linux/drivers/net/ethernet/intel/igb/igb_main.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/* Copyright(c) 2007 - 2018 Intel Corporation. */
   3
   4#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   5
   6#include <linux/module.h>
   7#include <linux/types.h>
   8#include <linux/init.h>
   9#include <linux/bitops.h>
  10#include <linux/vmalloc.h>
  11#include <linux/pagemap.h>
  12#include <linux/netdevice.h>
  13#include <linux/ipv6.h>
  14#include <linux/slab.h>
  15#include <net/checksum.h>
  16#include <net/ip6_checksum.h>
  17#include <net/pkt_sched.h>
  18#include <net/pkt_cls.h>
  19#include <linux/net_tstamp.h>
  20#include <linux/mii.h>
  21#include <linux/ethtool.h>
  22#include <linux/if.h>
  23#include <linux/if_vlan.h>
  24#include <linux/pci.h>
  25#include <linux/delay.h>
  26#include <linux/interrupt.h>
  27#include <linux/ip.h>
  28#include <linux/tcp.h>
  29#include <linux/sctp.h>
  30#include <linux/if_ether.h>
  31#include <linux/aer.h>
  32#include <linux/prefetch.h>
  33#include <linux/pm_runtime.h>
  34#include <linux/etherdevice.h>
  35#ifdef CONFIG_IGB_DCA
  36#include <linux/dca.h>
  37#endif
  38#include <linux/i2c.h>
  39#include "igb.h"
  40
  41#define MAJ 5
  42#define MIN 6
  43#define BUILD 0
  44#define DRV_VERSION __stringify(MAJ) "." __stringify(MIN) "." \
  45__stringify(BUILD) "-k"
  46
  47enum queue_mode {
  48        QUEUE_MODE_STRICT_PRIORITY,
  49        QUEUE_MODE_STREAM_RESERVATION,
  50};
  51
  52enum tx_queue_prio {
  53        TX_QUEUE_PRIO_HIGH,
  54        TX_QUEUE_PRIO_LOW,
  55};
  56
  57char igb_driver_name[] = "igb";
  58char igb_driver_version[] = DRV_VERSION;
  59static const char igb_driver_string[] =
  60                                "Intel(R) Gigabit Ethernet Network Driver";
  61static const char igb_copyright[] =
  62                                "Copyright (c) 2007-2014 Intel Corporation.";
  63
  64static const struct e1000_info *igb_info_tbl[] = {
  65        [board_82575] = &e1000_82575_info,
  66};
  67
  68static const struct pci_device_id igb_pci_tbl[] = {
  69        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
  70        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
  71        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
  72        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
  73        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
  74        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
  75        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
  76        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
  77        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
  78        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
  79        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
  80        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
  81        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
  82        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
  83        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
  84        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
  85        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
  86        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
  87        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
  88        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
  89        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
  90        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
  91        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
  92        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
  93        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
  94        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
  95        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
  96        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
  97        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
  98        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
  99        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
 100        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
 101        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
 102        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
 103        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
 104        /* required last entry */
 105        {0, }
 106};
 107
 108MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
 109
 110static int igb_setup_all_tx_resources(struct igb_adapter *);
 111static int igb_setup_all_rx_resources(struct igb_adapter *);
 112static void igb_free_all_tx_resources(struct igb_adapter *);
 113static void igb_free_all_rx_resources(struct igb_adapter *);
 114static void igb_setup_mrqc(struct igb_adapter *);
 115static int igb_probe(struct pci_dev *, const struct pci_device_id *);
 116static void igb_remove(struct pci_dev *pdev);
 117static int igb_sw_init(struct igb_adapter *);
 118int igb_open(struct net_device *);
 119int igb_close(struct net_device *);
 120static void igb_configure(struct igb_adapter *);
 121static void igb_configure_tx(struct igb_adapter *);
 122static void igb_configure_rx(struct igb_adapter *);
 123static void igb_clean_all_tx_rings(struct igb_adapter *);
 124static void igb_clean_all_rx_rings(struct igb_adapter *);
 125static void igb_clean_tx_ring(struct igb_ring *);
 126static void igb_clean_rx_ring(struct igb_ring *);
 127static void igb_set_rx_mode(struct net_device *);
 128static void igb_update_phy_info(struct timer_list *);
 129static void igb_watchdog(struct timer_list *);
 130static void igb_watchdog_task(struct work_struct *);
 131static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
 132static void igb_get_stats64(struct net_device *dev,
 133                            struct rtnl_link_stats64 *stats);
 134static int igb_change_mtu(struct net_device *, int);
 135static int igb_set_mac(struct net_device *, void *);
 136static void igb_set_uta(struct igb_adapter *adapter, bool set);
 137static irqreturn_t igb_intr(int irq, void *);
 138static irqreturn_t igb_intr_msi(int irq, void *);
 139static irqreturn_t igb_msix_other(int irq, void *);
 140static irqreturn_t igb_msix_ring(int irq, void *);
 141#ifdef CONFIG_IGB_DCA
 142static void igb_update_dca(struct igb_q_vector *);
 143static void igb_setup_dca(struct igb_adapter *);
 144#endif /* CONFIG_IGB_DCA */
 145static int igb_poll(struct napi_struct *, int);
 146static bool igb_clean_tx_irq(struct igb_q_vector *, int);
 147static int igb_clean_rx_irq(struct igb_q_vector *, int);
 148static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
 149static void igb_tx_timeout(struct net_device *);
 150static void igb_reset_task(struct work_struct *);
 151static void igb_vlan_mode(struct net_device *netdev,
 152                          netdev_features_t features);
 153static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
 154static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
 155static void igb_restore_vlan(struct igb_adapter *);
 156static void igb_rar_set_index(struct igb_adapter *, u32);
 157static void igb_ping_all_vfs(struct igb_adapter *);
 158static void igb_msg_task(struct igb_adapter *);
 159static void igb_vmm_control(struct igb_adapter *);
 160static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
 161static void igb_flush_mac_table(struct igb_adapter *);
 162static int igb_available_rars(struct igb_adapter *, u8);
 163static void igb_set_default_mac_filter(struct igb_adapter *);
 164static int igb_uc_sync(struct net_device *, const unsigned char *);
 165static int igb_uc_unsync(struct net_device *, const unsigned char *);
 166static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
 167static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
 168static int igb_ndo_set_vf_vlan(struct net_device *netdev,
 169                               int vf, u16 vlan, u8 qos, __be16 vlan_proto);
 170static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
 171static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
 172                                   bool setting);
 173static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf,
 174                                bool setting);
 175static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
 176                                 struct ifla_vf_info *ivi);
 177static void igb_check_vf_rate_limit(struct igb_adapter *);
 178static void igb_nfc_filter_exit(struct igb_adapter *adapter);
 179static void igb_nfc_filter_restore(struct igb_adapter *adapter);
 180
 181#ifdef CONFIG_PCI_IOV
 182static int igb_vf_configure(struct igb_adapter *adapter, int vf);
 183static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
 184static int igb_disable_sriov(struct pci_dev *dev);
 185static int igb_pci_disable_sriov(struct pci_dev *dev);
 186#endif
 187
 188static int igb_suspend(struct device *);
 189static int igb_resume(struct device *);
 190static int igb_runtime_suspend(struct device *dev);
 191static int igb_runtime_resume(struct device *dev);
 192static int igb_runtime_idle(struct device *dev);
 193static const struct dev_pm_ops igb_pm_ops = {
 194        SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
 195        SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
 196                        igb_runtime_idle)
 197};
 198static void igb_shutdown(struct pci_dev *);
 199static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
 200#ifdef CONFIG_IGB_DCA
 201static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
 202static struct notifier_block dca_notifier = {
 203        .notifier_call  = igb_notify_dca,
 204        .next           = NULL,
 205        .priority       = 0
 206};
 207#endif
 208#ifdef CONFIG_PCI_IOV
 209static unsigned int max_vfs;
 210module_param(max_vfs, uint, 0);
 211MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
 212#endif /* CONFIG_PCI_IOV */
 213
 214static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
 215                     pci_channel_state_t);
 216static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
 217static void igb_io_resume(struct pci_dev *);
 218
 219static const struct pci_error_handlers igb_err_handler = {
 220        .error_detected = igb_io_error_detected,
 221        .slot_reset = igb_io_slot_reset,
 222        .resume = igb_io_resume,
 223};
 224
 225static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
 226
 227static struct pci_driver igb_driver = {
 228        .name     = igb_driver_name,
 229        .id_table = igb_pci_tbl,
 230        .probe    = igb_probe,
 231        .remove   = igb_remove,
 232#ifdef CONFIG_PM
 233        .driver.pm = &igb_pm_ops,
 234#endif
 235        .shutdown = igb_shutdown,
 236        .sriov_configure = igb_pci_sriov_configure,
 237        .err_handler = &igb_err_handler
 238};
 239
 240MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 241MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
 242MODULE_LICENSE("GPL v2");
 243MODULE_VERSION(DRV_VERSION);
 244
 245#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
 246static int debug = -1;
 247module_param(debug, int, 0);
 248MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 249
 250struct igb_reg_info {
 251        u32 ofs;
 252        char *name;
 253};
 254
 255static const struct igb_reg_info igb_reg_info_tbl[] = {
 256
 257        /* General Registers */
 258        {E1000_CTRL, "CTRL"},
 259        {E1000_STATUS, "STATUS"},
 260        {E1000_CTRL_EXT, "CTRL_EXT"},
 261
 262        /* Interrupt Registers */
 263        {E1000_ICR, "ICR"},
 264
 265        /* RX Registers */
 266        {E1000_RCTL, "RCTL"},
 267        {E1000_RDLEN(0), "RDLEN"},
 268        {E1000_RDH(0), "RDH"},
 269        {E1000_RDT(0), "RDT"},
 270        {E1000_RXDCTL(0), "RXDCTL"},
 271        {E1000_RDBAL(0), "RDBAL"},
 272        {E1000_RDBAH(0), "RDBAH"},
 273
 274        /* TX Registers */
 275        {E1000_TCTL, "TCTL"},
 276        {E1000_TDBAL(0), "TDBAL"},
 277        {E1000_TDBAH(0), "TDBAH"},
 278        {E1000_TDLEN(0), "TDLEN"},
 279        {E1000_TDH(0), "TDH"},
 280        {E1000_TDT(0), "TDT"},
 281        {E1000_TXDCTL(0), "TXDCTL"},
 282        {E1000_TDFH, "TDFH"},
 283        {E1000_TDFT, "TDFT"},
 284        {E1000_TDFHS, "TDFHS"},
 285        {E1000_TDFPC, "TDFPC"},
 286
 287        /* List Terminator */
 288        {}
 289};
 290
 291/* igb_regdump - register printout routine */
 292static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
 293{
 294        int n = 0;
 295        char rname[16];
 296        u32 regs[8];
 297
 298        switch (reginfo->ofs) {
 299        case E1000_RDLEN(0):
 300                for (n = 0; n < 4; n++)
 301                        regs[n] = rd32(E1000_RDLEN(n));
 302                break;
 303        case E1000_RDH(0):
 304                for (n = 0; n < 4; n++)
 305                        regs[n] = rd32(E1000_RDH(n));
 306                break;
 307        case E1000_RDT(0):
 308                for (n = 0; n < 4; n++)
 309                        regs[n] = rd32(E1000_RDT(n));
 310                break;
 311        case E1000_RXDCTL(0):
 312                for (n = 0; n < 4; n++)
 313                        regs[n] = rd32(E1000_RXDCTL(n));
 314                break;
 315        case E1000_RDBAL(0):
 316                for (n = 0; n < 4; n++)
 317                        regs[n] = rd32(E1000_RDBAL(n));
 318                break;
 319        case E1000_RDBAH(0):
 320                for (n = 0; n < 4; n++)
 321                        regs[n] = rd32(E1000_RDBAH(n));
 322                break;
 323        case E1000_TDBAL(0):
 324                for (n = 0; n < 4; n++)
 325                        regs[n] = rd32(E1000_RDBAL(n));
 326                break;
 327        case E1000_TDBAH(0):
 328                for (n = 0; n < 4; n++)
 329                        regs[n] = rd32(E1000_TDBAH(n));
 330                break;
 331        case E1000_TDLEN(0):
 332                for (n = 0; n < 4; n++)
 333                        regs[n] = rd32(E1000_TDLEN(n));
 334                break;
 335        case E1000_TDH(0):
 336                for (n = 0; n < 4; n++)
 337                        regs[n] = rd32(E1000_TDH(n));
 338                break;
 339        case E1000_TDT(0):
 340                for (n = 0; n < 4; n++)
 341                        regs[n] = rd32(E1000_TDT(n));
 342                break;
 343        case E1000_TXDCTL(0):
 344                for (n = 0; n < 4; n++)
 345                        regs[n] = rd32(E1000_TXDCTL(n));
 346                break;
 347        default:
 348                pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
 349                return;
 350        }
 351
 352        snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
 353        pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
 354                regs[2], regs[3]);
 355}
 356
 357/* igb_dump - Print registers, Tx-rings and Rx-rings */
 358static void igb_dump(struct igb_adapter *adapter)
 359{
 360        struct net_device *netdev = adapter->netdev;
 361        struct e1000_hw *hw = &adapter->hw;
 362        struct igb_reg_info *reginfo;
 363        struct igb_ring *tx_ring;
 364        union e1000_adv_tx_desc *tx_desc;
 365        struct my_u0 { u64 a; u64 b; } *u0;
 366        struct igb_ring *rx_ring;
 367        union e1000_adv_rx_desc *rx_desc;
 368        u32 staterr;
 369        u16 i, n;
 370
 371        if (!netif_msg_hw(adapter))
 372                return;
 373
 374        /* Print netdevice Info */
 375        if (netdev) {
 376                dev_info(&adapter->pdev->dev, "Net device Info\n");
 377                pr_info("Device Name     state            trans_start\n");
 378                pr_info("%-15s %016lX %016lX\n", netdev->name,
 379                        netdev->state, dev_trans_start(netdev));
 380        }
 381
 382        /* Print Registers */
 383        dev_info(&adapter->pdev->dev, "Register Dump\n");
 384        pr_info(" Register Name   Value\n");
 385        for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
 386             reginfo->name; reginfo++) {
 387                igb_regdump(hw, reginfo);
 388        }
 389
 390        /* Print TX Ring Summary */
 391        if (!netdev || !netif_running(netdev))
 392                goto exit;
 393
 394        dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
 395        pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
 396        for (n = 0; n < adapter->num_tx_queues; n++) {
 397                struct igb_tx_buffer *buffer_info;
 398                tx_ring = adapter->tx_ring[n];
 399                buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
 400                pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
 401                        n, tx_ring->next_to_use, tx_ring->next_to_clean,
 402                        (u64)dma_unmap_addr(buffer_info, dma),
 403                        dma_unmap_len(buffer_info, len),
 404                        buffer_info->next_to_watch,
 405                        (u64)buffer_info->time_stamp);
 406        }
 407
 408        /* Print TX Rings */
 409        if (!netif_msg_tx_done(adapter))
 410                goto rx_ring_summary;
 411
 412        dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
 413
 414        /* Transmit Descriptor Formats
 415         *
 416         * Advanced Transmit Descriptor
 417         *   +--------------------------------------------------------------+
 418         * 0 |         Buffer Address [63:0]                                |
 419         *   +--------------------------------------------------------------+
 420         * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
 421         *   +--------------------------------------------------------------+
 422         *   63      46 45    40 39 38 36 35 32 31   24             15       0
 423         */
 424
 425        for (n = 0; n < adapter->num_tx_queues; n++) {
 426                tx_ring = adapter->tx_ring[n];
 427                pr_info("------------------------------------\n");
 428                pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
 429                pr_info("------------------------------------\n");
 430                pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
 431
 432                for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
 433                        const char *next_desc;
 434                        struct igb_tx_buffer *buffer_info;
 435                        tx_desc = IGB_TX_DESC(tx_ring, i);
 436                        buffer_info = &tx_ring->tx_buffer_info[i];
 437                        u0 = (struct my_u0 *)tx_desc;
 438                        if (i == tx_ring->next_to_use &&
 439                            i == tx_ring->next_to_clean)
 440                                next_desc = " NTC/U";
 441                        else if (i == tx_ring->next_to_use)
 442                                next_desc = " NTU";
 443                        else if (i == tx_ring->next_to_clean)
 444                                next_desc = " NTC";
 445                        else
 446                                next_desc = "";
 447
 448                        pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
 449                                i, le64_to_cpu(u0->a),
 450                                le64_to_cpu(u0->b),
 451                                (u64)dma_unmap_addr(buffer_info, dma),
 452                                dma_unmap_len(buffer_info, len),
 453                                buffer_info->next_to_watch,
 454                                (u64)buffer_info->time_stamp,
 455                                buffer_info->skb, next_desc);
 456
 457                        if (netif_msg_pktdata(adapter) && buffer_info->skb)
 458                                print_hex_dump(KERN_INFO, "",
 459                                        DUMP_PREFIX_ADDRESS,
 460                                        16, 1, buffer_info->skb->data,
 461                                        dma_unmap_len(buffer_info, len),
 462                                        true);
 463                }
 464        }
 465
 466        /* Print RX Rings Summary */
 467rx_ring_summary:
 468        dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
 469        pr_info("Queue [NTU] [NTC]\n");
 470        for (n = 0; n < adapter->num_rx_queues; n++) {
 471                rx_ring = adapter->rx_ring[n];
 472                pr_info(" %5d %5X %5X\n",
 473                        n, rx_ring->next_to_use, rx_ring->next_to_clean);
 474        }
 475
 476        /* Print RX Rings */
 477        if (!netif_msg_rx_status(adapter))
 478                goto exit;
 479
 480        dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
 481
 482        /* Advanced Receive Descriptor (Read) Format
 483         *    63                                           1        0
 484         *    +-----------------------------------------------------+
 485         *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
 486         *    +----------------------------------------------+------+
 487         *  8 |       Header Buffer Address [63:1]           |  DD  |
 488         *    +-----------------------------------------------------+
 489         *
 490         *
 491         * Advanced Receive Descriptor (Write-Back) Format
 492         *
 493         *   63       48 47    32 31  30      21 20 17 16   4 3     0
 494         *   +------------------------------------------------------+
 495         * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
 496         *   | Checksum   Ident  |   |           |    | Type | Type |
 497         *   +------------------------------------------------------+
 498         * 8 | VLAN Tag | Length | Extended Error | Extended Status |
 499         *   +------------------------------------------------------+
 500         *   63       48 47    32 31            20 19               0
 501         */
 502
 503        for (n = 0; n < adapter->num_rx_queues; n++) {
 504                rx_ring = adapter->rx_ring[n];
 505                pr_info("------------------------------------\n");
 506                pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
 507                pr_info("------------------------------------\n");
 508                pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
 509                pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
 510
 511                for (i = 0; i < rx_ring->count; i++) {
 512                        const char *next_desc;
 513                        struct igb_rx_buffer *buffer_info;
 514                        buffer_info = &rx_ring->rx_buffer_info[i];
 515                        rx_desc = IGB_RX_DESC(rx_ring, i);
 516                        u0 = (struct my_u0 *)rx_desc;
 517                        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 518
 519                        if (i == rx_ring->next_to_use)
 520                                next_desc = " NTU";
 521                        else if (i == rx_ring->next_to_clean)
 522                                next_desc = " NTC";
 523                        else
 524                                next_desc = "";
 525
 526                        if (staterr & E1000_RXD_STAT_DD) {
 527                                /* Descriptor Done */
 528                                pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
 529                                        "RWB", i,
 530                                        le64_to_cpu(u0->a),
 531                                        le64_to_cpu(u0->b),
 532                                        next_desc);
 533                        } else {
 534                                pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
 535                                        "R  ", i,
 536                                        le64_to_cpu(u0->a),
 537                                        le64_to_cpu(u0->b),
 538                                        (u64)buffer_info->dma,
 539                                        next_desc);
 540
 541                                if (netif_msg_pktdata(adapter) &&
 542                                    buffer_info->dma && buffer_info->page) {
 543                                        print_hex_dump(KERN_INFO, "",
 544                                          DUMP_PREFIX_ADDRESS,
 545                                          16, 1,
 546                                          page_address(buffer_info->page) +
 547                                                      buffer_info->page_offset,
 548                                          igb_rx_bufsz(rx_ring), true);
 549                                }
 550                        }
 551                }
 552        }
 553
 554exit:
 555        return;
 556}
 557
 558/**
 559 *  igb_get_i2c_data - Reads the I2C SDA data bit
 560 *  @hw: pointer to hardware structure
 561 *  @i2cctl: Current value of I2CCTL register
 562 *
 563 *  Returns the I2C data bit value
 564 **/
 565static int igb_get_i2c_data(void *data)
 566{
 567        struct igb_adapter *adapter = (struct igb_adapter *)data;
 568        struct e1000_hw *hw = &adapter->hw;
 569        s32 i2cctl = rd32(E1000_I2CPARAMS);
 570
 571        return !!(i2cctl & E1000_I2C_DATA_IN);
 572}
 573
 574/**
 575 *  igb_set_i2c_data - Sets the I2C data bit
 576 *  @data: pointer to hardware structure
 577 *  @state: I2C data value (0 or 1) to set
 578 *
 579 *  Sets the I2C data bit
 580 **/
 581static void igb_set_i2c_data(void *data, int state)
 582{
 583        struct igb_adapter *adapter = (struct igb_adapter *)data;
 584        struct e1000_hw *hw = &adapter->hw;
 585        s32 i2cctl = rd32(E1000_I2CPARAMS);
 586
 587        if (state)
 588                i2cctl |= E1000_I2C_DATA_OUT;
 589        else
 590                i2cctl &= ~E1000_I2C_DATA_OUT;
 591
 592        i2cctl &= ~E1000_I2C_DATA_OE_N;
 593        i2cctl |= E1000_I2C_CLK_OE_N;
 594        wr32(E1000_I2CPARAMS, i2cctl);
 595        wrfl();
 596
 597}
 598
 599/**
 600 *  igb_set_i2c_clk - Sets the I2C SCL clock
 601 *  @data: pointer to hardware structure
 602 *  @state: state to set clock
 603 *
 604 *  Sets the I2C clock line to state
 605 **/
 606static void igb_set_i2c_clk(void *data, int state)
 607{
 608        struct igb_adapter *adapter = (struct igb_adapter *)data;
 609        struct e1000_hw *hw = &adapter->hw;
 610        s32 i2cctl = rd32(E1000_I2CPARAMS);
 611
 612        if (state) {
 613                i2cctl |= E1000_I2C_CLK_OUT;
 614                i2cctl &= ~E1000_I2C_CLK_OE_N;
 615        } else {
 616                i2cctl &= ~E1000_I2C_CLK_OUT;
 617                i2cctl &= ~E1000_I2C_CLK_OE_N;
 618        }
 619        wr32(E1000_I2CPARAMS, i2cctl);
 620        wrfl();
 621}
 622
 623/**
 624 *  igb_get_i2c_clk - Gets the I2C SCL clock state
 625 *  @data: pointer to hardware structure
 626 *
 627 *  Gets the I2C clock state
 628 **/
 629static int igb_get_i2c_clk(void *data)
 630{
 631        struct igb_adapter *adapter = (struct igb_adapter *)data;
 632        struct e1000_hw *hw = &adapter->hw;
 633        s32 i2cctl = rd32(E1000_I2CPARAMS);
 634
 635        return !!(i2cctl & E1000_I2C_CLK_IN);
 636}
 637
 638static const struct i2c_algo_bit_data igb_i2c_algo = {
 639        .setsda         = igb_set_i2c_data,
 640        .setscl         = igb_set_i2c_clk,
 641        .getsda         = igb_get_i2c_data,
 642        .getscl         = igb_get_i2c_clk,
 643        .udelay         = 5,
 644        .timeout        = 20,
 645};
 646
 647/**
 648 *  igb_get_hw_dev - return device
 649 *  @hw: pointer to hardware structure
 650 *
 651 *  used by hardware layer to print debugging information
 652 **/
 653struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
 654{
 655        struct igb_adapter *adapter = hw->back;
 656        return adapter->netdev;
 657}
 658
 659/**
 660 *  igb_init_module - Driver Registration Routine
 661 *
 662 *  igb_init_module is the first routine called when the driver is
 663 *  loaded. All it does is register with the PCI subsystem.
 664 **/
 665static int __init igb_init_module(void)
 666{
 667        int ret;
 668
 669        pr_info("%s - version %s\n",
 670               igb_driver_string, igb_driver_version);
 671        pr_info("%s\n", igb_copyright);
 672
 673#ifdef CONFIG_IGB_DCA
 674        dca_register_notify(&dca_notifier);
 675#endif
 676        ret = pci_register_driver(&igb_driver);
 677        return ret;
 678}
 679
 680module_init(igb_init_module);
 681
 682/**
 683 *  igb_exit_module - Driver Exit Cleanup Routine
 684 *
 685 *  igb_exit_module is called just before the driver is removed
 686 *  from memory.
 687 **/
 688static void __exit igb_exit_module(void)
 689{
 690#ifdef CONFIG_IGB_DCA
 691        dca_unregister_notify(&dca_notifier);
 692#endif
 693        pci_unregister_driver(&igb_driver);
 694}
 695
 696module_exit(igb_exit_module);
 697
 698#define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
 699/**
 700 *  igb_cache_ring_register - Descriptor ring to register mapping
 701 *  @adapter: board private structure to initialize
 702 *
 703 *  Once we know the feature-set enabled for the device, we'll cache
 704 *  the register offset the descriptor ring is assigned to.
 705 **/
 706static void igb_cache_ring_register(struct igb_adapter *adapter)
 707{
 708        int i = 0, j = 0;
 709        u32 rbase_offset = adapter->vfs_allocated_count;
 710
 711        switch (adapter->hw.mac.type) {
 712        case e1000_82576:
 713                /* The queues are allocated for virtualization such that VF 0
 714                 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
 715                 * In order to avoid collision we start at the first free queue
 716                 * and continue consuming queues in the same sequence
 717                 */
 718                if (adapter->vfs_allocated_count) {
 719                        for (; i < adapter->rss_queues; i++)
 720                                adapter->rx_ring[i]->reg_idx = rbase_offset +
 721                                                               Q_IDX_82576(i);
 722                }
 723                /* Fall through */
 724        case e1000_82575:
 725        case e1000_82580:
 726        case e1000_i350:
 727        case e1000_i354:
 728        case e1000_i210:
 729        case e1000_i211:
 730                /* Fall through */
 731        default:
 732                for (; i < adapter->num_rx_queues; i++)
 733                        adapter->rx_ring[i]->reg_idx = rbase_offset + i;
 734                for (; j < adapter->num_tx_queues; j++)
 735                        adapter->tx_ring[j]->reg_idx = rbase_offset + j;
 736                break;
 737        }
 738}
 739
 740u32 igb_rd32(struct e1000_hw *hw, u32 reg)
 741{
 742        struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
 743        u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
 744        u32 value = 0;
 745
 746        if (E1000_REMOVED(hw_addr))
 747                return ~value;
 748
 749        value = readl(&hw_addr[reg]);
 750
 751        /* reads should not return all F's */
 752        if (!(~value) && (!reg || !(~readl(hw_addr)))) {
 753                struct net_device *netdev = igb->netdev;
 754                hw->hw_addr = NULL;
 755                netdev_err(netdev, "PCIe link lost\n");
 756                WARN(1, "igb: Failed to read reg 0x%x!\n", reg);
 757        }
 758
 759        return value;
 760}
 761
 762/**
 763 *  igb_write_ivar - configure ivar for given MSI-X vector
 764 *  @hw: pointer to the HW structure
 765 *  @msix_vector: vector number we are allocating to a given ring
 766 *  @index: row index of IVAR register to write within IVAR table
 767 *  @offset: column offset of in IVAR, should be multiple of 8
 768 *
 769 *  This function is intended to handle the writing of the IVAR register
 770 *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
 771 *  each containing an cause allocation for an Rx and Tx ring, and a
 772 *  variable number of rows depending on the number of queues supported.
 773 **/
 774static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
 775                           int index, int offset)
 776{
 777        u32 ivar = array_rd32(E1000_IVAR0, index);
 778
 779        /* clear any bits that are currently set */
 780        ivar &= ~((u32)0xFF << offset);
 781
 782        /* write vector and valid bit */
 783        ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
 784
 785        array_wr32(E1000_IVAR0, index, ivar);
 786}
 787
 788#define IGB_N0_QUEUE -1
 789static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
 790{
 791        struct igb_adapter *adapter = q_vector->adapter;
 792        struct e1000_hw *hw = &adapter->hw;
 793        int rx_queue = IGB_N0_QUEUE;
 794        int tx_queue = IGB_N0_QUEUE;
 795        u32 msixbm = 0;
 796
 797        if (q_vector->rx.ring)
 798                rx_queue = q_vector->rx.ring->reg_idx;
 799        if (q_vector->tx.ring)
 800                tx_queue = q_vector->tx.ring->reg_idx;
 801
 802        switch (hw->mac.type) {
 803        case e1000_82575:
 804                /* The 82575 assigns vectors using a bitmask, which matches the
 805                 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
 806                 * or more queues to a vector, we write the appropriate bits
 807                 * into the MSIXBM register for that vector.
 808                 */
 809                if (rx_queue > IGB_N0_QUEUE)
 810                        msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
 811                if (tx_queue > IGB_N0_QUEUE)
 812                        msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
 813                if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
 814                        msixbm |= E1000_EIMS_OTHER;
 815                array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
 816                q_vector->eims_value = msixbm;
 817                break;
 818        case e1000_82576:
 819                /* 82576 uses a table that essentially consists of 2 columns
 820                 * with 8 rows.  The ordering is column-major so we use the
 821                 * lower 3 bits as the row index, and the 4th bit as the
 822                 * column offset.
 823                 */
 824                if (rx_queue > IGB_N0_QUEUE)
 825                        igb_write_ivar(hw, msix_vector,
 826                                       rx_queue & 0x7,
 827                                       (rx_queue & 0x8) << 1);
 828                if (tx_queue > IGB_N0_QUEUE)
 829                        igb_write_ivar(hw, msix_vector,
 830                                       tx_queue & 0x7,
 831                                       ((tx_queue & 0x8) << 1) + 8);
 832                q_vector->eims_value = BIT(msix_vector);
 833                break;
 834        case e1000_82580:
 835        case e1000_i350:
 836        case e1000_i354:
 837        case e1000_i210:
 838        case e1000_i211:
 839                /* On 82580 and newer adapters the scheme is similar to 82576
 840                 * however instead of ordering column-major we have things
 841                 * ordered row-major.  So we traverse the table by using
 842                 * bit 0 as the column offset, and the remaining bits as the
 843                 * row index.
 844                 */
 845                if (rx_queue > IGB_N0_QUEUE)
 846                        igb_write_ivar(hw, msix_vector,
 847                                       rx_queue >> 1,
 848                                       (rx_queue & 0x1) << 4);
 849                if (tx_queue > IGB_N0_QUEUE)
 850                        igb_write_ivar(hw, msix_vector,
 851                                       tx_queue >> 1,
 852                                       ((tx_queue & 0x1) << 4) + 8);
 853                q_vector->eims_value = BIT(msix_vector);
 854                break;
 855        default:
 856                BUG();
 857                break;
 858        }
 859
 860        /* add q_vector eims value to global eims_enable_mask */
 861        adapter->eims_enable_mask |= q_vector->eims_value;
 862
 863        /* configure q_vector to set itr on first interrupt */
 864        q_vector->set_itr = 1;
 865}
 866
 867/**
 868 *  igb_configure_msix - Configure MSI-X hardware
 869 *  @adapter: board private structure to initialize
 870 *
 871 *  igb_configure_msix sets up the hardware to properly
 872 *  generate MSI-X interrupts.
 873 **/
 874static void igb_configure_msix(struct igb_adapter *adapter)
 875{
 876        u32 tmp;
 877        int i, vector = 0;
 878        struct e1000_hw *hw = &adapter->hw;
 879
 880        adapter->eims_enable_mask = 0;
 881
 882        /* set vector for other causes, i.e. link changes */
 883        switch (hw->mac.type) {
 884        case e1000_82575:
 885                tmp = rd32(E1000_CTRL_EXT);
 886                /* enable MSI-X PBA support*/
 887                tmp |= E1000_CTRL_EXT_PBA_CLR;
 888
 889                /* Auto-Mask interrupts upon ICR read. */
 890                tmp |= E1000_CTRL_EXT_EIAME;
 891                tmp |= E1000_CTRL_EXT_IRCA;
 892
 893                wr32(E1000_CTRL_EXT, tmp);
 894
 895                /* enable msix_other interrupt */
 896                array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
 897                adapter->eims_other = E1000_EIMS_OTHER;
 898
 899                break;
 900
 901        case e1000_82576:
 902        case e1000_82580:
 903        case e1000_i350:
 904        case e1000_i354:
 905        case e1000_i210:
 906        case e1000_i211:
 907                /* Turn on MSI-X capability first, or our settings
 908                 * won't stick.  And it will take days to debug.
 909                 */
 910                wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
 911                     E1000_GPIE_PBA | E1000_GPIE_EIAME |
 912                     E1000_GPIE_NSICR);
 913
 914                /* enable msix_other interrupt */
 915                adapter->eims_other = BIT(vector);
 916                tmp = (vector++ | E1000_IVAR_VALID) << 8;
 917
 918                wr32(E1000_IVAR_MISC, tmp);
 919                break;
 920        default:
 921                /* do nothing, since nothing else supports MSI-X */
 922                break;
 923        } /* switch (hw->mac.type) */
 924
 925        adapter->eims_enable_mask |= adapter->eims_other;
 926
 927        for (i = 0; i < adapter->num_q_vectors; i++)
 928                igb_assign_vector(adapter->q_vector[i], vector++);
 929
 930        wrfl();
 931}
 932
 933/**
 934 *  igb_request_msix - Initialize MSI-X interrupts
 935 *  @adapter: board private structure to initialize
 936 *
 937 *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
 938 *  kernel.
 939 **/
 940static int igb_request_msix(struct igb_adapter *adapter)
 941{
 942        struct net_device *netdev = adapter->netdev;
 943        int i, err = 0, vector = 0, free_vector = 0;
 944
 945        err = request_irq(adapter->msix_entries[vector].vector,
 946                          igb_msix_other, 0, netdev->name, adapter);
 947        if (err)
 948                goto err_out;
 949
 950        for (i = 0; i < adapter->num_q_vectors; i++) {
 951                struct igb_q_vector *q_vector = adapter->q_vector[i];
 952
 953                vector++;
 954
 955                q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
 956
 957                if (q_vector->rx.ring && q_vector->tx.ring)
 958                        sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
 959                                q_vector->rx.ring->queue_index);
 960                else if (q_vector->tx.ring)
 961                        sprintf(q_vector->name, "%s-tx-%u", netdev->name,
 962                                q_vector->tx.ring->queue_index);
 963                else if (q_vector->rx.ring)
 964                        sprintf(q_vector->name, "%s-rx-%u", netdev->name,
 965                                q_vector->rx.ring->queue_index);
 966                else
 967                        sprintf(q_vector->name, "%s-unused", netdev->name);
 968
 969                err = request_irq(adapter->msix_entries[vector].vector,
 970                                  igb_msix_ring, 0, q_vector->name,
 971                                  q_vector);
 972                if (err)
 973                        goto err_free;
 974        }
 975
 976        igb_configure_msix(adapter);
 977        return 0;
 978
 979err_free:
 980        /* free already assigned IRQs */
 981        free_irq(adapter->msix_entries[free_vector++].vector, adapter);
 982
 983        vector--;
 984        for (i = 0; i < vector; i++) {
 985                free_irq(adapter->msix_entries[free_vector++].vector,
 986                         adapter->q_vector[i]);
 987        }
 988err_out:
 989        return err;
 990}
 991
 992/**
 993 *  igb_free_q_vector - Free memory allocated for specific interrupt vector
 994 *  @adapter: board private structure to initialize
 995 *  @v_idx: Index of vector to be freed
 996 *
 997 *  This function frees the memory allocated to the q_vector.
 998 **/
 999static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
1000{
1001        struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1002
1003        adapter->q_vector[v_idx] = NULL;
1004
1005        /* igb_get_stats64() might access the rings on this vector,
1006         * we must wait a grace period before freeing it.
1007         */
1008        if (q_vector)
1009                kfree_rcu(q_vector, rcu);
1010}
1011
1012/**
1013 *  igb_reset_q_vector - Reset config for interrupt vector
1014 *  @adapter: board private structure to initialize
1015 *  @v_idx: Index of vector to be reset
1016 *
1017 *  If NAPI is enabled it will delete any references to the
1018 *  NAPI struct. This is preparation for igb_free_q_vector.
1019 **/
1020static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1021{
1022        struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1023
1024        /* Coming from igb_set_interrupt_capability, the vectors are not yet
1025         * allocated. So, q_vector is NULL so we should stop here.
1026         */
1027        if (!q_vector)
1028                return;
1029
1030        if (q_vector->tx.ring)
1031                adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1032
1033        if (q_vector->rx.ring)
1034                adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1035
1036        netif_napi_del(&q_vector->napi);
1037
1038}
1039
1040static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1041{
1042        int v_idx = adapter->num_q_vectors;
1043
1044        if (adapter->flags & IGB_FLAG_HAS_MSIX)
1045                pci_disable_msix(adapter->pdev);
1046        else if (adapter->flags & IGB_FLAG_HAS_MSI)
1047                pci_disable_msi(adapter->pdev);
1048
1049        while (v_idx--)
1050                igb_reset_q_vector(adapter, v_idx);
1051}
1052
1053/**
1054 *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1055 *  @adapter: board private structure to initialize
1056 *
1057 *  This function frees the memory allocated to the q_vectors.  In addition if
1058 *  NAPI is enabled it will delete any references to the NAPI struct prior
1059 *  to freeing the q_vector.
1060 **/
1061static void igb_free_q_vectors(struct igb_adapter *adapter)
1062{
1063        int v_idx = adapter->num_q_vectors;
1064
1065        adapter->num_tx_queues = 0;
1066        adapter->num_rx_queues = 0;
1067        adapter->num_q_vectors = 0;
1068
1069        while (v_idx--) {
1070                igb_reset_q_vector(adapter, v_idx);
1071                igb_free_q_vector(adapter, v_idx);
1072        }
1073}
1074
1075/**
1076 *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1077 *  @adapter: board private structure to initialize
1078 *
1079 *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1080 *  MSI-X interrupts allocated.
1081 */
1082static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1083{
1084        igb_free_q_vectors(adapter);
1085        igb_reset_interrupt_capability(adapter);
1086}
1087
1088/**
1089 *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1090 *  @adapter: board private structure to initialize
1091 *  @msix: boolean value of MSIX capability
1092 *
1093 *  Attempt to configure interrupts using the best available
1094 *  capabilities of the hardware and kernel.
