linux/drivers/net/ethernet/intel/e1000e/netdev.c
<<
>>
Prefs
   1/*******************************************************************************
   2
   3  Intel PRO/1000 Linux driver
   4  Copyright(c) 1999 - 2013 Intel Corporation.
   5
   6  This program is free software; you can redistribute it and/or modify it
   7  under the terms and conditions of the GNU General Public License,
   8  version 2, as published by the Free Software Foundation.
   9
  10  This program is distributed in the hope it will be useful, but WITHOUT
  11  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13  more details.
  14
  15  You should have received a copy of the GNU General Public License along with
  16  this program; if not, write to the Free Software Foundation, Inc.,
  17  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
  18
  19  The full GNU General Public License is included in this distribution in
  20  the file called "COPYING".
  21
  22  Contact Information:
  23  Linux NICS <linux.nics@intel.com>
  24  e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  25  Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
  26
  27*******************************************************************************/
  28
  29#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  30
  31#include <linux/module.h>
  32#include <linux/types.h>
  33#include <linux/init.h>
  34#include <linux/pci.h>
  35#include <linux/vmalloc.h>
  36#include <linux/pagemap.h>
  37#include <linux/delay.h>
  38#include <linux/netdevice.h>
  39#include <linux/interrupt.h>
  40#include <linux/tcp.h>
  41#include <linux/ipv6.h>
  42#include <linux/slab.h>
  43#include <net/checksum.h>
  44#include <net/ip6_checksum.h>
  45#include <linux/ethtool.h>
  46#include <linux/if_vlan.h>
  47#include <linux/cpu.h>
  48#include <linux/smp.h>
  49#include <linux/pm_qos.h>
  50#include <linux/pm_runtime.h>
  51#include <linux/aer.h>
  52#include <linux/prefetch.h>
  53
  54#include "e1000.h"
  55
  56#define DRV_EXTRAVERSION "-k"
  57
  58#define DRV_VERSION "2.3.2" DRV_EXTRAVERSION
  59char e1000e_driver_name[] = "e1000e";
  60const char e1000e_driver_version[] = DRV_VERSION;
  61
  62#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
  63static int debug = -1;
  64module_param(debug, int, 0);
  65MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
  66
  67static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state);
  68
  69static const struct e1000_info *e1000_info_tbl[] = {
  70        [board_82571]           = &e1000_82571_info,
  71        [board_82572]           = &e1000_82572_info,
  72        [board_82573]           = &e1000_82573_info,
  73        [board_82574]           = &e1000_82574_info,
  74        [board_82583]           = &e1000_82583_info,
  75        [board_80003es2lan]     = &e1000_es2_info,
  76        [board_ich8lan]         = &e1000_ich8_info,
  77        [board_ich9lan]         = &e1000_ich9_info,
  78        [board_ich10lan]        = &e1000_ich10_info,
  79        [board_pchlan]          = &e1000_pch_info,
  80        [board_pch2lan]         = &e1000_pch2_info,
  81        [board_pch_lpt]         = &e1000_pch_lpt_info,
  82};
  83
  84struct e1000_reg_info {
  85        u32 ofs;
  86        char *name;
  87};
  88
  89static const struct e1000_reg_info e1000_reg_info_tbl[] = {
  90        /* General Registers */
  91        {E1000_CTRL, "CTRL"},
  92        {E1000_STATUS, "STATUS"},
  93        {E1000_CTRL_EXT, "CTRL_EXT"},
  94
  95        /* Interrupt Registers */
  96        {E1000_ICR, "ICR"},
  97
  98        /* Rx Registers */
  99        {E1000_RCTL, "RCTL"},
 100        {E1000_RDLEN(0), "RDLEN"},
 101        {E1000_RDH(0), "RDH"},
 102        {E1000_RDT(0), "RDT"},
 103        {E1000_RDTR, "RDTR"},
 104        {E1000_RXDCTL(0), "RXDCTL"},
 105        {E1000_ERT, "ERT"},
 106        {E1000_RDBAL(0), "RDBAL"},
 107        {E1000_RDBAH(0), "RDBAH"},
 108        {E1000_RDFH, "RDFH"},
 109        {E1000_RDFT, "RDFT"},
 110        {E1000_RDFHS, "RDFHS"},
 111        {E1000_RDFTS, "RDFTS"},
 112        {E1000_RDFPC, "RDFPC"},
 113
 114        /* Tx Registers */
 115        {E1000_TCTL, "TCTL"},
 116        {E1000_TDBAL(0), "TDBAL"},
 117        {E1000_TDBAH(0), "TDBAH"},
 118        {E1000_TDLEN(0), "TDLEN"},
 119        {E1000_TDH(0), "TDH"},
 120        {E1000_TDT(0), "TDT"},
 121        {E1000_TIDV, "TIDV"},
 122        {E1000_TXDCTL(0), "TXDCTL"},
 123        {E1000_TADV, "TADV"},
 124        {E1000_TARC(0), "TARC"},
 125        {E1000_TDFH, "TDFH"},
 126        {E1000_TDFT, "TDFT"},
 127        {E1000_TDFHS, "TDFHS"},
 128        {E1000_TDFTS, "TDFTS"},
 129        {E1000_TDFPC, "TDFPC"},
 130
 131        /* List Terminator */
 132        {0, NULL}
 133};
 134
 135/**
 136 * e1000_regdump - register printout routine
 137 * @hw: pointer to the HW structure
 138 * @reginfo: pointer to the register info table
 139 **/
 140static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
 141{
 142        int n = 0;
 143        char rname[16];
 144        u32 regs[8];
 145
 146        switch (reginfo->ofs) {
 147        case E1000_RXDCTL(0):
 148                for (n = 0; n < 2; n++)
 149                        regs[n] = __er32(hw, E1000_RXDCTL(n));
 150                break;
 151        case E1000_TXDCTL(0):
 152                for (n = 0; n < 2; n++)
 153                        regs[n] = __er32(hw, E1000_TXDCTL(n));
 154                break;
 155        case E1000_TARC(0):
 156                for (n = 0; n < 2; n++)
 157                        regs[n] = __er32(hw, E1000_TARC(n));
 158                break;
 159        default:
 160                pr_info("%-15s %08x\n",
 161                        reginfo->name, __er32(hw, reginfo->ofs));
 162                return;
 163        }
 164
 165        snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
 166        pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
 167}
 168
 169static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
 170                                 struct e1000_buffer *bi)
 171{
 172        int i;
 173        struct e1000_ps_page *ps_page;
 174
 175        for (i = 0; i < adapter->rx_ps_pages; i++) {
 176                ps_page = &bi->ps_pages[i];
 177
 178                if (ps_page->page) {
 179                        pr_info("packet dump for ps_page %d:\n", i);
 180                        print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
 181                                       16, 1, page_address(ps_page->page),
 182                                       PAGE_SIZE, true);
 183                }
 184        }
 185}
 186
 187/**
 188 * e1000e_dump - Print registers, Tx-ring and Rx-ring
 189 * @adapter: board private structure
 190 **/
 191static void e1000e_dump(struct e1000_adapter *adapter)
 192{
 193        struct net_device *netdev = adapter->netdev;
 194        struct e1000_hw *hw = &adapter->hw;
 195        struct e1000_reg_info *reginfo;
 196        struct e1000_ring *tx_ring = adapter->tx_ring;
 197        struct e1000_tx_desc *tx_desc;
 198        struct my_u0 {
 199                __le64 a;
 200                __le64 b;
 201        } *u0;
 202        struct e1000_buffer *buffer_info;
 203        struct e1000_ring *rx_ring = adapter->rx_ring;
 204        union e1000_rx_desc_packet_split *rx_desc_ps;
 205        union e1000_rx_desc_extended *rx_desc;
 206        struct my_u1 {
 207                __le64 a;
 208                __le64 b;
 209                __le64 c;
 210                __le64 d;
 211        } *u1;
 212        u32 staterr;
 213        int i = 0;
 214
 215        if (!netif_msg_hw(adapter))
 216                return;
 217
 218        /* Print netdevice Info */
 219        if (netdev) {
 220                dev_info(&adapter->pdev->dev, "Net device Info\n");
 221                pr_info("Device Name     state            trans_start      last_rx\n");
 222                pr_info("%-15s %016lX %016lX %016lX\n", netdev->name,
 223                        netdev->state, netdev->trans_start, netdev->last_rx);
 224        }
 225
 226        /* Print Registers */
 227        dev_info(&adapter->pdev->dev, "Register Dump\n");
 228        pr_info(" Register Name   Value\n");
 229        for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
 230             reginfo->name; reginfo++) {
 231                e1000_regdump(hw, reginfo);
 232        }
 233
 234        /* Print Tx Ring Summary */
 235        if (!netdev || !netif_running(netdev))
 236                return;
 237
 238        dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
 239        pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
 240        buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
 241        pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
 242                0, tx_ring->next_to_use, tx_ring->next_to_clean,
 243                (unsigned long long)buffer_info->dma,
 244                buffer_info->length,
 245                buffer_info->next_to_watch,
 246                (unsigned long long)buffer_info->time_stamp);
 247
 248        /* Print Tx Ring */
 249        if (!netif_msg_tx_done(adapter))
 250                goto rx_ring_summary;
 251
 252        dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
 253
 254        /* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
 255         *
 256         * Legacy Transmit Descriptor
 257         *   +--------------------------------------------------------------+
 258         * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
 259         *   +--------------------------------------------------------------+
 260         * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
 261         *   +--------------------------------------------------------------+
 262         *   63       48 47        36 35    32 31     24 23    16 15        0
 263         *
 264         * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
 265         *   63      48 47    40 39       32 31             16 15    8 7      0
 266         *   +----------------------------------------------------------------+
 267         * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
 268         *   +----------------------------------------------------------------+
 269         * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
 270         *   +----------------------------------------------------------------+
 271         *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
 272         *
 273         * Extended Data Descriptor (DTYP=0x1)
 274         *   +----------------------------------------------------------------+
 275         * 0 |                     Buffer Address [63:0]                      |
 276         *   +----------------------------------------------------------------+
 277         * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
 278         *   +----------------------------------------------------------------+
 279         *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
 280         */
 281        pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
 282        pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
 283        pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
 284        for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
 285                const char *next_desc;
 286                tx_desc = E1000_TX_DESC(*tx_ring, i);
 287                buffer_info = &tx_ring->buffer_info[i];
 288                u0 = (struct my_u0 *)tx_desc;
 289                if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
 290                        next_desc = " NTC/U";
 291                else if (i == tx_ring->next_to_use)
 292                        next_desc = " NTU";
 293                else if (i == tx_ring->next_to_clean)
 294                        next_desc = " NTC";
 295                else
 296                        next_desc = "";
 297                pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
 298                        (!(le64_to_cpu(u0->b) & (1 << 29)) ? 'l' :
 299                         ((le64_to_cpu(u0->b) & (1 << 20)) ? 'd' : 'c')),
 300                        i,
 301                        (unsigned long long)le64_to_cpu(u0->a),
 302                        (unsigned long long)le64_to_cpu(u0->b),
 303                        (unsigned long long)buffer_info->dma,
 304                        buffer_info->length, buffer_info->next_to_watch,
 305                        (unsigned long long)buffer_info->time_stamp,
 306                        buffer_info->skb, next_desc);
 307
 308                if (netif_msg_pktdata(adapter) && buffer_info->skb)
 309                        print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
 310                                       16, 1, buffer_info->skb->data,
 311                                       buffer_info->skb->len, true);
 312        }
 313
 314        /* Print Rx Ring Summary */
 315rx_ring_summary:
 316        dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
 317        pr_info("Queue [NTU] [NTC]\n");
 318        pr_info(" %5d %5X %5X\n",
 319                0, rx_ring->next_to_use, rx_ring->next_to_clean);
 320
 321        /* Print Rx Ring */
 322        if (!netif_msg_rx_status(adapter))
 323                return;
 324
 325        dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
 326        switch (adapter->rx_ps_pages) {
 327        case 1:
 328        case 2:
 329        case 3:
 330                /* [Extended] Packet Split Receive Descriptor Format
 331                 *
 332                 *    +-----------------------------------------------------+
 333                 *  0 |                Buffer Address 0 [63:0]              |
 334                 *    +-----------------------------------------------------+
 335                 *  8 |                Buffer Address 1 [63:0]              |
 336                 *    +-----------------------------------------------------+
 337                 * 16 |                Buffer Address 2 [63:0]              |
 338                 *    +-----------------------------------------------------+
 339                 * 24 |                Buffer Address 3 [63:0]              |
 340                 *    +-----------------------------------------------------+
 341                 */
 342                pr_info("R  [desc]      [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma       ] [bi->skb] <-- Ext Pkt Split format\n");
 343                /* [Extended] Receive Descriptor (Write-Back) Format
 344                 *
 345                 *   63       48 47    32 31     13 12    8 7    4 3        0
 346                 *   +------------------------------------------------------+
 347                 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
 348                 *   | Checksum | Ident  |         | Queue |      |  Type   |
 349                 *   +------------------------------------------------------+
 350                 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
 351                 *   +------------------------------------------------------+
 352                 *   63       48 47    32 31            20 19               0
 353                 */
 354                pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
 355                for (i = 0; i < rx_ring->count; i++) {
 356                        const char *next_desc;
 357                        buffer_info = &rx_ring->buffer_info[i];
 358                        rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
 359                        u1 = (struct my_u1 *)rx_desc_ps;
 360                        staterr =
 361                            le32_to_cpu(rx_desc_ps->wb.middle.status_error);
 362
 363                        if (i == rx_ring->next_to_use)
 364                                next_desc = " NTU";
 365                        else if (i == rx_ring->next_to_clean)
 366                                next_desc = " NTC";
 367                        else
 368                                next_desc = "";
 369
 370                        if (staterr & E1000_RXD_STAT_DD) {
 371                                /* Descriptor Done */
 372                                pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
 373                                        "RWB", i,
 374                                        (unsigned long long)le64_to_cpu(u1->a),
 375                                        (unsigned long long)le64_to_cpu(u1->b),
 376                                        (unsigned long long)le64_to_cpu(u1->c),
 377                                        (unsigned long long)le64_to_cpu(u1->d),
 378                                        buffer_info->skb, next_desc);
 379                        } else {
 380                                pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
 381                                        "R  ", i,
 382                                        (unsigned long long)le64_to_cpu(u1->a),
 383                                        (unsigned long long)le64_to_cpu(u1->b),
 384                                        (unsigned long long)le64_to_cpu(u1->c),
 385                                        (unsigned long long)le64_to_cpu(u1->d),
 386                                        (unsigned long long)buffer_info->dma,
 387                                        buffer_info->skb, next_desc);
 388
 389                                if (netif_msg_pktdata(adapter))
 390                                        e1000e_dump_ps_pages(adapter,
 391                                                             buffer_info);
 392                        }
 393                }
 394                break;
 395        default:
 396        case 0:
 397                /* Extended Receive Descriptor (Read) Format
 398                 *
 399                 *   +-----------------------------------------------------+
 400                 * 0 |                Buffer Address [63:0]                |
 401                 *   +-----------------------------------------------------+
 402                 * 8 |                      Reserved                       |
 403                 *   +-----------------------------------------------------+
 404                 */
 405                pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
 406                /* Extended Receive Descriptor (Write-Back) Format
 407                 *
 408                 *   63       48 47    32 31    24 23            4 3        0
 409                 *   +------------------------------------------------------+
 410                 *   |     RSS Hash      |        |               |         |
 411                 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
 412                 *   | Packet   | IP     |        |               |  Type   |
 413                 *   | Checksum | Ident  |        |               |         |
 414                 *   +------------------------------------------------------+
 415                 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
 416                 *   +------------------------------------------------------+
 417                 *   63       48 47    32 31            20 19               0
 418                 */
 419                pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
 420
 421                for (i = 0; i < rx_ring->count; i++) {
 422                        const char *next_desc;
 423
 424                        buffer_info = &rx_ring->buffer_info[i];
 425                        rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
 426                        u1 = (struct my_u1 *)rx_desc;
 427                        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 428
 429                        if (i == rx_ring->next_to_use)
 430                                next_desc = " NTU";
 431                        else if (i == rx_ring->next_to_clean)
 432                                next_desc = " NTC";
 433                        else
 434                                next_desc = "";
 435
 436                        if (staterr & E1000_RXD_STAT_DD) {
 437                                /* Descriptor Done */
 438                                pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
 439                                        "RWB", i,
 440                                        (unsigned long long)le64_to_cpu(u1->a),
 441                                        (unsigned long long)le64_to_cpu(u1->b),
 442                                        buffer_info->skb, next_desc);
 443                        } else {
 444                                pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
 445                                        "R  ", i,
 446                                        (unsigned long long)le64_to_cpu(u1->a),
 447                                        (unsigned long long)le64_to_cpu(u1->b),
 448                                        (unsigned long long)buffer_info->dma,
 449                                        buffer_info->skb, next_desc);
 450
 451                                if (netif_msg_pktdata(adapter) &&
 452                                    buffer_info->skb)
 453                                        print_hex_dump(KERN_INFO, "",
 454                                                       DUMP_PREFIX_ADDRESS, 16,
 455                                                       1,
 456                                                       buffer_info->skb->data,
 457                                                       adapter->rx_buffer_len,
 458                                                       true);
 459                        }
 460                }
 461        }
 462}
 463
 464/**
 465 * e1000_desc_unused - calculate if we have unused descriptors
 466 **/
 467static int e1000_desc_unused(struct e1000_ring *ring)
 468{
 469        if (ring->next_to_clean > ring->next_to_use)
 470                return ring->next_to_clean - ring->next_to_use - 1;
 471
 472        return ring->count + ring->next_to_clean - ring->next_to_use - 1;
 473}
 474
 475/**
 476 * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
 477 * @adapter: board private structure
 478 * @hwtstamps: time stamp structure to update
 479 * @systim: unsigned 64bit system time value.
 480 *
 481 * Convert the system time value stored in the RX/TXSTMP registers into a
 482 * hwtstamp which can be used by the upper level time stamping functions.
 483 *
 484 * The 'systim_lock' spinlock is used to protect the consistency of the
 485 * system time value. This is needed because reading the 64 bit time
 486 * value involves reading two 32 bit registers. The first read latches the
 487 * value.
 488 **/
 489static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
 490                                      struct skb_shared_hwtstamps *hwtstamps,
 491                                      u64 systim)
 492{
 493        u64 ns;
 494        unsigned long flags;
 495
 496        spin_lock_irqsave(&adapter->systim_lock, flags);
 497        ns = timecounter_cyc2time(&adapter->tc, systim);
 498        spin_unlock_irqrestore(&adapter->systim_lock, flags);
 499
 500        memset(hwtstamps, 0, sizeof(*hwtstamps));
 501        hwtstamps->hwtstamp = ns_to_ktime(ns);
 502}
 503
 504/**
 505 * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
 506 * @adapter: board private structure
 507 * @status: descriptor extended error and status field
 508 * @skb: particular skb to include time stamp
 509 *
 510 * If the time stamp is valid, convert it into the timecounter ns value
 511 * and store that result into the shhwtstamps structure which is passed
 512 * up the network stack.
 513 **/
 514static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
 515                               struct sk_buff *skb)
 516{
 517        struct e1000_hw *hw = &adapter->hw;
 518        u64 rxstmp;
 519
 520        if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
 521            !(status & E1000_RXDEXT_STATERR_TST) ||
 522            !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
 523                return;
 524
 525        /* The Rx time stamp registers contain the time stamp.  No other
 526         * received packet will be time stamped until the Rx time stamp
 527         * registers are read.  Because only one packet can be time stamped
 528         * at a time, the register values must belong to this packet and
 529         * therefore none of the other additional attributes need to be
 530         * compared.
 531         */
 532        rxstmp = (u64)er32(RXSTMPL);
 533        rxstmp |= (u64)er32(RXSTMPH) << 32;
 534        e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
 535
 536        adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
 537}
 538
 539/**
 540 * e1000_receive_skb - helper function to handle Rx indications
 541 * @adapter: board private structure
 542 * @staterr: descriptor extended error and status field as written by hardware
 543 * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
 544 * @skb: pointer to sk_buff to be indicated to stack
 545 **/
 546static void e1000_receive_skb(struct e1000_adapter *adapter,
 547                              struct net_device *netdev, struct sk_buff *skb,
 548                              u32 staterr, __le16 vlan)
 549{
 550        u16 tag = le16_to_cpu(vlan);
 551
 552        e1000e_rx_hwtstamp(adapter, staterr, skb);
 553
 554        skb->protocol = eth_type_trans(skb, netdev);
 555
 556        if (staterr & E1000_RXD_STAT_VP)
 557                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
 558
 559        napi_gro_receive(&adapter->napi, skb);
 560}
 561
 562/**
 563 * e1000_rx_checksum - Receive Checksum Offload
 564 * @adapter: board private structure
 565 * @status_err: receive descriptor status and error fields
 566 * @csum: receive descriptor csum field
 567 * @sk_buff: socket buffer with received data
 568 **/
 569static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
 570                              struct sk_buff *skb)
 571{
 572        u16 status = (u16)status_err;
 573        u8 errors = (u8)(status_err >> 24);
 574
 575        skb_checksum_none_assert(skb);
 576
 577        /* Rx checksum disabled */
 578        if (!(adapter->netdev->features & NETIF_F_RXCSUM))
 579                return;
 580
 581        /* Ignore Checksum bit is set */
 582        if (status & E1000_RXD_STAT_IXSM)
 583                return;
 584
 585        /* TCP/UDP checksum error bit or IP checksum error bit is set */
 586        if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
 587                /* let the stack verify checksum errors */
 588                adapter->hw_csum_err++;
 589                return;
 590        }
 591
 592        /* TCP/UDP Checksum has not been calculated */
 593        if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
 594                return;
 595
 596        /* It must be a TCP or UDP packet with a valid checksum */
 597        skb->ip_summed = CHECKSUM_UNNECESSARY;
 598        adapter->hw_csum_good++;
 599}
 600
 601static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
 602{
 603        struct e1000_adapter *adapter = rx_ring->adapter;
 604        struct e1000_hw *hw = &adapter->hw;
 605        s32 ret_val = __ew32_prepare(hw);
 606
 607        writel(i, rx_ring->tail);
 608
 609        if (unlikely(!ret_val && (i != readl(rx_ring->tail)))) {
 610                u32 rctl = er32(RCTL);
 611                ew32(RCTL, rctl & ~E1000_RCTL_EN);
 612                e_err("ME firmware caused invalid RDT - resetting\n");
 613                schedule_work(&adapter->reset_task);
 614        }
 615}
 616
 617static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
 618{
 619        struct e1000_adapter *adapter = tx_ring->adapter;
 620        struct e1000_hw *hw = &adapter->hw;
 621        s32 ret_val = __ew32_prepare(hw);
 622
 623        writel(i, tx_ring->tail);
 624
 625        if (unlikely(!ret_val && (i != readl(tx_ring->tail)))) {
 626                u32 tctl = er32(TCTL);
 627                ew32(TCTL, tctl & ~E1000_TCTL_EN);
 628                e_err("ME firmware caused invalid TDT - resetting\n");
 629                schedule_work(&adapter->reset_task);
 630        }
 631}
 632
 633/**
 634 * e1000_alloc_rx_buffers - Replace used receive buffers
 635 * @rx_ring: Rx descriptor ring
 636 **/
 637static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
 638                                   int cleaned_count, gfp_t gfp)
 639{
 640        struct e1000_adapter *adapter = rx_ring->adapter;
 641        struct net_device *netdev = adapter->netdev;
 642        struct pci_dev *pdev = adapter->pdev;
 643        union e1000_rx_desc_extended *rx_desc;
 644        struct e1000_buffer *buffer_info;
 645        struct sk_buff *skb;
 646        unsigned int i;
 647        unsigned int bufsz = adapter->rx_buffer_len;
 648
 649        i = rx_ring->next_to_use;
 650        buffer_info = &rx_ring->buffer_info[i];
 651
 652        while (cleaned_count--) {
 653                skb = buffer_info->skb;
 654                if (skb) {
 655                        skb_trim(skb, 0);
 656                        goto map_skb;
 657                }
 658
 659                skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
 660                if (!skb) {
 661                        /* Better luck next round */
 662                        adapter->alloc_rx_buff_failed++;
 663                        break;
 664                }
 665
 666                buffer_info->skb = skb;
 667map_skb:
 668                buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
 669                                                  adapter->rx_buffer_len,
 670                                                  DMA_FROM_DEVICE);
 671                if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
 672                        dev_err(&pdev->dev, "Rx DMA map failed\n");
 673                        adapter->rx_dma_failed++;
 674                        break;
 675                }
 676
 677                rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
 678                rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
 679
 680                if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
 681                        /* Force memory writes to complete before letting h/w
 682                         * know there are new descriptors to fetch.  (Only
 683                         * applicable for weak-ordered memory model archs,
 684                         * such as IA-64).
 685                         */
 686                        wmb();
 687                        if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
 688                                e1000e_update_rdt_wa(rx_ring, i);
 689                        else
 690                                writel(i, rx_ring->tail);
 691                }
 692                i++;
 693                if (i == rx_ring->count)
 694                        i = 0;
 695                buffer_info = &rx_ring->buffer_info[i];
 696        }
 697
 698        rx_ring->next_to_use = i;
 699}
 700
 701/**
 702 * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
 703 * @rx_ring: Rx descriptor ring
 704 **/
 705static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
 706                                      int cleaned_count, gfp_t gfp)
 707{
 708        struct e1000_adapter *adapter = rx_ring->adapter;
 709        struct net_device *netdev = adapter->netdev;
 710        struct pci_dev *pdev = adapter->pdev;
 711        union e1000_rx_desc_packet_split *rx_desc;
 712        struct e1000_buffer *buffer_info;
 713        struct e1000_ps_page *ps_page;
 714        struct sk_buff *skb;
 715        unsigned int i, j;
 716
 717        i = rx_ring->next_to_use;
 718        buffer_info = &rx_ring->buffer_info[i];
 719
 720        while (cleaned_count--) {
 721                rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
 722
 723                for (j = 0; j < PS_PAGE_BUFFERS; j++) {
 724                        ps_page = &buffer_info->ps_pages[j];
 725                        if (j >= adapter->rx_ps_pages) {
 726                                /* all unused desc entries get hw null ptr */
 727                                rx_desc->read.buffer_addr[j + 1] =
 728                                    ~cpu_to_le64(0);
 729                                continue;
 730                        }
 731                        if (!ps_page->page) {
 732                                ps_page->page = alloc_page(gfp);
 733                                if (!ps_page->page) {
 734                                        adapter->alloc_rx_buff_failed++;
 735                                        goto no_buffers;
 736                                }
 737                                ps_page->dma = dma_map_page(&pdev->dev,
 738                                                            ps_page->page,
 739                                                            0, PAGE_SIZE,
 740                                                            DMA_FROM_DEVICE);
 741                                if (dma_mapping_error(&pdev->dev,
 742                                                      ps_page->dma)) {
 743                                        dev_err(&adapter->pdev->dev,
 744                                                "Rx DMA page map failed\n");
 745                                        adapter->rx_dma_failed++;
 746                                        goto no_buffers;
 747                                }
 748                        }
 749                        /* Refresh the desc even if buffer_addrs
 750                         * didn't change because each write-back
 751                         * erases this info.
 752                         */
 753                        rx_desc->read.buffer_addr[j + 1] =
 754                            cpu_to_le64(ps_page->dma);
 755                }
 756
 757                skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
 758                                                  gfp);
 759
 760                if (!skb) {
 761                        adapter->alloc_rx_buff_failed++;
 762                        break;
 763                }
 764
 765                buffer_info->skb = skb;
 766                buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
 767                                                  adapter->rx_ps_bsize0,
 768                                                  DMA_FROM_DEVICE);
 769                if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
 770                        dev_err(&pdev->dev, "Rx DMA map failed\n");
 771                        adapter->rx_dma_failed++;
 772                        /* cleanup skb */
 773                        dev_kfree_skb_any(skb);
 774                        buffer_info->skb = NULL;
 775                        break;
 776                }
 777
 778                rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
 779
 780                if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
 781                        /* Force memory writes to complete before letting h/w
 782                         * know there are new descriptors to fetch.  (Only
 783                         * applicable for weak-ordered memory model archs,
 784                         * such as IA-64).
 785                         */
 786                        wmb();
 787                        if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
 788                                e1000e_update_rdt_wa(rx_ring, i << 1);
 789                        else
 790                                writel(i << 1, rx_ring->tail);
 791                }
 792
 793                i++;
 794                if (i == rx_ring->count)
 795                        i = 0;
 796                buffer_info = &rx_ring->buffer_info[i];
 797        }
 798
 799no_buffers:
 800        rx_ring->next_to_use = i;
 801}
 802
 803/**
 804 * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
 805 * @rx_ring: Rx descriptor ring
 806 * @cleaned_count: number of buffers to allocate this pass
 807 **/
 808
 809static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
 810                                         int cleaned_count, gfp_t gfp)
 811{
 812        struct e1000_adapter *adapter = rx_ring->adapter;
 813        struct net_device *netdev = adapter->netdev;
 814        struct pci_dev *pdev = adapter->pdev;
 815        union e1000_rx_desc_extended *rx_desc;
 816        struct e1000_buffer *buffer_info;
 817        struct sk_buff *skb;
 818        unsigned int i;
 819        unsigned int bufsz = 256 - 16;  /* for skb_reserve */
 820
 821        i = rx_ring->next_to_use;
 822        buffer_info = &rx_ring->buffer_info[i];
 823
 824        while (cleaned_count--) {
 825                skb = buffer_info->skb;
 826                if (skb) {
 827                        skb_trim(skb, 0);
 828                        goto check_page;
 829                }
 830
 831                skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
 832                if (unlikely(!skb)) {
 833                        /* Better luck next round */
 834                        adapter->alloc_rx_buff_failed++;
 835                        break;
 836                }
 837
 838                buffer_info->skb = skb;
 839check_page:
 840                /* allocate a new page if necessary */
 841                if (!buffer_info->page) {
 842                        buffer_info->page = alloc_page(gfp);
 843                        if (unlikely(!buffer_info->page)) {
 844                                adapter->alloc_rx_buff_failed++;
 845                                break;
 846                        }
 847                }
 848
 849                if (!buffer_info->dma) {
 850                        buffer_info->dma = dma_map_page(&pdev->dev,
 851                                                        buffer_info->page, 0,
 852                                                        PAGE_SIZE,
 853                                                        DMA_FROM_DEVICE);
 854                        if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
 855                                adapter->alloc_rx_buff_failed++;
 856                                break;
 857                        }
 858                }
 859
 860                rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
 861                rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
 862
 863                if (unlikely(++i == rx_ring->count))
 864                        i = 0;
 865                buffer_info = &rx_ring->buffer_info[i];
 866        }
 867
 868        if (likely(rx_ring->next_to_use != i)) {
 869                rx_ring->next_to_use = i;
 870                if (unlikely(i-- == 0))
 871                        i = (rx_ring->count - 1);
 872
 873                /* Force memory writes to complete before letting h/w
 874                 * know there are new descriptors to fetch.  (Only
 875                 * applicable for weak-ordered memory model archs,
 876                 * such as IA-64).
