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/bpf.h>
  34#include <linux/bpf_trace.h>
  35#include <linux/pm_runtime.h>
  36#include <linux/etherdevice.h>
  37#ifdef CONFIG_IGB_DCA
  38#include <linux/dca.h>
  39#endif
  40#include <linux/i2c.h>
  41#include "igb.h"
  42
  43enum queue_mode {
  44        QUEUE_MODE_STRICT_PRIORITY,
  45        QUEUE_MODE_STREAM_RESERVATION,
  46};
  47
  48enum tx_queue_prio {
  49        TX_QUEUE_PRIO_HIGH,
  50        TX_QUEUE_PRIO_LOW,
  51};
  52
  53char igb_driver_name[] = "igb";
  54static const char igb_driver_string[] =
  55                                "Intel(R) Gigabit Ethernet Network Driver";
  56static const char igb_copyright[] =
  57                                "Copyright (c) 2007-2014 Intel Corporation.";
  58
  59static const struct e1000_info *igb_info_tbl[] = {
  60        [board_82575] = &e1000_82575_info,
  61};
  62
  63static const struct pci_device_id igb_pci_tbl[] = {
  64        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_1GBPS) },
  65        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_SGMII) },
  66        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I354_BACKPLANE_2_5GBPS) },
  67        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I211_COPPER), board_82575 },
  68        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER), board_82575 },
  69        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_FIBER), board_82575 },
  70        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES), board_82575 },
  71        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SGMII), board_82575 },
  72        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_COPPER_FLASHLESS), board_82575 },
  73        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I210_SERDES_FLASHLESS), board_82575 },
  74        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_COPPER), board_82575 },
  75        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_FIBER), board_82575 },
  76        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SERDES), board_82575 },
  77        { PCI_VDEVICE(INTEL, E1000_DEV_ID_I350_SGMII), board_82575 },
  78        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER), board_82575 },
  79        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_FIBER), board_82575 },
  80        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_QUAD_FIBER), board_82575 },
  81        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SERDES), board_82575 },
  82        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_SGMII), board_82575 },
  83        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82580_COPPER_DUAL), board_82575 },
  84        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SGMII), board_82575 },
  85        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SERDES), board_82575 },
  86        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_BACKPLANE), board_82575 },
  87        { PCI_VDEVICE(INTEL, E1000_DEV_ID_DH89XXCC_SFP), board_82575 },
  88        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576), board_82575 },
  89        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS), board_82575 },
  90        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_NS_SERDES), board_82575 },
  91        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_FIBER), board_82575 },
  92        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES), board_82575 },
  93        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_SERDES_QUAD), board_82575 },
  94        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER_ET2), board_82575 },
  95        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82576_QUAD_COPPER), board_82575 },
  96        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_COPPER), board_82575 },
  97        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575EB_FIBER_SERDES), board_82575 },
  98        { PCI_VDEVICE(INTEL, E1000_DEV_ID_82575GB_QUAD_COPPER), board_82575 },
  99        /* required last entry */
 100        {0, }
 101};
 102
 103MODULE_DEVICE_TABLE(pci, igb_pci_tbl);
 104
 105static int igb_setup_all_tx_resources(struct igb_adapter *);
 106static int igb_setup_all_rx_resources(struct igb_adapter *);
 107static void igb_free_all_tx_resources(struct igb_adapter *);
 108static void igb_free_all_rx_resources(struct igb_adapter *);
 109static void igb_setup_mrqc(struct igb_adapter *);
 110static int igb_probe(struct pci_dev *, const struct pci_device_id *);
 111static void igb_remove(struct pci_dev *pdev);
 112static int igb_sw_init(struct igb_adapter *);
 113int igb_open(struct net_device *);
 114int igb_close(struct net_device *);
 115static void igb_configure(struct igb_adapter *);
 116static void igb_configure_tx(struct igb_adapter *);
 117static void igb_configure_rx(struct igb_adapter *);
 118static void igb_clean_all_tx_rings(struct igb_adapter *);
 119static void igb_clean_all_rx_rings(struct igb_adapter *);
 120static void igb_clean_tx_ring(struct igb_ring *);
 121static void igb_clean_rx_ring(struct igb_ring *);
 122static void igb_set_rx_mode(struct net_device *);
 123static void igb_update_phy_info(struct timer_list *);
 124static void igb_watchdog(struct timer_list *);
 125static void igb_watchdog_task(struct work_struct *);
 126static netdev_tx_t igb_xmit_frame(struct sk_buff *skb, struct net_device *);
 127static void igb_get_stats64(struct net_device *dev,
 128                            struct rtnl_link_stats64 *stats);
 129static int igb_change_mtu(struct net_device *, int);
 130static int igb_set_mac(struct net_device *, void *);
 131static void igb_set_uta(struct igb_adapter *adapter, bool set);
 132static irqreturn_t igb_intr(int irq, void *);
 133static irqreturn_t igb_intr_msi(int irq, void *);
 134static irqreturn_t igb_msix_other(int irq, void *);
 135static irqreturn_t igb_msix_ring(int irq, void *);
 136#ifdef CONFIG_IGB_DCA
 137static void igb_update_dca(struct igb_q_vector *);
 138static void igb_setup_dca(struct igb_adapter *);
 139#endif /* CONFIG_IGB_DCA */
 140static int igb_poll(struct napi_struct *, int);
 141static bool igb_clean_tx_irq(struct igb_q_vector *, int);
 142static int igb_clean_rx_irq(struct igb_q_vector *, int);
 143static int igb_ioctl(struct net_device *, struct ifreq *, int cmd);
 144static void igb_tx_timeout(struct net_device *, unsigned int txqueue);
 145static void igb_reset_task(struct work_struct *);
 146static void igb_vlan_mode(struct net_device *netdev,
 147                          netdev_features_t features);
 148static int igb_vlan_rx_add_vid(struct net_device *, __be16, u16);
 149static int igb_vlan_rx_kill_vid(struct net_device *, __be16, u16);
 150static void igb_restore_vlan(struct igb_adapter *);
 151static void igb_rar_set_index(struct igb_adapter *, u32);
 152static void igb_ping_all_vfs(struct igb_adapter *);
 153static void igb_msg_task(struct igb_adapter *);
 154static void igb_vmm_control(struct igb_adapter *);
 155static int igb_set_vf_mac(struct igb_adapter *, int, unsigned char *);
 156static void igb_flush_mac_table(struct igb_adapter *);
 157static int igb_available_rars(struct igb_adapter *, u8);
 158static void igb_set_default_mac_filter(struct igb_adapter *);
 159static int igb_uc_sync(struct net_device *, const unsigned char *);
 160static int igb_uc_unsync(struct net_device *, const unsigned char *);
 161static void igb_restore_vf_multicasts(struct igb_adapter *adapter);
 162static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
 163static int igb_ndo_set_vf_vlan(struct net_device *netdev,
 164                               int vf, u16 vlan, u8 qos, __be16 vlan_proto);
 165static int igb_ndo_set_vf_bw(struct net_device *, int, int, int);
 166static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
 167                                   bool setting);
 168static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf,
 169                                bool setting);
 170static int igb_ndo_get_vf_config(struct net_device *netdev, int vf,
 171                                 struct ifla_vf_info *ivi);
 172static void igb_check_vf_rate_limit(struct igb_adapter *);
 173static void igb_nfc_filter_exit(struct igb_adapter *adapter);
 174static void igb_nfc_filter_restore(struct igb_adapter *adapter);
 175
 176#ifdef CONFIG_PCI_IOV
 177static int igb_vf_configure(struct igb_adapter *adapter, int vf);
 178static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs);
 179static int igb_disable_sriov(struct pci_dev *dev);
 180static int igb_pci_disable_sriov(struct pci_dev *dev);
 181#endif
 182
 183static int igb_suspend(struct device *);
 184static int igb_resume(struct device *);
 185static int igb_runtime_suspend(struct device *dev);
 186static int igb_runtime_resume(struct device *dev);
 187static int igb_runtime_idle(struct device *dev);
 188static const struct dev_pm_ops igb_pm_ops = {
 189        SET_SYSTEM_SLEEP_PM_OPS(igb_suspend, igb_resume)
 190        SET_RUNTIME_PM_OPS(igb_runtime_suspend, igb_runtime_resume,
 191                        igb_runtime_idle)
 192};
 193static void igb_shutdown(struct pci_dev *);
 194static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs);
 195#ifdef CONFIG_IGB_DCA
 196static int igb_notify_dca(struct notifier_block *, unsigned long, void *);
 197static struct notifier_block dca_notifier = {
 198        .notifier_call  = igb_notify_dca,
 199        .next           = NULL,
 200        .priority       = 0
 201};
 202#endif
 203#ifdef CONFIG_PCI_IOV
 204static unsigned int max_vfs;
 205module_param(max_vfs, uint, 0);
 206MODULE_PARM_DESC(max_vfs, "Maximum number of virtual functions to allocate per physical function");
 207#endif /* CONFIG_PCI_IOV */
 208
 209static pci_ers_result_t igb_io_error_detected(struct pci_dev *,
 210                     pci_channel_state_t);
 211static pci_ers_result_t igb_io_slot_reset(struct pci_dev *);
 212static void igb_io_resume(struct pci_dev *);
 213
 214static const struct pci_error_handlers igb_err_handler = {
 215        .error_detected = igb_io_error_detected,
 216        .slot_reset = igb_io_slot_reset,
 217        .resume = igb_io_resume,
 218};
 219
 220static void igb_init_dmac(struct igb_adapter *adapter, u32 pba);
 221
 222static struct pci_driver igb_driver = {
 223        .name     = igb_driver_name,
 224        .id_table = igb_pci_tbl,
 225        .probe    = igb_probe,
 226        .remove   = igb_remove,
 227#ifdef CONFIG_PM
 228        .driver.pm = &igb_pm_ops,
 229#endif
 230        .shutdown = igb_shutdown,
 231        .sriov_configure = igb_pci_sriov_configure,
 232        .err_handler = &igb_err_handler
 233};
 234
 235MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 236MODULE_DESCRIPTION("Intel(R) Gigabit Ethernet Network Driver");
 237MODULE_LICENSE("GPL v2");
 238
 239#define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
 240static int debug = -1;
 241module_param(debug, int, 0);
 242MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 243
 244struct igb_reg_info {
 245        u32 ofs;
 246        char *name;
 247};
 248
 249static const struct igb_reg_info igb_reg_info_tbl[] = {
 250
 251        /* General Registers */
 252        {E1000_CTRL, "CTRL"},
 253        {E1000_STATUS, "STATUS"},
 254        {E1000_CTRL_EXT, "CTRL_EXT"},
 255
 256        /* Interrupt Registers */
 257        {E1000_ICR, "ICR"},
 258
 259        /* RX Registers */
 260        {E1000_RCTL, "RCTL"},
 261        {E1000_RDLEN(0), "RDLEN"},
 262        {E1000_RDH(0), "RDH"},
 263        {E1000_RDT(0), "RDT"},
 264        {E1000_RXDCTL(0), "RXDCTL"},
 265        {E1000_RDBAL(0), "RDBAL"},
 266        {E1000_RDBAH(0), "RDBAH"},
 267
 268        /* TX Registers */
 269        {E1000_TCTL, "TCTL"},
 270        {E1000_TDBAL(0), "TDBAL"},
 271        {E1000_TDBAH(0), "TDBAH"},
 272        {E1000_TDLEN(0), "TDLEN"},
 273        {E1000_TDH(0), "TDH"},
 274        {E1000_TDT(0), "TDT"},
 275        {E1000_TXDCTL(0), "TXDCTL"},
 276        {E1000_TDFH, "TDFH"},
 277        {E1000_TDFT, "TDFT"},
 278        {E1000_TDFHS, "TDFHS"},
 279        {E1000_TDFPC, "TDFPC"},
 280
 281        /* List Terminator */
 282        {}
 283};
 284
 285/* igb_regdump - register printout routine */
 286static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
 287{
 288        int n = 0;
 289        char rname[16];
 290        u32 regs[8];
 291
 292        switch (reginfo->ofs) {
 293        case E1000_RDLEN(0):
 294                for (n = 0; n < 4; n++)
 295                        regs[n] = rd32(E1000_RDLEN(n));
 296                break;
 297        case E1000_RDH(0):
 298                for (n = 0; n < 4; n++)
 299                        regs[n] = rd32(E1000_RDH(n));
 300                break;
 301        case E1000_RDT(0):
 302                for (n = 0; n < 4; n++)
 303                        regs[n] = rd32(E1000_RDT(n));
 304                break;
 305        case E1000_RXDCTL(0):
 306                for (n = 0; n < 4; n++)
 307                        regs[n] = rd32(E1000_RXDCTL(n));
 308                break;
 309        case E1000_RDBAL(0):
 310                for (n = 0; n < 4; n++)
 311                        regs[n] = rd32(E1000_RDBAL(n));
 312                break;
 313        case E1000_RDBAH(0):
 314                for (n = 0; n < 4; n++)
 315                        regs[n] = rd32(E1000_RDBAH(n));
 316                break;
 317        case E1000_TDBAL(0):
 318                for (n = 0; n < 4; n++)
 319                        regs[n] = rd32(E1000_TDBAL(n));
 320                break;
 321        case E1000_TDBAH(0):
 322                for (n = 0; n < 4; n++)
 323                        regs[n] = rd32(E1000_TDBAH(n));
 324                break;
 325        case E1000_TDLEN(0):
 326                for (n = 0; n < 4; n++)
 327                        regs[n] = rd32(E1000_TDLEN(n));
 328                break;
 329        case E1000_TDH(0):
 330                for (n = 0; n < 4; n++)
 331                        regs[n] = rd32(E1000_TDH(n));
 332                break;
 333        case E1000_TDT(0):
 334                for (n = 0; n < 4; n++)
 335                        regs[n] = rd32(E1000_TDT(n));
 336                break;
 337        case E1000_TXDCTL(0):
 338                for (n = 0; n < 4; n++)
 339                        regs[n] = rd32(E1000_TXDCTL(n));
 340                break;
 341        default:
 342                pr_info("%-15s %08x\n", reginfo->name, rd32(reginfo->ofs));
 343                return;
 344        }
 345
 346        snprintf(rname, 16, "%s%s", reginfo->name, "[0-3]");
 347        pr_info("%-15s %08x %08x %08x %08x\n", rname, regs[0], regs[1],
 348                regs[2], regs[3]);
 349}
 350
 351/* igb_dump - Print registers, Tx-rings and Rx-rings */
 352static void igb_dump(struct igb_adapter *adapter)
 353{
 354        struct net_device *netdev = adapter->netdev;
 355        struct e1000_hw *hw = &adapter->hw;
 356        struct igb_reg_info *reginfo;
 357        struct igb_ring *tx_ring;
 358        union e1000_adv_tx_desc *tx_desc;
 359        struct my_u0 { __le64 a; __le64 b; } *u0;
 360        struct igb_ring *rx_ring;
 361        union e1000_adv_rx_desc *rx_desc;
 362        u32 staterr;
 363        u16 i, n;
 364
 365        if (!netif_msg_hw(adapter))
 366                return;
 367
 368        /* Print netdevice Info */
 369        if (netdev) {
 370                dev_info(&adapter->pdev->dev, "Net device Info\n");
 371                pr_info("Device Name     state            trans_start\n");
 372                pr_info("%-15s %016lX %016lX\n", netdev->name,
 373                        netdev->state, dev_trans_start(netdev));
 374        }
 375
 376        /* Print Registers */
 377        dev_info(&adapter->pdev->dev, "Register Dump\n");
 378        pr_info(" Register Name   Value\n");
 379        for (reginfo = (struct igb_reg_info *)igb_reg_info_tbl;
 380             reginfo->name; reginfo++) {
 381                igb_regdump(hw, reginfo);
 382        }
 383
 384        /* Print TX Ring Summary */
 385        if (!netdev || !netif_running(netdev))
 386                goto exit;
 387
 388        dev_info(&adapter->pdev->dev, "TX Rings Summary\n");
 389        pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
 390        for (n = 0; n < adapter->num_tx_queues; n++) {
 391                struct igb_tx_buffer *buffer_info;
 392                tx_ring = adapter->tx_ring[n];
 393                buffer_info = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
 394                pr_info(" %5d %5X %5X %016llX %04X %p %016llX\n",
 395                        n, tx_ring->next_to_use, tx_ring->next_to_clean,
 396                        (u64)dma_unmap_addr(buffer_info, dma),
 397                        dma_unmap_len(buffer_info, len),
 398                        buffer_info->next_to_watch,
 399                        (u64)buffer_info->time_stamp);
 400        }
 401
 402        /* Print TX Rings */
 403        if (!netif_msg_tx_done(adapter))
 404                goto rx_ring_summary;
 405
 406        dev_info(&adapter->pdev->dev, "TX Rings Dump\n");
 407
 408        /* Transmit Descriptor Formats
 409         *
 410         * Advanced Transmit Descriptor
 411         *   +--------------------------------------------------------------+
 412         * 0 |         Buffer Address [63:0]                                |
 413         *   +--------------------------------------------------------------+
 414         * 8 | PAYLEN  | PORTS  |CC|IDX | STA | DCMD  |DTYP|MAC|RSV| DTALEN |
 415         *   +--------------------------------------------------------------+
 416         *   63      46 45    40 39 38 36 35 32 31   24             15       0
 417         */
 418
 419        for (n = 0; n < adapter->num_tx_queues; n++) {
 420                tx_ring = adapter->tx_ring[n];
 421                pr_info("------------------------------------\n");
 422                pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
 423                pr_info("------------------------------------\n");
 424                pr_info("T [desc]     [address 63:0  ] [PlPOCIStDDM Ln] [bi->dma       ] leng  ntw timestamp        bi->skb\n");
 425
 426                for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
 427                        const char *next_desc;
 428                        struct igb_tx_buffer *buffer_info;
 429                        tx_desc = IGB_TX_DESC(tx_ring, i);
 430                        buffer_info = &tx_ring->tx_buffer_info[i];
 431                        u0 = (struct my_u0 *)tx_desc;
 432                        if (i == tx_ring->next_to_use &&
 433                            i == tx_ring->next_to_clean)
 434                                next_desc = " NTC/U";
 435                        else if (i == tx_ring->next_to_use)
 436                                next_desc = " NTU";
 437                        else if (i == tx_ring->next_to_clean)
 438                                next_desc = " NTC";
 439                        else
 440                                next_desc = "";
 441
 442                        pr_info("T [0x%03X]    %016llX %016llX %016llX %04X  %p %016llX %p%s\n",
 443                                i, le64_to_cpu(u0->a),
 444                                le64_to_cpu(u0->b),
 445                                (u64)dma_unmap_addr(buffer_info, dma),
 446                                dma_unmap_len(buffer_info, len),
 447                                buffer_info->next_to_watch,
 448                                (u64)buffer_info->time_stamp,
 449                                buffer_info->skb, next_desc);
 450
 451                        if (netif_msg_pktdata(adapter) && buffer_info->skb)
 452                                print_hex_dump(KERN_INFO, "",
 453                                        DUMP_PREFIX_ADDRESS,
 454                                        16, 1, buffer_info->skb->data,
 455                                        dma_unmap_len(buffer_info, len),
 456                                        true);
 457                }
 458        }
 459
 460        /* Print RX Rings Summary */
 461rx_ring_summary:
 462        dev_info(&adapter->pdev->dev, "RX Rings Summary\n");
 463        pr_info("Queue [NTU] [NTC]\n");
 464        for (n = 0; n < adapter->num_rx_queues; n++) {
 465                rx_ring = adapter->rx_ring[n];
 466                pr_info(" %5d %5X %5X\n",
 467                        n, rx_ring->next_to_use, rx_ring->next_to_clean);
 468        }
 469
 470        /* Print RX Rings */
 471        if (!netif_msg_rx_status(adapter))
 472                goto exit;
 473
 474        dev_info(&adapter->pdev->dev, "RX Rings Dump\n");
 475
 476        /* Advanced Receive Descriptor (Read) Format
 477         *    63                                           1        0
 478         *    +-----------------------------------------------------+
 479         *  0 |       Packet Buffer Address [63:1]           |A0/NSE|
 480         *    +----------------------------------------------+------+
 481         *  8 |       Header Buffer Address [63:1]           |  DD  |
 482         *    +-----------------------------------------------------+
 483         *
 484         *
 485         * Advanced Receive Descriptor (Write-Back) Format
 486         *
 487         *   63       48 47    32 31  30      21 20 17 16   4 3     0
 488         *   +------------------------------------------------------+
 489         * 0 | Packet     IP     |SPH| HDR_LEN   | RSV|Packet|  RSS |
 490         *   | Checksum   Ident  |   |           |    | Type | Type |
 491         *   +------------------------------------------------------+
 492         * 8 | VLAN Tag | Length | Extended Error | Extended Status |
 493         *   +------------------------------------------------------+
 494         *   63       48 47    32 31            20 19               0
 495         */
 496
 497        for (n = 0; n < adapter->num_rx_queues; n++) {
 498                rx_ring = adapter->rx_ring[n];
 499                pr_info("------------------------------------\n");
 500                pr_info("RX QUEUE INDEX = %d\n", rx_ring->queue_index);
 501                pr_info("------------------------------------\n");
 502                pr_info("R  [desc]      [ PktBuf     A0] [  HeadBuf   DD] [bi->dma       ] [bi->skb] <-- Adv Rx Read format\n");
 503                pr_info("RWB[desc]      [PcsmIpSHl PtRs] [vl er S cks ln] ---------------- [bi->skb] <-- Adv Rx Write-Back format\n");
 504
 505                for (i = 0; i < rx_ring->count; i++) {
 506                        const char *next_desc;
 507                        struct igb_rx_buffer *buffer_info;
 508                        buffer_info = &rx_ring->rx_buffer_info[i];
 509                        rx_desc = IGB_RX_DESC(rx_ring, i);
 510                        u0 = (struct my_u0 *)rx_desc;
 511                        staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 512
 513                        if (i == rx_ring->next_to_use)
 514                                next_desc = " NTU";
 515                        else if (i == rx_ring->next_to_clean)
 516                                next_desc = " NTC";
 517                        else
 518                                next_desc = "";
 519
 520                        if (staterr & E1000_RXD_STAT_DD) {
 521                                /* Descriptor Done */
 522                                pr_info("%s[0x%03X]     %016llX %016llX ---------------- %s\n",
 523                                        "RWB", i,
 524                                        le64_to_cpu(u0->a),
 525                                        le64_to_cpu(u0->b),
 526                                        next_desc);
 527                        } else {
 528                                pr_info("%s[0x%03X]     %016llX %016llX %016llX %s\n",
 529                                        "R  ", i,
 530                                        le64_to_cpu(u0->a),
 531                                        le64_to_cpu(u0->b),
 532                                        (u64)buffer_info->dma,
 533                                        next_desc);
 534
 535                                if (netif_msg_pktdata(adapter) &&
 536                                    buffer_info->dma && buffer_info->page) {
 537                                        print_hex_dump(KERN_INFO, "",
 538                                          DUMP_PREFIX_ADDRESS,
 539                                          16, 1,
 540                                          page_address(buffer_info->page) +
 541                                                      buffer_info->page_offset,
 542                                          igb_rx_bufsz(rx_ring), true);
 543                                }
 544                        }
 545                }
 546        }
 547
 548exit:
 549        return;
 550}
 551
 552/**
 553 *  igb_get_i2c_data - Reads the I2C SDA data bit
 554 *  @data: opaque pointer to adapter struct
 555 *
 556 *  Returns the I2C data bit value
 557 **/
 558static int igb_get_i2c_data(void *data)
 559{
 560        struct igb_adapter *adapter = (struct igb_adapter *)data;
 561        struct e1000_hw *hw = &adapter->hw;
 562        s32 i2cctl = rd32(E1000_I2CPARAMS);
 563
 564        return !!(i2cctl & E1000_I2C_DATA_IN);
 565}
 566
 567/**
 568 *  igb_set_i2c_data - Sets the I2C data bit
 569 *  @data: pointer to hardware structure
 570 *  @state: I2C data value (0 or 1) to set
 571 *
 572 *  Sets the I2C data bit
 573 **/
 574static void igb_set_i2c_data(void *data, int state)
 575{
 576        struct igb_adapter *adapter = (struct igb_adapter *)data;
 577        struct e1000_hw *hw = &adapter->hw;
 578        s32 i2cctl = rd32(E1000_I2CPARAMS);
 579
 580        if (state) {
 581                i2cctl |= E1000_I2C_DATA_OUT | E1000_I2C_DATA_OE_N;
 582        } else {
 583                i2cctl &= ~E1000_I2C_DATA_OE_N;
 584                i2cctl &= ~E1000_I2C_DATA_OUT;
 585        }
 586
 587        wr32(E1000_I2CPARAMS, i2cctl);
 588        wrfl();
 589}
 590
 591/**
 592 *  igb_set_i2c_clk - Sets the I2C SCL clock
 593 *  @data: pointer to hardware structure
 594 *  @state: state to set clock
 595 *
 596 *  Sets the I2C clock line to state
 597 **/
 598static void igb_set_i2c_clk(void *data, int state)
 599{
 600        struct igb_adapter *adapter = (struct igb_adapter *)data;
 601        struct e1000_hw *hw = &adapter->hw;
 602        s32 i2cctl = rd32(E1000_I2CPARAMS);
 603
 604        if (state) {
 605                i2cctl |= E1000_I2C_CLK_OUT | E1000_I2C_CLK_OE_N;
 606        } else {
 607                i2cctl &= ~E1000_I2C_CLK_OUT;
 608                i2cctl &= ~E1000_I2C_CLK_OE_N;
 609        }
 610        wr32(E1000_I2CPARAMS, i2cctl);
 611        wrfl();
 612}
 613
 614/**
 615 *  igb_get_i2c_clk - Gets the I2C SCL clock state
 616 *  @data: pointer to hardware structure
 617 *
 618 *  Gets the I2C clock state
 619 **/
 620static int igb_get_i2c_clk(void *data)
 621{
 622        struct igb_adapter *adapter = (struct igb_adapter *)data;
 623        struct e1000_hw *hw = &adapter->hw;
 624        s32 i2cctl = rd32(E1000_I2CPARAMS);
 625
 626        return !!(i2cctl & E1000_I2C_CLK_IN);
 627}
 628
 629static const struct i2c_algo_bit_data igb_i2c_algo = {
 630        .setsda         = igb_set_i2c_data,
 631        .setscl         = igb_set_i2c_clk,
 632        .getsda         = igb_get_i2c_data,
 633        .getscl         = igb_get_i2c_clk,
 634        .udelay         = 5,
 635        .timeout        = 20,
 636};
 637
 638/**
 639 *  igb_get_hw_dev - return device
 640 *  @hw: pointer to hardware structure
 641 *
 642 *  used by hardware layer to print debugging information
 643 **/
 644struct net_device *igb_get_hw_dev(struct e1000_hw *hw)
 645{
 646        struct igb_adapter *adapter = hw->back;
 647        return adapter->netdev;
 648}
 649
 650/**
 651 *  igb_init_module - Driver Registration Routine
 652 *
 653 *  igb_init_module is the first routine called when the driver is
 654 *  loaded. All it does is register with the PCI subsystem.
 655 **/
 656static int __init igb_init_module(void)
 657{
 658        int ret;
 659
 660        pr_info("%s\n", igb_driver_string);
 661        pr_info("%s\n", igb_copyright);
 662
 663#ifdef CONFIG_IGB_DCA
 664        dca_register_notify(&dca_notifier);
 665#endif
 666        ret = pci_register_driver(&igb_driver);
 667        return ret;
 668}
 669
 670module_init(igb_init_module);
 671
 672/**
 673 *  igb_exit_module - Driver Exit Cleanup Routine
 674 *
 675 *  igb_exit_module is called just before the driver is removed
 676 *  from memory.
 677 **/
 678static void __exit igb_exit_module(void)
 679{
 680#ifdef CONFIG_IGB_DCA
 681        dca_unregister_notify(&dca_notifier);
 682#endif
 683        pci_unregister_driver(&igb_driver);
 684}
 685
 686module_exit(igb_exit_module);
 687
 688#define Q_IDX_82576(i) (((i & 0x1) << 3) + (i >> 1))
 689/**
 690 *  igb_cache_ring_register - Descriptor ring to register mapping
 691 *  @adapter: board private structure to initialize
 692 *
 693 *  Once we know the feature-set enabled for the device, we'll cache
 694 *  the register offset the descriptor ring is assigned to.
 695 **/
 696static void igb_cache_ring_register(struct igb_adapter *adapter)
 697{
 698        int i = 0, j = 0;
 699        u32 rbase_offset = adapter->vfs_allocated_count;
 700
 701        switch (adapter->hw.mac.type) {
 702        case e1000_82576:
 703                /* The queues are allocated for virtualization such that VF 0
 704                 * is allocated queues 0 and 8, VF 1 queues 1 and 9, etc.
 705                 * In order to avoid collision we start at the first free queue
 706                 * and continue consuming queues in the same sequence
 707                 */
 708                if (adapter->vfs_allocated_count) {
 709                        for (; i < adapter->rss_queues; i++)
 710                                adapter->rx_ring[i]->reg_idx = rbase_offset +
 711                                                               Q_IDX_82576(i);
 712                }
 713                fallthrough;
 714        case e1000_82575:
 715        case e1000_82580:
 716        case e1000_i350:
 717        case e1000_i354:
 718        case e1000_i210:
 719        case e1000_i211:
 720        default:
 721                for (; i < adapter->num_rx_queues; i++)
 722                        adapter->rx_ring[i]->reg_idx = rbase_offset + i;
 723                for (; j < adapter->num_tx_queues; j++)
 724                        adapter->tx_ring[j]->reg_idx = rbase_offset + j;
 725                break;
 726        }
 727}
 728
 729u32 igb_rd32(struct e1000_hw *hw, u32 reg)
 730{
 731        struct igb_adapter *igb = container_of(hw, struct igb_adapter, hw);
 732        u8 __iomem *hw_addr = READ_ONCE(hw->hw_addr);
 733        u32 value = 0;
 734
 735        if (E1000_REMOVED(hw_addr))
 736                return ~value;
 737
 738        value = readl(&hw_addr[reg]);
 739
 740        /* reads should not return all F's */
 741        if (!(~value) && (!reg || !(~readl(hw_addr)))) {
 742                struct net_device *netdev = igb->netdev;
 743                hw->hw_addr = NULL;
 744                netdev_err(netdev, "PCIe link lost\n");
 745                WARN(pci_device_is_present(igb->pdev),
 746                     "igb: Failed to read reg 0x%x!\n", reg);
 747        }
 748
 749        return value;
 750}
 751
 752/**
 753 *  igb_write_ivar - configure ivar for given MSI-X vector
 754 *  @hw: pointer to the HW structure
 755 *  @msix_vector: vector number we are allocating to a given ring
 756 *  @index: row index of IVAR register to write within IVAR table
 757 *  @offset: column offset of in IVAR, should be multiple of 8
 758 *
 759 *  This function is intended to handle the writing of the IVAR register
 760 *  for adapters 82576 and newer.  The IVAR table consists of 2 columns,
 761 *  each containing an cause allocation for an Rx and Tx ring, and a
 762 *  variable number of rows depending on the number of queues supported.
 763 **/
 764static void igb_write_ivar(struct e1000_hw *hw, int msix_vector,
 765                           int index, int offset)
 766{
 767        u32 ivar = array_rd32(E1000_IVAR0, index);
 768
 769        /* clear any bits that are currently set */
 770        ivar &= ~((u32)0xFF << offset);
 771
 772        /* write vector and valid bit */
 773        ivar |= (msix_vector | E1000_IVAR_VALID) << offset;
 774
 775        array_wr32(E1000_IVAR0, index, ivar);
 776}
 777
 778#define IGB_N0_QUEUE -1
 779static void igb_assign_vector(struct igb_q_vector *q_vector, int msix_vector)
 780{
 781        struct igb_adapter *adapter = q_vector->adapter;
 782        struct e1000_hw *hw = &adapter->hw;
 783        int rx_queue = IGB_N0_QUEUE;
 784        int tx_queue = IGB_N0_QUEUE;
 785        u32 msixbm = 0;
 786
 787        if (q_vector->rx.ring)
 788                rx_queue = q_vector->rx.ring->reg_idx;
 789        if (q_vector->tx.ring)
 790                tx_queue = q_vector->tx.ring->reg_idx;
 791
 792        switch (hw->mac.type) {
 793        case e1000_82575:
 794                /* The 82575 assigns vectors using a bitmask, which matches the
 795                 * bitmask for the EICR/EIMS/EIMC registers.  To assign one
 796                 * or more queues to a vector, we write the appropriate bits
 797                 * into the MSIXBM register for that vector.
 798                 */
 799                if (rx_queue > IGB_N0_QUEUE)
 800                        msixbm = E1000_EICR_RX_QUEUE0 << rx_queue;
 801                if (tx_queue > IGB_N0_QUEUE)
 802                        msixbm |= E1000_EICR_TX_QUEUE0 << tx_queue;
 803                if (!(adapter->flags & IGB_FLAG_HAS_MSIX) && msix_vector == 0)
 804                        msixbm |= E1000_EIMS_OTHER;
 805                array_wr32(E1000_MSIXBM(0), msix_vector, msixbm);
 806                q_vector->eims_value = msixbm;
 807                break;
 808        case e1000_82576:
 809                /* 82576 uses a table that essentially consists of 2 columns
 810                 * with 8 rows.  The ordering is column-major so we use the
 811                 * lower 3 bits as the row index, and the 4th bit as the
 812                 * column offset.
 813                 */
 814                if (rx_queue > IGB_N0_QUEUE)
 815                        igb_write_ivar(hw, msix_vector,
 816                                       rx_queue & 0x7,
 817                                       (rx_queue & 0x8) << 1);
 818                if (tx_queue > IGB_N0_QUEUE)
 819                        igb_write_ivar(hw, msix_vector,
 820                                       tx_queue & 0x7,
 821                                       ((tx_queue & 0x8) << 1) + 8);
 822                q_vector->eims_value = BIT(msix_vector);
 823                break;
 824        case e1000_82580:
 825        case e1000_i350:
 826        case e1000_i354:
 827        case e1000_i210:
 828        case e1000_i211:
 829                /* On 82580 and newer adapters the scheme is similar to 82576
 830                 * however instead of ordering column-major we have things
 831                 * ordered row-major.  So we traverse the table by using
 832                 * bit 0 as the column offset, and the remaining bits as the
 833                 * row index.
 834                 */
 835                if (rx_queue > IGB_N0_QUEUE)
 836                        igb_write_ivar(hw, msix_vector,
 837                                       rx_queue >> 1,
 838                                       (rx_queue & 0x1) << 4);
 839                if (tx_queue > IGB_N0_QUEUE)
 840                        igb_write_ivar(hw, msix_vector,
 841                                       tx_queue >> 1,
 842                                       ((tx_queue & 0x1) << 4) + 8);
 843                q_vector->eims_value = BIT(msix_vector);
 844                break;
 845        default:
 846                BUG();
 847                break;
 848        }
 849
 850        /* add q_vector eims value to global eims_enable_mask */
 851        adapter->eims_enable_mask |= q_vector->eims_value;
 852
 853        /* configure q_vector to set itr on first interrupt */
 854        q_vector->set_itr = 1;
 855}
 856
 857/**
 858 *  igb_configure_msix - Configure MSI-X hardware
 859 *  @adapter: board private structure to initialize
 860 *
 861 *  igb_configure_msix sets up the hardware to properly
 862 *  generate MSI-X interrupts.
 863 **/
 864static void igb_configure_msix(struct igb_adapter *adapter)
 865{
 866        u32 tmp;
 867        int i, vector = 0;
 868        struct e1000_hw *hw = &adapter->hw;
 869
 870        adapter->eims_enable_mask = 0;
 871
 872        /* set vector for other causes, i.e. link changes */
 873        switch (hw->mac.type) {
 874        case e1000_82575:
 875                tmp = rd32(E1000_CTRL_EXT);
 876                /* enable MSI-X PBA support*/
 877                tmp |= E1000_CTRL_EXT_PBA_CLR;
 878
 879                /* Auto-Mask interrupts upon ICR read. */
 880                tmp |= E1000_CTRL_EXT_EIAME;
 881                tmp |= E1000_CTRL_EXT_IRCA;
 882
 883                wr32(E1000_CTRL_EXT, tmp);
 884
 885                /* enable msix_other interrupt */
 886                array_wr32(E1000_MSIXBM(0), vector++, E1000_EIMS_OTHER);
 887                adapter->eims_other = E1000_EIMS_OTHER;
 888
 889                break;
 890
 891        case e1000_82576:
 892        case e1000_82580:
 893        case e1000_i350:
 894        case e1000_i354:
 895        case e1000_i210:
 896        case e1000_i211:
 897                /* Turn on MSI-X capability first, or our settings
 898                 * won't stick.  And it will take days to debug.
 899                 */
 900                wr32(E1000_GPIE, E1000_GPIE_MSIX_MODE |
 901                     E1000_GPIE_PBA | E1000_GPIE_EIAME |
 902                     E1000_GPIE_NSICR);
 903
 904                /* enable msix_other interrupt */
 905                adapter->eims_other = BIT(vector);
 906                tmp = (vector++ | E1000_IVAR_VALID) << 8;
 907
 908                wr32(E1000_IVAR_MISC, tmp);
 909                break;
 910        default:
 911                /* do nothing, since nothing else supports MSI-X */
 912                break;
 913        } /* switch (hw->mac.type) */
 914
 915        adapter->eims_enable_mask |= adapter->eims_other;
 916
 917        for (i = 0; i < adapter->num_q_vectors; i++)
 918                igb_assign_vector(adapter->q_vector[i], vector++);
 919
 920        wrfl();
 921}
 922
 923/**
 924 *  igb_request_msix - Initialize MSI-X interrupts
 925 *  @adapter: board private structure to initialize
 926 *
 927 *  igb_request_msix allocates MSI-X vectors and requests interrupts from the
 928 *  kernel.