1095 **/
1096static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1097{
1098        int err;
1099        int numvecs, i;
1100
1101        if (!msix)
1102                goto msi_only;
1103        adapter->flags |= IGB_FLAG_HAS_MSIX;
1104
1105        /* Number of supported queues. */
1106        adapter->num_rx_queues = adapter->rss_queues;
1107        if (adapter->vfs_allocated_count)
1108                adapter->num_tx_queues = 1;
1109        else
1110                adapter->num_tx_queues = adapter->rss_queues;
1111
1112        /* start with one vector for every Rx queue */
1113        numvecs = adapter->num_rx_queues;
1114
1115        /* if Tx handler is separate add 1 for every Tx queue */
1116        if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1117                numvecs += adapter->num_tx_queues;
1118
1119        /* store the number of vectors reserved for queues */
1120        adapter->num_q_vectors = numvecs;
1121
1122        /* add 1 vector for link status interrupts */
1123        numvecs++;
1124        for (i = 0; i < numvecs; i++)
1125                adapter->msix_entries[i].entry = i;
1126
1127        err = pci_enable_msix_range(adapter->pdev,
1128                                    adapter->msix_entries,
1129                                    numvecs,
1130                                    numvecs);
1131        if (err > 0)
1132                return;
1133
1134        igb_reset_interrupt_capability(adapter);
1135
1136        /* If we can't do MSI-X, try MSI */
1137msi_only:
1138        adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1139#ifdef CONFIG_PCI_IOV
1140        /* disable SR-IOV for non MSI-X configurations */
1141        if (adapter->vf_data) {
1142                struct e1000_hw *hw = &adapter->hw;
1143                /* disable iov and allow time for transactions to clear */
1144                pci_disable_sriov(adapter->pdev);
1145                msleep(500);
1146
1147                kfree(adapter->vf_mac_list);
1148                adapter->vf_mac_list = NULL;
1149                kfree(adapter->vf_data);
1150                adapter->vf_data = NULL;
1151                wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1152                wrfl();
1153                msleep(100);
1154                dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1155        }
1156#endif
1157        adapter->vfs_allocated_count = 0;
1158        adapter->rss_queues = 1;
1159        adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1160        adapter->num_rx_queues = 1;
1161        adapter->num_tx_queues = 1;
1162        adapter->num_q_vectors = 1;
1163        if (!pci_enable_msi(adapter->pdev))
1164                adapter->flags |= IGB_FLAG_HAS_MSI;
1165}
1166
1167static void igb_add_ring(struct igb_ring *ring,
1168                         struct igb_ring_container *head)
1169{
1170        head->ring = ring;
1171        head->count++;
1172}
1173
1174/**
1175 *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1176 *  @adapter: board private structure to initialize
1177 *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1178 *  @v_idx: index of vector in adapter struct
1179 *  @txr_count: total number of Tx rings to allocate
1180 *  @txr_idx: index of first Tx ring to allocate
1181 *  @rxr_count: total number of Rx rings to allocate
1182 *  @rxr_idx: index of first Rx ring to allocate
1183 *
1184 *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1185 **/
1186static int igb_alloc_q_vector(struct igb_adapter *adapter,
1187                              int v_count, int v_idx,
1188                              int txr_count, int txr_idx,
1189                              int rxr_count, int rxr_idx)
1190{
1191        struct igb_q_vector *q_vector;
1192        struct igb_ring *ring;
1193        int ring_count;
1194        size_t size;
1195
1196        /* igb only supports 1 Tx and/or 1 Rx queue per vector */
1197        if (txr_count > 1 || rxr_count > 1)
1198                return -ENOMEM;
1199
1200        ring_count = txr_count + rxr_count;
1201        size = struct_size(q_vector, ring, ring_count);
1202
1203        /* allocate q_vector and rings */
1204        q_vector = adapter->q_vector[v_idx];
1205        if (!q_vector) {
1206                q_vector = kzalloc(size, GFP_KERNEL);
1207        } else if (size > ksize(q_vector)) {
1208                kfree_rcu(q_vector, rcu);
1209                q_vector = kzalloc(size, GFP_KERNEL);
1210        } else {
1211                memset(q_vector, 0, size);
1212        }
1213        if (!q_vector)
1214                return -ENOMEM;
1215
1216        /* initialize NAPI */
1217        netif_napi_add(adapter->netdev, &q_vector->napi,
1218                       igb_poll, 64);
1219
1220        /* tie q_vector and adapter together */
1221        adapter->q_vector[v_idx] = q_vector;
1222        q_vector->adapter = adapter;
1223
1224        /* initialize work limits */
1225        q_vector->tx.work_limit = adapter->tx_work_limit;
1226
1227        /* initialize ITR configuration */
1228        q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1229        q_vector->itr_val = IGB_START_ITR;
1230
1231        /* initialize pointer to rings */
1232        ring = q_vector->ring;
1233
1234        /* intialize ITR */
1235        if (rxr_count) {
1236                /* rx or rx/tx vector */
1237                if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1238                        q_vector->itr_val = adapter->rx_itr_setting;
1239        } else {
1240                /* tx only vector */
1241                if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1242                        q_vector->itr_val = adapter->tx_itr_setting;
1243        }
1244
1245        if (txr_count) {
1246                /* assign generic ring traits */
1247                ring->dev = &adapter->pdev->dev;
1248                ring->netdev = adapter->netdev;
1249
1250                /* configure backlink on ring */
1251                ring->q_vector = q_vector;
1252
1253                /* update q_vector Tx values */
1254                igb_add_ring(ring, &q_vector->tx);
1255
1256                /* For 82575, context index must be unique per ring. */
1257                if (adapter->hw.mac.type == e1000_82575)
1258                        set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1259
1260                /* apply Tx specific ring traits */
1261                ring->count = adapter->tx_ring_count;
1262                ring->queue_index = txr_idx;
1263
1264                ring->cbs_enable = false;
1265                ring->idleslope = 0;
1266                ring->sendslope = 0;
1267                ring->hicredit = 0;
1268                ring->locredit = 0;
1269
1270                u64_stats_init(&ring->tx_syncp);
1271                u64_stats_init(&ring->tx_syncp2);
1272
1273                /* assign ring to adapter */
1274                adapter->tx_ring[txr_idx] = ring;
1275
1276                /* push pointer to next ring */
1277                ring++;
1278        }
1279
1280        if (rxr_count) {
1281                /* assign generic ring traits */
1282                ring->dev = &adapter->pdev->dev;
1283                ring->netdev = adapter->netdev;
1284
1285                /* configure backlink on ring */
1286                ring->q_vector = q_vector;
1287
1288                /* update q_vector Rx values */
1289                igb_add_ring(ring, &q_vector->rx);
1290
1291                /* set flag indicating ring supports SCTP checksum offload */
1292                if (adapter->hw.mac.type >= e1000_82576)
1293                        set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1294
1295                /* On i350, i354, i210, and i211, loopback VLAN packets
1296                 * have the tag byte-swapped.
1297                 */
1298                if (adapter->hw.mac.type >= e1000_i350)
1299                        set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1300
1301                /* apply Rx specific ring traits */
1302                ring->count = adapter->rx_ring_count;
1303                ring->queue_index = rxr_idx;
1304
1305                u64_stats_init(&ring->rx_syncp);
1306
1307                /* assign ring to adapter */
1308                adapter->rx_ring[rxr_idx] = ring;
1309        }
1310
1311        return 0;
1312}
1313
1314
1315/**
1316 *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1317 *  @adapter: board private structure to initialize
1318 *
1319 *  We allocate one q_vector per queue interrupt.  If allocation fails we
1320 *  return -ENOMEM.
1321 **/
1322static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1323{
1324        int q_vectors = adapter->num_q_vectors;
1325        int rxr_remaining = adapter->num_rx_queues;
1326        int txr_remaining = adapter->num_tx_queues;
1327        int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1328        int err;
1329
1330        if (q_vectors >= (rxr_remaining + txr_remaining)) {
1331                for (; rxr_remaining; v_idx++) {
1332                        err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1333                                                 0, 0, 1, rxr_idx);
1334
1335                        if (err)
1336                                goto err_out;
1337
1338                        /* update counts and index */
1339                        rxr_remaining--;
1340                        rxr_idx++;
1341                }
1342        }
1343
1344        for (; v_idx < q_vectors; v_idx++) {
1345                int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1346                int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1347
1348                err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1349                                         tqpv, txr_idx, rqpv, rxr_idx);
1350
1351                if (err)
1352                        goto err_out;
1353
1354                /* update counts and index */
1355                rxr_remaining -= rqpv;
1356                txr_remaining -= tqpv;
1357                rxr_idx++;
1358                txr_idx++;
1359        }
1360
1361        return 0;
1362
1363err_out:
1364        adapter->num_tx_queues = 0;
1365        adapter->num_rx_queues = 0;
1366        adapter->num_q_vectors = 0;
1367
1368        while (v_idx--)
1369                igb_free_q_vector(adapter, v_idx);
1370
1371        return -ENOMEM;
1372}
1373
1374/**
1375 *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1376 *  @adapter: board private structure to initialize
1377 *  @msix: boolean value of MSIX capability
1378 *
1379 *  This function initializes the interrupts and allocates all of the queues.
1380 **/
1381static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1382{
1383        struct pci_dev *pdev = adapter->pdev;
1384        int err;
1385
1386        igb_set_interrupt_capability(adapter, msix);
1387
1388        err = igb_alloc_q_vectors(adapter);
1389        if (err) {
1390                dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1391                goto err_alloc_q_vectors;
1392        }
1393
1394        igb_cache_ring_register(adapter);
1395
1396        return 0;
1397
1398err_alloc_q_vectors:
1399        igb_reset_interrupt_capability(adapter);
1400        return err;
1401}
1402
1403/**
1404 *  igb_request_irq - initialize interrupts
1405 *  @adapter: board private structure to initialize
1406 *
1407 *  Attempts to configure interrupts using the best available
1408 *  capabilities of the hardware and kernel.
1409 **/
1410static int igb_request_irq(struct igb_adapter *adapter)
1411{
1412        struct net_device *netdev = adapter->netdev;
1413        struct pci_dev *pdev = adapter->pdev;
1414        int err = 0;
1415
1416        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1417                err = igb_request_msix(adapter);
1418                if (!err)
1419                        goto request_done;
1420                /* fall back to MSI */
1421                igb_free_all_tx_resources(adapter);
1422                igb_free_all_rx_resources(adapter);
1423
1424                igb_clear_interrupt_scheme(adapter);
1425                err = igb_init_interrupt_scheme(adapter, false);
1426                if (err)
1427                        goto request_done;
1428
1429                igb_setup_all_tx_resources(adapter);
1430                igb_setup_all_rx_resources(adapter);
1431                igb_configure(adapter);
1432        }
1433
1434        igb_assign_vector(adapter->q_vector[0], 0);
1435
1436        if (adapter->flags & IGB_FLAG_HAS_MSI) {
1437                err = request_irq(pdev->irq, igb_intr_msi, 0,
1438                                  netdev->name, adapter);
1439                if (!err)
1440                        goto request_done;
1441
1442                /* fall back to legacy interrupts */
1443                igb_reset_interrupt_capability(adapter);
1444                adapter->flags &= ~IGB_FLAG_HAS_MSI;
1445        }
1446
1447        err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1448                          netdev->name, adapter);
1449
1450        if (err)
1451                dev_err(&pdev->dev, "Error %d getting interrupt\n",
1452                        err);
1453
1454request_done:
1455        return err;
1456}
1457
1458static void igb_free_irq(struct igb_adapter *adapter)
1459{
1460        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1461                int vector = 0, i;
1462
1463                free_irq(adapter->msix_entries[vector++].vector, adapter);
1464
1465                for (i = 0; i < adapter->num_q_vectors; i++)
1466                        free_irq(adapter->msix_entries[vector++].vector,
1467                                 adapter->q_vector[i]);
1468        } else {
1469                free_irq(adapter->pdev->irq, adapter);
1470        }
1471}
1472
1473/**
1474 *  igb_irq_disable - Mask off interrupt generation on the NIC
1475 *  @adapter: board private structure
1476 **/
1477static void igb_irq_disable(struct igb_adapter *adapter)
1478{
1479        struct e1000_hw *hw = &adapter->hw;
1480
1481        /* we need to be careful when disabling interrupts.  The VFs are also
1482         * mapped into these registers and so clearing the bits can cause
1483         * issues on the VF drivers so we only need to clear what we set
1484         */
1485        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1486                u32 regval = rd32(E1000_EIAM);
1487
1488                wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1489                wr32(E1000_EIMC, adapter->eims_enable_mask);
1490                regval = rd32(E1000_EIAC);
1491                wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1492        }
1493
1494        wr32(E1000_IAM, 0);
1495        wr32(E1000_IMC, ~0);
1496        wrfl();
1497        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1498                int i;
1499
1500                for (i = 0; i < adapter->num_q_vectors; i++)
1501                        synchronize_irq(adapter->msix_entries[i].vector);
1502        } else {
1503                synchronize_irq(adapter->pdev->irq);
1504        }
1505}
1506
1507/**
1508 *  igb_irq_enable - Enable default interrupt generation settings
1509 *  @adapter: board private structure
1510 **/
1511static void igb_irq_enable(struct igb_adapter *adapter)
1512{
1513        struct e1000_hw *hw = &adapter->hw;
1514
1515        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1516                u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1517                u32 regval = rd32(E1000_EIAC);
1518
1519                wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1520                regval = rd32(E1000_EIAM);
1521                wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1522                wr32(E1000_EIMS, adapter->eims_enable_mask);
1523                if (adapter->vfs_allocated_count) {
1524                        wr32(E1000_MBVFIMR, 0xFF);
1525                        ims |= E1000_IMS_VMMB;
1526                }
1527                wr32(E1000_IMS, ims);
1528        } else {
1529                wr32(E1000_IMS, IMS_ENABLE_MASK |
1530                                E1000_IMS_DRSTA);
1531                wr32(E1000_IAM, IMS_ENABLE_MASK |
1532                                E1000_IMS_DRSTA);
1533        }
1534}
1535
1536static void igb_update_mng_vlan(struct igb_adapter *adapter)
1537{
1538        struct e1000_hw *hw = &adapter->hw;
1539        u16 pf_id = adapter->vfs_allocated_count;
1540        u16 vid = adapter->hw.mng_cookie.vlan_id;
1541        u16 old_vid = adapter->mng_vlan_id;
1542
1543        if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1544                /* add VID to filter table */
1545                igb_vfta_set(hw, vid, pf_id, true, true);
1546                adapter->mng_vlan_id = vid;
1547        } else {
1548                adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1549        }
1550
1551        if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1552            (vid != old_vid) &&
1553            !test_bit(old_vid, adapter->active_vlans)) {
1554                /* remove VID from filter table */
1555                igb_vfta_set(hw, vid, pf_id, false, true);
1556        }
1557}
1558
1559/**
1560 *  igb_release_hw_control - release control of the h/w to f/w
1561 *  @adapter: address of board private structure
1562 *
1563 *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1564 *  For ASF and Pass Through versions of f/w this means that the
1565 *  driver is no longer loaded.
1566 **/
1567static void igb_release_hw_control(struct igb_adapter *adapter)
1568{
1569        struct e1000_hw *hw = &adapter->hw;
1570        u32 ctrl_ext;
1571
1572        /* Let firmware take over control of h/w */
1573        ctrl_ext = rd32(E1000_CTRL_EXT);
1574        wr32(E1000_CTRL_EXT,
1575                        ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1576}
1577
1578/**
1579 *  igb_get_hw_control - get control of the h/w from f/w
1580 *  @adapter: address of board private structure
1581 *
1582 *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1583 *  For ASF and Pass Through versions of f/w this means that
1584 *  the driver is loaded.
1585 **/
1586static void igb_get_hw_control(struct igb_adapter *adapter)
1587{
1588        struct e1000_hw *hw = &adapter->hw;
1589        u32 ctrl_ext;
1590
1591        /* Let firmware know the driver has taken over */
1592        ctrl_ext = rd32(E1000_CTRL_EXT);
1593        wr32(E1000_CTRL_EXT,
1594                        ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1595}
1596
1597static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1598{
1599        struct net_device *netdev = adapter->netdev;
1600        struct e1000_hw *hw = &adapter->hw;
1601
1602        WARN_ON(hw->mac.type != e1000_i210);
1603
1604        if (enable)
1605                adapter->flags |= IGB_FLAG_FQTSS;
1606        else
1607                adapter->flags &= ~IGB_FLAG_FQTSS;
1608
1609        if (netif_running(netdev))
1610                schedule_work(&adapter->reset_task);
1611}
1612
1613static bool is_fqtss_enabled(struct igb_adapter *adapter)
1614{
1615        return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1616}
1617
1618static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1619                                   enum tx_queue_prio prio)
1620{
1621        u32 val;
1622
1623        WARN_ON(hw->mac.type != e1000_i210);
1624        WARN_ON(queue < 0 || queue > 4);
1625
1626        val = rd32(E1000_I210_TXDCTL(queue));
1627
1628        if (prio == TX_QUEUE_PRIO_HIGH)
1629                val |= E1000_TXDCTL_PRIORITY;
1630        else
1631                val &= ~E1000_TXDCTL_PRIORITY;
1632
1633        wr32(E1000_I210_TXDCTL(queue), val);
1634}
1635
1636static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1637{
1638        u32 val;
1639
1640        WARN_ON(hw->mac.type != e1000_i210);
1641        WARN_ON(queue < 0 || queue > 1);
1642
1643        val = rd32(E1000_I210_TQAVCC(queue));
1644
1645        if (mode == QUEUE_MODE_STREAM_RESERVATION)
1646                val |= E1000_TQAVCC_QUEUEMODE;
1647        else
1648                val &= ~E1000_TQAVCC_QUEUEMODE;
1649
1650        wr32(E1000_I210_TQAVCC(queue), val);
1651}
1652
1653static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1654{
1655        int i;
1656
1657        for (i = 0; i < adapter->num_tx_queues; i++) {
1658                if (adapter->tx_ring[i]->cbs_enable)
1659                        return true;
1660        }
1661
1662        return false;
1663}
1664
1665static bool is_any_txtime_enabled(struct igb_adapter *adapter)
1666{
1667        int i;
1668
1669        for (i = 0; i < adapter->num_tx_queues; i++) {
1670                if (adapter->tx_ring[i]->launchtime_enable)
1671                        return true;
1672        }
1673
1674        return false;
1675}
1676
1677/**
1678 *  igb_config_tx_modes - Configure "Qav Tx mode" features on igb
1679 *  @adapter: pointer to adapter struct
1680 *  @queue: queue number
1681 *
1682 *  Configure CBS and Launchtime for a given hardware queue.
1683 *  Parameters are retrieved from the correct Tx ring, so
1684 *  igb_save_cbs_params() and igb_save_txtime_params() should be used
1685 *  for setting those correctly prior to this function being called.
1686 **/
1687static void igb_config_tx_modes(struct igb_adapter *adapter, int queue)
1688{
1689        struct igb_ring *ring = adapter->tx_ring[queue];
1690        struct net_device *netdev = adapter->netdev;
1691        struct e1000_hw *hw = &adapter->hw;
1692        u32 tqavcc, tqavctrl;
1693        u16 value;
1694
1695        WARN_ON(hw->mac.type != e1000_i210);
1696        WARN_ON(queue < 0 || queue > 1);
1697
1698        /* If any of the Qav features is enabled, configure queues as SR and
1699         * with HIGH PRIO. If none is, then configure them with LOW PRIO and
1700         * as SP.
1701         */
1702        if (ring->cbs_enable || ring->launchtime_enable) {
1703                set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1704                set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1705        } else {
1706                set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1707                set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1708        }
1709
1710        /* If CBS is enabled, set DataTranARB and config its parameters. */
1711        if (ring->cbs_enable || queue == 0) {
1712                /* i210 does not allow the queue 0 to be in the Strict
1713                 * Priority mode while the Qav mode is enabled, so,
1714                 * instead of disabling strict priority mode, we give
1715                 * queue 0 the maximum of credits possible.
1716                 *
1717                 * See section 8.12.19 of the i210 datasheet, "Note:
1718                 * Queue0 QueueMode must be set to 1b when
1719                 * TransmitMode is set to Qav."
1720                 */
1721                if (queue == 0 && !ring->cbs_enable) {
1722                        /* max "linkspeed" idleslope in kbps */
1723                        ring->idleslope = 1000000;
1724                        ring->hicredit = ETH_FRAME_LEN;
1725                }
1726
1727                /* Always set data transfer arbitration to credit-based
1728                 * shaper algorithm on TQAVCTRL if CBS is enabled for any of
1729                 * the queues.
1730                 */
1731                tqavctrl = rd32(E1000_I210_TQAVCTRL);
1732                tqavctrl |= E1000_TQAVCTRL_DATATRANARB;
1733                wr32(E1000_I210_TQAVCTRL, tqavctrl);
1734
1735                /* According to i210 datasheet section 7.2.7.7, we should set
1736                 * the 'idleSlope' field from TQAVCC register following the
1737                 * equation:
1738                 *
1739                 * For 100 Mbps link speed:
1740                 *
1741                 *     value = BW * 0x7735 * 0.2                          (E1)
1742                 *
1743                 * For 1000Mbps link speed:
1744                 *
1745                 *     value = BW * 0x7735 * 2                            (E2)
1746                 *
1747                 * E1 and E2 can be merged into one equation as shown below.
1748                 * Note that 'link-speed' is in Mbps.
1749                 *
1750                 *     value = BW * 0x7735 * 2 * link-speed
1751                 *                           --------------               (E3)
1752                 *                                1000
1753                 *
1754                 * 'BW' is the percentage bandwidth out of full link speed
1755                 * which can be found with the following equation. Note that
1756                 * idleSlope here is the parameter from this function which
1757                 * is in kbps.
1758                 *
1759                 *     BW =     idleSlope
1760                 *          -----------------                             (E4)
1761                 *          link-speed * 1000
1762                 *
1763                 * That said, we can come up with a generic equation to
1764                 * calculate the value we should set it TQAVCC register by
1765                 * replacing 'BW' in E3 by E4. The resulting equation is:
1766                 *
1767                 * value =     idleSlope     * 0x7735 * 2 * link-speed
1768                 *         -----------------            --------------    (E5)
1769                 *         link-speed * 1000                 1000
1770                 *
1771                 * 'link-speed' is present in both sides of the fraction so
1772                 * it is canceled out. The final equation is the following:
1773                 *
1774                 *     value = idleSlope * 61034
1775                 *             -----------------                          (E6)
1776                 *                  1000000
1777                 *
1778                 * NOTE: For i210, given the above, we can see that idleslope
1779                 *       is represented in 16.38431 kbps units by the value at
1780                 *       the TQAVCC register (1Gbps / 61034), which reduces
1781                 *       the granularity for idleslope increments.
1782                 *       For instance, if you want to configure a 2576kbps
1783                 *       idleslope, the value to be written on the register
1784                 *       would have to be 157.23. If rounded down, you end
1785                 *       up with less bandwidth available than originally
1786                 *       required (~2572 kbps). If rounded up, you end up
1787                 *       with a higher bandwidth (~2589 kbps). Below the
1788                 *       approach we take is to always round up the
1789                 *       calculated value, so the resulting bandwidth might
1790                 *       be slightly higher for some configurations.
1791                 */
1792                value = DIV_ROUND_UP_ULL(ring->idleslope * 61034ULL, 1000000);
1793
1794                tqavcc = rd32(E1000_I210_TQAVCC(queue));
1795                tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1796                tqavcc |= value;
1797                wr32(E1000_I210_TQAVCC(queue), tqavcc);
1798
1799                wr32(E1000_I210_TQAVHC(queue),
1800                     0x80000000 + ring->hicredit * 0x7735);
1801        } else {
1802
1803                /* Set idleSlope to zero. */
1804                tqavcc = rd32(E1000_I210_TQAVCC(queue));
1805                tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1806                wr32(E1000_I210_TQAVCC(queue), tqavcc);
1807
1808                /* Set hiCredit to zero. */
1809                wr32(E1000_I210_TQAVHC(queue), 0);
1810
1811                /* If CBS is not enabled for any queues anymore, then return to
1812                 * the default state of Data Transmission Arbitration on
1813                 * TQAVCTRL.
1814                 */
1815                if (!is_any_cbs_enabled(adapter)) {
1816                        tqavctrl = rd32(E1000_I210_TQAVCTRL);
1817                        tqavctrl &= ~E1000_TQAVCTRL_DATATRANARB;
1818                        wr32(E1000_I210_TQAVCTRL, tqavctrl);
1819                }
1820        }
1821
1822        /* If LaunchTime is enabled, set DataTranTIM. */
1823        if (ring->launchtime_enable) {
1824                /* Always set DataTranTIM on TQAVCTRL if LaunchTime is enabled
1825                 * for any of the SR queues, and configure fetchtime delta.
1826                 * XXX NOTE:
1827                 *     - LaunchTime will be enabled for all SR queues.
1828                 *     - A fixed offset can be added relative to the launch
1829                 *       time of all packets if configured at reg LAUNCH_OS0.
1830                 *       We are keeping it as 0 for now (default value).
1831                 */
1832                tqavctrl = rd32(E1000_I210_TQAVCTRL);
1833                tqavctrl |= E1000_TQAVCTRL_DATATRANTIM |
1834                       E1000_TQAVCTRL_FETCHTIME_DELTA;
1835                wr32(E1000_I210_TQAVCTRL, tqavctrl);
1836        } else {
1837                /* If Launchtime is not enabled for any SR queues anymore,
1838                 * then clear DataTranTIM on TQAVCTRL and clear fetchtime delta,
1839                 * effectively disabling Launchtime.
1840                 */
1841                if (!is_any_txtime_enabled(adapter)) {
1842                        tqavctrl = rd32(E1000_I210_TQAVCTRL);
1843                        tqavctrl &= ~E1000_TQAVCTRL_DATATRANTIM;
1844                        tqavctrl &= ~E1000_TQAVCTRL_FETCHTIME_DELTA;
1845                        wr32(E1000_I210_TQAVCTRL, tqavctrl);
1846                }
1847        }
1848
1849        /* XXX: In i210 controller the sendSlope and loCredit parameters from
1850         * CBS are not configurable by software so we don't do any 'controller
1851         * configuration' in respect to these parameters.
1852         */
1853
1854        netdev_dbg(netdev, "Qav Tx mode: cbs %s, launchtime %s, queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1855                   ring->cbs_enable ? "enabled" : "disabled",
1856                   ring->launchtime_enable ? "enabled" : "disabled",
1857                   queue,
1858                   ring->idleslope, ring->sendslope,
1859                   ring->hicredit, ring->locredit);
1860}
1861
1862static int igb_save_txtime_params(struct igb_adapter *adapter, int queue,
1863                                  bool enable)
1864{
1865        struct igb_ring *ring;
1866
1867        if (queue < 0 || queue > adapter->num_tx_queues)
1868                return -EINVAL;
1869
1870        ring = adapter->tx_ring[queue];
1871        ring->launchtime_enable = enable;
1872
1873        return 0;
1874}
1875
1876static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1877                               bool enable, int idleslope, int sendslope,
1878                               int hicredit, int locredit)
1879{
1880        struct igb_ring *ring;
1881
1882        if (queue < 0 || queue > adapter->num_tx_queues)
1883                return -EINVAL;
1884
1885        ring = adapter->tx_ring[queue];
1886
1887        ring->cbs_enable = enable;
1888        ring->idleslope = idleslope;
1889        ring->sendslope = sendslope;
1890        ring->hicredit = hicredit;
1891        ring->locredit = locredit;
1892
1893        return 0;
1894}
1895
1896/**
1897 *  igb_setup_tx_mode - Switch to/from Qav Tx mode when applicable
1898 *  @adapter: pointer to adapter struct
1899 *
1900 *  Configure TQAVCTRL register switching the controller's Tx mode
1901 *  if FQTSS mode is enabled or disabled. Additionally, will issue
1902 *  a call to igb_config_tx_modes() per queue so any previously saved
1903 *  Tx parameters are applied.
1904 **/
1905static void igb_setup_tx_mode(struct igb_adapter *adapter)
1906{
1907        struct net_device *netdev = adapter->netdev;
1908        struct e1000_hw *hw = &adapter->hw;
1909        u32 val;
1910
1911        /* Only i210 controller supports changing the transmission mode. */
1912        if (hw->mac.type != e1000_i210)
1913                return;
1914
1915        if (is_fqtss_enabled(adapter)) {
1916                int i, max_queue;
1917
1918                /* Configure TQAVCTRL register: set transmit mode to 'Qav',
1919                 * set data fetch arbitration to 'round robin', set SP_WAIT_SR
1920                 * so SP queues wait for SR ones.
1921                 */
1922                val = rd32(E1000_I210_TQAVCTRL);
1923                val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_SP_WAIT_SR;
1924                val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1925                wr32(E1000_I210_TQAVCTRL, val);
1926
1927                /* Configure Tx and Rx packet buffers sizes as described in
1928                 * i210 datasheet section 7.2.7.7.
1929                 */
1930                val = rd32(E1000_TXPBS);
1931                val &= ~I210_TXPBSIZE_MASK;
1932                val |= I210_TXPBSIZE_PB0_8KB | I210_TXPBSIZE_PB1_8KB |
1933                        I210_TXPBSIZE_PB2_4KB | I210_TXPBSIZE_PB3_4KB;
1934                wr32(E1000_TXPBS, val);
1935
1936                val = rd32(E1000_RXPBS);
1937                val &= ~I210_RXPBSIZE_MASK;
1938                val |= I210_RXPBSIZE_PB_30KB;
1939                wr32(E1000_RXPBS, val);
1940
1941                /* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1942                 * register should not exceed the buffer size programmed in
1943                 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1944                 * so according to the datasheet we should set MAX_TPKT_SIZE to
1945                 * 4kB / 64.
1946                 *
1947                 * However, when we do so, no frame from queue 2 and 3 are
1948                 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1949                 * or _equal_ to the buffer size programmed in TXPBS. For this
1950                 * reason, we set set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1951                 */
1952                val = (4096 - 1) / 64;
1953                wr32(E1000_I210_DTXMXPKTSZ, val);
1954
1955                /* Since FQTSS mode is enabled, apply any CBS configuration
1956                 * previously set. If no previous CBS configuration has been
1957                 * done, then the initial configuration is applied, which means
1958                 * CBS is disabled.
1959                 */
1960                max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1961                            adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1962
1963                for (i = 0; i < max_queue; i++) {
1964                        igb_config_tx_modes(adapter, i);
1965                }
1966        } else {
1967                wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1968                wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1969                wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1970
1971                val = rd32(E1000_I210_TQAVCTRL);
1972                /* According to Section 8.12.21, the other flags we've set when
1973                 * enabling FQTSS are not relevant when disabling FQTSS so we
1974                 * don't set they here.
1975                 */
1976                val &= ~E1000_TQAVCTRL_XMIT_MODE;
1977                wr32(E1000_I210_TQAVCTRL, val);
1978        }
1979
1980        netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1981                   "enabled" : "disabled");
1982}
1983
1984/**
1985 *  igb_configure - configure the hardware for RX and TX
1986 *  @adapter: private board structure
1987 **/
1988static void igb_configure(struct igb_adapter *adapter)
1989{
1990        struct net_device *netdev = adapter->netdev;
1991        int i;
1992
1993        igb_get_hw_control(adapter);
1994        igb_set_rx_mode(netdev);
1995        igb_setup_tx_mode(adapter);
1996
1997        igb_restore_vlan(adapter);
1998
1999        igb_setup_tctl(adapter);
2000        igb_setup_mrqc(adapter);
2001        igb_setup_rctl(adapter);
2002
2003        igb_nfc_filter_restore(adapter);
2004        igb_configure_tx(adapter);
2005        igb_configure_rx(adapter);
2006
2007        igb_rx_fifo_flush_82575(&adapter->hw);
2008
2009        /* call igb_desc_unused which always leaves
2010         * at least 1 descriptor unused to make sure
2011         * next_to_use != next_to_clean
2012         */
2013        for (i = 0; i < adapter->num_rx_queues; i++) {
2014                struct igb_ring *ring = adapter->rx_ring[i];
2015                igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
2016        }
2017}
2018
2019/**
2020 *  igb_power_up_link - Power up the phy/serdes link
2021 *  @adapter: address of board private structure
2022 **/
2023void igb_power_up_link(struct igb_adapter *adapter)
2024{
2025        igb_reset_phy(&adapter->hw);
2026
2027        if (adapter->hw.phy.media_type == e1000_media_type_copper)
2028                igb_power_up_phy_copper(&adapter->hw);
2029        else
2030                igb_power_up_serdes_link_82575(&adapter->hw);
2031
2032        igb_setup_link(&adapter->hw);
2033}
2034
2035/**
2036 *  igb_power_down_link - Power down the phy/serdes link
2037 *  @adapter: address of board private structure
2038 */
2039static void igb_power_down_link(struct igb_adapter *adapter)
2040{
2041        if (adapter->hw.phy.media_type == e1000_media_type_copper)
2042                igb_power_down_phy_copper_82575(&adapter->hw);
2043        else
2044                igb_shutdown_serdes_link_82575(&adapter->hw);
2045}
2046
2047/**
2048 * Detect and switch function for Media Auto Sense
2049 * @adapter: address of the board private structure
2050 **/
2051static void igb_check_swap_media(struct igb_adapter *adapter)
2052{
2053        struct e1000_hw *hw = &adapter->hw;
2054        u32 ctrl_ext, connsw;
2055        bool swap_now = false;
2056
2057        ctrl_ext = rd32(E1000_CTRL_EXT);
2058        connsw = rd32(E1000_CONNSW);
2059
2060        /* need to live swap if current media is copper and we have fiber/serdes
2061         * to go to.
2062         */
2063
2064        if ((hw->phy.media_type == e1000_media_type_copper) &&
2065            (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
2066                swap_now = true;
2067        } else if (!(connsw & E1000_CONNSW_SERDESD)) {
2068                /* copper signal takes time to appear */
2069                if (adapter->copper_tries < 4) {
2070                        adapter->copper_tries++;
2071                        connsw |= E1000_CONNSW_AUTOSENSE_CONF;
2072                        wr32(E1000_CONNSW, connsw);
2073                        return;
2074                } else {
2075                        adapter->copper_tries = 0;
2076                        if ((connsw & E1000_CONNSW_PHYSD) &&
2077                            (!(connsw & E1000_CONNSW_PHY_PDN))) {
2078                                swap_now = true;
2079                                connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
2080                                wr32(E1000_CONNSW, connsw);
2081                        }
2082                }
2083        }
2084
2085        if (!swap_now)
2086                return;
2087
2088        switch (hw->phy.media_type) {
2089        case e1000_media_type_copper:
2090                netdev_info(adapter->netdev,
2091                        "MAS: changing media to fiber/serdes\n");
2092                ctrl_ext |=
2093                        E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2094                adapter->flags |= IGB_FLAG_MEDIA_RESET;
2095                adapter->copper_tries = 0;
2096                break;
2097        case e1000_media_type_internal_serdes:
2098        case e1000_media_type_fiber:
2099                netdev_info(adapter->netdev,
2100                        "MAS: changing media to copper\n");
2101                ctrl_ext &=
2102                        ~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2103                adapter->flags |= IGB_FLAG_MEDIA_RESET;
2104                break;
2105        default:
2106                /* shouldn't get here during regular operation */
2107                netdev_err(adapter->netdev,
2108                        "AMS: Invalid media type found, returning\n");
2109                break;
2110        }
2111        wr32(E1000_CTRL_EXT, ctrl_ext);
2112}
2113
2114/**
2115 *  igb_up - Open the interface and prepare it to handle traffic
2116 *  @adapter: board private structure
2117 **/
2118int igb_up(struct igb_adapter *adapter)
2119{
2120        struct e1000_hw *hw = &adapter->hw;
2121        int i;
2122
2123        /* hardware has been reset, we need to reload some things */
2124        igb_configure(adapter);
2125
2126        clear_bit(__IGB_DOWN, &adapter->state);
2127
2128        for (i = 0; i < adapter->num_q_vectors; i++)
2129                napi_enable(&(adapter->q_vector[i]->napi));
2130
2131        if (adapter->flags & IGB_FLAG_HAS_MSIX)
2132                igb_configure_msix(adapter);
2133        else
2134                igb_assign_vector(adapter->q_vector[0], 0);
2135
2136        /* Clear any pending interrupts. */
2137        rd32(E1000_TSICR);
2138        rd32(E1000_ICR);
2139        igb_irq_enable(adapter);
2140
2141        /* notify VFs that reset has been completed */
2142        if (adapter->vfs_allocated_count) {
2143                u32 reg_data = rd32(E1000_CTRL_EXT);
2144
2145                reg_data |= E1000_CTRL_EXT_PFRSTD;
2146                wr32(E1000_CTRL_EXT, reg_data);
2147        }
2148
2149        netif_tx_start_all_queues(adapter->netdev);
2150
2151        /* start the watchdog. */
2152        hw->mac.get_link_status = 1;
2153        schedule_work(&adapter->watchdog_task);
2154
2155        if ((adapter->flags & IGB_FLAG_EEE) &&
2156            (!hw->dev_spec._82575.eee_disable))
2157                adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2158
2159        return 0;
2160}
2161
2162void igb_down(struct igb_adapter *adapter)
2163{
2164        struct net_device *netdev = adapter->netdev;
2165        struct e1000_hw *hw = &adapter->hw;
2166        u32 tctl, rctl;
2167        int i;
2168
2169        /* signal that we're down so the interrupt handler does not
2170         * reschedule our watchdog timer
2171         */
2172        set_bit(__IGB_DOWN, &adapter->state);
2173
2174        /* disable receives in the hardware */
2175        rctl = rd32(E1000_RCTL);
2176        wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2177        /* flush and sleep below */
2178
2179        igb_nfc_filter_exit(adapter);
2180
2181        netif_carrier_off(netdev);
2182        netif_tx_stop_all_queues(netdev);
2183
2184        /* disable transmits in the hardware */
2185        tctl = rd32(E1000_TCTL);
2186        tctl &= ~E1000_TCTL_EN;
2187        wr32(E1000_TCTL, tctl);
2188        /* flush both disables and wait for them to finish */
2189        wrfl();
2190        usleep_range(10000, 11000);
2191
2192        igb_irq_disable(adapter);
2193
2194        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2195
2196        for (i = 0; i < adapter->num_q_vectors; i++) {
2197                if (adapter->q_vector[i]) {
2198                        napi_synchronize(&adapter->q_vector[i]->napi);
2199                        napi_disable(&adapter->q_vector[i]->napi);
2200                }
2201        }
2202
2203        del_timer_sync(&adapter->watchdog_timer);
2204        del_timer_sync(&adapter->phy_info_timer);
2205
2206        /* record the stats before reset*/
2207        spin_lock(&adapter->stats64_lock);
2208        igb_update_stats(adapter);
2209        spin_unlock(&adapter->stats64_lock);
2210
2211        adapter->link_speed = 0;
2212        adapter->link_duplex = 0;
2213
2214        if (!pci_channel_offline(adapter->pdev))
2215                igb_reset(adapter);
2216
2217        /* clear VLAN promisc flag so VFTA will be updated if necessary */
2218        adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2219
2220        igb_clean_all_tx_rings(adapter);
2221        igb_clean_all_rx_rings(adapter);
2222#ifdef CONFIG_IGB_DCA
2223
2224        /* since we reset the hardware DCA settings were cleared */
2225        igb_setup_dca(adapter);
2226#endif
2227}
2228
2229void igb_reinit_locked(struct igb_adapter *adapter)
2230{
2231        WARN_ON(in_interrupt());
2232        while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2233                usleep_range(1000, 2000);
2234        igb_down(adapter);
2235        igb_up(adapter);
2236        clear_bit(__IGB_RESETTING, &adapter->state);
2237}
2238
2239/** igb_enable_mas - Media Autosense re-enable after swap
2240 *
2241 * @adapter: adapter struct
2242 **/
2243static void igb_enable_mas(struct igb_adapter *adapter)
2244{
2245        struct e1000_hw *hw = &adapter->hw;
2246        u32 connsw = rd32(E1000_CONNSW);
2247
2248        /* configure for SerDes media detect */
2249        if ((hw->phy.media_type == e1000_media_type_copper) &&
2250            (!(connsw & E1000_CONNSW_SERDESD))) {
2251                connsw |= E1000_CONNSW_ENRGSRC;
2252                connsw |= E1000_CONNSW_AUTOSENSE_EN;
2253                wr32(E1000_CONNSW, connsw);
2254                wrfl();
2255        }
2256}
2257
2258void igb_reset(struct igb_adapter *adapter)
2259{
2260        struct pci_dev *pdev = adapter->pdev;
2261        struct e1000_hw *hw = &adapter->hw;
2262        struct e1000_mac_info *mac = &hw->mac;
2263        struct e1000_fc_info *fc = &hw->fc;
2264        u32 pba, hwm;
2265
2266        /* Repartition Pba for greater than 9k mtu
2267         * To take effect CTRL.RST is required.
2268         */
2269        switch (mac->type) {
2270        case e1000_i350:
2271        case e1000_i354:
2272        case e1000_82580:
2273                pba = rd32(E1000_RXPBS);
2274                pba = igb_rxpbs_adjust_82580(pba);
2275                break;
2276        case e1000_82576:
2277                pba = rd32(E1000_RXPBS);
2278                pba &= E1000_RXPBS_SIZE_MASK_82576;
2279                break;
2280        case e1000_82575:
2281        case e1000_i210:
2282        case e1000_i211:
2283        default:
2284                pba = E1000_PBA_34K;
2285                break;
2286        }
2287
2288        if (mac->type == e1000_82575) {
2289                u32 min_rx_space, min_tx_space, needed_tx_space;
2290
2291                /* write Rx PBA so that hardware can report correct Tx PBA */
2292                wr32(E1000_PBA, pba);
2293
2294                /* To maintain wire speed transmits, the Tx FIFO should be
2295                 * large enough to accommodate two full transmit packets,
2296                 * rounded up to the next 1KB and expressed in KB.  Likewise,
2297                 * the Rx FIFO should be large enough to accommodate at least
2298                 * one full receive packet and is similarly rounded up and
2299                 * expressed in KB.
2300                 */
2301                min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2302
2303                /* The Tx FIFO also stores 16 bytes of information about the Tx
2304                 * but don't include Ethernet FCS because hardware appends it.
2305                 * We only need to round down to the nearest 512 byte block
2306                 * count since the value we care about is 2 frames, not 1.
