1 /* 2 * Copyright (c) 2001-2013, 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_ifpoll.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 68 #include <bus/pci/pcivar.h> 69 #include <bus/pci/pcireg.h> 70 71 #include <dev/netif/ig_hal/e1000_api.h> 72 #include <dev/netif/ig_hal/e1000_82575.h> 73 #include <dev/netif/ig_hal/e1000_dragonfly.h> 74 #include <dev/netif/igb/if_igb.h> 75 76 #ifdef IGB_RSS_DEBUG 77 #define IGB_RSS_DPRINTF(sc, lvl, fmt, ...) \ 78 do { \ 79 if (sc->rss_debug >= lvl) \ 80 if_printf(&sc->arpcom.ac_if, fmt, __VA_ARGS__); \ 81 } while (0) 82 #else /* !IGB_RSS_DEBUG */ 83 #define IGB_RSS_DPRINTF(sc, lvl, fmt, ...) ((void)0) 84 #endif /* IGB_RSS_DEBUG */ 85 86 #define IGB_NAME "Intel(R) PRO/1000 " 87 #define IGB_DEVICE(id) \ 88 { IGB_VENDOR_ID, E1000_DEV_ID_##id, IGB_NAME #id } 89 #define IGB_DEVICE_NULL { 0, 0, NULL } 90 91 static struct igb_device { 92 uint16_t vid; 93 uint16_t did; 94 const char *desc; 95 } igb_devices[] = { 96 IGB_DEVICE(82575EB_COPPER), 97 IGB_DEVICE(82575EB_FIBER_SERDES), 98 IGB_DEVICE(82575GB_QUAD_COPPER), 99 IGB_DEVICE(82576), 100 IGB_DEVICE(82576_NS), 101 IGB_DEVICE(82576_NS_SERDES), 102 IGB_DEVICE(82576_FIBER), 103 IGB_DEVICE(82576_SERDES), 104 IGB_DEVICE(82576_SERDES_QUAD), 105 IGB_DEVICE(82576_QUAD_COPPER), 106 IGB_DEVICE(82576_QUAD_COPPER_ET2), 107 IGB_DEVICE(82576_VF), 108 IGB_DEVICE(82580_COPPER), 109 IGB_DEVICE(82580_FIBER), 110 IGB_DEVICE(82580_SERDES), 111 IGB_DEVICE(82580_SGMII), 112 IGB_DEVICE(82580_COPPER_DUAL), 113 IGB_DEVICE(82580_QUAD_FIBER), 114 IGB_DEVICE(DH89XXCC_SERDES), 115 IGB_DEVICE(DH89XXCC_SGMII), 116 IGB_DEVICE(DH89XXCC_SFP), 117 IGB_DEVICE(DH89XXCC_BACKPLANE), 118 IGB_DEVICE(I350_COPPER), 119 IGB_DEVICE(I350_FIBER), 120 IGB_DEVICE(I350_SERDES), 121 IGB_DEVICE(I350_SGMII), 122 IGB_DEVICE(I350_VF), 123 IGB_DEVICE(I210_COPPER), 124 IGB_DEVICE(I210_COPPER_IT), 125 IGB_DEVICE(I210_COPPER_OEM1), 126 IGB_DEVICE(I210_COPPER_FLASHLESS), 127 IGB_DEVICE(I210_SERDES_FLASHLESS), 128 IGB_DEVICE(I210_FIBER), 129 IGB_DEVICE(I210_SERDES), 130 IGB_DEVICE(I210_SGMII), 131 IGB_DEVICE(I211_COPPER), 132 IGB_DEVICE(I354_BACKPLANE_1GBPS), 133 IGB_DEVICE(I354_BACKPLANE_2_5GBPS), 134 IGB_DEVICE(I354_SGMII), 135 136 /* required last entry */ 137 IGB_DEVICE_NULL 138 }; 139 140 static int igb_probe(device_t); 141 static int igb_attach(device_t); 142 static int igb_detach(device_t); 143 static int igb_shutdown(device_t); 144 static int igb_suspend(device_t); 145 static int igb_resume(device_t); 146 147 static boolean_t igb_is_valid_ether_addr(const uint8_t *); 148 static void igb_setup_ifp(struct igb_softc *); 149 static boolean_t igb_txcsum_ctx(struct igb_tx_ring *, struct mbuf *); 150 static int igb_tso_pullup(struct igb_tx_ring *, struct mbuf **); 151 static void igb_tso_ctx(struct igb_tx_ring *, struct mbuf *, uint32_t *); 152 static void igb_add_sysctl(struct igb_softc *); 153 static void igb_add_intr_rate_sysctl(struct igb_softc *, int, 154 const char *, const char *); 155 static int igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS); 156 static int igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS); 157 static int igb_sysctl_tx_wreg_nsegs(SYSCTL_HANDLER_ARGS); 158 static int igb_sysctl_rx_wreg_nsegs(SYSCTL_HANDLER_ARGS); 159 static void igb_set_ring_inuse(struct igb_softc *, boolean_t); 160 static int igb_get_rxring_inuse(const struct igb_softc *, boolean_t); 161 static int igb_get_txring_inuse(const struct igb_softc *, boolean_t); 162 static void igb_set_timer_cpuid(struct igb_softc *, boolean_t); 163 #ifdef IFPOLL_ENABLE 164 static int igb_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS); 165 static int igb_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS); 166 #endif 167 168 static void igb_vf_init_stats(struct igb_softc *); 169 static void igb_reset(struct igb_softc *, boolean_t); 170 static void igb_update_stats_counters(struct igb_softc *); 171 static void igb_update_vf_stats_counters(struct igb_softc *); 172 static void igb_update_link_status(struct igb_softc *); 173 static void igb_init_tx_unit(struct igb_softc *); 174 static void igb_init_rx_unit(struct igb_softc *); 175 static void igb_init_dmac(struct igb_softc *, uint32_t); 176 177 static void igb_set_vlan(struct igb_softc *); 178 static void igb_set_multi(struct igb_softc *); 179 static void igb_set_promisc(struct igb_softc *); 180 static void igb_disable_promisc(struct igb_softc *); 181 182 static int igb_alloc_rings(struct igb_softc *); 183 static void igb_free_rings(struct igb_softc *); 184 static int igb_create_tx_ring(struct igb_tx_ring *); 185 static int igb_create_rx_ring(struct igb_rx_ring *); 186 static void igb_free_tx_ring(struct igb_tx_ring *); 187 static void igb_free_rx_ring(struct igb_rx_ring *); 188 static void igb_destroy_tx_ring(struct igb_tx_ring *, int); 189 static void igb_destroy_rx_ring(struct igb_rx_ring *, int); 190 static void igb_init_tx_ring(struct igb_tx_ring *); 191 static int igb_init_rx_ring(struct igb_rx_ring *); 192 static int igb_newbuf(struct igb_rx_ring *, int, boolean_t); 193 static int igb_encap(struct igb_tx_ring *, struct mbuf **, int *, int *); 194 static void igb_rx_refresh(struct igb_rx_ring *, int); 195 static void igb_setup_serialize(struct igb_softc *); 196 197 static void igb_stop(struct igb_softc *); 198 static void igb_init(void *); 199 static int igb_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *); 200 static void igb_media_status(struct ifnet *, struct ifmediareq *); 201 static int igb_media_change(struct ifnet *); 202 static void igb_timer(void *); 203 static void igb_watchdog(struct ifaltq_subque *); 204 static void igb_start(struct ifnet *, struct ifaltq_subque *); 205 #ifdef IFPOLL_ENABLE 206 static void igb_npoll(struct ifnet *, struct ifpoll_info *); 207 static void igb_npoll_rx(struct ifnet *, void *, int); 208 static void igb_npoll_tx(struct ifnet *, void *, int); 209 static void igb_npoll_status(struct ifnet *); 210 #endif 211 static void igb_serialize(struct ifnet *, enum ifnet_serialize); 212 static void igb_deserialize(struct ifnet *, enum ifnet_serialize); 213 static int igb_tryserialize(struct ifnet *, enum ifnet_serialize); 214 #ifdef INVARIANTS 215 static void igb_serialize_assert(struct ifnet *, enum ifnet_serialize, 216 boolean_t); 217 #endif 218 219 static void igb_intr(void *); 220 static void igb_intr_shared(void *); 221 static void igb_rxeof(struct igb_rx_ring *, int); 222 static void igb_txeof(struct igb_tx_ring *, int); 223 static void igb_set_eitr(struct igb_softc *, int, int); 224 static void igb_enable_intr(struct igb_softc *); 225 static void igb_disable_intr(struct igb_softc *); 226 static void igb_init_unshared_intr(struct igb_softc *); 227 static void igb_init_intr(struct igb_softc *); 228 static int igb_setup_intr(struct igb_softc *); 229 static void igb_set_txintr_mask(struct igb_tx_ring *, int *, int); 230 static void igb_set_rxintr_mask(struct igb_rx_ring *, int *, int); 231 static void igb_set_intr_mask(struct igb_softc *); 232 static int igb_alloc_intr(struct igb_softc *); 233 static void igb_free_intr(struct igb_softc *); 234 static void igb_teardown_intr(struct igb_softc *, int); 235 static void igb_alloc_msix(struct igb_softc *); 236 static void igb_free_msix(struct igb_softc *, boolean_t); 237 static void igb_msix_rx_conf(struct igb_softc *, int, int *, int); 238 static void igb_msix_tx_conf(struct igb_softc *, int, int *, int); 239 static void igb_msix_rx(void *); 240 static void igb_msix_tx(void *); 241 static void igb_msix_status(void *); 242 static void igb_msix_rxtx(void *); 243 244 /* Management and WOL Support */ 245 static void igb_get_mgmt(struct igb_softc *); 246 static void igb_rel_mgmt(struct igb_softc *); 247 static void igb_get_hw_control(struct igb_softc *); 248 static void igb_rel_hw_control(struct igb_softc *); 249 static void igb_enable_wol(device_t); 250 251 static device_method_t igb_methods[] = { 252 /* Device interface */ 253 DEVMETHOD(device_probe, igb_probe), 254 DEVMETHOD(device_attach, igb_attach), 255 DEVMETHOD(device_detach, igb_detach), 256 DEVMETHOD(device_shutdown, igb_shutdown), 257 DEVMETHOD(device_suspend, igb_suspend), 258 DEVMETHOD(device_resume, igb_resume), 259 DEVMETHOD_END 260 }; 261 262 static driver_t igb_driver = { 263 "igb", 264 igb_methods, 265 sizeof(struct igb_softc), 266 }; 267 268 static devclass_t igb_devclass; 269 270 DECLARE_DUMMY_MODULE(if_igb); 271 MODULE_DEPEND(igb, ig_hal, 1, 1, 1); 272 DRIVER_MODULE(if_igb, pci, igb_driver, igb_devclass, NULL, NULL); 273 274 static int igb_rxd = IGB_DEFAULT_RXD; 275 static int igb_txd = IGB_DEFAULT_TXD; 276 static int igb_rxr = 0; 277 static int igb_txr = 0; 278 static int igb_msi_enable = 1; 279 static int igb_msix_enable = 1; 280 static int igb_msix_agg_rxtx = 1; 281 static int igb_eee_disabled = 1; /* Energy Efficient Ethernet */ 282 283 static char igb_flowctrl[IFM_ETH_FC_STRLEN] = IFM_ETH_FC_RXPAUSE; 284 285 /* 286 * DMA Coalescing, only for i350 - default to off, 287 * this feature is for power savings 288 */ 289 static int igb_dma_coalesce = 0; 290 291 TUNABLE_INT("hw.igb.rxd", &igb_rxd); 292 TUNABLE_INT("hw.igb.txd", &igb_txd); 293 TUNABLE_INT("hw.igb.rxr", &igb_rxr); 294 TUNABLE_INT("hw.igb.txr", &igb_txr); 295 TUNABLE_INT("hw.igb.msi.enable", &igb_msi_enable); 296 TUNABLE_INT("hw.igb.msix.enable", &igb_msix_enable); 297 TUNABLE_INT("hw.igb.msix.agg_rxtx", &igb_msix_agg_rxtx); 298 TUNABLE_STR("hw.igb.flow_ctrl", igb_flowctrl, sizeof(igb_flowctrl)); 299 300 /* i350 specific */ 301 TUNABLE_INT("hw.igb.eee_disabled", &igb_eee_disabled); 302 TUNABLE_INT("hw.igb.dma_coalesce", &igb_dma_coalesce); 303 304 static __inline void 305 igb_rxcsum(uint32_t staterr, struct mbuf *mp) 306 { 307 /* Ignore Checksum bit is set */ 308 if (staterr & E1000_RXD_STAT_IXSM) 309 return; 310 311 if ((staterr & (E1000_RXD_STAT_IPCS | E1000_RXDEXT_STATERR_IPE)) == 312 E1000_RXD_STAT_IPCS) 313 mp->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; 314 315 if (staterr & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)) { 316 if ((staterr & E1000_RXDEXT_STATERR_TCPE) == 0) { 317 mp->m_pkthdr.csum_flags |= CSUM_DATA_VALID | 318 CSUM_PSEUDO_HDR | CSUM_FRAG_NOT_CHECKED; 319 mp->m_pkthdr.csum_data = htons(0xffff); 320 } 321 } 322 } 323 324 static __inline struct pktinfo * 325 igb_rssinfo(struct mbuf *m, struct pktinfo *pi, 326 uint32_t hash, uint32_t hashtype, uint32_t staterr) 327 { 328 switch (hashtype) { 329 case E1000_RXDADV_RSSTYPE_IPV4_TCP: 330 pi->pi_netisr = NETISR_IP; 331 pi->pi_flags = 0; 332 pi->pi_l3proto = IPPROTO_TCP; 333 break; 334 335 case E1000_RXDADV_RSSTYPE_IPV4: 336 if (staterr & E1000_RXD_STAT_IXSM) 337 return NULL; 338 339 if ((staterr & 340 (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) == 341 E1000_RXD_STAT_TCPCS) { 342 pi->pi_netisr = NETISR_IP; 343 pi->pi_flags = 0; 344 pi->pi_l3proto = IPPROTO_UDP; 345 break; 346 } 347 /* FALL THROUGH */ 348 default: 349 return NULL; 350 } 351 352 m->m_flags |= M_HASH; 353 m->m_pkthdr.hash = toeplitz_hash(hash); 354 return pi; 355 } 356 357 static int 358 igb_probe(device_t dev) 359 { 360 const struct igb_device *d; 361 uint16_t vid, did; 362 363 vid = pci_get_vendor(dev); 364 did = pci_get_device(dev); 365 366 for (d = igb_devices; d->desc != NULL; ++d) { 367 if (vid == d->vid && did == d->did) { 368 device_set_desc(dev, d->desc); 369 return 0; 370 } 371 } 372 return ENXIO; 373 } 374 375 static int 376 igb_attach(device_t dev) 377 { 378 struct igb_softc *sc = device_get_softc(dev); 379 uint16_t eeprom_data; 380 int error = 0, ring_max; 381 char flowctrl[IFM_ETH_FC_STRLEN]; 382 #ifdef IFPOLL_ENABLE 383 int offset, offset_def; 384 #endif 385 386 #ifdef notyet 387 /* SYSCTL stuff */ 388 SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), 389 SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), 390 OID_AUTO, "nvm", CTLTYPE_INT|CTLFLAG_RW, adapter, 0, 391 igb_sysctl_nvm_info, "I", "NVM Information"); 392 #endif 393 394 ifmedia_init(&sc->media, IFM_IMASK | IFM_ETH_FCMASK, 395 igb_media_change, igb_media_status); 396 callout_init_mp(&sc->timer); 397 lwkt_serialize_init(&sc->main_serialize); 398 399 if_initname(&sc->arpcom.ac_if, device_get_name(dev), 400 device_get_unit(dev)); 401 sc->dev = sc->osdep.dev = dev; 402 403 /* Enable bus mastering */ 404 pci_enable_busmaster(dev); 405 406 /* 407 * Determine hardware and mac type 408 */ 409 sc->hw.vendor_id = pci_get_vendor(dev); 410 sc->hw.device_id = pci_get_device(dev); 411 sc->hw.revision_id = pci_read_config(dev, PCIR_REVID, 1); 412 sc->hw.subsystem_vendor_id = pci_read_config(dev, PCIR_SUBVEND_0, 2); 413 sc->hw.subsystem_device_id = pci_read_config(dev, PCIR_SUBDEV_0, 2); 414 415 if (e1000_set_mac_type(&sc->hw)) 416 return ENXIO; 417 418 /* Are we a VF device? */ 419 if (sc->hw.mac.type == e1000_vfadapt || 420 sc->hw.mac.type == e1000_vfadapt_i350) 421 sc->vf_ifp = 1; 422 else 423 sc->vf_ifp = 0; 424 425 /* 426 * Configure total supported RX/TX ring count 427 */ 428 switch (sc->hw.mac.type) { 429 case e1000_82575: 430 ring_max = IGB_MAX_RING_82575; 431 break; 432 433 case e1000_82576: 434 ring_max = IGB_MAX_RING_82576; 435 break; 436 437 case e1000_82580: 438 ring_max = IGB_MAX_RING_82580; 439 break; 440 441 case e1000_i350: 442 ring_max = IGB_MAX_RING_I350; 443 break; 444 445 case e1000_i354: 446 ring_max = IGB_MAX_RING_I354; 447 break; 448 449 case e1000_i210: 450 ring_max = IGB_MAX_RING_I210; 451 break; 452 453 case e1000_i211: 454 ring_max = IGB_MAX_RING_I211; 455 break; 456 457 default: 458 ring_max = IGB_MIN_RING; 459 break; 460 } 461 462 sc->rx_ring_cnt = device_getenv_int(dev, "rxr", igb_rxr); 463 sc->rx_ring_cnt = if_ring_count2(sc->rx_ring_cnt, ring_max); 464 #ifdef IGB_RSS_DEBUG 465 sc->rx_ring_cnt = device_getenv_int(dev, "rxr_debug", sc->rx_ring_cnt); 466 #endif 467 sc->rx_ring_inuse = sc->rx_ring_cnt; 468 469 sc->tx_ring_cnt = device_getenv_int(dev, "txr", igb_txr); 470 sc->tx_ring_cnt = if_ring_count2(sc->tx_ring_cnt, ring_max); 471 #ifdef IGB_TSS_DEBUG 472 sc->tx_ring_cnt = device_getenv_int(dev, "txr_debug", sc->tx_ring_cnt); 473 #endif 474 sc->tx_ring_inuse = sc->tx_ring_cnt; 475 476 /* Setup flow control. */ 477 device_getenv_string(dev, "flow_ctrl", flowctrl, sizeof(flowctrl), 478 igb_flowctrl); 479 sc->ifm_flowctrl = ifmedia_str2ethfc(flowctrl); 480 481 /* 482 * Allocate IO memory 483 */ 484 sc->mem_rid = PCIR_BAR(0); 485 sc->mem_res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->mem_rid, 486 RF_ACTIVE); 487 if (sc->mem_res == NULL) { 488 device_printf(dev, "Unable to allocate bus resource: memory\n"); 489 error = ENXIO; 490 goto failed; 491 } 492 sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->mem_res); 493 sc->osdep.mem_bus_space_handle = rman_get_bushandle(sc->mem_res); 494 495 sc->hw.hw_addr = (uint8_t *)&sc->osdep.mem_bus_space_handle; 496 497 /* Save PCI command register for Shared Code */ 498 sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2); 499 sc->hw.back = &sc->osdep; 500 501 /* Do Shared Code initialization */ 502 if (e1000_setup_init_funcs(&sc->hw, TRUE)) { 503 device_printf(dev, "Setup of Shared code failed\n"); 504 error = ENXIO; 505 goto failed; 506 } 507 508 e1000_get_bus_info(&sc->hw); 509 510 sc->hw.mac.autoneg = DO_AUTO_NEG; 511 sc->hw.phy.autoneg_wait_to_complete = FALSE; 512 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT; 513 514 /* Copper options */ 515 if (sc->hw.phy.media_type == e1000_media_type_copper) { 516 sc->hw.phy.mdix = AUTO_ALL_MODES; 517 sc->hw.phy.disable_polarity_correction = FALSE; 518 sc->hw.phy.ms_type = IGB_MASTER_SLAVE; 519 } 520 521 /* Set the frame limits assuming standard ethernet sized frames. */ 522 sc->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN; 523 524 /* Allocate RX/TX rings */ 525 error = igb_alloc_rings(sc); 526 if (error) 527 goto failed; 528 529 #ifdef IFPOLL_ENABLE 530 /* 531 * NPOLLING RX CPU offset 532 */ 533 if (sc->rx_ring_cnt == ncpus2) { 534 offset = 0; 535 } else { 536 offset_def = (sc->rx_ring_cnt * device_get_unit(dev)) % ncpus2; 537 offset = device_getenv_int(dev, "npoll.rxoff", offset_def); 538 if (offset >= ncpus2 || 539 offset % sc->rx_ring_cnt != 0) { 540 device_printf(dev, "invalid npoll.rxoff %d, use %d\n", 541 offset, offset_def); 542 offset = offset_def; 543 } 544 } 545 sc->rx_npoll_off = offset; 546 547 /* 548 * NPOLLING TX CPU offset 549 */ 550 if (sc->tx_ring_cnt == ncpus2) { 551 offset = 0; 552 } else { 553 offset_def = (sc->tx_ring_cnt * device_get_unit(dev)) % ncpus2; 554 offset = device_getenv_int(dev, "npoll.txoff", offset_def); 555 if (offset >= ncpus2 || 556 offset % sc->tx_ring_cnt != 0) { 557 device_printf(dev, "invalid npoll.txoff %d, use %d\n", 558 offset, offset_def); 559 offset = offset_def; 560 } 561 } 562 sc->tx_npoll_off = offset; 563 #endif 564 565 /* Allocate interrupt */ 566 error = igb_alloc_intr(sc); 567 if (error) 568 goto failed; 569 570 /* Setup serializes */ 571 igb_setup_serialize(sc); 572 573 /* Allocate the appropriate stats memory */ 574 if (sc->vf_ifp) { 575 sc->stats = kmalloc(sizeof(struct e1000_vf_stats), M_DEVBUF, 576 M_WAITOK | M_ZERO); 577 igb_vf_init_stats(sc); 578 } else { 579 sc->stats = kmalloc(sizeof(struct e1000_hw_stats), M_DEVBUF, 580 M_WAITOK | M_ZERO); 581 } 582 583 /* Allocate multicast array memory. */ 584 sc->mta = kmalloc(ETHER_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES, 585 M_DEVBUF, M_WAITOK); 586 587 /* Some adapter-specific advanced features */ 588 if (sc->hw.mac.type >= e1000_i350) { 589 #ifdef notyet 590 igb_set_sysctl_value(adapter, "dma_coalesce", 591 "configure dma coalesce", 592 &adapter->dma_coalesce, igb_dma_coalesce); 593 igb_set_sysctl_value(adapter, "eee_disabled", 594 "enable Energy Efficient Ethernet", 595 &adapter->hw.dev_spec._82575.eee_disable, 596 igb_eee_disabled); 597 #else 598 sc->dma_coalesce = igb_dma_coalesce; 599 sc->hw.dev_spec._82575.eee_disable = igb_eee_disabled; 600 #endif 601 if (sc->hw.phy.media_type == e1000_media_type_copper) { 602 if (sc->hw.mac.type == e1000_i354) 603 e1000_set_eee_i354(&sc->hw, TRUE, TRUE); 604 else 605 e1000_set_eee_i350(&sc->hw, TRUE, TRUE); 606 } 607 } 608 609 /* 610 * Start from a known state, this is important in reading the nvm and 611 * mac from that. 612 */ 613 e1000_reset_hw(&sc->hw); 614 615 /* Make sure we have a good EEPROM before we read from it */ 616 if (sc->hw.mac.type != e1000_i210 && sc->hw.mac.type != e1000_i211 && 617 e1000_validate_nvm_checksum(&sc->hw) < 0) { 618 /* 619 * Some PCI-E parts fail the first check due to 620 * the link being in sleep state, call it again, 621 * if it fails a second time its a real issue. 