serialize_enter(&txr->tx_serialize);
2929                         igb_txeof(txr);
2930                         if (!ifq_is_empty(&ifp->if_snd))
2931                                 if_devstart(ifp);
2932                         lwkt_serialize_exit(&txr->tx_serialize);
2933                 }
2934                 break;
2935         }
2936 }
2937
2938 #endif /* DEVICE_POLLING */
2939
2940 static void
2941 igb_intr(void *xsc)
2942 {
2943         struct igb_softc *sc = xsc;
2944         struct ifnet *ifp = &sc->arpcom.ac_if;
2945         uint32_t eicr;
2946
2947         ASSERT_SERIALIZED(&sc->main_serialize);
2948
2949         eicr = E1000_READ_REG(&sc->hw, E1000_EICR);
2950
2951         if (eicr == 0)
2952                 return;
2953
2954         if (ifp->if_flags & IFF_RUNNING) {
2955                 struct igb_tx_ring *txr;
2956                 int i;
2957
2958                 for (i = 0; i < sc->rx_ring_inuse; ++i) {
2959                         struct igb_rx_ring *rxr = &sc->rx_rings[i];
2960
2961                         if (eicr & rxr->rx_intr_mask) {
2962                                 lwkt_serialize_enter(&rxr->rx_serialize);
2963                                 igb_rxeof(rxr, -1);
2964                                 lwkt_serialize_exit(&rxr->rx_serialize);
2965                         }
2966                 }
2967
2968                 txr = &sc->tx_rings[0];
2969                 if (eicr & txr->tx_intr_mask) {
2970                         lwkt_serialize_enter(&txr->tx_serialize);
2971                         igb_txeof(txr);
2972                         if (!ifq_is_empty(&ifp->if_snd))
2973                                 if_devstart(ifp);
2974                         lwkt_serialize_exit(&txr->tx_serialize);
2975                 }
2976         }
2977
2978         if (eicr & E1000_EICR_OTHER) {
2979                 uint32_t icr = E1000_READ_REG(&sc->hw, E1000_ICR);
2980
2981                 /* Link status change */
2982                 if (icr & E1000_ICR_LSC) {
2983                         sc->hw.mac.get_link_status = 1;
2984                         igb_update_link_status(sc);
2985                 }
2986         }
2987
2988         /*
2989          * Reading EICR has the side effect to clear interrupt mask,
2990          * so all interrupts need to be enabled here.
2991          */
2992         E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask);
2993 }
2994
2995 static void
2996 igb_intr_shared(void *xsc)
2997 {
2998         struct igb_softc *sc = xsc;
2999         struct ifnet *ifp = &sc->arpcom.ac_if;
3000         uint32_t reg_icr;
3001
3002         ASSERT_SERIALIZED(&sc->main_serialize);
3003
3004         reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR);
3005
3006         /* Hot eject?  */
3007         if (reg_icr == 0xffffffff)
3008                 return;
3009
3010         /* Definitely not our interrupt.  */
3011         if (reg_icr == 0x0)
3012                 return;
3013
3014         if ((reg_icr & E1000_ICR_INT_ASSERTED) == 0)
3015                 return;
3016
3017         if (ifp->if_flags & IFF_RUNNING) {
3018                 if (reg_icr &
3019                     (E1000_ICR_RXT0 | E1000_ICR_RXDMT0 | E1000_ICR_RXO)) {
3020                         int i;
3021
3022                         for (i = 0; i < sc->rx_ring_inuse; ++i) {
3023                                 struct igb_rx_ring *rxr = &sc->rx_rings[i];
3024
3025                                 lwkt_serialize_enter(&rxr->rx_serialize);
3026                                 igb_rxeof(rxr, -1);
3027                                 lwkt_serialize_exit(&rxr->rx_serialize);
3028                         }
3029                 }
3030
3031                 if (reg_icr & E1000_ICR_TXDW) {
3032                         struct igb_tx_ring *txr = &sc->tx_rings[0];
3033
3034                         lwkt_serialize_enter(&txr->tx_serialize);
3035                         igb_txeof(txr);
3036                         if (!ifq_is_empty(&ifp->if_snd))
3037                                 if_devstart(ifp);
3038                         lwkt_serialize_exit(&txr->tx_serialize);
3039                 }
3040         }
3041
3042         /* Link status change */
3043         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
3044                 sc->hw.mac.get_link_status = 1;
3045                 igb_update_link_status(sc);
3046         }
3047
3048         if (reg_icr & E1000_ICR_RXO)
3049                 sc->rx_overruns++;
3050 }
3051
3052 static int
3053 igb_encap(struct igb_tx_ring *txr, struct mbuf **m_headp)
3054 {
3055         bus_dma_segment_t segs[IGB_MAX_SCATTER];
3056         bus_dmamap_t map;
3057         struct igb_tx_buf *tx_buf, *tx_buf_mapped;
3058         union e1000_adv_tx_desc *txd = NULL;
3059         struct mbuf *m_head = *m_headp;
3060         uint32_t olinfo_status = 0, cmd_type_len = 0, cmd_rs = 0;
3061         int maxsegs, nsegs, i, j, error, last = 0;
3062         uint32_t hdrlen = 0;
3063
3064         if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3065                 error = igb_tso_pullup(txr, m_headp);
3066                 if (error)
3067                         return error;
3068                 m_head = *m_headp;
3069         }
3070
3071         /* Set basic descriptor constants */
3072         cmd_type_len |= E1000_ADVTXD_DTYP_DATA;
3073         cmd_type_len |= E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT;
3074         if (m_head->m_flags & M_VLANTAG)
3075                 cmd_type_len |= E1000_ADVTXD_DCMD_VLE;
3076
3077         /*
3078          * Map the packet for DMA.
3079          */
3080         tx_buf = &txr->tx_buf[txr->next_avail_desc];
3081         tx_buf_mapped = tx_buf;
3082         map = tx_buf->map;
3083
3084         maxsegs = txr->tx_avail - IGB_TX_RESERVED;
3085         KASSERT(maxsegs >= txr->spare_desc, ("not enough spare TX desc\n"));
3086         if (maxsegs > IGB_MAX_SCATTER)
3087                 maxsegs = IGB_MAX_SCATTER;
3088
3089         error = bus_dmamap_load_mbuf_defrag(txr->tx_tag, map, m_headp,
3090             segs, maxsegs, &nsegs, BUS_DMA_NOWAIT);
3091         if (error) {
3092                 if (error == ENOBUFS)
3093                         txr->sc->mbuf_defrag_failed++;
3094                 else
3095                         txr->sc->no_tx_dma_setup++;
3096
3097                 m_freem(*m_headp);
3098                 *m_headp = NULL;
3099                 return error;
3100         }
3101         bus_dmamap_sync(txr->tx_tag, map, BUS_DMASYNC_PREWRITE);
3102
3103         m_head = *m_headp;
3104
3105 #if 0
3106         /*
3107          * Set up the context descriptor:
3108          * used when any hardware offload is done.