2307                 */
2308                min_tx_space = adapter->max_frame_size;
2309                min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2310                min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2311
2312                /* upper 16 bits has Tx packet buffer allocation size in KB */
2313                needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2314
2315                /* If current Tx allocation is less than the min Tx FIFO size,
2316                 * and the min Tx FIFO size is less than the current Rx FIFO
2317                 * allocation, take space away from current Rx allocation.
2318                 */
2319                if (needed_tx_space < pba) {
2320                        pba -= needed_tx_space;
2321
2322                        /* if short on Rx space, Rx wins and must trump Tx
2323                         * adjustment
2324                         */
2325                        if (pba < min_rx_space)
2326                                pba = min_rx_space;
2327                }
2328
2329                /* adjust PBA for jumbo frames */
2330                wr32(E1000_PBA, pba);
2331        }
2332
2333        /* flow control settings
2334         * The high water mark must be low enough to fit one full frame
2335         * after transmitting the pause frame.  As such we must have enough
2336         * space to allow for us to complete our current transmit and then
2337         * receive the frame that is in progress from the link partner.
2338         * Set it to:
2339         * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2340         */
2341        hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2342
2343        fc->high_water = hwm & 0xFFFFFFF0;      /* 16-byte granularity */
2344        fc->low_water = fc->high_water - 16;
2345        fc->pause_time = 0xFFFF;
2346        fc->send_xon = 1;
2347        fc->current_mode = fc->requested_mode;
2348
2349        /* disable receive for all VFs and wait one second */
2350        if (adapter->vfs_allocated_count) {
2351                int i;
2352
2353                for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2354                        adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2355
2356                /* ping all the active vfs to let them know we are going down */
2357                igb_ping_all_vfs(adapter);
2358
2359                /* disable transmits and receives */
2360                wr32(E1000_VFRE, 0);
2361                wr32(E1000_VFTE, 0);
2362        }
2363
2364        /* Allow time for pending master requests to run */
2365        hw->mac.ops.reset_hw(hw);
2366        wr32(E1000_WUC, 0);
2367
2368        if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2369                /* need to resetup here after media swap */
2370                adapter->ei.get_invariants(hw);
2371                adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2372        }
2373        if ((mac->type == e1000_82575) &&
2374            (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2375                igb_enable_mas(adapter);
2376        }
2377        if (hw->mac.ops.init_hw(hw))
2378                dev_err(&pdev->dev, "Hardware Error\n");
2379
2380        /* RAR registers were cleared during init_hw, clear mac table */
2381        igb_flush_mac_table(adapter);
2382        __dev_uc_unsync(adapter->netdev, NULL);
2383
2384        /* Recover default RAR entry */
2385        igb_set_default_mac_filter(adapter);
2386
2387        /* Flow control settings reset on hardware reset, so guarantee flow
2388         * control is off when forcing speed.
2389         */
2390        if (!hw->mac.autoneg)
2391                igb_force_mac_fc(hw);
2392
2393        igb_init_dmac(adapter, pba);
2394#ifdef CONFIG_IGB_HWMON
2395        /* Re-initialize the thermal sensor on i350 devices. */
2396        if (!test_bit(__IGB_DOWN, &adapter->state)) {
2397                if (mac->type == e1000_i350 && hw->bus.func == 0) {
2398                        /* If present, re-initialize the external thermal sensor
2399                         * interface.
2400                         */
2401                        if (adapter->ets)
2402                                mac->ops.init_thermal_sensor_thresh(hw);
2403                }
2404        }
2405#endif
2406        /* Re-establish EEE setting */
2407        if (hw->phy.media_type == e1000_media_type_copper) {
2408                switch (mac->type) {
2409                case e1000_i350:
2410                case e1000_i210:
2411                case e1000_i211:
2412                        igb_set_eee_i350(hw, true, true);
2413                        break;
2414                case e1000_i354:
2415                        igb_set_eee_i354(hw, true, true);
2416                        break;
2417                default:
2418                        break;
2419                }
2420        }
2421        if (!netif_running(adapter->netdev))
2422                igb_power_down_link(adapter);
2423
2424        igb_update_mng_vlan(adapter);
2425
2426        /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2427        wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2428
2429        /* Re-enable PTP, where applicable. */
2430        if (adapter->ptp_flags & IGB_PTP_ENABLED)
2431                igb_ptp_reset(adapter);
2432
2433        igb_get_phy_info(hw);
2434}
2435
2436static netdev_features_t igb_fix_features(struct net_device *netdev,
2437        netdev_features_t features)
2438{
2439        /* Since there is no support for separate Rx/Tx vlan accel
2440         * enable/disable make sure Tx flag is always in same state as Rx.
2441         */
2442        if (features & NETIF_F_HW_VLAN_CTAG_RX)
2443                features |= NETIF_F_HW_VLAN_CTAG_TX;
2444        else
2445                features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2446
2447        return features;
2448}
2449
2450static int igb_set_features(struct net_device *netdev,
2451        netdev_features_t features)
2452{
2453        netdev_features_t changed = netdev->features ^ features;
2454        struct igb_adapter *adapter = netdev_priv(netdev);
2455
2456        if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2457                igb_vlan_mode(netdev, features);
2458
2459        if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2460                return 0;
2461
2462        if (!(features & NETIF_F_NTUPLE)) {
2463                struct hlist_node *node2;
2464                struct igb_nfc_filter *rule;
2465
2466                spin_lock(&adapter->nfc_lock);
2467                hlist_for_each_entry_safe(rule, node2,
2468                                          &adapter->nfc_filter_list, nfc_node) {
2469                        igb_erase_filter(adapter, rule);
2470                        hlist_del(&rule->nfc_node);
2471                        kfree(rule);
2472                }
2473                spin_unlock(&adapter->nfc_lock);
2474                adapter->nfc_filter_count = 0;
2475        }
2476
2477        netdev->features = features;
2478
2479        if (netif_running(netdev))
2480                igb_reinit_locked(adapter);
2481        else
2482                igb_reset(adapter);
2483
2484        return 1;
2485}
2486
2487static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2488                           struct net_device *dev,
2489                           const unsigned char *addr, u16 vid,
2490                           u16 flags,
2491                           struct netlink_ext_ack *extack)
2492{
2493        /* guarantee we can provide a unique filter for the unicast address */
2494        if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2495                struct igb_adapter *adapter = netdev_priv(dev);
2496                int vfn = adapter->vfs_allocated_count;
2497
2498                if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2499                        return -ENOMEM;
2500        }
2501
2502        return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2503}
2504
2505#define IGB_MAX_MAC_HDR_LEN     127
2506#define IGB_MAX_NETWORK_HDR_LEN 511
2507
2508static netdev_features_t
2509igb_features_check(struct sk_buff *skb, struct net_device *dev,
2510                   netdev_features_t features)
2511{
2512        unsigned int network_hdr_len, mac_hdr_len;
2513
2514        /* Make certain the headers can be described by a context descriptor */
2515        mac_hdr_len = skb_network_header(skb) - skb->data;
2516        if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2517                return features & ~(NETIF_F_HW_CSUM |
2518                                    NETIF_F_SCTP_CRC |
2519                                    NETIF_F_HW_VLAN_CTAG_TX |
2520                                    NETIF_F_TSO |
2521                                    NETIF_F_TSO6);
2522
2523        network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2524        if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2525                return features & ~(NETIF_F_HW_CSUM |
2526                                    NETIF_F_SCTP_CRC |
2527                                    NETIF_F_TSO |
2528                                    NETIF_F_TSO6);
2529
2530        /* We can only support IPV4 TSO in tunnels if we can mangle the
2531         * inner IP ID field, so strip TSO if MANGLEID is not supported.
2532         */
2533        if (skb->encapsulation && !(features & NETIF_F_TSO_MANGLEID))
2534                features &= ~NETIF_F_TSO;
2535
2536        return features;
2537}
2538
2539static void igb_offload_apply(struct igb_adapter *adapter, s32 queue)
2540{
2541        if (!is_fqtss_enabled(adapter)) {
2542                enable_fqtss(adapter, true);
2543                return;
2544        }
2545
2546        igb_config_tx_modes(adapter, queue);
2547
2548        if (!is_any_cbs_enabled(adapter) && !is_any_txtime_enabled(adapter))
2549                enable_fqtss(adapter, false);
2550}
2551
2552static int igb_offload_cbs(struct igb_adapter *adapter,
2553                           struct tc_cbs_qopt_offload *qopt)
2554{
2555        struct e1000_hw *hw = &adapter->hw;
2556        int err;
2557
2558        /* CBS offloading is only supported by i210 controller. */
2559        if (hw->mac.type != e1000_i210)
2560                return -EOPNOTSUPP;
2561
2562        /* CBS offloading is only supported by queue 0 and queue 1. */
2563        if (qopt->queue < 0 || qopt->queue > 1)
2564                return -EINVAL;
2565
2566        err = igb_save_cbs_params(adapter, qopt->queue, qopt->enable,
2567                                  qopt->idleslope, qopt->sendslope,
2568                                  qopt->hicredit, qopt->locredit);
2569        if (err)
2570                return err;
2571
2572        igb_offload_apply(adapter, qopt->queue);
2573
2574        return 0;
2575}
2576
2577#define ETHER_TYPE_FULL_MASK ((__force __be16)~0)
2578#define VLAN_PRIO_FULL_MASK (0x07)
2579
2580static int igb_parse_cls_flower(struct igb_adapter *adapter,
2581                                struct flow_cls_offload *f,
2582                                int traffic_class,
2583                                struct igb_nfc_filter *input)
2584{
2585        struct flow_rule *rule = flow_cls_offload_flow_rule(f);
2586        struct flow_dissector *dissector = rule->match.dissector;
2587        struct netlink_ext_ack *extack = f->common.extack;
2588
2589        if (dissector->used_keys &
2590            ~(BIT(FLOW_DISSECTOR_KEY_BASIC) |
2591              BIT(FLOW_DISSECTOR_KEY_CONTROL) |
2592              BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
2593              BIT(FLOW_DISSECTOR_KEY_VLAN))) {
2594                NL_SET_ERR_MSG_MOD(extack,
2595                                   "Unsupported key used, only BASIC, CONTROL, ETH_ADDRS and VLAN are supported");
2596                return -EOPNOTSUPP;
2597        }
2598
2599        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
2600                struct flow_match_eth_addrs match;
2601
2602                flow_rule_match_eth_addrs(rule, &match);
2603                if (!is_zero_ether_addr(match.mask->dst)) {
2604                        if (!is_broadcast_ether_addr(match.mask->dst)) {
2605                                NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for destination MAC address");
2606                                return -EINVAL;
2607                        }
2608
2609                        input->filter.match_flags |=
2610                                IGB_FILTER_FLAG_DST_MAC_ADDR;
2611                        ether_addr_copy(input->filter.dst_addr, match.key->dst);
2612                }
2613
2614                if (!is_zero_ether_addr(match.mask->src)) {
2615                        if (!is_broadcast_ether_addr(match.mask->src)) {
2616                                NL_SET_ERR_MSG_MOD(extack, "Only full masks are supported for source MAC address");
2617                                return -EINVAL;
2618                        }
2619
2620                        input->filter.match_flags |=
2621                                IGB_FILTER_FLAG_SRC_MAC_ADDR;
2622                        ether_addr_copy(input->filter.src_addr, match.key->src);
2623                }
2624        }
2625
2626        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
2627                struct flow_match_basic match;
2628
2629                flow_rule_match_basic(rule, &match);
2630                if (match.mask->n_proto) {
2631                        if (match.mask->n_proto != ETHER_TYPE_FULL_MASK) {
2632                                NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for EtherType filter");
2633                                return -EINVAL;
2634                        }
2635
2636                        input->filter.match_flags |= IGB_FILTER_FLAG_ETHER_TYPE;
2637                        input->filter.etype = match.key->n_proto;
2638                }
2639        }
2640
2641        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
2642                struct flow_match_vlan match;
2643
2644                flow_rule_match_vlan(rule, &match);
2645                if (match.mask->vlan_priority) {
2646                        if (match.mask->vlan_priority != VLAN_PRIO_FULL_MASK) {
2647                                NL_SET_ERR_MSG_MOD(extack, "Only full mask is supported for VLAN priority");
2648                                return -EINVAL;
2649                        }
2650
2651                        input->filter.match_flags |= IGB_FILTER_FLAG_VLAN_TCI;
2652                        input->filter.vlan_tci = match.key->vlan_priority;
2653                }
2654        }
2655
2656        input->action = traffic_class;
2657        input->cookie = f->cookie;
2658
2659        return 0;
2660}
2661
2662static int igb_configure_clsflower(struct igb_adapter *adapter,
2663                                   struct flow_cls_offload *cls_flower)
2664{
2665        struct netlink_ext_ack *extack = cls_flower->common.extack;
2666        struct igb_nfc_filter *filter, *f;
2667        int err, tc;
2668
2669        tc = tc_classid_to_hwtc(adapter->netdev, cls_flower->classid);
2670        if (tc < 0) {
2671                NL_SET_ERR_MSG_MOD(extack, "Invalid traffic class");
2672                return -EINVAL;
2673        }
2674
2675        filter = kzalloc(sizeof(*filter), GFP_KERNEL);
2676        if (!filter)
2677                return -ENOMEM;
2678
2679        err = igb_parse_cls_flower(adapter, cls_flower, tc, filter);
2680        if (err < 0)
2681                goto err_parse;
2682
2683        spin_lock(&adapter->nfc_lock);
2684
2685        hlist_for_each_entry(f, &adapter->nfc_filter_list, nfc_node) {
2686                if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2687                        err = -EEXIST;
2688                        NL_SET_ERR_MSG_MOD(extack,
2689                                           "This filter is already set in ethtool");
2690                        goto err_locked;
2691                }
2692        }
2693
2694        hlist_for_each_entry(f, &adapter->cls_flower_list, nfc_node) {
2695                if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2696                        err = -EEXIST;
2697                        NL_SET_ERR_MSG_MOD(extack,
2698                                           "This filter is already set in cls_flower");
2699                        goto err_locked;
2700                }
2701        }
2702
2703        err = igb_add_filter(adapter, filter);
2704        if (err < 0) {
2705                NL_SET_ERR_MSG_MOD(extack, "Could not add filter to the adapter");
2706                goto err_locked;
2707        }
2708
2709        hlist_add_head(&filter->nfc_node, &adapter->cls_flower_list);
2710
2711        spin_unlock(&adapter->nfc_lock);
2712
2713        return 0;
2714
2715err_locked:
2716        spin_unlock(&adapter->nfc_lock);
2717
2718err_parse:
2719        kfree(filter);
2720
2721        return err;
2722}
2723
2724static int igb_delete_clsflower(struct igb_adapter *adapter,
2725                                struct flow_cls_offload *cls_flower)
2726{
2727        struct igb_nfc_filter *filter;
2728        int err;
2729
2730        spin_lock(&adapter->nfc_lock);
2731
2732        hlist_for_each_entry(filter, &adapter->cls_flower_list, nfc_node)
2733                if (filter->cookie == cls_flower->cookie)
2734                        break;
2735
2736        if (!filter) {
2737                err = -ENOENT;
2738                goto out;
2739        }
2740
2741        err = igb_erase_filter(adapter, filter);
2742        if (err < 0)
2743                goto out;
2744
2745        hlist_del(&filter->nfc_node);
2746        kfree(filter);
2747
2748out:
2749        spin_unlock(&adapter->nfc_lock);
2750
2751        return err;
2752}
2753
2754static int igb_setup_tc_cls_flower(struct igb_adapter *adapter,
2755                                   struct flow_cls_offload *cls_flower)
2756{
2757        switch (cls_flower->command) {
2758        case FLOW_CLS_REPLACE:
2759                return igb_configure_clsflower(adapter, cls_flower);
2760        case FLOW_CLS_DESTROY:
2761                return igb_delete_clsflower(adapter, cls_flower);
2762        case FLOW_CLS_STATS:
2763                return -EOPNOTSUPP;
2764        default:
2765                return -EOPNOTSUPP;
2766        }
2767}
2768
2769static int igb_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
2770                                 void *cb_priv)
2771{
2772        struct igb_adapter *adapter = cb_priv;
2773
2774        if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
2775                return -EOPNOTSUPP;
2776
2777        switch (type) {
2778        case TC_SETUP_CLSFLOWER:
2779                return igb_setup_tc_cls_flower(adapter, type_data);
2780
2781        default:
2782                return -EOPNOTSUPP;
2783        }
2784}
2785
2786static int igb_offload_txtime(struct igb_adapter *adapter,
2787                              struct tc_etf_qopt_offload *qopt)
2788{
2789        struct e1000_hw *hw = &adapter->hw;
2790        int err;
2791
2792        /* Launchtime offloading is only supported by i210 controller. */
2793        if (hw->mac.type != e1000_i210)
2794                return -EOPNOTSUPP;
2795
2796        /* Launchtime offloading is only supported by queues 0 and 1. */
2797        if (qopt->queue < 0 || qopt->queue > 1)
2798                return -EINVAL;
2799
2800        err = igb_save_txtime_params(adapter, qopt->queue, qopt->enable);
2801        if (err)
2802                return err;
2803
2804        igb_offload_apply(adapter, qopt->queue);
2805
2806        return 0;
2807}
2808
2809static LIST_HEAD(igb_block_cb_list);
2810
2811static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2812                        void *type_data)
2813{
2814        struct igb_adapter *adapter = netdev_priv(dev);
2815
2816        switch (type) {
2817        case TC_SETUP_QDISC_CBS:
2818                return igb_offload_cbs(adapter, type_data);
2819        case TC_SETUP_BLOCK:
2820                return flow_block_cb_setup_simple(type_data,
2821                                                  &igb_block_cb_list,
2822                                                  igb_setup_tc_block_cb,
2823                                                  adapter, adapter, true);
2824
2825        case TC_SETUP_QDISC_ETF:
2826                return igb_offload_txtime(adapter, type_data);
2827
2828        default:
2829                return -EOPNOTSUPP;
2830        }
2831}
2832
2833static const struct net_device_ops igb_netdev_ops = {
2834        .ndo_open               = igb_open,
2835        .ndo_stop               = igb_close,
2836        .ndo_start_xmit         = igb_xmit_frame,
2837        .ndo_get_stats64        = igb_get_stats64,
2838        .ndo_set_rx_mode        = igb_set_rx_mode,
2839        .ndo_set_mac_address    = igb_set_mac,
2840        .ndo_change_mtu         = igb_change_mtu,
2841        .ndo_do_ioctl           = igb_ioctl,
2842        .ndo_tx_timeout         = igb_tx_timeout,
2843        .ndo_validate_addr      = eth_validate_addr,
2844        .ndo_vlan_rx_add_vid    = igb_vlan_rx_add_vid,
2845        .ndo_vlan_rx_kill_vid   = igb_vlan_rx_kill_vid,
2846        .ndo_set_vf_mac         = igb_ndo_set_vf_mac,
2847        .ndo_set_vf_vlan        = igb_ndo_set_vf_vlan,
2848        .ndo_set_vf_rate        = igb_ndo_set_vf_bw,
2849        .ndo_set_vf_spoofchk    = igb_ndo_set_vf_spoofchk,
2850        .ndo_set_vf_trust       = igb_ndo_set_vf_trust,
2851        .ndo_get_vf_config      = igb_ndo_get_vf_config,
2852        .ndo_fix_features       = igb_fix_features,
2853        .ndo_set_features       = igb_set_features,
2854        .ndo_fdb_add            = igb_ndo_fdb_add,
2855        .ndo_features_check     = igb_features_check,
2856        .ndo_setup_tc           = igb_setup_tc,
2857};
2858
2859/**
2860 * igb_set_fw_version - Configure version string for ethtool
2861 * @adapter: adapter struct
2862 **/
2863void igb_set_fw_version(struct igb_adapter *adapter)
2864{
2865        struct e1000_hw *hw = &adapter->hw;
2866        struct e1000_fw_version fw;
2867
2868        igb_get_fw_version(hw, &fw);
2869
2870        switch (hw->mac.type) {
2871        case e1000_i210:
2872        case e1000_i211:
2873                if (!(igb_get_flash_presence_i210(hw))) {
2874                        snprintf(adapter->fw_version,
2875                                 sizeof(adapter->fw_version),
2876                                 "%2d.%2d-%d",
2877                                 fw.invm_major, fw.invm_minor,
2878                                 fw.invm_img_type);
2879                        break;
2880                }
2881                /* fall through */
2882        default:
2883                /* if option is rom valid, display its version too */
2884                if (fw.or_valid) {
2885                        snprintf(adapter->fw_version,
2886                                 sizeof(adapter->fw_version),
2887                                 "%d.%d, 0x%08x, %d.%d.%d",
2888                                 fw.eep_major, fw.eep_minor, fw.etrack_id,
2889                                 fw.or_major, fw.or_build, fw.or_patch);
2890                /* no option rom */
2891                } else if (fw.etrack_id != 0X0000) {
2892                        snprintf(adapter->fw_version,
2893                            sizeof(adapter->fw_version),
2894                            "%d.%d, 0x%08x",
2895                            fw.eep_major, fw.eep_minor, fw.etrack_id);
2896                } else {
2897                snprintf(adapter->fw_version,
2898                    sizeof(adapter->fw_version),
2899                    "%d.%d.%d",
2900                    fw.eep_major, fw.eep_minor, fw.eep_build);
2901                }
2902                break;
2903        }
2904}
2905
2906/**
2907 * igb_init_mas - init Media Autosense feature if enabled in the NVM
2908 *
2909 * @adapter: adapter struct
2910 **/
2911static void igb_init_mas(struct igb_adapter *adapter)
2912{
2913        struct e1000_hw *hw = &adapter->hw;
2914        u16 eeprom_data;
2915
2916        hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
2917        switch (hw->bus.func) {
2918        case E1000_FUNC_0:
2919                if (eeprom_data & IGB_MAS_ENABLE_0) {
2920                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
2921                        netdev_info(adapter->netdev,
2922                                "MAS: Enabling Media Autosense for port %d\n",
2923                                hw->bus.func);
2924                }
2925                break;
2926        case E1000_FUNC_1:
2927                if (eeprom_data & IGB_MAS_ENABLE_1) {
2928                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
2929                        netdev_info(adapter->netdev,
2930                                "MAS: Enabling Media Autosense for port %d\n",
2931                                hw->bus.func);
2932                }
2933                break;
2934        case E1000_FUNC_2:
2935                if (eeprom_data & IGB_MAS_ENABLE_2) {
2936                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
2937                        netdev_info(adapter->netdev,
2938                                "MAS: Enabling Media Autosense for port %d\n",
2939                                hw->bus.func);
2940                }
2941                break;
2942        case E1000_FUNC_3:
2943                if (eeprom_data & IGB_MAS_ENABLE_3) {
2944                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
2945                        netdev_info(adapter->netdev,
2946                                "MAS: Enabling Media Autosense for port %d\n",
2947                                hw->bus.func);
2948                }
2949                break;
2950        default:
2951                /* Shouldn't get here */
2952                netdev_err(adapter->netdev,
2953                        "MAS: Invalid port configuration, returning\n");
2954                break;
2955        }
2956}
2957
2958/**
2959 *  igb_init_i2c - Init I2C interface
2960 *  @adapter: pointer to adapter structure
2961 **/
2962static s32 igb_init_i2c(struct igb_adapter *adapter)
2963{
2964        s32 status = 0;
2965
2966        /* I2C interface supported on i350 devices */
2967        if (adapter->hw.mac.type != e1000_i350)
2968                return 0;
2969
2970        /* Initialize the i2c bus which is controlled by the registers.
2971         * This bus will use the i2c_algo_bit structue that implements
2972         * the protocol through toggling of the 4 bits in the register.
2973         */
2974        adapter->i2c_adap.owner = THIS_MODULE;
2975        adapter->i2c_algo = igb_i2c_algo;
2976        adapter->i2c_algo.data = adapter;
2977        adapter->i2c_adap.algo_data = &adapter->i2c_algo;
2978        adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
2979        strlcpy(adapter->i2c_adap.name, "igb BB",
2980                sizeof(adapter->i2c_adap.name));
2981        status = i2c_bit_add_bus(&adapter->i2c_adap);
2982        return status;
2983}
2984
2985/**
2986 *  igb_probe - Device Initialization Routine
2987 *  @pdev: PCI device information struct
2988 *  @ent: entry in igb_pci_tbl
2989 *
2990 *  Returns 0 on success, negative on failure
2991 *
2992 *  igb_probe initializes an adapter identified by a pci_dev structure.
2993 *  The OS initialization, configuring of the adapter private structure,
2994 *  and a hardware reset occur.
2995 **/
2996static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2997{
2998        struct net_device *netdev;
2999        struct igb_adapter *adapter;
3000        struct e1000_hw *hw;
3001        u16 eeprom_data = 0;
3002        s32 ret_val;
3003        static int global_quad_port_a; /* global quad port a indication */
3004        const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
3005        int err, pci_using_dac;
3006        u8 part_str[E1000_PBANUM_LENGTH];
3007
3008        /* Catch broken hardware that put the wrong VF device ID in
3009         * the PCIe SR-IOV capability.
3010         */
3011        if (pdev->is_virtfn) {
3012                WARN(1, KERN_ERR "%s (%hx:%hx) should not be a VF!\n",
3013                        pci_name(pdev), pdev->vendor, pdev->device);
3014                return -EINVAL;
3015        }
3016
3017        err = pci_enable_device_mem(pdev);
3018        if (err)
3019                return err;
3020
3021        pci_using_dac = 0;
3022        err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
3023        if (!err) {
3024                pci_using_dac = 1;
3025        } else {
3026                err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
3027                if (err) {
3028                        dev_err(&pdev->dev,
3029                                "No usable DMA configuration, aborting\n");
3030                        goto err_dma;
3031                }
3032        }
3033
3034        err = pci_request_mem_regions(pdev, igb_driver_name);
3035        if (err)
3036                goto err_pci_reg;
3037
3038        pci_enable_pcie_error_reporting(pdev);
3039
3040        pci_set_master(pdev);
3041        pci_save_state(pdev);
3042
3043        err = -ENOMEM;
3044        netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
3045                                   IGB_MAX_TX_QUEUES);
3046        if (!netdev)
3047                goto err_alloc_etherdev;
3048
3049        SET_NETDEV_DEV(netdev, &pdev->dev);
3050
3051        pci_set_drvdata(pdev, netdev);
3052        adapter = netdev_priv(netdev);
3053        adapter->netdev = netdev;
3054        adapter->pdev = pdev;
3055        hw = &adapter->hw;
3056        hw->back = adapter;
3057        adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
3058
3059        err = -EIO;
3060        adapter->io_addr = pci_iomap(pdev, 0, 0);
3061        if (!adapter->io_addr)
3062                goto err_ioremap;
3063        /* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
3064        hw->hw_addr = adapter->io_addr;
3065
3066        netdev->netdev_ops = &igb_netdev_ops;
3067        igb_set_ethtool_ops(netdev);
3068        netdev->watchdog_timeo = 5 * HZ;
3069
3070        strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
3071
3072        netdev->mem_start = pci_resource_start(pdev, 0);
3073        netdev->mem_end = pci_resource_end(pdev, 0);
3074
3075        /* PCI config space info */
3076        hw->vendor_id = pdev->vendor;
3077        hw->device_id = pdev->device;
3078        hw->revision_id = pdev->revision;
3079        hw->subsystem_vendor_id = pdev->subsystem_vendor;
3080        hw->subsystem_device_id = pdev->subsystem_device;
3081
3082        /* Copy the default MAC, PHY and NVM function pointers */
3083        memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
3084        memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
3085        memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
3086        /* Initialize skew-specific constants */
3087        err = ei->get_invariants(hw);
3088        if (err)
3089                goto err_sw_init;
3090
3091        /* setup the private structure */
3092        err = igb_sw_init(adapter);
3093        if (err)
3094                goto err_sw_init;
3095
3096        igb_get_bus_info_pcie(hw);
3097
3098        hw->phy.autoneg_wait_to_complete = false;
3099
3100        /* Copper options */
3101        if (hw->phy.media_type == e1000_media_type_copper) {
3102                hw->phy.mdix = AUTO_ALL_MODES;
3103                hw->phy.disable_polarity_correction = false;
3104                hw->phy.ms_type = e1000_ms_hw_default;
3105        }
3106
3107        if (igb_check_reset_block(hw))
3108                dev_info(&pdev->dev,
3109                        "PHY reset is blocked due to SOL/IDER session.\n");
3110
3111        /* features is initialized to 0 in allocation, it might have bits
3112         * set by igb_sw_init so we should use an or instead of an
3113         * assignment.
3114         */
3115        netdev->features |= NETIF_F_SG |
3116                            NETIF_F_TSO |
3117                            NETIF_F_TSO6 |
3118                            NETIF_F_RXHASH |
3119                            NETIF_F_RXCSUM |
3120                            NETIF_F_HW_CSUM;
3121
3122        if (hw->mac.type >= e1000_82576)
3123                netdev->features |= NETIF_F_SCTP_CRC;
3124
3125        if (hw->mac.type >= e1000_i350)
3126                netdev->features |= NETIF_F_HW_TC;
3127
3128#define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
3129                                  NETIF_F_GSO_GRE_CSUM | \
3130                                  NETIF_F_GSO_IPXIP4 | \
3131                                  NETIF_F_GSO_IPXIP6 | \
3132                                  NETIF_F_GSO_UDP_TUNNEL | \
3133                                  NETIF_F_GSO_UDP_TUNNEL_CSUM)
3134
3135        netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
3136        netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
3137
3138        /* copy netdev features into list of user selectable features */
3139        netdev->hw_features |= netdev->features |
3140                               NETIF_F_HW_VLAN_CTAG_RX |
3141                               NETIF_F_HW_VLAN_CTAG_TX |
3142                               NETIF_F_RXALL;
3143
3144        if (hw->mac.type >= e1000_i350)
3145                netdev->hw_features |= NETIF_F_NTUPLE;
3146
3147        if (pci_using_dac)
3148                netdev->features |= NETIF_F_HIGHDMA;
3149
3150        netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
3151        netdev->mpls_features |= NETIF_F_HW_CSUM;
3152        netdev->hw_enc_features |= netdev->vlan_features;
3153
3154        /* set this bit last since it cannot be part of vlan_features */
3155        netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
3156                            NETIF_F_HW_VLAN_CTAG_RX |
3157                            NETIF_F_HW_VLAN_CTAG_TX;
3158
3159        netdev->priv_flags |= IFF_SUPP_NOFCS;
3160
3161        netdev->priv_flags |= IFF_UNICAST_FLT;
3162
3163        /* MTU range: 68 - 9216 */
3164        netdev->min_mtu = ETH_MIN_MTU;
3165        netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
3166
3167        adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
3168
3169        /* before reading the NVM, reset the controller to put the device in a
3170         * known good starting state
3171         */
3172        hw->mac.ops.reset_hw(hw);
3173
3174        /* make sure the NVM is good , i211/i210 parts can have special NVM
3175         * that doesn't contain a checksum
3176         */
3177        switch (hw->mac.type) {
3178        case e1000_i210:
3179        case e1000_i211:
3180                if (igb_get_flash_presence_i210(hw)) {
3181                        if (hw->nvm.ops.validate(hw) < 0) {
3182                                dev_err(&pdev->dev,
3183                                        "The NVM Checksum Is Not Valid\n");
3184                                err = -EIO;
3185                                goto err_eeprom;
3186                        }
3187                }
3188                break;
3189        default:
3190                if (hw->nvm.ops.validate(hw) < 0) {
3191                        dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
3192                        err = -EIO;
3193                        goto err_eeprom;
3194                }
3195                break;
3196        }
3197
3198        if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
3199                /* copy the MAC address out of the NVM */
3200                if (hw->mac.ops.read_mac_addr(hw))
3201                        dev_err(&pdev->dev, "NVM Read Error\n");
3202        }
3203
3204        memcpy(netdev->dev_addr, hw->mac.addr, netdev->addr_len);
3205
3206        if (!is_valid_ether_addr(netdev->dev_addr)) {
3207                dev_err(&pdev->dev, "Invalid MAC Address\n");
3208                err = -EIO;
3209                goto err_eeprom;
3210        }
3211
3212        igb_set_default_mac_filter(adapter);
3213
3214        /* get firmware version for ethtool -i */
3215        igb_set_fw_version(adapter);
3216
3217        /* configure RXPBSIZE and TXPBSIZE */
3218        if (hw->mac.type == e1000_i210) {
3219                wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
3220                wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
3221        }
3222
3223        timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
3224        timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
3225
3226        INIT_WORK(&adapter->reset_task, igb_reset_task);
3227        INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
3228
3229        /* Initialize link properties that are user-changeable */
3230        adapter->fc_autoneg = true;
3231        hw->mac.autoneg = true;
3232        hw->phy.autoneg_advertised = 0x2f;
3233
3234        hw->fc.requested_mode = e1000_fc_default;
3235        hw->fc.current_mode = e1000_fc_default;
3236
3237        igb_validate_mdi_setting(hw);
3238
3239        /* By default, support wake on port A */
3240        if (hw->bus.func == 0)
3241                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3242
3243        /* Check the NVM for wake support on non-port A ports */
3244        if (hw->mac.type >= e1000_82580)
3245                hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
3246                                 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
3247                                 &eeprom_data);
3248        else if (hw->bus.func == 1)
3249                hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3250
3251        if (eeprom_data & IGB_EEPROM_APME)
3252                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3253
3254        /* now that we have the eeprom settings, apply the special cases where
3255         * the eeprom may be wrong or the board simply won't support wake on
3256         * lan on a particular port
3257         */
3258        switch (pdev->device) {
3259        case E1000_DEV_ID_82575GB_QUAD_COPPER:
3260                adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3261                break;
3262        case E1000_DEV_ID_82575EB_FIBER_SERDES:
3263        case E1000_DEV_ID_82576_FIBER:
3264        case E1000_DEV_ID_82576_SERDES:
3265                /* Wake events only supported on port A for dual fiber
3266                 * regardless of eeprom setting
3267                 */
3268                if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
3269                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3270                break;
3271        case E1000_DEV_ID_82576_QUAD_COPPER:
3272        case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
3273                /* if quad port adapter, disable WoL on all but port A */
3274                if (global_quad_port_a != 0)
3275                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3276                else
3277                        adapter->flags |= IGB_FLAG_QUAD_PORT_A;
3278                /* Reset for multiple quad port adapters */
3279                if (++global_quad_port_a == 4)
3280                        global_quad_port_a = 0;
3281                break;
3282        default:
3283                /* If the device can't wake, don't set software support */
3284                if (!device_can_wakeup(&adapter->pdev->dev))
3285                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3286        }
3287
3288        /* initialize the wol settings based on the eeprom settings */
3289        if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
3290                adapter->wol |= E1000_WUFC_MAG;
3291
3292        /* Some vendors want WoL disabled by default, but still supported */
3293        if ((hw->mac.type == e1000_i350) &&
3294            (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
3295                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3296                adapter->wol = 0;
3297        }
3298
3299        /* Some vendors want the ability to Use the EEPROM setting as
3300         * enable/disable only, and not for capability
3301         */
3302        if (((hw->mac.type == e1000_i350) ||
3303             (hw->mac.type == e1000_i354)) &&
3304            (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
3305                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3306                adapter->wol = 0;
3307        }
3308        if (hw->mac.type == e1000_i350) {
3309                if (((pdev->subsystem_device == 0x5001) ||
3310                     (pdev->subsystem_device == 0x5002)) &&
3311                                (hw->bus.func == 0)) {
3312                        adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3313                        adapter->wol = 0;
3314                }
3315                if (pdev->subsystem_device == 0x1F52)
3316                        adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3317        }
3318
3319        device_set_wakeup_enable(&adapter->pdev->dev,
3320                                 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3321
3322        /* reset the hardware with the new settings */
3323        igb_reset(adapter);
3324
3325        /* Init the I2C interface */
3326        err = igb_init_i2c(adapter);
3327        if (err) {
3328                dev_err(&pdev->dev, "failed to init i2c interface\n");
3329                goto err_eeprom;
3330        }
3331
3332        /* let the f/w know that the h/w is now under the control of the
3333         * driver.
3334         */
3335        igb_get_hw_control(adapter);
3336
3337        strcpy(netdev->name, "eth%d");
3338        err = register_netdev(netdev);
3339        if (err)
3340                goto err_register;
3341
3342        /* carrier off reporting is important to ethtool even BEFORE open */
3343        netif_carrier_off(netdev);
3344
3345#ifdef CONFIG_IGB_DCA
3346        if (dca_add_requester(&pdev->dev) == 0) {
3347                adapter->flags |= IGB_FLAG_DCA_ENABLED;
3348                dev_info(&pdev->dev, "DCA enabled\n");
3349                igb_setup_dca(adapter);
3350        }
3351
3352#endif
3353#ifdef CONFIG_IGB_HWMON
3354        /* Initialize the thermal sensor on i350 devices. */
3355        if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3356                u16 ets_word;
3357
3358                /* Read the NVM to determine if this i350 device supports an
3359                 * external thermal sensor.
3360                 */
3361                hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3362                if (ets_word != 0x0000 && ets_word != 0xFFFF)
3363                        adapter->ets = true;
3364                else
3365                        adapter->ets = false;
3366                if (igb_sysfs_init(adapter))
3367                        dev_err(&pdev->dev,
3368                                "failed to allocate sysfs resources\n");
3369        } else {
3370                adapter->ets = false;
3371        }
3372#endif
3373        /* Check if Media Autosense is enabled */
3374        adapter->ei = *ei;
3375        if (hw->dev_spec._82575.mas_capable)
3376                igb_init_mas(adapter);
3377
3378        /* do hw tstamp init after resetting */
3379        igb_ptp_init(adapter);
3380
3381        dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3382        /* print bus type/speed/width info, not applicable to i354 */
3383        if (hw->mac.type != e1000_i354) {
3384                dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3385                         netdev->name,
3386                         ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3387                          (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3388                           "unknown"),
3389                         ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3390                          "Width x4" :
3391                          (hw->bus.width == e1000_bus_width_pcie_x2) ?
3392                          "Width x2" :
3393                          (hw->bus.width == e1000_bus_width_pcie_x1) ?