622 */ 623 if (e1000_validate_nvm_checksum(&sc->hw) < 0) { 624 device_printf(dev, 625 "The EEPROM Checksum Is Not Valid\n"); 626 error = EIO; 627 goto failed; 628 } 629 } 630 631 /* Copy the permanent MAC address out of the EEPROM */ 632 if (e1000_read_mac_addr(&sc->hw) < 0) { 633 device_printf(dev, "EEPROM read error while reading MAC" 634 " address\n"); 635 error = EIO; 636 goto failed; 637 } 638 if (!igb_is_valid_ether_addr(sc->hw.mac.addr)) { 639 device_printf(dev, "Invalid MAC address\n"); 640 error = EIO; 641 goto failed; 642 } 643 644 /* Setup OS specific network interface */ 645 igb_setup_ifp(sc); 646 647 /* Add sysctl tree, must after igb_setup_ifp() */ 648 igb_add_sysctl(sc); 649 650 /* Now get a good starting state */ 651 igb_reset(sc, FALSE); 652 653 /* Initialize statistics */ 654 igb_update_stats_counters(sc); 655 656 sc->hw.mac.get_link_status = 1; 657 igb_update_link_status(sc); 658 659 /* Indicate SOL/IDER usage */ 660 if (e1000_check_reset_block(&sc->hw)) { 661 device_printf(dev, 662 "PHY reset is blocked due to SOL/IDER session.\n"); 663 } 664 665 /* Determine if we have to control management hardware */ 666 if (e1000_enable_mng_pass_thru(&sc->hw)) 667 sc->flags |= IGB_FLAG_HAS_MGMT; 668 669 /* 670 * Setup Wake-on-Lan 671 */ 672 /* APME bit in EEPROM is mapped to WUC.APME */ 673 eeprom_data = E1000_READ_REG(&sc->hw, E1000_WUC) & E1000_WUC_APME; 674 if (eeprom_data) 675 sc->wol = E1000_WUFC_MAG; 676 /* XXX disable WOL */ 677 sc->wol = 0; 678 679 #ifdef notyet 680 /* Register for VLAN events */ 681 adapter->vlan_attach = EVENTHANDLER_REGISTER(vlan_config, 682 igb_register_vlan, adapter, EVENTHANDLER_PRI_FIRST); 683 adapter->vlan_detach = EVENTHANDLER_REGISTER(vlan_unconfig, 684 igb_unregister_vlan, adapter, EVENTHANDLER_PRI_FIRST); 685 #endif 686 687 #ifdef notyet 688 igb_add_hw_stats(adapter); 689 #endif 690 691 /* 692 * Disable interrupt to prevent spurious interrupts (line based 693 * interrupt, MSI or even MSI-X), which had been observed on 694 * several types of LOMs, from being handled. 695 */ 696 igb_disable_intr(sc); 697 698 error = igb_setup_intr(sc); 699 if (error) { 700 ether_ifdetach(&sc->arpcom.ac_if); 701 goto failed; 702 } 703 return 0; 704 705 failed: 706 igb_detach(dev); 707 return error; 708 } 709 710 static int 711 igb_detach(device_t dev) 712 { 713 struct igb_softc *sc = device_get_softc(dev); 714 715 if (device_is_attached(dev)) { 716 struct ifnet *ifp = &sc->arpcom.ac_if; 717 718 ifnet_serialize_all(ifp); 719 720 igb_stop(sc); 721 722 e1000_phy_hw_reset(&sc->hw); 723 724 /* Give control back to firmware */ 725 igb_rel_mgmt(sc); 726 igb_rel_hw_control(sc); 727 728 if (sc->wol) { 729 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN); 730 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol); 731 igb_enable_wol(dev); 732 } 733 734 igb_teardown_intr(sc, sc->intr_cnt); 735 736 ifnet_deserialize_all(ifp); 737 738 ether_ifdetach(ifp); 739 } else if (sc->mem_res != NULL) { 740 igb_rel_hw_control(sc); 741 } 742 743 ifmedia_removeall(&sc->media); 744 bus_generic_detach(dev); 745 746 igb_free_intr(sc); 747 748 if (sc->msix_mem_res != NULL) { 749 bus_release_resource(dev, SYS_RES_MEMORY, sc->msix_mem_rid, 750 sc->msix_mem_res); 751 } 752 if (sc->mem_res != NULL) { 753 bus_release_resource(dev, SYS_RES_MEMORY, sc->mem_rid, 754 sc->mem_res); 755 } 756 757 igb_free_rings(sc); 758 759 if (sc->mta != NULL) 760 kfree(sc->mta, M_DEVBUF); 761 if (sc->stats != NULL) 762 kfree(sc->stats, M_DEVBUF); 763 if (sc->serializes != NULL) 764 kfree(sc->serializes, M_DEVBUF); 765 766 return 0; 767 } 768 769 static int 770 igb_shutdown(device_t dev) 771 { 772 return igb_suspend(dev); 773 } 774 775 static int 776 igb_suspend(device_t dev) 777 { 778 struct igb_softc *sc = device_get_softc(dev); 779 struct ifnet *ifp = &sc->arpcom.ac_if; 780 781 ifnet_serialize_all(ifp); 782 783 igb_stop(sc); 784 785 igb_rel_mgmt(sc); 786 igb_rel_hw_control(sc); 787 788 if (sc->wol) { 789 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN); 790 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol); 791 igb_enable_wol(dev); 792 } 793 794 ifnet_deserialize_all(ifp); 795 796 return bus_generic_suspend(dev); 797 } 798 799 static int 800 igb_resume(device_t dev) 801 { 802 struct igb_softc *sc = device_get_softc(dev); 803 struct ifnet *ifp = &sc->arpcom.ac_if; 804 int i; 805 806 ifnet_serialize_all(ifp); 807 808 igb_init(sc); 809 igb_get_mgmt(sc); 810 811 for (i = 0; i < sc->tx_ring_inuse; ++i) 812 ifsq_devstart_sched(sc->tx_rings[i].ifsq); 813 814 ifnet_deserialize_all(ifp); 815 816 return bus_generic_resume(dev); 817 } 818 819 static int 820 igb_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr) 821 { 822 struct igb_softc *sc = ifp->if_softc; 823 struct ifreq *ifr = (struct ifreq *)data; 824 int max_frame_size, mask, reinit; 825 int error = 0; 826 827 ASSERT_IFNET_SERIALIZED_ALL(ifp); 828 829 switch (command) { 830 case SIOCSIFMTU: 831 max_frame_size = 9234; 832 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN - 833 ETHER_CRC_LEN) { 834 error = EINVAL; 835 break; 836 } 837 838 ifp->if_mtu = ifr->ifr_mtu; 839 sc->max_frame_size = ifp->if_mtu + ETHER_HDR_LEN + 840 ETHER_CRC_LEN; 841 842 if (ifp->if_flags & IFF_RUNNING) 843 igb_init(sc); 844 break; 845 846 case SIOCSIFFLAGS: 847 if (ifp->if_flags & IFF_UP) { 848 if (ifp->if_flags & IFF_RUNNING) { 849 if ((ifp->if_flags ^ sc->if_flags) & 850 (IFF_PROMISC | IFF_ALLMULTI)) { 851 igb_disable_promisc(sc); 852 igb_set_promisc(sc); 853 } 854 } else { 855 igb_init(sc); 856 } 857 } else if (ifp->if_flags & IFF_RUNNING) { 858 igb_stop(sc); 859 } 860 sc->if_flags = ifp->if_flags; 861 break; 862 863 case SIOCADDMULTI: 864 case SIOCDELMULTI: 865 if (ifp->if_flags & IFF_RUNNING) { 866 igb_disable_intr(sc); 867 igb_set_multi(sc); 868 #ifdef IFPOLL_ENABLE 869 if (!(ifp->if_flags & IFF_NPOLLING)) 870 #endif 871 igb_enable_intr(sc); 872 } 873 break; 874 875 case SIOCSIFMEDIA: 876 /* Check SOL/IDER usage */ 877 if (e1000_check_reset_block(&sc->hw)) { 878 if_printf(ifp, "Media change is " 879 "blocked due to SOL/IDER session.\n"); 880 break; 881 } 882 /* FALL THROUGH */ 883 884 case SIOCGIFMEDIA: 885 error = ifmedia_ioctl(ifp, ifr, &sc->media, command); 886 break; 887 888 case SIOCSIFCAP: 889 reinit = 0; 890 mask = ifr->ifr_reqcap ^ ifp->if_capenable; 891 if (mask & IFCAP_RXCSUM) { 892 ifp->if_capenable ^= IFCAP_RXCSUM; 893 reinit = 1; 894 } 895 if (mask & IFCAP_VLAN_HWTAGGING) { 896 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; 897 reinit = 1; 898 } 899 if (mask & IFCAP_TXCSUM) { 900 ifp->if_capenable ^= IFCAP_TXCSUM; 901 if (ifp->if_capenable & IFCAP_TXCSUM) 902 ifp->if_hwassist |= IGB_CSUM_FEATURES; 903 else 904 ifp->if_hwassist &= ~IGB_CSUM_FEATURES; 905 } 906 if (mask & IFCAP_TSO) { 907 ifp->if_capenable ^= IFCAP_TSO; 908 if (ifp->if_capenable & IFCAP_TSO) 909 ifp->if_hwassist |= CSUM_TSO; 910 else 911 ifp->if_hwassist &= ~CSUM_TSO; 912 } 913 if (mask & IFCAP_RSS) 914 ifp->if_capenable ^= IFCAP_RSS; 915 if (reinit && (ifp->if_flags & IFF_RUNNING)) 916 igb_init(sc); 917 break; 918 919 default: 920 error = ether_ioctl(ifp, command, data); 921 break; 922 } 923 return error; 924 } 925 926 static void 927 igb_init(void *xsc) 928 { 929 struct igb_softc *sc = xsc; 930 struct ifnet *ifp = &sc->arpcom.ac_if; 931 boolean_t polling; 932 int i; 933 934 ASSERT_IFNET_SERIALIZED_ALL(ifp); 935 936 igb_stop(sc); 937 938 /* Get the latest mac address, User can use a LAA */ 939 bcopy(IF_LLADDR(ifp), sc->hw.mac.addr, ETHER_ADDR_LEN); 940 941 /* Put the address into the Receive Address Array */ 942 e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0); 943 944 igb_reset(sc, FALSE); 945 igb_update_link_status(sc); 946 947 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN); 948 949 /* Configure for OS presence */ 950 igb_get_mgmt(sc); 951 952 polling = FALSE; 953 #ifdef IFPOLL_ENABLE 954 if (ifp->if_flags & IFF_NPOLLING) 955 polling = TRUE; 956 #endif 957 958 /* Configured used RX/TX rings */ 959 igb_set_ring_inuse(sc, polling); 960 ifq_set_subq_mask(&ifp->if_snd, sc->tx_ring_inuse - 1); 961 962 /* Initialize interrupt */ 963 igb_init_intr(sc); 964 965 /* Prepare transmit descriptors and buffers */ 966 for (i = 0; i < sc->tx_ring_inuse; ++i) 967 igb_init_tx_ring(&sc->tx_rings[i]); 968 igb_init_tx_unit(sc); 969 970 /* Setup Multicast table */ 971 igb_set_multi(sc); 972 973 #if 0 974 /* 975 * Figure out the desired mbuf pool 976 * for doing jumbo/packetsplit 977 */ 978 if (adapter->max_frame_size <= 2048) 979 adapter->rx_mbuf_sz = MCLBYTES; 980 else if (adapter->max_frame_size <= 4096) 981 adapter->rx_mbuf_sz = MJUMPAGESIZE; 982 else 983 adapter->rx_mbuf_sz = MJUM9BYTES; 984 #endif 985 986 /* Prepare receive descriptors and buffers */ 987 for (i = 0; i < sc->rx_ring_inuse; ++i) { 988 int error; 989 990 error = igb_init_rx_ring(&sc->rx_rings[i]); 991 if (error) { 992 if_printf(ifp, "Could not setup receive structures\n"); 993 igb_stop(sc); 994 return; 995 } 996 } 997 igb_init_rx_unit(sc); 998 999 /* Enable VLAN support */ 1000 if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) 1001 igb_set_vlan(sc); 1002 1003 /* Don't lose promiscuous settings */ 1004 igb_set_promisc(sc); 1005 1006 ifp->if_flags |= IFF_RUNNING; 1007 for (i = 0; i < sc->tx_ring_inuse; ++i) { 1008 ifsq_clr_oactive(sc->tx_rings[i].ifsq); 1009 ifsq_watchdog_start(&sc->tx_rings[i].tx_watchdog); 1010 } 1011 1012 igb_set_timer_cpuid(sc, polling); 1013 callout_reset_bycpu(&sc->timer, hz, igb_timer, sc, sc->timer_cpuid); 1014 e1000_clear_hw_cntrs_base_generic(&sc->hw); 1015 1016 /* This clears any pending interrupts */ 1017 E1000_READ_REG(&sc->hw, E1000_ICR); 1018 1019 /* 1020 * Only enable interrupts if we are not polling, make sure 1021 * they are off otherwise. 1022 */ 1023 if (polling) { 1024 igb_disable_intr(sc); 1025 } else { 1026 igb_enable_intr(sc); 1027 E1000_WRITE_REG(&sc->hw, E1000_ICS, E1000_ICS_LSC); 1028 } 1029 1030 /* Set Energy Efficient Ethernet */ 1031 if (sc->hw.phy.media_type == e1000_media_type_copper) { 1032 if (sc->hw.mac.type == e1000_i354) 1033 e1000_set_eee_i354(&sc->hw, TRUE, TRUE); 1034 else 1035 e1000_set_eee_i350(&sc->hw, TRUE, TRUE); 1036 } 1037 } 1038 1039 static void 1040 igb_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) 1041 { 1042 struct igb_softc *sc = ifp->if_softc; 1043 1044 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1045 1046 if ((ifp->if_flags & IFF_RUNNING) == 0) 1047 sc->hw.mac.get_link_status = 1; 1048 igb_update_link_status(sc); 1049 1050 ifmr->ifm_status = IFM_AVALID; 1051 ifmr->ifm_active = IFM_ETHER; 1052 1053 if (!sc->link_active) { 1054 if (sc->hw.mac.autoneg) 1055 ifmr->ifm_active |= IFM_NONE; 1056 else 1057 ifmr->ifm_active |= sc->media.ifm_media; 1058 return; 1059 } 1060 1061 ifmr->ifm_status |= IFM_ACTIVE; 1062 if (sc->ifm_flowctrl & IFM_ETH_FORCEPAUSE) 1063 ifmr->ifm_active |= sc->ifm_flowctrl; 1064 1065 switch (sc->link_speed) { 1066 case 10: 1067 ifmr->ifm_active |= IFM_10_T; 1068 break; 1069 1070 case 100: 1071 /* 1072 * Support for 100Mb SFP - these are Fiber 1073 * but the media type appears as serdes 1074 */ 1075 if (sc->hw.phy.media_type == e1000_media_type_fiber || 1076 sc->hw.phy.media_type == e1000_media_type_internal_serdes) 1077 ifmr->ifm_active |= IFM_100_FX; 1078 else 1079 ifmr->ifm_active |= IFM_100_TX; 1080 break; 1081 1082 case 1000: 1083 if (sc->hw.phy.media_type == e1000_media_type_fiber || 1084 sc->hw.phy.media_type == e1000_media_type_internal_serdes) 1085 ifmr->ifm_active |= IFM_1000_SX; 1086 else 1087 ifmr->ifm_active |= IFM_1000_T; 1088 break; 1089 1090 case 2500: 1091 ifmr->ifm_active |= IFM_2500_SX; 1092 break; 1093 } 1094 1095 if (sc->link_duplex == FULL_DUPLEX) 1096 ifmr->ifm_active |= IFM_FDX; 1097 else 1098 ifmr->ifm_active |= IFM_HDX; 1099 1100 if (sc->link_duplex == FULL_DUPLEX) 1101 ifmr->ifm_active |= e1000_fc2ifmedia(sc->hw.fc.current_mode); 1102 } 1103 1104 static int 1105 igb_media_change(struct ifnet *ifp) 1106 { 1107 struct igb_softc *sc = ifp->if_softc; 1108 struct ifmedia *ifm = &sc->media; 1109 1110 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1111 1112 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) 1113 return EINVAL; 1114 1115 switch (IFM_SUBTYPE(ifm->ifm_media)) { 1116 case IFM_AUTO: 1117 sc->hw.mac.autoneg = DO_AUTO_NEG; 1118 sc->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT; 1119 break; 1120 1121 case IFM_1000_SX: 1122 case IFM_1000_T: 1123 sc->hw.mac.autoneg = DO_AUTO_NEG; 1124 sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL; 1125 break; 1126 1127 case IFM_100_TX: 1128 if (IFM_OPTIONS(ifm->ifm_media) & IFM_FDX) { 1129 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL; 1130 } else { 1131 if (IFM_OPTIONS(ifm->ifm_media) & 1132 (IFM_ETH_RXPAUSE | IFM_ETH_TXPAUSE)) { 1133 if (bootverbose) { 1134 if_printf(ifp, "Flow control is not " 1135 "allowed for half-duplex\n"); 1136 } 1137 return EINVAL; 1138 } 1139 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF; 1140 } 1141 sc->hw.mac.autoneg = FALSE; 1142 sc->hw.phy.autoneg_advertised = 0; 1143 break; 1144 1145 case IFM_10_T: 1146 if (IFM_OPTIONS(ifm->ifm_media) & IFM_FDX) { 1147 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL; 1148 } else { 1149 if (IFM_OPTIONS(ifm->ifm_media) & 1150 (IFM_ETH_RXPAUSE | IFM_ETH_TXPAUSE)) { 1151 if (bootverbose) { 1152 if_printf(ifp, "Flow control is not " 1153 "allowed for half-duplex\n"); 1154 } 1155 return EINVAL; 1156 } 1157 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF; 1158 } 1159 sc->hw.mac.autoneg = FALSE; 1160 sc->hw.phy.autoneg_advertised = 0; 1161 break; 1162 1163 default: 1164 if (bootverbose) { 1165 if_printf(ifp, "Unsupported media type %d\n", 1166 IFM_SUBTYPE(ifm->ifm_media)); 1167 } 1168 return EINVAL; 1169 } 1170 sc->ifm_flowctrl = ifm->ifm_media & IFM_ETH_FCMASK; 1171 1172 if (ifp->if_flags & IFF_RUNNING) 1173 igb_init(sc); 1174 1175 return 0; 1176 } 1177 1178 static void 1179 igb_set_promisc(struct igb_softc *sc) 1180 { 1181 struct ifnet *ifp = &sc->arpcom.ac_if; 1182 struct e1000_hw *hw = &sc->hw; 1183 uint32_t reg; 1184 1185 if (sc->vf_ifp) { 1186 e1000_promisc_set_vf(hw, e1000_promisc_enabled); 1187 return; 1188 } 1189 1190 reg = E1000_READ_REG(hw, E1000_RCTL); 1191 if (ifp->if_flags & IFF_PROMISC) { 1192 reg |= (E1000_RCTL_UPE | E1000_RCTL_MPE); 1193 E1000_WRITE_REG(hw, E1000_RCTL, reg); 1194 } else if (ifp->if_flags & IFF_ALLMULTI) { 1195 reg |= E1000_RCTL_MPE; 1196 reg &= ~E1000_RCTL_UPE; 1197 E1000_WRITE_REG(hw, E1000_RCTL, reg); 1198 } 1199 } 1200 1201 static void 1202 igb_disable_promisc(struct igb_softc *sc) 1203 { 1204 struct e1000_hw *hw = &sc->hw; 1205 struct ifnet *ifp = &sc->arpcom.ac_if; 1206 uint32_t reg; 1207 int mcnt = 0; 1208 1209 if (sc->vf_ifp) { 1210 e1000_promisc_set_vf(hw, e1000_promisc_disabled); 1211 return; 1212 } 1213 reg = E1000_READ_REG(hw, E1000_RCTL); 1214 reg &= ~E1000_RCTL_UPE; 1215 if (ifp->if_flags & IFF_ALLMULTI) { 1216 mcnt = MAX_NUM_MULTICAST_ADDRESSES; 1217 } else { 1218 struct ifmultiaddr *ifma; 1219 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1220 if (ifma->ifma_addr->sa_family != AF_LINK) 1221 continue; 1222 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) 1223 break; 1224 mcnt++; 1225 } 1226 } 1227 /* Don't disable if in MAX groups */ 1228 if (mcnt < MAX_NUM_MULTICAST_ADDRESSES) 1229 reg &= ~E1000_RCTL_MPE; 1230 E1000_WRITE_REG(hw, E1000_RCTL, reg); 1231 } 1232 1233 static void 1234 igb_set_multi(struct igb_softc *sc) 1235 { 1236 struct ifnet *ifp = &sc->arpcom.ac_if; 1237 struct ifmultiaddr *ifma; 1238 uint32_t reg_rctl = 0; 1239 uint8_t *mta; 1240 int mcnt = 0; 1241 1242 mta = sc->mta; 1243 bzero(mta, ETH_ADDR_LEN * MAX_NUM_MULTICAST_ADDRESSES); 1244 1245 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1246 if (ifma->ifma_addr->sa_family != AF_LINK) 1247 continue; 1248 1249 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES) 1250 break; 1251 1252 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 1253 &mta[mcnt * ETH_ADDR_LEN], ETH_ADDR_LEN); 1254 mcnt++; 1255 } 1256 1257 if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) { 1258 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL); 1259 reg_rctl |= E1000_RCTL_MPE; 1260 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl); 1261 } else { 1262 e1000_update_mc_addr_list(&sc->hw, mta, mcnt); 1263 } 1264 } 1265 1266 static void 1267 igb_timer(void *xsc) 1268 { 1269 struct igb_softc *sc = xsc; 1270 1271 lwkt_serialize_enter(&sc->main_serialize); 1272 1273 igb_update_link_status(sc); 1274 igb_update_stats_counters(sc); 1275 1276 callout_reset_bycpu(&sc->timer, hz, igb_timer, sc, sc->timer_cpuid); 1277 1278 lwkt_serialize_exit(&sc->main_serialize); 1279 } 1280 1281 static void 1282 igb_update_link_status(struct igb_softc *sc) 1283 { 1284 struct ifnet *ifp = &sc->arpcom.ac_if; 1285 struct e1000_hw *hw = &sc->hw; 1286 uint32_t link_check, thstat, ctrl; 1287 1288 link_check = thstat = ctrl = 0; 1289 1290 /* Get the cached link value or read for real */ 1291 switch (hw->phy.media_type) { 1292 case e1000_media_type_copper: 1293 if (hw->mac.get_link_status) { 1294 /* Do the work to read phy */ 1295 e1000_check_for_link(hw); 1296 link_check = !hw->mac.get_link_status; 1297 } else { 1298 link_check = TRUE; 1299 } 1300 break; 1301 1302 case e1000_media_type_fiber: 1303 e1000_check_for_link(hw); 1304 link_check = E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU; 1305 break; 1306 1307 case e1000_media_type_internal_serdes: 1308 e1000_check_for_link(hw); 1309 link_check = hw->mac.serdes_has_link; 1310 break; 1311 1312 /* VF device is type_unknown */ 1313 case e1000_media_type_unknown: 1314 e1000_check_for_link(hw); 1315 link_check = !hw->mac.get_link_status; 1316 /* Fall thru */ 1317 default: 1318 break; 1319 } 1320 1321 /* Check for thermal downshift or shutdown */ 1322 if (hw->mac.type == e1000_i350) { 1323 thstat = E1000_READ_REG(hw, E1000_THSTAT); 1324 ctrl = E1000_READ_REG(hw, E1000_CTRL_EXT); 1325 } 1326 1327 /* Now we check if a transition has happened */ 1328 if (link_check && sc->link_active == 0) { 1329 e1000_get_speed_and_duplex(hw, 1330 &sc->link_speed, &sc->link_duplex); 1331 if (bootverbose) { 1332 char flowctrl[IFM_ETH_FC_STRLEN]; 1333 1334 /* Get the flow control for display */ 1335 e1000_fc2str(hw->fc.current_mode, flowctrl, 1336 sizeof(flowctrl)); 1337 1338 if_printf(ifp, "Link is up %d Mbps %s, " 1339 "Flow control: %s\n", 1340 sc->link_speed, 1341 sc->link_duplex == FULL_DUPLEX ? 