1 /* 2 * Copyright (c) 2004 Joerg Sonnenberger <joerg@bec.de>. All rights reserved. 3 * 4 * Copyright (c) 2001-2008, Intel Corporation 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions are met: 9 * 10 * 1. Redistributions of source code must retain the above copyright notice, 11 * this list of conditions and the following disclaimer. 12 * 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * 3. Neither the name of the Intel Corporation nor the names of its 18 * contributors may be used to endorse or promote products derived from 19 * this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 22 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE 25 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 31 * POSSIBILITY OF SUCH DAMAGE. 32 * 33 * 34 * Copyright (c) 2005 The DragonFly Project. All rights reserved. 35 * 36 * This code is derived from software contributed to The DragonFly Project 37 * by Matthew Dillon <dillon@backplane.com> 38 * 39 * Redistribution and use in source and binary forms, with or without 40 * modification, are permitted provided that the following conditions 41 * are met: 42 * 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in 47 * the documentation and/or other materials provided with the 48 * distribution. 49 * 3. Neither the name of The DragonFly Project nor the names of its 50 * contributors may be used to endorse or promote products derived 51 * from this software without specific, prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 54 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 55 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 56 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 57 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 58 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 59 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 60 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 61 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 62 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 63 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 */ 66 67 #include "opt_ifpoll.h" 68 #include "opt_emx.h" 69 70 #include <sys/param.h> 71 #include <sys/bus.h> 72 #include <sys/endian.h> 73 #include <sys/interrupt.h> 74 #include <sys/kernel.h> 75 #include <sys/ktr.h> 76 #include <sys/malloc.h> 77 #include <sys/mbuf.h> 78 #include <sys/proc.h> 79 #include <sys/rman.h> 80 #include <sys/serialize.h> 81 #include <sys/serialize2.h> 82 #include <sys/socket.h> 83 #include <sys/sockio.h> 84 #include <sys/sysctl.h> 85 #include <sys/systm.h> 86 87 #include <net/bpf.h> 88 #include <net/ethernet.h> 89 #include <net/if.h> 90 #include <net/if_arp.h> 91 #include <net/if_dl.h> 92 #include <net/if_media.h> 93 #include <net/ifq_var.h> 94 #include <net/toeplitz.h> 95 #include <net/toeplitz2.h> 96 #include <net/vlan/if_vlan_var.h> 97 #include <net/vlan/if_vlan_ether.h> 98 #include <net/if_poll.h> 99 100 #include <netinet/in_systm.h> 101 #include <netinet/in.h> 102 #include <netinet/ip.h> 103 #include <netinet/tcp.h> 104 #include <netinet/udp.h> 105 106 #include <bus/pci/pcivar.h> 107 #include <bus/pci/pcireg.h> 108 109 #include <dev/netif/ig_hal/e1000_api.h> 110 #include <dev/netif/ig_hal/e1000_82571.h> 111 #include <dev/netif/emx/if_emx.h> 112 113 #ifdef EMX_RSS_DEBUG 114 #define EMX_RSS_DPRINTF(sc, lvl, fmt, ...) \ 115 do { \ 116 if (sc->rss_debug >= lvl) \ 117 if_printf(&sc->arpcom.ac_if, fmt, __VA_ARGS__); \ 118 } while (0) 119 #else /* !EMX_RSS_DEBUG */ 120 #define EMX_RSS_DPRINTF(sc, lvl, fmt, ...) ((void)0) 121 #endif /* EMX_RSS_DEBUG */ 122 123 #define EMX_TX_SERIALIZE 1 124 #define EMX_RX_SERIALIZE 3 125 126 #define EMX_NAME "Intel(R) PRO/1000 " 127 128 #define EMX_DEVICE(id) \ 129 { EMX_VENDOR_ID, E1000_DEV_ID_##id, EMX_NAME #id } 130 #define EMX_DEVICE_NULL { 0, 0, NULL } 131 132 static const struct emx_device { 133 uint16_t vid; 134 uint16_t did; 135 const char *desc; 136 } emx_devices[] = { 137 EMX_DEVICE(82571EB_COPPER), 138 EMX_DEVICE(82571EB_FIBER), 139 EMX_DEVICE(82571EB_SERDES), 140 EMX_DEVICE(82571EB_SERDES_DUAL), 141 EMX_DEVICE(82571EB_SERDES_QUAD), 142 EMX_DEVICE(82571EB_QUAD_COPPER), 143 EMX_DEVICE(82571EB_QUAD_COPPER_BP), 144 EMX_DEVICE(82571EB_QUAD_COPPER_LP), 145 EMX_DEVICE(82571EB_QUAD_FIBER), 146 EMX_DEVICE(82571PT_QUAD_COPPER), 147 148 EMX_DEVICE(82572EI_COPPER), 149 EMX_DEVICE(82572EI_FIBER), 150 EMX_DEVICE(82572EI_SERDES), 151 EMX_DEVICE(82572EI), 152 153 EMX_DEVICE(82573E), 154 EMX_DEVICE(82573E_IAMT), 155 EMX_DEVICE(82573L), 156 157 EMX_DEVICE(80003ES2LAN_COPPER_SPT), 158 EMX_DEVICE(80003ES2LAN_SERDES_SPT), 159 EMX_DEVICE(80003ES2LAN_COPPER_DPT), 160 EMX_DEVICE(80003ES2LAN_SERDES_DPT), 161 162 EMX_DEVICE(82574L), 163 EMX_DEVICE(82574LA), 164 165 /* required last entry */ 166 EMX_DEVICE_NULL 167 }; 168 169 static int emx_probe(device_t); 170 static int emx_attach(device_t); 171 static int emx_detach(device_t); 172 static int emx_shutdown(device_t); 173 static int emx_suspend(device_t); 174 static int emx_resume(device_t); 175 176 static void emx_init(void *); 177 static void emx_stop(struct emx_softc *); 178 static int emx_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *); 179 static void emx_start(struct ifnet *, struct ifaltq_subque *); 180 #ifdef IFPOLL_ENABLE 181 static void emx_npoll(struct ifnet *, struct ifpoll_info *); 182 static void emx_npoll_status(struct ifnet *); 183 static void emx_npoll_tx(struct ifnet *, void *, int); 184 static void emx_npoll_rx(struct ifnet *, void *, int); 185 #endif 186 static void emx_watchdog(struct ifaltq_subque *); 187 static void emx_media_status(struct ifnet *, struct ifmediareq *); 188 static int emx_media_change(struct ifnet *); 189 static void emx_timer(void *); 190 static void emx_serialize(struct ifnet *, enum ifnet_serialize); 191 static void emx_deserialize(struct ifnet *, enum ifnet_serialize); 192 static int emx_tryserialize(struct ifnet *, enum ifnet_serialize); 193 #ifdef INVARIANTS 194 static void emx_serialize_assert(struct ifnet *, enum ifnet_serialize, 195 boolean_t); 196 #endif 197 198 static void emx_intr(void *); 199 static void emx_intr_mask(void *); 200 static void emx_intr_body(struct emx_softc *, boolean_t); 201 static void emx_rxeof(struct emx_rxdata *, int); 202 static void emx_txeof(struct emx_txdata *); 203 static void emx_tx_collect(struct emx_txdata *); 204 static void emx_tx_purge(struct emx_softc *); 205 static void emx_enable_intr(struct emx_softc *); 206 static void emx_disable_intr(struct emx_softc *); 207 208 static int emx_dma_alloc(struct emx_softc *); 209 static void emx_dma_free(struct emx_softc *); 210 static void emx_init_tx_ring(struct emx_txdata *); 211 static int emx_init_rx_ring(struct emx_rxdata *); 212 static void emx_free_tx_ring(struct emx_txdata *); 213 static void emx_free_rx_ring(struct emx_rxdata *); 214 static int emx_create_tx_ring(struct emx_txdata *); 215 static int emx_create_rx_ring(struct emx_rxdata *); 216 static void emx_destroy_tx_ring(struct emx_txdata *, int); 217 static void emx_destroy_rx_ring(struct emx_rxdata *, int); 218 static int emx_newbuf(struct emx_rxdata *, int, int); 219 static int emx_encap(struct emx_txdata *, struct mbuf **, int *, int *); 220 static int emx_txcsum(struct emx_txdata *, struct mbuf *, 221 uint32_t *, uint32_t *); 222 static int emx_tso_pullup(struct emx_txdata *, struct mbuf **); 223 static int emx_tso_setup(struct emx_txdata *, struct mbuf *, 224 uint32_t *, uint32_t *); 225 static int emx_get_txring_inuse(const struct emx_softc *, boolean_t); 226 227 static int emx_is_valid_eaddr(const uint8_t *); 228 static int emx_reset(struct emx_softc *); 229 static void emx_setup_ifp(struct emx_softc *); 230 static void emx_init_tx_unit(struct emx_softc *); 231 static void emx_init_rx_unit(struct emx_softc *); 232 static void emx_update_stats(struct emx_softc *); 233 static void emx_set_promisc(struct emx_softc *); 234 static void emx_disable_promisc(struct emx_softc *); 235 static void emx_set_multi(struct emx_softc *); 236 static void emx_update_link_status(struct emx_softc *); 237 static void emx_smartspeed(struct emx_softc *); 238 static void emx_set_itr(struct emx_softc *, uint32_t); 239 static void emx_disable_aspm(struct emx_softc *); 240 241 static void emx_print_debug_info(struct emx_softc *); 242 static void emx_print_nvm_info(struct emx_softc *); 243 static void emx_print_hw_stats(struct emx_softc *); 244 245 static int emx_sysctl_stats(SYSCTL_HANDLER_ARGS); 246 static int emx_sysctl_debug_info(SYSCTL_HANDLER_ARGS); 247 static int emx_sysctl_int_throttle(SYSCTL_HANDLER_ARGS); 248 static int emx_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS); 249 static int emx_sysctl_tx_wreg_nsegs(SYSCTL_HANDLER_ARGS); 250 #ifdef IFPOLL_ENABLE 251 static int emx_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS); 252 static int emx_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS); 253 #endif 254 static void emx_add_sysctl(struct emx_softc *); 255 256 static void emx_serialize_skipmain(struct emx_softc *); 257 static void emx_deserialize_skipmain(struct emx_softc *); 258 259 /* Management and WOL Support */ 260 static void emx_get_mgmt(struct emx_softc *); 261 static void emx_rel_mgmt(struct emx_softc *); 262 static void emx_get_hw_control(struct emx_softc *); 263 static void emx_rel_hw_control(struct emx_softc *); 264 static void emx_enable_wol(device_t); 265 266 static device_method_t emx_methods[] = { 267 /* Device interface */ 268 DEVMETHOD(device_probe, emx_probe), 269 DEVMETHOD(device_attach, emx_attach), 270 DEVMETHOD(device_detach, emx_detach), 271 DEVMETHOD(device_shutdown, emx_shutdown), 272 DEVMETHOD(device_suspend, emx_suspend), 273 DEVMETHOD(device_resume, emx_resume), 274 DEVMETHOD_END 275 }; 276 277 static driver_t emx_driver = { 278 "emx", 279 emx_methods, 280 sizeof(struct emx_softc), 281 }; 282 283 static devclass_t emx_devclass; 284 285 DECLARE_DUMMY_MODULE(if_emx); 286 MODULE_DEPEND(emx, ig_hal, 1, 1, 1); 287 DRIVER_MODULE(if_emx, pci, emx_driver, emx_devclass, NULL, NULL); 288 289 /* 290 * Tunables 291 */ 292 static int emx_int_throttle_ceil = EMX_DEFAULT_ITR; 293 static int emx_rxd = EMX_DEFAULT_RXD; 294 static int emx_txd = EMX_DEFAULT_TXD; 295 static int emx_smart_pwr_down = 0; 296 static int emx_rxr = 0; 297 static int emx_txr = 1; 298 299 /* Controls whether promiscuous also shows bad packets */ 300 static int emx_debug_sbp = 0; 301 302 static int emx_82573_workaround = 1; 303 static int emx_msi_enable = 1; 304 305 TUNABLE_INT("hw.emx.int_throttle_ceil", &emx_int_throttle_ceil); 306 TUNABLE_INT("hw.emx.rxd", &emx_rxd); 307 TUNABLE_INT("hw.emx.rxr", &emx_rxr); 308 TUNABLE_INT("hw.emx.txd", &emx_txd); 309 TUNABLE_INT("hw.emx.txr", &emx_txr); 310 TUNABLE_INT("hw.emx.smart_pwr_down", &emx_smart_pwr_down); 311 TUNABLE_INT("hw.emx.sbp", &emx_debug_sbp); 312 TUNABLE_INT("hw.emx.82573_workaround", &emx_82573_workaround); 313 TUNABLE_INT("hw.emx.msi.enable", &emx_msi_enable); 314 315 /* Global used in WOL setup with multiport cards */ 316 static int emx_global_quad_port_a = 0; 317 318 /* Set this to one to display debug statistics */ 319 static int emx_display_debug_stats = 0; 320 321 #if !defined(KTR_IF_EMX) 322 #define KTR_IF_EMX KTR_ALL 323 #endif 324 KTR_INFO_MASTER(if_emx); 325 KTR_INFO(KTR_IF_EMX, if_emx, intr_beg, 0, "intr begin"); 326 KTR_INFO(KTR_IF_EMX, if_emx, intr_end, 1, "intr end"); 327 KTR_INFO(KTR_IF_EMX, if_emx, pkt_receive, 4, "rx packet"); 328 KTR_INFO(KTR_IF_EMX, if_emx, pkt_txqueue, 5, "tx packet"); 329 KTR_INFO(KTR_IF_EMX, if_emx, pkt_txclean, 6, "tx clean"); 330 #define logif(name) KTR_LOG(if_emx_ ## name) 331 332 static __inline void 333 emx_setup_rxdesc(emx_rxdesc_t *rxd, const struct emx_rxbuf *rxbuf) 334 { 335 rxd->rxd_bufaddr = htole64(rxbuf->paddr); 336 /* DD bit must be cleared */ 337 rxd->rxd_staterr = 0; 338 } 339 340 static __inline void 341 emx_rxcsum(uint32_t staterr, struct mbuf *mp) 342 { 343 /* Ignore Checksum bit is set */ 344 if (staterr & E1000_RXD_STAT_IXSM) 345 return; 346 347 if ((staterr & (E1000_RXD_STAT_IPCS | E1000_RXDEXT_STATERR_IPE)) == 348 E1000_RXD_STAT_IPCS) 349 mp->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID; 350 351 if ((staterr & (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) == 352 E1000_RXD_STAT_TCPCS) { 353 mp->m_pkthdr.csum_flags |= CSUM_DATA_VALID | 354 CSUM_PSEUDO_HDR | 355 CSUM_FRAG_NOT_CHECKED; 356 mp->m_pkthdr.csum_data = htons(0xffff); 357 } 358 } 359 360 static __inline struct pktinfo * 361 emx_rssinfo(struct mbuf *m, struct pktinfo *pi, 362 uint32_t mrq, uint32_t hash, uint32_t staterr) 363 { 364 switch (mrq & EMX_RXDMRQ_RSSTYPE_MASK) { 365 case EMX_RXDMRQ_IPV4_TCP: 366 pi->pi_netisr = NETISR_IP; 367 pi->pi_flags = 0; 368 pi->pi_l3proto = IPPROTO_TCP; 369 break; 370 371 case EMX_RXDMRQ_IPV6_TCP: 372 pi->pi_netisr = NETISR_IPV6; 373 pi->pi_flags = 0; 374 pi->pi_l3proto = IPPROTO_TCP; 375 break; 376 377 case EMX_RXDMRQ_IPV4: 378 if (staterr & E1000_RXD_STAT_IXSM) 379 return NULL; 380 381 if ((staterr & 382 (E1000_RXD_STAT_TCPCS | E1000_RXDEXT_STATERR_TCPE)) == 383 E1000_RXD_STAT_TCPCS) { 384 pi->pi_netisr = NETISR_IP; 385 pi->pi_flags = 0; 386 pi->pi_l3proto = IPPROTO_UDP; 387 break; 388 } 389 /* FALL THROUGH */ 390 default: 391 return NULL; 392 } 393 394 m->m_flags |= M_HASH; 395 m->m_pkthdr.hash = toeplitz_hash(hash); 396 return pi; 397 } 398 399 static int 400 emx_probe(device_t dev) 401 { 402 const struct emx_device *d; 403 uint16_t vid, did; 404 405 vid = pci_get_vendor(dev); 406 did = pci_get_device(dev); 407 408 for (d = emx_devices; d->desc != NULL; ++d) { 409 if (vid == d->vid && did == d->did) { 410 device_set_desc(dev, d->desc); 411 device_set_async_attach(dev, TRUE); 412 return 0; 413 } 414 } 415 return ENXIO; 416 } 417 418 static int 419 emx_attach(device_t dev) 420 { 421 struct emx_softc *sc = device_get_softc(dev); 422 struct ifnet *ifp = &sc->arpcom.ac_if; 423 int error = 0, i, throttle, msi_enable, tx_ring_max; 424 u_int intr_flags; 425 uint16_t eeprom_data, device_id, apme_mask; 426 driver_intr_t *intr_func; 427 #ifdef IFPOLL_ENABLE 428 int offset, offset_def; 429 #endif 430 431 /* 432 * Setup RX rings 433 */ 434 for (i = 0; i < EMX_NRX_RING; ++i) { 435 sc->rx_data[i].sc = sc; 436 sc->rx_data[i].idx = i; 437 } 438 439 /* 440 * Setup TX ring 441 */ 442 for (i = 0; i < EMX_NTX_RING; ++i) { 443 sc->tx_data[i].sc = sc; 444 sc->tx_data[i].idx = i; 445 } 446 447 /* 448 * Initialize serializers 449 */ 450 lwkt_serialize_init(&sc->main_serialize); 451 for (i = 0; i < EMX_NTX_RING; ++i) 452 lwkt_serialize_init(&sc->tx_data[i].tx_serialize); 453 for (i = 0; i < EMX_NRX_RING; ++i) 454 lwkt_serialize_init(&sc->rx_data[i].rx_serialize); 455 456 /* 457 * Initialize serializer array 458 */ 459 i = 0; 460 sc->serializes[i++] = &sc->main_serialize; 461 462 KKASSERT(i == EMX_TX_SERIALIZE); 463 sc->serializes[i++] = &sc->tx_data[0].tx_serialize; 464 sc->serializes[i++] = &sc->tx_data[1].tx_serialize; 465 466 KKASSERT(i == EMX_RX_SERIALIZE); 467 sc->serializes[i++] = &sc->rx_data[0].rx_serialize; 468 sc->serializes[i++] = &sc->rx_data[1].rx_serialize; 469 KKASSERT(i == EMX_NSERIALIZE); 470 471 callout_init_mp(&sc->timer); 472 473 sc->dev = sc->osdep.dev = dev; 474 475 /* 476 * Determine hardware and mac type 477 */ 478 sc->hw.vendor_id = pci_get_vendor(dev); 479 sc->hw.device_id = pci_get_device(dev); 480 sc->hw.revision_id = pci_get_revid(dev); 481 sc->hw.subsystem_vendor_id = pci_get_subvendor(dev); 482 sc->hw.subsystem_device_id = pci_get_subdevice(dev); 483 484 if (e1000_set_mac_type(&sc->hw)) 