 929 **/
 930static int igb_request_msix(struct igb_adapter *adapter)
 931{
 932        unsigned int num_q_vectors = adapter->num_q_vectors;
 933        struct net_device *netdev = adapter->netdev;
 934        int i, err = 0, vector = 0, free_vector = 0;
 935
 936        err = request_irq(adapter->msix_entries[vector].vector,
 937                          igb_msix_other, 0, netdev->name, adapter);
 938        if (err)
 939                goto err_out;
 940
 941        if (num_q_vectors > MAX_Q_VECTORS) {
 942                num_q_vectors = MAX_Q_VECTORS;
 943                dev_warn(&adapter->pdev->dev,
 944                         "The number of queue vectors (%d) is higher than max allowed (%d)\n",
 945                         adapter->num_q_vectors, MAX_Q_VECTORS);
 946        }
 947        for (i = 0; i < num_q_vectors; i++) {
 948                struct igb_q_vector *q_vector = adapter->q_vector[i];
 949
 950                vector++;
 951
 952                q_vector->itr_register = adapter->io_addr + E1000_EITR(vector);
 953
 954                if (q_vector->rx.ring && q_vector->tx.ring)
 955                        sprintf(q_vector->name, "%s-TxRx-%u", netdev->name,
 956                                q_vector->rx.ring->queue_index);
 957                else if (q_vector->tx.ring)
 958                        sprintf(q_vector->name, "%s-tx-%u", netdev->name,
 959                                q_vector->tx.ring->queue_index);
 960                else if (q_vector->rx.ring)
 961                        sprintf(q_vector->name, "%s-rx-%u", netdev->name,
 962                                q_vector->rx.ring->queue_index);
 963                else
 964                        sprintf(q_vector->name, "%s-unused", netdev->name);
 965
 966                err = request_irq(adapter->msix_entries[vector].vector,
 967                                  igb_msix_ring, 0, q_vector->name,
 968                                  q_vector);
 969                if (err)
 970                        goto err_free;
 971        }
 972
 973        igb_configure_msix(adapter);
 974        return 0;
 975
 976err_free:
 977        /* free already assigned IRQs */
 978        free_irq(adapter->msix_entries[free_vector++].vector, adapter);
 979
 980        vector--;
 981        for (i = 0; i < vector; i++) {
 982                free_irq(adapter->msix_entries[free_vector++].vector,
 983                         adapter->q_vector[i]);
 984        }
 985err_out:
 986        return err;
 987}
 988
 989/**
 990 *  igb_free_q_vector - Free memory allocated for specific interrupt vector
 991 *  @adapter: board private structure to initialize
 992 *  @v_idx: Index of vector to be freed
 993 *
 994 *  This function frees the memory allocated to the q_vector.
 995 **/
 996static void igb_free_q_vector(struct igb_adapter *adapter, int v_idx)
 997{
 998        struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
 999
1000        adapter->q_vector[v_idx] = NULL;
1001
1002        /* igb_get_stats64() might access the rings on this vector,
1003         * we must wait a grace period before freeing it.
1004         */
1005        if (q_vector)
1006                kfree_rcu(q_vector, rcu);
1007}
1008
1009/**
1010 *  igb_reset_q_vector - Reset config for interrupt vector
1011 *  @adapter: board private structure to initialize
1012 *  @v_idx: Index of vector to be reset
1013 *
1014 *  If NAPI is enabled it will delete any references to the
1015 *  NAPI struct. This is preparation for igb_free_q_vector.
1016 **/
1017static void igb_reset_q_vector(struct igb_adapter *adapter, int v_idx)
1018{
1019        struct igb_q_vector *q_vector = adapter->q_vector[v_idx];
1020
1021        /* Coming from igb_set_interrupt_capability, the vectors are not yet
1022         * allocated. So, q_vector is NULL so we should stop here.
1023         */
1024        if (!q_vector)
1025                return;
1026
1027        if (q_vector->tx.ring)
1028                adapter->tx_ring[q_vector->tx.ring->queue_index] = NULL;
1029
1030        if (q_vector->rx.ring)
1031                adapter->rx_ring[q_vector->rx.ring->queue_index] = NULL;
1032
1033        netif_napi_del(&q_vector->napi);
1034
1035}
1036
1037static void igb_reset_interrupt_capability(struct igb_adapter *adapter)
1038{
1039        int v_idx = adapter->num_q_vectors;
1040
1041        if (adapter->flags & IGB_FLAG_HAS_MSIX)
1042                pci_disable_msix(adapter->pdev);
1043        else if (adapter->flags & IGB_FLAG_HAS_MSI)
1044                pci_disable_msi(adapter->pdev);
1045
1046        while (v_idx--)
1047                igb_reset_q_vector(adapter, v_idx);
1048}
1049
1050/**
1051 *  igb_free_q_vectors - Free memory allocated for interrupt vectors
1052 *  @adapter: board private structure to initialize
1053 *
1054 *  This function frees the memory allocated to the q_vectors.  In addition if
1055 *  NAPI is enabled it will delete any references to the NAPI struct prior
1056 *  to freeing the q_vector.
1057 **/
1058static void igb_free_q_vectors(struct igb_adapter *adapter)
1059{
1060        int v_idx = adapter->num_q_vectors;
1061
1062        adapter->num_tx_queues = 0;
1063        adapter->num_rx_queues = 0;
1064        adapter->num_q_vectors = 0;
1065
1066        while (v_idx--) {
1067                igb_reset_q_vector(adapter, v_idx);
1068                igb_free_q_vector(adapter, v_idx);
1069        }
1070}
1071
1072/**
1073 *  igb_clear_interrupt_scheme - reset the device to a state of no interrupts
1074 *  @adapter: board private structure to initialize
1075 *
1076 *  This function resets the device so that it has 0 Rx queues, Tx queues, and
1077 *  MSI-X interrupts allocated.
1078 */
1079static void igb_clear_interrupt_scheme(struct igb_adapter *adapter)
1080{
1081        igb_free_q_vectors(adapter);
1082        igb_reset_interrupt_capability(adapter);
1083}
1084
1085/**
1086 *  igb_set_interrupt_capability - set MSI or MSI-X if supported
1087 *  @adapter: board private structure to initialize
1088 *  @msix: boolean value of MSIX capability
1089 *
1090 *  Attempt to configure interrupts using the best available
1091 *  capabilities of the hardware and kernel.
1092 **/
1093static void igb_set_interrupt_capability(struct igb_adapter *adapter, bool msix)
1094{
1095        int err;
1096        int numvecs, i;
1097
1098        if (!msix)
1099                goto msi_only;
1100        adapter->flags |= IGB_FLAG_HAS_MSIX;
1101
1102        /* Number of supported queues. */
1103        adapter->num_rx_queues = adapter->rss_queues;
1104        if (adapter->vfs_allocated_count)
1105                adapter->num_tx_queues = 1;
1106        else
1107                adapter->num_tx_queues = adapter->rss_queues;
1108
1109        /* start with one vector for every Rx queue */
1110        numvecs = adapter->num_rx_queues;
1111
1112        /* if Tx handler is separate add 1 for every Tx queue */
1113        if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
1114                numvecs += adapter->num_tx_queues;
1115
1116        /* store the number of vectors reserved for queues */
1117        adapter->num_q_vectors = numvecs;
1118
1119        /* add 1 vector for link status interrupts */
1120        numvecs++;
1121        for (i = 0; i < numvecs; i++)
1122                adapter->msix_entries[i].entry = i;
1123
1124        err = pci_enable_msix_range(adapter->pdev,
1125                                    adapter->msix_entries,
1126                                    numvecs,
1127                                    numvecs);
1128        if (err > 0)
1129                return;
1130
1131        igb_reset_interrupt_capability(adapter);
1132
1133        /* If we can't do MSI-X, try MSI */
1134msi_only:
1135        adapter->flags &= ~IGB_FLAG_HAS_MSIX;
1136#ifdef CONFIG_PCI_IOV
1137        /* disable SR-IOV for non MSI-X configurations */
1138        if (adapter->vf_data) {
1139                struct e1000_hw *hw = &adapter->hw;
1140                /* disable iov and allow time for transactions to clear */
1141                pci_disable_sriov(adapter->pdev);
1142                msleep(500);
1143
1144                kfree(adapter->vf_mac_list);
1145                adapter->vf_mac_list = NULL;
1146                kfree(adapter->vf_data);
1147                adapter->vf_data = NULL;
1148                wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
1149                wrfl();
1150                msleep(100);
1151                dev_info(&adapter->pdev->dev, "IOV Disabled\n");
1152        }
1153#endif
1154        adapter->vfs_allocated_count = 0;
1155        adapter->rss_queues = 1;
1156        adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
1157        adapter->num_rx_queues = 1;
1158        adapter->num_tx_queues = 1;
1159        adapter->num_q_vectors = 1;
1160        if (!pci_enable_msi(adapter->pdev))
1161                adapter->flags |= IGB_FLAG_HAS_MSI;
1162}
1163
1164static void igb_add_ring(struct igb_ring *ring,
1165                         struct igb_ring_container *head)
1166{
1167        head->ring = ring;
1168        head->count++;
1169}
1170
1171/**
1172 *  igb_alloc_q_vector - Allocate memory for a single interrupt vector
1173 *  @adapter: board private structure to initialize
1174 *  @v_count: q_vectors allocated on adapter, used for ring interleaving
1175 *  @v_idx: index of vector in adapter struct
1176 *  @txr_count: total number of Tx rings to allocate
1177 *  @txr_idx: index of first Tx ring to allocate
1178 *  @rxr_count: total number of Rx rings to allocate
1179 *  @rxr_idx: index of first Rx ring to allocate
1180 *
1181 *  We allocate one q_vector.  If allocation fails we return -ENOMEM.
1182 **/
1183static int igb_alloc_q_vector(struct igb_adapter *adapter,
1184                              int v_count, int v_idx,
1185                              int txr_count, int txr_idx,
1186                              int rxr_count, int rxr_idx)
1187{
1188        struct igb_q_vector *q_vector;
1189        struct igb_ring *ring;
1190        int ring_count;
1191        size_t size;
1192
1193        /* igb only supports 1 Tx and/or 1 Rx queue per vector */
1194        if (txr_count > 1 || rxr_count > 1)
1195                return -ENOMEM;
1196
1197        ring_count = txr_count + rxr_count;
1198        size = struct_size(q_vector, ring, ring_count);
1199
1200        /* allocate q_vector and rings */
1201        q_vector = adapter->q_vector[v_idx];
1202        if (!q_vector) {
1203                q_vector = kzalloc(size, GFP_KERNEL);
1204        } else if (size > ksize(q_vector)) {
1205                kfree_rcu(q_vector, rcu);
1206                q_vector = kzalloc(size, GFP_KERNEL);
1207        } else {
1208                memset(q_vector, 0, size);
1209        }
1210        if (!q_vector)
1211                return -ENOMEM;
1212
1213        /* initialize NAPI */
1214        netif_napi_add(adapter->netdev, &q_vector->napi,
1215                       igb_poll, 64);
1216
1217        /* tie q_vector and adapter together */
1218        adapter->q_vector[v_idx] = q_vector;
1219        q_vector->adapter = adapter;
1220
1221        /* initialize work limits */
1222        q_vector->tx.work_limit = adapter->tx_work_limit;
1223
1224        /* initialize ITR configuration */
1225        q_vector->itr_register = adapter->io_addr + E1000_EITR(0);
1226        q_vector->itr_val = IGB_START_ITR;
1227
1228        /* initialize pointer to rings */
1229        ring = q_vector->ring;
1230
1231        /* intialize ITR */
1232        if (rxr_count) {
1233                /* rx or rx/tx vector */
1234                if (!adapter->rx_itr_setting || adapter->rx_itr_setting > 3)
1235                        q_vector->itr_val = adapter->rx_itr_setting;
1236        } else {
1237                /* tx only vector */
1238                if (!adapter->tx_itr_setting || adapter->tx_itr_setting > 3)
1239                        q_vector->itr_val = adapter->tx_itr_setting;
1240        }
1241
1242        if (txr_count) {
1243                /* assign generic ring traits */
1244                ring->dev = &adapter->pdev->dev;
1245                ring->netdev = adapter->netdev;
1246
1247                /* configure backlink on ring */
1248                ring->q_vector = q_vector;
1249
1250                /* update q_vector Tx values */
1251                igb_add_ring(ring, &q_vector->tx);
1252
1253                /* For 82575, context index must be unique per ring. */
1254                if (adapter->hw.mac.type == e1000_82575)
1255                        set_bit(IGB_RING_FLAG_TX_CTX_IDX, &ring->flags);
1256
1257                /* apply Tx specific ring traits */
1258                ring->count = adapter->tx_ring_count;
1259                ring->queue_index = txr_idx;
1260
1261                ring->cbs_enable = false;
1262                ring->idleslope = 0;
1263                ring->sendslope = 0;
1264                ring->hicredit = 0;
1265                ring->locredit = 0;
1266
1267                u64_stats_init(&ring->tx_syncp);
1268                u64_stats_init(&ring->tx_syncp2);
1269
1270                /* assign ring to adapter */
1271                adapter->tx_ring[txr_idx] = ring;
1272
1273                /* push pointer to next ring */
1274                ring++;
1275        }
1276
1277        if (rxr_count) {
1278                /* assign generic ring traits */
1279                ring->dev = &adapter->pdev->dev;
1280                ring->netdev = adapter->netdev;
1281
1282                /* configure backlink on ring */
1283                ring->q_vector = q_vector;
1284
1285                /* update q_vector Rx values */
1286                igb_add_ring(ring, &q_vector->rx);
1287
1288                /* set flag indicating ring supports SCTP checksum offload */
1289                if (adapter->hw.mac.type >= e1000_82576)
1290                        set_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags);
1291
1292                /* On i350, i354, i210, and i211, loopback VLAN packets
1293                 * have the tag byte-swapped.
1294                 */
1295                if (adapter->hw.mac.type >= e1000_i350)
1296                        set_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &ring->flags);
1297
1298                /* apply Rx specific ring traits */
1299                ring->count = adapter->rx_ring_count;
1300                ring->queue_index = rxr_idx;
1301
1302                u64_stats_init(&ring->rx_syncp);
1303
1304                /* assign ring to adapter */
1305                adapter->rx_ring[rxr_idx] = ring;
1306        }
1307
1308        return 0;
1309}
1310
1311
1312/**
1313 *  igb_alloc_q_vectors - Allocate memory for interrupt vectors
1314 *  @adapter: board private structure to initialize
1315 *
1316 *  We allocate one q_vector per queue interrupt.  If allocation fails we
1317 *  return -ENOMEM.
1318 **/
1319static int igb_alloc_q_vectors(struct igb_adapter *adapter)
1320{
1321        int q_vectors = adapter->num_q_vectors;
1322        int rxr_remaining = adapter->num_rx_queues;
1323        int txr_remaining = adapter->num_tx_queues;
1324        int rxr_idx = 0, txr_idx = 0, v_idx = 0;
1325        int err;
1326
1327        if (q_vectors >= (rxr_remaining + txr_remaining)) {
1328                for (; rxr_remaining; v_idx++) {
1329                        err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1330                                                 0, 0, 1, rxr_idx);
1331
1332                        if (err)
1333                                goto err_out;
1334
1335                        /* update counts and index */
1336                        rxr_remaining--;
1337                        rxr_idx++;
1338                }
1339        }
1340
1341        for (; v_idx < q_vectors; v_idx++) {
1342                int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
1343                int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
1344
1345                err = igb_alloc_q_vector(adapter, q_vectors, v_idx,
1346                                         tqpv, txr_idx, rqpv, rxr_idx);
1347
1348                if (err)
1349                        goto err_out;
1350
1351                /* update counts and index */
1352                rxr_remaining -= rqpv;
1353                txr_remaining -= tqpv;
1354                rxr_idx++;
1355                txr_idx++;
1356        }
1357
1358        return 0;
1359
1360err_out:
1361        adapter->num_tx_queues = 0;
1362        adapter->num_rx_queues = 0;
1363        adapter->num_q_vectors = 0;
1364
1365        while (v_idx--)
1366                igb_free_q_vector(adapter, v_idx);
1367
1368        return -ENOMEM;
1369}
1370
1371/**
1372 *  igb_init_interrupt_scheme - initialize interrupts, allocate queues/vectors
1373 *  @adapter: board private structure to initialize
1374 *  @msix: boolean value of MSIX capability
1375 *
1376 *  This function initializes the interrupts and allocates all of the queues.
1377 **/
1378static int igb_init_interrupt_scheme(struct igb_adapter *adapter, bool msix)
1379{
1380        struct pci_dev *pdev = adapter->pdev;
1381        int err;
1382
1383        igb_set_interrupt_capability(adapter, msix);
1384
1385        err = igb_alloc_q_vectors(adapter);
1386        if (err) {
1387                dev_err(&pdev->dev, "Unable to allocate memory for vectors\n");
1388                goto err_alloc_q_vectors;
1389        }
1390
1391        igb_cache_ring_register(adapter);
1392
1393        return 0;
1394
1395err_alloc_q_vectors:
1396        igb_reset_interrupt_capability(adapter);
1397        return err;
1398}
1399
1400/**
1401 *  igb_request_irq - initialize interrupts
1402 *  @adapter: board private structure to initialize
1403 *
1404 *  Attempts to configure interrupts using the best available
1405 *  capabilities of the hardware and kernel.
1406 **/
1407static int igb_request_irq(struct igb_adapter *adapter)
1408{
1409        struct net_device *netdev = adapter->netdev;
1410        struct pci_dev *pdev = adapter->pdev;
1411        int err = 0;
1412
1413        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1414                err = igb_request_msix(adapter);
1415                if (!err)
1416                        goto request_done;
1417                /* fall back to MSI */
1418                igb_free_all_tx_resources(adapter);
1419                igb_free_all_rx_resources(adapter);
1420
1421                igb_clear_interrupt_scheme(adapter);
1422                err = igb_init_interrupt_scheme(adapter, false);
1423                if (err)
1424                        goto request_done;
1425
1426                igb_setup_all_tx_resources(adapter);
1427                igb_setup_all_rx_resources(adapter);
1428                igb_configure(adapter);
1429        }
1430
1431        igb_assign_vector(adapter->q_vector[0], 0);
1432
1433        if (adapter->flags & IGB_FLAG_HAS_MSI) {
1434                err = request_irq(pdev->irq, igb_intr_msi, 0,
1435                                  netdev->name, adapter);
1436                if (!err)
1437                        goto request_done;
1438
1439                /* fall back to legacy interrupts */
1440                igb_reset_interrupt_capability(adapter);
1441                adapter->flags &= ~IGB_FLAG_HAS_MSI;
1442        }
1443
1444        err = request_irq(pdev->irq, igb_intr, IRQF_SHARED,
1445                          netdev->name, adapter);
1446
1447        if (err)
1448                dev_err(&pdev->dev, "Error %d getting interrupt\n",
1449                        err);
1450
1451request_done:
1452        return err;
1453}
1454
1455static void igb_free_irq(struct igb_adapter *adapter)
1456{
1457        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1458                int vector = 0, i;
1459
1460                free_irq(adapter->msix_entries[vector++].vector, adapter);
1461
1462                for (i = 0; i < adapter->num_q_vectors; i++)
1463                        free_irq(adapter->msix_entries[vector++].vector,
1464                                 adapter->q_vector[i]);
1465        } else {
1466                free_irq(adapter->pdev->irq, adapter);
1467        }
1468}
1469
1470/**
1471 *  igb_irq_disable - Mask off interrupt generation on the NIC
1472 *  @adapter: board private structure
1473 **/
1474static void igb_irq_disable(struct igb_adapter *adapter)
1475{
1476        struct e1000_hw *hw = &adapter->hw;
1477
1478        /* we need to be careful when disabling interrupts.  The VFs are also
1479         * mapped into these registers and so clearing the bits can cause
1480         * issues on the VF drivers so we only need to clear what we set
1481         */
1482        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1483                u32 regval = rd32(E1000_EIAM);
1484
1485                wr32(E1000_EIAM, regval & ~adapter->eims_enable_mask);
1486                wr32(E1000_EIMC, adapter->eims_enable_mask);
1487                regval = rd32(E1000_EIAC);
1488                wr32(E1000_EIAC, regval & ~adapter->eims_enable_mask);
1489        }
1490
1491        wr32(E1000_IAM, 0);
1492        wr32(E1000_IMC, ~0);
1493        wrfl();
1494        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1495                int i;
1496
1497                for (i = 0; i < adapter->num_q_vectors; i++)
1498                        synchronize_irq(adapter->msix_entries[i].vector);
1499        } else {
1500                synchronize_irq(adapter->pdev->irq);
1501        }
1502}
1503
1504/**
1505 *  igb_irq_enable - Enable default interrupt generation settings
1506 *  @adapter: board private structure
1507 **/
1508static void igb_irq_enable(struct igb_adapter *adapter)
1509{
1510        struct e1000_hw *hw = &adapter->hw;
1511
1512        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
1513                u32 ims = E1000_IMS_LSC | E1000_IMS_DOUTSYNC | E1000_IMS_DRSTA;
1514                u32 regval = rd32(E1000_EIAC);
1515
1516                wr32(E1000_EIAC, regval | adapter->eims_enable_mask);
1517                regval = rd32(E1000_EIAM);
1518                wr32(E1000_EIAM, regval | adapter->eims_enable_mask);
1519                wr32(E1000_EIMS, adapter->eims_enable_mask);
1520                if (adapter->vfs_allocated_count) {
1521                        wr32(E1000_MBVFIMR, 0xFF);
1522                        ims |= E1000_IMS_VMMB;
1523                }
1524                wr32(E1000_IMS, ims);
1525        } else {
1526                wr32(E1000_IMS, IMS_ENABLE_MASK |
1527                                E1000_IMS_DRSTA);
1528                wr32(E1000_IAM, IMS_ENABLE_MASK |
1529                                E1000_IMS_DRSTA);
1530        }
1531}
1532
1533static void igb_update_mng_vlan(struct igb_adapter *adapter)
1534{
1535        struct e1000_hw *hw = &adapter->hw;
1536        u16 pf_id = adapter->vfs_allocated_count;
1537        u16 vid = adapter->hw.mng_cookie.vlan_id;
1538        u16 old_vid = adapter->mng_vlan_id;
1539
1540        if (hw->mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
1541                /* add VID to filter table */
1542                igb_vfta_set(hw, vid, pf_id, true, true);
1543                adapter->mng_vlan_id = vid;
1544        } else {
1545                adapter->mng_vlan_id = IGB_MNG_VLAN_NONE;
1546        }
1547
1548        if ((old_vid != (u16)IGB_MNG_VLAN_NONE) &&
1549            (vid != old_vid) &&
1550            !test_bit(old_vid, adapter->active_vlans)) {
1551                /* remove VID from filter table */
1552                igb_vfta_set(hw, vid, pf_id, false, true);
1553        }
1554}
1555
1556/**
1557 *  igb_release_hw_control - release control of the h/w to f/w
1558 *  @adapter: address of board private structure
1559 *
1560 *  igb_release_hw_control resets CTRL_EXT:DRV_LOAD bit.
1561 *  For ASF and Pass Through versions of f/w this means that the
1562 *  driver is no longer loaded.
1563 **/
1564static void igb_release_hw_control(struct igb_adapter *adapter)
1565{
1566        struct e1000_hw *hw = &adapter->hw;
1567        u32 ctrl_ext;
1568
1569        /* Let firmware take over control of h/w */
1570        ctrl_ext = rd32(E1000_CTRL_EXT);
1571        wr32(E1000_CTRL_EXT,
1572                        ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1573}
1574
1575/**
1576 *  igb_get_hw_control - get control of the h/w from f/w
1577 *  @adapter: address of board private structure
1578 *
1579 *  igb_get_hw_control sets CTRL_EXT:DRV_LOAD bit.
1580 *  For ASF and Pass Through versions of f/w this means that
1581 *  the driver is loaded.
1582 **/
1583static void igb_get_hw_control(struct igb_adapter *adapter)
1584{
1585        struct e1000_hw *hw = &adapter->hw;
1586        u32 ctrl_ext;
1587
1588        /* Let firmware know the driver has taken over */
1589        ctrl_ext = rd32(E1000_CTRL_EXT);
1590        wr32(E1000_CTRL_EXT,
1591                        ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1592}
1593
1594static void enable_fqtss(struct igb_adapter *adapter, bool enable)
1595{
1596        struct net_device *netdev = adapter->netdev;
1597        struct e1000_hw *hw = &adapter->hw;
1598
1599        WARN_ON(hw->mac.type != e1000_i210);
1600
1601        if (enable)
1602                adapter->flags |= IGB_FLAG_FQTSS;
1603        else
1604                adapter->flags &= ~IGB_FLAG_FQTSS;
1605
1606        if (netif_running(netdev))
1607                schedule_work(&adapter->reset_task);
1608}
1609
1610static bool is_fqtss_enabled(struct igb_adapter *adapter)
1611{
1612        return (adapter->flags & IGB_FLAG_FQTSS) ? true : false;
1613}
1614
1615static void set_tx_desc_fetch_prio(struct e1000_hw *hw, int queue,
1616                                   enum tx_queue_prio prio)
1617{
1618        u32 val;
1619
1620        WARN_ON(hw->mac.type != e1000_i210);
1621        WARN_ON(queue < 0 || queue > 4);
1622
1623        val = rd32(E1000_I210_TXDCTL(queue));
1624
1625        if (prio == TX_QUEUE_PRIO_HIGH)
1626                val |= E1000_TXDCTL_PRIORITY;
1627        else
1628                val &= ~E1000_TXDCTL_PRIORITY;
1629
1630        wr32(E1000_I210_TXDCTL(queue), val);
1631}
1632
1633static void set_queue_mode(struct e1000_hw *hw, int queue, enum queue_mode mode)
1634{
1635        u32 val;
1636
1637        WARN_ON(hw->mac.type != e1000_i210);
1638        WARN_ON(queue < 0 || queue > 1);
1639
1640        val = rd32(E1000_I210_TQAVCC(queue));
1641
1642        if (mode == QUEUE_MODE_STREAM_RESERVATION)
1643                val |= E1000_TQAVCC_QUEUEMODE;
1644        else
1645                val &= ~E1000_TQAVCC_QUEUEMODE;
1646
1647        wr32(E1000_I210_TQAVCC(queue), val);
1648}
1649
1650static bool is_any_cbs_enabled(struct igb_adapter *adapter)
1651{
1652        int i;
1653
1654        for (i = 0; i < adapter->num_tx_queues; i++) {
1655                if (adapter->tx_ring[i]->cbs_enable)
1656                        return true;
1657        }
1658
1659        return false;
1660}
1661
1662static bool is_any_txtime_enabled(struct igb_adapter *adapter)
1663{
1664        int i;
1665
1666        for (i = 0; i < adapter->num_tx_queues; i++) {
1667                if (adapter->tx_ring[i]->launchtime_enable)
1668                        return true;
1669        }
1670
1671        return false;
1672}
1673
1674/**
1675 *  igb_config_tx_modes - Configure "Qav Tx mode" features on igb
1676 *  @adapter: pointer to adapter struct
1677 *  @queue: queue number
1678 *
1679 *  Configure CBS and Launchtime for a given hardware queue.
1680 *  Parameters are retrieved from the correct Tx ring, so
1681 *  igb_save_cbs_params() and igb_save_txtime_params() should be used
1682 *  for setting those correctly prior to this function being called.
1683 **/
1684static void igb_config_tx_modes(struct igb_adapter *adapter, int queue)
1685{
1686        struct net_device *netdev = adapter->netdev;
1687        struct e1000_hw *hw = &adapter->hw;
1688        struct igb_ring *ring;
1689        u32 tqavcc, tqavctrl;
1690        u16 value;
1691
1692        WARN_ON(hw->mac.type != e1000_i210);
1693        WARN_ON(queue < 0 || queue > 1);
1694        ring = adapter->tx_ring[queue];
1695
1696        /* If any of the Qav features is enabled, configure queues as SR and
1697         * with HIGH PRIO. If none is, then configure them with LOW PRIO and
1698         * as SP.
1699         */
1700        if (ring->cbs_enable || ring->launchtime_enable) {
1701                set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_HIGH);
1702                set_queue_mode(hw, queue, QUEUE_MODE_STREAM_RESERVATION);
1703        } else {
1704                set_tx_desc_fetch_prio(hw, queue, TX_QUEUE_PRIO_LOW);
1705                set_queue_mode(hw, queue, QUEUE_MODE_STRICT_PRIORITY);
1706        }
1707
1708        /* If CBS is enabled, set DataTranARB and config its parameters. */
1709        if (ring->cbs_enable || queue == 0) {
1710                /* i210 does not allow the queue 0 to be in the Strict
1711                 * Priority mode while the Qav mode is enabled, so,
1712                 * instead of disabling strict priority mode, we give
1713                 * queue 0 the maximum of credits possible.
1714                 *
1715                 * See section 8.12.19 of the i210 datasheet, "Note:
1716                 * Queue0 QueueMode must be set to 1b when
1717                 * TransmitMode is set to Qav."
1718                 */
1719                if (queue == 0 && !ring->cbs_enable) {
1720                        /* max "linkspeed" idleslope in kbps */
1721                        ring->idleslope = 1000000;
1722                        ring->hicredit = ETH_FRAME_LEN;
1723                }
1724
1725                /* Always set data transfer arbitration to credit-based
1726                 * shaper algorithm on TQAVCTRL if CBS is enabled for any of
1727                 * the queues.
1728                 */
1729                tqavctrl = rd32(E1000_I210_TQAVCTRL);
1730                tqavctrl |= E1000_TQAVCTRL_DATATRANARB;
1731                wr32(E1000_I210_TQAVCTRL, tqavctrl);
1732
1733                /* According to i210 datasheet section 7.2.7.7, we should set
1734                 * the 'idleSlope' field from TQAVCC register following the
1735                 * equation:
1736                 *
1737                 * For 100 Mbps link speed:
1738                 *
1739                 *     value = BW * 0x7735 * 0.2                          (E1)
1740                 *
1741                 * For 1000Mbps link speed:
1742                 *
1743                 *     value = BW * 0x7735 * 2                            (E2)
1744                 *
1745                 * E1 and E2 can be merged into one equation as shown below.
1746                 * Note that 'link-speed' is in Mbps.
1747                 *
1748                 *     value = BW * 0x7735 * 2 * link-speed
1749                 *                           --------------               (E3)
1750                 *                                1000
1751                 *
1752                 * 'BW' is the percentage bandwidth out of full link speed
1753                 * which can be found with the following equation. Note that
1754                 * idleSlope here is the parameter from this function which
1755                 * is in kbps.
1756                 *
1757                 *     BW =     idleSlope
1758                 *          -----------------                             (E4)
1759                 *          link-speed * 1000
1760                 *
1761                 * That said, we can come up with a generic equation to
1762                 * calculate the value we should set it TQAVCC register by
1763                 * replacing 'BW' in E3 by E4. The resulting equation is:
1764                 *
1765                 * value =     idleSlope     * 0x7735 * 2 * link-speed
1766                 *         -----------------            --------------    (E5)
1767                 *         link-speed * 1000                 1000
1768                 *
1769                 * 'link-speed' is present in both sides of the fraction so
1770                 * it is canceled out. The final equation is the following:
1771                 *
1772                 *     value = idleSlope * 61034
1773                 *             -----------------                          (E6)
1774                 *                  1000000
1775                 *
1776                 * NOTE: For i210, given the above, we can see that idleslope
1777                 *       is represented in 16.38431 kbps units by the value at
1778                 *       the TQAVCC register (1Gbps / 61034), which reduces
1779                 *       the granularity for idleslope increments.
1780                 *       For instance, if you want to configure a 2576kbps
1781                 *       idleslope, the value to be written on the register
1782                 *       would have to be 157.23. If rounded down, you end
1783                 *       up with less bandwidth available than originally
1784                 *       required (~2572 kbps). If rounded up, you end up
1785                 *       with a higher bandwidth (~2589 kbps). Below the
1786                 *       approach we take is to always round up the
1787                 *       calculated value, so the resulting bandwidth might
1788                 *       be slightly higher for some configurations.
1789                 */
1790                value = DIV_ROUND_UP_ULL(ring->idleslope * 61034ULL, 1000000);
1791
1792                tqavcc = rd32(E1000_I210_TQAVCC(queue));
1793                tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1794                tqavcc |= value;
1795                wr32(E1000_I210_TQAVCC(queue), tqavcc);
1796
1797                wr32(E1000_I210_TQAVHC(queue),
1798                     0x80000000 + ring->hicredit * 0x7735);
1799        } else {
1800
1801                /* Set idleSlope to zero. */
1802                tqavcc = rd32(E1000_I210_TQAVCC(queue));
1803                tqavcc &= ~E1000_TQAVCC_IDLESLOPE_MASK;
1804                wr32(E1000_I210_TQAVCC(queue), tqavcc);
1805
1806                /* Set hiCredit to zero. */
1807                wr32(E1000_I210_TQAVHC(queue), 0);
1808
1809                /* If CBS is not enabled for any queues anymore, then return to
1810                 * the default state of Data Transmission Arbitration on
1811                 * TQAVCTRL.
1812                 */
1813                if (!is_any_cbs_enabled(adapter)) {
1814                        tqavctrl = rd32(E1000_I210_TQAVCTRL);
1815                        tqavctrl &= ~E1000_TQAVCTRL_DATATRANARB;
1816                        wr32(E1000_I210_TQAVCTRL, tqavctrl);
1817                }
1818        }
1819
1820        /* If LaunchTime is enabled, set DataTranTIM. */
1821        if (ring->launchtime_enable) {
1822                /* Always set DataTranTIM on TQAVCTRL if LaunchTime is enabled
1823                 * for any of the SR queues, and configure fetchtime delta.
1824                 * XXX NOTE:
1825                 *     - LaunchTime will be enabled for all SR queues.
1826                 *     - A fixed offset can be added relative to the launch
1827                 *       time of all packets if configured at reg LAUNCH_OS0.
1828                 *       We are keeping it as 0 for now (default value).
1829                 */
1830                tqavctrl = rd32(E1000_I210_TQAVCTRL);
1831                tqavctrl |= E1000_TQAVCTRL_DATATRANTIM |
1832                       E1000_TQAVCTRL_FETCHTIME_DELTA;
1833                wr32(E1000_I210_TQAVCTRL, tqavctrl);
1834        } else {
1835                /* If Launchtime is not enabled for any SR queues anymore,
1836                 * then clear DataTranTIM on TQAVCTRL and clear fetchtime delta,
1837                 * effectively disabling Launchtime.
1838                 */
1839                if (!is_any_txtime_enabled(adapter)) {
1840                        tqavctrl = rd32(E1000_I210_TQAVCTRL);
1841                        tqavctrl &= ~E1000_TQAVCTRL_DATATRANTIM;
1842                        tqavctrl &= ~E1000_TQAVCTRL_FETCHTIME_DELTA;
1843                        wr32(E1000_I210_TQAVCTRL, tqavctrl);
1844                }
1845        }
1846
1847        /* XXX: In i210 controller the sendSlope and loCredit parameters from
1848         * CBS are not configurable by software so we don't do any 'controller
1849         * configuration' in respect to these parameters.
1850         */
1851
1852        netdev_dbg(netdev, "Qav Tx mode: cbs %s, launchtime %s, queue %d idleslope %d sendslope %d hiCredit %d locredit %d\n",
1853                   ring->cbs_enable ? "enabled" : "disabled",
1854                   ring->launchtime_enable ? "enabled" : "disabled",
1855                   queue,
1856                   ring->idleslope, ring->sendslope,
1857                   ring->hicredit, ring->locredit);
1858}
1859
1860static int igb_save_txtime_params(struct igb_adapter *adapter, int queue,
1861                                  bool enable)
1862{
1863        struct igb_ring *ring;
1864
1865        if (queue < 0 || queue > adapter->num_tx_queues)
1866                return -EINVAL;
1867
1868        ring = adapter->tx_ring[queue];
1869        ring->launchtime_enable = enable;
1870
1871        return 0;
1872}
1873
1874static int igb_save_cbs_params(struct igb_adapter *adapter, int queue,
1875                               bool enable, int idleslope, int sendslope,
1876                               int hicredit, int locredit)
1877{
1878        struct igb_ring *ring;
1879
1880        if (queue < 0 || queue > adapter->num_tx_queues)
1881                return -EINVAL;
1882
1883        ring = adapter->tx_ring[queue];
1884
1885        ring->cbs_enable = enable;
1886        ring->idleslope = idleslope;
1887        ring->sendslope = sendslope;
1888        ring->hicredit = hicredit;
1889        ring->locredit = locredit;
1890
1891        return 0;
1892}
1893
1894/**
1895 *  igb_setup_tx_mode - Switch to/from Qav Tx mode when applicable
1896 *  @adapter: pointer to adapter struct
1897 *
1898 *  Configure TQAVCTRL register switching the controller's Tx mode
1899 *  if FQTSS mode is enabled or disabled. Additionally, will issue
1900 *  a call to igb_config_tx_modes() per queue so any previously saved
1901 *  Tx parameters are applied.
1902 **/
1903static void igb_setup_tx_mode(struct igb_adapter *adapter)
1904{
1905        struct net_device *netdev = adapter->netdev;
1906        struct e1000_hw *hw = &adapter->hw;
1907        u32 val;
1908
1909        /* Only i210 controller supports changing the transmission mode. */
1910        if (hw->mac.type != e1000_i210)
1911                return;
1912
1913        if (is_fqtss_enabled(adapter)) {
1914                int i, max_queue;
1915
1916                /* Configure TQAVCTRL register: set transmit mode to 'Qav',
1917                 * set data fetch arbitration to 'round robin', set SP_WAIT_SR
1918                 * so SP queues wait for SR ones.
1919                 */
1920                val = rd32(E1000_I210_TQAVCTRL);
1921                val |= E1000_TQAVCTRL_XMIT_MODE | E1000_TQAVCTRL_SP_WAIT_SR;
1922                val &= ~E1000_TQAVCTRL_DATAFETCHARB;
1923                wr32(E1000_I210_TQAVCTRL, val);
1924
1925                /* Configure Tx and Rx packet buffers sizes as described in
1926                 * i210 datasheet section 7.2.7.7.
1927                 */
1928                val = rd32(E1000_TXPBS);
1929                val &= ~I210_TXPBSIZE_MASK;
1930                val |= I210_TXPBSIZE_PB0_6KB | I210_TXPBSIZE_PB1_6KB |
1931                        I210_TXPBSIZE_PB2_6KB | I210_TXPBSIZE_PB3_6KB;
1932                wr32(E1000_TXPBS, val);
1933
1934                val = rd32(E1000_RXPBS);
1935                val &= ~I210_RXPBSIZE_MASK;
1936                val |= I210_RXPBSIZE_PB_30KB;
1937                wr32(E1000_RXPBS, val);
1938
1939                /* Section 8.12.9 states that MAX_TPKT_SIZE from DTXMXPKTSZ
1940                 * register should not exceed the buffer size programmed in
1941                 * TXPBS. The smallest buffer size programmed in TXPBS is 4kB
1942                 * so according to the datasheet we should set MAX_TPKT_SIZE to
1943                 * 4kB / 64.
1944                 *
1945                 * However, when we do so, no frame from queue 2 and 3 are
1946                 * transmitted.  It seems the MAX_TPKT_SIZE should not be great
1947                 * or _equal_ to the buffer size programmed in TXPBS. For this
1948                 * reason, we set set MAX_ TPKT_SIZE to (4kB - 1) / 64.
