sys/dev/netif/em/if_em.c

   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  * $DragonFly: src/sys/dev/netif/em/if_em.c,v 1.80 2008/09/17 08:51:29 sephe Exp $
  67  */
  68 /*
  69  * SERIALIZATION API RULES:
  70  *
  71  * - If the driver uses the same serializer for the interrupt as for the
  72  *   ifnet, most of the serialization will be done automatically for the
  73  *   driver.
  74  *
  75  * - ifmedia entry points will be serialized by the ifmedia code using the
  76  *   ifnet serializer.
  77  *
  78  * - if_* entry points except for if_input will be serialized by the IF
  79  *   and protocol layers.
  80  *
  81  * - The device driver must be sure to serialize access from timeout code
  82  *   installed by the device driver.
  83  *
  84  * - The device driver typically holds the serializer at the time it wishes
  85  *   to call if_input.
  86  *
  87  * - We must call lwkt_serialize_handler_enable() prior to enabling the
  88  *   hardware interrupt and lwkt_serialize_handler_disable() after disabling
  89  *   the hardware interrupt in order to avoid handler execution races from
  90  *   scheduled interrupt threads.
  91  *
  92  *   NOTE!  Since callers into the device driver hold the ifnet serializer,
  93  *   the device driver may be holding a serializer at the time it calls
  94  *   if_input even if it is not serializer-aware.
  95  */
  96
  97 #include "opt_polling.h"
  98 #include "opt_serializer.h"
  99
 100 #include <sys/param.h>
 101 #include <sys/bus.h>
 102 #include <sys/endian.h>
 103 #include <sys/interrupt.h>
 104 #include <sys/kernel.h>
 105 #include <sys/ktr.h>
 106 #include <sys/malloc.h>
 107 #include <sys/mbuf.h>
 108 #include <sys/proc.h>
 109 #include <sys/rman.h>
 110 #include <sys/serialize.h>
 111 #include <sys/socket.h>
 112 #include <sys/sockio.h>
 113 #include <sys/sysctl.h>
 114 #include <sys/systm.h>
 115
 116 #include <net/bpf.h>
 117 #include <net/ethernet.h>
 118 #include <net/if.h>
 119 #include <net/if_arp.h>
 120 #include <net/if_dl.h>
 121 #include <net/if_media.h>
 122 #include <net/ifq_var.h>
 123 #include <net/vlan/if_vlan_var.h>
 124 #include <net/vlan/if_vlan_ether.h>
 125
 126 #include <netinet/in_systm.h>
 127 #include <netinet/in.h>
 128 #include <netinet/ip.h>
 129 #include <netinet/tcp.h>
 130 #include <netinet/udp.h>
 131
 132 #include <bus/pci/pcivar.h>
 133 #include <bus/pci/pcireg.h>
 134
 135 #include <dev/netif/ig_hal/e1000_api.h>
 136 #include <dev/netif/ig_hal/e1000_82571.h>
 137 #include <dev/netif/em/if_em.h>
 138
 139 #define EM_NAME "Intel(R) PRO/1000 Network Connection "
 140 #define EM_VER  " 6.9.6"
 141
 142 #define EM_DEVICE(id)   \
 143         { EM_VENDOR_ID, E1000_DEV_ID_##id, EM_NAME #id EM_VER }
 144 #define EM_DEVICE_NULL  { 0, 0, NULL }
 145
 146 static const struct em_vendor_info em_vendor_info_array[] = {
 147         EM_DEVICE(82540EM),
 148         EM_DEVICE(82540EM_LOM),
 149         EM_DEVICE(82540EP),
 150         EM_DEVICE(82540EP_LOM),
 151         EM_DEVICE(82540EP_LP),
 152
 153         EM_DEVICE(82541EI),
 154         EM_DEVICE(82541ER),
 155         EM_DEVICE(82541ER_LOM),
 156         EM_DEVICE(82541EI_MOBILE),
 157         EM_DEVICE(82541GI),
 158         EM_DEVICE(82541GI_LF),
 159         EM_DEVICE(82541GI_MOBILE),
 160
 161         EM_DEVICE(82542),
 162
 163         EM_DEVICE(82543GC_FIBER),
 164         EM_DEVICE(82543GC_COPPER),
 165
 166         EM_DEVICE(82544EI_COPPER),
 167         EM_DEVICE(82544EI_FIBER),
 168         EM_DEVICE(82544GC_COPPER),
 169         EM_DEVICE(82544GC_LOM),
 170
 171         EM_DEVICE(82545EM_COPPER),
 172         EM_DEVICE(82545EM_FIBER),
 173         EM_DEVICE(82545GM_COPPER),
 174         EM_DEVICE(82545GM_FIBER),
 175         EM_DEVICE(82545GM_SERDES),
 176
 177         EM_DEVICE(82546EB_COPPER),
 178         EM_DEVICE(82546EB_FIBER),
 179         EM_DEVICE(82546EB_QUAD_COPPER),
 180         EM_DEVICE(82546GB_COPPER),
 181         EM_DEVICE(82546GB_FIBER),
 182         EM_DEVICE(82546GB_SERDES),
 183         EM_DEVICE(82546GB_PCIE),
 184         EM_DEVICE(82546GB_QUAD_COPPER),
 185         EM_DEVICE(82546GB_QUAD_COPPER_KSP3),
 186
 187         EM_DEVICE(82547EI),
 188         EM_DEVICE(82547EI_MOBILE),
 189         EM_DEVICE(82547GI),
 190
 191         EM_DEVICE(82571EB_COPPER),
 192         EM_DEVICE(82571EB_FIBER),
 193         EM_DEVICE(82571EB_SERDES),
 194         EM_DEVICE(82571EB_SERDES_DUAL),
 195         EM_DEVICE(82571EB_SERDES_QUAD),
 196         EM_DEVICE(82571EB_QUAD_COPPER),
 197         EM_DEVICE(82571EB_QUAD_COPPER_LP),
 198         EM_DEVICE(82571EB_QUAD_FIBER),
 199         EM_DEVICE(82571PT_QUAD_COPPER),
 200
 201         EM_DEVICE(82572EI_COPPER),
 202         EM_DEVICE(82572EI_FIBER),
 203         EM_DEVICE(82572EI_SERDES),
 204         EM_DEVICE(82572EI),
 205
 206         EM_DEVICE(82573E),
 207         EM_DEVICE(82573E_IAMT),
 208         EM_DEVICE(82573L),
 209
 210         EM_DEVICE(80003ES2LAN_COPPER_SPT),
 211         EM_DEVICE(80003ES2LAN_SERDES_SPT),
 212         EM_DEVICE(80003ES2LAN_COPPER_DPT),
 213         EM_DEVICE(80003ES2LAN_SERDES_DPT),
 214
 215         EM_DEVICE(ICH8_IGP_M_AMT),
 216         EM_DEVICE(ICH8_IGP_AMT),
 217         EM_DEVICE(ICH8_IGP_C),
 218         EM_DEVICE(ICH8_IFE),
 219         EM_DEVICE(ICH8_IFE_GT),
 220         EM_DEVICE(ICH8_IFE_G),
 221         EM_DEVICE(ICH8_IGP_M),
 222
 223         EM_DEVICE(ICH9_IGP_M_AMT),
 224         EM_DEVICE(ICH9_IGP_AMT),
 225         EM_DEVICE(ICH9_IGP_C),
 226         EM_DEVICE(ICH9_IGP_M),
 227         EM_DEVICE(ICH9_IGP_M_V),
 228         EM_DEVICE(ICH9_IFE),
 229         EM_DEVICE(ICH9_IFE_GT),
 230         EM_DEVICE(ICH9_IFE_G),
 231         EM_DEVICE(ICH9_BM),
 232
 233         EM_DEVICE(82574L),
 234
 235         EM_DEVICE(ICH10_R_BM_LM),
 236         EM_DEVICE(ICH10_R_BM_LF),
 237         EM_DEVICE(ICH10_R_BM_V),
 238         EM_DEVICE(ICH10_D_BM_LM),
 239         EM_DEVICE(ICH10_D_BM_LF),
 240
 241         /* required last entry */
 242         EM_DEVICE_NULL
 243 };
 244
 245 static int      em_probe(device_t);
 246 static int      em_attach(device_t);
 247 static int      em_detach(device_t);
 248 static int      em_shutdown(device_t);
 249 static int      em_suspend(device_t);
 250 static int      em_resume(device_t);
 251
 252 static void     em_init(void *);
 253 static void     em_stop(struct adapter *);
 254 static int      em_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
 255 static void     em_start(struct ifnet *);
 256 #ifdef DEVICE_POLLING
 257 static void     em_poll(struct ifnet *, enum poll_cmd, int);
 258 #endif
 259 static void     em_watchdog(struct ifnet *);
 260 static void     em_media_status(struct ifnet *, struct ifmediareq *);
 261 static int      em_media_change(struct ifnet *);
 262 static void     em_timer(void *);
 263
 264 static void     em_intr(void *);
 265 static void     em_rxeof(struct adapter *, int);
 266 static void     em_txeof(struct adapter *);
 267 static void     em_tx_collect(struct adapter *);
 268 static void     em_tx_purge(struct adapter *);
 269 static void     em_enable_intr(struct adapter *);
 270 static void     em_disable_intr(struct adapter *);
 271
 272 static int      em_dma_malloc(struct adapter *, bus_size_t,
 273                     struct em_dma_alloc *);
 274 static void     em_dma_free(struct adapter *, struct em_dma_alloc *);
 275 static void     em_init_tx_ring(struct adapter *);
 276 static int      em_init_rx_ring(struct adapter *);
 277 static int      em_create_tx_ring(struct adapter *);
 278 static int      em_create_rx_ring(struct adapter *);
 279 static void     em_destroy_tx_ring(struct adapter *, int);
 280 static void     em_destroy_rx_ring(struct adapter *, int);
 281 static int      em_newbuf(struct adapter *, int, int);
 282 static int      em_encap(struct adapter *, struct mbuf **);
 283 static void     em_rxcsum(struct adapter *, struct e1000_rx_desc *,
 284                     struct mbuf *);
 285 static int      em_txcsum_pullup(struct adapter *, struct mbuf **);
 286 static int      em_txcsum(struct adapter *, struct mbuf *,
 287                     uint32_t *, uint32_t *);
 288
 289 static int      em_get_hw_info(struct adapter *);
 290 static int      em_is_valid_eaddr(const uint8_t *);
 291 static int      em_alloc_pci_res(struct adapter *);
 292 static void     em_free_pci_res(struct adapter *);
 293 static int      em_hw_init(struct adapter *);
 294 static void     em_setup_ifp(struct adapter *);
 295 static void     em_init_tx_unit(struct adapter *);
 296 static void     em_init_rx_unit(struct adapter *);
 297 static void     em_update_stats(struct adapter *);
 298 static void     em_set_promisc(struct adapter *);
 299 static void     em_disable_promisc(struct adapter *);
 300 static void     em_set_multi(struct adapter *);
 301 static void     em_update_link_status(struct adapter *);
 302 static void     em_smartspeed(struct adapter *);
 303
 304 /* Hardware workarounds */
 305 static int      em_82547_fifo_workaround(struct adapter *, int);
 306 static void     em_82547_update_fifo_head(struct adapter *, int);
 307 static int      em_82547_tx_fifo_reset(struct adapter *);
 308 static void     em_82547_move_tail(void *);
 309 static void     em_82547_move_tail_serialized(struct adapter *);
 310 static uint32_t em_82544_fill_desc(bus_addr_t, uint32_t, PDESC_ARRAY);
 311
 312 static void     em_print_debug_info(struct adapter *);
 313 static void     em_print_nvm_info(struct adapter *);
 314 static void     em_print_hw_stats(struct adapter *);
 315
 316 static int      em_sysctl_stats(SYSCTL_HANDLER_ARGS);
 317 static int      em_sysctl_debug_info(SYSCTL_HANDLER_ARGS);
 318 static int      em_sysctl_int_throttle(SYSCTL_HANDLER_ARGS);
 319 static int      em_sysctl_int_tx_nsegs(SYSCTL_HANDLER_ARGS);
 320 static void     em_add_sysctl(struct adapter *adapter);
 321
 322 /* Management and WOL Support */
 323 static void     em_get_mgmt(struct adapter *);
 324 static void     em_rel_mgmt(struct adapter *);
 325 static void     em_get_hw_control(struct adapter *);
 326 static void     em_rel_hw_control(struct adapter *);
 327 static void     em_enable_wol(device_t);
 328
 329 static device_method_t em_methods[] = {
 330         /* Device interface */
 331         DEVMETHOD(device_probe,         em_probe),
 332         DEVMETHOD(device_attach,        em_attach),
 333         DEVMETHOD(device_detach,        em_detach),
 334         DEVMETHOD(device_shutdown,      em_shutdown),
 335         DEVMETHOD(device_suspend,       em_suspend),
 336         DEVMETHOD(device_resume,        em_resume),
 337         { 0, 0 }
 338 };
 339
 340 static driver_t em_driver = {
 341         "em",
 342         em_methods,
 343         sizeof(struct adapter),
 344 };
 345
 346 static devclass_t em_devclass;
 347
 348 DECLARE_DUMMY_MODULE(if_em);
 349 MODULE_DEPEND(em, ig_hal, 1, 1, 1);
 350 DRIVER_MODULE(if_em, pci, em_driver, em_devclass, 0, 0);
 351
 352 /*
 353  * Tunables
 354  */
 355 static int      em_int_throttle_ceil = EM_DEFAULT_ITR;
 356 static int      em_rxd = EM_DEFAULT_RXD;
 357 static int      em_txd = EM_DEFAULT_TXD;
 358 static int      em_smart_pwr_down = FALSE;
 359
 360 /* Controls whether promiscuous also shows bad packets */
 361 static int      em_debug_sbp = FALSE;
 362
 363 static int      em_82573_workaround = TRUE;
 364
 365 TUNABLE_INT("hw.em.int_throttle_ceil", &em_int_throttle_ceil);
 366 TUNABLE_INT("hw.em.rxd", &em_rxd);
 367 TUNABLE_INT("hw.em.txd", &em_txd);
 368 TUNABLE_INT("hw.em.smart_pwr_down", &em_smart_pwr_down);
 369 TUNABLE_INT("hw.em.sbp", &em_debug_sbp);
 370 TUNABLE_INT("hw.em.82573_workaround", &em_82573_workaround);
 371
 372 /* Global used in WOL setup with multiport cards */
 373 static int      em_global_quad_port_a = 0;
 374
 375 /* Set this to one to display debug statistics */
 376 static int      em_display_debug_stats = 0;
 377
 378 #if !defined(KTR_IF_EM)
 379 #define KTR_IF_EM       KTR_ALL
 380 #endif
 381 KTR_INFO_MASTER(if_em);
 382 KTR_INFO(KTR_IF_EM, if_em, intr_beg, 0, "intr begin", 0);
 383 KTR_INFO(KTR_IF_EM, if_em, intr_end, 1, "intr end", 0);
 384 KTR_INFO(KTR_IF_EM, if_em, pkt_receive, 4, "rx packet", 0);
 385 KTR_INFO(KTR_IF_EM, if_em, pkt_txqueue, 5, "tx packet", 0);
 386 KTR_INFO(KTR_IF_EM, if_em, pkt_txclean, 6, "tx clean", 0);
 387 #define logif(name)     KTR_LOG(if_em_ ## name)
 388
 389 static int
 390 em_probe(device_t dev)
 391 {
 392         const struct em_vendor_info *ent;
 393         uint16_t vid, did;
 394
 395         vid = pci_get_vendor(dev);
 396         did = pci_get_device(dev);
 397
 398         for (ent = em_vendor_info_array; ent->desc != NULL; ++ent) {
 399                 if (vid == ent->vendor_id && did == ent->device_id) {
 400                         device_set_desc(dev, ent->desc);
 401                         device_set_async_attach(dev, TRUE);
 402                         return (0);
 403                 }
 404         }
 405         return (ENXIO);
 406 }
 407
 408 static int
 409 em_attach(device_t dev)
 410 {
 411         struct adapter *adapter = device_get_softc(dev);
 412         struct ifnet *ifp = &adapter->arpcom.ac_if;
 413         int tsize, rsize;
 414         int error = 0;
 415         uint16_t eeprom_data, device_id;
 416
 417         adapter->dev = adapter->osdep.dev = dev;
 418
 419         callout_init(&adapter->timer);
 420         callout_init(&adapter->tx_fifo_timer);
 421
 422         /* Determine hardware and mac info */
 423         error = em_get_hw_info(adapter);
 424         if (error) {
 425                 device_printf(dev, "Identify hardware failed\n");
 426                 goto fail;
 427         }
 428
 429         /* Setup PCI resources */
 430         error = em_alloc_pci_res(adapter);
 431         if (error) {
 432                 device_printf(dev, "Allocation of PCI resources failed\n");
 433                 goto fail;
 434         }
 435
 436         /*
 437          * For ICH8 and family we need to map the flash memory,
 438          * and this must happen after the MAC is identified.
 439          */
 440         if (adapter->hw.mac.type == e1000_ich8lan ||
 441             adapter->hw.mac.type == e1000_ich10lan ||
 442             adapter->hw.mac.type == e1000_ich9lan) {
 443                 adapter->flash_rid = EM_BAR_FLASH;
 444
 445                 adapter->flash = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
 446                                         &adapter->flash_rid, RF_ACTIVE);
 447                 if (adapter->flash == NULL) {
 448                         device_printf(dev, "Mapping of Flash failed\n");
 449                         error = ENXIO;
 450                         goto fail;
 451                 }
 452                 adapter->osdep.flash_bus_space_tag =
 453                     rman_get_bustag(adapter->flash);
 454                 adapter->osdep.flash_bus_space_handle =
 455                     rman_get_bushandle(adapter->flash);
 456
 457                 /*
 458                  * This is used in the shared code
 459                  * XXX this goof is actually not used.
 460                  */
 461                 adapter->hw.flash_address = (uint8_t *)adapter->flash;
 462         }
 463
 464         /* Do Shared Code initialization */
 465         if (e1000_setup_init_funcs(&adapter->hw, TRUE)) {
 466                 device_printf(dev, "Setup of Shared code failed\n");
 467                 error = ENXIO;
 468                 goto fail;
 469         }
 470
 471         e1000_get_bus_info(&adapter->hw);
 472
 473         /*
 474          * Validate number of transmit and receive descriptors.  It
 475          * must not exceed hardware maximum, and must be multiple
 476          * of E1000_DBA_ALIGN.