3394                          "Width x1" : "unknown"), netdev->dev_addr);
3395        }
3396
3397        if ((hw->mac.type >= e1000_i210 ||
3398             igb_get_flash_presence_i210(hw))) {
3399                ret_val = igb_read_part_string(hw, part_str,
3400                                               E1000_PBANUM_LENGTH);
3401        } else {
3402                ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3403        }
3404
3405        if (ret_val)
3406                strcpy(part_str, "Unknown");
3407        dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3408        dev_info(&pdev->dev,
3409                "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3410                (adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3411                (adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3412                adapter->num_rx_queues, adapter->num_tx_queues);
3413        if (hw->phy.media_type == e1000_media_type_copper) {
3414                switch (hw->mac.type) {
3415                case e1000_i350:
3416                case e1000_i210:
3417                case e1000_i211:
3418                        /* Enable EEE for internal copper PHY devices */
3419                        err = igb_set_eee_i350(hw, true, true);
3420                        if ((!err) &&
3421                            (!hw->dev_spec._82575.eee_disable)) {
3422                                adapter->eee_advert =
3423                                        MDIO_EEE_100TX | MDIO_EEE_1000T;
3424                                adapter->flags |= IGB_FLAG_EEE;
3425                        }
3426                        break;
3427                case e1000_i354:
3428                        if ((rd32(E1000_CTRL_EXT) &
3429                            E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3430                                err = igb_set_eee_i354(hw, true, true);
3431                                if ((!err) &&
3432                                        (!hw->dev_spec._82575.eee_disable)) {
3433                                        adapter->eee_advert =
3434                                           MDIO_EEE_100TX | MDIO_EEE_1000T;
3435                                        adapter->flags |= IGB_FLAG_EEE;
3436                                }
3437                        }
3438                        break;
3439                default:
3440                        break;
3441                }
3442        }
3443
3444        dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NEVER_SKIP);
3445
3446        pm_runtime_put_noidle(&pdev->dev);
3447        return 0;
3448
3449err_register:
3450        igb_release_hw_control(adapter);
3451        memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3452err_eeprom:
3453        if (!igb_check_reset_block(hw))
3454                igb_reset_phy(hw);
3455
3456        if (hw->flash_address)
3457                iounmap(hw->flash_address);
3458err_sw_init:
3459        kfree(adapter->mac_table);
3460        kfree(adapter->shadow_vfta);
3461        igb_clear_interrupt_scheme(adapter);
3462#ifdef CONFIG_PCI_IOV
3463        igb_disable_sriov(pdev);
3464#endif
3465        pci_iounmap(pdev, adapter->io_addr);
3466err_ioremap:
3467        free_netdev(netdev);
3468err_alloc_etherdev:
3469        pci_release_mem_regions(pdev);
3470err_pci_reg:
3471err_dma:
3472        pci_disable_device(pdev);
3473        return err;
3474}
3475
3476#ifdef CONFIG_PCI_IOV
3477static int igb_disable_sriov(struct pci_dev *pdev)
3478{
3479        struct net_device *netdev = pci_get_drvdata(pdev);
3480        struct igb_adapter *adapter = netdev_priv(netdev);
3481        struct e1000_hw *hw = &adapter->hw;
3482
3483        /* reclaim resources allocated to VFs */
3484        if (adapter->vf_data) {
3485                /* disable iov and allow time for transactions to clear */
3486                if (pci_vfs_assigned(pdev)) {
3487                        dev_warn(&pdev->dev,
3488                                 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3489                        return -EPERM;
3490                } else {
3491                        pci_disable_sriov(pdev);
3492                        msleep(500);
3493                }
3494
3495                kfree(adapter->vf_mac_list);
3496                adapter->vf_mac_list = NULL;
3497                kfree(adapter->vf_data);
3498                adapter->vf_data = NULL;
3499                adapter->vfs_allocated_count = 0;
3500                wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3501                wrfl();
3502                msleep(100);
3503                dev_info(&pdev->dev, "IOV Disabled\n");
3504
3505                /* Re-enable DMA Coalescing flag since IOV is turned off */
3506                adapter->flags |= IGB_FLAG_DMAC;
3507        }
3508
3509        return 0;
3510}
3511
3512static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3513{
3514        struct net_device *netdev = pci_get_drvdata(pdev);
3515        struct igb_adapter *adapter = netdev_priv(netdev);
3516        int old_vfs = pci_num_vf(pdev);
3517        struct vf_mac_filter *mac_list;
3518        int err = 0;
3519        int num_vf_mac_filters, i;
3520
3521        if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3522                err = -EPERM;
3523                goto out;
3524        }
3525        if (!num_vfs)
3526                goto out;
3527
3528        if (old_vfs) {
3529                dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3530                         old_vfs, max_vfs);
3531                adapter->vfs_allocated_count = old_vfs;
3532        } else
3533                adapter->vfs_allocated_count = num_vfs;
3534
3535        adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3536                                sizeof(struct vf_data_storage), GFP_KERNEL);
3537
3538        /* if allocation failed then we do not support SR-IOV */
3539        if (!adapter->vf_data) {
3540                adapter->vfs_allocated_count = 0;
3541                err = -ENOMEM;
3542                goto out;
3543        }
3544
3545        /* Due to the limited number of RAR entries calculate potential
3546         * number of MAC filters available for the VFs. Reserve entries
3547         * for PF default MAC, PF MAC filters and at least one RAR entry
3548         * for each VF for VF MAC.
3549         */
3550        num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3551                             (1 + IGB_PF_MAC_FILTERS_RESERVED +
3552                              adapter->vfs_allocated_count);
3553
3554        adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3555                                       sizeof(struct vf_mac_filter),
3556                                       GFP_KERNEL);
3557
3558        mac_list = adapter->vf_mac_list;
3559        INIT_LIST_HEAD(&adapter->vf_macs.l);
3560
3561        if (adapter->vf_mac_list) {
3562                /* Initialize list of VF MAC filters */
3563                for (i = 0; i < num_vf_mac_filters; i++) {
3564                        mac_list->vf = -1;
3565                        mac_list->free = true;
3566                        list_add(&mac_list->l, &adapter->vf_macs.l);
3567                        mac_list++;
3568                }
3569        } else {
3570                /* If we could not allocate memory for the VF MAC filters
3571                 * we can continue without this feature but warn user.
3572                 */
3573                dev_err(&pdev->dev,
3574                        "Unable to allocate memory for VF MAC filter list\n");
3575        }
3576
3577        /* only call pci_enable_sriov() if no VFs are allocated already */
3578        if (!old_vfs) {
3579                err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3580                if (err)
3581                        goto err_out;
3582        }
3583        dev_info(&pdev->dev, "%d VFs allocated\n",
3584                 adapter->vfs_allocated_count);
3585        for (i = 0; i < adapter->vfs_allocated_count; i++)
3586                igb_vf_configure(adapter, i);
3587
3588        /* DMA Coalescing is not supported in IOV mode. */
3589        adapter->flags &= ~IGB_FLAG_DMAC;
3590        goto out;
3591
3592err_out:
3593        kfree(adapter->vf_mac_list);
3594        adapter->vf_mac_list = NULL;
3595        kfree(adapter->vf_data);
3596        adapter->vf_data = NULL;
3597        adapter->vfs_allocated_count = 0;
3598out:
3599        return err;
3600}
3601
3602#endif
3603/**
3604 *  igb_remove_i2c - Cleanup  I2C interface
3605 *  @adapter: pointer to adapter structure
3606 **/
3607static void igb_remove_i2c(struct igb_adapter *adapter)
3608{
3609        /* free the adapter bus structure */
3610        i2c_del_adapter(&adapter->i2c_adap);
3611}
3612
3613/**
3614 *  igb_remove - Device Removal Routine
3615 *  @pdev: PCI device information struct
3616 *
3617 *  igb_remove is called by the PCI subsystem to alert the driver
3618 *  that it should release a PCI device.  The could be caused by a
3619 *  Hot-Plug event, or because the driver is going to be removed from
3620 *  memory.
3621 **/
3622static void igb_remove(struct pci_dev *pdev)
3623{
3624        struct net_device *netdev = pci_get_drvdata(pdev);
3625        struct igb_adapter *adapter = netdev_priv(netdev);
3626        struct e1000_hw *hw = &adapter->hw;
3627
3628        pm_runtime_get_noresume(&pdev->dev);
3629#ifdef CONFIG_IGB_HWMON
3630        igb_sysfs_exit(adapter);
3631#endif
3632        igb_remove_i2c(adapter);
3633        igb_ptp_stop(adapter);
3634        /* The watchdog timer may be rescheduled, so explicitly
3635         * disable watchdog from being rescheduled.
3636         */
3637        set_bit(__IGB_DOWN, &adapter->state);
3638        del_timer_sync(&adapter->watchdog_timer);
3639        del_timer_sync(&adapter->phy_info_timer);
3640
3641        cancel_work_sync(&adapter->reset_task);
3642        cancel_work_sync(&adapter->watchdog_task);
3643
3644#ifdef CONFIG_IGB_DCA
3645        if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3646                dev_info(&pdev->dev, "DCA disabled\n");
3647                dca_remove_requester(&pdev->dev);
3648                adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3649                wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3650        }
3651#endif
3652
3653        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
3654         * would have already happened in close and is redundant.
3655         */
3656        igb_release_hw_control(adapter);
3657
3658#ifdef CONFIG_PCI_IOV
3659        igb_disable_sriov(pdev);
3660#endif
3661
3662        unregister_netdev(netdev);
3663
3664        igb_clear_interrupt_scheme(adapter);
3665
3666        pci_iounmap(pdev, adapter->io_addr);
3667        if (hw->flash_address)
3668                iounmap(hw->flash_address);
3669        pci_release_mem_regions(pdev);
3670
3671        kfree(adapter->mac_table);
3672        kfree(adapter->shadow_vfta);
3673        free_netdev(netdev);
3674
3675        pci_disable_pcie_error_reporting(pdev);
3676
3677        pci_disable_device(pdev);
3678}
3679
3680/**
3681 *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3682 *  @adapter: board private structure to initialize
3683 *
3684 *  This function initializes the vf specific data storage and then attempts to
3685 *  allocate the VFs.  The reason for ordering it this way is because it is much
3686 *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3687 *  the memory for the VFs.
3688 **/
3689static void igb_probe_vfs(struct igb_adapter *adapter)
3690{
3691#ifdef CONFIG_PCI_IOV
3692        struct pci_dev *pdev = adapter->pdev;
3693        struct e1000_hw *hw = &adapter->hw;
3694
3695        /* Virtualization features not supported on i210 family. */
3696        if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3697                return;
3698
3699        /* Of the below we really only want the effect of getting
3700         * IGB_FLAG_HAS_MSIX set (if available), without which
3701         * igb_enable_sriov() has no effect.
3702         */
3703        igb_set_interrupt_capability(adapter, true);
3704        igb_reset_interrupt_capability(adapter);
3705
3706        pci_sriov_set_totalvfs(pdev, 7);
3707        igb_enable_sriov(pdev, max_vfs);
3708
3709#endif /* CONFIG_PCI_IOV */
3710}
3711
3712unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3713{
3714        struct e1000_hw *hw = &adapter->hw;
3715        unsigned int max_rss_queues;
3716
3717        /* Determine the maximum number of RSS queues supported. */
3718        switch (hw->mac.type) {
3719        case e1000_i211:
3720                max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3721                break;
3722        case e1000_82575:
3723        case e1000_i210:
3724                max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3725                break;
3726        case e1000_i350:
3727                /* I350 cannot do RSS and SR-IOV at the same time */
3728                if (!!adapter->vfs_allocated_count) {
3729                        max_rss_queues = 1;
3730                        break;
3731                }
3732                /* fall through */
3733        case e1000_82576:
3734                if (!!adapter->vfs_allocated_count) {
3735                        max_rss_queues = 2;
3736                        break;
3737                }
3738                /* fall through */
3739        case e1000_82580:
3740        case e1000_i354:
3741        default:
3742                max_rss_queues = IGB_MAX_RX_QUEUES;
3743                break;
3744        }
3745
3746        return max_rss_queues;
3747}
3748
3749static void igb_init_queue_configuration(struct igb_adapter *adapter)
3750{
3751        u32 max_rss_queues;
3752
3753        max_rss_queues = igb_get_max_rss_queues(adapter);
3754        adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3755
3756        igb_set_flag_queue_pairs(adapter, max_rss_queues);
3757}
3758
3759void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3760                              const u32 max_rss_queues)
3761{
3762        struct e1000_hw *hw = &adapter->hw;
3763
3764        /* Determine if we need to pair queues. */
3765        switch (hw->mac.type) {
3766        case e1000_82575:
3767        case e1000_i211:
3768                /* Device supports enough interrupts without queue pairing. */
3769                break;
3770        case e1000_82576:
3771        case e1000_82580:
3772        case e1000_i350:
3773        case e1000_i354:
3774        case e1000_i210:
3775        default:
3776                /* If rss_queues > half of max_rss_queues, pair the queues in
3777                 * order to conserve interrupts due to limited supply.
3778                 */
3779                if (adapter->rss_queues > (max_rss_queues / 2))
3780                        adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3781                else
3782                        adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3783                break;
3784        }
3785}
3786
3787/**
3788 *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3789 *  @adapter: board private structure to initialize
3790 *
3791 *  igb_sw_init initializes the Adapter private data structure.
3792 *  Fields are initialized based on PCI device information and
3793 *  OS network device settings (MTU size).
3794 **/
3795static int igb_sw_init(struct igb_adapter *adapter)
3796{
3797        struct e1000_hw *hw = &adapter->hw;
3798        struct net_device *netdev = adapter->netdev;
3799        struct pci_dev *pdev = adapter->pdev;
3800
3801        pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3802
3803        /* set default ring sizes */
3804        adapter->tx_ring_count = IGB_DEFAULT_TXD;
3805        adapter->rx_ring_count = IGB_DEFAULT_RXD;
3806
3807        /* set default ITR values */
3808        adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3809        adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3810
3811        /* set default work limits */
3812        adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
3813
3814        adapter->max_frame_size = netdev->mtu + ETH_HLEN + ETH_FCS_LEN +
3815                                  VLAN_HLEN;
3816        adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
3817
3818        spin_lock_init(&adapter->nfc_lock);
3819        spin_lock_init(&adapter->stats64_lock);
3820#ifdef CONFIG_PCI_IOV
3821        switch (hw->mac.type) {
3822        case e1000_82576:
3823        case e1000_i350:
3824                if (max_vfs > 7) {
3825                        dev_warn(&pdev->dev,
3826                                 "Maximum of 7 VFs per PF, using max\n");
3827                        max_vfs = adapter->vfs_allocated_count = 7;
3828                } else
3829                        adapter->vfs_allocated_count = max_vfs;
3830                if (adapter->vfs_allocated_count)
3831                        dev_warn(&pdev->dev,
3832                                 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
3833                break;
3834        default:
3835                break;
3836        }
3837#endif /* CONFIG_PCI_IOV */
3838
3839        /* Assume MSI-X interrupts, will be checked during IRQ allocation */
3840        adapter->flags |= IGB_FLAG_HAS_MSIX;
3841
3842        adapter->mac_table = kcalloc(hw->mac.rar_entry_count,
3843                                     sizeof(struct igb_mac_addr),
3844                                     GFP_KERNEL);
3845        if (!adapter->mac_table)
3846                return -ENOMEM;
3847
3848        igb_probe_vfs(adapter);
3849
3850        igb_init_queue_configuration(adapter);
3851
3852        /* Setup and initialize a copy of the hw vlan table array */
3853        adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
3854                                       GFP_KERNEL);
3855        if (!adapter->shadow_vfta)
3856                return -ENOMEM;
3857
3858        /* This call may decrease the number of queues */
3859        if (igb_init_interrupt_scheme(adapter, true)) {
3860                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
3861                return -ENOMEM;
3862        }
3863
3864        /* Explicitly disable IRQ since the NIC can be in any state. */
3865        igb_irq_disable(adapter);
3866
3867        if (hw->mac.type >= e1000_i350)
3868                adapter->flags &= ~IGB_FLAG_DMAC;
3869
3870        set_bit(__IGB_DOWN, &adapter->state);
3871        return 0;
3872}
3873
3874/**
3875 *  igb_open - Called when a network interface is made active
3876 *  @netdev: network interface device structure
3877 *
3878 *  Returns 0 on success, negative value on failure
3879 *
3880 *  The open entry point is called when a network interface is made
3881 *  active by the system (IFF_UP).  At this point all resources needed
3882 *  for transmit and receive operations are allocated, the interrupt
3883 *  handler is registered with the OS, the watchdog timer is started,
3884 *  and the stack is notified that the interface is ready.
3885 **/
3886static int __igb_open(struct net_device *netdev, bool resuming)
3887{
3888        struct igb_adapter *adapter = netdev_priv(netdev);
3889        struct e1000_hw *hw = &adapter->hw;
3890        struct pci_dev *pdev = adapter->pdev;
3891        int err;
3892        int i;
3893
3894        /* disallow open during test */
3895        if (test_bit(__IGB_TESTING, &adapter->state)) {
3896                WARN_ON(resuming);
3897                return -EBUSY;
3898        }
3899
3900        if (!resuming)
3901                pm_runtime_get_sync(&pdev->dev);
3902
3903        netif_carrier_off(netdev);
3904
3905        /* allocate transmit descriptors */
3906        err = igb_setup_all_tx_resources(adapter);
3907        if (err)
3908                goto err_setup_tx;
3909
3910        /* allocate receive descriptors */
3911        err = igb_setup_all_rx_resources(adapter);
3912        if (err)
3913                goto err_setup_rx;
3914
3915        igb_power_up_link(adapter);
3916
3917        /* before we allocate an interrupt, we must be ready to handle it.
3918         * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
3919         * as soon as we call pci_request_irq, so we have to setup our
3920         * clean_rx handler before we do so.
3921         */
3922        igb_configure(adapter);
3923
3924        err = igb_request_irq(adapter);
3925        if (err)
3926                goto err_req_irq;
3927
3928        /* Notify the stack of the actual queue counts. */
3929        err = netif_set_real_num_tx_queues(adapter->netdev,
3930                                           adapter->num_tx_queues);
3931        if (err)
3932                goto err_set_queues;
3933
3934        err = netif_set_real_num_rx_queues(adapter->netdev,
3935                                           adapter->num_rx_queues);
3936        if (err)
3937                goto err_set_queues;
3938
3939        /* From here on the code is the same as igb_up() */
3940        clear_bit(__IGB_DOWN, &adapter->state);
3941
3942        for (i = 0; i < adapter->num_q_vectors; i++)
3943                napi_enable(&(adapter->q_vector[i]->napi));
3944
3945        /* Clear any pending interrupts. */
3946        rd32(E1000_TSICR);
3947        rd32(E1000_ICR);
3948
3949        igb_irq_enable(adapter);
3950
3951        /* notify VFs that reset has been completed */
3952        if (adapter->vfs_allocated_count) {
3953                u32 reg_data = rd32(E1000_CTRL_EXT);
3954
3955                reg_data |= E1000_CTRL_EXT_PFRSTD;
3956                wr32(E1000_CTRL_EXT, reg_data);
3957        }
3958
3959        netif_tx_start_all_queues(netdev);
3960
3961        if (!resuming)
3962                pm_runtime_put(&pdev->dev);
3963
3964        /* start the watchdog. */
3965        hw->mac.get_link_status = 1;
3966        schedule_work(&adapter->watchdog_task);
3967
3968        return 0;
3969
3970err_set_queues:
3971        igb_free_irq(adapter);
3972err_req_irq:
3973        igb_release_hw_control(adapter);
3974        igb_power_down_link(adapter);
3975        igb_free_all_rx_resources(adapter);
3976err_setup_rx:
3977        igb_free_all_tx_resources(adapter);
3978err_setup_tx:
3979        igb_reset(adapter);
3980        if (!resuming)
3981                pm_runtime_put(&pdev->dev);
3982
3983        return err;
3984}
3985
3986int igb_open(struct net_device *netdev)
3987{
3988        return __igb_open(netdev, false);
3989}
3990
3991/**
3992 *  igb_close - Disables a network interface
3993 *  @netdev: network interface device structure
3994 *
3995 *  Returns 0, this is not allowed to fail
3996 *
3997 *  The close entry point is called when an interface is de-activated
3998 *  by the OS.  The hardware is still under the driver's control, but
3999 *  needs to be disabled.  A global MAC reset is issued to stop the
4000 *  hardware, and all transmit and receive resources are freed.
4001 **/
4002static int __igb_close(struct net_device *netdev, bool suspending)
4003{
4004        struct igb_adapter *adapter = netdev_priv(netdev);
4005        struct pci_dev *pdev = adapter->pdev;
4006
4007        WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
4008
4009        if (!suspending)
4010                pm_runtime_get_sync(&pdev->dev);
4011
4012        igb_down(adapter);
4013        igb_free_irq(adapter);
4014
4015        igb_free_all_tx_resources(adapter);
4016        igb_free_all_rx_resources(adapter);
4017
4018        if (!suspending)
4019                pm_runtime_put_sync(&pdev->dev);
4020        return 0;
4021}
4022
4023int igb_close(struct net_device *netdev)
4024{
4025        if (netif_device_present(netdev) || netdev->dismantle)
4026                return __igb_close(netdev, false);
4027        return 0;
4028}
4029
4030/**
4031 *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
4032 *  @tx_ring: tx descriptor ring (for a specific queue) to setup
4033 *
4034 *  Return 0 on success, negative on failure
4035 **/
4036int igb_setup_tx_resources(struct igb_ring *tx_ring)
4037{
4038        struct device *dev = tx_ring->dev;
4039        int size;
4040
4041        size = sizeof(struct igb_tx_buffer) * tx_ring->count;
4042
4043        tx_ring->tx_buffer_info = vmalloc(size);
4044        if (!tx_ring->tx_buffer_info)
4045                goto err;
4046
4047        /* round up to nearest 4K */
4048        tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
4049        tx_ring->size = ALIGN(tx_ring->size, 4096);
4050
4051        tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4052                                           &tx_ring->dma, GFP_KERNEL);
4053        if (!tx_ring->desc)
4054                goto err;
4055
4056        tx_ring->next_to_use = 0;
4057        tx_ring->next_to_clean = 0;
4058
4059        return 0;
4060
4061err:
4062        vfree(tx_ring->tx_buffer_info);
4063        tx_ring->tx_buffer_info = NULL;
4064        dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4065        return -ENOMEM;
4066}
4067
4068/**
4069 *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
4070 *                               (Descriptors) for all queues
4071 *  @adapter: board private structure
4072 *
4073 *  Return 0 on success, negative on failure
4074 **/
4075static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
4076{
4077        struct pci_dev *pdev = adapter->pdev;
4078        int i, err = 0;
4079
4080        for (i = 0; i < adapter->num_tx_queues; i++) {
4081                err = igb_setup_tx_resources(adapter->tx_ring[i]);
4082                if (err) {
4083                        dev_err(&pdev->dev,
4084                                "Allocation for Tx Queue %u failed\n", i);
4085                        for (i--; i >= 0; i--)
4086                                igb_free_tx_resources(adapter->tx_ring[i]);
4087                        break;
4088                }
4089        }
4090
4091        return err;
4092}
4093
4094/**
4095 *  igb_setup_tctl - configure the transmit control registers
4096 *  @adapter: Board private structure
4097 **/
4098void igb_setup_tctl(struct igb_adapter *adapter)
4099{
4100        struct e1000_hw *hw = &adapter->hw;
4101        u32 tctl;
4102
4103        /* disable queue 0 which is enabled by default on 82575 and 82576 */
4104        wr32(E1000_TXDCTL(0), 0);
4105
4106        /* Program the Transmit Control Register */
4107        tctl = rd32(E1000_TCTL);
4108        tctl &= ~E1000_TCTL_CT;
4109        tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
4110                (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
4111
4112        igb_config_collision_dist(hw);
4113
4114        /* Enable transmits */
4115        tctl |= E1000_TCTL_EN;
4116
4117        wr32(E1000_TCTL, tctl);
4118}
4119
4120/**
4121 *  igb_configure_tx_ring - Configure transmit ring after Reset
4122 *  @adapter: board private structure
4123 *  @ring: tx ring to configure
4124 *
4125 *  Configure a transmit ring after a reset.
4126 **/
4127void igb_configure_tx_ring(struct igb_adapter *adapter,
4128                           struct igb_ring *ring)
4129{
4130        struct e1000_hw *hw = &adapter->hw;
4131        u32 txdctl = 0;
4132        u64 tdba = ring->dma;
4133        int reg_idx = ring->reg_idx;
4134
4135        wr32(E1000_TDLEN(reg_idx),
4136             ring->count * sizeof(union e1000_adv_tx_desc));
4137        wr32(E1000_TDBAL(reg_idx),
4138             tdba & 0x00000000ffffffffULL);
4139        wr32(E1000_TDBAH(reg_idx), tdba >> 32);
4140
4141        ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
4142        wr32(E1000_TDH(reg_idx), 0);
4143        writel(0, ring->tail);
4144
4145        txdctl |= IGB_TX_PTHRESH;
4146        txdctl |= IGB_TX_HTHRESH << 8;
4147        txdctl |= IGB_TX_WTHRESH << 16;
4148
4149        /* reinitialize tx_buffer_info */
4150        memset(ring->tx_buffer_info, 0,
4151               sizeof(struct igb_tx_buffer) * ring->count);
4152
4153        txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
4154        wr32(E1000_TXDCTL(reg_idx), txdctl);
4155}
4156
4157/**
4158 *  igb_configure_tx - Configure transmit Unit after Reset
4159 *  @adapter: board private structure
4160 *
4161 *  Configure the Tx unit of the MAC after a reset.
4162 **/
4163static void igb_configure_tx(struct igb_adapter *adapter)
4164{
4165        struct e1000_hw *hw = &adapter->hw;
4166        int i;
4167
4168        /* disable the queues */
4169        for (i = 0; i < adapter->num_tx_queues; i++)
4170                wr32(E1000_TXDCTL(adapter->tx_ring[i]->reg_idx), 0);
4171
4172        wrfl();
4173        usleep_range(10000, 20000);
4174
4175        for (i = 0; i < adapter->num_tx_queues; i++)
4176                igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
4177}
4178
4179/**
4180 *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
4181 *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
4182 *
4183 *  Returns 0 on success, negative on failure
4184 **/
4185int igb_setup_rx_resources(struct igb_ring *rx_ring)
4186{
4187        struct device *dev = rx_ring->dev;
4188        int size;
4189
4190        size = sizeof(struct igb_rx_buffer) * rx_ring->count;
4191
4192        rx_ring->rx_buffer_info = vmalloc(size);
4193        if (!rx_ring->rx_buffer_info)
4194                goto err;
4195
4196        /* Round up to nearest 4K */
4197        rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
4198        rx_ring->size = ALIGN(rx_ring->size, 4096);
4199
4200        rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4201                                           &rx_ring->dma, GFP_KERNEL);
4202        if (!rx_ring->desc)
4203                goto err;
4204
4205        rx_ring->next_to_alloc = 0;
4206        rx_ring->next_to_clean = 0;
4207        rx_ring->next_to_use = 0;
4208
4209        return 0;
4210
4211err:
4212        vfree(rx_ring->rx_buffer_info);
4213        rx_ring->rx_buffer_info = NULL;
4214        dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4215        return -ENOMEM;
4216}
4217
4218/**
4219 *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
4220 *                               (Descriptors) for all queues
4221 *  @adapter: board private structure
4222 *
4223 *  Return 0 on success, negative on failure
4224 **/
4225static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
4226{
4227        struct pci_dev *pdev = adapter->pdev;
4228        int i, err = 0;
4229
4230        for (i = 0; i < adapter->num_rx_queues; i++) {
4231                err = igb_setup_rx_resources(adapter->rx_ring[i]);
4232                if (err) {
4233                        dev_err(&pdev->dev,
4234                                "Allocation for Rx Queue %u failed\n", i);
4235                        for (i--; i >= 0; i--)
4236                                igb_free_rx_resources(adapter->rx_ring[i]);
4237                        break;
4238                }
4239        }
4240
4241        return err;
4242}
4243
4244/**
4245 *  igb_setup_mrqc - configure the multiple receive queue control registers
4246 *  @adapter: Board private structure
4247 **/
4248static void igb_setup_mrqc(struct igb_adapter *adapter)
4249{
4250        struct e1000_hw *hw = &adapter->hw;
4251        u32 mrqc, rxcsum;
4252        u32 j, num_rx_queues;
4253        u32 rss_key[10];
4254
4255        netdev_rss_key_fill(rss_key, sizeof(rss_key));
4256        for (j = 0; j < 10; j++)
4257                wr32(E1000_RSSRK(j), rss_key[j]);
4258
4259        num_rx_queues = adapter->rss_queues;
4260
4261        switch (hw->mac.type) {
4262        case e1000_82576:
4263                /* 82576 supports 2 RSS queues for SR-IOV */
4264                if (adapter->vfs_allocated_count)
4265                        num_rx_queues = 2;
4266                break;
4267        default:
4268                break;
4269        }
4270
4271        if (adapter->rss_indir_tbl_init != num_rx_queues) {
4272                for (j = 0; j < IGB_RETA_SIZE; j++)
4273                        adapter->rss_indir_tbl[j] =
4274                        (j * num_rx_queues) / IGB_RETA_SIZE;
4275                adapter->rss_indir_tbl_init = num_rx_queues;
4276        }
4277        igb_write_rss_indir_tbl(adapter);
4278
4279        /* Disable raw packet checksumming so that RSS hash is placed in
4280         * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
4281         * offloads as they are enabled by default
4282         */
4283        rxcsum = rd32(E1000_RXCSUM);
4284        rxcsum |= E1000_RXCSUM_PCSD;
4285
4286        if (adapter->hw.mac.type >= e1000_82576)
4287                /* Enable Receive Checksum Offload for SCTP */
4288                rxcsum |= E1000_RXCSUM_CRCOFL;
4289
4290        /* Don't need to set TUOFL or IPOFL, they default to 1 */
4291        wr32(E1000_RXCSUM, rxcsum);
4292
4293        /* Generate RSS hash based on packet types, TCP/UDP
4294         * port numbers and/or IPv4/v6 src and dst addresses
4295         */
4296        mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
4297               E1000_MRQC_RSS_FIELD_IPV4_TCP |
4298               E1000_MRQC_RSS_FIELD_IPV6 |
4299               E1000_MRQC_RSS_FIELD_IPV6_TCP |
4300               E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
4301
4302        if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
4303                mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
4304        if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
4305                mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
4306
4307        /* If VMDq is enabled then we set the appropriate mode for that, else
4308         * we default to RSS so that an RSS hash is calculated per packet even
4309         * if we are only using one queue
4310         */
4311        if (adapter->vfs_allocated_count) {
4312                if (hw->mac.type > e1000_82575) {
4313                        /* Set the default pool for the PF's first queue */
4314                        u32 vtctl = rd32(E1000_VT_CTL);
4315
4316                        vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
4317                                   E1000_VT_CTL_DISABLE_DEF_POOL);
4318                        vtctl |= adapter->vfs_allocated_count <<
4319                                E1000_VT_CTL_DEFAULT_POOL_SHIFT;
4320                        wr32(E1000_VT_CTL, vtctl);
4321                }
4322                if (adapter->rss_queues > 1)
4323                        mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
4324                else
4325                        mrqc |= E1000_MRQC_ENABLE_VMDQ;
4326        } else {
4327                if (hw->mac.type != e1000_i211)
4328                        mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4329        }
4330        igb_vmm_control(adapter);
4331
4332        wr32(E1000_MRQC, mrqc);
4333}
4334
4335/**
4336 *  igb_setup_rctl - configure the receive control registers
4337 *  @adapter: Board private structure
4338 **/
4339void igb_setup_rctl(struct igb_adapter *adapter)
4340{
4341        struct e1000_hw *hw = &adapter->hw;
4342        u32 rctl;
4343
4344        rctl = rd32(E1000_RCTL);
4345
4346        rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4347        rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4348
4349        rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4350                (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4351
4352        /* enable stripping of CRC. It's unlikely this will break BMC
4353         * redirection as it did with e1000. Newer features require
4354         * that the HW strips the CRC.
4355         */
4356        rctl |= E1000_RCTL_SECRC;
4357
4358        /* disable store bad packets and clear size bits. */
4359        rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4360
4361        /* enable LPE to allow for reception of jumbo frames */
4362        rctl |= E1000_RCTL_LPE;
4363
4364        /* disable queue 0 to prevent tail write w/o re-config */
4365        wr32(E1000_RXDCTL(0), 0);
4366
4367        /* Attention!!!  For SR-IOV PF driver operations you must enable
4368         * queue drop for all VF and PF queues to prevent head of line blocking
4369         * if an un-trusted VF does not provide descriptors to hardware.
4370         */
4371        if (adapter->vfs_allocated_count) {
4372                /* set all queue drop enable bits */
4373                wr32(E1000_QDE, ALL_QUEUES);
4374        }
4375
4376        /* This is useful for sniffing bad packets. */
4377        if (adapter->netdev->features & NETIF_F_RXALL) {
4378                /* UPE and MPE will be handled by normal PROMISC logic
4379                 * in e1000e_set_rx_mode
4380                 */
4381                rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4382                         E1000_RCTL_BAM | /* RX All Bcast Pkts */
4383                         E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4384
4385                rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4386                          E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4387                /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4388                 * and that breaks VLANs.
4389                 */
4390        }
4391
4392        wr32(E1000_RCTL, rctl);
4393}
4394
4395static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4396                                   int vfn)
4397{
4398        struct e1000_hw *hw = &adapter->hw;
4399        u32 vmolr;
4400
4401        if (size > MAX_JUMBO_FRAME_SIZE)
4402                size = MAX_JUMBO_FRAME_SIZE;
4403
4404        vmolr = rd32(E1000_VMOLR(vfn));
4405        vmolr &= ~E1000_VMOLR_RLPML_MASK;
4406        vmolr |= size | E1000_VMOLR_LPE;
4407        wr32(E1000_VMOLR(vfn), vmolr);
4408
4409        return 0;
4410}
4411
4412static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4413                                         int vfn, bool enable)
4414{
4415        struct e1000_hw *hw = &adapter->hw;
4416        u32 val, reg;
4417
4418        if (hw->mac.type < e1000_82576)
4419                return;
4420
4421        if (hw->mac.type == e1000_i350)
4422                reg = E1000_DVMOLR(vfn);
4423        else
4424                reg = E1000_VMOLR(vfn);
4425
4426        val = rd32(reg);
4427        if (enable)
4428                val |= E1000_VMOLR_STRVLAN;
4429        else
4430                val &= ~(E1000_VMOLR_STRVLAN);
4431        wr32(reg, val);
4432}
4433
4434static inline void igb_set_vmolr(struct igb_adapter *adapter,
4435                                 int vfn, bool aupe)
4436{
4437        struct e1000_hw *hw = &adapter->hw;
4438        u32 vmolr;
4439
4440        /* This register exists only on 82576 and newer so if we are older then
4441         * we should exit and do nothing
4442         */
4443        if (hw->mac.type < e1000_82576)
4444                return;
4445
4446        vmolr = rd32(E1000_VMOLR(vfn));
4447        if (aupe)
4448                vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4449        else
4450                vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4451
4452        /* clear all bits that might not be set */
4453        vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4454
4455        if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4456                vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4457        /* for VMDq only allow the VFs and pool 0 to accept broadcast and
4458         * multicast packets
4459         */
4460        if (vfn <= adapter->vfs_allocated_count)
4461                vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4462
4463        wr32(E1000_VMOLR(vfn), vmolr);
4464}
4465
4466/**
4467 *  igb_configure_rx_ring - Configure a receive ring after Reset
4468 *  @adapter: board private structure
4469 *  @ring: receive ring to be configured
4470 *
4471 *  Configure the Rx unit of the MAC after a reset.
4472 **/
4473void igb_configure_rx_ring(struct igb_adapter *adapter,
4474                           struct igb_ring *ring)
4475{
4476        struct e1000_hw *hw = &adapter->hw;
4477        union e1000_adv_rx_desc *rx_desc;
4478        u64 rdba = ring->dma;
4479        int reg_idx = ring->reg_idx;
4480        u32 srrctl = 0, rxdctl = 0;
4481
4482        /* disable the queue */
4483        wr32(E1000_RXDCTL(reg_idx), 0);
4484
4485        /* Set DMA base address registers */
4486        wr32(E1000_RDBAL(reg_idx),
4487             rdba & 0x00000000ffffffffULL);
4488        wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4489        wr32(E1000_RDLEN(reg_idx),
4490             ring->count * sizeof(union e1000_adv_rx_desc));
4491
4492        /* initialize head and tail */
4493        ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4494        wr32(E1000_RDH(reg_idx), 0);
4495        writel(0, ring->tail);
4496
4497        /* set descriptor configuration */
4498        srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4499        if (ring_uses_large_buffer(ring))
4500                srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4501        else
4502                srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4503        srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4504        if (hw->mac.type >= e1000_82580)
4505                srrctl |= E1000_SRRCTL_TIMESTAMP;
4506        /* Only set Drop Enable if we are supporting multiple queues */
4507        if (adapter->vfs_allocated_count || adapter->num_rx_queues > 1)
4508                srrctl |= E1000_SRRCTL_DROP_EN;
4509
4510        wr32(E1000_SRRCTL(reg_idx), srrctl);
4511
4512        /* set filtering for VMDQ pools */
4513        igb_set_vmolr(adapter, reg_idx & 0x7, true);
4514
4515        rxdctl |= IGB_RX_PTHRESH;
4516        rxdctl |= IGB_RX_HTHRESH << 8;
4517        rxdctl |= IGB_RX_WTHRESH << 16;
4518
4519        /* initialize rx_buffer_info */
4520        memset(ring->rx_buffer_info, 0,
4521               sizeof(struct igb_rx_buffer) * ring->count);
4522
4523        /* initialize Rx descriptor 0 */
4524        rx_desc = IGB_RX_DESC(ring, 0);
4525        rx_desc->wb.upper.length = 0;
4526
4527        /* enable receive descriptor fetching */
4528        rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4529        wr32(E1000_RXDCTL(reg_idx), rxdctl);
4530}
4531
4532static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4533                                  struct igb_ring *rx_ring)
4534{
4535        /* set build_skb and buffer size flags */
4536        clear_ring_build_skb_enabled(rx_ring);
4537        clear_ring_uses_large_buffer(rx_ring);
4538
4539        if (adapter->flags & IGB_FLAG_RX_LEGACY)
4540                return;
4541
4542        set_ring_build_skb_enabled(rx_ring);
4543
4544#if (PAGE_SIZE < 8192)
4545        if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4546                return;
4547
4548        set_ring_uses_large_buffer(rx_ring);
4549#endif
4550}
4551
4552/**
4553 *  igb_configure_rx - Configure receive Unit after Reset
4554 *  @adapter: board private structure
4555 *
4556 *  Configure the Rx unit of the MAC after a reset.
4557 **/
4558static void igb_configure_rx(struct igb_adapter *adapter)
4559{
4560        int i;
4561
4562        /* set the correct pool for the PF default MAC address in entry 0 */
4563        igb_set_default_mac_filter(adapter);
4564
4565        /* Setup the HW Rx Head and Tail Descriptor Pointers and
4566         * the Base and Length of the Rx Descriptor Ring
4567         */
4568        for (i = 0; i < adapter->num_rx_queues; i++) {
4569                struct igb_ring *rx_ring = adapter->rx_ring[i];
4570
4571                igb_set_rx_buffer_len(adapter, rx_ring);
4572                igb_configure_rx_ring(adapter, rx_ring);
4573        }
4574}
4575
4576/**
4577 *  igb_free_tx_resources - Free Tx Resources per Queue
4578 *  @tx_ring: Tx descriptor ring for a specific queue
4579 *
4580 *  Free all transmit software resources
4581 **/
4582void igb_free_tx_resources(struct igb_ring *tx_ring)
4583{
4584        igb_clean_tx_ring(tx_ring);
4585
4586        vfree(tx_ring->tx_buffer_info);
4587        tx_ring->tx_buffer_info = NULL;
4588
4589        /* if not set, then don't free */
4590        if (!tx_ring->desc)
4591                return;
4592
4593        dma_free_coherent(tx_ring->dev, tx_ring->size,
4594                          tx_ring->desc, tx_ring->dma);
4595
4596        tx_ring->desc = NULL;
4597}
4598
4599/**
4600 *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4601 *  @adapter: board private structure
4602 *
4603 *  Free all transmit software resources
4604 **/
4605static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4606{
4607        int i;
4608
4609        for (i = 0; i < adapter->num_tx_queues; i++)
4610                if (adapter->tx_ring[i])
4611                        igb_free_tx_resources(adapter->tx_ring[i]);
4612}
4613
4614/**
4615 *  igb_clean_tx_ring - Free Tx Buffers
4616 *  @tx_ring: ring to be cleaned
4617 **/
4618static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4619{
4620        u16 i = tx_ring->next_to_clean;
4621        struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4622
4623        while (i != tx_ring->next_to_use) {
4624                union e1000_adv_tx_desc *eop_desc, *tx_desc;
4625
4626                /* Free all the Tx ring sk_buffs */
4627                dev_kfree_skb_any(tx_buffer->skb);
4628
4629                /* unmap skb header data */
4630                dma_unmap_single(tx_ring->dev,
4631                                 dma_unmap_addr(tx_buffer, dma),
4632                                 dma_unmap_len(tx_buffer, len),
4633                                 DMA_TO_DEVICE);
4634
4635                /* check for eop_desc to determine the end of the packet */
4636                eop_desc = tx_buffer->next_to_watch;
4637                tx_desc = IGB_TX_DESC(tx_ring, i);
4638
4639                /* unmap remaining buffers */
4640                while (tx_desc != eop_desc) {
4641                        tx_buffer++;
4642                        tx_desc++;
4643                        i++;
4644                        if (unlikely(i == tx_ring->count)) {
4645                                i = 0;
4646                                tx_buffer = tx_ring->tx_buffer_info;
4647                                tx_desc = IGB_TX_DESC(tx_ring, 0);
4648                        }
4649
4650                        /* unmap any remaining paged data */
4651                        if (dma_unmap_len(tx_buffer, len))
4652                                dma_unmap_page(tx_ring->dev,
4653                                               dma_unmap_addr(tx_buffer, dma),
4654                                               dma_unmap_len(tx_buffer, len),
4655                                               DMA_TO_DEVICE);
4656                }
4657
4658                /* move us one more past the eop_desc for start of next pkt */
4659                tx_buffer++;
4660                i++;
4661                if (unlikely(i == tx_ring->count)) {
4662                        i = 0;
4663                        tx_buffer = tx_ring->tx_buffer_info;
4664                }
4665        }
4666
4667        /* reset BQL for queue */
4668        netdev_tx_reset_queue(txring_txq(tx_ring));
4669
4670        /* reset next_to_use and next_to_clean */
4671        tx_ring->next_to_use = 0;
4672        tx_ring->next_to_clean = 0;
4673}
4674
4675/**
4676 *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4677 *  @adapter: board private structure
4678 **/
4679static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4680{
4681        int i;
4682
4683        for (i = 0; i < adapter->num_tx_queues; i++)
4684                if (adapter->tx_ring[i])
4685                        igb_clean_tx_ring(adapter->tx_ring[i]);
4686}
4687
4688/**
4689 *  igb_free_rx_resources - Free Rx Resources
4690 *  @rx_ring: ring to clean the resources from
4691 *
4692 *  Free all receive software resources
4693 **/
4694void igb_free_rx_resources(struct igb_ring *rx_ring)
4695{
4696        igb_clean_rx_ring(rx_ring);
4697
4698        vfree(rx_ring->rx_buffer_info);
4699        rx_ring->rx_buffer_info = NULL;
4700
4701        /* if not set, then don't free */
4702        if (!rx_ring->desc)
4703                return;
4704
4705        dma_free_coherent(rx_ring->dev, rx_ring->size,
4706                          rx_ring->desc, rx_ring->dma);
4707
4708        rx_ring->desc = NULL;
4709}
4710
4711/**
4712 *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4713 *  @adapter: board private structure
4714 *
4715 *  Free all receive software resources
4716 **/
4717static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4718{
4719        int i;
4720
4721        for (i = 0; i < adapter->num_rx_queues; i++)
4722                if (adapter->rx_ring[i])
4723                        igb_free_rx_resources(adapter->rx_ring[i]);
4724}
4725
4726/**
4727 *  igb_clean_rx_ring - Free Rx Buffers per Queue
4728 *  @rx_ring: ring to free buffers from
4729 **/
4730static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4731{
4732        u16 i = rx_ring->next_to_clean;
4733
4734        if (rx_ring->skb)
4735                dev_kfree_skb(rx_ring->skb);
4736        rx_ring->skb = NULL;
4737
4738        /* Free all the Rx ring sk_buffs */
4739        while (i != rx_ring->next_to_alloc) {
4740                struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4741
4742                /* Invalidate cache lines that may have been written to by
4743                 * device so that we avoid corrupting memory.