485 return ENXIO; 486 487 /* Enable bus mastering */ 488 pci_enable_busmaster(dev); 489 490 /* 491 * Allocate IO memory 492 */ 493 sc->memory_rid = EMX_BAR_MEM; 494 sc->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY, 495 &sc->memory_rid, RF_ACTIVE); 496 if (sc->memory == NULL) { 497 device_printf(dev, "Unable to allocate bus resource: memory\n"); 498 error = ENXIO; 499 goto fail; 500 } 501 sc->osdep.mem_bus_space_tag = rman_get_bustag(sc->memory); 502 sc->osdep.mem_bus_space_handle = rman_get_bushandle(sc->memory); 503 504 /* XXX This is quite goofy, it is not actually used */ 505 sc->hw.hw_addr = (uint8_t *)&sc->osdep.mem_bus_space_handle; 506 507 /* 508 * Don't enable MSI-X on 82574, see: 509 * 82574 specification update errata #15 510 * 511 * Don't enable MSI on 82571/82572, see: 512 * 82571/82572 specification update errata #63 513 */ 514 msi_enable = emx_msi_enable; 515 if (msi_enable && 516 (sc->hw.mac.type == e1000_82571 || 517 sc->hw.mac.type == e1000_82572)) 518 msi_enable = 0; 519 520 /* 521 * Allocate interrupt 522 */ 523 sc->intr_type = pci_alloc_1intr(dev, msi_enable, 524 &sc->intr_rid, &intr_flags); 525 526 if (sc->intr_type == PCI_INTR_TYPE_LEGACY) { 527 int unshared; 528 529 unshared = device_getenv_int(dev, "irq.unshared", 0); 530 if (!unshared) { 531 sc->flags |= EMX_FLAG_SHARED_INTR; 532 if (bootverbose) 533 device_printf(dev, "IRQ shared\n"); 534 } else { 535 intr_flags &= ~RF_SHAREABLE; 536 if (bootverbose) 537 device_printf(dev, "IRQ unshared\n"); 538 } 539 } 540 541 sc->intr_res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &sc->intr_rid, 542 intr_flags); 543 if (sc->intr_res == NULL) { 544 device_printf(dev, "Unable to allocate bus resource: " 545 "interrupt\n"); 546 error = ENXIO; 547 goto fail; 548 } 549 550 /* Save PCI command register for Shared Code */ 551 sc->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2); 552 sc->hw.back = &sc->osdep; 553 554 /* Do Shared Code initialization */ 555 if (e1000_setup_init_funcs(&sc->hw, TRUE)) { 556 device_printf(dev, "Setup of Shared code failed\n"); 557 error = ENXIO; 558 goto fail; 559 } 560 e1000_get_bus_info(&sc->hw); 561 562 sc->hw.mac.autoneg = EMX_DO_AUTO_NEG; 563 sc->hw.phy.autoneg_wait_to_complete = FALSE; 564 sc->hw.phy.autoneg_advertised = EMX_AUTONEG_ADV_DEFAULT; 565 566 /* 567 * Interrupt throttle rate 568 */ 569 throttle = device_getenv_int(dev, "int_throttle_ceil", 570 emx_int_throttle_ceil); 571 if (throttle == 0) { 572 sc->int_throttle_ceil = 0; 573 } else { 574 if (throttle < 0) 575 throttle = EMX_DEFAULT_ITR; 576 577 /* Recalculate the tunable value to get the exact frequency. */ 578 throttle = 1000000000 / 256 / throttle; 579 580 /* Upper 16bits of ITR is reserved and should be zero */ 581 if (throttle & 0xffff0000) 582 throttle = 1000000000 / 256 / EMX_DEFAULT_ITR; 583 584 sc->int_throttle_ceil = 1000000000 / 256 / throttle; 585 } 586 587 e1000_init_script_state_82541(&sc->hw, TRUE); 588 e1000_set_tbi_compatibility_82543(&sc->hw, TRUE); 589 590 /* Copper options */ 591 if (sc->hw.phy.media_type == e1000_media_type_copper) { 592 sc->hw.phy.mdix = EMX_AUTO_ALL_MODES; 593 sc->hw.phy.disable_polarity_correction = FALSE; 594 sc->hw.phy.ms_type = EMX_MASTER_SLAVE; 595 } 596 597 /* Set the frame limits assuming standard ethernet sized frames. */ 598 sc->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN; 599 sc->min_frame_size = ETHER_MIN_LEN; 600 601 /* This controls when hardware reports transmit completion status. */ 602 sc->hw.mac.report_tx_early = 1; 603 604 /* Calculate # of RX rings */ 605 sc->rx_ring_cnt = device_getenv_int(dev, "rxr", emx_rxr); 606 sc->rx_ring_cnt = if_ring_count2(sc->rx_ring_cnt, EMX_NRX_RING); 607 608 /* 609 * Calculate # of TX rings 610 * 611 * NOTE: 612 * Don't enable multiple TX queues on 82574; it always gives 613 * watchdog timeout when multiple TCP streams are received. 614 */ 615 tx_ring_max = 1; 616 if (sc->hw.mac.type == e1000_82571 || 617 sc->hw.mac.type == e1000_82572 || 618 sc->hw.mac.type == e1000_80003es2lan) 619 tx_ring_max = EMX_NTX_RING; 620 sc->tx_ring_cnt = device_getenv_int(dev, "txr", emx_txr); 621 sc->tx_ring_cnt = if_ring_count2(sc->tx_ring_cnt, tx_ring_max); 622 623 /* Allocate RX/TX rings' busdma(9) stuffs */ 624 error = emx_dma_alloc(sc); 625 if (error) 626 goto fail; 627 628 /* Allocate multicast array memory. */ 629 sc->mta = kmalloc(ETH_ADDR_LEN * EMX_MCAST_ADDR_MAX, 630 M_DEVBUF, M_WAITOK); 631 632 /* Indicate SOL/IDER usage */ 633 if (e1000_check_reset_block(&sc->hw)) { 634 device_printf(dev, 635 "PHY reset is blocked due to SOL/IDER session.\n"); 636 } 637 638 /* 639 * Start from a known state, this is important in reading the 640 * nvm and mac from that. 641 */ 642 e1000_reset_hw(&sc->hw); 643 644 /* Make sure we have a good EEPROM before we read from it */ 645 if (e1000_validate_nvm_checksum(&sc->hw) < 0) { 646 /* 647 * Some PCI-E parts fail the first check due to 648 * the link being in sleep state, call it again, 649 * if it fails a second time its a real issue. 650 */ 651 if (e1000_validate_nvm_checksum(&sc->hw) < 0) { 652 device_printf(dev, 653 "The EEPROM Checksum Is Not Valid\n"); 654 error = EIO; 655 goto fail; 656 } 657 } 658 659 /* Copy the permanent MAC address out of the EEPROM */ 660 if (e1000_read_mac_addr(&sc->hw) < 0) { 661 device_printf(dev, "EEPROM read error while reading MAC" 662 " address\n"); 663 error = EIO; 664 goto fail; 665 } 666 if (!emx_is_valid_eaddr(sc->hw.mac.addr)) { 667 device_printf(dev, "Invalid MAC address\n"); 668 error = EIO; 669 goto fail; 670 } 671 672 /* Determine if we have to control management hardware */ 673 if (e1000_enable_mng_pass_thru(&sc->hw)) 674 sc->flags |= EMX_FLAG_HAS_MGMT; 675 676 /* 677 * Setup Wake-on-Lan 678 */ 679 apme_mask = EMX_EEPROM_APME; 680 eeprom_data = 0; 681 switch (sc->hw.mac.type) { 682 case e1000_82573: 683 sc->flags |= EMX_FLAG_HAS_AMT; 684 /* FALL THROUGH */ 685 686 case e1000_82571: 687 case e1000_82572: 688 case e1000_80003es2lan: 689 if (sc->hw.bus.func == 1) { 690 e1000_read_nvm(&sc->hw, 691 NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data); 692 } else { 693 e1000_read_nvm(&sc->hw, 694 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data); 695 } 696 break; 697 698 default: 699 e1000_read_nvm(&sc->hw, 700 NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data); 701 break; 702 } 703 if (eeprom_data & apme_mask) 704 sc->wol = E1000_WUFC_MAG | E1000_WUFC_MC; 705 706 /* 707 * We have the eeprom settings, now apply the special cases 708 * where the eeprom may be wrong or the board won't support 709 * wake on lan on a particular port 710 */ 711 device_id = pci_get_device(dev); 712 switch (device_id) { 713 case E1000_DEV_ID_82571EB_FIBER: 714 /* 715 * Wake events only supported on port A for dual fiber 716 * regardless of eeprom setting 717 */ 718 if (E1000_READ_REG(&sc->hw, E1000_STATUS) & 719 E1000_STATUS_FUNC_1) 720 sc->wol = 0; 721 break; 722 723 case E1000_DEV_ID_82571EB_QUAD_COPPER: 724 case E1000_DEV_ID_82571EB_QUAD_FIBER: 725 case E1000_DEV_ID_82571EB_QUAD_COPPER_LP: 726 /* if quad port sc, disable WoL on all but port A */ 727 if (emx_global_quad_port_a != 0) 728 sc->wol = 0; 729 /* Reset for multiple quad port adapters */ 730 if (++emx_global_quad_port_a == 4) 731 emx_global_quad_port_a = 0; 732 break; 733 } 734 735 /* XXX disable wol */ 736 sc->wol = 0; 737 738 #ifdef IFPOLL_ENABLE 739 /* 740 * NPOLLING RX CPU offset 741 */ 742 if (sc->rx_ring_cnt == ncpus2) { 743 offset = 0; 744 } else { 745 offset_def = (sc->rx_ring_cnt * device_get_unit(dev)) % ncpus2; 746 offset = device_getenv_int(dev, "npoll.rxoff", offset_def); 747 if (offset >= ncpus2 || 748 offset % sc->rx_ring_cnt != 0) { 749 device_printf(dev, "invalid npoll.rxoff %d, use %d\n", 750 offset, offset_def); 751 offset = offset_def; 752 } 753 } 754 sc->rx_npoll_off = offset; 755 756 /* 757 * NPOLLING TX CPU offset 758 */ 759 if (sc->tx_ring_cnt == ncpus2) { 760 offset = 0; 761 } else { 762 offset_def = (sc->tx_ring_cnt * device_get_unit(dev)) % ncpus2; 763 offset = device_getenv_int(dev, "npoll.txoff", offset_def); 764 if (offset >= ncpus2 || 765 offset % sc->tx_ring_cnt != 0) { 766 device_printf(dev, "invalid npoll.txoff %d, use %d\n", 767 offset, offset_def); 768 offset = offset_def; 769 } 770 } 771 sc->tx_npoll_off = offset; 772 #endif 773 774 /* Setup OS specific network interface */ 775 emx_setup_ifp(sc); 776 777 /* Add sysctl tree, must after em_setup_ifp() */ 778 emx_add_sysctl(sc); 779 780 /* Reset the hardware */ 781 error = emx_reset(sc); 782 if (error) { 783 device_printf(dev, "Unable to reset the hardware\n"); 784 goto fail; 785 } 786 787 /* Initialize statistics */ 788 emx_update_stats(sc); 789 790 sc->hw.mac.get_link_status = 1; 791 emx_update_link_status(sc); 792 793 /* Non-AMT based hardware can now take control from firmware */ 794 if ((sc->flags & (EMX_FLAG_HAS_MGMT | EMX_FLAG_HAS_AMT)) == 795 EMX_FLAG_HAS_MGMT) 796 emx_get_hw_control(sc); 797 798 /* 799 * Missing Interrupt Following ICR read: 800 * 801 * 82571/82572 specification update errata #76 802 * 82573 specification update errata #31 803 * 82574 specification update errata #12 804 */ 805 intr_func = emx_intr; 806 if ((sc->flags & EMX_FLAG_SHARED_INTR) && 807 (sc->hw.mac.type == e1000_82571 || 808 sc->hw.mac.type == e1000_82572 || 809 sc->hw.mac.type == e1000_82573 || 810 sc->hw.mac.type == e1000_82574)) 811 intr_func = emx_intr_mask; 812 813 error = bus_setup_intr(dev, sc->intr_res, INTR_MPSAFE, intr_func, sc, 814 &sc->intr_tag, &sc->main_serialize); 815 if (error) { 816 device_printf(dev, "Failed to register interrupt handler"); 817 ether_ifdetach(&sc->arpcom.ac_if); 818 goto fail; 819 } 820 821 sc->tx_ring_inuse = emx_get_txring_inuse(sc, FALSE); 822 for (i = 0; i < sc->tx_ring_cnt; ++i) { 823 struct ifaltq_subque *ifsq = ifq_get_subq(&ifp->if_snd, i); 824 struct emx_txdata *tdata = &sc->tx_data[i]; 825 826 ifsq_set_cpuid(ifsq, rman_get_cpuid(sc->intr_res)); 827 ifsq_set_priv(ifsq, tdata); 828 tdata->ifsq = ifsq; 829 830 ifsq_watchdog_init(&tdata->tx_watchdog, ifsq, emx_watchdog); 831 } 832 return (0); 833 fail: 834 emx_detach(dev); 835 return (error); 836 } 837 838 static int 839 emx_detach(device_t dev) 840 { 841 struct emx_softc *sc = device_get_softc(dev); 842 843 if (device_is_attached(dev)) { 844 struct ifnet *ifp = &sc->arpcom.ac_if; 845 846 ifnet_serialize_all(ifp); 847 848 emx_stop(sc); 849 850 e1000_phy_hw_reset(&sc->hw); 851 852 emx_rel_mgmt(sc); 853 emx_rel_hw_control(sc); 854 855 if (sc->wol) { 856 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN); 857 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol); 858 emx_enable_wol(dev); 859 } 860 861 bus_teardown_intr(dev, sc->intr_res, sc->intr_tag); 862 863 ifnet_deserialize_all(ifp); 864 865 ether_ifdetach(ifp); 866 } else if (sc->memory != NULL) { 867 emx_rel_hw_control(sc); 868 } 869 bus_generic_detach(dev); 870 871 if (sc->intr_res != NULL) { 872 bus_release_resource(dev, SYS_RES_IRQ, sc->intr_rid, 873 sc->intr_res); 874 } 875 876 if (sc->intr_type == PCI_INTR_TYPE_MSI) 877 pci_release_msi(dev); 878 879 if (sc->memory != NULL) { 880 bus_release_resource(dev, SYS_RES_MEMORY, sc->memory_rid, 881 sc->memory); 882 } 883 884 emx_dma_free(sc); 885 886 /* Free sysctl tree */ 887 if (sc->sysctl_tree != NULL) 888 sysctl_ctx_free(&sc->sysctl_ctx); 889 890 if (sc->mta != NULL) 891 kfree(sc->mta, M_DEVBUF); 892 893 return (0); 894 } 895 896 static int 897 emx_shutdown(device_t dev) 898 { 899 return emx_suspend(dev); 900 } 901 902 static int 903 emx_suspend(device_t dev) 904 { 905 struct emx_softc *sc = device_get_softc(dev); 906 struct ifnet *ifp = &sc->arpcom.ac_if; 907 908 ifnet_serialize_all(ifp); 909 910 emx_stop(sc); 911 912 emx_rel_mgmt(sc); 913 emx_rel_hw_control(sc); 914 915 if (sc->wol) { 916 E1000_WRITE_REG(&sc->hw, E1000_WUC, E1000_WUC_PME_EN); 917 E1000_WRITE_REG(&sc->hw, E1000_WUFC, sc->wol); 918 emx_enable_wol(dev); 919 } 920 921 ifnet_deserialize_all(ifp); 922 923 return bus_generic_suspend(dev); 924 } 925 926 static int 927 emx_resume(device_t dev) 928 { 929 struct emx_softc *sc = device_get_softc(dev); 930 struct ifnet *ifp = &sc->arpcom.ac_if; 931 int i; 932 933 ifnet_serialize_all(ifp); 934 935 emx_init(sc); 936 emx_get_mgmt(sc); 937 for (i = 0; i < sc->tx_ring_inuse; ++i) 938 ifsq_devstart_sched(sc->tx_data[i].ifsq); 939 940 ifnet_deserialize_all(ifp); 941 942 return bus_generic_resume(dev); 943 } 944 945 static void 946 emx_start(struct ifnet *ifp, struct ifaltq_subque *ifsq) 947 { 948 struct emx_softc *sc = ifp->if_softc; 949 struct emx_txdata *tdata = ifsq_get_priv(ifsq); 950 struct mbuf *m_head; 951 int idx = -1, nsegs = 0; 952 953 KKASSERT(tdata->ifsq == ifsq); 954 ASSERT_SERIALIZED(&tdata->tx_serialize); 955 956 if ((ifp->if_flags & IFF_RUNNING) == 0 || ifsq_is_oactive(ifsq)) 957 return; 958 959 if (!sc->link_active || (tdata->tx_flags & EMX_TXFLAG_ENABLED) == 0) { 960 ifsq_purge(ifsq); 961 return; 962 } 963 964 while (!ifsq_is_empty(ifsq)) { 965 /* Now do we at least have a minimal? */ 966 if (EMX_IS_OACTIVE(tdata)) { 967 emx_tx_collect(tdata); 968 if (EMX_IS_OACTIVE(tdata)) { 969 ifsq_set_oactive(ifsq); 970 break; 971 } 972 } 973 974 logif(pkt_txqueue); 975 m_head = ifsq_dequeue(ifsq, NULL); 976 if (m_head == NULL) 977 break; 978 979 if (emx_encap(tdata, &m_head, &nsegs, &idx)) { 980 IFNET_STAT_INC(ifp, oerrors, 1); 981 emx_tx_collect(tdata); 982 continue; 983 } 984 985 if (nsegs >= tdata->tx_wreg_nsegs) { 986 E1000_WRITE_REG(&sc->hw, E1000_TDT(tdata->idx), idx); 987 nsegs = 0; 988 idx = -1; 989 } 990 991 /* Send a copy of the frame to the BPF listener */ 992 ETHER_BPF_MTAP(ifp, m_head); 993 994 /* Set timeout in case hardware has problems transmitting. */ 995 tdata->tx_watchdog.wd_timer = EMX_TX_TIMEOUT; 996 } 997 if (idx >= 0) 998 E1000_WRITE_REG(&sc->hw, E1000_TDT(tdata->idx), idx); 999 } 1000 1001 static int 1002 emx_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr) 1003 { 1004 struct emx_softc *sc = ifp->if_softc; 1005 struct ifreq *ifr = (struct ifreq *)data; 1006 uint16_t eeprom_data = 0; 1007 int max_frame_size, mask, reinit; 1008 int error = 0; 1009 1010 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1011 1012 switch (command) { 1013 case SIOCSIFMTU: 1014 switch (sc->hw.mac.type) { 1015 case e1000_82573: 1016 /* 1017 * 82573 only supports jumbo frames 1018 * if ASPM is disabled. 1019 */ 1020 e1000_read_nvm(&sc->hw, NVM_INIT_3GIO_3, 1, 1021 &eeprom_data); 1022 if (eeprom_data & NVM_WORD1A_ASPM_MASK) { 1023 max_frame_size = ETHER_MAX_LEN; 1024 break; 1025 } 1026 /* FALL THROUGH */ 1027 1028 /* Limit Jumbo Frame size */ 1029 case e1000_82571: 1030 case e1000_82572: 1031 case e1000_82574: 1032 case e1000_80003es2lan: 1033 max_frame_size = 9234; 1034 break; 1035 1036 default: 1037 max_frame_size = MAX_JUMBO_FRAME_SIZE; 1038 break; 1039 } 1040 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN - 1041 ETHER_CRC_LEN) { 1042 error = EINVAL; 1043 break; 1044 } 1045 1046 ifp->if_mtu = ifr->ifr_mtu; 1047 sc->max_frame_size = ifp->if_mtu + ETHER_HDR_LEN + 1048 ETHER_CRC_LEN; 1049 1050 if (ifp->if_flags & IFF_RUNNING) 1051 emx_init(sc); 1052 break; 1053 1054 case SIOCSIFFLAGS: 1055 if (ifp->if_flags & IFF_UP) { 1056 if ((ifp->if_flags & IFF_RUNNING)) { 1057 if ((ifp->if_flags ^ sc->if_flags) & 1058 (IFF_PROMISC | IFF_ALLMULTI)) { 1059 emx_disable_promisc(sc); 1060 emx_set_promisc(sc); 1061 } 1062 } else { 1063 emx_init(sc); 1064 } 1065 } else if (ifp->if_flags & IFF_RUNNING) { 1066 emx_stop(sc); 1067 } 1068 sc->if_flags = ifp->if_flags; 1069 break; 1070 1071 case SIOCADDMULTI: 1072 case SIOCDELMULTI: 1073 if (ifp->if_flags & IFF_RUNNING) { 1074 emx_disable_intr(sc); 1075 emx_set_multi(sc); 1076 #ifdef IFPOLL_ENABLE 1077 if (!