 477          */
 478         if ((em_txd * sizeof(struct e1000_tx_desc)) % EM_DBA_ALIGN != 0 ||
 479             (adapter->hw.mac.type >= e1000_82544 && em_txd > EM_MAX_TXD) ||
 480             (adapter->hw.mac.type < e1000_82544 && em_txd > EM_MAX_TXD_82543) ||
 481             em_txd < EM_MIN_TXD) {
 482                 device_printf(dev, "Using %d TX descriptors instead of %d!\n",
 483                     EM_DEFAULT_TXD, em_txd);
 484                 adapter->num_tx_desc = EM_DEFAULT_TXD;
 485         } else {
 486                 adapter->num_tx_desc = em_txd;
 487         }
 488         if ((em_rxd * sizeof(struct e1000_rx_desc)) % EM_DBA_ALIGN != 0 ||
 489             (adapter->hw.mac.type >= e1000_82544 && em_rxd > EM_MAX_RXD) ||
 490             (adapter->hw.mac.type < e1000_82544 && em_rxd > EM_MAX_RXD_82543) ||
 491             em_rxd < EM_MIN_RXD) {
 492                 device_printf(dev, "Using %d RX descriptors instead of %d!\n",
 493                     EM_DEFAULT_RXD, em_rxd);
 494                 adapter->num_rx_desc = EM_DEFAULT_RXD;
 495         } else {
 496                 adapter->num_rx_desc = em_rxd;
 497         }
 498
 499         adapter->hw.mac.autoneg = DO_AUTO_NEG;
 500         adapter->hw.phy.autoneg_wait_to_complete = FALSE;
 501         adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
 502         adapter->rx_buffer_len = MCLBYTES;
 503
 504         /*
 505          * Interrupt throttle rate
 506          */
 507         if (em_int_throttle_ceil == 0) {
 508                 adapter->int_throttle_ceil = 0;
 509         } else {
 510                 int throttle = em_int_throttle_ceil;
 511
 512                 if (throttle < 0)
 513                         throttle = EM_DEFAULT_ITR;
 514
 515                 /* Recalculate the tunable value to get the exact frequency. */
 516                 throttle = 1000000000 / 256 / throttle;
 517
 518                 /* Upper 16bits of ITR is reserved and should be zero */
 519                 if (throttle & 0xffff0000)
 520                         throttle = 1000000000 / 256 / EM_DEFAULT_ITR;
 521
 522                 adapter->int_throttle_ceil = 1000000000 / 256 / throttle;
 523         }
 524
 525         e1000_init_script_state_82541(&adapter->hw, TRUE);
 526         e1000_set_tbi_compatibility_82543(&adapter->hw, TRUE);
 527
 528         /* Copper options */
 529         if (adapter->hw.phy.media_type == e1000_media_type_copper) {
 530                 adapter->hw.phy.mdix = AUTO_ALL_MODES;
 531                 adapter->hw.phy.disable_polarity_correction = FALSE;
 532                 adapter->hw.phy.ms_type = EM_MASTER_SLAVE;
 533         }
 534
 535         /* Set the frame limits assuming standard ethernet sized frames. */
 536         adapter->max_frame_size = ETHERMTU + ETHER_HDR_LEN + ETHER_CRC_LEN;
 537         adapter->min_frame_size = ETH_ZLEN + ETHER_CRC_LEN;
 538
 539         /* This controls when hardware reports transmit completion status. */
 540         adapter->hw.mac.report_tx_early = 1;
 541
 542         /*
 543          * Create top level busdma tag
 544          */
 545         error = bus_dma_tag_create(NULL, 1, 0,
 546                         BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR,
 547                         NULL, NULL,
 548                         BUS_SPACE_MAXSIZE_32BIT, 0, BUS_SPACE_MAXSIZE_32BIT,
 549                         0, &adapter->parent_dtag);
 550         if (error) {
 551                 device_printf(dev, "could not create top level DMA tag\n");
 552                 goto fail;
 553         }
 554
 555         /*
 556          * Allocate Transmit Descriptor ring
 557          */
 558         tsize = roundup2(adapter->num_tx_desc * sizeof(struct e1000_tx_desc),
 559                          EM_DBA_ALIGN);
 560         error = em_dma_malloc(adapter, tsize, &adapter->txdma);
 561         if (error) {
 562                 device_printf(dev, "Unable to allocate tx_desc memory\n");
 563                 goto fail;
 564         }
 565         adapter->tx_desc_base = adapter->txdma.dma_vaddr;
 566
 567         /*
 568          * Allocate Receive Descriptor ring
 569          */
 570         rsize = roundup2(adapter->num_rx_desc * sizeof(struct e1000_rx_desc),
 571                          EM_DBA_ALIGN);
 572         error = em_dma_malloc(adapter, rsize, &adapter->rxdma);
 573         if (error) {
 574                 device_printf(dev, "Unable to allocate rx_desc memory\n");
 575                 goto fail;
 576         }
 577         adapter->rx_desc_base = adapter->rxdma.dma_vaddr;
 578
 579         /* Make sure we have a good EEPROM before we read from it */
 580         if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
 581                 /*
 582                  * Some PCI-E parts fail the first check due to
 583                  * the link being in sleep state, call it again,
 584                  * if it fails a second time its a real issue.
 585                  */
 586                 if (e1000_validate_nvm_checksum(&adapter->hw) < 0) {
 587                         device_printf(dev,
 588                             "The EEPROM Checksum Is Not Valid\n");
 589                         error = EIO;
 590                         goto fail;
 591                 }
 592         }
 593
 594         /* Initialize the hardware */
 595         error = em_hw_init(adapter);
 596         if (error) {
 597                 device_printf(dev, "Unable to initialize the hardware\n");
 598                 goto fail;
 599         }
 600
 601         /* Copy the permanent MAC address out of the EEPROM */
 602         if (e1000_read_mac_addr(&adapter->hw) < 0) {
 603                 device_printf(dev, "EEPROM read error while reading MAC"
 604                     " address\n");
 605                 error = EIO;
 606                 goto fail;
 607         }
 608         if (!em_is_valid_eaddr(adapter->hw.mac.addr)) {
 609                 device_printf(dev, "Invalid MAC address\n");
 610                 error = EIO;
 611                 goto fail;
 612         }
 613
 614         /* Allocate transmit descriptors and buffers */
 615         error = em_create_tx_ring(adapter);
 616         if (error) {
 617                 device_printf(dev, "Could not setup transmit structures\n");
 618                 goto fail;
 619         }
 620
 621         /* Allocate receive descriptors and buffers */
 622         error = em_create_rx_ring(adapter);
 623         if (error) {
 624                 device_printf(dev, "Could not setup receive structures\n");
 625                 goto fail;
 626         }
 627
 628         /* Manually turn off all interrupts */
 629         E1000_WRITE_REG(&adapter->hw, E1000_IMC, 0xffffffff);
 630
 631         /* Setup OS specific network interface */
 632         em_setup_ifp(adapter);
 633
 634         /* Add sysctl tree, must after em_setup_ifp() */
 635         em_add_sysctl(adapter);
 636
 637         /* Initialize statistics */
 638         em_update_stats(adapter);
 639
 640         adapter->hw.mac.get_link_status = 1;
 641         em_update_link_status(adapter);
 642
 643         /* Indicate SOL/IDER usage */
 644         if (e1000_check_reset_block(&adapter->hw)) {
 645                 device_printf(dev,
 646                     "PHY reset is blocked due to SOL/IDER session.\n");
 647         }
 648
 649         /* Determine if we have to control management hardware */
 650         adapter->has_manage = e1000_enable_mng_pass_thru(&adapter->hw);
 651
 652         /*
 653          * Setup Wake-on-Lan
 654          */
 655         switch (adapter->hw.mac.type) {
 656         case e1000_82542:
 657         case e1000_82543:
 658                 break;
 659
 660         case e1000_82546:
 661         case e1000_82546_rev_3:
 662         case e1000_82571:
 663         case e1000_80003es2lan:
 664                 if (adapter->hw.bus.func == 1) {
 665                         e1000_read_nvm(&adapter->hw,
 666                             NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
 667                 } else {
 668                         e1000_read_nvm(&adapter->hw,
 669                             NVM_INIT_CONTROL3_PORT_A, 1, &eeprom_data);
 670                 }
 671                 eeprom_data &= EM_EEPROM_APME;
 672                 break;
 673
 674         default:
 675                 /* APME bit in EEPROM is mapped to WUC.APME */
 676                 eeprom_data =
 677                     E1000_READ_REG(&adapter->hw, E1000_WUC) & E1000_WUC_APME;
 678                 break;
 679         }
 680         if (eeprom_data)
 681                 adapter->wol = E1000_WUFC_MAG;
 682         /*
 683          * We have the eeprom settings, now apply the special cases
 684          * where the eeprom may be wrong or the board won't support
 685          * wake on lan on a particular port
 686          */
 687         device_id = pci_get_device(dev);
 688         switch (device_id) {
 689         case E1000_DEV_ID_82546GB_PCIE:
 690                 adapter->wol = 0;
 691                 break;
 692
 693         case E1000_DEV_ID_82546EB_FIBER:
 694         case E1000_DEV_ID_82546GB_FIBER:
 695         case E1000_DEV_ID_82571EB_FIBER:
 696                 /*
 697                  * Wake events only supported on port A for dual fiber
 698                  * regardless of eeprom setting
 699                  */
 700                 if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
 701                     E1000_STATUS_FUNC_1)
 702                         adapter->wol = 0;
 703                 break;
 704
 705         case E1000_DEV_ID_82546GB_QUAD_COPPER_KSP3:
 706         case E1000_DEV_ID_82571EB_QUAD_COPPER:
 707         case E1000_DEV_ID_82571EB_QUAD_FIBER:
 708         case E1000_DEV_ID_82571EB_QUAD_COPPER_LP:
 709                 /* if quad port adapter, disable WoL on all but port A */
 710                 if (em_global_quad_port_a != 0)
 711                         adapter->wol = 0;
 712                 /* Reset for multiple quad port adapters */
 713                 if (++em_global_quad_port_a == 4)
 714                         em_global_quad_port_a = 0;
 715                 break;
 716         }
 717
 718         /* XXX disable wol */
 719         adapter->wol = 0;
 720
 721         /* Do we need workaround for 82544 PCI-X adapter? */
 722         if (adapter->hw.bus.type == e1000_bus_type_pcix &&
 723             adapter->hw.mac.type == e1000_82544)
 724                 adapter->pcix_82544 = TRUE;
 725         else
 726                 adapter->pcix_82544 = FALSE;
 727
 728         if (adapter->pcix_82544) {
 729                 /*
 730                  * 82544 on PCI-X may split one TX segment
 731                  * into two TX descs, so we double its number
 732                  * of spare TX desc here.
 733                  */
 734                 adapter->spare_tx_desc = 2 * EM_TX_SPARE;
 735         } else {
 736                 adapter->spare_tx_desc = EM_TX_SPARE;
 737         }
 738
 739         /*
 740          * Keep following relationship between spare_tx_desc, oact_tx_desc
 741          * and tx_int_nsegs:
 742          * (spare_tx_desc + EM_TX_RESERVED) <=
 743          * oact_tx_desc <= EM_TX_OACTIVE_MAX <= tx_int_nsegs
 744          */
 745         adapter->oact_tx_desc = adapter->num_tx_desc / 8;
 746         if (adapter->oact_tx_desc > EM_TX_OACTIVE_MAX)
 747                 adapter->oact_tx_desc = EM_TX_OACTIVE_MAX;
 748         if (adapter->oact_tx_desc < adapter->spare_tx_desc + EM_TX_RESERVED)
 749                 adapter->oact_tx_desc = adapter->spare_tx_desc + EM_TX_RESERVED;
 750
 751         adapter->tx_int_nsegs = adapter->num_tx_desc / 16;
 752         if (adapter->tx_int_nsegs < adapter->oact_tx_desc)
 753                 adapter->tx_int_nsegs = adapter->oact_tx_desc;
 754
 755         error = bus_setup_intr(dev, adapter->intr_res, INTR_MPSAFE,
 756                                em_intr, adapter, &adapter->intr_tag,
 757                                ifp->if_serializer);
 758         if (error) {
 759                 device_printf(dev, "Failed to register interrupt handler");
 760                 ether_ifdetach(&adapter->arpcom.ac_if);
 761                 goto fail;
 762         }
 763
 764         ifp->if_cpuid = ithread_cpuid(rman_get_start(adapter->intr_res));
 765         KKASSERT(ifp->if_cpuid >= 0 && ifp->if_cpuid < ncpus);
 766         return (0);
 767 fail:
 768         em_detach(dev);
 769         return (error);
 770 }
 771
 772 static int
 773 em_detach(device_t dev)
 774 {
 775         struct adapter *adapter = device_get_softc(dev);
 776
 777         if (device_is_attached(dev)) {
 778                 struct ifnet *ifp = &adapter->arpcom.ac_if;
 779
 780                 lwkt_serialize_enter(ifp->if_serializer);
 781
 782                 adapter->in_detach = 1;
 783                 em_stop(adapter);
 784
 785                 e1000_phy_hw_reset(&adapter->hw);
 786
 787                 em_rel_mgmt(adapter);
 788
 789                 if ((adapter->hw.mac.type == e1000_82573 ||
 790                      adapter->hw.mac.type == e1000_ich8lan ||
 791                      adapter->hw.mac.type == e1000_ich10lan ||
 792                      adapter->hw.mac.type == e1000_ich9lan) &&
 793                     e1000_check_mng_mode(&adapter->hw))
 794                         em_rel_hw_control(adapter);
 795
 796                 if (adapter->wol) {
 797                         E1000_WRITE_REG(&adapter->hw, E1000_WUC,
 798                                         E1000_WUC_PME_EN);
 799                         E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
 800                         em_enable_wol(dev);
 801                 }
 802
 803                 bus_teardown_intr(dev, adapter->intr_res, adapter->intr_tag);
 804
 805                 lwkt_serialize_exit(ifp->if_serializer);
 806
 807                 ether_ifdetach(ifp);
 808         }
 809         bus_generic_detach(dev);
 810
 811         em_free_pci_res(adapter);
 812
 813         em_destroy_tx_ring(adapter, adapter->num_tx_desc);
 814         em_destroy_rx_ring(adapter, adapter->num_rx_desc);
 815
 816         /* Free Transmit Descriptor ring */
 817         if (adapter->tx_desc_base)
 818                 em_dma_free(adapter, &adapter->txdma);
 819
 820         /* Free Receive Descriptor ring */
 821         if (adapter->rx_desc_base)
 822                 em_dma_free(adapter, &adapter->rxdma);
 823
 824         /* Free top level busdma tag */
 825         if (adapter->parent_dtag != NULL)
 826                 bus_dma_tag_destroy(adapter->parent_dtag);
 827
 828         /* Free sysctl tree */
 829         if (adapter->sysctl_tree != NULL)
 830                 sysctl_ctx_free(&adapter->sysctl_ctx);
 831
 832         return (0);
 833 }
 834
 835 static int
 836 em_shutdown(device_t dev)
 837 {
 838         return em_suspend(dev);
 839 }
 840
 841 static int
 842 em_suspend(device_t dev)
 843 {
 844         struct adapter *adapter = device_get_softc(dev);
 845         struct ifnet *ifp = &adapter->arpcom.ac_if;
 846
 847         lwkt_serialize_enter(ifp->if_serializer);
 848
 849         em_stop(adapter);
 850
 851         em_rel_mgmt(adapter);
 852
 853         if ((adapter->hw.mac.type == e1000_82573 ||
 854              adapter->hw.mac.type == e1000_ich8lan ||
 855              adapter->hw.mac.type == e1000_ich10lan ||
 856              adapter->hw.mac.type == e1000_ich9lan) &&
 857             e1000_check_mng_mode(&adapter->hw))
 858                 em_rel_hw_control(adapter);
 859
 860         if (adapter->wol) {
 861                 E1000_WRITE_REG(&adapter->hw, E1000_WUC, E1000_WUC_PME_EN);
 862                 E1000_WRITE_REG(&adapter->hw, E1000_WUFC, adapter->wol);
 863                 em_enable_wol(dev);
 864         }
 865
 866         lwkt_serialize_exit(ifp->if_serializer);
 867
 868         return bus_generic_suspend(dev);
 869 }
 870
 871 static int
 872 em_resume(device_t dev)
 873 {
 874         struct adapter *adapter = device_get_softc(dev);
 875         struct ifnet *ifp = &adapter->arpcom.ac_if;
 876
 877         lwkt_serialize_enter(ifp->if_serializer);
 878
 879         em_init(adapter);
 880         em_get_mgmt(adapter);
 881         if_devstart(ifp);
 882
 883         lwkt_serialize_exit(ifp->if_serializer);
 884
 885         return bus_generic_resume(dev);
 886 }
 887
 888 static void
 889 em_start(struct ifnet *ifp)
 890 {
 891         struct adapter *adapter = ifp->if_softc;
 892         struct mbuf *m_head;
 893
 894         ASSERT_SERIALIZED(ifp->if_serializer);
 895
 896         if ((ifp->if_flags & (IFF_RUNNING | IFF_OACTIVE)) != IFF_RUNNING)
 897                 return;
 898
 899         if (!adapter->link_active) {
 900                 ifq_purge(&ifp->if_snd);
 901                 return;
 902         }
 903
 904         while (!ifq_is_empty(&ifp->if_snd)) {
 905                 /* Now do we at least have a minimal? */
 906                 if (EM_IS_OACTIVE(adapter)) {
 907                         em_tx_collect(adapter);
 908                         if (EM_IS_OACTIVE(adapter)) {
 909                                 ifp->if_flags |= IFF_OACTIVE;
 910                                 adapter->no_tx_desc_avail1++;
 911                                 break;
 912                         }
 913                 }
 914
 915                 logif(pkt_txqueue);
 916                 m_head = ifq_dequeue(&ifp->if_snd, NULL);
 917                 if (m_head == NULL)
 918                         break;
 919
 920                 if (em_encap(adapter, &m_head)) {
 921                         ifp->if_oerrors++;
 922                         em_tx_collect(adapter);
 923                         continue;
 924                 }
 925
 926                 /* Send a copy of the frame to the BPF listener */
 927                 ETHER_BPF_MTAP(ifp, m_head);
 928
 929                 /* Set timeout in case hardware has problems transmitting. */
 930                 ifp->if_timer = EM_TX_TIMEOUT;
 931         }
 932 }
 933
 934 static int
 935 em_ioctl(struct ifnet *ifp, u_long command, caddr_t data, struct ucred *cr)
 936 {
 937         struct adapter *adapter = ifp->if_softc;
 938         struct ifreq *ifr = (struct ifreq *)data;
 939         uint16_t eeprom_data = 0;
 940         int max_frame_size, mask, reinit;
 941         int error = 0;
 942
 943         if (adapter->in_detach)
 944                 return (error);
 945
 946         ASSERT_SERIALIZED(ifp->if_serializer);
 947
 948         switch (command) {
 949         case SIOCSIFMTU:
 950                 switch (adapter->hw.mac.type) {
 951                 case e1000_82573:
 952                         /*
 953                          * 82573 only supports jumbo frames
 954                          * if ASPM is disabled.
 955                          */
 956                         e1000_read_nvm(&adapter->hw,
 957                             NVM_INIT_3GIO_3, 1, &eeprom_data);
 958                         if (eeprom_data & NVM_WORD1A_ASPM_MASK) {
 959                                 max_frame_size = ETHER_MAX_LEN;
 960                                 break;
 961                         }
 962                         /* FALL THROUGH */
 963
 964                 /* Limit Jumbo Frame size */
 965                 case e1000_82571:
 966                 case e1000_82572:
 967                 case e1000_ich9lan:
 968                 case e1000_ich10lan:
 969                 case e1000_82574:
 970                 case e1000_80003es2lan:
 971                         max_frame_size = 9234;
 972                         break;
 973
 974                 /* Adapters that do not support jumbo frames */
 975                 case e1000_82542:
 976                 case e1000_ich8lan:
 977                         max_frame_size = ETHER_MAX_LEN;
 978                         break;
 979
 980                 default:
 981                         max_frame_size = MAX_JUMBO_FRAME_SIZE;
 982                         break;
 983                 }
 984                 if (ifr->ifr_mtu > max_frame_size - ETHER_HDR_LEN -
 985                     ETHER_CRC_LEN) {
 986                         error = EINVAL;
 987                         break;
 988                 }
 989
 990                 ifp->if_mtu = ifr->ifr_mtu;
 991                 adapter->max_frame_size =
 992                     ifp->if_mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
 993
 994                 if (ifp->if_flags & IFF_RUNNING)
 995                         em_init(adapter);
 996                 break;
 997
 998         case SIOCSIFFLAGS:
 999                 if (ifp->if_flags & IFF_UP) {
1000                         if ((ifp->if_flags & IFF_RUNNING)) {
1001                                 if ((ifp->if_flags ^ adapter->if_flags) &
1002                                     (IFF_PROMISC | IFF_ALLMULTI)) {
1003                                         em_disable_promisc(adapter);
1004                                         em_set_promisc(adapter);
1005                                 }
1006                         } else {
1007                                 em_init(adapter);
1008                         }
1009                 } else if (ifp->if_flags & IFF_RUNNING) {
1010                         em_stop(adapter);
1011                 }
1012                 adapter->if_flags = ifp->if_flags;
1013                 break;
1014
1015         case SIOCADDMULTI:
1016         case SIOCDELMULTI:
1017                 if (ifp->if_flags & IFF_RUNNING) {
1018                         em_disable_intr(adapter);
1019                         em_set_multi(adapter);
1020                         if (adapter->hw.mac.type == e1000_82542 &&
1021                             adapter->hw.revision_id == E1000_REVISION_2)
1022                                 em_init_rx_unit(adapter);
1023 #ifdef DEVICE_POLLING
1024                         if (!(ifp->if_flags & IFF_POLLING))
1025 #endif
1026                                 em_enable_intr(adapter);
1027                 }
1028                 break;
1029
1030         case SIOCSIFMEDIA:
1031                 /* Check SOL/IDER usage */
1032                 if (e1000_check_reset_block(&adapter->hw)) {
1033                         device_printf(adapter->dev, "Media change is"
1034                             " blocked due to SOL/IDER session.\n");
1035                         break;
1036                 }
1037                 /* FALL THROUGH */
1038
1039         case SIOCGIFMEDIA:
1040                 error = ifmedia_ioctl(ifp, ifr, &adapter->media, command);
1041                 break;
1042
1043         case SIOCSIFCAP:
1044                 reinit = 0;
1045                 mask = ifr->ifr_reqcap ^ ifp->if_capenable;
1046                 if (mask & IFCAP_HWCSUM) {
1047                         ifp->if_capenable ^= (mask & IFCAP_HWCSUM);
1048                         reinit = 1;
1049                 }
1050                 if (mask & IFCAP_VLAN_HWTAGGING) {
1051                         ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING;
1052                         reinit = 1;
1053                 }
1054                 if (reinit && (ifp->if_flags & IFF_RUNNING))
1055                         em_init(adapter);
1056                 break;
1057
1058         default:
1059                 error = ether_ioctl(ifp, command, data);
1060                 break;
1061         }
1062         return (error);
1063 }
1064
1065 static void
1066 em_watchdog(struct ifnet *ifp)
1067 {
1068         struct adapter *adapter = ifp->if_softc;
1069
1070         ASSERT_SERIALIZED(ifp->if_serializer);
1071
1072         /*
1073          * The timer is set to 5 every time start queues a packet.