4744                 */
4745                dma_sync_single_range_for_cpu(rx_ring->dev,
4746                                              buffer_info->dma,
4747                                              buffer_info->page_offset,
4748                                              igb_rx_bufsz(rx_ring),
4749                                              DMA_FROM_DEVICE);
4750
4751                /* free resources associated with mapping */
4752                dma_unmap_page_attrs(rx_ring->dev,
4753                                     buffer_info->dma,
4754                                     igb_rx_pg_size(rx_ring),
4755                                     DMA_FROM_DEVICE,
4756                                     IGB_RX_DMA_ATTR);
4757                __page_frag_cache_drain(buffer_info->page,
4758                                        buffer_info->pagecnt_bias);
4759
4760                i++;
4761                if (i == rx_ring->count)
4762                        i = 0;
4763        }
4764
4765        rx_ring->next_to_alloc = 0;
4766        rx_ring->next_to_clean = 0;
4767        rx_ring->next_to_use = 0;
4768}
4769
4770/**
4771 *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
4772 *  @adapter: board private structure
4773 **/
4774static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
4775{
4776        int i;
4777
4778        for (i = 0; i < adapter->num_rx_queues; i++)
4779                if (adapter->rx_ring[i])
4780                        igb_clean_rx_ring(adapter->rx_ring[i]);
4781}
4782
4783/**
4784 *  igb_set_mac - Change the Ethernet Address of the NIC
4785 *  @netdev: network interface device structure
4786 *  @p: pointer to an address structure
4787 *
4788 *  Returns 0 on success, negative on failure
4789 **/
4790static int igb_set_mac(struct net_device *netdev, void *p)
4791{
4792        struct igb_adapter *adapter = netdev_priv(netdev);
4793        struct e1000_hw *hw = &adapter->hw;
4794        struct sockaddr *addr = p;
4795
4796        if (!is_valid_ether_addr(addr->sa_data))
4797                return -EADDRNOTAVAIL;
4798
4799        memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
4800        memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
4801
4802        /* set the correct pool for the new PF MAC address in entry 0 */
4803        igb_set_default_mac_filter(adapter);
4804
4805        return 0;
4806}
4807
4808/**
4809 *  igb_write_mc_addr_list - write multicast addresses to MTA
4810 *  @netdev: network interface device structure
4811 *
4812 *  Writes multicast address list to the MTA hash table.
4813 *  Returns: -ENOMEM on failure
4814 *           0 on no addresses written
4815 *           X on writing X addresses to MTA
4816 **/
4817static int igb_write_mc_addr_list(struct net_device *netdev)
4818{
4819        struct igb_adapter *adapter = netdev_priv(netdev);
4820        struct e1000_hw *hw = &adapter->hw;
4821        struct netdev_hw_addr *ha;
4822        u8  *mta_list;
4823        int i;
4824
4825        if (netdev_mc_empty(netdev)) {
4826                /* nothing to program, so clear mc list */
4827                igb_update_mc_addr_list(hw, NULL, 0);
4828                igb_restore_vf_multicasts(adapter);
4829                return 0;
4830        }
4831
4832        mta_list = kcalloc(netdev_mc_count(netdev), 6, GFP_ATOMIC);
4833        if (!mta_list)
4834                return -ENOMEM;
4835
4836        /* The shared function expects a packed array of only addresses. */
4837        i = 0;
4838        netdev_for_each_mc_addr(ha, netdev)
4839                memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
4840
4841        igb_update_mc_addr_list(hw, mta_list, i);
4842        kfree(mta_list);
4843
4844        return netdev_mc_count(netdev);
4845}
4846
4847static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
4848{
4849        struct e1000_hw *hw = &adapter->hw;
4850        u32 i, pf_id;
4851
4852        switch (hw->mac.type) {
4853        case e1000_i210:
4854        case e1000_i211:
4855        case e1000_i350:
4856                /* VLAN filtering needed for VLAN prio filter */
4857                if (adapter->netdev->features & NETIF_F_NTUPLE)
4858                        break;
4859                /* fall through */
4860        case e1000_82576:
4861        case e1000_82580:
4862        case e1000_i354:
4863                /* VLAN filtering needed for pool filtering */
4864                if (adapter->vfs_allocated_count)
4865                        break;
4866                /* fall through */
4867        default:
4868                return 1;
4869        }
4870
4871        /* We are already in VLAN promisc, nothing to do */
4872        if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
4873                return 0;
4874
4875        if (!adapter->vfs_allocated_count)
4876                goto set_vfta;
4877
4878        /* Add PF to all active pools */
4879        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4880
4881        for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4882                u32 vlvf = rd32(E1000_VLVF(i));
4883
4884                vlvf |= BIT(pf_id);
4885                wr32(E1000_VLVF(i), vlvf);
4886        }
4887
4888set_vfta:
4889        /* Set all bits in the VLAN filter table array */
4890        for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
4891                hw->mac.ops.write_vfta(hw, i, ~0U);
4892
4893        /* Set flag so we don't redo unnecessary work */
4894        adapter->flags |= IGB_FLAG_VLAN_PROMISC;
4895
4896        return 0;
4897}
4898
4899#define VFTA_BLOCK_SIZE 8
4900static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
4901{
4902        struct e1000_hw *hw = &adapter->hw;
4903        u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
4904        u32 vid_start = vfta_offset * 32;
4905        u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
4906        u32 i, vid, word, bits, pf_id;
4907
4908        /* guarantee that we don't scrub out management VLAN */
4909        vid = adapter->mng_vlan_id;
4910        if (vid >= vid_start && vid < vid_end)
4911                vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4912
4913        if (!adapter->vfs_allocated_count)
4914                goto set_vfta;
4915
4916        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
4917
4918        for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
4919                u32 vlvf = rd32(E1000_VLVF(i));
4920
4921                /* pull VLAN ID from VLVF */
4922                vid = vlvf & VLAN_VID_MASK;
4923
4924                /* only concern ourselves with a certain range */
4925                if (vid < vid_start || vid >= vid_end)
4926                        continue;
4927
4928                if (vlvf & E1000_VLVF_VLANID_ENABLE) {
4929                        /* record VLAN ID in VFTA */
4930                        vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
4931
4932                        /* if PF is part of this then continue */
4933                        if (test_bit(vid, adapter->active_vlans))
4934                                continue;
4935                }
4936
4937                /* remove PF from the pool */
4938                bits = ~BIT(pf_id);
4939                bits &= rd32(E1000_VLVF(i));
4940                wr32(E1000_VLVF(i), bits);
4941        }
4942
4943set_vfta:
4944        /* extract values from active_vlans and write back to VFTA */
4945        for (i = VFTA_BLOCK_SIZE; i--;) {
4946                vid = (vfta_offset + i) * 32;
4947                word = vid / BITS_PER_LONG;
4948                bits = vid % BITS_PER_LONG;
4949
4950                vfta[i] |= adapter->active_vlans[word] >> bits;
4951
4952                hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
4953        }
4954}
4955
4956static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
4957{
4958        u32 i;
4959
4960        /* We are not in VLAN promisc, nothing to do */
4961        if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
4962                return;
4963
4964        /* Set flag so we don't redo unnecessary work */
4965        adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
4966
4967        for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
4968                igb_scrub_vfta(adapter, i);
4969}
4970
4971/**
4972 *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
4973 *  @netdev: network interface device structure
4974 *
4975 *  The set_rx_mode entry point is called whenever the unicast or multicast
4976 *  address lists or the network interface flags are updated.  This routine is
4977 *  responsible for configuring the hardware for proper unicast, multicast,
4978 *  promiscuous mode, and all-multi behavior.
4979 **/
4980static void igb_set_rx_mode(struct net_device *netdev)
4981{
4982        struct igb_adapter *adapter = netdev_priv(netdev);
4983        struct e1000_hw *hw = &adapter->hw;
4984        unsigned int vfn = adapter->vfs_allocated_count;
4985        u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
4986        int count;
4987
4988        /* Check for Promiscuous and All Multicast modes */
4989        if (netdev->flags & IFF_PROMISC) {
4990                rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
4991                vmolr |= E1000_VMOLR_MPME;
4992
4993                /* enable use of UTA filter to force packets to default pool */
4994                if (hw->mac.type == e1000_82576)
4995                        vmolr |= E1000_VMOLR_ROPE;
4996        } else {
4997                if (netdev->flags & IFF_ALLMULTI) {
4998                        rctl |= E1000_RCTL_MPE;
4999                        vmolr |= E1000_VMOLR_MPME;
5000                } else {
5001                        /* Write addresses to the MTA, if the attempt fails
5002                         * then we should just turn on promiscuous mode so
5003                         * that we can at least receive multicast traffic
5004                         */
5005                        count = igb_write_mc_addr_list(netdev);
5006                        if (count < 0) {
5007                                rctl |= E1000_RCTL_MPE;
5008                                vmolr |= E1000_VMOLR_MPME;
5009                        } else if (count) {
5010                                vmolr |= E1000_VMOLR_ROMPE;
5011                        }
5012                }
5013        }
5014
5015        /* Write addresses to available RAR registers, if there is not
5016         * sufficient space to store all the addresses then enable
5017         * unicast promiscuous mode
5018         */
5019        if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
5020                rctl |= E1000_RCTL_UPE;
5021                vmolr |= E1000_VMOLR_ROPE;
5022        }
5023
5024        /* enable VLAN filtering by default */
5025        rctl |= E1000_RCTL_VFE;
5026
5027        /* disable VLAN filtering for modes that require it */
5028        if ((netdev->flags & IFF_PROMISC) ||
5029            (netdev->features & NETIF_F_RXALL)) {
5030                /* if we fail to set all rules then just clear VFE */
5031                if (igb_vlan_promisc_enable(adapter))
5032                        rctl &= ~E1000_RCTL_VFE;
5033        } else {
5034                igb_vlan_promisc_disable(adapter);
5035        }
5036
5037        /* update state of unicast, multicast, and VLAN filtering modes */
5038        rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
5039                                     E1000_RCTL_VFE);
5040        wr32(E1000_RCTL, rctl);
5041
5042#if (PAGE_SIZE < 8192)
5043        if (!adapter->vfs_allocated_count) {
5044                if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5045                        rlpml = IGB_MAX_FRAME_BUILD_SKB;
5046        }
5047#endif
5048        wr32(E1000_RLPML, rlpml);
5049
5050        /* In order to support SR-IOV and eventually VMDq it is necessary to set
5051         * the VMOLR to enable the appropriate modes.  Without this workaround
5052         * we will have issues with VLAN tag stripping not being done for frames
5053         * that are only arriving because we are the default pool
5054         */
5055        if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
5056                return;
5057
5058        /* set UTA to appropriate mode */
5059        igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
5060
5061        vmolr |= rd32(E1000_VMOLR(vfn)) &
5062                 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
5063
5064        /* enable Rx jumbo frames, restrict as needed to support build_skb */
5065        vmolr &= ~E1000_VMOLR_RLPML_MASK;
5066#if (PAGE_SIZE < 8192)
5067        if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5068                vmolr |= IGB_MAX_FRAME_BUILD_SKB;
5069        else
5070#endif
5071                vmolr |= MAX_JUMBO_FRAME_SIZE;
5072        vmolr |= E1000_VMOLR_LPE;
5073
5074        wr32(E1000_VMOLR(vfn), vmolr);
5075
5076        igb_restore_vf_multicasts(adapter);
5077}
5078
5079static void igb_check_wvbr(struct igb_adapter *adapter)
5080{
5081        struct e1000_hw *hw = &adapter->hw;
5082        u32 wvbr = 0;
5083
5084        switch (hw->mac.type) {
5085        case e1000_82576:
5086        case e1000_i350:
5087                wvbr = rd32(E1000_WVBR);
5088                if (!wvbr)
5089                        return;
5090                break;
5091        default:
5092                break;
5093        }
5094
5095        adapter->wvbr |= wvbr;
5096}
5097
5098#define IGB_STAGGERED_QUEUE_OFFSET 8
5099
5100static void igb_spoof_check(struct igb_adapter *adapter)
5101{
5102        int j;
5103
5104        if (!adapter->wvbr)
5105                return;
5106
5107        for (j = 0; j < adapter->vfs_allocated_count; j++) {
5108                if (adapter->wvbr & BIT(j) ||
5109                    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
5110                        dev_warn(&adapter->pdev->dev,
5111                                "Spoof event(s) detected on VF %d\n", j);
5112                        adapter->wvbr &=
5113                                ~(BIT(j) |
5114                                  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
5115                }
5116        }
5117}
5118
5119/* Need to wait a few seconds after link up to get diagnostic information from
5120 * the phy
5121 */
5122static void igb_update_phy_info(struct timer_list *t)
5123{
5124        struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
5125        igb_get_phy_info(&adapter->hw);
5126}
5127
5128/**
5129 *  igb_has_link - check shared code for link and determine up/down
5130 *  @adapter: pointer to driver private info
5131 **/
5132bool igb_has_link(struct igb_adapter *adapter)
5133{
5134        struct e1000_hw *hw = &adapter->hw;
5135        bool link_active = false;
5136
5137        /* get_link_status is set on LSC (link status) interrupt or
5138         * rx sequence error interrupt.  get_link_status will stay
5139         * false until the e1000_check_for_link establishes link
5140         * for copper adapters ONLY
5141         */
5142        switch (hw->phy.media_type) {
5143        case e1000_media_type_copper:
5144                if (!hw->mac.get_link_status)
5145                        return true;
5146                /* fall through */
5147        case e1000_media_type_internal_serdes:
5148                hw->mac.ops.check_for_link(hw);
5149                link_active = !hw->mac.get_link_status;
5150                break;
5151        default:
5152        case e1000_media_type_unknown:
5153                break;
5154        }
5155
5156        if (((hw->mac.type == e1000_i210) ||
5157             (hw->mac.type == e1000_i211)) &&
5158             (hw->phy.id == I210_I_PHY_ID)) {
5159                if (!netif_carrier_ok(adapter->netdev)) {
5160                        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5161                } else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
5162                        adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
5163                        adapter->link_check_timeout = jiffies;
5164                }
5165        }
5166
5167        return link_active;
5168}
5169
5170static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
5171{
5172        bool ret = false;
5173        u32 ctrl_ext, thstat;
5174
5175        /* check for thermal sensor event on i350 copper only */
5176        if (hw->mac.type == e1000_i350) {
5177                thstat = rd32(E1000_THSTAT);
5178                ctrl_ext = rd32(E1000_CTRL_EXT);
5179
5180                if ((hw->phy.media_type == e1000_media_type_copper) &&
5181                    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
5182                        ret = !!(thstat & event);
5183        }
5184
5185        return ret;
5186}
5187
5188/**
5189 *  igb_check_lvmmc - check for malformed packets received
5190 *  and indicated in LVMMC register
5191 *  @adapter: pointer to adapter
5192 **/
5193static void igb_check_lvmmc(struct igb_adapter *adapter)
5194{
5195        struct e1000_hw *hw = &adapter->hw;
5196        u32 lvmmc;
5197
5198        lvmmc = rd32(E1000_LVMMC);
5199        if (lvmmc) {
5200                if (unlikely(net_ratelimit())) {
5201                        netdev_warn(adapter->netdev,
5202                                    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
5203                                    lvmmc);
5204                }
5205        }
5206}
5207
5208/**
5209 *  igb_watchdog - Timer Call-back
5210 *  @data: pointer to adapter cast into an unsigned long
5211 **/
5212static void igb_watchdog(struct timer_list *t)
5213{
5214        struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5215        /* Do the rest outside of interrupt context */
5216        schedule_work(&adapter->watchdog_task);
5217}
5218
5219static void igb_watchdog_task(struct work_struct *work)
5220{
5221        struct igb_adapter *adapter = container_of(work,
5222                                                   struct igb_adapter,
5223                                                   watchdog_task);
5224        struct e1000_hw *hw = &adapter->hw;
5225        struct e1000_phy_info *phy = &hw->phy;
5226        struct net_device *netdev = adapter->netdev;
5227        u32 link;
5228        int i;
5229        u32 connsw;
5230        u16 phy_data, retry_count = 20;
5231
5232        link = igb_has_link(adapter);
5233
5234        if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
5235                if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
5236                        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5237                else
5238                        link = false;
5239        }
5240
5241        /* Force link down if we have fiber to swap to */
5242        if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5243                if (hw->phy.media_type == e1000_media_type_copper) {
5244                        connsw = rd32(E1000_CONNSW);
5245                        if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
5246                                link = 0;
5247                }
5248        }
5249        if (link) {
5250                /* Perform a reset if the media type changed. */
5251                if (hw->dev_spec._82575.media_changed) {
5252                        hw->dev_spec._82575.media_changed = false;
5253                        adapter->flags |= IGB_FLAG_MEDIA_RESET;
5254                        igb_reset(adapter);
5255                }
5256                /* Cancel scheduled suspend requests. */
5257                pm_runtime_resume(netdev->dev.parent);
5258
5259                if (!netif_carrier_ok(netdev)) {
5260                        u32 ctrl;
5261
5262                        hw->mac.ops.get_speed_and_duplex(hw,
5263                                                         &adapter->link_speed,
5264                                                         &adapter->link_duplex);
5265
5266                        ctrl = rd32(E1000_CTRL);
5267                        /* Links status message must follow this format */
5268                        netdev_info(netdev,
5269                               "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5270                               netdev->name,
5271                               adapter->link_speed,
5272                               adapter->link_duplex == FULL_DUPLEX ?
5273                               "Full" : "Half",
5274                               (ctrl & E1000_CTRL_TFCE) &&
5275                               (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
5276                               (ctrl & E1000_CTRL_RFCE) ?  "RX" :
5277                               (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
5278
5279                        /* disable EEE if enabled */
5280                        if ((adapter->flags & IGB_FLAG_EEE) &&
5281                                (adapter->link_duplex == HALF_DUPLEX)) {
5282                                dev_info(&adapter->pdev->dev,
5283                                "EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
5284                                adapter->hw.dev_spec._82575.eee_disable = true;
5285                                adapter->flags &= ~IGB_FLAG_EEE;
5286                        }
5287
5288                        /* check if SmartSpeed worked */
5289                        igb_check_downshift(hw);
5290                        if (phy->speed_downgraded)
5291                                netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
5292
5293                        /* check for thermal sensor event */
5294                        if (igb_thermal_sensor_event(hw,
5295                            E1000_THSTAT_LINK_THROTTLE))
5296                                netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
5297
5298                        /* adjust timeout factor according to speed/duplex */
5299                        adapter->tx_timeout_factor = 1;
5300                        switch (adapter->link_speed) {
5301                        case SPEED_10:
5302                                adapter->tx_timeout_factor = 14;
5303                                break;
5304                        case SPEED_100:
5305                                /* maybe add some timeout factor ? */
5306                                break;
5307                        }
5308
5309                        if (adapter->link_speed != SPEED_1000)
5310                                goto no_wait;
5311
5312                        /* wait for Remote receiver status OK */
5313retry_read_status:
5314                        if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
5315                                              &phy_data)) {
5316                                if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
5317                                    retry_count) {
5318                                        msleep(100);
5319                                        retry_count--;
5320                                        goto retry_read_status;
5321                                } else if (!retry_count) {
5322                                        dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
5323                                }
5324                        } else {
5325                                dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
5326                        }
5327no_wait:
5328                        netif_carrier_on(netdev);
5329
5330                        igb_ping_all_vfs(adapter);
5331                        igb_check_vf_rate_limit(adapter);
5332
5333                        /* link state has changed, schedule phy info update */
5334                        if (!test_bit(__IGB_DOWN, &adapter->state))
5335                                mod_timer(&adapter->phy_info_timer,
5336                                          round_jiffies(jiffies + 2 * HZ));
5337                }
5338        } else {
5339                if (netif_carrier_ok(netdev)) {
5340                        adapter->link_speed = 0;
5341                        adapter->link_duplex = 0;
5342
5343                        /* check for thermal sensor event */
5344                        if (igb_thermal_sensor_event(hw,
5345                            E1000_THSTAT_PWR_DOWN)) {
5346                                netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5347                        }
5348
5349                        /* Links status message must follow this format */
5350                        netdev_info(netdev, "igb: %s NIC Link is Down\n",
5351                               netdev->name);
5352                        netif_carrier_off(netdev);
5353
5354                        igb_ping_all_vfs(adapter);
5355
5356                        /* link state has changed, schedule phy info update */
5357                        if (!test_bit(__IGB_DOWN, &adapter->state))
5358                                mod_timer(&adapter->phy_info_timer,
5359                                          round_jiffies(jiffies + 2 * HZ));
5360
5361                        /* link is down, time to check for alternate media */
5362                        if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5363                                igb_check_swap_media(adapter);
5364                                if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5365                                        schedule_work(&adapter->reset_task);
5366                                        /* return immediately */
5367                                        return;
5368                                }
5369                        }
5370                        pm_schedule_suspend(netdev->dev.parent,
5371                                            MSEC_PER_SEC * 5);
5372
5373                /* also check for alternate media here */
5374                } else if (!netif_carrier_ok(netdev) &&
5375                           (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5376                        igb_check_swap_media(adapter);
5377                        if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5378                                schedule_work(&adapter->reset_task);
5379                                /* return immediately */
5380                                return;
5381                        }
5382                }
5383        }
5384
5385        spin_lock(&adapter->stats64_lock);
5386        igb_update_stats(adapter);
5387        spin_unlock(&adapter->stats64_lock);
5388
5389        for (i = 0; i < adapter->num_tx_queues; i++) {
5390                struct igb_ring *tx_ring = adapter->tx_ring[i];
5391                if (!netif_carrier_ok(netdev)) {
5392                        /* We've lost link, so the controller stops DMA,
5393                         * but we've got queued Tx work that's never going
5394                         * to get done, so reset controller to flush Tx.
5395                         * (Do the reset outside of interrupt context).
5396                         */
5397                        if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5398                                adapter->tx_timeout_count++;
5399                                schedule_work(&adapter->reset_task);
5400                                /* return immediately since reset is imminent */
5401                                return;
5402                        }
5403                }
5404
5405                /* Force detection of hung controller every watchdog period */
5406                set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5407        }
5408
5409        /* Cause software interrupt to ensure Rx ring is cleaned */
5410        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5411                u32 eics = 0;
5412
5413                for (i = 0; i < adapter->num_q_vectors; i++)
5414                        eics |= adapter->q_vector[i]->eims_value;
5415                wr32(E1000_EICS, eics);
5416        } else {
5417                wr32(E1000_ICS, E1000_ICS_RXDMT0);
5418        }
5419
5420        igb_spoof_check(adapter);
5421        igb_ptp_rx_hang(adapter);
5422        igb_ptp_tx_hang(adapter);
5423
5424        /* Check LVMMC register on i350/i354 only */
5425        if ((adapter->hw.mac.type == e1000_i350) ||
5426            (adapter->hw.mac.type == e1000_i354))
5427                igb_check_lvmmc(adapter);
5428
5429        /* Reset the timer */
5430        if (!test_bit(__IGB_DOWN, &adapter->state)) {
5431                if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5432                        mod_timer(&adapter->watchdog_timer,
5433                                  round_jiffies(jiffies +  HZ));
5434                else
5435                        mod_timer(&adapter->watchdog_timer,
5436                                  round_jiffies(jiffies + 2 * HZ));
5437        }
5438}
5439
5440enum latency_range {
5441        lowest_latency = 0,
5442        low_latency = 1,
5443        bulk_latency = 2,
5444        latency_invalid = 255
5445};
5446
5447/**
5448 *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5449 *  @q_vector: pointer to q_vector
5450 *
5451 *  Stores a new ITR value based on strictly on packet size.  This
5452 *  algorithm is less sophisticated than that used in igb_update_itr,
5453 *  due to the difficulty of synchronizing statistics across multiple
5454 *  receive rings.  The divisors and thresholds used by this function
5455 *  were determined based on theoretical maximum wire speed and testing
5456 *  data, in order to minimize response time while increasing bulk
5457 *  throughput.
5458 *  This functionality is controlled by ethtool's coalescing settings.
5459 *  NOTE:  This function is called only when operating in a multiqueue
5460 *         receive environment.
5461 **/
5462static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5463{
5464        int new_val = q_vector->itr_val;
5465        int avg_wire_size = 0;
5466        struct igb_adapter *adapter = q_vector->adapter;
5467        unsigned int packets;
5468
5469        /* For non-gigabit speeds, just fix the interrupt rate at 4000
5470         * ints/sec - ITR timer value of 120 ticks.
5471         */
5472        if (adapter->link_speed != SPEED_1000) {
5473                new_val = IGB_4K_ITR;
5474                goto set_itr_val;
5475        }
5476
5477        packets = q_vector->rx.total_packets;
5478        if (packets)
5479                avg_wire_size = q_vector->rx.total_bytes / packets;
5480
5481        packets = q_vector->tx.total_packets;
5482        if (packets)
5483                avg_wire_size = max_t(u32, avg_wire_size,
5484                                      q_vector->tx.total_bytes / packets);
5485
5486        /* if avg_wire_size isn't set no work was done */
5487        if (!avg_wire_size)
5488                goto clear_counts;
5489
5490        /* Add 24 bytes to size to account for CRC, preamble, and gap */
5491        avg_wire_size += 24;
5492
5493        /* Don't starve jumbo frames */
5494        avg_wire_size = min(avg_wire_size, 3000);
5495
5496        /* Give a little boost to mid-size frames */
5497        if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5498                new_val = avg_wire_size / 3;
5499        else
5500                new_val = avg_wire_size / 2;
5501
5502        /* conservative mode (itr 3) eliminates the lowest_latency setting */
5503        if (new_val < IGB_20K_ITR &&
5504            ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5505             (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5506                new_val = IGB_20K_ITR;
5507
5508set_itr_val:
5509        if (new_val != q_vector->itr_val) {
5510                q_vector->itr_val = new_val;
5511                q_vector->set_itr = 1;
5512        }
5513clear_counts:
5514        q_vector->rx.total_bytes = 0;
5515        q_vector->rx.total_packets = 0;
5516        q_vector->tx.total_bytes = 0;
5517        q_vector->tx.total_packets = 0;
5518}
5519
5520/**
5521 *  igb_update_itr - update the dynamic ITR value based on statistics
5522 *  @q_vector: pointer to q_vector
5523 *  @ring_container: ring info to update the itr for
5524 *
5525 *  Stores a new ITR value based on packets and byte
5526 *  counts during the last interrupt.  The advantage of per interrupt
5527 *  computation is faster updates and more accurate ITR for the current
5528 *  traffic pattern.  Constants in this function were computed
5529 *  based on theoretical maximum wire speed and thresholds were set based
5530 *  on testing data as well as attempting to minimize response time
5531 *  while increasing bulk throughput.
5532 *  This functionality is controlled by ethtool's coalescing settings.
5533 *  NOTE:  These calculations are only valid when operating in a single-
5534 *         queue environment.
5535 **/
5536static void igb_update_itr(struct igb_q_vector *q_vector,
5537                           struct igb_ring_container *ring_container)
5538{
5539        unsigned int packets = ring_container->total_packets;
5540        unsigned int bytes = ring_container->total_bytes;
5541        u8 itrval = ring_container->itr;
5542
5543        /* no packets, exit with status unchanged */
5544        if (packets == 0)
5545                return;
5546
5547        switch (itrval) {
5548        case lowest_latency:
5549                /* handle TSO and jumbo frames */
5550                if (bytes/packets > 8000)
5551                        itrval = bulk_latency;
5552                else if ((packets < 5) && (bytes > 512))
5553                        itrval = low_latency;
5554                break;
5555        case low_latency:  /* 50 usec aka 20000 ints/s */
5556                if (bytes > 10000) {
5557                        /* this if handles the TSO accounting */
5558                        if (bytes/packets > 8000)
5559                                itrval = bulk_latency;
5560                        else if ((packets < 10) || ((bytes/packets) > 1200))
5561                                itrval = bulk_latency;
5562                        else if ((packets > 35))
5563                                itrval = lowest_latency;
5564                } else if (bytes/packets > 2000) {
5565                        itrval = bulk_latency;
5566                } else if (packets <= 2 && bytes < 512) {
5567                        itrval = lowest_latency;
5568                }
5569                break;
5570        case bulk_latency: /* 250 usec aka 4000 ints/s */
5571                if (bytes > 25000) {
5572                        if (packets > 35)
5573                                itrval = low_latency;
5574                } else if (bytes < 1500) {
5575                        itrval = low_latency;
5576                }
5577                break;
5578        }
5579
5580        /* clear work counters since we have the values we need */
5581        ring_container->total_bytes = 0;
5582        ring_container->total_packets = 0;
5583
5584        /* write updated itr to ring container */
5585        ring_container->itr = itrval;
5586}
5587
5588static void igb_set_itr(struct igb_q_vector *q_vector)
5589{
5590        struct igb_adapter *adapter = q_vector->adapter;
5591        u32 new_itr = q_vector->itr_val;
5592        u8 current_itr = 0;
5593
5594        /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5595        if (adapter->link_speed != SPEED_1000) {
5596                current_itr = 0;
5597                new_itr = IGB_4K_ITR;
5598                goto set_itr_now;
5599        }
5600
5601        igb_update_itr(q_vector, &q_vector->tx);
5602        igb_update_itr(q_vector, &q_vector->rx);
5603
5604        current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5605
5606        /* conservative mode (itr 3) eliminates the lowest_latency setting */
5607        if (current_itr == lowest_latency &&
5608            ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5609             (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5610                current_itr = low_latency;
5611
5612        switch (current_itr) {
5613        /* counts and packets in update_itr are dependent on these numbers */
5614        case lowest_latency:
5615                new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5616                break;
5617        case low_latency:
5618                new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5619                break;
5620        case bulk_latency:
5621                new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5622                break;
5623        default:
5624                break;
5625        }
5626
5627set_itr_now:
5628        if (new_itr != q_vector->itr_val) {
5629                /* this attempts to bias the interrupt rate towards Bulk
5630                 * by adding intermediate steps when interrupt rate is
5631                 * increasing
5632                 */
5633                new_itr = new_itr > q_vector->itr_val ?
5634                          max((new_itr * q_vector->itr_val) /
5635                          (new_itr + (q_vector->itr_val >> 2)),
5636                          new_itr) : new_itr;
5637                /* Don't write the value here; it resets the adapter's
5638                 * internal timer, and causes us to delay far longer than
5639                 * we should between interrupts.  Instead, we write the ITR
5640                 * value at the beginning of the next interrupt so the timing
5641                 * ends up being correct.
5642                 */
5643                q_vector->itr_val = new_itr;
5644                q_vector->set_itr = 1;
5645        }
5646}
5647
5648static void igb_tx_ctxtdesc(struct igb_ring *tx_ring,
5649                            struct igb_tx_buffer *first,
5650                            u32 vlan_macip_lens, u32 type_tucmd,
5651                            u32 mss_l4len_idx)
5652{
5653        struct e1000_adv_tx_context_desc *context_desc;
5654        u16 i = tx_ring->next_to_use;
5655        struct timespec64 ts;
5656
5657        context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5658
5659        i++;
5660        tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5661
5662        /* set bits to identify this as an advanced context descriptor */
5663        type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5664
5665        /* For 82575, context index must be unique per ring. */
5666        if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5667                mss_l4len_idx |= tx_ring->reg_idx << 4;
5668
5669        context_desc->vlan_macip_lens   = cpu_to_le32(vlan_macip_lens);
5670        context_desc->type_tucmd_mlhl   = cpu_to_le32(type_tucmd);
5671        context_desc->mss_l4len_idx     = cpu_to_le32(mss_l4len_idx);
5672
5673        /* We assume there is always a valid tx time available. Invalid times
5674         * should have been handled by the upper layers.
5675         */
5676        if (tx_ring->launchtime_enable) {
5677                ts = ns_to_timespec64(first->skb->tstamp);
5678                first->skb->tstamp = 0;
5679                context_desc->seqnum_seed = cpu_to_le32(ts.tv_nsec / 32);
5680        } else {
5681                context_desc->seqnum_seed = 0;
5682        }
5683}
5684
5685static int igb_tso(struct igb_ring *tx_ring,
5686                   struct igb_tx_buffer *first,
5687                   u8 *hdr_len)
5688{
5689        u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5690        struct sk_buff *skb = first->skb;
5691        union {
5692                struct iphdr *v4;
5693                struct ipv6hdr *v6;
5694                unsigned char *hdr;
5695        } ip;
5696        union {
5697                struct tcphdr *tcp;
5698                unsigned char *hdr;
5699        } l4;
5700        u32 paylen, l4_offset;
5701        int err;
5702
5703        if (skb->ip_summed != CHECKSUM_PARTIAL)
5704                return 0;
5705
5706        if (!skb_is_gso(skb))
5707                return 0;
5708
5709        err = skb_cow_head(skb, 0);
5710        if (err < 0)
5711                return err;
5712
5713        ip.hdr = skb_network_header(skb);
5714        l4.hdr = skb_checksum_start(skb);
5715
5716        /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5717        type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5718
5719        /* initialize outer IP header fields */
5720        if (ip.v4->version == 4) {
5721                unsigned char *csum_start = skb_checksum_start(skb);
5722                unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5723
5724                /* IP header will have to cancel out any data that
5725                 * is not a part of the outer IP header
5726                 */
5727                ip.v4->check = csum_fold(csum_partial(trans_start,
5728                                                      csum_start - trans_start,
5729                                                      0));
5730                type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5731
5732                ip.v4->tot_len = 0;
5733                first->tx_flags |= IGB_TX_FLAGS_TSO |
5734                                   IGB_TX_FLAGS_CSUM |
5735                                   IGB_TX_FLAGS_IPV4;
5736        } else {
5737                ip.v6->payload_len = 0;
5738                first->tx_flags |= IGB_TX_FLAGS_TSO |
5739                                   IGB_TX_FLAGS_CSUM;
5740        }
5741
5742        /* determine offset of inner transport header */
5743        l4_offset = l4.hdr - skb->data;
5744
5745        /* compute length of segmentation header */
5746        *hdr_len = (l4.tcp->doff * 4) + l4_offset;
5747
5748        /* remove payload length from inner checksum */
5749        paylen = skb->len - l4_offset;
5750        csum_replace_by_diff(&l4.tcp->check, htonl(paylen));
5751
5752        /* update gso size and bytecount with header size */
5753        first->gso_segs = skb_shinfo(skb)->gso_segs;
5754        first->bytecount += (first->gso_segs - 1) * *hdr_len;
5755
5756        /* MSS L4LEN IDX */
5757        mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
5758        mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
5759
5760        /* VLAN MACLEN IPLEN */
5761        vlan_macip_lens = l4.hdr - ip.hdr;
5762        vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
5763        vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5764
5765        igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens,
5766                        type_tucmd, mss_l4len_idx);
5767
5768        return 1;
5769}
5770
5771static inline bool igb_ipv6_csum_is_sctp(struct sk_buff *skb)
5772{
5773        unsigned int offset = 0;
5774
5775        ipv6_find_hdr(skb, &offset, IPPROTO_SCTP, NULL, NULL);
5776
5777        return offset == skb_checksum_start_offset(skb);
5778}
5779
5780static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
5781{
5782        struct sk_buff *skb = first->skb;
5783        u32 vlan_macip_lens = 0;
5784        u32 type_tucmd = 0;
5785
5786        if (skb->ip_summed != CHECKSUM_PARTIAL) {
5787csum_failed:
5788                if (!(first->tx_flags & IGB_TX_FLAGS_VLAN) &&
5789                    !tx_ring->launchtime_enable)
5790                        return;
5791                goto no_csum;
5792        }
5793
5794        switch (skb->csum_offset) {
5795        case offsetof(struct tcphdr, check):
5796                type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
5797                /* fall through */
5798        case offsetof(struct udphdr, check):
5799                break;
5800        case offsetof(struct sctphdr, checksum):
5801                /* validate that this is actually an SCTP request */
5802                if (((first->protocol == htons(ETH_P_IP)) &&
5803                     (ip_hdr(skb)->protocol == IPPROTO_SCTP)) ||
5804                    ((first->protocol == htons(ETH_P_IPV6)) &&
5805                     igb_ipv6_csum_is_sctp(skb))) {
5806                        type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
5807                        break;
5808                }
5809                /* fall through */
5810        default:
5811                skb_checksum_help(skb);
5812                goto csum_failed;
5813        }
5814
5815        /* update TX checksum flag */
5816        first->tx_flags |= IGB_TX_FLAGS_CSUM;
5817        vlan_macip_lens = skb_checksum_start_offset(skb) -
5818                          skb_network_offset(skb);
5819no_csum:
5820        vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
5821        vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5822
5823        igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens, type_tucmd, 0);
5824}
5825
5826#define IGB_SET_FLAG(_input, _flag, _result) \
5827        ((_flag <= _result) ? \
5828         ((u32)(_input & _flag) * (_result / _flag)) : \
5829         ((u32)(_input & _flag) / (_flag / _result)))
5830
5831static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
5832{
5833        /* set type for advanced descriptor with frame checksum insertion */
5834        u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
5835                       E1000_ADVTXD_DCMD_DEXT |
5836                       E1000_ADVTXD_DCMD_IFCS;
5837
5838        /* set HW vlan bit if vlan is present */
5839        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
5840                                 (E1000_ADVTXD_DCMD_VLE));
5841
5842        /* set segmentation bits for TSO */
5843        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
5844                                 (E1000_ADVTXD_DCMD_TSE));
5845
5846        /* set timestamp bit if present */
5847        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
5848                                 (E1000_ADVTXD_MAC_TSTAMP));
5849
5850        /* insert frame checksum */
5851        cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
5852
5853        return cmd_type;
5854}
5855
5856static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
5857                                 union e1000_adv_tx_desc *tx_desc,
5858                                 u32 tx_flags, unsigned int paylen)
5859{
5860        u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
5861
5862        /* 82575 requires a unique index per ring */
5863        if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5864                olinfo_status |= tx_ring->reg_idx << 4;
5865
5866        /* insert L4 checksum */
5867        olinfo_status |= IGB_SET_FLAG(tx_flags,
5868                                      IGB_TX_FLAGS_CSUM,
5869                                      (E1000_TXD_POPTS_TXSM << 8));
5870
5871        /* insert IPv4 checksum */
5872        olinfo_status |= IGB_SET_FLAG(tx_flags,
5873                                      IGB_TX_FLAGS_IPV4,
5874                                      (E1000_TXD_POPTS_IXSM << 8));
5875
5876        tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
5877}
5878
5879static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5880{
5881        struct net_device *netdev = tx_ring->netdev;
5882
5883        netif_stop_subqueue(netdev, tx_ring->queue_index);
5884
5885        /* Herbert's original patch had:
5886         *  smp_mb__after_netif_stop_queue();
5887         * but since that doesn't exist yet, just open code it.