(ifp->if_flags & IFF_NPOLLING)) 1078 #endif 1079 emx_enable_intr(sc); 1080 } 1081 break; 1082 1083 case SIOCSIFMEDIA: 1084 /* Check SOL/IDER usage */ 1085 if (e1000_check_reset_block(&sc->hw)) { 1086 device_printf(sc->dev, "Media change is" 1087 " blocked due to SOL/IDER session.\n"); 1088 break; 1089 } 1090 /* FALL THROUGH */ 1091 1092 case SIOCGIFMEDIA: 1093 error = ifmedia_ioctl(ifp, ifr, &sc->media, command); 1094 break; 1095 1096 case SIOCSIFCAP: 1097 reinit = 0; 1098 mask = ifr->ifr_reqcap ^ ifp->if_capenable; 1099 if (mask & IFCAP_RXCSUM) { 1100 ifp->if_capenable ^= IFCAP_RXCSUM; 1101 reinit = 1; 1102 } 1103 if (mask & IFCAP_VLAN_HWTAGGING) { 1104 ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; 1105 reinit = 1; 1106 } 1107 if (mask & IFCAP_TXCSUM) { 1108 ifp->if_capenable ^= IFCAP_TXCSUM; 1109 if (ifp->if_capenable & IFCAP_TXCSUM) 1110 ifp->if_hwassist |= EMX_CSUM_FEATURES; 1111 else 1112 ifp->if_hwassist &= ~EMX_CSUM_FEATURES; 1113 } 1114 if (mask & IFCAP_TSO) { 1115 ifp->if_capenable ^= IFCAP_TSO; 1116 if (ifp->if_capenable & IFCAP_TSO) 1117 ifp->if_hwassist |= CSUM_TSO; 1118 else 1119 ifp->if_hwassist &= ~CSUM_TSO; 1120 } 1121 if (mask & IFCAP_RSS) 1122 ifp->if_capenable ^= IFCAP_RSS; 1123 if (reinit && (ifp->if_flags & IFF_RUNNING)) 1124 emx_init(sc); 1125 break; 1126 1127 default: 1128 error = ether_ioctl(ifp, command, data); 1129 break; 1130 } 1131 return (error); 1132 } 1133 1134 static void 1135 emx_watchdog(struct ifaltq_subque *ifsq) 1136 { 1137 struct emx_txdata *tdata = ifsq_get_priv(ifsq); 1138 struct ifnet *ifp = ifsq_get_ifp(ifsq); 1139 struct emx_softc *sc = ifp->if_softc; 1140 int i; 1141 1142 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1143 1144 /* 1145 * The timer is set to 5 every time start queues a packet. 1146 * Then txeof keeps resetting it as long as it cleans at 1147 * least one descriptor. 1148 * Finally, anytime all descriptors are clean the timer is 1149 * set to 0. 1150 */ 1151 1152 if (E1000_READ_REG(&sc->hw, E1000_TDT(tdata->idx)) == 1153 E1000_READ_REG(&sc->hw, E1000_TDH(tdata->idx))) { 1154 /* 1155 * If we reach here, all TX jobs are completed and 1156 * the TX engine should have been idled for some time. 1157 * We don't need to call ifsq_devstart_sched() here. 1158 */ 1159 ifsq_clr_oactive(ifsq); 1160 tdata->tx_watchdog.wd_timer = 0; 1161 return; 1162 } 1163 1164 /* 1165 * If we are in this routine because of pause frames, then 1166 * don't reset the hardware. 1167 */ 1168 if (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_TXOFF) { 1169 tdata->tx_watchdog.wd_timer = EMX_TX_TIMEOUT; 1170 return; 1171 } 1172 1173 if_printf(ifp, "TX %d watchdog timeout -- resetting\n", tdata->idx); 1174 1175 IFNET_STAT_INC(ifp, oerrors, 1); 1176 1177 emx_init(sc); 1178 for (i = 0; i < sc->tx_ring_inuse; ++i) 1179 ifsq_devstart_sched(sc->tx_data[i].ifsq); 1180 } 1181 1182 static void 1183 emx_init(void *xsc) 1184 { 1185 struct emx_softc *sc = xsc; 1186 struct ifnet *ifp = &sc->arpcom.ac_if; 1187 device_t dev = sc->dev; 1188 boolean_t polling; 1189 int i; 1190 1191 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1192 1193 emx_stop(sc); 1194 1195 /* Get the latest mac address, User can use a LAA */ 1196 bcopy(IF_LLADDR(ifp), sc->hw.mac.addr, ETHER_ADDR_LEN); 1197 1198 /* Put the address into the Receive Address Array */ 1199 e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0); 1200 1201 /* 1202 * With the 82571 sc, RAR[0] may be overwritten 1203 * when the other port is reset, we make a duplicate 1204 * in RAR[14] for that eventuality, this assures 1205 * the interface continues to function. 1206 */ 1207 if (sc->hw.mac.type == e1000_82571) { 1208 e1000_set_laa_state_82571(&sc->hw, TRUE); 1209 e1000_rar_set(&sc->hw, sc->hw.mac.addr, 1210 E1000_RAR_ENTRIES - 1); 1211 } 1212 1213 /* Initialize the hardware */ 1214 if (emx_reset(sc)) { 1215 device_printf(dev, "Unable to reset the hardware\n"); 1216 /* XXX emx_stop()? */ 1217 return; 1218 } 1219 emx_update_link_status(sc); 1220 1221 /* Setup VLAN support, basic and offload if available */ 1222 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN); 1223 1224 if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) { 1225 uint32_t ctrl; 1226 1227 ctrl = E1000_READ_REG(&sc->hw, E1000_CTRL); 1228 ctrl |= E1000_CTRL_VME; 1229 E1000_WRITE_REG(&sc->hw, E1000_CTRL, ctrl); 1230 } 1231 1232 /* Configure for OS presence */ 1233 emx_get_mgmt(sc); 1234 1235 polling = FALSE; 1236 #ifdef IFPOLL_ENABLE 1237 if (ifp->if_flags & IFF_NPOLLING) 1238 polling = TRUE; 1239 #endif 1240 sc->tx_ring_inuse = emx_get_txring_inuse(sc, polling); 1241 ifq_set_subq_mask(&ifp->if_snd, sc->tx_ring_inuse - 1); 1242 1243 /* Prepare transmit descriptors and buffers */ 1244 for (i = 0; i < sc->tx_ring_inuse; ++i) 1245 emx_init_tx_ring(&sc->tx_data[i]); 1246 emx_init_tx_unit(sc); 1247 1248 /* Setup Multicast table */ 1249 emx_set_multi(sc); 1250 1251 /* Prepare receive descriptors and buffers */ 1252 for (i = 0; i < sc->rx_ring_cnt; ++i) { 1253 if (emx_init_rx_ring(&sc->rx_data[i])) { 1254 device_printf(dev, 1255 "Could not setup receive structures\n"); 1256 emx_stop(sc); 1257 return; 1258 } 1259 } 1260 emx_init_rx_unit(sc); 1261 1262 /* Don't lose promiscuous settings */ 1263 emx_set_promisc(sc); 1264 1265 ifp->if_flags |= IFF_RUNNING; 1266 for (i = 0; i < sc->tx_ring_inuse; ++i) { 1267 ifsq_clr_oactive(sc->tx_data[i].ifsq); 1268 ifsq_watchdog_start(&sc->tx_data[i].tx_watchdog); 1269 } 1270 1271 callout_reset(&sc->timer, hz, emx_timer, sc); 1272 e1000_clear_hw_cntrs_base_generic(&sc->hw); 1273 1274 /* MSI/X configuration for 82574 */ 1275 if (sc->hw.mac.type == e1000_82574) { 1276 int tmp; 1277 1278 tmp = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT); 1279 tmp |= E1000_CTRL_EXT_PBA_CLR; 1280 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, tmp); 1281 /* 1282 * XXX MSIX 1283 * Set the IVAR - interrupt vector routing. 1284 * Each nibble represents a vector, high bit 1285 * is enable, other 3 bits are the MSIX table 1286 * entry, we map RXQ0 to 0, TXQ0 to 1, and 1287 * Link (other) to 2, hence the magic number. 1288 */ 1289 E1000_WRITE_REG(&sc->hw, E1000_IVAR, 0x800A0908); 1290 } 1291 1292 /* 1293 * Only enable interrupts if we are not polling, make sure 1294 * they are off otherwise. 1295 */ 1296 if (polling) 1297 emx_disable_intr(sc); 1298 else 1299 emx_enable_intr(sc); 1300 1301 /* AMT based hardware can now take control from firmware */ 1302 if ((sc->flags & (EMX_FLAG_HAS_MGMT | EMX_FLAG_HAS_AMT)) == 1303 (EMX_FLAG_HAS_MGMT | EMX_FLAG_HAS_AMT)) 1304 emx_get_hw_control(sc); 1305 } 1306 1307 static void 1308 emx_intr(void *xsc) 1309 { 1310 emx_intr_body(xsc, TRUE); 1311 } 1312 1313 static void 1314 emx_intr_body(struct emx_softc *sc, boolean_t chk_asserted) 1315 { 1316 struct ifnet *ifp = &sc->arpcom.ac_if; 1317 uint32_t reg_icr; 1318 1319 logif(intr_beg); 1320 ASSERT_SERIALIZED(&sc->main_serialize); 1321 1322 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR); 1323 1324 if (chk_asserted && (reg_icr & E1000_ICR_INT_ASSERTED) == 0) { 1325 logif(intr_end); 1326 return; 1327 } 1328 1329 /* 1330 * XXX: some laptops trigger several spurious interrupts 1331 * on emx(4) when in the resume cycle. The ICR register 1332 * reports all-ones value in this case. Processing such 1333 * interrupts would lead to a freeze. I don't know why. 1334 */ 1335 if (reg_icr == 0xffffffff) { 1336 logif(intr_end); 1337 return; 1338 } 1339 1340 if (ifp->if_flags & IFF_RUNNING) { 1341 if (reg_icr & 1342 (E1000_ICR_RXT0 | E1000_ICR_RXDMT0 | E1000_ICR_RXO)) { 1343 int i; 1344 1345 for (i = 0; i < sc->rx_ring_cnt; ++i) { 1346 lwkt_serialize_enter( 1347 &sc->rx_data[i].rx_serialize); 1348 emx_rxeof(&sc->rx_data[i], -1); 1349 lwkt_serialize_exit( 1350 &sc->rx_data[i].rx_serialize); 1351 } 1352 } 1353 if (reg_icr & E1000_ICR_TXDW) { 1354 struct emx_txdata *tdata = &sc->tx_data[0]; 1355 1356 lwkt_serialize_enter(&tdata->tx_serialize); 1357 emx_txeof(tdata); 1358 if (!ifsq_is_empty(tdata->ifsq)) 1359 ifsq_devstart(tdata->ifsq); 1360 lwkt_serialize_exit(&tdata->tx_serialize); 1361 } 1362 } 1363 1364 /* Link status change */ 1365 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) { 1366 emx_serialize_skipmain(sc); 1367 1368 callout_stop(&sc->timer); 1369 sc->hw.mac.get_link_status = 1; 1370 emx_update_link_status(sc); 1371 1372 /* Deal with TX cruft when link lost */ 1373 emx_tx_purge(sc); 1374 1375 callout_reset(&sc->timer, hz, emx_timer, sc); 1376 1377 emx_deserialize_skipmain(sc); 1378 } 1379 1380 if (reg_icr & E1000_ICR_RXO) 1381 sc->rx_overruns++; 1382 1383 logif(intr_end); 1384 } 1385 1386 static void 1387 emx_intr_mask(void *xsc) 1388 { 1389 struct emx_softc *sc = xsc; 1390 1391 E1000_WRITE_REG(&sc->hw, E1000_IMC, 0xffffffff); 1392 /* 1393 * NOTE: 1394 * ICR.INT_ASSERTED bit will never be set if IMS is 0, 1395 * so don't check it. 1396 */ 1397 emx_intr_body(sc, FALSE); 1398 E1000_WRITE_REG(&sc->hw, E1000_IMS, IMS_ENABLE_MASK); 1399 } 1400 1401 static void 1402 emx_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) 1403 { 1404 struct emx_softc *sc = ifp->if_softc; 1405 1406 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1407 1408 emx_update_link_status(sc); 1409 1410 ifmr->ifm_status = IFM_AVALID; 1411 ifmr->ifm_active = IFM_ETHER; 1412 1413 if (!sc->link_active) 1414 return; 1415 1416 ifmr->ifm_status |= IFM_ACTIVE; 1417 1418 if (sc->hw.phy.media_type == e1000_media_type_fiber || 1419 sc->hw.phy.media_type == e1000_media_type_internal_serdes) { 1420 ifmr->ifm_active |= IFM_1000_SX | IFM_FDX; 1421 } else { 1422 switch (sc->link_speed) { 1423 case 10: 1424 ifmr->ifm_active |= IFM_10_T; 1425 break; 1426 case 100: 1427 ifmr->ifm_active |= IFM_100_TX; 1428 break; 1429 1430 case 1000: 1431 ifmr->ifm_active |= IFM_1000_T; 1432 break; 1433 } 1434 if (sc->link_duplex == FULL_DUPLEX) 1435 ifmr->ifm_active |= IFM_FDX; 1436 else 1437 ifmr->ifm_active |= IFM_HDX; 1438 } 1439 } 1440 1441 static int 1442 emx_media_change(struct ifnet *ifp) 1443 { 1444 struct emx_softc *sc = ifp->if_softc; 1445 struct ifmedia *ifm = &sc->media; 1446 1447 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1448 1449 if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) 1450 return (EINVAL); 1451 1452 switch (IFM_SUBTYPE(ifm->ifm_media)) { 1453 case IFM_AUTO: 1454 sc->hw.mac.autoneg = EMX_DO_AUTO_NEG; 1455 sc->hw.phy.autoneg_advertised = EMX_AUTONEG_ADV_DEFAULT; 1456 break; 1457 1458 case IFM_1000_LX: 1459 case IFM_1000_SX: 1460 case IFM_1000_T: 1461 sc->hw.mac.autoneg = EMX_DO_AUTO_NEG; 1462 sc->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL; 1463 break; 1464 1465 case IFM_100_TX: 1466 sc->hw.mac.autoneg = FALSE; 1467 sc->hw.phy.autoneg_advertised = 0; 1468 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) 1469 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL; 1470 else 1471 sc->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF; 1472 break; 1473 1474 case IFM_10_T: 1475 sc->hw.mac.autoneg = FALSE; 1476 sc->hw.phy.autoneg_advertised = 0; 1477 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX) 1478 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL; 1479 else 1480 sc->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF; 1481 break; 1482 1483 default: 1484 if_printf(ifp, "Unsupported media type\n"); 1485 break; 1486 } 1487 1488 emx_init(sc); 1489 1490 return (0); 1491 } 1492 1493 static int 1494 emx_encap(struct emx_txdata *tdata, struct mbuf **m_headp, 1495 int *segs_used, int *idx) 1496 { 1497 bus_dma_segment_t segs[EMX_MAX_SCATTER]; 1498 bus_dmamap_t map; 1499 struct emx_txbuf *tx_buffer, *tx_buffer_mapped; 1500 struct e1000_tx_desc *ctxd = NULL; 1501 struct mbuf *m_head = *m_headp; 1502 uint32_t txd_upper, txd_lower, cmd = 0; 1503 int maxsegs, nsegs, i, j, first, last = 0, error; 1504 1505 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 1506 error = emx_tso_pullup(tdata, m_headp); 1507 if (error) 1508 return error; 1509 m_head = *m_headp; 1510 } 1511 1512 txd_upper = txd_lower = 0; 1513 1514 /* 1515 * Capture the first descriptor index, this descriptor 1516 * will have the index of the EOP which is the only one 1517 * that now gets a DONE bit writeback. 1518 */ 1519 first = tdata->next_avail_tx_desc; 1520 tx_buffer = &tdata->tx_buf[first]; 1521 tx_buffer_mapped = tx_buffer; 1522 map = tx_buffer->map; 1523 1524 maxsegs = tdata->num_tx_desc_avail - EMX_TX_RESERVED; 1525 KASSERT(maxsegs >= tdata->spare_tx_desc, ("not enough spare TX desc")); 1526 if (maxsegs > EMX_MAX_SCATTER) 1527 maxsegs = EMX_MAX_SCATTER; 1528 1529 error = bus_dmamap_load_mbuf_defrag(tdata->txtag, map, m_headp, 1530 segs, maxsegs, &nsegs, BUS_DMA_NOWAIT); 1531 if (error) { 1532 m_freem(*m_headp); 1533 *m_headp = NULL; 1534 return error; 1535 } 1536 bus_dmamap_sync(tdata->txtag, map, BUS_DMASYNC_PREWRITE); 1537 1538 m_head = *m_headp; 1539 tdata->tx_nsegs += nsegs; 1540 *segs_used += nsegs; 1541 1542 if (m_head->m_pkthdr.csum_flags & CSUM_TSO) { 1543 /* TSO will consume one TX desc */ 1544 i = emx_tso_setup(tdata, m_head, &txd_upper, &txd_lower); 1545 tdata->tx_nsegs += i; 1546 *segs_used += i; 1547 } else if (m_head->m_pkthdr.csum_flags & EMX_CSUM_FEATURES) { 1548 /* TX csum offloading will consume one TX desc */ 1549 i = emx_txcsum(tdata, m_head, &txd_upper, &txd_lower); 1550 tdata->tx_nsegs += i; 1551 *segs_used += i; 1552 } 1553 i = tdata->next_avail_tx_desc; 1554 1555 /* Set up our transmit descriptors */ 1556 for (j = 0; j < nsegs; j++) { 1557 tx_buffer = &tdata->tx_buf[i]; 1558 ctxd = &tdata->tx_desc_base[i]; 1559 1560 ctxd->buffer_addr = htole64(segs[j].ds_addr); 1561 ctxd->lower.data = htole32(E1000_TXD_CMD_IFCS | 1562 txd_lower | segs[j].ds_len); 1563 ctxd->upper.data = htole32(txd_upper); 1564 1565 last = i; 1566 if (++i == tdata->num_tx_desc) 1567 i = 0; 1568 } 1569 1570 tdata->next_avail_tx_desc = i; 1571 1572 KKASSERT(tdata->num_tx_desc_avail > nsegs); 1573 tdata->num_tx_desc_avail -= nsegs; 1574 1575 /* Handle VLAN tag */ 1576 if (m_head->m_flags & M_VLANTAG) { 1577 /* Set the vlan id. */ 1578 ctxd->upper.fields.special = 1579 htole16(m_head->m_pkthdr.ether_vlantag); 1580 1581 /* Tell hardware to add tag */ 1582 ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE); 1583 } 1584 1585 tx_buffer->m_head = m_head; 1586 tx_buffer_mapped->map = tx_buffer->map; 1587 tx_buffer->map = map; 1588 1589 if (tdata->tx_nsegs >= tdata->tx_intr_nsegs) { 1590 tdata->tx_nsegs = 0; 1591 1592 /* 1593 * Report Status (RS) is turned on 1594 * every tx_intr_nsegs descriptors. 1595 */ 1596 cmd = E1000_TXD_CMD_RS; 1597 1598 /* 1599 * Keep track of the descriptor, which will 1600 * be written back by hardware. 