1074          * Then txeof keeps resetting it as long as it cleans at
1075          * least one descriptor.
1076          * Finally, anytime all descriptors are clean the timer is
1077          * set to 0.
1078          */
1079
1080         if (E1000_READ_REG(&adapter->hw, E1000_TDT(0)) ==
1081             E1000_READ_REG(&adapter->hw, E1000_TDH(0))) {
1082                 /*
1083                  * If we reach here, all TX jobs are completed and
1084                  * the TX engine should have been idled for some time.
1085                  * We don't need to call if_devstart() here.
1086                  */
1087                 ifp->if_flags &= ~IFF_OACTIVE;
1088                 ifp->if_timer = 0;
1089                 return;
1090         }
1091
1092         /*
1093          * If we are in this routine because of pause frames, then
1094          * don't reset the hardware.
1095          */
1096         if (E1000_READ_REG(&adapter->hw, E1000_STATUS) &
1097             E1000_STATUS_TXOFF) {
1098                 ifp->if_timer = EM_TX_TIMEOUT;
1099                 return;
1100         }
1101
1102         if (e1000_check_for_link(&adapter->hw) == 0)
1103                 if_printf(ifp, "watchdog timeout -- resetting\n");
1104
1105         ifp->if_oerrors++;
1106         adapter->watchdog_events++;
1107
1108         em_init(adapter);
1109
1110         if (!ifq_is_empty(&ifp->if_snd))
1111                 if_devstart(ifp);
1112 }
1113
1114 static void
1115 em_init(void *xsc)
1116 {
1117         struct adapter *adapter = xsc;
1118         struct ifnet *ifp = &adapter->arpcom.ac_if;
1119         device_t dev = adapter->dev;
1120         uint32_t pba;
1121
1122         ASSERT_SERIALIZED(ifp->if_serializer);
1123
1124         em_stop(adapter);
1125
1126         /*
1127          * Packet Buffer Allocation (PBA)
1128          * Writing PBA sets the receive portion of the buffer
1129          * the remainder is used for the transmit buffer.
1130          *
1131          * Devices before the 82547 had a Packet Buffer of 64K.
1132          *   Default allocation: PBA=48K for Rx, leaving 16K for Tx.
1133          * After the 82547 the buffer was reduced to 40K.
1134          *   Default allocation: PBA=30K for Rx, leaving 10K for Tx.
1135          *   Note: default does not leave enough room for Jumbo Frame >10k.
1136          */
1137         switch (adapter->hw.mac.type) {
1138         case e1000_82547:
1139         case e1000_82547_rev_2: /* 82547: Total Packet Buffer is 40K */
1140                 if (adapter->max_frame_size > 8192)
1141                         pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
1142                 else
1143                         pba = E1000_PBA_30K; /* 30K for Rx, 10K for Tx */
1144                 adapter->tx_fifo_head = 0;
1145                 adapter->tx_head_addr = pba << EM_TX_HEAD_ADDR_SHIFT;
1146                 adapter->tx_fifo_size =
1147                     (E1000_PBA_40K - pba) << EM_PBA_BYTES_SHIFT;
1148                 break;
1149
1150         /* Total Packet Buffer on these is 48K */
1151         case e1000_82571:
1152         case e1000_82572:
1153         case e1000_80003es2lan:
1154                 pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
1155                 break;
1156
1157         case e1000_82573: /* 82573: Total Packet Buffer is 32K */
1158                 pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
1159                 break;
1160
1161         case e1000_82574:
1162                 pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
1163                 break;
1164
1165         case e1000_ich9lan:
1166         case e1000_ich10lan:
1167 #define E1000_PBA_10K   0x000A
1168                 pba = E1000_PBA_10K;
1169                 break;
1170
1171         case e1000_ich8lan:
1172                 pba = E1000_PBA_8K;
1173                 break;
1174
1175         default:
1176                 /* Devices before 82547 had a Packet Buffer of 64K.   */
1177                 if (adapter->max_frame_size > 8192)
1178                         pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
1179                 else
1180                         pba = E1000_PBA_48K; /* 48K for Rx, 16K for Tx */
1181         }
1182
1183         E1000_WRITE_REG(&adapter->hw, E1000_PBA, pba);
1184
1185         /* Get the latest mac address, User can use a LAA */
1186         bcopy(IF_LLADDR(ifp), adapter->hw.mac.addr, ETHER_ADDR_LEN);
1187
1188         /* Put the address into the Receive Address Array */
1189         e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1190
1191         /*
1192          * With the 82571 adapter, RAR[0] may be overwritten
1193          * when the other port is reset, we make a duplicate
1194          * in RAR[14] for that eventuality, this assures
1195          * the interface continues to function.
1196          */
1197         if (adapter->hw.mac.type == e1000_82571) {
1198                 e1000_set_laa_state_82571(&adapter->hw, TRUE);
1199                 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr,
1200                     E1000_RAR_ENTRIES - 1);
1201         }
1202
1203         /* Initialize the hardware */
1204         if (em_hw_init(adapter)) {
1205                 device_printf(dev, "Unable to initialize the hardware\n");
1206                 /* XXX em_stop()? */
1207                 return;
1208         }
1209         em_update_link_status(adapter);
1210
1211         /* Setup VLAN support, basic and offload if available */
1212         E1000_WRITE_REG(&adapter->hw, E1000_VET, ETHERTYPE_VLAN);
1213
1214         if (ifp->if_capenable & IFCAP_VLAN_HWTAGGING) {
1215                 uint32_t ctrl;
1216
1217                 ctrl = E1000_READ_REG(&adapter->hw, E1000_CTRL);
1218                 ctrl |= E1000_CTRL_VME;
1219                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL, ctrl);
1220         }
1221
1222         /* Set hardware offload abilities */
1223         if (ifp->if_capenable & IFCAP_TXCSUM)
1224                 ifp->if_hwassist = EM_CSUM_FEATURES;
1225         else
1226                 ifp->if_hwassist = 0;
1227
1228         /* Configure for OS presence */
1229         em_get_mgmt(adapter);
1230
1231         /* Prepare transmit descriptors and buffers */
1232         em_init_tx_ring(adapter);
1233         em_init_tx_unit(adapter);
1234
1235         /* Setup Multicast table */
1236         em_set_multi(adapter);
1237
1238         /* Prepare receive descriptors and buffers */
1239         if (em_init_rx_ring(adapter)) {
1240                 device_printf(dev, "Could not setup receive structures\n");
1241                 em_stop(adapter);
1242                 return;
1243         }
1244         em_init_rx_unit(adapter);
1245
1246         /* Don't lose promiscuous settings */
1247         em_set_promisc(adapter);
1248
1249         ifp->if_flags |= IFF_RUNNING;
1250         ifp->if_flags &= ~IFF_OACTIVE;
1251
1252         callout_reset(&adapter->timer, hz, em_timer, adapter);
1253         e1000_clear_hw_cntrs_base_generic(&adapter->hw);
1254
1255         /* MSI/X configuration for 82574 */
1256         if (adapter->hw.mac.type == e1000_82574) {
1257                 int tmp;
1258
1259                 tmp = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
1260                 tmp |= E1000_CTRL_EXT_PBA_CLR;
1261                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT, tmp);
1262                 /*
1263                  * Set the IVAR - interrupt vector routing.
1264                  * Each nibble represents a vector, high bit
1265                  * is enable, other 3 bits are the MSIX table
1266                  * entry, we map RXQ0 to 0, TXQ0 to 1, and
1267                  * Link (other) to 2, hence the magic number.
1268                  */
1269                 E1000_WRITE_REG(&adapter->hw, E1000_IVAR, 0x800A0908);
1270         }
1271
1272 #ifdef DEVICE_POLLING
1273         /*
1274          * Only enable interrupts if we are not polling, make sure
1275          * they are off otherwise.
1276          */
1277         if (ifp->if_flags & IFF_POLLING)
1278                 em_disable_intr(adapter);
1279         else
1280 #endif /* DEVICE_POLLING */
1281                 em_enable_intr(adapter);
1282
1283         /* Don't reset the phy next time init gets called */
1284         adapter->hw.phy.reset_disable = TRUE;
1285 }
1286
1287 #ifdef DEVICE_POLLING
1288
1289 static void
1290 em_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
1291 {
1292         struct adapter *adapter = ifp->if_softc;
1293         uint32_t reg_icr;
1294
1295         ASSERT_SERIALIZED(ifp->if_serializer);
1296
1297         switch (cmd) {
1298         case POLL_REGISTER:
1299                 em_disable_intr(adapter);
1300                 break;
1301
1302         case POLL_DEREGISTER:
1303                 em_enable_intr(adapter);
1304                 break;
1305
1306         case POLL_AND_CHECK_STATUS:
1307                 reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1308                 if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1309                         callout_stop(&adapter->timer);
1310                         adapter->hw.mac.get_link_status = 1;
1311                         em_update_link_status(adapter);
1312                         callout_reset(&adapter->timer, hz, em_timer, adapter);
1313                 }
1314                 /* FALL THROUGH */
1315         case POLL_ONLY:
1316                 if (ifp->if_flags & IFF_RUNNING) {
1317                         em_rxeof(adapter, count);
1318                         em_txeof(adapter);
1319
1320                         if (!ifq_is_empty(&ifp->if_snd))
1321                                 if_devstart(ifp);
1322                 }
1323                 break;
1324         }
1325 }
1326
1327 #endif /* DEVICE_POLLING */
1328
1329 static void
1330 em_intr(void *xsc)
1331 {
1332         struct adapter *adapter = xsc;
1333         struct ifnet *ifp = &adapter->arpcom.ac_if;
1334         uint32_t reg_icr;
1335
1336         logif(intr_beg);
1337         ASSERT_SERIALIZED(ifp->if_serializer);
1338
1339         reg_icr = E1000_READ_REG(&adapter->hw, E1000_ICR);
1340
1341         if ((adapter->hw.mac.type >= e1000_82571 &&
1342              (reg_icr & E1000_ICR_INT_ASSERTED) == 0) ||
1343             reg_icr == 0) {
1344                 logif(intr_end);
1345                 return;
1346         }
1347
1348         /*
1349          * XXX: some laptops trigger several spurious interrupts
1350          * on em(4) when in the resume cycle. The ICR register
1351          * reports all-ones value in this case. Processing such
1352          * interrupts would lead to a freeze. I don't know why.
1353          */
1354         if (reg_icr == 0xffffffff) {
1355                 logif(intr_end);
1356                 return;
1357         }
1358
1359         if (ifp->if_flags & IFF_RUNNING) {
1360                 if (reg_icr &
1361                     (E1000_IMS_RXT0 | E1000_IMS_RXDMT0 | E1000_ICR_RXO))
1362                         em_rxeof(adapter, -1);
1363                 if (reg_icr & E1000_IMS_TXDW) {
1364                         em_txeof(adapter);
1365                         if (!ifq_is_empty(&ifp->if_snd))
1366                                 if_devstart(ifp);
1367                 }
1368         }
1369
1370         /* Link status change */
1371         if (reg_icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
1372                 callout_stop(&adapter->timer);
1373                 adapter->hw.mac.get_link_status = 1;
1374                 em_update_link_status(adapter);
1375
1376                 /* Deal with TX cruft when link lost */
1377                 em_tx_purge(adapter);
1378
1379                 callout_reset(&adapter->timer, hz, em_timer, adapter);
1380         }
1381
1382         if (reg_icr & E1000_ICR_RXO)
1383                 adapter->rx_overruns++;
1384
1385         logif(intr_end);
1386 }
1387
1388 static void
1389 em_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
1390 {
1391         struct adapter *adapter = ifp->if_softc;
1392         u_char fiber_type = IFM_1000_SX;
1393
1394         ASSERT_SERIALIZED(ifp->if_serializer);
1395
1396         em_update_link_status(adapter);
1397
1398         ifmr->ifm_status = IFM_AVALID;
1399         ifmr->ifm_active = IFM_ETHER;
1400
1401         if (!adapter->link_active)
1402                 return;
1403
1404         ifmr->ifm_status |= IFM_ACTIVE;
1405
1406         if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
1407             adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
1408                 if (adapter->hw.mac.type == e1000_82545)
1409                         fiber_type = IFM_1000_LX;
1410                 ifmr->ifm_active |= fiber_type | IFM_FDX;
1411         } else {
1412                 switch (adapter->link_speed) {
1413                 case 10:
1414                         ifmr->ifm_active |= IFM_10_T;
1415                         break;
1416                 case 100:
1417                         ifmr->ifm_active |= IFM_100_TX;
1418                         break;
1419
1420                 case 1000:
1421                         ifmr->ifm_active |= IFM_1000_T;
1422                         break;
1423                 }
1424                 if (adapter->link_duplex == FULL_DUPLEX)
1425                         ifmr->ifm_active |= IFM_FDX;
1426                 else
1427                         ifmr->ifm_active |= IFM_HDX;
1428         }
1429 }
1430
1431 static int
1432 em_media_change(struct ifnet *ifp)
1433 {
1434         struct adapter *adapter = ifp->if_softc;
1435         struct ifmedia *ifm = &adapter->media;
1436
1437         ASSERT_SERIALIZED(ifp->if_serializer);
1438
1439         if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER)
1440                 return (EINVAL);
1441
1442         switch (IFM_SUBTYPE(ifm->ifm_media)) {
1443         case IFM_AUTO:
1444                 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1445                 adapter->hw.phy.autoneg_advertised = AUTONEG_ADV_DEFAULT;
1446                 break;
1447
1448         case IFM_1000_LX:
1449         case IFM_1000_SX:
1450         case IFM_1000_T:
1451                 adapter->hw.mac.autoneg = DO_AUTO_NEG;
1452                 adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
1453                 break;
1454
1455         case IFM_100_TX:
1456                 adapter->hw.mac.autoneg = FALSE;
1457                 adapter->hw.phy.autoneg_advertised = 0;
1458                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1459                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_FULL;
1460                 else
1461                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_100_HALF;
1462                 break;
1463
1464         case IFM_10_T:
1465                 adapter->hw.mac.autoneg = FALSE;
1466                 adapter->hw.phy.autoneg_advertised = 0;
1467                 if ((ifm->ifm_media & IFM_GMASK) == IFM_FDX)
1468                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_FULL;
1469                 else
1470                         adapter->hw.mac.forced_speed_duplex = ADVERTISE_10_HALF;
1471                 break;
1472
1473         default:
1474                 if_printf(ifp, "Unsupported media type\n");
1475                 break;
1476         }
1477
1478         /*
1479          * As the speed/duplex settings my have changed we need to
1480          * reset the PHY.
1481          */
1482         adapter->hw.phy.reset_disable = FALSE;
1483
1484         em_init(adapter);
1485
1486         return (0);
1487 }
1488
1489 static int
1490 em_encap(struct adapter *adapter, struct mbuf **m_headp)
1491 {
1492         bus_dma_segment_t segs[EM_MAX_SCATTER];
1493         bus_dmamap_t map;
1494         struct em_buffer *tx_buffer, *tx_buffer_mapped;
1495         struct e1000_tx_desc *ctxd = NULL;
1496         struct mbuf *m_head = *m_headp;
1497         uint32_t txd_upper, txd_lower, txd_used, cmd = 0;
1498         int maxsegs, nsegs, i, j, first, last = 0, error;
1499
1500         if (__predict_false(m_head->m_len < EM_TXCSUM_MINHL) &&
1501             (m_head->m_flags & EM_CSUM_FEATURES)) {
1502                 /*
1503                  * Make sure that ethernet header and ip.ip_hl are in
1504                  * contiguous memory, since if TXCSUM is enabled, later
1505                  * TX context descriptor's setup need to access ip.ip_hl.
1506                  */
1507                 error = em_txcsum_pullup(adapter, m_headp);
1508                 if (error) {
1509                         KKASSERT(*m_headp == NULL);
1510                         return error;
1511                 }
1512                 m_head = *m_headp;
1513         }
1514
1515         txd_upper = txd_lower = 0;
1516         txd_used = 0;
1517
1518         /*
1519          * Capture the first descriptor index, this descriptor
1520          * will have the index of the EOP which is the only one
1521          * that now gets a DONE bit writeback.
1522          */
1523         first = adapter->next_avail_tx_desc;
1524         tx_buffer = &adapter->tx_buffer_area[first];
1525         tx_buffer_mapped = tx_buffer;
1526         map = tx_buffer->map;
1527
1528         maxsegs = adapter->num_tx_desc_avail - EM_TX_RESERVED;
1529         KASSERT(maxsegs >= adapter->spare_tx_desc,
1530                 ("not enough spare TX desc\n"));
1531         if (adapter->pcix_82544) {
1532                 /* Half it; see the comment in em_attach() */
1533                 maxsegs >>= 1;
1534         }
1535         if (maxsegs > EM_MAX_SCATTER)
1536                 maxsegs = EM_MAX_SCATTER;
1537
1538         error = bus_dmamap_load_mbuf_defrag(adapter->txtag, map, m_headp,
1539                         segs, maxsegs, &nsegs, BUS_DMA_NOWAIT);
1540         if (error) {
1541                 if (error == ENOBUFS)
1542                         adapter->mbuf_alloc_failed++;
1543                 else
1544                         adapter->no_tx_dma_setup++;
1545
1546                 m_freem(*m_headp);
1547                 *m_headp = NULL;
1548                 return error;
1549         }
1550         bus_dmamap_sync(adapter->txtag, map, BUS_DMASYNC_PREWRITE);
1551
1552         m_head = *m_headp;
1553         adapter->tx_nsegs += nsegs;
1554
1555         if (m_head->m_pkthdr.csum_flags & EM_CSUM_FEATURES) {
1556                 /* TX csum offloading will consume one TX desc */
1557                 adapter->tx_nsegs += em_txcsum(adapter, m_head,
1558                                                &txd_upper, &txd_lower);
1559         }
1560         i = adapter->next_avail_tx_desc;
1561
1562         /* Set up our transmit descriptors */
1563         for (j = 0; j < nsegs; j++) {
1564                 /* If adapter is 82544 and on PCIX bus */
1565                 if(adapter->pcix_82544) {
1566                         DESC_ARRAY desc_array;
1567                         uint32_t array_elements, counter;
1568
1569                         /*
1570                          * Check the Address and Length combination and
1571                          * split the data accordingly
1572                          */
1573                         array_elements = em_82544_fill_desc(segs[j].ds_addr,
1574                                                 segs[j].ds_len, &desc_array);
1575                         for (counter = 0; counter < array_elements; counter++) {
1576                                 KKASSERT(txd_used < adapter->num_tx_desc_avail);
1577
1578                                 tx_buffer = &adapter->tx_buffer_area[i];
1579                                 ctxd = &adapter->tx_desc_base[i];
1580
1581                                 ctxd->buffer_addr = htole64(
1582                                     desc_array.descriptor[counter].address);
1583                                 ctxd->lower.data = htole32(
1584                                     E1000_TXD_CMD_IFCS | txd_lower |
1585                                     desc_array.descriptor[counter].length);
1586                                 ctxd->upper.data = htole32(txd_upper);
1587
1588                                 last = i;
1589                                 if (++i == adapter->num_tx_desc)
1590                                         i = 0;
1591
1592                                 tx_buffer->m_head = NULL;
1593                                 txd_used++;
1594                         }
1595                 } else {
1596                         tx_buffer = &adapter->tx_buffer_area[i];
1597                         ctxd = &adapter->tx_desc_base[i];
1598
1599                         ctxd->buffer_addr = htole64(segs[j].ds_addr);
1600                         ctxd->lower.data = htole32(E1000_TXD_CMD_IFCS |
1601                                                    txd_lower | segs[j].ds_len);
1602                         ctxd->upper.data = htole32(txd_upper);
1603
1604                         last = i;
1605                         if (++i == adapter->num_tx_desc)
1606                                 i = 0;
1607
1608                         tx_buffer->m_head = NULL;
1609                 }
1610         }
1611
1612         adapter->next_avail_tx_desc = i;
1613         if (adapter->pcix_82544) {
1614                 KKASSERT(adapter->num_tx_desc_avail > txd_used);
1615                 adapter->num_tx_desc_avail -= txd_used;
1616         } else {
1617                 KKASSERT(adapter->num_tx_desc_avail > nsegs);
1618                 adapter->num_tx_desc_avail -= nsegs;
1619         }
1620
1621         /* Handle VLAN tag */
1622         if (m_head->m_flags & M_VLANTAG) {
1623                 /* Set the vlan id. */
1624                 ctxd->upper.fields.special =
1625                     htole16(m_head->m_pkthdr.ether_vlantag);
1626
1627                 /* Tell hardware to add tag */
1628                 ctxd->lower.data |= htole32(E1000_TXD_CMD_VLE);
1629         }
1630
1631         tx_buffer->m_head = m_head;
1632         tx_buffer_mapped->map = tx_buffer->map;
1633         tx_buffer->map = map;
1634
1635         if (adapter->tx_nsegs >= adapter->tx_int_nsegs) {
1636                 adapter->tx_nsegs = 0;
1637                 cmd = E1000_TXD_CMD_RS;
1638
1639                 adapter->tx_dd[adapter->tx_dd_tail] = last;
1640                 EM_INC_TXDD_IDX(adapter->tx_dd_tail);
1641                 KKASSERT(adapter->tx_dd_tail != adapter->tx_dd_head);
1642         }
1643
1644         /*
1645          * Last Descriptor of Packet needs End Of Packet (EOP)
1646          * and Report Status (RS)
1647          */
1648         ctxd->lower.data |= htole32(E1000_TXD_CMD_EOP | cmd);
1649
1650         /*
1651          * Keep track in the first buffer which descriptor will be
1652          * written back
1653          */
1654         tx_buffer = &adapter->tx_buffer_area[first];
1655
1656         /*
1657          * Advance the Transmit Descriptor Tail (TDT), this tells the E1000
1658          * that this frame is available to transmit.