1342 "Full Duplex" : "Half Duplex", 1343 flowctrl); 1344 } 1345 if (sc->ifm_flowctrl & IFM_ETH_FORCEPAUSE) 1346 e1000_force_flowctrl(hw, sc->ifm_flowctrl); 1347 sc->link_active = 1; 1348 1349 ifp->if_baudrate = sc->link_speed * 1000000; 1350 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) && 1351 (thstat & E1000_THSTAT_LINK_THROTTLE)) 1352 if_printf(ifp, "Link: thermal downshift\n"); 1353 /* Delay Link Up for Phy update */ 1354 if ((hw->mac.type == e1000_i210 || 1355 hw->mac.type == e1000_i211) && 1356 hw->phy.id == I210_I_PHY_ID) 1357 msec_delay(IGB_I210_LINK_DELAY); 1358 /* 1359 * Reset if the media type changed. 1360 * Support AutoMediaDetect for Marvell M88 PHY in i354. 1361 */ 1362 if (hw->dev_spec._82575.media_changed) { 1363 hw->dev_spec._82575.media_changed = FALSE; 1364 igb_reset(sc, TRUE); 1365 } 1366 /* This can sleep */ 1367 ifp->if_link_state = LINK_STATE_UP; 1368 if_link_state_change(ifp); 1369 } else if (!link_check && sc->link_active == 1) { 1370 ifp->if_baudrate = sc->link_speed = 0; 1371 sc->link_duplex = 0; 1372 if (bootverbose) 1373 if_printf(ifp, "Link is Down\n"); 1374 if ((ctrl & E1000_CTRL_EXT_LINK_MODE_GMII) && 1375 (thstat & E1000_THSTAT_PWR_DOWN)) 1376 if_printf(ifp, "Link: thermal shutdown\n"); 1377 sc->link_active = 0; 1378 /* This can sleep */ 1379 ifp->if_link_state = LINK_STATE_DOWN; 1380 if_link_state_change(ifp); 1381 } 1382 } 1383 1384 static void 1385 igb_stop(struct igb_softc *sc) 1386 { 1387 struct ifnet *ifp = &sc->arpcom.ac_if; 1388 int i; 1389 1390 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1391 1392 igb_disable_intr(sc); 1393 1394 callout_stop(&sc->timer); 1395 1396 ifp->if_flags &= ~IFF_RUNNING; 1397 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1398 ifsq_clr_oactive(sc->tx_rings[i].ifsq); 1399 ifsq_watchdog_stop(&sc->tx_rings[i].tx_watchdog); 1400 sc->tx_rings[i].tx_flags &= ~IGB_TXFLAG_ENABLED; 1401 } 1402 1403 e1000_reset_hw(&sc->hw); 1404 E1000_WRITE_REG(&sc->hw, E1000_WUC, 0); 1405 1406 e1000_led_off(&sc->hw); 1407 e1000_cleanup_led(&sc->hw); 1408 1409 for (i = 0; i < sc->tx_ring_cnt; ++i) 1410 igb_free_tx_ring(&sc->tx_rings[i]); 1411 for (i = 0; i < sc->rx_ring_cnt; ++i) 1412 igb_free_rx_ring(&sc->rx_rings[i]); 1413 } 1414 1415 static void 1416 igb_reset(struct igb_softc *sc, boolean_t media_reset) 1417 { 1418 struct ifnet *ifp = &sc->arpcom.ac_if; 1419 struct e1000_hw *hw = &sc->hw; 1420 struct e1000_fc_info *fc = &hw->fc; 1421 uint32_t pba = 0; 1422 uint16_t hwm; 1423 1424 /* Let the firmware know the OS is in control */ 1425 igb_get_hw_control(sc); 1426 1427 /* 1428 * Packet Buffer Allocation (PBA) 1429 * Writing PBA sets the receive portion of the buffer 1430 * the remainder is used for the transmit buffer. 1431 */ 1432 switch (hw->mac.type) { 1433 case e1000_82575: 1434 pba = E1000_PBA_32K; 1435 break; 1436 1437 case e1000_82576: 1438 case e1000_vfadapt: 1439 pba = E1000_READ_REG(hw, E1000_RXPBS); 1440 pba &= E1000_RXPBS_SIZE_MASK_82576; 1441 break; 1442 1443 case e1000_82580: 1444 case e1000_i350: 1445 case e1000_i354: 1446 case e1000_vfadapt_i350: 1447 pba = E1000_READ_REG(hw, E1000_RXPBS); 1448 pba = e1000_rxpbs_adjust_82580(pba); 1449 break; 1450 1451 case e1000_i210: 1452 case e1000_i211: 1453 pba = E1000_PBA_34K; 1454 break; 1455 1456 default: 1457 break; 1458 } 1459 1460 /* Special needs in case of Jumbo frames */ 1461 if (hw->mac.type == e1000_82575 && ifp->if_mtu > ETHERMTU) { 1462 uint32_t tx_space, min_tx, min_rx; 1463 1464 pba = E1000_READ_REG(hw, E1000_PBA); 1465 tx_space = pba >> 16; 1466 pba &= 0xffff; 1467 1468 min_tx = (sc->max_frame_size + 1469 sizeof(struct e1000_tx_desc) - ETHER_CRC_LEN) * 2; 1470 min_tx = roundup2(min_tx, 1024); 1471 min_tx >>= 10; 1472 min_rx = sc->max_frame_size; 1473 min_rx = roundup2(min_rx, 1024); 1474 min_rx >>= 10; 1475 if (tx_space < min_tx && (min_tx - tx_space) < pba) { 1476 pba = pba - (min_tx - tx_space); 1477 /* 1478 * if short on rx space, rx wins 1479 * and must trump tx adjustment 1480 */ 1481 if (pba < min_rx) 1482 pba = min_rx; 1483 } 1484 E1000_WRITE_REG(hw, E1000_PBA, pba); 1485 } 1486 1487 /* 1488 * These parameters control the automatic generation (Tx) and 1489 * response (Rx) to Ethernet PAUSE frames. 1490 * - High water mark should allow for at least two frames to be 1491 * received after sending an XOFF. 1492 * - Low water mark works best when it is very near the high water mark. 1493 * This allows the receiver to restart by sending XON when it has 1494 * drained a bit. 1495 */ 1496 hwm = min(((pba << 10) * 9 / 10), 1497 ((pba << 10) - 2 * sc->max_frame_size)); 1498 1499 if (hw->mac.type < e1000_82576) { 1500 fc->high_water = hwm & 0xFFF8; /* 8-byte granularity */ 1501 fc->low_water = fc->high_water - 8; 1502 } else { 1503 fc->high_water = hwm & 0xFFF0; /* 16-byte granularity */ 1504 fc->low_water = fc->high_water - 16; 1505 } 1506 fc->pause_time = IGB_FC_PAUSE_TIME; 1507 fc->send_xon = TRUE; 1508 fc->requested_mode = e1000_ifmedia2fc(sc->ifm_flowctrl); 1509 1510 /* Issue a global reset */ 1511 e1000_reset_hw(hw); 1512 E1000_WRITE_REG(hw, E1000_WUC, 0); 1513 1514 /* Reset for AutoMediaDetect */ 1515 if (media_reset) { 1516 e1000_setup_init_funcs(hw, TRUE); 1517 e1000_get_bus_info(hw); 1518 } 1519 1520 if (e1000_init_hw(hw) < 0) 1521 if_printf(ifp, "Hardware Initialization Failed\n"); 1522 1523 /* Setup DMA Coalescing */ 1524 igb_init_dmac(sc, pba); 1525 1526 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN); 1527 e1000_get_phy_info(hw); 1528 e1000_check_for_link(hw); 1529 } 1530 1531 static void 1532 igb_setup_ifp(struct igb_softc *sc) 1533 { 1534 struct ifnet *ifp = &sc->arpcom.ac_if; 1535 int i; 1536 1537 ifp->if_softc = sc; 1538 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 1539 ifp->if_init = igb_init; 1540 ifp->if_ioctl = igb_ioctl; 1541 ifp->if_start = igb_start; 1542 ifp->if_serialize = igb_serialize; 1543 ifp->if_deserialize = igb_deserialize; 1544 ifp->if_tryserialize = igb_tryserialize; 1545 #ifdef INVARIANTS 1546 ifp->if_serialize_assert = igb_serialize_assert; 1547 #endif 1548 #ifdef IFPOLL_ENABLE 1549 ifp->if_npoll = igb_npoll; 1550 #endif 1551 1552 ifp->if_nmbclusters = sc->rx_ring_cnt * sc->rx_rings[0].num_rx_desc; 1553 1554 ifq_set_maxlen(&ifp->if_snd, sc->tx_rings[0].num_tx_desc - 1); 1555 ifq_set_ready(&ifp->if_snd); 1556 ifq_set_subq_cnt(&ifp->if_snd, sc->tx_ring_cnt); 1557 1558 ifp->if_mapsubq = ifq_mapsubq_mask; 1559 ifq_set_subq_mask(&ifp->if_snd, 0); 1560 1561 ether_ifattach(ifp, sc->hw.mac.addr, NULL); 1562 1563 ifp->if_capabilities = 1564 IFCAP_HWCSUM | IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_TSO; 1565 if (IGB_ENABLE_HWRSS(sc)) 1566 ifp->if_capabilities |= IFCAP_RSS; 1567 ifp->if_capenable = ifp->if_capabilities; 1568 ifp->if_hwassist = IGB_CSUM_FEATURES | CSUM_TSO; 1569 1570 /* 1571 * Tell the upper layer(s) we support long frames 1572 */ 1573 ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header); 1574 1575 /* Setup TX rings and subqueues */ 1576 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1577 struct ifaltq_subque *ifsq = ifq_get_subq(&ifp->if_snd, i); 1578 struct igb_tx_ring *txr = &sc->tx_rings[i]; 1579 1580 ifsq_set_cpuid(ifsq, txr->tx_intr_cpuid); 1581 ifsq_set_priv(ifsq, txr); 1582 ifsq_set_hw_serialize(ifsq, &txr->tx_serialize); 1583 txr->ifsq = ifsq; 1584 1585 ifsq_watchdog_init(&txr->tx_watchdog, ifsq, igb_watchdog); 1586 } 1587 1588 /* 1589 * Specify the media types supported by this adapter and register 1590 * callbacks to update media and link information 1591 */ 1592 if (sc->hw.phy.media_type == e1000_media_type_fiber || 1593 sc->hw.phy.media_type == e1000_media_type_internal_serdes) { 1594 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX | IFM_FDX, 1595 0, NULL); 1596 } else { 1597 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T, 0, NULL); 1598 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 1599 0, NULL); 1600 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX, 0, NULL); 1601 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 1602 0, NULL); 1603 if (sc->hw.phy.type != e1000_phy_ife) { 1604 ifmedia_add(&sc->media, 1605 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL); 1606 } 1607 } 1608 ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL); 1609 ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO | sc->ifm_flowctrl); 1610 } 1611 1612 static void 1613 igb_add_sysctl(struct igb_softc *sc) 1614 { 1615 struct sysctl_ctx_list *ctx; 1616 struct sysctl_oid *tree; 1617 #if defined(IGB_RSS_DEBUG) || defined(IGB_TSS_DEBUG) 1618 char node[32]; 1619 int i; 1620 #endif 1621 1622 ctx = device_get_sysctl_ctx(sc->dev); 1623 tree = device_get_sysctl_tree(sc->dev); 1624 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1625 OID_AUTO, "rxr", CTLFLAG_RD, &sc->rx_ring_cnt, 0, "# of RX rings"); 1626 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1627 OID_AUTO, "rxr_inuse", CTLFLAG_RD, &sc->rx_ring_inuse, 0, 1628 "# of RX rings used"); 1629 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1630 OID_AUTO, "txr", CTLFLAG_RD, &sc->tx_ring_cnt, 0, "# of TX rings"); 1631 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1632 OID_AUTO, "txr_inuse", CTLFLAG_RD, &sc->tx_ring_inuse, 0, 1633 "# of TX rings used"); 1634 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1635 OID_AUTO, "rxd", CTLFLAG_RD, &sc->rx_rings[0].num_rx_desc, 0, 1636 "# of RX descs"); 1637 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1638 OID_AUTO, "txd", CTLFLAG_RD, &sc->tx_rings[0].num_tx_desc, 0, 1639 "# of TX descs"); 1640 1641 #define IGB_ADD_INTR_RATE_SYSCTL(sc, use, name) \ 1642 do { \ 1643 igb_add_intr_rate_sysctl(sc, IGB_INTR_USE_##use, #name "_intr_rate", \ 1644 #use " interrupt rate"); \ 1645 } while (0) 1646 1647 IGB_ADD_INTR_RATE_SYSCTL(sc, RXTX, rxtx); 1648 IGB_ADD_INTR_RATE_SYSCTL(sc, RX, rx); 1649 IGB_ADD_INTR_RATE_SYSCTL(sc, TX, tx); 1650 IGB_ADD_INTR_RATE_SYSCTL(sc, STATUS, sts); 1651 1652 #undef IGB_ADD_INTR_RATE_SYSCTL 1653 1654 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), 1655 OID_AUTO, "tx_intr_nsegs", CTLTYPE_INT | CTLFLAG_RW, 1656 sc, 0, igb_sysctl_tx_intr_nsegs, "I", 1657 "# of segments per TX interrupt"); 1658 1659 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), 1660 OID_AUTO, "tx_wreg_nsegs", CTLTYPE_INT | CTLFLAG_RW, 1661 sc, 0, igb_sysctl_tx_wreg_nsegs, "I", 1662 "# of segments sent before write to hardware register"); 1663 1664 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), 1665 OID_AUTO, "rx_wreg_nsegs", CTLTYPE_INT | CTLFLAG_RW, 1666 sc, 0, igb_sysctl_rx_wreg_nsegs, "I", 1667 "# of segments received before write to hardware register"); 1668 1669 #ifdef IFPOLL_ENABLE 1670 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), 1671 OID_AUTO, "npoll_rxoff", CTLTYPE_INT|CTLFLAG_RW, 1672 sc, 0, igb_sysctl_npoll_rxoff, "I", "NPOLLING RX cpu offset"); 1673 SYSCTL_ADD_PROC(ctx, SYSCTL_CHILDREN(tree), 1674 OID_AUTO, "npoll_txoff", CTLTYPE_INT|CTLFLAG_RW, 1675 sc, 0, igb_sysctl_npoll_txoff, "I", "NPOLLING TX cpu offset"); 1676 #endif 1677 1678 #ifdef IGB_RSS_DEBUG 1679 SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(tree), 1680 OID_AUTO, "rss_debug", CTLFLAG_RW, &sc->rss_debug, 0, 1681 "RSS debug level"); 1682 for (i = 0; i < sc->rx_ring_cnt; ++i) { 1683 ksnprintf(node, sizeof(node), "rx%d_pkt", i); 1684 SYSCTL_ADD_ULONG(ctx, 1685 SYSCTL_CHILDREN(tree), OID_AUTO, node, 1686 CTLFLAG_RW, &sc->rx_rings[i].rx_packets, "RXed packets"); 1687 } 1688 #endif 1689 #ifdef IGB_TSS_DEBUG 1690 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1691 ksnprintf(node, sizeof(node), "tx%d_pkt", i); 1692 SYSCTL_ADD_ULONG(ctx, 1693 SYSCTL_CHILDREN(tree), OID_AUTO, node, 1694 CTLFLAG_RW, &sc->tx_rings[i].tx_packets, "TXed packets"); 1695 } 1696 #endif 1697 } 1698 1699 static int 1700 igb_alloc_rings(struct igb_softc *sc) 1701 { 1702 int error, i; 1703 1704 /* 1705 * Create top level busdma tag 1706 */ 1707 error = bus_dma_tag_create(NULL, 1, 0, 1708 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, 1709 BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 0, 1710 &sc->parent_tag); 1711 if (error) { 1712 device_printf(sc->dev, "could not create top level DMA tag\n"); 1713 return error; 1714 } 1715 1716 /* 1717 * Allocate TX descriptor rings and buffers 1718 */ 1719 sc->tx_rings = kmalloc_cachealign( 1720 sizeof(struct igb_tx_ring) * sc->tx_ring_cnt, 1721 M_DEVBUF, M_WAITOK | M_ZERO); 1722 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1723 struct igb_tx_ring *txr = &sc->tx_rings[i]; 1724 1725 /* Set up some basics */ 1726 txr->sc = sc; 1727 txr->me = i; 1728 lwkt_serialize_init(&txr->tx_serialize); 1729 1730 error = igb_create_tx_ring(txr); 1731 if (error) 1732 return error; 1733 } 1734 1735 /* 1736 * Allocate RX descriptor rings and buffers 1737 */ 1738 sc->rx_rings = kmalloc_cachealign( 1739 sizeof(struct igb_rx_ring) * sc->rx_ring_cnt, 1740 M_DEVBUF, M_WAITOK | M_ZERO); 1741 for (i = 0; i < sc->rx_ring_cnt; ++i) { 1742 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 1743 1744 /* Set up some basics */ 1745 rxr->sc = sc; 1746 rxr->me = i; 1747 lwkt_serialize_init(&rxr->rx_serialize); 1748 1749 error = igb_create_rx_ring(rxr); 1750 if (error) 1751 return error; 1752 } 1753 1754 return 0; 1755 } 1756 1757 static void 1758 igb_free_rings(struct igb_softc *sc) 1759 { 1760 int i; 1761 1762 if (sc->tx_rings != NULL) { 1763 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1764 struct igb_tx_ring *txr = &sc->tx_rings[i]; 1765 1766 igb_destroy_tx_ring(txr, txr->num_tx_desc); 1767 } 1768 kfree(sc->tx_rings, M_DEVBUF); 1769 } 1770 1771 if (sc->rx_rings != NULL) { 1772 for (i = 0; i < sc->rx_ring_cnt; ++i) { 1773 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 1774 1775 igb_destroy_rx_ring(rxr, rxr->num_rx_desc); 1776 } 1777 kfree(sc->rx_rings, M_DEVBUF); 1778 } 1779 } 1780 1781 static int 1782 igb_create_tx_ring(struct igb_tx_ring *txr) 1783 { 1784 int tsize, error, i, ntxd; 1785 1786 /* 1787 * Validate number of transmit descriptors. It must not exceed 1788 * hardware maximum, and must be multiple of IGB_DBA_ALIGN. 1789 */ 1790 ntxd = device_getenv_int(txr->sc->dev, "txd", igb_txd); 1791 if ((ntxd * sizeof(struct e1000_tx_desc)) % IGB_DBA_ALIGN != 0 || 1792 ntxd > IGB_MAX_TXD || ntxd < IGB_MIN_TXD) { 1793 device_printf(txr->sc->dev, 1794 "Using %d TX descriptors instead of %d!