 877                 */
 878                wmb();
 879                if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
 880                        e1000e_update_rdt_wa(rx_ring, i);
 881                else
 882                        writel(i, rx_ring->tail);
 883        }
 884}
 885
 886static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
 887                                 struct sk_buff *skb)
 888{
 889        if (netdev->features & NETIF_F_RXHASH)
 890                skb->rxhash = le32_to_cpu(rss);
 891}
 892
 893/**
 894 * e1000_clean_rx_irq - Send received data up the network stack
 895 * @rx_ring: Rx descriptor ring
 896 *
 897 * the return value indicates whether actual cleaning was done, there
 898 * is no guarantee that everything was cleaned
 899 **/
 900static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
 901                               int work_to_do)
 902{
 903        struct e1000_adapter *adapter = rx_ring->adapter;
 904        struct net_device *netdev = adapter->netdev;
 905        struct pci_dev *pdev = adapter->pdev;
 906        struct e1000_hw *hw = &adapter->hw;
 907        union e1000_rx_desc_extended *rx_desc, *next_rxd;
 908        struct e1000_buffer *buffer_info, *next_buffer;
 909        u32 length, staterr;
 910        unsigned int i;
 911        int cleaned_count = 0;
 912        bool cleaned = false;
 913        unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 914
 915        i = rx_ring->next_to_clean;
 916        rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
 917        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 918        buffer_info = &rx_ring->buffer_info[i];
 919
 920        while (staterr & E1000_RXD_STAT_DD) {
 921                struct sk_buff *skb;
 922
 923                if (*work_done >= work_to_do)
 924                        break;
 925                (*work_done)++;
 926                rmb();  /* read descriptor and rx_buffer_info after status DD */
 927
 928                skb = buffer_info->skb;
 929                buffer_info->skb = NULL;
 930
 931                prefetch(skb->data - NET_IP_ALIGN);
 932
 933                i++;
 934                if (i == rx_ring->count)
 935                        i = 0;
 936                next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
 937                prefetch(next_rxd);
 938
 939                next_buffer = &rx_ring->buffer_info[i];
 940
 941                cleaned = true;
 942                cleaned_count++;
 943                dma_unmap_single(&pdev->dev, buffer_info->dma,
 944                                 adapter->rx_buffer_len, DMA_FROM_DEVICE);
 945                buffer_info->dma = 0;
 946
 947                length = le16_to_cpu(rx_desc->wb.upper.length);
 948
 949                /* !EOP means multiple descriptors were used to store a single
 950                 * packet, if that's the case we need to toss it.  In fact, we
 951                 * need to toss every packet with the EOP bit clear and the
 952                 * next frame that _does_ have the EOP bit set, as it is by
 953                 * definition only a frame fragment
 954                 */
 955                if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
 956                        adapter->flags2 |= FLAG2_IS_DISCARDING;
 957
 958                if (adapter->flags2 & FLAG2_IS_DISCARDING) {
 959                        /* All receives must fit into a single buffer */
 960                        e_dbg("Receive packet consumed multiple buffers\n");
 961                        /* recycle */
 962                        buffer_info->skb = skb;
 963                        if (staterr & E1000_RXD_STAT_EOP)
 964                                adapter->flags2 &= ~FLAG2_IS_DISCARDING;
 965                        goto next_desc;
 966                }
 967
 968                if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
 969                             !(netdev->features & NETIF_F_RXALL))) {
 970                        /* recycle */
 971                        buffer_info->skb = skb;
 972                        goto next_desc;
 973                }
 974
 975                /* adjust length to remove Ethernet CRC */
 976                if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
 977                        /* If configured to store CRC, don't subtract FCS,
 978                         * but keep the FCS bytes out of the total_rx_bytes
 979                         * counter
 980                         */
 981                        if (netdev->features & NETIF_F_RXFCS)
 982                                total_rx_bytes -= 4;
 983                        else
 984                                length -= 4;
 985                }
 986
 987                total_rx_bytes += length;
 988                total_rx_packets++;
 989
 990                /* code added for copybreak, this should improve
 991                 * performance for small packets with large amounts
 992                 * of reassembly being done in the stack
 993                 */
 994                if (length < copybreak) {
 995                        struct sk_buff *new_skb =
 996                            netdev_alloc_skb_ip_align(netdev, length);
 997                        if (new_skb) {
 998                                skb_copy_to_linear_data_offset(new_skb,
 999                                                               -NET_IP_ALIGN,
1000                                                               (skb->data -
1001                                                                NET_IP_ALIGN),
1002                                                               (length +
1003                                                                NET_IP_ALIGN));
1004                                /* save the skb in buffer_info as good */
1005                                buffer_info->skb = skb;
1006                                skb = new_skb;
1007                        }
1008                        /* else just continue with the old one */
1009                }
1010                /* end copybreak code */
1011                skb_put(skb, length);
1012
1013                /* Receive Checksum Offload */
1014                e1000_rx_checksum(adapter, staterr, skb);
1015
1016                e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1017
1018                e1000_receive_skb(adapter, netdev, skb, staterr,
1019                                  rx_desc->wb.upper.vlan);
1020
1021next_desc:
1022                rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1023
1024                /* return some buffers to hardware, one at a time is too slow */
1025                if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1026                        adapter->alloc_rx_buf(rx_ring, cleaned_count,
1027                                              GFP_ATOMIC);
1028                        cleaned_count = 0;
1029                }
1030
1031                /* use prefetched values */
1032                rx_desc = next_rxd;
1033                buffer_info = next_buffer;
1034
1035                staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1036        }
1037        rx_ring->next_to_clean = i;
1038
1039        cleaned_count = e1000_desc_unused(rx_ring);
1040        if (cleaned_count)
1041                adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1042
1043        adapter->total_rx_bytes += total_rx_bytes;
1044        adapter->total_rx_packets += total_rx_packets;
1045        return cleaned;
1046}
1047
1048static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1049                            struct e1000_buffer *buffer_info)
1050{
1051        struct e1000_adapter *adapter = tx_ring->adapter;
1052
1053        if (buffer_info->dma) {
1054                if (buffer_info->mapped_as_page)
1055                        dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1056                                       buffer_info->length, DMA_TO_DEVICE);
1057                else
1058                        dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1059                                         buffer_info->length, DMA_TO_DEVICE);
1060                buffer_info->dma = 0;
1061        }
1062        if (buffer_info->skb) {
1063                dev_kfree_skb_any(buffer_info->skb);
1064                buffer_info->skb = NULL;
1065        }
1066        buffer_info->time_stamp = 0;
1067}
1068
1069static void e1000_print_hw_hang(struct work_struct *work)
1070{
1071        struct e1000_adapter *adapter = container_of(work,
1072                                                     struct e1000_adapter,
1073                                                     print_hang_task);
1074        struct net_device *netdev = adapter->netdev;
1075        struct e1000_ring *tx_ring = adapter->tx_ring;
1076        unsigned int i = tx_ring->next_to_clean;
1077        unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1078        struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1079        struct e1000_hw *hw = &adapter->hw;
1080        u16 phy_status, phy_1000t_status, phy_ext_status;
1081        u16 pci_status;
1082
1083        if (test_bit(__E1000_DOWN, &adapter->state))
1084                return;
1085
1086        if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1087                /* May be block on write-back, flush and detect again
1088                 * flush pending descriptor writebacks to memory
1089                 */
1090                ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1091                /* execute the writes immediately */
1092                e1e_flush();
1093                /* Due to rare timing issues, write to TIDV again to ensure
1094                 * the write is successful
1095                 */
1096                ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1097                /* execute the writes immediately */
1098                e1e_flush();
1099                adapter->tx_hang_recheck = true;
1100                return;
1101        }
1102        /* Real hang detected */
1103        adapter->tx_hang_recheck = false;
1104        netif_stop_queue(netdev);
1105
1106        e1e_rphy(hw, MII_BMSR, &phy_status);
1107        e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1108        e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1109
1110        pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1111
1112        /* detected Hardware unit hang */
1113        e_err("Detected Hardware Unit Hang:\n"
1114              "  TDH                  <%x>\n"
1115              "  TDT                  <%x>\n"
1116              "  next_to_use          <%x>\n"
1117              "  next_to_clean        <%x>\n"
1118              "buffer_info[next_to_clean]:\n"
1119              "  time_stamp           <%lx>\n"
1120              "  next_to_watch        <%x>\n"
1121              "  jiffies              <%lx>\n"
1122              "  next_to_watch.status <%x>\n"
1123              "MAC Status             <%x>\n"
1124              "PHY Status             <%x>\n"
1125              "PHY 1000BASE-T Status  <%x>\n"
1126              "PHY Extended Status    <%x>\n"
1127              "PCI Status             <%x>\n",
1128              readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1129              tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1130              eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1131              phy_status, phy_1000t_status, phy_ext_status, pci_status);
1132
1133        /* Suggest workaround for known h/w issue */
1134        if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1135                e_err("Try turning off Tx pause (flow control) via ethtool\n");
1136}
1137
1138/**
1139 * e1000e_tx_hwtstamp_work - check for Tx time stamp
1140 * @work: pointer to work struct
1141 *
1142 * This work function polls the TSYNCTXCTL valid bit to determine when a
1143 * timestamp has been taken for the current stored skb.  The timestamp must
1144 * be for this skb because only one such packet is allowed in the queue.
1145 */
1146static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1147{
1148        struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1149                                                     tx_hwtstamp_work);
1150        struct e1000_hw *hw = &adapter->hw;
1151
1152        if (!adapter->tx_hwtstamp_skb)
1153                return;
1154
1155        if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1156                struct skb_shared_hwtstamps shhwtstamps;
1157                u64 txstmp;
1158
1159                txstmp = er32(TXSTMPL);
1160                txstmp |= (u64)er32(TXSTMPH) << 32;
1161
1162                e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1163
1164                skb_tstamp_tx(adapter->tx_hwtstamp_skb, &shhwtstamps);
1165                dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1166                adapter->tx_hwtstamp_skb = NULL;
1167        } else {
1168                /* reschedule to check later */
1169                schedule_work(&adapter->tx_hwtstamp_work);
1170        }
1171}
1172
1173/**
1174 * e1000_clean_tx_irq - Reclaim resources after transmit completes
1175 * @tx_ring: Tx descriptor ring
1176 *
1177 * the return value indicates whether actual cleaning was done, there
1178 * is no guarantee that everything was cleaned
1179 **/
1180static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1181{
1182        struct e1000_adapter *adapter = tx_ring->adapter;
1183        struct net_device *netdev = adapter->netdev;
1184        struct e1000_hw *hw = &adapter->hw;
1185        struct e1000_tx_desc *tx_desc, *eop_desc;
1186        struct e1000_buffer *buffer_info;
1187        unsigned int i, eop;
1188        unsigned int count = 0;
1189        unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1190        unsigned int bytes_compl = 0, pkts_compl = 0;
1191
1192        i = tx_ring->next_to_clean;
1193        eop = tx_ring->buffer_info[i].next_to_watch;
1194        eop_desc = E1000_TX_DESC(*tx_ring, eop);
1195
1196        while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1197               (count < tx_ring->count)) {
1198                bool cleaned = false;
1199                rmb();          /* read buffer_info after eop_desc */
1200                for (; !cleaned; count++) {
1201                        tx_desc = E1000_TX_DESC(*tx_ring, i);
1202                        buffer_info = &tx_ring->buffer_info[i];
1203                        cleaned = (i == eop);
1204
1205                        if (cleaned) {
1206                                total_tx_packets += buffer_info->segs;
1207                                total_tx_bytes += buffer_info->bytecount;
1208                                if (buffer_info->skb) {
1209                                        bytes_compl += buffer_info->skb->len;
1210                                        pkts_compl++;
1211                                }
1212                        }
1213
1214                        e1000_put_txbuf(tx_ring, buffer_info);
1215                        tx_desc->upper.data = 0;
1216
1217                        i++;
1218                        if (i == tx_ring->count)
1219                                i = 0;
1220                }
1221
1222                if (i == tx_ring->next_to_use)
1223                        break;
1224                eop = tx_ring->buffer_info[i].next_to_watch;
1225                eop_desc = E1000_TX_DESC(*tx_ring, eop);
1226        }
1227
1228        tx_ring->next_to_clean = i;
1229
1230        netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1231
1232#define TX_WAKE_THRESHOLD 32
1233        if (count && netif_carrier_ok(netdev) &&
1234            e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1235                /* Make sure that anybody stopping the queue after this
1236                 * sees the new next_to_clean.
1237                 */
1238                smp_mb();
1239
1240                if (netif_queue_stopped(netdev) &&
1241                    !(test_bit(__E1000_DOWN, &adapter->state))) {
1242                        netif_wake_queue(netdev);
1243                        ++adapter->restart_queue;
1244                }
1245        }
1246
1247        if (adapter->detect_tx_hung) {
1248                /* Detect a transmit hang in hardware, this serializes the
1249                 * check with the clearing of time_stamp and movement of i
1250                 */
1251                adapter->detect_tx_hung = false;
1252                if (tx_ring->buffer_info[i].time_stamp &&
1253                    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1254                               + (adapter->tx_timeout_factor * HZ)) &&
1255                    !(er32(STATUS) & E1000_STATUS_TXOFF))
1256                        schedule_work(&adapter->print_hang_task);
1257                else
1258                        adapter->tx_hang_recheck = false;
1259        }
1260        adapter->total_tx_bytes += total_tx_bytes;
1261        adapter->total_tx_packets += total_tx_packets;
1262        return count < tx_ring->count;
1263}
1264
1265/**
1266 * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1267 * @rx_ring: Rx descriptor ring
1268 *
1269 * the return value indicates whether actual cleaning was done, there
1270 * is no guarantee that everything was cleaned
1271 **/
1272static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1273                                  int work_to_do)
1274{
1275        struct e1000_adapter *adapter = rx_ring->adapter;
1276        struct e1000_hw *hw = &adapter->hw;
1277        union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1278        struct net_device *netdev = adapter->netdev;
1279        struct pci_dev *pdev = adapter->pdev;
1280        struct e1000_buffer *buffer_info, *next_buffer;
1281        struct e1000_ps_page *ps_page;
1282        struct sk_buff *skb;
1283        unsigned int i, j;
1284        u32 length, staterr;
1285        int cleaned_count = 0;
1286        bool cleaned = false;
1287        unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1288
1289        i = rx_ring->next_to_clean;
1290        rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1291        staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1292        buffer_info = &rx_ring->buffer_info[i];
1293
1294        while (staterr & E1000_RXD_STAT_DD) {
1295                if (*work_done >= work_to_do)
1296                        break;
1297                (*work_done)++;
1298                skb = buffer_info->skb;
1299                rmb();  /* read descriptor and rx_buffer_info after status DD */
1300
1301                /* in the packet split case this is header only */
1302                prefetch(skb->data - NET_IP_ALIGN);
1303
1304                i++;
1305                if (i == rx_ring->count)
1306                        i = 0;
1307                next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1308                prefetch(next_rxd);
1309
1310                next_buffer = &rx_ring->buffer_info[i];
1311
1312                cleaned = true;
1313                cleaned_count++;
1314                dma_unmap_single(&pdev->dev, buffer_info->dma,
1315                                 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1316                buffer_info->dma = 0;
1317
1318                /* see !EOP comment in other Rx routine */
1319                if (!(staterr & E1000_RXD_STAT_EOP))
1320                        adapter->flags2 |= FLAG2_IS_DISCARDING;
1321
1322                if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1323                        e_dbg("Packet Split buffers didn't pick up the full packet\n");
1324                        dev_kfree_skb_irq(skb);
1325                        if (staterr & E1000_RXD_STAT_EOP)
1326                                adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1327                        goto next_desc;
1328                }
1329
1330                if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1331                             !(netdev->features & NETIF_F_RXALL))) {
1332                        dev_kfree_skb_irq(skb);
1333                        goto next_desc;
1334                }
1335
1336                length = le16_to_cpu(rx_desc->wb.middle.length0);
1337
1338                if (!length) {
1339                        e_dbg("Last part of the packet spanning multiple descriptors\n");
1340                        dev_kfree_skb_irq(skb);
1341                        goto next_desc;
1342                }
1343
1344                /* Good Receive */
1345                skb_put(skb, length);
1346
1347                {
1348                        /* this looks ugly, but it seems compiler issues make
1349                         * it more efficient than reusing j
1350                         */
1351                        int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1352
1353                        /* page alloc/put takes too long and effects small
1354                         * packet throughput, so unsplit small packets and
1355                         * save the alloc/put only valid in softirq (napi)
1356                         * context to call kmap_*
1357                         */
1358                        if (l1 && (l1 <= copybreak) &&
1359                            ((length + l1) <= adapter->rx_ps_bsize0)) {
1360                                u8 *vaddr;
1361
1362                                ps_page = &buffer_info->ps_pages[0];
1363
1364                                /* there is no documentation about how to call
1365                                 * kmap_atomic, so we can't hold the mapping
1366                                 * very long
1367                                 */
1368                                dma_sync_single_for_cpu(&pdev->dev,
1369                                                        ps_page->dma,
1370                                                        PAGE_SIZE,
1371                                                        DMA_FROM_DEVICE);
1372                                vaddr = kmap_atomic(ps_page->page);
1373                                memcpy(skb_tail_pointer(skb), vaddr, l1);
1374                                kunmap_atomic(vaddr);
1375                                dma_sync_single_for_device(&pdev->dev,
1376                                                           ps_page->dma,
1377                                                           PAGE_SIZE,
1378                                                           DMA_FROM_DEVICE);
1379
1380                                /* remove the CRC */
1381                                if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1382                                        if (!(netdev->features & NETIF_F_RXFCS))
1383                                                l1 -= 4;
1384                                }
1385
1386                                skb_put(skb, l1);
1387                                goto copydone;
1388                        }       /* if */
1389                }
1390
1391                for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1392                        length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1393                        if (!length)
1394                                break;
1395
1396                        ps_page = &buffer_info->ps_pages[j];
1397                        dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1398                                       DMA_FROM_DEVICE);
1399                        ps_page->dma = 0;
1400                        skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1401                        ps_page->page = NULL;
1402                        skb->len += length;
1403                        skb->data_len += length;
1404                        skb->truesize += PAGE_SIZE;
1405                }
1406
1407                /* strip the ethernet crc, problem is we're using pages now so
1408                 * this whole operation can get a little cpu intensive
1409                 */
1410                if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1411                        if (!(netdev->features & NETIF_F_RXFCS))
1412                                pskb_trim(skb, skb->len - 4);
1413                }
1414
1415copydone:
1416                total_rx_bytes += skb->len;
1417                total_rx_packets++;
1418
1419                e1000_rx_checksum(adapter, staterr, skb);
1420
1421                e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1422
1423                if (rx_desc->wb.upper.header_status &
1424                    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1425                        adapter->rx_hdr_split++;
1426
1427                e1000_receive_skb(adapter, netdev, skb, staterr,
1428                                  rx_desc->wb.middle.vlan);
1429
1430next_desc:
1431                rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1432                buffer_info->skb = NULL;
1433
1434                /* return some buffers to hardware, one at a time is too slow */
1435                if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1436                        adapter->alloc_rx_buf(rx_ring, cleaned_count,
1437                                              GFP_ATOMIC);
1438                        cleaned_count = 0;
1439                }
1440
1441                /* use prefetched values */
1442                rx_desc = next_rxd;
1443                buffer_info = next_buffer;
1444
1445                staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1446        }
1447        rx_ring->next_to_clean = i;
1448
1449        cleaned_count = e1000_desc_unused(rx_ring);
1450        if (cleaned_count)
1451                adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1452
1453        adapter->total_rx_bytes += total_rx_bytes;
1454        adapter->total_rx_packets += total_rx_packets;
1455        return cleaned;
1456}
1457
1458/**
1459 * e1000_consume_page - helper function
1460 **/
1461static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1462                               u16 length)
1463{
1464        bi->page = NULL;
1465        skb->len += length;
1466        skb->data_len += length;
1467        skb->truesize += PAGE_SIZE;
1468}
1469
1470/**
1471 * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1472 * @adapter: board private structure
1473 *
1474 * the return value indicates whether actual cleaning was done, there
1475 * is no guarantee that everything was cleaned
1476 **/
1477static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1478                                     int work_to_do)
1479{
1480        struct e1000_adapter *adapter = rx_ring->adapter;
1481        struct net_device *netdev = adapter->netdev;
1482        struct pci_dev *pdev = adapter->pdev;
1483        union e1000_rx_desc_extended *rx_desc, *next_rxd;
1484        struct e1000_buffer *buffer_info, *next_buffer;
1485        u32 length, staterr;
1486        unsigned int i;
1487        int cleaned_count = 0;
1488        bool cleaned = false;
1489        unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1490        struct skb_shared_info *shinfo;
1491
1492        i = rx_ring->next_to_clean;
1493        rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1494        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1495        buffer_info = &rx_ring->buffer_info[i];
1496
1497        while (staterr & E1000_RXD_STAT_DD) {
1498                struct sk_buff *skb;
1499
1500                if (*work_done >= work_to_do)
1501                        break;
1502                (*work_done)++;
1503                rmb();  /* read descriptor and rx_buffer_info after status DD */
1504
1505                skb = buffer_info->skb;
1506                buffer_info->skb = NULL;
1507
1508                ++i;
1509                if (i == rx_ring->count)
1510                        i = 0;
1511                next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1512                prefetch(next_rxd);
1513
1514                next_buffer = &rx_ring->buffer_info[i];
1515
1516                cleaned = true;
1517                cleaned_count++;
1518                dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1519                               DMA_FROM_DEVICE);
1520                buffer_info->dma = 0;
1521
1522                length = le16_to_cpu(rx_desc->wb.upper.length);
1523
1524                /* errors is only valid for DD + EOP descriptors */
1525                if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1526                             ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1527                              !(netdev->features & NETIF_F_RXALL)))) {
1528                        /* recycle both page and skb */
1529                        buffer_info->skb = skb;
1530                        /* an error means any chain goes out the window too */
1531                        if (rx_ring->rx_skb_top)
1532                                dev_kfree_skb_irq(rx_ring->rx_skb_top);
1533                        rx_ring->rx_skb_top = NULL;
1534                        goto next_desc;
1535                }
1536#define rxtop (rx_ring->rx_skb_top)
1537                if (!(staterr & E1000_RXD_STAT_EOP)) {
1538                        /* this descriptor is only the beginning (or middle) */
1539                        if (!rxtop) {
1540                                /* this is the beginning of a chain */
1541                                rxtop = skb;
1542                                skb_fill_page_desc(rxtop, 0, buffer_info->page,
1543                                                   0, length);
1544                        } else {
1545                                /* this is the middle of a chain */
1546                                shinfo = skb_shinfo(rxtop);
1547                                skb_fill_page_desc(rxtop, shinfo->nr_frags,
1548                                                   buffer_info->page, 0,
1549                                                   length);
1550                                /* re-use the skb, only consumed the page */
1551                                buffer_info->skb = skb;
1552                        }
1553                        e1000_consume_page(buffer_info, rxtop, length);
1554                        goto next_desc;
1555                } else {
1556                        if (rxtop) {
1557                                /* end of the chain */
1558                                shinfo = skb_shinfo(rxtop);
1559                                skb_fill_page_desc(rxtop, shinfo->nr_frags,
1560                                                   buffer_info->page, 0,
1561                                                   length);
1562                                /* re-use the current skb, we only consumed the
1563                                 * page
1564                                 */
1565                                buffer_info->skb = skb;
1566                                skb = rxtop;
1567                                rxtop = NULL;
1568                                e1000_consume_page(buffer_info, skb, length);
1569                        } else {
1570                                /* no chain, got EOP, this buf is the packet
1571                                 * copybreak to save the put_page/alloc_page
1572                                 */
1573                                if (length <= copybreak &&
1574                                    skb_tailroom(skb) >= length) {
1575                                        u8 *vaddr;
1576                                        vaddr = kmap_atomic(buffer_info->page);
1577                                        memcpy(skb_tail_pointer(skb), vaddr,
1578                                               length);
1579                                        kunmap_atomic(vaddr);
1580                                        /* re-use the page, so don't erase
1581                                         * buffer_info->page
1582                                         */
1583                                        skb_put(skb, length);
1584                                } else {
1585                                        skb_fill_page_desc(skb, 0,
1586                                                           buffer_info->page, 0,
1587                                                           length);
1588                                        e1000_consume_page(buffer_info, skb,
1589                                                           length);
1590                                }
1591                        }
1592                }
1593
1594                /* Receive Checksum Offload */
1595                e1000_rx_checksum(adapter, staterr, skb);
1596
1597                e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1598
1599                /* probably a little skewed due to removing CRC */
1600                total_rx_bytes += skb->len;
1601                total_rx_packets++;
1602
1603                /* eth type trans needs skb->data to point to something */
1604                if (!pskb_may_pull(skb, ETH_HLEN)) {
1605                        e_err("pskb_may_pull failed.\n");
1606                        dev_kfree_skb_irq(skb);
1607                        goto next_desc;
1608                }
1609
1610                e1000_receive_skb(adapter, netdev, skb, staterr,
1611                                  rx_desc->wb.upper.vlan);
1612
1613next_desc:
1614                rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1615
1616                /* return some buffers to hardware, one at a time is too slow */
1617                if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1618                        adapter->alloc_rx_buf(rx_ring, cleaned_count,
1619                                              GFP_ATOMIC);
1620                        cleaned_count = 0;
1621                }
1622
1623                /* use prefetched values */
1624                rx_desc = next_rxd;
1625                buffer_info = next_buffer;
1626
1627                staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1628        }
1629        rx_ring->next_to_clean = i;
1630
1631        cleaned_count = e1000_desc_unused(rx_ring);
1632        if (cleaned_count)
1633                adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1634
1635        adapter->total_rx_bytes += total_rx_bytes;
1636        adapter->total_rx_packets += total_rx_packets;
1637        return cleaned;
1638}
1639
1640/**
1641 * e1000_clean_rx_ring - Free Rx Buffers per Queue
1642 * @rx_ring: Rx descriptor ring
1643 **/
1644static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1645{
1646        struct e1000_adapter *adapter = rx_ring->adapter;
1647        struct e1000_buffer *buffer_info;
1648        struct e1000_ps_page *ps_page;
1649        struct pci_dev *pdev = adapter->pdev;
1650        unsigned int i, j;
1651
1652        /* Free all the Rx ring sk_buffs */
1653        for (i = 0; i < rx_ring->count; i++) {
1654                buffer_info = &rx_ring->buffer_info[i];
1655                if (buffer_info->dma) {
1656                        if (adapter->clean_rx == e1000_clean_rx_irq)
1657                                dma_unmap_single(&pdev->dev, buffer_info->dma,
1658                                                 adapter->rx_buffer_len,
1659                                                 DMA_FROM_DEVICE);
1660                        else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1661                                dma_unmap_page(&pdev->dev, buffer_info->dma,
1662                                               PAGE_SIZE, DMA_FROM_DEVICE);
1663                        else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1664                                dma_unmap_single(&pdev->dev, buffer_info->dma,
1665                                                 adapter->rx_ps_bsize0,
1666                                                 DMA_FROM_DEVICE);
1667                        buffer_info->dma = 0;
1668                }
1669
1670                if (buffer_info->page) {
1671                        put_page(buffer_info->page);
1672                        buffer_info->page = NULL;
1673                }
1674
1675                if (buffer_info->skb) {
1676                        dev_kfree_skb(buffer_info->skb);
1677                        buffer_info->skb = NULL;
1678                }
1679
1680                for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1681                        ps_page = &buffer_info->ps_pages[j];
1682                        if (!ps_page->page)
1683                                break;
1684                        dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1685                                       DMA_FROM_DEVICE);
1686                        ps_page->dma = 0;
1687                        put_page(ps_page->page);
1688                        ps_page->page = NULL;
1689                }
1690        }
1691
1692        /* there also may be some cached data from a chained receive */
1693        if (rx_ring->rx_skb_top) {
1694                dev_kfree_skb(rx_ring->rx_skb_top);
1695                rx_ring->rx_skb_top = NULL;
1696        }
1697
1698        /* Zero out the descriptor ring */
1699        memset(rx_ring->desc, 0, rx_ring->size);
1700
1701        rx_ring->next_to_clean = 0;
1702        rx_ring->next_to_use = 0;
1703        adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1704
1705        writel(0, rx_ring->head);
1706        if (rx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
1707                e1000e_update_rdt_wa(rx_ring, 0);
1708        else
1709                writel(0, rx_ring->tail);
1710}
1711
1712static void e1000e_downshift_workaround(struct work_struct *work)
1713{
1714        struct e1000_adapter *adapter = container_of(work,
1715                                                     struct e1000_adapter,
1716                                                     downshift_task);
1717
1718        if (test_bit(__E1000_DOWN, &adapter->state))
1719                return;
1720
1721        e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1722}
1723
1724/**
1725 * e1000_intr_msi - Interrupt Handler
1726 * @irq: interrupt number
1727 * @data: pointer to a network interface device structure
1728 **/
1729static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1730{
1731        struct net_device *netdev = data;
1732        struct e1000_adapter *adapter = netdev_priv(netdev);
1733        struct e1000_hw *hw = &adapter->hw;
1734        u32 icr = er32(ICR);
1735
1736        /* read ICR disables interrupts using IAM */
1737        if (icr & E1000_ICR_LSC) {
1738                hw->mac.get_link_status = true;
1739                /* ICH8 workaround-- Call gig speed drop workaround on cable
1740                 * disconnect (LSC) before accessing any PHY registers
1741                 */
1742                if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1743                    (!(er32(STATUS) & E1000_STATUS_LU)))
1744                        schedule_work(&adapter->downshift_task);
1745
1746                /* 80003ES2LAN workaround-- For packet buffer work-around on
1747                 * link down event; disable receives here in the ISR and reset
1748                 * adapter in watchdog
1749                 */
1750                if (netif_carrier_ok(netdev) &&
1751                    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1752                        /* disable receives */
1753                        u32 rctl = er32(RCTL);
1754                        ew32(RCTL, rctl & ~E1000_RCTL_EN);
1755                        adapter->flags |= FLAG_RESTART_NOW;
1756                }
1757                /* guard against interrupt when we're going down */
1758                if (!test_bit(__E1000_DOWN, &adapter->state))
1759                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
1760        }
1761
1762        /* Reset on uncorrectable ECC error */
1763        if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1764                u32 pbeccsts = er32(PBECCSTS);
1765
1766                adapter->corr_errors +=
1767                    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1768                adapter->uncorr_errors +=
1769                    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1770                    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1771
1772                /* Do the reset outside of interrupt context */
1773                schedule_work(&adapter->reset_task);
1774
1775                /* return immediately since reset is imminent */
1776                return IRQ_HANDLED;
1777        }
1778
1779        if (napi_schedule_prep(&adapter->napi)) {
1780                adapter->total_tx_bytes = 0;
1781                adapter->total_tx_packets = 0;
1782                adapter->total_rx_bytes = 0;
1783                adapter->total_rx_packets = 0;
1784                __napi_schedule(&adapter->napi);
1785        }
1786
1787        return IRQ_HANDLED;
1788}
1789
1790/**
1791 * e1000_intr - Interrupt Handler
1792 * @irq: interrupt number
1793 * @data: pointer to a network interface device structure
1794 **/
1795static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1796{
1797        struct net_device *netdev = data;
1798        struct e1000_adapter *adapter = netdev_priv(netdev);
1799        struct e1000_hw *hw = &adapter->hw;
1800        u32 rctl, icr = er32(ICR);
1801
1802        if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1803                return IRQ_NONE;        /* Not our interrupt */
1804
1805        /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1806         * not set, then the adapter didn't send an interrupt
1807         */
1808        if (!(icr & E1000_ICR_INT_ASSERTED))
1809                return IRQ_NONE;
1810
1811        /* Interrupt Auto-Mask...upon reading ICR,
1812         * interrupts are masked.  No need for the
1813         * IMC write
1814         */
1815
1816        if (icr & E1000_ICR_LSC) {
1817                hw->mac.get_link_status = true;
1818                /* ICH8 workaround-- Call gig speed drop workaround on cable
1819                 * disconnect (LSC) before accessing any PHY registers
1820                 */
1821                if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1822                    (!(er32(STATUS) & E1000_STATUS_LU)))
1823                        schedule_work(&adapter->downshift_task);
1824
1825                /* 80003ES2LAN workaround--
1826                 * For packet buffer work-around on link down event;
1827                 * disable receives here in the ISR and
1828                 * reset adapter in watchdog
1829                 */
1830                if (netif_carrier_ok(netdev) &&
1831                    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1832                        /* disable receives */
1833                        rctl = er32(RCTL);
1834                        ew32(RCTL, rctl & ~E1000_RCTL_EN);
1835                        adapter->flags |= FLAG_RESTART_NOW;
1836                }
1837                /* guard against interrupt when we're going down */
1838                if (!test_bit(__E1000_DOWN, &adapter->state))
1839                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
1840        }
1841
1842        /* Reset on uncorrectable ECC error */
1843        if ((icr & E1000_ICR_ECCER) && (hw->mac.type == e1000_pch_lpt)) {
1844                u32 pbeccsts = er32(PBECCSTS);
1845
1846                adapter->corr_errors +=
1847                    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1848                adapter->uncorr_errors +=
1849                    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
1850                    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
1851
1852                /* Do the reset outside of interrupt context */
1853                schedule_work(&adapter->reset_task);
1854
1855                /* return immediately since reset is imminent */
1856                return IRQ_HANDLED;
1857        }
1858
1859        if (napi_schedule_prep(&adapter->napi)) {
1860                adapter->total_tx_bytes = 0;
1861                adapter->total_tx_packets = 0;
1862                adapter->total_rx_bytes = 0;
1863                adapter->total_rx_packets = 0;
1864                __napi_schedule(&adapter->napi);
1865        }
1866
1867        return IRQ_HANDLED;
1868}
1869
1870static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1871{
1872        struct net_device *netdev = data;
1873        struct e1000_adapter *adapter = netdev_priv(netdev);
1874        struct e1000_hw *hw = &adapter->hw;
1875        u32 icr = er32(ICR);
1876
1877        if (!(icr & E1000_ICR_INT_ASSERTED)) {
1878                if (!test_bit(__E1000_DOWN, &adapter->state))
1879                        ew32(IMS, E1000_IMS_OTHER);
1880                return IRQ_NONE;
1881        }
1882
1883        if (icr & adapter->eiac_mask)
1884                ew32(ICS, (icr & adapter->eiac_mask));
1885
1886        if (icr & E1000_ICR_OTHER) {
1887                if (!(icr & E1000_ICR_LSC))
1888                        goto no_link_interrupt;
1889                hw->mac.get_link_status = true;
1890                /* guard against interrupt when we're going down */
1891                if (!test_bit(__E1000_DOWN, &adapter->state))
1892                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
1893        }
1894
1895no_link_interrupt:
1896        if (!test_bit(__E1000_DOWN, &adapter->state))
1897                ew32(IMS, E1000_IMS_LSC | E1000_IMS_OTHER);
1898
1899        return IRQ_HANDLED;
1900}
1901
1902static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1903{
1904        struct net_device *netdev = data;
1905        struct e1000_adapter *adapter = netdev_priv(netdev);
1906        struct e1000_hw *hw = &adapter->hw;
1907        struct e1000_ring *tx_ring = adapter->tx_ring;
1908
1909        adapter->total_tx_bytes = 0;
1910        adapter->total_tx_packets = 0;
1911
1912        if (!e1000_clean_tx_irq(tx_ring))
1913                /* Ring was not completely cleaned, so fire another interrupt */
1914                ew32(ICS, tx_ring->ims_val);
1915
1916        return IRQ_HANDLED;
1917}
1918
1919static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1920{
1921        struct net_device *netdev = data;
1922        struct e1000_adapter *adapter = netdev_priv(netdev);
1923        struct e1000_ring *rx_ring = adapter->rx_ring;
1924
1925        /* Write the ITR value calculated at the end of the
1926         * previous interrupt.