3109          * This includes CSUM, VLAN, and TSO. It
3110          * will use the first descriptor.
3111          */
3112         if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3113                 if (igb_tso_setup(txr, m_head, &hdrlen)) {
3114                         cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
3115                         olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
3116                         olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3117                 } else
3118                         return (ENXIO);
3119         } else if (igb_tx_ctx_setup(txr, m_head))
3120                 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3121 #else
3122         if (m_head->m_pkthdr.csum_flags & CSUM_TSO) {
3123                 igb_tso_ctx(txr, m_head, &hdrlen);
3124                 cmd_type_len |= E1000_ADVTXD_DCMD_TSE;
3125                 olinfo_status |= E1000_TXD_POPTS_IXSM << 8;
3126                 olinfo_status |= E1000_TXD_POPTS_TXSM << 8;
3127                 txr->tx_nsegs++;
3128         } else if (igb_txcsum_ctx(txr, m_head)) {
3129                 if (m_head->m_pkthdr.csum_flags & CSUM_IP)
3130                         olinfo_status |= (E1000_TXD_POPTS_IXSM << 8);
3131                 if (m_head->m_pkthdr.csum_flags & (CSUM_UDP | CSUM_TCP))
3132                         olinfo_status |= (E1000_TXD_POPTS_TXSM << 8);
3133                 txr->tx_nsegs++;
3134         }
3135 #endif
3136
3137         txr->tx_nsegs += nsegs;
3138         if (txr->tx_nsegs >= txr->intr_nsegs) {
3139                 /*
3140                  * Report Status (RS) is turned on every intr_nsegs
3141                  * descriptors (roughly).
3142                  */
3143                 txr->tx_nsegs = 0;
3144                 cmd_rs = E1000_ADVTXD_DCMD_RS;
3145         }
3146
3147         /* Calculate payload length */
3148         olinfo_status |= ((m_head->m_pkthdr.len - hdrlen)
3149             << E1000_ADVTXD_PAYLEN_SHIFT);
3150
3151         /* 82575 needs the queue index added */
3152         if (txr->sc->hw.mac.type == e1000_82575)
3153                 olinfo_status |= txr->me << 4;
3154
3155         /* Set up our transmit descriptors */
3156         i = txr->next_avail_desc;
3157         for (j = 0; j < nsegs; j++) {
3158                 bus_size_t seg_len;
3159                 bus_addr_t seg_addr;
3160
3161                 tx_buf = &txr->tx_buf[i];
3162                 txd = (union e1000_adv_tx_desc *)&txr->tx_base[i];
3163                 seg_addr = segs[j].ds_addr;
3164                 seg_len = segs[j].ds_len;
3165
3166                 txd->read.buffer_addr = htole64(seg_addr);
3167                 txd->read.cmd_type_len = htole32(cmd_type_len | seg_len);
3168                 txd->read.olinfo_status = htole32(olinfo_status);
3169                 last = i;
3170                 if (++i == txr->num_tx_desc)
3171                         i = 0;
3172                 tx_buf->m_head = NULL;
3173         }
3174
3175         KASSERT(txr->tx_avail > nsegs, ("invalid avail TX desc\n"));
3176         txr->next_avail_desc = i;
3177         txr->tx_avail -= nsegs;
3178
3179         tx_buf->m_head = m_head;
3180         tx_buf_mapped->map = tx_buf->map;
3181         tx_buf->map = map;
3182
3183         /*
3184          * Last Descriptor of Packet needs End Of Packet (EOP)
3185          */
3186         txd->read.cmd_type_len |= htole32(E1000_ADVTXD_DCMD_EOP | cmd_rs);
3187
3188         /*
3189          * Advance the Transmit Descriptor Tail (TDT), this tells the E1000
3190          * that this frame is available to transmit.
3191          */
3192         E1000_WRITE_REG(&txr->sc->hw, E1000_TDT(txr->me), i);
3193         ++txr->tx_packets;
3194
3195         return 0;
3196 }
3197
3198 static void
3199 igb_start(struct ifnet *ifp)
3200 {
3201         struct igb_softc *sc = ifp->if_softc;
3202         struct igb_tx_ring *txr = &sc->tx_rings[0];
3203         struct mbuf *m_head;
3204
3205         ASSERT_SERIALIZED(&txr->tx_serialize);
3206
3207         if ((ifp->if_flags & (IFF_RUNNING | IFF_OACTIVE)) != IFF_RUNNING)
3208                 return;
3209
3210         if (!sc->link_active) {
3211                 ifq_purge(&ifp->if_snd);
3212                 return;
3213         }
3214
3215         if (!IGB_IS_NOT_OACTIVE(txr))
3216                 igb_txeof(txr);
3217
3218         while (!ifq_is_empty(&ifp->if_snd)) {
3219                 if (IGB_IS_OACTIVE(txr)) {
3220                         ifp->if_flags |= IFF_OACTIVE;
3221                         /* Set watchdog on */
3222                         ifp->if_timer = 5;
3223                         break;
3224                 }
3225
3226                 m_head = ifq_dequeue(&ifp->if_snd, NULL);
3227                 if (m_head == NULL)
3228                         break;
3229
3230                 if (igb_encap(txr, &m_head)) {
3231                         ifp->if_oerrors++;
3232                         continue;
3233                 }
3234
3235                 /* Send a copy of the frame to the BPF listener */
3236                 ETHER_BPF_MTAP(ifp, m_head);
3237         }
3238 }
3239
3240 static void
3241 igb_watchdog(struct ifnet *ifp)
3242 {
3243         struct igb_softc *sc = ifp->if_softc;
3244         struct igb_tx_ring *txr = &sc->tx_rings[0];
3245
3246         ASSERT_IFNET_SERIALIZED_ALL(ifp);
3247
3248         /*
3249          * If flow control has paused us since last checking
3250          * it invalidates the watchdog timing, so dont run it.