1949                 */
1950                val = (4096 - 1) / 64;
1951                wr32(E1000_I210_DTXMXPKTSZ, val);
1952
1953                /* Since FQTSS mode is enabled, apply any CBS configuration
1954                 * previously set. If no previous CBS configuration has been
1955                 * done, then the initial configuration is applied, which means
1956                 * CBS is disabled.
1957                 */
1958                max_queue = (adapter->num_tx_queues < I210_SR_QUEUES_NUM) ?
1959                            adapter->num_tx_queues : I210_SR_QUEUES_NUM;
1960
1961                for (i = 0; i < max_queue; i++) {
1962                        igb_config_tx_modes(adapter, i);
1963                }
1964        } else {
1965                wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
1966                wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
1967                wr32(E1000_I210_DTXMXPKTSZ, I210_DTXMXPKTSZ_DEFAULT);
1968
1969                val = rd32(E1000_I210_TQAVCTRL);
1970                /* According to Section 8.12.21, the other flags we've set when
1971                 * enabling FQTSS are not relevant when disabling FQTSS so we
1972                 * don't set they here.
1973                 */
1974                val &= ~E1000_TQAVCTRL_XMIT_MODE;
1975                wr32(E1000_I210_TQAVCTRL, val);
1976        }
1977
1978        netdev_dbg(netdev, "FQTSS %s\n", (is_fqtss_enabled(adapter)) ?
1979                   "enabled" : "disabled");
1980}
1981
1982/**
1983 *  igb_configure - configure the hardware for RX and TX
1984 *  @adapter: private board structure
1985 **/
1986static void igb_configure(struct igb_adapter *adapter)
1987{
1988        struct net_device *netdev = adapter->netdev;
1989        int i;
1990
1991        igb_get_hw_control(adapter);
1992        igb_set_rx_mode(netdev);
1993        igb_setup_tx_mode(adapter);
1994
1995        igb_restore_vlan(adapter);
1996
1997        igb_setup_tctl(adapter);
1998        igb_setup_mrqc(adapter);
1999        igb_setup_rctl(adapter);
2000
2001        igb_nfc_filter_restore(adapter);
2002        igb_configure_tx(adapter);
2003        igb_configure_rx(adapter);
2004
2005        igb_rx_fifo_flush_82575(&adapter->hw);
2006
2007        /* call igb_desc_unused which always leaves
2008         * at least 1 descriptor unused to make sure
2009         * next_to_use != next_to_clean
2010         */
2011        for (i = 0; i < adapter->num_rx_queues; i++) {
2012                struct igb_ring *ring = adapter->rx_ring[i];
2013                igb_alloc_rx_buffers(ring, igb_desc_unused(ring));
2014        }
2015}
2016
2017/**
2018 *  igb_power_up_link - Power up the phy/serdes link
2019 *  @adapter: address of board private structure
2020 **/
2021void igb_power_up_link(struct igb_adapter *adapter)
2022{
2023        igb_reset_phy(&adapter->hw);
2024
2025        if (adapter->hw.phy.media_type == e1000_media_type_copper)
2026                igb_power_up_phy_copper(&adapter->hw);
2027        else
2028                igb_power_up_serdes_link_82575(&adapter->hw);
2029
2030        igb_setup_link(&adapter->hw);
2031}
2032
2033/**
2034 *  igb_power_down_link - Power down the phy/serdes link
2035 *  @adapter: address of board private structure
2036 */
2037static void igb_power_down_link(struct igb_adapter *adapter)
2038{
2039        if (adapter->hw.phy.media_type == e1000_media_type_copper)
2040                igb_power_down_phy_copper_82575(&adapter->hw);
2041        else
2042                igb_shutdown_serdes_link_82575(&adapter->hw);
2043}
2044
2045/**
2046 * igb_check_swap_media -  Detect and switch function for Media Auto Sense
2047 * @adapter: address of the board private structure
2048 **/
2049static void igb_check_swap_media(struct igb_adapter *adapter)
2050{
2051        struct e1000_hw *hw = &adapter->hw;
2052        u32 ctrl_ext, connsw;
2053        bool swap_now = false;
2054
2055        ctrl_ext = rd32(E1000_CTRL_EXT);
2056        connsw = rd32(E1000_CONNSW);
2057
2058        /* need to live swap if current media is copper and we have fiber/serdes
2059         * to go to.
2060         */
2061
2062        if ((hw->phy.media_type == e1000_media_type_copper) &&
2063            (!(connsw & E1000_CONNSW_AUTOSENSE_EN))) {
2064                swap_now = true;
2065        } else if ((hw->phy.media_type != e1000_media_type_copper) &&
2066                   !(connsw & E1000_CONNSW_SERDESD)) {
2067                /* copper signal takes time to appear */
2068                if (adapter->copper_tries < 4) {
2069                        adapter->copper_tries++;
2070                        connsw |= E1000_CONNSW_AUTOSENSE_CONF;
2071                        wr32(E1000_CONNSW, connsw);
2072                        return;
2073                } else {
2074                        adapter->copper_tries = 0;
2075                        if ((connsw & E1000_CONNSW_PHYSD) &&
2076                            (!(connsw & E1000_CONNSW_PHY_PDN))) {
2077                                swap_now = true;
2078                                connsw &= ~E1000_CONNSW_AUTOSENSE_CONF;
2079                                wr32(E1000_CONNSW, connsw);
2080                        }
2081                }
2082        }
2083
2084        if (!swap_now)
2085                return;
2086
2087        switch (hw->phy.media_type) {
2088        case e1000_media_type_copper:
2089                netdev_info(adapter->netdev,
2090                        "MAS: changing media to fiber/serdes\n");
2091                ctrl_ext |=
2092                        E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2093                adapter->flags |= IGB_FLAG_MEDIA_RESET;
2094                adapter->copper_tries = 0;
2095                break;
2096        case e1000_media_type_internal_serdes:
2097        case e1000_media_type_fiber:
2098                netdev_info(adapter->netdev,
2099                        "MAS: changing media to copper\n");
2100                ctrl_ext &=
2101                        ~E1000_CTRL_EXT_LINK_MODE_PCIE_SERDES;
2102                adapter->flags |= IGB_FLAG_MEDIA_RESET;
2103                break;
2104        default:
2105                /* shouldn't get here during regular operation */
2106                netdev_err(adapter->netdev,
2107                        "AMS: Invalid media type found, returning\n");
2108                break;
2109        }
2110        wr32(E1000_CTRL_EXT, ctrl_ext);
2111}
2112
2113/**
2114 *  igb_up - Open the interface and prepare it to handle traffic
2115 *  @adapter: board private structure
2116 **/
2117int igb_up(struct igb_adapter *adapter)
2118{
2119        struct e1000_hw *hw = &adapter->hw;
2120        int i;
2121
2122        /* hardware has been reset, we need to reload some things */
2123        igb_configure(adapter);
2124
2125        clear_bit(__IGB_DOWN, &adapter->state);
2126
2127        for (i = 0; i < adapter->num_q_vectors; i++)
2128                napi_enable(&(adapter->q_vector[i]->napi));
2129
2130        if (adapter->flags & IGB_FLAG_HAS_MSIX)
2131                igb_configure_msix(adapter);
2132        else
2133                igb_assign_vector(adapter->q_vector[0], 0);
2134
2135        /* Clear any pending interrupts. */
2136        rd32(E1000_TSICR);
2137        rd32(E1000_ICR);
2138        igb_irq_enable(adapter);
2139
2140        /* notify VFs that reset has been completed */
2141        if (adapter->vfs_allocated_count) {
2142                u32 reg_data = rd32(E1000_CTRL_EXT);
2143
2144                reg_data |= E1000_CTRL_EXT_PFRSTD;
2145                wr32(E1000_CTRL_EXT, reg_data);
2146        }
2147
2148        netif_tx_start_all_queues(adapter->netdev);
2149
2150        /* start the watchdog. */
2151        hw->mac.get_link_status = 1;
2152        schedule_work(&adapter->watchdog_task);
2153
2154        if ((adapter->flags & IGB_FLAG_EEE) &&
2155            (!hw->dev_spec._82575.eee_disable))
2156                adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
2157
2158        return 0;
2159}
2160
2161void igb_down(struct igb_adapter *adapter)
2162{
2163        struct net_device *netdev = adapter->netdev;
2164        struct e1000_hw *hw = &adapter->hw;
2165        u32 tctl, rctl;
2166        int i;
2167
2168        /* signal that we're down so the interrupt handler does not
2169         * reschedule our watchdog timer
2170         */
2171        set_bit(__IGB_DOWN, &adapter->state);
2172
2173        /* disable receives in the hardware */
2174        rctl = rd32(E1000_RCTL);
2175        wr32(E1000_RCTL, rctl & ~E1000_RCTL_EN);
2176        /* flush and sleep below */
2177
2178        igb_nfc_filter_exit(adapter);
2179
2180        netif_carrier_off(netdev);
2181        netif_tx_stop_all_queues(netdev);
2182
2183        /* disable transmits in the hardware */
2184        tctl = rd32(E1000_TCTL);
2185        tctl &= ~E1000_TCTL_EN;
2186        wr32(E1000_TCTL, tctl);
2187        /* flush both disables and wait for them to finish */
2188        wrfl();
2189        usleep_range(10000, 11000);
2190
2191        igb_irq_disable(adapter);
2192
2193        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
2194
2195        for (i = 0; i < adapter->num_q_vectors; i++) {
2196                if (adapter->q_vector[i]) {
2197                        napi_synchronize(&adapter->q_vector[i]->napi);
2198                        napi_disable(&adapter->q_vector[i]->napi);
2199                }
2200        }
2201
2202        del_timer_sync(&adapter->watchdog_timer);
2203        del_timer_sync(&adapter->phy_info_timer);
2204
2205        /* record the stats before reset*/
2206        spin_lock(&adapter->stats64_lock);
2207        igb_update_stats(adapter);
2208        spin_unlock(&adapter->stats64_lock);
2209
2210        adapter->link_speed = 0;
2211        adapter->link_duplex = 0;
2212
2213        if (!pci_channel_offline(adapter->pdev))
2214                igb_reset(adapter);
2215
2216        /* clear VLAN promisc flag so VFTA will be updated if necessary */
2217        adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
2218
2219        igb_clean_all_tx_rings(adapter);
2220        igb_clean_all_rx_rings(adapter);
2221#ifdef CONFIG_IGB_DCA
2222
2223        /* since we reset the hardware DCA settings were cleared */
2224        igb_setup_dca(adapter);
2225#endif
2226}
2227
2228void igb_reinit_locked(struct igb_adapter *adapter)
2229{
2230        while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
2231                usleep_range(1000, 2000);
2232        igb_down(adapter);
2233        igb_up(adapter);
2234        clear_bit(__IGB_RESETTING, &adapter->state);
2235}
2236
2237/** igb_enable_mas - Media Autosense re-enable after swap
2238 *
2239 * @adapter: adapter struct
2240 **/
2241static void igb_enable_mas(struct igb_adapter *adapter)
2242{
2243        struct e1000_hw *hw = &adapter->hw;
2244        u32 connsw = rd32(E1000_CONNSW);
2245
2246        /* configure for SerDes media detect */
2247        if ((hw->phy.media_type == e1000_media_type_copper) &&
2248            (!(connsw & E1000_CONNSW_SERDESD))) {
2249                connsw |= E1000_CONNSW_ENRGSRC;
2250                connsw |= E1000_CONNSW_AUTOSENSE_EN;
2251                wr32(E1000_CONNSW, connsw);
2252                wrfl();
2253        }
2254}
2255
2256void igb_reset(struct igb_adapter *adapter)
2257{
2258        struct pci_dev *pdev = adapter->pdev;
2259        struct e1000_hw *hw = &adapter->hw;
2260        struct e1000_mac_info *mac = &hw->mac;
2261        struct e1000_fc_info *fc = &hw->fc;
2262        u32 pba, hwm;
2263
2264        /* Repartition Pba for greater than 9k mtu
2265         * To take effect CTRL.RST is required.
2266         */
2267        switch (mac->type) {
2268        case e1000_i350:
2269        case e1000_i354:
2270        case e1000_82580:
2271                pba = rd32(E1000_RXPBS);
2272                pba = igb_rxpbs_adjust_82580(pba);
2273                break;
2274        case e1000_82576:
2275                pba = rd32(E1000_RXPBS);
2276                pba &= E1000_RXPBS_SIZE_MASK_82576;
2277                break;
2278        case e1000_82575:
2279        case e1000_i210:
2280        case e1000_i211:
2281        default:
2282                pba = E1000_PBA_34K;
2283                break;
2284        }
2285
2286        if (mac->type == e1000_82575) {
2287                u32 min_rx_space, min_tx_space, needed_tx_space;
2288
2289                /* write Rx PBA so that hardware can report correct Tx PBA */
2290                wr32(E1000_PBA, pba);
2291
2292                /* To maintain wire speed transmits, the Tx FIFO should be
2293                 * large enough to accommodate two full transmit packets,
2294                 * rounded up to the next 1KB and expressed in KB.  Likewise,
2295                 * the Rx FIFO should be large enough to accommodate at least
2296                 * one full receive packet and is similarly rounded up and
2297                 * expressed in KB.
2298                 */
2299                min_rx_space = DIV_ROUND_UP(MAX_JUMBO_FRAME_SIZE, 1024);
2300
2301                /* The Tx FIFO also stores 16 bytes of information about the Tx
2302                 * but don't include Ethernet FCS because hardware appends it.
2303                 * We only need to round down to the nearest 512 byte block
2304                 * count since the value we care about is 2 frames, not 1.
2305                 */
2306                min_tx_space = adapter->max_frame_size;
2307                min_tx_space += sizeof(union e1000_adv_tx_desc) - ETH_FCS_LEN;
2308                min_tx_space = DIV_ROUND_UP(min_tx_space, 512);
2309
2310                /* upper 16 bits has Tx packet buffer allocation size in KB */
2311                needed_tx_space = min_tx_space - (rd32(E1000_PBA) >> 16);
2312
2313                /* If current Tx allocation is less than the min Tx FIFO size,
2314                 * and the min Tx FIFO size is less than the current Rx FIFO
2315                 * allocation, take space away from current Rx allocation.
2316                 */
2317                if (needed_tx_space < pba) {
2318                        pba -= needed_tx_space;
2319
2320                        /* if short on Rx space, Rx wins and must trump Tx
2321                         * adjustment
2322                         */
2323                        if (pba < min_rx_space)
2324                                pba = min_rx_space;
2325                }
2326
2327                /* adjust PBA for jumbo frames */
2328                wr32(E1000_PBA, pba);
2329        }
2330
2331        /* flow control settings
2332         * The high water mark must be low enough to fit one full frame
2333         * after transmitting the pause frame.  As such we must have enough
2334         * space to allow for us to complete our current transmit and then
2335         * receive the frame that is in progress from the link partner.
2336         * Set it to:
2337         * - the full Rx FIFO size minus one full Tx plus one full Rx frame
2338         */
2339        hwm = (pba << 10) - (adapter->max_frame_size + MAX_JUMBO_FRAME_SIZE);
2340
2341        fc->high_water = hwm & 0xFFFFFFF0;      /* 16-byte granularity */
2342        fc->low_water = fc->high_water - 16;
2343        fc->pause_time = 0xFFFF;
2344        fc->send_xon = 1;
2345        fc->current_mode = fc->requested_mode;
2346
2347        /* disable receive for all VFs and wait one second */
2348        if (adapter->vfs_allocated_count) {
2349                int i;
2350
2351                for (i = 0 ; i < adapter->vfs_allocated_count; i++)
2352                        adapter->vf_data[i].flags &= IGB_VF_FLAG_PF_SET_MAC;
2353
2354                /* ping all the active vfs to let them know we are going down */
2355                igb_ping_all_vfs(adapter);
2356
2357                /* disable transmits and receives */
2358                wr32(E1000_VFRE, 0);
2359                wr32(E1000_VFTE, 0);
2360        }
2361
2362        /* Allow time for pending master requests to run */
2363        hw->mac.ops.reset_hw(hw);
2364        wr32(E1000_WUC, 0);
2365
2366        if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
2367                /* need to resetup here after media swap */
2368                adapter->ei.get_invariants(hw);
2369                adapter->flags &= ~IGB_FLAG_MEDIA_RESET;
2370        }
2371        if ((mac->type == e1000_82575 || mac->type == e1000_i350) &&
2372            (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
2373                igb_enable_mas(adapter);
2374        }
2375        if (hw->mac.ops.init_hw(hw))
2376                dev_err(&pdev->dev, "Hardware Error\n");
2377
2378        /* RAR registers were cleared during init_hw, clear mac table */
2379        igb_flush_mac_table(adapter);
2380        __dev_uc_unsync(adapter->netdev, NULL);
2381
2382        /* Recover default RAR entry */
2383        igb_set_default_mac_filter(adapter);
2384
2385        /* Flow control settings reset on hardware reset, so guarantee flow
2386         * control is off when forcing speed.
2387         */
2388        if (!hw->mac.autoneg)
2389                igb_force_mac_fc(hw);
2390
2391        igb_init_dmac(adapter, pba);
2392#ifdef CONFIG_IGB_HWMON
2393        /* Re-initialize the thermal sensor on i350 devices. */
2394        if (!test_bit(__IGB_DOWN, &adapter->state)) {
2395                if (mac->type == e1000_i350 && hw->bus.func == 0) {
2396                        /* If present, re-initialize the external thermal sensor
2397                         * interface.
2398                         */
2399                        if (adapter->ets)
2400                                mac->ops.init_thermal_sensor_thresh(hw);
2401                }
2402        }
2403#endif
2404        /* Re-establish EEE setting */
2405        if (hw->phy.media_type == e1000_media_type_copper) {
2406                switch (mac->type) {
2407                case e1000_i350:
2408                case e1000_i210:
2409                case e1000_i211:
2410                        igb_set_eee_i350(hw, true, true);
2411                        break;
2412                case e1000_i354:
2413                        igb_set_eee_i354(hw, true, true);
2414                        break;
2415                default:
2416                        break;
2417                }
2418        }
2419        if (!netif_running(adapter->netdev))
2420                igb_power_down_link(adapter);
2421
2422        igb_update_mng_vlan(adapter);
2423
2424        /* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
2425        wr32(E1000_VET, ETHERNET_IEEE_VLAN_TYPE);
2426
2427        /* Re-enable PTP, where applicable. */
2428        if (adapter->ptp_flags & IGB_PTP_ENABLED)
2429                igb_ptp_reset(adapter);
2430
2431        igb_get_phy_info(hw);
2432}
2433
2434static netdev_features_t igb_fix_features(struct net_device *netdev,
2435        netdev_features_t features)
2436{
2437        /* Since there is no support for separate Rx/Tx vlan accel
2438         * enable/disable make sure Tx flag is always in same state as Rx.
2439         */
2440        if (features & NETIF_F_HW_VLAN_CTAG_RX)
2441                features |= NETIF_F_HW_VLAN_CTAG_TX;
2442        else
2443                features &= ~NETIF_F_HW_VLAN_CTAG_TX;
2444
2445        return features;
2446}
2447
2448static int igb_set_features(struct net_device *netdev,
2449        netdev_features_t features)
2450{
2451        netdev_features_t changed = netdev->features ^ features;
2452        struct igb_adapter *adapter = netdev_priv(netdev);
2453
2454        if (changed & NETIF_F_HW_VLAN_CTAG_RX)
2455                igb_vlan_mode(netdev, features);
2456
2457        if (!(changed & (NETIF_F_RXALL | NETIF_F_NTUPLE)))
2458                return 0;
2459
2460        if (!(features & NETIF_F_NTUPLE)) {
2461                struct hlist_node *node2;
2462                struct igb_nfc_filter *rule;
2463
2464                spin_lock(&adapter->nfc_lock);
2465                hlist_for_each_entry_safe(rule, node2,
2466                                          &adapter->nfc_filter_list, nfc_node) {
2467                        igb_erase_filter(adapter, rule);
2468                        hlist_del(&rule->nfc_node);
2469                        kfree(rule);
2470                }
2471                spin_unlock(&adapter->nfc_lock);
2472                adapter->nfc_filter_count = 0;
2473        }
2474
2475        netdev->features = features;
2476
2477        if (netif_running(netdev))
2478                igb_reinit_locked(adapter);
2479        else
2480                igb_reset(adapter);
2481
2482        return 1;
2483}
2484
2485static int igb_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
2486                           struct net_device *dev,
2487                           const unsigned char *addr, u16 vid,
2488                           u16 flags,
2489                           struct netlink_ext_ack *extack)
2490{
2491        /* guarantee we can provide a unique filter for the unicast address */
2492        if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr)) {
2493                struct igb_adapter *adapter = netdev_priv(dev);
2494                int vfn = adapter->vfs_allocated_count;
2495
2496                if (netdev_uc_count(dev) >= igb_available_rars(adapter, vfn))
2497                        return -ENOMEM;
2498        }
2499
2500        return ndo_dflt_fdb_add(ndm, tb, dev, addr, vid, flags);
2501}
2502
2503#define IGB_MAX_MAC_HDR_LEN     127
2504#define IGB_MAX_NETWORK_HDR_LEN 511
2505
2506static netdev_features_t
2507igb_features_check(struct sk_buff *skb, struct net_device *dev,
2508                   netdev_features_t features)
2509{
2510        unsigned int network_hdr_len, mac_hdr_len;
2511
2512        /* Make certain the headers can be described by a context descriptor */
2513        mac_hdr_len = skb_network_header(skb) - skb->data;
2514        if (unlikely(mac_hdr_len > IGB_MAX_MAC_HDR_LEN))
2515                return features & ~(NETIF_F_HW_CSUM |
2516                                    NETIF_F_SCTP_CRC |
2517                                    NETIF_F_GSO_UDP_L4 |
2518                                    NETIF_F_HW_VLAN_CTAG_TX |
2519                                    NETIF_F_TSO |
2520                                    NETIF_F_TSO6);
2521
2522        network_hdr_len = skb_checksum_start(skb) - skb_network_header(skb);
2523        if (unlikely(network_hdr_len >  IGB_MAX_NETWORK_HDR_LEN))
2524                return features & ~(NETIF_F_HW_CSUM |
2525                                    NETIF_F_SCTP_CRC |
2526                                    NETIF_F_GSO_UDP_L4 |
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 =
2653                                (__force __be16)match.key->vlan_priority;
2654                }
2655        }
2656
2657        input->action = traffic_class;
2658        input->cookie = f->cookie;
2659
2660        return 0;
2661}
2662
2663static int igb_configure_clsflower(struct igb_adapter *adapter,
2664                                   struct flow_cls_offload *cls_flower)
2665{
2666        struct netlink_ext_ack *extack = cls_flower->common.extack;
2667        struct igb_nfc_filter *filter, *f;
2668        int err, tc;
2669
2670        tc = tc_classid_to_hwtc(adapter->netdev, cls_flower->classid);
2671        if (tc < 0) {
2672                NL_SET_ERR_MSG_MOD(extack, "Invalid traffic class");
2673                return -EINVAL;
2674        }
2675
2676        filter = kzalloc(sizeof(*filter), GFP_KERNEL);
2677        if (!filter)
2678                return -ENOMEM;
2679
2680        err = igb_parse_cls_flower(adapter, cls_flower, tc, filter);
2681        if (err < 0)
2682                goto err_parse;
2683
2684        spin_lock(&adapter->nfc_lock);
2685
2686        hlist_for_each_entry(f, &adapter->nfc_filter_list, nfc_node) {
2687                if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2688                        err = -EEXIST;
2689                        NL_SET_ERR_MSG_MOD(extack,
2690                                           "This filter is already set in ethtool");
2691                        goto err_locked;
2692                }
2693        }
2694
2695        hlist_for_each_entry(f, &adapter->cls_flower_list, nfc_node) {
2696                if (!memcmp(&f->filter, &filter->filter, sizeof(f->filter))) {
2697                        err = -EEXIST;
2698                        NL_SET_ERR_MSG_MOD(extack,
2699                                           "This filter is already set in cls_flower");
2700                        goto err_locked;
2701                }
2702        }
2703
2704        err = igb_add_filter(adapter, filter);
2705        if (err < 0) {
2706                NL_SET_ERR_MSG_MOD(extack, "Could not add filter to the adapter");
2707                goto err_locked;
2708        }
2709
2710        hlist_add_head(&filter->nfc_node, &adapter->cls_flower_list);
2711
2712        spin_unlock(&adapter->nfc_lock);
2713
2714        return 0;
2715
2716err_locked:
2717        spin_unlock(&adapter->nfc_lock);
2718
2719err_parse:
2720        kfree(filter);
2721
2722        return err;
2723}
2724
2725static int igb_delete_clsflower(struct igb_adapter *adapter,
2726                                struct flow_cls_offload *cls_flower)
2727{
2728        struct igb_nfc_filter *filter;
2729        int err;
2730
2731        spin_lock(&adapter->nfc_lock);
2732
2733        hlist_for_each_entry(filter, &adapter->cls_flower_list, nfc_node)
2734                if (filter->cookie == cls_flower->cookie)
2735                        break;
2736
2737        if (!filter) {
2738                err = -ENOENT;
2739                goto out;
2740        }
2741
2742        err = igb_erase_filter(adapter, filter);
2743        if (err < 0)
2744                goto out;
2745
2746        hlist_del(&filter->nfc_node);
2747        kfree(filter);
2748
2749out:
2750        spin_unlock(&adapter->nfc_lock);
2751
2752        return err;
2753}
2754
2755static int igb_setup_tc_cls_flower(struct igb_adapter *adapter,
2756                                   struct flow_cls_offload *cls_flower)
2757{
2758        switch (cls_flower->command) {
2759        case FLOW_CLS_REPLACE:
2760                return igb_configure_clsflower(adapter, cls_flower);
2761        case FLOW_CLS_DESTROY:
2762                return igb_delete_clsflower(adapter, cls_flower);
2763        case FLOW_CLS_STATS:
2764                return -EOPNOTSUPP;
2765        default:
2766                return -EOPNOTSUPP;
2767        }
2768}
2769
2770static int igb_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
2771                                 void *cb_priv)
2772{
2773        struct igb_adapter *adapter = cb_priv;
2774
2775        if (!tc_cls_can_offload_and_chain0(adapter->netdev, type_data))
2776                return -EOPNOTSUPP;
2777
2778        switch (type) {
2779        case TC_SETUP_CLSFLOWER:
2780                return igb_setup_tc_cls_flower(adapter, type_data);
2781
2782        default:
2783                return -EOPNOTSUPP;
2784        }
2785}
2786
2787static int igb_offload_txtime(struct igb_adapter *adapter,
2788                              struct tc_etf_qopt_offload *qopt)
2789{
2790        struct e1000_hw *hw = &adapter->hw;
2791        int err;
2792
2793        /* Launchtime offloading is only supported by i210 controller. */
2794        if (hw->mac.type != e1000_i210)
2795                return -EOPNOTSUPP;
2796
2797        /* Launchtime offloading is only supported by queues 0 and 1. */
2798        if (qopt->queue < 0 || qopt->queue > 1)
2799                return -EINVAL;
2800
2801        err = igb_save_txtime_params(adapter, qopt->queue, qopt->enable);
2802        if (err)
2803                return err;
2804
2805        igb_offload_apply(adapter, qopt->queue);
2806
2807        return 0;
2808}
2809
2810static LIST_HEAD(igb_block_cb_list);
2811
2812static int igb_setup_tc(struct net_device *dev, enum tc_setup_type type,
2813                        void *type_data)
2814{
2815        struct igb_adapter *adapter = netdev_priv(dev);
2816
2817        switch (type) {
2818        case TC_SETUP_QDISC_CBS:
2819                return igb_offload_cbs(adapter, type_data);
2820        case TC_SETUP_BLOCK:
2821                return flow_block_cb_setup_simple(type_data,
2822                                                  &igb_block_cb_list,
2823                                                  igb_setup_tc_block_cb,
2824                                                  adapter, adapter, true);
2825
2826        case TC_SETUP_QDISC_ETF:
2827                return igb_offload_txtime(adapter, type_data);
2828
2829        default:
2830                return -EOPNOTSUPP;
2831        }
2832}
2833
2834static int igb_xdp_setup(struct net_device *dev, struct netdev_bpf *bpf)
2835{
2836        int i, frame_size = dev->mtu + IGB_ETH_PKT_HDR_PAD;
2837        struct igb_adapter *adapter = netdev_priv(dev);
2838        struct bpf_prog *prog = bpf->prog, *old_prog;
2839        bool running = netif_running(dev);
2840        bool need_reset;
2841
2842        /* verify igb ring attributes are sufficient for XDP */
2843        for (i = 0; i < adapter->num_rx_queues; i++) {
2844                struct igb_ring *ring = adapter->rx_ring[i];
2845
2846                if (frame_size > igb_rx_bufsz(ring)) {
2847                        NL_SET_ERR_MSG_MOD(bpf->extack,
2848                                           "The RX buffer size is too small for the frame size");
2849                        netdev_warn(dev, "XDP RX buffer size %d is too small for the frame size %d\n",
2850                                    igb_rx_bufsz(ring), frame_size);
2851                        return -EINVAL;
2852                }
2853        }
2854
2855        old_prog = xchg(&adapter->xdp_prog, prog);
2856        need_reset = (!!prog != !!old_prog);
2857
2858        /* device is up and bpf is added/removed, must setup the RX queues */
2859        if (need_reset && running) {
2860                igb_close(dev);
2861        } else {
2862                for (i = 0; i < adapter->num_rx_queues; i++)
2863                        (void)xchg(&adapter->rx_ring[i]->xdp_prog,
2864                            adapter->xdp_prog);
2865        }
2866
2867        if (old_prog)
2868                bpf_prog_put(old_prog);
2869
2870        /* bpf is just replaced, RXQ and MTU are already setup */
2871        if (!need_reset)
2872                return 0;
2873
2874        if (running)
2875                igb_open(dev);
2876
2877        return 0;
2878}
2879
2880static int igb_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2881{
2882        switch (xdp->command) {
2883        case XDP_SETUP_PROG:
2884                return igb_xdp_setup(dev, xdp);
2885        default:
2886                return -EINVAL;
2887        }
2888}
2889
2890static void igb_xdp_ring_update_tail(struct igb_ring *ring)
2891{
2892        /* Force memory writes to complete before letting h/w know there
2893         * are new descriptors to fetch.
2894         */
2895        wmb();
2896        writel(ring->next_to_use, ring->tail);
2897}
2898
2899static struct igb_ring *igb_xdp_tx_queue_mapping(struct igb_adapter *adapter)
2900{
2901        unsigned int r_idx = smp_processor_id();
2902
2903        if (r_idx >= adapter->num_tx_queues)
2904                r_idx = r_idx % adapter->num_tx_queues;
2905
2906        return adapter->tx_ring[r_idx];
2907}
2908
2909static int igb_xdp_xmit_back(struct igb_adapter *adapter, struct xdp_buff *xdp)
2910{
2911        struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
2912        int cpu = smp_processor_id();
2913        struct igb_ring *tx_ring;
2914        struct netdev_queue *nq;
2915        u32 ret;
2916
2917        if (unlikely(!xdpf))
2918                return IGB_XDP_CONSUMED;
2919
2920        /* During program transitions its possible adapter->xdp_prog is assigned
2921         * but ring has not been configured yet. In this case simply abort xmit.
2922         */
2923        tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2924        if (unlikely(!tx_ring))
2925                return IGB_XDP_CONSUMED;
2926
2927        nq = txring_txq(tx_ring);
2928        __netif_tx_lock(nq, cpu);
2929        /* Avoid transmit queue timeout since we share it with the slow path */
2930        txq_trans_cond_update(nq);
2931        ret = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2932        __netif_tx_unlock(nq);
2933
2934        return ret;
2935}
2936
2937static int igb_xdp_xmit(struct net_device *dev, int n,
2938                        struct xdp_frame **frames, u32 flags)
2939{
2940        struct igb_adapter *adapter = netdev_priv(dev);
2941        int cpu = smp_processor_id();
2942        struct igb_ring *tx_ring;
2943        struct netdev_queue *nq;
2944        int nxmit = 0;
2945        int i;
2946
2947        if (unlikely(test_bit(__IGB_DOWN, &adapter->state)))
2948                return -ENETDOWN;
2949
2950        if (unlikely(flags & ~XDP_XMIT_FLAGS_MASK))
2951                return -EINVAL;
2952
2953        /* During program transitions its possible adapter->xdp_prog is assigned
2954         * but ring has not been configured yet. In this case simply abort xmit.
2955         */
2956        tx_ring = adapter->xdp_prog ? igb_xdp_tx_queue_mapping(adapter) : NULL;
2957        if (unlikely(!tx_ring))
2958                return -ENXIO;
2959
2960        nq = txring_txq(tx_ring);
2961        __netif_tx_lock(nq, cpu);
2962
2963        /* Avoid transmit queue timeout since we share it with the slow path */
2964        txq_trans_cond_update(nq);
2965
2966        for (i = 0; i < n; i++) {
2967                struct xdp_frame *xdpf = frames[i];
2968                int err;
2969
2970                err = igb_xmit_xdp_ring(adapter, tx_ring, xdpf);
2971                if (err != IGB_XDP_TX)
2972                        break;
2973                nxmit++;
2974        }
2975
2976        __netif_tx_unlock(nq);
2977
2978        if (unlikely(flags & XDP_XMIT_FLUSH))
2979                igb_xdp_ring_update_tail(tx_ring);
2980
2981        return nxmit;
2982}
2983
2984static const struct net_device_ops igb_netdev_ops = {
2985        .ndo_open               = igb_open,
2986        .ndo_stop               = igb_close,
2987        .ndo_start_xmit         = igb_xmit_frame,
2988        .ndo_get_stats64        = igb_get_stats64,
2989        .ndo_set_rx_mode        = igb_set_rx_mode,
2990        .ndo_set_mac_address    = igb_set_mac,
2991        .ndo_change_mtu         = igb_change_mtu,
2992        .ndo_eth_ioctl          = igb_ioctl,
2993        .ndo_tx_timeout         = igb_tx_timeout,
2994        .ndo_validate_addr      = eth_validate_addr,
2995        .ndo_vlan_rx_add_vid    = igb_vlan_rx_add_vid,
2996        .ndo_vlan_rx_kill_vid   = igb_vlan_rx_kill_vid,
2997        .ndo_set_vf_mac         = igb_ndo_set_vf_mac,
2998        .ndo_set_vf_vlan        = igb_ndo_set_vf_vlan,
2999        .ndo_set_vf_rate        = igb_ndo_set_vf_bw,
3000        .ndo_set_vf_spoofchk    = igb_ndo_set_vf_spoofchk,
3001        .ndo_set_vf_trust       = igb_ndo_set_vf_trust,
3002        .ndo_get_vf_config      = igb_ndo_get_vf_config,
3003        .ndo_fix_features       = igb_fix_features,
3004        .ndo_set_features       = igb_set_features,
3005        .ndo_fdb_add            = igb_ndo_fdb_add,
3006        .ndo_features_check     = igb_features_check,
3007        .ndo_setup_tc           = igb_setup_tc,
3008        .ndo_bpf                = igb_xdp,
3009        .ndo_xdp_xmit           = igb_xdp_xmit,
3010};
3011
3012/**
3013 * igb_set_fw_version - Configure version string for ethtool
3014 * @adapter: adapter struct
3015 **/
3016void igb_set_fw_version(struct igb_adapter *adapter)
3017{
3018        struct e1000_hw *hw = &adapter->hw;
3019        struct e1000_fw_version fw;
3020
3021        igb_get_fw_version(hw, &fw);
3022
3023        switch (hw->mac.type) {
3024        case e1000_i210:
3025        case e1000_i211:
3026                if (!(igb_get_flash_presence_i210(hw))) {
3027                        snprintf(adapter->fw_version,
3028                                 sizeof(adapter->fw_version),
3029                                 "%2d.%2d-%d",
3030                                 fw.invm_major, fw.invm_minor,
3031                                 fw.invm_img_type);
3032                        break;
3033                }
3034                fallthrough;
3035        default:
3036                /* if option is rom valid, display its version too */
3037                if (fw.or_valid) {
3038                        snprintf(adapter->fw_version,
3039                                 sizeof(adapter->fw_version),
3040                                 "%d.%d, 0x%08x, %d.%d.%d",
3041                                 fw.eep_major, fw.eep_minor, fw.etrack_id,
3042                                 fw.or_major, fw.or_build, fw.or_patch);
3043                /* no option rom */
3044                } else if (fw.etrack_id != 0X0000) {
3045                        snprintf(adapter->fw_version,
3046                            sizeof(adapter->fw_version),
3047                            "%d.%d, 0x%08x",
3048                            fw.eep_major, fw.eep_minor, fw.etrack_id);
3049                } else {
3050                snprintf(adapter->fw_version,
3051                    sizeof(adapter->fw_version),
3052                    "%d.%d.%d",
3053                    fw.eep_major, fw.eep_minor, fw.eep_build);
3054                }
3055                break;
3056        }
3057}
3058
3059/**
3060 * igb_init_mas - init Media Autosense feature if enabled in the NVM
3061 *
3062 * @adapter: adapter struct
3063 **/
3064static void igb_init_mas(struct igb_adapter *adapter)
3065{
3066        struct e1000_hw *hw = &adapter->hw;
3067        u16 eeprom_data;
3068
3069        hw->nvm.ops.read(hw, NVM_COMPAT, 1, &eeprom_data);
3070        switch (hw->bus.func) {
3071        case E1000_FUNC_0:
3072                if (eeprom_data & IGB_MAS_ENABLE_0) {
3073                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
3074                        netdev_info(adapter->netdev,
3075                                "MAS: Enabling Media Autosense for port %d\n",
3076                                hw->bus.func);
3077                }
3078                break;
3079        case E1000_FUNC_1:
3080                if (eeprom_data & IGB_MAS_ENABLE_1) {
3081                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
3082                        netdev_info(adapter->netdev,
3083                                "MAS: Enabling Media Autosense for port %d\n",
3084                                hw->bus.func);
3085                }
3086                break;
3087        case E1000_FUNC_2:
3088                if (eeprom_data & IGB_MAS_ENABLE_2) {
3089                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
3090                        netdev_info(adapter->netdev,
3091                                "MAS: Enabling Media Autosense for port %d\n",
3092                                hw->bus.func);
3093                }
3094                break;
3095        case E1000_FUNC_3:
3096                if (eeprom_data & IGB_MAS_ENABLE_3) {
3097                        adapter->flags |= IGB_FLAG_MAS_ENABLE;
3098                        netdev_info(adapter->netdev,
3099                                "MAS: Enabling Media Autosense for port %d\n",
3100                                hw->bus.func);
3101                }
3102                break;
3103        default:
3104                /* Shouldn't get here */
3105                netdev_err(adapter->netdev,
3106                        "MAS: Invalid port configuration, returning\n");
3107                break;
3108        }
3109}
3110
3111/**
3112 *  igb_init_i2c - Init I2C interface
3113 *  @adapter: pointer to adapter structure
3114 **/
3115static s32 igb_init_i2c(struct igb_adapter *adapter)
3116{
3117        struct e1000_hw *hw = &adapter->hw;
3118        s32 status = 0;
3119        s32 i2cctl;
3120
3121        /* I2C interface supported on i350 devices */
3122        if (adapter->hw.mac.type != e1000_i350)
3123                return 0;
3124
3125        i2cctl = rd32(E1000_I2CPARAMS);
3126        i2cctl |= E1000_I2CBB_EN
3127                | E1000_I2C_CLK_OUT | E1000_I2C_CLK_OE_N
3128                | E1000_I2C_DATA_OUT | E1000_I2C_DATA_OE_N;
3129        wr32(E1000_I2CPARAMS, i2cctl);
3130        wrfl();
3131
3132        /* Initialize the i2c bus which is controlled by the registers.