1927         */
1928        if (rx_ring->set_itr) {
1929                writel(1000000000 / (rx_ring->itr_val * 256),
1930                       rx_ring->itr_register);
1931                rx_ring->set_itr = 0;
1932        }
1933
1934        if (napi_schedule_prep(&adapter->napi)) {
1935                adapter->total_rx_bytes = 0;
1936                adapter->total_rx_packets = 0;
1937                __napi_schedule(&adapter->napi);
1938        }
1939        return IRQ_HANDLED;
1940}
1941
1942/**
1943 * e1000_configure_msix - Configure MSI-X hardware
1944 *
1945 * e1000_configure_msix sets up the hardware to properly
1946 * generate MSI-X interrupts.
1947 **/
1948static void e1000_configure_msix(struct e1000_adapter *adapter)
1949{
1950        struct e1000_hw *hw = &adapter->hw;
1951        struct e1000_ring *rx_ring = adapter->rx_ring;
1952        struct e1000_ring *tx_ring = adapter->tx_ring;
1953        int vector = 0;
1954        u32 ctrl_ext, ivar = 0;
1955
1956        adapter->eiac_mask = 0;
1957
1958        /* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1959        if (hw->mac.type == e1000_82574) {
1960                u32 rfctl = er32(RFCTL);
1961                rfctl |= E1000_RFCTL_ACK_DIS;
1962                ew32(RFCTL, rfctl);
1963        }
1964
1965        /* Configure Rx vector */
1966        rx_ring->ims_val = E1000_IMS_RXQ0;
1967        adapter->eiac_mask |= rx_ring->ims_val;
1968        if (rx_ring->itr_val)
1969                writel(1000000000 / (rx_ring->itr_val * 256),
1970                       rx_ring->itr_register);
1971        else
1972                writel(1, rx_ring->itr_register);
1973        ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1974
1975        /* Configure Tx vector */
1976        tx_ring->ims_val = E1000_IMS_TXQ0;
1977        vector++;
1978        if (tx_ring->itr_val)
1979                writel(1000000000 / (tx_ring->itr_val * 256),
1980                       tx_ring->itr_register);
1981        else
1982                writel(1, tx_ring->itr_register);
1983        adapter->eiac_mask |= tx_ring->ims_val;
1984        ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
1985
1986        /* set vector for Other Causes, e.g. link changes */
1987        vector++;
1988        ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
1989        if (rx_ring->itr_val)
1990                writel(1000000000 / (rx_ring->itr_val * 256),
1991                       hw->hw_addr + E1000_EITR_82574(vector));
1992        else
1993                writel(1, hw->hw_addr + E1000_EITR_82574(vector));
1994
1995        /* Cause Tx interrupts on every write back */
1996        ivar |= (1 << 31);
1997
1998        ew32(IVAR, ivar);
1999
2000        /* enable MSI-X PBA support */
2001        ctrl_ext = er32(CTRL_EXT);
2002        ctrl_ext |= E1000_CTRL_EXT_PBA_CLR;
2003
2004        /* Auto-Mask Other interrupts upon ICR read */
2005        ew32(IAM, ~E1000_EIAC_MASK_82574 | E1000_IMS_OTHER);
2006        ctrl_ext |= E1000_CTRL_EXT_EIAME;
2007        ew32(CTRL_EXT, ctrl_ext);
2008        e1e_flush();
2009}
2010
2011void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2012{
2013        if (adapter->msix_entries) {
2014                pci_disable_msix(adapter->pdev);
2015                kfree(adapter->msix_entries);
2016                adapter->msix_entries = NULL;
2017        } else if (adapter->flags & FLAG_MSI_ENABLED) {
2018                pci_disable_msi(adapter->pdev);
2019                adapter->flags &= ~FLAG_MSI_ENABLED;
2020        }
2021}
2022
2023/**
2024 * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2025 *
2026 * Attempt to configure interrupts using the best available
2027 * capabilities of the hardware and kernel.
2028 **/
2029void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2030{
2031        int err;
2032        int i;
2033
2034        switch (adapter->int_mode) {
2035        case E1000E_INT_MODE_MSIX:
2036                if (adapter->flags & FLAG_HAS_MSIX) {
2037                        adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2038                        adapter->msix_entries = kcalloc(adapter->num_vectors,
2039                                                        sizeof(struct
2040                                                               msix_entry),
2041                                                        GFP_KERNEL);
2042                        if (adapter->msix_entries) {
2043                                for (i = 0; i < adapter->num_vectors; i++)
2044                                        adapter->msix_entries[i].entry = i;
2045
2046                                err = pci_enable_msix(adapter->pdev,
2047                                                      adapter->msix_entries,
2048                                                      adapter->num_vectors);
2049                                if (err == 0)
2050                                        return;
2051                        }
2052                        /* MSI-X failed, so fall through and try MSI */
2053                        e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2054                        e1000e_reset_interrupt_capability(adapter);
2055                }
2056                adapter->int_mode = E1000E_INT_MODE_MSI;
2057                /* Fall through */
2058        case E1000E_INT_MODE_MSI:
2059                if (!pci_enable_msi(adapter->pdev)) {
2060                        adapter->flags |= FLAG_MSI_ENABLED;
2061                } else {
2062                        adapter->int_mode = E1000E_INT_MODE_LEGACY;
2063                        e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2064                }
2065                /* Fall through */
2066        case E1000E_INT_MODE_LEGACY:
2067                /* Don't do anything; this is the system default */
2068                break;
2069        }
2070
2071        /* store the number of vectors being used */
2072        adapter->num_vectors = 1;
2073}
2074
2075/**
2076 * e1000_request_msix - Initialize MSI-X interrupts
2077 *
2078 * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2079 * kernel.
2080 **/
2081static int e1000_request_msix(struct e1000_adapter *adapter)
2082{
2083        struct net_device *netdev = adapter->netdev;
2084        int err = 0, vector = 0;
2085
2086        if (strlen(netdev->name) < (IFNAMSIZ - 5))
2087                snprintf(adapter->rx_ring->name,
2088                         sizeof(adapter->rx_ring->name) - 1,
2089                         "%s-rx-0", netdev->name);
2090        else
2091                memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2092        err = request_irq(adapter->msix_entries[vector].vector,
2093                          e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2094                          netdev);
2095        if (err)
2096                return err;
2097        adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2098            E1000_EITR_82574(vector);
2099        adapter->rx_ring->itr_val = adapter->itr;
2100        vector++;
2101
2102        if (strlen(netdev->name) < (IFNAMSIZ - 5))
2103                snprintf(adapter->tx_ring->name,
2104                         sizeof(adapter->tx_ring->name) - 1,
2105                         "%s-tx-0", netdev->name);
2106        else
2107                memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2108        err = request_irq(adapter->msix_entries[vector].vector,
2109                          e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2110                          netdev);
2111        if (err)
2112                return err;
2113        adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2114            E1000_EITR_82574(vector);
2115        adapter->tx_ring->itr_val = adapter->itr;
2116        vector++;
2117
2118        err = request_irq(adapter->msix_entries[vector].vector,
2119                          e1000_msix_other, 0, netdev->name, netdev);
2120        if (err)
2121                return err;
2122
2123        e1000_configure_msix(adapter);
2124
2125        return 0;
2126}
2127
2128/**
2129 * e1000_request_irq - initialize interrupts
2130 *
2131 * Attempts to configure interrupts using the best available
2132 * capabilities of the hardware and kernel.
2133 **/
2134static int e1000_request_irq(struct e1000_adapter *adapter)
2135{
2136        struct net_device *netdev = adapter->netdev;
2137        int err;
2138
2139        if (adapter->msix_entries) {
2140                err = e1000_request_msix(adapter);
2141                if (!err)
2142                        return err;
2143                /* fall back to MSI */
2144                e1000e_reset_interrupt_capability(adapter);
2145                adapter->int_mode = E1000E_INT_MODE_MSI;
2146                e1000e_set_interrupt_capability(adapter);
2147        }
2148        if (adapter->flags & FLAG_MSI_ENABLED) {
2149                err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2150                                  netdev->name, netdev);
2151                if (!err)
2152                        return err;
2153
2154                /* fall back to legacy interrupt */
2155                e1000e_reset_interrupt_capability(adapter);
2156                adapter->int_mode = E1000E_INT_MODE_LEGACY;
2157        }
2158
2159        err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2160                          netdev->name, netdev);
2161        if (err)
2162                e_err("Unable to allocate interrupt, Error: %d\n", err);
2163
2164        return err;
2165}
2166
2167static void e1000_free_irq(struct e1000_adapter *adapter)
2168{
2169        struct net_device *netdev = adapter->netdev;
2170
2171        if (adapter->msix_entries) {
2172                int vector = 0;
2173
2174                free_irq(adapter->msix_entries[vector].vector, netdev);
2175                vector++;
2176
2177                free_irq(adapter->msix_entries[vector].vector, netdev);
2178                vector++;
2179
2180                /* Other Causes interrupt vector */
2181                free_irq(adapter->msix_entries[vector].vector, netdev);
2182                return;
2183        }
2184
2185        free_irq(adapter->pdev->irq, netdev);
2186}
2187
2188/**
2189 * e1000_irq_disable - Mask off interrupt generation on the NIC
2190 **/
2191static void e1000_irq_disable(struct e1000_adapter *adapter)
2192{
2193        struct e1000_hw *hw = &adapter->hw;
2194
2195        ew32(IMC, ~0);
2196        if (adapter->msix_entries)
2197                ew32(EIAC_82574, 0);
2198        e1e_flush();
2199
2200        if (adapter->msix_entries) {
2201                int i;
2202                for (i = 0; i < adapter->num_vectors; i++)
2203                        synchronize_irq(adapter->msix_entries[i].vector);
2204        } else {
2205                synchronize_irq(adapter->pdev->irq);
2206        }
2207}
2208
2209/**
2210 * e1000_irq_enable - Enable default interrupt generation settings
2211 **/
2212static void e1000_irq_enable(struct e1000_adapter *adapter)
2213{
2214        struct e1000_hw *hw = &adapter->hw;
2215
2216        if (adapter->msix_entries) {
2217                ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2218                ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER | E1000_IMS_LSC);
2219        } else if (hw->mac.type == e1000_pch_lpt) {
2220                ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2221        } else {
2222                ew32(IMS, IMS_ENABLE_MASK);
2223        }
2224        e1e_flush();
2225}
2226
2227/**
2228 * e1000e_get_hw_control - get control of the h/w from f/w
2229 * @adapter: address of board private structure
2230 *
2231 * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2232 * For ASF and Pass Through versions of f/w this means that
2233 * the driver is loaded. For AMT version (only with 82573)
2234 * of the f/w this means that the network i/f is open.
2235 **/
2236void e1000e_get_hw_control(struct e1000_adapter *adapter)
2237{
2238        struct e1000_hw *hw = &adapter->hw;
2239        u32 ctrl_ext;
2240        u32 swsm;
2241
2242        /* Let firmware know the driver has taken over */
2243        if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2244                swsm = er32(SWSM);
2245                ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2246        } else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2247                ctrl_ext = er32(CTRL_EXT);
2248                ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2249        }
2250}
2251
2252/**
2253 * e1000e_release_hw_control - release control of the h/w to f/w
2254 * @adapter: address of board private structure
2255 *
2256 * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2257 * For ASF and Pass Through versions of f/w this means that the
2258 * driver is no longer loaded. For AMT version (only with 82573) i
2259 * of the f/w this means that the network i/f is closed.
2260 *
2261 **/
2262void e1000e_release_hw_control(struct e1000_adapter *adapter)
2263{
2264        struct e1000_hw *hw = &adapter->hw;
2265        u32 ctrl_ext;
2266        u32 swsm;
2267
2268        /* Let firmware taken over control of h/w */
2269        if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2270                swsm = er32(SWSM);
2271                ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2272        } else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2273                ctrl_ext = er32(CTRL_EXT);
2274                ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2275        }
2276}
2277
2278/**
2279 * e1000_alloc_ring_dma - allocate memory for a ring structure
2280 **/
2281static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2282                                struct e1000_ring *ring)
2283{
2284        struct pci_dev *pdev = adapter->pdev;
2285
2286        ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2287                                        GFP_KERNEL);
2288        if (!ring->desc)
2289                return -ENOMEM;
2290
2291        return 0;
2292}
2293
2294/**
2295 * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2296 * @tx_ring: Tx descriptor ring
2297 *
2298 * Return 0 on success, negative on failure
2299 **/
2300int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2301{
2302        struct e1000_adapter *adapter = tx_ring->adapter;
2303        int err = -ENOMEM, size;
2304
2305        size = sizeof(struct e1000_buffer) * tx_ring->count;
2306        tx_ring->buffer_info = vzalloc(size);
2307        if (!tx_ring->buffer_info)
2308                goto err;
2309
2310        /* round up to nearest 4K */
2311        tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2312        tx_ring->size = ALIGN(tx_ring->size, 4096);
2313
2314        err = e1000_alloc_ring_dma(adapter, tx_ring);
2315        if (err)
2316                goto err;
2317
2318        tx_ring->next_to_use = 0;
2319        tx_ring->next_to_clean = 0;
2320
2321        return 0;
2322err:
2323        vfree(tx_ring->buffer_info);
2324        e_err("Unable to allocate memory for the transmit descriptor ring\n");
2325        return err;
2326}
2327
2328/**
2329 * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2330 * @rx_ring: Rx descriptor ring
2331 *
2332 * Returns 0 on success, negative on failure
2333 **/
2334int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2335{
2336        struct e1000_adapter *adapter = rx_ring->adapter;
2337        struct e1000_buffer *buffer_info;
2338        int i, size, desc_len, err = -ENOMEM;
2339
2340        size = sizeof(struct e1000_buffer) * rx_ring->count;
2341        rx_ring->buffer_info = vzalloc(size);
2342        if (!rx_ring->buffer_info)
2343                goto err;
2344
2345        for (i = 0; i < rx_ring->count; i++) {
2346                buffer_info = &rx_ring->buffer_info[i];
2347                buffer_info->ps_pages = kcalloc(PS_PAGE_BUFFERS,
2348                                                sizeof(struct e1000_ps_page),
2349                                                GFP_KERNEL);
2350                if (!buffer_info->ps_pages)
2351                        goto err_pages;
2352        }
2353
2354        desc_len = sizeof(union e1000_rx_desc_packet_split);
2355
2356        /* Round up to nearest 4K */
2357        rx_ring->size = rx_ring->count * desc_len;
2358        rx_ring->size = ALIGN(rx_ring->size, 4096);
2359
2360        err = e1000_alloc_ring_dma(adapter, rx_ring);
2361        if (err)
2362                goto err_pages;
2363
2364        rx_ring->next_to_clean = 0;
2365        rx_ring->next_to_use = 0;
2366        rx_ring->rx_skb_top = NULL;
2367
2368        return 0;
2369
2370err_pages:
2371        for (i = 0; i < rx_ring->count; i++) {
2372                buffer_info = &rx_ring->buffer_info[i];
2373                kfree(buffer_info->ps_pages);
2374        }
2375err:
2376        vfree(rx_ring->buffer_info);
2377        e_err("Unable to allocate memory for the receive descriptor ring\n");
2378        return err;
2379}
2380
2381/**
2382 * e1000_clean_tx_ring - Free Tx Buffers
2383 * @tx_ring: Tx descriptor ring
2384 **/
2385static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2386{
2387        struct e1000_adapter *adapter = tx_ring->adapter;
2388        struct e1000_buffer *buffer_info;
2389        unsigned long size;
2390        unsigned int i;
2391
2392        for (i = 0; i < tx_ring->count; i++) {
2393                buffer_info = &tx_ring->buffer_info[i];
2394                e1000_put_txbuf(tx_ring, buffer_info);
2395        }
2396
2397        netdev_reset_queue(adapter->netdev);
2398        size = sizeof(struct e1000_buffer) * tx_ring->count;
2399        memset(tx_ring->buffer_info, 0, size);
2400
2401        memset(tx_ring->desc, 0, tx_ring->size);
2402
2403        tx_ring->next_to_use = 0;
2404        tx_ring->next_to_clean = 0;
2405
2406        writel(0, tx_ring->head);
2407        if (tx_ring->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2408                e1000e_update_tdt_wa(tx_ring, 0);
2409        else
2410                writel(0, tx_ring->tail);
2411}
2412
2413/**
2414 * e1000e_free_tx_resources - Free Tx Resources per Queue
2415 * @tx_ring: Tx descriptor ring
2416 *
2417 * Free all transmit software resources
2418 **/
2419void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2420{
2421        struct e1000_adapter *adapter = tx_ring->adapter;
2422        struct pci_dev *pdev = adapter->pdev;
2423
2424        e1000_clean_tx_ring(tx_ring);
2425
2426        vfree(tx_ring->buffer_info);
2427        tx_ring->buffer_info = NULL;
2428
2429        dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2430                          tx_ring->dma);
2431        tx_ring->desc = NULL;
2432}
2433
2434/**
2435 * e1000e_free_rx_resources - Free Rx Resources
2436 * @rx_ring: Rx descriptor ring
2437 *
2438 * Free all receive software resources
2439 **/
2440void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2441{
2442        struct e1000_adapter *adapter = rx_ring->adapter;
2443        struct pci_dev *pdev = adapter->pdev;
2444        int i;
2445
2446        e1000_clean_rx_ring(rx_ring);
2447
2448        for (i = 0; i < rx_ring->count; i++)
2449                kfree(rx_ring->buffer_info[i].ps_pages);
2450
2451        vfree(rx_ring->buffer_info);
2452        rx_ring->buffer_info = NULL;
2453
2454        dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2455                          rx_ring->dma);
2456        rx_ring->desc = NULL;
2457}
2458
2459/**
2460 * e1000_update_itr - update the dynamic ITR value based on statistics
2461 * @adapter: pointer to adapter
2462 * @itr_setting: current adapter->itr
2463 * @packets: the number of packets during this measurement interval
2464 * @bytes: the number of bytes during this measurement interval
2465 *
2466 *      Stores a new ITR value based on packets and byte
2467 *      counts during the last interrupt.  The advantage of per interrupt
2468 *      computation is faster updates and more accurate ITR for the current
2469 *      traffic pattern.  Constants in this function were computed
2470 *      based on theoretical maximum wire speed and thresholds were set based
2471 *      on testing data as well as attempting to minimize response time
2472 *      while increasing bulk throughput.  This functionality is controlled
2473 *      by the InterruptThrottleRate module parameter.
2474 **/
2475static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2476{
2477        unsigned int retval = itr_setting;
2478
2479        if (packets == 0)
2480                return itr_setting;
2481
2482        switch (itr_setting) {
2483        case lowest_latency:
2484                /* handle TSO and jumbo frames */
2485                if (bytes / packets > 8000)
2486                        retval = bulk_latency;
2487                else if ((packets < 5) && (bytes > 512))
2488                        retval = low_latency;
2489                break;
2490        case low_latency:       /* 50 usec aka 20000 ints/s */
2491                if (bytes > 10000) {
2492                        /* this if handles the TSO accounting */
2493                        if (bytes / packets > 8000)
2494                                retval = bulk_latency;
2495                        else if ((packets < 10) || ((bytes / packets) > 1200))
2496                                retval = bulk_latency;
2497                        else if ((packets > 35))
2498                                retval = lowest_latency;
2499                } else if (bytes / packets > 2000) {
2500                        retval = bulk_latency;
2501                } else if (packets <= 2 && bytes < 512) {
2502                        retval = lowest_latency;
2503                }
2504                break;
2505        case bulk_latency:      /* 250 usec aka 4000 ints/s */
2506                if (bytes > 25000) {
2507                        if (packets > 35)
2508                                retval = low_latency;
2509                } else if (bytes < 6000) {
2510                        retval = low_latency;
2511                }
2512                break;
2513        }
2514
2515        return retval;
2516}
2517
2518static void e1000_set_itr(struct e1000_adapter *adapter)
2519{
2520        u16 current_itr;
2521        u32 new_itr = adapter->itr;
2522
2523        /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2524        if (adapter->link_speed != SPEED_1000) {
2525                current_itr = 0;
2526                new_itr = 4000;
2527                goto set_itr_now;
2528        }
2529
2530        if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2531                new_itr = 0;
2532                goto set_itr_now;
2533        }
2534
2535        adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2536                                           adapter->total_tx_packets,
2537                                           adapter->total_tx_bytes);
2538        /* conservative mode (itr 3) eliminates the lowest_latency setting */
2539        if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2540                adapter->tx_itr = low_latency;
2541
2542        adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2543                                           adapter->total_rx_packets,
2544                                           adapter->total_rx_bytes);
2545        /* conservative mode (itr 3) eliminates the lowest_latency setting */
2546        if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2547                adapter->rx_itr = low_latency;
2548
2549        current_itr = max(adapter->rx_itr, adapter->tx_itr);
2550
2551        /* counts and packets in update_itr are dependent on these numbers */
2552        switch (current_itr) {
2553        case lowest_latency:
2554                new_itr = 70000;
2555                break;
2556        case low_latency:
2557                new_itr = 20000;        /* aka hwitr = ~200 */
2558                break;
2559        case bulk_latency:
2560                new_itr = 4000;
2561                break;
2562        default:
2563                break;
2564        }
2565
2566set_itr_now:
2567        if (new_itr != adapter->itr) {
2568                /* this attempts to bias the interrupt rate towards Bulk
2569                 * by adding intermediate steps when interrupt rate is
2570                 * increasing
2571                 */
2572                new_itr = new_itr > adapter->itr ?
2573                    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2574                adapter->itr = new_itr;
2575                adapter->rx_ring->itr_val = new_itr;
2576                if (adapter->msix_entries)
2577                        adapter->rx_ring->set_itr = 1;
2578                else
2579                        e1000e_write_itr(adapter, new_itr);
2580        }
2581}
2582
2583/**
2584 * e1000e_write_itr - write the ITR value to the appropriate registers
2585 * @adapter: address of board private structure
2586 * @itr: new ITR value to program
2587 *
2588 * e1000e_write_itr determines if the adapter is in MSI-X mode
2589 * and, if so, writes the EITR registers with the ITR value.
2590 * Otherwise, it writes the ITR value into the ITR register.