3251          */
3252         if (sc->pause_frames) {
3253                 sc->pause_frames = 0;
3254                 ifp->if_timer = 5;
3255                 return;
3256         }
3257
3258         if_printf(ifp, "Watchdog timeout -- resetting\n");
3259         if_printf(ifp, "Queue(%d) tdh = %d, hw tdt = %d\n", txr->me,
3260             E1000_READ_REG(&sc->hw, E1000_TDH(txr->me)),
3261             E1000_READ_REG(&sc->hw, E1000_TDT(txr->me)));
3262         if_printf(ifp, "TX(%d) desc avail = %d, "
3263             "Next TX to Clean = %d\n",
3264             txr->me, txr->tx_avail, txr->next_to_clean);
3265
3266         ifp->if_oerrors++;
3267         sc->watchdog_events++;
3268
3269         igb_init(sc);
3270         if (!ifq_is_empty(&ifp->if_snd))
3271                 if_devstart(ifp);
3272 }
3273
3274 static void
3275 igb_set_eitr(struct igb_softc *sc, int idx, int rate)
3276 {
3277         uint32_t eitr = 0;
3278
3279         if (rate > 0) {
3280                 if (sc->hw.mac.type == e1000_82575) {
3281                         eitr = 1000000000 / 256 / rate;
3282                         /*
3283                          * NOTE:
3284                          * Document is wrong on the 2 bits left shift
3285                          */
3286                 } else {
3287                         eitr = 1000000 / rate;
3288                         eitr <<= IGB_EITR_INTVL_SHIFT;
3289                 }
3290
3291                 if (eitr == 0) {
3292                         /* Don't disable it */
3293                         eitr = 1 << IGB_EITR_INTVL_SHIFT;
3294                 } else if (eitr > IGB_EITR_INTVL_MASK) {
3295                         /* Don't allow it to be too large */
3296                         eitr = IGB_EITR_INTVL_MASK;
3297                 }
3298         }
3299         if (sc->hw.mac.type == e1000_82575)
3300                 eitr |= eitr << 16;
3301         else
3302                 eitr |= E1000_EITR_CNT_IGNR;
3303         E1000_WRITE_REG(&sc->hw, E1000_EITR(idx), eitr);
3304 }
3305
3306 static int
3307 igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS)
3308 {
3309         struct igb_softc *sc = (void *)arg1;
3310         struct ifnet *ifp = &sc->arpcom.ac_if;
3311         int error, intr_rate;
3312
3313         intr_rate = sc->intr_rate;
3314         error = sysctl_handle_int(oidp, &intr_rate, 0, req);
3315         if (error || req->newptr == NULL)
3316                 return error;
3317         if (intr_rate < 0)
3318                 return EINVAL;
3319
3320         ifnet_serialize_all(ifp);
3321
3322         sc->intr_rate = intr_rate;
3323         if (ifp->if_flags & IFF_RUNNING)
3324                 igb_set_eitr(sc, 0, sc->intr_rate);
3325
3326         if (bootverbose)
3327                 if_printf(ifp, "interrupt rate set to %d/sec\n", sc->intr_rate);
3328
3329         ifnet_deserialize_all(ifp);
3330
3331         return 0;
3332 }
3333
3334 static int
3335 igb_sysctl_msix_rate(SYSCTL_HANDLER_ARGS)
3336 {
3337         struct igb_msix_data *msix = (void *)arg1;
3338         struct igb_softc *sc = msix->msix_sc;
3339         struct ifnet *ifp = &sc->arpcom.ac_if;
3340         int error, msix_rate;
3341
3342         msix_rate = msix->msix_rate;
3343         error = sysctl_handle_int(oidp, &msix_rate, 0, req);
3344         if (error || req->newptr == NULL)
3345                 return error;
3346         if (msix_rate < 0)
3347                 return EINVAL;
3348
3349         lwkt_serialize_enter(msix->msix_serialize);
3350
3351         msix->msix_rate = msix_rate;
3352         if (ifp->if_flags & IFF_RUNNING)
3353                 igb_set_eitr(sc, msix->msix_vector, msix->msix_rate);
3354
3355         if (bootverbose) {
3356                 if_printf(ifp, "%s set to %d/sec\n", msix->msix_rate_desc,
3357                     msix->msix_rate);
3358         }
3359
3360         lwkt_serialize_exit(msix->msix_serialize);
3361
3362         return 0;
3363 }
3364
3365 static int
3366 igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS)
3367 {
3368         struct igb_softc *sc = (void *)arg1;
3369         struct ifnet *ifp = &sc->arpcom.ac_if;
3370         struct igb_tx_ring *txr = &sc->tx_rings[0];
3371         int error, nsegs;
3372
3373         nsegs = txr->intr_nsegs;
3374         error = sysctl_handle_int(oidp, &nsegs, 0, req);
3375         if (error || req->newptr == NULL)
3376                 return error;
3377         if (nsegs <= 0)
3378                 return EINVAL;
3379
3380         ifnet_serialize_all(ifp);
3381
3382         if (nsegs >= txr->num_tx_desc - txr->oact_lo_desc ||
3383             nsegs >= txr->oact_hi_desc - IGB_MAX_SCATTER) {
3384                 error = EINVAL;
3385         } else {
3386                 error = 0;
3387                 txr->intr_nsegs = nsegs;
3388         }
3389
3390         ifnet_deserialize_all(ifp);
3391
3392         return error;
3393 }
3394
3395 static void
3396 igb_init_intr(struct igb_softc *sc)
3397 {
3398         igb_set_intr_mask(sc);
3399
3400         if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0)
3401                 igb_init_unshared_intr(sc);
3402
3403         if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
3404                 igb_set_eitr(sc, 0, sc->intr_rate);
3405         } else {
3406                 int i;
3407
3408                 for (i = 0; i < sc->msix_cnt; ++i)
3409                         igb_set_eitr(sc, i, sc->msix_data[i].msix_rate);
3410         }
3411 }
3412
3413 static void
3414 igb_init_unshared_intr(struct igb_softc *sc)
3415 {
3416         struct e1000_hw *hw = &sc->hw;
3417         const struct igb_rx_ring *rxr;
3418         const struct igb_tx_ring *txr;
3419         uint32_t ivar, index;
3420         int i;
3421
3422         /*
3423          * Enable extended mode
3424          */
3425         if (sc->hw.mac.type != e1000_82575) {
3426                 uint32_t gpie;
3427                 int ivar_max;
3428
3429                 gpie = E1000_GPIE_NSICR;
3430                 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3431                         gpie |= E1000_GPIE_MSIX_MODE |
3432                             E1000_GPIE_EIAME |
3433                             E1000_GPIE_PBA;
3434                 }
3435                 E1000_WRITE_REG(hw, E1000_GPIE, gpie);
3436
3437                 /*
3438                  * Clear IVARs
3439                  */