3133         * This bus will use the i2c_algo_bit structure that implements
3134         * the protocol through toggling of the 4 bits in the register.
3135         */
3136        adapter->i2c_adap.owner = THIS_MODULE;
3137        adapter->i2c_algo = igb_i2c_algo;
3138        adapter->i2c_algo.data = adapter;
3139        adapter->i2c_adap.algo_data = &adapter->i2c_algo;
3140        adapter->i2c_adap.dev.parent = &adapter->pdev->dev;
3141        strlcpy(adapter->i2c_adap.name, "igb BB",
3142                sizeof(adapter->i2c_adap.name));
3143        status = i2c_bit_add_bus(&adapter->i2c_adap);
3144        return status;
3145}
3146
3147/**
3148 *  igb_probe - Device Initialization Routine
3149 *  @pdev: PCI device information struct
3150 *  @ent: entry in igb_pci_tbl
3151 *
3152 *  Returns 0 on success, negative on failure
3153 *
3154 *  igb_probe initializes an adapter identified by a pci_dev structure.
3155 *  The OS initialization, configuring of the adapter private structure,
3156 *  and a hardware reset occur.
3157 **/
3158static int igb_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
3159{
3160        struct net_device *netdev;
3161        struct igb_adapter *adapter;
3162        struct e1000_hw *hw;
3163        u16 eeprom_data = 0;
3164        s32 ret_val;
3165        static int global_quad_port_a; /* global quad port a indication */
3166        const struct e1000_info *ei = igb_info_tbl[ent->driver_data];
3167        u8 part_str[E1000_PBANUM_LENGTH];
3168        int err;
3169
3170        /* Catch broken hardware that put the wrong VF device ID in
3171         * the PCIe SR-IOV capability.
3172         */
3173        if (pdev->is_virtfn) {
3174                WARN(1, KERN_ERR "%s (%x:%x) should not be a VF!\n",
3175                        pci_name(pdev), pdev->vendor, pdev->device);
3176                return -EINVAL;
3177        }
3178
3179        err = pci_enable_device_mem(pdev);
3180        if (err)
3181                return err;
3182
3183        err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
3184        if (err) {
3185                dev_err(&pdev->dev,
3186                        "No usable DMA configuration, aborting\n");
3187                goto err_dma;
3188        }
3189
3190        err = pci_request_mem_regions(pdev, igb_driver_name);
3191        if (err)
3192                goto err_pci_reg;
3193
3194        pci_enable_pcie_error_reporting(pdev);
3195
3196        pci_set_master(pdev);
3197        pci_save_state(pdev);
3198
3199        err = -ENOMEM;
3200        netdev = alloc_etherdev_mq(sizeof(struct igb_adapter),
3201                                   IGB_MAX_TX_QUEUES);
3202        if (!netdev)
3203                goto err_alloc_etherdev;
3204
3205        SET_NETDEV_DEV(netdev, &pdev->dev);
3206
3207        pci_set_drvdata(pdev, netdev);
3208        adapter = netdev_priv(netdev);
3209        adapter->netdev = netdev;
3210        adapter->pdev = pdev;
3211        hw = &adapter->hw;
3212        hw->back = adapter;
3213        adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
3214
3215        err = -EIO;
3216        adapter->io_addr = pci_iomap(pdev, 0, 0);
3217        if (!adapter->io_addr)
3218                goto err_ioremap;
3219        /* hw->hw_addr can be altered, we'll use adapter->io_addr for unmap */
3220        hw->hw_addr = adapter->io_addr;
3221
3222        netdev->netdev_ops = &igb_netdev_ops;
3223        igb_set_ethtool_ops(netdev);
3224        netdev->watchdog_timeo = 5 * HZ;
3225
3226        strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
3227
3228        netdev->mem_start = pci_resource_start(pdev, 0);
3229        netdev->mem_end = pci_resource_end(pdev, 0);
3230
3231        /* PCI config space info */
3232        hw->vendor_id = pdev->vendor;
3233        hw->device_id = pdev->device;
3234        hw->revision_id = pdev->revision;
3235        hw->subsystem_vendor_id = pdev->subsystem_vendor;
3236        hw->subsystem_device_id = pdev->subsystem_device;
3237
3238        /* Copy the default MAC, PHY and NVM function pointers */
3239        memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
3240        memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
3241        memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
3242        /* Initialize skew-specific constants */
3243        err = ei->get_invariants(hw);
3244        if (err)
3245                goto err_sw_init;
3246
3247        /* setup the private structure */
3248        err = igb_sw_init(adapter);
3249        if (err)
3250                goto err_sw_init;
3251
3252        igb_get_bus_info_pcie(hw);
3253
3254        hw->phy.autoneg_wait_to_complete = false;
3255
3256        /* Copper options */
3257        if (hw->phy.media_type == e1000_media_type_copper) {
3258                hw->phy.mdix = AUTO_ALL_MODES;
3259                hw->phy.disable_polarity_correction = false;
3260                hw->phy.ms_type = e1000_ms_hw_default;
3261        }
3262
3263        if (igb_check_reset_block(hw))
3264                dev_info(&pdev->dev,
3265                        "PHY reset is blocked due to SOL/IDER session.\n");
3266
3267        /* features is initialized to 0 in allocation, it might have bits
3268         * set by igb_sw_init so we should use an or instead of an
3269         * assignment.
3270         */
3271        netdev->features |= NETIF_F_SG |
3272                            NETIF_F_TSO |
3273                            NETIF_F_TSO6 |
3274                            NETIF_F_RXHASH |
3275                            NETIF_F_RXCSUM |
3276                            NETIF_F_HW_CSUM;
3277
3278        if (hw->mac.type >= e1000_82576)
3279                netdev->features |= NETIF_F_SCTP_CRC | NETIF_F_GSO_UDP_L4;
3280
3281        if (hw->mac.type >= e1000_i350)
3282                netdev->features |= NETIF_F_HW_TC;
3283
3284#define IGB_GSO_PARTIAL_FEATURES (NETIF_F_GSO_GRE | \
3285                                  NETIF_F_GSO_GRE_CSUM | \
3286                                  NETIF_F_GSO_IPXIP4 | \
3287                                  NETIF_F_GSO_IPXIP6 | \
3288                                  NETIF_F_GSO_UDP_TUNNEL | \
3289                                  NETIF_F_GSO_UDP_TUNNEL_CSUM)
3290
3291        netdev->gso_partial_features = IGB_GSO_PARTIAL_FEATURES;
3292        netdev->features |= NETIF_F_GSO_PARTIAL | IGB_GSO_PARTIAL_FEATURES;
3293
3294        /* copy netdev features into list of user selectable features */
3295        netdev->hw_features |= netdev->features |
3296                               NETIF_F_HW_VLAN_CTAG_RX |
3297                               NETIF_F_HW_VLAN_CTAG_TX |
3298                               NETIF_F_RXALL;
3299
3300        if (hw->mac.type >= e1000_i350)
3301                netdev->hw_features |= NETIF_F_NTUPLE;
3302
3303        netdev->features |= NETIF_F_HIGHDMA;
3304
3305        netdev->vlan_features |= netdev->features | NETIF_F_TSO_MANGLEID;
3306        netdev->mpls_features |= NETIF_F_HW_CSUM;
3307        netdev->hw_enc_features |= netdev->vlan_features;
3308
3309        /* set this bit last since it cannot be part of vlan_features */
3310        netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER |
3311                            NETIF_F_HW_VLAN_CTAG_RX |
3312                            NETIF_F_HW_VLAN_CTAG_TX;
3313
3314        netdev->priv_flags |= IFF_SUPP_NOFCS;
3315
3316        netdev->priv_flags |= IFF_UNICAST_FLT;
3317
3318        /* MTU range: 68 - 9216 */
3319        netdev->min_mtu = ETH_MIN_MTU;
3320        netdev->max_mtu = MAX_STD_JUMBO_FRAME_SIZE;
3321
3322        adapter->en_mng_pt = igb_enable_mng_pass_thru(hw);
3323
3324        /* before reading the NVM, reset the controller to put the device in a
3325         * known good starting state
3326         */
3327        hw->mac.ops.reset_hw(hw);
3328
3329        /* make sure the NVM is good , i211/i210 parts can have special NVM
3330         * that doesn't contain a checksum
3331         */
3332        switch (hw->mac.type) {
3333        case e1000_i210:
3334        case e1000_i211:
3335                if (igb_get_flash_presence_i210(hw)) {
3336                        if (hw->nvm.ops.validate(hw) < 0) {
3337                                dev_err(&pdev->dev,
3338                                        "The NVM Checksum Is Not Valid\n");
3339                                err = -EIO;
3340                                goto err_eeprom;
3341                        }
3342                }
3343                break;
3344        default:
3345                if (hw->nvm.ops.validate(hw) < 0) {
3346                        dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
3347                        err = -EIO;
3348                        goto err_eeprom;
3349                }
3350                break;
3351        }
3352
3353        if (eth_platform_get_mac_address(&pdev->dev, hw->mac.addr)) {
3354                /* copy the MAC address out of the NVM */
3355                if (hw->mac.ops.read_mac_addr(hw))
3356                        dev_err(&pdev->dev, "NVM Read Error\n");
3357        }
3358
3359        eth_hw_addr_set(netdev, hw->mac.addr);
3360
3361        if (!is_valid_ether_addr(netdev->dev_addr)) {
3362                dev_err(&pdev->dev, "Invalid MAC Address\n");
3363                err = -EIO;
3364                goto err_eeprom;
3365        }
3366
3367        igb_set_default_mac_filter(adapter);
3368
3369        /* get firmware version for ethtool -i */
3370        igb_set_fw_version(adapter);
3371
3372        /* configure RXPBSIZE and TXPBSIZE */
3373        if (hw->mac.type == e1000_i210) {
3374                wr32(E1000_RXPBS, I210_RXPBSIZE_DEFAULT);
3375                wr32(E1000_TXPBS, I210_TXPBSIZE_DEFAULT);
3376        }
3377
3378        timer_setup(&adapter->watchdog_timer, igb_watchdog, 0);
3379        timer_setup(&adapter->phy_info_timer, igb_update_phy_info, 0);
3380
3381        INIT_WORK(&adapter->reset_task, igb_reset_task);
3382        INIT_WORK(&adapter->watchdog_task, igb_watchdog_task);
3383
3384        /* Initialize link properties that are user-changeable */
3385        adapter->fc_autoneg = true;
3386        hw->mac.autoneg = true;
3387        hw->phy.autoneg_advertised = 0x2f;
3388
3389        hw->fc.requested_mode = e1000_fc_default;
3390        hw->fc.current_mode = e1000_fc_default;
3391
3392        igb_validate_mdi_setting(hw);
3393
3394        /* By default, support wake on port A */
3395        if (hw->bus.func == 0)
3396                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3397
3398        /* Check the NVM for wake support on non-port A ports */
3399        if (hw->mac.type >= e1000_82580)
3400                hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_A +
3401                                 NVM_82580_LAN_FUNC_OFFSET(hw->bus.func), 1,
3402                                 &eeprom_data);
3403        else if (hw->bus.func == 1)
3404                hw->nvm.ops.read(hw, NVM_INIT_CONTROL3_PORT_B, 1, &eeprom_data);
3405
3406        if (eeprom_data & IGB_EEPROM_APME)
3407                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3408
3409        /* now that we have the eeprom settings, apply the special cases where
3410         * the eeprom may be wrong or the board simply won't support wake on
3411         * lan on a particular port
3412         */
3413        switch (pdev->device) {
3414        case E1000_DEV_ID_82575GB_QUAD_COPPER:
3415                adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3416                break;
3417        case E1000_DEV_ID_82575EB_FIBER_SERDES:
3418        case E1000_DEV_ID_82576_FIBER:
3419        case E1000_DEV_ID_82576_SERDES:
3420                /* Wake events only supported on port A for dual fiber
3421                 * regardless of eeprom setting
3422                 */
3423                if (rd32(E1000_STATUS) & E1000_STATUS_FUNC_1)
3424                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3425                break;
3426        case E1000_DEV_ID_82576_QUAD_COPPER:
3427        case E1000_DEV_ID_82576_QUAD_COPPER_ET2:
3428                /* if quad port adapter, disable WoL on all but port A */
3429                if (global_quad_port_a != 0)
3430                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3431                else
3432                        adapter->flags |= IGB_FLAG_QUAD_PORT_A;
3433                /* Reset for multiple quad port adapters */
3434                if (++global_quad_port_a == 4)
3435                        global_quad_port_a = 0;
3436                break;
3437        default:
3438                /* If the device can't wake, don't set software support */
3439                if (!device_can_wakeup(&adapter->pdev->dev))
3440                        adapter->flags &= ~IGB_FLAG_WOL_SUPPORTED;
3441        }
3442
3443        /* initialize the wol settings based on the eeprom settings */
3444        if (adapter->flags & IGB_FLAG_WOL_SUPPORTED)
3445                adapter->wol |= E1000_WUFC_MAG;
3446
3447        /* Some vendors want WoL disabled by default, but still supported */
3448        if ((hw->mac.type == e1000_i350) &&
3449            (pdev->subsystem_vendor == PCI_VENDOR_ID_HP)) {
3450                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3451                adapter->wol = 0;
3452        }
3453
3454        /* Some vendors want the ability to Use the EEPROM setting as
3455         * enable/disable only, and not for capability
3456         */
3457        if (((hw->mac.type == e1000_i350) ||
3458             (hw->mac.type == e1000_i354)) &&
3459            (pdev->subsystem_vendor == PCI_VENDOR_ID_DELL)) {
3460                adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3461                adapter->wol = 0;
3462        }
3463        if (hw->mac.type == e1000_i350) {
3464                if (((pdev->subsystem_device == 0x5001) ||
3465                     (pdev->subsystem_device == 0x5002)) &&
3466                                (hw->bus.func == 0)) {
3467                        adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3468                        adapter->wol = 0;
3469                }
3470                if (pdev->subsystem_device == 0x1F52)
3471                        adapter->flags |= IGB_FLAG_WOL_SUPPORTED;
3472        }
3473
3474        device_set_wakeup_enable(&adapter->pdev->dev,
3475                                 adapter->flags & IGB_FLAG_WOL_SUPPORTED);
3476
3477        /* reset the hardware with the new settings */
3478        igb_reset(adapter);
3479
3480        /* Init the I2C interface */
3481        err = igb_init_i2c(adapter);
3482        if (err) {
3483                dev_err(&pdev->dev, "failed to init i2c interface\n");
3484                goto err_eeprom;
3485        }
3486
3487        /* let the f/w know that the h/w is now under the control of the
3488         * driver.
3489         */
3490        igb_get_hw_control(adapter);
3491
3492        strcpy(netdev->name, "eth%d");
3493        err = register_netdev(netdev);
3494        if (err)
3495                goto err_register;
3496
3497        /* carrier off reporting is important to ethtool even BEFORE open */
3498        netif_carrier_off(netdev);
3499
3500#ifdef CONFIG_IGB_DCA
3501        if (dca_add_requester(&pdev->dev) == 0) {
3502                adapter->flags |= IGB_FLAG_DCA_ENABLED;
3503                dev_info(&pdev->dev, "DCA enabled\n");
3504                igb_setup_dca(adapter);
3505        }
3506
3507#endif
3508#ifdef CONFIG_IGB_HWMON
3509        /* Initialize the thermal sensor on i350 devices. */
3510        if (hw->mac.type == e1000_i350 && hw->bus.func == 0) {
3511                u16 ets_word;
3512
3513                /* Read the NVM to determine if this i350 device supports an
3514                 * external thermal sensor.
3515                 */
3516                hw->nvm.ops.read(hw, NVM_ETS_CFG, 1, &ets_word);
3517                if (ets_word != 0x0000 && ets_word != 0xFFFF)
3518                        adapter->ets = true;
3519                else
3520                        adapter->ets = false;
3521                if (igb_sysfs_init(adapter))
3522                        dev_err(&pdev->dev,
3523                                "failed to allocate sysfs resources\n");
3524        } else {
3525                adapter->ets = false;
3526        }
3527#endif
3528        /* Check if Media Autosense is enabled */
3529        adapter->ei = *ei;
3530        if (hw->dev_spec._82575.mas_capable)
3531                igb_init_mas(adapter);
3532
3533        /* do hw tstamp init after resetting */
3534        igb_ptp_init(adapter);
3535
3536        dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
3537        /* print bus type/speed/width info, not applicable to i354 */
3538        if (hw->mac.type != e1000_i354) {
3539                dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
3540                         netdev->name,
3541                         ((hw->bus.speed == e1000_bus_speed_2500) ? "2.5Gb/s" :
3542                          (hw->bus.speed == e1000_bus_speed_5000) ? "5.0Gb/s" :
3543                           "unknown"),
3544                         ((hw->bus.width == e1000_bus_width_pcie_x4) ?
3545                          "Width x4" :
3546                          (hw->bus.width == e1000_bus_width_pcie_x2) ?
3547                          "Width x2" :
3548                          (hw->bus.width == e1000_bus_width_pcie_x1) ?
3549                          "Width x1" : "unknown"), netdev->dev_addr);
3550        }
3551
3552        if ((hw->mac.type == e1000_82576 &&
3553             rd32(E1000_EECD) & E1000_EECD_PRES) ||
3554            (hw->mac.type >= e1000_i210 ||
3555             igb_get_flash_presence_i210(hw))) {
3556                ret_val = igb_read_part_string(hw, part_str,
3557                                               E1000_PBANUM_LENGTH);
3558        } else {
3559                ret_val = -E1000_ERR_INVM_VALUE_NOT_FOUND;
3560        }
3561
3562        if (ret_val)
3563                strcpy(part_str, "Unknown");
3564        dev_info(&pdev->dev, "%s: PBA No: %s\n", netdev->name, part_str);
3565        dev_info(&pdev->dev,
3566                "Using %s interrupts. %d rx queue(s), %d tx queue(s)\n",
3567                (adapter->flags & IGB_FLAG_HAS_MSIX) ? "MSI-X" :
3568                (adapter->flags & IGB_FLAG_HAS_MSI) ? "MSI" : "legacy",
3569                adapter->num_rx_queues, adapter->num_tx_queues);
3570        if (hw->phy.media_type == e1000_media_type_copper) {
3571                switch (hw->mac.type) {
3572                case e1000_i350:
3573                case e1000_i210:
3574                case e1000_i211:
3575                        /* Enable EEE for internal copper PHY devices */
3576                        err = igb_set_eee_i350(hw, true, true);
3577                        if ((!err) &&
3578                            (!hw->dev_spec._82575.eee_disable)) {
3579                                adapter->eee_advert =
3580                                        MDIO_EEE_100TX | MDIO_EEE_1000T;
3581                                adapter->flags |= IGB_FLAG_EEE;
3582                        }
3583                        break;
3584                case e1000_i354:
3585                        if ((rd32(E1000_CTRL_EXT) &
3586                            E1000_CTRL_EXT_LINK_MODE_SGMII)) {
3587                                err = igb_set_eee_i354(hw, true, true);
3588                                if ((!err) &&
3589                                        (!hw->dev_spec._82575.eee_disable)) {
3590                                        adapter->eee_advert =
3591                                           MDIO_EEE_100TX | MDIO_EEE_1000T;
3592                                        adapter->flags |= IGB_FLAG_EEE;
3593                                }
3594                        }
3595                        break;
3596                default:
3597                        break;
3598                }
3599        }
3600
3601        dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_NO_DIRECT_COMPLETE);
3602
3603        pm_runtime_put_noidle(&pdev->dev);
3604        return 0;
3605
3606err_register:
3607        igb_release_hw_control(adapter);
3608        memset(&adapter->i2c_adap, 0, sizeof(adapter->i2c_adap));
3609err_eeprom:
3610        if (!igb_check_reset_block(hw))
3611                igb_reset_phy(hw);
3612
3613        if (hw->flash_address)
3614                iounmap(hw->flash_address);
3615err_sw_init:
3616        kfree(adapter->mac_table);
3617        kfree(adapter->shadow_vfta);
3618        igb_clear_interrupt_scheme(adapter);
3619#ifdef CONFIG_PCI_IOV
3620        igb_disable_sriov(pdev);
3621#endif
3622        pci_iounmap(pdev, adapter->io_addr);
3623err_ioremap:
3624        free_netdev(netdev);
3625err_alloc_etherdev:
3626        pci_disable_pcie_error_reporting(pdev);
3627        pci_release_mem_regions(pdev);
3628err_pci_reg:
3629err_dma:
3630        pci_disable_device(pdev);
3631        return err;
3632}
3633
3634#ifdef CONFIG_PCI_IOV
3635static int igb_disable_sriov(struct pci_dev *pdev)
3636{
3637        struct net_device *netdev = pci_get_drvdata(pdev);
3638        struct igb_adapter *adapter = netdev_priv(netdev);
3639        struct e1000_hw *hw = &adapter->hw;
3640
3641        /* reclaim resources allocated to VFs */
3642        if (adapter->vf_data) {
3643                /* disable iov and allow time for transactions to clear */
3644                if (pci_vfs_assigned(pdev)) {
3645                        dev_warn(&pdev->dev,
3646                                 "Cannot deallocate SR-IOV virtual functions while they are assigned - VFs will not be deallocated\n");
3647                        return -EPERM;
3648                } else {
3649                        pci_disable_sriov(pdev);
3650                        msleep(500);
3651                }
3652
3653                kfree(adapter->vf_mac_list);
3654                adapter->vf_mac_list = NULL;
3655                kfree(adapter->vf_data);
3656                adapter->vf_data = NULL;
3657                adapter->vfs_allocated_count = 0;
3658                wr32(E1000_IOVCTL, E1000_IOVCTL_REUSE_VFQ);
3659                wrfl();
3660                msleep(100);
3661                dev_info(&pdev->dev, "IOV Disabled\n");
3662
3663                /* Re-enable DMA Coalescing flag since IOV is turned off */
3664                adapter->flags |= IGB_FLAG_DMAC;
3665        }
3666
3667        return 0;
3668}
3669
3670static int igb_enable_sriov(struct pci_dev *pdev, int num_vfs)
3671{
3672        struct net_device *netdev = pci_get_drvdata(pdev);
3673        struct igb_adapter *adapter = netdev_priv(netdev);
3674        int old_vfs = pci_num_vf(pdev);
3675        struct vf_mac_filter *mac_list;
3676        int err = 0;
3677        int num_vf_mac_filters, i;
3678
3679        if (!(adapter->flags & IGB_FLAG_HAS_MSIX) || num_vfs > 7) {
3680                err = -EPERM;
3681                goto out;
3682        }
3683        if (!num_vfs)
3684                goto out;
3685
3686        if (old_vfs) {
3687                dev_info(&pdev->dev, "%d pre-allocated VFs found - override max_vfs setting of %d\n",
3688                         old_vfs, max_vfs);
3689                adapter->vfs_allocated_count = old_vfs;
3690        } else
3691                adapter->vfs_allocated_count = num_vfs;
3692
3693        adapter->vf_data = kcalloc(adapter->vfs_allocated_count,
3694                                sizeof(struct vf_data_storage), GFP_KERNEL);
3695
3696        /* if allocation failed then we do not support SR-IOV */
3697        if (!adapter->vf_data) {
3698                adapter->vfs_allocated_count = 0;
3699                err = -ENOMEM;
3700                goto out;
3701        }
3702
3703        /* Due to the limited number of RAR entries calculate potential
3704         * number of MAC filters available for the VFs. Reserve entries
3705         * for PF default MAC, PF MAC filters and at least one RAR entry
3706         * for each VF for VF MAC.
3707         */
3708        num_vf_mac_filters = adapter->hw.mac.rar_entry_count -
3709                             (1 + IGB_PF_MAC_FILTERS_RESERVED +
3710                              adapter->vfs_allocated_count);
3711
3712        adapter->vf_mac_list = kcalloc(num_vf_mac_filters,
3713                                       sizeof(struct vf_mac_filter),
3714                                       GFP_KERNEL);
3715
3716        mac_list = adapter->vf_mac_list;
3717        INIT_LIST_HEAD(&adapter->vf_macs.l);
3718
3719        if (adapter->vf_mac_list) {
3720                /* Initialize list of VF MAC filters */
3721                for (i = 0; i < num_vf_mac_filters; i++) {
3722                        mac_list->vf = -1;
3723                        mac_list->free = true;
3724                        list_add(&mac_list->l, &adapter->vf_macs.l);
3725                        mac_list++;
3726                }
3727        } else {
3728                /* If we could not allocate memory for the VF MAC filters
3729                 * we can continue without this feature but warn user.
3730                 */
3731                dev_err(&pdev->dev,
3732                        "Unable to allocate memory for VF MAC filter list\n");
3733        }
3734
3735        /* only call pci_enable_sriov() if no VFs are allocated already */
3736        if (!old_vfs) {
3737                err = pci_enable_sriov(pdev, adapter->vfs_allocated_count);
3738                if (err)
3739                        goto err_out;
3740        }
3741        dev_info(&pdev->dev, "%d VFs allocated\n",
3742                 adapter->vfs_allocated_count);
3743        for (i = 0; i < adapter->vfs_allocated_count; i++)
3744                igb_vf_configure(adapter, i);
3745
3746        /* DMA Coalescing is not supported in IOV mode. */
3747        adapter->flags &= ~IGB_FLAG_DMAC;
3748        goto out;
3749
3750err_out:
3751        kfree(adapter->vf_mac_list);
3752        adapter->vf_mac_list = NULL;
3753        kfree(adapter->vf_data);
3754        adapter->vf_data = NULL;
3755        adapter->vfs_allocated_count = 0;
3756out:
3757        return err;
3758}
3759
3760#endif
3761/**
3762 *  igb_remove_i2c - Cleanup  I2C interface
3763 *  @adapter: pointer to adapter structure
3764 **/
3765static void igb_remove_i2c(struct igb_adapter *adapter)
3766{
3767        /* free the adapter bus structure */
3768        i2c_del_adapter(&adapter->i2c_adap);
3769}
3770
3771/**
3772 *  igb_remove - Device Removal Routine
3773 *  @pdev: PCI device information struct
3774 *
3775 *  igb_remove is called by the PCI subsystem to alert the driver
3776 *  that it should release a PCI device.  The could be caused by a
3777 *  Hot-Plug event, or because the driver is going to be removed from
3778 *  memory.
3779 **/
3780static void igb_remove(struct pci_dev *pdev)
3781{
3782        struct net_device *netdev = pci_get_drvdata(pdev);
3783        struct igb_adapter *adapter = netdev_priv(netdev);
3784        struct e1000_hw *hw = &adapter->hw;
3785
3786        pm_runtime_get_noresume(&pdev->dev);
3787#ifdef CONFIG_IGB_HWMON
3788        igb_sysfs_exit(adapter);
3789#endif
3790        igb_remove_i2c(adapter);
3791        igb_ptp_stop(adapter);
3792        /* The watchdog timer may be rescheduled, so explicitly
3793         * disable watchdog from being rescheduled.
3794         */
3795        set_bit(__IGB_DOWN, &adapter->state);
3796        del_timer_sync(&adapter->watchdog_timer);
3797        del_timer_sync(&adapter->phy_info_timer);
3798
3799        cancel_work_sync(&adapter->reset_task);
3800        cancel_work_sync(&adapter->watchdog_task);
3801
3802#ifdef CONFIG_IGB_DCA
3803        if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
3804                dev_info(&pdev->dev, "DCA disabled\n");
3805                dca_remove_requester(&pdev->dev);
3806                adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
3807                wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
3808        }
3809#endif
3810
3811        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
3812         * would have already happened in close and is redundant.
3813         */
3814        igb_release_hw_control(adapter);
3815
3816#ifdef CONFIG_PCI_IOV
3817        igb_disable_sriov(pdev);
3818#endif
3819
3820        unregister_netdev(netdev);
3821
3822        igb_clear_interrupt_scheme(adapter);
3823
3824        pci_iounmap(pdev, adapter->io_addr);
3825        if (hw->flash_address)
3826                iounmap(hw->flash_address);
3827        pci_release_mem_regions(pdev);
3828
3829        kfree(adapter->mac_table);
3830        kfree(adapter->shadow_vfta);
3831        free_netdev(netdev);
3832
3833        pci_disable_pcie_error_reporting(pdev);
3834
3835        pci_disable_device(pdev);
3836}
3837
3838/**
3839 *  igb_probe_vfs - Initialize vf data storage and add VFs to pci config space
3840 *  @adapter: board private structure to initialize
3841 *
3842 *  This function initializes the vf specific data storage and then attempts to
3843 *  allocate the VFs.  The reason for ordering it this way is because it is much
3844 *  mor expensive time wise to disable SR-IOV than it is to allocate and free
3845 *  the memory for the VFs.
3846 **/
3847static void igb_probe_vfs(struct igb_adapter *adapter)
3848{
3849#ifdef CONFIG_PCI_IOV
3850        struct pci_dev *pdev = adapter->pdev;
3851        struct e1000_hw *hw = &adapter->hw;
3852
3853        /* Virtualization features not supported on i210 family. */
3854        if ((hw->mac.type == e1000_i210) || (hw->mac.type == e1000_i211))
3855                return;
3856
3857        /* Of the below we really only want the effect of getting
3858         * IGB_FLAG_HAS_MSIX set (if available), without which
3859         * igb_enable_sriov() has no effect.
3860         */
3861        igb_set_interrupt_capability(adapter, true);
3862        igb_reset_interrupt_capability(adapter);
3863
3864        pci_sriov_set_totalvfs(pdev, 7);
3865        igb_enable_sriov(pdev, max_vfs);
3866
3867#endif /* CONFIG_PCI_IOV */
3868}
3869
3870unsigned int igb_get_max_rss_queues(struct igb_adapter *adapter)
3871{
3872        struct e1000_hw *hw = &adapter->hw;
3873        unsigned int max_rss_queues;
3874
3875        /* Determine the maximum number of RSS queues supported. */
3876        switch (hw->mac.type) {
3877        case e1000_i211:
3878                max_rss_queues = IGB_MAX_RX_QUEUES_I211;
3879                break;
3880        case e1000_82575:
3881        case e1000_i210:
3882                max_rss_queues = IGB_MAX_RX_QUEUES_82575;
3883                break;
3884        case e1000_i350:
3885                /* I350 cannot do RSS and SR-IOV at the same time */
3886                if (!!adapter->vfs_allocated_count) {
3887                        max_rss_queues = 1;
3888                        break;
3889                }
3890                fallthrough;
3891        case e1000_82576:
3892                if (!!adapter->vfs_allocated_count) {
3893                        max_rss_queues = 2;
3894                        break;
3895                }
3896                fallthrough;
3897        case e1000_82580:
3898        case e1000_i354:
3899        default:
3900                max_rss_queues = IGB_MAX_RX_QUEUES;
3901                break;
3902        }
3903
3904        return max_rss_queues;
3905}
3906
3907static void igb_init_queue_configuration(struct igb_adapter *adapter)
3908{
3909        u32 max_rss_queues;
3910
3911        max_rss_queues = igb_get_max_rss_queues(adapter);
3912        adapter->rss_queues = min_t(u32, max_rss_queues, num_online_cpus());
3913
3914        igb_set_flag_queue_pairs(adapter, max_rss_queues);
3915}
3916
3917void igb_set_flag_queue_pairs(struct igb_adapter *adapter,
3918                              const u32 max_rss_queues)
3919{
3920        struct e1000_hw *hw = &adapter->hw;
3921
3922        /* Determine if we need to pair queues. */
3923        switch (hw->mac.type) {
3924        case e1000_82575:
3925        case e1000_i211:
3926                /* Device supports enough interrupts without queue pairing. */
3927                break;
3928        case e1000_82576:
3929        case e1000_82580:
3930        case e1000_i350:
3931        case e1000_i354:
3932        case e1000_i210:
3933        default:
3934                /* If rss_queues > half of max_rss_queues, pair the queues in
3935                 * order to conserve interrupts due to limited supply.
3936                 */
3937                if (adapter->rss_queues > (max_rss_queues / 2))
3938                        adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
3939                else
3940                        adapter->flags &= ~IGB_FLAG_QUEUE_PAIRS;
3941                break;
3942        }
3943}
3944
3945/**
3946 *  igb_sw_init - Initialize general software structures (struct igb_adapter)
3947 *  @adapter: board private structure to initialize
3948 *
3949 *  igb_sw_init initializes the Adapter private data structure.
3950 *  Fields are initialized based on PCI device information and
3951 *  OS network device settings (MTU size).
3952 **/
3953static int igb_sw_init(struct igb_adapter *adapter)
3954{
3955        struct e1000_hw *hw = &adapter->hw;
3956        struct net_device *netdev = adapter->netdev;
3957        struct pci_dev *pdev = adapter->pdev;
3958
3959        pci_read_config_word(pdev, PCI_COMMAND, &hw->bus.pci_cmd_word);
3960
3961        /* set default ring sizes */
3962        adapter->tx_ring_count = IGB_DEFAULT_TXD;
3963        adapter->rx_ring_count = IGB_DEFAULT_RXD;
3964
3965        /* set default ITR values */
3966        adapter->rx_itr_setting = IGB_DEFAULT_ITR;
3967        adapter->tx_itr_setting = IGB_DEFAULT_ITR;
3968
3969        /* set default work limits */
3970        adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;
3971
3972        adapter->max_frame_size = netdev->mtu + IGB_ETH_PKT_HDR_PAD;
3973        adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
3974
3975        spin_lock_init(&adapter->nfc_lock);
3976        spin_lock_init(&adapter->stats64_lock);
3977#ifdef CONFIG_PCI_IOV
3978        switch (hw->mac.type) {
3979        case e1000_82576:
3980        case e1000_i350:
3981                if (max_vfs > 7) {
3982                        dev_warn(&pdev->dev,
3983                                 "Maximum of 7 VFs per PF, using max\n");
3984                        max_vfs = adapter->vfs_allocated_count = 7;
3985                } else
3986                        adapter->vfs_allocated_count = max_vfs;
3987                if (adapter->vfs_allocated_count)
3988                        dev_warn(&pdev->dev,
3989                                 "Enabling SR-IOV VFs using the module parameter is deprecated - please use the pci sysfs interface.\n");
3990                break;
3991        default:
3992                break;
3993        }
3994#endif /* CONFIG_PCI_IOV */
3995
3996        /* Assume MSI-X interrupts, will be checked during IRQ allocation */
3997        adapter->flags |= IGB_FLAG_HAS_MSIX;
3998
3999        adapter->mac_table = kcalloc(hw->mac.rar_entry_count,
4000                                     sizeof(struct igb_mac_addr),
4001                                     GFP_KERNEL);
4002        if (!adapter->mac_table)
4003                return -ENOMEM;
4004
4005        igb_probe_vfs(adapter);
4006
4007        igb_init_queue_configuration(adapter);
4008
4009        /* Setup and initialize a copy of the hw vlan table array */
4010        adapter->shadow_vfta = kcalloc(E1000_VLAN_FILTER_TBL_SIZE, sizeof(u32),
4011                                       GFP_KERNEL);
4012        if (!adapter->shadow_vfta)
4013                return -ENOMEM;
4014
4015        /* This call may decrease the number of queues */
4016        if (igb_init_interrupt_scheme(adapter, true)) {
4017                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
4018                return -ENOMEM;
4019        }
4020
4021        /* Explicitly disable IRQ since the NIC can be in any state. */
4022        igb_irq_disable(adapter);
4023
4024        if (hw->mac.type >= e1000_i350)
4025                adapter->flags &= ~IGB_FLAG_DMAC;
4026
4027        set_bit(__IGB_DOWN, &adapter->state);
4028        return 0;
4029}
4030
4031/**
4032 *  __igb_open - Called when a network interface is made active
4033 *  @netdev: network interface device structure
4034 *  @resuming: indicates whether we are in a resume call
4035 *
4036 *  Returns 0 on success, negative value on failure
4037 *
4038 *  The open entry point is called when a network interface is made
4039 *  active by the system (IFF_UP).  At this point all resources needed
4040 *  for transmit and receive operations are allocated, the interrupt
4041 *  handler is registered with the OS, the watchdog timer is started,
4042 *  and the stack is notified that the interface is ready.
4043 **/
4044static int __igb_open(struct net_device *netdev, bool resuming)
4045{
4046        struct igb_adapter *adapter = netdev_priv(netdev);
4047        struct e1000_hw *hw = &adapter->hw;
4048        struct pci_dev *pdev = adapter->pdev;
4049        int err;
4050        int i;
4051
4052        /* disallow open during test */
4053        if (test_bit(__IGB_TESTING, &adapter->state)) {
4054                WARN_ON(resuming);
4055                return -EBUSY;
4056        }
4057
4058        if (!resuming)
4059                pm_runtime_get_sync(&pdev->dev);
4060
4061        netif_carrier_off(netdev);
4062
4063        /* allocate transmit descriptors */
4064        err = igb_setup_all_tx_resources(adapter);
4065        if (err)
4066                goto err_setup_tx;
4067
4068        /* allocate receive descriptors */
4069        err = igb_setup_all_rx_resources(adapter);
4070        if (err)
4071                goto err_setup_rx;
4072
4073        igb_power_up_link(adapter);
4074
4075        /* before we allocate an interrupt, we must be ready to handle it.
4076         * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4077         * as soon as we call pci_request_irq, so we have to setup our
4078         * clean_rx handler before we do so.