\n", 1795 IGB_DEFAULT_TXD, ntxd); 1796 txr->num_tx_desc = IGB_DEFAULT_TXD; 1797 } else { 1798 txr->num_tx_desc = ntxd; 1799 } 1800 1801 /* 1802 * Allocate TX descriptor ring 1803 */ 1804 tsize = roundup2(txr->num_tx_desc * sizeof(union e1000_adv_tx_desc), 1805 IGB_DBA_ALIGN); 1806 txr->txdma.dma_vaddr = bus_dmamem_coherent_any(txr->sc->parent_tag, 1807 IGB_DBA_ALIGN, tsize, BUS_DMA_WAITOK, 1808 &txr->txdma.dma_tag, &txr->txdma.dma_map, &txr->txdma.dma_paddr); 1809 if (txr->txdma.dma_vaddr == NULL) { 1810 device_printf(txr->sc->dev, 1811 "Unable to allocate TX Descriptor memory\n"); 1812 return ENOMEM; 1813 } 1814 txr->tx_base = txr->txdma.dma_vaddr; 1815 bzero(txr->tx_base, tsize); 1816 1817 tsize = __VM_CACHELINE_ALIGN( 1818 sizeof(struct igb_tx_buf) * txr->num_tx_desc); 1819 txr->tx_buf = kmalloc_cachealign(tsize, M_DEVBUF, M_WAITOK | M_ZERO); 1820 1821 /* 1822 * Allocate TX head write-back buffer 1823 */ 1824 txr->tx_hdr = bus_dmamem_coherent_any(txr->sc->parent_tag, 1825 __VM_CACHELINE_SIZE, __VM_CACHELINE_SIZE, BUS_DMA_WAITOK, 1826 &txr->tx_hdr_dtag, &txr->tx_hdr_dmap, &txr->tx_hdr_paddr); 1827 if (txr->tx_hdr == NULL) { 1828 device_printf(txr->sc->dev, 1829 "Unable to allocate TX head write-back buffer\n"); 1830 return ENOMEM; 1831 } 1832 1833 /* 1834 * Create DMA tag for TX buffers 1835 */ 1836 error = bus_dma_tag_create(txr->sc->parent_tag, 1837 1, 0, /* alignment, bounds */ 1838 BUS_SPACE_MAXADDR, /* lowaddr */ 1839 BUS_SPACE_MAXADDR, /* highaddr */ 1840 NULL, NULL, /* filter, filterarg */ 1841 IGB_TSO_SIZE, /* maxsize */ 1842 IGB_MAX_SCATTER, /* nsegments */ 1843 PAGE_SIZE, /* maxsegsize */ 1844 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW | 1845 BUS_DMA_ONEBPAGE, /* flags */ 1846 &txr->tx_tag); 1847 if (error) { 1848 device_printf(txr->sc->dev, "Unable to allocate TX DMA tag\n"); 1849 kfree(txr->tx_buf, M_DEVBUF); 1850 txr->tx_buf = NULL; 1851 return error; 1852 } 1853 1854 /* 1855 * Create DMA maps for TX buffers 1856 */ 1857 for (i = 0; i < txr->num_tx_desc; ++i) { 1858 struct igb_tx_buf *txbuf = &txr->tx_buf[i]; 1859 1860 error = bus_dmamap_create(txr->tx_tag, 1861 BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE, &txbuf->map); 1862 if (error) { 1863 device_printf(txr->sc->dev, 1864 "Unable to create TX DMA map\n"); 1865 igb_destroy_tx_ring(txr, i); 1866 return error; 1867 } 1868 } 1869 1870 if (txr->sc->hw.mac.type == e1000_82575) 1871 txr->tx_flags |= IGB_TXFLAG_TSO_IPLEN0; 1872 1873 /* 1874 * Initialize various watermark 1875 */ 1876 txr->intr_nsegs = txr->num_tx_desc / 16; 1877 txr->wreg_nsegs = IGB_DEF_TXWREG_NSEGS; 1878 1879 return 0; 1880 } 1881 1882 static void 1883 igb_free_tx_ring(struct igb_tx_ring *txr) 1884 { 1885 int i; 1886 1887 for (i = 0; i < txr->num_tx_desc; ++i) { 1888 struct igb_tx_buf *txbuf = &txr->tx_buf[i]; 1889 1890 if (txbuf->m_head != NULL) { 1891 bus_dmamap_unload(txr->tx_tag, txbuf->map); 1892 m_freem(txbuf->m_head); 1893 txbuf->m_head = NULL; 1894 } 1895 } 1896 } 1897 1898 static void 1899 igb_destroy_tx_ring(struct igb_tx_ring *txr, int ndesc) 1900 { 1901 int i; 1902 1903 if (txr->txdma.dma_vaddr != NULL) { 1904 bus_dmamap_unload(txr->txdma.dma_tag, txr->txdma.dma_map); 1905 bus_dmamem_free(txr->txdma.dma_tag, txr->txdma.dma_vaddr, 1906 txr->txdma.dma_map); 1907 bus_dma_tag_destroy(txr->txdma.dma_tag); 1908 txr->txdma.dma_vaddr = NULL; 1909 } 1910 1911 if (txr->tx_hdr != NULL) { 1912 bus_dmamap_unload(txr->tx_hdr_dtag, txr->tx_hdr_dmap); 1913 bus_dmamem_free(txr->tx_hdr_dtag, txr->tx_hdr, 1914 txr->tx_hdr_dmap); 1915 bus_dma_tag_destroy(txr->tx_hdr_dtag); 1916 txr->tx_hdr = NULL; 1917 } 1918 1919 if (txr->tx_buf == NULL) 1920 return; 1921 1922 for (i = 0; i < ndesc; ++i) { 1923 struct igb_tx_buf *txbuf = &txr->tx_buf[i]; 1924 1925 KKASSERT(txbuf->m_head == NULL); 1926 bus_dmamap_destroy(txr->tx_tag, txbuf->map); 1927 } 1928 bus_dma_tag_destroy(txr->tx_tag); 1929 1930 kfree(txr->tx_buf, M_DEVBUF); 1931 txr->tx_buf = NULL; 1932 } 1933 1934 static void 1935 igb_init_tx_ring(struct igb_tx_ring *txr) 1936 { 1937 /* Clear the old descriptor contents */ 1938 bzero(txr->tx_base, 1939 sizeof(union e1000_adv_tx_desc) * txr->num_tx_desc); 1940 1941 /* Clear TX head write-back buffer */ 1942 *(txr->tx_hdr) = 0; 1943 1944 /* Reset indices */ 1945 txr->next_avail_desc = 0; 1946 txr->next_to_clean = 0; 1947 txr->tx_nsegs = 0; 1948 1949 /* Set number of descriptors available */ 1950 txr->tx_avail = txr->num_tx_desc; 1951 1952 /* Enable this TX ring */ 1953 txr->tx_flags |= IGB_TXFLAG_ENABLED; 1954 } 1955 1956 static void 1957 igb_init_tx_unit(struct igb_softc *sc) 1958 { 1959 struct e1000_hw *hw = &sc->hw; 1960 uint32_t tctl; 1961 int i; 1962 1963 /* Setup the Tx Descriptor Rings */ 1964 for (i = 0; i < sc->tx_ring_inuse; ++i) { 1965 struct igb_tx_ring *txr = &sc->tx_rings[i]; 1966 uint64_t bus_addr = txr->txdma.dma_paddr; 1967 uint64_t hdr_paddr = txr->tx_hdr_paddr; 1968 uint32_t txdctl = 0; 1969 uint32_t dca_txctrl; 1970 1971 E1000_WRITE_REG(hw, E1000_TDLEN(i), 1972 txr->num_tx_desc * sizeof(struct e1000_tx_desc)); 1973 E1000_WRITE_REG(hw, E1000_TDBAH(i), 1974 (uint32_t)(bus_addr >> 32)); 1975 E1000_WRITE_REG(hw, E1000_TDBAL(i), 1976 (uint32_t)bus_addr); 1977 1978 /* Setup the HW Tx Head and Tail descriptor pointers */ 1979 E1000_WRITE_REG(hw, E1000_TDT(i), 0); 1980 E1000_WRITE_REG(hw, E1000_TDH(i), 0); 1981 1982 dca_txctrl = E1000_READ_REG(hw, E1000_DCA_TXCTRL(i)); 1983 dca_txctrl &= ~E1000_DCA_TXCTRL_TX_WB_RO_EN; 1984 E1000_WRITE_REG(hw, E1000_DCA_TXCTRL(i), dca_txctrl); 1985 1986 /* 1987 * Don't set WB_on_EITR: 1988 * - 82575 does not have it 1989 * - It almost has no effect on 82576, see: 1990 * 82576 specification update errata #26 1991 * - It causes unnecessary bus traffic 1992 */ 1993 E1000_WRITE_REG(hw, E1000_TDWBAH(i), 1994 (uint32_t)(hdr_paddr >> 32)); 1995 E1000_WRITE_REG(hw, E1000_TDWBAL(i), 1996 ((uint32_t)hdr_paddr) | E1000_TX_HEAD_WB_ENABLE); 1997 1998 /* 1999 * WTHRESH is ignored by the hardware, since header 2000 * write back mode is used. 2001 */ 2002 txdctl |= IGB_TX_PTHRESH; 2003 txdctl |= IGB_TX_HTHRESH << 8; 2004 txdctl |= IGB_TX_WTHRESH << 16; 2005 txdctl |= E1000_TXDCTL_QUEUE_ENABLE; 2006 E1000_WRITE_REG(hw, E1000_TXDCTL(i), txdctl); 2007 } 2008 2009 if (sc->vf_ifp) 2010 return; 2011 2012 e1000_config_collision_dist(hw); 2013 2014 /* Program the Transmit Control Register */ 2015 tctl = E1000_READ_REG(hw, E1000_TCTL); 2016 tctl &= ~E1000_TCTL_CT; 2017 tctl |= (E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN | 2018 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT)); 2019 2020 /* This write will effectively turn on the transmit unit. */ 2021 E1000_WRITE_REG(hw, E1000_TCTL, tctl); 2022 } 2023 2024 static boolean_t 2025 igb_txcsum_ctx(struct igb_tx_ring *txr, struct mbuf *mp) 2026 { 2027 struct e1000_adv_tx_context_desc *TXD; 2028 uint32_t vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx; 2029 int ehdrlen, ctxd, ip_hlen = 0; 2030 boolean_t offload = TRUE; 2031 2032 if ((mp->m_pkthdr.csum_flags & IGB_CSUM_FEATURES) == 0) 2033 offload = FALSE; 2034 2035 vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0; 2036 2037 ctxd = txr->next_avail_desc; 2038 TXD = (struct e1000_adv_tx_context_desc *)&txr->tx_base[ctxd]; 2039 2040 /* 2041 * In advanced descriptors the vlan tag must 2042 * be placed into the context descriptor, thus 2043 * we need to be here just for that setup. 2044 */ 2045 if (mp->m_flags & M_VLANTAG) { 2046 uint16_t vlantag; 2047 2048 vlantag = htole16(mp->m_pkthdr.ether_vlantag); 2049 vlan_macip_lens |= (vlantag << E1000_ADVTXD_VLAN_SHIFT); 2050 } else if (!offload) { 2051 return FALSE; 2052 } 2053 2054 ehdrlen = mp->m_pkthdr.csum_lhlen; 2055 KASSERT(ehdrlen > 0, ("invalid ether hlen")); 2056 2057 /* Set the ether header length */ 2058 vlan_macip_lens |= ehdrlen << E1000_ADVTXD_MACLEN_SHIFT; 2059 if (mp->m_pkthdr.csum_flags & CSUM_IP) { 2060 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4; 2061 ip_hlen = mp->m_pkthdr.csum_iphlen; 2062 KASSERT(ip_hlen > 0, ("invalid ip hlen")); 2063 } 2064 vlan_macip_lens |= ip_hlen; 2065 2066 type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT; 2067 if (mp->m_pkthdr.csum_flags & CSUM_TCP) 2068 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP; 2069 else if (mp->m_pkthdr.csum_flags & CSUM_UDP) 2070 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_UDP; 2071 2072 /* 2073 * 82575 needs the TX context index added; the queue 2074 * index is used as TX context index here. 2075 */ 2076 if (txr->sc->hw.mac.type == e1000_82575) 2077 mss_l4len_idx = txr->me << 4; 2078 2079 /* Now copy bits into descriptor */ 2080 TXD->vlan_macip_lens = htole32(vlan_macip_lens); 2081 TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl); 2082 TXD->seqnum_seed = htole32(0); 2083 TXD->mss_l4len_idx = htole32(mss_l4len_idx); 2084 2085 /* We've consumed the first desc, adjust counters */ 2086 if (++ctxd == txr->num_tx_desc) 2087 ctxd = 0; 2088 txr->next_avail_desc = ctxd; 2089 --txr->tx_avail; 2090 2091 return offload; 2092 } 2093 2094 static void 2095 igb_txeof(struct igb_tx_ring *txr, int hdr) 2096 { 2097 int first, avail; 2098 2099 if (txr->tx_avail == txr->num_tx_desc) 2100 return; 2101 2102 first = txr->next_to_clean; 2103 if (first == hdr) 2104 return; 2105 2106 avail = txr->tx_avail; 2107 while (first != hdr) { 2108 struct igb_tx_buf *txbuf = &txr->tx_buf[first]; 2109 2110 ++avail; 2111 if (txbuf->m_head) { 2112 bus_dmamap_unload(txr->tx_tag, txbuf->map); 2113 m_freem(txbuf->m_head); 2114 txbuf->m_head = NULL; 2115 } 2116 if (++first == txr->num_tx_desc) 2117 first = 0; 2118 } 2119 txr->next_to_clean = first; 2120 txr->tx_avail = avail; 2121 2122 /* 2123 * If we have a minimum free, clear OACTIVE 2124 * to tell the stack that it is OK to send packets. 2125 */ 2126 if (txr->tx_avail > IGB_MAX_SCATTER + IGB_TX_RESERVED) { 2127 ifsq_clr_oactive(txr->ifsq); 2128 2129 /* 2130 * We have enough TX descriptors, turn off 2131 * the watchdog. We allow small amount of 2132 * packets (roughly intr_nsegs) pending on 2133 * the transmit ring. 2134 */ 2135 txr->tx_watchdog.wd_timer = 0; 2136 } 2137 } 2138 2139 static int 2140 igb_create_rx_ring(struct igb_rx_ring *rxr) 2141 { 2142 int rsize, i, error, nrxd; 2143 2144 /* 2145 * Validate number of receive descriptors. It must not exceed 2146 * hardware maximum, and must be multiple of IGB_DBA_ALIGN. 2147 */ 2148 nrxd = device_getenv_int(rxr->sc->dev, "rxd", igb_rxd); 2149 if ((nrxd * sizeof(struct e1000_rx_desc)) % IGB_DBA_ALIGN != 0 || 2150 nrxd > IGB_MAX_RXD || nrxd < IGB_MIN_RXD) { 2151 device_printf(rxr->sc->dev, 2152 "Using %d RX descriptors instead of %d!\n", 2153 IGB_DEFAULT_RXD, nrxd); 2154 rxr->num_rx_desc = IGB_DEFAULT_RXD; 2155 } else { 2156 rxr->num_rx_desc = nrxd; 2157 } 2158 2159 /* 2160 * Allocate RX descriptor ring 2161 */ 2162 rsize = roundup2(rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc), 2163 IGB_DBA_ALIGN); 2164 rxr->rxdma.dma_vaddr = bus_dmamem_coherent_any(rxr->sc->parent_tag, 2165 IGB_DBA_ALIGN, rsize, BUS_DMA_WAITOK, 2166 &rxr->rxdma.dma_tag, &rxr->rxdma.dma_map, 2167 &rxr->rxdma.dma_paddr); 2168 if (rxr->rxdma.dma_vaddr == NULL) { 2169 device_printf(rxr->sc->dev, 2170 "Unable to allocate RxDescriptor memory\n"); 2171 return ENOMEM; 2172 } 2173 rxr->rx_base = rxr->rxdma.dma_vaddr; 2174 bzero(rxr->rx_base, rsize); 2175 2176 rsize = __VM_CACHELINE_ALIGN( 2177 sizeof(struct igb_rx_buf) * rxr->num_rx_desc); 2178 rxr->rx_buf = kmalloc_cachealign(rsize, M_DEVBUF, M_WAITOK | M_ZERO); 2179 2180 /* 2181 * Create DMA tag for RX buffers 2182 */ 2183 error = bus_dma_tag_create(rxr->sc->parent_tag, 2184 1, 0, /* alignment, bounds */ 2185 BUS_SPACE_MAXADDR, /* lowaddr */ 2186 BUS_SPACE_MAXADDR, /* highaddr */ 2187 NULL, NULL, /* filter, filterarg */ 2188 MCLBYTES, /* maxsize */ 2189 1, /* nsegments */ 2190 MCLBYTES, /* maxsegsize */ 2191 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW, /* flags */ 2192 &rxr->rx_tag); 2193 if (error) { 2194 device_printf(rxr->sc->dev, 2195 "Unable to create RX payload DMA tag\n"); 2196 kfree(rxr->rx_buf, M_DEVBUF); 2197 rxr->rx_buf = NULL; 2198 return error; 2199 } 2200 2201 /* 2202 * Create spare DMA map for RX buffers 2203 */ 2204 error = bus_dmamap_create(rxr->rx_tag, BUS_DMA_WAITOK, 2205 &rxr->rx_sparemap); 2206 if (error) { 2207 device_printf(rxr->sc->dev, 2208 "Unable to create spare RX DMA maps\n"); 2209 bus_dma_tag_destroy(rxr->rx_tag); 2210 kfree(rxr->rx_buf, M_DEVBUF); 2211 rxr->rx_buf = NULL; 2212 return error; 2213 } 2214 2215 /* 2216 * Create DMA maps for RX buffers 2217 */ 2218 for (i = 0; i < rxr->num_rx_desc; i++) { 2219 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i]; 2220 2221 error = bus_dmamap_create(rxr->rx_tag, 2222 BUS_DMA_WAITOK, &rxbuf->map); 2223 if (error) { 2224 device_printf(rxr->sc->dev, 2225 "Unable to create RX DMA maps\n"); 2226 igb_destroy_rx_ring(rxr, i); 2227 return error; 2228 } 2229 } 2230 2231 /* 2232 * Initialize various watermark 2233 */ 2234 rxr->wreg_nsegs = IGB_DEF_RXWREG_NSEGS; 2235 2236 return 0; 2237 } 2238 2239 static void 2240 igb_free_rx_ring(struct igb_rx_ring *rxr) 2241 { 2242 int i; 2243 2244 for (i = 0; i < rxr->num_rx_desc; ++i) { 2245 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i]; 2246 2247 if (rxbuf->m_head != NULL) { 2248 bus_dmamap_unload(rxr->rx_tag, rxbuf->map); 2249 m_freem(rxbuf->m_head); 2250 rxbuf->m_head = NULL; 2251 } 2252 } 2253 2254 if (rxr->fmp != NULL) 2255 m_freem(rxr->fmp); 2256 rxr->fmp = NULL; 2257 rxr->lmp = NULL; 2258 } 2259 2260 static void 2261 igb_destroy_rx_ring(struct igb_rx_ring *rxr, int ndesc) 2262 { 2263 int i; 2264 2265 if (rxr->rxdma.dma_vaddr != NULL) { 2266 bus_dmamap_unload(rxr->rxdma.dma_tag, rxr->rxdma.dma_map); 2267 bus_dmamem_free(rxr->rxdma.dma_tag, rxr->rxdma.dma_vaddr, 2268 rxr->rxdma.dma_map); 2269 bus_dma_tag_destroy(rxr->rxdma.dma_tag); 2270 rxr->rxdma.dma_vaddr = NULL; 2271 } 2272 2273 if (rxr->rx_buf == NULL) 2274 return; 2275 2276 for (i = 0; i < ndesc; ++i) { 2277 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i]; 2278 2279 KKASSERT(rxbuf->m_head == NULL); 2280 bus_dmamap_destroy(rxr->rx_tag, rxbuf->map); 2281 } 2282 bus_dmamap_destroy(rxr->rx_tag, rxr->rx_sparemap); 2283 bus_dma_tag_destroy(rxr->rx_tag); 2284 2285 kfree(rxr->rx_buf, M_DEVBUF); 2286 rxr->rx_buf = NULL; 2287 } 2288 2289 static void 2290 igb_setup_rxdesc(union e1000_adv_rx_desc *rxd, const struct igb_rx_buf *rxbuf) 2291 { 2292 rxd->read.pkt_addr = htole64(rxbuf->paddr); 2293 rxd->wb.upper.status_error = 0; 2294 } 2295 2296 static int 2297 igb_newbuf(struct igb_rx_ring *rxr, int i, boolean_t wait) 2298 { 2299 struct mbuf *m; 2300 bus_dma_segment_t seg; 2301 bus_dmamap_t map; 2302 struct igb_rx_buf *rxbuf; 2303 int error, nseg; 2304 2305 m = m_getcl(wait ? M_WAITOK : M_NOWAIT, MT_DATA, M_PKTHDR); 2306 if (m == NULL) { 2307 if (wait) { 2308 if_printf(&rxr->sc->arpcom.ac_if, 2309 "Unable to allocate RX mbuf\n"); 2310 } 2311 return ENOBUFS; 2312 } 2313 m->m_len = m->m_pkthdr.len = MCLBYTES; 2314 2315 if (rxr->sc->max_frame_size <= MCLBYTES - ETHER_ALIGN) 2316 m_adj(m, ETHER_ALIGN); 2317 2318 error = bus_dmamap_load_mbuf_segment(rxr->rx_tag, 2319 rxr->rx_sparemap, m, &seg, 1, &nseg, BUS_DMA_NOWAIT); 2320 if (error) { 2321 m_freem(m); 2322 if (wait) { 2323 if_printf(&rxr->sc->arpcom.ac_if, 2324 "Unable to load RX mbuf\n"); 2325 } 2326 return error; 2327 } 2328 2329 rxbuf = &rxr->rx_buf[i]; 2330 if (rxbuf->m_head != NULL) 2331 bus_dmamap_unload(rxr->rx_tag, rxbuf->map); 2332 2333 map = rxbuf->map; 2334 rxbuf->map = rxr->rx_sparemap; 2335 rxr->rx_sparemap = map; 2336 2337 rxbuf->m_head = m; 2338 rxbuf->paddr = seg.ds_addr; 2339 2340 igb_setup_rxdesc(&rxr->rx_base[i], rxbuf); 2341 return 0; 2342 } 2343 2344 static int 2345 igb_init_rx_ring(struct igb_rx_ring *rxr) 2346 { 2347 int i; 2348 2349 /* Clear the ring contents */ 2350 bzero(rxr->rx_base, 2351 rxr->num_rx_desc * sizeof(union e1000_adv_rx_desc)); 2352 2353 /* Now replenish the ring mbufs */ 2354 for (i = 0; i < rxr->num_rx_desc; ++i) { 2355 int error; 2356 2357 error = igb_newbuf(rxr, i, TRUE); 2358 if (error) 2359 return error; 2360 } 2361 2362 /* Setup our descriptor indices */ 2363 rxr->next_to_check = 0; 2364 2365 rxr->fmp = NULL; 2366 rxr->lmp = NULL; 2367 rxr->discard = FALSE; 2368 2369 return 0; 2370 } 2371 2372 static void 2373 igb_init_rx_unit(struct igb_softc *sc) 2374 { 2375 struct ifnet *ifp = &sc->arpcom.ac_if; 2376 struct e1000_hw *hw = &sc->hw; 2377 uint32_t rctl, rxcsum, srrctl = 0; 2378 int i; 2379 2380 /* 2381 * Make sure receives are disabled while setting 2382 * up the descriptor ring 2383 */ 2384 rctl = E1000_READ_REG(hw, E1000_RCTL); 2385 E1000_WRITE_REG(hw, E1000_RCTL, rctl & ~E1000_RCTL_EN); 2386 2387 #if 0 2388 /* 2389 ** Set up for header split 2390 */ 2391 if (igb_header_split) { 2392 /* Use a standard mbuf for the header */ 2393 srrctl |= IGB_HDR_BUF << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT; 2394 srrctl |= E1000_SRRCTL_DESCTYPE_HDR_SPLIT_ALWAYS; 2395 } else 2396 #endif 2397 srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF; 2398 2399 /* 2400 ** Set up for jumbo frames 2401 */ 2402 if (ifp->if_mtu > ETHERMTU) { 2403 rctl |= E1000_RCTL_LPE; 2404 #if 0 2405 if (adapter->rx_mbuf_sz == MJUMPAGESIZE) { 2406 srrctl |= 4096 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 2407 rctl |= E1000_RCTL_SZ_4096 | E1000_RCTL_BSEX; 2408 } else if (adapter->rx_mbuf_sz > MJUMPAGESIZE) { 2409 srrctl |= 8192 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 2410 rctl |= E1000_RCTL_SZ_8192 | E1000_RCTL_BSEX; 2411 } 2412 /* Set maximum packet len */ 2413 psize = adapter->max_frame_size; 2414 /* are we on a vlan? */ 2415 if (adapter->ifp->if_vlantrunk != NULL) 2416 psize += VLAN_TAG_SIZE; 2417 E1000_WRITE_REG(&adapter->hw, E1000_RLPML, psize); 2418 #else 2419 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 2420 rctl |= E1000_RCTL_SZ_2048; 2421 #endif 2422 } else { 2423 rctl &= ~E1000_RCTL_LPE; 2424 srrctl |= 2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT; 2425 rctl |= E1000_RCTL_SZ_2048; 2426 } 2427 2428 /* Setup the Base and Length of the Rx Descriptor Rings */ 2429 for (i = 0; i < sc->rx_ring_inuse; ++i) { 2430 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 2431 uint64_t bus_addr = rxr->rxdma.dma_paddr; 2432 uint32_t rxdctl; 2433 2434 E1000_WRITE_REG(hw, E1000_RDLEN(i), 2435 rxr->num_rx_desc * sizeof(struct e1000_rx_desc)); 2436 E1000_WRITE_REG(hw, E1000_RDBAH(i), 2437 (uint32_t)(bus_addr >> 32)); 2438 E1000_WRITE_REG(hw, E1000_RDBAL(i), 2439 (uint32_t)bus_addr); 2440 E1000_WRITE_REG(hw, E1000_SRRCTL(i), srrctl); 2441 /* Enable this Queue */ 2442 rxdctl = E1000_READ_REG(hw, E1000_RXDCTL(i)); 2443 rxdctl |= E1000_RXDCTL_QUEUE_ENABLE; 2444 rxdctl &= 0xFFF00000; 2445 rxdctl |= IGB_RX_PTHRESH; 2446 rxdctl |= IGB_RX_HTHRESH << 8; 2447 /* 2448 * Don't set WTHRESH to a value above 1 on 82576, see: 2449 * 82576 specification update errata #26 2450 */ 2451 rxdctl |= IGB_RX_WTHRESH << 16; 2452 E1000_WRITE_REG(hw, E1000_RXDCTL(i), rxdctl); 2453 } 2454 2455 rxcsum = E1000_READ_REG(&sc->hw, E1000_RXCSUM); 2456 rxcsum &= ~(E1000_RXCSUM_PCSS_MASK | E1000_RXCSUM_IPPCSE); 2457 2458 /* 2459 * Receive Checksum Offload for TCP and UDP 2460 * 2461 * Checksum offloading is also enabled if multiple receive 2462 * queue is to be supported, since we need it to figure out 2463 * fragments. 