3440                 switch (sc->hw.mac.type) {
3441                 case e1000_82580:
3442                         ivar_max = IGB_MAX_IVAR_82580;
3443                         break;
3444
3445                 case e1000_i350:
3446                         ivar_max = IGB_MAX_IVAR_I350;
3447                         break;
3448
3449                 case e1000_vfadapt:
3450                 case e1000_vfadapt_i350:
3451                         ivar_max = IGB_MAX_IVAR_VF;
3452                         break;
3453
3454                 case e1000_82576:
3455                         ivar_max = IGB_MAX_IVAR_82576;
3456                         break;
3457
3458                 default:
3459                         panic("unknown mac type %d\n", sc->hw.mac.type);
3460                 }
3461                 for (i = 0; i < ivar_max; ++i)
3462                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, i, 0);
3463                 E1000_WRITE_REG(hw, E1000_IVAR_MISC, 0);
3464         } else {
3465                 uint32_t tmp;
3466
3467                 KASSERT(sc->intr_type != PCI_INTR_TYPE_MSIX,
3468                     ("82575 w/ MSI-X"));
3469                 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT);
3470                 tmp |= E1000_CTRL_EXT_IRCA;
3471                 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp);
3472         }
3473
3474         /*
3475          * Map TX/RX interrupts to EICR
3476          */
3477         switch (sc->hw.mac.type) {
3478         case e1000_82580:
3479         case e1000_i350:
3480         case e1000_vfadapt:
3481         case e1000_vfadapt_i350:
3482                 /* RX entries */
3483                 for (i = 0; i < sc->rx_ring_inuse; ++i) {
3484                         rxr = &sc->rx_rings[i];
3485
3486                         index = i >> 1;
3487                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3488
3489                         if (i & 1) {
3490                                 ivar &= 0xff00ffff;
3491                                 ivar |=
3492                                 (rxr->rx_intr_bit | E1000_IVAR_VALID) << 16;
3493                         } else {
3494                                 ivar &= 0xffffff00;
3495                                 ivar |=
3496                                 (rxr->rx_intr_bit | E1000_IVAR_VALID);
3497                         }
3498                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3499                 }
3500                 /* TX entries */
3501                 for (i = 0; i < sc->tx_ring_cnt; ++i) {
3502                         txr = &sc->tx_rings[i];
3503
3504                         index = i >> 1;
3505                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3506
3507                         if (i & 1) {
3508                                 ivar &= 0x00ffffff;
3509                                 ivar |=
3510                                 (txr->tx_intr_bit | E1000_IVAR_VALID) << 24;
3511                         } else {
3512                                 ivar &= 0xffff00ff;
3513                                 ivar |=
3514                                 (txr->tx_intr_bit | E1000_IVAR_VALID) << 8;
3515                         }
3516                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3517                 }
3518                 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3519                         ivar = (sc->sts_intr_bit | E1000_IVAR_VALID) << 8;
3520                         E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
3521                 }
3522                 break;
3523
3524         case e1000_82576:
3525                 /* RX entries */
3526                 for (i = 0; i < sc->rx_ring_inuse; ++i) {
3527                         rxr = &sc->rx_rings[i];
3528
3529                         index = i & 0x7; /* Each IVAR has two entries */
3530                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3531
3532                         if (i < 8) {
3533                                 ivar &= 0xffffff00;
3534                                 ivar |=
3535                                 (rxr->rx_intr_bit | E1000_IVAR_VALID);
3536                         } else {
3537                                 ivar &= 0xff00ffff;
3538                                 ivar |=
3539                                 (rxr->rx_intr_bit | E1000_IVAR_VALID) << 16;
3540                         }
3541                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3542                 }
3543                 /* TX entries */
3544                 for (i = 0; i < sc->tx_ring_cnt; ++i) {
3545                         txr = &sc->tx_rings[i];
3546
3547                         index = i & 0x7; /* Each IVAR has two entries */
3548                         ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index);
3549
3550                         if (i < 8) {
3551                                 ivar &= 0xffff00ff;
3552                                 ivar |=
3553                                 (txr->tx_intr_bit | E1000_IVAR_VALID) << 8;
3554                         } else {
3555                                 ivar &= 0x00ffffff;
3556                                 ivar |=
3557                                 (txr->tx_intr_bit | E1000_IVAR_VALID) << 24;
3558                         }
3559                         E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar);
3560                 }
3561                 if (sc->intr_type == PCI_INTR_TYPE_MSIX) {
3562                         ivar = (sc->sts_intr_bit | E1000_IVAR_VALID) << 8;
3563                         E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar);
3564                 }
3565                 break;
3566
3567         case e1000_82575:
3568                 /*
3569                  * Enable necessary interrupt bits.
3570                  *
3571                  * The name of the register is confusing; in addition to
3572                  * configuring the first vector of MSI-X, it also configures
3573                  * which bits of EICR could be set by the hardware even when
3574                  * MSI or line interrupt is used; it thus controls interrupt
3575                  * generation.  It MUST be configured explicitly; the default
3576                  * value mentioned in the datasheet is wrong: RX queue0 and
3577                  * TX queue0 are NOT enabled by default.