1659          */
1660         if (adapter->hw.mac.type == e1000_82547 &&
1661             adapter->link_duplex == HALF_DUPLEX) {
1662                 em_82547_move_tail_serialized(adapter);
1663         } else {
1664                 E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), i);
1665                 if (adapter->hw.mac.type == e1000_82547) {
1666                         em_82547_update_fifo_head(adapter,
1667                             m_head->m_pkthdr.len);
1668                 }
1669         }
1670         return (0);
1671 }
1672
1673 /*
1674  * 82547 workaround to avoid controller hang in half-duplex environment.
1675  * The workaround is to avoid queuing a large packet that would span
1676  * the internal Tx FIFO ring boundary.  We need to reset the FIFO pointers
1677  * in this case.  We do that only when FIFO is quiescent.
1678  */
1679 static void
1680 em_82547_move_tail_serialized(struct adapter *adapter)
1681 {
1682         struct e1000_tx_desc *tx_desc;
1683         uint16_t hw_tdt, sw_tdt, length = 0;
1684         bool eop = 0;
1685
1686         ASSERT_SERIALIZED(adapter->arpcom.ac_if.if_serializer);
1687
1688         hw_tdt = E1000_READ_REG(&adapter->hw, E1000_TDT(0));
1689         sw_tdt = adapter->next_avail_tx_desc;
1690
1691         while (hw_tdt != sw_tdt) {
1692                 tx_desc = &adapter->tx_desc_base[hw_tdt];
1693                 length += tx_desc->lower.flags.length;
1694                 eop = tx_desc->lower.data & E1000_TXD_CMD_EOP;
1695                 if (++hw_tdt == adapter->num_tx_desc)
1696                         hw_tdt = 0;
1697
1698                 if (eop) {
1699                         if (em_82547_fifo_workaround(adapter, length)) {
1700                                 adapter->tx_fifo_wrk_cnt++;
1701                                 callout_reset(&adapter->tx_fifo_timer, 1,
1702                                         em_82547_move_tail, adapter);
1703                                 break;
1704                         }
1705                         E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), hw_tdt);
1706                         em_82547_update_fifo_head(adapter, length);
1707                         length = 0;
1708                 }
1709         }
1710 }
1711
1712 static void
1713 em_82547_move_tail(void *xsc)
1714 {
1715         struct adapter *adapter = xsc;
1716         struct ifnet *ifp = &adapter->arpcom.ac_if;
1717
1718         lwkt_serialize_enter(ifp->if_serializer);
1719         em_82547_move_tail_serialized(adapter);
1720         lwkt_serialize_exit(ifp->if_serializer);
1721 }
1722
1723 static int
1724 em_82547_fifo_workaround(struct adapter *adapter, int len)
1725 {
1726         int fifo_space, fifo_pkt_len;
1727
1728         fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1729
1730         if (adapter->link_duplex == HALF_DUPLEX) {
1731                 fifo_space = adapter->tx_fifo_size - adapter->tx_fifo_head;
1732
1733                 if (fifo_pkt_len >= (EM_82547_PKT_THRESH + fifo_space)) {
1734                         if (em_82547_tx_fifo_reset(adapter))
1735                                 return (0);
1736                         else
1737                                 return (1);
1738                 }
1739         }
1740         return (0);
1741 }
1742
1743 static void
1744 em_82547_update_fifo_head(struct adapter *adapter, int len)
1745 {
1746         int fifo_pkt_len = roundup2(len + EM_FIFO_HDR, EM_FIFO_HDR);
1747
1748         /* tx_fifo_head is always 16 byte aligned */
1749         adapter->tx_fifo_head += fifo_pkt_len;
1750         if (adapter->tx_fifo_head >= adapter->tx_fifo_size)
1751                 adapter->tx_fifo_head -= adapter->tx_fifo_size;
1752 }
1753
1754 static int
1755 em_82547_tx_fifo_reset(struct adapter *adapter)
1756 {
1757         uint32_t tctl;
1758
1759         if ((E1000_READ_REG(&adapter->hw, E1000_TDT(0)) ==
1760              E1000_READ_REG(&adapter->hw, E1000_TDH(0))) &&
1761             (E1000_READ_REG(&adapter->hw, E1000_TDFT) ==
1762              E1000_READ_REG(&adapter->hw, E1000_TDFH)) &&
1763             (E1000_READ_REG(&adapter->hw, E1000_TDFTS) ==
1764              E1000_READ_REG(&adapter->hw, E1000_TDFHS)) &&
1765             (E1000_READ_REG(&adapter->hw, E1000_TDFPC) == 0)) {
1766                 /* Disable TX unit */
1767                 tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
1768                 E1000_WRITE_REG(&adapter->hw, E1000_TCTL,
1769                     tctl & ~E1000_TCTL_EN);
1770
1771                 /* Reset FIFO pointers */
1772                 E1000_WRITE_REG(&adapter->hw, E1000_TDFT,
1773                     adapter->tx_head_addr);
1774                 E1000_WRITE_REG(&adapter->hw, E1000_TDFH,
1775                     adapter->tx_head_addr);
1776                 E1000_WRITE_REG(&adapter->hw, E1000_TDFTS,
1777                     adapter->tx_head_addr);
1778                 E1000_WRITE_REG(&adapter->hw, E1000_TDFHS,
1779                     adapter->tx_head_addr);
1780
1781                 /* Re-enable TX unit */
1782                 E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
1783                 E1000_WRITE_FLUSH(&adapter->hw);
1784
1785                 adapter->tx_fifo_head = 0;
1786                 adapter->tx_fifo_reset_cnt++;
1787
1788                 return (TRUE);
1789         } else {
1790                 return (FALSE);
1791         }
1792 }
1793
1794 static void
1795 em_set_promisc(struct adapter *adapter)
1796 {
1797         struct ifnet *ifp = &adapter->arpcom.ac_if;
1798         uint32_t reg_rctl;
1799
1800         reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1801
1802         if (ifp->if_flags & IFF_PROMISC) {
1803                 reg_rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1804                 /* Turn this on if you want to see bad packets */
1805                 if (em_debug_sbp)
1806                         reg_rctl |= E1000_RCTL_SBP;
1807                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1808         } else if (ifp->if_flags & IFF_ALLMULTI) {
1809                 reg_rctl |= E1000_RCTL_MPE;
1810                 reg_rctl &= ~E1000_RCTL_UPE;
1811                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1812         }
1813 }
1814
1815 static void
1816 em_disable_promisc(struct adapter *adapter)
1817 {
1818         uint32_t reg_rctl;
1819
1820         reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1821
1822         reg_rctl &= ~E1000_RCTL_UPE;
1823         reg_rctl &= ~E1000_RCTL_MPE;
1824         reg_rctl &= ~E1000_RCTL_SBP;
1825         E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1826 }
1827
1828 static void
1829 em_set_multi(struct adapter *adapter)
1830 {
1831         struct ifnet *ifp = &adapter->arpcom.ac_if;
1832         struct ifmultiaddr *ifma;
1833         uint32_t reg_rctl = 0;
1834         uint8_t  mta[512]; /* Largest MTS is 4096 bits */
1835         int mcnt = 0;
1836
1837         if (adapter->hw.mac.type == e1000_82542 &&
1838             adapter->hw.revision_id == E1000_REVISION_2) {
1839                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1840                 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1841                         e1000_pci_clear_mwi(&adapter->hw);
1842                 reg_rctl |= E1000_RCTL_RST;
1843                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1844                 msec_delay(5);
1845         }
1846
1847         LIST_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) {
1848                 if (ifma->ifma_addr->sa_family != AF_LINK)
1849                         continue;
1850
1851                 if (mcnt == MAX_NUM_MULTICAST_ADDRESSES)
1852                         break;
1853
1854                 bcopy(LLADDR((struct sockaddr_dl *)ifma->ifma_addr),
1855                     &mta[mcnt * ETHER_ADDR_LEN], ETHER_ADDR_LEN);
1856                 mcnt++;
1857         }
1858
1859         if (mcnt >= MAX_NUM_MULTICAST_ADDRESSES) {
1860                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1861                 reg_rctl |= E1000_RCTL_MPE;
1862                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1863         } else {
1864                 e1000_update_mc_addr_list(&adapter->hw, mta,
1865                     mcnt, 1, adapter->hw.mac.rar_entry_count);
1866         }
1867
1868         if (adapter->hw.mac.type == e1000_82542 &&
1869             adapter->hw.revision_id == E1000_REVISION_2) {
1870                 reg_rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
1871                 reg_rctl &= ~E1000_RCTL_RST;
1872                 E1000_WRITE_REG(&adapter->hw, E1000_RCTL, reg_rctl);
1873                 msec_delay(5);
1874                 if (adapter->hw.bus.pci_cmd_word & CMD_MEM_WRT_INVALIDATE)
1875                         e1000_pci_set_mwi(&adapter->hw);
1876         }
1877 }
1878
1879 /*
1880  * This routine checks for link status and updates statistics.
1881  */
1882 static void
1883 em_timer(void *xsc)
1884 {
1885         struct adapter *adapter = xsc;
1886         struct ifnet *ifp = &adapter->arpcom.ac_if;
1887
1888         lwkt_serialize_enter(ifp->if_serializer);
1889
1890         em_update_link_status(adapter);
1891         em_update_stats(adapter);
1892
1893         /* Reset LAA into RAR[0] on 82571 */
1894         if (e1000_get_laa_state_82571(&adapter->hw) == TRUE)
1895                 e1000_rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
1896
1897         if (em_display_debug_stats && (ifp->if_flags & IFF_RUNNING))
1898                 em_print_hw_stats(adapter);
1899
1900         em_smartspeed(adapter);
1901
1902         callout_reset(&adapter->timer, hz, em_timer, adapter);
1903
1904         lwkt_serialize_exit(ifp->if_serializer);
1905 }
1906
1907 static void
1908 em_update_link_status(struct adapter *adapter)
1909 {
1910         struct e1000_hw *hw = &adapter->hw;
1911         struct ifnet *ifp = &adapter->arpcom.ac_if;
1912         device_t dev = adapter->dev;
1913         uint32_t link_check = 0;
1914
1915         /* Get the cached link value or read phy for real */
1916         switch (hw->phy.media_type) {
1917         case e1000_media_type_copper:
1918                 if (hw->mac.get_link_status) {
1919                         /* Do the work to read phy */
1920                         e1000_check_for_link(hw);
1921                         link_check = !hw->mac.get_link_status;
1922                         if (link_check) /* ESB2 fix */
1923                                 e1000_cfg_on_link_up(hw);
1924                 } else {
1925                         link_check = TRUE;
1926                 }
1927                 break;
1928
1929         case e1000_media_type_fiber:
1930                 e1000_check_for_link(hw);
1931                 link_check =
1932                         E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU;
1933                 break;
1934
1935         case e1000_media_type_internal_serdes:
1936                 e1000_check_for_link(hw);
1937                 link_check = adapter->hw.mac.serdes_has_link;
1938                 break;
1939
1940         case e1000_media_type_unknown:
1941         default:
1942                 break;
1943         }
1944
1945         /* Now check for a transition */
1946         if (link_check && adapter->link_active == 0) {
1947                 e1000_get_speed_and_duplex(hw, &adapter->link_speed,
1948                     &adapter->link_duplex);
1949
1950                 /*
1951                  * Check if we should enable/disable SPEED_MODE bit on
1952                  * 82571/82572
1953                  */
1954                 if (hw->mac.type == e1000_82571 ||
1955                     hw->mac.type == e1000_82572) {
1956                         int tarc0;
1957
1958                         tarc0 = E1000_READ_REG(hw, E1000_TARC(0));
1959                         if (adapter->link_speed != SPEED_1000)
1960                                 tarc0 &= ~SPEED_MODE_BIT;
1961                         else
1962                                 tarc0 |= SPEED_MODE_BIT;
1963                         E1000_WRITE_REG(hw, E1000_TARC(0), tarc0);
1964                 }
1965                 if (bootverbose) {
1966                         device_printf(dev, "Link is up %d Mbps %s\n",
1967                             adapter->link_speed,
1968                             ((adapter->link_duplex == FULL_DUPLEX) ?
1969                             "Full Duplex" : "Half Duplex"));
1970                 }
1971                 adapter->link_active = 1;
1972                 adapter->smartspeed = 0;
1973                 ifp->if_baudrate = adapter->link_speed * 1000000;
1974                 ifp->if_link_state = LINK_STATE_UP;
1975                 if_link_state_change(ifp);
1976         } else if (!link_check && adapter->link_active == 1) {
1977                 ifp->if_baudrate = adapter->link_speed = 0;
1978                 adapter->link_duplex = 0;
1979                 if (bootverbose)
1980                         device_printf(dev, "Link is Down\n");
1981                 adapter->link_active = 0;
1982 #if 0
1983                 /* Link down, disable watchdog */
1984                 if->if_timer = 0;
1985 #endif
1986                 ifp->if_link_state = LINK_STATE_DOWN;
1987                 if_link_state_change(ifp);
1988         }
1989 }
1990
1991 static void
1992 em_stop(struct adapter *adapter)
1993 {
1994         struct ifnet *ifp = &adapter->arpcom.ac_if;
1995         int i;
1996
1997         ASSERT_SERIALIZED(ifp->if_serializer);
1998
1999         em_disable_intr(adapter);
2000
2001         callout_stop(&adapter->timer);
2002         callout_stop(&adapter->tx_fifo_timer);
2003
2004         ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
2005         ifp->if_timer = 0;
2006
2007         e1000_reset_hw(&adapter->hw);
2008         if (adapter->hw.mac.type >= e1000_82544)
2009                 E1000_WRITE_REG(&adapter->hw, E1000_WUC, 0);
2010
2011         for (i = 0; i < adapter->num_tx_desc; i++) {
2012                 struct em_buffer *tx_buffer = &adapter->tx_buffer_area[i];
2013
2014                 if (tx_buffer->m_head != NULL) {
2015                         bus_dmamap_unload(adapter->txtag, tx_buffer->map);
2016                         m_freem(tx_buffer->m_head);
2017                         tx_buffer->m_head = NULL;
2018                 }
2019         }
2020
2021         for (i = 0; i < adapter->num_rx_desc; i++) {
2022                 struct em_buffer *rx_buffer = &adapter->rx_buffer_area[i];
2023
2024                 if (rx_buffer->m_head != NULL) {
2025                         bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
2026                         m_freem(rx_buffer->m_head);
2027                         rx_buffer->m_head = NULL;
2028                 }
2029         }
2030
2031         if (adapter->fmp != NULL)
2032                 m_freem(adapter->fmp);
2033         adapter->fmp = NULL;
2034         adapter->lmp = NULL;
2035
2036         adapter->csum_flags = 0;
2037         adapter->csum_ehlen = 0;
2038         adapter->csum_iphlen = 0;
2039
2040         adapter->tx_dd_head = 0;
2041         adapter->tx_dd_tail = 0;
2042         adapter->tx_nsegs = 0;
2043 }
2044
2045 static int
2046 em_get_hw_info(struct adapter *adapter)
2047 {
2048         device_t dev = adapter->dev;
2049
2050         /* Save off the information about this board */
2051         adapter->hw.vendor_id = pci_get_vendor(dev);
2052         adapter->hw.device_id = pci_get_device(dev);
2053         adapter->hw.revision_id = pci_get_revid(dev);
2054         adapter->hw.subsystem_vendor_id = pci_get_subvendor(dev);
2055         adapter->hw.subsystem_device_id = pci_get_subdevice(dev);
2056
2057         /* Do Shared Code Init and Setup */
2058         if (e1000_set_mac_type(&adapter->hw))
2059                 return ENXIO;
2060         return 0;
2061 }
2062
2063 static int
2064 em_alloc_pci_res(struct adapter *adapter)
2065 {
2066         device_t dev = adapter->dev;
2067         int val, rid, error = E1000_SUCCESS;
2068
2069         /* Enable bus mastering */
2070         pci_enable_busmaster(dev);
2071
2072         adapter->memory_rid = EM_BAR_MEM;
2073         adapter->memory = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
2074                                 &adapter->memory_rid, RF_ACTIVE);
2075         if (adapter->memory == NULL) {
2076                 device_printf(dev, "Unable to allocate bus resource: memory\n");
2077                 return (ENXIO);
2078         }
2079         adapter->osdep.mem_bus_space_tag =
2080             rman_get_bustag(adapter->memory);
2081         adapter->osdep.mem_bus_space_handle =
2082             rman_get_bushandle(adapter->memory);
2083
2084         /* XXX This is quite goofy, it is not actually used */
2085         adapter->hw.hw_addr = (uint8_t *)&adapter->osdep.mem_bus_space_handle;
2086
2087         /* Only older adapters use IO mapping */
2088         if (adapter->hw.mac.type > e1000_82543 &&
2089             adapter->hw.mac.type < e1000_82571) {
2090                 /* Figure our where our IO BAR is ? */
2091                 for (rid = PCIR_BAR(0); rid < PCIR_CARDBUSCIS;) {
2092                         val = pci_read_config(dev, rid, 4);
2093                         if (EM_BAR_TYPE(val) == EM_BAR_TYPE_IO) {
2094                                 adapter->io_rid = rid;
2095                                 break;
2096                         }
2097                         rid += 4;
2098                         /* check for 64bit BAR */
2099                         if (EM_BAR_MEM_TYPE(val) == EM_BAR_MEM_TYPE_64BIT)
2100                                 rid += 4;
2101                 }
2102                 if (rid >= PCIR_CARDBUSCIS) {
2103                         device_printf(dev, "Unable to locate IO BAR\n");
2104                         return (ENXIO);
2105                 }
2106                 adapter->ioport = bus_alloc_resource_any(dev, SYS_RES_IOPORT,
2107                                         &adapter->io_rid, RF_ACTIVE);
2108                 if (adapter->ioport == NULL) {
2109                         device_printf(dev, "Unable to allocate bus resource: "
2110                             "ioport\n");
2111                         return (ENXIO);
2112                 }
2113                 adapter->hw.io_base = 0;
2114                 adapter->osdep.io_bus_space_tag =
2115                     rman_get_bustag(adapter->ioport);
2116                 adapter->osdep.io_bus_space_handle =
2117                     rman_get_bushandle(adapter->ioport);
2118         }
2119
2120         adapter->intr_rid = 0;
2121         adapter->intr_res = bus_alloc_resource_any(dev, SYS_RES_IRQ,
2122                                 &adapter->intr_rid,
2123                                 RF_SHAREABLE | RF_ACTIVE);
2124         if (adapter->intr_res == NULL) {
2125                 device_printf(dev, "Unable to allocate bus resource: "
2126                     "interrupt\n");
2127                 return (ENXIO);
2128         }
2129
2130         adapter->hw.bus.pci_cmd_word = pci_read_config(dev, PCIR_COMMAND, 2);
2131         adapter->hw.back = &adapter->osdep;
2132         return (error);
2133 }
2134
2135 static void
2136 em_free_pci_res(struct adapter *adapter)
2137 {
2138         device_t dev = adapter->dev;
2139
2140         if (adapter->intr_res != NULL) {
2141                 bus_release_resource(dev, SYS_RES_IRQ,
2142                     adapter->intr_rid, adapter->intr_res);
2143         }
2144
2145         if (adapter->memory != NULL) {
2146                 bus_release_resource(dev, SYS_RES_MEMORY,
2147                     adapter->memory_rid, adapter->memory);
2148         }
2149
2150         if (adapter->flash != NULL) {
2151                 bus_release_resource(dev, SYS_RES_MEMORY,
2152                     adapter->flash_rid, adapter->flash);
2153         }
2154
2155         if (adapter->ioport != NULL) {
2156                 bus_release_resource(dev, SYS_RES_IOPORT,
2157                     adapter->io_rid, adapter->ioport);
2158         }
2159 }
2160
2161 static int
2162 em_hw_init(struct adapter *adapter)
2163 {
2164         device_t dev = adapter->dev;
2165         uint16_t rx_buffer_size;
2166
2167         /* Issue a global reset */
2168         e1000_reset_hw(&adapter->hw);
2169
2170         /* Get control from any management/hw control */
2171         if ((adapter->hw.mac.type == e1000_82573 ||
2172              adapter->hw.mac.type == e1000_ich8lan ||
2173              adapter->hw.mac.type == e1000_ich10lan ||
2174              adapter->hw.mac.type == e1000_ich9lan) &&
2175             e1000_check_mng_mode(&adapter->hw))
2176                 em_get_hw_control(adapter);
2177
2178         /* When hardware is reset, fifo_head is also reset */
2179         adapter->tx_fifo_head = 0;
2180
2181         /* Set up smart power down as default off on newer adapters. */
2182         if (!em_smart_pwr_down &&
2183             (adapter->hw.mac.type == e1000_82571 ||
2184              adapter->hw.mac.type == e1000_82572)) {
2185                 uint16_t phy_tmp = 0;
2186
2187                 /* Speed up time to link by disabling smart power down. */
2188                 e1000_read_phy_reg(&adapter->hw,
2189                     IGP02E1000_PHY_POWER_MGMT, &phy_tmp);
2190                 phy_tmp &= ~IGP02E1000_PM_SPD;
2191                 e1000_write_phy_reg(&adapter->hw,
2192                     IGP02E1000_PHY_POWER_MGMT, phy_tmp);
2193         }
2194
2195         /*
2196          * These parameters control the automatic generation (Tx) and
2197          * response (Rx) to Ethernet PAUSE frames.