2591 **/
2592void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2593{
2594        struct e1000_hw *hw = &adapter->hw;
2595        u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2596
2597        if (adapter->msix_entries) {
2598                int vector;
2599
2600                for (vector = 0; vector < adapter->num_vectors; vector++)
2601                        writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2602        } else {
2603                ew32(ITR, new_itr);
2604        }
2605}
2606
2607/**
2608 * e1000_alloc_queues - Allocate memory for all rings
2609 * @adapter: board private structure to initialize
2610 **/
2611static int e1000_alloc_queues(struct e1000_adapter *adapter)
2612{
2613        int size = sizeof(struct e1000_ring);
2614
2615        adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2616        if (!adapter->tx_ring)
2617                goto err;
2618        adapter->tx_ring->count = adapter->tx_ring_count;
2619        adapter->tx_ring->adapter = adapter;
2620
2621        adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2622        if (!adapter->rx_ring)
2623                goto err;
2624        adapter->rx_ring->count = adapter->rx_ring_count;
2625        adapter->rx_ring->adapter = adapter;
2626
2627        return 0;
2628err:
2629        e_err("Unable to allocate memory for queues\n");
2630        kfree(adapter->rx_ring);
2631        kfree(adapter->tx_ring);
2632        return -ENOMEM;
2633}
2634
2635/**
2636 * e1000e_poll - NAPI Rx polling callback
2637 * @napi: struct associated with this polling callback
2638 * @weight: number of packets driver is allowed to process this poll
2639 **/
2640static int e1000e_poll(struct napi_struct *napi, int weight)
2641{
2642        struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2643                                                     napi);
2644        struct e1000_hw *hw = &adapter->hw;
2645        struct net_device *poll_dev = adapter->netdev;
2646        int tx_cleaned = 1, work_done = 0;
2647
2648        adapter = netdev_priv(poll_dev);
2649
2650        if (!adapter->msix_entries ||
2651            (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2652                tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2653
2654        adapter->clean_rx(adapter->rx_ring, &work_done, weight);
2655
2656        if (!tx_cleaned)
2657                work_done = weight;
2658
2659        /* If weight not fully consumed, exit the polling mode */
2660        if (work_done < weight) {
2661                if (adapter->itr_setting & 3)
2662                        e1000_set_itr(adapter);
2663                napi_complete(napi);
2664                if (!test_bit(__E1000_DOWN, &adapter->state)) {
2665                        if (adapter->msix_entries)
2666                                ew32(IMS, adapter->rx_ring->ims_val);
2667                        else
2668                                e1000_irq_enable(adapter);
2669                }
2670        }
2671
2672        return work_done;
2673}
2674
2675static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2676                                 __always_unused __be16 proto, u16 vid)
2677{
2678        struct e1000_adapter *adapter = netdev_priv(netdev);
2679        struct e1000_hw *hw = &adapter->hw;
2680        u32 vfta, index;
2681
2682        /* don't update vlan cookie if already programmed */
2683        if ((adapter->hw.mng_cookie.status &
2684             E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2685            (vid == adapter->mng_vlan_id))
2686                return 0;
2687
2688        /* add VID to filter table */
2689        if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2690                index = (vid >> 5) & 0x7F;
2691                vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2692                vfta |= (1 << (vid & 0x1F));
2693                hw->mac.ops.write_vfta(hw, index, vfta);
2694        }
2695
2696        set_bit(vid, adapter->active_vlans);
2697
2698        return 0;
2699}
2700
2701static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2702                                  __always_unused __be16 proto, u16 vid)
2703{
2704        struct e1000_adapter *adapter = netdev_priv(netdev);
2705        struct e1000_hw *hw = &adapter->hw;
2706        u32 vfta, index;
2707
2708        if ((adapter->hw.mng_cookie.status &
2709             E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2710            (vid == adapter->mng_vlan_id)) {
2711                /* release control to f/w */
2712                e1000e_release_hw_control(adapter);
2713                return 0;
2714        }
2715
2716        /* remove VID from filter table */
2717        if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2718                index = (vid >> 5) & 0x7F;
2719                vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2720                vfta &= ~(1 << (vid & 0x1F));
2721                hw->mac.ops.write_vfta(hw, index, vfta);
2722        }
2723
2724        clear_bit(vid, adapter->active_vlans);
2725
2726        return 0;
2727}
2728
2729/**
2730 * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2731 * @adapter: board private structure to initialize
2732 **/
2733static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2734{
2735        struct net_device *netdev = adapter->netdev;
2736        struct e1000_hw *hw = &adapter->hw;
2737        u32 rctl;
2738
2739        if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2740                /* disable VLAN receive filtering */
2741                rctl = er32(RCTL);
2742                rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2743                ew32(RCTL, rctl);
2744
2745                if (adapter->mng_vlan_id != (u16)E1000_MNG_VLAN_NONE) {
2746                        e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2747                                               adapter->mng_vlan_id);
2748                        adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2749                }
2750        }
2751}
2752
2753/**
2754 * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2755 * @adapter: board private structure to initialize
2756 **/
2757static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2758{
2759        struct e1000_hw *hw = &adapter->hw;
2760        u32 rctl;
2761
2762        if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2763                /* enable VLAN receive filtering */
2764                rctl = er32(RCTL);
2765                rctl |= E1000_RCTL_VFE;
2766                rctl &= ~E1000_RCTL_CFIEN;
2767                ew32(RCTL, rctl);
2768        }
2769}
2770
2771/**
2772 * e1000e_vlan_strip_enable - helper to disable HW VLAN stripping
2773 * @adapter: board private structure to initialize
2774 **/
2775static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2776{
2777        struct e1000_hw *hw = &adapter->hw;
2778        u32 ctrl;
2779
2780        /* disable VLAN tag insert/strip */
2781        ctrl = er32(CTRL);
2782        ctrl &= ~E1000_CTRL_VME;
2783        ew32(CTRL, ctrl);
2784}
2785
2786/**
2787 * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2788 * @adapter: board private structure to initialize
2789 **/
2790static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2791{
2792        struct e1000_hw *hw = &adapter->hw;
2793        u32 ctrl;
2794
2795        /* enable VLAN tag insert/strip */
2796        ctrl = er32(CTRL);
2797        ctrl |= E1000_CTRL_VME;
2798        ew32(CTRL, ctrl);
2799}
2800
2801static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2802{
2803        struct net_device *netdev = adapter->netdev;
2804        u16 vid = adapter->hw.mng_cookie.vlan_id;
2805        u16 old_vid = adapter->mng_vlan_id;
2806
2807        if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2808                e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2809                adapter->mng_vlan_id = vid;
2810        }
2811
2812        if ((old_vid != (u16)E1000_MNG_VLAN_NONE) && (vid != old_vid))
2813                e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2814}
2815
2816static void e1000_restore_vlan(struct e1000_adapter *adapter)
2817{
2818        u16 vid;
2819
2820        e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2821
2822        for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2823            e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2824}
2825
2826static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2827{
2828        struct e1000_hw *hw = &adapter->hw;
2829        u32 manc, manc2h, mdef, i, j;
2830
2831        if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2832                return;
2833
2834        manc = er32(MANC);
2835
2836        /* enable receiving management packets to the host. this will probably
2837         * generate destination unreachable messages from the host OS, but
2838         * the packets will be handled on SMBUS
2839         */
2840        manc |= E1000_MANC_EN_MNG2HOST;
2841        manc2h = er32(MANC2H);
2842
2843        switch (hw->mac.type) {
2844        default:
2845                manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2846                break;
2847        case e1000_82574:
2848        case e1000_82583:
2849                /* Check if IPMI pass-through decision filter already exists;
2850                 * if so, enable it.
2851                 */
2852                for (i = 0, j = 0; i < 8; i++) {
2853                        mdef = er32(MDEF(i));
2854
2855                        /* Ignore filters with anything other than IPMI ports */
2856                        if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2857                                continue;
2858
2859                        /* Enable this decision filter in MANC2H */
2860                        if (mdef)
2861                                manc2h |= (1 << i);
2862
2863                        j |= mdef;
2864                }
2865
2866                if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2867                        break;
2868
2869                /* Create new decision filter in an empty filter */
2870                for (i = 0, j = 0; i < 8; i++)
2871                        if (er32(MDEF(i)) == 0) {
2872                                ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2873                                               E1000_MDEF_PORT_664));
2874                                manc2h |= (1 << 1);
2875                                j++;
2876                                break;
2877                        }
2878
2879                if (!j)
2880                        e_warn("Unable to create IPMI pass-through filter\n");
2881                break;
2882        }
2883
2884        ew32(MANC2H, manc2h);
2885        ew32(MANC, manc);
2886}
2887
2888/**
2889 * e1000_configure_tx - Configure Transmit Unit after Reset
2890 * @adapter: board private structure
2891 *
2892 * Configure the Tx unit of the MAC after a reset.
2893 **/
2894static void e1000_configure_tx(struct e1000_adapter *adapter)
2895{
2896        struct e1000_hw *hw = &adapter->hw;
2897        struct e1000_ring *tx_ring = adapter->tx_ring;
2898        u64 tdba;
2899        u32 tdlen, tarc;
2900
2901        /* Setup the HW Tx Head and Tail descriptor pointers */
2902        tdba = tx_ring->dma;
2903        tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2904        ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2905        ew32(TDBAH(0), (tdba >> 32));
2906        ew32(TDLEN(0), tdlen);
2907        ew32(TDH(0), 0);
2908        ew32(TDT(0), 0);
2909        tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2910        tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2911
2912        /* Set the Tx Interrupt Delay register */
2913        ew32(TIDV, adapter->tx_int_delay);
2914        /* Tx irq moderation */
2915        ew32(TADV, adapter->tx_abs_int_delay);
2916
2917        if (adapter->flags2 & FLAG2_DMA_BURST) {
2918                u32 txdctl = er32(TXDCTL(0));
2919                txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2920                            E1000_TXDCTL_WTHRESH);
2921                /* set up some performance related parameters to encourage the
2922                 * hardware to use the bus more efficiently in bursts, depends
2923                 * on the tx_int_delay to be enabled,
2924                 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2925                 * hthresh = 1 ==> prefetch when one or more available
2926                 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2927                 * BEWARE: this seems to work but should be considered first if
2928                 * there are Tx hangs or other Tx related bugs
2929                 */
2930                txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2931                ew32(TXDCTL(0), txdctl);
2932        }
2933        /* erratum work around: set txdctl the same for both queues */
2934        ew32(TXDCTL(1), er32(TXDCTL(0)));
2935
2936        if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2937                tarc = er32(TARC(0));
2938                /* set the speed mode bit, we'll clear it if we're not at
2939                 * gigabit link later
2940                 */
2941#define SPEED_MODE_BIT (1 << 21)
2942                tarc |= SPEED_MODE_BIT;
2943                ew32(TARC(0), tarc);
2944        }
2945
2946        /* errata: program both queues to unweighted RR */
2947        if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2948                tarc = er32(TARC(0));
2949                tarc |= 1;
2950                ew32(TARC(0), tarc);
2951                tarc = er32(TARC(1));
2952                tarc |= 1;
2953                ew32(TARC(1), tarc);
2954        }
2955
2956        /* Setup Transmit Descriptor Settings for eop descriptor */
2957        adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2958
2959        /* only set IDE if we are delaying interrupts using the timers */
2960        if (adapter->tx_int_delay)
2961                adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2962
2963        /* enable Report Status bit */
2964        adapter->txd_cmd |= E1000_TXD_CMD_RS;
2965
2966        hw->mac.ops.config_collision_dist(hw);
2967}
2968
2969/**
2970 * e1000_setup_rctl - configure the receive control registers
2971 * @adapter: Board private structure
2972 **/
2973#define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
2974                           (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
2975static void e1000_setup_rctl(struct e1000_adapter *adapter)
2976{
2977        struct e1000_hw *hw = &adapter->hw;
2978        u32 rctl, rfctl;
2979        u32 pages = 0;
2980
2981        /* Workaround Si errata on PCHx - configure jumbo frame flow */
2982        if (hw->mac.type >= e1000_pch2lan) {
2983                s32 ret_val;
2984
2985                if (adapter->netdev->mtu > ETH_DATA_LEN)
2986                        ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
2987                else
2988                        ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
2989
2990                if (ret_val)
2991                        e_dbg("failed to enable jumbo frame workaround mode\n");
2992        }
2993
2994        /* Program MC offset vector base */
2995        rctl = er32(RCTL);
2996        rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
2997        rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
2998            E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
2999            (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3000
3001        /* Do not Store bad packets */
3002        rctl &= ~E1000_RCTL_SBP;
3003
3004        /* Enable Long Packet receive */
3005        if (adapter->netdev->mtu <= ETH_DATA_LEN)
3006                rctl &= ~E1000_RCTL_LPE;
3007        else
3008                rctl |= E1000_RCTL_LPE;
3009
3010        /* Some systems expect that the CRC is included in SMBUS traffic. The
3011         * hardware strips the CRC before sending to both SMBUS (BMC) and to
3012         * host memory when this is enabled
3013         */
3014        if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3015                rctl |= E1000_RCTL_SECRC;
3016
3017        /* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3018        if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3019                u16 phy_data;
3020
3021                e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3022                phy_data &= 0xfff8;
3023                phy_data |= (1 << 2);
3024                e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3025
3026                e1e_rphy(hw, 22, &phy_data);
3027                phy_data &= 0x0fff;
3028                phy_data |= (1 << 14);
3029                e1e_wphy(hw, 0x10, 0x2823);
3030                e1e_wphy(hw, 0x11, 0x0003);
3031                e1e_wphy(hw, 22, phy_data);
3032        }
3033
3034        /* Setup buffer sizes */
3035        rctl &= ~E1000_RCTL_SZ_4096;
3036        rctl |= E1000_RCTL_BSEX;
3037        switch (adapter->rx_buffer_len) {
3038        case 2048:
3039        default:
3040                rctl |= E1000_RCTL_SZ_2048;
3041                rctl &= ~E1000_RCTL_BSEX;
3042                break;
3043        case 4096:
3044                rctl |= E1000_RCTL_SZ_4096;
3045                break;
3046        case 8192:
3047                rctl |= E1000_RCTL_SZ_8192;
3048                break;
3049        case 16384:
3050                rctl |= E1000_RCTL_SZ_16384;
3051                break;
3052        }
3053
3054        /* Enable Extended Status in all Receive Descriptors */
3055        rfctl = er32(RFCTL);
3056        rfctl |= E1000_RFCTL_EXTEN;
3057        ew32(RFCTL, rfctl);
3058
3059        /* 82571 and greater support packet-split where the protocol
3060         * header is placed in skb->data and the packet data is
3061         * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3062         * In the case of a non-split, skb->data is linearly filled,
3063         * followed by the page buffers.  Therefore, skb->data is
3064         * sized to hold the largest protocol header.
3065         *
3066         * allocations using alloc_page take too long for regular MTU
3067         * so only enable packet split for jumbo frames
3068         *
3069         * Using pages when the page size is greater than 16k wastes
3070         * a lot of memory, since we allocate 3 pages at all times
3071         * per packet.
3072         */
3073        pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3074        if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3075                adapter->rx_ps_pages = pages;
3076        else
3077                adapter->rx_ps_pages = 0;
3078
3079        if (adapter->rx_ps_pages) {
3080                u32 psrctl = 0;
3081
3082                /* Enable Packet split descriptors */
3083                rctl |= E1000_RCTL_DTYP_PS;
3084
3085                psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3086
3087                switch (adapter->rx_ps_pages) {
3088                case 3:
3089                        psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3090                        /* fall-through */
3091                case 2:
3092                        psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3093                        /* fall-through */
3094                case 1:
3095                        psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3096                        break;
3097                }
3098
3099                ew32(PSRCTL, psrctl);
3100        }
3101
3102        /* This is useful for sniffing bad packets. */
3103        if (adapter->netdev->features & NETIF_F_RXALL) {
3104                /* UPE and MPE will be handled by normal PROMISC logic
3105                 * in e1000e_set_rx_mode
3106                 */
3107                rctl |= (E1000_RCTL_SBP |       /* Receive bad packets */
3108                         E1000_RCTL_BAM |       /* RX All Bcast Pkts */
3109                         E1000_RCTL_PMCF);      /* RX All MAC Ctrl Pkts */
3110
3111                rctl &= ~(E1000_RCTL_VFE |      /* Disable VLAN filter */
3112                          E1000_RCTL_DPF |      /* Allow filtered pause */
3113                          E1000_RCTL_CFIEN);    /* Dis VLAN CFIEN Filter */
3114                /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3115                 * and that breaks VLANs.
3116                 */
3117        }
3118
3119        ew32(RCTL, rctl);
3120        /* just started the receive unit, no need to restart */
3121        adapter->flags &= ~FLAG_RESTART_NOW;
3122}
3123
3124/**
3125 * e1000_configure_rx - Configure Receive Unit after Reset
3126 * @adapter: board private structure
3127 *
3128 * Configure the Rx unit of the MAC after a reset.
3129 **/
3130static void e1000_configure_rx(struct e1000_adapter *adapter)
3131{
3132        struct e1000_hw *hw = &adapter->hw;
3133        struct e1000_ring *rx_ring = adapter->rx_ring;
3134        u64 rdba;
3135        u32 rdlen, rctl, rxcsum, ctrl_ext;
3136
3137        if (adapter->rx_ps_pages) {
3138                /* this is a 32 byte descriptor */
3139                rdlen = rx_ring->count *
3140                    sizeof(union e1000_rx_desc_packet_split);
3141                adapter->clean_rx = e1000_clean_rx_irq_ps;
3142                adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3143        } else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3144                rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3145                adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3146                adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3147        } else {
3148                rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3149                adapter->clean_rx = e1000_clean_rx_irq;
3150                adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3151        }
3152
3153        /* disable receives while setting up the descriptors */
3154        rctl = er32(RCTL);
3155        if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3156                ew32(RCTL, rctl & ~E1000_RCTL_EN);
3157        e1e_flush();
3158        usleep_range(10000, 20000);
3159
3160        if (adapter->flags2 & FLAG2_DMA_BURST) {
3161                /* set the writeback threshold (only takes effect if the RDTR
3162                 * is set). set GRAN=1 and write back up to 0x4 worth, and
3163                 * enable prefetching of 0x20 Rx descriptors
3164                 * granularity = 01
3165                 * wthresh = 04,
3166                 * hthresh = 04,
3167                 * pthresh = 0x20
3168                 */
3169                ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3170                ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3171
3172                /* override the delay timers for enabling bursting, only if
3173                 * the value was not set by the user via module options
3174                 */
3175                if (adapter->rx_int_delay == DEFAULT_RDTR)
3176                        adapter->rx_int_delay = BURST_RDTR;
3177                if (adapter->rx_abs_int_delay == DEFAULT_RADV)
3178                        adapter->rx_abs_int_delay = BURST_RADV;
3179        }
3180
3181        /* set the Receive Delay Timer Register */
3182        ew32(RDTR, adapter->rx_int_delay);
3183
3184        /* irq moderation */
3185        ew32(RADV, adapter->rx_abs_int_delay);
3186        if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3187                e1000e_write_itr(adapter, adapter->itr);
3188
3189        ctrl_ext = er32(CTRL_EXT);
3190        /* Auto-Mask interrupts upon ICR access */
3191        ctrl_ext |= E1000_CTRL_EXT_IAME;
3192        ew32(IAM, 0xffffffff);
3193        ew32(CTRL_EXT, ctrl_ext);
3194        e1e_flush();
3195
3196        /* Setup the HW Rx Head and Tail Descriptor Pointers and
3197         * the Base and Length of the Rx Descriptor Ring
3198         */
3199        rdba = rx_ring->dma;
3200        ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3201        ew32(RDBAH(0), (rdba >> 32));
3202        ew32(RDLEN(0), rdlen);
3203        ew32(RDH(0), 0);
3204        ew32(RDT(0), 0);
3205        rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3206        rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3207
3208        /* Enable Receive Checksum Offload for TCP and UDP */
3209        rxcsum = er32(RXCSUM);
3210        if (adapter->netdev->features & NETIF_F_RXCSUM)
3211                rxcsum |= E1000_RXCSUM_TUOFL;
3212        else
3213                rxcsum &= ~E1000_RXCSUM_TUOFL;
3214        ew32(RXCSUM, rxcsum);
3215
3216        /* With jumbo frames, excessive C-state transition latencies result
3217         * in dropped transactions.
3218         */
3219        if (adapter->netdev->mtu > ETH_DATA_LEN) {
3220                u32 lat =
3221                    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3222                     adapter->max_frame_size) * 8 / 1000;
3223
3224                if (adapter->flags & FLAG_IS_ICH) {
3225                        u32 rxdctl = er32(RXDCTL(0));
3226                        ew32(RXDCTL(0), rxdctl | 0x3);
3227                }
3228
3229                pm_qos_update_request(&adapter->netdev->pm_qos_req, lat);
3230        } else {
3231                pm_qos_update_request(&adapter->netdev->pm_qos_req,
3232                                      PM_QOS_DEFAULT_VALUE);
3233        }
3234
3235        /* Enable Receives */
3236        ew32(RCTL, rctl);
3237}
3238
3239/**
3240 * e1000e_write_mc_addr_list - write multicast addresses to MTA
3241 * @netdev: network interface device structure
3242 *
3243 * Writes multicast address list to the MTA hash table.
3244 * Returns: -ENOMEM on failure
3245 *                0 on no addresses written
3246 *                X on writing X addresses to MTA
3247 */
3248static int e1000e_write_mc_addr_list(struct net_device *netdev)
3249{
3250        struct e1000_adapter *adapter = netdev_priv(netdev);
3251        struct e1000_hw *hw = &adapter->hw;
3252        struct netdev_hw_addr *ha;
3253        u8 *mta_list;
3254        int i;
3255
3256        if (netdev_mc_empty(netdev)) {
3257                /* nothing to program, so clear mc list */
3258                hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3259                return 0;
3260        }
3261
3262        mta_list = kzalloc(netdev_mc_count(netdev) * ETH_ALEN, GFP_ATOMIC);
3263        if (!mta_list)
3264                return -ENOMEM;
3265
3266        /* update_mc_addr_list expects a packed array of only addresses. */
3267        i = 0;
3268        netdev_for_each_mc_addr(ha, netdev)
3269            memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3270
3271        hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3272        kfree(mta_list);
3273
3274        return netdev_mc_count(netdev);
3275}
3276
3277/**
3278 * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3279 * @netdev: network interface device structure
3280 *
3281 * Writes unicast address list to the RAR table.
3282 * Returns: -ENOMEM on failure/insufficient address space
3283 *                0 on no addresses written
3284 *                X on writing X addresses to the RAR table
3285 **/
3286static int e1000e_write_uc_addr_list(struct net_device *netdev)
3287{
3288        struct e1000_adapter *adapter = netdev_priv(netdev);
3289        struct e1000_hw *hw = &adapter->hw;
3290        unsigned int rar_entries = hw->mac.rar_entry_count;
3291        int count = 0;
3292
3293        /* save a rar entry for our hardware address */
3294        rar_entries--;
3295
3296        /* save a rar entry for the LAA workaround */
3297        if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3298                rar_entries--;
3299
3300        /* return ENOMEM indicating insufficient memory for addresses */
3301        if (netdev_uc_count(netdev) > rar_entries)
3302                return -ENOMEM;
3303
3304        if (!netdev_uc_empty(netdev) && rar_entries) {
3305                struct netdev_hw_addr *ha;
3306
3307                /* write the addresses in reverse order to avoid write
3308                 * combining
3309                 */
3310                netdev_for_each_uc_addr(ha, netdev) {
3311                        if (!rar_entries)
3312                                break;
3313                        hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3314                        count++;
3315                }
3316        }
3317
3318        /* zero out the remaining RAR entries not used above */
3319        for (; rar_entries > 0; rar_entries--) {
3320                ew32(RAH(rar_entries), 0);
3321                ew32(RAL(rar_entries), 0);
3322        }
3323        e1e_flush();
3324
3325        return count;
3326}
3327
3328/**
3329 * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3330 * @netdev: network interface device structure
3331 *
3332 * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3333 * address list or the network interface flags are updated.  This routine is
3334 * responsible for configuring the hardware for proper unicast, multicast,
3335 * promiscuous mode, and all-multi behavior.
3336 **/
3337static void e1000e_set_rx_mode(struct net_device *netdev)
3338{
3339        struct e1000_adapter *adapter = netdev_priv(netdev);
3340        struct e1000_hw *hw = &adapter->hw;
3341        u32 rctl;
3342
3343        /* Check for Promiscuous and All Multicast modes */
3344        rctl = er32(RCTL);
3345
3346        /* clear the affected bits */
3347        rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3348
3349        if (netdev->flags & IFF_PROMISC) {
3350                rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3351                /* Do not hardware filter VLANs in promisc mode */
3352                e1000e_vlan_filter_disable(adapter);
3353        } else {
3354                int count;
3355
3356                if (netdev->flags & IFF_ALLMULTI) {
3357                        rctl |= E1000_RCTL_MPE;
3358                } else {
3359                        /* Write addresses to the MTA, if the attempt fails
3360                         * then we should just turn on promiscuous mode so
3361                         * that we can at least receive multicast traffic
3362                         */
3363                        count = e1000e_write_mc_addr_list(netdev);
3364                        if (count < 0)
3365                                rctl |= E1000_RCTL_MPE;
3366                }
3367                e1000e_vlan_filter_enable(adapter);
3368                /* Write addresses to available RAR registers, if there is not
3369                 * sufficient space to store all the addresses then enable
3370                 * unicast promiscuous mode
3371                 */
3372                count = e1000e_write_uc_addr_list(netdev);
3373                if (count < 0)
3374                        rctl |= E1000_RCTL_UPE;
3375        }
3376
3377        ew32(RCTL, rctl);
3378
3379        if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3380                e1000e_vlan_strip_enable(adapter);
3381        else
3382                e1000e_vlan_strip_disable(adapter);
3383}
3384
3385static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3386{
3387        struct e1000_hw *hw = &adapter->hw;
3388        u32 mrqc, rxcsum;
3389        int i;
3390        static const u32 rsskey[10] = {
3391                0xda565a6d, 0xc20e5b25, 0x3d256741, 0xb08fa343, 0xcb2bcad0,
3392                0xb4307bae, 0xa32dcb77, 0x0cf23080, 0x3bb7426a, 0xfa01acbe
3393        };
3394
3395        /* Fill out hash function seed */
3396        for (i = 0; i < 10; i++)
3397                ew32(RSSRK(i), rsskey[i]);
3398
3399        /* Direct all traffic to queue 0 */
3400        for (i = 0; i < 32; i++)
3401                ew32(RETA(i), 0);
3402
3403        /* Disable raw packet checksumming so that RSS hash is placed in
3404         * descriptor on writeback.
3405         */
3406        rxcsum = er32(RXCSUM);
3407        rxcsum |= E1000_RXCSUM_PCSD;
3408
3409        ew32(RXCSUM, rxcsum);
3410
3411        mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3412                E1000_MRQC_RSS_FIELD_IPV4_TCP |
3413                E1000_MRQC_RSS_FIELD_IPV6 |
3414                E1000_MRQC_RSS_FIELD_IPV6_TCP |
3415                E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3416
3417        ew32(MRQC, mrqc);
3418}
3419
3420/**
3421 * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3422 * @adapter: board private structure
3423 * @timinca: pointer to returned time increment attributes
3424 *
3425 * Get attributes for incrementing the System Time Register SYSTIML/H at
3426 * the default base frequency, and set the cyclecounter shift value.
3427 **/
3428s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3429{
3430        struct e1000_hw *hw = &adapter->hw;
3431        u32 incvalue, incperiod, shift;
3432
3433        /* Make sure clock is enabled on I217 before checking the frequency */
3434        if ((hw->mac.type == e1000_pch_lpt) &&
3435            !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3436            !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3437                u32 fextnvm7 = er32(FEXTNVM7);
3438
3439                if (!(fextnvm7 & (1 << 0))) {
3440                        ew32(FEXTNVM7, fextnvm7 | (1 << 0));
3441                        e1e_flush();
3442                }
3443        }
3444
3445        switch (hw->mac.type) {
3446        case e1000_pch2lan:
3447        case e1000_pch_lpt:
3448                /* On I217, the clock frequency is 25MHz or 96MHz as
3449                 * indicated by the System Clock Frequency Indication
3450                 */
3451                if ((hw->mac.type != e1000_pch_lpt) ||
3452                    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI)) {
3453                        /* Stable 96MHz frequency */
3454                        incperiod = INCPERIOD_96MHz;
3455                        incvalue = INCVALUE_96MHz;
3456                        shift = INCVALUE_SHIFT_96MHz;
3457                        adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHz;
3458                        break;
3459                }
3460                /* fall-through */
3461        case e1000_82574:
3462        case e1000_82583:
3463                /* Stable 25MHz frequency */
3464                incperiod = INCPERIOD_25MHz;
3465                incvalue = INCVALUE_25MHz;
3466                shift = INCVALUE_SHIFT_25MHz;
3467                adapter->cc.shift = shift;
3468                break;
3469        default:
3470                return -EINVAL;
3471        }
3472
3473        *timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3474                    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3475
3476        return 0;
3477}
3478
3479/**
3480 * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3481 * @adapter: board private structure
3482 *
3483 * Outgoing time stamping can be enabled and disabled. Play nice and
3484 * disable it when requested, although it shouldn't cause any overhead
3485 * when no packet needs it. At most one packet in the queue may be
3486 * marked for time stamping, otherwise it would be impossible to tell
3487 * for sure to which packet the hardware time stamp belongs.
3488 *
3489 * Incoming time stamping has to be configured via the hardware filters.
3490 * Not all combinations are supported, in particular event type has to be
3491 * specified. Matching the kind of event packet is not supported, with the
3492 * exception of "all V2 events regardless of level 2 or 4".
3493 **/
3494static int e1000e_config_hwtstamp(struct e1000_adapter *adapter)
3495{
3496        struct e1000_hw *hw = &adapter->hw;
3497        struct hwtstamp_config *config = &adapter->hwtstamp_config;
3498        u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3499        u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3500        u32 rxmtrl = 0;
3501        u16 rxudp = 0;
3502        bool is_l4 = false;
3503        bool is_l2 = false;
3504        u32 regval;
3505        s32 ret_val;
3506
3507        if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3508                return -EINVAL;
3509
3510        /* flags reserved for future extensions - must be zero */
3511        if (config->flags)
3512                return -EINVAL;
3513
3514        switch (config->tx_type) {
3515        case HWTSTAMP_TX_OFF:
3516                tsync_tx_ctl = 0;
3517                break;
3518        case HWTSTAMP_TX_ON:
3519                break;
3520        default:
3521                return -ERANGE;
3522        }
3523
3524        switch (config->rx_filter) {
3525        case HWTSTAMP_FILTER_NONE:
3526                tsync_rx_ctl = 0;
3527                break;
3528        case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3529                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3530                rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3531                is_l4 = true;
3532                break;
3533        case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3534                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3535                rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3536                is_l4 = true;
3537                break;
3538        case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3539                /* Also time stamps V2 L2 Path Delay Request/Response */
3540                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3541                rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3542                is_l2 = true;
3543                break;
3544        case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3545                /* Also time stamps V2 L2 Path Delay Request/Response. */
3546                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3547                rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3548                is_l2 = true;
3549                break;
3550        case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3551                /* Hardware cannot filter just V2 L4 Sync messages;
3552                 * fall-through to V2 (both L2 and L4) Sync.