3578                  */
3579                 E1000_WRITE_REG(&sc->hw, E1000_MSIXBM(0), sc->intr_mask);
3580                 break;
3581
3582         default:
3583                 panic("unknown mac type %d\n", sc->hw.mac.type);
3584         }
3585 }
3586
3587 static int
3588 igb_setup_intr(struct igb_softc *sc)
3589 {
3590         struct ifnet *ifp = &sc->arpcom.ac_if;
3591         int error;
3592
3593         if (sc->intr_type == PCI_INTR_TYPE_MSIX)
3594                 return igb_msix_setup(sc);
3595
3596         error = bus_setup_intr(sc->dev, sc->intr_res, INTR_MPSAFE,
3597             (sc->flags & IGB_FLAG_SHARED_INTR) ? igb_intr_shared : igb_intr,
3598             sc, &sc->intr_tag, &sc->main_serialize);
3599         if (error) {
3600                 device_printf(sc->dev, "Failed to register interrupt handler");
3601                 return error;
3602         }
3603
3604         ifp->if_cpuid = rman_get_cpuid(sc->intr_res);
3605         KKASSERT(ifp->if_cpuid >= 0 && ifp->if_cpuid < ncpus);
3606
3607         return 0;
3608 }
3609
3610 static void
3611 igb_set_txintr_mask(struct igb_tx_ring *txr, int *intr_bit0, int intr_bitmax)
3612 {
3613         if (txr->sc->hw.mac.type == e1000_82575) {
3614                 txr->tx_intr_bit = 0;   /* unused */
3615                 switch (txr->me) {
3616                 case 0:
3617                         txr->tx_intr_mask = E1000_EICR_TX_QUEUE0;
3618                         break;
3619                 case 1:
3620                         txr->tx_intr_mask = E1000_EICR_TX_QUEUE1;
3621                         break;
3622                 case 2:
3623                         txr->tx_intr_mask = E1000_EICR_TX_QUEUE2;
3624                         break;
3625                 case 3:
3626                         txr->tx_intr_mask = E1000_EICR_TX_QUEUE3;
3627                         break;
3628                 default:
3629                         panic("unsupported # of TX ring, %d\n", txr->me);
3630                 }
3631         } else {
3632                 int intr_bit = *intr_bit0;
3633
3634                 txr->tx_intr_bit = intr_bit % intr_bitmax;
3635                 txr->tx_intr_mask = 1 << txr->tx_intr_bit;
3636
3637                 *intr_bit0 = intr_bit + 1;
3638         }
3639 }
3640
3641 static void
3642 igb_set_rxintr_mask(struct igb_rx_ring *rxr, int *intr_bit0, int intr_bitmax)
3643 {
3644         if (rxr->sc->hw.mac.type == e1000_82575) {
3645                 rxr->rx_intr_bit = 0;   /* unused */
3646                 switch (rxr->me) {
3647                 case 0:
3648                         rxr->rx_intr_mask = E1000_EICR_RX_QUEUE0;
3649                         break;
3650                 case 1:
3651                         rxr->rx_intr_mask = E1000_EICR_RX_QUEUE1;
3652                         break;
3653                 case 2:
3654                         rxr->rx_intr_mask = E1000_EICR_RX_QUEUE2;
3655                         break;
3656                 case 3:
3657                         rxr->rx_intr_mask = E1000_EICR_RX_QUEUE3;
3658                         break;
3659                 default:
3660                         panic("unsupported # of RX ring, %d\n", rxr->me);
3661                 }
3662         } else {
3663                 int intr_bit = *intr_bit0;
3664
3665                 rxr->rx_intr_bit = intr_bit % intr_bitmax;
3666                 rxr->rx_intr_mask = 1 << rxr->rx_intr_bit;
3667
3668                 *intr_bit0 = intr_bit + 1;
3669         }
3670 }
3671
3672 static void
3673 igb_serialize(struct ifnet *ifp, enum ifnet_serialize slz)
3674 {
3675         struct igb_softc *sc = ifp->if_softc;
3676
3677         ifnet_serialize_array_enter(sc->serializes, sc->serialize_cnt,
3678             sc->tx_serialize, sc->rx_serialize, slz);
3679 }
3680
3681 static void
3682 igb_deserialize(struct ifnet *ifp, enum ifnet_serialize slz)
3683 {
3684         struct igb_softc *sc = ifp->if_softc;
3685
3686         ifnet_serialize_array_exit(sc->serializes, sc->serialize_cnt,
3687             sc->tx_serialize, sc->rx_serialize, slz);
3688 }
3689
3690 static int
3691 igb_tryserialize(struct ifnet *ifp, enum ifnet_serialize slz)
3692 {
3693         struct igb_softc *sc = ifp->if_softc;
3694
3695         return ifnet_serialize_array_try(sc->serializes, sc->serialize_cnt,
3696             sc->tx_serialize, sc->rx_serialize, slz);
3697 }
3698
3699 #ifdef INVARIANTS
3700
3701 static void
3702 igb_serialize_assert(struct ifnet *ifp, enum ifnet_serialize slz,
3703     boolean_t serialized)
3704 {
3705         struct igb_softc *sc = ifp->if_softc;
3706
3707         ifnet_serialize_array_assert(sc->serializes, sc->serialize_cnt,
3708             sc->tx_serialize, sc->rx_serialize, slz, serialized);
3709 }
3710
3711 #endif  /* INVARIANTS */
3712
3713 static void
3714 igb_set_intr_mask(struct igb_softc *sc)
3715 {
3716         int i;
3717
3718         sc->intr_mask = sc->sts_intr_mask;
3719         for (i = 0; i < sc->rx_ring_inuse; ++i)
3720                 sc->intr_mask |= sc->rx_rings[i].rx_intr_mask;
3721         for (i = 0; i < sc->tx_ring_cnt; ++i)
3722                 sc->intr_mask |= sc->tx_rings[i].tx_intr_mask;
3723         if (bootverbose)
3724                 device_printf(sc->dev, "intr mask 0x%08x\n", sc->intr_mask);
3725 }
3726
3727 static int
3728 igb_alloc_intr(struct igb_softc *sc)
3729 {
3730         int i, intr_bit, intr_bitmax;
3731         u_int intr_flags;
3732
3733         igb_msix_try_alloc(sc);
3734         if (sc->intr_type == PCI_INTR_TYPE_MSIX)
3735                 goto done;
3736
3737         /*
3738          * Allocate MSI/legacy interrupt resource
3739          */
3740         sc->intr_type = pci_alloc_1intr(sc->dev, igb_msi_enable,
3741             &sc->intr_rid, &intr_flags);
3742
3743         if (sc->intr_type == PCI_INTR_TYPE_LEGACY) {
3744                 int unshared;
3745
3746                 unshared = device_getenv_int(sc->dev, "irq.unshared", 0);
3747                 if (!unshared) {
3748                         sc->flags |= IGB_FLAG_SHARED_INTR;
3749                         if (bootverbose)
3750                                 device_printf(sc->dev, "IRQ shared\n");
3751                 } else {
3752                         intr_flags &= ~RF_SHAREABLE;
3753                         if (bootverbose)
3754                                 device_printf(sc->dev, "IRQ unshared\n");
3755                 }
3756         }
3757
3758         sc->intr_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ,
3759             &sc->intr_rid, intr_flags);
3760         if (sc->intr_res == NULL) {