4079         */
4080        igb_configure(adapter);
4081
4082        err = igb_request_irq(adapter);
4083        if (err)
4084                goto err_req_irq;
4085
4086        /* Notify the stack of the actual queue counts. */
4087        err = netif_set_real_num_tx_queues(adapter->netdev,
4088                                           adapter->num_tx_queues);
4089        if (err)
4090                goto err_set_queues;
4091
4092        err = netif_set_real_num_rx_queues(adapter->netdev,
4093                                           adapter->num_rx_queues);
4094        if (err)
4095                goto err_set_queues;
4096
4097        /* From here on the code is the same as igb_up() */
4098        clear_bit(__IGB_DOWN, &adapter->state);
4099
4100        for (i = 0; i < adapter->num_q_vectors; i++)
4101                napi_enable(&(adapter->q_vector[i]->napi));
4102
4103        /* Clear any pending interrupts. */
4104        rd32(E1000_TSICR);
4105        rd32(E1000_ICR);
4106
4107        igb_irq_enable(adapter);
4108
4109        /* notify VFs that reset has been completed */
4110        if (adapter->vfs_allocated_count) {
4111                u32 reg_data = rd32(E1000_CTRL_EXT);
4112
4113                reg_data |= E1000_CTRL_EXT_PFRSTD;
4114                wr32(E1000_CTRL_EXT, reg_data);
4115        }
4116
4117        netif_tx_start_all_queues(netdev);
4118
4119        if (!resuming)
4120                pm_runtime_put(&pdev->dev);
4121
4122        /* start the watchdog. */
4123        hw->mac.get_link_status = 1;
4124        schedule_work(&adapter->watchdog_task);
4125
4126        return 0;
4127
4128err_set_queues:
4129        igb_free_irq(adapter);
4130err_req_irq:
4131        igb_release_hw_control(adapter);
4132        igb_power_down_link(adapter);
4133        igb_free_all_rx_resources(adapter);
4134err_setup_rx:
4135        igb_free_all_tx_resources(adapter);
4136err_setup_tx:
4137        igb_reset(adapter);
4138        if (!resuming)
4139                pm_runtime_put(&pdev->dev);
4140
4141        return err;
4142}
4143
4144int igb_open(struct net_device *netdev)
4145{
4146        return __igb_open(netdev, false);
4147}
4148
4149/**
4150 *  __igb_close - Disables a network interface
4151 *  @netdev: network interface device structure
4152 *  @suspending: indicates we are in a suspend call
4153 *
4154 *  Returns 0, this is not allowed to fail
4155 *
4156 *  The close entry point is called when an interface is de-activated
4157 *  by the OS.  The hardware is still under the driver's control, but
4158 *  needs to be disabled.  A global MAC reset is issued to stop the
4159 *  hardware, and all transmit and receive resources are freed.
4160 **/
4161static int __igb_close(struct net_device *netdev, bool suspending)
4162{
4163        struct igb_adapter *adapter = netdev_priv(netdev);
4164        struct pci_dev *pdev = adapter->pdev;
4165
4166        WARN_ON(test_bit(__IGB_RESETTING, &adapter->state));
4167
4168        if (!suspending)
4169                pm_runtime_get_sync(&pdev->dev);
4170
4171        igb_down(adapter);
4172        igb_free_irq(adapter);
4173
4174        igb_free_all_tx_resources(adapter);
4175        igb_free_all_rx_resources(adapter);
4176
4177        if (!suspending)
4178                pm_runtime_put_sync(&pdev->dev);
4179        return 0;
4180}
4181
4182int igb_close(struct net_device *netdev)
4183{
4184        if (netif_device_present(netdev) || netdev->dismantle)
4185                return __igb_close(netdev, false);
4186        return 0;
4187}
4188
4189/**
4190 *  igb_setup_tx_resources - allocate Tx resources (Descriptors)
4191 *  @tx_ring: tx descriptor ring (for a specific queue) to setup
4192 *
4193 *  Return 0 on success, negative on failure
4194 **/
4195int igb_setup_tx_resources(struct igb_ring *tx_ring)
4196{
4197        struct device *dev = tx_ring->dev;
4198        int size;
4199
4200        size = sizeof(struct igb_tx_buffer) * tx_ring->count;
4201
4202        tx_ring->tx_buffer_info = vmalloc(size);
4203        if (!tx_ring->tx_buffer_info)
4204                goto err;
4205
4206        /* round up to nearest 4K */
4207        tx_ring->size = tx_ring->count * sizeof(union e1000_adv_tx_desc);
4208        tx_ring->size = ALIGN(tx_ring->size, 4096);
4209
4210        tx_ring->desc = dma_alloc_coherent(dev, tx_ring->size,
4211                                           &tx_ring->dma, GFP_KERNEL);
4212        if (!tx_ring->desc)
4213                goto err;
4214
4215        tx_ring->next_to_use = 0;
4216        tx_ring->next_to_clean = 0;
4217
4218        return 0;
4219
4220err:
4221        vfree(tx_ring->tx_buffer_info);
4222        tx_ring->tx_buffer_info = NULL;
4223        dev_err(dev, "Unable to allocate memory for the Tx descriptor ring\n");
4224        return -ENOMEM;
4225}
4226
4227/**
4228 *  igb_setup_all_tx_resources - wrapper to allocate Tx resources
4229 *                               (Descriptors) for all queues
4230 *  @adapter: board private structure
4231 *
4232 *  Return 0 on success, negative on failure
4233 **/
4234static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
4235{
4236        struct pci_dev *pdev = adapter->pdev;
4237        int i, err = 0;
4238
4239        for (i = 0; i < adapter->num_tx_queues; i++) {
4240                err = igb_setup_tx_resources(adapter->tx_ring[i]);
4241                if (err) {
4242                        dev_err(&pdev->dev,
4243                                "Allocation for Tx Queue %u failed\n", i);
4244                        for (i--; i >= 0; i--)
4245                                igb_free_tx_resources(adapter->tx_ring[i]);
4246                        break;
4247                }
4248        }
4249
4250        return err;
4251}
4252
4253/**
4254 *  igb_setup_tctl - configure the transmit control registers
4255 *  @adapter: Board private structure
4256 **/
4257void igb_setup_tctl(struct igb_adapter *adapter)
4258{
4259        struct e1000_hw *hw = &adapter->hw;
4260        u32 tctl;
4261
4262        /* disable queue 0 which is enabled by default on 82575 and 82576 */
4263        wr32(E1000_TXDCTL(0), 0);
4264
4265        /* Program the Transmit Control Register */
4266        tctl = rd32(E1000_TCTL);
4267        tctl &= ~E1000_TCTL_CT;
4268        tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
4269                (E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
4270
4271        igb_config_collision_dist(hw);
4272
4273        /* Enable transmits */
4274        tctl |= E1000_TCTL_EN;
4275
4276        wr32(E1000_TCTL, tctl);
4277}
4278
4279/**
4280 *  igb_configure_tx_ring - Configure transmit ring after Reset
4281 *  @adapter: board private structure
4282 *  @ring: tx ring to configure
4283 *
4284 *  Configure a transmit ring after a reset.
4285 **/
4286void igb_configure_tx_ring(struct igb_adapter *adapter,
4287                           struct igb_ring *ring)
4288{
4289        struct e1000_hw *hw = &adapter->hw;
4290        u32 txdctl = 0;
4291        u64 tdba = ring->dma;
4292        int reg_idx = ring->reg_idx;
4293
4294        wr32(E1000_TDLEN(reg_idx),
4295             ring->count * sizeof(union e1000_adv_tx_desc));
4296        wr32(E1000_TDBAL(reg_idx),
4297             tdba & 0x00000000ffffffffULL);
4298        wr32(E1000_TDBAH(reg_idx), tdba >> 32);
4299
4300        ring->tail = adapter->io_addr + E1000_TDT(reg_idx);
4301        wr32(E1000_TDH(reg_idx), 0);
4302        writel(0, ring->tail);
4303
4304        txdctl |= IGB_TX_PTHRESH;
4305        txdctl |= IGB_TX_HTHRESH << 8;
4306        txdctl |= IGB_TX_WTHRESH << 16;
4307
4308        /* reinitialize tx_buffer_info */
4309        memset(ring->tx_buffer_info, 0,
4310               sizeof(struct igb_tx_buffer) * ring->count);
4311
4312        txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
4313        wr32(E1000_TXDCTL(reg_idx), txdctl);
4314}
4315
4316/**
4317 *  igb_configure_tx - Configure transmit Unit after Reset
4318 *  @adapter: board private structure
4319 *
4320 *  Configure the Tx unit of the MAC after a reset.
4321 **/
4322static void igb_configure_tx(struct igb_adapter *adapter)
4323{
4324        struct e1000_hw *hw = &adapter->hw;
4325        int i;
4326
4327        /* disable the queues */
4328        for (i = 0; i < adapter->num_tx_queues; i++)
4329                wr32(E1000_TXDCTL(adapter->tx_ring[i]->reg_idx), 0);
4330
4331        wrfl();
4332        usleep_range(10000, 20000);
4333
4334        for (i = 0; i < adapter->num_tx_queues; i++)
4335                igb_configure_tx_ring(adapter, adapter->tx_ring[i]);
4336}
4337
4338/**
4339 *  igb_setup_rx_resources - allocate Rx resources (Descriptors)
4340 *  @rx_ring: Rx descriptor ring (for a specific queue) to setup
4341 *
4342 *  Returns 0 on success, negative on failure
4343 **/
4344int igb_setup_rx_resources(struct igb_ring *rx_ring)
4345{
4346        struct igb_adapter *adapter = netdev_priv(rx_ring->netdev);
4347        struct device *dev = rx_ring->dev;
4348        int size, res;
4349
4350        /* XDP RX-queue info */
4351        if (xdp_rxq_info_is_reg(&rx_ring->xdp_rxq))
4352                xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4353        res = xdp_rxq_info_reg(&rx_ring->xdp_rxq, rx_ring->netdev,
4354                               rx_ring->queue_index, 0);
4355        if (res < 0) {
4356                dev_err(dev, "Failed to register xdp_rxq index %u\n",
4357                        rx_ring->queue_index);
4358                return res;
4359        }
4360
4361        size = sizeof(struct igb_rx_buffer) * rx_ring->count;
4362
4363        rx_ring->rx_buffer_info = vmalloc(size);
4364        if (!rx_ring->rx_buffer_info)
4365                goto err;
4366
4367        /* Round up to nearest 4K */
4368        rx_ring->size = rx_ring->count * sizeof(union e1000_adv_rx_desc);
4369        rx_ring->size = ALIGN(rx_ring->size, 4096);
4370
4371        rx_ring->desc = dma_alloc_coherent(dev, rx_ring->size,
4372                                           &rx_ring->dma, GFP_KERNEL);
4373        if (!rx_ring->desc)
4374                goto err;
4375
4376        rx_ring->next_to_alloc = 0;
4377        rx_ring->next_to_clean = 0;
4378        rx_ring->next_to_use = 0;
4379
4380        rx_ring->xdp_prog = adapter->xdp_prog;
4381
4382        return 0;
4383
4384err:
4385        xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4386        vfree(rx_ring->rx_buffer_info);
4387        rx_ring->rx_buffer_info = NULL;
4388        dev_err(dev, "Unable to allocate memory for the Rx descriptor ring\n");
4389        return -ENOMEM;
4390}
4391
4392/**
4393 *  igb_setup_all_rx_resources - wrapper to allocate Rx resources
4394 *                               (Descriptors) for all queues
4395 *  @adapter: board private structure
4396 *
4397 *  Return 0 on success, negative on failure
4398 **/
4399static int igb_setup_all_rx_resources(struct igb_adapter *adapter)
4400{
4401        struct pci_dev *pdev = adapter->pdev;
4402        int i, err = 0;
4403
4404        for (i = 0; i < adapter->num_rx_queues; i++) {
4405                err = igb_setup_rx_resources(adapter->rx_ring[i]);
4406                if (err) {
4407                        dev_err(&pdev->dev,
4408                                "Allocation for Rx Queue %u failed\n", i);
4409                        for (i--; i >= 0; i--)
4410                                igb_free_rx_resources(adapter->rx_ring[i]);
4411                        break;
4412                }
4413        }
4414
4415        return err;
4416}
4417
4418/**
4419 *  igb_setup_mrqc - configure the multiple receive queue control registers
4420 *  @adapter: Board private structure
4421 **/
4422static void igb_setup_mrqc(struct igb_adapter *adapter)
4423{
4424        struct e1000_hw *hw = &adapter->hw;
4425        u32 mrqc, rxcsum;
4426        u32 j, num_rx_queues;
4427        u32 rss_key[10];
4428
4429        netdev_rss_key_fill(rss_key, sizeof(rss_key));
4430        for (j = 0; j < 10; j++)
4431                wr32(E1000_RSSRK(j), rss_key[j]);
4432
4433        num_rx_queues = adapter->rss_queues;
4434
4435        switch (hw->mac.type) {
4436        case e1000_82576:
4437                /* 82576 supports 2 RSS queues for SR-IOV */
4438                if (adapter->vfs_allocated_count)
4439                        num_rx_queues = 2;
4440                break;
4441        default:
4442                break;
4443        }
4444
4445        if (adapter->rss_indir_tbl_init != num_rx_queues) {
4446                for (j = 0; j < IGB_RETA_SIZE; j++)
4447                        adapter->rss_indir_tbl[j] =
4448                        (j * num_rx_queues) / IGB_RETA_SIZE;
4449                adapter->rss_indir_tbl_init = num_rx_queues;
4450        }
4451        igb_write_rss_indir_tbl(adapter);
4452
4453        /* Disable raw packet checksumming so that RSS hash is placed in
4454         * descriptor on writeback.  No need to enable TCP/UDP/IP checksum
4455         * offloads as they are enabled by default
4456         */
4457        rxcsum = rd32(E1000_RXCSUM);
4458        rxcsum |= E1000_RXCSUM_PCSD;
4459
4460        if (adapter->hw.mac.type >= e1000_82576)
4461                /* Enable Receive Checksum Offload for SCTP */
4462                rxcsum |= E1000_RXCSUM_CRCOFL;
4463
4464        /* Don't need to set TUOFL or IPOFL, they default to 1 */
4465        wr32(E1000_RXCSUM, rxcsum);
4466
4467        /* Generate RSS hash based on packet types, TCP/UDP
4468         * port numbers and/or IPv4/v6 src and dst addresses
4469         */
4470        mrqc = E1000_MRQC_RSS_FIELD_IPV4 |
4471               E1000_MRQC_RSS_FIELD_IPV4_TCP |
4472               E1000_MRQC_RSS_FIELD_IPV6 |
4473               E1000_MRQC_RSS_FIELD_IPV6_TCP |
4474               E1000_MRQC_RSS_FIELD_IPV6_TCP_EX;
4475
4476        if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV4_UDP)
4477                mrqc |= E1000_MRQC_RSS_FIELD_IPV4_UDP;
4478        if (adapter->flags & IGB_FLAG_RSS_FIELD_IPV6_UDP)
4479                mrqc |= E1000_MRQC_RSS_FIELD_IPV6_UDP;
4480
4481        /* If VMDq is enabled then we set the appropriate mode for that, else
4482         * we default to RSS so that an RSS hash is calculated per packet even
4483         * if we are only using one queue
4484         */
4485        if (adapter->vfs_allocated_count) {
4486                if (hw->mac.type > e1000_82575) {
4487                        /* Set the default pool for the PF's first queue */
4488                        u32 vtctl = rd32(E1000_VT_CTL);
4489
4490                        vtctl &= ~(E1000_VT_CTL_DEFAULT_POOL_MASK |
4491                                   E1000_VT_CTL_DISABLE_DEF_POOL);
4492                        vtctl |= adapter->vfs_allocated_count <<
4493                                E1000_VT_CTL_DEFAULT_POOL_SHIFT;
4494                        wr32(E1000_VT_CTL, vtctl);
4495                }
4496                if (adapter->rss_queues > 1)
4497                        mrqc |= E1000_MRQC_ENABLE_VMDQ_RSS_MQ;
4498                else
4499                        mrqc |= E1000_MRQC_ENABLE_VMDQ;
4500        } else {
4501                mrqc |= E1000_MRQC_ENABLE_RSS_MQ;
4502        }
4503        igb_vmm_control(adapter);
4504
4505        wr32(E1000_MRQC, mrqc);
4506}
4507
4508/**
4509 *  igb_setup_rctl - configure the receive control registers
4510 *  @adapter: Board private structure
4511 **/
4512void igb_setup_rctl(struct igb_adapter *adapter)
4513{
4514        struct e1000_hw *hw = &adapter->hw;
4515        u32 rctl;
4516
4517        rctl = rd32(E1000_RCTL);
4518
4519        rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
4520        rctl &= ~(E1000_RCTL_LBM_TCVR | E1000_RCTL_LBM_MAC);
4521
4522        rctl |= E1000_RCTL_EN | E1000_RCTL_BAM | E1000_RCTL_RDMTS_HALF |
4523                (hw->mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
4524
4525        /* enable stripping of CRC. It's unlikely this will break BMC
4526         * redirection as it did with e1000. Newer features require
4527         * that the HW strips the CRC.
4528         */
4529        rctl |= E1000_RCTL_SECRC;
4530
4531        /* disable store bad packets and clear size bits. */
4532        rctl &= ~(E1000_RCTL_SBP | E1000_RCTL_SZ_256);
4533
4534        /* enable LPE to allow for reception of jumbo frames */
4535        rctl |= E1000_RCTL_LPE;
4536
4537        /* disable queue 0 to prevent tail write w/o re-config */
4538        wr32(E1000_RXDCTL(0), 0);
4539
4540        /* Attention!!!  For SR-IOV PF driver operations you must enable
4541         * queue drop for all VF and PF queues to prevent head of line blocking
4542         * if an un-trusted VF does not provide descriptors to hardware.
4543         */
4544        if (adapter->vfs_allocated_count) {
4545                /* set all queue drop enable bits */
4546                wr32(E1000_QDE, ALL_QUEUES);
4547        }
4548
4549        /* This is useful for sniffing bad packets. */
4550        if (adapter->netdev->features & NETIF_F_RXALL) {
4551                /* UPE and MPE will be handled by normal PROMISC logic
4552                 * in e1000e_set_rx_mode
4553                 */
4554                rctl |= (E1000_RCTL_SBP | /* Receive bad packets */
4555                         E1000_RCTL_BAM | /* RX All Bcast Pkts */
4556                         E1000_RCTL_PMCF); /* RX All MAC Ctrl Pkts */
4557
4558                rctl &= ~(E1000_RCTL_DPF | /* Allow filtered pause */
4559                          E1000_RCTL_CFIEN); /* Dis VLAN CFIEN Filter */
4560                /* Do not mess with E1000_CTRL_VME, it affects transmit as well,
4561                 * and that breaks VLANs.
4562                 */
4563        }
4564
4565        wr32(E1000_RCTL, rctl);
4566}
4567
4568static inline int igb_set_vf_rlpml(struct igb_adapter *adapter, int size,
4569                                   int vfn)
4570{
4571        struct e1000_hw *hw = &adapter->hw;
4572        u32 vmolr;
4573
4574        if (size > MAX_JUMBO_FRAME_SIZE)
4575                size = MAX_JUMBO_FRAME_SIZE;
4576
4577        vmolr = rd32(E1000_VMOLR(vfn));
4578        vmolr &= ~E1000_VMOLR_RLPML_MASK;
4579        vmolr |= size | E1000_VMOLR_LPE;
4580        wr32(E1000_VMOLR(vfn), vmolr);
4581
4582        return 0;
4583}
4584
4585static inline void igb_set_vf_vlan_strip(struct igb_adapter *adapter,
4586                                         int vfn, bool enable)
4587{
4588        struct e1000_hw *hw = &adapter->hw;
4589        u32 val, reg;
4590
4591        if (hw->mac.type < e1000_82576)
4592                return;
4593
4594        if (hw->mac.type == e1000_i350)
4595                reg = E1000_DVMOLR(vfn);
4596        else
4597                reg = E1000_VMOLR(vfn);
4598
4599        val = rd32(reg);
4600        if (enable)
4601                val |= E1000_VMOLR_STRVLAN;
4602        else
4603                val &= ~(E1000_VMOLR_STRVLAN);
4604        wr32(reg, val);
4605}
4606
4607static inline void igb_set_vmolr(struct igb_adapter *adapter,
4608                                 int vfn, bool aupe)
4609{
4610        struct e1000_hw *hw = &adapter->hw;
4611        u32 vmolr;
4612
4613        /* This register exists only on 82576 and newer so if we are older then
4614         * we should exit and do nothing
4615         */
4616        if (hw->mac.type < e1000_82576)
4617                return;
4618
4619        vmolr = rd32(E1000_VMOLR(vfn));
4620        if (aupe)
4621                vmolr |= E1000_VMOLR_AUPE; /* Accept untagged packets */
4622        else
4623                vmolr &= ~(E1000_VMOLR_AUPE); /* Tagged packets ONLY */
4624
4625        /* clear all bits that might not be set */
4626        vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
4627
4628        if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
4629                vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
4630        /* for VMDq only allow the VFs and pool 0 to accept broadcast and
4631         * multicast packets
4632         */
4633        if (vfn <= adapter->vfs_allocated_count)
4634                vmolr |= E1000_VMOLR_BAM; /* Accept broadcast */
4635
4636        wr32(E1000_VMOLR(vfn), vmolr);
4637}
4638
4639/**
4640 *  igb_setup_srrctl - configure the split and replication receive control
4641 *                     registers
4642 *  @adapter: Board private structure
4643 *  @ring: receive ring to be configured
4644 **/
4645void igb_setup_srrctl(struct igb_adapter *adapter, struct igb_ring *ring)
4646{
4647        struct e1000_hw *hw = &adapter->hw;
4648        int reg_idx = ring->reg_idx;
4649        u32 srrctl = 0;
4650
4651        srrctl = IGB_RX_HDR_LEN << E1000_SRRCTL_BSIZEHDRSIZE_SHIFT;
4652        if (ring_uses_large_buffer(ring))
4653                srrctl |= IGB_RXBUFFER_3072 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4654        else
4655                srrctl |= IGB_RXBUFFER_2048 >> E1000_SRRCTL_BSIZEPKT_SHIFT;
4656        srrctl |= E1000_SRRCTL_DESCTYPE_ADV_ONEBUF;
4657        if (hw->mac.type >= e1000_82580)
4658                srrctl |= E1000_SRRCTL_TIMESTAMP;
4659        /* Only set Drop Enable if VFs allocated, or we are supporting multiple
4660         * queues and rx flow control is disabled
4661         */
4662        if (adapter->vfs_allocated_count ||
4663            (!(hw->fc.current_mode & e1000_fc_rx_pause) &&
4664             adapter->num_rx_queues > 1))
4665                srrctl |= E1000_SRRCTL_DROP_EN;
4666
4667        wr32(E1000_SRRCTL(reg_idx), srrctl);
4668}
4669
4670/**
4671 *  igb_configure_rx_ring - Configure a receive ring after Reset
4672 *  @adapter: board private structure
4673 *  @ring: receive ring to be configured
4674 *
4675 *  Configure the Rx unit of the MAC after a reset.
4676 **/
4677void igb_configure_rx_ring(struct igb_adapter *adapter,
4678                           struct igb_ring *ring)
4679{
4680        struct e1000_hw *hw = &adapter->hw;
4681        union e1000_adv_rx_desc *rx_desc;
4682        u64 rdba = ring->dma;
4683        int reg_idx = ring->reg_idx;
4684        u32 rxdctl = 0;
4685
4686        xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
4687        WARN_ON(xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
4688                                           MEM_TYPE_PAGE_SHARED, NULL));
4689
4690        /* disable the queue */
4691        wr32(E1000_RXDCTL(reg_idx), 0);
4692
4693        /* Set DMA base address registers */
4694        wr32(E1000_RDBAL(reg_idx),
4695             rdba & 0x00000000ffffffffULL);
4696        wr32(E1000_RDBAH(reg_idx), rdba >> 32);
4697        wr32(E1000_RDLEN(reg_idx),
4698             ring->count * sizeof(union e1000_adv_rx_desc));
4699
4700        /* initialize head and tail */
4701        ring->tail = adapter->io_addr + E1000_RDT(reg_idx);
4702        wr32(E1000_RDH(reg_idx), 0);
4703        writel(0, ring->tail);
4704
4705        /* set descriptor configuration */
4706        igb_setup_srrctl(adapter, ring);
4707
4708        /* set filtering for VMDQ pools */
4709        igb_set_vmolr(adapter, reg_idx & 0x7, true);
4710
4711        rxdctl |= IGB_RX_PTHRESH;
4712        rxdctl |= IGB_RX_HTHRESH << 8;
4713        rxdctl |= IGB_RX_WTHRESH << 16;
4714
4715        /* initialize rx_buffer_info */
4716        memset(ring->rx_buffer_info, 0,
4717               sizeof(struct igb_rx_buffer) * ring->count);
4718
4719        /* initialize Rx descriptor 0 */
4720        rx_desc = IGB_RX_DESC(ring, 0);
4721        rx_desc->wb.upper.length = 0;
4722
4723        /* enable receive descriptor fetching */
4724        rxdctl |= E1000_RXDCTL_QUEUE_ENABLE;
4725        wr32(E1000_RXDCTL(reg_idx), rxdctl);
4726}
4727
4728static void igb_set_rx_buffer_len(struct igb_adapter *adapter,
4729                                  struct igb_ring *rx_ring)
4730{
4731        /* set build_skb and buffer size flags */
4732        clear_ring_build_skb_enabled(rx_ring);
4733        clear_ring_uses_large_buffer(rx_ring);
4734
4735        if (adapter->flags & IGB_FLAG_RX_LEGACY)
4736                return;
4737
4738        set_ring_build_skb_enabled(rx_ring);
4739
4740#if (PAGE_SIZE < 8192)
4741        if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
4742                return;
4743
4744        set_ring_uses_large_buffer(rx_ring);
4745#endif
4746}
4747
4748/**
4749 *  igb_configure_rx - Configure receive Unit after Reset
4750 *  @adapter: board private structure
4751 *
4752 *  Configure the Rx unit of the MAC after a reset.
4753 **/
4754static void igb_configure_rx(struct igb_adapter *adapter)
4755{
4756        int i;
4757
4758        /* set the correct pool for the PF default MAC address in entry 0 */
4759        igb_set_default_mac_filter(adapter);
4760
4761        /* Setup the HW Rx Head and Tail Descriptor Pointers and
4762         * the Base and Length of the Rx Descriptor Ring
4763         */
4764        for (i = 0; i < adapter->num_rx_queues; i++) {
4765                struct igb_ring *rx_ring = adapter->rx_ring[i];
4766
4767                igb_set_rx_buffer_len(adapter, rx_ring);
4768                igb_configure_rx_ring(adapter, rx_ring);
4769        }
4770}
4771
4772/**
4773 *  igb_free_tx_resources - Free Tx Resources per Queue
4774 *  @tx_ring: Tx descriptor ring for a specific queue
4775 *
4776 *  Free all transmit software resources
4777 **/
4778void igb_free_tx_resources(struct igb_ring *tx_ring)
4779{
4780        igb_clean_tx_ring(tx_ring);
4781
4782        vfree(tx_ring->tx_buffer_info);
4783        tx_ring->tx_buffer_info = NULL;
4784
4785        /* if not set, then don't free */
4786        if (!tx_ring->desc)
4787                return;
4788
4789        dma_free_coherent(tx_ring->dev, tx_ring->size,
4790                          tx_ring->desc, tx_ring->dma);
4791
4792        tx_ring->desc = NULL;
4793}
4794
4795/**
4796 *  igb_free_all_tx_resources - Free Tx Resources for All Queues
4797 *  @adapter: board private structure
4798 *
4799 *  Free all transmit software resources
4800 **/
4801static void igb_free_all_tx_resources(struct igb_adapter *adapter)
4802{
4803        int i;
4804
4805        for (i = 0; i < adapter->num_tx_queues; i++)
4806                if (adapter->tx_ring[i])
4807                        igb_free_tx_resources(adapter->tx_ring[i]);
4808}
4809
4810/**
4811 *  igb_clean_tx_ring - Free Tx Buffers
4812 *  @tx_ring: ring to be cleaned
4813 **/
4814static void igb_clean_tx_ring(struct igb_ring *tx_ring)
4815{
4816        u16 i = tx_ring->next_to_clean;
4817        struct igb_tx_buffer *tx_buffer = &tx_ring->tx_buffer_info[i];
4818
4819        while (i != tx_ring->next_to_use) {
4820                union e1000_adv_tx_desc *eop_desc, *tx_desc;
4821
4822                /* Free all the Tx ring sk_buffs or xdp frames */
4823                if (tx_buffer->type == IGB_TYPE_SKB)
4824                        dev_kfree_skb_any(tx_buffer->skb);
4825                else
4826                        xdp_return_frame(tx_buffer->xdpf);
4827
4828                /* unmap skb header data */
4829                dma_unmap_single(tx_ring->dev,
4830                                 dma_unmap_addr(tx_buffer, dma),
4831                                 dma_unmap_len(tx_buffer, len),
4832                                 DMA_TO_DEVICE);
4833
4834                /* check for eop_desc to determine the end of the packet */
4835                eop_desc = tx_buffer->next_to_watch;
4836                tx_desc = IGB_TX_DESC(tx_ring, i);
4837
4838                /* unmap remaining buffers */
4839                while (tx_desc != eop_desc) {
4840                        tx_buffer++;
4841                        tx_desc++;
4842                        i++;
4843                        if (unlikely(i == tx_ring->count)) {
4844                                i = 0;
4845                                tx_buffer = tx_ring->tx_buffer_info;
4846                                tx_desc = IGB_TX_DESC(tx_ring, 0);
4847                        }
4848
4849                        /* unmap any remaining paged data */
4850                        if (dma_unmap_len(tx_buffer, len))
4851                                dma_unmap_page(tx_ring->dev,
4852                                               dma_unmap_addr(tx_buffer, dma),
4853                                               dma_unmap_len(tx_buffer, len),
4854                                               DMA_TO_DEVICE);
4855                }
4856
4857                tx_buffer->next_to_watch = NULL;
4858
4859                /* move us one more past the eop_desc for start of next pkt */
4860                tx_buffer++;
4861                i++;
4862                if (unlikely(i == tx_ring->count)) {
4863                        i = 0;
4864                        tx_buffer = tx_ring->tx_buffer_info;
4865                }
4866        }
4867
4868        /* reset BQL for queue */
4869        netdev_tx_reset_queue(txring_txq(tx_ring));
4870
4871        /* reset next_to_use and next_to_clean */
4872        tx_ring->next_to_use = 0;
4873        tx_ring->next_to_clean = 0;
4874}
4875
4876/**
4877 *  igb_clean_all_tx_rings - Free Tx Buffers for all queues
4878 *  @adapter: board private structure
4879 **/
4880static void igb_clean_all_tx_rings(struct igb_adapter *adapter)
4881{
4882        int i;
4883
4884        for (i = 0; i < adapter->num_tx_queues; i++)
4885                if (adapter->tx_ring[i])
4886                        igb_clean_tx_ring(adapter->tx_ring[i]);
4887}
4888
4889/**
4890 *  igb_free_rx_resources - Free Rx Resources
4891 *  @rx_ring: ring to clean the resources from
4892 *
4893 *  Free all receive software resources
4894 **/
4895void igb_free_rx_resources(struct igb_ring *rx_ring)
4896{
4897        igb_clean_rx_ring(rx_ring);
4898
4899        rx_ring->xdp_prog = NULL;
4900        xdp_rxq_info_unreg(&rx_ring->xdp_rxq);
4901        vfree(rx_ring->rx_buffer_info);
4902        rx_ring->rx_buffer_info = NULL;
4903
4904        /* if not set, then don't free */
4905        if (!rx_ring->desc)
4906                return;
4907
4908        dma_free_coherent(rx_ring->dev, rx_ring->size,
4909                          rx_ring->desc, rx_ring->dma);
4910
4911        rx_ring->desc = NULL;
4912}
4913
4914/**
4915 *  igb_free_all_rx_resources - Free Rx Resources for All Queues
4916 *  @adapter: board private structure
4917 *
4918 *  Free all receive software resources
4919 **/
4920static void igb_free_all_rx_resources(struct igb_adapter *adapter)
4921{
4922        int i;
4923
4924        for (i = 0; i < adapter->num_rx_queues; i++)
4925                if (adapter->rx_ring[i])
4926                        igb_free_rx_resources(adapter->rx_ring[i]);
4927}
4928
4929/**
4930 *  igb_clean_rx_ring - Free Rx Buffers per Queue
4931 *  @rx_ring: ring to free buffers from
4932 **/
4933static void igb_clean_rx_ring(struct igb_ring *rx_ring)
4934{
4935        u16 i = rx_ring->next_to_clean;
4936
4937        dev_kfree_skb(rx_ring->skb);
4938        rx_ring->skb = NULL;
4939
4940        /* Free all the Rx ring sk_buffs */
4941        while (i != rx_ring->next_to_alloc) {
4942                struct igb_rx_buffer *buffer_info = &rx_ring->rx_buffer_info[i];
4943
4944                /* Invalidate cache lines that may have been written to by
4945                 * device so that we avoid corrupting memory.
4946                 */
4947                dma_sync_single_range_for_cpu(rx_ring->dev,
4948                                              buffer_info->dma,
4949                                              buffer_info->page_offset,
4950                                              igb_rx_bufsz(rx_ring),
4951                                              DMA_FROM_DEVICE);
4952
4953                /* free resources associated with mapping */
4954                dma_unmap_page_attrs(rx_ring->dev,
4955                                     buffer_info->dma,
4956                                     igb_rx_pg_size(rx_ring),
4957                                     DMA_FROM_DEVICE,
4958                                     IGB_RX_DMA_ATTR);
4959                __page_frag_cache_drain(buffer_info->page,
4960                                        buffer_info->pagecnt_bias);
4961
4962                i++;
4963                if (i == rx_ring->count)
4964                        i = 0;
4965        }
4966
4967        rx_ring->next_to_alloc = 0;
4968        rx_ring->next_to_clean = 0;
4969        rx_ring->next_to_use = 0;
4970}
4971
4972/**
4973 *  igb_clean_all_rx_rings - Free Rx Buffers for all queues
4974 *  @adapter: board private structure
4975 **/
4976static void igb_clean_all_rx_rings(struct igb_adapter *adapter)
4977{
4978        int i;
4979
4980        for (i = 0; i < adapter->num_rx_queues; i++)
4981                if (adapter->rx_ring[i])
4982                        igb_clean_rx_ring(adapter->rx_ring[i]);
4983}
4984
4985/**
4986 *  igb_set_mac - Change the Ethernet Address of the NIC
4987 *  @netdev: network interface device structure
4988 *  @p: pointer to an address structure
4989 *
4990 *  Returns 0 on success, negative on failure
4991 **/
4992static int igb_set_mac(struct net_device *netdev, void *p)
4993{
4994        struct igb_adapter *adapter = netdev_priv(netdev);
4995        struct e1000_hw *hw = &adapter->hw;
4996        struct sockaddr *addr = p;
4997
4998        if (!is_valid_ether_addr(addr->sa_data))
4999                return -EADDRNOTAVAIL;
5000
5001        eth_hw_addr_set(netdev, addr->sa_data);
5002        memcpy(hw->mac.addr, addr->sa_data, netdev->addr_len);
5003
5004        /* set the correct pool for the new PF MAC address in entry 0 */
5005        igb_set_default_mac_filter(adapter);
5006
5007        return 0;
5008}
5009
5010/**
5011 *  igb_write_mc_addr_list - write multicast addresses to MTA
5012 *  @netdev: network interface device structure
5013 *
5014 *  Writes multicast address list to the MTA hash table.
5015 *  Returns: -ENOMEM on failure
5016 *           0 on no addresses written
5017 *           X on writing X addresses to MTA
5018 **/
5019static int igb_write_mc_addr_list(struct net_device *netdev)
5020{
5021        struct igb_adapter *adapter = netdev_priv(netdev);
5022        struct e1000_hw *hw = &adapter->hw;
5023        struct netdev_hw_addr *ha;
5024        u8  *mta_list;
5025        int i;
5026
5027        if (netdev_mc_empty(netdev)) {
5028                /* nothing to program, so clear mc list */
5029                igb_update_mc_addr_list(hw, NULL, 0);
5030                igb_restore_vf_multicasts(adapter);
5031                return 0;
5032        }
5033
5034        mta_list = kcalloc(netdev_mc_count(netdev), 6, GFP_ATOMIC);
5035        if (!mta_list)
5036                return -ENOMEM;
5037
5038        /* The shared function expects a packed array of only addresses. */
5039        i = 0;
5040        netdev_for_each_mc_addr(ha, netdev)
5041                memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
5042
5043        igb_update_mc_addr_list(hw, mta_list, i);
5044        kfree(mta_list);
5045
5046        return netdev_mc_count(netdev);
5047}
5048
5049static int igb_vlan_promisc_enable(struct igb_adapter *adapter)
5050{
5051        struct e1000_hw *hw = &adapter->hw;
5052        u32 i, pf_id;
5053
5054        switch (hw->mac.type) {
5055        case e1000_i210:
5056        case e1000_i211:
5057        case e1000_i350:
5058                /* VLAN filtering needed for VLAN prio filter */
5059                if (adapter->netdev->features & NETIF_F_NTUPLE)
5060                        break;
5061                fallthrough;
5062        case e1000_82576:
5063        case e1000_82580:
5064        case e1000_i354:
5065                /* VLAN filtering needed for pool filtering */
5066                if (adapter->vfs_allocated_count)
5067                        break;
5068                fallthrough;
5069        default:
5070                return 1;
5071        }
5072
5073        /* We are already in VLAN promisc, nothing to do */
5074        if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
5075                return 0;
5076
5077        if (!adapter->vfs_allocated_count)
5078                goto set_vfta;
5079
5080        /* Add PF to all active pools */
5081        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5082
5083        for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5084                u32 vlvf = rd32(E1000_VLVF(i));
5085
5086                vlvf |= BIT(pf_id);
5087                wr32(E1000_VLVF(i), vlvf);
5088        }
5089
5090set_vfta:
5091        /* Set all bits in the VLAN filter table array */
5092        for (i = E1000_VLAN_FILTER_TBL_SIZE; i--;)
5093                hw->mac.ops.write_vfta(hw, i, ~0U);
5094
5095        /* Set flag so we don't redo unnecessary work */
5096        adapter->flags |= IGB_FLAG_VLAN_PROMISC;
5097
5098        return 0;
5099}
5100
5101#define VFTA_BLOCK_SIZE 8
5102static void igb_scrub_vfta(struct igb_adapter *adapter, u32 vfta_offset)
5103{
5104        struct e1000_hw *hw = &adapter->hw;
5105        u32 vfta[VFTA_BLOCK_SIZE] = { 0 };
5106        u32 vid_start = vfta_offset * 32;
5107        u32 vid_end = vid_start + (VFTA_BLOCK_SIZE * 32);
5108        u32 i, vid, word, bits, pf_id;
5109
5110        /* guarantee that we don't scrub out management VLAN */
5111        vid = adapter->mng_vlan_id;
5112        if (vid >= vid_start && vid < vid_end)
5113                vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5114
5115        if (!adapter->vfs_allocated_count)
5116                goto set_vfta;
5117
5118        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
5119
5120        for (i = E1000_VLVF_ARRAY_SIZE; --i;) {
5121                u32 vlvf = rd32(E1000_VLVF(i));
5122
5123                /* pull VLAN ID from VLVF */
5124                vid = vlvf & VLAN_VID_MASK;
5125
5126                /* only concern ourselves with a certain range */
5127                if (vid < vid_start || vid >= vid_end)
5128                        continue;
5129
5130                if (vlvf & E1000_VLVF_VLANID_ENABLE) {
5131                        /* record VLAN ID in VFTA */
5132                        vfta[(vid - vid_start) / 32] |= BIT(vid % 32);
5133
5134                        /* if PF is part of this then continue */
5135                        if (test_bit(vid, adapter->active_vlans))
5136                                continue;
5137                }
5138
5139                /* remove PF from the pool */
5140                bits = ~BIT(pf_id);
5141                bits &= rd32(E1000_VLVF(i));
5142                wr32(E1000_VLVF(i), bits);
5143        }
5144
5145set_vfta:
5146        /* extract values from active_vlans and write back to VFTA */
5147        for (i = VFTA_BLOCK_SIZE; i--;) {
5148                vid = (vfta_offset + i) * 32;
5149                word = vid / BITS_PER_LONG;
5150                bits = vid % BITS_PER_LONG;
5151
5152                vfta[i] |= adapter->active_vlans[word] >> bits;
5153
5154                hw->mac.ops.write_vfta(hw, vfta_offset + i, vfta[i]);
5155        }
5156}
5157
5158static void igb_vlan_promisc_disable(struct igb_adapter *adapter)
5159{
5160        u32 i;
5161
5162        /* We are not in VLAN promisc, nothing to do */
5163        if (!(adapter->flags & IGB_FLAG_VLAN_PROMISC))
5164                return;
5165
5166        /* Set flag so we don't redo unnecessary work */
5167        adapter->flags &= ~IGB_FLAG_VLAN_PROMISC;
5168
5169        for (i = 0; i < E1000_VLAN_FILTER_TBL_SIZE; i += VFTA_BLOCK_SIZE)
5170                igb_scrub_vfta(adapter, i);
5171}
5172
5173/**
5174 *  igb_set_rx_mode - Secondary Unicast, Multicast and Promiscuous mode set
5175 *  @netdev: network interface device structure
5176 *
5177 *  The set_rx_mode entry point is called whenever the unicast or multicast
5178 *  address lists or the network interface flags are updated.  This routine is
5179 *  responsible for configuring the hardware for proper unicast, multicast,
5180 *  promiscuous mode, and all-multi behavior.