3553                 */
3554        case HWTSTAMP_FILTER_PTP_V2_SYNC:
3555                /* Also time stamps V2 Path Delay Request/Response. */
3556                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3557                rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3558                is_l2 = true;
3559                is_l4 = true;
3560                break;
3561        case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3562                /* Hardware cannot filter just V2 L4 Delay Request messages;
3563                 * fall-through to V2 (both L2 and L4) Delay Request.
3564                 */
3565        case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3566                /* Also time stamps V2 Path Delay Request/Response. */
3567                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3568                rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3569                is_l2 = true;
3570                is_l4 = true;
3571                break;
3572        case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3573        case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3574                /* Hardware cannot filter just V2 L4 or L2 Event messages;
3575                 * fall-through to all V2 (both L2 and L4) Events.
3576                 */
3577        case HWTSTAMP_FILTER_PTP_V2_EVENT:
3578                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3579                config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3580                is_l2 = true;
3581                is_l4 = true;
3582                break;
3583        case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3584                /* For V1, the hardware can only filter Sync messages or
3585                 * Delay Request messages but not both so fall-through to
3586                 * time stamp all packets.
3587                 */
3588        case HWTSTAMP_FILTER_ALL:
3589                is_l2 = true;
3590                is_l4 = true;
3591                tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3592                config->rx_filter = HWTSTAMP_FILTER_ALL;
3593                break;
3594        default:
3595                return -ERANGE;
3596        }
3597
3598        /* enable/disable Tx h/w time stamping */
3599        regval = er32(TSYNCTXCTL);
3600        regval &= ~E1000_TSYNCTXCTL_ENABLED;
3601        regval |= tsync_tx_ctl;
3602        ew32(TSYNCTXCTL, regval);
3603        if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3604            (regval & E1000_TSYNCTXCTL_ENABLED)) {
3605                e_err("Timesync Tx Control register not set as expected\n");
3606                return -EAGAIN;
3607        }
3608
3609        /* enable/disable Rx h/w time stamping */
3610        regval = er32(TSYNCRXCTL);
3611        regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3612        regval |= tsync_rx_ctl;
3613        ew32(TSYNCRXCTL, regval);
3614        if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3615                                 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3616            (regval & (E1000_TSYNCRXCTL_ENABLED |
3617                       E1000_TSYNCRXCTL_TYPE_MASK))) {
3618                e_err("Timesync Rx Control register not set as expected\n");
3619                return -EAGAIN;
3620        }
3621
3622        /* L2: define ethertype filter for time stamped packets */
3623        if (is_l2)
3624                rxmtrl |= ETH_P_1588;
3625
3626        /* define which PTP packets get time stamped */
3627        ew32(RXMTRL, rxmtrl);
3628
3629        /* Filter by destination port */
3630        if (is_l4) {
3631                rxudp = PTP_EV_PORT;
3632                cpu_to_be16s(&rxudp);
3633        }
3634        ew32(RXUDP, rxudp);
3635
3636        e1e_flush();
3637
3638        /* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3639        er32(RXSTMPH);
3640        er32(TXSTMPH);
3641
3642        /* Get and set the System Time Register SYSTIM base frequency */
3643        ret_val = e1000e_get_base_timinca(adapter, &regval);
3644        if (ret_val)
3645                return ret_val;
3646        ew32(TIMINCA, regval);
3647
3648        /* reset the ns time counter */
3649        timecounter_init(&adapter->tc, &adapter->cc,
3650                         ktime_to_ns(ktime_get_real()));
3651
3652        return 0;
3653}
3654
3655/**
3656 * e1000_configure - configure the hardware for Rx and Tx
3657 * @adapter: private board structure
3658 **/
3659static void e1000_configure(struct e1000_adapter *adapter)
3660{
3661        struct e1000_ring *rx_ring = adapter->rx_ring;
3662
3663        e1000e_set_rx_mode(adapter->netdev);
3664
3665        e1000_restore_vlan(adapter);
3666        e1000_init_manageability_pt(adapter);
3667
3668        e1000_configure_tx(adapter);
3669
3670        if (adapter->netdev->features & NETIF_F_RXHASH)
3671                e1000e_setup_rss_hash(adapter);
3672        e1000_setup_rctl(adapter);
3673        e1000_configure_rx(adapter);
3674        adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3675}
3676
3677/**
3678 * e1000e_power_up_phy - restore link in case the phy was powered down
3679 * @adapter: address of board private structure
3680 *
3681 * The phy may be powered down to save power and turn off link when the
3682 * driver is unloaded and wake on lan is not enabled (among others)
3683 * *** this routine MUST be followed by a call to e1000e_reset ***
3684 **/
3685void e1000e_power_up_phy(struct e1000_adapter *adapter)
3686{
3687        if (adapter->hw.phy.ops.power_up)
3688                adapter->hw.phy.ops.power_up(&adapter->hw);
3689
3690        adapter->hw.mac.ops.setup_link(&adapter->hw);
3691}
3692
3693/**
3694 * e1000_power_down_phy - Power down the PHY
3695 *
3696 * Power down the PHY so no link is implied when interface is down.
3697 * The PHY cannot be powered down if management or WoL is active.
3698 */
3699static void e1000_power_down_phy(struct e1000_adapter *adapter)
3700{
3701        /* WoL is enabled */
3702        if (adapter->wol)
3703                return;
3704
3705        if (adapter->hw.phy.ops.power_down)
3706                adapter->hw.phy.ops.power_down(&adapter->hw);
3707}
3708
3709/**
3710 * e1000e_reset - bring the hardware into a known good state
3711 *
3712 * This function boots the hardware and enables some settings that
3713 * require a configuration cycle of the hardware - those cannot be
3714 * set/changed during runtime. After reset the device needs to be
3715 * properly configured for Rx, Tx etc.
3716 */
3717void e1000e_reset(struct e1000_adapter *adapter)
3718{
3719        struct e1000_mac_info *mac = &adapter->hw.mac;
3720        struct e1000_fc_info *fc = &adapter->hw.fc;
3721        struct e1000_hw *hw = &adapter->hw;
3722        u32 tx_space, min_tx_space, min_rx_space;
3723        u32 pba = adapter->pba;
3724        u16 hwm;
3725
3726        /* reset Packet Buffer Allocation to default */
3727        ew32(PBA, pba);
3728
3729        if (adapter->max_frame_size > ETH_FRAME_LEN + ETH_FCS_LEN) {
3730                /* To maintain wire speed transmits, the Tx FIFO should be
3731                 * large enough to accommodate two full transmit packets,
3732                 * rounded up to the next 1KB and expressed in KB.  Likewise,
3733                 * the Rx FIFO should be large enough to accommodate at least
3734                 * one full receive packet and is similarly rounded up and
3735                 * expressed in KB.
3736                 */
3737                pba = er32(PBA);
3738                /* upper 16 bits has Tx packet buffer allocation size in KB */
3739                tx_space = pba >> 16;
3740                /* lower 16 bits has Rx packet buffer allocation size in KB */
3741                pba &= 0xffff;
3742                /* the Tx fifo also stores 16 bytes of information about the Tx
3743                 * but don't include ethernet FCS because hardware appends it
3744                 */
3745                min_tx_space = (adapter->max_frame_size +
3746                                sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3747                min_tx_space = ALIGN(min_tx_space, 1024);
3748                min_tx_space >>= 10;
3749                /* software strips receive CRC, so leave room for it */
3750                min_rx_space = adapter->max_frame_size;
3751                min_rx_space = ALIGN(min_rx_space, 1024);
3752                min_rx_space >>= 10;
3753
3754                /* If current Tx allocation is less than the min Tx FIFO size,
3755                 * and the min Tx FIFO size is less than the current Rx FIFO
3756                 * allocation, take space away from current Rx allocation
3757                 */
3758                if ((tx_space < min_tx_space) &&
3759                    ((min_tx_space - tx_space) < pba)) {
3760                        pba -= min_tx_space - tx_space;
3761
3762                        /* if short on Rx space, Rx wins and must trump Tx
3763                         * adjustment
3764                         */
3765                        if (pba < min_rx_space)
3766                                pba = min_rx_space;
3767                }
3768
3769                ew32(PBA, pba);
3770        }
3771
3772        /* flow control settings
3773         *
3774         * The high water mark must be low enough to fit one full frame
3775         * (or the size used for early receive) above it in the Rx FIFO.
3776         * Set it to the lower of:
3777         * - 90% of the Rx FIFO size, and
3778         * - the full Rx FIFO size minus one full frame
3779         */
3780        if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
3781                fc->pause_time = 0xFFFF;
3782        else
3783                fc->pause_time = E1000_FC_PAUSE_TIME;
3784        fc->send_xon = true;
3785        fc->current_mode = fc->requested_mode;
3786
3787        switch (hw->mac.type) {
3788        case e1000_ich9lan:
3789        case e1000_ich10lan:
3790                if (adapter->netdev->mtu > ETH_DATA_LEN) {
3791                        pba = 14;
3792                        ew32(PBA, pba);
3793                        fc->high_water = 0x2800;
3794                        fc->low_water = fc->high_water - 8;
3795                        break;
3796                }
3797                /* fall-through */
3798        default:
3799                hwm = min(((pba << 10) * 9 / 10),
3800                          ((pba << 10) - adapter->max_frame_size));
3801
3802                fc->high_water = hwm & E1000_FCRTH_RTH; /* 8-byte granularity */
3803                fc->low_water = fc->high_water - 8;
3804                break;
3805        case e1000_pchlan:
3806                /* Workaround PCH LOM adapter hangs with certain network
3807                 * loads.  If hangs persist, try disabling Tx flow control.
3808                 */
3809                if (adapter->netdev->mtu > ETH_DATA_LEN) {
3810                        fc->high_water = 0x3500;
3811                        fc->low_water = 0x1500;
3812                } else {
3813                        fc->high_water = 0x5000;
3814                        fc->low_water = 0x3000;
3815                }
3816                fc->refresh_time = 0x1000;
3817                break;
3818        case e1000_pch2lan:
3819        case e1000_pch_lpt:
3820                fc->refresh_time = 0x0400;
3821
3822                if (adapter->netdev->mtu <= ETH_DATA_LEN) {
3823                        fc->high_water = 0x05C20;
3824                        fc->low_water = 0x05048;
3825                        fc->pause_time = 0x0650;
3826                        break;
3827                }
3828
3829                fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
3830                fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
3831                break;
3832        }
3833
3834        /* Alignment of Tx data is on an arbitrary byte boundary with the
3835         * maximum size per Tx descriptor limited only to the transmit
3836         * allocation of the packet buffer minus 96 bytes with an upper
3837         * limit of 24KB due to receive synchronization limitations.
3838         */
3839        adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
3840                                       24 << 10);
3841
3842        /* Disable Adaptive Interrupt Moderation if 2 full packets cannot
3843         * fit in receive buffer.
3844         */
3845        if (adapter->itr_setting & 0x3) {
3846                if ((adapter->max_frame_size * 2) > (pba << 10)) {
3847                        if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
3848                                dev_info(&adapter->pdev->dev,
3849                                         "Interrupt Throttle Rate off\n");
3850                                adapter->flags2 |= FLAG2_DISABLE_AIM;
3851                                e1000e_write_itr(adapter, 0);
3852                        }
3853                } else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
3854                        dev_info(&adapter->pdev->dev,
3855                                 "Interrupt Throttle Rate on\n");
3856                        adapter->flags2 &= ~FLAG2_DISABLE_AIM;
3857                        adapter->itr = 20000;
3858                        e1000e_write_itr(adapter, adapter->itr);
3859                }
3860        }
3861
3862        /* Allow time for pending master requests to run */
3863        mac->ops.reset_hw(hw);
3864
3865        /* For parts with AMT enabled, let the firmware know
3866         * that the network interface is in control
3867         */
3868        if (adapter->flags & FLAG_HAS_AMT)
3869                e1000e_get_hw_control(adapter);
3870
3871        ew32(WUC, 0);
3872
3873        if (mac->ops.init_hw(hw))
3874                e_err("Hardware Error\n");
3875
3876        e1000_update_mng_vlan(adapter);
3877
3878        /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
3879        ew32(VET, ETH_P_8021Q);
3880
3881        e1000e_reset_adaptive(hw);
3882
3883        /* initialize systim and reset the ns time counter */
3884        e1000e_config_hwtstamp(adapter);
3885
3886        /* Set EEE advertisement as appropriate */
3887        if (adapter->flags2 & FLAG2_HAS_EEE) {
3888                s32 ret_val;
3889                u16 adv_addr;
3890
3891                switch (hw->phy.type) {
3892                case e1000_phy_82579:
3893                        adv_addr = I82579_EEE_ADVERTISEMENT;
3894                        break;
3895                case e1000_phy_i217:
3896                        adv_addr = I217_EEE_ADVERTISEMENT;
3897                        break;
3898                default:
3899                        dev_err(&adapter->pdev->dev,
3900                                "Invalid PHY type setting EEE advertisement\n");
3901                        return;
3902                }
3903
3904                ret_val = hw->phy.ops.acquire(hw);
3905                if (ret_val) {
3906                        dev_err(&adapter->pdev->dev,
3907                                "EEE advertisement - unable to acquire PHY\n");
3908                        return;
3909                }
3910
3911                e1000_write_emi_reg_locked(hw, adv_addr,
3912                                           hw->dev_spec.ich8lan.eee_disable ?
3913                                           0 : adapter->eee_advert);
3914
3915                hw->phy.ops.release(hw);
3916        }
3917
3918        if (!netif_running(adapter->netdev) &&
3919            !test_bit(__E1000_TESTING, &adapter->state)) {
3920                e1000_power_down_phy(adapter);
3921                return;
3922        }
3923
3924        e1000_get_phy_info(hw);
3925
3926        if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
3927            !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
3928                u16 phy_data = 0;
3929                /* speed up time to link by disabling smart power down, ignore
3930                 * the return value of this function because there is nothing
3931                 * different we would do if it failed
3932                 */
3933                e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
3934                phy_data &= ~IGP02E1000_PM_SPD;
3935                e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
3936        }
3937}
3938
3939int e1000e_up(struct e1000_adapter *adapter)
3940{
3941        struct e1000_hw *hw = &adapter->hw;
3942
3943        /* hardware has been reset, we need to reload some things */
3944        e1000_configure(adapter);
3945
3946        clear_bit(__E1000_DOWN, &adapter->state);
3947
3948        if (adapter->msix_entries)
3949                e1000_configure_msix(adapter);
3950        e1000_irq_enable(adapter);
3951
3952        netif_start_queue(adapter->netdev);
3953
3954        /* fire a link change interrupt to start the watchdog */
3955        if (adapter->msix_entries)
3956                ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
3957        else
3958                ew32(ICS, E1000_ICS_LSC);
3959
3960        return 0;
3961}
3962
3963static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
3964{
3965        struct e1000_hw *hw = &adapter->hw;
3966
3967        if (!(adapter->flags2 & FLAG2_DMA_BURST))
3968                return;
3969
3970        /* flush pending descriptor writebacks to memory */
3971        ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3972        ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3973
3974        /* execute the writes immediately */
3975        e1e_flush();
3976
3977        /* due to rare timing issues, write to TIDV/RDTR again to ensure the
3978         * write is successful
3979         */
3980        ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
3981        ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
3982
3983        /* execute the writes immediately */
3984        e1e_flush();
3985}
3986
3987static void e1000e_update_stats(struct e1000_adapter *adapter);
3988
3989void e1000e_down(struct e1000_adapter *adapter)
3990{
3991        struct net_device *netdev = adapter->netdev;
3992        struct e1000_hw *hw = &adapter->hw;
3993        u32 tctl, rctl;
3994
3995        /* signal that we're down so the interrupt handler does not
3996         * reschedule our watchdog timer
3997         */
3998        set_bit(__E1000_DOWN, &adapter->state);
3999
4000        /* disable receives in the hardware */
4001        rctl = er32(RCTL);
4002        if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4003                ew32(RCTL, rctl & ~E1000_RCTL_EN);
4004        /* flush and sleep below */
4005
4006        netif_stop_queue(netdev);
4007
4008        /* disable transmits in the hardware */
4009        tctl = er32(TCTL);
4010        tctl &= ~E1000_TCTL_EN;
4011        ew32(TCTL, tctl);
4012
4013        /* flush both disables and wait for them to finish */
4014        e1e_flush();
4015        usleep_range(10000, 20000);
4016
4017        e1000_irq_disable(adapter);
4018
4019        napi_synchronize(&adapter->napi);
4020
4021        del_timer_sync(&adapter->watchdog_timer);
4022        del_timer_sync(&adapter->phy_info_timer);
4023
4024        netif_carrier_off(netdev);
4025
4026        spin_lock(&adapter->stats64_lock);
4027        e1000e_update_stats(adapter);
4028        spin_unlock(&adapter->stats64_lock);
4029
4030        e1000e_flush_descriptors(adapter);
4031        e1000_clean_tx_ring(adapter->tx_ring);
4032        e1000_clean_rx_ring(adapter->rx_ring);
4033
4034        adapter->link_speed = 0;
4035        adapter->link_duplex = 0;
4036
4037        if (!pci_channel_offline(adapter->pdev))
4038                e1000e_reset(adapter);
4039
4040        /* TODO: for power management, we could drop the link and
4041         * pci_disable_device here.
4042         */
4043}
4044
4045void e1000e_reinit_locked(struct e1000_adapter *adapter)
4046{
4047        might_sleep();
4048        while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4049                usleep_range(1000, 2000);
4050        e1000e_down(adapter);
4051        e1000e_up(adapter);
4052        clear_bit(__E1000_RESETTING, &adapter->state);
4053}
4054
4055/**
4056 * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4057 * @cc: cyclecounter structure
4058 **/
4059static cycle_t e1000e_cyclecounter_read(const struct cyclecounter *cc)
4060{
4061        struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4062                                                     cc);
4063        struct e1000_hw *hw = &adapter->hw;
4064        cycle_t systim;
4065
4066        /* latch SYSTIMH on read of SYSTIML */
4067        systim = (cycle_t)er32(SYSTIML);
4068        systim |= (cycle_t)er32(SYSTIMH) << 32;
4069
4070        return systim;
4071}
4072
4073/**
4074 * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4075 * @adapter: board private structure to initialize
4076 *
4077 * e1000_sw_init initializes the Adapter private data structure.
4078 * Fields are initialized based on PCI device information and
4079 * OS network device settings (MTU size).
4080 **/
4081static int e1000_sw_init(struct e1000_adapter *adapter)
4082{
4083        struct net_device *netdev = adapter->netdev;
4084
4085        adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN;
4086        adapter->rx_ps_bsize0 = 128;
4087        adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN;
4088        adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4089        adapter->tx_ring_count = E1000_DEFAULT_TXD;
4090        adapter->rx_ring_count = E1000_DEFAULT_RXD;
4091
4092        spin_lock_init(&adapter->stats64_lock);
4093
4094        e1000e_set_interrupt_capability(adapter);
4095
4096        if (e1000_alloc_queues(adapter))
4097                return -ENOMEM;
4098
4099        /* Setup hardware time stamping cyclecounter */
4100        if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4101                adapter->cc.read = e1000e_cyclecounter_read;
4102                adapter->cc.mask = CLOCKSOURCE_MASK(64);
4103                adapter->cc.mult = 1;
4104                /* cc.shift set in e1000e_get_base_tininca() */
4105
4106                spin_lock_init(&adapter->systim_lock);
4107                INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4108        }
4109
4110        /* Explicitly disable IRQ since the NIC can be in any state. */
4111        e1000_irq_disable(adapter);
4112
4113        set_bit(__E1000_DOWN, &adapter->state);
4114        return 0;
4115}
4116
4117/**
4118 * e1000_intr_msi_test - Interrupt Handler
4119 * @irq: interrupt number
4120 * @data: pointer to a network interface device structure
4121 **/
4122static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4123{
4124        struct net_device *netdev = data;
4125        struct e1000_adapter *adapter = netdev_priv(netdev);
4126        struct e1000_hw *hw = &adapter->hw;
4127        u32 icr = er32(ICR);
4128
4129        e_dbg("icr is %08X\n", icr);
4130        if (icr & E1000_ICR_RXSEQ) {
4131                adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4132                /* Force memory writes to complete before acknowledging the
4133                 * interrupt is handled.
4134                 */
4135                wmb();
4136        }
4137
4138        return IRQ_HANDLED;
4139}
4140
4141/**
4142 * e1000_test_msi_interrupt - Returns 0 for successful test
4143 * @adapter: board private struct
4144 *
4145 * code flow taken from tg3.c
4146 **/
4147static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4148{
4149        struct net_device *netdev = adapter->netdev;
4150        struct e1000_hw *hw = &adapter->hw;
4151        int err;
4152
4153        /* poll_enable hasn't been called yet, so don't need disable */
4154        /* clear any pending events */
4155        er32(ICR);
4156
4157        /* free the real vector and request a test handler */
4158        e1000_free_irq(adapter);
4159        e1000e_reset_interrupt_capability(adapter);
4160
4161        /* Assume that the test fails, if it succeeds then the test
4162         * MSI irq handler will unset this flag
4163         */
4164        adapter->flags |= FLAG_MSI_TEST_FAILED;
4165
4166        err = pci_enable_msi(adapter->pdev);
4167        if (err)
4168                goto msi_test_failed;
4169
4170        err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4171                          netdev->name, netdev);
4172        if (err) {
4173                pci_disable_msi(adapter->pdev);
4174                goto msi_test_failed;
4175        }
4176
4177        /* Force memory writes to complete before enabling and firing an
4178         * interrupt.
4179         */
4180        wmb();
4181
4182        e1000_irq_enable(adapter);
4183
4184        /* fire an unusual interrupt on the test handler */
4185        ew32(ICS, E1000_ICS_RXSEQ);
4186        e1e_flush();
4187        msleep(100);
4188
4189        e1000_irq_disable(adapter);
4190
4191        rmb();                  /* read flags after interrupt has been fired */
4192
4193        if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4194                adapter->int_mode = E1000E_INT_MODE_LEGACY;
4195                e_info("MSI interrupt test failed, using legacy interrupt.\n");
4196        } else {
4197                e_dbg("MSI interrupt test succeeded!\n");
4198        }
4199
4200        free_irq(adapter->pdev->irq, netdev);
4201        pci_disable_msi(adapter->pdev);
4202
4203msi_test_failed:
4204        e1000e_set_interrupt_capability(adapter);
4205        return e1000_request_irq(adapter);
4206}
4207
4208/**
4209 * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4210 * @adapter: board private struct
4211 *
4212 * code flow taken from tg3.c, called with e1000 interrupts disabled.
4213 **/
4214static int e1000_test_msi(struct e1000_adapter *adapter)
4215{
4216        int err;
4217        u16 pci_cmd;
4218
4219        if (!(adapter->flags & FLAG_MSI_ENABLED))
4220                return 0;
4221
4222        /* disable SERR in case the MSI write causes a master abort */
4223        pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4224        if (pci_cmd & PCI_COMMAND_SERR)
4225                pci_write_config_word(adapter->pdev, PCI_COMMAND,
4226                                      pci_cmd & ~PCI_COMMAND_SERR);
4227
4228        err = e1000_test_msi_interrupt(adapter);
4229
4230        /* re-enable SERR */
4231        if (pci_cmd & PCI_COMMAND_SERR) {
4232                pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4233                pci_cmd |= PCI_COMMAND_SERR;
4234                pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4235        }
4236
4237        return err;
4238}
4239
4240/**
4241 * e1000_open - Called when a network interface is made active
4242 * @netdev: network interface device structure
4243 *
4244 * Returns 0 on success, negative value on failure
4245 *
4246 * The open entry point is called when a network interface is made
4247 * active by the system (IFF_UP).  At this point all resources needed
4248 * for transmit and receive operations are allocated, the interrupt
4249 * handler is registered with the OS, the watchdog timer is started,
4250 * and the stack is notified that the interface is ready.
4251 **/
4252static int e1000_open(struct net_device *netdev)
4253{
4254        struct e1000_adapter *adapter = netdev_priv(netdev);
4255        struct e1000_hw *hw = &adapter->hw;
4256        struct pci_dev *pdev = adapter->pdev;
4257        int err;
4258
4259        /* disallow open during test */
4260        if (test_bit(__E1000_TESTING, &adapter->state))
4261                return -EBUSY;
4262
4263        pm_runtime_get_sync(&pdev->dev);
4264
4265        netif_carrier_off(netdev);
4266
4267        /* allocate transmit descriptors */
4268        err = e1000e_setup_tx_resources(adapter->tx_ring);
4269        if (err)
4270                goto err_setup_tx;
4271
4272        /* allocate receive descriptors */
4273        err = e1000e_setup_rx_resources(adapter->rx_ring);
4274        if (err)
4275                goto err_setup_rx;
4276
4277        /* If AMT is enabled, let the firmware know that the network
4278         * interface is now open and reset the part to a known state.
4279         */
4280        if (adapter->flags & FLAG_HAS_AMT) {
4281                e1000e_get_hw_control(adapter);
4282                e1000e_reset(adapter);
4283        }
4284
4285        e1000e_power_up_phy(adapter);
4286
4287        adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4288        if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4289                e1000_update_mng_vlan(adapter);
4290
4291        /* DMA latency requirement to workaround jumbo issue */
4292        pm_qos_add_request(&adapter->netdev->pm_qos_req, PM_QOS_CPU_DMA_LATENCY,
4293                           PM_QOS_DEFAULT_VALUE);
4294
4295        /* before we allocate an interrupt, we must be ready to handle it.
4296         * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4297         * as soon as we call pci_request_irq, so we have to setup our
4298         * clean_rx handler before we do so.
4299         */
4300        e1000_configure(adapter);
4301
4302        err = e1000_request_irq(adapter);
4303        if (err)
4304                goto err_req_irq;
4305
4306        /* Work around PCIe errata with MSI interrupts causing some chipsets to
4307         * ignore e1000e MSI messages, which means we need to test our MSI
4308         * interrupt now
4309         */
4310        if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4311                err = e1000_test_msi(adapter);
4312                if (err) {
4313                        e_err("Interrupt allocation failed\n");
4314                        goto err_req_irq;
4315                }
4316        }
4317
4318        /* From here on the code is the same as e1000e_up() */
4319        clear_bit(__E1000_DOWN, &adapter->state);
4320
4321        napi_enable(&adapter->napi);
4322
4323        e1000_irq_enable(adapter);
4324
4325        adapter->tx_hang_recheck = false;
4326        netif_start_queue(netdev);
4327
4328        adapter->idle_check = true;
4329        hw->mac.get_link_status = true;
4330        pm_runtime_put(&pdev->dev);
4331
4332        /* fire a link status change interrupt to start the watchdog */
4333        if (adapter->msix_entries)
4334                ew32(ICS, E1000_ICS_LSC | E1000_ICR_OTHER);
4335        else
4336                ew32(ICS, E1000_ICS_LSC);
4337
4338        return 0;
4339
4340err_req_irq:
4341        e1000e_release_hw_control(adapter);
4342        e1000_power_down_phy(adapter);
4343        e1000e_free_rx_resources(adapter->rx_ring);
4344err_setup_rx:
4345        e1000e_free_tx_resources(adapter->tx_ring);
4346err_setup_tx:
4347        e1000e_reset(adapter);
4348        pm_runtime_put_sync(&pdev->dev);
4349
4350        return err;
4351}
4352
4353/**
4354 * e1000_close - Disables a network interface
4355 * @netdev: network interface device structure
4356 *
4357 * Returns 0, this is not allowed to fail
4358 *
4359 * The close entry point is called when an interface is de-activated
4360 * by the OS.  The hardware is still under the drivers control, but
4361 * needs to be disabled.  A global MAC reset is issued to stop the
4362 * hardware, and all transmit and receive resources are freed.
4363 **/
4364static int e1000_close(struct net_device *netdev)
4365{
4366        struct e1000_adapter *adapter = netdev_priv(netdev);
4367        struct pci_dev *pdev = adapter->pdev;
4368        int count = E1000_CHECK_RESET_COUNT;
4369
4370        while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4371                usleep_range(10000, 20000);
4372
4373        WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4374
4375        pm_runtime_get_sync(&pdev->dev);
4376
4377        if (!test_bit(__E1000_DOWN, &adapter->state)) {
4378                e1000e_down(adapter);
4379                e1000_free_irq(adapter);
4380        }
4381
4382        napi_disable(&adapter->napi);
4383
4384        e1000_power_down_phy(adapter);
4385
4386        e1000e_free_tx_resources(adapter->tx_ring);
4387        e1000e_free_rx_resources(adapter->rx_ring);
4388
4389        /* kill manageability vlan ID if supported, but not if a vlan with
4390         * the same ID is registered on the host OS (let 8021q kill it)
4391         */
4392        if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4393                e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4394                                       adapter->mng_vlan_id);
4395
4396        /* If AMT is enabled, let the firmware know that the network
4397         * interface is now closed
4398         */
4399        if ((adapter->flags & FLAG_HAS_AMT) &&
4400            !test_bit(__E1000_TESTING, &adapter->state))
4401                e1000e_release_hw_control(adapter);
4402
4403        pm_qos_remove_request(&adapter->netdev->pm_qos_req);
4404
4405        pm_runtime_put_sync(&pdev->dev);
4406
4407        return 0;
4408}
4409
4410/**
4411 * e1000_set_mac - Change the Ethernet Address of the NIC
4412 * @netdev: network interface device structure
4413 * @p: pointer to an address structure
4414 *
4415 * Returns 0 on success, negative on failure
4416 **/
4417static int e1000_set_mac(struct net_device *netdev, void *p)
4418{
4419        struct e1000_adapter *adapter = netdev_priv(netdev);
4420        struct e1000_hw *hw = &adapter->hw;
4421        struct sockaddr *addr = p;
4422
4423        if (!is_valid_ether_addr(addr->sa_data))
4424                return -EADDRNOTAVAIL;
4425
4426        memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4427        memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4428
4429        hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4430
4431        if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4432                /* activate the work around */
4433                e1000e_set_laa_state_82571(&adapter->hw, 1);
4434
4435                /* Hold a copy of the LAA in RAR[14] This is done so that
4436                 * between the time RAR[0] gets clobbered  and the time it
4437                 * gets fixed (in e1000_watchdog), the actual LAA is in one
4438                 * of the RARs and no incoming packets directed to this port
4439                 * are dropped. Eventually the LAA will be in RAR[0] and
4440                 * RAR[14]
4441                 */
4442                hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4443                                    adapter->hw.mac.rar_entry_count - 1);
4444        }
4445
4446        return 0;
4447}
4448
4449/**
4450 * e1000e_update_phy_task - work thread to update phy
4451 * @work: pointer to our work struct
4452 *
4453 * this worker thread exists because we must acquire a
4454 * semaphore to read the phy, which we could msleep while
4455 * waiting for it, and we can't msleep in a timer.