3761                 device_printf(sc->dev, "Unable to allocate bus resource: "
3762                     "interrupt\n");
3763                 return ENXIO;
3764         }
3765
3766         /*
3767          * Setup MSI/legacy interrupt mask
3768          */
3769         switch (sc->hw.mac.type) {
3770         case e1000_82575:
3771                 intr_bitmax = IGB_MAX_TXRXINT_82575;
3772                 break;
3773         case e1000_82580:
3774                 intr_bitmax = IGB_MAX_TXRXINT_82580;
3775                 break;
3776         case e1000_i350:
3777                 intr_bitmax = IGB_MAX_TXRXINT_I350;
3778                 break;
3779         case e1000_82576:
3780                 intr_bitmax = IGB_MAX_TXRXINT_82576;
3781                 break;
3782         default:
3783                 intr_bitmax = IGB_MIN_TXRXINT;
3784                 break;
3785         }
3786         intr_bit = 0;
3787         for (i = 0; i < sc->tx_ring_cnt; ++i)
3788                 igb_set_txintr_mask(&sc->tx_rings[i], &intr_bit, intr_bitmax);
3789         for (i = 0; i < sc->rx_ring_cnt; ++i)
3790                 igb_set_rxintr_mask(&sc->rx_rings[i], &intr_bit, intr_bitmax);
3791         sc->sts_intr_bit = 0;
3792         sc->sts_intr_mask = E1000_EICR_OTHER;
3793
3794         /* Initialize interrupt rate */
3795         sc->intr_rate = IGB_INTR_RATE;
3796 done:
3797         igb_set_ring_inuse(sc, FALSE);
3798         igb_set_intr_mask(sc);
3799         return 0;
3800 }
3801
3802 static void
3803 igb_free_intr(struct igb_softc *sc)
3804 {
3805         if (sc->intr_type != PCI_INTR_TYPE_MSIX) {
3806                 if (sc->intr_res != NULL) {
3807                         bus_release_resource(sc->dev, SYS_RES_IRQ, sc->intr_rid,
3808                             sc->intr_res);
3809                 }
3810                 if (sc->intr_type == PCI_INTR_TYPE_MSI)
3811                         pci_release_msi(sc->dev);
3812         } else {
3813                 igb_msix_free(sc, TRUE);
3814         }
3815 }
3816
3817 static void
3818 igb_teardown_intr(struct igb_softc *sc)
3819 {
3820         if (sc->intr_type != PCI_INTR_TYPE_MSIX)
3821                 bus_teardown_intr(sc->dev, sc->intr_res, sc->intr_tag);
3822         else
3823                 igb_msix_teardown(sc, sc->msix_cnt);
3824 }
3825
3826 static void
3827 igb_msix_try_alloc(struct igb_softc *sc)
3828 {
3829         int msix_enable, msix_cnt, msix_cnt2, alloc_cnt;
3830         int i, x, error;
3831         struct igb_msix_data *msix;
3832         boolean_t aggregate, setup = FALSE;
3833
3834         /*
3835          * Don't enable MSI-X on 82575, see:
3836          * 82575 specification update errata #25
3837          */
3838         if (sc->hw.mac.type == e1000_82575)
3839                 return;
3840
3841         /* Don't enable MSI-X on VF */
3842         if (sc->vf_ifp)
3843                 return;
3844
3845         msix_enable = device_getenv_int(sc->dev, "msix.enable",
3846             igb_msix_enable);
3847         if (!msix_enable)
3848                 return;
3849
3850         msix_cnt = pci_msix_count(sc->dev);
3851 #ifdef IGB_MSIX_DEBUG
3852         msix_cnt = device_getenv_int(sc->dev, "msix.count", msix_cnt);
3853 #endif
3854         if (msix_cnt <= 1) {
3855                 /* One MSI-X model does not make sense */
3856                 return;
3857         }
3858
3859         i = 0;
3860         while ((1 << (i + 1)) <= msix_cnt)
3861                 ++i;
3862         msix_cnt2 = 1 << i;
3863
3864         if (bootverbose) {
3865                 device_printf(sc->dev, "MSI-X count %d/%d\n",
3866                     msix_cnt2, msix_cnt);
3867         }
3868
3869         KKASSERT(msix_cnt2 <= msix_cnt);
3870         if (msix_cnt == msix_cnt2) {
3871                 /* We need at least one MSI-X for link status */
3872                 msix_cnt2 >>= 1;
3873                 if (msix_cnt2 <= 1) {
3874                         /* One MSI-X for RX/TX does not make sense */
3875                         device_printf(sc->dev, "not enough MSI-X for TX/RX, "
3876                             "MSI-X count %d/%d\n", msix_cnt2, msix_cnt);
3877                         return;
3878                 }
3879                 KKASSERT(msix_cnt > msix_cnt2);
3880
3881                 if (bootverbose) {
3882                         device_printf(sc->dev, "MSI-X count fixup %d/%d\n",
3883                             msix_cnt2, msix_cnt);
3884                 }
3885         }
3886
3887         sc->rx_ring_msix = sc->rx_ring_cnt;
3888         if (sc->rx_ring_msix > msix_cnt2)
3889                 sc->rx_ring_msix = msix_cnt2;
3890
3891         if (msix_cnt >= sc->tx_ring_cnt + sc->rx_ring_msix + 1) {
3892                 /*
3893                  * Independent TX/RX MSI-X
3894                  */
3895                 aggregate = FALSE;
3896                 if (bootverbose)
3897                         device_printf(sc->dev, "independent TX/RX MSI-X\n");
3898                 alloc_cnt = sc->tx_ring_cnt + sc->rx_ring_msix;
3899         } else {
3900                 /*
3901                  * Aggregate TX/RX MSI-X
3902                  */
3903                 aggregate = TRUE;
3904                 if (bootverbose)
3905                         device_printf(sc->dev, "aggregate TX/RX MSI-X\n");
3906                 alloc_cnt = msix_cnt2;
3907                 if (alloc_cnt > ncpus2)
3908                         alloc_cnt = ncpus2;
3909                 if (sc->rx_ring_msix > alloc_cnt)
3910                         sc->rx_ring_msix = alloc_cnt;
3911         }
3912         ++alloc_cnt;    /* For link status */
3913
3914         if (bootverbose) {
3915                 device_printf(sc->dev, "MSI-X alloc %d, RX ring %d\n",
3916                     alloc_cnt, sc->rx_ring_msix);
3917         }
3918
3919         sc->msix_mem_rid = PCIR_BAR(IGB_MSIX_BAR);
3920         sc->msix_mem_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY,
3921             &sc->msix_mem_rid, RF_ACTIVE);
3922         if (sc->msix_mem_res == NULL) {
3923                 device_printf(sc->dev, "Unable to map MSI-X table\n");
3924                 return;
3925         }
3926
3927         sc->msix_cnt = alloc_cnt;
3928         sc->msix_data = kmalloc(sizeof(struct igb_msix_data) * sc->msix_cnt,
3929             M_DEVBUF, M_WAITOK | M_ZERO);
3930         for (x = 0; x < sc->msix_cnt; ++x) {
3931                 msix = &sc->msix_data[x];
3932
3933                 lwkt_serialize_init(&msix->msix_serialize0);
3934                 msix->msix_sc = sc;
3935                 msix->msix_rid = -1;
3936                 msix->msix_vector = x;