5181 **/
5182static void igb_set_rx_mode(struct net_device *netdev)
5183{
5184        struct igb_adapter *adapter = netdev_priv(netdev);
5185        struct e1000_hw *hw = &adapter->hw;
5186        unsigned int vfn = adapter->vfs_allocated_count;
5187        u32 rctl = 0, vmolr = 0, rlpml = MAX_JUMBO_FRAME_SIZE;
5188        int count;
5189
5190        /* Check for Promiscuous and All Multicast modes */
5191        if (netdev->flags & IFF_PROMISC) {
5192                rctl |= E1000_RCTL_UPE | E1000_RCTL_MPE;
5193                vmolr |= E1000_VMOLR_MPME;
5194
5195                /* enable use of UTA filter to force packets to default pool */
5196                if (hw->mac.type == e1000_82576)
5197                        vmolr |= E1000_VMOLR_ROPE;
5198        } else {
5199                if (netdev->flags & IFF_ALLMULTI) {
5200                        rctl |= E1000_RCTL_MPE;
5201                        vmolr |= E1000_VMOLR_MPME;
5202                } else {
5203                        /* Write addresses to the MTA, if the attempt fails
5204                         * then we should just turn on promiscuous mode so
5205                         * that we can at least receive multicast traffic
5206                         */
5207                        count = igb_write_mc_addr_list(netdev);
5208                        if (count < 0) {
5209                                rctl |= E1000_RCTL_MPE;
5210                                vmolr |= E1000_VMOLR_MPME;
5211                        } else if (count) {
5212                                vmolr |= E1000_VMOLR_ROMPE;
5213                        }
5214                }
5215        }
5216
5217        /* Write addresses to available RAR registers, if there is not
5218         * sufficient space to store all the addresses then enable
5219         * unicast promiscuous mode
5220         */
5221        if (__dev_uc_sync(netdev, igb_uc_sync, igb_uc_unsync)) {
5222                rctl |= E1000_RCTL_UPE;
5223                vmolr |= E1000_VMOLR_ROPE;
5224        }
5225
5226        /* enable VLAN filtering by default */
5227        rctl |= E1000_RCTL_VFE;
5228
5229        /* disable VLAN filtering for modes that require it */
5230        if ((netdev->flags & IFF_PROMISC) ||
5231            (netdev->features & NETIF_F_RXALL)) {
5232                /* if we fail to set all rules then just clear VFE */
5233                if (igb_vlan_promisc_enable(adapter))
5234                        rctl &= ~E1000_RCTL_VFE;
5235        } else {
5236                igb_vlan_promisc_disable(adapter);
5237        }
5238
5239        /* update state of unicast, multicast, and VLAN filtering modes */
5240        rctl |= rd32(E1000_RCTL) & ~(E1000_RCTL_UPE | E1000_RCTL_MPE |
5241                                     E1000_RCTL_VFE);
5242        wr32(E1000_RCTL, rctl);
5243
5244#if (PAGE_SIZE < 8192)
5245        if (!adapter->vfs_allocated_count) {
5246                if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5247                        rlpml = IGB_MAX_FRAME_BUILD_SKB;
5248        }
5249#endif
5250        wr32(E1000_RLPML, rlpml);
5251
5252        /* In order to support SR-IOV and eventually VMDq it is necessary to set
5253         * the VMOLR to enable the appropriate modes.  Without this workaround
5254         * we will have issues with VLAN tag stripping not being done for frames
5255         * that are only arriving because we are the default pool
5256         */
5257        if ((hw->mac.type < e1000_82576) || (hw->mac.type > e1000_i350))
5258                return;
5259
5260        /* set UTA to appropriate mode */
5261        igb_set_uta(adapter, !!(vmolr & E1000_VMOLR_ROPE));
5262
5263        vmolr |= rd32(E1000_VMOLR(vfn)) &
5264                 ~(E1000_VMOLR_ROPE | E1000_VMOLR_MPME | E1000_VMOLR_ROMPE);
5265
5266        /* enable Rx jumbo frames, restrict as needed to support build_skb */
5267        vmolr &= ~E1000_VMOLR_RLPML_MASK;
5268#if (PAGE_SIZE < 8192)
5269        if (adapter->max_frame_size <= IGB_MAX_FRAME_BUILD_SKB)
5270                vmolr |= IGB_MAX_FRAME_BUILD_SKB;
5271        else
5272#endif
5273                vmolr |= MAX_JUMBO_FRAME_SIZE;
5274        vmolr |= E1000_VMOLR_LPE;
5275
5276        wr32(E1000_VMOLR(vfn), vmolr);
5277
5278        igb_restore_vf_multicasts(adapter);
5279}
5280
5281static void igb_check_wvbr(struct igb_adapter *adapter)
5282{
5283        struct e1000_hw *hw = &adapter->hw;
5284        u32 wvbr = 0;
5285
5286        switch (hw->mac.type) {
5287        case e1000_82576:
5288        case e1000_i350:
5289                wvbr = rd32(E1000_WVBR);
5290                if (!wvbr)
5291                        return;
5292                break;
5293        default:
5294                break;
5295        }
5296
5297        adapter->wvbr |= wvbr;
5298}
5299
5300#define IGB_STAGGERED_QUEUE_OFFSET 8
5301
5302static void igb_spoof_check(struct igb_adapter *adapter)
5303{
5304        int j;
5305
5306        if (!adapter->wvbr)
5307                return;
5308
5309        for (j = 0; j < adapter->vfs_allocated_count; j++) {
5310                if (adapter->wvbr & BIT(j) ||
5311                    adapter->wvbr & BIT(j + IGB_STAGGERED_QUEUE_OFFSET)) {
5312                        dev_warn(&adapter->pdev->dev,
5313                                "Spoof event(s) detected on VF %d\n", j);
5314                        adapter->wvbr &=
5315                                ~(BIT(j) |
5316                                  BIT(j + IGB_STAGGERED_QUEUE_OFFSET));
5317                }
5318        }
5319}
5320
5321/* Need to wait a few seconds after link up to get diagnostic information from
5322 * the phy
5323 */
5324static void igb_update_phy_info(struct timer_list *t)
5325{
5326        struct igb_adapter *adapter = from_timer(adapter, t, phy_info_timer);
5327        igb_get_phy_info(&adapter->hw);
5328}
5329
5330/**
5331 *  igb_has_link - check shared code for link and determine up/down
5332 *  @adapter: pointer to driver private info
5333 **/
5334bool igb_has_link(struct igb_adapter *adapter)
5335{
5336        struct e1000_hw *hw = &adapter->hw;
5337        bool link_active = false;
5338
5339        /* get_link_status is set on LSC (link status) interrupt or
5340         * rx sequence error interrupt.  get_link_status will stay
5341         * false until the e1000_check_for_link establishes link
5342         * for copper adapters ONLY
5343         */
5344        switch (hw->phy.media_type) {
5345        case e1000_media_type_copper:
5346                if (!hw->mac.get_link_status)
5347                        return true;
5348                fallthrough;
5349        case e1000_media_type_internal_serdes:
5350                hw->mac.ops.check_for_link(hw);
5351                link_active = !hw->mac.get_link_status;
5352                break;
5353        default:
5354        case e1000_media_type_unknown:
5355                break;
5356        }
5357
5358        if (((hw->mac.type == e1000_i210) ||
5359             (hw->mac.type == e1000_i211)) &&
5360             (hw->phy.id == I210_I_PHY_ID)) {
5361                if (!netif_carrier_ok(adapter->netdev)) {
5362                        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5363                } else if (!(adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)) {
5364                        adapter->flags |= IGB_FLAG_NEED_LINK_UPDATE;
5365                        adapter->link_check_timeout = jiffies;
5366                }
5367        }
5368
5369        return link_active;
5370}
5371
5372static bool igb_thermal_sensor_event(struct e1000_hw *hw, u32 event)
5373{
5374        bool ret = false;
5375        u32 ctrl_ext, thstat;
5376
5377        /* check for thermal sensor event on i350 copper only */
5378        if (hw->mac.type == e1000_i350) {
5379                thstat = rd32(E1000_THSTAT);
5380                ctrl_ext = rd32(E1000_CTRL_EXT);
5381
5382                if ((hw->phy.media_type == e1000_media_type_copper) &&
5383                    !(ctrl_ext & E1000_CTRL_EXT_LINK_MODE_SGMII))
5384                        ret = !!(thstat & event);
5385        }
5386
5387        return ret;
5388}
5389
5390/**
5391 *  igb_check_lvmmc - check for malformed packets received
5392 *  and indicated in LVMMC register
5393 *  @adapter: pointer to adapter
5394 **/
5395static void igb_check_lvmmc(struct igb_adapter *adapter)
5396{
5397        struct e1000_hw *hw = &adapter->hw;
5398        u32 lvmmc;
5399
5400        lvmmc = rd32(E1000_LVMMC);
5401        if (lvmmc) {
5402                if (unlikely(net_ratelimit())) {
5403                        netdev_warn(adapter->netdev,
5404                                    "malformed Tx packet detected and dropped, LVMMC:0x%08x\n",
5405                                    lvmmc);
5406                }
5407        }
5408}
5409
5410/**
5411 *  igb_watchdog - Timer Call-back
5412 *  @t: pointer to timer_list containing our private info pointer
5413 **/
5414static void igb_watchdog(struct timer_list *t)
5415{
5416        struct igb_adapter *adapter = from_timer(adapter, t, watchdog_timer);
5417        /* Do the rest outside of interrupt context */
5418        schedule_work(&adapter->watchdog_task);
5419}
5420
5421static void igb_watchdog_task(struct work_struct *work)
5422{
5423        struct igb_adapter *adapter = container_of(work,
5424                                                   struct igb_adapter,
5425                                                   watchdog_task);
5426        struct e1000_hw *hw = &adapter->hw;
5427        struct e1000_phy_info *phy = &hw->phy;
5428        struct net_device *netdev = adapter->netdev;
5429        u32 link;
5430        int i;
5431        u32 connsw;
5432        u16 phy_data, retry_count = 20;
5433
5434        link = igb_has_link(adapter);
5435
5436        if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE) {
5437                if (time_after(jiffies, (adapter->link_check_timeout + HZ)))
5438                        adapter->flags &= ~IGB_FLAG_NEED_LINK_UPDATE;
5439                else
5440                        link = false;
5441        }
5442
5443        /* Force link down if we have fiber to swap to */
5444        if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5445                if (hw->phy.media_type == e1000_media_type_copper) {
5446                        connsw = rd32(E1000_CONNSW);
5447                        if (!(connsw & E1000_CONNSW_AUTOSENSE_EN))
5448                                link = 0;
5449                }
5450        }
5451        if (link) {
5452                /* Perform a reset if the media type changed. */
5453                if (hw->dev_spec._82575.media_changed) {
5454                        hw->dev_spec._82575.media_changed = false;
5455                        adapter->flags |= IGB_FLAG_MEDIA_RESET;
5456                        igb_reset(adapter);
5457                }
5458                /* Cancel scheduled suspend requests. */
5459                pm_runtime_resume(netdev->dev.parent);
5460
5461                if (!netif_carrier_ok(netdev)) {
5462                        u32 ctrl;
5463
5464                        hw->mac.ops.get_speed_and_duplex(hw,
5465                                                         &adapter->link_speed,
5466                                                         &adapter->link_duplex);
5467
5468                        ctrl = rd32(E1000_CTRL);
5469                        /* Links status message must follow this format */
5470                        netdev_info(netdev,
5471                               "igb: %s NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5472                               netdev->name,
5473                               adapter->link_speed,
5474                               adapter->link_duplex == FULL_DUPLEX ?
5475                               "Full" : "Half",
5476                               (ctrl & E1000_CTRL_TFCE) &&
5477                               (ctrl & E1000_CTRL_RFCE) ? "RX/TX" :
5478                               (ctrl & E1000_CTRL_RFCE) ?  "RX" :
5479                               (ctrl & E1000_CTRL_TFCE) ?  "TX" : "None");
5480
5481                        /* disable EEE if enabled */
5482                        if ((adapter->flags & IGB_FLAG_EEE) &&
5483                                (adapter->link_duplex == HALF_DUPLEX)) {
5484                                dev_info(&adapter->pdev->dev,
5485                                "EEE Disabled: unsupported at half duplex. Re-enable using ethtool when at full duplex.\n");
5486                                adapter->hw.dev_spec._82575.eee_disable = true;
5487                                adapter->flags &= ~IGB_FLAG_EEE;
5488                        }
5489
5490                        /* check if SmartSpeed worked */
5491                        igb_check_downshift(hw);
5492                        if (phy->speed_downgraded)
5493                                netdev_warn(netdev, "Link Speed was downgraded by SmartSpeed\n");
5494
5495                        /* check for thermal sensor event */
5496                        if (igb_thermal_sensor_event(hw,
5497                            E1000_THSTAT_LINK_THROTTLE))
5498                                netdev_info(netdev, "The network adapter link speed was downshifted because it overheated\n");
5499
5500                        /* adjust timeout factor according to speed/duplex */
5501                        adapter->tx_timeout_factor = 1;
5502                        switch (adapter->link_speed) {
5503                        case SPEED_10:
5504                                adapter->tx_timeout_factor = 14;
5505                                break;
5506                        case SPEED_100:
5507                                /* maybe add some timeout factor ? */
5508                                break;
5509                        }
5510
5511                        if (adapter->link_speed != SPEED_1000 ||
5512                            !hw->phy.ops.read_reg)
5513                                goto no_wait;
5514
5515                        /* wait for Remote receiver status OK */
5516retry_read_status:
5517                        if (!igb_read_phy_reg(hw, PHY_1000T_STATUS,
5518                                              &phy_data)) {
5519                                if (!(phy_data & SR_1000T_REMOTE_RX_STATUS) &&
5520                                    retry_count) {
5521                                        msleep(100);
5522                                        retry_count--;
5523                                        goto retry_read_status;
5524                                } else if (!retry_count) {
5525                                        dev_err(&adapter->pdev->dev, "exceed max 2 second\n");
5526                                }
5527                        } else {
5528                                dev_err(&adapter->pdev->dev, "read 1000Base-T Status Reg\n");
5529                        }
5530no_wait:
5531                        netif_carrier_on(netdev);
5532
5533                        igb_ping_all_vfs(adapter);
5534                        igb_check_vf_rate_limit(adapter);
5535
5536                        /* link state has changed, schedule phy info update */
5537                        if (!test_bit(__IGB_DOWN, &adapter->state))
5538                                mod_timer(&adapter->phy_info_timer,
5539                                          round_jiffies(jiffies + 2 * HZ));
5540                }
5541        } else {
5542                if (netif_carrier_ok(netdev)) {
5543                        adapter->link_speed = 0;
5544                        adapter->link_duplex = 0;
5545
5546                        /* check for thermal sensor event */
5547                        if (igb_thermal_sensor_event(hw,
5548                            E1000_THSTAT_PWR_DOWN)) {
5549                                netdev_err(netdev, "The network adapter was stopped because it overheated\n");
5550                        }
5551
5552                        /* Links status message must follow this format */
5553                        netdev_info(netdev, "igb: %s NIC Link is Down\n",
5554                               netdev->name);
5555                        netif_carrier_off(netdev);
5556
5557                        igb_ping_all_vfs(adapter);
5558
5559                        /* link state has changed, schedule phy info update */
5560                        if (!test_bit(__IGB_DOWN, &adapter->state))
5561                                mod_timer(&adapter->phy_info_timer,
5562                                          round_jiffies(jiffies + 2 * HZ));
5563
5564                        /* link is down, time to check for alternate media */
5565                        if (adapter->flags & IGB_FLAG_MAS_ENABLE) {
5566                                igb_check_swap_media(adapter);
5567                                if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5568                                        schedule_work(&adapter->reset_task);
5569                                        /* return immediately */
5570                                        return;
5571                                }
5572                        }
5573                        pm_schedule_suspend(netdev->dev.parent,
5574                                            MSEC_PER_SEC * 5);
5575
5576                /* also check for alternate media here */
5577                } else if (!netif_carrier_ok(netdev) &&
5578                           (adapter->flags & IGB_FLAG_MAS_ENABLE)) {
5579                        igb_check_swap_media(adapter);
5580                        if (adapter->flags & IGB_FLAG_MEDIA_RESET) {
5581                                schedule_work(&adapter->reset_task);
5582                                /* return immediately */
5583                                return;
5584                        }
5585                }
5586        }
5587
5588        spin_lock(&adapter->stats64_lock);
5589        igb_update_stats(adapter);
5590        spin_unlock(&adapter->stats64_lock);
5591
5592        for (i = 0; i < adapter->num_tx_queues; i++) {
5593                struct igb_ring *tx_ring = adapter->tx_ring[i];
5594                if (!netif_carrier_ok(netdev)) {
5595                        /* We've lost link, so the controller stops DMA,
5596                         * but we've got queued Tx work that's never going
5597                         * to get done, so reset controller to flush Tx.
5598                         * (Do the reset outside of interrupt context).
5599                         */
5600                        if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
5601                                adapter->tx_timeout_count++;
5602                                schedule_work(&adapter->reset_task);
5603                                /* return immediately since reset is imminent */
5604                                return;
5605                        }
5606                }
5607
5608                /* Force detection of hung controller every watchdog period */
5609                set_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
5610        }
5611
5612        /* Cause software interrupt to ensure Rx ring is cleaned */
5613        if (adapter->flags & IGB_FLAG_HAS_MSIX) {
5614                u32 eics = 0;
5615
5616                for (i = 0; i < adapter->num_q_vectors; i++)
5617                        eics |= adapter->q_vector[i]->eims_value;
5618                wr32(E1000_EICS, eics);
5619        } else {
5620                wr32(E1000_ICS, E1000_ICS_RXDMT0);
5621        }
5622
5623        igb_spoof_check(adapter);
5624        igb_ptp_rx_hang(adapter);
5625        igb_ptp_tx_hang(adapter);
5626
5627        /* Check LVMMC register on i350/i354 only */
5628        if ((adapter->hw.mac.type == e1000_i350) ||
5629            (adapter->hw.mac.type == e1000_i354))
5630                igb_check_lvmmc(adapter);
5631
5632        /* Reset the timer */
5633        if (!test_bit(__IGB_DOWN, &adapter->state)) {
5634                if (adapter->flags & IGB_FLAG_NEED_LINK_UPDATE)
5635                        mod_timer(&adapter->watchdog_timer,
5636                                  round_jiffies(jiffies +  HZ));
5637                else
5638                        mod_timer(&adapter->watchdog_timer,
5639                                  round_jiffies(jiffies + 2 * HZ));
5640        }
5641}
5642
5643enum latency_range {
5644        lowest_latency = 0,
5645        low_latency = 1,
5646        bulk_latency = 2,
5647        latency_invalid = 255
5648};
5649
5650/**
5651 *  igb_update_ring_itr - update the dynamic ITR value based on packet size
5652 *  @q_vector: pointer to q_vector
5653 *
5654 *  Stores a new ITR value based on strictly on packet size.  This
5655 *  algorithm is less sophisticated than that used in igb_update_itr,
5656 *  due to the difficulty of synchronizing statistics across multiple
5657 *  receive rings.  The divisors and thresholds used by this function
5658 *  were determined based on theoretical maximum wire speed and testing
5659 *  data, in order to minimize response time while increasing bulk
5660 *  throughput.
5661 *  This functionality is controlled by ethtool's coalescing settings.
5662 *  NOTE:  This function is called only when operating in a multiqueue
5663 *         receive environment.
5664 **/
5665static void igb_update_ring_itr(struct igb_q_vector *q_vector)
5666{
5667        int new_val = q_vector->itr_val;
5668        int avg_wire_size = 0;
5669        struct igb_adapter *adapter = q_vector->adapter;
5670        unsigned int packets;
5671
5672        /* For non-gigabit speeds, just fix the interrupt rate at 4000
5673         * ints/sec - ITR timer value of 120 ticks.
5674         */
5675        if (adapter->link_speed != SPEED_1000) {
5676                new_val = IGB_4K_ITR;
5677                goto set_itr_val;
5678        }
5679
5680        packets = q_vector->rx.total_packets;
5681        if (packets)
5682                avg_wire_size = q_vector->rx.total_bytes / packets;
5683
5684        packets = q_vector->tx.total_packets;
5685        if (packets)
5686                avg_wire_size = max_t(u32, avg_wire_size,
5687                                      q_vector->tx.total_bytes / packets);
5688
5689        /* if avg_wire_size isn't set no work was done */
5690        if (!avg_wire_size)
5691                goto clear_counts;
5692
5693        /* Add 24 bytes to size to account for CRC, preamble, and gap */
5694        avg_wire_size += 24;
5695
5696        /* Don't starve jumbo frames */
5697        avg_wire_size = min(avg_wire_size, 3000);
5698
5699        /* Give a little boost to mid-size frames */
5700        if ((avg_wire_size > 300) && (avg_wire_size < 1200))
5701                new_val = avg_wire_size / 3;
5702        else
5703                new_val = avg_wire_size / 2;
5704
5705        /* conservative mode (itr 3) eliminates the lowest_latency setting */
5706        if (new_val < IGB_20K_ITR &&
5707            ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5708             (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5709                new_val = IGB_20K_ITR;
5710
5711set_itr_val:
5712        if (new_val != q_vector->itr_val) {
5713                q_vector->itr_val = new_val;
5714                q_vector->set_itr = 1;
5715        }
5716clear_counts:
5717        q_vector->rx.total_bytes = 0;
5718        q_vector->rx.total_packets = 0;
5719        q_vector->tx.total_bytes = 0;
5720        q_vector->tx.total_packets = 0;
5721}
5722
5723/**
5724 *  igb_update_itr - update the dynamic ITR value based on statistics
5725 *  @q_vector: pointer to q_vector
5726 *  @ring_container: ring info to update the itr for
5727 *
5728 *  Stores a new ITR value based on packets and byte
5729 *  counts during the last interrupt.  The advantage of per interrupt
5730 *  computation is faster updates and more accurate ITR for the current
5731 *  traffic pattern.  Constants in this function were computed
5732 *  based on theoretical maximum wire speed and thresholds were set based
5733 *  on testing data as well as attempting to minimize response time
5734 *  while increasing bulk throughput.
5735 *  This functionality is controlled by ethtool's coalescing settings.
5736 *  NOTE:  These calculations are only valid when operating in a single-
5737 *         queue environment.
5738 **/
5739static void igb_update_itr(struct igb_q_vector *q_vector,
5740                           struct igb_ring_container *ring_container)
5741{
5742        unsigned int packets = ring_container->total_packets;
5743        unsigned int bytes = ring_container->total_bytes;
5744        u8 itrval = ring_container->itr;
5745
5746        /* no packets, exit with status unchanged */
5747        if (packets == 0)
5748                return;
5749
5750        switch (itrval) {
5751        case lowest_latency:
5752                /* handle TSO and jumbo frames */
5753                if (bytes/packets > 8000)
5754                        itrval = bulk_latency;
5755                else if ((packets < 5) && (bytes > 512))
5756                        itrval = low_latency;
5757                break;
5758        case low_latency:  /* 50 usec aka 20000 ints/s */
5759                if (bytes > 10000) {
5760                        /* this if handles the TSO accounting */
5761                        if (bytes/packets > 8000)
5762                                itrval = bulk_latency;
5763                        else if ((packets < 10) || ((bytes/packets) > 1200))
5764                                itrval = bulk_latency;
5765                        else if ((packets > 35))
5766                                itrval = lowest_latency;
5767                } else if (bytes/packets > 2000) {
5768                        itrval = bulk_latency;
5769                } else if (packets <= 2 && bytes < 512) {
5770                        itrval = lowest_latency;
5771                }
5772                break;
5773        case bulk_latency: /* 250 usec aka 4000 ints/s */
5774                if (bytes > 25000) {
5775                        if (packets > 35)
5776                                itrval = low_latency;
5777                } else if (bytes < 1500) {
5778                        itrval = low_latency;
5779                }
5780                break;
5781        }
5782
5783        /* clear work counters since we have the values we need */
5784        ring_container->total_bytes = 0;
5785        ring_container->total_packets = 0;
5786
5787        /* write updated itr to ring container */
5788        ring_container->itr = itrval;
5789}
5790
5791static void igb_set_itr(struct igb_q_vector *q_vector)
5792{
5793        struct igb_adapter *adapter = q_vector->adapter;
5794        u32 new_itr = q_vector->itr_val;
5795        u8 current_itr = 0;
5796
5797        /* for non-gigabit speeds, just fix the interrupt rate at 4000 */
5798        if (adapter->link_speed != SPEED_1000) {
5799                current_itr = 0;
5800                new_itr = IGB_4K_ITR;
5801                goto set_itr_now;
5802        }
5803
5804        igb_update_itr(q_vector, &q_vector->tx);
5805        igb_update_itr(q_vector, &q_vector->rx);
5806
5807        current_itr = max(q_vector->rx.itr, q_vector->tx.itr);
5808
5809        /* conservative mode (itr 3) eliminates the lowest_latency setting */
5810        if (current_itr == lowest_latency &&
5811            ((q_vector->rx.ring && adapter->rx_itr_setting == 3) ||
5812             (!q_vector->rx.ring && adapter->tx_itr_setting == 3)))
5813                current_itr = low_latency;
5814
5815        switch (current_itr) {
5816        /* counts and packets in update_itr are dependent on these numbers */
5817        case lowest_latency:
5818                new_itr = IGB_70K_ITR; /* 70,000 ints/sec */
5819                break;
5820        case low_latency:
5821                new_itr = IGB_20K_ITR; /* 20,000 ints/sec */
5822                break;
5823        case bulk_latency:
5824                new_itr = IGB_4K_ITR;  /* 4,000 ints/sec */
5825                break;
5826        default:
5827                break;
5828        }
5829
5830set_itr_now:
5831        if (new_itr != q_vector->itr_val) {
5832                /* this attempts to bias the interrupt rate towards Bulk
5833                 * by adding intermediate steps when interrupt rate is
5834                 * increasing
5835                 */
5836                new_itr = new_itr > q_vector->itr_val ?
5837                          max((new_itr * q_vector->itr_val) /
5838                          (new_itr + (q_vector->itr_val >> 2)),
5839                          new_itr) : new_itr;
5840                /* Don't write the value here; it resets the adapter's
5841                 * internal timer, and causes us to delay far longer than
5842                 * we should between interrupts.  Instead, we write the ITR
5843                 * value at the beginning of the next interrupt so the timing
5844                 * ends up being correct.
5845                 */
5846                q_vector->itr_val = new_itr;
5847                q_vector->set_itr = 1;
5848        }
5849}
5850
5851static void igb_tx_ctxtdesc(struct igb_ring *tx_ring,
5852                            struct igb_tx_buffer *first,
5853                            u32 vlan_macip_lens, u32 type_tucmd,
5854                            u32 mss_l4len_idx)
5855{
5856        struct e1000_adv_tx_context_desc *context_desc;
5857        u16 i = tx_ring->next_to_use;
5858        struct timespec64 ts;
5859
5860        context_desc = IGB_TX_CTXTDESC(tx_ring, i);
5861
5862        i++;
5863        tx_ring->next_to_use = (i < tx_ring->count) ? i : 0;
5864
5865        /* set bits to identify this as an advanced context descriptor */
5866        type_tucmd |= E1000_TXD_CMD_DEXT | E1000_ADVTXD_DTYP_CTXT;
5867
5868        /* For 82575, context index must be unique per ring. */
5869        if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
5870                mss_l4len_idx |= tx_ring->reg_idx << 4;
5871
5872        context_desc->vlan_macip_lens   = cpu_to_le32(vlan_macip_lens);
5873        context_desc->type_tucmd_mlhl   = cpu_to_le32(type_tucmd);
5874        context_desc->mss_l4len_idx     = cpu_to_le32(mss_l4len_idx);
5875
5876        /* We assume there is always a valid tx time available. Invalid times
5877         * should have been handled by the upper layers.
5878         */
5879        if (tx_ring->launchtime_enable) {
5880                ts = ktime_to_timespec64(first->skb->tstamp);
5881                skb_txtime_consumed(first->skb);
5882                context_desc->seqnum_seed = cpu_to_le32(ts.tv_nsec / 32);
5883        } else {
5884                context_desc->seqnum_seed = 0;
5885        }
5886}
5887
5888static int igb_tso(struct igb_ring *tx_ring,
5889                   struct igb_tx_buffer *first,
5890                   u8 *hdr_len)
5891{
5892        u32 vlan_macip_lens, type_tucmd, mss_l4len_idx;
5893        struct sk_buff *skb = first->skb;
5894        union {
5895                struct iphdr *v4;
5896                struct ipv6hdr *v6;
5897                unsigned char *hdr;
5898        } ip;
5899        union {
5900                struct tcphdr *tcp;
5901                struct udphdr *udp;
5902                unsigned char *hdr;
5903        } l4;
5904        u32 paylen, l4_offset;
5905        int err;
5906
5907        if (skb->ip_summed != CHECKSUM_PARTIAL)
5908                return 0;
5909
5910        if (!skb_is_gso(skb))
5911                return 0;
5912
5913        err = skb_cow_head(skb, 0);
5914        if (err < 0)
5915                return err;
5916
5917        ip.hdr = skb_network_header(skb);
5918        l4.hdr = skb_checksum_start(skb);
5919
5920        /* ADV DTYP TUCMD MKRLOC/ISCSIHEDLEN */
5921        type_tucmd = (skb_shinfo(skb)->gso_type & SKB_GSO_UDP_L4) ?
5922                      E1000_ADVTXD_TUCMD_L4T_UDP : E1000_ADVTXD_TUCMD_L4T_TCP;
5923
5924        /* initialize outer IP header fields */
5925        if (ip.v4->version == 4) {
5926                unsigned char *csum_start = skb_checksum_start(skb);
5927                unsigned char *trans_start = ip.hdr + (ip.v4->ihl * 4);
5928
5929                /* IP header will have to cancel out any data that
5930                 * is not a part of the outer IP header
5931                 */
5932                ip.v4->check = csum_fold(csum_partial(trans_start,
5933                                                      csum_start - trans_start,
5934                                                      0));
5935                type_tucmd |= E1000_ADVTXD_TUCMD_IPV4;
5936
5937                ip.v4->tot_len = 0;
5938                first->tx_flags |= IGB_TX_FLAGS_TSO |
5939                                   IGB_TX_FLAGS_CSUM |
5940                                   IGB_TX_FLAGS_IPV4;
5941        } else {
5942                ip.v6->payload_len = 0;
5943                first->tx_flags |= IGB_TX_FLAGS_TSO |
5944                                   IGB_TX_FLAGS_CSUM;
5945        }
5946
5947        /* determine offset of inner transport header */
5948        l4_offset = l4.hdr - skb->data;
5949
5950        /* remove payload length from inner checksum */
5951        paylen = skb->len - l4_offset;
5952        if (type_tucmd & E1000_ADVTXD_TUCMD_L4T_TCP) {
5953                /* compute length of segmentation header */
5954                *hdr_len = (l4.tcp->doff * 4) + l4_offset;
5955                csum_replace_by_diff(&l4.tcp->check,
5956                        (__force __wsum)htonl(paylen));
5957        } else {
5958                /* compute length of segmentation header */
5959                *hdr_len = sizeof(*l4.udp) + l4_offset;
5960                csum_replace_by_diff(&l4.udp->check,
5961                                     (__force __wsum)htonl(paylen));
5962        }
5963
5964        /* update gso size and bytecount with header size */
5965        first->gso_segs = skb_shinfo(skb)->gso_segs;
5966        first->bytecount += (first->gso_segs - 1) * *hdr_len;
5967
5968        /* MSS L4LEN IDX */
5969        mss_l4len_idx = (*hdr_len - l4_offset) << E1000_ADVTXD_L4LEN_SHIFT;
5970        mss_l4len_idx |= skb_shinfo(skb)->gso_size << E1000_ADVTXD_MSS_SHIFT;
5971
5972        /* VLAN MACLEN IPLEN */
5973        vlan_macip_lens = l4.hdr - ip.hdr;
5974        vlan_macip_lens |= (ip.hdr - skb->data) << E1000_ADVTXD_MACLEN_SHIFT;
5975        vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
5976
5977        igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens,
5978                        type_tucmd, mss_l4len_idx);
5979
5980        return 1;
5981}
5982
5983static void igb_tx_csum(struct igb_ring *tx_ring, struct igb_tx_buffer *first)
5984{
5985        struct sk_buff *skb = first->skb;
5986        u32 vlan_macip_lens = 0;
5987        u32 type_tucmd = 0;
5988
5989        if (skb->ip_summed != CHECKSUM_PARTIAL) {
5990csum_failed:
5991                if (!(first->tx_flags & IGB_TX_FLAGS_VLAN) &&
5992                    !tx_ring->launchtime_enable)
5993                        return;
5994                goto no_csum;
5995        }
5996
5997        switch (skb->csum_offset) {
5998        case offsetof(struct tcphdr, check):
5999                type_tucmd = E1000_ADVTXD_TUCMD_L4T_TCP;
6000                fallthrough;
6001        case offsetof(struct udphdr, check):
6002                break;
6003        case offsetof(struct sctphdr, checksum):
6004                /* validate that this is actually an SCTP request */
6005                if (skb_csum_is_sctp(skb)) {
6006                        type_tucmd = E1000_ADVTXD_TUCMD_L4T_SCTP;
6007                        break;
6008                }
6009                fallthrough;
6010        default:
6011                skb_checksum_help(skb);
6012                goto csum_failed;
6013        }
6014
6015        /* update TX checksum flag */
6016        first->tx_flags |= IGB_TX_FLAGS_CSUM;
6017        vlan_macip_lens = skb_checksum_start_offset(skb) -
6018                          skb_network_offset(skb);
6019no_csum:
6020        vlan_macip_lens |= skb_network_offset(skb) << E1000_ADVTXD_MACLEN_SHIFT;
6021        vlan_macip_lens |= first->tx_flags & IGB_TX_FLAGS_VLAN_MASK;
6022
6023        igb_tx_ctxtdesc(tx_ring, first, vlan_macip_lens, type_tucmd, 0);
6024}
6025
6026#define IGB_SET_FLAG(_input, _flag, _result) \
6027        ((_flag <= _result) ? \
6028         ((u32)(_input & _flag) * (_result / _flag)) : \
6029         ((u32)(_input & _flag) / (_flag / _result)))
6030
6031static u32 igb_tx_cmd_type(struct sk_buff *skb, u32 tx_flags)
6032{
6033        /* set type for advanced descriptor with frame checksum insertion */
6034        u32 cmd_type = E1000_ADVTXD_DTYP_DATA |
6035                       E1000_ADVTXD_DCMD_DEXT |
6036                       E1000_ADVTXD_DCMD_IFCS;
6037
6038        /* set HW vlan bit if vlan is present */
6039        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_VLAN,
6040                                 (E1000_ADVTXD_DCMD_VLE));
6041
6042        /* set segmentation bits for TSO */
6043        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSO,
6044                                 (E1000_ADVTXD_DCMD_TSE));
6045
6046        /* set timestamp bit if present */
6047        cmd_type |= IGB_SET_FLAG(tx_flags, IGB_TX_FLAGS_TSTAMP,
6048                                 (E1000_ADVTXD_MAC_TSTAMP));
6049
6050        /* insert frame checksum */
6051        cmd_type ^= IGB_SET_FLAG(skb->no_fcs, 1, E1000_ADVTXD_DCMD_IFCS);
6052
6053        return cmd_type;
6054}
6055
6056static void igb_tx_olinfo_status(struct igb_ring *tx_ring,
6057                                 union e1000_adv_tx_desc *tx_desc,
6058                                 u32 tx_flags, unsigned int paylen)
6059{
6060        u32 olinfo_status = paylen << E1000_ADVTXD_PAYLEN_SHIFT;
6061
6062        /* 82575 requires a unique index per ring */
6063        if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6064                olinfo_status |= tx_ring->reg_idx << 4;
6065
6066        /* insert L4 checksum */
6067        olinfo_status |= IGB_SET_FLAG(tx_flags,
6068                                      IGB_TX_FLAGS_CSUM,
6069                                      (E1000_TXD_POPTS_TXSM << 8));
6070
6071        /* insert IPv4 checksum */
6072        olinfo_status |= IGB_SET_FLAG(tx_flags,
6073                                      IGB_TX_FLAGS_IPV4,
6074                                      (E1000_TXD_POPTS_IXSM << 8));
6075
6076        tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6077}
6078
6079static int __igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6080{
6081        struct net_device *netdev = tx_ring->netdev;
6082
6083        netif_stop_subqueue(netdev, tx_ring->queue_index);
6084
6085        /* Herbert's original patch had:
6086         *  smp_mb__after_netif_stop_queue();
6087         * but since that doesn't exist yet, just open code it.