4456 **/
4457static void e1000e_update_phy_task(struct work_struct *work)
4458{
4459        struct e1000_adapter *adapter = container_of(work,
4460                                                     struct e1000_adapter,
4461                                                     update_phy_task);
4462
4463        if (test_bit(__E1000_DOWN, &adapter->state))
4464                return;
4465
4466        e1000_get_phy_info(&adapter->hw);
4467}
4468
4469/**
4470 * e1000_update_phy_info - timre call-back to update PHY info
4471 * @data: pointer to adapter cast into an unsigned long
4472 *
4473 * Need to wait a few seconds after link up to get diagnostic information from
4474 * the phy
4475 **/
4476static void e1000_update_phy_info(unsigned long data)
4477{
4478        struct e1000_adapter *adapter = (struct e1000_adapter *)data;
4479
4480        if (test_bit(__E1000_DOWN, &adapter->state))
4481                return;
4482
4483        schedule_work(&adapter->update_phy_task);
4484}
4485
4486/**
4487 * e1000e_update_phy_stats - Update the PHY statistics counters
4488 * @adapter: board private structure
4489 *
4490 * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4491 **/
4492static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4493{
4494        struct e1000_hw *hw = &adapter->hw;
4495        s32 ret_val;
4496        u16 phy_data;
4497
4498        ret_val = hw->phy.ops.acquire(hw);
4499        if (ret_val)
4500                return;
4501
4502        /* A page set is expensive so check if already on desired page.
4503         * If not, set to the page with the PHY status registers.
4504         */
4505        hw->phy.addr = 1;
4506        ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4507                                           &phy_data);
4508        if (ret_val)
4509                goto release;
4510        if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4511                ret_val = hw->phy.ops.set_page(hw,
4512                                               HV_STATS_PAGE << IGP_PAGE_SHIFT);
4513                if (ret_val)
4514                        goto release;
4515        }
4516
4517        /* Single Collision Count */
4518        hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4519        ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4520        if (!ret_val)
4521                adapter->stats.scc += phy_data;
4522
4523        /* Excessive Collision Count */
4524        hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4525        ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4526        if (!ret_val)
4527                adapter->stats.ecol += phy_data;
4528
4529        /* Multiple Collision Count */
4530        hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4531        ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4532        if (!ret_val)
4533                adapter->stats.mcc += phy_data;
4534
4535        /* Late Collision Count */
4536        hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4537        ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4538        if (!ret_val)
4539                adapter->stats.latecol += phy_data;
4540
4541        /* Collision Count - also used for adaptive IFS */
4542        hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4543        ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4544        if (!ret_val)
4545                hw->mac.collision_delta = phy_data;
4546
4547        /* Defer Count */
4548        hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4549        ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4550        if (!ret_val)
4551                adapter->stats.dc += phy_data;
4552
4553        /* Transmit with no CRS */
4554        hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4555        ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4556        if (!ret_val)
4557                adapter->stats.tncrs += phy_data;
4558
4559release:
4560        hw->phy.ops.release(hw);
4561}
4562
4563/**
4564 * e1000e_update_stats - Update the board statistics counters
4565 * @adapter: board private structure
4566 **/
4567static void e1000e_update_stats(struct e1000_adapter *adapter)
4568{
4569        struct net_device *netdev = adapter->netdev;
4570        struct e1000_hw *hw = &adapter->hw;
4571        struct pci_dev *pdev = adapter->pdev;
4572
4573        /* Prevent stats update while adapter is being reset, or if the pci
4574         * connection is down.
4575         */
4576        if (adapter->link_speed == 0)
4577                return;
4578        if (pci_channel_offline(pdev))
4579                return;
4580
4581        adapter->stats.crcerrs += er32(CRCERRS);
4582        adapter->stats.gprc += er32(GPRC);
4583        adapter->stats.gorc += er32(GORCL);
4584        er32(GORCH);            /* Clear gorc */
4585        adapter->stats.bprc += er32(BPRC);
4586        adapter->stats.mprc += er32(MPRC);
4587        adapter->stats.roc += er32(ROC);
4588
4589        adapter->stats.mpc += er32(MPC);
4590
4591        /* Half-duplex statistics */
4592        if (adapter->link_duplex == HALF_DUPLEX) {
4593                if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4594                        e1000e_update_phy_stats(adapter);
4595                } else {
4596                        adapter->stats.scc += er32(SCC);
4597                        adapter->stats.ecol += er32(ECOL);
4598                        adapter->stats.mcc += er32(MCC);
4599                        adapter->stats.latecol += er32(LATECOL);
4600                        adapter->stats.dc += er32(DC);
4601
4602                        hw->mac.collision_delta = er32(COLC);
4603
4604                        if ((hw->mac.type != e1000_82574) &&
4605                            (hw->mac.type != e1000_82583))
4606                                adapter->stats.tncrs += er32(TNCRS);
4607                }
4608                adapter->stats.colc += hw->mac.collision_delta;
4609        }
4610
4611        adapter->stats.xonrxc += er32(XONRXC);
4612        adapter->stats.xontxc += er32(XONTXC);
4613        adapter->stats.xoffrxc += er32(XOFFRXC);
4614        adapter->stats.xofftxc += er32(XOFFTXC);
4615        adapter->stats.gptc += er32(GPTC);
4616        adapter->stats.gotc += er32(GOTCL);
4617        er32(GOTCH);            /* Clear gotc */
4618        adapter->stats.rnbc += er32(RNBC);
4619        adapter->stats.ruc += er32(RUC);
4620
4621        adapter->stats.mptc += er32(MPTC);
4622        adapter->stats.bptc += er32(BPTC);
4623
4624        /* used for adaptive IFS */
4625
4626        hw->mac.tx_packet_delta = er32(TPT);
4627        adapter->stats.tpt += hw->mac.tx_packet_delta;
4628
4629        adapter->stats.algnerrc += er32(ALGNERRC);
4630        adapter->stats.rxerrc += er32(RXERRC);
4631        adapter->stats.cexterr += er32(CEXTERR);
4632        adapter->stats.tsctc += er32(TSCTC);
4633        adapter->stats.tsctfc += er32(TSCTFC);
4634
4635        /* Fill out the OS statistics structure */
4636        netdev->stats.multicast = adapter->stats.mprc;
4637        netdev->stats.collisions = adapter->stats.colc;
4638
4639        /* Rx Errors */
4640
4641        /* RLEC on some newer hardware can be incorrect so build
4642         * our own version based on RUC and ROC
4643         */
4644        netdev->stats.rx_errors = adapter->stats.rxerrc +
4645            adapter->stats.crcerrs + adapter->stats.algnerrc +
4646            adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
4647        netdev->stats.rx_length_errors = adapter->stats.ruc +
4648            adapter->stats.roc;
4649        netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
4650        netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
4651        netdev->stats.rx_missed_errors = adapter->stats.mpc;
4652
4653        /* Tx Errors */
4654        netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
4655        netdev->stats.tx_aborted_errors = adapter->stats.ecol;
4656        netdev->stats.tx_window_errors = adapter->stats.latecol;
4657        netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
4658
4659        /* Tx Dropped needs to be maintained elsewhere */
4660
4661        /* Management Stats */
4662        adapter->stats.mgptc += er32(MGTPTC);
4663        adapter->stats.mgprc += er32(MGTPRC);
4664        adapter->stats.mgpdc += er32(MGTPDC);
4665
4666        /* Correctable ECC Errors */
4667        if (hw->mac.type == e1000_pch_lpt) {
4668                u32 pbeccsts = er32(PBECCSTS);
4669                adapter->corr_errors +=
4670                    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
4671                adapter->uncorr_errors +=
4672                    (pbeccsts & E1000_PBECCSTS_UNCORR_ERR_CNT_MASK) >>
4673                    E1000_PBECCSTS_UNCORR_ERR_CNT_SHIFT;
4674        }
4675}
4676
4677/**
4678 * e1000_phy_read_status - Update the PHY register status snapshot
4679 * @adapter: board private structure
4680 **/
4681static void e1000_phy_read_status(struct e1000_adapter *adapter)
4682{
4683        struct e1000_hw *hw = &adapter->hw;
4684        struct e1000_phy_regs *phy = &adapter->phy_regs;
4685
4686        if ((er32(STATUS) & E1000_STATUS_LU) &&
4687            (adapter->hw.phy.media_type == e1000_media_type_copper)) {
4688                int ret_val;
4689
4690                pm_runtime_get_sync(&adapter->pdev->dev);
4691                ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
4692                ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
4693                ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
4694                ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
4695                ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
4696                ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
4697                ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
4698                ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
4699                if (ret_val)
4700                        e_warn("Error reading PHY register\n");
4701                pm_runtime_put_sync(&adapter->pdev->dev);
4702        } else {
4703                /* Do not read PHY registers if link is not up
4704                 * Set values to typical power-on defaults
4705                 */
4706                phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
4707                phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
4708                             BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
4709                             BMSR_ERCAP);
4710                phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
4711                                  ADVERTISE_ALL | ADVERTISE_CSMA);
4712                phy->lpa = 0;
4713                phy->expansion = EXPANSION_ENABLENPAGE;
4714                phy->ctrl1000 = ADVERTISE_1000FULL;
4715                phy->stat1000 = 0;
4716                phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
4717        }
4718}
4719
4720static void e1000_print_link_info(struct e1000_adapter *adapter)
4721{
4722        struct e1000_hw *hw = &adapter->hw;
4723        u32 ctrl = er32(CTRL);
4724
4725        /* Link status message must follow this format for user tools */
4726        pr_info("%s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
4727                adapter->netdev->name, adapter->link_speed,
4728                adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
4729                (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
4730                (ctrl & E1000_CTRL_RFCE) ? "Rx" :
4731                (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
4732}
4733
4734static bool e1000e_has_link(struct e1000_adapter *adapter)
4735{
4736        struct e1000_hw *hw = &adapter->hw;
4737        bool link_active = false;
4738        s32 ret_val = 0;
4739
4740        /* get_link_status is set on LSC (link status) interrupt or
4741         * Rx sequence error interrupt.  get_link_status will stay
4742         * false until the check_for_link establishes link
4743         * for copper adapters ONLY
4744         */
4745        switch (hw->phy.media_type) {
4746        case e1000_media_type_copper:
4747                if (hw->mac.get_link_status) {
4748                        ret_val = hw->mac.ops.check_for_link(hw);
4749                        link_active = !hw->mac.get_link_status;
4750                } else {
4751                        link_active = true;
4752                }
4753                break;
4754        case e1000_media_type_fiber:
4755                ret_val = hw->mac.ops.check_for_link(hw);
4756                link_active = !!(er32(STATUS) & E1000_STATUS_LU);
4757                break;
4758        case e1000_media_type_internal_serdes:
4759                ret_val = hw->mac.ops.check_for_link(hw);
4760                link_active = adapter->hw.mac.serdes_has_link;
4761                break;
4762        default:
4763        case e1000_media_type_unknown:
4764                break;
4765        }
4766
4767        if ((ret_val == E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
4768            (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
4769                /* See e1000_kmrn_lock_loss_workaround_ich8lan() */
4770                e_info("Gigabit has been disabled, downgrading speed\n");
4771        }
4772
4773        return link_active;
4774}
4775
4776static void e1000e_enable_receives(struct e1000_adapter *adapter)
4777{
4778        /* make sure the receive unit is started */
4779        if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
4780            (adapter->flags & FLAG_RESTART_NOW)) {
4781                struct e1000_hw *hw = &adapter->hw;
4782                u32 rctl = er32(RCTL);
4783                ew32(RCTL, rctl | E1000_RCTL_EN);
4784                adapter->flags &= ~FLAG_RESTART_NOW;
4785        }
4786}
4787
4788static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
4789{
4790        struct e1000_hw *hw = &adapter->hw;
4791
4792        /* With 82574 controllers, PHY needs to be checked periodically
4793         * for hung state and reset, if two calls return true
4794         */
4795        if (e1000_check_phy_82574(hw))
4796                adapter->phy_hang_count++;
4797        else
4798                adapter->phy_hang_count = 0;
4799
4800        if (adapter->phy_hang_count > 1) {
4801                adapter->phy_hang_count = 0;
4802                schedule_work(&adapter->reset_task);
4803        }
4804}
4805
4806/**
4807 * e1000_watchdog - Timer Call-back
4808 * @data: pointer to adapter cast into an unsigned long
4809 **/
4810static void e1000_watchdog(unsigned long data)
4811{
4812        struct e1000_adapter *adapter = (struct e1000_adapter *)data;
4813
4814        /* Do the rest outside of interrupt context */
4815        schedule_work(&adapter->watchdog_task);
4816
4817        /* TODO: make this use queue_delayed_work() */
4818}
4819
4820static void e1000_watchdog_task(struct work_struct *work)
4821{
4822        struct e1000_adapter *adapter = container_of(work,
4823                                                     struct e1000_adapter,
4824                                                     watchdog_task);
4825        struct net_device *netdev = adapter->netdev;
4826        struct e1000_mac_info *mac = &adapter->hw.mac;
4827        struct e1000_phy_info *phy = &adapter->hw.phy;
4828        struct e1000_ring *tx_ring = adapter->tx_ring;
4829        struct e1000_hw *hw = &adapter->hw;
4830        u32 link, tctl;
4831
4832        if (test_bit(__E1000_DOWN, &adapter->state))
4833                return;
4834
4835        link = e1000e_has_link(adapter);
4836        if ((netif_carrier_ok(netdev)) && link) {
4837                /* Cancel scheduled suspend requests. */
4838                pm_runtime_resume(netdev->dev.parent);
4839
4840                e1000e_enable_receives(adapter);
4841                goto link_up;
4842        }
4843
4844        if ((e1000e_enable_tx_pkt_filtering(hw)) &&
4845            (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
4846                e1000_update_mng_vlan(adapter);
4847
4848        if (link) {
4849                if (!netif_carrier_ok(netdev)) {
4850                        bool txb2b = true;
4851
4852                        /* Cancel scheduled suspend requests. */
4853                        pm_runtime_resume(netdev->dev.parent);
4854
4855                        /* update snapshot of PHY registers on LSC */
4856                        e1000_phy_read_status(adapter);
4857                        mac->ops.get_link_up_info(&adapter->hw,
4858                                                  &adapter->link_speed,
4859                                                  &adapter->link_duplex);
4860                        e1000_print_link_info(adapter);
4861
4862                        /* check if SmartSpeed worked */
4863                        e1000e_check_downshift(hw);
4864                        if (phy->speed_downgraded)
4865                                netdev_warn(netdev,
4866                                            "Link Speed was downgraded by SmartSpeed\n");
4867
4868                        /* On supported PHYs, check for duplex mismatch only
4869                         * if link has autonegotiated at 10/100 half
4870                         */
4871                        if ((hw->phy.type == e1000_phy_igp_3 ||
4872                             hw->phy.type == e1000_phy_bm) &&
4873                            (hw->mac.autoneg == true) &&
4874                            (adapter->link_speed == SPEED_10 ||
4875                             adapter->link_speed == SPEED_100) &&
4876                            (adapter->link_duplex == HALF_DUPLEX)) {
4877                                u16 autoneg_exp;
4878
4879                                e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
4880
4881                                if (!(autoneg_exp & EXPANSION_NWAY))
4882                                        e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
4883                        }
4884
4885                        /* adjust timeout factor according to speed/duplex */
4886                        adapter->tx_timeout_factor = 1;
4887                        switch (adapter->link_speed) {
4888                        case SPEED_10:
4889                                txb2b = false;
4890                                adapter->tx_timeout_factor = 16;
4891                                break;
4892                        case SPEED_100:
4893                                txb2b = false;
4894                                adapter->tx_timeout_factor = 10;
4895                                break;
4896                        }
4897
4898                        /* workaround: re-program speed mode bit after
4899                         * link-up event
4900                         */
4901                        if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
4902                            !txb2b) {
4903                                u32 tarc0;
4904                                tarc0 = er32(TARC(0));
4905                                tarc0 &= ~SPEED_MODE_BIT;
4906                                ew32(TARC(0), tarc0);
4907                        }
4908
4909                        /* disable TSO for pcie and 10/100 speeds, to avoid
4910                         * some hardware issues
4911                         */
4912                        if (!(adapter->flags & FLAG_TSO_FORCE)) {
4913                                switch (adapter->link_speed) {
4914                                case SPEED_10:
4915                                case SPEED_100:
4916                                        e_info("10/100 speed: disabling TSO\n");
4917                                        netdev->features &= ~NETIF_F_TSO;
4918                                        netdev->features &= ~NETIF_F_TSO6;
4919                                        break;
4920                                case SPEED_1000:
4921                                        netdev->features |= NETIF_F_TSO;
4922                                        netdev->features |= NETIF_F_TSO6;
4923                                        break;
4924                                default:
4925                                        /* oops */
4926                                        break;
4927                                }
4928                        }
4929
4930                        /* enable transmits in the hardware, need to do this
4931                         * after setting TARC(0)
4932                         */
4933                        tctl = er32(TCTL);
4934                        tctl |= E1000_TCTL_EN;
4935                        ew32(TCTL, tctl);
4936
4937                        /* Perform any post-link-up configuration before
4938                         * reporting link up.
4939                         */
4940                        if (phy->ops.cfg_on_link_up)
4941                                phy->ops.cfg_on_link_up(hw);
4942
4943                        netif_carrier_on(netdev);
4944
4945                        if (!test_bit(__E1000_DOWN, &adapter->state))
4946                                mod_timer(&adapter->phy_info_timer,
4947                                          round_jiffies(jiffies + 2 * HZ));
4948                }
4949        } else {
4950                if (netif_carrier_ok(netdev)) {
4951                        adapter->link_speed = 0;
4952                        adapter->link_duplex = 0;
4953                        /* Link status message must follow this format */
4954                        pr_info("%s NIC Link is Down\n", adapter->netdev->name);
4955                        netif_carrier_off(netdev);
4956                        if (!test_bit(__E1000_DOWN, &adapter->state))
4957                                mod_timer(&adapter->phy_info_timer,
4958                                          round_jiffies(jiffies + 2 * HZ));
4959
4960                        /* The link is lost so the controller stops DMA.
4961                         * If there is queued Tx work that cannot be done
4962                         * or if on an 8000ES2LAN which requires a Rx packet
4963                         * buffer work-around on link down event, reset the
4964                         * controller to flush the Tx/Rx packet buffers.
4965                         * (Do the reset outside of interrupt context).
4966                         */
4967                        if ((adapter->flags & FLAG_RX_NEEDS_RESTART) ||
4968                            (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
4969                                adapter->flags |= FLAG_RESTART_NOW;
4970                        else
4971                                pm_schedule_suspend(netdev->dev.parent,
4972                                                    LINK_TIMEOUT);
4973                }
4974        }
4975
4976link_up:
4977        spin_lock(&adapter->stats64_lock);
4978        e1000e_update_stats(adapter);
4979
4980        mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
4981        adapter->tpt_old = adapter->stats.tpt;
4982        mac->collision_delta = adapter->stats.colc - adapter->colc_old;
4983        adapter->colc_old = adapter->stats.colc;
4984
4985        adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
4986        adapter->gorc_old = adapter->stats.gorc;
4987        adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
4988        adapter->gotc_old = adapter->stats.gotc;
4989        spin_unlock(&adapter->stats64_lock);
4990
4991        if (adapter->flags & FLAG_RESTART_NOW) {
4992                schedule_work(&adapter->reset_task);
4993                /* return immediately since reset is imminent */
4994                return;
4995        }
4996
4997        e1000e_update_adaptive(&adapter->hw);
4998
4999        /* Simple mode for Interrupt Throttle Rate (ITR) */
5000        if (adapter->itr_setting == 4) {
5001                /* Symmetric Tx/Rx gets a reduced ITR=2000;
5002                 * Total asymmetrical Tx or Rx gets ITR=8000;
5003                 * everyone else is between 2000-8000.
5004                 */
5005                u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5006                u32 dif = (adapter->gotc > adapter->gorc ?
5007                           adapter->gotc - adapter->gorc :
5008                           adapter->gorc - adapter->gotc) / 10000;
5009                u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5010
5011                e1000e_write_itr(adapter, itr);
5012        }
5013
5014        /* Cause software interrupt to ensure Rx ring is cleaned */
5015        if (adapter->msix_entries)
5016                ew32(ICS, adapter->rx_ring->ims_val);
5017        else
5018                ew32(ICS, E1000_ICS_RXDMT0);
5019
5020        /* flush pending descriptors to memory before detecting Tx hang */
5021        e1000e_flush_descriptors(adapter);
5022
5023        /* Force detection of hung controller every watchdog period */
5024        adapter->detect_tx_hung = true;
5025
5026        /* With 82571 controllers, LAA may be overwritten due to controller
5027         * reset from the other port. Set the appropriate LAA in RAR[0]
5028         */
5029        if (e1000e_get_laa_state_82571(hw))
5030                hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5031
5032        if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5033                e1000e_check_82574_phy_workaround(adapter);
5034
5035        /* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5036        if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5037                if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5038                    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5039                        er32(RXSTMPH);
5040                        adapter->rx_hwtstamp_cleared++;
5041                } else {
5042                        adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5043                }
5044        }
5045
5046        /* Reset the timer */
5047        if (!test_bit(__E1000_DOWN, &adapter->state))
5048                mod_timer(&adapter->watchdog_timer,
5049                          round_jiffies(jiffies + 2 * HZ));
5050}
5051
5052#define E1000_TX_FLAGS_CSUM             0x00000001
5053#define E1000_TX_FLAGS_VLAN             0x00000002
5054#define E1000_TX_FLAGS_TSO              0x00000004
5055#define E1000_TX_FLAGS_IPV4             0x00000008
5056#define E1000_TX_FLAGS_NO_FCS           0x00000010
5057#define E1000_TX_FLAGS_HWTSTAMP         0x00000020
5058#define E1000_TX_FLAGS_VLAN_MASK        0xffff0000
5059#define E1000_TX_FLAGS_VLAN_SHIFT       16
5060
5061static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb)
5062{
5063        struct e1000_context_desc *context_desc;
5064        struct e1000_buffer *buffer_info;
5065        unsigned int i;
5066        u32 cmd_length = 0;
5067        u16 ipcse = 0, mss;
5068        u8 ipcss, ipcso, tucss, tucso, hdr_len;
5069
5070        if (!skb_is_gso(skb))
5071                return 0;
5072
5073        if (skb_header_cloned(skb)) {
5074                int err = pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
5075
5076                if (err)
5077                        return err;
5078        }
5079
5080        hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5081        mss = skb_shinfo(skb)->gso_size;
5082        if (skb->protocol == htons(ETH_P_IP)) {
5083                struct iphdr *iph = ip_hdr(skb);
5084                iph->tot_len = 0;
5085                iph->check = 0;
5086                tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5087                                                         0, IPPROTO_TCP, 0);
5088                cmd_length = E1000_TXD_CMD_IP;
5089                ipcse = skb_transport_offset(skb) - 1;
5090        } else if (skb_is_gso_v6(skb)) {
5091                ipv6_hdr(skb)->payload_len = 0;
5092                tcp_hdr(skb)->check = ~csum_ipv6_magic(&ipv6_hdr(skb)->saddr,
5093                                                       &ipv6_hdr(skb)->daddr,
5094                                                       0, IPPROTO_TCP, 0);
5095                ipcse = 0;
5096        }
5097        ipcss = skb_network_offset(skb);
5098        ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5099        tucss = skb_transport_offset(skb);
5100        tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5101
5102        cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5103                       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5104
5105        i = tx_ring->next_to_use;
5106        context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5107        buffer_info = &tx_ring->buffer_info[i];
5108
5109        context_desc->lower_setup.ip_fields.ipcss = ipcss;
5110        context_desc->lower_setup.ip_fields.ipcso = ipcso;
5111        context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5112        context_desc->upper_setup.tcp_fields.tucss = tucss;
5113        context_desc->upper_setup.tcp_fields.tucso = tucso;
5114        context_desc->upper_setup.tcp_fields.tucse = 0;
5115        context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5116        context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5117        context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5118
5119        buffer_info->time_stamp = jiffies;
5120        buffer_info->next_to_watch = i;
5121
5122        i++;
5123        if (i == tx_ring->count)
5124                i = 0;
5125        tx_ring->next_to_use = i;
5126
5127        return 1;
5128}
5129
5130static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb)
5131{
5132        struct e1000_adapter *adapter = tx_ring->adapter;
5133        struct e1000_context_desc *context_desc;
5134        struct e1000_buffer *buffer_info;
5135        unsigned int i;
5136        u8 css;
5137        u32 cmd_len = E1000_TXD_CMD_DEXT;
5138        __be16 protocol;
5139
5140        if (skb->ip_summed != CHECKSUM_PARTIAL)
5141                return 0;
5142
5143        if (skb->protocol == cpu_to_be16(ETH_P_8021Q))
5144                protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
5145        else
5146                protocol = skb->protocol;
5147
5148        switch (protocol) {
5149        case cpu_to_be16(ETH_P_IP):
5150                if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5151                        cmd_len |= E1000_TXD_CMD_TCP;
5152                break;
5153        case cpu_to_be16(ETH_P_IPV6):
5154                /* XXX not handling all IPV6 headers */
5155                if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5156                        cmd_len |= E1000_TXD_CMD_TCP;
5157                break;
5158        default:
5159                if (unlikely(net_ratelimit()))
5160                        e_warn("checksum_partial proto=%x!\n",
5161                               be16_to_cpu(protocol));
5162                break;
5163        }
5164
5165        css = skb_checksum_start_offset(skb);
5166
5167        i = tx_ring->next_to_use;
5168        buffer_info = &tx_ring->buffer_info[i];
5169        context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5170
5171        context_desc->lower_setup.ip_config = 0;
5172        context_desc->upper_setup.tcp_fields.tucss = css;
5173        context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5174        context_desc->upper_setup.tcp_fields.tucse = 0;
5175        context_desc->tcp_seg_setup.data = 0;
5176        context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5177
5178        buffer_info->time_stamp = jiffies;
5179        buffer_info->next_to_watch = i;
5180
5181        i++;
5182        if (i == tx_ring->count)
5183                i = 0;
5184        tx_ring->next_to_use = i;
5185
5186        return 1;
5187}
5188
5189static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5190                        unsigned int first, unsigned int max_per_txd,
5191                        unsigned int nr_frags)
5192{
5193        struct e1000_adapter *adapter = tx_ring->adapter;
5194        struct pci_dev *pdev = adapter->pdev;
5195        struct e1000_buffer *buffer_info;
5196        unsigned int len = skb_headlen(skb);
5197        unsigned int offset = 0, size, count = 0, i;
5198        unsigned int f, bytecount, segs;
5199
5200        i = tx_ring->next_to_use;
5201
5202        while (len) {
5203                buffer_info = &tx_ring->buffer_info[i];
5204                size = min(len, max_per_txd);
5205
5206                buffer_info->length = size;
5207                buffer_info->time_stamp = jiffies;
5208                buffer_info->next_to_watch = i;
5209                buffer_info->dma = dma_map_single(&pdev->dev,
5210                                                  skb->data + offset,
5211                                                  size, DMA_TO_DEVICE);
5212                buffer_info->mapped_as_page = false;
5213                if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5214                        goto dma_error;
5215
5216                len -= size;
5217                offset += size;
5218                count++;
5219
5220                if (len) {
5221                        i++;
5222                        if (i == tx_ring->count)
5223                                i = 0;
5224                }
5225        }
5226
5227        for (f = 0; f < nr_frags; f++) {
5228                const struct skb_frag_struct *frag;
5229
5230                frag = &skb_shinfo(skb)->frags[f];
5231                len = skb_frag_size(frag);
5232                offset = 0;
5233
5234                while (len) {
5235                        i++;
5236                        if (i == tx_ring->count)
5237                                i = 0;
5238
5239                        buffer_info = &tx_ring->buffer_info[i];
5240                        size = min(len, max_per_txd);
5241
5242                        buffer_info->length = size;
5243                        buffer_info->time_stamp = jiffies;
5244                        buffer_info->next_to_watch = i;
5245                        buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5246                                                            offset, size,
5247                                                            DMA_TO_DEVICE);
5248                        buffer_info->mapped_as_page = true;
5249                        if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5250                                goto dma_error;
5251
5252                        len -= size;
5253                        offset += size;
5254                        count++;
5255                }
5256        }
5257
5258        segs = skb_shinfo(skb)->gso_segs ? : 1;
5259        /* multiply data chunks by size of headers */
5260        bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5261
5262        tx_ring->buffer_info[i].skb = skb;
5263        tx_ring->buffer_info[i].segs = segs;
5264        tx_ring->buffer_info[i].bytecount = bytecount;
5265        tx_ring->buffer_info[first].next_to_watch = i;
5266
5267        return count;
5268
5269dma_error:
5270        dev_err(&pdev->dev, "Tx DMA map failed\n");
5271        buffer_info->dma = 0;
5272        if (count)
5273                count--;
5274
5275        while (count--) {
5276                if (i == 0)
5277                        i += tx_ring->count;
5278                i--;
5279                buffer_info = &tx_ring->buffer_info[i];
5280                e1000_put_txbuf(tx_ring, buffer_info);
5281        }
5282
5283        return 0;
5284}
5285
5286static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5287{
5288        struct e1000_adapter *adapter = tx_ring->adapter;
5289        struct e1000_tx_desc *tx_desc = NULL;
5290        struct e1000_buffer *buffer_info;
5291        u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5292        unsigned int i;
5293
5294        if (tx_flags & E1000_TX_FLAGS_TSO) {
5295                txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5296                    E1000_TXD_CMD_TSE;
5297                txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5298
5299                if (tx_flags & E1000_TX_FLAGS_IPV4)
5300                        txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5301        }
5302
5303        if (tx_flags & E1000_TX_FLAGS_CSUM) {
5304                txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5305                txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5306        }
5307
5308        if (tx_flags & E1000_TX_FLAGS_VLAN) {
5309                txd_lower |= E1000_TXD_CMD_VLE;
5310                txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5311        }
5312
5313        if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5314                txd_lower &= ~(E1000_TXD_CMD_IFCS);
5315
5316        if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5317                txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5318                txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5319        }
5320
5321        i = tx_ring->next_to_use;
5322
5323        do {
5324                buffer_info = &tx_ring->buffer_info[i];
5325                tx_desc = E1000_TX_DESC(*tx_ring, i);
5326                tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5327                tx_desc->lower.data = cpu_to_le32(txd_lower |
5328                                                  buffer_info->length);
5329                tx_desc->upper.data = cpu_to_le32(txd_upper);
5330
5331                i++;
5332                if (i == tx_ring->count)
5333                        i = 0;
5334        } while (--count > 0);
5335
5336        tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5337
5338        /* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5339        if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5340                tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5341
5342        /* Force memory writes to complete before letting h/w
5343         * know there are new descriptors to fetch.  (Only
5344         * applicable for weak-ordered memory model archs,
5345         * such as IA-64).