1601 */ 1602 tdata->tx_dd[tdata->tx_dd_tail] = last; 1603 EMX_INC_TXDD_IDX(tdata->tx_dd_tail); 1604 KKASSERT(tdata->tx_dd_tail != tdata->tx_dd_head); 1605 } 1606 1607 /* 1608 * Last Descriptor of Packet needs End Of Packet (EOP) 1609 */ 1610 ctxd->lower.data |= htole32(E1000_TXD_CMD_EOP | cmd); 1611 1612 /* 1613 * Defer TDT updating, until enough descriptors are setup 1614 */ 1615 *idx = i; 1616 1617 #ifdef EMX_TSS_DEBUG 1618 tdata->tx_pkts++; 1619 #endif 1620 1621 return (0); 1622 } 1623 1624 static void 1625 emx_set_promisc(struct emx_softc *sc) 1626 { 1627 struct ifnet *ifp = &sc->arpcom.ac_if; 1628 uint32_t reg_rctl; 1629 1630 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL); 1631 1632 if (ifp->if_flags & IFF_PROMISC) { 1633 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE); 1634 /* Turn this on if you want to see bad packets */ 1635 if (emx_debug_sbp) 1636 reg_rctl |= E1000_RCTL_SBP; 1637 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl); 1638 } else if (ifp->if_flags & IFF_ALLMULTI) { 1639 reg_rctl |= E1000_RCTL_MPE; 1640 reg_rctl &= ~E1000_RCTL_UPE; 1641 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl); 1642 } 1643 } 1644 1645 static void 1646 emx_disable_promisc(struct emx_softc *sc) 1647 { 1648 uint32_t reg_rctl; 1649 1650 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL); 1651 1652 reg_rctl &= ~E1000_RCTL_UPE; 1653 reg_rctl &= ~E1000_RCTL_MPE; 1654 reg_rctl &= ~E1000_RCTL_SBP; 1655 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl); 1656 } 1657 1658 static void 1659 emx_set_multi(struct emx_softc *sc) 1660 { 1661 struct ifnet *ifp = &sc->arpcom.ac_if; 1662 struct ifmultiaddr *ifma; 1663 uint32_t reg_rctl = 0; 1664 uint8_t *mta; 1665 int mcnt = 0; 1666 1667 mta = sc->mta; 1668 bzero(mta, ETH_ADDR_LEN * EMX_MCAST_ADDR_MAX); 1669 1670 TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { 1671 if (ifma->ifma_addr->sa_family != AF_LINK) 1672 continue; 1673 1674 if (mcnt == EMX_MCAST_ADDR_MAX) 1675 break; 1676 1677 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr), 1678 &mta[mcnt * ETHER_ADDR_LEN], ETHER_ADDR_LEN); 1679 mcnt++; 1680 } 1681 1682 if (mcnt >= EMX_MCAST_ADDR_MAX) { 1683 reg_rctl = E1000_READ_REG(&sc->hw, E1000_RCTL); 1684 reg_rctl |= E1000_RCTL_MPE; 1685 E1000_WRITE_REG(&sc->hw, E1000_RCTL, reg_rctl); 1686 } else { 1687 e1000_update_mc_addr_list(&sc->hw, mta, mcnt); 1688 } 1689 } 1690 1691 /* 1692 * This routine checks for link status and updates statistics. 1693 */ 1694 static void 1695 emx_timer(void *xsc) 1696 { 1697 struct emx_softc *sc = xsc; 1698 struct ifnet *ifp = &sc->arpcom.ac_if; 1699 1700 lwkt_serialize_enter(&sc->main_serialize); 1701 1702 emx_update_link_status(sc); 1703 emx_update_stats(sc); 1704 1705 /* Reset LAA into RAR[0] on 82571 */ 1706 if (e1000_get_laa_state_82571(&sc->hw) == TRUE) 1707 e1000_rar_set(&sc->hw, sc->hw.mac.addr, 0); 1708 1709 if (emx_display_debug_stats && (ifp->if_flags & IFF_RUNNING)) 1710 emx_print_hw_stats(sc); 1711 1712 emx_smartspeed(sc); 1713 1714 callout_reset(&sc->timer, hz, emx_timer, sc); 1715 1716 lwkt_serialize_exit(&sc->main_serialize); 1717 } 1718 1719 static void 1720 emx_update_link_status(struct emx_softc *sc) 1721 { 1722 struct e1000_hw *hw = &sc->hw; 1723 struct ifnet *ifp = &sc->arpcom.ac_if; 1724 device_t dev = sc->dev; 1725 uint32_t link_check = 0; 1726 1727 /* Get the cached link value or read phy for real */ 1728 switch (hw->phy.media_type) { 1729 case e1000_media_type_copper: 1730 if (hw->mac.get_link_status) { 1731 /* Do the work to read phy */ 1732 e1000_check_for_link(hw); 1733 link_check = !hw->mac.get_link_status; 1734 if (link_check) /* ESB2 fix */ 1735 e1000_cfg_on_link_up(hw); 1736 } else { 1737 link_check = TRUE; 1738 } 1739 break; 1740 1741 case e1000_media_type_fiber: 1742 e1000_check_for_link(hw); 1743 link_check = E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU; 1744 break; 1745 1746 case e1000_media_type_internal_serdes: 1747 e1000_check_for_link(hw); 1748 link_check = sc->hw.mac.serdes_has_link; 1749 break; 1750 1751 case e1000_media_type_unknown: 1752 default: 1753 break; 1754 } 1755 1756 /* Now check for a transition */ 1757 if (link_check && sc->link_active == 0) { 1758 e1000_get_speed_and_duplex(hw, &sc->link_speed, 1759 &sc->link_duplex); 1760 1761 /* 1762 * Check if we should enable/disable SPEED_MODE bit on 1763 * 82571EB/82572EI 1764 */ 1765 if (sc->link_speed != SPEED_1000 && 1766 (hw->mac.type == e1000_82571 || 1767 hw->mac.type == e1000_82572)) { 1768 int tarc0; 1769 1770 tarc0 = E1000_READ_REG(hw, E1000_TARC(0)); 1771 tarc0 &= ~EMX_TARC_SPEED_MODE; 1772 E1000_WRITE_REG(hw, E1000_TARC(0), tarc0); 1773 } 1774 if (bootverbose) { 1775 device_printf(dev, "Link is up %d Mbps %s\n", 1776 sc->link_speed, 1777 ((sc->link_duplex == FULL_DUPLEX) ? 1778 "Full Duplex" : "Half Duplex")); 1779 } 1780 sc->link_active = 1; 1781 sc->smartspeed = 0; 1782 ifp->if_baudrate = sc->link_speed * 1000000; 1783 ifp->if_link_state = LINK_STATE_UP; 1784 if_link_state_change(ifp); 1785 } else if (!link_check && sc->link_active == 1) { 1786 ifp->if_baudrate = sc->link_speed = 0; 1787 sc->link_duplex = 0; 1788 if (bootverbose) 1789 device_printf(dev, "Link is Down\n"); 1790 sc->link_active = 0; 1791 ifp->if_link_state = LINK_STATE_DOWN; 1792 if_link_state_change(ifp); 1793 } 1794 } 1795 1796 static void 1797 emx_stop(struct emx_softc *sc) 1798 { 1799 struct ifnet *ifp = &sc->arpcom.ac_if; 1800 int i; 1801 1802 ASSERT_IFNET_SERIALIZED_ALL(ifp); 1803 1804 emx_disable_intr(sc); 1805 1806 callout_stop(&sc->timer); 1807 1808 ifp->if_flags &= ~IFF_RUNNING; 1809 for (i = 0; i < sc->tx_ring_cnt; ++i) { 1810 struct emx_txdata *tdata = &sc->tx_data[i]; 1811 1812 ifsq_clr_oactive(tdata->ifsq); 1813 ifsq_watchdog_stop(&tdata->tx_watchdog); 1814 tdata->tx_flags &= ~EMX_TXFLAG_ENABLED; 1815 } 1816 1817 /* 1818 * Disable multiple receive queues. 1819 * 1820 * NOTE: 1821 * We should disable multiple receive queues before 1822 * resetting the hardware. 1823 */ 1824 E1000_WRITE_REG(&sc->hw, E1000_MRQC, 0); 1825 1826 e1000_reset_hw(&sc->hw); 1827 E1000_WRITE_REG(&sc->hw, E1000_WUC, 0); 1828 1829 for (i = 0; i < sc->tx_ring_cnt; ++i) 1830 emx_free_tx_ring(&sc->tx_data[i]); 1831 for (i = 0; i < sc->rx_ring_cnt; ++i) 1832 emx_free_rx_ring(&sc->rx_data[i]); 1833 } 1834 1835 static int 1836 emx_reset(struct emx_softc *sc) 1837 { 1838 device_t dev = sc->dev; 1839 uint16_t rx_buffer_size; 1840 uint32_t pba; 1841 1842 /* Set up smart power down as default off on newer adapters. */ 1843 if (!emx_smart_pwr_down && 1844 (sc->hw.mac.type == e1000_82571 || 1845 sc->hw.mac.type == e1000_82572)) { 1846 uint16_t phy_tmp = 0; 1847 1848 /* Speed up time to link by disabling smart power down. */ 1849 e1000_read_phy_reg(&sc->hw, 1850 IGP02E1000_PHY_POWER_MGMT, &phy_tmp); 1851 phy_tmp &= ~IGP02E1000_PM_SPD; 1852 e1000_write_phy_reg(&sc->hw, 1853 IGP02E1000_PHY_POWER_MGMT, phy_tmp); 1854 } 1855 1856 /* 1857 * Packet Buffer Allocation (PBA) 1858 * Writing PBA sets the receive portion of the buffer 1859 * the remainder is used for the transmit buffer. 1860 */ 1861 switch (sc->hw.mac.type) { 1862 /* Total Packet Buffer on these is 48K */ 1863 case e1000_82571: 1864 case e1000_82572: 1865 case e1000_80003es2lan: 1866 pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */ 1867 break; 1868 1869 case e1000_82573: /* 82573: Total Packet Buffer is 32K */ 1870 pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */ 1871 break; 1872 1873 case e1000_82574: 1874 pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */ 1875 break; 1876 1877 default: 1878 /* Devices before 82547 had a Packet Buffer of 64K. */ 1879 if (sc->max_frame_size > 8192) 1880 pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */ 1881 else 1882 pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */ 1883 } 1884 E1000_WRITE_REG(&sc->hw, E1000_PBA, pba); 1885 1886 /* 1887 * These parameters control the automatic generation (Tx) and 1888 * response (Rx) to Ethernet PAUSE frames. 1889 * - High water mark should allow for at least two frames to be 1890 * received after sending an XOFF. 1891 * - Low water mark works best when it is very near the high water mark. 1892 * This allows the receiver to restart by sending XON when it has 1893 * drained a bit. Here we use an arbitary value of 1500 which will 1894 * restart after one full frame is pulled from the buffer. There 1895 * could be several smaller frames in the buffer and if so they will 1896 * not trigger the XON until their total number reduces the buffer 1897 * by 1500. 1898 * - The pause time is fairly large at 1000 x 512ns = 512 usec. 1899 */ 1900 rx_buffer_size = (E1000_READ_REG(&sc->hw, E1000_PBA) & 0xffff) << 10; 1901 1902 sc->hw.fc.high_water = rx_buffer_size - 1903 roundup2(sc->max_frame_size, 1024); 1904 sc->hw.fc.low_water = sc->hw.fc.high_water - 1500; 1905 1906 if (sc->hw.mac.type == e1000_80003es2lan) 1907 sc->hw.fc.pause_time = 0xFFFF; 1908 else 1909 sc->hw.fc.pause_time = EMX_FC_PAUSE_TIME; 1910 sc->hw.fc.send_xon = TRUE; 1911 sc->hw.fc.requested_mode = e1000_fc_full; 1912 1913 /* Issue a global reset */ 1914 e1000_reset_hw(&sc->hw); 1915 E1000_WRITE_REG(&sc->hw, E1000_WUC, 0); 1916 emx_disable_aspm(sc); 1917 1918 if (e1000_init_hw(&sc->hw) < 0) { 1919 device_printf(dev, "Hardware Initialization Failed\n"); 1920 return (EIO); 1921 } 1922 1923 E1000_WRITE_REG(&sc->hw, E1000_VET, ETHERTYPE_VLAN); 1924 e1000_get_phy_info(&sc->hw); 1925 e1000_check_for_link(&sc->hw); 1926 1927 return (0); 1928 } 1929 1930 static void 1931 emx_setup_ifp(struct emx_softc *sc) 1932 { 1933 struct ifnet *ifp = &sc->arpcom.ac_if; 1934 1935 if_initname(ifp, device_get_name(sc->dev), 1936 device_get_unit(sc->dev)); 1937 ifp->if_softc = sc; 1938 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 1939 ifp->if_init = emx_init; 1940 ifp->if_ioctl = emx_ioctl; 1941 ifp->if_start = emx_start; 1942 #ifdef IFPOLL_ENABLE 1943 ifp->if_npoll = emx_npoll; 1944 #endif 1945 ifp->if_serialize = emx_serialize; 1946 ifp->if_deserialize = emx_deserialize; 1947 ifp->if_tryserialize = emx_tryserialize; 1948 #ifdef INVARIANTS 1949 ifp->if_serialize_assert = emx_serialize_assert; 1950 #endif 1951 1952 ifq_set_maxlen(&ifp->if_snd, sc->tx_data[0].num_tx_desc - 1); 1953 ifq_set_ready(&ifp->if_snd); 1954 ifq_set_subq_cnt(&ifp->if_snd, sc->tx_ring_cnt); 1955 1956 ifp->if_mapsubq = ifq_mapsubq_mask; 1957 ifq_set_subq_mask(&ifp->if_snd, 0); 1958 1959 ether_ifattach(ifp, sc->hw.mac.addr, NULL); 1960 1961 ifp->if_capabilities = IFCAP_HWCSUM | 1962 IFCAP_VLAN_HWTAGGING | 1963 IFCAP_VLAN_MTU | 1964 IFCAP_TSO; 1965 if (sc->rx_ring_cnt > 1) 1966 ifp->if_capabilities |= IFCAP_RSS; 1967 ifp->if_capenable = ifp->if_capabilities; 1968 ifp->if_hwassist = EMX_CSUM_FEATURES | CSUM_TSO; 1969 1970 /* 1971 * Tell the upper layer(s) we support long frames. 1972 */ 1973 ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header); 1974 1975 /* 1976 * Specify the media types supported by this sc and register 1977 * callbacks to update media and link information 1978 */ 1979 ifmedia_init(&sc->media, IFM_IMASK, 1980 emx_media_change, emx_media_status); 1981 if (sc->hw.phy.media_type == e1000_media_type_fiber || 1982 sc->hw.phy.media_type == e1000_media_type_internal_serdes) { 1983 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX | IFM_FDX, 1984 0, NULL); 1985 ifmedia_add(&sc->media, IFM_ETHER | IFM_1000_SX, 0, NULL); 1986 } else { 1987 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T, 0, NULL); 1988 ifmedia_add(&sc->media, IFM_ETHER | IFM_10_T | IFM_FDX, 1989 0, NULL); 1990 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX, 0, NULL); 1991 ifmedia_add(&sc->media, IFM_ETHER | IFM_100_TX | IFM_FDX, 1992 0, NULL); 1993 if (sc->hw.phy.type != e1000_phy_ife) { 1994 ifmedia_add(&sc->media, 1995 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL); 1996 ifmedia_add(&sc->media, 1997 IFM_ETHER | IFM_1000_T, 0, NULL); 1998 } 1999 } 2000 ifmedia_add(&sc->media, IFM_ETHER | IFM_AUTO, 0, NULL); 2001 ifmedia_set(&sc->media, IFM_ETHER | IFM_AUTO); 2002 } 2003 2004 /* 2005 * Workaround for SmartSpeed on 82541 and 82547 controllers 2006 */ 2007 static void 2008 emx_smartspeed(struct emx_softc *sc) 2009 { 2010 uint16_t phy_tmp; 2011 2012 if (sc->link_active || sc->hw.phy.type != e1000_phy_igp || 2013 sc->hw.mac.autoneg == 0 || 2014 (sc->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0) 2015 return; 2016 2017 if (sc->smartspeed == 0) { 2018 /* 2019 * If Master/Slave config fault is asserted twice, 2020 * we assume back-to-back 2021 */ 2022 e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp); 2023 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT)) 2024 return; 2025 e1000_read_phy_reg(&sc->hw, PHY_1000T_STATUS, &phy_tmp); 2026 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) { 2027 e1000_read_phy_reg(&sc->hw, 2028 PHY_1000T_CTRL, &phy_tmp); 2029 if (phy_tmp & CR_1000T_MS_ENABLE) { 2030 phy_tmp &= ~CR_1000T_MS_ENABLE; 2031 e1000_write_phy_reg(&sc->hw, 2032 PHY_1000T_CTRL, phy_tmp); 2033 sc->smartspeed++; 2034 if (sc->hw.mac.autoneg && 2035 !e1000_phy_setup_autoneg(&sc->hw) && 2036 !e1000_read_phy_reg(&sc->hw, 2037 PHY_CONTROL, &phy_tmp)) { 2038 phy_tmp |= MII_CR_AUTO_NEG_EN | 2039 MII_CR_RESTART_AUTO_NEG; 2040 e1000_write_phy_reg(&sc->hw, 2041 PHY_CONTROL, phy_tmp); 2042 } 2043 } 2044 } 2045 return; 2046 } else if (sc->smartspeed == EMX_SMARTSPEED_DOWNSHIFT) { 2047 /* If still no link, perhaps using 2/3 pair cable */ 2048 e1000_read_phy_reg(&sc->hw, PHY_1000T_CTRL, &phy_tmp); 2049 phy_tmp |= CR_1000T_MS_ENABLE; 2050 e1000_write_phy_reg(&sc->hw, PHY_1000T_CTRL, phy_tmp); 2051 if (sc->hw.mac.autoneg && 2052 !e1000_phy_setup_autoneg(&sc->hw) && 2053 !e1000_read_phy_reg(&sc->hw, PHY_CONTROL, &phy_tmp)) { 2054 phy_tmp |= MII_CR_AUTO_NEG_EN | MII_CR_RESTART_AUTO_NEG; 2055 e1000_write_phy_reg(&sc->hw, PHY_CONTROL, phy_tmp); 2056 } 2057 } 2058 2059 /* Restart process after EMX_SMARTSPEED_MAX iterations */ 2060 if (sc->smartspeed++ == EMX_SMARTSPEED_MAX) 2061 sc->smartspeed = 0; 2062 } 2063 2064 static int 2065 emx_create_tx_ring(struct emx_txdata *tdata) 2066 { 2067 device_t dev = tdata->sc->dev; 2068 struct emx_txbuf *tx_buffer; 2069 int error, i, tsize, ntxd; 2070 2071 /* 2072 * Validate number of transmit descriptors. It must not exceed 2073 * hardware maximum, and must be multiple of E1000_DBA_ALIGN. 2074 */ 2075 ntxd = device_getenv_int(dev, "txd", emx_txd); 2076 if ((ntxd * sizeof(struct e1000_tx_desc)) % EMX_DBA_ALIGN != 0 || 2077 ntxd > EMX_MAX_TXD || ntxd < EMX_MIN_TXD) { 2078 device_printf(dev, "Using %d TX descriptors instead of %d!\n", 2079 EMX_DEFAULT_TXD, ntxd); 2080 tdata->num_tx_desc = EMX_DEFAULT_TXD; 2081 } else { 2082 tdata->num_tx_desc = ntxd; 2083 } 2084 2085 /* 2086 * Allocate Transmit Descriptor ring 2087 */ 2088 tsize = roundup2(tdata->num_tx_desc * sizeof(struct e1000_tx_desc), 2089 EMX_DBA_ALIGN); 2090 tdata->tx_desc_base = bus_dmamem_coherent_any(tdata->sc->parent_dtag, 2091 EMX_DBA_ALIGN, tsize, BUS_DMA_WAITOK, 2092 &tdata->tx_desc_dtag, &tdata->tx_desc_dmap, 2093 &tdata->tx_desc_paddr); 2094 if (tdata->tx_desc_base == NULL) { 2095 device_printf(dev, "Unable to allocate tx_desc memory\n"); 2096 return ENOMEM; 2097 } 2098 2099 tsize = __VM_CACHELINE_ALIGN( 2100 sizeof(struct emx_txbuf) * tdata->num_tx_desc); 2101 tdata->tx_buf = kmalloc_cachealign(tsize, M_DEVBUF, M_WAITOK | M_ZERO); 2102 2103 /* 2104 * Create DMA tags for tx buffers 2105 */ 2106 error = bus_dma_tag_create(tdata->sc->parent_dtag, /* parent */ 2107 1, 0, /* alignment, bounds */ 2108 BUS_SPACE_MAXADDR, /* lowaddr */ 2109 BUS_SPACE_MAXADDR, /* highaddr */ 2110 NULL, NULL, /* filter, filterarg */ 2111 EMX_TSO_SIZE, /* maxsize */ 2112 EMX_MAX_SCATTER, /* nsegments */ 2113 EMX_MAX_SEGSIZE, /* maxsegsize */ 2114 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW | 2115 BUS_DMA_ONEBPAGE, /* flags */ 2116 &tdata->txtag); 2117 if (error) { 2118 device_printf(dev, "Unable to allocate TX DMA tag\n"); 2119 kfree(tdata->tx_buf, M_DEVBUF); 2120 tdata->tx_buf = NULL; 2121 return error; 2122 } 2123 2124 /* 2125 * Create DMA maps for tx buffers 2126 */ 2127 for (i = 0; i < tdata->num_tx_desc; i++) { 2128 tx_buffer = &tdata->tx_buf[i]; 2129 2130 error = bus_dmamap_create(tdata->txtag, 2131 BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE, 2132 &tx_buffer->map); 2133 if (error) { 2134 device_printf(dev, "Unable to create TX DMA map\n"); 2135 emx_destroy_tx_ring(tdata, i); 2136 return error; 2137 } 2138 } 2139 2140 /* 2141 * Setup TX parameters 2142 */ 2143 tdata->spare_tx_desc = EMX_TX_SPARE; 2144 tdata->tx_wreg_nsegs = EMX_DEFAULT_TXWREG; 2145 2146 /* 2147 * Keep following relationship between spare_tx_desc, oact_tx_desc 2148 * and tx_intr_nsegs: 2149 * (spare_tx_desc + EMX_TX_RESERVED) <= 2150 * oact_tx_desc <= EMX_TX_OACTIVE_MAX <= tx_intr_nsegs 2151 */ 2152 tdata->oact_tx_desc = tdata->num_tx_desc / 8; 2153 if (tdata->oact_tx_desc > EMX_TX_OACTIVE_MAX) 2154 tdata->oact_tx_desc = EMX_TX_OACTIVE_MAX; 2155 if (tdata->oact_tx_desc < tdata->spare_tx_desc + EMX_TX_RESERVED) 2156 tdata->oact_tx_desc = tdata->spare_tx_desc + EMX_TX_RESERVED; 2157 2158 tdata->tx_intr_nsegs = tdata->num_tx_desc / 16; 2159 if (tdata->tx_intr_nsegs < tdata->oact_tx_desc) 2160 tdata->tx_intr_nsegs = tdata->oact_tx_desc; 2161 2162 /* 2163 * Pullup extra 4bytes into the first data segment, see: 2164 * 82571/82572 specification update errata #7 2165 * 2166 * NOTE: 2167 * 4bytes instead of 2bytes, which are mentioned in the errata, 2168 * are pulled; mainly to keep rest of the data properly aligned. 