2464 */ 2465 if ((ifp->if_capenable & IFCAP_RXCSUM) || IGB_ENABLE_HWRSS(sc)) { 2466 /* 2467 * NOTE: 2468 * PCSD must be enabled to enable multiple 2469 * receive queues. 2470 */ 2471 rxcsum |= E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL | 2472 E1000_RXCSUM_PCSD; 2473 } else { 2474 rxcsum &= ~(E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL | 2475 E1000_RXCSUM_PCSD); 2476 } 2477 E1000_WRITE_REG(&sc->hw, E1000_RXCSUM, rxcsum); 2478 2479 if (IGB_ENABLE_HWRSS(sc)) { 2480 uint8_t key[IGB_NRSSRK * IGB_RSSRK_SIZE]; 2481 uint32_t reta_shift; 2482 int j, r; 2483 2484 /* 2485 * NOTE: 2486 * When we reach here, RSS has already been disabled 2487 * in igb_stop(), so we could safely configure RSS key 2488 * and redirect table. 2489 */ 2490 2491 /* 2492 * Configure RSS key 2493 */ 2494 toeplitz_get_key(key, sizeof(key)); 2495 for (i = 0; i < IGB_NRSSRK; ++i) { 2496 uint32_t rssrk; 2497 2498 rssrk = IGB_RSSRK_VAL(key, i); 2499 IGB_RSS_DPRINTF(sc, 1, "rssrk%d 0x%08x\n", i, rssrk); 2500 2501 E1000_WRITE_REG(hw, E1000_RSSRK(i), rssrk); 2502 } 2503 2504 /* 2505 * Configure RSS redirect table in following fashion: 2506 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)] 2507 */ 2508 reta_shift = IGB_RETA_SHIFT; 2509 if (hw->mac.type == e1000_82575) 2510 reta_shift = IGB_RETA_SHIFT_82575; 2511 2512 r = 0; 2513 for (j = 0; j < IGB_NRETA; ++j) { 2514 uint32_t reta = 0; 2515 2516 for (i = 0; i < IGB_RETA_SIZE; ++i) { 2517 uint32_t q; 2518 2519 q = (r % sc->rx_ring_inuse) << reta_shift; 2520 reta |= q << (8 * i); 2521 ++r; 2522 } 2523 IGB_RSS_DPRINTF(sc, 1, "reta 0x%08x\n", reta); 2524 E1000_WRITE_REG(hw, E1000_RETA(j), reta); 2525 } 2526 2527 /* 2528 * Enable multiple receive queues. 2529 * Enable IPv4 RSS standard hash functions. 2530 * Disable RSS interrupt on 82575 2531 */ 2532 E1000_WRITE_REG(&sc->hw, E1000_MRQC, 2533 E1000_MRQC_ENABLE_RSS_4Q | 2534 E1000_MRQC_RSS_FIELD_IPV4_TCP | 2535 E1000_MRQC_RSS_FIELD_IPV4); 2536 } 2537 2538 /* Setup the Receive Control Register */ 2539 rctl &= ~(3 << E1000_RCTL_MO_SHIFT); 2540 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO | 2541 E1000_RCTL_RDMTS_HALF | 2542 (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT); 2543 /* Strip CRC bytes. */ 2544 rctl |= E1000_RCTL_SECRC; 2545 /* Make sure VLAN Filters are off */ 2546 rctl &= ~E1000_RCTL_VFE; 2547 /* Don't store bad packets */ 2548 rctl &= ~E1000_RCTL_SBP; 2549 2550 /* Enable Receives */ 2551 E1000_WRITE_REG(hw, E1000_RCTL, rctl); 2552 2553 /* 2554 * Setup the HW Rx Head and Tail Descriptor Pointers 2555 * - needs to be after enable 2556 */ 2557 for (i = 0; i < sc->rx_ring_inuse; ++i) { 2558 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 2559 2560 E1000_WRITE_REG(hw, E1000_RDH(i), rxr->next_to_check); 2561 E1000_WRITE_REG(hw, E1000_RDT(i), rxr->num_rx_desc - 1); 2562 } 2563 } 2564 2565 static void 2566 igb_rx_refresh(struct igb_rx_ring *rxr, int i) 2567 { 2568 if (--i < 0) 2569 i = rxr->num_rx_desc - 1; 2570 E1000_WRITE_REG(&rxr->sc->hw, E1000_RDT(rxr->me), i); 2571 } 2572 2573 static void 2574 igb_rxeof(struct igb_rx_ring *rxr, int count) 2575 { 2576 struct ifnet *ifp = &rxr->sc->arpcom.ac_if; 2577 union e1000_adv_rx_desc *cur; 2578 uint32_t staterr; 2579 int i, ncoll = 0, cpuid = mycpuid; 2580 2581 i = rxr->next_to_check; 2582 cur = &rxr->rx_base[i]; 2583 staterr = le32toh(cur->wb.upper.status_error); 2584 2585 if ((staterr & E1000_RXD_STAT_DD) == 0) 2586 return; 2587 2588 while ((staterr & E1000_RXD_STAT_DD) && count != 0) { 2589 struct pktinfo *pi = NULL, pi0; 2590 struct igb_rx_buf *rxbuf = &rxr->rx_buf[i]; 2591 struct mbuf *m = NULL; 2592 boolean_t eop; 2593 2594 eop = (staterr & E1000_RXD_STAT_EOP) ? TRUE : FALSE; 2595 if (eop) 2596 --count; 2597 2598 ++ncoll; 2599 if ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) == 0 && 2600 !rxr->discard) { 2601 struct mbuf *mp = rxbuf->m_head; 2602 uint32_t hash, hashtype; 2603 uint16_t vlan; 2604 int len; 2605 2606 len = le16toh(cur->wb.upper.length); 2607 if ((rxr->sc->hw.mac.type == e1000_i350 || 2608 rxr->sc->hw.mac.type == e1000_i354) && 2609 (staterr & E1000_RXDEXT_STATERR_LB)) 2610 vlan = be16toh(cur->wb.upper.vlan); 2611 else 2612 vlan = le16toh(cur->wb.upper.vlan); 2613 2614 hash = le32toh(cur->wb.lower.hi_dword.rss); 2615 hashtype = le32toh(cur->wb.lower.lo_dword.data) & 2616 E1000_RXDADV_RSSTYPE_MASK; 2617 2618 IGB_RSS_DPRINTF(rxr->sc, 10, 2619 "ring%d, hash 0x%08x, hashtype %u\n", 2620 rxr->me, hash, hashtype); 2621 2622 bus_dmamap_sync(rxr->rx_tag, rxbuf->map, 2623 BUS_DMASYNC_POSTREAD); 2624 2625 if (igb_newbuf(rxr, i, FALSE) != 0) { 2626 IFNET_STAT_INC(ifp, iqdrops, 1); 2627 goto discard; 2628 } 2629 2630 mp->m_len = len; 2631 if (rxr->fmp == NULL) { 2632 mp->m_pkthdr.len = len; 2633 rxr->fmp = mp; 2634 rxr->lmp = mp; 2635 } else { 2636 rxr->lmp->m_next = mp; 2637 rxr->lmp = rxr->lmp->m_next; 2638 rxr->fmp->m_pkthdr.len += len; 2639 } 2640 2641 if (eop) { 2642 m = rxr->fmp; 2643 rxr->fmp = NULL; 2644 rxr->lmp = NULL; 2645 2646 m->m_pkthdr.rcvif = ifp; 2647 IFNET_STAT_INC(ifp, ipackets, 1); 2648 2649 if (ifp->if_capenable & IFCAP_RXCSUM) 2650 igb_rxcsum(staterr, m); 2651 2652 if (staterr & E1000_RXD_STAT_VP) { 2653 m->m_pkthdr.ether_vlantag = vlan; 2654 m->m_flags |= M_VLANTAG; 2655 } 2656 2657 if (ifp->if_capenable & IFCAP_RSS) { 2658 pi = igb_rssinfo(m, &pi0, 2659 hash, hashtype, staterr); 2660 } 2661 #ifdef IGB_RSS_DEBUG 2662 rxr->rx_packets++; 2663 #endif 2664 } 2665 } else { 2666 IFNET_STAT_INC(ifp, ierrors, 1); 2667 discard: 2668 igb_setup_rxdesc(cur, rxbuf); 2669 if (!eop) 2670 rxr->discard = TRUE; 2671 else 2672 rxr->discard = FALSE; 2673 if (rxr->fmp != NULL) { 2674 m_freem(rxr->fmp); 2675 rxr->fmp = NULL; 2676 rxr->lmp = NULL; 2677 } 2678 m = NULL; 2679 } 2680 2681 if (m != NULL) 2682 ifp->if_input(ifp, m, pi, cpuid); 2683 2684 /* Advance our pointers to the next descriptor. */ 2685 if (++i == rxr->num_rx_desc) 2686 i = 0; 2687 2688 if (ncoll >= rxr->wreg_nsegs) { 2689 igb_rx_refresh(rxr, i); 2690 ncoll = 0; 2691 } 2692 2693 cur = &rxr->rx_base[i]; 2694 staterr = le32toh(cur->wb.upper.status_error); 2695 } 2696 rxr->next_to_check = i; 2697 2698 if (ncoll > 0) 2699 igb_rx_refresh(rxr, i); 2700 } 2701 2702 2703 static void 2704 igb_set_vlan(struct igb_softc *sc) 2705 { 2706 struct e1000_hw *hw = &sc->hw; 2707 uint32_t reg; 2708 #if 0 2709 struct ifnet *ifp = sc->arpcom.ac_if; 2710 #endif 2711 2712 if (sc->vf_ifp) { 2713 e1000_rlpml_set_vf(hw, sc->max_frame_size + VLAN_TAG_SIZE); 2714 return; 2715 } 2716 2717 reg = E1000_READ_REG(hw, E1000_CTRL); 2718 reg |= E1000_CTRL_VME; 2719 E1000_WRITE_REG(hw, E1000_CTRL, reg); 2720 2721 #if 0 2722 /* Enable the Filter Table */ 2723 if (ifp->if_capenable & IFCAP_VLAN_HWFILTER) { 2724 reg = E1000_READ_REG(hw, E1000_RCTL); 2725 reg &= ~E1000_RCTL_CFIEN; 2726 reg |= E1000_RCTL_VFE; 2727 E1000_WRITE_REG(hw, E1000_RCTL, reg); 2728 } 2729 #endif 2730 2731 /* Update the frame size */ 2732 E1000_WRITE_REG(&sc->hw, E1000_RLPML, 2733 sc->max_frame_size + VLAN_TAG_SIZE); 2734 2735 #if 0 2736 /* Don't bother with table if no vlans */ 2737 if ((adapter->num_vlans == 0) || 2738 ((ifp->if_capenable & IFCAP_VLAN_HWFILTER) == 0)) 2739 return; 2740 /* 2741 ** A soft reset zero's out the VFTA, so 2742 ** we need to repopulate it now. 2743 */ 2744 for (int i = 0; i < IGB_VFTA_SIZE; i++) 2745 if (adapter->shadow_vfta[i] != 0) { 2746 if (adapter->vf_ifp) 2747 e1000_vfta_set_vf(hw, 2748 adapter->shadow_vfta[i], TRUE); 2749 else 2750 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, 2751 i, adapter->shadow_vfta[i]); 2752 } 2753 #endif 2754 } 2755 2756 static void 2757 igb_enable_intr(struct igb_softc *sc) 2758 { 2759 int i; 2760 2761 for (i = 0; i < sc->intr_cnt; ++i) 2762 lwkt_serialize_handler_enable(sc->intr_data[i].intr_serialize); 2763 2764 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) { 2765 if (sc->intr_type == PCI_INTR_TYPE_MSIX) 2766 E1000_WRITE_REG(&sc->hw, E1000_EIAC, sc->intr_mask); 2767 else 2768 E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0); 2769 E1000_WRITE_REG(&sc->hw, E1000_EIAM, sc->intr_mask); 2770 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask); 2771 E1000_WRITE_REG(&sc->hw, E1000_IMS, E1000_IMS_LSC); 2772 } else { 2773 E1000_WRITE_REG(&sc->hw, E1000_IMS, IMS_ENABLE_MASK); 2774 } 2775 E1000_WRITE_FLUSH(&sc->hw); 2776 } 2777 2778 static void 2779 igb_disable_intr(struct igb_softc *sc) 2780 { 2781 int i; 2782 2783 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) { 2784 E1000_WRITE_REG(&sc->hw, E1000_EIMC, 0xffffffff); 2785 E1000_WRITE_REG(&sc->hw, E1000_EIAC, 0); 2786 } 2787 E1000_WRITE_REG(&sc->hw, E1000_IMC, 0xffffffff); 2788 E1000_WRITE_FLUSH(&sc->hw); 2789 2790 for (i = 0; i < sc->intr_cnt; ++i) 2791 lwkt_serialize_handler_disable(sc->intr_data[i].intr_serialize); 2792 } 2793 2794 /* 2795 * Bit of a misnomer, what this really means is 2796 * to enable OS management of the system... aka 2797 * to disable special hardware management features 2798 */ 2799 static void 2800 igb_get_mgmt(struct igb_softc *sc) 2801 { 2802 if (sc->flags & IGB_FLAG_HAS_MGMT) { 2803 int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H); 2804 int manc = E1000_READ_REG(&sc->hw, E1000_MANC); 2805 2806 /* disable hardware interception of ARP */ 2807 manc &= ~E1000_MANC_ARP_EN; 2808 2809 /* enable receiving management packets to the host */ 2810 manc |= E1000_MANC_EN_MNG2HOST; 2811 manc2h |= 1 << 5; /* Mng Port 623 */ 2812 manc2h |= 1 << 6; /* Mng Port 664 */ 2813 E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h); 2814 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc); 2815 } 2816 } 2817 2818 /* 2819 * Give control back to hardware management controller 2820 * if there is one. 2821 */ 2822 static void 2823 igb_rel_mgmt(struct igb_softc *sc) 2824 { 2825 if (sc->flags & IGB_FLAG_HAS_MGMT) { 2826 int manc = E1000_READ_REG(&sc->hw, E1000_MANC); 2827 2828 /* Re-enable hardware interception of ARP */ 2829 manc |= E1000_MANC_ARP_EN; 2830 manc &= ~E1000_MANC_EN_MNG2HOST; 2831 2832 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc); 2833 } 2834 } 2835 2836 /* 2837 * Sets CTRL_EXT:DRV_LOAD bit. 2838 * 2839 * For ASF and Pass Through versions of f/w this means that 2840 * the driver is loaded. 2841 */ 2842 static void 2843 igb_get_hw_control(struct igb_softc *sc) 2844 { 2845 uint32_t ctrl_ext; 2846 2847 if (sc->vf_ifp) 2848 return; 2849 2850 /* Let firmware know the driver has taken over */ 2851 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT); 2852 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, 2853 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD); 2854 } 2855 2856 /* 2857 * Resets CTRL_EXT:DRV_LOAD bit. 2858 * 2859 * For ASF and Pass Through versions of f/w this means that the 2860 * driver is no longer loaded. 2861 */ 2862 static void 2863 igb_rel_hw_control(struct igb_softc *sc) 2864 { 2865 uint32_t ctrl_ext; 2866 2867 if (sc->vf_ifp) 2868 return; 2869 2870 /* Let firmware taken over control of h/w */ 2871 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT); 2872 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, 2873 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD); 2874 } 2875 2876 static boolean_t 2877 igb_is_valid_ether_addr(const uint8_t *addr) 2878 { 2879 uint8_t zero_addr[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 }; 2880 2881 if ((addr[0] & 1) || !bcmp(addr, zero_addr, ETHER_ADDR_LEN)) 2882 return FALSE; 2883 return TRUE; 2884 } 2885 2886 /* 2887 * Enable PCI Wake On Lan capability 2888 */ 2889 static void 2890 igb_enable_wol(device_t dev) 2891 { 2892 uint16_t cap, status; 2893 uint8_t id; 2894 2895 /* First find the capabilities pointer*/ 2896 cap = pci_read_config(dev, PCIR_CAP_PTR, 2); 2897 2898 /* Read the PM Capabilities */ 2899 id = pci_read_config(dev, cap, 1); 2900 if (id != PCIY_PMG) /* Something wrong */ 2901 return; 2902 2903 /* 2904 * OK, we have the power capabilities, 2905 * so now get the status register 2906 */ 2907 cap += PCIR_POWER_STATUS; 2908 status = pci_read_config(dev, cap, 2); 2909 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE; 2910 pci_write_config(dev, cap, status, 2); 2911 } 2912 2913 static void 2914 igb_update_stats_counters(struct igb_softc *sc) 2915 { 2916 struct e1000_hw *hw = &sc->hw; 2917 struct e1000_hw_stats *stats; 2918 struct ifnet *ifp = &sc->arpcom.ac_if; 2919 2920 /* 2921 * The virtual function adapter has only a 2922 * small controlled set of stats, do only 2923 * those and return. 2924 */ 2925 if (sc->vf_ifp) { 2926 igb_update_vf_stats_counters(sc); 2927 return; 2928 } 2929 stats = sc->stats; 2930 2931 if (sc->hw.phy.media_type == e1000_media_type_copper || 2932 (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) { 2933 stats->symerrs += 2934 E1000_READ_REG(hw,E1000_SYMERRS); 2935 stats->sec += E1000_READ_REG(hw, E1000_SEC); 2936 } 2937 2938 stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS); 2939 stats->mpc += E1000_READ_REG(hw, E1000_MPC); 2940 stats->scc += E1000_READ_REG(hw, E1000_SCC); 2941 stats->ecol += E1000_READ_REG(hw, E1000_ECOL); 2942 2943 stats->mcc += E1000_READ_REG(hw, E1000_MCC); 2944 stats->latecol += E1000_READ_REG(hw, E1000_LATECOL); 2945 stats->colc += E1000_READ_REG(hw, E1000_COLC); 2946 stats->dc += E1000_READ_REG(hw, E1000_DC); 2947 stats->rlec += E1000_READ_REG(hw, E1000_RLEC); 2948 stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC); 2949 stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC); 2950 2951 /* 2952 * For watchdog management we need to know if we have been 2953 * paused during the last interval, so capture that here. 2954 */ 2955 sc->pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC); 2956 stats->xoffrxc += sc->pause_frames; 2957 stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC); 2958 stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC); 2959 stats->prc64 += E1000_READ_REG(hw, E1000_PRC64); 2960 stats->prc127 += E1000_READ_REG(hw, E1000_PRC127); 2961 stats->prc255 += E1000_READ_REG(hw, E1000_PRC255); 2962 stats->prc511 += E1000_READ_REG(hw, E1000_PRC511); 2963 stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023); 2964 stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522); 2965 stats->gprc += E1000_READ_REG(hw, E1000_GPRC); 2966 stats->bprc += E1000_READ_REG(hw, E1000_BPRC); 2967 stats->mprc += E1000_READ_REG(hw, E1000_MPRC); 2968 stats->gptc += E1000_READ_REG(hw, E1000_GPTC); 2969 2970 /* For the 64-bit byte counters the low dword must be read first. */ 2971 /* Both registers clear on the read of the high dword */ 2972 2973 stats->gorc += E1000_READ_REG(hw, E1000_GORCL) + 2974 ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32); 2975 stats->gotc += E1000_READ_REG(hw, E1000_GOTCL) + 2976 ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32); 2977 2978 stats->rnbc += E1000_READ_REG(hw, E1000_RNBC); 2979 stats->ruc += E1000_READ_REG(hw, E1000_RUC); 2980 stats->rfc += E1000_READ_REG(hw, E1000_RFC); 2981 stats->roc += E1000_READ_REG(hw, E1000_ROC); 2982 stats->rjc += E1000_READ_REG(hw, E1000_RJC); 2983 2984 stats->tor += E1000_READ_REG(hw, E1000_TORH); 2985 stats->tot += E1000_READ_REG(hw, E1000_TOTH); 2986 2987 stats->tpr += E1000_READ_REG(hw, E1000_TPR); 2988 stats->tpt += E1000_READ_REG(hw, E1000_TPT); 2989 stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64); 2990 stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127); 2991 stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255); 2992 stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511); 2993 stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023); 2994 stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522); 2995 stats->mptc += E1000_READ_REG(hw, E1000_MPTC); 2996 stats->bptc += E1000_READ_REG(hw, E1000_BPTC); 2997 2998 /* Interrupt Counts */ 2999 3000 stats->iac += E1000_READ_REG(hw, E1000_IAC); 3001 stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC); 3002 stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC); 3003 stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC); 3004 stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC); 3005 stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC); 3006 stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC); 3007 stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC); 3008 stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC); 3009 3010 /* Host to Card Statistics */ 3011 3012 stats->cbtmpc += E1000_READ_REG(hw, E1000_CBTMPC); 3013 stats->htdpmc += E1000_READ_REG(hw, E1000_HTDPMC); 3014 stats->cbrdpc += E1000_READ_REG(hw, E1000_CBRDPC); 3015 stats->cbrmpc += E1000_READ_REG(hw, E1000_CBRMPC); 3016 stats->rpthc += E1000_READ_REG(hw, E1000_RPTHC); 3017 stats->hgptc += E1000_READ_REG(hw, E1000_HGPTC); 3018 stats->htcbdpc += E1000_READ_REG(hw, E1000_HTCBDPC); 3019 stats->hgorc += (E1000_READ_REG(hw, E1000_HGORCL) + 3020 ((uint64_t)E1000_READ_REG(hw, E1000_HGORCH) << 32)); 3021 stats->hgotc += (E1000_READ_REG(hw, E1000_HGOTCL) + 3022 ((uint64_t)E1000_READ_REG(hw, E1000_HGOTCH) << 32)); 3023 stats->lenerrs += E1000_READ_REG(hw, E1000_LENERRS); 3024 stats->scvpc += E1000_READ_REG(hw, E1000_SCVPC); 3025 stats->hrmpc += E1000_READ_REG(hw, E1000_HRMPC); 3026 3027 stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC); 3028 stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC); 3029 stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS); 3030 stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR); 3031 stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC); 3032 stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC); 3033 3034 IFNET_STAT_SET(ifp, collisions, stats->colc); 3035 3036 /* Rx Errors */ 3037 IFNET_STAT_SET(ifp, ierrors, 3038 stats->rxerrc + stats->crcerrs + stats->algnerrc + 3039 stats->ruc + stats->roc + stats->mpc + stats->cexterr); 3040 3041 /* Tx Errors */ 3042 IFNET_STAT_SET(ifp, oerrors, 3043 stats->ecol + stats->latecol + sc->watchdog_events); 3044 3045 /* Driver specific counters */ 3046 sc->device_control = E1000_READ_REG(hw, E1000_CTRL); 3047 sc->rx_control = E1000_READ_REG(hw, E1000_RCTL); 3048 sc->int_mask = E1000_READ_REG(hw, E1000_IMS); 3049 sc->eint_mask = E1000_READ_REG(hw, E1000_EIMS); 3050 sc->packet_buf_alloc_tx = 3051 ((E1000_READ_REG(hw, E1000_PBA) & 0xffff0000) >> 16); 3052 sc->packet_buf_alloc_rx = 3053 (E1000_READ_REG(hw, E1000_PBA) & 0xffff); 3054 } 3055 3056 static void 3057 igb_vf_init_stats(struct igb_softc *sc) 3058 { 3059 struct e1000_hw *hw = &sc->hw; 3060 struct e1000_vf_stats *stats; 3061 3062 stats = sc->stats; 3063 stats->last_gprc = E1000_READ_REG(hw, E1000_VFGPRC); 3064 stats->last_gorc = E1000_READ_REG(hw, E1000_VFGORC); 3065 stats->last_gptc = E1000_READ_REG(hw, E1000_VFGPTC); 3066 stats->last_gotc = E1000_READ_REG(hw, E1000_VFGOTC); 3067 stats->last_mprc = E1000_READ_REG(hw, E1000_VFMPRC); 3068 } 3069 3070 static void 3071 igb_update_vf_stats_counters(struct igb_softc *sc) 3072 { 3073 struct e1000_hw *hw = &sc->hw; 3074 struct e1000_vf_stats *stats; 3075 3076 if (sc->link_speed == 0) 3077 return; 3078 3079 stats = sc->stats; 3080 UPDATE_VF_REG(E1000_VFGPRC, stats->last_gprc, stats->gprc); 3081 UPDATE_VF_REG(E1000_VFGORC, stats->last_gorc, stats->gorc); 3082 UPDATE_VF_REG(E1000_VFGPTC, stats->last_gptc, stats->gptc); 3083 UPDATE_VF_REG(E1000_VFGOTC, stats->last_gotc, stats->gotc); 3084 UPDATE_VF_REG(E1000_VFMPRC, stats->last_mprc, stats->mprc); 3085 } 3086 3087 #ifdef IFPOLL_ENABLE 3088 3089 static void 3090 igb_npoll_status(struct ifnet *ifp) 3091 { 3092 struct igb_softc *sc = ifp->if_softc; 3093 uint32_t reg_icr; 3094 3095 ASSERT_SERIALIZED(&sc->main_serialize); 3096 3097 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR); 3098 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) { 3099 sc->hw.mac.get_link_status = 1; 3100 igb_update_link_status(sc); 3101 } 3102 } 3103 3104 static void 3105 igb_npoll_tx(struct ifnet *ifp, void *arg, int cycle __unused) 3106 { 3107 struct igb_tx_ring *txr = arg; 3108 3109 ASSERT_SERIALIZED(&txr->tx_serialize); 3110 3111 igb_txeof(txr, *(txr->tx_hdr)); 3112 if (!ifsq_is_empty(txr->ifsq)) 3113 ifsq_devstart(txr->ifsq); 3114 } 3115 3116 static void 3117 igb_npoll_rx(struct ifnet *ifp __unused, void *arg, int cycle) 3118 { 3119 struct igb_rx_ring *rxr = arg; 3120 3121 ASSERT_SERIALIZED(&rxr->rx_serialize); 3122 3123 igb_rxeof(rxr, cycle); 3124 } 3125 3126 static void 3127 igb_npoll(struct ifnet *ifp, struct ifpoll_info *info) 3128 { 3129 struct igb_softc *sc = ifp->if_softc; 3130 int i, txr_cnt, rxr_cnt; 3131 3132 ASSERT_IFNET_SERIALIZED_ALL(ifp); 3133 3134 if (info) { 3135 int off; 3136 3137 info->ifpi_status.status_func = igb_npoll_status; 3138 info->ifpi_status.serializer = &sc->main_serialize; 3139 3140 txr_cnt = igb_get_txring_inuse(sc, TRUE); 3141 off = sc->tx_npoll_off; 3142 for (i = 0; i < txr_cnt; ++i) { 3143 struct igb_tx_ring *txr = &sc->tx_rings[i]; 3144 int idx = i + off; 3145 3146 KKASSERT(idx < ncpus2); 3147 info->ifpi_tx[idx].poll_func = igb_npoll_tx; 3148 info->ifpi_tx[idx].arg = txr; 3149 info->ifpi_tx[idx].serializer = &txr->tx_serialize; 3150 ifsq_set_cpuid(txr->ifsq, idx); 3151 } 3152 3153 rxr_cnt = igb_get_rxring_inuse(sc, TRUE); 3154 off = sc->rx_npoll_off; 3155 for (i = 0; i < rxr_cnt; ++i) { 3156 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 3157 int idx = i + off; 3158 3159 KKASSERT(idx < ncpus2); 3160 info->ifpi_rx[idx].poll_func = igb_npoll_rx; 3161 info->ifpi_rx[idx].arg = rxr; 3162 info->ifpi_rx[idx].serializer = &rxr->rx_serialize; 3163 } 3164 3165 if (ifp->if_flags & IFF_RUNNING) { 3166 if (rxr_cnt == sc->rx_ring_inuse && 3167 txr_cnt == sc->tx_ring_inuse) { 3168 igb_set_timer_cpuid(sc, TRUE); 3169 igb_disable_intr(sc); 3170 } else { 3171 igb_init(sc); 3172 } 3173 } 3174 } else { 3175 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3176 struct igb_tx_ring *txr = &sc->tx_rings[i]; 3177 3178 ifsq_set_cpuid(txr->ifsq, txr->tx_intr_cpuid); 3179 } 3180 3181 if (ifp->if_flags & IFF_RUNNING) { 3182 txr_cnt = igb_get_txring_inuse(sc, FALSE); 3183 rxr_cnt = igb_get_rxring_inuse(sc, FALSE); 3184 3185 if (rxr_cnt == sc->rx_ring_inuse && 3186 txr_cnt == sc->tx_ring_inuse) { 3187 igb_set_timer_cpuid(sc, FALSE); 3188 igb_enable_intr(sc); 3189 } else { 3190 igb_init(sc); 3191 } 3192 } 3193 } 3194 } 3195 3196 #endif /* IFPOLL_ENABLE */ 3197 3198 static void 3199 igb_intr(void *xsc) 3200 { 3201 struct igb_softc *sc = xsc; 3202 struct ifnet *ifp = &sc->arpcom.ac_if; 3203 uint32_t eicr; 3204 3205 ASSERT_SERIALIZED(&sc->main_serialize); 3206 3207 eicr = E1000_READ_REG(&sc->hw, E1000_EICR); 3208 3209 if (eicr == 0) 3210 return; 3211 3212 if (ifp->if_flags & IFF_RUNNING) { 3213 struct igb_tx_ring *txr = &sc->tx_rings[0]; 3214 int i; 3215 3216 for (i = 0; i < sc->rx_ring_inuse; ++i) { 3217 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 3218 3219 if (eicr & rxr->rx_intr_mask) { 3220 lwkt_serialize_enter(&rxr->rx_serialize); 3221 igb_rxeof(rxr, -1); 3222 lwkt_serialize_exit(&rxr->rx_serialize); 3223 } 3224 } 3225 3226 if (eicr & txr->tx_intr_mask) { 3227 lwkt_serialize_enter(&txr->tx_serialize); 3228 igb_txeof(txr, *(txr->tx_hdr)); 3229 if (!ifsq_is_empty(txr->ifsq)) 3230 ifsq_devstart(txr->ifsq); 3231 lwkt_serialize_exit(&txr->tx_serialize); 3232 } 3233 } 3234 3235 if (eicr & E1000_EICR_OTHER) { 3236 uint32_t icr = E1000_READ_REG(&sc->hw, E1000_ICR); 3237 3238 /* Link status change */ 3239 if (icr & E1000_ICR_LSC) { 3240 sc->hw.mac.get_link_status = 1; 3241 igb_update_link_status(sc); 3242 } 3243 } 3244 3245 /* 3246 * Reading EICR has the side effect to clear interrupt mask, 3247 * so all interrupts need to be enabled here. 3248 */ 3249 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->intr_mask); 3250 } 3251 3252 static void 3253 igb_intr_shared(void *xsc) 3254 { 3255 struct igb_softc *sc = xsc; 3256 struct ifnet *ifp = &sc->arpcom.ac_if; 3257 uint32_t reg_icr; 3258 3259 ASSERT_SERIALIZED(&sc->main_serialize); 3260 3261 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR); 3262 3263 /* Hot eject? */ 3264 if (reg_icr == 0xffffffff) 3265 return; 3266 3267 /* Definitely not our interrupt. */ 3268 if (reg_icr == 0x0) 3269 return; 3270 3271 if ((reg_icr & E1000_ICR_INT_ASSERTED) == 0) 3272 return; 3273 3274 if (ifp->if_flags & IFF_RUNNING) { 3275 if (reg_icr & 3276 (E1000_ICR_RXT0 | E1000_ICR_RXDMT0 | E1000_ICR_RXO)) { 3277 int i; 3278 3279 for (i = 0; i < sc->rx_ring_inuse; ++i) { 3280 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 3281 3282 lwkt_serialize_enter(&rxr->rx_serialize); 3283 igb_rxeof(rxr, -1); 3284 lwkt_serialize_exit(&rxr->rx_serialize); 3285 } 3286 } 3287 3288 if (reg_icr & E1000_ICR_TXDW) { 3289 struct igb_tx_ring *txr = &sc->tx_rings[0]; 3290 3291 lwkt_serialize_enter(&txr->tx_serialize); 3292 igb_txeof(txr, *(txr->tx_hdr)); 3293 if (!ifsq_is_empty(txr->ifsq)) 3294 ifsq_devstart(txr->ifsq); 3295 lwkt_serialize_exit(&txr->tx_serialize); 3296 } 3297 } 3298 3299 /* Link status change */ 3300 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) { 3301 sc->hw.mac.get_link_status = 1; 3302 igb_update_link_status(sc); 3303 } 3304 3305 if (reg_icr & E1000_ICR_RXO) 3306 sc->rx_overruns++; 3307 } 3308 3309 static int 3310 igb_encap(struct igb_tx_ring *txr, struct mbuf **m_headp, 3311 int *segs_used, int *idx) 3312 { 3313 bus_dma_segment_t segs[IGB_MAX_SCATTER]; 3314 bus_dmamap_t map; 3315 struct igb_tx_buf *tx_buf, *tx_buf_mapped; 3316 union e1000_adv_tx_desc *txd = NULL; 3317 struct mbuf *m_head = *m_headp; 3318 uint32_t olinfo_status = 0, cmd_type_len = 0, cmd_rs = 0; 3319 int maxsegs, nsegs, i, j, error; 3320 uint32_t hdrlen = 0; 3321 3322 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 3323 error = igb_tso_pullup(txr, m_headp); 3324 if (error) 3325 return error; 3326 m_head = *m_headp; 3327 } 3328 3329 /* Set basic descriptor constants */ 3330 cmd_type_len |= E1000_ADVTXD_DTYP_DATA; 3331 cmd_type_len |= E1000_ADVTXD_DCMD_IFCS | E1000_ADVTXD_DCMD_DEXT; 3332 if (m_head->m_flags & M_VLANTAG) 3333 cmd_type_len |= E1000_ADVTXD_DCMD_VLE; 3334 3335 /* 3336 * Map the packet for DMA. 3337 */ 3338 tx_buf = &txr->tx_buf[txr->next_avail_desc]; 3339 tx_buf_mapped = tx_buf; 3340 map = tx_buf->map; 3341 3342 maxsegs = txr->tx_avail - IGB_TX_RESERVED; 3343 if (maxsegs > IGB_MAX_SCATTER) 3344 maxsegs = IGB_MAX_SCATTER; 3345 3346 error = bus_dmamap_load_mbuf_defrag(txr->tx_tag, map, m_headp, 3347 segs, maxsegs, &nsegs, BUS_DMA_NOWAIT); 3348 if (error) { 3349 if (error == ENOBUFS) 3350 txr->sc->mbuf_defrag_failed++; 3351 else 3352 txr->sc->no_tx_dma_setup++; 3353 3354 m_freem(*m_headp); 3355 *m_headp = NULL; 3356 return error; 3357 } 3358 bus_dmamap_sync(txr->tx_tag, map, BUS_DMASYNC_PREWRITE); 3359 3360 m_head = *m_headp; 3361 3362 /* 3363 * Set up the TX context descriptor, if any hardware offloading is 3364 * needed. This includes CSUM, VLAN, and TSO. It will consume one 3365 * TX descriptor. 3366 * 3367 * Unlike these chips' predecessors (em/emx), TX context descriptor 3368 * will _not_ interfere TX data fetching pipelining. 3369 */ 3370 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 3371 igb_tso_ctx(txr, m_head, &hdrlen); 3372 cmd_type_len |= E1000_ADVTXD_DCMD_TSE; 3373 olinfo_status |= E1000_TXD_POPTS_IXSM << 8; 3374 olinfo_status |= E1000_TXD_POPTS_TXSM << 8; 3375 txr->tx_nsegs++; 3376 (*segs_used)++; 3377 } else if (igb_txcsum_ctx(txr, m_head)) { 3378 if (m_head->m_pkthdr.csum_flags & CSUM_IP) 3379 olinfo_status |= (E1000_TXD_POPTS_IXSM << 8); 3380 if (m_head->m_pkthdr.csum_flags & (CSUM_UDP | CSUM_TCP)) 3381 olinfo_status |= (E1000_TXD_POPTS_TXSM << 8); 3382 txr->tx_nsegs++; 3383 (*segs_used)++; 3384 } 3385 3386 *segs_used += nsegs; 3387 txr->tx_nsegs += nsegs; 3388 if (txr->tx_nsegs >= txr->intr_nsegs) { 3389 /* 3390 * Report Status (RS) is turned on every intr_nsegs 3391 * descriptors (roughly). 3392 */ 3393 txr->tx_nsegs = 0; 3394 cmd_rs = E1000_ADVTXD_DCMD_RS; 3395 } 3396 3397 /* Calculate payload length */ 3398 olinfo_status |= ((m_head->m_pkthdr.len - hdrlen) 3399 << E1000_ADVTXD_PAYLEN_SHIFT); 3400 3401 /* 3402 * 82575 needs the TX context index added; the queue 3403 * index is used as TX context index here. 3404 */ 3405 if (txr->sc->hw.mac.type == e1000_82575) 3406 olinfo_status |= txr->me << 4; 3407 3408 /* Set up our transmit descriptors */ 3409 i = txr->next_avail_desc; 3410 for (j = 0; j < nsegs; j++) { 3411 bus_size_t seg_len; 3412 bus_addr_t seg_addr; 3413 3414 tx_buf = &txr->tx_buf[i]; 3415 txd = (union e1000_adv_tx_desc *)&txr->tx_base[i]; 3416 seg_addr = segs[j].ds_addr; 3417 seg_len = segs[j].ds_len; 3418 3419 txd->read.buffer_addr = htole64(seg_addr); 3420 txd->read.cmd_type_len = htole32(cmd_type_len | seg_len); 3421 txd->read.olinfo_status = htole32(olinfo_status); 3422 if (++i == txr->num_tx_desc) 3423 i = 0; 3424 tx_buf->m_head = NULL; 3425 } 3426 3427 KASSERT(txr->tx_avail > nsegs, ("invalid avail TX desc\n")); 3428 txr->next_avail_desc = i; 3429 txr->tx_avail -= nsegs; 3430 3431 tx_buf->m_head = m_head; 3432 tx_buf_mapped->map = tx_buf->map; 3433 tx_buf->map = map; 3434 3435 /* 3436 * Last Descriptor of Packet needs End Of Packet (EOP) 3437 */ 3438 txd->read.cmd_type_len |= htole32(E1000_ADVTXD_DCMD_EOP | cmd_rs); 3439 3440 /* 3441 * Defer TDT updating, until enough descrptors are setup 3442 */ 3443 *idx = i; 3444 #ifdef IGB_TSS_DEBUG 3445 ++txr->tx_packets; 3446 #endif 3447 3448 return 0; 3449 } 3450 3451 static void 3452 igb_start(struct ifnet *ifp, struct ifaltq_subque *ifsq) 3453 { 3454 struct igb_softc *sc = ifp->if_softc; 3455 struct igb_tx_ring *txr = ifsq_get_priv(ifsq); 3456 struct mbuf *m_head; 3457 int idx = -1, nsegs = 0; 3458 3459 KKASSERT(txr->ifsq == ifsq); 3460 ASSERT_SERIALIZED(&txr->tx_serialize); 3461 3462 if ((ifp->if_flags & IFF_RUNNING) == 0 || ifsq_is_oactive(ifsq)) 3463 return; 3464 3465 if (!sc->link_active || (txr->tx_flags & IGB_TXFLAG_ENABLED) == 0) { 3466 ifsq_purge(ifsq); 3467 return; 3468 } 3469 3470 while (!ifsq_is_empty(ifsq)) { 3471 if (txr->tx_avail <= IGB_MAX_SCATTER + IGB_TX_RESERVED) { 3472 ifsq_set_oactive(ifsq); 3473 /* Set watchdog on */ 3474 txr->tx_watchdog.wd_timer = 5; 3475 break; 3476 } 3477 3478 m_head = ifsq_dequeue(ifsq); 3479 if (m_head == NULL) 3480 break; 3481 3482 if (igb_encap(txr, &m_head, &nsegs, &idx)) { 3483 IFNET_STAT_INC(ifp, oerrors, 1); 3484 continue; 3485 } 3486 3487 /* 3488 * TX interrupt are aggressively aggregated, so increasing 3489 * opackets at TX interrupt time will make the opackets 3490 * statistics vastly inaccurate; we do the opackets increment 3491 * now. 3492 */ 3493 IFNET_STAT_INC(ifp, opackets, 1); 3494 3495 if (nsegs >= txr->wreg_nsegs) { 3496 E1000_WRITE_REG(&txr->sc->hw, E1000_TDT(txr->me), idx); 3497 idx = -1; 3498 nsegs = 0; 3499 } 3500 3501 /* Send a copy of the frame to the BPF listener */ 3502 ETHER_BPF_MTAP(ifp, m_head); 3503 } 3504 if (idx >= 0) 3505 E1000_WRITE_REG(&txr->sc->hw, E1000_TDT(txr->me), idx); 3506 } 3507 3508 static void 3509 igb_watchdog(struct ifaltq_subque *ifsq) 3510 { 3511 struct igb_tx_ring *txr = ifsq_get_priv(ifsq); 3512 struct ifnet *ifp = ifsq_get_ifp(ifsq); 3513 struct igb_softc *sc = ifp->if_softc; 3514 int i; 3515 3516 KKASSERT(txr->ifsq == ifsq); 3517 ASSERT_IFNET_SERIALIZED_ALL(ifp); 3518 3519 /* 3520 * If flow control has paused us since last checking 3521 * it invalidates the watchdog timing, so dont run it. 3522 */ 3523 if (sc->pause_frames) { 3524 sc->pause_frames = 0; 3525 txr->tx_watchdog.wd_timer = 5; 3526 return; 3527 } 3528 3529 if_printf(ifp, "Watchdog timeout -- resetting\n"); 3530 if_printf(ifp, "Queue(%d) tdh = %d, hw tdt = %d\n", txr->me, 3531 E1000_READ_REG(&sc->hw, E1000_TDH(txr->me)), 3532 E1000_READ_REG(&sc->hw, E1000_TDT(txr->me))); 3533 if_printf(ifp, "TX(%d) desc avail = %d, " 3534 "Next TX to Clean = %d\n", 3535 txr->me, txr->tx_avail, txr->next_to_clean); 3536 3537 IFNET_STAT_INC(ifp, oerrors, 1); 3538 sc->watchdog_events++; 3539 3540 igb_init(sc); 3541 for (i = 0; i < sc->tx_ring_inuse; ++i) 3542 ifsq_devstart_sched(sc->tx_rings[i].ifsq); 3543 } 3544 3545 static void 3546 igb_set_eitr(struct igb_softc *sc, int idx, int rate) 3547 { 3548 uint32_t eitr = 0; 3549 3550 if (rate > 0) { 3551 if (sc->hw.mac.type == e1000_82575) { 3552 eitr = 1000000000 / 256 / rate; 3553 /* 3554 * NOTE: 3555 * Document is wrong on the 2 bits left shift 3556 */ 3557 } else { 3558 eitr = 1000000 / rate; 3559 eitr <<= IGB_EITR_INTVL_SHIFT; 3560 } 3561 3562 if (eitr == 0) { 3563 /* Don't disable it */ 3564 eitr = 1 << IGB_EITR_INTVL_SHIFT; 3565 } else if (eitr > IGB_EITR_INTVL_MASK) { 3566 /* Don't allow it to be too large */ 3567 eitr = IGB_EITR_INTVL_MASK; 3568 } 3569 } 3570 if (sc->hw.mac.type == e1000_82575) 3571 eitr |= eitr << 16; 3572 else 3573 eitr |= E1000_EITR_CNT_IGNR; 3574 E1000_WRITE_REG(&sc->hw, E1000_EITR(idx), eitr); 3575 } 3576 3577 static void 3578 igb_add_intr_rate_sysctl(struct igb_softc *sc, int use, 3579 const char *name, const char *desc) 3580 { 3581 int i; 3582 3583 for (i = 0; i < sc->intr_cnt; ++i) { 3584 if (sc->intr_data[i].intr_use == use) { 3585 SYSCTL_ADD_PROC(device_get_sysctl_ctx(sc->dev), 3586 SYSCTL_CHILDREN(device_get_sysctl_tree(sc->dev)), 3587 OID_AUTO, name, CTLTYPE_INT | CTLFLAG_RW, 3588 sc, use, igb_sysctl_intr_rate, "I", desc); 3589 break; 3590 } 3591 } 3592 } 3593 3594 static int 3595 igb_sysctl_intr_rate(SYSCTL_HANDLER_ARGS) 3596 { 3597 struct igb_softc *sc = (void *)arg1; 3598 int use = arg2; 3599 struct ifnet *ifp = &sc->arpcom.ac_if; 3600 int error, rate, i; 3601 struct igb_intr_data *intr; 3602 3603 rate = 0; 3604 for (i = 0; i < sc->intr_cnt; ++i) { 3605 intr = &sc->intr_data[i]; 3606 if (intr->intr_use == use) { 3607 rate = intr->intr_rate; 3608 break; 3609 } 3610 } 3611 3612 error = sysctl_handle_int(oidp, &rate, 0, req); 3613 if (error || req->newptr == NULL) 3614 return error; 3615 if (rate <= 0) 3616 return EINVAL; 3617 3618 ifnet_serialize_all(ifp); 3619 3620 for (i = 0; i < sc->intr_cnt; ++i) { 3621 intr = &sc->intr_data[i]; 3622 if (intr->intr_use == use && intr->intr_rate != rate) { 3623 intr->intr_rate = rate; 3624 if (ifp->if_flags & IFF_RUNNING) 3625 igb_set_eitr(sc, i, rate); 3626 } 3627 } 3628 3629 ifnet_deserialize_all(ifp); 3630 3631 return error; 3632 } 3633 3634 static int 3635 igb_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS) 3636 { 3637 struct igb_softc *sc = (void *)arg1; 3638 struct ifnet *ifp = &sc->arpcom.ac_if; 3639 struct igb_tx_ring *txr = &sc->tx_rings[0]; 3640 int error, nsegs; 3641 3642 nsegs = txr->intr_nsegs; 3643 error = sysctl_handle_int(oidp, &nsegs, 0, req); 3644 if (error || req->newptr == NULL) 3645 return error; 3646 if (nsegs <= 0) 3647 return EINVAL; 3648 3649 ifnet_serialize_all(ifp); 3650 3651 if (nsegs >= txr->num_tx_desc - IGB_MAX_SCATTER - IGB_TX_RESERVED) { 3652 error = EINVAL; 3653 } else { 3654 int i; 