5346         */
5347        wmb();
5348
5349        tx_ring->next_to_use = i;
5350
5351        if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5352                e1000e_update_tdt_wa(tx_ring, i);
5353        else
5354                writel(i, tx_ring->tail);
5355
5356        /* we need this if more than one processor can write to our tail
5357         * at a time, it synchronizes IO on IA64/Altix systems
5358         */
5359        mmiowb();
5360}
5361
5362#define MINIMUM_DHCP_PACKET_SIZE 282
5363static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5364                                    struct sk_buff *skb)
5365{
5366        struct e1000_hw *hw = &adapter->hw;
5367        u16 length, offset;
5368
5369        if (vlan_tx_tag_present(skb) &&
5370            !((vlan_tx_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5371              (adapter->hw.mng_cookie.status &
5372               E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5373                return 0;
5374
5375        if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5376                return 0;
5377
5378        if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5379                return 0;
5380
5381        {
5382                const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5383                struct udphdr *udp;
5384
5385                if (ip->protocol != IPPROTO_UDP)
5386                        return 0;
5387
5388                udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5389                if (ntohs(udp->dest) != 67)
5390                        return 0;
5391
5392                offset = (u8 *)udp + 8 - skb->data;
5393                length = skb->len - offset;
5394                return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5395        }
5396
5397        return 0;
5398}
5399
5400static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5401{
5402        struct e1000_adapter *adapter = tx_ring->adapter;
5403
5404        netif_stop_queue(adapter->netdev);
5405        /* Herbert's original patch had:
5406         *  smp_mb__after_netif_stop_queue();
5407         * but since that doesn't exist yet, just open code it.
5408         */
5409        smp_mb();
5410
5411        /* We need to check again in a case another CPU has just
5412         * made room available.
5413         */
5414        if (e1000_desc_unused(tx_ring) < size)
5415                return -EBUSY;
5416
5417        /* A reprieve! */
5418        netif_start_queue(adapter->netdev);
5419        ++adapter->restart_queue;
5420        return 0;
5421}
5422
5423static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5424{
5425        BUG_ON(size > tx_ring->count);
5426
5427        if (e1000_desc_unused(tx_ring) >= size)
5428                return 0;
5429        return __e1000_maybe_stop_tx(tx_ring, size);
5430}
5431
5432static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5433                                    struct net_device *netdev)
5434{
5435        struct e1000_adapter *adapter = netdev_priv(netdev);
5436        struct e1000_ring *tx_ring = adapter->tx_ring;
5437        unsigned int first;
5438        unsigned int tx_flags = 0;
5439        unsigned int len = skb_headlen(skb);
5440        unsigned int nr_frags;
5441        unsigned int mss;
5442        int count = 0;
5443        int tso;
5444        unsigned int f;
5445
5446        if (test_bit(__E1000_DOWN, &adapter->state)) {
5447                dev_kfree_skb_any(skb);
5448                return NETDEV_TX_OK;
5449        }
5450
5451        if (skb->len <= 0) {
5452                dev_kfree_skb_any(skb);
5453                return NETDEV_TX_OK;
5454        }
5455
5456        /* The minimum packet size with TCTL.PSP set is 17 bytes so
5457         * pad skb in order to meet this minimum size requirement
5458         */
5459        if (unlikely(skb->len < 17)) {
5460                if (skb_pad(skb, 17 - skb->len))
5461                        return NETDEV_TX_OK;
5462                skb->len = 17;
5463                skb_set_tail_pointer(skb, 17);
5464        }
5465
5466        mss = skb_shinfo(skb)->gso_size;
5467        if (mss) {
5468                u8 hdr_len;
5469
5470                /* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5471                 * points to just header, pull a few bytes of payload from
5472                 * frags into skb->data
5473                 */
5474                hdr_len = skb_transport_offset(skb) + tcp_hdrlen(skb);
5475                /* we do this workaround for ES2LAN, but it is un-necessary,
5476                 * avoiding it could save a lot of cycles
5477                 */
5478                if (skb->data_len && (hdr_len == len)) {
5479                        unsigned int pull_size;
5480
5481                        pull_size = min_t(unsigned int, 4, skb->data_len);
5482                        if (!__pskb_pull_tail(skb, pull_size)) {
5483                                e_err("__pskb_pull_tail failed.\n");
5484                                dev_kfree_skb_any(skb);
5485                                return NETDEV_TX_OK;
5486                        }
5487                        len = skb_headlen(skb);
5488                }
5489        }
5490
5491        /* reserve a descriptor for the offload context */
5492        if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5493                count++;
5494        count++;
5495
5496        count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5497
5498        nr_frags = skb_shinfo(skb)->nr_frags;
5499        for (f = 0; f < nr_frags; f++)
5500                count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5501                                      adapter->tx_fifo_limit);
5502
5503        if (adapter->hw.mac.tx_pkt_filtering)
5504                e1000_transfer_dhcp_info(adapter, skb);
5505
5506        /* need: count + 2 desc gap to keep tail from touching
5507         * head, otherwise try next time
5508         */
5509        if (e1000_maybe_stop_tx(tx_ring, count + 2))
5510                return NETDEV_TX_BUSY;
5511
5512        if (vlan_tx_tag_present(skb)) {
5513                tx_flags |= E1000_TX_FLAGS_VLAN;
5514                tx_flags |= (vlan_tx_tag_get(skb) << E1000_TX_FLAGS_VLAN_SHIFT);
5515        }
5516
5517        first = tx_ring->next_to_use;
5518
5519        tso = e1000_tso(tx_ring, skb);
5520        if (tso < 0) {
5521                dev_kfree_skb_any(skb);
5522                return NETDEV_TX_OK;
5523        }
5524
5525        if (tso)
5526                tx_flags |= E1000_TX_FLAGS_TSO;
5527        else if (e1000_tx_csum(tx_ring, skb))
5528                tx_flags |= E1000_TX_FLAGS_CSUM;
5529
5530        /* Old method was to assume IPv4 packet by default if TSO was enabled.
5531         * 82571 hardware supports TSO capabilities for IPv6 as well...
5532         * no longer assume, we must.
5533         */
5534        if (skb->protocol == htons(ETH_P_IP))
5535                tx_flags |= E1000_TX_FLAGS_IPV4;
5536
5537        if (unlikely(skb->no_fcs))
5538                tx_flags |= E1000_TX_FLAGS_NO_FCS;
5539
5540        /* if count is 0 then mapping error has occurred */
5541        count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5542                             nr_frags);
5543        if (count) {
5544                if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5545                             !adapter->tx_hwtstamp_skb)) {
5546                        skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5547                        tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5548                        adapter->tx_hwtstamp_skb = skb_get(skb);
5549                        schedule_work(&adapter->tx_hwtstamp_work);
5550                } else {
5551                        skb_tx_timestamp(skb);
5552                }
5553
5554                netdev_sent_queue(netdev, skb->len);
5555                e1000_tx_queue(tx_ring, tx_flags, count);
5556                /* Make sure there is space in the ring for the next send. */
5557                e1000_maybe_stop_tx(tx_ring,
5558                                    (MAX_SKB_FRAGS *
5559                                     DIV_ROUND_UP(PAGE_SIZE,
5560                                                  adapter->tx_fifo_limit) + 2));
5561        } else {
5562                dev_kfree_skb_any(skb);
5563                tx_ring->buffer_info[first].time_stamp = 0;
5564                tx_ring->next_to_use = first;
5565        }
5566
5567        return NETDEV_TX_OK;
5568}
5569
5570/**
5571 * e1000_tx_timeout - Respond to a Tx Hang
5572 * @netdev: network interface device structure
5573 **/
5574static void e1000_tx_timeout(struct net_device *netdev)
5575{
5576        struct e1000_adapter *adapter = netdev_priv(netdev);
5577
5578        /* Do the reset outside of interrupt context */
5579        adapter->tx_timeout_count++;
5580        schedule_work(&adapter->reset_task);
5581}
5582
5583static void e1000_reset_task(struct work_struct *work)
5584{
5585        struct e1000_adapter *adapter;
5586        adapter = container_of(work, struct e1000_adapter, reset_task);
5587
5588        /* don't run the task if already down */
5589        if (test_bit(__E1000_DOWN, &adapter->state))
5590                return;
5591
5592        if (!(adapter->flags & FLAG_RESTART_NOW)) {
5593                e1000e_dump(adapter);
5594                e_err("Reset adapter unexpectedly\n");
5595        }
5596        e1000e_reinit_locked(adapter);
5597}
5598
5599/**
5600 * e1000_get_stats64 - Get System Network Statistics
5601 * @netdev: network interface device structure
5602 * @stats: rtnl_link_stats64 pointer
5603 *
5604 * Returns the address of the device statistics structure.
5605 **/
5606struct rtnl_link_stats64 *e1000e_get_stats64(struct net_device *netdev,
5607                                             struct rtnl_link_stats64 *stats)
5608{
5609        struct e1000_adapter *adapter = netdev_priv(netdev);
5610
5611        memset(stats, 0, sizeof(struct rtnl_link_stats64));
5612        spin_lock(&adapter->stats64_lock);
5613        e1000e_update_stats(adapter);
5614        /* Fill out the OS statistics structure */
5615        stats->rx_bytes = adapter->stats.gorc;
5616        stats->rx_packets = adapter->stats.gprc;
5617        stats->tx_bytes = adapter->stats.gotc;
5618        stats->tx_packets = adapter->stats.gptc;
5619        stats->multicast = adapter->stats.mprc;
5620        stats->collisions = adapter->stats.colc;
5621
5622        /* Rx Errors */
5623
5624        /* RLEC on some newer hardware can be incorrect so build
5625         * our own version based on RUC and ROC
5626         */
5627        stats->rx_errors = adapter->stats.rxerrc +
5628            adapter->stats.crcerrs + adapter->stats.algnerrc +
5629            adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5630        stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
5631        stats->rx_crc_errors = adapter->stats.crcerrs;
5632        stats->rx_frame_errors = adapter->stats.algnerrc;
5633        stats->rx_missed_errors = adapter->stats.mpc;
5634
5635        /* Tx Errors */
5636        stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5637        stats->tx_aborted_errors = adapter->stats.ecol;
5638        stats->tx_window_errors = adapter->stats.latecol;
5639        stats->tx_carrier_errors = adapter->stats.tncrs;
5640
5641        /* Tx Dropped needs to be maintained elsewhere */
5642
5643        spin_unlock(&adapter->stats64_lock);
5644        return stats;
5645}
5646
5647/**
5648 * e1000_change_mtu - Change the Maximum Transfer Unit
5649 * @netdev: network interface device structure
5650 * @new_mtu: new value for maximum frame size
5651 *
5652 * Returns 0 on success, negative on failure
5653 **/
5654static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
5655{
5656        struct e1000_adapter *adapter = netdev_priv(netdev);
5657        int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
5658
5659        /* Jumbo frame support */
5660        if ((max_frame > ETH_FRAME_LEN + ETH_FCS_LEN) &&
5661            !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
5662                e_err("Jumbo Frames not supported.\n");
5663                return -EINVAL;
5664        }
5665
5666        /* Supported frame sizes */
5667        if ((new_mtu < ETH_ZLEN + ETH_FCS_LEN + VLAN_HLEN) ||
5668            (max_frame > adapter->max_hw_frame_size)) {
5669                e_err("Unsupported MTU setting\n");
5670                return -EINVAL;
5671        }
5672
5673        /* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
5674        if ((adapter->hw.mac.type >= e1000_pch2lan) &&
5675            !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
5676            (new_mtu > ETH_DATA_LEN)) {
5677                e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
5678                return -EINVAL;
5679        }
5680
5681        while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
5682                usleep_range(1000, 2000);
5683        /* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
5684        adapter->max_frame_size = max_frame;
5685        e_info("changing MTU from %d to %d\n", netdev->mtu, new_mtu);
5686        netdev->mtu = new_mtu;
5687        if (netif_running(netdev))
5688                e1000e_down(adapter);
5689
5690        /* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
5691         * means we reserve 2 more, this pushes us to allocate from the next
5692         * larger slab size.
5693         * i.e. RXBUFFER_2048 --> size-4096 slab
5694         * However with the new *_jumbo_rx* routines, jumbo receives will use
5695         * fragmented skbs
5696         */
5697
5698        if (max_frame <= 2048)
5699                adapter->rx_buffer_len = 2048;
5700        else
5701                adapter->rx_buffer_len = 4096;
5702
5703        /* adjust allocation if LPE protects us, and we aren't using SBP */
5704        if ((max_frame == ETH_FRAME_LEN + ETH_FCS_LEN) ||
5705            (max_frame == ETH_FRAME_LEN + VLAN_HLEN + ETH_FCS_LEN))
5706                adapter->rx_buffer_len = ETH_FRAME_LEN + VLAN_HLEN
5707                    + ETH_FCS_LEN;
5708
5709        if (netif_running(netdev))
5710                e1000e_up(adapter);
5711        else
5712                e1000e_reset(adapter);
5713
5714        clear_bit(__E1000_RESETTING, &adapter->state);
5715
5716        return 0;
5717}
5718
5719static int e1000_mii_ioctl(struct net_device *netdev, struct ifreq *ifr,
5720                           int cmd)
5721{
5722        struct e1000_adapter *adapter = netdev_priv(netdev);
5723        struct mii_ioctl_data *data = if_mii(ifr);
5724
5725        if (adapter->hw.phy.media_type != e1000_media_type_copper)
5726                return -EOPNOTSUPP;
5727
5728        switch (cmd) {
5729        case SIOCGMIIPHY:
5730                data->phy_id = adapter->hw.phy.addr;
5731                break;
5732        case SIOCGMIIREG:
5733                e1000_phy_read_status(adapter);
5734
5735                switch (data->reg_num & 0x1F) {
5736                case MII_BMCR:
5737                        data->val_out = adapter->phy_regs.bmcr;
5738                        break;
5739                case MII_BMSR:
5740                        data->val_out = adapter->phy_regs.bmsr;
5741                        break;
5742                case MII_PHYSID1:
5743                        data->val_out = (adapter->hw.phy.id >> 16);
5744                        break;
5745                case MII_PHYSID2:
5746                        data->val_out = (adapter->hw.phy.id & 0xFFFF);
5747                        break;
5748                case MII_ADVERTISE:
5749                        data->val_out = adapter->phy_regs.advertise;
5750                        break;
5751                case MII_LPA:
5752                        data->val_out = adapter->phy_regs.lpa;
5753                        break;
5754                case MII_EXPANSION:
5755                        data->val_out = adapter->phy_regs.expansion;
5756                        break;
5757                case MII_CTRL1000:
5758                        data->val_out = adapter->phy_regs.ctrl1000;
5759                        break;
5760                case MII_STAT1000:
5761                        data->val_out = adapter->phy_regs.stat1000;
5762                        break;
5763                case MII_ESTATUS:
5764                        data->val_out = adapter->phy_regs.estatus;
5765                        break;
5766                default:
5767                        return -EIO;
5768                }
5769                break;
5770        case SIOCSMIIREG:
5771        default:
5772                return -EOPNOTSUPP;
5773        }
5774        return 0;
5775}
5776
5777/**
5778 * e1000e_hwtstamp_ioctl - control hardware time stamping
5779 * @netdev: network interface device structure
5780 * @ifreq: interface request
5781 *
5782 * Outgoing time stamping can be enabled and disabled. Play nice and
5783 * disable it when requested, although it shouldn't cause any overhead
5784 * when no packet needs it. At most one packet in the queue may be
5785 * marked for time stamping, otherwise it would be impossible to tell
5786 * for sure to which packet the hardware time stamp belongs.
5787 *
5788 * Incoming time stamping has to be configured via the hardware filters.
5789 * Not all combinations are supported, in particular event type has to be
5790 * specified. Matching the kind of event packet is not supported, with the
5791 * exception of "all V2 events regardless of level 2 or 4".
5792 **/
5793static int e1000e_hwtstamp_ioctl(struct net_device *netdev, struct ifreq *ifr)
5794{
5795        struct e1000_adapter *adapter = netdev_priv(netdev);
5796        struct hwtstamp_config config;
5797        int ret_val;
5798
5799        if (copy_from_user(&config, ifr->ifr_data, sizeof(config)))
5800                return -EFAULT;
5801
5802        adapter->hwtstamp_config = config;
5803
5804        ret_val = e1000e_config_hwtstamp(adapter);
5805        if (ret_val)
5806                return ret_val;
5807
5808        config = adapter->hwtstamp_config;
5809
5810        switch (config.rx_filter) {
5811        case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
5812        case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
5813        case HWTSTAMP_FILTER_PTP_V2_SYNC:
5814        case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
5815        case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
5816        case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
5817                /* With V2 type filters which specify a Sync or Delay Request,
5818                 * Path Delay Request/Response messages are also time stamped
5819                 * by hardware so notify the caller the requested packets plus
5820                 * some others are time stamped.
5821                 */
5822                config.rx_filter = HWTSTAMP_FILTER_SOME;
5823                break;
5824        default:
5825                break;
5826        }
5827
5828        return copy_to_user(ifr->ifr_data, &config,
5829                            sizeof(config)) ? -EFAULT : 0;
5830}
5831
5832static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
5833{
5834        switch (cmd) {
5835        case SIOCGMIIPHY:
5836        case SIOCGMIIREG:
5837        case SIOCSMIIREG:
5838                return e1000_mii_ioctl(netdev, ifr, cmd);
5839        case SIOCSHWTSTAMP:
5840                return e1000e_hwtstamp_ioctl(netdev, ifr);
5841        default:
5842                return -EOPNOTSUPP;
5843        }
5844}
5845
5846static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
5847{
5848        struct e1000_hw *hw = &adapter->hw;
5849        u32 i, mac_reg;
5850        u16 phy_reg, wuc_enable;
5851        int retval;
5852
5853        /* copy MAC RARs to PHY RARs */
5854        e1000_copy_rx_addrs_to_phy_ich8lan(hw);
5855
5856        retval = hw->phy.ops.acquire(hw);
5857        if (retval) {
5858                e_err("Could not acquire PHY\n");
5859                return retval;
5860        }
5861
5862        /* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
5863        retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5864        if (retval)
5865                goto release;
5866
5867        /* copy MAC MTA to PHY MTA - only needed for pchlan */
5868        for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
5869                mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
5870                hw->phy.ops.write_reg_page(hw, BM_MTA(i),
5871                                           (u16)(mac_reg & 0xFFFF));
5872                hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
5873                                           (u16)((mac_reg >> 16) & 0xFFFF));
5874        }
5875
5876        /* configure PHY Rx Control register */
5877        hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
5878        mac_reg = er32(RCTL);
5879        if (mac_reg & E1000_RCTL_UPE)
5880                phy_reg |= BM_RCTL_UPE;
5881        if (mac_reg & E1000_RCTL_MPE)
5882                phy_reg |= BM_RCTL_MPE;
5883        phy_reg &= ~(BM_RCTL_MO_MASK);
5884        if (mac_reg & E1000_RCTL_MO_3)
5885                phy_reg |= (((mac_reg & E1000_RCTL_MO_3) >> E1000_RCTL_MO_SHIFT)
5886                            << BM_RCTL_MO_SHIFT);
5887        if (mac_reg & E1000_RCTL_BAM)
5888                phy_reg |= BM_RCTL_BAM;
5889        if (mac_reg & E1000_RCTL_PMCF)
5890                phy_reg |= BM_RCTL_PMCF;
5891        mac_reg = er32(CTRL);
5892        if (mac_reg & E1000_CTRL_RFCE)
5893                phy_reg |= BM_RCTL_RFCE;
5894        hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
5895
5896        /* enable PHY wakeup in MAC register */
5897        ew32(WUFC, wufc);
5898        ew32(WUC, E1000_WUC_PHY_WAKE | E1000_WUC_PME_EN);
5899
5900        /* configure and enable PHY wakeup in PHY registers */
5901        hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
5902        hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, E1000_WUC_PME_EN);
5903
5904        /* activate PHY wakeup */
5905        wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
5906        retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
5907        if (retval)
5908                e_err("Could not set PHY Host Wakeup bit\n");
5909release:
5910        hw->phy.ops.release(hw);
5911
5912        return retval;
5913}
5914
5915static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
5916{
5917        struct net_device *netdev = pci_get_drvdata(pdev);
5918        struct e1000_adapter *adapter = netdev_priv(netdev);
5919        struct e1000_hw *hw = &adapter->hw;
5920        u32 ctrl, ctrl_ext, rctl, status;
5921        /* Runtime suspend should only enable wakeup for link changes */
5922        u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
5923        int retval = 0;
5924
5925        netif_device_detach(netdev);
5926
5927        if (netif_running(netdev)) {
5928                int count = E1000_CHECK_RESET_COUNT;
5929
5930                while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
5931                        usleep_range(10000, 20000);
5932
5933                WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
5934                e1000e_down(adapter);
5935                e1000_free_irq(adapter);
5936        }
5937        e1000e_reset_interrupt_capability(adapter);
5938
5939        status = er32(STATUS);
5940        if (status & E1000_STATUS_LU)
5941                wufc &= ~E1000_WUFC_LNKC;
5942
5943        if (wufc) {
5944                e1000_setup_rctl(adapter);
5945                e1000e_set_rx_mode(netdev);
5946
5947                /* turn on all-multi mode if wake on multicast is enabled */
5948                if (wufc & E1000_WUFC_MC) {
5949                        rctl = er32(RCTL);
5950                        rctl |= E1000_RCTL_MPE;
5951                        ew32(RCTL, rctl);
5952                }
5953
5954                ctrl = er32(CTRL);
5955                ctrl |= E1000_CTRL_ADVD3WUC;
5956                if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
5957                        ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
5958                ew32(CTRL, ctrl);
5959
5960                if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
5961                    adapter->hw.phy.media_type ==
5962                    e1000_media_type_internal_serdes) {
5963                        /* keep the laser running in D3 */
5964                        ctrl_ext = er32(CTRL_EXT);
5965                        ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
5966                        ew32(CTRL_EXT, ctrl_ext);
5967                }
5968
5969                if (adapter->flags & FLAG_IS_ICH)
5970                        e1000_suspend_workarounds_ich8lan(&adapter->hw);
5971
5972                /* Allow time for pending master requests to run */
5973                e1000e_disable_pcie_master(&adapter->hw);
5974
5975                if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
5976                        /* enable wakeup by the PHY */
5977                        retval = e1000_init_phy_wakeup(adapter, wufc);
5978                        if (retval)
5979                                return retval;
5980                } else {
5981                        /* enable wakeup by the MAC */
5982                        ew32(WUFC, wufc);
5983                        ew32(WUC, E1000_WUC_PME_EN);
5984                }
5985        } else {
5986                ew32(WUC, 0);
5987                ew32(WUFC, 0);
5988        }
5989
5990        if (adapter->hw.phy.type == e1000_phy_igp_3)
5991                e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
5992
5993        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
5994         * would have already happened in close and is redundant.
5995         */
5996        e1000e_release_hw_control(adapter);
5997
5998        /* The pci-e switch on some quad port adapters will report a
5999         * correctable error when the MAC transitions from D0 to D3.  To
6000         * prevent this we need to mask off the correctable errors on the
6001         * downstream port of the pci-e switch.
6002         */
6003        if (adapter->flags & FLAG_IS_QUAD_PORT) {
6004                struct pci_dev *us_dev = pdev->bus->self;
6005                u16 devctl;
6006
6007                pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6008                pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6009                                           (devctl & ~PCI_EXP_DEVCTL_CERE));
6010
6011                pci_save_state(pdev);
6012                pci_prepare_to_sleep(pdev);
6013
6014                pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6015        }
6016
6017        return 0;
6018}
6019
6020#ifdef CONFIG_PCIEASPM
6021static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6022{
6023        pci_disable_link_state_locked(pdev, state);
6024}
6025#else
6026static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6027{
6028        u16 aspm_ctl = 0;
6029
6030        if (state & PCIE_LINK_STATE_L0S)
6031                aspm_ctl |= PCI_EXP_LNKCTL_ASPM_L0S;
6032        if (state & PCIE_LINK_STATE_L1)
6033                aspm_ctl |= PCI_EXP_LNKCTL_ASPM_L1;
6034
6035        /* Both device and parent should have the same ASPM setting.
6036         * Disable ASPM in downstream component first and then upstream.
6037         */
6038        pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_ctl);
6039
6040        if (pdev->bus->self)
6041                pcie_capability_clear_word(pdev->bus->self, PCI_EXP_LNKCTL,
6042                                           aspm_ctl);
6043}
6044#endif
6045static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6046{
6047        dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6048                 (state & PCIE_LINK_STATE_L0S) ? "L0s" : "",
6049                 (state & PCIE_LINK_STATE_L1) ? "L1" : "");
6050
6051        __e1000e_disable_aspm(pdev, state);
6052}
6053
6054#ifdef CONFIG_PM
6055static bool e1000e_pm_ready(struct e1000_adapter *adapter)
6056{
6057        return !!adapter->tx_ring->buffer_info;
6058}
6059
6060static int __e1000_resume(struct pci_dev *pdev)
6061{
6062        struct net_device *netdev = pci_get_drvdata(pdev);
6063        struct e1000_adapter *adapter = netdev_priv(netdev);
6064        struct e1000_hw *hw = &adapter->hw;
6065        u16 aspm_disable_flag = 0;
6066        u32 err;
6067
6068        if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6069                aspm_disable_flag = PCIE_LINK_STATE_L0S;
6070        if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6071                aspm_disable_flag |= PCIE_LINK_STATE_L1;
6072        if (aspm_disable_flag)
6073                e1000e_disable_aspm(pdev, aspm_disable_flag);
6074
6075        pci_set_master(pdev);
6076
6077        e1000e_set_interrupt_capability(adapter);
6078        if (netif_running(netdev)) {
6079                err = e1000_request_irq(adapter);
6080                if (err)
6081                        return err;
6082        }
6083
6084        if (hw->mac.type >= e1000_pch2lan)
6085                e1000_resume_workarounds_pchlan(&adapter->hw);
6086
6087        e1000e_power_up_phy(adapter);
6088
6089        /* report the system wakeup cause from S3/S4 */
6090        if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6091                u16 phy_data;
6092
6093                e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6094                if (phy_data) {
6095                        e_info("PHY Wakeup cause - %s\n",
6096                               phy_data & E1000_WUS_EX ? "Unicast Packet" :
6097                               phy_data & E1000_WUS_MC ? "Multicast Packet" :
6098                               phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6099                               phy_data & E1000_WUS_MAG ? "Magic Packet" :
6100                               phy_data & E1000_WUS_LNKC ?
6101                               "Link Status Change" : "other");
6102                }
6103                e1e_wphy(&adapter->hw, BM_WUS, ~0);
6104        } else {
6105                u32 wus = er32(WUS);
6106                if (wus) {
6107                        e_info("MAC Wakeup cause - %s\n",
6108                               wus & E1000_WUS_EX ? "Unicast Packet" :
6109                               wus & E1000_WUS_MC ? "Multicast Packet" :
6110                               wus & E1000_WUS_BC ? "Broadcast Packet" :
6111                               wus & E1000_WUS_MAG ? "Magic Packet" :
6112                               wus & E1000_WUS_LNKC ? "Link Status Change" :
6113                               "other");
6114                }
6115                ew32(WUS, ~0);
6116        }
6117
6118        e1000e_reset(adapter);
6119
6120        e1000_init_manageability_pt(adapter);
6121
6122        if (netif_running(netdev))
6123                e1000e_up(adapter);
6124
6125        netif_device_attach(netdev);
6126
6127        /* If the controller has AMT, do not set DRV_LOAD until the interface
6128         * is up.  For all other cases, let the f/w know that the h/w is now
6129         * under the control of the driver.