5888         */
5889        smp_mb();
5890
5891        /* We need to check again in a case another CPU has just
5892         * made room available.
5893         */
5894        if (igb_desc_unused(tx_ring) < size)
5895                return -EBUSY;
5896
5897        /* A reprieve! */
5898        netif_wake_subqueue(netdev, tx_ring->queue_index);
5899
5900        u64_stats_update_begin(&tx_ring->tx_syncp2);
5901        tx_ring->tx_stats.restart_queue2++;
5902        u64_stats_update_end(&tx_ring->tx_syncp2);
5903
5904        return 0;
5905}
5906
5907static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
5908{
5909        if (igb_desc_unused(tx_ring) >= size)
5910                return 0;
5911        return __igb_maybe_stop_tx(tx_ring, size);
5912}
5913
5914static int igb_tx_map(struct igb_ring *tx_ring,
5915                      struct igb_tx_buffer *first,
5916                      const u8 hdr_len)
5917{
5918        struct sk_buff *skb = first->skb;
5919        struct igb_tx_buffer *tx_buffer;
5920        union e1000_adv_tx_desc *tx_desc;
5921        struct skb_frag_struct *frag;
5922        dma_addr_t dma;
5923        unsigned int data_len, size;
5924        u32 tx_flags = first->tx_flags;
5925        u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
5926        u16 i = tx_ring->next_to_use;
5927
5928        tx_desc = IGB_TX_DESC(tx_ring, i);
5929
5930        igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
5931
5932        size = skb_headlen(skb);
5933        data_len = skb->data_len;
5934
5935        dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
5936
5937        tx_buffer = first;
5938
5939        for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
5940                if (dma_mapping_error(tx_ring->dev, dma))
5941                        goto dma_error;
5942
5943                /* record length, and DMA address */
5944                dma_unmap_len_set(tx_buffer, len, size);
5945                dma_unmap_addr_set(tx_buffer, dma, dma);
5946
5947                tx_desc->read.buffer_addr = cpu_to_le64(dma);
5948
5949                while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
5950                        tx_desc->read.cmd_type_len =
5951                                cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
5952
5953                        i++;
5954                        tx_desc++;
5955                        if (i == tx_ring->count) {
5956                                tx_desc = IGB_TX_DESC(tx_ring, 0);
5957                                i = 0;
5958                        }
5959                        tx_desc->read.olinfo_status = 0;
5960
5961                        dma += IGB_MAX_DATA_PER_TXD;
5962                        size -= IGB_MAX_DATA_PER_TXD;
5963
5964                        tx_desc->read.buffer_addr = cpu_to_le64(dma);
5965                }
5966
5967                if (likely(!data_len))
5968                        break;
5969
5970                tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
5971
5972                i++;
5973                tx_desc++;
5974                if (i == tx_ring->count) {
5975                        tx_desc = IGB_TX_DESC(tx_ring, 0);
5976                        i = 0;
5977                }
5978                tx_desc->read.olinfo_status = 0;
5979
5980                size = skb_frag_size(frag);
5981                data_len -= size;
5982
5983                dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
5984                                       size, DMA_TO_DEVICE);
5985
5986                tx_buffer = &tx_ring->tx_buffer_info[i];
5987        }
5988
5989        /* write last descriptor with RS and EOP bits */
5990        cmd_type |= size | IGB_TXD_DCMD;
5991        tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
5992
5993        netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
5994
5995        /* set the timestamp */
5996        first->time_stamp = jiffies;
5997
5998        skb_tx_timestamp(skb);
5999
6000        /* Force memory writes to complete before letting h/w know there
6001         * are new descriptors to fetch.  (Only applicable for weak-ordered
6002         * memory model archs, such as IA-64).
6003         *
6004         * We also need this memory barrier to make certain all of the
6005         * status bits have been updated before next_to_watch is written.
6006         */
6007        dma_wmb();
6008
6009        /* set next_to_watch value indicating a packet is present */
6010        first->next_to_watch = tx_desc;
6011
6012        i++;
6013        if (i == tx_ring->count)
6014                i = 0;
6015
6016        tx_ring->next_to_use = i;
6017
6018        /* Make sure there is space in the ring for the next send. */
6019        igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6020
6021        if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
6022                writel(i, tx_ring->tail);
6023        }
6024        return 0;
6025
6026dma_error:
6027        dev_err(tx_ring->dev, "TX DMA map failed\n");
6028        tx_buffer = &tx_ring->tx_buffer_info[i];
6029
6030        /* clear dma mappings for failed tx_buffer_info map */
6031        while (tx_buffer != first) {
6032                if (dma_unmap_len(tx_buffer, len))
6033                        dma_unmap_page(tx_ring->dev,
6034                                       dma_unmap_addr(tx_buffer, dma),
6035                                       dma_unmap_len(tx_buffer, len),
6036                                       DMA_TO_DEVICE);
6037                dma_unmap_len_set(tx_buffer, len, 0);
6038
6039                if (i-- == 0)
6040                        i += tx_ring->count;
6041                tx_buffer = &tx_ring->tx_buffer_info[i];
6042        }
6043
6044        if (dma_unmap_len(tx_buffer, len))
6045                dma_unmap_single(tx_ring->dev,
6046                                 dma_unmap_addr(tx_buffer, dma),
6047                                 dma_unmap_len(tx_buffer, len),
6048                                 DMA_TO_DEVICE);
6049        dma_unmap_len_set(tx_buffer, len, 0);
6050
6051        dev_kfree_skb_any(tx_buffer->skb);
6052        tx_buffer->skb = NULL;
6053
6054        tx_ring->next_to_use = i;
6055
6056        return -1;
6057}
6058
6059netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
6060                                struct igb_ring *tx_ring)
6061{
6062        struct igb_tx_buffer *first;
6063        int tso;
6064        u32 tx_flags = 0;
6065        unsigned short f;
6066        u16 count = TXD_USE_COUNT(skb_headlen(skb));
6067        __be16 protocol = vlan_get_protocol(skb);
6068        u8 hdr_len = 0;
6069
6070        /* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
6071         *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
6072         *       + 2 desc gap to keep tail from touching head,
6073         *       + 1 desc for context descriptor,
6074         * otherwise try next time
6075         */
6076        for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6077                count += TXD_USE_COUNT(skb_shinfo(skb)->frags[f].size);
6078
6079        if (igb_maybe_stop_tx(tx_ring, count + 3)) {
6080                /* this is a hard error */
6081                return NETDEV_TX_BUSY;
6082        }
6083
6084        /* record the location of the first descriptor for this packet */
6085        first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6086        first->skb = skb;
6087        first->bytecount = skb->len;
6088        first->gso_segs = 1;
6089
6090        if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6091                struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6092
6093                if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
6094                    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
6095                                           &adapter->state)) {
6096                        skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6097                        tx_flags |= IGB_TX_FLAGS_TSTAMP;
6098
6099                        adapter->ptp_tx_skb = skb_get(skb);
6100                        adapter->ptp_tx_start = jiffies;
6101                        if (adapter->hw.mac.type == e1000_82576)
6102                                schedule_work(&adapter->ptp_tx_work);
6103                } else {
6104                        adapter->tx_hwtstamp_skipped++;
6105                }
6106        }
6107
6108        if (skb_vlan_tag_present(skb)) {
6109                tx_flags |= IGB_TX_FLAGS_VLAN;
6110                tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
6111        }
6112
6113        /* record initial flags and protocol */
6114        first->tx_flags = tx_flags;
6115        first->protocol = protocol;
6116
6117        tso = igb_tso(tx_ring, first, &hdr_len);
6118        if (tso < 0)
6119                goto out_drop;
6120        else if (!tso)
6121                igb_tx_csum(tx_ring, first);
6122
6123        if (igb_tx_map(tx_ring, first, hdr_len))
6124                goto cleanup_tx_tstamp;
6125
6126        return NETDEV_TX_OK;
6127
6128out_drop:
6129        dev_kfree_skb_any(first->skb);
6130        first->skb = NULL;
6131cleanup_tx_tstamp:
6132        if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
6133                struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6134
6135                dev_kfree_skb_any(adapter->ptp_tx_skb);
6136                adapter->ptp_tx_skb = NULL;
6137                if (adapter->hw.mac.type == e1000_82576)
6138                        cancel_work_sync(&adapter->ptp_tx_work);
6139                clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
6140        }
6141
6142        return NETDEV_TX_OK;
6143}
6144
6145static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
6146                                                    struct sk_buff *skb)
6147{
6148        unsigned int r_idx = skb->queue_mapping;
6149
6150        if (r_idx >= adapter->num_tx_queues)
6151                r_idx = r_idx % adapter->num_tx_queues;
6152
6153        return adapter->tx_ring[r_idx];
6154}
6155
6156static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
6157                                  struct net_device *netdev)
6158{
6159        struct igb_adapter *adapter = netdev_priv(netdev);
6160
6161        /* The minimum packet size with TCTL.PSP set is 17 so pad the skb
6162         * in order to meet this minimum size requirement.
6163         */
6164        if (skb_put_padto(skb, 17))
6165                return NETDEV_TX_OK;
6166
6167        return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
6168}
6169
6170/**
6171 *  igb_tx_timeout - Respond to a Tx Hang
6172 *  @netdev: network interface device structure
6173 **/
6174static void igb_tx_timeout(struct net_device *netdev)
6175{
6176        struct igb_adapter *adapter = netdev_priv(netdev);
6177        struct e1000_hw *hw = &adapter->hw;
6178
6179        /* Do the reset outside of interrupt context */
6180        adapter->tx_timeout_count++;
6181
6182        if (hw->mac.type >= e1000_82580)
6183                hw->dev_spec._82575.global_device_reset = true;
6184
6185        schedule_work(&adapter->reset_task);
6186        wr32(E1000_EICS,
6187             (adapter->eims_enable_mask & ~adapter->eims_other));
6188}
6189
6190static void igb_reset_task(struct work_struct *work)
6191{
6192        struct igb_adapter *adapter;
6193        adapter = container_of(work, struct igb_adapter, reset_task);
6194
6195        igb_dump(adapter);
6196        netdev_err(adapter->netdev, "Reset adapter\n");
6197        igb_reinit_locked(adapter);
6198}
6199
6200/**
6201 *  igb_get_stats64 - Get System Network Statistics
6202 *  @netdev: network interface device structure
6203 *  @stats: rtnl_link_stats64 pointer
6204 **/
6205static void igb_get_stats64(struct net_device *netdev,
6206                            struct rtnl_link_stats64 *stats)
6207{
6208        struct igb_adapter *adapter = netdev_priv(netdev);
6209
6210        spin_lock(&adapter->stats64_lock);
6211        igb_update_stats(adapter);
6212        memcpy(stats, &adapter->stats64, sizeof(*stats));
6213        spin_unlock(&adapter->stats64_lock);
6214}
6215
6216/**
6217 *  igb_change_mtu - Change the Maximum Transfer Unit
6218 *  @netdev: network interface device structure
6219 *  @new_mtu: new value for maximum frame size
6220 *
6221 *  Returns 0 on success, negative on failure
6222 **/
6223static int igb_change_mtu(struct net_device *netdev, int new_mtu)
6224{
6225        struct igb_adapter *adapter = netdev_priv(netdev);
6226        struct pci_dev *pdev = adapter->pdev;
6227        int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
6228
6229        /* adjust max frame to be at least the size of a standard frame */
6230        if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
6231                max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
6232
6233        while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
6234                usleep_range(1000, 2000);
6235
6236        /* igb_down has a dependency on max_frame_size */
6237        adapter->max_frame_size = max_frame;
6238
6239        if (netif_running(netdev))
6240                igb_down(adapter);
6241
6242        dev_info(&pdev->dev, "changing MTU from %d to %d\n",
6243                 netdev->mtu, new_mtu);
6244        netdev->mtu = new_mtu;
6245
6246        if (netif_running(netdev))
6247                igb_up(adapter);
6248        else
6249                igb_reset(adapter);
6250
6251        clear_bit(__IGB_RESETTING, &adapter->state);
6252
6253        return 0;
6254}
6255
6256/**
6257 *  igb_update_stats - Update the board statistics counters
6258 *  @adapter: board private structure
6259 **/
6260void igb_update_stats(struct igb_adapter *adapter)
6261{
6262        struct rtnl_link_stats64 *net_stats = &adapter->stats64;
6263        struct e1000_hw *hw = &adapter->hw;
6264        struct pci_dev *pdev = adapter->pdev;
6265        u32 reg, mpc;
6266        int i;
6267        u64 bytes, packets;
6268        unsigned int start;
6269        u64 _bytes, _packets;
6270
6271        /* Prevent stats update while adapter is being reset, or if the pci
6272         * connection is down.
6273         */
6274        if (adapter->link_speed == 0)
6275                return;
6276        if (pci_channel_offline(pdev))
6277                return;
6278
6279        bytes = 0;
6280        packets = 0;
6281
6282        rcu_read_lock();
6283        for (i = 0; i < adapter->num_rx_queues; i++) {
6284                struct igb_ring *ring = adapter->rx_ring[i];
6285                u32 rqdpc = rd32(E1000_RQDPC(i));
6286                if (hw->mac.type >= e1000_i210)
6287                        wr32(E1000_RQDPC(i), 0);
6288
6289                if (rqdpc) {
6290                        ring->rx_stats.drops += rqdpc;
6291                        net_stats->rx_fifo_errors += rqdpc;
6292                }
6293
6294                do {
6295                        start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
6296                        _bytes = ring->rx_stats.bytes;
6297                        _packets = ring->rx_stats.packets;
6298                } while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
6299                bytes += _bytes;
6300                packets += _packets;
6301        }
6302
6303        net_stats->rx_bytes = bytes;
6304        net_stats->rx_packets = packets;
6305
6306        bytes = 0;
6307        packets = 0;
6308        for (i = 0; i < adapter->num_tx_queues; i++) {
6309                struct igb_ring *ring = adapter->tx_ring[i];
6310                do {
6311                        start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
6312                        _bytes = ring->tx_stats.bytes;
6313                        _packets = ring->tx_stats.packets;
6314                } while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
6315                bytes += _bytes;
6316                packets += _packets;
6317        }
6318        net_stats->tx_bytes = bytes;
6319        net_stats->tx_packets = packets;
6320        rcu_read_unlock();
6321
6322        /* read stats registers */
6323        adapter->stats.crcerrs += rd32(E1000_CRCERRS);
6324        adapter->stats.gprc += rd32(E1000_GPRC);
6325        adapter->stats.gorc += rd32(E1000_GORCL);
6326        rd32(E1000_GORCH); /* clear GORCL */
6327        adapter->stats.bprc += rd32(E1000_BPRC);
6328        adapter->stats.mprc += rd32(E1000_MPRC);
6329        adapter->stats.roc += rd32(E1000_ROC);
6330
6331        adapter->stats.prc64 += rd32(E1000_PRC64);
6332        adapter->stats.prc127 += rd32(E1000_PRC127);
6333        adapter->stats.prc255 += rd32(E1000_PRC255);
6334        adapter->stats.prc511 += rd32(E1000_PRC511);
6335        adapter->stats.prc1023 += rd32(E1000_PRC1023);
6336        adapter->stats.prc1522 += rd32(E1000_PRC1522);
6337        adapter->stats.symerrs += rd32(E1000_SYMERRS);
6338        adapter->stats.sec += rd32(E1000_SEC);
6339
6340        mpc = rd32(E1000_MPC);
6341        adapter->stats.mpc += mpc;
6342        net_stats->rx_fifo_errors += mpc;
6343        adapter->stats.scc += rd32(E1000_SCC);
6344        adapter->stats.ecol += rd32(E1000_ECOL);
6345        adapter->stats.mcc += rd32(E1000_MCC);
6346        adapter->stats.latecol += rd32(E1000_LATECOL);
6347        adapter->stats.dc += rd32(E1000_DC);
6348        adapter->stats.rlec += rd32(E1000_RLEC);
6349        adapter->stats.xonrxc += rd32(E1000_XONRXC);
6350        adapter->stats.xontxc += rd32(E1000_XONTXC);
6351        adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6352        adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6353        adapter->stats.fcruc += rd32(E1000_FCRUC);
6354        adapter->stats.gptc += rd32(E1000_GPTC);
6355        adapter->stats.gotc += rd32(E1000_GOTCL);
6356        rd32(E1000_GOTCH); /* clear GOTCL */
6357        adapter->stats.rnbc += rd32(E1000_RNBC);
6358        adapter->stats.ruc += rd32(E1000_RUC);
6359        adapter->stats.rfc += rd32(E1000_RFC);
6360        adapter->stats.rjc += rd32(E1000_RJC);
6361        adapter->stats.tor += rd32(E1000_TORH);
6362        adapter->stats.tot += rd32(E1000_TOTH);
6363        adapter->stats.tpr += rd32(E1000_TPR);
6364
6365        adapter->stats.ptc64 += rd32(E1000_PTC64);
6366        adapter->stats.ptc127 += rd32(E1000_PTC127);
6367        adapter->stats.ptc255 += rd32(E1000_PTC255);
6368        adapter->stats.ptc511 += rd32(E1000_PTC511);
6369        adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6370        adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6371
6372        adapter->stats.mptc += rd32(E1000_MPTC);
6373        adapter->stats.bptc += rd32(E1000_BPTC);
6374
6375        adapter->stats.tpt += rd32(E1000_TPT);
6376        adapter->stats.colc += rd32(E1000_COLC);
6377
6378        adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6379        /* read internal phy specific stats */
6380        reg = rd32(E1000_CTRL_EXT);
6381        if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6382                adapter->stats.rxerrc += rd32(E1000_RXERRC);
6383
6384                /* this stat has invalid values on i210/i211 */
6385                if ((hw->mac.type != e1000_i210) &&
6386                    (hw->mac.type != e1000_i211))
6387                        adapter->stats.tncrs += rd32(E1000_TNCRS);
6388        }
6389
6390        adapter->stats.tsctc += rd32(E1000_TSCTC);
6391        adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6392
6393        adapter->stats.iac += rd32(E1000_IAC);
6394        adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6395        adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6396        adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6397        adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6398        adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6399        adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6400        adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6401        adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6402
6403        /* Fill out the OS statistics structure */
6404        net_stats->multicast = adapter->stats.mprc;
6405        net_stats->collisions = adapter->stats.colc;
6406
6407        /* Rx Errors */
6408
6409        /* RLEC on some newer hardware can be incorrect so build
6410         * our own version based on RUC and ROC
6411         */
6412        net_stats->rx_errors = adapter->stats.rxerrc +
6413                adapter->stats.crcerrs + adapter->stats.algnerrc +
6414                adapter->stats.ruc + adapter->stats.roc +
6415                adapter->stats.cexterr;
6416        net_stats->rx_length_errors = adapter->stats.ruc +
6417                                      adapter->stats.roc;
6418        net_stats->rx_crc_errors = adapter->stats.crcerrs;
6419        net_stats->rx_frame_errors = adapter->stats.algnerrc;
6420        net_stats->rx_missed_errors = adapter->stats.mpc;
6421
6422        /* Tx Errors */
6423        net_stats->tx_errors = adapter->stats.ecol +
6424                               adapter->stats.latecol;
6425        net_stats->tx_aborted_errors = adapter->stats.ecol;
6426        net_stats->tx_window_errors = adapter->stats.latecol;
6427        net_stats->tx_carrier_errors = adapter->stats.tncrs;
6428
6429        /* Tx Dropped needs to be maintained elsewhere */
6430
6431        /* Management Stats */
6432        adapter->stats.mgptc += rd32(E1000_MGTPTC);
6433        adapter->stats.mgprc += rd32(E1000_MGTPRC);
6434        adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6435
6436        /* OS2BMC Stats */
6437        reg = rd32(E1000_MANC);
6438        if (reg & E1000_MANC_EN_BMC2OS) {
6439                adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6440                adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6441                adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6442                adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6443        }
6444}
6445
6446static void igb_tsync_interrupt(struct igb_adapter *adapter)
6447{
6448        struct e1000_hw *hw = &adapter->hw;
6449        struct ptp_clock_event event;
6450        struct timespec64 ts;
6451        u32 ack = 0, tsauxc, sec, nsec, tsicr = rd32(E1000_TSICR);
6452
6453        if (tsicr & TSINTR_SYS_WRAP) {
6454                event.type = PTP_CLOCK_PPS;
6455                if (adapter->ptp_caps.pps)
6456                        ptp_clock_event(adapter->ptp_clock, &event);
6457                ack |= TSINTR_SYS_WRAP;
6458        }
6459
6460        if (tsicr & E1000_TSICR_TXTS) {
6461                /* retrieve hardware timestamp */
6462                schedule_work(&adapter->ptp_tx_work);
6463                ack |= E1000_TSICR_TXTS;
6464        }
6465
6466        if (tsicr & TSINTR_TT0) {
6467                spin_lock(&adapter->tmreg_lock);
6468                ts = timespec64_add(adapter->perout[0].start,
6469                                    adapter->perout[0].period);
6470                /* u32 conversion of tv_sec is safe until y2106 */
6471                wr32(E1000_TRGTTIML0, ts.tv_nsec);
6472                wr32(E1000_TRGTTIMH0, (u32)ts.tv_sec);
6473                tsauxc = rd32(E1000_TSAUXC);
6474                tsauxc |= TSAUXC_EN_TT0;
6475                wr32(E1000_TSAUXC, tsauxc);
6476                adapter->perout[0].start = ts;
6477                spin_unlock(&adapter->tmreg_lock);
6478                ack |= TSINTR_TT0;
6479        }
6480
6481        if (tsicr & TSINTR_TT1) {
6482                spin_lock(&adapter->tmreg_lock);
6483                ts = timespec64_add(adapter->perout[1].start,
6484                                    adapter->perout[1].period);
6485                wr32(E1000_TRGTTIML1, ts.tv_nsec);
6486                wr32(E1000_TRGTTIMH1, (u32)ts.tv_sec);
6487                tsauxc = rd32(E1000_TSAUXC);
6488                tsauxc |= TSAUXC_EN_TT1;
6489                wr32(E1000_TSAUXC, tsauxc);
6490                adapter->perout[1].start = ts;
6491                spin_unlock(&adapter->tmreg_lock);
6492                ack |= TSINTR_TT1;
6493        }
6494
6495        if (tsicr & TSINTR_AUTT0) {
6496                nsec = rd32(E1000_AUXSTMPL0);
6497                sec  = rd32(E1000_AUXSTMPH0);
6498                event.type = PTP_CLOCK_EXTTS;
6499                event.index = 0;
6500                event.timestamp = sec * 1000000000ULL + nsec;
6501                ptp_clock_event(adapter->ptp_clock, &event);
6502                ack |= TSINTR_AUTT0;
6503        }
6504
6505        if (tsicr & TSINTR_AUTT1) {
6506                nsec = rd32(E1000_AUXSTMPL1);
6507                sec  = rd32(E1000_AUXSTMPH1);
6508                event.type = PTP_CLOCK_EXTTS;
6509                event.index = 1;
6510                event.timestamp = sec * 1000000000ULL + nsec;
6511                ptp_clock_event(adapter->ptp_clock, &event);
6512                ack |= TSINTR_AUTT1;
6513        }
6514
6515        /* acknowledge the interrupts */
6516        wr32(E1000_TSICR, ack);
6517}
6518
6519static irqreturn_t igb_msix_other(int irq, void *data)
6520{
6521        struct igb_adapter *adapter = data;
6522        struct e1000_hw *hw = &adapter->hw;
6523        u32 icr = rd32(E1000_ICR);
6524        /* reading ICR causes bit 31 of EICR to be cleared */
6525
6526        if (icr & E1000_ICR_DRSTA)
6527                schedule_work(&adapter->reset_task);
6528
6529        if (icr & E1000_ICR_DOUTSYNC) {
6530                /* HW is reporting DMA is out of sync */
6531                adapter->stats.doosync++;
6532                /* The DMA Out of Sync is also indication of a spoof event
6533                 * in IOV mode. Check the Wrong VM Behavior register to
6534                 * see if it is really a spoof event.
6535                 */
6536                igb_check_wvbr(adapter);
6537        }
6538
6539        /* Check for a mailbox event */
6540        if (icr & E1000_ICR_VMMB)
6541                igb_msg_task(adapter);
6542
6543        if (icr & E1000_ICR_LSC) {
6544                hw->mac.get_link_status = 1;
6545                /* guard against interrupt when we're going down */
6546                if (!test_bit(__IGB_DOWN, &adapter->state))
6547                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
6548        }
6549
6550        if (icr & E1000_ICR_TS)
6551                igb_tsync_interrupt(adapter);
6552
6553        wr32(E1000_EIMS, adapter->eims_other);
6554
6555        return IRQ_HANDLED;
6556}
6557
6558static void igb_write_itr(struct igb_q_vector *q_vector)
6559{
6560        struct igb_adapter *adapter = q_vector->adapter;
6561        u32 itr_val = q_vector->itr_val & 0x7FFC;
6562
6563        if (!q_vector->set_itr)
6564                return;
6565
6566        if (!itr_val)
6567                itr_val = 0x4;
6568
6569        if (adapter->hw.mac.type == e1000_82575)
6570                itr_val |= itr_val << 16;
6571        else
6572                itr_val |= E1000_EITR_CNT_IGNR;
6573
6574        writel(itr_val, q_vector->itr_register);
6575        q_vector->set_itr = 0;
6576}
6577
6578static irqreturn_t igb_msix_ring(int irq, void *data)
6579{
6580        struct igb_q_vector *q_vector = data;
6581
6582        /* Write the ITR value calculated from the previous interrupt. */
6583        igb_write_itr(q_vector);
6584
6585        napi_schedule(&q_vector->napi);
6586
6587        return IRQ_HANDLED;
6588}
6589
6590#ifdef CONFIG_IGB_DCA
6591static void igb_update_tx_dca(struct igb_adapter *adapter,
6592                              struct igb_ring *tx_ring,
6593                              int cpu)
6594{
6595        struct e1000_hw *hw = &adapter->hw;
6596        u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
6597
6598        if (hw->mac.type != e1000_82575)
6599                txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
6600
6601        /* We can enable relaxed ordering for reads, but not writes when
6602         * DCA is enabled.  This is due to a known issue in some chipsets
6603         * which will cause the DCA tag to be cleared.
6604         */
6605        txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
6606                  E1000_DCA_TXCTRL_DATA_RRO_EN |
6607                  E1000_DCA_TXCTRL_DESC_DCA_EN;
6608
6609        wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
6610}
6611
6612static void igb_update_rx_dca(struct igb_adapter *adapter,
6613                              struct igb_ring *rx_ring,
6614                              int cpu)
6615{
6616        struct e1000_hw *hw = &adapter->hw;
6617        u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
6618
6619        if (hw->mac.type != e1000_82575)
6620                rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
6621
6622        /* We can enable relaxed ordering for reads, but not writes when
6623         * DCA is enabled.  This is due to a known issue in some chipsets
6624         * which will cause the DCA tag to be cleared.
6625         */
6626        rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
6627                  E1000_DCA_RXCTRL_DESC_DCA_EN;
6628
6629        wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
6630}
6631
6632static void igb_update_dca(struct igb_q_vector *q_vector)
6633{
6634        struct igb_adapter *adapter = q_vector->adapter;
6635        int cpu = get_cpu();
6636
6637        if (q_vector->cpu == cpu)
6638                goto out_no_update;
6639
6640        if (q_vector->tx.ring)
6641                igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
6642
6643        if (q_vector->rx.ring)
6644                igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
6645
6646        q_vector->cpu = cpu;
6647out_no_update:
6648        put_cpu();
6649}
6650
6651static void igb_setup_dca(struct igb_adapter *adapter)
6652{
6653        struct e1000_hw *hw = &adapter->hw;
6654        int i;
6655
6656        if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
6657                return;
6658
6659        /* Always use CB2 mode, difference is masked in the CB driver. */
6660        wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
6661
6662        for (i = 0; i < adapter->num_q_vectors; i++) {
6663                adapter->q_vector[i]->cpu = -1;
6664                igb_update_dca(adapter->q_vector[i]);
6665        }
6666}
6667
6668static int __igb_notify_dca(struct device *dev, void *data)
6669{
6670        struct net_device *netdev = dev_get_drvdata(dev);
6671        struct igb_adapter *adapter = netdev_priv(netdev);
6672        struct pci_dev *pdev = adapter->pdev;
6673        struct e1000_hw *hw = &adapter->hw;
6674        unsigned long event = *(unsigned long *)data;
6675
6676        switch (event) {
6677        case DCA_PROVIDER_ADD:
6678                /* if already enabled, don't do it again */
6679                if (adapter->flags & IGB_FLAG_DCA_ENABLED)
6680                        break;
6681                if (dca_add_requester(dev) == 0) {
6682                        adapter->flags |= IGB_FLAG_DCA_ENABLED;
6683                        dev_info(&pdev->dev, "DCA enabled\n");
6684                        igb_setup_dca(adapter);
6685                        break;
6686                }
6687                /* Fall Through - since DCA is disabled. */
6688        case DCA_PROVIDER_REMOVE:
6689                if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
6690                        /* without this a class_device is left
6691                         * hanging around in the sysfs model
6692                         */
6693                        dca_remove_requester(dev);
6694                        dev_info(&pdev->dev, "DCA disabled\n");
6695                        adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
6696                        wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
6697                }
6698                break;
6699        }
6700
6701        return 0;
6702}
6703
6704static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
6705                          void *p)
6706{
6707        int ret_val;
6708
6709        ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
6710                                         __igb_notify_dca);
6711
6712        return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
6713}
6714#endif /* CONFIG_IGB_DCA */
6715
6716#ifdef CONFIG_PCI_IOV
6717static int igb_vf_configure(struct igb_adapter *adapter, int vf)
6718{
6719        unsigned char mac_addr[ETH_ALEN];
6720
6721        eth_zero_addr(mac_addr);
6722        igb_set_vf_mac(adapter, vf, mac_addr);
6723
6724        /* By default spoof check is enabled for all VFs */
6725        adapter->vf_data[vf].spoofchk_enabled = true;
6726
6727        /* By default VFs are not trusted */
6728        adapter->vf_data[vf].trusted = false;
6729
6730        return 0;
6731}
6732
6733#endif
6734static void igb_ping_all_vfs(struct igb_adapter *adapter)
6735{
6736        struct e1000_hw *hw = &adapter->hw;
6737        u32 ping;
6738        int i;
6739
6740        for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
6741                ping = E1000_PF_CONTROL_MSG;
6742                if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
6743                        ping |= E1000_VT_MSGTYPE_CTS;
6744                igb_write_mbx(hw, &ping, 1, i);
6745        }
6746}
6747
6748static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
6749{
6750        struct e1000_hw *hw = &adapter->hw;
6751        u32 vmolr = rd32(E1000_VMOLR(vf));
6752        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6753
6754        vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
6755                            IGB_VF_FLAG_MULTI_PROMISC);
6756        vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6757
6758        if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
6759                vmolr |= E1000_VMOLR_MPME;
6760                vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
6761                *msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
6762        } else {
6763                /* if we have hashes and we are clearing a multicast promisc
6764                 * flag we need to write the hashes to the MTA as this step
6765                 * was previously skipped
6766                 */
6767                if (vf_data->num_vf_mc_hashes > 30) {
6768                        vmolr |= E1000_VMOLR_MPME;
6769                } else if (vf_data->num_vf_mc_hashes) {
6770                        int j;
6771
6772                        vmolr |= E1000_VMOLR_ROMPE;
6773                        for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6774                                igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6775                }
6776        }
6777
6778        wr32(E1000_VMOLR(vf), vmolr);
6779
6780        /* there are flags left unprocessed, likely not supported */
6781        if (*msgbuf & E1000_VT_MSGINFO_MASK)
6782                return -EINVAL;
6783
6784        return 0;
6785}
6786
6787static int igb_set_vf_multicasts(struct igb_adapter *adapter,
6788                                  u32 *msgbuf, u32 vf)
6789{
6790        int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
6791        u16 *hash_list = (u16 *)&msgbuf[1];
6792        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
6793        int i;
6794
6795        /* salt away the number of multicast addresses assigned
6796         * to this VF for later use to restore when the PF multi cast
6797         * list changes
6798         */
6799        vf_data->num_vf_mc_hashes = n;
6800
6801        /* only up to 30 hash values supported */
6802        if (n > 30)
6803                n = 30;
6804
6805        /* store the hashes for later use */
6806        for (i = 0; i < n; i++)
6807                vf_data->vf_mc_hashes[i] = hash_list[i];
6808
6809        /* Flush and reset the mta with the new values */
6810        igb_set_rx_mode(adapter->netdev);
6811
6812        return 0;
6813}
6814
6815static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
6816{
6817        struct e1000_hw *hw = &adapter->hw;
6818        struct vf_data_storage *vf_data;
6819        int i, j;
6820
6821        for (i = 0; i < adapter->vfs_allocated_count; i++) {
6822                u32 vmolr = rd32(E1000_VMOLR(i));
6823
6824                vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
6825
6826                vf_data = &adapter->vf_data[i];
6827
6828                if ((vf_data->num_vf_mc_hashes > 30) ||
6829                    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
6830                        vmolr |= E1000_VMOLR_MPME;
6831                } else if (vf_data->num_vf_mc_hashes) {
6832                        vmolr |= E1000_VMOLR_ROMPE;
6833                        for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
6834                                igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
6835                }
6836                wr32(E1000_VMOLR(i), vmolr);
6837        }
6838}
6839
6840static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
6841{
6842        struct e1000_hw *hw = &adapter->hw;
6843        u32 pool_mask, vlvf_mask, i;
6844
6845        /* create mask for VF and other pools */
6846        pool_mask = E1000_VLVF_POOLSEL_MASK;
6847        vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
6848
6849        /* drop PF from pool bits */
6850        pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
6851                             adapter->vfs_allocated_count);
6852
6853        /* Find the vlan filter for this id */
6854        for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
6855                u32 vlvf = rd32(E1000_VLVF(i));
6856                u32 vfta_mask, vid, vfta;
6857
6858                /* remove the vf from the pool */
6859                if (!(vlvf & vlvf_mask))
6860                        continue;
6861
6862                /* clear out bit from VLVF */
6863                vlvf ^= vlvf_mask;
6864
6865                /* if other pools are present, just remove ourselves */
6866                if (vlvf & pool_mask)
6867                        goto update_vlvfb;
6868
6869                /* if PF is present, leave VFTA */
6870                if (vlvf & E1000_VLVF_POOLSEL_MASK)
6871                        goto update_vlvf;
6872
6873                vid = vlvf & E1000_VLVF_VLANID_MASK;
6874                vfta_mask = BIT(vid % 32);
6875
6876                /* clear bit from VFTA */
6877                vfta = adapter->shadow_vfta[vid / 32];
6878                if (vfta & vfta_mask)
6879                        hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
6880update_vlvf:
6881                /* clear pool selection enable */
6882                if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6883                        vlvf &= E1000_VLVF_POOLSEL_MASK;
6884                else
6885                        vlvf = 0;
6886update_vlvfb:
6887                /* clear pool bits */
6888                wr32(E1000_VLVF(i), vlvf);
6889        }
6890}
6891
6892static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
6893{
6894        u32 vlvf;
6895        int idx;
6896
6897        /* short cut the special case */
6898        if (vlan == 0)
6899                return 0;
6900
6901        /* Search for the VLAN id in the VLVF entries */
6902        for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
6903                vlvf = rd32(E1000_VLVF(idx));
6904                if ((vlvf & VLAN_VID_MASK) == vlan)
6905                        break;
6906        }
6907
6908        return idx;
6909}
6910
6911static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
6912{
6913        struct e1000_hw *hw = &adapter->hw;
6914        u32 bits, pf_id;
6915        int idx;
6916
6917        idx = igb_find_vlvf_entry(hw, vid);
6918        if (!idx)
6919                return;
6920
6921        /* See if any other pools are set for this VLAN filter
6922         * entry other than the PF.
6923         */
6924        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
6925        bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
6926        bits &= rd32(E1000_VLVF(idx));
6927
6928        /* Disable the filter so this falls into the default pool. */
6929        if (!bits) {
6930                if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
6931                        wr32(E1000_VLVF(idx), BIT(pf_id));
6932                else
6933                        wr32(E1000_VLVF(idx), 0);
6934        }
6935}
6936
6937static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
6938                           bool add, u32 vf)
6939{
6940        int pf_id = adapter->vfs_allocated_count;
6941        struct e1000_hw *hw = &adapter->hw;
6942        int err;
6943
6944        /* If VLAN overlaps with one the PF is currently monitoring make
6945         * sure that we are able to allocate a VLVF entry.  This may be
6946         * redundant but it guarantees PF will maintain visibility to
6947         * the VLAN.
6948         */
6949        if (add && test_bit(vid, adapter->active_vlans)) {
6950                err = igb_vfta_set(hw, vid, pf_id, true, false);
6951                if (err)
6952                        return err;
6953        }
6954
6955        err = igb_vfta_set(hw, vid, vf, add, false);
6956
6957        if (add && !err)
6958                return err;
6959
6960        /* If we failed to add the VF VLAN or we are removing the VF VLAN
6961         * we may need to drop the PF pool bit in order to allow us to free
6962         * up the VLVF resources.
6963         */
6964        if (test_bit(vid, adapter->active_vlans) ||
6965            (adapter->flags & IGB_FLAG_VLAN_PROMISC))
6966                igb_update_pf_vlvf(adapter, vid);
6967
6968        return err;
6969}
6970
6971static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
6972{
6973        struct e1000_hw *hw = &adapter->hw;
6974
6975        if (vid)
6976                wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
6977        else
6978                wr32(E1000_VMVIR(vf), 0);
6979}
6980
6981static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
6982                                u16 vlan, u8 qos)
6983{
6984        int err;
6985
6986        err = igb_set_vf_vlan(adapter, vlan, true, vf);
6987        if (err)
6988                return err;
6989
6990        igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
6991        igb_set_vmolr(adapter, vf, !vlan);
6992
6993        /* revoke access to previous VLAN */
6994        if (vlan != adapter->vf_data[vf].pf_vlan)
6995                igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
6996                                false, vf);
6997
6998        adapter->vf_data[vf].pf_vlan = vlan;
6999        adapter->vf_data[vf].pf_qos = qos;
7000        igb_set_vf_vlan_strip(adapter, vf, true);
7001        dev_info(&adapter->pdev->dev,
7002                 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
7003        if (test_bit(__IGB_DOWN, &adapter->state)) {
7004                dev_warn(&adapter->pdev->dev,
7005                         "The VF VLAN has been set, but the PF device is not up.\n");
7006                dev_warn(&adapter->pdev->dev,
7007                         "Bring the PF device up before attempting to use the VF device.\n");
7008        }
7009
7010        return err;
7011}
7012
7013static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
7014{
7015        /* Restore tagless access via VLAN 0 */
7016        igb_set_vf_vlan(adapter, 0, true, vf);
7017
7018        igb_set_vmvir(adapter, 0, vf);
7019        igb_set_vmolr(adapter, vf, true);
7020
7021        /* Remove any PF assigned VLAN */
7022        if (adapter->vf_data[vf].pf_vlan)
7023                igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7024                                false, vf);
7025
7026        adapter->vf_data[vf].pf_vlan = 0;
7027        adapter->vf_data[vf].pf_qos = 0;
7028        igb_set_vf_vlan_strip(adapter, vf, false);
7029
7030        return 0;
7031}
7032
7033static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
7034                               u16 vlan, u8 qos, __be16 vlan_proto)
7035{
7036        struct igb_adapter *adapter = netdev_priv(netdev);
7037
7038        if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
7039                return -EINVAL;
7040
7041        if (vlan_proto != htons(ETH_P_8021Q))
7042                return -EPROTONOSUPPORT;
7043
7044        return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
7045                               igb_disable_port_vlan(adapter, vf);
7046}
7047
7048static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7049{
7050        int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7051        int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
7052        int ret;
7053
7054        if (adapter->vf_data[vf].pf_vlan)
7055                return -1;
7056
7057        /* VLAN 0 is a special case, don't allow it to be removed */
7058        if (!vid && !add)
7059                return 0;
7060
7061        ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
7062        if (!ret)
7063                igb_set_vf_vlan_strip(adapter, vf, !!vid);
7064        return ret;
7065}
7066
7067static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
7068{
7069        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7070
7071        /* clear flags - except flag that indicates PF has set the MAC */
7072        vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
7073        vf_data->last_nack = jiffies;
7074
7075        /* reset vlans for device */
7076        igb_clear_vf_vfta(adapter, vf);
7077        igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
7078        igb_set_vmvir(adapter, vf_data->pf_vlan |
7079                               (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
7080        igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
7081        igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
7082
7083        /* reset multicast table array for vf */
7084        adapter->vf_data[vf].num_vf_mc_hashes = 0;
7085
7086        /* Flush and reset the mta with the new values */
7087        igb_set_rx_mode(adapter->netdev);
7088}
7089
7090static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
7091{
7092        unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7093
7094        /* clear mac address as we were hotplug removed/added */
7095        if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
7096                eth_zero_addr(vf_mac);
7097
7098        /* process remaining reset events */
7099        igb_vf_reset(adapter, vf);
7100}
7101
7102static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
7103{
7104        struct e1000_hw *hw = &adapter->hw;
7105        unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7106        u32 reg, msgbuf[3];
7107        u8 *addr = (u8 *)(&msgbuf[1]);
7108
7109        /* process all the same items cleared in a function level reset */
7110        igb_vf_reset(adapter, vf);
7111
7112        /* set vf mac address */
7113        igb_set_vf_mac(adapter, vf, vf_mac);
7114
7115        /* enable transmit and receive for vf */
7116        reg = rd32(E1000_VFTE);
7117        wr32(E1000_VFTE, reg | BIT(vf));
7118        reg = rd32(E1000_VFRE);
7119        wr32(E1000_VFRE, reg | BIT(vf));
7120
7121        adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
7122
7123        /* reply to reset with ack and vf mac address */
7124        if (!is_zero_ether_addr(vf_mac)) {
7125                msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
7126                memcpy(addr, vf_mac, ETH_ALEN);
7127        } else {
7128                msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
7129        }
7130        igb_write_mbx(hw, msgbuf, 3, vf);
7131}
7132
7133static void igb_flush_mac_table(struct igb_adapter *adapter)
7134{
7135        struct e1000_hw *hw = &adapter->hw;
7136        int i;
7137
7138        for (i = 0; i < hw->mac.rar_entry_count; i++) {
7139                adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
7140                memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
7141                adapter->mac_table[i].queue = 0;
7142                igb_rar_set_index(adapter, i);
7143        }
7144}
7145
7146static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
7147{
7148        struct e1000_hw *hw = &adapter->hw;
7149        /* do not count rar entries reserved for VFs MAC addresses */
7150        int rar_entries = hw->mac.rar_entry_count -
7151                          adapter->vfs_allocated_count;
7152        int i, count = 0;
7153
7154        for (i = 0; i < rar_entries; i++) {
7155                /* do not count default entries */
7156                if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
7157                        continue;
7158
7159                /* do not count "in use" entries for different queues */
7160                if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
7161                    (adapter->mac_table[i].queue != queue))
7162                        continue;
7163
7164                count++;
7165        }
7166
7167        return count;
7168}
7169
7170/* Set default MAC address for the PF in the first RAR entry */
7171static void igb_set_default_mac_filter(struct igb_adapter *adapter)
7172{
7173        struct igb_mac_addr *mac_table = &adapter->mac_table[0];
7174
7175        ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
7176        mac_table->queue = adapter->vfs_allocated_count;
7177        mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7178
7179        igb_rar_set_index(adapter, 0);
7180}
7181
7182/* If the filter to be added and an already existing filter express
7183 * the same address and address type, it should be possible to only
7184 * override the other configurations, for example the queue to steer
7185 * traffic.