2198          * - High water mark should allow for at least two frames to be
2199          *   received after sending an XOFF.
2200          * - Low water mark works best when it is very near the high water mark.
2201          *   This allows the receiver to restart by sending XON when it has
2202          *   drained a bit. Here we use an arbitary value of 1500 which will
2203          *   restart after one full frame is pulled from the buffer. There
2204          *   could be several smaller frames in the buffer and if so they will
2205          *   not trigger the XON until their total number reduces the buffer
2206          *   by 1500.
2207          * - The pause time is fairly large at 1000 x 512ns = 512 usec.
2208          */
2209         rx_buffer_size =
2210                 (E1000_READ_REG(&adapter->hw, E1000_PBA) & 0xffff) << 10;
2211
2212         adapter->hw.fc.high_water = rx_buffer_size -
2213                                     roundup2(adapter->max_frame_size, 1024);
2214         adapter->hw.fc.low_water = adapter->hw.fc.high_water - 1500;
2215
2216         if (adapter->hw.mac.type == e1000_80003es2lan)
2217                 adapter->hw.fc.pause_time = 0xFFFF;
2218         else
2219                 adapter->hw.fc.pause_time = EM_FC_PAUSE_TIME;
2220         adapter->hw.fc.send_xon = TRUE;
2221         adapter->hw.fc.requested_mode = e1000_fc_full;
2222
2223         if (e1000_init_hw(&adapter->hw) < 0) {
2224                 device_printf(dev, "Hardware Initialization Failed\n");
2225                 return (EIO);
2226         }
2227
2228         e1000_check_for_link(&adapter->hw);
2229
2230         return (0);
2231 }
2232
2233 static void
2234 em_setup_ifp(struct adapter *adapter)
2235 {
2236         struct ifnet *ifp = &adapter->arpcom.ac_if;
2237
2238         if_initname(ifp, device_get_name(adapter->dev),
2239                     device_get_unit(adapter->dev));
2240         ifp->if_softc = adapter;
2241         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
2242         ifp->if_init =  em_init;
2243         ifp->if_ioctl = em_ioctl;
2244         ifp->if_start = em_start;
2245 #ifdef DEVICE_POLLING
2246         ifp->if_poll = em_poll;
2247 #endif
2248         ifp->if_watchdog = em_watchdog;
2249         ifq_set_maxlen(&ifp->if_snd, adapter->num_tx_desc - 1);
2250         ifq_set_ready(&ifp->if_snd);
2251
2252         ether_ifattach(ifp, adapter->hw.mac.addr, NULL);
2253
2254         if (adapter->hw.mac.type >= e1000_82543)
2255                 ifp->if_capabilities = IFCAP_HWCSUM;
2256
2257         ifp->if_capabilities |= IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU;
2258         ifp->if_capenable = ifp->if_capabilities;
2259
2260         if (ifp->if_capenable & IFCAP_TXCSUM)
2261                 ifp->if_hwassist = EM_CSUM_FEATURES;
2262
2263         /*
2264          * Tell the upper layer(s) we support long frames.
2265          */
2266         ifp->if_data.ifi_hdrlen = sizeof(struct ether_vlan_header);
2267
2268         /*
2269          * Specify the media types supported by this adapter and register
2270          * callbacks to update media and link information
2271          */
2272         ifmedia_init(&adapter->media, IFM_IMASK,
2273                      em_media_change, em_media_status);
2274         if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
2275             adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
2276                 u_char fiber_type = IFM_1000_SX; /* default type */
2277
2278                 if (adapter->hw.mac.type == e1000_82545)
2279                         fiber_type = IFM_1000_LX;
2280                 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type | IFM_FDX,
2281                             0, NULL);
2282                 ifmedia_add(&adapter->media, IFM_ETHER | fiber_type, 0, NULL);
2283         } else {
2284                 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T, 0, NULL);
2285                 ifmedia_add(&adapter->media, IFM_ETHER | IFM_10_T | IFM_FDX,
2286                             0, NULL);
2287                 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX,
2288                             0, NULL);
2289                 ifmedia_add(&adapter->media, IFM_ETHER | IFM_100_TX | IFM_FDX,
2290                             0, NULL);
2291                 if (adapter->hw.phy.type != e1000_phy_ife) {
2292                         ifmedia_add(&adapter->media,
2293                                 IFM_ETHER | IFM_1000_T | IFM_FDX, 0, NULL);
2294                         ifmedia_add(&adapter->media,
2295                                 IFM_ETHER | IFM_1000_T, 0, NULL);
2296                 }
2297         }
2298         ifmedia_add(&adapter->media, IFM_ETHER | IFM_AUTO, 0, NULL);
2299         ifmedia_set(&adapter->media, IFM_ETHER | IFM_AUTO);
2300 }
2301
2302
2303 /*
2304  * Workaround for SmartSpeed on 82541 and 82547 controllers
2305  */
2306 static void
2307 em_smartspeed(struct adapter *adapter)
2308 {
2309         uint16_t phy_tmp;
2310
2311         if (adapter->link_active || adapter->hw.phy.type != e1000_phy_igp ||
2312             adapter->hw.mac.autoneg == 0 ||
2313             (adapter->hw.phy.autoneg_advertised & ADVERTISE_1000_FULL) == 0)
2314                 return;
2315
2316         if (adapter->smartspeed == 0) {
2317                 /*
2318                  * If Master/Slave config fault is asserted twice,
2319                  * we assume back-to-back
2320                  */
2321                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2322                 if (!(phy_tmp & SR_1000T_MS_CONFIG_FAULT))
2323                         return;
2324                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_STATUS, &phy_tmp);
2325                 if (phy_tmp & SR_1000T_MS_CONFIG_FAULT) {
2326                         e1000_read_phy_reg(&adapter->hw,
2327                             PHY_1000T_CTRL, &phy_tmp);
2328                         if (phy_tmp & CR_1000T_MS_ENABLE) {
2329                                 phy_tmp &= ~CR_1000T_MS_ENABLE;
2330                                 e1000_write_phy_reg(&adapter->hw,
2331                                     PHY_1000T_CTRL, phy_tmp);
2332                                 adapter->smartspeed++;
2333                                 if (adapter->hw.mac.autoneg &&
2334                                     !e1000_phy_setup_autoneg(&adapter->hw) &&
2335                                     !e1000_read_phy_reg(&adapter->hw,
2336                                      PHY_CONTROL, &phy_tmp)) {
2337                                         phy_tmp |= MII_CR_AUTO_NEG_EN |
2338                                                    MII_CR_RESTART_AUTO_NEG;
2339                                         e1000_write_phy_reg(&adapter->hw,
2340                                             PHY_CONTROL, phy_tmp);
2341                                 }
2342                         }
2343                 }
2344                 return;
2345         } else if (adapter->smartspeed == EM_SMARTSPEED_DOWNSHIFT) {
2346                 /* If still no link, perhaps using 2/3 pair cable */
2347                 e1000_read_phy_reg(&adapter->hw, PHY_1000T_CTRL, &phy_tmp);
2348                 phy_tmp |= CR_1000T_MS_ENABLE;
2349                 e1000_write_phy_reg(&adapter->hw, PHY_1000T_CTRL, phy_tmp);
2350                 if (adapter->hw.mac.autoneg &&
2351                     !e1000_phy_setup_autoneg(&adapter->hw) &&
2352                     !e1000_read_phy_reg(&adapter->hw, PHY_CONTROL, &phy_tmp)) {
2353                         phy_tmp |= MII_CR_AUTO_NEG_EN | MII_CR_RESTART_AUTO_NEG;
2354                         e1000_write_phy_reg(&adapter->hw, PHY_CONTROL, phy_tmp);
2355                 }
2356         }
2357
2358         /* Restart process after EM_SMARTSPEED_MAX iterations */
2359         if (adapter->smartspeed++ == EM_SMARTSPEED_MAX)
2360                 adapter->smartspeed = 0;
2361 }
2362
2363 static int
2364 em_dma_malloc(struct adapter *adapter, bus_size_t size,
2365               struct em_dma_alloc *dma)
2366 {
2367         dma->dma_vaddr = bus_dmamem_coherent_any(adapter->parent_dtag,
2368                                 EM_DBA_ALIGN, size, BUS_DMA_WAITOK,
2369                                 &dma->dma_tag, &dma->dma_map,
2370                                 &dma->dma_paddr);
2371         if (dma->dma_vaddr == NULL)
2372                 return ENOMEM;
2373         else
2374                 return 0;
2375 }
2376
2377 static void
2378 em_dma_free(struct adapter *adapter, struct em_dma_alloc *dma)
2379 {
2380         if (dma->dma_tag == NULL)
2381                 return;
2382         bus_dmamap_unload(dma->dma_tag, dma->dma_map);
2383         bus_dmamem_free(dma->dma_tag, dma->dma_vaddr, dma->dma_map);
2384         bus_dma_tag_destroy(dma->dma_tag);
2385 }
2386
2387 static int
2388 em_create_tx_ring(struct adapter *adapter)
2389 {
2390         device_t dev = adapter->dev;
2391         struct em_buffer *tx_buffer;
2392         int error, i;
2393
2394         adapter->tx_buffer_area =
2395                 kmalloc(sizeof(struct em_buffer) * adapter->num_tx_desc,
2396                         M_DEVBUF, M_WAITOK | M_ZERO);
2397
2398         /*
2399          * Create DMA tags for tx buffers
2400          */
2401         error = bus_dma_tag_create(adapter->parent_dtag, /* parent */
2402                         1, 0,                   /* alignment, bounds */
2403                         BUS_SPACE_MAXADDR,      /* lowaddr */
2404                         BUS_SPACE_MAXADDR,      /* highaddr */
2405                         NULL, NULL,             /* filter, filterarg */
2406                         EM_TSO_SIZE,            /* maxsize */
2407                         EM_MAX_SCATTER,         /* nsegments */
2408                         EM_MAX_SEGSIZE,         /* maxsegsize */
2409                         BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW |
2410                         BUS_DMA_ONEBPAGE,       /* flags */
2411                         &adapter->txtag);
2412         if (error) {
2413                 device_printf(dev, "Unable to allocate TX DMA tag\n");
2414                 kfree(adapter->tx_buffer_area, M_DEVBUF);
2415                 adapter->tx_buffer_area = NULL;
2416                 return error;
2417         }
2418
2419         /*
2420          * Create DMA maps for tx buffers
2421          */
2422         for (i = 0; i < adapter->num_tx_desc; i++) {
2423                 tx_buffer = &adapter->tx_buffer_area[i];
2424
2425                 error = bus_dmamap_create(adapter->txtag,
2426                                           BUS_DMA_WAITOK | BUS_DMA_ONEBPAGE,
2427                                           &tx_buffer->map);
2428                 if (error) {
2429                         device_printf(dev, "Unable to create TX DMA map\n");
2430                         em_destroy_tx_ring(adapter, i);
2431                         return error;
2432                 }
2433         }
2434         return (0);
2435 }
2436
2437 static void
2438 em_init_tx_ring(struct adapter *adapter)
2439 {
2440         /* Clear the old ring contents */
2441         bzero(adapter->tx_desc_base,
2442             (sizeof(struct e1000_tx_desc)) * adapter->num_tx_desc);
2443
2444         /* Reset state */
2445         adapter->next_avail_tx_desc = 0;
2446         adapter->next_tx_to_clean = 0;
2447         adapter->num_tx_desc_avail = adapter->num_tx_desc;
2448 }
2449
2450 static void
2451 em_init_tx_unit(struct adapter *adapter)
2452 {
2453         uint32_t tctl, tarc, tipg = 0;
2454         uint64_t bus_addr;
2455
2456         /* Setup the Base and Length of the Tx Descriptor Ring */
2457         bus_addr = adapter->txdma.dma_paddr;
2458         E1000_WRITE_REG(&adapter->hw, E1000_TDLEN(0),
2459             adapter->num_tx_desc * sizeof(struct e1000_tx_desc));
2460         E1000_WRITE_REG(&adapter->hw, E1000_TDBAH(0),
2461             (uint32_t)(bus_addr >> 32));
2462         E1000_WRITE_REG(&adapter->hw, E1000_TDBAL(0),
2463             (uint32_t)bus_addr);
2464         /* Setup the HW Tx Head and Tail descriptor pointers */
2465         E1000_WRITE_REG(&adapter->hw, E1000_TDT(0), 0);
2466         E1000_WRITE_REG(&adapter->hw, E1000_TDH(0), 0);
2467
2468         /* Set the default values for the Tx Inter Packet Gap timer */
2469         switch (adapter->hw.mac.type) {
2470         case e1000_82542:
2471                 tipg = DEFAULT_82542_TIPG_IPGT;
2472                 tipg |= DEFAULT_82542_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2473                 tipg |= DEFAULT_82542_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2474                 break;
2475
2476         case e1000_80003es2lan:
2477                 tipg = DEFAULT_82543_TIPG_IPGR1;
2478                 tipg |= DEFAULT_80003ES2LAN_TIPG_IPGR2 <<
2479                     E1000_TIPG_IPGR2_SHIFT;
2480                 break;
2481
2482         default:
2483                 if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
2484                     adapter->hw.phy.media_type ==
2485                     e1000_media_type_internal_serdes)
2486                         tipg = DEFAULT_82543_TIPG_IPGT_FIBER;
2487                 else
2488                         tipg = DEFAULT_82543_TIPG_IPGT_COPPER;
2489                 tipg |= DEFAULT_82543_TIPG_IPGR1 << E1000_TIPG_IPGR1_SHIFT;
2490                 tipg |= DEFAULT_82543_TIPG_IPGR2 << E1000_TIPG_IPGR2_SHIFT;
2491                 break;
2492         }
2493
2494         E1000_WRITE_REG(&adapter->hw, E1000_TIPG, tipg);
2495
2496         /* NOTE: 0 is not allowed for TIDV */
2497         E1000_WRITE_REG(&adapter->hw, E1000_TIDV, 1);
2498         if(adapter->hw.mac.type >= e1000_82540)
2499                 E1000_WRITE_REG(&adapter->hw, E1000_TADV, 0);
2500
2501         if (adapter->hw.mac.type == e1000_82571 ||
2502             adapter->hw.mac.type == e1000_82572) {
2503                 tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
2504                 tarc |= SPEED_MODE_BIT;
2505                 E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2506         } else if (adapter->hw.mac.type == e1000_80003es2lan) {
2507                 tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(0));
2508                 tarc |= 1;
2509                 E1000_WRITE_REG(&adapter->hw, E1000_TARC(0), tarc);
2510                 tarc = E1000_READ_REG(&adapter->hw, E1000_TARC(1));
2511                 tarc |= 1;
2512                 E1000_WRITE_REG(&adapter->hw, E1000_TARC(1), tarc);
2513         }
2514
2515         /* Program the Transmit Control Register */
2516         tctl = E1000_READ_REG(&adapter->hw, E1000_TCTL);
2517         tctl &= ~E1000_TCTL_CT;
2518         tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC | E1000_TCTL_EN |
2519                 (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2520
2521         if (adapter->hw.mac.type >= e1000_82571)
2522                 tctl |= E1000_TCTL_MULR;
2523
2524         /* This write will effectively turn on the transmit unit. */
2525         E1000_WRITE_REG(&adapter->hw, E1000_TCTL, tctl);
2526 }
2527
2528 static void
2529 em_destroy_tx_ring(struct adapter *adapter, int ndesc)
2530 {
2531         struct em_buffer *tx_buffer;
2532         int i;
2533
2534         if (adapter->tx_buffer_area == NULL)
2535                 return;
2536
2537         for (i = 0; i < ndesc; i++) {
2538                 tx_buffer = &adapter->tx_buffer_area[i];
2539
2540                 KKASSERT(tx_buffer->m_head == NULL);
2541                 bus_dmamap_destroy(adapter->txtag, tx_buffer->map);
2542         }
2543         bus_dma_tag_destroy(adapter->txtag);
2544
2545         kfree(adapter->tx_buffer_area, M_DEVBUF);
2546         adapter->tx_buffer_area = NULL;
2547 }
2548
2549 /*
2550  * The offload context needs to be set when we transfer the first
2551  * packet of a particular protocol (TCP/UDP).  This routine has been
2552  * enhanced to deal with inserted VLAN headers.
2553  *
2554  * If the new packet's ether header length, ip header length and
2555  * csum offloading type are same as the previous packet, we should
2556  * avoid allocating a new csum context descriptor; mainly to take
2557  * advantage of the pipeline effect of the TX data read request.
2558  *
2559  * This function returns number of TX descrptors allocated for
2560  * csum context.
2561  */
2562 static int
2563 em_txcsum(struct adapter *adapter, struct mbuf *mp,
2564           uint32_t *txd_upper, uint32_t *txd_lower)
2565 {
2566         struct e1000_context_desc *TXD;
2567         struct em_buffer *tx_buffer;
2568         struct ether_vlan_header *eh;
2569         struct ip *ip;
2570         int curr_txd, ehdrlen, csum_flags;
2571         uint32_t cmd, hdr_len, ip_hlen;
2572         uint16_t etype;
2573
2574         /*
2575          * Determine where frame payload starts.
2576          * Jump over vlan headers if already present,
2577          * helpful for QinQ too.
2578          */
2579         KASSERT(mp->m_len >= ETHER_HDR_LEN,
2580                 ("em_txcsum_pullup is not called (eh)?\n"));
2581         eh = mtod(mp, struct ether_vlan_header *);
2582         if (eh->evl_encap_proto == htons(ETHERTYPE_VLAN)) {
2583                 KASSERT(mp->m_len >= ETHER_HDR_LEN + EVL_ENCAPLEN,
2584                         ("em_txcsum_pullup is not called (evh)?\n"));
2585                 etype = ntohs(eh->evl_proto);
2586                 ehdrlen = ETHER_HDR_LEN + EVL_ENCAPLEN;
2587         } else {
2588                 etype = ntohs(eh->evl_encap_proto);
2589                 ehdrlen = ETHER_HDR_LEN;
2590         }
2591
2592         /*
2593          * We only support TCP/UDP for IPv4 for the moment.
2594          * TODO: Support SCTP too when it hits the tree.