6130         */
6131        if (!(adapter->flags & FLAG_HAS_AMT))
6132                e1000e_get_hw_control(adapter);
6133
6134        return 0;
6135}
6136
6137#ifdef CONFIG_PM_SLEEP
6138static int e1000_suspend(struct device *dev)
6139{
6140        struct pci_dev *pdev = to_pci_dev(dev);
6141
6142        return __e1000_shutdown(pdev, false);
6143}
6144
6145static int e1000_resume(struct device *dev)
6146{
6147        struct pci_dev *pdev = to_pci_dev(dev);
6148        struct net_device *netdev = pci_get_drvdata(pdev);
6149        struct e1000_adapter *adapter = netdev_priv(netdev);
6150
6151        if (e1000e_pm_ready(adapter))
6152                adapter->idle_check = true;
6153
6154        return __e1000_resume(pdev);
6155}
6156#endif /* CONFIG_PM_SLEEP */
6157
6158#ifdef CONFIG_PM_RUNTIME
6159static int e1000_runtime_suspend(struct device *dev)
6160{
6161        struct pci_dev *pdev = to_pci_dev(dev);
6162        struct net_device *netdev = pci_get_drvdata(pdev);
6163        struct e1000_adapter *adapter = netdev_priv(netdev);
6164
6165        if (!e1000e_pm_ready(adapter))
6166                return 0;
6167
6168        return __e1000_shutdown(pdev, true);
6169}
6170
6171static int e1000_idle(struct device *dev)
6172{
6173        struct pci_dev *pdev = to_pci_dev(dev);
6174        struct net_device *netdev = pci_get_drvdata(pdev);
6175        struct e1000_adapter *adapter = netdev_priv(netdev);
6176
6177        if (!e1000e_pm_ready(adapter))
6178                return 0;
6179
6180        if (adapter->idle_check) {
6181                adapter->idle_check = false;
6182                if (!e1000e_has_link(adapter))
6183                        pm_schedule_suspend(dev, MSEC_PER_SEC);
6184        }
6185
6186        return -EBUSY;
6187}
6188
6189static int e1000_runtime_resume(struct device *dev)
6190{
6191        struct pci_dev *pdev = to_pci_dev(dev);
6192        struct net_device *netdev = pci_get_drvdata(pdev);
6193        struct e1000_adapter *adapter = netdev_priv(netdev);
6194
6195        if (!e1000e_pm_ready(adapter))
6196                return 0;
6197
6198        adapter->idle_check = !dev->power.runtime_auto;
6199        return __e1000_resume(pdev);
6200}
6201#endif /* CONFIG_PM_RUNTIME */
6202#endif /* CONFIG_PM */
6203
6204static void e1000_shutdown(struct pci_dev *pdev)
6205{
6206        __e1000_shutdown(pdev, false);
6207}
6208
6209#ifdef CONFIG_NET_POLL_CONTROLLER
6210
6211static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
6212{
6213        struct net_device *netdev = data;
6214        struct e1000_adapter *adapter = netdev_priv(netdev);
6215
6216        if (adapter->msix_entries) {
6217                int vector, msix_irq;
6218
6219                vector = 0;
6220                msix_irq = adapter->msix_entries[vector].vector;
6221                disable_irq(msix_irq);
6222                e1000_intr_msix_rx(msix_irq, netdev);
6223                enable_irq(msix_irq);
6224
6225                vector++;
6226                msix_irq = adapter->msix_entries[vector].vector;
6227                disable_irq(msix_irq);
6228                e1000_intr_msix_tx(msix_irq, netdev);
6229                enable_irq(msix_irq);
6230
6231                vector++;
6232                msix_irq = adapter->msix_entries[vector].vector;
6233                disable_irq(msix_irq);
6234                e1000_msix_other(msix_irq, netdev);
6235                enable_irq(msix_irq);
6236        }
6237
6238        return IRQ_HANDLED;
6239}
6240
6241/**
6242 * e1000_netpoll
6243 * @netdev: network interface device structure
6244 *
6245 * Polling 'interrupt' - used by things like netconsole to send skbs
6246 * without having to re-enable interrupts. It's not called while
6247 * the interrupt routine is executing.
6248 */
6249static void e1000_netpoll(struct net_device *netdev)
6250{
6251        struct e1000_adapter *adapter = netdev_priv(netdev);
6252
6253        switch (adapter->int_mode) {
6254        case E1000E_INT_MODE_MSIX:
6255                e1000_intr_msix(adapter->pdev->irq, netdev);
6256                break;
6257        case E1000E_INT_MODE_MSI:
6258                disable_irq(adapter->pdev->irq);
6259                e1000_intr_msi(adapter->pdev->irq, netdev);
6260                enable_irq(adapter->pdev->irq);
6261                break;
6262        default:                /* E1000E_INT_MODE_LEGACY */
6263                disable_irq(adapter->pdev->irq);
6264                e1000_intr(adapter->pdev->irq, netdev);
6265                enable_irq(adapter->pdev->irq);
6266                break;
6267        }
6268}
6269#endif
6270
6271/**
6272 * e1000_io_error_detected - called when PCI error is detected
6273 * @pdev: Pointer to PCI device
6274 * @state: The current pci connection state
6275 *
6276 * This function is called after a PCI bus error affecting
6277 * this device has been detected.
6278 */
6279static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
6280                                                pci_channel_state_t state)
6281{
6282        struct net_device *netdev = pci_get_drvdata(pdev);
6283        struct e1000_adapter *adapter = netdev_priv(netdev);
6284
6285        netif_device_detach(netdev);
6286
6287        if (state == pci_channel_io_perm_failure)
6288                return PCI_ERS_RESULT_DISCONNECT;
6289
6290        if (netif_running(netdev))
6291                e1000e_down(adapter);
6292        pci_disable_device(pdev);
6293
6294        /* Request a slot slot reset. */
6295        return PCI_ERS_RESULT_NEED_RESET;
6296}
6297
6298/**
6299 * e1000_io_slot_reset - called after the pci bus has been reset.
6300 * @pdev: Pointer to PCI device
6301 *
6302 * Restart the card from scratch, as if from a cold-boot. Implementation
6303 * resembles the first-half of the e1000_resume routine.
6304 */
6305static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
6306{
6307        struct net_device *netdev = pci_get_drvdata(pdev);
6308        struct e1000_adapter *adapter = netdev_priv(netdev);
6309        struct e1000_hw *hw = &adapter->hw;
6310        u16 aspm_disable_flag = 0;
6311        int err;
6312        pci_ers_result_t result;
6313
6314        if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6315                aspm_disable_flag = PCIE_LINK_STATE_L0S;
6316        if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6317                aspm_disable_flag |= PCIE_LINK_STATE_L1;
6318        if (aspm_disable_flag)
6319                e1000e_disable_aspm(pdev, aspm_disable_flag);
6320
6321        err = pci_enable_device_mem(pdev);
6322        if (err) {
6323                dev_err(&pdev->dev,
6324                        "Cannot re-enable PCI device after reset.\n");
6325                result = PCI_ERS_RESULT_DISCONNECT;
6326        } else {
6327                pdev->state_saved = true;
6328                pci_restore_state(pdev);
6329                pci_set_master(pdev);
6330
6331                pci_enable_wake(pdev, PCI_D3hot, 0);
6332                pci_enable_wake(pdev, PCI_D3cold, 0);
6333
6334                e1000e_reset(adapter);
6335                ew32(WUS, ~0);
6336                result = PCI_ERS_RESULT_RECOVERED;
6337        }
6338
6339        pci_cleanup_aer_uncorrect_error_status(pdev);
6340
6341        return result;
6342}
6343
6344/**
6345 * e1000_io_resume - called when traffic can start flowing again.
6346 * @pdev: Pointer to PCI device
6347 *
6348 * This callback is called when the error recovery driver tells us that
6349 * its OK to resume normal operation. Implementation resembles the
6350 * second-half of the e1000_resume routine.
6351 */
6352static void e1000_io_resume(struct pci_dev *pdev)
6353{
6354        struct net_device *netdev = pci_get_drvdata(pdev);
6355        struct e1000_adapter *adapter = netdev_priv(netdev);
6356
6357        e1000_init_manageability_pt(adapter);
6358
6359        if (netif_running(netdev)) {
6360                if (e1000e_up(adapter)) {
6361                        dev_err(&pdev->dev,
6362                                "can't bring device back up after reset\n");
6363                        return;
6364                }
6365        }
6366
6367        netif_device_attach(netdev);
6368
6369        /* If the controller has AMT, do not set DRV_LOAD until the interface
6370         * is up.  For all other cases, let the f/w know that the h/w is now
6371         * under the control of the driver.
6372         */
6373        if (!(adapter->flags & FLAG_HAS_AMT))
6374                e1000e_get_hw_control(adapter);
6375}
6376
6377static void e1000_print_device_info(struct e1000_adapter *adapter)
6378{
6379        struct e1000_hw *hw = &adapter->hw;
6380        struct net_device *netdev = adapter->netdev;
6381        u32 ret_val;
6382        u8 pba_str[E1000_PBANUM_LENGTH];
6383
6384        /* print bus type/speed/width info */
6385        e_info("(PCI Express:2.5GT/s:%s) %pM\n",
6386               /* bus width */
6387               ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
6388                "Width x1"),
6389               /* MAC address */
6390               netdev->dev_addr);
6391        e_info("Intel(R) PRO/%s Network Connection\n",
6392               (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
6393        ret_val = e1000_read_pba_string_generic(hw, pba_str,
6394                                                E1000_PBANUM_LENGTH);
6395        if (ret_val)
6396                strlcpy((char *)pba_str, "Unknown", sizeof(pba_str));
6397        e_info("MAC: %d, PHY: %d, PBA No: %s\n",
6398               hw->mac.type, hw->phy.type, pba_str);
6399}
6400
6401static void e1000_eeprom_checks(struct e1000_adapter *adapter)
6402{
6403        struct e1000_hw *hw = &adapter->hw;
6404        int ret_val;
6405        u16 buf = 0;
6406
6407        if (hw->mac.type != e1000_82573)
6408                return;
6409
6410        ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
6411        le16_to_cpus(&buf);
6412        if (!ret_val && (!(buf & (1 << 0)))) {
6413                /* Deep Smart Power Down (DSPD) */
6414                dev_warn(&adapter->pdev->dev,
6415                         "Warning: detected DSPD enabled in EEPROM\n");
6416        }
6417}
6418
6419static int e1000_set_features(struct net_device *netdev,
6420                              netdev_features_t features)
6421{
6422        struct e1000_adapter *adapter = netdev_priv(netdev);
6423        netdev_features_t changed = features ^ netdev->features;
6424
6425        if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
6426                adapter->flags |= FLAG_TSO_FORCE;
6427
6428        if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
6429                         NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
6430                         NETIF_F_RXALL)))
6431                return 0;
6432
6433        if (changed & NETIF_F_RXFCS) {
6434                if (features & NETIF_F_RXFCS) {
6435                        adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6436                } else {
6437                        /* We need to take it back to defaults, which might mean
6438                         * stripping is still disabled at the adapter level.
6439                         */
6440                        if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
6441                                adapter->flags2 |= FLAG2_CRC_STRIPPING;
6442                        else
6443                                adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
6444                }
6445        }
6446
6447        netdev->features = features;
6448
6449        if (netif_running(netdev))
6450                e1000e_reinit_locked(adapter);
6451        else
6452                e1000e_reset(adapter);
6453
6454        return 0;
6455}
6456
6457static const struct net_device_ops e1000e_netdev_ops = {
6458        .ndo_open               = e1000_open,
6459        .ndo_stop               = e1000_close,
6460        .ndo_start_xmit         = e1000_xmit_frame,
6461        .ndo_get_stats64        = e1000e_get_stats64,
6462        .ndo_set_rx_mode        = e1000e_set_rx_mode,
6463        .ndo_set_mac_address    = e1000_set_mac,
6464        .ndo_change_mtu         = e1000_change_mtu,
6465        .ndo_do_ioctl           = e1000_ioctl,
6466        .ndo_tx_timeout         = e1000_tx_timeout,
6467        .ndo_validate_addr      = eth_validate_addr,
6468
6469        .ndo_vlan_rx_add_vid    = e1000_vlan_rx_add_vid,
6470        .ndo_vlan_rx_kill_vid   = e1000_vlan_rx_kill_vid,
6471#ifdef CONFIG_NET_POLL_CONTROLLER
6472        .ndo_poll_controller    = e1000_netpoll,
6473#endif
6474        .ndo_set_features = e1000_set_features,
6475};
6476
6477/**
6478 * e1000_probe - Device Initialization Routine
6479 * @pdev: PCI device information struct
6480 * @ent: entry in e1000_pci_tbl
6481 *
6482 * Returns 0 on success, negative on failure
6483 *
6484 * e1000_probe initializes an adapter identified by a pci_dev structure.
6485 * The OS initialization, configuring of the adapter private structure,
6486 * and a hardware reset occur.
6487 **/
6488static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
6489{
6490        struct net_device *netdev;
6491        struct e1000_adapter *adapter;
6492        struct e1000_hw *hw;
6493        const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
6494        resource_size_t mmio_start, mmio_len;
6495        resource_size_t flash_start, flash_len;
6496        static int cards_found;
6497        u16 aspm_disable_flag = 0;
6498        int bars, i, err, pci_using_dac;
6499        u16 eeprom_data = 0;
6500        u16 eeprom_apme_mask = E1000_EEPROM_APME;
6501
6502        if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
6503                aspm_disable_flag = PCIE_LINK_STATE_L0S;
6504        if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
6505                aspm_disable_flag |= PCIE_LINK_STATE_L1;
6506        if (aspm_disable_flag)
6507                e1000e_disable_aspm(pdev, aspm_disable_flag);
6508
6509        err = pci_enable_device_mem(pdev);
6510        if (err)
6511                return err;
6512
6513        pci_using_dac = 0;
6514        err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(64));
6515        if (!err) {
6516                err = dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
6517                if (!err)
6518                        pci_using_dac = 1;
6519        } else {
6520                err = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
6521                if (err) {
6522                        err = dma_set_coherent_mask(&pdev->dev,
6523                                                    DMA_BIT_MASK(32));
6524                        if (err) {
6525                                dev_err(&pdev->dev,
6526                                        "No usable DMA configuration, aborting\n");
6527                                goto err_dma;
6528                        }
6529                }
6530        }
6531
6532        bars = pci_select_bars(pdev, IORESOURCE_MEM);
6533        err = pci_request_selected_regions_exclusive(pdev, bars,
6534                                                     e1000e_driver_name);
6535        if (err)
6536                goto err_pci_reg;
6537
6538        /* AER (Advanced Error Reporting) hooks */
6539        pci_enable_pcie_error_reporting(pdev);
6540
6541        pci_set_master(pdev);
6542        /* PCI config space info */
6543        err = pci_save_state(pdev);
6544        if (err)
6545                goto err_alloc_etherdev;
6546
6547        err = -ENOMEM;
6548        netdev = alloc_etherdev(sizeof(struct e1000_adapter));
6549        if (!netdev)
6550                goto err_alloc_etherdev;
6551
6552        SET_NETDEV_DEV(netdev, &pdev->dev);
6553
6554        netdev->irq = pdev->irq;
6555
6556        pci_set_drvdata(pdev, netdev);
6557        adapter = netdev_priv(netdev);
6558        hw = &adapter->hw;
6559        adapter->netdev = netdev;
6560        adapter->pdev = pdev;
6561        adapter->ei = ei;
6562        adapter->pba = ei->pba;
6563        adapter->flags = ei->flags;
6564        adapter->flags2 = ei->flags2;
6565        adapter->hw.adapter = adapter;
6566        adapter->hw.mac.type = ei->mac;
6567        adapter->max_hw_frame_size = ei->max_hw_frame_size;
6568        adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
6569
6570        mmio_start = pci_resource_start(pdev, 0);
6571        mmio_len = pci_resource_len(pdev, 0);
6572
6573        err = -EIO;
6574        adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
6575        if (!adapter->hw.hw_addr)
6576                goto err_ioremap;
6577
6578        if ((adapter->flags & FLAG_HAS_FLASH) &&
6579            (pci_resource_flags(pdev, 1) & IORESOURCE_MEM)) {
6580                flash_start = pci_resource_start(pdev, 1);
6581                flash_len = pci_resource_len(pdev, 1);
6582                adapter->hw.flash_address = ioremap(flash_start, flash_len);
6583                if (!adapter->hw.flash_address)
6584                        goto err_flashmap;
6585        }
6586
6587        /* Set default EEE advertisement */
6588        if (adapter->flags2 & FLAG2_HAS_EEE)
6589                adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
6590
6591        /* construct the net_device struct */
6592        netdev->netdev_ops = &e1000e_netdev_ops;
6593        e1000e_set_ethtool_ops(netdev);
6594        netdev->watchdog_timeo = 5 * HZ;
6595        netif_napi_add(netdev, &adapter->napi, e1000e_poll, 64);
6596        strlcpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
6597
6598        netdev->mem_start = mmio_start;
6599        netdev->mem_end = mmio_start + mmio_len;
6600
6601        adapter->bd_number = cards_found++;
6602
6603        e1000e_check_options(adapter);
6604
6605        /* setup adapter struct */
6606        err = e1000_sw_init(adapter);
6607        if (err)
6608                goto err_sw_init;
6609
6610        memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
6611        memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
6612        memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
6613
6614        err = ei->get_variants(adapter);
6615        if (err)
6616                goto err_hw_init;
6617
6618        if ((adapter->flags & FLAG_IS_ICH) &&
6619            (adapter->flags & FLAG_READ_ONLY_NVM))
6620                e1000e_write_protect_nvm_ich8lan(&adapter->hw);
6621
6622        hw->mac.ops.get_bus_info(&adapter->hw);
6623
6624        adapter->hw.phy.autoneg_wait_to_complete = 0;
6625
6626        /* Copper options */
6627        if (adapter->hw.phy.media_type == e1000_media_type_copper) {
6628                adapter->hw.phy.mdix = AUTO_ALL_MODES;
6629                adapter->hw.phy.disable_polarity_correction = 0;
6630                adapter->hw.phy.ms_type = e1000_ms_hw_default;
6631        }
6632
6633        if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
6634                dev_info(&pdev->dev,
6635                         "PHY reset is blocked due to SOL/IDER session.\n");
6636
6637        /* Set initial default active device features */
6638        netdev->features = (NETIF_F_SG |
6639                            NETIF_F_HW_VLAN_CTAG_RX |
6640                            NETIF_F_HW_VLAN_CTAG_TX |
6641                            NETIF_F_TSO |
6642                            NETIF_F_TSO6 |
6643                            NETIF_F_RXHASH |
6644                            NETIF_F_RXCSUM |
6645                            NETIF_F_HW_CSUM);
6646
6647        /* Set user-changeable features (subset of all device features) */
6648        netdev->hw_features = netdev->features;
6649        netdev->hw_features |= NETIF_F_RXFCS;
6650        netdev->priv_flags |= IFF_SUPP_NOFCS;
6651        netdev->hw_features |= NETIF_F_RXALL;
6652
6653        if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
6654                netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6655
6656        netdev->vlan_features |= (NETIF_F_SG |
6657                                  NETIF_F_TSO |
6658                                  NETIF_F_TSO6 |
6659                                  NETIF_F_HW_CSUM);
6660
6661        netdev->priv_flags |= IFF_UNICAST_FLT;
6662
6663        if (pci_using_dac) {
6664                netdev->features |= NETIF_F_HIGHDMA;
6665                netdev->vlan_features |= NETIF_F_HIGHDMA;
6666        }
6667
6668        if (e1000e_enable_mng_pass_thru(&adapter->hw))
6669                adapter->flags |= FLAG_MNG_PT_ENABLED;
6670
6671        /* before reading the NVM, reset the controller to
6672         * put the device in a known good starting state
6673         */
6674        adapter->hw.mac.ops.reset_hw(&adapter->hw);
6675
6676        /* systems with ASPM and others may see the checksum fail on the first
6677         * attempt. Let's give it a few tries
6678         */
6679        for (i = 0;; i++) {
6680                if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
6681                        break;
6682                if (i == 2) {
6683                        dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
6684                        err = -EIO;
6685                        goto err_eeprom;
6686                }
6687        }
6688
6689        e1000_eeprom_checks(adapter);
6690
6691        /* copy the MAC address */
6692        if (e1000e_read_mac_addr(&adapter->hw))
6693                dev_err(&pdev->dev,
6694                        "NVM Read Error while reading MAC address\n");
6695
6696        memcpy(netdev->dev_addr, adapter->hw.mac.addr, netdev->addr_len);
6697
6698        if (!is_valid_ether_addr(netdev->dev_addr)) {
6699                dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
6700                        netdev->dev_addr);
6701                err = -EIO;
6702                goto err_eeprom;
6703        }
6704
6705        init_timer(&adapter->watchdog_timer);
6706        adapter->watchdog_timer.function = e1000_watchdog;
6707        adapter->watchdog_timer.data = (unsigned long)adapter;
6708
6709        init_timer(&adapter->phy_info_timer);
6710        adapter->phy_info_timer.function = e1000_update_phy_info;
6711        adapter->phy_info_timer.data = (unsigned long)adapter;
6712
6713        INIT_WORK(&adapter->reset_task, e1000_reset_task);
6714        INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
6715        INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
6716        INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
6717        INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
6718
6719        /* Initialize link parameters. User can change them with ethtool */
6720        adapter->hw.mac.autoneg = 1;
6721        adapter->fc_autoneg = true;
6722        adapter->hw.fc.requested_mode = e1000_fc_default;
6723        adapter->hw.fc.current_mode = e1000_fc_default;
6724        adapter->hw.phy.autoneg_advertised = 0x2f;
6725
6726        /* ring size defaults */
6727        adapter->rx_ring->count = E1000_DEFAULT_RXD;
6728        adapter->tx_ring->count = E1000_DEFAULT_TXD;
6729
6730        /* Initial Wake on LAN setting - If APM wake is enabled in
6731         * the EEPROM, enable the ACPI Magic Packet filter
6732         */
6733        if (adapter->flags & FLAG_APME_IN_WUC) {
6734                /* APME bit in EEPROM is mapped to WUC.APME */
6735                eeprom_data = er32(WUC);
6736                eeprom_apme_mask = E1000_WUC_APME;
6737                if ((hw->mac.type > e1000_ich10lan) &&
6738                    (eeprom_data & E1000_WUC_PHY_WAKE))
6739                        adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
6740        } else if (adapter->flags & FLAG_APME_IN_CTRL3) {
6741                if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
6742                    (adapter->hw.bus.func == 1))
6743                        e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_B,
6744                                       1, &eeprom_data);
6745                else
6746                        e1000_read_nvm(&adapter->hw, NVM_INIT_CONTROL3_PORT_A,
6747                                       1, &eeprom_data);
6748        }
6749
6750        /* fetch WoL from EEPROM */
6751        if (eeprom_data & eeprom_apme_mask)
6752                adapter->eeprom_wol |= E1000_WUFC_MAG;
6753
6754        /* now that we have the eeprom settings, apply the special cases
6755         * where the eeprom may be wrong or the board simply won't support
6756         * wake on lan on a particular port
6757         */
6758        if (!(adapter->flags & FLAG_HAS_WOL))
6759                adapter->eeprom_wol = 0;
6760
6761        /* initialize the wol settings based on the eeprom settings */
6762        adapter->wol = adapter->eeprom_wol;
6763
6764        /* make sure adapter isn't asleep if manageability is enabled */
6765        if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
6766            (hw->mac.ops.check_mng_mode(hw)))
6767                device_wakeup_enable(&pdev->dev);
6768
6769        /* save off EEPROM version number */
6770        e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
6771
6772        /* reset the hardware with the new settings */
6773        e1000e_reset(adapter);
6774
6775        /* If the controller has AMT, do not set DRV_LOAD until the interface
6776         * is up.  For all other cases, let the f/w know that the h/w is now
6777         * under the control of the driver.
6778         */
6779        if (!(adapter->flags & FLAG_HAS_AMT))
6780                e1000e_get_hw_control(adapter);
6781
6782        strlcpy(netdev->name, "eth%d", sizeof(netdev->name));
6783        err = register_netdev(netdev);
6784        if (err)
6785                goto err_register;
6786
6787        /* carrier off reporting is important to ethtool even BEFORE open */
6788        netif_carrier_off(netdev);
6789
6790        /* init PTP hardware clock */
6791        e1000e_ptp_init(adapter);
6792
6793        e1000_print_device_info(adapter);
6794
6795        if (pci_dev_run_wake(pdev))
6796                pm_runtime_put_noidle(&pdev->dev);
6797
6798        return 0;
6799
6800err_register:
6801        if (!(adapter->flags & FLAG_HAS_AMT))
6802                e1000e_release_hw_control(adapter);
6803err_eeprom:
6804        if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
6805                e1000_phy_hw_reset(&adapter->hw);
6806err_hw_init:
6807        kfree(adapter->tx_ring);
6808        kfree(adapter->rx_ring);
6809err_sw_init:
6810        if (adapter->hw.flash_address)
6811                iounmap(adapter->hw.flash_address);
6812        e1000e_reset_interrupt_capability(adapter);
6813err_flashmap:
6814        iounmap(adapter->hw.hw_addr);
6815err_ioremap:
6816        free_netdev(netdev);
6817err_alloc_etherdev:
6818        pci_release_selected_regions(pdev,
6819                                     pci_select_bars(pdev, IORESOURCE_MEM));
6820err_pci_reg:
6821err_dma:
6822        pci_disable_device(pdev);
6823        return err;
6824}
6825
6826/**
6827 * e1000_remove - Device Removal Routine
6828 * @pdev: PCI device information struct
6829 *
6830 * e1000_remove is called by the PCI subsystem to alert the driver
6831 * that it should release a PCI device.  The could be caused by a
6832 * Hot-Plug event, or because the driver is going to be removed from
6833 * memory.
6834 **/
6835static void e1000_remove(struct pci_dev *pdev)
6836{
6837        struct net_device *netdev = pci_get_drvdata(pdev);
6838        struct e1000_adapter *adapter = netdev_priv(netdev);
6839        bool down = test_bit(__E1000_DOWN, &adapter->state);
6840
6841        e1000e_ptp_remove(adapter);
6842
6843        /* The timers may be rescheduled, so explicitly disable them
6844         * from being rescheduled.
6845         */
6846        if (!down)
6847                set_bit(__E1000_DOWN, &adapter->state);
6848        del_timer_sync(&adapter->watchdog_timer);
6849        del_timer_sync(&adapter->phy_info_timer);
6850
6851        cancel_work_sync(&adapter->reset_task);
6852        cancel_work_sync(&adapter->watchdog_task);
6853        cancel_work_sync(&adapter->downshift_task);
6854        cancel_work_sync(&adapter->update_phy_task);
6855        cancel_work_sync(&adapter->print_hang_task);
6856
6857        if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
6858                cancel_work_sync(&adapter->tx_hwtstamp_work);
6859                if (adapter->tx_hwtstamp_skb) {
6860                        dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
6861                        adapter->tx_hwtstamp_skb = NULL;
6862                }
6863        }
6864
6865        if (!(netdev->flags & IFF_UP))
6866                e1000_power_down_phy(adapter);
6867
6868        /* Don't lie to e1000_close() down the road. */
6869        if (!down)
6870                clear_bit(__E1000_DOWN, &adapter->state);
6871        unregister_netdev(netdev);
6872
6873        if (pci_dev_run_wake(pdev))
6874                pm_runtime_get_noresume(&pdev->dev);
6875
6876        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
6877         * would have already happened in close and is redundant.
6878         */
6879        e1000e_release_hw_control(adapter);
6880
6881        e1000e_reset_interrupt_capability(adapter);
6882        kfree(adapter->tx_ring);
6883        kfree(adapter->rx_ring);
6884
6885        iounmap(adapter->hw.hw_addr);
6886        if (adapter->hw.flash_address)
6887                iounmap(adapter->hw.flash_address);
6888        pci_release_selected_regions(pdev,
6889                                     pci_select_bars(pdev, IORESOURCE_MEM));
6890
6891        free_netdev(netdev);
6892
6893        /* AER disable */
6894        pci_disable_pcie_error_reporting(pdev);
6895
6896        pci_disable_device(pdev);
6897}
6898
6899/* PCI Error Recovery (ERS) */
6900static const struct pci_error_handlers e1000_err_handler = {
6901        .error_detected = e1000_io_error_detected,
6902        .slot_reset = e1000_io_slot_reset,
6903        .resume = e1000_io_resume,
6904};
6905
6906static DEFINE_PCI_DEVICE_TABLE(e1000_pci_tbl) = {
6907        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
6908        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
6909        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
6910        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
6911          board_82571 },
6912        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
6913        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
6914        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
6915        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
6916        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
6917
6918        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
6919        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
6920        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
6921        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
6922
6923        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
6924        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
6925        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
6926
6927        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
6928        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
6929        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
6930
6931        { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
6932          board_80003es2lan },
6933        { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
6934          board_80003es2lan },
6935        { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
6936          board_80003es2lan },
6937        { PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
6938          board_80003es2lan },
6939
6940        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
6941        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
6942        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
6943        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
6944        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
6945        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
6946        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
6947        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
6948
6949        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
6950        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
6951        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
6952        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
6953        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
6954        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
6955        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
6956        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
6957        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
6958
6959        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
6960        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
6961        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
6962
6963        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
6964        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
6965        { PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
6966
6967        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
6968        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
6969        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
6970        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
6971
6972        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
6973        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
6974
6975        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
6976        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
6977        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
6978        { PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
6979
6980        { 0, 0, 0, 0, 0, 0, 0 } /* terminate list */
6981};
6982MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
6983
6984#ifdef CONFIG_PM
6985static const struct dev_pm_ops e1000_pm_ops = {
6986        SET_SYSTEM_SLEEP_PM_OPS(e1000_suspend, e1000_resume)
6987        SET_RUNTIME_PM_OPS(e1000_runtime_suspend, e1000_runtime_resume,
6988                           e1000_idle)
6989};
6990#endif
6991
6992/* PCI Device API Driver */
6993static struct pci_driver e1000_driver = {
6994        .name     = e1000e_driver_name,
6995        .id_table = e1000_pci_tbl,
6996        .probe    = e1000_probe,
6997        .remove   = e1000_remove,
6998#ifdef CONFIG_PM
6999        .driver   = {
7000                .pm = &e1000_pm_ops,
7001        },
7002#endif
7003        .shutdown = e1000_shutdown,
7004        .err_handler = &e1000_err_handler
7005};
7006
7007/**
7008 * e1000_init_module - Driver Registration Routine
7009 *
7010 * e1000_init_module is the first routine called when the driver is
7011 * loaded. All it does is register with the PCI subsystem.
7012 **/
7013static int __init e1000_init_module(void)
7014{
7015        int ret;
7016        pr_info("Intel(R) PRO/1000 Network Driver - %s\n",
7017                e1000e_driver_version);
7018        pr_info("Copyright(c) 1999 - 2013 Intel Corporation.\n");
7019        ret = pci_register_driver(&e1000_driver);
7020
7021        return ret;
7022}
7023module_init(e1000_init_module);
7024
7025/**
7026 * e1000_exit_module - Driver Exit Cleanup Routine
7027 *
7028 * e1000_exit_module is called just before the driver is removed
7029 * from memory.
7030 **/
7031static void __exit e1000_exit_module(void)
7032{
7033        pci_unregister_driver(&e1000_driver);
7034}
7035module_exit(e1000_exit_module);
7036
7037MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
7038MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7039MODULE_LICENSE("GPL");
7040MODULE_VERSION(DRV_VERSION);
7041
7042/* netdev.c */
7043