7186 */
7187static bool igb_mac_entry_can_be_used(const struct igb_mac_addr *entry,
7188                                      const u8 *addr, const u8 flags)
7189{
7190        if (!(entry->state & IGB_MAC_STATE_IN_USE))
7191                return true;
7192
7193        if ((entry->state & IGB_MAC_STATE_SRC_ADDR) !=
7194            (flags & IGB_MAC_STATE_SRC_ADDR))
7195                return false;
7196
7197        if (!ether_addr_equal(addr, entry->addr))
7198                return false;
7199
7200        return true;
7201}
7202
7203/* Add a MAC filter for 'addr' directing matching traffic to 'queue',
7204 * 'flags' is used to indicate what kind of match is made, match is by
7205 * default for the destination address, if matching by source address
7206 * is desired the flag IGB_MAC_STATE_SRC_ADDR can be used.
7207 */
7208static int igb_add_mac_filter_flags(struct igb_adapter *adapter,
7209                                    const u8 *addr, const u8 queue,
7210                                    const u8 flags)
7211{
7212        struct e1000_hw *hw = &adapter->hw;
7213        int rar_entries = hw->mac.rar_entry_count -
7214                          adapter->vfs_allocated_count;
7215        int i;
7216
7217        if (is_zero_ether_addr(addr))
7218                return -EINVAL;
7219
7220        /* Search for the first empty entry in the MAC table.
7221         * Do not touch entries at the end of the table reserved for the VF MAC
7222         * addresses.
7223         */
7224        for (i = 0; i < rar_entries; i++) {
7225                if (!igb_mac_entry_can_be_used(&adapter->mac_table[i],
7226                                               addr, flags))
7227                        continue;
7228
7229                ether_addr_copy(adapter->mac_table[i].addr, addr);
7230                adapter->mac_table[i].queue = queue;
7231                adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE | flags;
7232
7233                igb_rar_set_index(adapter, i);
7234                return i;
7235        }
7236
7237        return -ENOSPC;
7238}
7239
7240static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7241                              const u8 queue)
7242{
7243        return igb_add_mac_filter_flags(adapter, addr, queue, 0);
7244}
7245
7246/* Remove a MAC filter for 'addr' directing matching traffic to
7247 * 'queue', 'flags' is used to indicate what kind of match need to be
7248 * removed, match is by default for the destination address, if
7249 * matching by source address is to be removed the flag
7250 * IGB_MAC_STATE_SRC_ADDR can be used.
7251 */
7252static int igb_del_mac_filter_flags(struct igb_adapter *adapter,
7253                                    const u8 *addr, const u8 queue,
7254                                    const u8 flags)
7255{
7256        struct e1000_hw *hw = &adapter->hw;
7257        int rar_entries = hw->mac.rar_entry_count -
7258                          adapter->vfs_allocated_count;
7259        int i;
7260
7261        if (is_zero_ether_addr(addr))
7262                return -EINVAL;
7263
7264        /* Search for matching entry in the MAC table based on given address
7265         * and queue. Do not touch entries at the end of the table reserved
7266         * for the VF MAC addresses.
7267         */
7268        for (i = 0; i < rar_entries; i++) {
7269                if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
7270                        continue;
7271                if ((adapter->mac_table[i].state & flags) != flags)
7272                        continue;
7273                if (adapter->mac_table[i].queue != queue)
7274                        continue;
7275                if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
7276                        continue;
7277
7278                /* When a filter for the default address is "deleted",
7279                 * we return it to its initial configuration
7280                 */
7281                if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT) {
7282                        adapter->mac_table[i].state =
7283                                IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7284                        adapter->mac_table[i].queue =
7285                                adapter->vfs_allocated_count;
7286                } else {
7287                        adapter->mac_table[i].state = 0;
7288                        adapter->mac_table[i].queue = 0;
7289                        memset(adapter->mac_table[i].addr, 0, ETH_ALEN);
7290                }
7291
7292                igb_rar_set_index(adapter, i);
7293                return 0;
7294        }
7295
7296        return -ENOENT;
7297}
7298
7299static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7300                              const u8 queue)
7301{
7302        return igb_del_mac_filter_flags(adapter, addr, queue, 0);
7303}
7304
7305int igb_add_mac_steering_filter(struct igb_adapter *adapter,
7306                                const u8 *addr, u8 queue, u8 flags)
7307{
7308        struct e1000_hw *hw = &adapter->hw;
7309
7310        /* In theory, this should be supported on 82575 as well, but
7311         * that part wasn't easily accessible during development.
7312         */
7313        if (hw->mac.type != e1000_i210)
7314                return -EOPNOTSUPP;
7315
7316        return igb_add_mac_filter_flags(adapter, addr, queue,
7317                                        IGB_MAC_STATE_QUEUE_STEERING | flags);
7318}
7319
7320int igb_del_mac_steering_filter(struct igb_adapter *adapter,
7321                                const u8 *addr, u8 queue, u8 flags)
7322{
7323        return igb_del_mac_filter_flags(adapter, addr, queue,
7324                                        IGB_MAC_STATE_QUEUE_STEERING | flags);
7325}
7326
7327static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
7328{
7329        struct igb_adapter *adapter = netdev_priv(netdev);
7330        int ret;
7331
7332        ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7333
7334        return min_t(int, ret, 0);
7335}
7336
7337static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
7338{
7339        struct igb_adapter *adapter = netdev_priv(netdev);
7340
7341        igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7342
7343        return 0;
7344}
7345
7346static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
7347                                 const u32 info, const u8 *addr)
7348{
7349        struct pci_dev *pdev = adapter->pdev;
7350        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7351        struct list_head *pos;
7352        struct vf_mac_filter *entry = NULL;
7353        int ret = 0;
7354
7355        switch (info) {
7356        case E1000_VF_MAC_FILTER_CLR:
7357                /* remove all unicast MAC filters related to the current VF */
7358                list_for_each(pos, &adapter->vf_macs.l) {
7359                        entry = list_entry(pos, struct vf_mac_filter, l);
7360                        if (entry->vf == vf) {
7361                                entry->vf = -1;
7362                                entry->free = true;
7363                                igb_del_mac_filter(adapter, entry->vf_mac, vf);
7364                        }
7365                }
7366                break;
7367        case E1000_VF_MAC_FILTER_ADD:
7368                if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7369                    !vf_data->trusted) {
7370                        dev_warn(&pdev->dev,
7371                                 "VF %d requested MAC filter but is administratively denied\n",
7372                                 vf);
7373                        return -EINVAL;
7374                }
7375                if (!is_valid_ether_addr(addr)) {
7376                        dev_warn(&pdev->dev,
7377                                 "VF %d attempted to set invalid MAC filter\n",
7378                                 vf);
7379                        return -EINVAL;
7380                }
7381
7382                /* try to find empty slot in the list */
7383                list_for_each(pos, &adapter->vf_macs.l) {
7384                        entry = list_entry(pos, struct vf_mac_filter, l);
7385                        if (entry->free)
7386                                break;
7387                }
7388
7389                if (entry && entry->free) {
7390                        entry->free = false;
7391                        entry->vf = vf;
7392                        ether_addr_copy(entry->vf_mac, addr);
7393
7394                        ret = igb_add_mac_filter(adapter, addr, vf);
7395                        ret = min_t(int, ret, 0);
7396                } else {
7397                        ret = -ENOSPC;
7398                }
7399
7400                if (ret == -ENOSPC)
7401                        dev_warn(&pdev->dev,
7402                                 "VF %d has requested MAC filter but there is no space for it\n",
7403                                 vf);
7404                break;
7405        default:
7406                ret = -EINVAL;
7407                break;
7408        }
7409
7410        return ret;
7411}
7412
7413static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
7414{
7415        struct pci_dev *pdev = adapter->pdev;
7416        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7417        u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
7418
7419        /* The VF MAC Address is stored in a packed array of bytes
7420         * starting at the second 32 bit word of the msg array
7421         */
7422        unsigned char *addr = (unsigned char *)&msg[1];
7423        int ret = 0;
7424
7425        if (!info) {
7426                if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7427                    !vf_data->trusted) {
7428                        dev_warn(&pdev->dev,
7429                                 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7430                                 vf);
7431                        return -EINVAL;
7432                }
7433
7434                if (!is_valid_ether_addr(addr)) {
7435                        dev_warn(&pdev->dev,
7436                                 "VF %d attempted to set invalid MAC\n",
7437                                 vf);
7438                        return -EINVAL;
7439                }
7440
7441                ret = igb_set_vf_mac(adapter, vf, addr);
7442        } else {
7443                ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7444        }
7445
7446        return ret;
7447}
7448
7449static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7450{
7451        struct e1000_hw *hw = &adapter->hw;
7452        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7453        u32 msg = E1000_VT_MSGTYPE_NACK;
7454
7455        /* if device isn't clear to send it shouldn't be reading either */
7456        if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7457            time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7458                igb_write_mbx(hw, &msg, 1, vf);
7459                vf_data->last_nack = jiffies;
7460        }
7461}
7462
7463static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7464{
7465        struct pci_dev *pdev = adapter->pdev;
7466        u32 msgbuf[E1000_VFMAILBOX_SIZE];
7467        struct e1000_hw *hw = &adapter->hw;
7468        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7469        s32 retval;
7470
7471        retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7472
7473        if (retval) {
7474                /* if receive failed revoke VF CTS stats and restart init */
7475                dev_err(&pdev->dev, "Error receiving message from VF\n");
7476                vf_data->flags &= ~IGB_VF_FLAG_CTS;
7477                if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7478                        goto unlock;
7479                goto out;
7480        }
7481
7482        /* this is a message we already processed, do nothing */
7483        if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7484                goto unlock;
7485
7486        /* until the vf completes a reset it should not be
7487         * allowed to start any configuration.
7488         */
7489        if (msgbuf[0] == E1000_VF_RESET) {
7490                /* unlocks mailbox */
7491                igb_vf_reset_msg(adapter, vf);
7492                return;
7493        }
7494
7495        if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7496                if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7497                        goto unlock;
7498                retval = -1;
7499                goto out;
7500        }
7501
7502        switch ((msgbuf[0] & 0xFFFF)) {
7503        case E1000_VF_SET_MAC_ADDR:
7504                retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7505                break;
7506        case E1000_VF_SET_PROMISC:
7507                retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7508                break;
7509        case E1000_VF_SET_MULTICAST:
7510                retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7511                break;
7512        case E1000_VF_SET_LPE:
7513                retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7514                break;
7515        case E1000_VF_SET_VLAN:
7516                retval = -1;
7517                if (vf_data->pf_vlan)
7518                        dev_warn(&pdev->dev,
7519                                 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7520                                 vf);
7521                else
7522                        retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7523                break;
7524        default:
7525                dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7526                retval = -1;
7527                break;
7528        }
7529
7530        msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7531out:
7532        /* notify the VF of the results of what it sent us */
7533        if (retval)
7534                msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7535        else
7536                msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7537
7538        /* unlocks mailbox */
7539        igb_write_mbx(hw, msgbuf, 1, vf);
7540        return;
7541
7542unlock:
7543        igb_unlock_mbx(hw, vf);
7544}
7545
7546static void igb_msg_task(struct igb_adapter *adapter)
7547{
7548        struct e1000_hw *hw = &adapter->hw;
7549        u32 vf;
7550
7551        for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7552                /* process any reset requests */
7553                if (!igb_check_for_rst(hw, vf))
7554                        igb_vf_reset_event(adapter, vf);
7555
7556                /* process any messages pending */
7557                if (!igb_check_for_msg(hw, vf))
7558                        igb_rcv_msg_from_vf(adapter, vf);
7559
7560                /* process any acks */
7561                if (!igb_check_for_ack(hw, vf))
7562                        igb_rcv_ack_from_vf(adapter, vf);
7563        }
7564}
7565
7566/**
7567 *  igb_set_uta - Set unicast filter table address
7568 *  @adapter: board private structure
7569 *  @set: boolean indicating if we are setting or clearing bits
7570 *
7571 *  The unicast table address is a register array of 32-bit registers.
7572 *  The table is meant to be used in a way similar to how the MTA is used
7573 *  however due to certain limitations in the hardware it is necessary to
7574 *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
7575 *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
7576 **/
7577static void igb_set_uta(struct igb_adapter *adapter, bool set)
7578{
7579        struct e1000_hw *hw = &adapter->hw;
7580        u32 uta = set ? ~0 : 0;
7581        int i;
7582
7583        /* we only need to do this if VMDq is enabled */
7584        if (!adapter->vfs_allocated_count)
7585                return;
7586
7587        for (i = hw->mac.uta_reg_count; i--;)
7588                array_wr32(E1000_UTA, i, uta);
7589}
7590
7591/**
7592 *  igb_intr_msi - Interrupt Handler
7593 *  @irq: interrupt number
7594 *  @data: pointer to a network interface device structure
7595 **/
7596static irqreturn_t igb_intr_msi(int irq, void *data)
7597{
7598        struct igb_adapter *adapter = data;
7599        struct igb_q_vector *q_vector = adapter->q_vector[0];
7600        struct e1000_hw *hw = &adapter->hw;
7601        /* read ICR disables interrupts using IAM */
7602        u32 icr = rd32(E1000_ICR);
7603
7604        igb_write_itr(q_vector);
7605
7606        if (icr & E1000_ICR_DRSTA)
7607                schedule_work(&adapter->reset_task);
7608
7609        if (icr & E1000_ICR_DOUTSYNC) {
7610                /* HW is reporting DMA is out of sync */
7611                adapter->stats.doosync++;
7612        }
7613
7614        if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7615                hw->mac.get_link_status = 1;
7616                if (!test_bit(__IGB_DOWN, &adapter->state))
7617                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
7618        }
7619
7620        if (icr & E1000_ICR_TS)
7621                igb_tsync_interrupt(adapter);
7622
7623        napi_schedule(&q_vector->napi);
7624
7625        return IRQ_HANDLED;
7626}
7627
7628/**
7629 *  igb_intr - Legacy Interrupt Handler
7630 *  @irq: interrupt number
7631 *  @data: pointer to a network interface device structure
7632 **/
7633static irqreturn_t igb_intr(int irq, void *data)
7634{
7635        struct igb_adapter *adapter = data;
7636        struct igb_q_vector *q_vector = adapter->q_vector[0];
7637        struct e1000_hw *hw = &adapter->hw;
7638        /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
7639         * need for the IMC write
7640         */
7641        u32 icr = rd32(E1000_ICR);
7642
7643        /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
7644         * not set, then the adapter didn't send an interrupt
7645         */
7646        if (!(icr & E1000_ICR_INT_ASSERTED))
7647                return IRQ_NONE;
7648
7649        igb_write_itr(q_vector);
7650
7651        if (icr & E1000_ICR_DRSTA)
7652                schedule_work(&adapter->reset_task);
7653
7654        if (icr & E1000_ICR_DOUTSYNC) {
7655                /* HW is reporting DMA is out of sync */
7656                adapter->stats.doosync++;
7657        }
7658
7659        if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7660                hw->mac.get_link_status = 1;
7661                /* guard against interrupt when we're going down */
7662                if (!test_bit(__IGB_DOWN, &adapter->state))
7663                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
7664        }
7665
7666        if (icr & E1000_ICR_TS)
7667                igb_tsync_interrupt(adapter);
7668
7669        napi_schedule(&q_vector->napi);
7670
7671        return IRQ_HANDLED;
7672}
7673
7674static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
7675{
7676        struct igb_adapter *adapter = q_vector->adapter;
7677        struct e1000_hw *hw = &adapter->hw;
7678
7679        if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
7680            (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
7681                if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
7682                        igb_set_itr(q_vector);
7683                else
7684                        igb_update_ring_itr(q_vector);
7685        }
7686
7687        if (!test_bit(__IGB_DOWN, &adapter->state)) {
7688                if (adapter->flags & IGB_FLAG_HAS_MSIX)
7689                        wr32(E1000_EIMS, q_vector->eims_value);
7690                else
7691                        igb_irq_enable(adapter);
7692        }
7693}
7694
7695/**
7696 *  igb_poll - NAPI Rx polling callback
7697 *  @napi: napi polling structure
7698 *  @budget: count of how many packets we should handle
7699 **/
7700static int igb_poll(struct napi_struct *napi, int budget)
7701{
7702        struct igb_q_vector *q_vector = container_of(napi,
7703                                                     struct igb_q_vector,
7704                                                     napi);
7705        bool clean_complete = true;
7706        int work_done = 0;
7707
7708#ifdef CONFIG_IGB_DCA
7709        if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
7710                igb_update_dca(q_vector);
7711#endif
7712        if (q_vector->tx.ring)
7713                clean_complete = igb_clean_tx_irq(q_vector, budget);
7714
7715        if (q_vector->rx.ring) {
7716                int cleaned = igb_clean_rx_irq(q_vector, budget);
7717
7718                work_done += cleaned;
7719                if (cleaned >= budget)
7720                        clean_complete = false;
7721        }
7722
7723        /* If all work not completed, return budget and keep polling */
7724        if (!clean_complete)
7725                return budget;
7726
7727        /* Exit the polling mode, but don't re-enable interrupts if stack might
7728         * poll us due to busy-polling
7729         */
7730        if (likely(napi_complete_done(napi, work_done)))
7731                igb_ring_irq_enable(q_vector);
7732
7733        return min(work_done, budget - 1);
7734}
7735
7736/**
7737 *  igb_clean_tx_irq - Reclaim resources after transmit completes
7738 *  @q_vector: pointer to q_vector containing needed info
7739 *  @napi_budget: Used to determine if we are in netpoll
7740 *
7741 *  returns true if ring is completely cleaned
7742 **/
7743static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
7744{
7745        struct igb_adapter *adapter = q_vector->adapter;
7746        struct igb_ring *tx_ring = q_vector->tx.ring;
7747        struct igb_tx_buffer *tx_buffer;
7748        union e1000_adv_tx_desc *tx_desc;
7749        unsigned int total_bytes = 0, total_packets = 0;
7750        unsigned int budget = q_vector->tx.work_limit;
7751        unsigned int i = tx_ring->next_to_clean;
7752
7753        if (test_bit(__IGB_DOWN, &adapter->state))
7754                return true;
7755
7756        tx_buffer = &tx_ring->tx_buffer_info[i];
7757        tx_desc = IGB_TX_DESC(tx_ring, i);
7758        i -= tx_ring->count;
7759
7760        do {
7761                union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
7762
7763                /* if next_to_watch is not set then there is no work pending */
7764                if (!eop_desc)
7765                        break;
7766
7767                /* prevent any other reads prior to eop_desc */
7768                smp_rmb();
7769
7770                /* if DD is not set pending work has not been completed */
7771                if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
7772                        break;
7773
7774                /* clear next_to_watch to prevent false hangs */
7775                tx_buffer->next_to_watch = NULL;
7776
7777                /* update the statistics for this packet */
7778                total_bytes += tx_buffer->bytecount;
7779                total_packets += tx_buffer->gso_segs;
7780
7781                /* free the skb */
7782                napi_consume_skb(tx_buffer->skb, napi_budget);
7783
7784                /* unmap skb header data */
7785                dma_unmap_single(tx_ring->dev,
7786                                 dma_unmap_addr(tx_buffer, dma),
7787                                 dma_unmap_len(tx_buffer, len),
7788                                 DMA_TO_DEVICE);
7789
7790                /* clear tx_buffer data */
7791                dma_unmap_len_set(tx_buffer, len, 0);
7792
7793                /* clear last DMA location and unmap remaining buffers */
7794                while (tx_desc != eop_desc) {
7795                        tx_buffer++;
7796                        tx_desc++;
7797                        i++;
7798                        if (unlikely(!i)) {
7799                                i -= tx_ring->count;
7800                                tx_buffer = tx_ring->tx_buffer_info;
7801                                tx_desc = IGB_TX_DESC(tx_ring, 0);
7802                        }
7803
7804                        /* unmap any remaining paged data */
7805                        if (dma_unmap_len(tx_buffer, len)) {
7806                                dma_unmap_page(tx_ring->dev,
7807                                               dma_unmap_addr(tx_buffer, dma),
7808                                               dma_unmap_len(tx_buffer, len),
7809                                               DMA_TO_DEVICE);
7810                                dma_unmap_len_set(tx_buffer, len, 0);
7811                        }
7812                }
7813
7814                /* move us one more past the eop_desc for start of next pkt */
7815                tx_buffer++;
7816                tx_desc++;
7817                i++;
7818                if (unlikely(!i)) {
7819                        i -= tx_ring->count;
7820                        tx_buffer = tx_ring->tx_buffer_info;
7821                        tx_desc = IGB_TX_DESC(tx_ring, 0);
7822                }
7823
7824                /* issue prefetch for next Tx descriptor */
7825                prefetch(tx_desc);
7826
7827                /* update budget accounting */
7828                budget--;
7829        } while (likely(budget));
7830
7831        netdev_tx_completed_queue(txring_txq(tx_ring),
7832                                  total_packets, total_bytes);
7833        i += tx_ring->count;
7834        tx_ring->next_to_clean = i;
7835        u64_stats_update_begin(&tx_ring->tx_syncp);
7836        tx_ring->tx_stats.bytes += total_bytes;
7837        tx_ring->tx_stats.packets += total_packets;
7838        u64_stats_update_end(&tx_ring->tx_syncp);
7839        q_vector->tx.total_bytes += total_bytes;
7840        q_vector->tx.total_packets += total_packets;
7841
7842        if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
7843                struct e1000_hw *hw = &adapter->hw;
7844
7845                /* Detect a transmit hang in hardware, this serializes the
7846                 * check with the clearing of time_stamp and movement of i
7847                 */
7848                clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
7849                if (tx_buffer->next_to_watch &&
7850                    time_after(jiffies, tx_buffer->time_stamp +
7851                               (adapter->tx_timeout_factor * HZ)) &&
7852                    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
7853
7854                        /* detected Tx unit hang */
7855                        dev_err(tx_ring->dev,
7856                                "Detected Tx Unit Hang\n"
7857                                "  Tx Queue             <%d>\n"
7858                                "  TDH                  <%x>\n"
7859                                "  TDT                  <%x>\n"
7860                                "  next_to_use          <%x>\n"
7861                                "  next_to_clean        <%x>\n"
7862                                "buffer_info[next_to_clean]\n"
7863                                "  time_stamp           <%lx>\n"
7864                                "  next_to_watch        <%p>\n"
7865                                "  jiffies              <%lx>\n"
7866                                "  desc.status          <%x>\n",
7867                                tx_ring->queue_index,
7868                                rd32(E1000_TDH(tx_ring->reg_idx)),
7869                                readl(tx_ring->tail),
7870                                tx_ring->next_to_use,
7871                                tx_ring->next_to_clean,
7872                                tx_buffer->time_stamp,
7873                                tx_buffer->next_to_watch,
7874                                jiffies,
7875                                tx_buffer->next_to_watch->wb.status);
7876                        netif_stop_subqueue(tx_ring->netdev,
7877                                            tx_ring->queue_index);
7878
7879                        /* we are about to reset, no point in enabling stuff */
7880                        return true;
7881                }
7882        }
7883
7884#define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
7885        if (unlikely(total_packets &&
7886            netif_carrier_ok(tx_ring->netdev) &&
7887            igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
7888                /* Make sure that anybody stopping the queue after this
7889                 * sees the new next_to_clean.
7890                 */
7891                smp_mb();
7892                if (__netif_subqueue_stopped(tx_ring->netdev,
7893                                             tx_ring->queue_index) &&
7894                    !(test_bit(__IGB_DOWN, &adapter->state))) {
7895                        netif_wake_subqueue(tx_ring->netdev,
7896                                            tx_ring->queue_index);
7897
7898                        u64_stats_update_begin(&tx_ring->tx_syncp);
7899                        tx_ring->tx_stats.restart_queue++;
7900                        u64_stats_update_end(&tx_ring->tx_syncp);
7901                }
7902        }
7903
7904        return !!budget;
7905}
7906
7907/**
7908 *  igb_reuse_rx_page - page flip buffer and store it back on the ring
7909 *  @rx_ring: rx descriptor ring to store buffers on
7910 *  @old_buff: donor buffer to have page reused
7911 *
7912 *  Synchronizes page for reuse by the adapter
7913 **/
7914static void igb_reuse_rx_page(struct igb_ring *rx_ring,
7915                              struct igb_rx_buffer *old_buff)
7916{
7917        struct igb_rx_buffer *new_buff;
7918        u16 nta = rx_ring->next_to_alloc;
7919
7920        new_buff = &rx_ring->rx_buffer_info[nta];
7921
7922        /* update, and store next to alloc */
7923        nta++;
7924        rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
7925
7926        /* Transfer page from old buffer to new buffer.
7927         * Move each member individually to avoid possible store
7928         * forwarding stalls.
7929         */
7930        new_buff->dma           = old_buff->dma;
7931        new_buff->page          = old_buff->page;
7932        new_buff->page_offset   = old_buff->page_offset;
7933        new_buff->pagecnt_bias  = old_buff->pagecnt_bias;
7934}
7935
7936static inline bool igb_page_is_reserved(struct page *page)
7937{
7938        return (page_to_nid(page) != numa_mem_id()) || page_is_pfmemalloc(page);
7939}
7940
7941static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer)
7942{
7943        unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
7944        struct page *page = rx_buffer->page;
7945
7946        /* avoid re-using remote pages */
7947        if (unlikely(igb_page_is_reserved(page)))
7948                return false;
7949
7950#if (PAGE_SIZE < 8192)
7951        /* if we are only owner of page we can reuse it */
7952        if (unlikely((page_ref_count(page) - pagecnt_bias) > 1))
7953                return false;
7954#else
7955#define IGB_LAST_OFFSET \
7956        (SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
7957
7958        if (rx_buffer->page_offset > IGB_LAST_OFFSET)
7959                return false;
7960#endif
7961
7962        /* If we have drained the page fragment pool we need to update
7963         * the pagecnt_bias and page count so that we fully restock the
7964         * number of references the driver holds.
7965         */
7966        if (unlikely(!pagecnt_bias)) {
7967                page_ref_add(page, USHRT_MAX);
7968                rx_buffer->pagecnt_bias = USHRT_MAX;
7969        }
7970
7971        return true;
7972}
7973
7974/**
7975 *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
7976 *  @rx_ring: rx descriptor ring to transact packets on
7977 *  @rx_buffer: buffer containing page to add
7978 *  @skb: sk_buff to place the data into
7979 *  @size: size of buffer to be added
7980 *
7981 *  This function will add the data contained in rx_buffer->page to the skb.
7982 **/
7983static void igb_add_rx_frag(struct igb_ring *rx_ring,
7984                            struct igb_rx_buffer *rx_buffer,
7985                            struct sk_buff *skb,
7986                            unsigned int size)
7987{
7988#if (PAGE_SIZE < 8192)
7989        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
7990#else
7991        unsigned int truesize = ring_uses_build_skb(rx_ring) ?
7992                                SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
7993                                SKB_DATA_ALIGN(size);
7994#endif
7995        skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
7996                        rx_buffer->page_offset, size, truesize);
7997#if (PAGE_SIZE < 8192)
7998        rx_buffer->page_offset ^= truesize;
7999#else
8000        rx_buffer->page_offset += truesize;
8001#endif
8002}
8003
8004static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
8005                                         struct igb_rx_buffer *rx_buffer,
8006                                         union e1000_adv_rx_desc *rx_desc,
8007                                         unsigned int size)
8008{
8009        void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
8010#if (PAGE_SIZE < 8192)
8011        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8012#else
8013        unsigned int truesize = SKB_DATA_ALIGN(size);
8014#endif
8015        unsigned int headlen;
8016        struct sk_buff *skb;
8017
8018        /* prefetch first cache line of first page */
8019        prefetch(va);
8020#if L1_CACHE_BYTES < 128
8021        prefetch(va + L1_CACHE_BYTES);
8022#endif
8023
8024        /* allocate a skb to store the frags */
8025        skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
8026        if (unlikely(!skb))
8027                return NULL;
8028
8029        if (unlikely(igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))) {
8030                igb_ptp_rx_pktstamp(rx_ring->q_vector, va, skb);
8031                va += IGB_TS_HDR_LEN;
8032                size -= IGB_TS_HDR_LEN;
8033        }
8034
8035        /* Determine available headroom for copy */
8036        headlen = size;
8037        if (headlen > IGB_RX_HDR_LEN)
8038                headlen = eth_get_headlen(skb->dev, va, IGB_RX_HDR_LEN);
8039
8040        /* align pull length to size of long to optimize memcpy performance */
8041        memcpy(__skb_put(skb, headlen), va, ALIGN(headlen, sizeof(long)));
8042
8043        /* update all of the pointers */
8044        size -= headlen;
8045        if (size) {
8046                skb_add_rx_frag(skb, 0, rx_buffer->page,
8047                                (va + headlen) - page_address(rx_buffer->page),
8048                                size, truesize);
8049#if (PAGE_SIZE < 8192)
8050                rx_buffer->page_offset ^= truesize;
8051#else
8052                rx_buffer->page_offset += truesize;
8053#endif
8054        } else {
8055                rx_buffer->pagecnt_bias++;
8056        }
8057
8058        return skb;
8059}
8060
8061static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
8062                                     struct igb_rx_buffer *rx_buffer,
8063                                     union e1000_adv_rx_desc *rx_desc,
8064                                     unsigned int size)
8065{
8066        void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
8067#if (PAGE_SIZE < 8192)
8068        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8069#else
8070        unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
8071                                SKB_DATA_ALIGN(IGB_SKB_PAD + size);
8072#endif
8073        struct sk_buff *skb;
8074
8075        /* prefetch first cache line of first page */
8076        prefetch(va);
8077#if L1_CACHE_BYTES < 128
8078        prefetch(va + L1_CACHE_BYTES);
8079#endif
8080
8081        /* build an skb around the page buffer */
8082        skb = build_skb(va - IGB_SKB_PAD, truesize);
8083        if (unlikely(!skb))
8084                return NULL;
8085
8086        /* update pointers within the skb to store the data */
8087        skb_reserve(skb, IGB_SKB_PAD);
8088        __skb_put(skb, size);
8089
8090        /* pull timestamp out of packet data */
8091        if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
8092                igb_ptp_rx_pktstamp(rx_ring->q_vector, skb->data, skb);
8093                __skb_pull(skb, IGB_TS_HDR_LEN);
8094        }
8095
8096        /* update buffer offset */
8097#if (PAGE_SIZE < 8192)
8098        rx_buffer->page_offset ^= truesize;
8099#else
8100        rx_buffer->page_offset += truesize;
8101#endif
8102
8103        return skb;
8104}
8105
8106static inline void igb_rx_checksum(struct igb_ring *ring,
8107                                   union e1000_adv_rx_desc *rx_desc,
8108                                   struct sk_buff *skb)
8109{
8110        skb_checksum_none_assert(skb);
8111
8112        /* Ignore Checksum bit is set */
8113        if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
8114                return;
8115
8116        /* Rx checksum disabled via ethtool */
8117        if (!(ring->netdev->features & NETIF_F_RXCSUM))
8118                return;
8119
8120        /* TCP/UDP checksum error bit is set */
8121        if (igb_test_staterr(rx_desc,
8122                             E1000_RXDEXT_STATERR_TCPE |
8123                             E1000_RXDEXT_STATERR_IPE)) {
8124                /* work around errata with sctp packets where the TCPE aka
8125                 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
8126                 * packets, (aka let the stack check the crc32c)
8127                 */
8128                if (!((skb->len == 60) &&
8129                      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
8130                        u64_stats_update_begin(&ring->rx_syncp);
8131                        ring->rx_stats.csum_err++;
8132                        u64_stats_update_end(&ring->rx_syncp);
8133                }
8134                /* let the stack verify checksum errors */
8135                return;
8136        }
8137        /* It must be a TCP or UDP packet with a valid checksum */
8138        if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
8139                                      E1000_RXD_STAT_UDPCS))
8140                skb->ip_summed = CHECKSUM_UNNECESSARY;
8141
8142        dev_dbg(ring->dev, "cksum success: bits %08X\n",
8143                le32_to_cpu(rx_desc->wb.upper.status_error));
8144}
8145
8146static inline void igb_rx_hash(struct igb_ring *ring,
8147                               union e1000_adv_rx_desc *rx_desc,
8148                               struct sk_buff *skb)
8149{
8150        if (ring->netdev->features & NETIF_F_RXHASH)
8151                skb_set_hash(skb,
8152                             le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
8153                             PKT_HASH_TYPE_L3);
8154}
8155
8156/**
8157 *  igb_is_non_eop - process handling of non-EOP buffers
8158 *  @rx_ring: Rx ring being processed
8159 *  @rx_desc: Rx descriptor for current buffer
8160 *  @skb: current socket buffer containing buffer in progress
8161 *
8162 *  This function updates next to clean.  If the buffer is an EOP buffer
8163 *  this function exits returning false, otherwise it will place the
8164 *  sk_buff in the next buffer to be chained and return true indicating
8165 *  that this is in fact a non-EOP buffer.
8166 **/
8167static bool igb_is_non_eop(struct igb_ring *rx_ring,
8168                           union e1000_adv_rx_desc *rx_desc)
8169{
8170        u32 ntc = rx_ring->next_to_clean + 1;
8171
8172        /* fetch, update, and store next to clean */
8173        ntc = (ntc < rx_ring->count) ? ntc : 0;
8174        rx_ring->next_to_clean = ntc;
8175
8176        prefetch(IGB_RX_DESC(rx_ring, ntc));
8177
8178        if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
8179                return false;
8180
8181        return true;
8182}
8183
8184/**
8185 *  igb_cleanup_headers - Correct corrupted or empty headers
8186 *  @rx_ring: rx descriptor ring packet is being transacted on
8187 *  @rx_desc: pointer to the EOP Rx descriptor
8188 *  @skb: pointer to current skb being fixed
8189 *
8190 *  Address the case where we are pulling data in on pages only
8191 *  and as such no data is present in the skb header.
8192 *
8193 *  In addition if skb is not at least 60 bytes we need to pad it so that
8194 *  it is large enough to qualify as a valid Ethernet frame.
8195 *
8196 *  Returns true if an error was encountered and skb was freed.
8197 **/
8198static bool igb_cleanup_headers(struct igb_ring *rx_ring,
8199                                union e1000_adv_rx_desc *rx_desc,
8200                                struct sk_buff *skb)
8201{
8202        if (unlikely((igb_test_staterr(rx_desc,
8203                                       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
8204                struct net_device *netdev = rx_ring->netdev;
8205                if (!(netdev->features & NETIF_F_RXALL)) {
8206                        dev_kfree_skb_any(skb);
8207                        return true;
8208                }
8209        }
8210
8211        /* if eth_skb_pad returns an error the skb was freed */
8212        if (eth_skb_pad(skb))
8213                return true;
8214
8215        return false;
8216}
8217
8218/**
8219 *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
8220 *  @rx_ring: rx descriptor ring packet is being transacted on
8221 *  @rx_desc: pointer to the EOP Rx descriptor
8222 *  @skb: pointer to current skb being populated
8223 *
8224 *  This function checks the ring, descriptor, and packet information in
8225 *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
8226 *  other fields within the skb.