2169 */ 2170 if (tdata->sc->hw.mac.type == e1000_82571 || 2171 tdata->sc->hw.mac.type == e1000_82572) 2172 tdata->tx_flags |= EMX_TXFLAG_TSO_PULLEX; 2173 2174 return (0); 2175 } 2176 2177 static void 2178 emx_init_tx_ring(struct emx_txdata *tdata) 2179 { 2180 /* Clear the old ring contents */ 2181 bzero(tdata->tx_desc_base, 2182 sizeof(struct e1000_tx_desc) * tdata->num_tx_desc); 2183 2184 /* Reset state */ 2185 tdata->next_avail_tx_desc = 0; 2186 tdata->next_tx_to_clean = 0; 2187 tdata->num_tx_desc_avail = tdata->num_tx_desc; 2188 2189 tdata->tx_flags |= EMX_TXFLAG_ENABLED; 2190 if (tdata->sc->tx_ring_inuse > 1) { 2191 tdata->tx_flags |= EMX_TXFLAG_FORCECTX; 2192 if (bootverbose) { 2193 if_printf(&tdata->sc->arpcom.ac_if, 2194 "TX %d force ctx setup\n", tdata->idx); 2195 } 2196 } 2197 } 2198 2199 static void 2200 emx_init_tx_unit(struct emx_softc *sc) 2201 { 2202 uint32_t tctl, tarc, tipg = 0; 2203 int i; 2204 2205 for (i = 0; i < sc->tx_ring_inuse; ++i) { 2206 struct emx_txdata *tdata = &sc->tx_data[i]; 2207 uint64_t bus_addr; 2208 2209 /* Setup the Base and Length of the Tx Descriptor Ring */ 2210 bus_addr = tdata->tx_desc_paddr; 2211 E1000_WRITE_REG(&sc->hw, E1000_TDLEN(i), 2212 tdata->num_tx_desc * sizeof(struct e1000_tx_desc)); 2213 E1000_WRITE_REG(&sc->hw, E1000_TDBAH(i), 2214 (uint32_t)(bus_addr >> 32)); 2215 E1000_WRITE_REG(&sc->hw, E1000_TDBAL(i), 2216 (uint32_t)bus_addr); 2217 /* Setup the HW Tx Head and Tail descriptor pointers */ 2218 E1000_WRITE_REG(&sc->hw, E1000_TDT(i), 0); 2219 E1000_WRITE_REG(&sc->hw, E1000_TDH(i), 0); 2220 } 2221 2222 /* Set the default values for the Tx Inter Packet Gap timer */ 2223 switch (sc->hw.mac.type) { 2224 case e1000_80003es2lan: 2225 tipg = DEFAULT_82543_TIPG_IPGR1; 2226 tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 << 2227 E1000_TIPG_IPGR2_SHIFT; 2228 break; 2229 2230 default: 2231 if (sc->hw.phy.media_type == e1000_media_type_fiber || 2232 sc->hw.phy.media_type == e1000_media_type_internal_serdes) 2233 tipg = DEFAULT_82543_TIPG_IPGT_FIBER; 2234 else 2235 tipg = DEFAULT_82543_TIPG_IPGT_COPPER; 2236 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT; 2237 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT; 2238 break; 2239 } 2240 2241 E1000_WRITE_REG(&sc->hw, E1000_TIPG, tipg); 2242 2243 /* NOTE: 0 is not allowed for TIDV */ 2244 E1000_WRITE_REG(&sc->hw, E1000_TIDV, 1); 2245 E1000_WRITE_REG(&sc->hw, E1000_TADV, 0); 2246 2247 if (sc->hw.mac.type == e1000_82571 || 2248 sc->hw.mac.type == e1000_82572) { 2249 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(0)); 2250 tarc |= EMX_TARC_SPEED_MODE; 2251 E1000_WRITE_REG(&sc->hw, E1000_TARC(0), tarc); 2252 } else if (sc->hw.mac.type == e1000_80003es2lan) { 2253 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(0)); 2254 tarc |= 1; 2255 E1000_WRITE_REG(&sc->hw, E1000_TARC(0), tarc); 2256 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(1)); 2257 tarc |= 1; 2258 E1000_WRITE_REG(&sc->hw, E1000_TARC(1), tarc); 2259 } 2260 2261 /* Program the Transmit Control Register */ 2262 tctl = E1000_READ_REG(&sc->hw, E1000_TCTL); 2263 tctl &= ~E1000_TCTL_CT; 2264 tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN | 2265 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT); 2266 tctl |= E1000_TCTL_MULR; 2267 2268 /* This write will effectively turn on the transmit unit. */ 2269 E1000_WRITE_REG(&sc->hw, E1000_TCTL, tctl); 2270 2271 if (sc->hw.mac.type == e1000_82571 || 2272 sc->hw.mac.type == e1000_82572 || 2273 sc->hw.mac.type == e1000_80003es2lan) { 2274 /* Bit 28 of TARC1 must be cleared when MULR is enabled */ 2275 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(1)); 2276 tarc &= ~(1 << 28); 2277 E1000_WRITE_REG(&sc->hw, E1000_TARC(1), tarc); 2278 } 2279 2280 if (sc->tx_ring_inuse > 1) { 2281 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(0)); 2282 tarc &= ~EMX_TARC_COUNT_MASK; 2283 tarc |= 1; 2284 E1000_WRITE_REG(&sc->hw, E1000_TARC(0), tarc); 2285 2286 tarc = E1000_READ_REG(&sc->hw, E1000_TARC(1)); 2287 tarc &= ~EMX_TARC_COUNT_MASK; 2288 tarc |= 1; 2289 E1000_WRITE_REG(&sc->hw, E1000_TARC(1), tarc); 2290 } 2291 } 2292 2293 static void 2294 emx_destroy_tx_ring(struct emx_txdata *tdata, int ndesc) 2295 { 2296 struct emx_txbuf *tx_buffer; 2297 int i; 2298 2299 /* Free Transmit Descriptor ring */ 2300 if (tdata->tx_desc_base) { 2301 bus_dmamap_unload(tdata->tx_desc_dtag, tdata->tx_desc_dmap); 2302 bus_dmamem_free(tdata->tx_desc_dtag, tdata->tx_desc_base, 2303 tdata->tx_desc_dmap); 2304 bus_dma_tag_destroy(tdata->tx_desc_dtag); 2305 2306 tdata->tx_desc_base = NULL; 2307 } 2308 2309 if (tdata->tx_buf == NULL) 2310 return; 2311 2312 for (i = 0; i < ndesc; i++) { 2313 tx_buffer = &tdata->tx_buf[i]; 2314 2315 KKASSERT(tx_buffer->m_head == NULL); 2316 bus_dmamap_destroy(tdata->txtag, tx_buffer->map); 2317 } 2318 bus_dma_tag_destroy(tdata->txtag); 2319 2320 kfree(tdata->tx_buf, M_DEVBUF); 2321 tdata->tx_buf = NULL; 2322 } 2323 2324 /* 2325 * The offload context needs to be set when we transfer the first 2326 * packet of a particular protocol (TCP/UDP). This routine has been 2327 * enhanced to deal with inserted VLAN headers. 2328 * 2329 * If the new packet's ether header length, ip header length and 2330 * csum offloading type are same as the previous packet, we should 2331 * avoid allocating a new csum context descriptor; mainly to take 2332 * advantage of the pipeline effect of the TX data read request. 2333 * 2334 * This function returns number of TX descrptors allocated for 2335 * csum context. 2336 */ 2337 static int 2338 emx_txcsum(struct emx_txdata *tdata, struct mbuf *mp, 2339 uint32_t *txd_upper, uint32_t *txd_lower) 2340 { 2341 struct e1000_context_desc *TXD; 2342 int curr_txd, ehdrlen, csum_flags; 2343 uint32_t cmd, hdr_len, ip_hlen; 2344 2345 csum_flags = mp->m_pkthdr.csum_flags & EMX_CSUM_FEATURES; 2346 ip_hlen = mp->m_pkthdr.csum_iphlen; 2347 ehdrlen = mp->m_pkthdr.csum_lhlen; 2348 2349 if ((tdata->tx_flags & EMX_TXFLAG_FORCECTX) == 0 && 2350 tdata->csum_lhlen == ehdrlen && tdata->csum_iphlen == ip_hlen && 2351 tdata->csum_flags == csum_flags) { 2352 /* 2353 * Same csum offload context as the previous packets; 2354 * just return. 2355 */ 2356 *txd_upper = tdata->csum_txd_upper; 2357 *txd_lower = tdata->csum_txd_lower; 2358 return 0; 2359 } 2360 2361 /* 2362 * Setup a new csum offload context. 2363 */ 2364 2365 curr_txd = tdata->next_avail_tx_desc; 2366 TXD = (struct e1000_context_desc *)&tdata->tx_desc_base[curr_txd]; 2367 2368 cmd = 0; 2369 2370 /* Setup of IP header checksum. */ 2371 if (csum_flags & CSUM_IP) { 2372 /* 2373 * Start offset for header checksum calculation. 2374 * End offset for header checksum calculation. 2375 * Offset of place to put the checksum. 2376 */ 2377 TXD->lower_setup.ip_fields.ipcss = ehdrlen; 2378 TXD->lower_setup.ip_fields.ipcse = 2379 htole16(ehdrlen + ip_hlen - 1); 2380 TXD->lower_setup.ip_fields.ipcso = 2381 ehdrlen + offsetof(struct ip, ip_sum); 2382 cmd |= E1000_TXD_CMD_IP; 2383 *txd_upper |= E1000_TXD_POPTS_IXSM << 8; 2384 } 2385 hdr_len = ehdrlen + ip_hlen; 2386 2387 if (csum_flags & CSUM_TCP) { 2388 /* 2389 * Start offset for payload checksum calculation. 2390 * End offset for payload checksum calculation. 2391 * Offset of place to put the checksum. 2392 */ 2393 TXD->upper_setup.tcp_fields.tucss = hdr_len; 2394 TXD->upper_setup.tcp_fields.tucse = htole16(0); 2395 TXD->upper_setup.tcp_fields.tucso = 2396 hdr_len + offsetof(struct tcphdr, th_sum); 2397 cmd |= E1000_TXD_CMD_TCP; 2398 *txd_upper |= E1000_TXD_POPTS_TXSM << 8; 2399 } else if (csum_flags & CSUM_UDP) { 2400 /* 2401 * Start offset for header checksum calculation. 2402 * End offset for header checksum calculation. 2403 * Offset of place to put the checksum. 2404 */ 2405 TXD->upper_setup.tcp_fields.tucss = hdr_len; 2406 TXD->upper_setup.tcp_fields.tucse = htole16(0); 2407 TXD->upper_setup.tcp_fields.tucso = 2408 hdr_len + offsetof(struct udphdr, uh_sum); 2409 *txd_upper |= E1000_TXD_POPTS_TXSM << 8; 2410 } 2411 2412 *txd_lower = E1000_TXD_CMD_DEXT | /* Extended descr type */ 2413 E1000_TXD_DTYP_D; /* Data descr */ 2414 2415 /* Save the information for this csum offloading context */ 2416 tdata->csum_lhlen = ehdrlen; 2417 tdata->csum_iphlen = ip_hlen; 2418 tdata->csum_flags = csum_flags; 2419 tdata->csum_txd_upper = *txd_upper; 2420 tdata->csum_txd_lower = *txd_lower; 2421 2422 TXD->tcp_seg_setup.data = htole32(0); 2423 TXD->cmd_and_length = 2424 htole32(E1000_TXD_CMD_IFCS | E1000_TXD_CMD_DEXT | cmd); 2425 2426 if (++curr_txd == tdata->num_tx_desc) 2427 curr_txd = 0; 2428 2429 KKASSERT(tdata->num_tx_desc_avail > 0); 2430 tdata->num_tx_desc_avail--; 2431 2432 tdata->next_avail_tx_desc = curr_txd; 2433 return 1; 2434 } 2435 2436 static void 2437 emx_txeof(struct emx_txdata *tdata) 2438 { 2439 struct ifnet *ifp = &tdata->sc->arpcom.ac_if; 2440 struct emx_txbuf *tx_buffer; 2441 int first, num_avail; 2442 2443 if (tdata->tx_dd_head == tdata->tx_dd_tail) 2444 return; 2445 2446 if (tdata->num_tx_desc_avail == tdata->num_tx_desc) 2447 return; 2448 2449 num_avail = tdata->num_tx_desc_avail; 2450 first = tdata->next_tx_to_clean; 2451 2452 while (tdata->tx_dd_head != tdata->tx_dd_tail) { 2453 int dd_idx = tdata->tx_dd[tdata->tx_dd_head]; 2454 struct e1000_tx_desc *tx_desc; 2455 2456 tx_desc = &tdata->tx_desc_base[dd_idx]; 2457 if (tx_desc->upper.fields.status & E1000_TXD_STAT_DD) { 2458 EMX_INC_TXDD_IDX(tdata->tx_dd_head); 2459 2460 if (++dd_idx == tdata->num_tx_desc) 2461 dd_idx = 0; 2462 2463 while (first != dd_idx) { 2464 logif(pkt_txclean); 2465 2466 num_avail++; 2467 2468 tx_buffer = &tdata->tx_buf[first]; 2469 if (tx_buffer->m_head) { 2470 IFNET_STAT_INC(ifp, opackets, 1); 2471 bus_dmamap_unload(tdata->txtag, 2472 tx_buffer->map); 2473 m_freem(tx_buffer->m_head); 2474 tx_buffer->m_head = NULL; 2475 } 2476 2477 if (++first == tdata->num_tx_desc) 2478 first = 0; 2479 } 2480 } else { 2481 break; 2482 } 2483 } 2484 tdata->next_tx_to_clean = first; 2485 tdata->num_tx_desc_avail = num_avail; 2486 2487 if (tdata->tx_dd_head == tdata->tx_dd_tail) { 2488 tdata->tx_dd_head = 0; 2489 tdata->tx_dd_tail = 0; 2490 } 2491 2492 if (!EMX_IS_OACTIVE(tdata)) { 2493 ifsq_clr_oactive(tdata->ifsq); 2494 2495 /* All clean, turn off the timer */ 2496 if (tdata->num_tx_desc_avail == tdata->num_tx_desc) 2497 tdata->tx_watchdog.wd_timer = 0; 2498 } 2499 } 2500 2501 static void 2502 emx_tx_collect(struct emx_txdata *tdata) 2503 { 2504 struct ifnet *ifp = &tdata->sc->arpcom.ac_if; 2505 struct emx_txbuf *tx_buffer; 2506 int tdh, first, num_avail, dd_idx = -1; 2507 2508 if (tdata->num_tx_desc_avail == tdata->num_tx_desc) 2509 return; 2510 2511 tdh = E1000_READ_REG(&tdata->sc->hw, E1000_TDH(tdata->idx)); 2512 if (tdh == tdata->next_tx_to_clean) 2513 return; 2514 2515 if (tdata->tx_dd_head != tdata->tx_dd_tail) 2516 dd_idx = tdata->tx_dd[tdata->tx_dd_head]; 2517 2518 num_avail = tdata->num_tx_desc_avail; 2519 first = tdata->next_tx_to_clean; 2520 2521 while (first != tdh) { 2522 logif(pkt_txclean); 2523 2524 num_avail++; 2525 2526 tx_buffer = &tdata->tx_buf[first]; 2527 if (tx_buffer->m_head) { 2528 IFNET_STAT_INC(ifp, opackets, 1); 2529 bus_dmamap_unload(tdata->txtag, 2530 tx_buffer->map); 2531 m_freem(tx_buffer->m_head); 2532 tx_buffer->m_head = NULL; 2533 } 2534 2535 if (first == dd_idx) { 2536 EMX_INC_TXDD_IDX(tdata->tx_dd_head); 2537 if (tdata->tx_dd_head == tdata->tx_dd_tail) { 2538 tdata->tx_dd_head = 0; 2539 tdata->tx_dd_tail = 0; 2540 dd_idx = -1; 2541 } else { 2542 dd_idx = tdata->tx_dd[tdata->tx_dd_head]; 2543 } 2544 } 2545 2546 if (++first == tdata->num_tx_desc) 2547 first = 0; 2548 } 2549 tdata->next_tx_to_clean = first; 2550 tdata->num_tx_desc_avail = num_avail; 2551 2552 if (!EMX_IS_OACTIVE(tdata)) { 2553 ifsq_clr_oactive(tdata->ifsq); 2554 2555 /* All clean, turn off the timer */ 2556 if (tdata->num_tx_desc_avail == tdata->num_tx_desc) 2557 tdata->tx_watchdog.wd_timer = 0; 2558 } 2559 } 2560 2561 /* 2562 * When Link is lost sometimes there is work still in the TX ring 2563 * which will result in a watchdog, rather than allow that do an 2564 * attempted cleanup and then reinit here. Note that this has been 2565 * seens mostly with fiber adapters. 2566 */ 2567 static void 2568 emx_tx_purge(struct emx_softc *sc) 2569 { 2570 int i; 2571 2572 if (sc->link_active) 2573 return; 2574 2575 for (i = 0; i < sc->tx_ring_inuse; ++i) { 2576 struct emx_txdata *tdata = &sc->tx_data[i]; 2577 2578 if (tdata->tx_watchdog.wd_timer) { 2579 emx_tx_collect(tdata); 2580 if (tdata->tx_watchdog.wd_timer) { 2581 if_printf(&sc->arpcom.ac_if, 2582 "Link lost, TX pending, reinit\n"); 2583 emx_init(sc); 2584 return; 2585 } 2586 } 2587 } 2588 } 2589 2590 static int 2591 emx_newbuf(struct emx_rxdata *rdata, int i, int init) 2592 { 2593 struct mbuf *m; 2594 bus_dma_segment_t seg; 2595 bus_dmamap_t map; 2596 struct emx_rxbuf *rx_buffer; 2597 int error, nseg; 2598 2599 m = m_getcl(init ? MB_WAIT : MB_DONTWAIT, MT_DATA, M_PKTHDR); 2600 if (m == NULL) { 2601 if (init) { 2602 if_printf(&rdata->sc->arpcom.ac_if, 2603 "Unable to allocate RX mbuf\n"); 2604 } 2605 return (ENOBUFS); 2606 } 2607 m->m_len = m->m_pkthdr.len = MCLBYTES; 2608 2609 if (rdata->sc->max_frame_size <= MCLBYTES - ETHER_ALIGN) 2610 m_adj(m, ETHER_ALIGN); 2611 2612 error = bus_dmamap_load_mbuf_segment(rdata->rxtag, 2613 rdata->rx_sparemap, m, 2614 &seg, 1, &nseg, BUS_DMA_NOWAIT); 2615 if (error) { 2616 m_freem(m); 2617 if (init) { 2618 if_printf(&rdata->sc->arpcom.ac_if, 2619 "Unable to load RX mbuf\n"); 2620 } 2621 return (error); 2622 } 2623 2624 rx_buffer = &rdata->rx_buf[i]; 2625 if (rx_buffer->m_head != NULL) 2626 bus_dmamap_unload(rdata->rxtag, rx_buffer->map); 2627 2628 map = rx_buffer->map; 2629 rx_buffer->map = rdata->rx_sparemap; 2630 rdata->rx_sparemap = map; 2631 2632 rx_buffer->m_head = m; 2633 rx_buffer->paddr = seg.ds_addr; 2634 2635 emx_setup_rxdesc(&rdata->rx_desc[i], rx_buffer); 2636 return (0); 2637 } 2638 2639 static int 2640 emx_create_rx_ring(struct emx_rxdata *rdata) 2641 { 2642 device_t dev = rdata->sc->dev; 2643 struct emx_rxbuf *rx_buffer; 2644 int i, error, rsize, nrxd; 2645 2646 /* 2647 * Validate number of receive descriptors. It must not exceed 2648 * hardware maximum, and must be multiple of E1000_DBA_ALIGN. 