6088         */
6089        smp_mb();
6090
6091        /* We need to check again in a case another CPU has just
6092         * made room available.
6093         */
6094        if (igb_desc_unused(tx_ring) < size)
6095                return -EBUSY;
6096
6097        /* A reprieve! */
6098        netif_wake_subqueue(netdev, tx_ring->queue_index);
6099
6100        u64_stats_update_begin(&tx_ring->tx_syncp2);
6101        tx_ring->tx_stats.restart_queue2++;
6102        u64_stats_update_end(&tx_ring->tx_syncp2);
6103
6104        return 0;
6105}
6106
6107static inline int igb_maybe_stop_tx(struct igb_ring *tx_ring, const u16 size)
6108{
6109        if (igb_desc_unused(tx_ring) >= size)
6110                return 0;
6111        return __igb_maybe_stop_tx(tx_ring, size);
6112}
6113
6114static int igb_tx_map(struct igb_ring *tx_ring,
6115                      struct igb_tx_buffer *first,
6116                      const u8 hdr_len)
6117{
6118        struct sk_buff *skb = first->skb;
6119        struct igb_tx_buffer *tx_buffer;
6120        union e1000_adv_tx_desc *tx_desc;
6121        skb_frag_t *frag;
6122        dma_addr_t dma;
6123        unsigned int data_len, size;
6124        u32 tx_flags = first->tx_flags;
6125        u32 cmd_type = igb_tx_cmd_type(skb, tx_flags);
6126        u16 i = tx_ring->next_to_use;
6127
6128        tx_desc = IGB_TX_DESC(tx_ring, i);
6129
6130        igb_tx_olinfo_status(tx_ring, tx_desc, tx_flags, skb->len - hdr_len);
6131
6132        size = skb_headlen(skb);
6133        data_len = skb->data_len;
6134
6135        dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
6136
6137        tx_buffer = first;
6138
6139        for (frag = &skb_shinfo(skb)->frags[0];; frag++) {
6140                if (dma_mapping_error(tx_ring->dev, dma))
6141                        goto dma_error;
6142
6143                /* record length, and DMA address */
6144                dma_unmap_len_set(tx_buffer, len, size);
6145                dma_unmap_addr_set(tx_buffer, dma, dma);
6146
6147                tx_desc->read.buffer_addr = cpu_to_le64(dma);
6148
6149                while (unlikely(size > IGB_MAX_DATA_PER_TXD)) {
6150                        tx_desc->read.cmd_type_len =
6151                                cpu_to_le32(cmd_type ^ IGB_MAX_DATA_PER_TXD);
6152
6153                        i++;
6154                        tx_desc++;
6155                        if (i == tx_ring->count) {
6156                                tx_desc = IGB_TX_DESC(tx_ring, 0);
6157                                i = 0;
6158                        }
6159                        tx_desc->read.olinfo_status = 0;
6160
6161                        dma += IGB_MAX_DATA_PER_TXD;
6162                        size -= IGB_MAX_DATA_PER_TXD;
6163
6164                        tx_desc->read.buffer_addr = cpu_to_le64(dma);
6165                }
6166
6167                if (likely(!data_len))
6168                        break;
6169
6170                tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type ^ size);
6171
6172                i++;
6173                tx_desc++;
6174                if (i == tx_ring->count) {
6175                        tx_desc = IGB_TX_DESC(tx_ring, 0);
6176                        i = 0;
6177                }
6178                tx_desc->read.olinfo_status = 0;
6179
6180                size = skb_frag_size(frag);
6181                data_len -= size;
6182
6183                dma = skb_frag_dma_map(tx_ring->dev, frag, 0,
6184                                       size, DMA_TO_DEVICE);
6185
6186                tx_buffer = &tx_ring->tx_buffer_info[i];
6187        }
6188
6189        /* write last descriptor with RS and EOP bits */
6190        cmd_type |= size | IGB_TXD_DCMD;
6191        tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6192
6193        netdev_tx_sent_queue(txring_txq(tx_ring), first->bytecount);
6194
6195        /* set the timestamp */
6196        first->time_stamp = jiffies;
6197
6198        skb_tx_timestamp(skb);
6199
6200        /* Force memory writes to complete before letting h/w know there
6201         * are new descriptors to fetch.  (Only applicable for weak-ordered
6202         * memory model archs, such as IA-64).
6203         *
6204         * We also need this memory barrier to make certain all of the
6205         * status bits have been updated before next_to_watch is written.
6206         */
6207        dma_wmb();
6208
6209        /* set next_to_watch value indicating a packet is present */
6210        first->next_to_watch = tx_desc;
6211
6212        i++;
6213        if (i == tx_ring->count)
6214                i = 0;
6215
6216        tx_ring->next_to_use = i;
6217
6218        /* Make sure there is space in the ring for the next send. */
6219        igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6220
6221        if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more()) {
6222                writel(i, tx_ring->tail);
6223        }
6224        return 0;
6225
6226dma_error:
6227        dev_err(tx_ring->dev, "TX DMA map failed\n");
6228        tx_buffer = &tx_ring->tx_buffer_info[i];
6229
6230        /* clear dma mappings for failed tx_buffer_info map */
6231        while (tx_buffer != first) {
6232                if (dma_unmap_len(tx_buffer, len))
6233                        dma_unmap_page(tx_ring->dev,
6234                                       dma_unmap_addr(tx_buffer, dma),
6235                                       dma_unmap_len(tx_buffer, len),
6236                                       DMA_TO_DEVICE);
6237                dma_unmap_len_set(tx_buffer, len, 0);
6238
6239                if (i-- == 0)
6240                        i += tx_ring->count;
6241                tx_buffer = &tx_ring->tx_buffer_info[i];
6242        }
6243
6244        if (dma_unmap_len(tx_buffer, len))
6245                dma_unmap_single(tx_ring->dev,
6246                                 dma_unmap_addr(tx_buffer, dma),
6247                                 dma_unmap_len(tx_buffer, len),
6248                                 DMA_TO_DEVICE);
6249        dma_unmap_len_set(tx_buffer, len, 0);
6250
6251        dev_kfree_skb_any(tx_buffer->skb);
6252        tx_buffer->skb = NULL;
6253
6254        tx_ring->next_to_use = i;
6255
6256        return -1;
6257}
6258
6259int igb_xmit_xdp_ring(struct igb_adapter *adapter,
6260                      struct igb_ring *tx_ring,
6261                      struct xdp_frame *xdpf)
6262{
6263        union e1000_adv_tx_desc *tx_desc;
6264        u32 len, cmd_type, olinfo_status;
6265        struct igb_tx_buffer *tx_buffer;
6266        dma_addr_t dma;
6267        u16 i;
6268
6269        len = xdpf->len;
6270
6271        if (unlikely(!igb_desc_unused(tx_ring)))
6272                return IGB_XDP_CONSUMED;
6273
6274        dma = dma_map_single(tx_ring->dev, xdpf->data, len, DMA_TO_DEVICE);
6275        if (dma_mapping_error(tx_ring->dev, dma))
6276                return IGB_XDP_CONSUMED;
6277
6278        /* record the location of the first descriptor for this packet */
6279        tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6280        tx_buffer->bytecount = len;
6281        tx_buffer->gso_segs = 1;
6282        tx_buffer->protocol = 0;
6283
6284        i = tx_ring->next_to_use;
6285        tx_desc = IGB_TX_DESC(tx_ring, i);
6286
6287        dma_unmap_len_set(tx_buffer, len, len);
6288        dma_unmap_addr_set(tx_buffer, dma, dma);
6289        tx_buffer->type = IGB_TYPE_XDP;
6290        tx_buffer->xdpf = xdpf;
6291
6292        tx_desc->read.buffer_addr = cpu_to_le64(dma);
6293
6294        /* put descriptor type bits */
6295        cmd_type = E1000_ADVTXD_DTYP_DATA |
6296                   E1000_ADVTXD_DCMD_DEXT |
6297                   E1000_ADVTXD_DCMD_IFCS;
6298        cmd_type |= len | IGB_TXD_DCMD;
6299        tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
6300
6301        olinfo_status = len << E1000_ADVTXD_PAYLEN_SHIFT;
6302        /* 82575 requires a unique index per ring */
6303        if (test_bit(IGB_RING_FLAG_TX_CTX_IDX, &tx_ring->flags))
6304                olinfo_status |= tx_ring->reg_idx << 4;
6305
6306        tx_desc->read.olinfo_status = cpu_to_le32(olinfo_status);
6307
6308        netdev_tx_sent_queue(txring_txq(tx_ring), tx_buffer->bytecount);
6309
6310        /* set the timestamp */
6311        tx_buffer->time_stamp = jiffies;
6312
6313        /* Avoid any potential race with xdp_xmit and cleanup */
6314        smp_wmb();
6315
6316        /* set next_to_watch value indicating a packet is present */
6317        i++;
6318        if (i == tx_ring->count)
6319                i = 0;
6320
6321        tx_buffer->next_to_watch = tx_desc;
6322        tx_ring->next_to_use = i;
6323
6324        /* Make sure there is space in the ring for the next send. */
6325        igb_maybe_stop_tx(tx_ring, DESC_NEEDED);
6326
6327        if (netif_xmit_stopped(txring_txq(tx_ring)) || !netdev_xmit_more())
6328                writel(i, tx_ring->tail);
6329
6330        return IGB_XDP_TX;
6331}
6332
6333netdev_tx_t igb_xmit_frame_ring(struct sk_buff *skb,
6334                                struct igb_ring *tx_ring)
6335{
6336        struct igb_tx_buffer *first;
6337        int tso;
6338        u32 tx_flags = 0;
6339        unsigned short f;
6340        u16 count = TXD_USE_COUNT(skb_headlen(skb));
6341        __be16 protocol = vlan_get_protocol(skb);
6342        u8 hdr_len = 0;
6343
6344        /* need: 1 descriptor per page * PAGE_SIZE/IGB_MAX_DATA_PER_TXD,
6345         *       + 1 desc for skb_headlen/IGB_MAX_DATA_PER_TXD,
6346         *       + 2 desc gap to keep tail from touching head,
6347         *       + 1 desc for context descriptor,
6348         * otherwise try next time
6349         */
6350        for (f = 0; f < skb_shinfo(skb)->nr_frags; f++)
6351                count += TXD_USE_COUNT(skb_frag_size(
6352                                                &skb_shinfo(skb)->frags[f]));
6353
6354        if (igb_maybe_stop_tx(tx_ring, count + 3)) {
6355                /* this is a hard error */
6356                return NETDEV_TX_BUSY;
6357        }
6358
6359        /* record the location of the first descriptor for this packet */
6360        first = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
6361        first->type = IGB_TYPE_SKB;
6362        first->skb = skb;
6363        first->bytecount = skb->len;
6364        first->gso_segs = 1;
6365
6366        if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) {
6367                struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6368
6369                if (adapter->tstamp_config.tx_type == HWTSTAMP_TX_ON &&
6370                    !test_and_set_bit_lock(__IGB_PTP_TX_IN_PROGRESS,
6371                                           &adapter->state)) {
6372                        skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
6373                        tx_flags |= IGB_TX_FLAGS_TSTAMP;
6374
6375                        adapter->ptp_tx_skb = skb_get(skb);
6376                        adapter->ptp_tx_start = jiffies;
6377                        if (adapter->hw.mac.type == e1000_82576)
6378                                schedule_work(&adapter->ptp_tx_work);
6379                } else {
6380                        adapter->tx_hwtstamp_skipped++;
6381                }
6382        }
6383
6384        if (skb_vlan_tag_present(skb)) {
6385                tx_flags |= IGB_TX_FLAGS_VLAN;
6386                tx_flags |= (skb_vlan_tag_get(skb) << IGB_TX_FLAGS_VLAN_SHIFT);
6387        }
6388
6389        /* record initial flags and protocol */
6390        first->tx_flags = tx_flags;
6391        first->protocol = protocol;
6392
6393        tso = igb_tso(tx_ring, first, &hdr_len);
6394        if (tso < 0)
6395                goto out_drop;
6396        else if (!tso)
6397                igb_tx_csum(tx_ring, first);
6398
6399        if (igb_tx_map(tx_ring, first, hdr_len))
6400                goto cleanup_tx_tstamp;
6401
6402        return NETDEV_TX_OK;
6403
6404out_drop:
6405        dev_kfree_skb_any(first->skb);
6406        first->skb = NULL;
6407cleanup_tx_tstamp:
6408        if (unlikely(tx_flags & IGB_TX_FLAGS_TSTAMP)) {
6409                struct igb_adapter *adapter = netdev_priv(tx_ring->netdev);
6410
6411                dev_kfree_skb_any(adapter->ptp_tx_skb);
6412                adapter->ptp_tx_skb = NULL;
6413                if (adapter->hw.mac.type == e1000_82576)
6414                        cancel_work_sync(&adapter->ptp_tx_work);
6415                clear_bit_unlock(__IGB_PTP_TX_IN_PROGRESS, &adapter->state);
6416        }
6417
6418        return NETDEV_TX_OK;
6419}
6420
6421static inline struct igb_ring *igb_tx_queue_mapping(struct igb_adapter *adapter,
6422                                                    struct sk_buff *skb)
6423{
6424        unsigned int r_idx = skb->queue_mapping;
6425
6426        if (r_idx >= adapter->num_tx_queues)
6427                r_idx = r_idx % adapter->num_tx_queues;
6428
6429        return adapter->tx_ring[r_idx];
6430}
6431
6432static netdev_tx_t igb_xmit_frame(struct sk_buff *skb,
6433                                  struct net_device *netdev)
6434{
6435        struct igb_adapter *adapter = netdev_priv(netdev);
6436
6437        /* The minimum packet size with TCTL.PSP set is 17 so pad the skb
6438         * in order to meet this minimum size requirement.
6439         */
6440        if (skb_put_padto(skb, 17))
6441                return NETDEV_TX_OK;
6442
6443        return igb_xmit_frame_ring(skb, igb_tx_queue_mapping(adapter, skb));
6444}
6445
6446/**
6447 *  igb_tx_timeout - Respond to a Tx Hang
6448 *  @netdev: network interface device structure
6449 *  @txqueue: number of the Tx queue that hung (unused)
6450 **/
6451static void igb_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
6452{
6453        struct igb_adapter *adapter = netdev_priv(netdev);
6454        struct e1000_hw *hw = &adapter->hw;
6455
6456        /* Do the reset outside of interrupt context */
6457        adapter->tx_timeout_count++;
6458
6459        if (hw->mac.type >= e1000_82580)
6460                hw->dev_spec._82575.global_device_reset = true;
6461
6462        schedule_work(&adapter->reset_task);
6463        wr32(E1000_EICS,
6464             (adapter->eims_enable_mask & ~adapter->eims_other));
6465}
6466
6467static void igb_reset_task(struct work_struct *work)
6468{
6469        struct igb_adapter *adapter;
6470        adapter = container_of(work, struct igb_adapter, reset_task);
6471
6472        rtnl_lock();
6473        /* If we're already down or resetting, just bail */
6474        if (test_bit(__IGB_DOWN, &adapter->state) ||
6475            test_bit(__IGB_RESETTING, &adapter->state)) {
6476                rtnl_unlock();
6477                return;
6478        }
6479
6480        igb_dump(adapter);
6481        netdev_err(adapter->netdev, "Reset adapter\n");
6482        igb_reinit_locked(adapter);
6483        rtnl_unlock();
6484}
6485
6486/**
6487 *  igb_get_stats64 - Get System Network Statistics
6488 *  @netdev: network interface device structure
6489 *  @stats: rtnl_link_stats64 pointer
6490 **/
6491static void igb_get_stats64(struct net_device *netdev,
6492                            struct rtnl_link_stats64 *stats)
6493{
6494        struct igb_adapter *adapter = netdev_priv(netdev);
6495
6496        spin_lock(&adapter->stats64_lock);
6497        igb_update_stats(adapter);
6498        memcpy(stats, &adapter->stats64, sizeof(*stats));
6499        spin_unlock(&adapter->stats64_lock);
6500}
6501
6502/**
6503 *  igb_change_mtu - Change the Maximum Transfer Unit
6504 *  @netdev: network interface device structure
6505 *  @new_mtu: new value for maximum frame size
6506 *
6507 *  Returns 0 on success, negative on failure
6508 **/
6509static int igb_change_mtu(struct net_device *netdev, int new_mtu)
6510{
6511        struct igb_adapter *adapter = netdev_priv(netdev);
6512        int max_frame = new_mtu + IGB_ETH_PKT_HDR_PAD;
6513
6514        if (adapter->xdp_prog) {
6515                int i;
6516
6517                for (i = 0; i < adapter->num_rx_queues; i++) {
6518                        struct igb_ring *ring = adapter->rx_ring[i];
6519
6520                        if (max_frame > igb_rx_bufsz(ring)) {
6521                                netdev_warn(adapter->netdev,
6522                                            "Requested MTU size is not supported with XDP. Max frame size is %d\n",
6523                                            max_frame);
6524                                return -EINVAL;
6525                        }
6526                }
6527        }
6528
6529        /* adjust max frame to be at least the size of a standard frame */
6530        if (max_frame < (ETH_FRAME_LEN + ETH_FCS_LEN))
6531                max_frame = ETH_FRAME_LEN + ETH_FCS_LEN;
6532
6533        while (test_and_set_bit(__IGB_RESETTING, &adapter->state))
6534                usleep_range(1000, 2000);
6535
6536        /* igb_down has a dependency on max_frame_size */
6537        adapter->max_frame_size = max_frame;
6538
6539        if (netif_running(netdev))
6540                igb_down(adapter);
6541
6542        netdev_dbg(netdev, "changing MTU from %d to %d\n",
6543                   netdev->mtu, new_mtu);
6544        netdev->mtu = new_mtu;
6545
6546        if (netif_running(netdev))
6547                igb_up(adapter);
6548        else
6549                igb_reset(adapter);
6550
6551        clear_bit(__IGB_RESETTING, &adapter->state);
6552
6553        return 0;
6554}
6555
6556/**
6557 *  igb_update_stats - Update the board statistics counters
6558 *  @adapter: board private structure
6559 **/
6560void igb_update_stats(struct igb_adapter *adapter)
6561{
6562        struct rtnl_link_stats64 *net_stats = &adapter->stats64;
6563        struct e1000_hw *hw = &adapter->hw;
6564        struct pci_dev *pdev = adapter->pdev;
6565        u32 reg, mpc;
6566        int i;
6567        u64 bytes, packets;
6568        unsigned int start;
6569        u64 _bytes, _packets;
6570
6571        /* Prevent stats update while adapter is being reset, or if the pci
6572         * connection is down.
6573         */
6574        if (adapter->link_speed == 0)
6575                return;
6576        if (pci_channel_offline(pdev))
6577                return;
6578
6579        bytes = 0;
6580        packets = 0;
6581
6582        rcu_read_lock();
6583        for (i = 0; i < adapter->num_rx_queues; i++) {
6584                struct igb_ring *ring = adapter->rx_ring[i];
6585                u32 rqdpc = rd32(E1000_RQDPC(i));
6586                if (hw->mac.type >= e1000_i210)
6587                        wr32(E1000_RQDPC(i), 0);
6588
6589                if (rqdpc) {
6590                        ring->rx_stats.drops += rqdpc;
6591                        net_stats->rx_fifo_errors += rqdpc;
6592                }
6593
6594                do {
6595                        start = u64_stats_fetch_begin_irq(&ring->rx_syncp);
6596                        _bytes = ring->rx_stats.bytes;
6597                        _packets = ring->rx_stats.packets;
6598                } while (u64_stats_fetch_retry_irq(&ring->rx_syncp, start));
6599                bytes += _bytes;
6600                packets += _packets;
6601        }
6602
6603        net_stats->rx_bytes = bytes;
6604        net_stats->rx_packets = packets;
6605
6606        bytes = 0;
6607        packets = 0;
6608        for (i = 0; i < adapter->num_tx_queues; i++) {
6609                struct igb_ring *ring = adapter->tx_ring[i];
6610                do {
6611                        start = u64_stats_fetch_begin_irq(&ring->tx_syncp);
6612                        _bytes = ring->tx_stats.bytes;
6613                        _packets = ring->tx_stats.packets;
6614                } while (u64_stats_fetch_retry_irq(&ring->tx_syncp, start));
6615                bytes += _bytes;
6616                packets += _packets;
6617        }
6618        net_stats->tx_bytes = bytes;
6619        net_stats->tx_packets = packets;
6620        rcu_read_unlock();
6621
6622        /* read stats registers */
6623        adapter->stats.crcerrs += rd32(E1000_CRCERRS);
6624        adapter->stats.gprc += rd32(E1000_GPRC);
6625        adapter->stats.gorc += rd32(E1000_GORCL);
6626        rd32(E1000_GORCH); /* clear GORCL */
6627        adapter->stats.bprc += rd32(E1000_BPRC);
6628        adapter->stats.mprc += rd32(E1000_MPRC);
6629        adapter->stats.roc += rd32(E1000_ROC);
6630
6631        adapter->stats.prc64 += rd32(E1000_PRC64);
6632        adapter->stats.prc127 += rd32(E1000_PRC127);
6633        adapter->stats.prc255 += rd32(E1000_PRC255);
6634        adapter->stats.prc511 += rd32(E1000_PRC511);
6635        adapter->stats.prc1023 += rd32(E1000_PRC1023);
6636        adapter->stats.prc1522 += rd32(E1000_PRC1522);
6637        adapter->stats.symerrs += rd32(E1000_SYMERRS);
6638        adapter->stats.sec += rd32(E1000_SEC);
6639
6640        mpc = rd32(E1000_MPC);
6641        adapter->stats.mpc += mpc;
6642        net_stats->rx_fifo_errors += mpc;
6643        adapter->stats.scc += rd32(E1000_SCC);
6644        adapter->stats.ecol += rd32(E1000_ECOL);
6645        adapter->stats.mcc += rd32(E1000_MCC);
6646        adapter->stats.latecol += rd32(E1000_LATECOL);
6647        adapter->stats.dc += rd32(E1000_DC);
6648        adapter->stats.rlec += rd32(E1000_RLEC);
6649        adapter->stats.xonrxc += rd32(E1000_XONRXC);
6650        adapter->stats.xontxc += rd32(E1000_XONTXC);
6651        adapter->stats.xoffrxc += rd32(E1000_XOFFRXC);
6652        adapter->stats.xofftxc += rd32(E1000_XOFFTXC);
6653        adapter->stats.fcruc += rd32(E1000_FCRUC);
6654        adapter->stats.gptc += rd32(E1000_GPTC);
6655        adapter->stats.gotc += rd32(E1000_GOTCL);
6656        rd32(E1000_GOTCH); /* clear GOTCL */
6657        adapter->stats.rnbc += rd32(E1000_RNBC);
6658        adapter->stats.ruc += rd32(E1000_RUC);
6659        adapter->stats.rfc += rd32(E1000_RFC);
6660        adapter->stats.rjc += rd32(E1000_RJC);
6661        adapter->stats.tor += rd32(E1000_TORH);
6662        adapter->stats.tot += rd32(E1000_TOTH);
6663        adapter->stats.tpr += rd32(E1000_TPR);
6664
6665        adapter->stats.ptc64 += rd32(E1000_PTC64);
6666        adapter->stats.ptc127 += rd32(E1000_PTC127);
6667        adapter->stats.ptc255 += rd32(E1000_PTC255);
6668        adapter->stats.ptc511 += rd32(E1000_PTC511);
6669        adapter->stats.ptc1023 += rd32(E1000_PTC1023);
6670        adapter->stats.ptc1522 += rd32(E1000_PTC1522);
6671
6672        adapter->stats.mptc += rd32(E1000_MPTC);
6673        adapter->stats.bptc += rd32(E1000_BPTC);
6674
6675        adapter->stats.tpt += rd32(E1000_TPT);
6676        adapter->stats.colc += rd32(E1000_COLC);
6677
6678        adapter->stats.algnerrc += rd32(E1000_ALGNERRC);
6679        /* read internal phy specific stats */
6680        reg = rd32(E1000_CTRL_EXT);
6681        if (!(reg & E1000_CTRL_EXT_LINK_MODE_MASK)) {
6682                adapter->stats.rxerrc += rd32(E1000_RXERRC);
6683
6684                /* this stat has invalid values on i210/i211 */
6685                if ((hw->mac.type != e1000_i210) &&
6686                    (hw->mac.type != e1000_i211))
6687                        adapter->stats.tncrs += rd32(E1000_TNCRS);
6688        }
6689
6690        adapter->stats.tsctc += rd32(E1000_TSCTC);
6691        adapter->stats.tsctfc += rd32(E1000_TSCTFC);
6692
6693        adapter->stats.iac += rd32(E1000_IAC);
6694        adapter->stats.icrxoc += rd32(E1000_ICRXOC);
6695        adapter->stats.icrxptc += rd32(E1000_ICRXPTC);
6696        adapter->stats.icrxatc += rd32(E1000_ICRXATC);
6697        adapter->stats.ictxptc += rd32(E1000_ICTXPTC);
6698        adapter->stats.ictxatc += rd32(E1000_ICTXATC);
6699        adapter->stats.ictxqec += rd32(E1000_ICTXQEC);
6700        adapter->stats.ictxqmtc += rd32(E1000_ICTXQMTC);
6701        adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
6702
6703        /* Fill out the OS statistics structure */
6704        net_stats->multicast = adapter->stats.mprc;
6705        net_stats->collisions = adapter->stats.colc;
6706
6707        /* Rx Errors */
6708
6709        /* RLEC on some newer hardware can be incorrect so build
6710         * our own version based on RUC and ROC
6711         */
6712        net_stats->rx_errors = adapter->stats.rxerrc +
6713                adapter->stats.crcerrs + adapter->stats.algnerrc +
6714                adapter->stats.ruc + adapter->stats.roc +
6715                adapter->stats.cexterr;
6716        net_stats->rx_length_errors = adapter->stats.ruc +
6717                                      adapter->stats.roc;
6718        net_stats->rx_crc_errors = adapter->stats.crcerrs;
6719        net_stats->rx_frame_errors = adapter->stats.algnerrc;
6720        net_stats->rx_missed_errors = adapter->stats.mpc;
6721
6722        /* Tx Errors */
6723        net_stats->tx_errors = adapter->stats.ecol +
6724                               adapter->stats.latecol;
6725        net_stats->tx_aborted_errors = adapter->stats.ecol;
6726        net_stats->tx_window_errors = adapter->stats.latecol;
6727        net_stats->tx_carrier_errors = adapter->stats.tncrs;
6728
6729        /* Tx Dropped needs to be maintained elsewhere */
6730
6731        /* Management Stats */
6732        adapter->stats.mgptc += rd32(E1000_MGTPTC);
6733        adapter->stats.mgprc += rd32(E1000_MGTPRC);
6734        adapter->stats.mgpdc += rd32(E1000_MGTPDC);
6735
6736        /* OS2BMC Stats */
6737        reg = rd32(E1000_MANC);
6738        if (reg & E1000_MANC_EN_BMC2OS) {
6739                adapter->stats.o2bgptc += rd32(E1000_O2BGPTC);
6740                adapter->stats.o2bspc += rd32(E1000_O2BSPC);
6741                adapter->stats.b2ospc += rd32(E1000_B2OSPC);
6742                adapter->stats.b2ogprc += rd32(E1000_B2OGPRC);
6743        }
6744}
6745
6746static void igb_perout(struct igb_adapter *adapter, int tsintr_tt)
6747{
6748        int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_PEROUT, tsintr_tt);
6749        struct e1000_hw *hw = &adapter->hw;
6750        struct timespec64 ts;
6751        u32 tsauxc;
6752
6753        if (pin < 0 || pin >= IGB_N_PEROUT)
6754                return;
6755
6756        spin_lock(&adapter->tmreg_lock);
6757
6758        if (hw->mac.type == e1000_82580 ||
6759            hw->mac.type == e1000_i354 ||
6760            hw->mac.type == e1000_i350) {
6761                s64 ns = timespec64_to_ns(&adapter->perout[pin].period);
6762                u32 systiml, systimh, level_mask, level, rem;
6763                u64 systim, now;
6764
6765                /* read systim registers in sequence */
6766                rd32(E1000_SYSTIMR);
6767                systiml = rd32(E1000_SYSTIML);
6768                systimh = rd32(E1000_SYSTIMH);
6769                systim = (((u64)(systimh & 0xFF)) << 32) | ((u64)systiml);
6770                now = timecounter_cyc2time(&adapter->tc, systim);
6771
6772                if (pin < 2) {
6773                        level_mask = (tsintr_tt == 1) ? 0x80000 : 0x40000;
6774                        level = (rd32(E1000_CTRL) & level_mask) ? 1 : 0;
6775                } else {
6776                        level_mask = (tsintr_tt == 1) ? 0x80 : 0x40;
6777                        level = (rd32(E1000_CTRL_EXT) & level_mask) ? 1 : 0;
6778                }
6779
6780                div_u64_rem(now, ns, &rem);
6781                systim = systim + (ns - rem);
6782
6783                /* synchronize pin level with rising/falling edges */
6784                div_u64_rem(now, ns << 1, &rem);
6785                if (rem < ns) {
6786                        /* first half of period */
6787                        if (level == 0) {
6788                                /* output is already low, skip this period */
6789                                systim += ns;
6790                                pr_notice("igb: periodic output on %s missed falling edge\n",
6791                                          adapter->sdp_config[pin].name);
6792                        }
6793                } else {
6794                        /* second half of period */
6795                        if (level == 1) {
6796                                /* output is already high, skip this period */
6797                                systim += ns;
6798                                pr_notice("igb: periodic output on %s missed rising edge\n",
6799                                          adapter->sdp_config[pin].name);
6800                        }
6801                }
6802
6803                /* for this chip family tv_sec is the upper part of the binary value,
6804                 * so not seconds
6805                 */
6806                ts.tv_nsec = (u32)systim;
6807                ts.tv_sec  = ((u32)(systim >> 32)) & 0xFF;
6808        } else {
6809                ts = timespec64_add(adapter->perout[pin].start,
6810                                    adapter->perout[pin].period);
6811        }
6812
6813        /* u32 conversion of tv_sec is safe until y2106 */
6814        wr32((tsintr_tt == 1) ? E1000_TRGTTIML1 : E1000_TRGTTIML0, ts.tv_nsec);
6815        wr32((tsintr_tt == 1) ? E1000_TRGTTIMH1 : E1000_TRGTTIMH0, (u32)ts.tv_sec);
6816        tsauxc = rd32(E1000_TSAUXC);
6817        tsauxc |= TSAUXC_EN_TT0;
6818        wr32(E1000_TSAUXC, tsauxc);
6819        adapter->perout[pin].start = ts;
6820
6821        spin_unlock(&adapter->tmreg_lock);
6822}
6823
6824static void igb_extts(struct igb_adapter *adapter, int tsintr_tt)
6825{
6826        int pin = ptp_find_pin(adapter->ptp_clock, PTP_PF_EXTTS, tsintr_tt);
6827        int auxstmpl = (tsintr_tt == 1) ? E1000_AUXSTMPL1 : E1000_AUXSTMPL0;
6828        int auxstmph = (tsintr_tt == 1) ? E1000_AUXSTMPH1 : E1000_AUXSTMPH0;
6829        struct e1000_hw *hw = &adapter->hw;
6830        struct ptp_clock_event event;
6831        struct timespec64 ts;
6832
6833        if (pin < 0 || pin >= IGB_N_EXTTS)
6834                return;
6835
6836        if (hw->mac.type == e1000_82580 ||
6837            hw->mac.type == e1000_i354 ||
6838            hw->mac.type == e1000_i350) {
6839                s64 ns = rd32(auxstmpl);
6840
6841                ns += ((s64)(rd32(auxstmph) & 0xFF)) << 32;
6842                ts = ns_to_timespec64(ns);
6843        } else {
6844                ts.tv_nsec = rd32(auxstmpl);
6845                ts.tv_sec  = rd32(auxstmph);
6846        }
6847
6848        event.type = PTP_CLOCK_EXTTS;
6849        event.index = tsintr_tt;
6850        event.timestamp = ts.tv_sec * 1000000000ULL + ts.tv_nsec;
6851        ptp_clock_event(adapter->ptp_clock, &event);
6852}
6853
6854static void igb_tsync_interrupt(struct igb_adapter *adapter)
6855{
6856        struct e1000_hw *hw = &adapter->hw;
6857        u32 ack = 0, tsicr = rd32(E1000_TSICR);
6858        struct ptp_clock_event event;
6859
6860        if (tsicr & TSINTR_SYS_WRAP) {
6861                event.type = PTP_CLOCK_PPS;
6862                if (adapter->ptp_caps.pps)
6863                        ptp_clock_event(adapter->ptp_clock, &event);
6864                ack |= TSINTR_SYS_WRAP;
6865        }
6866
6867        if (tsicr & E1000_TSICR_TXTS) {
6868                /* retrieve hardware timestamp */
6869                schedule_work(&adapter->ptp_tx_work);
6870                ack |= E1000_TSICR_TXTS;
6871        }
6872
6873        if (tsicr & TSINTR_TT0) {
6874                igb_perout(adapter, 0);
6875                ack |= TSINTR_TT0;
6876        }
6877
6878        if (tsicr & TSINTR_TT1) {
6879                igb_perout(adapter, 1);
6880                ack |= TSINTR_TT1;
6881        }
6882
6883        if (tsicr & TSINTR_AUTT0) {
6884                igb_extts(adapter, 0);
6885                ack |= TSINTR_AUTT0;
6886        }
6887
6888        if (tsicr & TSINTR_AUTT1) {
6889                igb_extts(adapter, 1);
6890                ack |= TSINTR_AUTT1;
6891        }
6892
6893        /* acknowledge the interrupts */
6894        wr32(E1000_TSICR, ack);
6895}
6896
6897static irqreturn_t igb_msix_other(int irq, void *data)
6898{
6899        struct igb_adapter *adapter = data;
6900        struct e1000_hw *hw = &adapter->hw;
6901        u32 icr = rd32(E1000_ICR);
6902        /* reading ICR causes bit 31 of EICR to be cleared */
6903
6904        if (icr & E1000_ICR_DRSTA)
6905                schedule_work(&adapter->reset_task);
6906
6907        if (icr & E1000_ICR_DOUTSYNC) {
6908                /* HW is reporting DMA is out of sync */
6909                adapter->stats.doosync++;
6910                /* The DMA Out of Sync is also indication of a spoof event
6911                 * in IOV mode. Check the Wrong VM Behavior register to
6912                 * see if it is really a spoof event.
6913                 */
6914                igb_check_wvbr(adapter);
6915        }
6916
6917        /* Check for a mailbox event */
6918        if (icr & E1000_ICR_VMMB)
6919                igb_msg_task(adapter);
6920
6921        if (icr & E1000_ICR_LSC) {
6922                hw->mac.get_link_status = 1;
6923                /* guard against interrupt when we're going down */
6924                if (!test_bit(__IGB_DOWN, &adapter->state))
6925                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
6926        }
6927
6928        if (icr & E1000_ICR_TS)
6929                igb_tsync_interrupt(adapter);
6930
6931        wr32(E1000_EIMS, adapter->eims_other);
6932
6933        return IRQ_HANDLED;
6934}
6935
6936static void igb_write_itr(struct igb_q_vector *q_vector)
6937{
6938        struct igb_adapter *adapter = q_vector->adapter;
6939        u32 itr_val = q_vector->itr_val & 0x7FFC;
6940
6941        if (!q_vector->set_itr)
6942                return;
6943
6944        if (!itr_val)
6945                itr_val = 0x4;
6946
6947        if (adapter->hw.mac.type == e1000_82575)
6948                itr_val |= itr_val << 16;
6949        else
6950                itr_val |= E1000_EITR_CNT_IGNR;
6951
6952        writel(itr_val, q_vector->itr_register);
6953        q_vector->set_itr = 0;
6954}
6955
6956static irqreturn_t igb_msix_ring(int irq, void *data)
6957{
6958        struct igb_q_vector *q_vector = data;
6959
6960        /* Write the ITR value calculated from the previous interrupt. */
6961        igb_write_itr(q_vector);
6962
6963        napi_schedule(&q_vector->napi);
6964
6965        return IRQ_HANDLED;
6966}
6967
6968#ifdef CONFIG_IGB_DCA
6969static void igb_update_tx_dca(struct igb_adapter *adapter,
6970                              struct igb_ring *tx_ring,
6971                              int cpu)
6972{
6973        struct e1000_hw *hw = &adapter->hw;
6974        u32 txctrl = dca3_get_tag(tx_ring->dev, cpu);
6975
6976        if (hw->mac.type != e1000_82575)
6977                txctrl <<= E1000_DCA_TXCTRL_CPUID_SHIFT;
6978
6979        /* We can enable relaxed ordering for reads, but not writes when
6980         * DCA is enabled.  This is due to a known issue in some chipsets
6981         * which will cause the DCA tag to be cleared.
6982         */
6983        txctrl |= E1000_DCA_TXCTRL_DESC_RRO_EN |
6984                  E1000_DCA_TXCTRL_DATA_RRO_EN |
6985                  E1000_DCA_TXCTRL_DESC_DCA_EN;
6986
6987        wr32(E1000_DCA_TXCTRL(tx_ring->reg_idx), txctrl);
6988}
6989
6990static void igb_update_rx_dca(struct igb_adapter *adapter,
6991                              struct igb_ring *rx_ring,
6992                              int cpu)
6993{
6994        struct e1000_hw *hw = &adapter->hw;
6995        u32 rxctrl = dca3_get_tag(&adapter->pdev->dev, cpu);
6996
6997        if (hw->mac.type != e1000_82575)
6998                rxctrl <<= E1000_DCA_RXCTRL_CPUID_SHIFT;
6999
7000        /* We can enable relaxed ordering for reads, but not writes when
7001         * DCA is enabled.  This is due to a known issue in some chipsets
7002         * which will cause the DCA tag to be cleared.