3655 3656 error = 0; 3657 for (i = 0; i < sc->tx_ring_cnt; ++i) 3658 sc->tx_rings[i].intr_nsegs = nsegs; 3659 } 3660 3661 ifnet_deserialize_all(ifp); 3662 3663 return error; 3664 } 3665 3666 static int 3667 igb_sysctl_rx_wreg_nsegs(SYSCTL_HANDLER_ARGS) 3668 { 3669 struct igb_softc *sc = (void *)arg1; 3670 struct ifnet *ifp = &sc->arpcom.ac_if; 3671 int error, nsegs, i; 3672 3673 nsegs = sc->rx_rings[0].wreg_nsegs; 3674 error = sysctl_handle_int(oidp, &nsegs, 0, req); 3675 if (error || req->newptr == NULL) 3676 return error; 3677 3678 ifnet_serialize_all(ifp); 3679 for (i = 0; i < sc->rx_ring_cnt; ++i) 3680 sc->rx_rings[i].wreg_nsegs = nsegs; 3681 ifnet_deserialize_all(ifp); 3682 3683 return 0; 3684 } 3685 3686 static int 3687 igb_sysctl_tx_wreg_nsegs(SYSCTL_HANDLER_ARGS) 3688 { 3689 struct igb_softc *sc = (void *)arg1; 3690 struct ifnet *ifp = &sc->arpcom.ac_if; 3691 int error, nsegs, i; 3692 3693 nsegs = sc->tx_rings[0].wreg_nsegs; 3694 error = sysctl_handle_int(oidp, &nsegs, 0, req); 3695 if (error || req->newptr == NULL) 3696 return error; 3697 3698 ifnet_serialize_all(ifp); 3699 for (i = 0; i < sc->tx_ring_cnt; ++i) 3700 sc->tx_rings[i].wreg_nsegs = nsegs; 3701 ifnet_deserialize_all(ifp); 3702 3703 return 0; 3704 } 3705 3706 #ifdef IFPOLL_ENABLE 3707 3708 static int 3709 igb_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS) 3710 { 3711 struct igb_softc *sc = (void *)arg1; 3712 struct ifnet *ifp = &sc->arpcom.ac_if; 3713 int error, off; 3714 3715 off = sc->rx_npoll_off; 3716 error = sysctl_handle_int(oidp, &off, 0, req); 3717 if (error || req->newptr == NULL) 3718 return error; 3719 if (off < 0) 3720 return EINVAL; 3721 3722 ifnet_serialize_all(ifp); 3723 if (off >= ncpus2 || off % sc->rx_ring_cnt != 0) { 3724 error = EINVAL; 3725 } else { 3726 error = 0; 3727 sc->rx_npoll_off = off; 3728 } 3729 ifnet_deserialize_all(ifp); 3730 3731 return error; 3732 } 3733 3734 static int 3735 igb_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS) 3736 { 3737 struct igb_softc *sc = (void *)arg1; 3738 struct ifnet *ifp = &sc->arpcom.ac_if; 3739 int error, off; 3740 3741 off = sc->tx_npoll_off; 3742 error = sysctl_handle_int(oidp, &off, 0, req); 3743 if (error || req->newptr == NULL) 3744 return error; 3745 if (off < 0) 3746 return EINVAL; 3747 3748 ifnet_serialize_all(ifp); 3749 if (off >= ncpus2 || off % sc->tx_ring_cnt != 0) { 3750 error = EINVAL; 3751 } else { 3752 error = 0; 3753 sc->tx_npoll_off = off; 3754 } 3755 ifnet_deserialize_all(ifp); 3756 3757 return error; 3758 } 3759 3760 #endif /* IFPOLL_ENABLE */ 3761 3762 static void 3763 igb_init_intr(struct igb_softc *sc) 3764 { 3765 int i; 3766 3767 igb_set_intr_mask(sc); 3768 3769 if ((sc->flags & IGB_FLAG_SHARED_INTR) == 0) 3770 igb_init_unshared_intr(sc); 3771 3772 for (i = 0; i < sc->intr_cnt; ++i) 3773 igb_set_eitr(sc, i, sc->intr_data[i].intr_rate); 3774 } 3775 3776 static void 3777 igb_init_unshared_intr(struct igb_softc *sc) 3778 { 3779 struct e1000_hw *hw = &sc->hw; 3780 const struct igb_rx_ring *rxr; 3781 const struct igb_tx_ring *txr; 3782 uint32_t ivar, index; 3783 int i; 3784 3785 /* 3786 * Enable extended mode 3787 */ 3788 if (sc->hw.mac.type != e1000_82575) { 3789 uint32_t gpie; 3790 int ivar_max; 3791 3792 gpie = E1000_GPIE_NSICR; 3793 if (sc->intr_type == PCI_INTR_TYPE_MSIX) { 3794 gpie |= E1000_GPIE_MSIX_MODE | 3795 E1000_GPIE_EIAME | 3796 E1000_GPIE_PBA; 3797 } 3798 E1000_WRITE_REG(hw, E1000_GPIE, gpie); 3799 3800 /* 3801 * Clear IVARs 3802 */ 3803 switch (sc->hw.mac.type) { 3804 case e1000_82576: 3805 ivar_max = IGB_MAX_IVAR_82576; 3806 break; 3807 3808 case e1000_82580: 3809 ivar_max = IGB_MAX_IVAR_82580; 3810 break; 3811 3812 case e1000_i350: 3813 ivar_max = IGB_MAX_IVAR_I350; 3814 break; 3815 3816 case e1000_i354: 3817 ivar_max = IGB_MAX_IVAR_I354; 3818 break; 3819 3820 case e1000_vfadapt: 3821 case e1000_vfadapt_i350: 3822 ivar_max = IGB_MAX_IVAR_VF; 3823 break; 3824 3825 case e1000_i210: 3826 ivar_max = IGB_MAX_IVAR_I210; 3827 break; 3828 3829 case e1000_i211: 3830 ivar_max = IGB_MAX_IVAR_I211; 3831 break; 3832 3833 default: 3834 panic("unknown mac type %d\n", sc->hw.mac.type); 3835 } 3836 for (i = 0; i < ivar_max; ++i) 3837 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, i, 0); 3838 E1000_WRITE_REG(hw, E1000_IVAR_MISC, 0); 3839 } else { 3840 uint32_t tmp; 3841 3842 KASSERT(sc->intr_type != PCI_INTR_TYPE_MSIX, 3843 ("82575 w/ MSI-X")); 3844 tmp = E1000_READ_REG(hw, E1000_CTRL_EXT); 3845 tmp |= E1000_CTRL_EXT_IRCA; 3846 E1000_WRITE_REG(hw, E1000_CTRL_EXT, tmp); 3847 } 3848 3849 /* 3850 * Map TX/RX interrupts to EICR 3851 */ 3852 switch (sc->hw.mac.type) { 3853 case e1000_82580: 3854 case e1000_i350: 3855 case e1000_i354: 3856 case e1000_vfadapt: 3857 case e1000_vfadapt_i350: 3858 case e1000_i210: 3859 case e1000_i211: 3860 /* RX entries */ 3861 for (i = 0; i < sc->rx_ring_inuse; ++i) { 3862 rxr = &sc->rx_rings[i]; 3863 3864 index = i >> 1; 3865 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 3866 3867 if (i & 1) { 3868 ivar &= 0xff00ffff; 3869 ivar |= 3870 (rxr->rx_intr_vec | E1000_IVAR_VALID) << 16; 3871 } else { 3872 ivar &= 0xffffff00; 3873 ivar |= 3874 (rxr->rx_intr_vec | E1000_IVAR_VALID); 3875 } 3876 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 3877 } 3878 /* TX entries */ 3879 for (i = 0; i < sc->tx_ring_inuse; ++i) { 3880 txr = &sc->tx_rings[i]; 3881 3882 index = i >> 1; 3883 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 3884 3885 if (i & 1) { 3886 ivar &= 0x00ffffff; 3887 ivar |= 3888 (txr->tx_intr_vec | E1000_IVAR_VALID) << 24; 3889 } else { 3890 ivar &= 0xffff00ff; 3891 ivar |= 3892 (txr->tx_intr_vec | E1000_IVAR_VALID) << 8; 3893 } 3894 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 3895 } 3896 if (sc->intr_type == PCI_INTR_TYPE_MSIX) { 3897 ivar = (sc->sts_msix_vec | E1000_IVAR_VALID) << 8; 3898 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar); 3899 } 3900 break; 3901 3902 case e1000_82576: 3903 /* RX entries */ 3904 for (i = 0; i < sc->rx_ring_inuse; ++i) { 3905 rxr = &sc->rx_rings[i]; 3906 3907 index = i & 0x7; /* Each IVAR has two entries */ 3908 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 3909 3910 if (i < 8) { 3911 ivar &= 0xffffff00; 3912 ivar |= 3913 (rxr->rx_intr_vec | E1000_IVAR_VALID); 3914 } else { 3915 ivar &= 0xff00ffff; 3916 ivar |= 3917 (rxr->rx_intr_vec | E1000_IVAR_VALID) << 16; 3918 } 3919 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 3920 } 3921 /* TX entries */ 3922 for (i = 0; i < sc->tx_ring_inuse; ++i) { 3923 txr = &sc->tx_rings[i]; 3924 3925 index = i & 0x7; /* Each IVAR has two entries */ 3926 ivar = E1000_READ_REG_ARRAY(hw, E1000_IVAR0, index); 3927 3928 if (i < 8) { 3929 ivar &= 0xffff00ff; 3930 ivar |= 3931 (txr->tx_intr_vec | E1000_IVAR_VALID) << 8; 3932 } else { 3933 ivar &= 0x00ffffff; 3934 ivar |= 3935 (txr->tx_intr_vec | E1000_IVAR_VALID) << 24; 3936 } 3937 E1000_WRITE_REG_ARRAY(hw, E1000_IVAR0, index, ivar); 3938 } 3939 if (sc->intr_type == PCI_INTR_TYPE_MSIX) { 3940 ivar = (sc->sts_msix_vec | E1000_IVAR_VALID) << 8; 3941 E1000_WRITE_REG(hw, E1000_IVAR_MISC, ivar); 3942 } 3943 break; 3944 3945 case e1000_82575: 3946 /* 3947 * Enable necessary interrupt bits. 3948 * 3949 * The name of the register is confusing; in addition to 3950 * configuring the first vector of MSI-X, it also configures 3951 * which bits of EICR could be set by the hardware even when 3952 * MSI or line interrupt is used; it thus controls interrupt 3953 * generation. It MUST be configured explicitly; the default 3954 * value mentioned in the datasheet is wrong: RX queue0 and 3955 * TX queue0 are NOT enabled by default. 3956 */ 3957 E1000_WRITE_REG(&sc->hw, E1000_MSIXBM(0), sc->intr_mask); 3958 break; 3959 3960 default: 3961 panic("unknown mac type %d\n", sc->hw.mac.type); 3962 } 3963 } 3964 3965 static int 3966 igb_setup_intr(struct igb_softc *sc) 3967 { 3968 int i; 3969 3970 for (i = 0; i < sc->intr_cnt; ++i) { 3971 struct igb_intr_data *intr = &sc->intr_data[i]; 3972 int error; 3973 3974 error = bus_setup_intr_descr(sc->dev, intr->intr_res, 3975 INTR_MPSAFE, intr->intr_func, intr->intr_funcarg, 3976 &intr->intr_hand, intr->intr_serialize, intr->intr_desc); 3977 if (error) { 3978 device_printf(sc->dev, "can't setup %dth intr\n", i); 3979 igb_teardown_intr(sc, i); 3980 return error; 3981 } 3982 } 3983 return 0; 3984 } 3985 3986 static void 3987 igb_set_txintr_mask(struct igb_tx_ring *txr, int *intr_vec0, int intr_vecmax) 3988 { 3989 if (txr->sc->hw.mac.type == e1000_82575) { 3990 txr->tx_intr_vec = 0; /* unused */ 3991 switch (txr->me) { 3992 case 0: 3993 txr->tx_intr_mask = E1000_EICR_TX_QUEUE0; 3994 break; 3995 case 1: 3996 txr->tx_intr_mask = E1000_EICR_TX_QUEUE1; 3997 break; 3998 case 2: 3999 txr->tx_intr_mask = E1000_EICR_TX_QUEUE2; 4000 break; 4001 case 3: 4002 txr->tx_intr_mask = E1000_EICR_TX_QUEUE3; 4003 break; 4004 default: 4005 panic("unsupported # of TX ring, %d\n", txr->me); 4006 } 4007 } else { 4008 int intr_vec = *intr_vec0; 4009 4010 txr->tx_intr_vec = intr_vec % intr_vecmax; 4011 txr->tx_intr_mask = 1 << txr->tx_intr_vec; 4012 4013 *intr_vec0 = intr_vec + 1; 4014 } 4015 } 4016 4017 static void 4018 igb_set_rxintr_mask(struct igb_rx_ring *rxr, int *intr_vec0, int intr_vecmax) 4019 { 4020 if (rxr->sc->hw.mac.type == e1000_82575) { 4021 rxr->rx_intr_vec = 0; /* unused */ 4022 switch (rxr->me) { 4023 case 0: 4024 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE0; 4025 break; 4026 case 1: 4027 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE1; 4028 break; 4029 case 2: 4030 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE2; 4031 break; 4032 case 3: 4033 rxr->rx_intr_mask = E1000_EICR_RX_QUEUE3; 4034 break; 4035 default: 4036 panic("unsupported # of RX ring, %d\n", rxr->me); 4037 } 4038 } else { 4039 int intr_vec = *intr_vec0; 4040 4041 rxr->rx_intr_vec = intr_vec % intr_vecmax; 4042 rxr->rx_intr_mask = 1 << rxr->rx_intr_vec; 4043 4044 *intr_vec0 = intr_vec + 1; 4045 } 4046 } 4047 4048 static void 4049 igb_serialize(struct ifnet *ifp, enum ifnet_serialize slz) 4050 { 4051 struct igb_softc *sc = ifp->if_softc; 4052 4053 ifnet_serialize_array_enter(sc->serializes, sc->serialize_cnt, slz); 4054 } 4055 4056 static void 4057 igb_deserialize(struct ifnet *ifp, enum ifnet_serialize slz) 4058 { 4059 struct igb_softc *sc = ifp->if_softc; 4060 4061 ifnet_serialize_array_exit(sc->serializes, sc->serialize_cnt, slz); 4062 } 4063 4064 static int 4065 igb_tryserialize(struct ifnet *ifp, enum ifnet_serialize slz) 4066 { 4067 struct igb_softc *sc = ifp->if_softc; 4068 4069 return ifnet_serialize_array_try(sc->serializes, sc->serialize_cnt, 4070 slz); 4071 } 4072 4073 #ifdef INVARIANTS 4074 4075 static void 4076 igb_serialize_assert(struct ifnet *ifp, enum ifnet_serialize slz, 4077 boolean_t serialized) 4078 { 4079 struct igb_softc *sc = ifp->if_softc; 4080 4081 ifnet_serialize_array_assert(sc->serializes, sc->serialize_cnt, 4082 slz, serialized); 4083 } 4084 4085 #endif /* INVARIANTS */ 4086 4087 static void 4088 igb_set_intr_mask(struct igb_softc *sc) 4089 { 4090 int i; 4091 4092 sc->intr_mask = sc->sts_intr_mask; 4093 for (i = 0; i < sc->rx_ring_inuse; ++i) 4094 sc->intr_mask |= sc->rx_rings[i].rx_intr_mask; 4095 for (i = 0; i < sc->tx_ring_inuse; ++i) 4096 sc->intr_mask |= sc->tx_rings[i].tx_intr_mask; 4097 if (bootverbose) { 4098 if_printf(&sc->arpcom.ac_if, "intr mask 0x%08x\n", 4099 sc->intr_mask); 4100 } 4101 } 4102 4103 static int 4104 igb_alloc_intr(struct igb_softc *sc) 4105 { 4106 struct igb_intr_data *intr; 4107 int i, intr_vec, intr_vecmax; 4108 u_int intr_flags; 4109 4110 igb_alloc_msix(sc); 4111 if (sc->intr_type == PCI_INTR_TYPE_MSIX) 4112 goto done; 4113 4114 if (sc->intr_data != NULL) 4115 kfree(sc->intr_data, M_DEVBUF); 4116 4117 sc->intr_cnt = 1; 4118 sc->intr_data = kmalloc(sizeof(struct igb_intr_data), M_DEVBUF, 4119 M_WAITOK | M_ZERO); 4120 intr = &sc->intr_data[0]; 4121 4122 /* 4123 * Allocate MSI/legacy interrupt resource 4124 */ 4125 sc->intr_type = pci_alloc_1intr(sc->dev, igb_msi_enable, 4126 &intr->intr_rid, &intr_flags); 4127 4128 if (sc->intr_type == PCI_INTR_TYPE_LEGACY) { 4129 int unshared; 4130 4131 unshared = device_getenv_int(sc->dev, "irq.unshared", 0); 4132 if (!unshared) { 4133 sc->flags |= IGB_FLAG_SHARED_INTR; 4134 if (bootverbose) 4135 device_printf(sc->dev, "IRQ shared\n"); 4136 } else { 4137 intr_flags &= ~RF_SHAREABLE; 4138 if (bootverbose) 4139 device_printf(sc->dev, "IRQ unshared\n"); 4140 } 4141 } 4142 4143 intr->intr_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, 4144 &intr->intr_rid, intr_flags); 4145 if (intr->intr_res == NULL) { 4146 device_printf(sc->dev, "Unable to allocate bus resource: " 4147 "interrupt\n"); 4148 return ENXIO; 4149 } 4150 4151 intr->intr_serialize = &sc->main_serialize; 4152 intr->intr_cpuid = rman_get_cpuid(intr->intr_res); 4153 intr->intr_func = (sc->flags & IGB_FLAG_SHARED_INTR) ? 4154 igb_intr_shared : igb_intr; 4155 intr->intr_funcarg = sc; 4156 intr->intr_rate = IGB_INTR_RATE; 4157 intr->intr_use = IGB_INTR_USE_RXTX; 4158 4159 for (i = 0; i < sc->tx_ring_cnt; ++i) 4160 sc->tx_rings[i].tx_intr_cpuid = intr->intr_cpuid; 4161 4162 /* 4163 * Setup MSI/legacy interrupt mask 4164 */ 4165 switch (sc->hw.mac.type) { 4166 case e1000_82575: 4167 intr_vecmax = IGB_MAX_TXRXINT_82575; 4168 break; 4169 4170 case e1000_82576: 4171 intr_vecmax = IGB_MAX_TXRXINT_82576; 4172 break; 4173 4174 case e1000_82580: 4175 intr_vecmax = IGB_MAX_TXRXINT_82580; 4176 break; 4177 4178 case e1000_i350: 4179 intr_vecmax = IGB_MAX_TXRXINT_I350; 4180 break; 4181 4182 case e1000_i354: 4183 intr_vecmax = IGB_MAX_TXRXINT_I354; 4184 break; 4185 4186 case e1000_i210: 4187 intr_vecmax = IGB_MAX_TXRXINT_I210; 4188 break; 4189 4190 case e1000_i211: 4191 intr_vecmax = IGB_MAX_TXRXINT_I211; 4192 break; 4193 4194 default: 4195 intr_vecmax = IGB_MIN_TXRXINT; 4196 break; 4197 } 4198 intr_vec = 0; 4199 for (i = 0; i < sc->tx_ring_cnt; ++i) 4200 igb_set_txintr_mask(&sc->tx_rings[i], &intr_vec, intr_vecmax); 4201 for (i = 0; i < sc->rx_ring_cnt; ++i) 4202 igb_set_rxintr_mask(&sc->rx_rings[i], &intr_vec, intr_vecmax); 4203 sc->sts_intr_mask = E1000_EICR_OTHER; 4204 done: 4205 igb_set_ring_inuse(sc, FALSE); 4206 igb_set_intr_mask(sc); 4207 return 0; 4208 } 4209 4210 static void 4211 igb_free_intr(struct igb_softc *sc) 4212 { 4213 if (sc->intr_data == NULL) 4214 return; 4215 4216 if (sc->intr_type != PCI_INTR_TYPE_MSIX) { 4217 struct igb_intr_data *intr = &sc->intr_data[0]; 4218 4219 KKASSERT(sc->intr_cnt == 1); 4220 if (intr->intr_res != NULL) { 4221 bus_release_resource(sc->dev, SYS_RES_IRQ, 4222 intr->intr_rid, intr->intr_res); 4223 } 4224 if (sc->intr_type == PCI_INTR_TYPE_MSI) 4225 pci_release_msi(sc->dev); 4226 4227 kfree(sc->intr_data, M_DEVBUF); 4228 } else { 4229 igb_free_msix(sc, TRUE); 4230 } 4231 } 4232 4233 static void 4234 igb_teardown_intr(struct igb_softc *sc, int intr_cnt) 4235 { 4236 int i; 4237 4238 if (sc->intr_data == NULL) 4239 return; 4240 4241 for (i = 0; i < intr_cnt; ++i) { 4242 struct igb_intr_data *intr = &sc->intr_data[i]; 4243 4244 bus_teardown_intr(sc->dev, intr->intr_res, intr->intr_hand); 4245 } 4246 } 4247 4248 static void 4249 igb_alloc_msix(struct igb_softc *sc) 4250 { 4251 int msix_enable, msix_cnt, msix_cnt2, alloc_cnt; 4252 int i, x, error; 4253 int offset, offset_def, agg_rxtx; 4254 struct igb_intr_data *intr; 4255 boolean_t aggregate, setup = FALSE; 4256 4257 /* 4258 * Don't enable MSI-X on 82575, see: 4259 * 82575 specification update errata #25 4260 */ 4261 if (sc->hw.mac.type == e1000_82575) 4262 return; 4263 4264 /* Don't enable MSI-X on VF */ 4265 if (sc->vf_ifp) 4266 return; 4267 4268 msix_enable = device_getenv_int(sc->dev, "msix.enable", 4269 igb_msix_enable); 4270 if (!msix_enable) 4271 return; 4272 4273 msix_cnt = pci_msix_count(sc->dev); 4274 #ifdef IGB_MSIX_DEBUG 4275 msix_cnt = device_getenv_int(sc->dev, "msix.count", msix_cnt); 4276 #endif 4277 if (msix_cnt <= 1) { 4278 /* One MSI-X model does not make sense */ 4279 return; 4280 } 4281 4282 i = 0; 4283 while ((1 << (i + 1)) <= msix_cnt) 4284 ++i; 4285 msix_cnt2 = 1 << i; 4286 4287 if (bootverbose) { 4288 device_printf(sc->dev, "MSI-X count %d/%d\n", 4289 msix_cnt2, msix_cnt); 4290 } 4291 4292 KKASSERT(msix_cnt2 <= msix_cnt); 4293 if (msix_cnt == msix_cnt2) { 4294 /* We need at least one MSI-X for link status */ 4295 msix_cnt2 >>= 1; 4296 if (msix_cnt2 <= 1) { 4297 /* One MSI-X for RX/TX does not make sense */ 4298 device_printf(sc->dev, "not enough MSI-X for TX/RX, " 4299 "MSI-X count %d/%d\n", msix_cnt2, msix_cnt); 4300 return; 4301 } 4302 KKASSERT(msix_cnt > msix_cnt2); 4303 4304 if (bootverbose) { 4305 device_printf(sc->dev, "MSI-X count fixup %d/%d\n", 4306 msix_cnt2, msix_cnt); 4307 } 4308 } 4309 4310 sc->rx_ring_msix = sc->rx_ring_cnt; 4311 if (sc->rx_ring_msix > msix_cnt2) 4312 sc->rx_ring_msix = msix_cnt2; 4313 4314 sc->tx_ring_msix = sc->tx_ring_cnt; 4315 if (sc->tx_ring_msix > msix_cnt2) 4316 sc->tx_ring_msix = msix_cnt2; 4317 4318 /* Allow user to force independent RX/TX MSI-X handling */ 4319 agg_rxtx = device_getenv_int(sc->dev, "msix.agg_rxtx", 4320 igb_msix_agg_rxtx); 4321 4322 if (!agg_rxtx && msix_cnt >= sc->tx_ring_msix + sc->rx_ring_msix + 1) { 4323 /* 4324 * Independent TX/RX MSI-X 4325 */ 4326 aggregate = FALSE; 4327 if (bootverbose) 4328 device_printf(sc->dev, "independent TX/RX MSI-X\n"); 4329 alloc_cnt = sc->tx_ring_msix + sc->rx_ring_msix; 4330 } else { 4331 /* 4332 * Aggregate TX/RX MSI-X 4333 */ 4334 aggregate = TRUE; 4335 if (bootverbose) 4336 device_printf(sc->dev, "aggregate TX/RX MSI-X\n"); 4337 alloc_cnt = msix_cnt2; 4338 if (alloc_cnt > ncpus2) 4339 alloc_cnt = ncpus2; 4340 if (sc->rx_ring_msix > alloc_cnt) 4341 sc->rx_ring_msix = alloc_cnt; 4342 if (sc->tx_ring_msix > alloc_cnt) 4343 sc->tx_ring_msix = alloc_cnt; 4344 } 4345 ++alloc_cnt; /* For link status */ 4346 4347 if (bootverbose) { 4348 device_printf(sc->dev, "MSI-X alloc %d, " 4349 "RX ring %d, TX ring %d\n", alloc_cnt, 4350 sc->rx_ring_msix, sc->tx_ring_msix); 4351 } 4352 4353 sc->msix_mem_rid = PCIR_BAR(IGB_MSIX_BAR); 4354 sc->msix_mem_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY, 4355 &sc->msix_mem_rid, RF_ACTIVE); 4356 if (sc->msix_mem_res == NULL) { 4357 sc->msix_mem_rid = PCIR_BAR(IGB_MSIX_BAR_ALT); 4358 sc->msix_mem_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY, 4359 &sc->msix_mem_rid, RF_ACTIVE); 4360 if (sc->msix_mem_res == NULL) { 4361 device_printf(sc->dev, "Unable to map MSI-X table\n"); 4362 return; 4363 } 4364 } 4365 4366 sc->intr_cnt = alloc_cnt; 4367 sc->intr_data = kmalloc(sizeof(struct igb_intr_data) * sc->intr_cnt, 4368 M_DEVBUF, M_WAITOK | M_ZERO); 4369 for (x = 0; x < sc->intr_cnt; ++x) { 4370 intr = &sc->intr_data[x]; 4371 intr->intr_rid = -1; 4372 intr->intr_rate = IGB_INTR_RATE; 4373 } 4374 4375 x = 0; 4376 if (!aggregate) { 4377 /* 4378 * RX rings 4379 */ 4380 if (sc->rx_ring_msix == ncpus2) { 4381 offset = 0; 4382 } else { 4383 offset_def = (sc->rx_ring_msix * 4384 device_get_unit(sc->dev)) % ncpus2; 4385 4386 offset = device_getenv_int(sc->dev, 4387 "msix.rxoff", offset_def); 4388 if (offset >= ncpus2 || 4389 offset % sc->rx_ring_msix != 0) { 4390 device_printf(sc->dev, 4391 "invalid msix.rxoff %d, use %d\n", 4392 offset, offset_def); 4393 offset = offset_def; 4394 } 4395 } 4396 igb_msix_rx_conf(sc, 0, &x, offset); 4397 4398 /* 4399 * TX rings 4400 */ 4401 if (sc->tx_ring_msix == ncpus2) { 4402 offset = 0; 4403 } else { 4404 offset_def = (sc->tx_ring_msix * 4405 device_get_unit(sc->dev)) % ncpus2; 4406 4407 offset = device_getenv_int(sc->dev, 4408 "msix.txoff", offset_def); 4409 if (offset >= ncpus2 || 4410 offset % sc->tx_ring_msix != 0) { 4411 device_printf(sc->dev, 4412 "invalid msix.txoff %d, use %d\n", 4413 offset, offset_def); 4414 offset = offset_def; 4415 } 4416 } 4417 igb_msix_tx_conf(sc, 0, &x, offset); 4418 } else { 4419 int ring_agg, ring_max; 4420 4421 ring_agg = sc->rx_ring_msix; 4422 if (ring_agg > sc->tx_ring_msix) 4423 ring_agg = sc->tx_ring_msix; 4424 4425 ring_max = sc->rx_ring_msix; 4426 if (ring_max < sc->tx_ring_msix) 4427 ring_max = sc->tx_ring_msix; 4428 4429 if (ring_max == ncpus2) { 4430 offset = 0; 4431 } else { 4432 offset_def = (ring_max * device_get_unit(sc->dev)) % 4433 ncpus2; 4434 4435 offset = device_getenv_int(sc->dev, "msix.off", 4436 offset_def); 4437 if (offset >= ncpus2 || offset % ring_max != 0) { 4438 device_printf(sc->dev, 4439 "invalid msix.off %d, use %d\n", 4440 offset, offset_def); 4441 offset = offset_def; 4442 } 4443 } 4444 4445 for (i = 0; i < ring_agg; ++i) { 4446 struct igb_tx_ring *txr = &sc->tx_rings[i]; 4447 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 4448 4449 KKASSERT(x < sc->intr_cnt); 4450 rxr->rx_intr_vec = x; 4451 rxr->rx_intr_mask = 1 << rxr->rx_intr_vec; 4452 rxr->rx_txr = txr; 4453 txr->tx_intr_vec = rxr->rx_intr_vec; 4454 txr->tx_intr_mask = rxr->rx_intr_mask; 4455 4456 intr = &sc->intr_data[x++]; 4457 4458 intr->intr_serialize = &rxr->rx_serialize; 4459 intr->intr_func = igb_msix_rxtx; 4460 intr->intr_funcarg = rxr; 4461 intr->intr_use = IGB_INTR_USE_RXTX; 4462 4463 intr->intr_cpuid = i + offset; 4464 KKASSERT(intr->intr_cpuid < ncpus2); 4465 txr->tx_intr_cpuid = intr->intr_cpuid; 4466 4467 ksnprintf(intr->intr_desc0, sizeof(intr->intr_desc0), 4468 "%s rxtx%d", device_get_nameunit(sc->dev), i); 4469 intr->intr_desc = intr->intr_desc0; 4470 } 4471 4472 if (ring_agg != ring_max) { 4473 if (ring_max == sc->tx_ring_msix) 4474 igb_msix_tx_conf(sc, i, &x, offset); 4475 else 4476 igb_msix_rx_conf(sc, i, &x, offset); 4477 } 4478 } 4479 4480 /* 4481 * Link status 4482 */ 4483 KKASSERT(x < sc->intr_cnt); 4484 sc->sts_msix_vec = x; 4485 sc->sts_intr_mask = 1 << sc->sts_msix_vec; 4486 4487 intr = &sc->intr_data[x++]; 4488 4489 intr->intr_serialize = &sc->main_serialize; 4490 intr->intr_func = igb_msix_status; 4491 intr->intr_funcarg = sc; 4492 intr->intr_cpuid = 0; 4493 intr->intr_use = IGB_INTR_USE_STATUS; 4494 4495 ksnprintf(intr->intr_desc0, sizeof(intr->intr_desc0), "%s sts", 4496 device_get_nameunit(sc->dev)); 4497 intr->intr_desc = intr->intr_desc0; 4498 4499 KKASSERT(x == sc->intr_cnt); 4500 4501 error = pci_setup_msix(sc->dev); 4502 if (error) { 4503 device_printf(sc->dev, "Setup MSI-X failed\n"); 4504 goto back; 4505 } 4506 setup = TRUE; 4507 4508 for (i = 0; i < sc->intr_cnt; ++i) { 4509 intr = &sc->intr_data[i]; 4510 4511 error = pci_alloc_msix_vector(sc->dev, i, &intr->intr_rid, 4512 intr->intr_cpuid); 4513 if (error) { 4514 device_printf(sc->dev, 4515 "Unable to allocate MSI-X %d on cpu%d\n", i, 4516 intr->intr_cpuid); 4517 goto back; 4518 } 4519 4520 intr->intr_res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, 4521 &intr->intr_rid, RF_ACTIVE); 4522 if (intr->intr_res == NULL) { 4523 device_printf(sc->dev, 4524 "Unable to allocate MSI-X %d resource\n", i); 4525 error = ENOMEM; 4526 goto back; 4527 } 4528 } 4529 4530 pci_enable_msix(sc->dev); 4531 sc->intr_type = PCI_INTR_TYPE_MSIX; 4532 back: 4533 if (error) 4534 igb_free_msix(sc, setup); 4535 } 4536 4537 static void 4538 igb_free_msix(struct igb_softc *sc, boolean_t setup) 4539 { 4540 int i; 4541 4542 KKASSERT(sc->intr_cnt > 1); 4543 4544 for (i = 0; i < sc->intr_cnt; ++i) { 4545 struct igb_intr_data *intr = &sc->intr_data[i]; 4546 4547 if (intr->intr_res != NULL) { 4548 bus_release_resource(sc->dev, SYS_RES_IRQ, 4549 intr->intr_rid, intr->intr_res); 4550 } 4551 if (intr->intr_rid >= 0) 4552 pci_release_msix_vector(sc->dev, intr->intr_rid); 4553 } 4554 if (setup) 4555 pci_teardown_msix(sc->dev); 4556 4557 sc->intr_cnt = 0; 4558 kfree(sc->intr_data, M_DEVBUF); 4559 sc->intr_data = NULL; 4560 } 4561 4562 static void 4563 igb_msix_rx(void *arg) 4564 { 4565 struct igb_rx_ring *rxr = arg; 4566 4567 ASSERT_SERIALIZED(&rxr->rx_serialize); 4568 igb_rxeof(rxr, -1); 4569 4570 E1000_WRITE_REG(&rxr->sc->hw, E1000_EIMS, rxr->rx_intr_mask); 4571 } 4572 4573 static void 4574 igb_msix_tx(void *arg) 4575 { 4576 struct igb_tx_ring *txr = arg; 4577 4578 ASSERT_SERIALIZED(&txr->tx_serialize); 4579 4580 igb_txeof(txr, *(txr->tx_hdr)); 4581 if (!ifsq_is_empty(txr->ifsq)) 4582 ifsq_devstart(txr->ifsq); 4583 4584 E1000_WRITE_REG(&txr->sc->hw, E1000_EIMS, txr->tx_intr_mask); 4585 } 4586 4587 static void 4588 igb_msix_status(void *arg) 4589 { 4590 struct igb_softc *sc = arg; 4591 uint32_t icr; 4592 4593 ASSERT_SERIALIZED(&sc->main_serialize); 4594 4595 icr = E1000_READ_REG(&sc->hw, E1000_ICR); 4596 if (icr & E1000_ICR_LSC) { 4597 sc->hw.mac.get_link_status = 1; 4598 igb_update_link_status(sc); 4599 } 4600 4601 E1000_WRITE_REG(&sc->hw, E1000_EIMS, sc->sts_intr_mask); 4602 } 4603 4604 static void 4605 igb_set_ring_inuse(struct igb_softc *sc, boolean_t polling) 4606 { 4607 sc->rx_ring_inuse = igb_get_rxring_inuse(sc, polling); 4608 sc->tx_ring_inuse = igb_get_txring_inuse(sc, polling); 4609 if (bootverbose) { 4610 if_printf(&sc->arpcom.ac_if, "RX rings %d/%d, TX rings %d/%d\n", 4611 sc->rx_ring_inuse, sc->rx_ring_cnt, 4612 sc->tx_ring_inuse, sc->tx_ring_cnt); 4613 } 4614 } 4615 4616 static int 4617 igb_get_rxring_inuse(const struct igb_softc *sc, boolean_t polling) 4618 { 4619 if (!IGB_ENABLE_HWRSS(sc)) 4620 return 1; 4621 4622 if (polling) 4623 return sc->rx_ring_cnt; 4624 else if (sc->intr_type != PCI_INTR_TYPE_MSIX) 4625 return IGB_MIN_RING_RSS; 4626 else 4627 return sc->rx_ring_msix; 4628 } 4629 4630 static int 4631 igb_get_txring_inuse(const struct igb_softc *sc, boolean_t polling) 4632 { 4633 if (!IGB_ENABLE_HWTSS(sc)) 4634 return 1; 4635 4636 if (polling) 4637 return sc->tx_ring_cnt; 4638 else if (sc->intr_type != PCI_INTR_TYPE_MSIX) 4639 return IGB_MIN_RING; 4640 else 4641 return sc->tx_ring_msix; 4642 } 4643 4644 static int 4645 igb_tso_pullup(struct igb_tx_ring *txr, struct mbuf **mp) 4646 { 4647 int hoff, iphlen, thoff; 4648 struct mbuf *m; 4649 4650 m = *mp; 4651 KASSERT(M_WRITABLE(m), ("TSO mbuf not writable")); 4652 4653 iphlen = m->m_pkthdr.csum_iphlen; 4654 thoff = m->m_pkthdr.csum_thlen; 4655 hoff = m->m_pkthdr.csum_lhlen; 4656 4657 KASSERT(iphlen > 0, ("invalid ip hlen")); 4658 KASSERT(thoff > 0, ("invalid tcp hlen")); 4659 KASSERT(hoff > 0, ("invalid ether hlen")); 4660 4661 if (__predict_false(m->m_len < hoff + iphlen + thoff)) { 4662 m = m_pullup(m, hoff + iphlen + thoff); 4663 if (m == NULL) { 4664 *mp = NULL; 4665 return ENOBUFS; 4666 } 4667 *mp = m; 4668 } 4669 if (txr->tx_flags & IGB_TXFLAG_TSO_IPLEN0) { 4670 struct ip *ip; 4671 4672 ip = mtodoff(m, struct ip *, hoff); 4673 ip->ip_len = 0; 4674 } 4675 4676 return 0; 4677 } 4678 4679 static void 4680 igb_tso_ctx(struct igb_tx_ring *txr, struct mbuf *m, uint32_t *hlen) 4681 { 4682 struct e1000_adv_tx_context_desc *TXD; 4683 uint32_t vlan_macip_lens, type_tucmd_mlhl, mss_l4len_idx; 4684 int hoff, ctxd, iphlen, thoff; 4685 4686 iphlen = m->m_pkthdr.csum_iphlen; 4687 thoff = m->m_pkthdr.csum_thlen; 4688 hoff = m->m_pkthdr.csum_lhlen; 4689 4690 vlan_macip_lens = type_tucmd_mlhl = mss_l4len_idx = 0; 4691 4692 ctxd = txr->next_avail_desc; 4693 TXD = (struct e1000_adv_tx_context_desc *)&txr->tx_base[ctxd]; 4694 4695 if (m->m_flags & M_VLANTAG) { 4696 uint16_t vlantag; 4697 4698 vlantag = htole16(m->m_pkthdr.ether_vlantag); 4699 vlan_macip_lens |= (vlantag << E1000_ADVTXD_VLAN_SHIFT); 4700 } 4701 4702 vlan_macip_lens |= (hoff << E1000_ADVTXD_MACLEN_SHIFT); 4703 vlan_macip_lens |= iphlen; 4704 4705 type_tucmd_mlhl |= E1000_ADVTXD_DCMD_DEXT | E1000_ADVTXD_DTYP_CTXT; 4706 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_L4T_TCP; 4707 type_tucmd_mlhl |= E1000_ADVTXD_TUCMD_IPV4; 4708 4709 mss_l4len_idx |= (m->m_pkthdr.tso_segsz << E1000_ADVTXD_MSS_SHIFT); 4710 mss_l4len_idx |= (thoff << E1000_ADVTXD_L4LEN_SHIFT); 4711 4712 /* 4713 * 82575 needs the TX context index added; the queue 4714 * index is used as TX context index here. 4715 */ 4716 if (txr->sc->hw.mac.type == e1000_82575) 4717 mss_l4len_idx |= txr->me << 4; 4718 4719 TXD->vlan_macip_lens = htole32(vlan_macip_lens); 4720 TXD->type_tucmd_mlhl = htole32(type_tucmd_mlhl); 4721 TXD->seqnum_seed = htole32(0); 4722 TXD->mss_l4len_idx = htole32(mss_l4len_idx); 4723 4724 /* We've consumed the first desc, adjust counters */ 4725 if (++ctxd == txr->num_tx_desc) 4726 ctxd = 0; 4727 txr->next_avail_desc = ctxd; 4728 --txr->tx_avail; 4729 4730 *hlen = hoff + iphlen + thoff; 4731 } 4732 4733 static void 4734 igb_setup_serialize(struct igb_softc *sc) 4735 { 4736 int i = 0, j; 4737 4738 /* Main + RX + TX */ 4739 sc->serialize_cnt = 1 + sc->rx_ring_cnt + sc->tx_ring_cnt; 4740 sc->serializes = 4741 kmalloc(sc->serialize_cnt * sizeof(struct lwkt_serialize *), 4742 M_DEVBUF, M_WAITOK | M_ZERO); 4743 4744 /* 4745 * Setup serializes 4746 * 4747 * NOTE: Order is critical 4748 */ 4749 4750 KKASSERT(i < sc->serialize_cnt); 4751 sc->serializes[i++] = &sc->main_serialize; 4752 4753 for (j = 0; j < sc->rx_ring_cnt; ++j) { 4754 KKASSERT(i < sc->serialize_cnt); 4755 sc->serializes[i++] = &sc->rx_rings[j].rx_serialize; 4756 } 4757 4758 for (j = 0; j < sc->tx_ring_cnt; ++j) { 4759 KKASSERT(i < sc->serialize_cnt); 4760 sc->serializes[i++] = &sc->tx_rings[j].tx_serialize; 4761 } 4762 4763 KKASSERT(i == sc->serialize_cnt); 4764 } 4765 4766 static void 4767 igb_msix_rx_conf(struct igb_softc *sc, int i, int *x0, int offset) 4768 { 4769 int x = *x0; 4770 4771 for (; i < sc->rx_ring_msix; ++i) { 4772 struct igb_rx_ring *rxr = &sc->rx_rings[i]; 4773 struct igb_intr_data *intr; 4774 4775 KKASSERT(x < sc->intr_cnt); 4776 rxr->rx_intr_vec = x; 4777 rxr->rx_intr_mask = 1 << rxr->rx_intr_vec; 4778 4779 intr = &sc->intr_data[x++]; 4780 4781 intr->intr_serialize = &rxr->rx_serialize; 4782 intr->intr_func = igb_msix_rx; 4783 intr->intr_funcarg = rxr; 4784 intr->intr_rate = IGB_MSIX_RX_RATE; 4785 intr->intr_use = IGB_INTR_USE_RX; 4786 4787 intr->intr_cpuid = i + offset; 4788 KKASSERT(intr->intr_cpuid < ncpus2); 4789 4790 ksnprintf(intr->intr_desc0, sizeof(intr->intr_desc0), "%s rx%d", 4791 device_get_nameunit(sc->dev), i); 4792 intr->intr_desc = intr->intr_desc0; 4793 } 4794 *x0 = x; 4795 } 4796 4797 static void 4798 igb_msix_tx_conf(struct igb_softc *sc, int i, int *x0, int offset) 4799 { 4800 int x = *x0; 4801 4802 for (; i < sc->tx_ring_msix; ++i) { 4803 struct igb_tx_ring *txr = &sc->tx_rings[i]; 4804 struct igb_intr_data *intr; 4805 4806 KKASSERT(x < sc->intr_cnt); 4807 txr->tx_intr_vec = x; 4808 txr->tx_intr_mask = 1 << txr->tx_intr_vec; 4809 4810 intr = &sc->intr_data[x++]; 4811 4812 intr->intr_serialize = &txr->tx_serialize; 4813 intr->intr_func = igb_msix_tx; 4814 intr->intr_funcarg = txr; 4815 intr->intr_rate = IGB_MSIX_TX_RATE; 4816 intr->intr_use = IGB_INTR_USE_TX; 4817 4818 intr->intr_cpuid = i + offset; 4819 KKASSERT(intr->intr_cpuid < ncpus2); 4820 txr->tx_intr_cpuid = intr->intr_cpuid; 4821 4822 ksnprintf(intr->intr_desc0, sizeof(intr->intr_desc0), "%s tx%d", 4823 device_get_nameunit(sc->dev), i); 4824 intr->intr_desc = intr->intr_desc0; 4825 } 4826 *x0 = x; 4827 } 4828 4829 static void 4830 igb_msix_rxtx(void *arg) 4831 { 4832 struct igb_rx_ring *rxr = arg; 4833 struct igb_tx_ring *txr; 4834 int hdr; 4835 4836 ASSERT_SERIALIZED(&rxr->rx_serialize); 4837 4838 igb_rxeof(rxr, -1); 4839 4840 /* 4841 * NOTE: 4842 * Since next_to_clean is only changed by igb_txeof(), 4843 * which is called only in interrupt handler, the 4844 * check w/o holding tx serializer is MPSAFE. 4845 */ 4846 txr = rxr->rx_txr; 4847 hdr = *(txr->tx_hdr); 4848 if (hdr != txr->next_to_clean) { 4849 lwkt_serialize_enter(&txr->tx_serialize); 4850 igb_txeof(txr, hdr); 4851 if (!ifsq_is_empty(txr->ifsq)) 4852 ifsq_devstart(txr->ifsq); 4853 lwkt_serialize_exit(&txr->tx_serialize); 4854 } 4855 4856 E1000_WRITE_REG(&rxr->sc->hw, E1000_EIMS, rxr->rx_intr_mask); 4857 } 4858 4859 static void 4860 igb_set_timer_cpuid(struct igb_softc *sc, boolean_t polling) 4861 { 4862 if (polling || sc->intr_type == PCI_INTR_TYPE_MSIX) 4863 sc->timer_cpuid = 0; /* XXX fixed */ 4864 else 4865 sc->timer_cpuid = rman_get_cpuid(sc->intr_data[0].intr_res); 4866 } 4867 4868 static void 4869 igb_init_dmac(struct igb_softc *sc, uint32_t pba) 4870 { 4871 struct e1000_hw *hw = &sc->hw; 4872 uint32_t reg; 4873 4874 if (hw->mac.type == e1000_i211) 4875 return; 4876 4877 if (hw->mac.type > e1000_82580) { 4878 uint32_t dmac; 4879 uint16_t hwm; 4880 4881 if (sc->dma_coalesce == 0) { /* Disabling it */ 4882 reg = ~E1000_DMACR_DMAC_EN; 4883 E1000_WRITE_REG(hw, E1000_DMACR, reg); 4884 return; 4885 } else { 4886 if_printf(&sc->arpcom.ac_if, 4887 "DMA Coalescing enabled\n"); 4888 } 4889 4890 /* Set starting threshold */ 4891 E1000_WRITE_REG(hw, E1000_DMCTXTH, 0); 4892 4893 hwm = 64 * pba - sc->max_frame_size / 16; 4894 if (hwm < 64 * (pba - 6)) 4895 hwm = 64 * (pba - 6); 4896 reg = E1000_READ_REG(hw, E1000_FCRTC); 4897 reg &= ~E1000_FCRTC_RTH_COAL_MASK; 4898 reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT) 4899 & E1000_FCRTC_RTH_COAL_MASK); 4900 E1000_WRITE_REG(hw, E1000_FCRTC, reg); 4901 4902 dmac = pba - sc->max_frame_size / 512; 4903 if (dmac < pba - 10) 4904 dmac = pba - 10; 4905 reg = E1000_READ_REG(hw, E1000_DMACR); 4906 reg &= ~E1000_DMACR_DMACTHR_MASK; 4907 reg |= ((dmac << E1000_DMACR_DMACTHR_SHIFT) 4908 & E1000_DMACR_DMACTHR_MASK); 4909 4910 /* transition to L0x or L1 if available..*/ 4911 reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK); 4912 4913 /* 4914 * Check if status is 2.5Gb backplane connection 4915 * before configuration of watchdog timer, which 4916 * is in msec values in 12.8usec intervals watchdog 4917 * timer = msec values in 32usec intervals for non 4918 * 2.5Gb connection. 4919 */ 4920 if (hw->mac.type == e1000_i354) { 4921 int status = E1000_READ_REG(hw, E1000_STATUS); 4922 4923 if ((status & E1000_STATUS_2P5_SKU) && 4924 !(status & E1000_STATUS_2P5_SKU_OVER)) 4925 reg |= ((sc->dma_coalesce * 5) >> 6); 4926 else 4927 reg |= (sc->dma_coalesce >> 5); 4928 } else { 4929 reg |= (sc->dma_coalesce >> 5); 4930 } 4931 4932 E1000_WRITE_REG(hw, E1000_DMACR, reg); 4933 4934 E1000_WRITE_REG(hw, E1000_DMCRTRH, 0); 4935 4936 /* Set the interval before transition */ 4937 reg = E1000_READ_REG(hw, E1000_DMCTLX); 4938 if (hw->mac.type == e1000_i350) 4939 reg |= IGB_DMCTLX_DCFLUSH_DIS; 4940 /* 4941 * In 2.5Gb connection, TTLX unit is 0.4 usec, which 4942 * is 0x4*2 = 0xA. But delay is still 4 usec. 4943 */ 4944 if (hw->mac.type == e1000_i354) { 4945 int status = E1000_READ_REG(hw, E1000_STATUS); 4946 4947 if ((status & E1000_STATUS_2P5_SKU) && 4948 !(status & E1000_STATUS_2P5_SKU_OVER)) 4949 reg |= 0xA; 4950 else 4951 reg |= 0x4; 4952 } else { 4953 reg |= 0x4; 4954 } 4955 E1000_WRITE_REG(hw, E1000_DMCTLX, reg); 4956 4957 /* Free space in tx packet buffer to wake from DMA coal */ 4958 E1000_WRITE_REG(hw, E1000_DMCTXTH, 4959 (IGB_TXPBSIZE - (2 * sc->max_frame_size)) >> 6); 4960 4961 /* Make low power state decision controlled by DMA coal */ 4962 reg = E1000_READ_REG(hw, E1000_PCIEMISC); 4963 reg &= ~E1000_PCIEMISC_LX_DECISION; 4964 E1000_WRITE_REG(hw, E1000_PCIEMISC, reg); 4965 } else if (hw->mac.type == e1000_82580) { 4966 reg = E1000_READ_REG(hw, E1000_PCIEMISC); 4967 E1000_WRITE_REG(hw, E1000_PCIEMISC, 4968 reg & ~E1000_PCIEMISC_LX_DECISION); 4969 E1000_WRITE_REG(hw, E1000_DMACR, 0); 4970 } 4971 } 4972