2649 */ 2650 nrxd = device_getenv_int(dev, "rxd", emx_rxd); 2651 if ((nrxd * sizeof(emx_rxdesc_t)) % EMX_DBA_ALIGN != 0 || 2652 nrxd > EMX_MAX_RXD || nrxd < EMX_MIN_RXD) { 2653 device_printf(dev, "Using %d RX descriptors instead of %d!\n", 2654 EMX_DEFAULT_RXD, nrxd); 2655 rdata->num_rx_desc = EMX_DEFAULT_RXD; 2656 } else { 2657 rdata->num_rx_desc = nrxd; 2658 } 2659 2660 /* 2661 * Allocate Receive Descriptor ring 2662 */ 2663 rsize = roundup2(rdata->num_rx_desc * sizeof(emx_rxdesc_t), 2664 EMX_DBA_ALIGN); 2665 rdata->rx_desc = bus_dmamem_coherent_any(rdata->sc->parent_dtag, 2666 EMX_DBA_ALIGN, rsize, BUS_DMA_WAITOK, 2667 &rdata->rx_desc_dtag, &rdata->rx_desc_dmap, 2668 &rdata->rx_desc_paddr); 2669 if (rdata->rx_desc == NULL) { 2670 device_printf(dev, "Unable to allocate rx_desc memory\n"); 2671 return ENOMEM; 2672 } 2673 2674 rsize = __VM_CACHELINE_ALIGN( 2675 sizeof(struct emx_rxbuf) * rdata->num_rx_desc); 2676 rdata->rx_buf = kmalloc_cachealign(rsize, M_DEVBUF, M_WAITOK | M_ZERO); 2677 2678 /* 2679 * Create DMA tag for rx buffers 2680 */ 2681 error = bus_dma_tag_create(rdata->sc->parent_dtag, /* parent */ 2682 1, 0, /* alignment, bounds */ 2683 BUS_SPACE_MAXADDR, /* lowaddr */ 2684 BUS_SPACE_MAXADDR, /* highaddr */ 2685 NULL, NULL, /* filter, filterarg */ 2686 MCLBYTES, /* maxsize */ 2687 1, /* nsegments */ 2688 MCLBYTES, /* maxsegsize */ 2689 BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW, /* flags */ 2690 &rdata->rxtag); 2691 if (error) { 2692 device_printf(dev, "Unable to allocate RX DMA tag\n"); 2693 kfree(rdata->rx_buf, M_DEVBUF); 2694 rdata->rx_buf = NULL; 2695 return error; 2696 } 2697 2698 /* 2699 * Create spare DMA map for rx buffers 2700 */ 2701 error = bus_dmamap_create(rdata->rxtag, BUS_DMA_WAITOK, 2702 &rdata->rx_sparemap); 2703 if (error) { 2704 device_printf(dev, "Unable to create spare RX DMA map\n"); 2705 bus_dma_tag_destroy(rdata->rxtag); 2706 kfree(rdata->rx_buf, M_DEVBUF); 2707 rdata->rx_buf = NULL; 2708 return error; 2709 } 2710 2711 /* 2712 * Create DMA maps for rx buffers 2713 */ 2714 for (i = 0; i < rdata->num_rx_desc; i++) { 2715 rx_buffer = &rdata->rx_buf[i]; 2716 2717 error = bus_dmamap_create(rdata->rxtag, BUS_DMA_WAITOK, 2718 &rx_buffer->map); 2719 if (error) { 2720 device_printf(dev, "Unable to create RX DMA map\n"); 2721 emx_destroy_rx_ring(rdata, i); 2722 return error; 2723 } 2724 } 2725 return (0); 2726 } 2727 2728 static void 2729 emx_free_rx_ring(struct emx_rxdata *rdata) 2730 { 2731 int i; 2732 2733 for (i = 0; i < rdata->num_rx_desc; i++) { 2734 struct emx_rxbuf *rx_buffer = &rdata->rx_buf[i]; 2735 2736 if (rx_buffer->m_head != NULL) { 2737 bus_dmamap_unload(rdata->rxtag, rx_buffer->map); 2738 m_freem(rx_buffer->m_head); 2739 rx_buffer->m_head = NULL; 2740 } 2741 } 2742 2743 if (rdata->fmp != NULL) 2744 m_freem(rdata->fmp); 2745 rdata->fmp = NULL; 2746 rdata->lmp = NULL; 2747 } 2748 2749 static void 2750 emx_free_tx_ring(struct emx_txdata *tdata) 2751 { 2752 int i; 2753 2754 for (i = 0; i < tdata->num_tx_desc; i++) { 2755 struct emx_txbuf *tx_buffer = &tdata->tx_buf[i]; 2756 2757 if (tx_buffer->m_head != NULL) { 2758 bus_dmamap_unload(tdata->txtag, tx_buffer->map); 2759 m_freem(tx_buffer->m_head); 2760 tx_buffer->m_head = NULL; 2761 } 2762 } 2763 2764 tdata->tx_flags &= ~EMX_TXFLAG_FORCECTX; 2765 2766 tdata->csum_flags = 0; 2767 tdata->csum_lhlen = 0; 2768 tdata->csum_iphlen = 0; 2769 tdata->csum_thlen = 0; 2770 tdata->csum_mss = 0; 2771 tdata->csum_pktlen = 0; 2772 2773 tdata->tx_dd_head = 0; 2774 tdata->tx_dd_tail = 0; 2775 tdata->tx_nsegs = 0; 2776 } 2777 2778 static int 2779 emx_init_rx_ring(struct emx_rxdata *rdata) 2780 { 2781 int i, error; 2782 2783 /* Reset descriptor ring */ 2784 bzero(rdata->rx_desc, sizeof(emx_rxdesc_t) * rdata->num_rx_desc); 2785 2786 /* Allocate new ones. */ 2787 for (i = 0; i < rdata->num_rx_desc; i++) { 2788 error = emx_newbuf(rdata, i, 1); 2789 if (error) 2790 return (error); 2791 } 2792 2793 /* Setup our descriptor pointers */ 2794 rdata->next_rx_desc_to_check = 0; 2795 2796 return (0); 2797 } 2798 2799 static void 2800 emx_init_rx_unit(struct emx_softc *sc) 2801 { 2802 struct ifnet *ifp = &sc->arpcom.ac_if; 2803 uint64_t bus_addr; 2804 uint32_t rctl, itr, rfctl; 2805 int i; 2806 2807 /* 2808 * Make sure receives are disabled while setting 2809 * up the descriptor ring 2810 */ 2811 rctl = E1000_READ_REG(&sc->hw, E1000_RCTL); 2812 E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl & ~E1000_RCTL_EN); 2813 2814 /* 2815 * Set the interrupt throttling rate. Value is calculated 2816 * as ITR = 1 / (INT_THROTTLE_CEIL * 256ns) 2817 */ 2818 if (sc->int_throttle_ceil) 2819 itr = 1000000000 / 256 / sc->int_throttle_ceil; 2820 else 2821 itr = 0; 2822 emx_set_itr(sc, itr); 2823 2824 /* Use extended RX descriptor */ 2825 rfctl = E1000_RFCTL_EXTEN; 2826 2827 /* Disable accelerated ackknowledge */ 2828 if (sc->hw.mac.type == e1000_82574) 2829 rfctl |= E1000_RFCTL_ACK_DIS; 2830 2831 E1000_WRITE_REG(&sc->hw, E1000_RFCTL, rfctl); 2832 2833 /* 2834 * Receive Checksum Offload for TCP and UDP 2835 * 2836 * Checksum offloading is also enabled if multiple receive 2837 * queue is to be supported, since we need it to figure out 2838 * packet type. 2839 */ 2840 if ((ifp->if_capenable & IFCAP_RXCSUM) || 2841 sc->rx_ring_cnt > 1) { 2842 uint32_t rxcsum; 2843 2844 rxcsum = E1000_READ_REG(&sc->hw, E1000_RXCSUM); 2845 2846 /* 2847 * NOTE: 2848 * PCSD must be enabled to enable multiple 2849 * receive queues. 2850 */ 2851 rxcsum |= E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL | 2852 E1000_RXCSUM_PCSD; 2853 E1000_WRITE_REG(&sc->hw, E1000_RXCSUM, rxcsum); 2854 } 2855 2856 /* 2857 * Configure multiple receive queue (RSS) 2858 */ 2859 if (sc->rx_ring_cnt > 1) { 2860 uint8_t key[EMX_NRSSRK * EMX_RSSRK_SIZE]; 2861 uint32_t reta; 2862 2863 KASSERT(sc->rx_ring_cnt == EMX_NRX_RING, 2864 ("invalid number of RX ring (%d)", sc->rx_ring_cnt)); 2865 2866 /* 2867 * NOTE: 2868 * When we reach here, RSS has already been disabled 2869 * in emx_stop(), so we could safely configure RSS key 2870 * and redirect table. 2871 */ 2872 2873 /* 2874 * Configure RSS key 2875 */ 2876 toeplitz_get_key(key, sizeof(key)); 2877 for (i = 0; i < EMX_NRSSRK; ++i) { 2878 uint32_t rssrk; 2879 2880 rssrk = EMX_RSSRK_VAL(key, i); 2881 EMX_RSS_DPRINTF(sc, 1, "rssrk%d 0x%08x\n", i, rssrk); 2882 2883 E1000_WRITE_REG(&sc->hw, E1000_RSSRK(i), rssrk); 2884 } 2885 2886 /* 2887 * Configure RSS redirect table in following fashion: 2888 * (hash & ring_cnt_mask) == rdr_table[(hash & rdr_table_mask)] 2889 */ 2890 reta = 0; 2891 for (i = 0; i < EMX_RETA_SIZE; ++i) { 2892 uint32_t q; 2893 2894 q = (i % sc->rx_ring_cnt) << EMX_RETA_RINGIDX_SHIFT; 2895 reta |= q << (8 * i); 2896 } 2897 EMX_RSS_DPRINTF(sc, 1, "reta 0x%08x\n", reta); 2898 2899 for (i = 0; i < EMX_NRETA; ++i) 2900 E1000_WRITE_REG(&sc->hw, E1000_RETA(i), reta); 2901 2902 /* 2903 * Enable multiple receive queues. 2904 * Enable IPv4 RSS standard hash functions. 2905 * Disable RSS interrupt. 2906 */ 2907 E1000_WRITE_REG(&sc->hw, E1000_MRQC, 2908 E1000_MRQC_ENABLE_RSS_2Q | 2909 E1000_MRQC_RSS_FIELD_IPV4_TCP | 2910 E1000_MRQC_RSS_FIELD_IPV4); 2911 } 2912 2913 /* 2914 * XXX TEMPORARY WORKAROUND: on some systems with 82573 2915 * long latencies are observed, like Lenovo X60. This 2916 * change eliminates the problem, but since having positive 2917 * values in RDTR is a known source of problems on other 2918 * platforms another solution is being sought. 2919 */ 2920 if (emx_82573_workaround && sc->hw.mac.type == e1000_82573) { 2921 E1000_WRITE_REG(&sc->hw, E1000_RADV, EMX_RADV_82573); 2922 E1000_WRITE_REG(&sc->hw, E1000_RDTR, EMX_RDTR_82573); 2923 } 2924 2925 for (i = 0; i < sc->rx_ring_cnt; ++i) { 2926 struct emx_rxdata *rdata = &sc->rx_data[i]; 2927 2928 /* 2929 * Setup the Base and Length of the Rx Descriptor Ring 2930 */ 2931 bus_addr = rdata->rx_desc_paddr; 2932 E1000_WRITE_REG(&sc->hw, E1000_RDLEN(i), 2933 rdata->num_rx_desc * sizeof(emx_rxdesc_t)); 2934 E1000_WRITE_REG(&sc->hw, E1000_RDBAH(i), 2935 (uint32_t)(bus_addr >> 32)); 2936 E1000_WRITE_REG(&sc->hw, E1000_RDBAL(i), 2937 (uint32_t)bus_addr); 2938 2939 /* 2940 * Setup the HW Rx Head and Tail Descriptor Pointers 2941 */ 2942 E1000_WRITE_REG(&sc->hw, E1000_RDH(i), 0); 2943 E1000_WRITE_REG(&sc->hw, E1000_RDT(i), 2944 sc->rx_data[i].num_rx_desc - 1); 2945 } 2946 2947 /* Setup the Receive Control Register */ 2948 rctl &= ~(3 << E1000_RCTL_MO_SHIFT); 2949 rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO | 2950 E1000_RCTL_RDMTS_HALF | E1000_RCTL_SECRC | 2951 (sc->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT); 2952 2953 /* Make sure VLAN Filters are off */ 2954 rctl &= ~E1000_RCTL_VFE; 2955 2956 /* Don't store bad paket */ 2957 rctl &= ~E1000_RCTL_SBP; 2958 2959 /* MCLBYTES */ 2960 rctl |= E1000_RCTL_SZ_2048; 2961 2962 if (ifp->if_mtu > ETHERMTU) 2963 rctl |= E1000_RCTL_LPE; 2964 else 2965 rctl &= ~E1000_RCTL_LPE; 2966 2967 /* Enable Receives */ 2968 E1000_WRITE_REG(&sc->hw, E1000_RCTL, rctl); 2969 } 2970 2971 static void 2972 emx_destroy_rx_ring(struct emx_rxdata *rdata, int ndesc) 2973 { 2974 struct emx_rxbuf *rx_buffer; 2975 int i; 2976 2977 /* Free Receive Descriptor ring */ 2978 if (rdata->rx_desc) { 2979 bus_dmamap_unload(rdata->rx_desc_dtag, rdata->rx_desc_dmap); 2980 bus_dmamem_free(rdata->rx_desc_dtag, rdata->rx_desc, 2981 rdata->rx_desc_dmap); 2982 bus_dma_tag_destroy(rdata->rx_desc_dtag); 2983 2984 rdata->rx_desc = NULL; 2985 } 2986 2987 if (rdata->rx_buf == NULL) 2988 return; 2989 2990 for (i = 0; i < ndesc; i++) { 2991 rx_buffer = &rdata->rx_buf[i]; 2992 2993 KKASSERT(rx_buffer->m_head == NULL); 2994 bus_dmamap_destroy(rdata->rxtag, rx_buffer->map); 2995 } 2996 bus_dmamap_destroy(rdata->rxtag, rdata->rx_sparemap); 2997 bus_dma_tag_destroy(rdata->rxtag); 2998 2999 kfree(rdata->rx_buf, M_DEVBUF); 3000 rdata->rx_buf = NULL; 3001 } 3002 3003 static void 3004 emx_rxeof(struct emx_rxdata *rdata, int count) 3005 { 3006 struct ifnet *ifp = &rdata->sc->arpcom.ac_if; 3007 uint32_t staterr; 3008 emx_rxdesc_t *current_desc; 3009 struct mbuf *mp; 3010 int i; 3011 3012 i = rdata->next_rx_desc_to_check; 3013 current_desc = &rdata->rx_desc[i]; 3014 staterr = le32toh(current_desc->rxd_staterr); 3015 3016 if (!(staterr & E1000_RXD_STAT_DD)) 3017 return; 3018 3019 while ((staterr & E1000_RXD_STAT_DD) && count != 0) { 3020 struct pktinfo *pi = NULL, pi0; 3021 struct emx_rxbuf *rx_buf = &rdata->rx_buf[i]; 3022 struct mbuf *m = NULL; 3023 int eop, len; 3024 3025 logif(pkt_receive); 3026 3027 mp = rx_buf->m_head; 3028 3029 /* 3030 * Can't defer bus_dmamap_sync(9) because TBI_ACCEPT 3031 * needs to access the last received byte in the mbuf. 3032 */ 3033 bus_dmamap_sync(rdata->rxtag, rx_buf->map, 3034 BUS_DMASYNC_POSTREAD); 3035 3036 len = le16toh(current_desc->rxd_length); 3037 if (staterr & E1000_RXD_STAT_EOP) { 3038 count--; 3039 eop = 1; 3040 } else { 3041 eop = 0; 3042 } 3043 3044 if (!(staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK)) { 3045 uint16_t vlan = 0; 3046 uint32_t mrq, rss_hash; 3047 3048 /* 3049 * Save several necessary information, 3050 * before emx_newbuf() destroy it. 3051 */ 3052 if ((staterr & E1000_RXD_STAT_VP) && eop) 3053 vlan = le16toh(current_desc->rxd_vlan); 3054 3055 mrq = le32toh(current_desc->rxd_mrq); 3056 rss_hash = le32toh(current_desc->rxd_rss); 3057 3058 EMX_RSS_DPRINTF(rdata->sc, 10, 3059 "ring%d, mrq 0x%08x, rss_hash 0x%08x\n", 3060 rdata->idx, mrq, rss_hash); 3061 3062 if (emx_newbuf(rdata, i, 0) != 0) { 3063 IFNET_STAT_INC(ifp, iqdrops, 1); 3064 goto discard; 3065 } 3066 3067 /* Assign correct length to the current fragment */ 3068 mp->m_len = len; 3069 3070 if (rdata->fmp == NULL) { 3071 mp->m_pkthdr.len = len; 3072 rdata->fmp = mp; /* Store the first mbuf */ 3073 rdata->lmp = mp; 3074 } else { 3075 /* 3076 * Chain mbuf's together 3077 */ 3078 rdata->lmp->m_next = mp; 3079 rdata->lmp = rdata->lmp->m_next; 3080 rdata->fmp->m_pkthdr.len += len; 3081 } 3082 3083 if (eop) { 3084 rdata->fmp->m_pkthdr.rcvif = ifp; 3085 IFNET_STAT_INC(ifp, ipackets, 1); 3086 3087 if (ifp->if_capenable & IFCAP_RXCSUM) 3088 emx_rxcsum(staterr, rdata->fmp); 3089 3090 if (staterr & E1000_RXD_STAT_VP) { 3091 rdata->fmp->m_pkthdr.ether_vlantag = 3092 vlan; 3093 rdata->fmp->m_flags |= M_VLANTAG; 3094 } 3095 m = rdata->fmp; 3096 rdata->fmp = NULL; 3097 rdata->lmp = NULL; 3098 3099 if (ifp->if_capenable & IFCAP_RSS) { 3100 pi = emx_rssinfo(m, &pi0, mrq, 3101 rss_hash, staterr); 3102 } 3103 #ifdef EMX_RSS_DEBUG 3104 rdata->rx_pkts++; 3105 #endif 3106 } 3107 } else { 3108 IFNET_STAT_INC(ifp, ierrors, 1); 3109 discard: 3110 emx_setup_rxdesc(current_desc, rx_buf); 3111 if (rdata->fmp != NULL) { 3112 m_freem(rdata->fmp); 3113 rdata->fmp = NULL; 3114 rdata->lmp = NULL; 3115 } 3116 m = NULL; 3117 } 3118 3119 if (m != NULL) 3120 ether_input_pkt(ifp, m, pi); 3121 3122 /* Advance our pointers to the next descriptor. */ 3123 if (++i == rdata->num_rx_desc) 3124 i = 0; 3125 3126 current_desc = &rdata->rx_desc[i]; 3127 staterr = le32toh(current_desc->rxd_staterr); 3128 } 3129 rdata->next_rx_desc_to_check = i; 3130 3131 /* Advance the E1000's Receive Queue "Tail Pointer". */ 3132 if (--i < 0) 3133 i = rdata->num_rx_desc - 1; 3134 E1000_WRITE_REG(&rdata->sc->hw, E1000_RDT(rdata->idx), i); 3135 } 3136 3137 static void 3138 emx_enable_intr(struct emx_softc *sc) 3139 { 3140 uint32_t ims_mask = IMS_ENABLE_MASK; 3141 3142 lwkt_serialize_handler_enable(&sc->main_serialize); 3143 3144 #if 0 3145 if (sc->hw.mac.type == e1000_82574) { 3146 E1000_WRITE_REG(hw, EMX_EIAC, EM_MSIX_MASK); 3147 ims_mask |= EM_MSIX_MASK; 3148 } 3149 #endif 3150 E1000_WRITE_REG(&sc->hw, E1000_IMS, ims_mask); 3151 } 3152 3153 static void 3154 emx_disable_intr(struct emx_softc *sc) 3155 { 3156 if (sc->hw.mac.type == e1000_82574) 3157 E1000_WRITE_REG(&sc->hw, EMX_EIAC, 0); 3158 E1000_WRITE_REG(&sc->hw, E1000_IMC, 0xffffffff); 3159 3160 lwkt_serialize_handler_disable(&sc->main_serialize); 3161 } 3162 3163 /* 3164 * Bit of a misnomer, what this really means is 3165 * to enable OS management of the system... aka 3166 * to disable special hardware management features 3167 */ 3168 static void 3169 emx_get_mgmt(struct emx_softc *sc) 3170 { 3171 /* A shared code workaround */ 3172 if (sc->flags & EMX_FLAG_HAS_MGMT) { 3173 int manc2h = E1000_READ_REG(&sc->hw, E1000_MANC2H); 3174 int manc = E1000_READ_REG(&sc->hw, E1000_MANC); 3175 3176 /* disable hardware interception of ARP */ 3177 manc &= ~(E1000_MANC_ARP_EN); 3178 3179 /* enable receiving management packets to the host */ 3180 manc |= E1000_MANC_EN_MNG2HOST; 3181 #define E1000_MNG2HOST_PORT_623 (1 << 5) 3182 #define E1000_MNG2HOST_PORT_664 (1 << 6) 3183 manc2h |= E1000_MNG2HOST_PORT_623; 3184 manc2h |= E1000_MNG2HOST_PORT_664; 3185 E1000_WRITE_REG(&sc->hw, E1000_MANC2H, manc2h); 3186 3187 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc); 3188 } 3189 } 3190 3191 /* 3192 * Give control back to hardware management 3193 * controller if there is one. 3194 */ 3195 static void 3196 emx_rel_mgmt(struct emx_softc *sc) 3197 { 3198 if (sc->flags & EMX_FLAG_HAS_MGMT) { 3199 int manc = E1000_READ_REG(&sc->hw, E1000_MANC); 3200 3201 /* re-enable hardware interception of ARP */ 3202 manc |= E1000_MANC_ARP_EN; 3203 manc &= ~E1000_MANC_EN_MNG2HOST; 3204 3205 E1000_WRITE_REG(&sc->hw, E1000_MANC, manc); 3206 } 3207 } 3208 3209 /* 3210 * emx_get_hw_control() sets {CTRL_EXT|FWSM}:DRV_LOAD bit. 3211 * For ASF and Pass Through versions of f/w this means that 3212 * the driver is loaded. For AMT version (only with 82573) 3213 * of the f/w this means that the network i/f is open. 3214 */ 3215 static void 3216 emx_get_hw_control(struct emx_softc *sc) 3217 { 3218 /* Let firmware know the driver has taken over */ 3219 if (sc->hw.mac.type == e1000_82573) { 3220 uint32_t swsm; 3221 3222 swsm = E1000_READ_REG(&sc->hw, E1000_SWSM); 3223 E1000_WRITE_REG(&sc->hw, E1000_SWSM, 3224 swsm | E1000_SWSM_DRV_LOAD); 3225 } else { 3226 uint32_t ctrl_ext; 3227 3228 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT); 3229 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, 3230 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD); 3231 } 3232 sc->flags |= EMX_FLAG_HW_CTRL; 3233 } 3234 3235 /* 3236 * emx_rel_hw_control() resets {CTRL_EXT|FWSM}:DRV_LOAD bit. 3237 * For ASF and Pass Through versions of f/w this means that the 3238 * driver is no longer loaded. For AMT version (only with 82573) 3239 * of the f/w this means that the network i/f is closed. 