3937                 msix->msix_mask = 1 << msix->msix_vector;
3938                 msix->msix_rate = IGB_INTR_RATE;
3939         }
3940
3941         x = 0;
3942         if (!aggregate) {
3943                 int offset, offset_def;
3944
3945                 if (sc->rx_ring_msix == ncpus2) {
3946                         offset = 0;
3947                 } else {
3948                         offset_def = (sc->rx_ring_msix *
3949                             device_get_unit(sc->dev)) % ncpus2;
3950
3951                         offset = device_getenv_int(sc->dev,
3952                             "msix.rxoff", offset_def);
3953                         if (offset >= ncpus2 ||
3954                             offset % sc->rx_ring_msix != 0) {
3955                                 device_printf(sc->dev,
3956                                     "invalid msix.rxoff %d, use %d\n",
3957                                     offset, offset_def);
3958                                 offset = offset_def;
3959                         }
3960                 }
3961
3962                 /* RX rings */
3963                 for (i = 0; i < sc->rx_ring_msix; ++i) {
3964                         struct igb_rx_ring *rxr = &sc->rx_rings[i];
3965
3966                         KKASSERT(x < sc->msix_cnt);
3967                         msix = &sc->msix_data[x++];
3968                         rxr->rx_intr_bit = msix->msix_vector;
3969                         rxr->rx_intr_mask = msix->msix_mask;
3970
3971                         msix->msix_serialize = &rxr->rx_serialize;
3972                         msix->msix_func = igb_msix_rx;
3973                         msix->msix_arg = rxr;
3974                         msix->msix_cpuid = i + offset;
3975                         KKASSERT(msix->msix_cpuid < ncpus2);
3976                         ksnprintf(msix->msix_desc, sizeof(msix->msix_desc),
3977                             "%s rx%d", device_get_nameunit(sc->dev), i);
3978                         msix->msix_rate = IGB_MSIX_RX_RATE;
3979                         ksnprintf(msix->msix_rate_desc,
3980                             sizeof(msix->msix_rate_desc),
3981                             "RX%d interrupt rate", i);
3982                 }
3983
3984                 offset_def = device_get_unit(sc->dev) % ncpus2;
3985                 offset = device_getenv_int(sc->dev, "msix.txoff", offset_def);
3986                 if (offset >= ncpus2) {
3987                         device_printf(sc->dev, "invalid msix.txoff %d, "
3988                             "use %d\n", offset, offset_def);
3989                         offset = offset_def;
3990                 }
3991
3992                 /* TX rings */
3993                 for (i = 0; i < sc->tx_ring_cnt; ++i) {
3994                         struct igb_tx_ring *txr = &sc->tx_rings[i];
3995
3996                         KKASSERT(x < sc->msix_cnt);
3997                         msix = &sc->msix_data[x++];
3998                         txr->tx_intr_bit = msix->msix_vector;
3999                         txr->tx_intr_mask = msix->msix_mask;
4000
4001                         msix->msix_serialize = &txr->tx_serialize;
4002                         msix->msix_func = igb_msix_tx;
4003                         msix->msix_arg = txr;
4004                         msix->msix_cpuid = i + offset;
4005                         sc->msix_tx_cpuid = msix->msix_cpuid; /* XXX */
4006                         KKASSERT(msix->msix_cpuid < ncpus2);
4007                         ksnprintf(msix->msix_desc, sizeof(msix->msix_desc),
4008                             "%s tx%d", device_get_nameunit(sc->dev), i);
4009                         msix->msix_rate = IGB_MSIX_TX_RATE;
4010                         ksnprintf(msix->msix_rate_desc,
4011                             sizeof(msix->msix_rate_desc),
4012                             "TX%d interrupt rate", i);
4013                 }
4014         } else {
4015                 /* TODO */
4016                 error = EOPNOTSUPP;
4017                 goto back;
4018         }
4019
4020         /*
4021          * Link status
4022          */
4023         KKASSERT(x < sc->msix_cnt);
4024         msix = &sc->msix_data[x++];
4025         sc->sts_intr_bit = msix->msix_vector;
4026         sc->sts_intr_mask = msix->msix_mask;
4027
4028         msix->msix_serialize = &sc->main_serialize;
4029         msix->msix_func = igb_msix_status;
4030         msix->msix_arg = sc;
4031         msix->msix_cpuid = 0; /* TODO tunable */
4032         ksnprintf(msix->msix_desc, sizeof(msix->msix_desc), "%s sts",
4033             device_get_nameunit(sc->dev));
4034         ksnprintf(msix->msix_rate_desc, sizeof(msix->msix_rate_desc),
4035             "status interrupt rate");
4036
4037         KKASSERT(x == sc->msix_cnt);
4038
4039         error = pci_setup_msix(sc->dev);
4040         if (error) {
4041                 device_printf(sc->dev, "Setup MSI-X failed\n");
4042                 goto back;
4043         }
4044         setup = TRUE;
4045
4046         for (i = 0; i < sc->msix_cnt; ++i) {
4047                 msix = &sc->msix_data[i];
4048
4049                 error = pci_alloc_msix_vector(sc->dev, msix->msix_vector,
4050                     &msix->msix_rid, msix->msix_cpuid);
4051                 if (error) {
4052                         device_printf(sc->dev,
4053                             "Unable to allocate MSI-X %d on cpu%d\n",
4054                             msix->msix_vector, msix->msix_cpuid);
4055                         goto back;
4056                 }
4057
4058                 msix->msix_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ,
4059                     &msix->msix_rid, RF_ACTIVE);
4060                 if (msix->msix_res == NULL) {
4061                         device_printf(sc->dev,
4062                             "Unable to allocate MSI-X %d resource\n",
4063                             msix->msix_vector);
4064                         error = ENOMEM;
4065                         goto back;
4066                 }
4067         }
4068
4069         pci_enable_msix(sc->dev);
4070         sc->intr_type = PCI_INTR_TYPE_MSIX;
4071 back:
4072         if (error)
4073                 igb_msix_free(sc, setup);
4074 }
4075
4076 static void
4077 igb_msix_free(struct igb_softc *sc, boolean_t setup)
4078 {
4079         int i;
4080
4081         KKASSERT(sc->msix_cnt > 1);
4082
4083         for (i = 0; i < sc->msix_cnt; ++i) {
4084                 struct igb_msix_data *msix = &sc->msix_data[i];
4085
4086                 if (msix->msix_res != NULL) {
4087                         bus_release_resource(sc->dev, SYS_RES_IRQ,