2595          */
2596         if (etype != ETHERTYPE_IP)
2597                 return 0;
2598
2599         KASSERT(mp->m_len >= ehdrlen + EM_IPVHL_SIZE,
2600                 ("em_txcsum_pullup is not called (eh+ip_vhl)?\n"));
2601
2602         /* NOTE: We could only safely access ip.ip_vhl part */
2603         ip = (struct ip *)(mp->m_data + ehdrlen);
2604         ip_hlen = ip->ip_hl << 2;
2605
2606         csum_flags = mp->m_pkthdr.csum_flags & EM_CSUM_FEATURES;
2607
2608         if (adapter->csum_ehlen == ehdrlen &&
2609             adapter->csum_iphlen == ip_hlen &&
2610             adapter->csum_flags == csum_flags) {
2611                 /*
2612                  * Same csum offload context as the previous packets;
2613                  * just return.
2614                  */
2615                 *txd_upper = adapter->csum_txd_upper;
2616                 *txd_lower = adapter->csum_txd_lower;
2617                 return 0;
2618         }
2619
2620         /*
2621          * Setup a new csum offload context.
2622          */
2623
2624         curr_txd = adapter->next_avail_tx_desc;
2625         tx_buffer = &adapter->tx_buffer_area[curr_txd];
2626         TXD = (struct e1000_context_desc *)&adapter->tx_desc_base[curr_txd];
2627
2628         cmd = 0;
2629
2630         /* Setup of IP header checksum. */
2631         if (csum_flags & CSUM_IP) {
2632                 /*
2633                  * Start offset for header checksum calculation.
2634                  * End offset for header checksum calculation.
2635                  * Offset of place to put the checksum.
2636                  */
2637                 TXD->lower_setup.ip_fields.ipcss = ehdrlen;
2638                 TXD->lower_setup.ip_fields.ipcse =
2639                     htole16(ehdrlen + ip_hlen - 1);
2640                 TXD->lower_setup.ip_fields.ipcso =
2641                     ehdrlen + offsetof(struct ip, ip_sum);
2642                 cmd |= E1000_TXD_CMD_IP;
2643                 *txd_upper |= E1000_TXD_POPTS_IXSM << 8;
2644         }
2645         hdr_len = ehdrlen + ip_hlen;
2646
2647         if (csum_flags & CSUM_TCP) {
2648                 /*
2649                  * Start offset for payload checksum calculation.
2650                  * End offset for payload checksum calculation.
2651                  * Offset of place to put the checksum.
2652                  */
2653                 TXD->upper_setup.tcp_fields.tucss = hdr_len;
2654                 TXD->upper_setup.tcp_fields.tucse = htole16(0);
2655                 TXD->upper_setup.tcp_fields.tucso =
2656                     hdr_len + offsetof(struct tcphdr, th_sum);
2657                 cmd |= E1000_TXD_CMD_TCP;
2658                 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
2659         } else if (csum_flags & CSUM_UDP) {
2660                 /*
2661                  * Start offset for header checksum calculation.
2662                  * End offset for header checksum calculation.
2663                  * Offset of place to put the checksum.
2664                  */
2665                 TXD->upper_setup.tcp_fields.tucss = hdr_len;
2666                 TXD->upper_setup.tcp_fields.tucse = htole16(0);
2667                 TXD->upper_setup.tcp_fields.tucso =
2668                     hdr_len + offsetof(struct udphdr, uh_sum);
2669                 *txd_upper |= E1000_TXD_POPTS_TXSM << 8;
2670         }
2671
2672         *txd_lower = E1000_TXD_CMD_DEXT |       /* Extended descr type */
2673                      E1000_TXD_DTYP_D;          /* Data descr */
2674
2675         /* Save the information for this csum offloading context */
2676         adapter->csum_ehlen = ehdrlen;
2677         adapter->csum_iphlen = ip_hlen;
2678         adapter->csum_flags = csum_flags;
2679         adapter->csum_txd_upper = *txd_upper;
2680         adapter->csum_txd_lower = *txd_lower;
2681
2682         TXD->tcp_seg_setup.data = htole32(0);
2683         TXD->cmd_and_length =
2684             htole32(E1000_TXD_CMD_IFCS | E1000_TXD_CMD_DEXT | cmd);
2685         tx_buffer->m_head = NULL;
2686
2687         if (++curr_txd == adapter->num_tx_desc)
2688                 curr_txd = 0;
2689
2690         KKASSERT(adapter->num_tx_desc_avail > 0);
2691         adapter->num_tx_desc_avail--;
2692
2693         adapter->next_avail_tx_desc = curr_txd;
2694         return 1;
2695 }
2696
2697 static int
2698 em_txcsum_pullup(struct adapter *adapter, struct mbuf **m0)
2699 {
2700         struct mbuf *m = *m0;
2701         struct ether_header *eh;
2702         int len;
2703
2704         adapter->tx_csum_try_pullup++;
2705
2706         len = ETHER_HDR_LEN + EM_IPVHL_SIZE;
2707
2708         if (__predict_false(!M_WRITABLE(m))) {
2709                 if (__predict_false(m->m_len < ETHER_HDR_LEN)) {
2710                         adapter->tx_csum_drop1++;
2711                         m_freem(m);
2712                         *m0 = NULL;
2713                         return ENOBUFS;
2714                 }
2715                 eh = mtod(m, struct ether_header *);
2716
2717                 if (eh->ether_type == htons(ETHERTYPE_VLAN))
2718                         len += EVL_ENCAPLEN;
2719
2720                 if (__predict_false(m->m_len < len)) {
2721                         adapter->tx_csum_drop2++;
2722                         m_freem(m);
2723                         *m0 = NULL;
2724                         return ENOBUFS;
2725                 }
2726                 return 0;
2727         }
2728
2729         if (__predict_false(m->m_len < ETHER_HDR_LEN)) {
2730                 adapter->tx_csum_pullup1++;
2731                 m = m_pullup(m, ETHER_HDR_LEN);
2732                 if (m == NULL) {
2733                         adapter->tx_csum_pullup1_failed++;
2734                         *m0 = NULL;
2735                         return ENOBUFS;
2736                 }
2737                 *m0 = m;
2738         }
2739         eh = mtod(m, struct ether_header *);
2740
2741         if (eh->ether_type == htons(ETHERTYPE_VLAN))
2742                 len += EVL_ENCAPLEN;
2743
2744         if (__predict_false(m->m_len < len)) {
2745                 adapter->tx_csum_pullup2++;
2746                 m = m_pullup(m, len);
2747                 if (m == NULL) {
2748                         adapter->tx_csum_pullup2_failed++;
2749                         *m0 = NULL;
2750                         return ENOBUFS;
2751                 }
2752                 *m0 = m;
2753         }
2754         return 0;
2755 }
2756
2757 static void
2758 em_txeof(struct adapter *adapter)
2759 {
2760         struct ifnet *ifp = &adapter->arpcom.ac_if;
2761         struct e1000_tx_desc *tx_desc;
2762         struct em_buffer *tx_buffer;
2763         int first, num_avail;
2764
2765         if (adapter->tx_dd_head == adapter->tx_dd_tail)
2766                 return;
2767
2768         if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
2769                 return;
2770
2771         num_avail = adapter->num_tx_desc_avail;
2772         first = adapter->next_tx_to_clean;
2773
2774         while (adapter->tx_dd_head != adapter->tx_dd_tail) {
2775                 int dd_idx = adapter->tx_dd[adapter->tx_dd_head];
2776
2777                 tx_desc = &adapter->tx_desc_base[dd_idx];
2778
2779                 if (tx_desc->upper.fields.status & E1000_TXD_STAT_DD) {
2780                         EM_INC_TXDD_IDX(adapter->tx_dd_head);
2781
2782                         if (++dd_idx == adapter->num_tx_desc)
2783                                 dd_idx = 0;
2784
2785                         while (first != dd_idx) {
2786                                 tx_buffer = &adapter->tx_buffer_area[first];
2787                                 tx_desc = &adapter->tx_desc_base[first];
2788
2789                                 tx_desc->upper.data = 0;
2790                                 tx_desc->lower.data = 0;
2791                                 tx_desc->buffer_addr = 0;
2792                                 num_avail++;
2793
2794                                 if (tx_buffer->m_head) {
2795                                         ifp->if_opackets++;
2796                                         bus_dmamap_unload(adapter->txtag,
2797                                                           tx_buffer->map);
2798                                         m_freem(tx_buffer->m_head);
2799                                         tx_buffer->m_head = NULL;
2800                                 }
2801
2802                                 if (++first == adapter->num_tx_desc)
2803                                         first = 0;
2804                         }
2805                 } else {
2806                         break;
2807                 }
2808         }
2809         adapter->next_tx_to_clean = first;
2810         adapter->num_tx_desc_avail = num_avail;
2811
2812         if (adapter->tx_dd_head == adapter->tx_dd_tail) {
2813                 adapter->tx_dd_head = 0;
2814                 adapter->tx_dd_tail = 0;
2815         }
2816
2817         if (!EM_IS_OACTIVE(adapter)) {
2818                 ifp->if_flags &= ~IFF_OACTIVE;
2819
2820                 /* All clean, turn off the timer */
2821                 if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
2822                         ifp->if_timer = 0;
2823         }
2824 }
2825
2826 static void
2827 em_tx_collect(struct adapter *adapter)
2828 {
2829         struct ifnet *ifp = &adapter->arpcom.ac_if;
2830         struct e1000_tx_desc *tx_desc;
2831         struct em_buffer *tx_buffer;
2832         int tdh, first, num_avail, dd_idx = -1;
2833
2834         if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
2835                 return;
2836
2837         tdh = E1000_READ_REG(&adapter->hw, E1000_TDH(0));
2838         if (tdh == adapter->next_tx_to_clean)
2839                 return;
2840
2841         if (adapter->tx_dd_head != adapter->tx_dd_tail)
2842                 dd_idx = adapter->tx_dd[adapter->tx_dd_head];
2843
2844         num_avail = adapter->num_tx_desc_avail;
2845         first = adapter->next_tx_to_clean;
2846
2847         while (first != tdh) {
2848                 tx_buffer = &adapter->tx_buffer_area[first];
2849                 tx_desc = &adapter->tx_desc_base[first];
2850
2851                 tx_desc->upper.data = 0;
2852                 tx_desc->lower.data = 0;
2853                 tx_desc->buffer_addr = 0;
2854                 num_avail++;
2855
2856                 if (tx_buffer->m_head) {
2857                         ifp->if_opackets++;
2858                         bus_dmamap_unload(adapter->txtag,
2859                                           tx_buffer->map);
2860                         m_freem(tx_buffer->m_head);
2861                         tx_buffer->m_head = NULL;
2862                 }
2863
2864                 if (first == dd_idx) {
2865                         EM_INC_TXDD_IDX(adapter->tx_dd_head);
2866                         if (adapter->tx_dd_head == adapter->tx_dd_tail) {
2867                                 adapter->tx_dd_head = 0;
2868                                 adapter->tx_dd_tail = 0;
2869                                 dd_idx = -1;
2870                         } else {
2871                                 dd_idx = adapter->tx_dd[adapter->tx_dd_head];
2872                         }
2873                 }
2874
2875                 if (++first == adapter->num_tx_desc)
2876                         first = 0;
2877         }
2878         adapter->next_tx_to_clean = first;
2879         adapter->num_tx_desc_avail = num_avail;
2880
2881         if (!EM_IS_OACTIVE(adapter)) {
2882                 ifp->if_flags &= ~IFF_OACTIVE;
2883
2884                 /* All clean, turn off the timer */
2885                 if (adapter->num_tx_desc_avail == adapter->num_tx_desc)
2886                         ifp->if_timer = 0;
2887         }
2888 }
2889
2890 /*
2891  * When Link is lost sometimes there is work still in the TX ring
2892  * which will result in a watchdog, rather than allow that do an
2893  * attempted cleanup and then reinit here.  Note that this has been
2894  * seens mostly with fiber adapters.
2895  */
2896 static void
2897 em_tx_purge(struct adapter *adapter)
2898 {
2899         struct ifnet *ifp = &adapter->arpcom.ac_if;
2900
2901         if (!adapter->link_active && ifp->if_timer) {
2902                 em_tx_collect(adapter);
2903                 if (ifp->if_timer) {
2904                         if_printf(ifp, "Link lost, TX pending, reinit\n");
2905                         ifp->if_timer = 0;
2906                         em_init(adapter);
2907                 }
2908         }
2909 }
2910
2911 static int
2912 em_newbuf(struct adapter *adapter, int i, int init)
2913 {
2914         struct mbuf *m;
2915         bus_dma_segment_t seg;
2916         bus_dmamap_t map;
2917         struct em_buffer *rx_buffer;
2918         int error, nseg;
2919
2920         m = m_getcl(init ? MB_WAIT : MB_DONTWAIT, MT_DATA, M_PKTHDR);
2921         if (m == NULL) {
2922                 adapter->mbuf_cluster_failed++;
2923                 if (init) {
2924                         if_printf(&adapter->arpcom.ac_if,
2925                                   "Unable to allocate RX mbuf\n");
2926                 }
2927                 return (ENOBUFS);
2928         }
2929         m->m_len = m->m_pkthdr.len = MCLBYTES;
2930
2931         if (adapter->max_frame_size <= MCLBYTES - ETHER_ALIGN)
2932                 m_adj(m, ETHER_ALIGN);
2933
2934         error = bus_dmamap_load_mbuf_segment(adapter->rxtag,
2935                         adapter->rx_sparemap, m,
2936                         &seg, 1, &nseg, BUS_DMA_NOWAIT);
2937         if (error) {
2938                 m_freem(m);
2939                 if (init) {
2940                         if_printf(&adapter->arpcom.ac_if,
2941                                   "Unable to load RX mbuf\n");
2942                 }
2943                 return (error);
2944         }
2945
2946         rx_buffer = &adapter->rx_buffer_area[i];
2947         if (rx_buffer->m_head != NULL)
2948                 bus_dmamap_unload(adapter->rxtag, rx_buffer->map);
2949
2950         map = rx_buffer->map;
2951         rx_buffer->map = adapter->rx_sparemap;
2952         adapter->rx_sparemap = map;
2953
2954         rx_buffer->m_head = m;
2955
2956         adapter->rx_desc_base[i].buffer_addr = htole64(seg.ds_addr);
2957         return (0);
2958 }
2959
2960 static int
2961 em_create_rx_ring(struct adapter *adapter)
2962 {
2963         device_t dev = adapter->dev;
2964         struct em_buffer *rx_buffer;
2965         int i, error;
2966
2967         adapter->rx_buffer_area =
2968                 kmalloc(sizeof(struct em_buffer) * adapter->num_rx_desc,
2969                         M_DEVBUF, M_WAITOK | M_ZERO);
2970
2971         /*
2972          * Create DMA tag for rx buffers
2973          */
2974         error = bus_dma_tag_create(adapter->parent_dtag, /* parent */
2975                         1, 0,                   /* alignment, bounds */
2976                         BUS_SPACE_MAXADDR,      /* lowaddr */
2977                         BUS_SPACE_MAXADDR,      /* highaddr */
2978                         NULL, NULL,             /* filter, filterarg */
2979                         MCLBYTES,               /* maxsize */
2980                         1,                      /* nsegments */
2981                         MCLBYTES,               /* maxsegsize */
2982                         BUS_DMA_WAITOK | BUS_DMA_ALLOCNOW, /* flags */
2983                         &adapter->rxtag);
2984         if (error) {
2985                 device_printf(dev, "Unable to allocate RX DMA tag\n");
2986                 kfree(adapter->rx_buffer_area, M_DEVBUF);
2987                 adapter->rx_buffer_area = NULL;
2988                 return error;
2989         }
2990
2991         /*
2992          * Create spare DMA map for rx buffers
2993          */
2994         error = bus_dmamap_create(adapter->rxtag, BUS_DMA_WAITOK,
2995                                   &adapter->rx_sparemap);
2996         if (error) {
2997                 device_printf(dev, "Unable to create spare RX DMA map\n");
2998                 bus_dma_tag_destroy(adapter->rxtag);
2999                 kfree(adapter->rx_buffer_area, M_DEVBUF);
3000                 adapter->rx_buffer_area = NULL;
3001                 return error;
3002         }
3003
3004         /*
3005          * Create DMA maps for rx buffers
3006          */
3007         for (i = 0; i < adapter->num_rx_desc; i++) {
3008                 rx_buffer = &adapter->rx_buffer_area[i];
3009
3010                 error = bus_dmamap_create(adapter->rxtag, BUS_DMA_WAITOK,
3011                                           &rx_buffer->map);
3012                 if (error) {
3013                         device_printf(dev, "Unable to create RX DMA map\n");
3014                         em_destroy_rx_ring(adapter, i);
3015                         return error;
3016                 }
3017         }
3018         return (0);
3019 }
3020
3021 static int
3022 em_init_rx_ring(struct adapter *adapter)
3023 {
3024         int i, error;
3025
3026         /* Reset descriptor ring */
3027         bzero(adapter->rx_desc_base,
3028             (sizeof(struct e1000_rx_desc)) * adapter->num_rx_desc);
3029
3030         /* Allocate new ones. */
3031         for (i = 0; i < adapter->num_rx_desc; i++) {
3032                 error = em_newbuf(adapter, i, 1);
3033                 if (error)
3034                         return (error);
3035         }
3036
3037         /* Setup our descriptor pointers */
3038         adapter->next_rx_desc_to_check = 0;
3039
3040         return (0);
3041 }
3042
3043 static void
3044 em_init_rx_unit(struct adapter *adapter)
3045 {
3046         struct ifnet *ifp = &adapter->arpcom.ac_if;
3047         uint64_t bus_addr;
3048         uint32_t rctl, rxcsum;
3049
3050         /*
3051          * Make sure receives are disabled while setting
3052          * up the descriptor ring
3053          */
3054         rctl = E1000_READ_REG(&adapter->hw, E1000_RCTL);
3055         E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl & ~E1000_RCTL_EN);
3056
3057         if (adapter->hw.mac.type >= e1000_82540) {
3058                 /*
3059                  * Set the interrupt throttling rate. Value is calculated
3060                  * as ITR = 1 / (INT_THROTTLE_CEIL * 256ns)
3061                  */
3062                 if (adapter->int_throttle_ceil) {
3063                         E1000_WRITE_REG(&adapter->hw, E1000_ITR,
3064                                 1000000000 / 256 / adapter->int_throttle_ceil);
3065                 } else {
3066                         E1000_WRITE_REG(&adapter->hw, E1000_ITR, 0);
3067                 }
3068         }
3069
3070         /* Disable accelerated ackknowledge */
3071         if (adapter->hw.mac.type == e1000_82574) {
3072                 E1000_WRITE_REG(&adapter->hw,
3073                     E1000_RFCTL, E1000_RFCTL_ACK_DIS);
3074         }
3075
3076         /* Setup the Base and Length of the Rx Descriptor Ring */
3077         bus_addr = adapter->rxdma.dma_paddr;
3078         E1000_WRITE_REG(&adapter->hw, E1000_RDLEN(0),
3079             adapter->num_rx_desc * sizeof(struct e1000_rx_desc));
3080         E1000_WRITE_REG(&adapter->hw, E1000_RDBAH(0),
3081             (uint32_t)(bus_addr >> 32));
3082         E1000_WRITE_REG(&adapter->hw, E1000_RDBAL(0),
3083             (uint32_t)bus_addr);
3084
3085         /* Setup the Receive Control Register */
3086         rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3087         rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_LBM_NO |
3088                 E1000_RCTL_RDMTS_HALF |
3089                 (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3090
3091         /* Make sure VLAN Filters are off */
3092         rctl &= ~E1000_RCTL_VFE;
3093
3094         if (e1000_tbi_sbp_enabled_82543(&adapter->hw))
3095                 rctl |= E1000_RCTL_SBP;
3096         else
3097                 rctl &= ~E1000_RCTL_SBP;
3098
3099         switch (adapter->rx_buffer_len) {
3100         default:
3101         case 2048:
3102                 rctl |= E1000_RCTL_SZ_2048;
3103                 break;
3104
3105         case 4096:
3106                 rctl |= E1000_RCTL_SZ_4096 |
3107                     E1000_RCTL_BSEX | E1000_RCTL_LPE;
3108                 break;
3109
3110         case 8192:
3111                 rctl |= E1000_RCTL_SZ_8192 |
3112                     E1000_RCTL_BSEX | E1000_RCTL_LPE;
3113                 break;
3114
3115         case 16384:
3116                 rctl |= E1000_RCTL_SZ_16384 |
3117                     E1000_RCTL_BSEX | E1000_RCTL_LPE;
3118                 break;
3119         }
3120
3121         if (ifp->if_mtu > ETHERMTU)
3122                 rctl |= E1000_RCTL_LPE;
3123         else
3124                 rctl &= ~E1000_RCTL_LPE;
3125
3126         /* Receive Checksum Offload for TCP and UDP */
3127         if (ifp->if_capenable & IFCAP_RXCSUM) {
3128                 rxcsum = E1000_READ_REG(&adapter->hw, E1000_RXCSUM);
3129                 rxcsum |= (E1000_RXCSUM_IPOFL | E1000_RXCSUM_TUOFL);
3130                 E1000_WRITE_REG(&adapter->hw, E1000_RXCSUM, rxcsum);
3131         }
3132
3133         /*
3134          * XXX TEMPORARY WORKAROUND: on some systems with 82573
3135          * long latencies are observed, like Lenovo X60. This
3136          * change eliminates the problem, but since having positive
3137          * values in RDTR is a known source of problems on other
3138          * platforms another solution is being sought.