8227 **/
8228static void igb_process_skb_fields(struct igb_ring *rx_ring,
8229                                   union e1000_adv_rx_desc *rx_desc,
8230                                   struct sk_buff *skb)
8231{
8232        struct net_device *dev = rx_ring->netdev;
8233
8234        igb_rx_hash(rx_ring, rx_desc, skb);
8235
8236        igb_rx_checksum(rx_ring, rx_desc, skb);
8237
8238        if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
8239            !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
8240                igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
8241
8242        if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
8243            igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
8244                u16 vid;
8245
8246                if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
8247                    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
8248                        vid = be16_to_cpu(rx_desc->wb.upper.vlan);
8249                else
8250                        vid = le16_to_cpu(rx_desc->wb.upper.vlan);
8251
8252                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
8253        }
8254
8255        skb_record_rx_queue(skb, rx_ring->queue_index);
8256
8257        skb->protocol = eth_type_trans(skb, rx_ring->netdev);
8258}
8259
8260static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
8261                                               const unsigned int size)
8262{
8263        struct igb_rx_buffer *rx_buffer;
8264
8265        rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
8266        prefetchw(rx_buffer->page);
8267
8268        /* we are reusing so sync this buffer for CPU use */
8269        dma_sync_single_range_for_cpu(rx_ring->dev,
8270                                      rx_buffer->dma,
8271                                      rx_buffer->page_offset,
8272                                      size,
8273                                      DMA_FROM_DEVICE);
8274
8275        rx_buffer->pagecnt_bias--;
8276
8277        return rx_buffer;
8278}
8279
8280static void igb_put_rx_buffer(struct igb_ring *rx_ring,
8281                              struct igb_rx_buffer *rx_buffer)
8282{
8283        if (igb_can_reuse_rx_page(rx_buffer)) {
8284                /* hand second half of page back to the ring */
8285                igb_reuse_rx_page(rx_ring, rx_buffer);
8286        } else {
8287                /* We are not reusing the buffer so unmap it and free
8288                 * any references we are holding to it
8289                 */
8290                dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
8291                                     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
8292                                     IGB_RX_DMA_ATTR);
8293                __page_frag_cache_drain(rx_buffer->page,
8294                                        rx_buffer->pagecnt_bias);
8295        }
8296
8297        /* clear contents of rx_buffer */
8298        rx_buffer->page = NULL;
8299}
8300
8301static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
8302{
8303        struct igb_ring *rx_ring = q_vector->rx.ring;
8304        struct sk_buff *skb = rx_ring->skb;
8305        unsigned int total_bytes = 0, total_packets = 0;
8306        u16 cleaned_count = igb_desc_unused(rx_ring);
8307
8308        while (likely(total_packets < budget)) {
8309                union e1000_adv_rx_desc *rx_desc;
8310                struct igb_rx_buffer *rx_buffer;
8311                unsigned int size;
8312
8313                /* return some buffers to hardware, one at a time is too slow */
8314                if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
8315                        igb_alloc_rx_buffers(rx_ring, cleaned_count);
8316                        cleaned_count = 0;
8317                }
8318
8319                rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
8320                size = le16_to_cpu(rx_desc->wb.upper.length);
8321                if (!size)
8322                        break;
8323
8324                /* This memory barrier is needed to keep us from reading
8325                 * any other fields out of the rx_desc until we know the
8326                 * descriptor has been written back
8327                 */
8328                dma_rmb();
8329
8330                rx_buffer = igb_get_rx_buffer(rx_ring, size);
8331
8332                /* retrieve a buffer from the ring */
8333                if (skb)
8334                        igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
8335                else if (ring_uses_build_skb(rx_ring))
8336                        skb = igb_build_skb(rx_ring, rx_buffer, rx_desc, size);
8337                else
8338                        skb = igb_construct_skb(rx_ring, rx_buffer,
8339                                                rx_desc, size);
8340
8341                /* exit if we failed to retrieve a buffer */
8342                if (!skb) {
8343                        rx_ring->rx_stats.alloc_failed++;
8344                        rx_buffer->pagecnt_bias++;
8345                        break;
8346                }
8347
8348                igb_put_rx_buffer(rx_ring, rx_buffer);
8349                cleaned_count++;
8350
8351                /* fetch next buffer in frame if non-eop */
8352                if (igb_is_non_eop(rx_ring, rx_desc))
8353                        continue;
8354
8355                /* verify the packet layout is correct */
8356                if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
8357                        skb = NULL;
8358                        continue;
8359                }
8360
8361                /* probably a little skewed due to removing CRC */
8362                total_bytes += skb->len;
8363
8364                /* populate checksum, timestamp, VLAN, and protocol */
8365                igb_process_skb_fields(rx_ring, rx_desc, skb);
8366
8367                napi_gro_receive(&q_vector->napi, skb);
8368
8369                /* reset skb pointer */
8370                skb = NULL;
8371
8372                /* update budget accounting */
8373                total_packets++;
8374        }
8375
8376        /* place incomplete frames back on ring for completion */
8377        rx_ring->skb = skb;
8378
8379        u64_stats_update_begin(&rx_ring->rx_syncp);
8380        rx_ring->rx_stats.packets += total_packets;
8381        rx_ring->rx_stats.bytes += total_bytes;
8382        u64_stats_update_end(&rx_ring->rx_syncp);
8383        q_vector->rx.total_packets += total_packets;
8384        q_vector->rx.total_bytes += total_bytes;
8385
8386        if (cleaned_count)
8387                igb_alloc_rx_buffers(rx_ring, cleaned_count);
8388
8389        return total_packets;
8390}
8391
8392static inline unsigned int igb_rx_offset(struct igb_ring *rx_ring)
8393{
8394        return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
8395}
8396
8397static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
8398                                  struct igb_rx_buffer *bi)
8399{
8400        struct page *page = bi->page;
8401        dma_addr_t dma;
8402
8403        /* since we are recycling buffers we should seldom need to alloc */
8404        if (likely(page))
8405                return true;
8406
8407        /* alloc new page for storage */
8408        page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
8409        if (unlikely(!page)) {
8410                rx_ring->rx_stats.alloc_failed++;
8411                return false;
8412        }
8413
8414        /* map page for use */
8415        dma = dma_map_page_attrs(rx_ring->dev, page, 0,
8416                                 igb_rx_pg_size(rx_ring),
8417                                 DMA_FROM_DEVICE,
8418                                 IGB_RX_DMA_ATTR);
8419
8420        /* if mapping failed free memory back to system since
8421         * there isn't much point in holding memory we can't use
8422         */
8423        if (dma_mapping_error(rx_ring->dev, dma)) {
8424                __free_pages(page, igb_rx_pg_order(rx_ring));
8425
8426                rx_ring->rx_stats.alloc_failed++;
8427                return false;
8428        }
8429
8430        bi->dma = dma;
8431        bi->page = page;
8432        bi->page_offset = igb_rx_offset(rx_ring);
8433        bi->pagecnt_bias = 1;
8434
8435        return true;
8436}
8437
8438/**
8439 *  igb_alloc_rx_buffers - Replace used receive buffers; packet split
8440 *  @adapter: address of board private structure
8441 **/
8442void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
8443{
8444        union e1000_adv_rx_desc *rx_desc;
8445        struct igb_rx_buffer *bi;
8446        u16 i = rx_ring->next_to_use;
8447        u16 bufsz;
8448
8449        /* nothing to do */
8450        if (!cleaned_count)
8451                return;
8452
8453        rx_desc = IGB_RX_DESC(rx_ring, i);
8454        bi = &rx_ring->rx_buffer_info[i];
8455        i -= rx_ring->count;
8456
8457        bufsz = igb_rx_bufsz(rx_ring);
8458
8459        do {
8460                if (!igb_alloc_mapped_page(rx_ring, bi))
8461                        break;
8462
8463                /* sync the buffer for use by the device */
8464                dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
8465                                                 bi->page_offset, bufsz,
8466                                                 DMA_FROM_DEVICE);
8467
8468                /* Refresh the desc even if buffer_addrs didn't change
8469                 * because each write-back erases this info.
8470                 */
8471                rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
8472
8473                rx_desc++;
8474                bi++;
8475                i++;
8476                if (unlikely(!i)) {
8477                        rx_desc = IGB_RX_DESC(rx_ring, 0);
8478                        bi = rx_ring->rx_buffer_info;
8479                        i -= rx_ring->count;
8480                }
8481
8482                /* clear the length for the next_to_use descriptor */
8483                rx_desc->wb.upper.length = 0;
8484
8485                cleaned_count--;
8486        } while (cleaned_count);
8487
8488        i += rx_ring->count;
8489
8490        if (rx_ring->next_to_use != i) {
8491                /* record the next descriptor to use */
8492                rx_ring->next_to_use = i;
8493
8494                /* update next to alloc since we have filled the ring */
8495                rx_ring->next_to_alloc = i;
8496
8497                /* Force memory writes to complete before letting h/w
8498                 * know there are new descriptors to fetch.  (Only
8499                 * applicable for weak-ordered memory model archs,
8500                 * such as IA-64).
8501                 */
8502                dma_wmb();
8503                writel(i, rx_ring->tail);
8504        }
8505}
8506
8507/**
8508 * igb_mii_ioctl -
8509 * @netdev:
8510 * @ifreq:
8511 * @cmd:
8512 **/
8513static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8514{
8515        struct igb_adapter *adapter = netdev_priv(netdev);
8516        struct mii_ioctl_data *data = if_mii(ifr);
8517
8518        if (adapter->hw.phy.media_type != e1000_media_type_copper)
8519                return -EOPNOTSUPP;
8520
8521        switch (cmd) {
8522        case SIOCGMIIPHY:
8523                data->phy_id = adapter->hw.phy.addr;
8524                break;
8525        case SIOCGMIIREG:
8526                if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
8527                                     &data->val_out))
8528                        return -EIO;
8529                break;
8530        case SIOCSMIIREG:
8531        default:
8532                return -EOPNOTSUPP;
8533        }
8534        return 0;
8535}
8536
8537/**
8538 * igb_ioctl -
8539 * @netdev:
8540 * @ifreq:
8541 * @cmd:
8542 **/
8543static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
8544{
8545        switch (cmd) {
8546        case SIOCGMIIPHY:
8547        case SIOCGMIIREG:
8548        case SIOCSMIIREG:
8549                return igb_mii_ioctl(netdev, ifr, cmd);
8550        case SIOCGHWTSTAMP:
8551                return igb_ptp_get_ts_config(netdev, ifr);
8552        case SIOCSHWTSTAMP:
8553                return igb_ptp_set_ts_config(netdev, ifr);
8554        default:
8555                return -EOPNOTSUPP;
8556        }
8557}
8558
8559void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8560{
8561        struct igb_adapter *adapter = hw->back;
8562
8563        pci_read_config_word(adapter->pdev, reg, value);
8564}
8565
8566void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
8567{
8568        struct igb_adapter *adapter = hw->back;
8569
8570        pci_write_config_word(adapter->pdev, reg, *value);
8571}
8572
8573s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8574{
8575        struct igb_adapter *adapter = hw->back;
8576
8577        if (pcie_capability_read_word(adapter->pdev, reg, value))
8578                return -E1000_ERR_CONFIG;
8579
8580        return 0;
8581}
8582
8583s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
8584{
8585        struct igb_adapter *adapter = hw->back;
8586
8587        if (pcie_capability_write_word(adapter->pdev, reg, *value))
8588                return -E1000_ERR_CONFIG;
8589
8590        return 0;
8591}
8592
8593static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
8594{
8595        struct igb_adapter *adapter = netdev_priv(netdev);
8596        struct e1000_hw *hw = &adapter->hw;
8597        u32 ctrl, rctl;
8598        bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
8599
8600        if (enable) {
8601                /* enable VLAN tag insert/strip */
8602                ctrl = rd32(E1000_CTRL);
8603                ctrl |= E1000_CTRL_VME;
8604                wr32(E1000_CTRL, ctrl);
8605
8606                /* Disable CFI check */
8607                rctl = rd32(E1000_RCTL);
8608                rctl &= ~E1000_RCTL_CFIEN;
8609                wr32(E1000_RCTL, rctl);
8610        } else {
8611                /* disable VLAN tag insert/strip */
8612                ctrl = rd32(E1000_CTRL);
8613                ctrl &= ~E1000_CTRL_VME;
8614                wr32(E1000_CTRL, ctrl);
8615        }
8616
8617        igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
8618}
8619
8620static int igb_vlan_rx_add_vid(struct net_device *netdev,
8621                               __be16 proto, u16 vid)
8622{
8623        struct igb_adapter *adapter = netdev_priv(netdev);
8624        struct e1000_hw *hw = &adapter->hw;
8625        int pf_id = adapter->vfs_allocated_count;
8626
8627        /* add the filter since PF can receive vlans w/o entry in vlvf */
8628        if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8629                igb_vfta_set(hw, vid, pf_id, true, !!vid);
8630
8631        set_bit(vid, adapter->active_vlans);
8632
8633        return 0;
8634}
8635
8636static int igb_vlan_rx_kill_vid(struct net_device *netdev,
8637                                __be16 proto, u16 vid)
8638{
8639        struct igb_adapter *adapter = netdev_priv(netdev);
8640        int pf_id = adapter->vfs_allocated_count;
8641        struct e1000_hw *hw = &adapter->hw;
8642
8643        /* remove VID from filter table */
8644        if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
8645                igb_vfta_set(hw, vid, pf_id, false, true);
8646
8647        clear_bit(vid, adapter->active_vlans);
8648
8649        return 0;
8650}
8651
8652static void igb_restore_vlan(struct igb_adapter *adapter)
8653{
8654        u16 vid = 1;
8655
8656        igb_vlan_mode(adapter->netdev, adapter->netdev->features);
8657        igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
8658
8659        for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
8660                igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
8661}
8662
8663int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
8664{
8665        struct pci_dev *pdev = adapter->pdev;
8666        struct e1000_mac_info *mac = &adapter->hw.mac;
8667
8668        mac->autoneg = 0;
8669
8670        /* Make sure dplx is at most 1 bit and lsb of speed is not set
8671         * for the switch() below to work
8672         */
8673        if ((spd & 1) || (dplx & ~1))
8674                goto err_inval;
8675
8676        /* Fiber NIC's only allow 1000 gbps Full duplex
8677         * and 100Mbps Full duplex for 100baseFx sfp
8678         */
8679        if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
8680                switch (spd + dplx) {
8681                case SPEED_10 + DUPLEX_HALF:
8682                case SPEED_10 + DUPLEX_FULL:
8683                case SPEED_100 + DUPLEX_HALF:
8684                        goto err_inval;
8685                default:
8686                        break;
8687                }
8688        }
8689
8690        switch (spd + dplx) {
8691        case SPEED_10 + DUPLEX_HALF:
8692                mac->forced_speed_duplex = ADVERTISE_10_HALF;
8693                break;
8694        case SPEED_10 + DUPLEX_FULL:
8695                mac->forced_speed_duplex = ADVERTISE_10_FULL;
8696                break;
8697        case SPEED_100 + DUPLEX_HALF:
8698                mac->forced_speed_duplex = ADVERTISE_100_HALF;
8699                break;
8700        case SPEED_100 + DUPLEX_FULL:
8701                mac->forced_speed_duplex = ADVERTISE_100_FULL;
8702                break;
8703        case SPEED_1000 + DUPLEX_FULL:
8704                mac->autoneg = 1;
8705                adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
8706                break;
8707        case SPEED_1000 + DUPLEX_HALF: /* not supported */
8708        default:
8709                goto err_inval;
8710        }
8711
8712        /* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
8713        adapter->hw.phy.mdix = AUTO_ALL_MODES;
8714
8715        return 0;
8716
8717err_inval:
8718        dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
8719        return -EINVAL;
8720}
8721
8722static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
8723                          bool runtime)
8724{
8725        struct net_device *netdev = pci_get_drvdata(pdev);
8726        struct igb_adapter *adapter = netdev_priv(netdev);
8727        struct e1000_hw *hw = &adapter->hw;
8728        u32 ctrl, rctl, status;
8729        u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
8730        bool wake;
8731
8732        rtnl_lock();
8733        netif_device_detach(netdev);
8734
8735        if (netif_running(netdev))
8736                __igb_close(netdev, true);
8737
8738        igb_ptp_suspend(adapter);
8739
8740        igb_clear_interrupt_scheme(adapter);
8741        rtnl_unlock();
8742
8743        status = rd32(E1000_STATUS);
8744        if (status & E1000_STATUS_LU)
8745                wufc &= ~E1000_WUFC_LNKC;
8746
8747        if (wufc) {
8748                igb_setup_rctl(adapter);
8749                igb_set_rx_mode(netdev);
8750
8751                /* turn on all-multi mode if wake on multicast is enabled */
8752                if (wufc & E1000_WUFC_MC) {
8753                        rctl = rd32(E1000_RCTL);
8754                        rctl |= E1000_RCTL_MPE;
8755                        wr32(E1000_RCTL, rctl);
8756                }
8757
8758                ctrl = rd32(E1000_CTRL);
8759                ctrl |= E1000_CTRL_ADVD3WUC;
8760                wr32(E1000_CTRL, ctrl);
8761
8762                /* Allow time for pending master requests to run */
8763                igb_disable_pcie_master(hw);
8764
8765                wr32(E1000_WUC, E1000_WUC_PME_EN);
8766                wr32(E1000_WUFC, wufc);
8767        } else {
8768                wr32(E1000_WUC, 0);
8769                wr32(E1000_WUFC, 0);
8770        }
8771
8772        wake = wufc || adapter->en_mng_pt;
8773        if (!wake)
8774                igb_power_down_link(adapter);
8775        else
8776                igb_power_up_link(adapter);
8777
8778        if (enable_wake)
8779                *enable_wake = wake;
8780
8781        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
8782         * would have already happened in close and is redundant.
8783         */
8784        igb_release_hw_control(adapter);
8785
8786        pci_disable_device(pdev);
8787
8788        return 0;
8789}
8790
8791static void igb_deliver_wake_packet(struct net_device *netdev)
8792{
8793        struct igb_adapter *adapter = netdev_priv(netdev);
8794        struct e1000_hw *hw = &adapter->hw;
8795        struct sk_buff *skb;
8796        u32 wupl;
8797
8798        wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
8799
8800        /* WUPM stores only the first 128 bytes of the wake packet.
8801         * Read the packet only if we have the whole thing.
8802         */
8803        if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
8804                return;
8805
8806        skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
8807        if (!skb)
8808                return;
8809
8810        skb_put(skb, wupl);
8811
8812        /* Ensure reads are 32-bit aligned */
8813        wupl = roundup(wupl, 4);
8814
8815        memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
8816
8817        skb->protocol = eth_type_trans(skb, netdev);
8818        netif_rx(skb);
8819}
8820
8821static int __maybe_unused igb_suspend(struct device *dev)
8822{
8823        return __igb_shutdown(to_pci_dev(dev), NULL, 0);
8824}
8825
8826static int __maybe_unused igb_resume(struct device *dev)
8827{
8828        struct pci_dev *pdev = to_pci_dev(dev);
8829        struct net_device *netdev = pci_get_drvdata(pdev);
8830        struct igb_adapter *adapter = netdev_priv(netdev);
8831        struct e1000_hw *hw = &adapter->hw;
8832        u32 err, val;
8833
8834        pci_set_power_state(pdev, PCI_D0);
8835        pci_restore_state(pdev);
8836        pci_save_state(pdev);
8837
8838        if (!pci_device_is_present(pdev))
8839                return -ENODEV;
8840        err = pci_enable_device_mem(pdev);
8841        if (err) {
8842                dev_err(&pdev->dev,
8843                        "igb: Cannot enable PCI device from suspend\n");
8844                return err;
8845        }
8846        pci_set_master(pdev);
8847
8848        pci_enable_wake(pdev, PCI_D3hot, 0);
8849        pci_enable_wake(pdev, PCI_D3cold, 0);
8850
8851        if (igb_init_interrupt_scheme(adapter, true)) {
8852                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8853                return -ENOMEM;
8854        }
8855
8856        igb_reset(adapter);
8857
8858        /* let the f/w know that the h/w is now under the control of the
8859         * driver.
8860         */
8861        igb_get_hw_control(adapter);
8862
8863        val = rd32(E1000_WUS);
8864        if (val & WAKE_PKT_WUS)
8865                igb_deliver_wake_packet(netdev);
8866
8867        wr32(E1000_WUS, ~0);
8868
8869        rtnl_lock();
8870        if (!err && netif_running(netdev))
8871                err = __igb_open(netdev, true);
8872
8873        if (!err)
8874                netif_device_attach(netdev);
8875        rtnl_unlock();
8876
8877        return err;
8878}
8879
8880static int __maybe_unused igb_runtime_idle(struct device *dev)
8881{
8882        struct pci_dev *pdev = to_pci_dev(dev);
8883        struct net_device *netdev = pci_get_drvdata(pdev);
8884        struct igb_adapter *adapter = netdev_priv(netdev);
8885
8886        if (!igb_has_link(adapter))
8887                pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
8888
8889        return -EBUSY;
8890}
8891
8892static int __maybe_unused igb_runtime_suspend(struct device *dev)
8893{
8894        return __igb_shutdown(to_pci_dev(dev), NULL, 1);
8895}
8896
8897static int __maybe_unused igb_runtime_resume(struct device *dev)
8898{
8899        return igb_resume(dev);
8900}
8901
8902static void igb_shutdown(struct pci_dev *pdev)
8903{
8904        bool wake;
8905
8906        __igb_shutdown(pdev, &wake, 0);
8907
8908        if (system_state == SYSTEM_POWER_OFF) {
8909                pci_wake_from_d3(pdev, wake);
8910                pci_set_power_state(pdev, PCI_D3hot);
8911        }
8912}
8913
8914#ifdef CONFIG_PCI_IOV
8915static int igb_sriov_reinit(struct pci_dev *dev)
8916{
8917        struct net_device *netdev = pci_get_drvdata(dev);
8918        struct igb_adapter *adapter = netdev_priv(netdev);
8919        struct pci_dev *pdev = adapter->pdev;
8920
8921        rtnl_lock();
8922
8923        if (netif_running(netdev))
8924                igb_close(netdev);
8925        else
8926                igb_reset(adapter);
8927
8928        igb_clear_interrupt_scheme(adapter);
8929
8930        igb_init_queue_configuration(adapter);
8931
8932        if (igb_init_interrupt_scheme(adapter, true)) {
8933                rtnl_unlock();
8934                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
8935                return -ENOMEM;
8936        }
8937
8938        if (netif_running(netdev))
8939                igb_open(netdev);
8940
8941        rtnl_unlock();
8942
8943        return 0;
8944}
8945
8946static int igb_pci_disable_sriov(struct pci_dev *dev)
8947{
8948        int err = igb_disable_sriov(dev);
8949
8950        if (!err)
8951                err = igb_sriov_reinit(dev);
8952
8953        return err;
8954}
8955
8956static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
8957{
8958        int err = igb_enable_sriov(dev, num_vfs);
8959
8960        if (err)
8961                goto out;
8962
8963        err = igb_sriov_reinit(dev);
8964        if (!err)
8965                return num_vfs;
8966
8967out:
8968        return err;
8969}
8970
8971#endif
8972static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
8973{
8974#ifdef CONFIG_PCI_IOV
8975        if (num_vfs == 0)
8976                return igb_pci_disable_sriov(dev);
8977        else
8978                return igb_pci_enable_sriov(dev, num_vfs);
8979#endif
8980        return 0;
8981}
8982
8983/**
8984 *  igb_io_error_detected - called when PCI error is detected
8985 *  @pdev: Pointer to PCI device
8986 *  @state: The current pci connection state
8987 *
8988 *  This function is called after a PCI bus error affecting
8989 *  this device has been detected.
8990 **/
8991static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
8992                                              pci_channel_state_t state)
8993{
8994        struct net_device *netdev = pci_get_drvdata(pdev);
8995        struct igb_adapter *adapter = netdev_priv(netdev);
8996
8997        netif_device_detach(netdev);
8998
8999        if (state == pci_channel_io_perm_failure)
9000                return PCI_ERS_RESULT_DISCONNECT;
9001
9002        if (netif_running(netdev))
9003                igb_down(adapter);
9004        pci_disable_device(pdev);
9005
9006        /* Request a slot slot reset. */
9007        return PCI_ERS_RESULT_NEED_RESET;
9008}
9009
9010/**
9011 *  igb_io_slot_reset - called after the pci bus has been reset.
9012 *  @pdev: Pointer to PCI device
9013 *
9014 *  Restart the card from scratch, as if from a cold-boot. Implementation
9015 *  resembles the first-half of the igb_resume routine.
9016 **/
9017static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
9018{
9019        struct net_device *netdev = pci_get_drvdata(pdev);
9020        struct igb_adapter *adapter = netdev_priv(netdev);
9021        struct e1000_hw *hw = &adapter->hw;
9022        pci_ers_result_t result;
9023
9024        if (pci_enable_device_mem(pdev)) {
9025                dev_err(&pdev->dev,
9026                        "Cannot re-enable PCI device after reset.\n");
9027                result = PCI_ERS_RESULT_DISCONNECT;
9028        } else {
9029                pci_set_master(pdev);
9030                pci_restore_state(pdev);
9031                pci_save_state(pdev);
9032
9033                pci_enable_wake(pdev, PCI_D3hot, 0);
9034                pci_enable_wake(pdev, PCI_D3cold, 0);
9035
9036                /* In case of PCI error, adapter lose its HW address
9037                 * so we should re-assign it here.
9038                 */
9039                hw->hw_addr = adapter->io_addr;
9040
9041                igb_reset(adapter);
9042                wr32(E1000_WUS, ~0);
9043                result = PCI_ERS_RESULT_RECOVERED;
9044        }
9045
9046        return result;
9047}
9048
9049/**
9050 *  igb_io_resume - called when traffic can start flowing again.
9051 *  @pdev: Pointer to PCI device
9052 *
9053 *  This callback is called when the error recovery driver tells us that
9054 *  its OK to resume normal operation. Implementation resembles the
9055 *  second-half of the igb_resume routine.
9056 */
9057static void igb_io_resume(struct pci_dev *pdev)
9058{
9059        struct net_device *netdev = pci_get_drvdata(pdev);
9060        struct igb_adapter *adapter = netdev_priv(netdev);
9061
9062        if (netif_running(netdev)) {
9063                if (igb_up(adapter)) {
9064                        dev_err(&pdev->dev, "igb_up failed after reset\n");
9065                        return;
9066                }
9067        }
9068
9069        netif_device_attach(netdev);
9070
9071        /* let the f/w know that the h/w is now under the control of the
9072         * driver.
9073         */
9074        igb_get_hw_control(adapter);
9075}
9076
9077/**
9078 *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
9079 *  @adapter: Pointer to adapter structure
9080 *  @index: Index of the RAR entry which need to be synced with MAC table
9081 **/
9082static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
9083{
9084        struct e1000_hw *hw = &adapter->hw;
9085        u32 rar_low, rar_high;
9086        u8 *addr = adapter->mac_table[index].addr;
9087
9088        /* HW expects these to be in network order when they are plugged
9089         * into the registers which are little endian.  In order to guarantee
9090         * that ordering we need to do an leXX_to_cpup here in order to be
9091         * ready for the byteswap that occurs with writel
9092         */
9093        rar_low = le32_to_cpup((__le32 *)(addr));
9094        rar_high = le16_to_cpup((__le16 *)(addr + 4));
9095
9096        /* Indicate to hardware the Address is Valid. */
9097        if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
9098                if (is_valid_ether_addr(addr))
9099                        rar_high |= E1000_RAH_AV;
9100
9101                if (adapter->mac_table[index].state & IGB_MAC_STATE_SRC_ADDR)
9102                        rar_high |= E1000_RAH_ASEL_SRC_ADDR;
9103
9104                switch (hw->mac.type) {
9105                case e1000_82575:
9106                case e1000_i210:
9107                        if (adapter->mac_table[index].state &
9108                            IGB_MAC_STATE_QUEUE_STEERING)
9109                                rar_high |= E1000_RAH_QSEL_ENABLE;
9110
9111                        rar_high |= E1000_RAH_POOL_1 *
9112                                    adapter->mac_table[index].queue;
9113                        break;
9114                default:
9115                        rar_high |= E1000_RAH_POOL_1 <<
9116                                    adapter->mac_table[index].queue;
9117                        break;
9118                }
9119        }
9120
9121        wr32(E1000_RAL(index), rar_low);
9122        wrfl();
9123        wr32(E1000_RAH(index), rar_high);
9124        wrfl();
9125}
9126
9127static int igb_set_vf_mac(struct igb_adapter *adapter,
9128                          int vf, unsigned char *mac_addr)
9129{
9130        struct e1000_hw *hw = &adapter->hw;
9131        /* VF MAC addresses start at end of receive addresses and moves
9132         * towards the first, as a result a collision should not be possible
9133         */
9134        int rar_entry = hw->mac.rar_entry_count - (vf + 1);
9135        unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
9136
9137        ether_addr_copy(vf_mac_addr, mac_addr);
9138        ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
9139        adapter->mac_table[rar_entry].queue = vf;
9140        adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
9141        igb_rar_set_index(adapter, rar_entry);
9142
9143        return 0;
9144}
9145
9146static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
9147{
9148        struct igb_adapter *adapter = netdev_priv(netdev);
9149
9150        if (vf >= adapter->vfs_allocated_count)
9151                return -EINVAL;
9152
9153        /* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
9154         * flag and allows to overwrite the MAC via VF netdev.  This
9155         * is necessary to allow libvirt a way to restore the original
9156         * MAC after unbinding vfio-pci and reloading igbvf after shutting
9157         * down a VM.
9158         */
9159        if (is_zero_ether_addr(mac)) {
9160                adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
9161                dev_info(&adapter->pdev->dev,
9162                         "remove administratively set MAC on VF %d\n",
9163                         vf);
9164        } else if (is_valid_ether_addr(mac)) {
9165                adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
9166                dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
9167                         mac, vf);
9168                dev_info(&adapter->pdev->dev,
9169                         "Reload the VF driver to make this change effective.");
9170                /* Generate additional warning if PF is down */
9171                if (test_bit(__IGB_DOWN, &adapter->state)) {
9172                        dev_warn(&adapter->pdev->dev,
9173                                 "The VF MAC address has been set, but the PF device is not up.\n");
9174                        dev_warn(&adapter->pdev->dev,
9175                                 "Bring the PF device up before attempting to use the VF device.\n");
9176                }
9177        } else {
9178                return -EINVAL;
9179        }
9180        return igb_set_vf_mac(adapter, vf, mac);
9181}
9182
9183static int igb_link_mbps(int internal_link_speed)
9184{
9185        switch (internal_link_speed) {
9186        case SPEED_100:
9187                return 100;
9188        case SPEED_1000:
9189                return 1000;
9190        default:
9191                return 0;
9192        }
9193}
9194
9195static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
9196                                  int link_speed)
9197{
9198        int rf_dec, rf_int;
9199        u32 bcnrc_val;
9200
9201        if (tx_rate != 0) {
9202                /* Calculate the rate factor values to set */
9203                rf_int = link_speed / tx_rate;
9204                rf_dec = (link_speed - (rf_int * tx_rate));
9205                rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
9206                         tx_rate;
9207
9208                bcnrc_val = E1000_RTTBCNRC_RS_ENA;
9209                bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
9210                              E1000_RTTBCNRC_RF_INT_MASK);
9211                bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
9212        } else {
9213                bcnrc_val = 0;
9214        }
9215
9216        wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
9217        /* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
9218         * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
9219         */
9220        wr32(E1000_RTTBCNRM, 0x14);
9221        wr32(E1000_RTTBCNRC, bcnrc_val);
9222}
9223
9224static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
9225{
9226        int actual_link_speed, i;
9227        bool reset_rate = false;
9228
9229        /* VF TX rate limit was not set or not supported */
9230        if ((adapter->vf_rate_link_speed == 0) ||
9231            (adapter->hw.mac.type != e1000_82576))
9232                return;
9233
9234        actual_link_speed = igb_link_mbps(adapter->link_speed);
9235        if (actual_link_speed != adapter->vf_rate_link_speed) {
9236                reset_rate = true;
9237                adapter->vf_rate_link_speed = 0;
9238                dev_info(&adapter->pdev->dev,
9239                         "Link speed has been changed. VF Transmit rate is disabled\n");
9240        }
9241
9242        for (i = 0; i < adapter->vfs_allocated_count; i++) {
9243                if (reset_rate)
9244                        adapter->vf_data[i].tx_rate = 0;
9245
9246                igb_set_vf_rate_limit(&adapter->hw, i,
9247                                      adapter->vf_data[i].tx_rate,
9248                                      actual_link_speed);
9249        }
9250}
9251
9252static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
9253                             int min_tx_rate, int max_tx_rate)
9254{
9255        struct igb_adapter *adapter = netdev_priv(netdev);
9256        struct e1000_hw *hw = &adapter->hw;
9257        int actual_link_speed;
9258
9259        if (hw->mac.type != e1000_82576)
9260                return -EOPNOTSUPP;
9261
9262        if (min_tx_rate)
9263                return -EINVAL;
9264
9265        actual_link_speed = igb_link_mbps(adapter->link_speed);
9266        if ((vf >= adapter->vfs_allocated_count) ||
9267            (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
9268            (max_tx_rate < 0) ||
9269            (max_tx_rate > actual_link_speed))
9270                return -EINVAL;
9271
9272        adapter->vf_rate_link_speed = actual_link_speed;
9273        adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
9274        igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
9275
9276        return 0;
9277}
9278
9279static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
9280                                   bool setting)
9281{
9282        struct igb_adapter *adapter = netdev_priv(netdev);
9283        struct e1000_hw *hw = &adapter->hw;
9284        u32 reg_val, reg_offset;
9285
9286        if (!adapter->vfs_allocated_count)
9287                return -EOPNOTSUPP;
9288
9289        if (vf >= adapter->vfs_allocated_count)
9290                return -EINVAL;
9291
9292        reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
9293        reg_val = rd32(reg_offset);
9294        if (setting)
9295                reg_val |= (BIT(vf) |
9296                            BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9297        else
9298                reg_val &= ~(BIT(vf) |
9299                             BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9300        wr32(reg_offset, reg_val);
9301
9302        adapter->vf_data[vf].spoofchk_enabled = setting;
9303        return 0;
9304}
9305
9306static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
9307{
9308        struct igb_adapter *adapter = netdev_priv(netdev);
9309
9310        if (vf >= adapter->vfs_allocated_count)
9311                return -EINVAL;
9312        if (adapter->vf_data[vf].trusted == setting)
9313                return 0;
9314
9315        adapter->vf_data[vf].trusted = setting;
9316
9317        dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
9318                 vf, setting ? "" : "not ");
9319        return 0;
9320}
9321
9322static int igb_ndo_get_vf_config(struct net_device *netdev,
9323                                 int vf, struct ifla_vf_info *ivi)
9324{
9325        struct igb_adapter *adapter = netdev_priv(netdev);
9326        if (vf >= adapter->vfs_allocated_count)
9327                return -EINVAL;
9328        ivi->vf = vf;
9329        memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
9330        ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
9331        ivi->min_tx_rate = 0;
9332        ivi->vlan = adapter->vf_data[vf].pf_vlan;
9333        ivi->qos = adapter->vf_data[vf].pf_qos;
9334        ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
9335        ivi->trusted = adapter->vf_data[vf].trusted;
9336        return 0;
9337}
9338
9339static void igb_vmm_control(struct igb_adapter *adapter)
9340{
9341        struct e1000_hw *hw = &adapter->hw;
9342        u32 reg;
9343
9344        switch (hw->mac.type) {
9345        case e1000_82575:
9346        case e1000_i210:
9347        case e1000_i211:
9348        case e1000_i354:
9349        default:
9350                /* replication is not supported for 82575 */
9351                return;
9352        case e1000_82576:
9353                /* notify HW that the MAC is adding vlan tags */
9354                reg = rd32(E1000_DTXCTL);
9355                reg |= E1000_DTXCTL_VLAN_ADDED;
9356                wr32(E1000_DTXCTL, reg);
9357                /* Fall through */
9358        case e1000_82580:
9359                /* enable replication vlan tag stripping */
9360                reg = rd32(E1000_RPLOLR);
9361                reg |= E1000_RPLOLR_STRVLAN;
9362                wr32(E1000_RPLOLR, reg);
9363                /* Fall through */
9364        case e1000_i350:
9365                /* none of the above registers are supported by i350 */
9366                break;
9367        }
9368
9369        if (adapter->vfs_allocated_count) {
9370                igb_vmdq_set_loopback_pf(hw, true);
9371                igb_vmdq_set_replication_pf(hw, true);
9372                igb_vmdq_set_anti_spoofing_pf(hw, true,
9373                                              adapter->vfs_allocated_count);
9374        } else {
9375                igb_vmdq_set_loopback_pf(hw, false);
9376                igb_vmdq_set_replication_pf(hw, false);
9377        }
9378}
9379
9380static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9381{
9382        struct e1000_hw *hw = &adapter->hw;
9383        u32 dmac_thr;
9384        u16 hwm;
9385
9386        if (hw->mac.type > e1000_82580) {
9387                if (adapter->flags & IGB_FLAG_DMAC) {
9388                        u32 reg;
9389
9390                        /* force threshold to 0. */
9391                        wr32(E1000_DMCTXTH, 0);
9392
9393                        /* DMA Coalescing high water mark needs to be greater
9394                         * than the Rx threshold. Set hwm to PBA - max frame
9395                         * size in 16B units, capping it at PBA - 6KB.
9396                         */
9397                        hwm = 64 * (pba - 6);
9398                        reg = rd32(E1000_FCRTC);
9399                        reg &= ~E1000_FCRTC_RTH_COAL_MASK;
9400                        reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
9401                                & E1000_FCRTC_RTH_COAL_MASK);
9402                        wr32(E1000_FCRTC, reg);
9403
9404                        /* Set the DMA Coalescing Rx threshold to PBA - 2 * max
9405                         * frame size, capping it at PBA - 10KB.
9406                         */
9407                        dmac_thr = pba - 10;
9408                        reg = rd32(E1000_DMACR);
9409                        reg &= ~E1000_DMACR_DMACTHR_MASK;
9410                        reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
9411                                & E1000_DMACR_DMACTHR_MASK);
9412
9413                        /* transition to L0x or L1 if available..*/
9414                        reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
9415
9416                        /* watchdog timer= +-1000 usec in 32usec intervals */
9417                        reg |= (1000 >> 5);
9418
9419                        /* Disable BMC-to-OS Watchdog Enable */
9420                        if (hw->mac.type != e1000_i354)
9421                                reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
9422
9423                        wr32(E1000_DMACR, reg);
9424
9425                        /* no lower threshold to disable
9426                         * coalescing(smart fifb)-UTRESH=0
9427                         */
9428                        wr32(E1000_DMCRTRH, 0);
9429
9430                        reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
9431
9432                        wr32(E1000_DMCTLX, reg);
9433
9434                        /* free space in tx packet buffer to wake from
9435                         * DMA coal
9436                         */
9437                        wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
9438                             (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
9439
9440                        /* make low power state decision controlled
9441                         * by DMA coal
9442                         */
9443                        reg = rd32(E1000_PCIEMISC);
9444                        reg &= ~E1000_PCIEMISC_LX_DECISION;
9445                        wr32(E1000_PCIEMISC, reg);
9446                } /* endif adapter->dmac is not disabled */
9447        } else if (hw->mac.type == e1000_82580) {
9448                u32 reg = rd32(E1000_PCIEMISC);
9449
9450                wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
9451                wr32(E1000_DMACR, 0);
9452        }
9453}
9454
9455/**
9456 *  igb_read_i2c_byte - Reads 8 bit word over I2C
9457 *  @hw: pointer to hardware structure
9458 *  @byte_offset: byte offset to read
9459 *  @dev_addr: device address
9460 *  @data: value read
9461 *
9462 *  Performs byte read operation over I2C interface at
9463 *  a specified device address.
9464 **/
9465s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9466                      u8 dev_addr, u8 *data)
9467{
9468        struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9469        struct i2c_client *this_client = adapter->i2c_client;
9470        s32 status;
9471        u16 swfw_mask = 0;
9472
9473        if (!this_client)
9474                return E1000_ERR_I2C;
9475
9476        swfw_mask = E1000_SWFW_PHY0_SM;
9477
9478        if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9479                return E1000_ERR_SWFW_SYNC;
9480
9481        status = i2c_smbus_read_byte_data(this_client, byte_offset);
9482        hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9483
9484        if (status < 0)
9485                return E1000_ERR_I2C;
9486        else {
9487                *data = status;
9488                return 0;
9489        }
9490}
9491
9492/**
9493 *  igb_write_i2c_byte - Writes 8 bit word over I2C
9494 *  @hw: pointer to hardware structure
9495 *  @byte_offset: byte offset to write
9496 *  @dev_addr: device address
9497 *  @data: value to write
9498 *
9499 *  Performs byte write operation over I2C interface at
9500 *  a specified device address.
9501 **/
9502s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9503                       u8 dev_addr, u8 data)
9504{
9505        struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9506        struct i2c_client *this_client = adapter->i2c_client;
9507        s32 status;
9508        u16 swfw_mask = E1000_SWFW_PHY0_SM;
9509
9510        if (!this_client)
9511                return E1000_ERR_I2C;
9512
9513        if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9514                return E1000_ERR_SWFW_SYNC;
9515        status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
9516        hw->mac.ops.release_swfw_sync(hw, swfw_mask);
9517
9518        if (status)
9519                return E1000_ERR_I2C;
9520        else
9521                return 0;
9522
9523}
9524
9525int igb_reinit_queues(struct igb_adapter *adapter)
9526{
9527        struct net_device *netdev = adapter->netdev;
9528        struct pci_dev *pdev = adapter->pdev;
9529        int err = 0;
9530
9531        if (netif_running(netdev))
9532                igb_close(netdev);
9533
9534        igb_reset_interrupt_capability(adapter);
9535
9536        if (igb_init_interrupt_scheme(adapter, true)) {
9537                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9538                return -ENOMEM;
9539        }
9540
9541        if (netif_running(netdev))
9542                err = igb_open(netdev);
9543
9544        return err;
9545}
9546
9547static void igb_nfc_filter_exit(struct igb_adapter *adapter)
9548{
9549        struct igb_nfc_filter *rule;
9550
9551        spin_lock(&adapter->nfc_lock);
9552
9553        hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9554                igb_erase_filter(adapter, rule);
9555
9556        hlist_for_each_entry(rule, &adapter->cls_flower_list, nfc_node)
9557                igb_erase_filter(adapter, rule);
9558
9559        spin_unlock(&adapter->nfc_lock);
9560}
9561
9562static void igb_nfc_filter_restore(struct igb_adapter *adapter)
9563{
9564        struct igb_nfc_filter *rule;
9565
9566        spin_lock(&adapter->nfc_lock);
9567
9568        hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
9569                igb_add_filter(adapter, rule);
9570
9571        spin_unlock(&adapter->nfc_lock);
9572}
9573/* igb_main.c */
9574