3240 */ 3241 static void 3242 emx_rel_hw_control(struct emx_softc *sc) 3243 { 3244 if ((sc->flags & EMX_FLAG_HW_CTRL) == 0) 3245 return; 3246 sc->flags &= ~EMX_FLAG_HW_CTRL; 3247 3248 /* Let firmware taken over control of h/w */ 3249 if (sc->hw.mac.type == e1000_82573) { 3250 uint32_t swsm; 3251 3252 swsm = E1000_READ_REG(&sc->hw, E1000_SWSM); 3253 E1000_WRITE_REG(&sc->hw, E1000_SWSM, 3254 swsm & ~E1000_SWSM_DRV_LOAD); 3255 } else { 3256 uint32_t ctrl_ext; 3257 3258 ctrl_ext = E1000_READ_REG(&sc->hw, E1000_CTRL_EXT); 3259 E1000_WRITE_REG(&sc->hw, E1000_CTRL_EXT, 3260 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD); 3261 } 3262 } 3263 3264 static int 3265 emx_is_valid_eaddr(const uint8_t *addr) 3266 { 3267 char zero_addr[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 }; 3268 3269 if ((addr[0] & 1) || !bcmp(addr, zero_addr, ETHER_ADDR_LEN)) 3270 return (FALSE); 3271 3272 return (TRUE); 3273 } 3274 3275 /* 3276 * Enable PCI Wake On Lan capability 3277 */ 3278 void 3279 emx_enable_wol(device_t dev) 3280 { 3281 uint16_t cap, status; 3282 uint8_t id; 3283 3284 /* First find the capabilities pointer*/ 3285 cap = pci_read_config(dev, PCIR_CAP_PTR, 2); 3286 3287 /* Read the PM Capabilities */ 3288 id = pci_read_config(dev, cap, 1); 3289 if (id != PCIY_PMG) /* Something wrong */ 3290 return; 3291 3292 /* 3293 * OK, we have the power capabilities, 3294 * so now get the status register 3295 */ 3296 cap += PCIR_POWER_STATUS; 3297 status = pci_read_config(dev, cap, 2); 3298 status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE; 3299 pci_write_config(dev, cap, status, 2); 3300 } 3301 3302 static void 3303 emx_update_stats(struct emx_softc *sc) 3304 { 3305 struct ifnet *ifp = &sc->arpcom.ac_if; 3306 3307 if (sc->hw.phy.media_type == e1000_media_type_copper || 3308 (E1000_READ_REG(&sc->hw, E1000_STATUS) & E1000_STATUS_LU)) { 3309 sc->stats.symerrs += E1000_READ_REG(&sc->hw, E1000_SYMERRS); 3310 sc->stats.sec += E1000_READ_REG(&sc->hw, E1000_SEC); 3311 } 3312 sc->stats.crcerrs += E1000_READ_REG(&sc->hw, E1000_CRCERRS); 3313 sc->stats.mpc += E1000_READ_REG(&sc->hw, E1000_MPC); 3314 sc->stats.scc += E1000_READ_REG(&sc->hw, E1000_SCC); 3315 sc->stats.ecol += E1000_READ_REG(&sc->hw, E1000_ECOL); 3316 3317 sc->stats.mcc += E1000_READ_REG(&sc->hw, E1000_MCC); 3318 sc->stats.latecol += E1000_READ_REG(&sc->hw, E1000_LATECOL); 3319 sc->stats.colc += E1000_READ_REG(&sc->hw, E1000_COLC); 3320 sc->stats.dc += E1000_READ_REG(&sc->hw, E1000_DC); 3321 sc->stats.rlec += E1000_READ_REG(&sc->hw, E1000_RLEC); 3322 sc->stats.xonrxc += E1000_READ_REG(&sc->hw, E1000_XONRXC); 3323 sc->stats.xontxc += E1000_READ_REG(&sc->hw, E1000_XONTXC); 3324 sc->stats.xoffrxc += E1000_READ_REG(&sc->hw, E1000_XOFFRXC); 3325 sc->stats.xofftxc += E1000_READ_REG(&sc->hw, E1000_XOFFTXC); 3326 sc->stats.fcruc += E1000_READ_REG(&sc->hw, E1000_FCRUC); 3327 sc->stats.prc64 += E1000_READ_REG(&sc->hw, E1000_PRC64); 3328 sc->stats.prc127 += E1000_READ_REG(&sc->hw, E1000_PRC127); 3329 sc->stats.prc255 += E1000_READ_REG(&sc->hw, E1000_PRC255); 3330 sc->stats.prc511 += E1000_READ_REG(&sc->hw, E1000_PRC511); 3331 sc->stats.prc1023 += E1000_READ_REG(&sc->hw, E1000_PRC1023); 3332 sc->stats.prc1522 += E1000_READ_REG(&sc->hw, E1000_PRC1522); 3333 sc->stats.gprc += E1000_READ_REG(&sc->hw, E1000_GPRC); 3334 sc->stats.bprc += E1000_READ_REG(&sc->hw, E1000_BPRC); 3335 sc->stats.mprc += E1000_READ_REG(&sc->hw, E1000_MPRC); 3336 sc->stats.gptc += E1000_READ_REG(&sc->hw, E1000_GPTC); 3337 3338 /* For the 64-bit byte counters the low dword must be read first. */ 3339 /* Both registers clear on the read of the high dword */ 3340 3341 sc->stats.gorc += E1000_READ_REG(&sc->hw, E1000_GORCH); 3342 sc->stats.gotc += E1000_READ_REG(&sc->hw, E1000_GOTCH); 3343 3344 sc->stats.rnbc += E1000_READ_REG(&sc->hw, E1000_RNBC); 3345 sc->stats.ruc += E1000_READ_REG(&sc->hw, E1000_RUC); 3346 sc->stats.rfc += E1000_READ_REG(&sc->hw, E1000_RFC); 3347 sc->stats.roc += E1000_READ_REG(&sc->hw, E1000_ROC); 3348 sc->stats.rjc += E1000_READ_REG(&sc->hw, E1000_RJC); 3349 3350 sc->stats.tor += E1000_READ_REG(&sc->hw, E1000_TORH); 3351 sc->stats.tot += E1000_READ_REG(&sc->hw, E1000_TOTH); 3352 3353 sc->stats.tpr += E1000_READ_REG(&sc->hw, E1000_TPR); 3354 sc->stats.tpt += E1000_READ_REG(&sc->hw, E1000_TPT); 3355 sc->stats.ptc64 += E1000_READ_REG(&sc->hw, E1000_PTC64); 3356 sc->stats.ptc127 += E1000_READ_REG(&sc->hw, E1000_PTC127); 3357 sc->stats.ptc255 += E1000_READ_REG(&sc->hw, E1000_PTC255); 3358 sc->stats.ptc511 += E1000_READ_REG(&sc->hw, E1000_PTC511); 3359 sc->stats.ptc1023 += E1000_READ_REG(&sc->hw, E1000_PTC1023); 3360 sc->stats.ptc1522 += E1000_READ_REG(&sc->hw, E1000_PTC1522); 3361 sc->stats.mptc += E1000_READ_REG(&sc->hw, E1000_MPTC); 3362 sc->stats.bptc += E1000_READ_REG(&sc->hw, E1000_BPTC); 3363 3364 sc->stats.algnerrc += E1000_READ_REG(&sc->hw, E1000_ALGNERRC); 3365 sc->stats.rxerrc += E1000_READ_REG(&sc->hw, E1000_RXERRC); 3366 sc->stats.tncrs += E1000_READ_REG(&sc->hw, E1000_TNCRS); 3367 sc->stats.cexterr += E1000_READ_REG(&sc->hw, E1000_CEXTERR); 3368 sc->stats.tsctc += E1000_READ_REG(&sc->hw, E1000_TSCTC); 3369 sc->stats.tsctfc += E1000_READ_REG(&sc->hw, E1000_TSCTFC); 3370 3371 IFNET_STAT_SET(ifp, collisions, sc->stats.colc); 3372 3373 /* Rx Errors */ 3374 IFNET_STAT_SET(ifp, ierrors, 3375 sc->stats.rxerrc + sc->stats.crcerrs + sc->stats.algnerrc + 3376 sc->stats.ruc + sc->stats.roc + sc->stats.mpc + sc->stats.cexterr); 3377 3378 /* Tx Errors */ 3379 IFNET_STAT_SET(ifp, oerrors, sc->stats.ecol + sc->stats.latecol); 3380 } 3381 3382 static void 3383 emx_print_debug_info(struct emx_softc *sc) 3384 { 3385 device_t dev = sc->dev; 3386 uint8_t *hw_addr = sc->hw.hw_addr; 3387 int i; 3388 3389 device_printf(dev, "Adapter hardware address = %p \n", hw_addr); 3390 device_printf(dev, "CTRL = 0x%x RCTL = 0x%x \n", 3391 E1000_READ_REG(&sc->hw, E1000_CTRL), 3392 E1000_READ_REG(&sc->hw, E1000_RCTL)); 3393 device_printf(dev, "Packet buffer = Tx=%dk Rx=%dk \n", 3394 ((E1000_READ_REG(&sc->hw, E1000_PBA) & 0xffff0000) >> 16),\ 3395 (E1000_READ_REG(&sc->hw, E1000_PBA) & 0xffff) ); 3396 device_printf(dev, "Flow control watermarks high = %d low = %d\n", 3397 sc->hw.fc.high_water, sc->hw.fc.low_water); 3398 device_printf(dev, "tx_int_delay = %d, tx_abs_int_delay = %d\n", 3399 E1000_READ_REG(&sc->hw, E1000_TIDV), 3400 E1000_READ_REG(&sc->hw, E1000_TADV)); 3401 device_printf(dev, "rx_int_delay = %d, rx_abs_int_delay = %d\n", 3402 E1000_READ_REG(&sc->hw, E1000_RDTR), 3403 E1000_READ_REG(&sc->hw, E1000_RADV)); 3404 3405 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3406 device_printf(dev, "hw %d tdh = %d, hw tdt = %d\n", i, 3407 E1000_READ_REG(&sc->hw, E1000_TDH(i)), 3408 E1000_READ_REG(&sc->hw, E1000_TDT(i))); 3409 } 3410 for (i = 0; i < sc->rx_ring_cnt; ++i) { 3411 device_printf(dev, "hw %d rdh = %d, hw rdt = %d\n", i, 3412 E1000_READ_REG(&sc->hw, E1000_RDH(i)), 3413 E1000_READ_REG(&sc->hw, E1000_RDT(i))); 3414 } 3415 3416 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3417 device_printf(dev, "TX %d Tx descriptors avail = %d\n", i, 3418 sc->tx_data[i].num_tx_desc_avail); 3419 device_printf(dev, "TX %d TSO segments = %lu\n", i, 3420 sc->tx_data[i].tso_segments); 3421 device_printf(dev, "TX %d TSO ctx reused = %lu\n", i, 3422 sc->tx_data[i].tso_ctx_reused); 3423 } 3424 } 3425 3426 static void 3427 emx_print_hw_stats(struct emx_softc *sc) 3428 { 3429 device_t dev = sc->dev; 3430 3431 device_printf(dev, "Excessive collisions = %lld\n", 3432 (long long)sc->stats.ecol); 3433 #if (DEBUG_HW > 0) /* Dont output these errors normally */ 3434 device_printf(dev, "Symbol errors = %lld\n", 3435 (long long)sc->stats.symerrs); 3436 #endif 3437 device_printf(dev, "Sequence errors = %lld\n", 3438 (long long)sc->stats.sec); 3439 device_printf(dev, "Defer count = %lld\n", 3440 (long long)sc->stats.dc); 3441 device_printf(dev, "Missed Packets = %lld\n", 3442 (long long)sc->stats.mpc); 3443 device_printf(dev, "Receive No Buffers = %lld\n", 3444 (long long)sc->stats.rnbc); 3445 /* RLEC is inaccurate on some hardware, calculate our own. */ 3446 device_printf(dev, "Receive Length Errors = %lld\n", 3447 ((long long)sc->stats.roc + (long long)sc->stats.ruc)); 3448 device_printf(dev, "Receive errors = %lld\n", 3449 (long long)sc->stats.rxerrc); 3450 device_printf(dev, "Crc errors = %lld\n", 3451 (long long)sc->stats.crcerrs); 3452 device_printf(dev, "Alignment errors = %lld\n", 3453 (long long)sc->stats.algnerrc); 3454 device_printf(dev, "Collision/Carrier extension errors = %lld\n", 3455 (long long)sc->stats.cexterr); 3456 device_printf(dev, "RX overruns = %ld\n", sc->rx_overruns); 3457 device_printf(dev, "XON Rcvd = %lld\n", 3458 (long long)sc->stats.xonrxc); 3459 device_printf(dev, "XON Xmtd = %lld\n", 3460 (long long)sc->stats.xontxc); 3461 device_printf(dev, "XOFF Rcvd = %lld\n", 3462 (long long)sc->stats.xoffrxc); 3463 device_printf(dev, "XOFF Xmtd = %lld\n", 3464 (long long)sc->stats.xofftxc); 3465 device_printf(dev, "Good Packets Rcvd = %lld\n", 3466 (long long)sc->stats.gprc); 3467 device_printf(dev, "Good Packets Xmtd = %lld\n", 3468 (long long)sc->stats.gptc); 3469 } 3470 3471 static void 3472 emx_print_nvm_info(struct emx_softc *sc) 3473 { 3474 uint16_t eeprom_data; 3475 int i, j, row = 0; 3476 3477 /* Its a bit crude, but it gets the job done */ 3478 kprintf("\nInterface EEPROM Dump:\n"); 3479 kprintf("Offset\n0x0000 "); 3480 for (i = 0, j = 0; i < 32; i++, j++) { 3481 if (j == 8) { /* Make the offset block */ 3482 j = 0; ++row; 3483 kprintf("\n0x00%x0 ",row); 3484 } 3485 e1000_read_nvm(&sc->hw, i, 1, &eeprom_data); 3486 kprintf("%04x ", eeprom_data); 3487 } 3488 kprintf("\n"); 3489 } 3490 3491 static int 3492 emx_sysctl_debug_info(SYSCTL_HANDLER_ARGS) 3493 { 3494 struct emx_softc *sc; 3495 struct ifnet *ifp; 3496 int error, result; 3497 3498 result = -1; 3499 error = sysctl_handle_int(oidp, &result, 0, req); 3500 if (error || !req->newptr) 3501 return (error); 3502 3503 sc = (struct emx_softc *)arg1; 3504 ifp = &sc->arpcom.ac_if; 3505 3506 ifnet_serialize_all(ifp); 3507 3508 if (result == 1) 3509 emx_print_debug_info(sc); 3510 3511 /* 3512 * This value will cause a hex dump of the 3513 * first 32 16-bit words of the EEPROM to 3514 * the screen. 3515 */ 3516 if (result == 2) 3517 emx_print_nvm_info(sc); 3518 3519 ifnet_deserialize_all(ifp); 3520 3521 return (error); 3522 } 3523 3524 static int 3525 emx_sysctl_stats(SYSCTL_HANDLER_ARGS) 3526 { 3527 int error, result; 3528 3529 result = -1; 3530 error = sysctl_handle_int(oidp, &result, 0, req); 3531 if (error || !req->newptr) 3532 return (error); 3533 3534 if (result == 1) { 3535 struct emx_softc *sc = (struct emx_softc *)arg1; 3536 struct ifnet *ifp = &sc->arpcom.ac_if; 3537 3538 ifnet_serialize_all(ifp); 3539 emx_print_hw_stats(sc); 3540 ifnet_deserialize_all(ifp); 3541 } 3542 return (error); 3543 } 3544 3545 static void 3546 emx_add_sysctl(struct emx_softc *sc) 3547 { 3548 #if defined(EMX_RSS_DEBUG) || defined(EMX_TSS_DEBUG) 3549 char pkt_desc[32]; 3550 int i; 3551 #endif 3552 3553 sysctl_ctx_init(&sc->sysctl_ctx); 3554 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, 3555 SYSCTL_STATIC_CHILDREN(_hw), OID_AUTO, 3556 device_get_nameunit(sc->dev), 3557 CTLFLAG_RD, 0, ""); 3558 if (sc->sysctl_tree == NULL) { 3559 device_printf(sc->dev, "can't add sysctl node\n"); 3560 return; 3561 } 3562 3563 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3564 OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, sc, 0, 3565 emx_sysctl_debug_info, "I", "Debug Information"); 3566 3567 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3568 OID_AUTO, "stats", CTLTYPE_INT|CTLFLAG_RW, sc, 0, 3569 emx_sysctl_stats, "I", "Statistics"); 3570 3571 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3572 OID_AUTO, "rxd", CTLFLAG_RD, &sc->rx_data[0].num_rx_desc, 0, 3573 "# of RX descs"); 3574 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3575 OID_AUTO, "txd", CTLFLAG_RD, &sc->tx_data[0].num_tx_desc, 0, 3576 "# of TX descs"); 3577 3578 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3579 OID_AUTO, "int_throttle_ceil", CTLTYPE_INT|CTLFLAG_RW, sc, 0, 3580 emx_sysctl_int_throttle, "I", "interrupt throttling rate"); 3581 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3582 OID_AUTO, "tx_intr_nsegs", CTLTYPE_INT|CTLFLAG_RW, sc, 0, 3583 emx_sysctl_tx_intr_nsegs, "I", "# segments per TX interrupt"); 3584 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3585 OID_AUTO, "tx_wreg_nsegs", CTLTYPE_INT|CTLFLAG_RW, sc, 0, 3586 emx_sysctl_tx_wreg_nsegs, "I", 3587 "# segments sent before write to hardware register"); 3588 3589 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3590 OID_AUTO, "rx_ring_cnt", CTLFLAG_RD, &sc->rx_ring_cnt, 0, 3591 "# of RX rings"); 3592 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3593 OID_AUTO, "tx_ring_cnt", CTLFLAG_RD, &sc->tx_ring_cnt, 0, 3594 "# of TX rings"); 3595 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3596 OID_AUTO, "tx_ring_inuse", CTLFLAG_RD, &sc->tx_ring_inuse, 0, 3597 "# of TX rings used"); 3598 3599 #ifdef IFPOLL_ENABLE 3600 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3601 OID_AUTO, "npoll_rxoff", CTLTYPE_INT|CTLFLAG_RW, 3602 sc, 0, emx_sysctl_npoll_rxoff, "I", 3603 "NPOLLING RX cpu offset"); 3604 SYSCTL_ADD_PROC(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3605 OID_AUTO, "npoll_txoff", CTLTYPE_INT|CTLFLAG_RW, 3606 sc, 0, emx_sysctl_npoll_txoff, "I", 3607 "NPOLLING TX cpu offset"); 3608 #endif 3609 3610 #ifdef EMX_RSS_DEBUG 3611 SYSCTL_ADD_INT(&sc->sysctl_ctx, SYSCTL_CHILDREN(sc->sysctl_tree), 3612 OID_AUTO, "rss_debug", CTLFLAG_RW, &sc->rss_debug, 3613 0, "RSS debug level"); 3614 for (i = 0; i < sc->rx_ring_cnt; ++i) { 3615 ksnprintf(pkt_desc, sizeof(pkt_desc), "rx%d_pkt", i); 3616 SYSCTL_ADD_ULONG(&sc->sysctl_ctx, 3617 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, 3618 pkt_desc, CTLFLAG_RW, &sc->rx_data[i].rx_pkts, 3619 "RXed packets"); 3620 } 3621 #endif 3622 #ifdef EMX_TSS_DEBUG 3623 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3624 ksnprintf(pkt_desc, sizeof(pkt_desc), "tx%d_pkt", i); 3625 SYSCTL_ADD_ULONG(&sc->sysctl_ctx, 3626 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, 3627 pkt_desc, CTLFLAG_RW, &sc->tx_data[i].tx_pkts, 3628 "TXed packets"); 3629 } 3630 #endif 3631 } 3632 3633 static int 3634 emx_sysctl_int_throttle(SYSCTL_HANDLER_ARGS) 3635 { 3636 struct emx_softc *sc = (void *)arg1; 3637 struct ifnet *ifp = &sc->arpcom.ac_if; 3638 int error, throttle; 3639 3640 throttle = sc->int_throttle_ceil; 3641 error = sysctl_handle_int(oidp, &throttle, 0, req); 3642 if (error || req->newptr == NULL) 3643 return error; 3644 if (throttle < 0 || throttle > 1000000000 / 256) 3645 return EINVAL; 3646 3647 if (throttle) { 3648 /* 3649 * Set the interrupt throttling rate in 256ns increments, 3650 * recalculate sysctl value assignment to get exact frequency. 