3139          */
3140         if (em_82573_workaround && adapter->hw.mac.type == e1000_82573) {
3141                 E1000_WRITE_REG(&adapter->hw, E1000_RADV, EM_RADV_82573);
3142                 E1000_WRITE_REG(&adapter->hw, E1000_RDTR, EM_RDTR_82573);
3143         }
3144
3145         /* Enable Receives */
3146         E1000_WRITE_REG(&adapter->hw, E1000_RCTL, rctl);
3147
3148         /*
3149          * Setup the HW Rx Head and Tail Descriptor Pointers
3150          */
3151         E1000_WRITE_REG(&adapter->hw, E1000_RDH(0), 0);
3152         E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), adapter->num_rx_desc - 1);
3153 }
3154
3155 static void
3156 em_destroy_rx_ring(struct adapter *adapter, int ndesc)
3157 {
3158         struct em_buffer *rx_buffer;
3159         int i;
3160
3161         if (adapter->rx_buffer_area == NULL)
3162                 return;
3163
3164         for (i = 0; i < ndesc; i++) {
3165                 rx_buffer = &adapter->rx_buffer_area[i];
3166
3167                 KKASSERT(rx_buffer->m_head == NULL);
3168                 bus_dmamap_destroy(adapter->rxtag, rx_buffer->map);
3169         }
3170         bus_dmamap_destroy(adapter->rxtag, adapter->rx_sparemap);
3171         bus_dma_tag_destroy(adapter->rxtag);
3172
3173         kfree(adapter->rx_buffer_area, M_DEVBUF);
3174         adapter->rx_buffer_area = NULL;
3175 }
3176
3177 static void
3178 em_rxeof(struct adapter *adapter, int count)
3179 {
3180         struct ifnet *ifp = &adapter->arpcom.ac_if;
3181         uint8_t status, accept_frame = 0, eop = 0;
3182         uint16_t len, desc_len, prev_len_adj;
3183         struct e1000_rx_desc *current_desc;
3184         struct mbuf *mp;
3185         int i;
3186         struct mbuf_chain chain[MAXCPU];
3187
3188         i = adapter->next_rx_desc_to_check;
3189         current_desc = &adapter->rx_desc_base[i];
3190
3191         if (!(current_desc->status & E1000_RXD_STAT_DD))
3192                 return;
3193
3194         ether_input_chain_init(chain);
3195
3196         while ((current_desc->status & E1000_RXD_STAT_DD) && count != 0) {
3197                 struct mbuf *m = NULL;
3198
3199                 logif(pkt_receive);
3200
3201                 mp = adapter->rx_buffer_area[i].m_head;
3202
3203                 /*
3204                  * Can't defer bus_dmamap_sync(9) because TBI_ACCEPT
3205                  * needs to access the last received byte in the mbuf.
3206                  */
3207                 bus_dmamap_sync(adapter->rxtag, adapter->rx_buffer_area[i].map,
3208                                 BUS_DMASYNC_POSTREAD);
3209
3210                 accept_frame = 1;
3211                 prev_len_adj = 0;
3212                 desc_len = le16toh(current_desc->length);
3213                 status = current_desc->status;
3214                 if (status & E1000_RXD_STAT_EOP) {
3215                         count--;
3216                         eop = 1;
3217                         if (desc_len < ETHER_CRC_LEN) {
3218                                 len = 0;
3219                                 prev_len_adj = ETHER_CRC_LEN - desc_len;
3220                         } else {
3221                                 len = desc_len - ETHER_CRC_LEN;
3222                         }
3223                 } else {
3224                         eop = 0;
3225                         len = desc_len;
3226                 }
3227
3228                 if (current_desc->errors & E1000_RXD_ERR_FRAME_ERR_MASK) {
3229                         uint8_t last_byte;
3230                         uint32_t pkt_len = desc_len;
3231
3232                         if (adapter->fmp != NULL)
3233                                 pkt_len += adapter->fmp->m_pkthdr.len;
3234
3235                         last_byte = *(mtod(mp, caddr_t) + desc_len - 1);
3236                         if (TBI_ACCEPT(&adapter->hw, status,
3237                             current_desc->errors, pkt_len, last_byte,
3238                             adapter->min_frame_size, adapter->max_frame_size)) {
3239                                 e1000_tbi_adjust_stats_82543(&adapter->hw,
3240                                     &adapter->stats, pkt_len,
3241                                     adapter->hw.mac.addr,
3242                                     adapter->max_frame_size);
3243                                 if (len > 0)
3244                                         len--;
3245                         } else {
3246                                 accept_frame = 0;
3247                         }
3248                 }
3249
3250                 if (accept_frame) {
3251                         if (em_newbuf(adapter, i, 0) != 0) {
3252                                 ifp->if_iqdrops++;
3253                                 goto discard;
3254                         }
3255
3256                         /* Assign correct length to the current fragment */
3257                         mp->m_len = len;
3258
3259                         if (adapter->fmp == NULL) {
3260                                 mp->m_pkthdr.len = len;
3261                                 adapter->fmp = mp; /* Store the first mbuf */
3262                                 adapter->lmp = mp;
3263                         } else {
3264                                 /*
3265                                  * Chain mbuf's together
3266                                  */
3267
3268                                 /*
3269                                  * Adjust length of previous mbuf in chain if
3270                                  * we received less than 4 bytes in the last
3271                                  * descriptor.
3272                                  */
3273                                 if (prev_len_adj > 0) {
3274                                         adapter->lmp->m_len -= prev_len_adj;
3275                                         adapter->fmp->m_pkthdr.len -=
3276                                             prev_len_adj;
3277                                 }
3278                                 adapter->lmp->m_next = mp;
3279                                 adapter->lmp = adapter->lmp->m_next;
3280                                 adapter->fmp->m_pkthdr.len += len;
3281                         }
3282
3283                         if (eop) {
3284                                 adapter->fmp->m_pkthdr.rcvif = ifp;
3285                                 ifp->if_ipackets++;
3286
3287                                 if (ifp->if_capenable & IFCAP_RXCSUM) {
3288                                         em_rxcsum(adapter, current_desc,
3289                                                   adapter->fmp);
3290                                 }
3291
3292                                 if (status & E1000_RXD_STAT_VP) {
3293                                         adapter->fmp->m_pkthdr.ether_vlantag =
3294                                             (le16toh(current_desc->special) &
3295                                             E1000_RXD_SPC_VLAN_MASK);
3296                                         adapter->fmp->m_flags |= M_VLANTAG;
3297                                 }
3298                                 m = adapter->fmp;
3299                                 adapter->fmp = NULL;
3300                                 adapter->lmp = NULL;
3301                         }
3302                 } else {
3303                         ifp->if_ierrors++;
3304 discard:
3305 #ifdef foo
3306                         /* Reuse loaded DMA map and just update mbuf chain */
3307                         mp = adapter->rx_buffer_area[i].m_head;
3308                         mp->m_len = mp->m_pkthdr.len = MCLBYTES;
3309                         mp->m_data = mp->m_ext.ext_buf;
3310                         mp->m_next = NULL;
3311                         if (adapter->max_frame_size <= (MCLBYTES - ETHER_ALIGN))
3312                                 m_adj(mp, ETHER_ALIGN);
3313 #endif
3314                         if (adapter->fmp != NULL) {
3315                                 m_freem(adapter->fmp);
3316                                 adapter->fmp = NULL;
3317                                 adapter->lmp = NULL;
3318                         }
3319                         m = NULL;
3320                 }
3321
3322                 /* Zero out the receive descriptors status. */
3323                 current_desc->status = 0;
3324
3325                 if (m != NULL)
3326                         ether_input_chain(ifp, m, chain);
3327
3328                 /* Advance our pointers to the next descriptor. */
3329                 if (++i == adapter->num_rx_desc)
3330                         i = 0;
3331                 current_desc = &adapter->rx_desc_base[i];
3332         }
3333         adapter->next_rx_desc_to_check = i;
3334
3335         ether_input_dispatch(chain);
3336
3337         /* Advance the E1000's Receive Queue #0  "Tail Pointer". */
3338         if (--i < 0)
3339                 i = adapter->num_rx_desc - 1;
3340         E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), i);
3341 }
3342
3343 static void
3344 em_rxcsum(struct adapter *adapter, struct e1000_rx_desc *rx_desc,
3345           struct mbuf *mp)
3346 {
3347         /* 82543 or newer only */
3348         if (adapter->hw.mac.type < e1000_82543 ||
3349             /* Ignore Checksum bit is set */
3350             (rx_desc->status & E1000_RXD_STAT_IXSM))
3351                 return;
3352
3353         if ((rx_desc->status & E1000_RXD_STAT_IPCS) &&
3354             !(rx_desc->errors & E1000_RXD_ERR_IPE)) {
3355                 /* IP Checksum Good */
3356                 mp->m_pkthdr.csum_flags |= CSUM_IP_CHECKED | CSUM_IP_VALID;
3357         }
3358
3359         if ((rx_desc->status & E1000_RXD_STAT_TCPCS) &&
3360             !(rx_desc->errors & E1000_RXD_ERR_TCPE)) {
3361                 mp->m_pkthdr.csum_flags |= CSUM_DATA_VALID |
3362                                            CSUM_PSEUDO_HDR |
3363                                            CSUM_FRAG_NOT_CHECKED;
3364                 mp->m_pkthdr.csum_data = htons(0xffff);
3365         }
3366 }
3367
3368 static void
3369 em_enable_intr(struct adapter *adapter)
3370 {
3371         lwkt_serialize_handler_enable(adapter->arpcom.ac_if.if_serializer);
3372         E1000_WRITE_REG(&adapter->hw, E1000_IMS, IMS_ENABLE_MASK);
3373 }
3374
3375 static void
3376 em_disable_intr(struct adapter *adapter)
3377 {
3378         uint32_t clear = 0xffffffff;
3379
3380         /*
3381          * The first version of 82542 had an errata where when link was forced
3382          * it would stay up even up even if the cable was disconnected.
3383          * Sequence errors were used to detect the disconnect and then the
3384          * driver would unforce the link.  This code in the in the ISR.  For
3385          * this to work correctly the Sequence error interrupt had to be
3386          * enabled all the time.
3387          */
3388         if (adapter->hw.mac.type == e1000_82542 &&
3389             adapter->hw.revision_id == E1000_REVISION_2)
3390                 clear &= ~E1000_IMC_RXSEQ;
3391
3392         E1000_WRITE_REG(&adapter->hw, E1000_IMC, clear);
3393
3394         lwkt_serialize_handler_disable(adapter->arpcom.ac_if.if_serializer);
3395 }
3396
3397 /*
3398  * Bit of a misnomer, what this really means is
3399  * to enable OS management of the system... aka
3400  * to disable special hardware management features
3401  */
3402 static void
3403 em_get_mgmt(struct adapter *adapter)
3404 {
3405         /* A shared code workaround */
3406 #define E1000_82542_MANC2H E1000_MANC2H
3407         if (adapter->has_manage) {
3408                 int manc2h = E1000_READ_REG(&adapter->hw, E1000_MANC2H);
3409                 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3410
3411                 /* disable hardware interception of ARP */
3412                 manc &= ~(E1000_MANC_ARP_EN);
3413
3414                 /* enable receiving management packets to the host */
3415                 if (adapter->hw.mac.type >= e1000_82571) {
3416                         manc |= E1000_MANC_EN_MNG2HOST;
3417 #define E1000_MNG2HOST_PORT_623 (1 << 5)
3418 #define E1000_MNG2HOST_PORT_664 (1 << 6)
3419                         manc2h |= E1000_MNG2HOST_PORT_623;
3420                         manc2h |= E1000_MNG2HOST_PORT_664;
3421                         E1000_WRITE_REG(&adapter->hw, E1000_MANC2H, manc2h);
3422                 }
3423
3424                 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3425         }
3426 }
3427
3428 /*
3429  * Give control back to hardware management
3430  * controller if there is one.
3431  */
3432 static void
3433 em_rel_mgmt(struct adapter *adapter)
3434 {
3435         if (adapter->has_manage) {
3436                 int manc = E1000_READ_REG(&adapter->hw, E1000_MANC);
3437
3438                 /* re-enable hardware interception of ARP */
3439                 manc |= E1000_MANC_ARP_EN;
3440
3441                 if (adapter->hw.mac.type >= e1000_82571)
3442                         manc &= ~E1000_MANC_EN_MNG2HOST;
3443
3444                 E1000_WRITE_REG(&adapter->hw, E1000_MANC, manc);
3445         }
3446 }
3447
3448 /*
3449  * em_get_hw_control() sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3450  * For ASF and Pass Through versions of f/w this means that
3451  * the driver is loaded.  For AMT version (only with 82573)
3452  * of the f/w this means that the network i/f is open.
3453  */
3454 static void
3455 em_get_hw_control(struct adapter *adapter)
3456 {
3457         uint32_t ctrl_ext, swsm;
3458
3459         /* Let firmware know the driver has taken over */
3460         switch (adapter->hw.mac.type) {
3461         case e1000_82573:
3462                 swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3463                 E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3464                     swsm | E1000_SWSM_DRV_LOAD);
3465                 break;
3466         case e1000_82571:
3467         case e1000_82572:
3468         case e1000_80003es2lan:
3469         case e1000_ich8lan:
3470         case e1000_ich9lan:
3471         case e1000_ich10lan:
3472                 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3473                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3474                     ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
3475                 break;
3476         default:
3477                 break;
3478         }
3479 }
3480
3481 /*
3482  * em_rel_hw_control() resets {CTRL_EXT|FWSM}:DRV_LOAD bit.
3483  * For ASF and Pass Through versions of f/w this means that the
3484  * driver is no longer loaded.  For AMT version (only with 82573)
3485  * of the f/w this means that the network i/f is closed.
3486  */
3487 static void
3488 em_rel_hw_control(struct adapter *adapter)
3489 {
3490         uint32_t ctrl_ext, swsm;
3491
3492         /* Let firmware taken over control of h/w */
3493         switch (adapter->hw.mac.type) {
3494         case e1000_82573:
3495                 swsm = E1000_READ_REG(&adapter->hw, E1000_SWSM);
3496                 E1000_WRITE_REG(&adapter->hw, E1000_SWSM,
3497                     swsm & ~E1000_SWSM_DRV_LOAD);
3498                 break;
3499
3500         case e1000_82571:
3501         case e1000_82572:
3502         case e1000_80003es2lan:
3503         case e1000_ich8lan:
3504         case e1000_ich9lan:
3505         case e1000_ich10lan:
3506                 ctrl_ext = E1000_READ_REG(&adapter->hw, E1000_CTRL_EXT);
3507                 E1000_WRITE_REG(&adapter->hw, E1000_CTRL_EXT,
3508                     ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
3509                 break;
3510
3511         default:
3512                 break;
3513         }
3514 }
3515
3516 static int
3517 em_is_valid_eaddr(const uint8_t *addr)
3518 {
3519         char zero_addr[ETHER_ADDR_LEN] = { 0, 0, 0, 0, 0, 0 };
3520
3521         if ((addr[0] & 1) || !bcmp(addr, zero_addr, ETHER_ADDR_LEN))
3522                 return (FALSE);
3523
3524         return (TRUE);
3525 }
3526
3527 /*
3528  * Enable PCI Wake On Lan capability
3529  */
3530 void
3531 em_enable_wol(device_t dev)
3532 {
3533         uint16_t cap, status;
3534         uint8_t id;
3535
3536         /* First find the capabilities pointer*/
3537         cap = pci_read_config(dev, PCIR_CAP_PTR, 2);
3538
3539         /* Read the PM Capabilities */
3540         id = pci_read_config(dev, cap, 1);
3541         if (id != PCIY_PMG)     /* Something wrong */
3542                 return;
3543
3544         /*
3545          * OK, we have the power capabilities,
3546          * so now get the status register
3547          */
3548         cap += PCIR_POWER_STATUS;
3549         status = pci_read_config(dev, cap, 2);
3550         status |= PCIM_PSTAT_PME | PCIM_PSTAT_PMEENABLE;
3551         pci_write_config(dev, cap, status, 2);
3552 }
3553
3554
3555 /*
3556  * 82544 Coexistence issue workaround.
3557  *    There are 2 issues.
3558  *       1. Transmit Hang issue.
3559  *    To detect this issue, following equation can be used...
3560  *        SIZE[3:0] + ADDR[2:0] = SUM[3:0].
3561  *        If SUM[3:0] is in between 1 to 4, we will have this issue.
3562  *
3563  *       2. DAC issue.
3564  *    To detect this issue, following equation can be used...
3565  *        SIZE[3:0] + ADDR[2:0] = SUM[3:0].
3566  *        If SUM[3:0] is in between 9 to c, we will have this issue.
3567  *
3568  *    WORKAROUND:
3569  *        Make sure we do not have ending address
3570  *        as 1,2,3,4(Hang) or 9,a,b,c (DAC)
3571  */
3572 static uint32_t
3573 em_82544_fill_desc(bus_addr_t address, uint32_t length, PDESC_ARRAY desc_array)
3574 {
3575         uint32_t safe_terminator;
3576
3577         /*
3578          * Since issue is sensitive to length and address.
3579          * Let us first check the address...