4088                             msix->msix_rid, msix->msix_res);
4089                 }
4090                 if (msix->msix_rid >= 0)
4091                         pci_release_msix_vector(sc->dev, msix->msix_rid);
4092         }
4093         if (setup)
4094                 pci_teardown_msix(sc->dev);
4095
4096         sc->msix_cnt = 0;
4097         kfree(sc->msix_data, M_DEVBUF);
4098         sc->msix_data = NULL;
4099 }
4100
4101 static int
4102 igb_msix_setup(struct igb_softc *sc)
4103 {
4104         struct ifnet *ifp = &sc->arpcom.ac_if;
4105         int i;
4106
4107         for (i = 0; i < sc->msix_cnt; ++i) {
4108                 struct igb_msix_data *msix = &sc->msix_data[i];
4109                 int error;
4110
4111                 error = bus_setup_intr_descr(sc->dev, msix->msix_res,
4112                     INTR_MPSAFE, msix->msix_func, msix->msix_arg,
4113                     &msix->msix_handle, msix->msix_serialize, msix->msix_desc);
4114                 if (error) {
4115                         device_printf(sc->dev, "could not set up %s "
4116                             "interrupt handler.\n", msix->msix_desc);
4117                         igb_msix_teardown(sc, i);
4118                         return error;
4119                 }
4120         }
4121         ifp->if_cpuid = sc->msix_tx_cpuid;
4122
4123         return 0;
4124 }
4125
4126 static void
4127 igb_msix_teardown(struct igb_softc *sc, int msix_cnt)
4128 {
4129         int i;
4130
4131         for (i = 0; i < msix_cnt; ++i) {
4132                 struct igb_msix_data *msix = &sc->msix_data[i];
4133
4134                 bus_teardown_intr(sc->dev, msix->msix_res, msix->msix_handle);
4135         }
4136 }
4137
4138 static void
4139 igb_msix_rx(void *arg)
4140 {
4141         struct igb_rx_ring *rxr = arg;
4142
4143         ASSERT_SERIALIZED(&rxr->rx_serialize);
4144         igb_rxeof(rxr, -1);
4145
4146         E1000_WRITE_REG(&rxr->sc->hw, E1000_EIMS, rxr->rx_intr_mask);
4147 }
4148
4149 static void
4150 igb_msix_tx(void *arg)
4151 {
4152         struct igb_tx_ring *txr = arg;
4153         struct ifnet *ifp = &txr->sc->arpcom.ac_if;
4154
4155         ASSERT_SERIALIZED(&txr->tx_serialize);
4156
4157         igb_txeof(txr);
4158         if (!ifq_is_empty(&ifp->if_snd))
4159                 if_devstart(ifp);
4160
4161         E1000_WRITE_REG(&txr->sc->hw, E1000_EIMS, txr->tx_intr_mask);
4162 }
4163
4164 static void
4165 igb_msix_status(void *arg)
4166 {
4167         struct igb_softc *sc = arg;
4168         uint32_t icr;
4169
4170         ASSERT_SERIALIZED(&sc->main_serialize);
4171
4172         icr = E1000_READ_REG(&sc->hw, E1000_ICR);
4173         if (icr & E1000_ICR_LSC) {
4174                 sc->hw.mac.get_link_status = 1;
4175                 igb_update_link_status(sc);
4176         }
4177
4178         E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->sts_intr_mask);
4179 }
4180
4181 static void
4182 igb_set_ring_inuse(struct igb_softc *sc, boolean_t polling)
4183 {
4184         if (!IGB_ENABLE_HWRSS(sc))
4185                 return;
4186
4187         if (sc->intr_type != PCI_INTR_TYPE_MSIX || polling)
4188                 sc->rx_ring_inuse = IGB_MIN_RING_RSS;
4189         else
4190                 sc->rx_ring_inuse = sc->rx_ring_msix;
4191         if (bootverbose) {
4192                 device_printf(sc->dev, "RX rings %d/%d\n",
4193                     sc->rx_ring_inuse, sc->rx_ring_cnt);
4194         }
4195 }
4196
4197 static int
4198 igb_tso_pullup(struct igb_tx_ring *txr, struct mbuf **mp)
4199 {
4200         int hoff, iphlen, thoff;
4201         struct mbuf *m;
4202
4203         m = *mp;
4204         KASSERT(M_WRITABLE(m), ("TSO mbuf not writable"));
4205
4206         iphlen = m->m_pkthdr.csum_iphlen;
4207         thoff = m->m_pkthdr.csum_thlen;
4208         hoff = m->m_pkthdr.csum_lhlen;
4209
4210         KASSERT(iphlen > 0, ("invalid ip hlen"));
4211         KASSERT(thoff > 0, ("invalid tcp hlen"));
4212         KASSERT(hoff > 0, ("invalid ether hlen"));
4213
4214         if (__predict_false(m->m_len < hoff + iphlen + thoff)) {
4215                 m = m_pullup(m, hoff + iphlen + thoff);
4216                 if (m == NULL) {
4217                         *mp = NULL;
4218                         return ENOBUFS;
4219                 }
4220                 *mp = m;
4221         }
4222         if (txr->sc->flags & IGB_FLAG_TSO_IPLEN0) {
4223                 struct ip *ip;
4224
4225                 ip = mtodoff(m, struct ip *, hoff);
4226                 ip->ip_len = 0;
4227         }
4228
4229         return 0;
4230 }
4231
4232 static void
4233 igb_tso_ctx(struct igb_tx_ring *txr, struct mbuf *m, uint32_t *hlen)
4234 {
4235         struct e1000_adv_tx_context_desc *TXD;
4236         uint32_t vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx;
4237         int hoff, ctxd, iphlen, thoff;
4238
4239         iphlen = m->m_pkthdr.csum_iphlen;
4240         thoff = m->m_pkthdr.csum_thlen;
4241         hoff = m->m_pkthdr.csum_lhlen;
4242
4243         vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0;
4244
4245         ctxd = txr->next_avail_desc;
4246         TXD = (struct e1000_adv_tx_context_desc *)&txr->tx_base[ctxd];
4247
4248         if (m->m_flags & M_VLANTAG) {
4249                 uint16_t vlantag;
4250
4251                 vlantag = htole16(m->m_pkthdr.ether_vlantag);
4252                 vlan_macip_lens |= (vlantag << E1000_ADVTXD_VLAN_SHIFT);
4253         }
4254
4255         vlan_macip_lens |= (hoff << E1000_ADVTXD_MACLEN_SHIFT);
4256         vlan_macip_lens |= iphlen;
4257
4258         type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
4259         type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP;
4260         type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4;
4261
4262         mss_l4len_idx |= (m->m_pkthdr.tso_segsz << E1000_ADVTXD_MSS_SHIFT);
4263         mss_l4len_idx |= (thoff << E1000_ADVTXD_L4LEN_SHIFT);
4264         /* 82575 needs the queue index added */
4265         if (txr->sc->hw.mac.type == e1000_82575)
4266                 mss_l4len_idx |= txr->me << 4;
4267
4268         TXD->vlan_macip_lens = htole32(vlan_macip_lens);
4269         TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl);
4270         TXD->seqnum_seed = htole32(0);
4271         TXD->mss_l4len_idx = htole32(mss_l4len_idx);
4272
4273         /* We've consumed the first desc, adjust counters */
4274         if (++ctxd == txr->num_tx_desc)
4275                 ctxd = 0;
4276         txr->next_avail_desc = ctxd;
4277         --txr->tx_avail;
4278
4279         *hlen = hoff + iphlen + thoff;
4280 }