3651 */ 3652 throttle = 1000000000 / 256 / throttle; 3653 3654 /* Upper 16bits of ITR is reserved and should be zero */ 3655 if (throttle & 0xffff0000) 3656 return EINVAL; 3657 } 3658 3659 ifnet_serialize_all(ifp); 3660 3661 if (throttle) 3662 sc->int_throttle_ceil = 1000000000 / 256 / throttle; 3663 else 3664 sc->int_throttle_ceil = 0; 3665 3666 if (ifp->if_flags & IFF_RUNNING) 3667 emx_set_itr(sc, throttle); 3668 3669 ifnet_deserialize_all(ifp); 3670 3671 if (bootverbose) { 3672 if_printf(ifp, "Interrupt moderation set to %d/sec\n", 3673 sc->int_throttle_ceil); 3674 } 3675 return 0; 3676 } 3677 3678 static int 3679 emx_sysctl_tx_intr_nsegs(SYSCTL_HANDLER_ARGS) 3680 { 3681 struct emx_softc *sc = (void *)arg1; 3682 struct ifnet *ifp = &sc->arpcom.ac_if; 3683 struct emx_txdata *tdata = &sc->tx_data[0]; 3684 int error, segs; 3685 3686 segs = tdata->tx_intr_nsegs; 3687 error = sysctl_handle_int(oidp, &segs, 0, req); 3688 if (error || req->newptr == NULL) 3689 return error; 3690 if (segs <= 0) 3691 return EINVAL; 3692 3693 ifnet_serialize_all(ifp); 3694 3695 /* 3696 * Don't allow tx_intr_nsegs to become: 3697 * o Less the oact_tx_desc 3698 * o Too large that no TX desc will cause TX interrupt to 3699 * be generated (OACTIVE will never recover) 3700 * o Too small that will cause tx_dd[] overflow 3701 */ 3702 if (segs < tdata->oact_tx_desc || 3703 segs >= tdata->num_tx_desc - tdata->oact_tx_desc || 3704 segs < tdata->num_tx_desc / EMX_TXDD_SAFE) { 3705 error = EINVAL; 3706 } else { 3707 int i; 3708 3709 error = 0; 3710 for (i = 0; i < sc->tx_ring_cnt; ++i) 3711 sc->tx_data[i].tx_intr_nsegs = segs; 3712 } 3713 3714 ifnet_deserialize_all(ifp); 3715 3716 return error; 3717 } 3718 3719 static int 3720 emx_sysctl_tx_wreg_nsegs(SYSCTL_HANDLER_ARGS) 3721 { 3722 struct emx_softc *sc = (void *)arg1; 3723 struct ifnet *ifp = &sc->arpcom.ac_if; 3724 int error, nsegs, i; 3725 3726 nsegs = sc->tx_data[0].tx_wreg_nsegs; 3727 error = sysctl_handle_int(oidp, &nsegs, 0, req); 3728 if (error || req->newptr == NULL) 3729 return error; 3730 3731 ifnet_serialize_all(ifp); 3732 for (i = 0; i < sc->tx_ring_cnt; ++i) 3733 sc->tx_data[i].tx_wreg_nsegs =nsegs; 3734 ifnet_deserialize_all(ifp); 3735 3736 return 0; 3737 } 3738 3739 #ifdef IFPOLL_ENABLE 3740 3741 static int 3742 emx_sysctl_npoll_rxoff(SYSCTL_HANDLER_ARGS) 3743 { 3744 struct emx_softc *sc = (void *)arg1; 3745 struct ifnet *ifp = &sc->arpcom.ac_if; 3746 int error, off; 3747 3748 off = sc->rx_npoll_off; 3749 error = sysctl_handle_int(oidp, &off, 0, req); 3750 if (error || req->newptr == NULL) 3751 return error; 3752 if (off < 0) 3753 return EINVAL; 3754 3755 ifnet_serialize_all(ifp); 3756 if (off >= ncpus2 || off % sc->rx_ring_cnt != 0) { 3757 error = EINVAL; 3758 } else { 3759 error = 0; 3760 sc->rx_npoll_off = off; 3761 } 3762 ifnet_deserialize_all(ifp); 3763 3764 return error; 3765 } 3766 3767 static int 3768 emx_sysctl_npoll_txoff(SYSCTL_HANDLER_ARGS) 3769 { 3770 struct emx_softc *sc = (void *)arg1; 3771 struct ifnet *ifp = &sc->arpcom.ac_if; 3772 int error, off; 3773 3774 off = sc->tx_npoll_off; 3775 error = sysctl_handle_int(oidp, &off, 0, req); 3776 if (error || req->newptr == NULL) 3777 return error; 3778 if (off < 0) 3779 return EINVAL; 3780 3781 ifnet_serialize_all(ifp); 3782 if (off >= ncpus2 || off % sc->tx_ring_cnt != 0) { 3783 error = EINVAL; 3784 } else { 3785 error = 0; 3786 sc->tx_npoll_off = off; 3787 } 3788 ifnet_deserialize_all(ifp); 3789 3790 return error; 3791 } 3792 3793 #endif /* IFPOLL_ENABLE */ 3794 3795 static int 3796 emx_dma_alloc(struct emx_softc *sc) 3797 { 3798 int error, i; 3799 3800 /* 3801 * Create top level busdma tag 3802 */ 3803 error = bus_dma_tag_create(NULL, 1, 0, 3804 BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, 3805 NULL, NULL, 3806 BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT, 3807 0, &sc->parent_dtag); 3808 if (error) { 3809 device_printf(sc->dev, "could not create top level DMA tag\n"); 3810 return error; 3811 } 3812 3813 /* 3814 * Allocate transmit descriptors ring and buffers 3815 */ 3816 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3817 error = emx_create_tx_ring(&sc->tx_data[i]); 3818 if (error) { 3819 device_printf(sc->dev, 3820 "Could not setup transmit structures\n"); 3821 return error; 3822 } 3823 } 3824 3825 /* 3826 * Allocate receive descriptors ring and buffers 3827 */ 3828 for (i = 0; i < sc->rx_ring_cnt; ++i) { 3829 error = emx_create_rx_ring(&sc->rx_data[i]); 3830 if (error) { 3831 device_printf(sc->dev, 3832 "Could not setup receive structures\n"); 3833 return error; 3834 } 3835 } 3836 return 0; 3837 } 3838 3839 static void 3840 emx_dma_free(struct emx_softc *sc) 3841 { 3842 int i; 3843 3844 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3845 emx_destroy_tx_ring(&sc->tx_data[i], 3846 sc->tx_data[i].num_tx_desc); 3847 } 3848 3849 for (i = 0; i < sc->rx_ring_cnt; ++i) { 3850 emx_destroy_rx_ring(&sc->rx_data[i], 3851 sc->rx_data[i].num_rx_desc); 3852 } 3853 3854 /* Free top level busdma tag */ 3855 if (sc->parent_dtag != NULL) 3856 bus_dma_tag_destroy(sc->parent_dtag); 3857 } 3858 3859 static void 3860 emx_serialize(struct ifnet *ifp, enum ifnet_serialize slz) 3861 { 3862 struct emx_softc *sc = ifp->if_softc; 3863 3864 ifnet_serialize_array_enter(sc->serializes, EMX_NSERIALIZE, 3865 EMX_TX_SERIALIZE, EMX_RX_SERIALIZE, slz); 3866 } 3867 3868 static void 3869 emx_deserialize(struct ifnet *ifp, enum ifnet_serialize slz) 3870 { 3871 struct emx_softc *sc = ifp->if_softc; 3872 3873 ifnet_serialize_array_exit(sc->serializes, EMX_NSERIALIZE, 3874 EMX_TX_SERIALIZE, EMX_RX_SERIALIZE, slz); 3875 } 3876 3877 static int 3878 emx_tryserialize(struct ifnet *ifp, enum ifnet_serialize slz) 3879 { 3880 struct emx_softc *sc = ifp->if_softc; 3881 3882 return ifnet_serialize_array_try(sc->serializes, EMX_NSERIALIZE, 3883 EMX_TX_SERIALIZE, EMX_RX_SERIALIZE, slz); 3884 } 3885 3886 static void 3887 emx_serialize_skipmain(struct emx_softc *sc) 3888 { 3889 lwkt_serialize_array_enter(sc->serializes, EMX_NSERIALIZE, 1); 3890 } 3891 3892 static void 3893 emx_deserialize_skipmain(struct emx_softc *sc) 3894 { 3895 lwkt_serialize_array_exit(sc->serializes, EMX_NSERIALIZE, 1); 3896 } 3897 3898 #ifdef INVARIANTS 3899 3900 static void 3901 emx_serialize_assert(struct ifnet *ifp, enum ifnet_serialize slz, 3902 boolean_t serialized) 3903 { 3904 struct emx_softc *sc = ifp->if_softc; 3905 3906 ifnet_serialize_array_assert(sc->serializes, EMX_NSERIALIZE, 3907 EMX_TX_SERIALIZE, EMX_RX_SERIALIZE, slz, serialized); 3908 } 3909 3910 #endif /* INVARIANTS */ 3911 3912 #ifdef IFPOLL_ENABLE 3913 3914 static void 3915 emx_npoll_status(struct ifnet *ifp) 3916 { 3917 struct emx_softc *sc = ifp->if_softc; 3918 uint32_t reg_icr; 3919 3920 ASSERT_SERIALIZED(&sc->main_serialize); 3921 3922 reg_icr = E1000_READ_REG(&sc->hw, E1000_ICR); 3923 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) { 3924 callout_stop(&sc->timer); 3925 sc->hw.mac.get_link_status = 1; 3926 emx_update_link_status(sc); 3927 callout_reset(&sc->timer, hz, emx_timer, sc); 3928 } 3929 } 3930 3931 static void 3932 emx_npoll_tx(struct ifnet *ifp, void *arg, int cycle __unused) 3933 { 3934 struct emx_txdata *tdata = arg; 3935 3936 ASSERT_SERIALIZED(&tdata->tx_serialize); 3937 3938 emx_txeof(tdata); 3939 if (!ifsq_is_empty(tdata->ifsq)) 3940 ifsq_devstart(tdata->ifsq); 3941 } 3942 3943 static void 3944 emx_npoll_rx(struct ifnet *ifp __unused, void *arg, int cycle) 3945 { 3946 struct emx_rxdata *rdata = arg; 3947 3948 ASSERT_SERIALIZED(&rdata->rx_serialize); 3949 3950 emx_rxeof(rdata, cycle); 3951 } 3952 3953 static void 3954 emx_npoll(struct ifnet *ifp, struct ifpoll_info *info) 3955 { 3956 struct emx_softc *sc = ifp->if_softc; 3957 int i, txr_cnt; 3958 3959 ASSERT_IFNET_SERIALIZED_ALL(ifp); 3960 3961 if (info) { 3962 int off; 3963 3964 info->ifpi_status.status_func = emx_npoll_status; 3965 info->ifpi_status.serializer = &sc->main_serialize; 3966 3967 txr_cnt = emx_get_txring_inuse(sc, TRUE); 3968 off = sc->tx_npoll_off; 3969 for (i = 0; i < txr_cnt; ++i) { 3970 struct emx_txdata *tdata = &sc->tx_data[i]; 3971 int idx = i + off; 3972 3973 KKASSERT(idx < ncpus2); 3974 info->ifpi_tx[idx].poll_func = emx_npoll_tx; 3975 info->ifpi_tx[idx].arg = tdata; 3976 info->ifpi_tx[idx].serializer = &tdata->tx_serialize; 3977 ifsq_set_cpuid(tdata->ifsq, idx); 3978 } 3979 3980 off = sc->rx_npoll_off; 3981 for (i = 0; i < sc->rx_ring_cnt; ++i) { 3982 struct emx_rxdata *rdata = &sc->rx_data[i]; 3983 int idx = i + off; 3984 3985 KKASSERT(idx < ncpus2); 3986 info->ifpi_rx[idx].poll_func = emx_npoll_rx; 3987 info->ifpi_rx[idx].arg = rdata; 3988 info->ifpi_rx[idx].serializer = &rdata->rx_serialize; 3989 } 3990 3991 if (ifp->if_flags & IFF_RUNNING) { 3992 if (txr_cnt == sc->tx_ring_inuse) 3993 emx_disable_intr(sc); 3994 else 3995 emx_init(sc); 3996 } 3997 } else { 3998 for (i = 0; i < sc->tx_ring_cnt; ++i) { 3999 struct emx_txdata *tdata = &sc->tx_data[i]; 4000 4001 ifsq_set_cpuid(tdata->ifsq, 4002 rman_get_cpuid(sc->intr_res)); 4003 } 4004 4005 if (ifp->if_flags & IFF_RUNNING) { 4006 txr_cnt = emx_get_txring_inuse(sc, FALSE); 4007 if (txr_cnt == sc->tx_ring_inuse) 4008 emx_enable_intr(sc); 4009 else 4010 emx_init(sc); 4011 } 4012 } 4013 } 4014 4015 #endif /* IFPOLL_ENABLE */ 4016 4017 static void 4018 emx_set_itr(struct emx_softc *sc, uint32_t itr) 4019 { 4020 E1000_WRITE_REG(&sc->hw, E1000_ITR, itr); 4021 if (sc->hw.mac.type == e1000_82574) { 4022 int i; 4023 4024 /* 4025 * When using MSIX interrupts we need to 4026 * throttle using the EITR register 4027 */ 4028 for (i = 0; i < 4; ++i) 4029 E1000_WRITE_REG(&sc->hw, E1000_EITR_82574(i), itr); 4030 } 4031 } 4032 4033 /* 4034 * Disable the L0s, 82574L Errata #20 4035 */ 4036 static void 4037 emx_disable_aspm(struct emx_softc *sc) 4038 { 4039 uint16_t link_cap, link_ctrl, disable; 4040 uint8_t pcie_ptr, reg; 4041 device_t dev = sc->dev; 4042 4043 switch (sc->hw.mac.type) { 4044 case e1000_82571: 4045 case e1000_82572: 4046 case e1000_82573: 4047 /* 4048 * 82573 specification update 4049 * errata #8 disable L0s 4050 * errata #41 disable L1 4051 * 4052 * 82571/82572 specification update 4053 # errata #13 disable L1 4054 * errata #68 disable L0s 4055 */ 4056 disable = PCIEM_LNKCTL_ASPM_L0S | PCIEM_LNKCTL_ASPM_L1; 4057 break; 4058 4059 case e1000_82574: 4060 /* 4061 * 82574 specification update errata #20 4062 * 4063 * There is no need to disable L1 4064 */ 4065 disable = PCIEM_LNKCTL_ASPM_L0S; 4066 break; 4067 4068 default: 4069 return; 4070 } 4071 4072 pcie_ptr = pci_get_pciecap_ptr(dev); 4073 if (pcie_ptr == 0) 4074 return; 4075 4076 link_cap = pci_read_config(dev, pcie_ptr + PCIER_LINKCAP, 2); 4077 if ((link_cap & PCIEM_LNKCAP_ASPM_MASK) == 0) 4078 return; 4079 4080 if (bootverbose) 4081 if_printf(&sc->arpcom.ac_if, "disable ASPM %#02x\n", disable); 4082 4083 reg = pcie_ptr + PCIER_LINKCTRL; 4084 link_ctrl = pci_read_config(dev, reg, 2); 4085 link_ctrl &= ~disable; 4086 pci_write_config(dev, reg, link_ctrl, 2); 4087 } 4088 4089 static int 4090 emx_tso_pullup(struct emx_txdata *tdata, struct mbuf **mp) 4091 { 4092 int iphlen, hoff, thoff, ex = 0; 4093 struct mbuf *m; 4094 struct ip *ip; 4095 4096 m = *mp; 4097 KASSERT(M_WRITABLE(m), ("TSO mbuf not writable")); 4098 4099 iphlen = m->m_pkthdr.csum_iphlen; 4100 thoff = m->m_pkthdr.csum_thlen; 4101 hoff = m->m_pkthdr.csum_lhlen; 4102 4103 KASSERT(iphlen > 0, ("invalid ip hlen")); 4104 KASSERT(thoff > 0, ("invalid tcp hlen")); 4105 KASSERT(hoff > 0, ("invalid ether hlen")); 4106 4107 if (tdata->tx_flags & EMX_TXFLAG_TSO_PULLEX) 4108 ex = 4; 4109 4110 if (m->m_len < hoff + iphlen + thoff + ex) { 4111 m = m_pullup(m, hoff + iphlen + thoff + ex); 4112 if (m == NULL) { 4113 *mp = NULL; 4114 return ENOBUFS; 4115 } 4116 *mp = m; 4117 } 4118 ip = mtodoff(m, struct ip *, hoff); 4119 ip->ip_len = 0; 4120 4121 return 0; 4122 } 4123 4124 static int 4125 emx_tso_setup(struct emx_txdata *tdata, struct mbuf *mp, 4126 uint32_t *txd_upper, uint32_t *txd_lower) 4127 { 4128 struct e1000_context_desc *TXD; 4129 int hoff, iphlen, thoff, hlen; 4130 int mss, pktlen, curr_txd; 4131 4132 #ifdef EMX_TSO_DEBUG 4133 tdata->tso_segments++; 4134 #endif 4135 4136 iphlen = mp->m_pkthdr.csum_iphlen; 4137 thoff = mp->m_pkthdr.csum_thlen; 4138 hoff = mp->m_pkthdr.csum_lhlen; 4139 mss = mp->m_pkthdr.tso_segsz; 4140 pktlen = mp->m_pkthdr.len; 4141 4142 if ((tdata->tx_flags & EMX_TXFLAG_FORCECTX) == 0 && 4143 tdata->csum_flags == CSUM_TSO && 4144 tdata->csum_iphlen == iphlen && 4145 tdata->csum_lhlen == hoff && 4146 tdata->csum_thlen == thoff && 4147 tdata->csum_mss == mss && 4148 tdata->csum_pktlen == pktlen) { 4149 *txd_upper = tdata->csum_txd_upper; 4150 *txd_lower = tdata->csum_txd_lower; 4151 #ifdef EMX_TSO_DEBUG 4152 tdata->tso_ctx_reused++; 4153 #endif 4154 return 0; 4155 } 4156 hlen = hoff + iphlen + thoff; 4157 4158 /* 4159 * Setup a new TSO context. 4160 */ 4161 4162 curr_txd = tdata->next_avail_tx_desc; 4163 TXD = (struct e1000_context_desc *)&tdata->tx_desc_base[curr_txd]; 4164 4165 *txd_lower = E1000_TXD_CMD_DEXT | /* Extended descr type */ 4166 E1000_TXD_DTYP_D | /* Data descr type */ 4167 E1000_TXD_CMD_TSE; /* Do TSE on this packet */ 4168 4169 /* IP and/or TCP header checksum calculation and insertion. */ 4170 *txd_upper = (E1000_TXD_POPTS_IXSM | E1000_TXD_POPTS_TXSM) << 8; 4171 4172 /* 4173 * Start offset for header checksum calculation. 4174 * End offset for header checksum calculation. 4175 * Offset of place put the checksum. 4176 */ 4177 TXD->lower_setup.ip_fields.ipcss = hoff; 4178 TXD->lower_setup.ip_fields.ipcse = htole16(hoff + iphlen - 1); 4179 TXD->lower_setup.ip_fields.ipcso = hoff + offsetof(struct ip, ip_sum); 4180 4181 /* 4182 * Start offset for payload checksum calculation. 4183 * End offset for payload checksum calculation. 4184 * Offset of place to put the checksum. 4185 */ 4186 TXD->upper_setup.tcp_fields.tucss = hoff + iphlen; 4187 TXD->upper_setup.tcp_fields.tucse = 0; 4188 TXD->upper_setup.tcp_fields.tucso = 4189 hoff + iphlen + offsetof(struct tcphdr, th_sum); 4190 4191 /* 4192 * Payload size per packet w/o any headers. 4193 * Length of all headers up to payload. 4194 */ 4195 TXD->tcp_seg_setup.fields.mss = htole16(mss); 4196 TXD->tcp_seg_setup.fields.hdr_len = hlen; 4197 TXD->cmd_and_length = htole32(E1000_TXD_CMD_IFCS | 4198 E1000_TXD_CMD_DEXT | /* Extended descr */ 4199 E1000_TXD_CMD_TSE | /* TSE context */ 4200 E1000_TXD_CMD_IP | /* Do IP csum */ 4201 E1000_TXD_CMD_TCP | /* Do TCP checksum */ 4202 (pktlen - hlen)); /* Total len */ 4203 4204 /* Save the information for this TSO context */ 4205 tdata->csum_flags = CSUM_TSO; 4206 tdata->csum_lhlen = hoff; 4207 tdata->csum_iphlen = iphlen; 4208 tdata->csum_thlen = thoff; 4209 tdata->csum_mss = mss; 4210 tdata->csum_pktlen = pktlen; 4211 tdata->csum_txd_upper = *txd_upper; 4212 tdata->csum_txd_lower = *txd_lower; 4213 4214 if (++curr_txd == tdata->num_tx_desc) 4215 curr_txd = 0; 4216 4217 KKASSERT(tdata->num_tx_desc_avail > 0); 4218 tdata->num_tx_desc_avail--; 4219 4220 tdata->next_avail_tx_desc = curr_txd; 4221 return 1; 4222 } 4223 4224 static int 4225 emx_get_txring_inuse(const struct emx_softc *sc, boolean_t polling) 4226 { 4227 if (polling) 4228 return sc->tx_ring_cnt; 4229 else 4230 return 1; 4231 } 4232