3580          */
3581         if (length <= 4) {
3582                 desc_array->descriptor[0].address = address;
3583                 desc_array->descriptor[0].length = length;
3584                 desc_array->elements = 1;
3585                 return (desc_array->elements);
3586         }
3587
3588         safe_terminator =
3589         (uint32_t)((((uint32_t)address & 0x7) + (length & 0xF)) & 0xF);
3590
3591         /* If it does not fall between 0x1 to 0x4 and 0x9 to 0xC then return */
3592         if (safe_terminator == 0 ||
3593             (safe_terminator > 4 && safe_terminator < 9) ||
3594             (safe_terminator > 0xC && safe_terminator <= 0xF)) {
3595                 desc_array->descriptor[0].address = address;
3596                 desc_array->descriptor[0].length = length;
3597                 desc_array->elements = 1;
3598                 return (desc_array->elements);
3599         }
3600
3601         desc_array->descriptor[0].address = address;
3602         desc_array->descriptor[0].length = length - 4;
3603         desc_array->descriptor[1].address = address + (length - 4);
3604         desc_array->descriptor[1].length = 4;
3605         desc_array->elements = 2;
3606         return (desc_array->elements);
3607 }
3608
3609 static void
3610 em_update_stats(struct adapter *adapter)
3611 {
3612         struct ifnet *ifp = &adapter->arpcom.ac_if;
3613
3614         if (adapter->hw.phy.media_type == e1000_media_type_copper ||
3615             (E1000_READ_REG(&adapter->hw, E1000_STATUS) & E1000_STATUS_LU)) {
3616                 adapter->stats.symerrs +=
3617                         E1000_READ_REG(&adapter->hw, E1000_SYMERRS);
3618                 adapter->stats.sec += E1000_READ_REG(&adapter->hw, E1000_SEC);
3619         }
3620         adapter->stats.crcerrs += E1000_READ_REG(&adapter->hw, E1000_CRCERRS);
3621         adapter->stats.mpc += E1000_READ_REG(&adapter->hw, E1000_MPC);
3622         adapter->stats.scc += E1000_READ_REG(&adapter->hw, E1000_SCC);
3623         adapter->stats.ecol += E1000_READ_REG(&adapter->hw, E1000_ECOL);
3624
3625         adapter->stats.mcc += E1000_READ_REG(&adapter->hw, E1000_MCC);
3626         adapter->stats.latecol += E1000_READ_REG(&adapter->hw, E1000_LATECOL);
3627         adapter->stats.colc += E1000_READ_REG(&adapter->hw, E1000_COLC);
3628         adapter->stats.dc += E1000_READ_REG(&adapter->hw, E1000_DC);
3629         adapter->stats.rlec += E1000_READ_REG(&adapter->hw, E1000_RLEC);
3630         adapter->stats.xonrxc += E1000_READ_REG(&adapter->hw, E1000_XONRXC);
3631         adapter->stats.xontxc += E1000_READ_REG(&adapter->hw, E1000_XONTXC);
3632         adapter->stats.xoffrxc += E1000_READ_REG(&adapter->hw, E1000_XOFFRXC);
3633         adapter->stats.xofftxc += E1000_READ_REG(&adapter->hw, E1000_XOFFTXC);
3634         adapter->stats.fcruc += E1000_READ_REG(&adapter->hw, E1000_FCRUC);
3635         adapter->stats.prc64 += E1000_READ_REG(&adapter->hw, E1000_PRC64);
3636         adapter->stats.prc127 += E1000_READ_REG(&adapter->hw, E1000_PRC127);
3637         adapter->stats.prc255 += E1000_READ_REG(&adapter->hw, E1000_PRC255);
3638         adapter->stats.prc511 += E1000_READ_REG(&adapter->hw, E1000_PRC511);
3639         adapter->stats.prc1023 += E1000_READ_REG(&adapter->hw, E1000_PRC1023);
3640         adapter->stats.prc1522 += E1000_READ_REG(&adapter->hw, E1000_PRC1522);
3641         adapter->stats.gprc += E1000_READ_REG(&adapter->hw, E1000_GPRC);
3642         adapter->stats.bprc += E1000_READ_REG(&adapter->hw, E1000_BPRC);
3643         adapter->stats.mprc += E1000_READ_REG(&adapter->hw, E1000_MPRC);
3644         adapter->stats.gptc += E1000_READ_REG(&adapter->hw, E1000_GPTC);
3645
3646         /* For the 64-bit byte counters the low dword must be read first. */
3647         /* Both registers clear on the read of the high dword */
3648
3649         adapter->stats.gorc += E1000_READ_REG(&adapter->hw, E1000_GORCH);
3650         adapter->stats.gotc += E1000_READ_REG(&adapter->hw, E1000_GOTCH);
3651
3652         adapter->stats.rnbc += E1000_READ_REG(&adapter->hw, E1000_RNBC);
3653         adapter->stats.ruc += E1000_READ_REG(&adapter->hw, E1000_RUC);
3654         adapter->stats.rfc += E1000_READ_REG(&adapter->hw, E1000_RFC);
3655         adapter->stats.roc += E1000_READ_REG(&adapter->hw, E1000_ROC);
3656         adapter->stats.rjc += E1000_READ_REG(&adapter->hw, E1000_RJC);
3657
3658         adapter->stats.tor += E1000_READ_REG(&adapter->hw, E1000_TORH);
3659         adapter->stats.tot += E1000_READ_REG(&adapter->hw, E1000_TOTH);
3660
3661         adapter->stats.tpr += E1000_READ_REG(&adapter->hw, E1000_TPR);
3662         adapter->stats.tpt += E1000_READ_REG(&adapter->hw, E1000_TPT);
3663         adapter->stats.ptc64 += E1000_READ_REG(&adapter->hw, E1000_PTC64);
3664         adapter->stats.ptc127 += E1000_READ_REG(&adapter->hw, E1000_PTC127);
3665         adapter->stats.ptc255 += E1000_READ_REG(&adapter->hw, E1000_PTC255);
3666         adapter->stats.ptc511 += E1000_READ_REG(&adapter->hw, E1000_PTC511);
3667         adapter->stats.ptc1023 += E1000_READ_REG(&adapter->hw, E1000_PTC1023);
3668         adapter->stats.ptc1522 += E1000_READ_REG(&adapter->hw, E1000_PTC1522);
3669         adapter->stats.mptc += E1000_READ_REG(&adapter->hw, E1000_MPTC);
3670         adapter->stats.bptc += E1000_READ_REG(&adapter->hw, E1000_BPTC);
3671
3672         if (adapter->hw.mac.type >= e1000_82543) {
3673                 adapter->stats.algnerrc +=
3674                 E1000_READ_REG(&adapter->hw, E1000_ALGNERRC);
3675                 adapter->stats.rxerrc +=
3676                 E1000_READ_REG(&adapter->hw, E1000_RXERRC);
3677                 adapter->stats.tncrs +=
3678                 E1000_READ_REG(&adapter->hw, E1000_TNCRS);
3679                 adapter->stats.cexterr +=
3680                 E1000_READ_REG(&adapter->hw, E1000_CEXTERR);
3681                 adapter->stats.tsctc +=
3682                 E1000_READ_REG(&adapter->hw, E1000_TSCTC);
3683                 adapter->stats.tsctfc +=
3684                 E1000_READ_REG(&adapter->hw, E1000_TSCTFC);
3685         }
3686
3687         ifp->if_collisions = adapter->stats.colc;
3688
3689         /* Rx Errors */
3690         ifp->if_ierrors =
3691             adapter->dropped_pkts + adapter->stats.rxerrc +
3692             adapter->stats.crcerrs + adapter->stats.algnerrc +
3693             adapter->stats.ruc + adapter->stats.roc +
3694             adapter->stats.mpc + adapter->stats.cexterr;
3695
3696         /* Tx Errors */
3697         ifp->if_oerrors =
3698             adapter->stats.ecol + adapter->stats.latecol +
3699             adapter->watchdog_events;
3700 }
3701
3702 static void
3703 em_print_debug_info(struct adapter *adapter)
3704 {
3705         device_t dev = adapter->dev;
3706         uint8_t *hw_addr = adapter->hw.hw_addr;
3707
3708         device_printf(dev, "Adapter hardware address = %p \n", hw_addr);
3709         device_printf(dev, "CTRL = 0x%x RCTL = 0x%x \n",
3710             E1000_READ_REG(&adapter->hw, E1000_CTRL),
3711             E1000_READ_REG(&adapter->hw, E1000_RCTL));
3712         device_printf(dev, "Packet buffer = Tx=%dk Rx=%dk \n",
3713             ((E1000_READ_REG(&adapter->hw, E1000_PBA) & 0xffff0000) >> 16),\
3714             (E1000_READ_REG(&adapter->hw, E1000_PBA) & 0xffff) );
3715         device_printf(dev, "Flow control watermarks high = %d low = %d\n",
3716             adapter->hw.fc.high_water,
3717             adapter->hw.fc.low_water);
3718         device_printf(dev, "tx_int_delay = %d, tx_abs_int_delay = %d\n",
3719             E1000_READ_REG(&adapter->hw, E1000_TIDV),
3720             E1000_READ_REG(&adapter->hw, E1000_TADV));
3721         device_printf(dev, "rx_int_delay = %d, rx_abs_int_delay = %d\n",
3722             E1000_READ_REG(&adapter->hw, E1000_RDTR),
3723             E1000_READ_REG(&adapter->hw, E1000_RADV));
3724         device_printf(dev, "fifo workaround = %lld, fifo_reset_count = %lld\n",
3725             (long long)adapter->tx_fifo_wrk_cnt,
3726             (long long)adapter->tx_fifo_reset_cnt);
3727         device_printf(dev, "hw tdh = %d, hw tdt = %d\n",
3728             E1000_READ_REG(&adapter->hw, E1000_TDH(0)),
3729             E1000_READ_REG(&adapter->hw, E1000_TDT(0)));
3730         device_printf(dev, "hw rdh = %d, hw rdt = %d\n",
3731             E1000_READ_REG(&adapter->hw, E1000_RDH(0)),
3732             E1000_READ_REG(&adapter->hw, E1000_RDT(0)));
3733         device_printf(dev, "Num Tx descriptors avail = %d\n",
3734             adapter->num_tx_desc_avail);
3735         device_printf(dev, "Tx Descriptors not avail1 = %ld\n",
3736             adapter->no_tx_desc_avail1);
3737         device_printf(dev, "Tx Descriptors not avail2 = %ld\n",
3738             adapter->no_tx_desc_avail2);
3739         device_printf(dev, "Std mbuf failed = %ld\n",
3740             adapter->mbuf_alloc_failed);
3741         device_printf(dev, "Std mbuf cluster failed = %ld\n",
3742             adapter->mbuf_cluster_failed);
3743         device_printf(dev, "Driver dropped packets = %ld\n",
3744             adapter->dropped_pkts);
3745         device_printf(dev, "Driver tx dma failure in encap = %ld\n",
3746             adapter->no_tx_dma_setup);
3747
3748         device_printf(dev, "TXCSUM try pullup = %lu\n",
3749             adapter->tx_csum_try_pullup);
3750         device_printf(dev, "TXCSUM m_pullup(eh) called = %lu\n",
3751             adapter->tx_csum_pullup1);
3752         device_printf(dev, "TXCSUM m_pullup(eh) failed = %lu\n",
3753             adapter->tx_csum_pullup1_failed);
3754         device_printf(dev, "TXCSUM m_pullup(eh+ip) called = %lu\n",
3755             adapter->tx_csum_pullup2);
3756         device_printf(dev, "TXCSUM m_pullup(eh+ip) failed = %lu\n",
3757             adapter->tx_csum_pullup2_failed);
3758         device_printf(dev, "TXCSUM non-writable(eh) droped = %lu\n",
3759             adapter->tx_csum_drop1);
3760         device_printf(dev, "TXCSUM non-writable(eh+ip) droped = %lu\n",
3761             adapter->tx_csum_drop2);
3762 }
3763
3764 static void
3765 em_print_hw_stats(struct adapter *adapter)
3766 {
3767         device_t dev = adapter->dev;
3768
3769         device_printf(dev, "Excessive collisions = %lld\n",
3770             (long long)adapter->stats.ecol);
3771 #if (DEBUG_HW > 0)  /* Dont output these errors normally */
3772         device_printf(dev, "Symbol errors = %lld\n",
3773             (long long)adapter->stats.symerrs);
3774 #endif
3775         device_printf(dev, "Sequence errors = %lld\n",
3776             (long long)adapter->stats.sec);
3777         device_printf(dev, "Defer count = %lld\n",
3778             (long long)adapter->stats.dc);
3779         device_printf(dev, "Missed Packets = %lld\n",
3780             (long long)adapter->stats.mpc);
3781         device_printf(dev, "Receive No Buffers = %lld\n",
3782             (long long)adapter->stats.rnbc);
3783         /* RLEC is inaccurate on some hardware, calculate our own. */
3784         device_printf(dev, "Receive Length Errors = %lld\n",
3785             ((long long)adapter->stats.roc + (long long)adapter->stats.ruc));
3786         device_printf(dev, "Receive errors = %lld\n",
3787             (long long)adapter->stats.rxerrc);
3788         device_printf(dev, "Crc errors = %lld\n",
3789             (long long)adapter->stats.crcerrs);
3790         device_printf(dev, "Alignment errors = %lld\n",
3791             (long long)adapter->stats.algnerrc);
3792         device_printf(dev, "Collision/Carrier extension errors = %lld\n",
3793             (long long)adapter->stats.cexterr);
3794         device_printf(dev, "RX overruns = %ld\n", adapter->rx_overruns);
3795         device_printf(dev, "watchdog timeouts = %ld\n",
3796             adapter->watchdog_events);
3797         device_printf(dev, "XON Rcvd = %lld\n",
3798             (long long)adapter->stats.xonrxc);
3799         device_printf(dev, "XON Xmtd = %lld\n",
3800             (long long)adapter->stats.xontxc);
3801         device_printf(dev, "XOFF Rcvd = %lld\n",
3802             (long long)adapter->stats.xoffrxc);
3803         device_printf(dev, "XOFF Xmtd = %lld\n",
3804             (long long)adapter->stats.xofftxc);
3805         device_printf(dev, "Good Packets Rcvd = %lld\n",
3806             (long long)adapter->stats.gprc);
3807         device_printf(dev, "Good Packets Xmtd = %lld\n",
3808             (long long)adapter->stats.gptc);
3809 }
3810
3811 static void
3812 em_print_nvm_info(struct adapter *adapter)
3813 {
3814         uint16_t        eeprom_data;
3815         int     i, j, row = 0;
3816
3817         /* Its a bit crude, but it gets the job done */
3818         kprintf("\nInterface EEPROM Dump:\n");
3819         kprintf("Offset\n0x0000  ");
3820         for (i = 0, j = 0; i < 32; i++, j++) {
3821                 if (j == 8) { /* Make the offset block */
3822                         j = 0; ++row;
3823                         kprintf("\n0x00%x0  ",row);
3824                 }
3825                 e1000_read_nvm(&adapter->hw, i, 1, &eeprom_data);
3826                 kprintf("%04x ", eeprom_data);
3827         }
3828         kprintf("\n");
3829 }
3830
3831 static int
3832 em_sysctl_debug_info(SYSCTL_HANDLER_ARGS)
3833 {
3834         struct adapter *adapter;
3835         struct ifnet *ifp;
3836         int error, result;
3837
3838         result = -1;
3839         error = sysctl_handle_int(oidp, &result, 0, req);
3840         if (error || !req->newptr)
3841                 return (error);
3842
3843         adapter = (struct adapter *)arg1;
3844         ifp = &adapter->arpcom.ac_if;
3845
3846         lwkt_serialize_enter(ifp->if_serializer);
3847
3848         if (result == 1)
3849                 em_print_debug_info(adapter);
3850
3851         /*
3852          * This value will cause a hex dump of the
3853          * first 32 16-bit words of the EEPROM to
3854          * the screen.
3855          */
3856         if (result == 2)
3857                 em_print_nvm_info(adapter);
3858
3859         lwkt_serialize_exit(ifp->if_serializer);
3860
3861         return (error);
3862 }
3863
3864 static int
3865 em_sysctl_stats(SYSCTL_HANDLER_ARGS)
3866 {
3867         int error, result;
3868
3869         result = -1;
3870         error = sysctl_handle_int(oidp, &result, 0, req);
3871         if (error || !req->newptr)
3872                 return (error);
3873
3874         if (result == 1) {
3875                 struct adapter *adapter = (struct adapter *)arg1;
3876                 struct ifnet *ifp = &adapter->arpcom.ac_if;
3877
3878                 lwkt_serialize_enter(ifp->if_serializer);
3879                 em_print_hw_stats(adapter);
3880                 lwkt_serialize_exit(ifp->if_serializer);
3881         }
3882         return (error);
3883 }
3884
3885 static void
3886 em_add_sysctl(struct adapter *adapter)
3887 {
3888 #ifdef PROFILE_SERIALIZER
3889         struct ifnet *ifp = &adapter->arpcom.ac_if;
3890 #endif
3891
3892         sysctl_ctx_init(&adapter->sysctl_ctx);
3893         adapter->sysctl_tree = SYSCTL_ADD_NODE(&adapter->sysctl_ctx,
3894                                         SYSCTL_STATIC_CHILDREN(_hw), OID_AUTO,
3895                                         device_get_nameunit(adapter->dev),
3896                                         CTLFLAG_RD, 0, "");
3897         if (adapter->sysctl_tree == NULL) {
3898                 device_printf(adapter->dev, "can't add sysctl node\n");
3899         } else {
3900                 SYSCTL_ADD_PROC(&adapter->sysctl_ctx,
3901                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3902                     OID_AUTO, "debug", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
3903                     em_sysctl_debug_info, "I", "Debug Information");
3904
3905                 SYSCTL_ADD_PROC(&adapter->sysctl_ctx,
3906                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3907                     OID_AUTO, "stats", CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
3908                     em_sysctl_stats, "I", "Statistics");
3909
3910                 SYSCTL_ADD_INT(&adapter->sysctl_ctx,
3911                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3912                     OID_AUTO, "rxd", CTLFLAG_RD,
3913                     &adapter->num_rx_desc, 0, NULL);
3914                 SYSCTL_ADD_INT(&adapter->sysctl_ctx,
3915                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3916                     OID_AUTO, "txd", CTLFLAG_RD,
3917                     &adapter->num_tx_desc, 0, NULL);
3918
3919 #ifdef PROFILE_SERIALIZER
3920                 SYSCTL_ADD_UINT(&adapter->sysctl_ctx,
3921                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3922                     OID_AUTO, "serializer_sleep", CTLFLAG_RW,
3923                     &ifp->if_serializer->sleep_cnt, 0, NULL);
3924                 SYSCTL_ADD_UINT(&adapter->sysctl_ctx,
3925                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3926                     OID_AUTO, "serializer_tryfail", CTLFLAG_RW,
3927                     &ifp->if_serializer->tryfail_cnt, 0, NULL);
3928                 SYSCTL_ADD_UINT(&adapter->sysctl_ctx,
3929                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3930                     OID_AUTO, "serializer_enter", CTLFLAG_RW,
3931                     &ifp->if_serializer->enter_cnt, 0, NULL);
3932                 SYSCTL_ADD_UINT(&adapter->sysctl_ctx,
3933                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3934                     OID_AUTO, "serializer_try", CTLFLAG_RW,
3935                     &ifp->if_serializer->try_cnt, 0, NULL);
3936 #endif
3937                 if (adapter->hw.mac.type >= e1000_82540) {
3938                         SYSCTL_ADD_PROC(&adapter->sysctl_ctx,
3939                             SYSCTL_CHILDREN(adapter->sysctl_tree),
3940                             OID_AUTO, "int_throttle_ceil",
3941                             CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
3942                             em_sysctl_int_throttle, "I",
3943                             "interrupt throttling rate");
3944                 }
3945                 SYSCTL_ADD_PROC(&adapter->sysctl_ctx,
3946                     SYSCTL_CHILDREN(adapter->sysctl_tree),
3947                     OID_AUTO, "int_tx_nsegs",
3948                     CTLTYPE_INT|CTLFLAG_RW, adapter, 0,
3949                     em_sysctl_int_tx_nsegs, "I",
3950                     "# segments per TX interrupt");
3951         }
3952 }
3953
3954 static int
3955 em_sysctl_int_throttle(SYSCTL_HANDLER_ARGS)
3956 {
3957         struct adapter *adapter = (void *)arg1;
3958         struct ifnet *ifp = &adapter->arpcom.ac_if;
3959         int error, throttle;
3960
3961         throttle = adapter->int_throttle_ceil;
3962         error = sysctl_handle_int(oidp, &throttle, 0, req);
3963         if (error || req->newptr == NULL)
3964                 return error;
3965         if (throttle < 0 || throttle > 1000000000 / 256)
3966                 return EINVAL;
3967
3968         if (throttle) {
3969                 /*
3970                  * Set the interrupt throttling rate in 256ns increments,
3971                  * recalculate sysctl value assignment to get exact frequency.
3972                  */
3973                 throttle = 1000000000 / 256 / throttle;
3974
3975                 /* Upper 16bits of ITR is reserved and should be zero */
3976                 if (throttle & 0xffff0000)
3977                         return EINVAL;
3978         }
3979
3980         lwkt_serialize_enter(ifp->if_serializer);
3981
3982         if (throttle)
3983                 adapter->int_throttle_ceil = 1000000000 / 256 / throttle;
3984         else
3985                 adapter->int_throttle_ceil = 0;
3986
3987         if (ifp->if_flags & IFF_RUNNING)
3988                 E1000_WRITE_REG(&adapter->hw, E1000_ITR, throttle);
3989
3990         lwkt_serialize_exit(ifp->if_serializer);
3991
3992         if (bootverbose) {
3993                 if_printf(ifp, "Interrupt moderation set to %d/sec\n",
3994                           adapter->int_throttle_ceil);
3995         }
3996         return 0;
3997 }
3998
3999 static int
4000 em_sysctl_int_tx_nsegs(SYSCTL_HANDLER_ARGS)
4001 {
4002         struct adapter *adapter = (void *)arg1;
4003         struct ifnet *ifp = &adapter->arpcom.ac_if;
4004         int error, segs;
4005
4006         segs = adapter->tx_int_nsegs;
4007         error = sysctl_handle_int(oidp, &segs, 0, req);
4008         if (error || req->newptr == NULL)
4009                 return error;
4010         if (segs <= 0)
4011                 return EINVAL;
4012
4013         lwkt_serialize_enter(ifp->if_serializer);
4014
4015         /*
4016          * Don't allow int_tx_nsegs to become:
4017          * o  Less the oact_tx_desc
4018          * o  Too large that no TX desc will cause TX interrupt to
4019          *    be generated (OACTIVE will never recover)
4020          * o  Too small that will cause tx_dd[] overflow
4021          */
4022         if (segs < adapter->oact_tx_desc ||
4023             segs >= adapter->num_tx_desc - adapter->oact_tx_desc ||
4024             segs < adapter->num_tx_desc / EM_TXDD_SAFE) {
4025                 error = EINVAL;
4026         } else {
4027                 error = 0;
4028                 adapter->tx_int_nsegs = segs;
4029         }
4030
4031         lwkt_serialize_exit(ifp->if_serializer);
4032
4033         return error;
4034 }