7003         */
7004        rxctrl |= E1000_DCA_RXCTRL_DESC_RRO_EN |
7005                  E1000_DCA_RXCTRL_DESC_DCA_EN;
7006
7007        wr32(E1000_DCA_RXCTRL(rx_ring->reg_idx), rxctrl);
7008}
7009
7010static void igb_update_dca(struct igb_q_vector *q_vector)
7011{
7012        struct igb_adapter *adapter = q_vector->adapter;
7013        int cpu = get_cpu();
7014
7015        if (q_vector->cpu == cpu)
7016                goto out_no_update;
7017
7018        if (q_vector->tx.ring)
7019                igb_update_tx_dca(adapter, q_vector->tx.ring, cpu);
7020
7021        if (q_vector->rx.ring)
7022                igb_update_rx_dca(adapter, q_vector->rx.ring, cpu);
7023
7024        q_vector->cpu = cpu;
7025out_no_update:
7026        put_cpu();
7027}
7028
7029static void igb_setup_dca(struct igb_adapter *adapter)
7030{
7031        struct e1000_hw *hw = &adapter->hw;
7032        int i;
7033
7034        if (!(adapter->flags & IGB_FLAG_DCA_ENABLED))
7035                return;
7036
7037        /* Always use CB2 mode, difference is masked in the CB driver. */
7038        wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_CB2);
7039
7040        for (i = 0; i < adapter->num_q_vectors; i++) {
7041                adapter->q_vector[i]->cpu = -1;
7042                igb_update_dca(adapter->q_vector[i]);
7043        }
7044}
7045
7046static int __igb_notify_dca(struct device *dev, void *data)
7047{
7048        struct net_device *netdev = dev_get_drvdata(dev);
7049        struct igb_adapter *adapter = netdev_priv(netdev);
7050        struct pci_dev *pdev = adapter->pdev;
7051        struct e1000_hw *hw = &adapter->hw;
7052        unsigned long event = *(unsigned long *)data;
7053
7054        switch (event) {
7055        case DCA_PROVIDER_ADD:
7056                /* if already enabled, don't do it again */
7057                if (adapter->flags & IGB_FLAG_DCA_ENABLED)
7058                        break;
7059                if (dca_add_requester(dev) == 0) {
7060                        adapter->flags |= IGB_FLAG_DCA_ENABLED;
7061                        dev_info(&pdev->dev, "DCA enabled\n");
7062                        igb_setup_dca(adapter);
7063                        break;
7064                }
7065                fallthrough; /* since DCA is disabled. */
7066        case DCA_PROVIDER_REMOVE:
7067                if (adapter->flags & IGB_FLAG_DCA_ENABLED) {
7068                        /* without this a class_device is left
7069                         * hanging around in the sysfs model
7070                         */
7071                        dca_remove_requester(dev);
7072                        dev_info(&pdev->dev, "DCA disabled\n");
7073                        adapter->flags &= ~IGB_FLAG_DCA_ENABLED;
7074                        wr32(E1000_DCA_CTRL, E1000_DCA_CTRL_DCA_MODE_DISABLE);
7075                }
7076                break;
7077        }
7078
7079        return 0;
7080}
7081
7082static int igb_notify_dca(struct notifier_block *nb, unsigned long event,
7083                          void *p)
7084{
7085        int ret_val;
7086
7087        ret_val = driver_for_each_device(&igb_driver.driver, NULL, &event,
7088                                         __igb_notify_dca);
7089
7090        return ret_val ? NOTIFY_BAD : NOTIFY_DONE;
7091}
7092#endif /* CONFIG_IGB_DCA */
7093
7094#ifdef CONFIG_PCI_IOV
7095static int igb_vf_configure(struct igb_adapter *adapter, int vf)
7096{
7097        unsigned char mac_addr[ETH_ALEN];
7098
7099        eth_zero_addr(mac_addr);
7100        igb_set_vf_mac(adapter, vf, mac_addr);
7101
7102        /* By default spoof check is enabled for all VFs */
7103        adapter->vf_data[vf].spoofchk_enabled = true;
7104
7105        /* By default VFs are not trusted */
7106        adapter->vf_data[vf].trusted = false;
7107
7108        return 0;
7109}
7110
7111#endif
7112static void igb_ping_all_vfs(struct igb_adapter *adapter)
7113{
7114        struct e1000_hw *hw = &adapter->hw;
7115        u32 ping;
7116        int i;
7117
7118        for (i = 0 ; i < adapter->vfs_allocated_count; i++) {
7119                ping = E1000_PF_CONTROL_MSG;
7120                if (adapter->vf_data[i].flags & IGB_VF_FLAG_CTS)
7121                        ping |= E1000_VT_MSGTYPE_CTS;
7122                igb_write_mbx(hw, &ping, 1, i);
7123        }
7124}
7125
7126static int igb_set_vf_promisc(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7127{
7128        struct e1000_hw *hw = &adapter->hw;
7129        u32 vmolr = rd32(E1000_VMOLR(vf));
7130        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7131
7132        vf_data->flags &= ~(IGB_VF_FLAG_UNI_PROMISC |
7133                            IGB_VF_FLAG_MULTI_PROMISC);
7134        vmolr &= ~(E1000_VMOLR_ROPE | E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7135
7136        if (*msgbuf & E1000_VF_SET_PROMISC_MULTICAST) {
7137                vmolr |= E1000_VMOLR_MPME;
7138                vf_data->flags |= IGB_VF_FLAG_MULTI_PROMISC;
7139                *msgbuf &= ~E1000_VF_SET_PROMISC_MULTICAST;
7140        } else {
7141                /* if we have hashes and we are clearing a multicast promisc
7142                 * flag we need to write the hashes to the MTA as this step
7143                 * was previously skipped
7144                 */
7145                if (vf_data->num_vf_mc_hashes > 30) {
7146                        vmolr |= E1000_VMOLR_MPME;
7147                } else if (vf_data->num_vf_mc_hashes) {
7148                        int j;
7149
7150                        vmolr |= E1000_VMOLR_ROMPE;
7151                        for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7152                                igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7153                }
7154        }
7155
7156        wr32(E1000_VMOLR(vf), vmolr);
7157
7158        /* there are flags left unprocessed, likely not supported */
7159        if (*msgbuf & E1000_VT_MSGINFO_MASK)
7160                return -EINVAL;
7161
7162        return 0;
7163}
7164
7165static int igb_set_vf_multicasts(struct igb_adapter *adapter,
7166                                  u32 *msgbuf, u32 vf)
7167{
7168        int n = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7169        u16 *hash_list = (u16 *)&msgbuf[1];
7170        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7171        int i;
7172
7173        /* salt away the number of multicast addresses assigned
7174         * to this VF for later use to restore when the PF multi cast
7175         * list changes
7176         */
7177        vf_data->num_vf_mc_hashes = n;
7178
7179        /* only up to 30 hash values supported */
7180        if (n > 30)
7181                n = 30;
7182
7183        /* store the hashes for later use */
7184        for (i = 0; i < n; i++)
7185                vf_data->vf_mc_hashes[i] = hash_list[i];
7186
7187        /* Flush and reset the mta with the new values */
7188        igb_set_rx_mode(adapter->netdev);
7189
7190        return 0;
7191}
7192
7193static void igb_restore_vf_multicasts(struct igb_adapter *adapter)
7194{
7195        struct e1000_hw *hw = &adapter->hw;
7196        struct vf_data_storage *vf_data;
7197        int i, j;
7198
7199        for (i = 0; i < adapter->vfs_allocated_count; i++) {
7200                u32 vmolr = rd32(E1000_VMOLR(i));
7201
7202                vmolr &= ~(E1000_VMOLR_ROMPE | E1000_VMOLR_MPME);
7203
7204                vf_data = &adapter->vf_data[i];
7205
7206                if ((vf_data->num_vf_mc_hashes > 30) ||
7207                    (vf_data->flags & IGB_VF_FLAG_MULTI_PROMISC)) {
7208                        vmolr |= E1000_VMOLR_MPME;
7209                } else if (vf_data->num_vf_mc_hashes) {
7210                        vmolr |= E1000_VMOLR_ROMPE;
7211                        for (j = 0; j < vf_data->num_vf_mc_hashes; j++)
7212                                igb_mta_set(hw, vf_data->vf_mc_hashes[j]);
7213                }
7214                wr32(E1000_VMOLR(i), vmolr);
7215        }
7216}
7217
7218static void igb_clear_vf_vfta(struct igb_adapter *adapter, u32 vf)
7219{
7220        struct e1000_hw *hw = &adapter->hw;
7221        u32 pool_mask, vlvf_mask, i;
7222
7223        /* create mask for VF and other pools */
7224        pool_mask = E1000_VLVF_POOLSEL_MASK;
7225        vlvf_mask = BIT(E1000_VLVF_POOLSEL_SHIFT + vf);
7226
7227        /* drop PF from pool bits */
7228        pool_mask &= ~BIT(E1000_VLVF_POOLSEL_SHIFT +
7229                             adapter->vfs_allocated_count);
7230
7231        /* Find the vlan filter for this id */
7232        for (i = E1000_VLVF_ARRAY_SIZE; i--;) {
7233                u32 vlvf = rd32(E1000_VLVF(i));
7234                u32 vfta_mask, vid, vfta;
7235
7236                /* remove the vf from the pool */
7237                if (!(vlvf & vlvf_mask))
7238                        continue;
7239
7240                /* clear out bit from VLVF */
7241                vlvf ^= vlvf_mask;
7242
7243                /* if other pools are present, just remove ourselves */
7244                if (vlvf & pool_mask)
7245                        goto update_vlvfb;
7246
7247                /* if PF is present, leave VFTA */
7248                if (vlvf & E1000_VLVF_POOLSEL_MASK)
7249                        goto update_vlvf;
7250
7251                vid = vlvf & E1000_VLVF_VLANID_MASK;
7252                vfta_mask = BIT(vid % 32);
7253
7254                /* clear bit from VFTA */
7255                vfta = adapter->shadow_vfta[vid / 32];
7256                if (vfta & vfta_mask)
7257                        hw->mac.ops.write_vfta(hw, vid / 32, vfta ^ vfta_mask);
7258update_vlvf:
7259                /* clear pool selection enable */
7260                if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7261                        vlvf &= E1000_VLVF_POOLSEL_MASK;
7262                else
7263                        vlvf = 0;
7264update_vlvfb:
7265                /* clear pool bits */
7266                wr32(E1000_VLVF(i), vlvf);
7267        }
7268}
7269
7270static int igb_find_vlvf_entry(struct e1000_hw *hw, u32 vlan)
7271{
7272        u32 vlvf;
7273        int idx;
7274
7275        /* short cut the special case */
7276        if (vlan == 0)
7277                return 0;
7278
7279        /* Search for the VLAN id in the VLVF entries */
7280        for (idx = E1000_VLVF_ARRAY_SIZE; --idx;) {
7281                vlvf = rd32(E1000_VLVF(idx));
7282                if ((vlvf & VLAN_VID_MASK) == vlan)
7283                        break;
7284        }
7285
7286        return idx;
7287}
7288
7289static void igb_update_pf_vlvf(struct igb_adapter *adapter, u32 vid)
7290{
7291        struct e1000_hw *hw = &adapter->hw;
7292        u32 bits, pf_id;
7293        int idx;
7294
7295        idx = igb_find_vlvf_entry(hw, vid);
7296        if (!idx)
7297                return;
7298
7299        /* See if any other pools are set for this VLAN filter
7300         * entry other than the PF.
7301         */
7302        pf_id = adapter->vfs_allocated_count + E1000_VLVF_POOLSEL_SHIFT;
7303        bits = ~BIT(pf_id) & E1000_VLVF_POOLSEL_MASK;
7304        bits &= rd32(E1000_VLVF(idx));
7305
7306        /* Disable the filter so this falls into the default pool. */
7307        if (!bits) {
7308                if (adapter->flags & IGB_FLAG_VLAN_PROMISC)
7309                        wr32(E1000_VLVF(idx), BIT(pf_id));
7310                else
7311                        wr32(E1000_VLVF(idx), 0);
7312        }
7313}
7314
7315static s32 igb_set_vf_vlan(struct igb_adapter *adapter, u32 vid,
7316                           bool add, u32 vf)
7317{
7318        int pf_id = adapter->vfs_allocated_count;
7319        struct e1000_hw *hw = &adapter->hw;
7320        int err;
7321
7322        /* If VLAN overlaps with one the PF is currently monitoring make
7323         * sure that we are able to allocate a VLVF entry.  This may be
7324         * redundant but it guarantees PF will maintain visibility to
7325         * the VLAN.
7326         */
7327        if (add && test_bit(vid, adapter->active_vlans)) {
7328                err = igb_vfta_set(hw, vid, pf_id, true, false);
7329                if (err)
7330                        return err;
7331        }
7332
7333        err = igb_vfta_set(hw, vid, vf, add, false);
7334
7335        if (add && !err)
7336                return err;
7337
7338        /* If we failed to add the VF VLAN or we are removing the VF VLAN
7339         * we may need to drop the PF pool bit in order to allow us to free
7340         * up the VLVF resources.
7341         */
7342        if (test_bit(vid, adapter->active_vlans) ||
7343            (adapter->flags & IGB_FLAG_VLAN_PROMISC))
7344                igb_update_pf_vlvf(adapter, vid);
7345
7346        return err;
7347}
7348
7349static void igb_set_vmvir(struct igb_adapter *adapter, u32 vid, u32 vf)
7350{
7351        struct e1000_hw *hw = &adapter->hw;
7352
7353        if (vid)
7354                wr32(E1000_VMVIR(vf), (vid | E1000_VMVIR_VLANA_DEFAULT));
7355        else
7356                wr32(E1000_VMVIR(vf), 0);
7357}
7358
7359static int igb_enable_port_vlan(struct igb_adapter *adapter, int vf,
7360                                u16 vlan, u8 qos)
7361{
7362        int err;
7363
7364        err = igb_set_vf_vlan(adapter, vlan, true, vf);
7365        if (err)
7366                return err;
7367
7368        igb_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
7369        igb_set_vmolr(adapter, vf, !vlan);
7370
7371        /* revoke access to previous VLAN */
7372        if (vlan != adapter->vf_data[vf].pf_vlan)
7373                igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7374                                false, vf);
7375
7376        adapter->vf_data[vf].pf_vlan = vlan;
7377        adapter->vf_data[vf].pf_qos = qos;
7378        igb_set_vf_vlan_strip(adapter, vf, true);
7379        dev_info(&adapter->pdev->dev,
7380                 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
7381        if (test_bit(__IGB_DOWN, &adapter->state)) {
7382                dev_warn(&adapter->pdev->dev,
7383                         "The VF VLAN has been set, but the PF device is not up.\n");
7384                dev_warn(&adapter->pdev->dev,
7385                         "Bring the PF device up before attempting to use the VF device.\n");
7386        }
7387
7388        return err;
7389}
7390
7391static int igb_disable_port_vlan(struct igb_adapter *adapter, int vf)
7392{
7393        /* Restore tagless access via VLAN 0 */
7394        igb_set_vf_vlan(adapter, 0, true, vf);
7395
7396        igb_set_vmvir(adapter, 0, vf);
7397        igb_set_vmolr(adapter, vf, true);
7398
7399        /* Remove any PF assigned VLAN */
7400        if (adapter->vf_data[vf].pf_vlan)
7401                igb_set_vf_vlan(adapter, adapter->vf_data[vf].pf_vlan,
7402                                false, vf);
7403
7404        adapter->vf_data[vf].pf_vlan = 0;
7405        adapter->vf_data[vf].pf_qos = 0;
7406        igb_set_vf_vlan_strip(adapter, vf, false);
7407
7408        return 0;
7409}
7410
7411static int igb_ndo_set_vf_vlan(struct net_device *netdev, int vf,
7412                               u16 vlan, u8 qos, __be16 vlan_proto)
7413{
7414        struct igb_adapter *adapter = netdev_priv(netdev);
7415
7416        if ((vf >= adapter->vfs_allocated_count) || (vlan > 4095) || (qos > 7))
7417                return -EINVAL;
7418
7419        if (vlan_proto != htons(ETH_P_8021Q))
7420                return -EPROTONOSUPPORT;
7421
7422        return (vlan || qos) ? igb_enable_port_vlan(adapter, vf, vlan, qos) :
7423                               igb_disable_port_vlan(adapter, vf);
7424}
7425
7426static int igb_set_vf_vlan_msg(struct igb_adapter *adapter, u32 *msgbuf, u32 vf)
7427{
7428        int add = (msgbuf[0] & E1000_VT_MSGINFO_MASK) >> E1000_VT_MSGINFO_SHIFT;
7429        int vid = (msgbuf[1] & E1000_VLVF_VLANID_MASK);
7430        int ret;
7431
7432        if (adapter->vf_data[vf].pf_vlan)
7433                return -1;
7434
7435        /* VLAN 0 is a special case, don't allow it to be removed */
7436        if (!vid && !add)
7437                return 0;
7438
7439        ret = igb_set_vf_vlan(adapter, vid, !!add, vf);
7440        if (!ret)
7441                igb_set_vf_vlan_strip(adapter, vf, !!vid);
7442        return ret;
7443}
7444
7445static inline void igb_vf_reset(struct igb_adapter *adapter, u32 vf)
7446{
7447        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7448
7449        /* clear flags - except flag that indicates PF has set the MAC */
7450        vf_data->flags &= IGB_VF_FLAG_PF_SET_MAC;
7451        vf_data->last_nack = jiffies;
7452
7453        /* reset vlans for device */
7454        igb_clear_vf_vfta(adapter, vf);
7455        igb_set_vf_vlan(adapter, vf_data->pf_vlan, true, vf);
7456        igb_set_vmvir(adapter, vf_data->pf_vlan |
7457                               (vf_data->pf_qos << VLAN_PRIO_SHIFT), vf);
7458        igb_set_vmolr(adapter, vf, !vf_data->pf_vlan);
7459        igb_set_vf_vlan_strip(adapter, vf, !!(vf_data->pf_vlan));
7460
7461        /* reset multicast table array for vf */
7462        adapter->vf_data[vf].num_vf_mc_hashes = 0;
7463
7464        /* Flush and reset the mta with the new values */
7465        igb_set_rx_mode(adapter->netdev);
7466}
7467
7468static void igb_vf_reset_event(struct igb_adapter *adapter, u32 vf)
7469{
7470        unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7471
7472        /* clear mac address as we were hotplug removed/added */
7473        if (!(adapter->vf_data[vf].flags & IGB_VF_FLAG_PF_SET_MAC))
7474                eth_zero_addr(vf_mac);
7475
7476        /* process remaining reset events */
7477        igb_vf_reset(adapter, vf);
7478}
7479
7480static void igb_vf_reset_msg(struct igb_adapter *adapter, u32 vf)
7481{
7482        struct e1000_hw *hw = &adapter->hw;
7483        unsigned char *vf_mac = adapter->vf_data[vf].vf_mac_addresses;
7484        u32 reg, msgbuf[3];
7485        u8 *addr = (u8 *)(&msgbuf[1]);
7486
7487        /* process all the same items cleared in a function level reset */
7488        igb_vf_reset(adapter, vf);
7489
7490        /* set vf mac address */
7491        igb_set_vf_mac(adapter, vf, vf_mac);
7492
7493        /* enable transmit and receive for vf */
7494        reg = rd32(E1000_VFTE);
7495        wr32(E1000_VFTE, reg | BIT(vf));
7496        reg = rd32(E1000_VFRE);
7497        wr32(E1000_VFRE, reg | BIT(vf));
7498
7499        adapter->vf_data[vf].flags |= IGB_VF_FLAG_CTS;
7500
7501        /* reply to reset with ack and vf mac address */
7502        if (!is_zero_ether_addr(vf_mac)) {
7503                msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_ACK;
7504                memcpy(addr, vf_mac, ETH_ALEN);
7505        } else {
7506                msgbuf[0] = E1000_VF_RESET | E1000_VT_MSGTYPE_NACK;
7507        }
7508        igb_write_mbx(hw, msgbuf, 3, vf);
7509}
7510
7511static void igb_flush_mac_table(struct igb_adapter *adapter)
7512{
7513        struct e1000_hw *hw = &adapter->hw;
7514        int i;
7515
7516        for (i = 0; i < hw->mac.rar_entry_count; i++) {
7517                adapter->mac_table[i].state &= ~IGB_MAC_STATE_IN_USE;
7518                eth_zero_addr(adapter->mac_table[i].addr);
7519                adapter->mac_table[i].queue = 0;
7520                igb_rar_set_index(adapter, i);
7521        }
7522}
7523
7524static int igb_available_rars(struct igb_adapter *adapter, u8 queue)
7525{
7526        struct e1000_hw *hw = &adapter->hw;
7527        /* do not count rar entries reserved for VFs MAC addresses */
7528        int rar_entries = hw->mac.rar_entry_count -
7529                          adapter->vfs_allocated_count;
7530        int i, count = 0;
7531
7532        for (i = 0; i < rar_entries; i++) {
7533                /* do not count default entries */
7534                if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT)
7535                        continue;
7536
7537                /* do not count "in use" entries for different queues */
7538                if ((adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE) &&
7539                    (adapter->mac_table[i].queue != queue))
7540                        continue;
7541
7542                count++;
7543        }
7544
7545        return count;
7546}
7547
7548/* Set default MAC address for the PF in the first RAR entry */
7549static void igb_set_default_mac_filter(struct igb_adapter *adapter)
7550{
7551        struct igb_mac_addr *mac_table = &adapter->mac_table[0];
7552
7553        ether_addr_copy(mac_table->addr, adapter->hw.mac.addr);
7554        mac_table->queue = adapter->vfs_allocated_count;
7555        mac_table->state = IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7556
7557        igb_rar_set_index(adapter, 0);
7558}
7559
7560/* If the filter to be added and an already existing filter express
7561 * the same address and address type, it should be possible to only
7562 * override the other configurations, for example the queue to steer
7563 * traffic.
7564 */
7565static bool igb_mac_entry_can_be_used(const struct igb_mac_addr *entry,
7566                                      const u8 *addr, const u8 flags)
7567{
7568        if (!(entry->state & IGB_MAC_STATE_IN_USE))
7569                return true;
7570
7571        if ((entry->state & IGB_MAC_STATE_SRC_ADDR) !=
7572            (flags & IGB_MAC_STATE_SRC_ADDR))
7573                return false;
7574
7575        if (!ether_addr_equal(addr, entry->addr))
7576                return false;
7577
7578        return true;
7579}
7580
7581/* Add a MAC filter for 'addr' directing matching traffic to 'queue',
7582 * 'flags' is used to indicate what kind of match is made, match is by
7583 * default for the destination address, if matching by source address
7584 * is desired the flag IGB_MAC_STATE_SRC_ADDR can be used.
7585 */
7586static int igb_add_mac_filter_flags(struct igb_adapter *adapter,
7587                                    const u8 *addr, const u8 queue,
7588                                    const u8 flags)
7589{
7590        struct e1000_hw *hw = &adapter->hw;
7591        int rar_entries = hw->mac.rar_entry_count -
7592                          adapter->vfs_allocated_count;
7593        int i;
7594
7595        if (is_zero_ether_addr(addr))
7596                return -EINVAL;
7597
7598        /* Search for the first empty entry in the MAC table.
7599         * Do not touch entries at the end of the table reserved for the VF MAC
7600         * addresses.
7601         */
7602        for (i = 0; i < rar_entries; i++) {
7603                if (!igb_mac_entry_can_be_used(&adapter->mac_table[i],
7604                                               addr, flags))
7605                        continue;
7606
7607                ether_addr_copy(adapter->mac_table[i].addr, addr);
7608                adapter->mac_table[i].queue = queue;
7609                adapter->mac_table[i].state |= IGB_MAC_STATE_IN_USE | flags;
7610
7611                igb_rar_set_index(adapter, i);
7612                return i;
7613        }
7614
7615        return -ENOSPC;
7616}
7617
7618static int igb_add_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7619                              const u8 queue)
7620{
7621        return igb_add_mac_filter_flags(adapter, addr, queue, 0);
7622}
7623
7624/* Remove a MAC filter for 'addr' directing matching traffic to
7625 * 'queue', 'flags' is used to indicate what kind of match need to be
7626 * removed, match is by default for the destination address, if
7627 * matching by source address is to be removed the flag
7628 * IGB_MAC_STATE_SRC_ADDR can be used.
7629 */
7630static int igb_del_mac_filter_flags(struct igb_adapter *adapter,
7631                                    const u8 *addr, const u8 queue,
7632                                    const u8 flags)
7633{
7634        struct e1000_hw *hw = &adapter->hw;
7635        int rar_entries = hw->mac.rar_entry_count -
7636                          adapter->vfs_allocated_count;
7637        int i;
7638
7639        if (is_zero_ether_addr(addr))
7640                return -EINVAL;
7641
7642        /* Search for matching entry in the MAC table based on given address
7643         * and queue. Do not touch entries at the end of the table reserved
7644         * for the VF MAC addresses.
7645         */
7646        for (i = 0; i < rar_entries; i++) {
7647                if (!(adapter->mac_table[i].state & IGB_MAC_STATE_IN_USE))
7648                        continue;
7649                if ((adapter->mac_table[i].state & flags) != flags)
7650                        continue;
7651                if (adapter->mac_table[i].queue != queue)
7652                        continue;
7653                if (!ether_addr_equal(adapter->mac_table[i].addr, addr))
7654                        continue;
7655
7656                /* When a filter for the default address is "deleted",
7657                 * we return it to its initial configuration
7658                 */
7659                if (adapter->mac_table[i].state & IGB_MAC_STATE_DEFAULT) {
7660                        adapter->mac_table[i].state =
7661                                IGB_MAC_STATE_DEFAULT | IGB_MAC_STATE_IN_USE;
7662                        adapter->mac_table[i].queue =
7663                                adapter->vfs_allocated_count;
7664                } else {
7665                        adapter->mac_table[i].state = 0;
7666                        adapter->mac_table[i].queue = 0;
7667                        eth_zero_addr(adapter->mac_table[i].addr);
7668                }
7669
7670                igb_rar_set_index(adapter, i);
7671                return 0;
7672        }
7673
7674        return -ENOENT;
7675}
7676
7677static int igb_del_mac_filter(struct igb_adapter *adapter, const u8 *addr,
7678                              const u8 queue)
7679{
7680        return igb_del_mac_filter_flags(adapter, addr, queue, 0);
7681}
7682
7683int igb_add_mac_steering_filter(struct igb_adapter *adapter,
7684                                const u8 *addr, u8 queue, u8 flags)
7685{
7686        struct e1000_hw *hw = &adapter->hw;
7687
7688        /* In theory, this should be supported on 82575 as well, but
7689         * that part wasn't easily accessible during development.
7690         */
7691        if (hw->mac.type != e1000_i210)
7692                return -EOPNOTSUPP;
7693
7694        return igb_add_mac_filter_flags(adapter, addr, queue,
7695                                        IGB_MAC_STATE_QUEUE_STEERING | flags);
7696}
7697
7698int igb_del_mac_steering_filter(struct igb_adapter *adapter,
7699                                const u8 *addr, u8 queue, u8 flags)
7700{
7701        return igb_del_mac_filter_flags(adapter, addr, queue,
7702                                        IGB_MAC_STATE_QUEUE_STEERING | flags);
7703}
7704
7705static int igb_uc_sync(struct net_device *netdev, const unsigned char *addr)
7706{
7707        struct igb_adapter *adapter = netdev_priv(netdev);
7708        int ret;
7709
7710        ret = igb_add_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7711
7712        return min_t(int, ret, 0);
7713}
7714
7715static int igb_uc_unsync(struct net_device *netdev, const unsigned char *addr)
7716{
7717        struct igb_adapter *adapter = netdev_priv(netdev);
7718
7719        igb_del_mac_filter(adapter, addr, adapter->vfs_allocated_count);
7720
7721        return 0;
7722}
7723
7724static int igb_set_vf_mac_filter(struct igb_adapter *adapter, const int vf,
7725                                 const u32 info, const u8 *addr)
7726{
7727        struct pci_dev *pdev = adapter->pdev;
7728        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7729        struct list_head *pos;
7730        struct vf_mac_filter *entry = NULL;
7731        int ret = 0;
7732
7733        if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7734            !vf_data->trusted) {
7735                dev_warn(&pdev->dev,
7736                         "VF %d requested MAC filter but is administratively denied\n",
7737                          vf);
7738                return -EINVAL;
7739        }
7740        if (!is_valid_ether_addr(addr)) {
7741                dev_warn(&pdev->dev,
7742                         "VF %d attempted to set invalid MAC filter\n",
7743                          vf);
7744                return -EINVAL;
7745        }
7746
7747        switch (info) {
7748        case E1000_VF_MAC_FILTER_CLR:
7749                /* remove all unicast MAC filters related to the current VF */
7750                list_for_each(pos, &adapter->vf_macs.l) {
7751                        entry = list_entry(pos, struct vf_mac_filter, l);
7752                        if (entry->vf == vf) {
7753                                entry->vf = -1;
7754                                entry->free = true;
7755                                igb_del_mac_filter(adapter, entry->vf_mac, vf);
7756                        }
7757                }
7758                break;
7759        case E1000_VF_MAC_FILTER_ADD:
7760                /* try to find empty slot in the list */
7761                list_for_each(pos, &adapter->vf_macs.l) {
7762                        entry = list_entry(pos, struct vf_mac_filter, l);
7763                        if (entry->free)
7764                                break;
7765                }
7766
7767                if (entry && entry->free) {
7768                        entry->free = false;
7769                        entry->vf = vf;
7770                        ether_addr_copy(entry->vf_mac, addr);
7771
7772                        ret = igb_add_mac_filter(adapter, addr, vf);
7773                        ret = min_t(int, ret, 0);
7774                } else {
7775                        ret = -ENOSPC;
7776                }
7777
7778                if (ret == -ENOSPC)
7779                        dev_warn(&pdev->dev,
7780                                 "VF %d has requested MAC filter but there is no space for it\n",
7781                                 vf);
7782                break;
7783        default:
7784                ret = -EINVAL;
7785                break;
7786        }
7787
7788        return ret;
7789}
7790
7791static int igb_set_vf_mac_addr(struct igb_adapter *adapter, u32 *msg, int vf)
7792{
7793        struct pci_dev *pdev = adapter->pdev;
7794        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7795        u32 info = msg[0] & E1000_VT_MSGINFO_MASK;
7796
7797        /* The VF MAC Address is stored in a packed array of bytes
7798         * starting at the second 32 bit word of the msg array
7799         */
7800        unsigned char *addr = (unsigned char *)&msg[1];
7801        int ret = 0;
7802
7803        if (!info) {
7804                if ((vf_data->flags & IGB_VF_FLAG_PF_SET_MAC) &&
7805                    !vf_data->trusted) {
7806                        dev_warn(&pdev->dev,
7807                                 "VF %d attempted to override administratively set MAC address\nReload the VF driver to resume operations\n",
7808                                 vf);
7809                        return -EINVAL;
7810                }
7811
7812                if (!is_valid_ether_addr(addr)) {
7813                        dev_warn(&pdev->dev,
7814                                 "VF %d attempted to set invalid MAC\n",
7815                                 vf);
7816                        return -EINVAL;
7817                }
7818
7819                ret = igb_set_vf_mac(adapter, vf, addr);
7820        } else {
7821                ret = igb_set_vf_mac_filter(adapter, vf, info, addr);
7822        }
7823
7824        return ret;
7825}
7826
7827static void igb_rcv_ack_from_vf(struct igb_adapter *adapter, u32 vf)
7828{
7829        struct e1000_hw *hw = &adapter->hw;
7830        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7831        u32 msg = E1000_VT_MSGTYPE_NACK;
7832
7833        /* if device isn't clear to send it shouldn't be reading either */
7834        if (!(vf_data->flags & IGB_VF_FLAG_CTS) &&
7835            time_after(jiffies, vf_data->last_nack + (2 * HZ))) {
7836                igb_write_mbx(hw, &msg, 1, vf);
7837                vf_data->last_nack = jiffies;
7838        }
7839}
7840
7841static void igb_rcv_msg_from_vf(struct igb_adapter *adapter, u32 vf)
7842{
7843        struct pci_dev *pdev = adapter->pdev;
7844        u32 msgbuf[E1000_VFMAILBOX_SIZE];
7845        struct e1000_hw *hw = &adapter->hw;
7846        struct vf_data_storage *vf_data = &adapter->vf_data[vf];
7847        s32 retval;
7848
7849        retval = igb_read_mbx(hw, msgbuf, E1000_VFMAILBOX_SIZE, vf, false);
7850
7851        if (retval) {
7852                /* if receive failed revoke VF CTS stats and restart init */
7853                dev_err(&pdev->dev, "Error receiving message from VF\n");
7854                vf_data->flags &= ~IGB_VF_FLAG_CTS;
7855                if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7856                        goto unlock;
7857                goto out;
7858        }
7859
7860        /* this is a message we already processed, do nothing */
7861        if (msgbuf[0] & (E1000_VT_MSGTYPE_ACK | E1000_VT_MSGTYPE_NACK))
7862                goto unlock;
7863
7864        /* until the vf completes a reset it should not be
7865         * allowed to start any configuration.
7866         */
7867        if (msgbuf[0] == E1000_VF_RESET) {
7868                /* unlocks mailbox */
7869                igb_vf_reset_msg(adapter, vf);
7870                return;
7871        }
7872
7873        if (!(vf_data->flags & IGB_VF_FLAG_CTS)) {
7874                if (!time_after(jiffies, vf_data->last_nack + (2 * HZ)))
7875                        goto unlock;
7876                retval = -1;
7877                goto out;
7878        }
7879
7880        switch ((msgbuf[0] & 0xFFFF)) {
7881        case E1000_VF_SET_MAC_ADDR:
7882                retval = igb_set_vf_mac_addr(adapter, msgbuf, vf);
7883                break;
7884        case E1000_VF_SET_PROMISC:
7885                retval = igb_set_vf_promisc(adapter, msgbuf, vf);
7886                break;
7887        case E1000_VF_SET_MULTICAST:
7888                retval = igb_set_vf_multicasts(adapter, msgbuf, vf);
7889                break;
7890        case E1000_VF_SET_LPE:
7891                retval = igb_set_vf_rlpml(adapter, msgbuf[1], vf);
7892                break;
7893        case E1000_VF_SET_VLAN:
7894                retval = -1;
7895                if (vf_data->pf_vlan)
7896                        dev_warn(&pdev->dev,
7897                                 "VF %d attempted to override administratively set VLAN tag\nReload the VF driver to resume operations\n",
7898                                 vf);
7899                else
7900                        retval = igb_set_vf_vlan_msg(adapter, msgbuf, vf);
7901                break;
7902        default:
7903                dev_err(&pdev->dev, "Unhandled Msg %08x\n", msgbuf[0]);
7904                retval = -1;
7905                break;
7906        }
7907
7908        msgbuf[0] |= E1000_VT_MSGTYPE_CTS;
7909out:
7910        /* notify the VF of the results of what it sent us */
7911        if (retval)
7912                msgbuf[0] |= E1000_VT_MSGTYPE_NACK;
7913        else
7914                msgbuf[0] |= E1000_VT_MSGTYPE_ACK;
7915
7916        /* unlocks mailbox */
7917        igb_write_mbx(hw, msgbuf, 1, vf);
7918        return;
7919
7920unlock:
7921        igb_unlock_mbx(hw, vf);
7922}
7923
7924static void igb_msg_task(struct igb_adapter *adapter)
7925{
7926        struct e1000_hw *hw = &adapter->hw;
7927        u32 vf;
7928
7929        for (vf = 0; vf < adapter->vfs_allocated_count; vf++) {
7930                /* process any reset requests */
7931                if (!igb_check_for_rst(hw, vf))
7932                        igb_vf_reset_event(adapter, vf);
7933
7934                /* process any messages pending */
7935                if (!igb_check_for_msg(hw, vf))
7936                        igb_rcv_msg_from_vf(adapter, vf);
7937
7938                /* process any acks */
7939                if (!igb_check_for_ack(hw, vf))
7940                        igb_rcv_ack_from_vf(adapter, vf);
7941        }
7942}
7943
7944/**
7945 *  igb_set_uta - Set unicast filter table address
7946 *  @adapter: board private structure
7947 *  @set: boolean indicating if we are setting or clearing bits
7948 *
7949 *  The unicast table address is a register array of 32-bit registers.
7950 *  The table is meant to be used in a way similar to how the MTA is used
7951 *  however due to certain limitations in the hardware it is necessary to
7952 *  set all the hash bits to 1 and use the VMOLR ROPE bit as a promiscuous
7953 *  enable bit to allow vlan tag stripping when promiscuous mode is enabled
7954 **/
7955static void igb_set_uta(struct igb_adapter *adapter, bool set)
7956{
7957        struct e1000_hw *hw = &adapter->hw;
7958        u32 uta = set ? ~0 : 0;
7959        int i;
7960
7961        /* we only need to do this if VMDq is enabled */
7962        if (!adapter->vfs_allocated_count)
7963                return;
7964
7965        for (i = hw->mac.uta_reg_count; i--;)
7966                array_wr32(E1000_UTA, i, uta);
7967}
7968
7969/**
7970 *  igb_intr_msi - Interrupt Handler
7971 *  @irq: interrupt number
7972 *  @data: pointer to a network interface device structure
7973 **/
7974static irqreturn_t igb_intr_msi(int irq, void *data)
7975{
7976        struct igb_adapter *adapter = data;
7977        struct igb_q_vector *q_vector = adapter->q_vector[0];
7978        struct e1000_hw *hw = &adapter->hw;
7979        /* read ICR disables interrupts using IAM */
7980        u32 icr = rd32(E1000_ICR);
7981
7982        igb_write_itr(q_vector);
7983
7984        if (icr & E1000_ICR_DRSTA)
7985                schedule_work(&adapter->reset_task);
7986
7987        if (icr & E1000_ICR_DOUTSYNC) {
7988                /* HW is reporting DMA is out of sync */
7989                adapter->stats.doosync++;
7990        }
7991
7992        if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
7993                hw->mac.get_link_status = 1;
7994                if (!test_bit(__IGB_DOWN, &adapter->state))
7995                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
7996        }
7997
7998        if (icr & E1000_ICR_TS)
7999                igb_tsync_interrupt(adapter);
8000
8001        napi_schedule(&q_vector->napi);
8002
8003        return IRQ_HANDLED;
8004}
8005
8006/**
8007 *  igb_intr - Legacy Interrupt Handler
8008 *  @irq: interrupt number
8009 *  @data: pointer to a network interface device structure
8010 **/
8011static irqreturn_t igb_intr(int irq, void *data)
8012{
8013        struct igb_adapter *adapter = data;
8014        struct igb_q_vector *q_vector = adapter->q_vector[0];
8015        struct e1000_hw *hw = &adapter->hw;
8016        /* Interrupt Auto-Mask...upon reading ICR, interrupts are masked.  No
8017         * need for the IMC write
8018         */
8019        u32 icr = rd32(E1000_ICR);
8020
8021        /* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
8022         * not set, then the adapter didn't send an interrupt
8023         */
8024        if (!(icr & E1000_ICR_INT_ASSERTED))
8025                return IRQ_NONE;
8026
8027        igb_write_itr(q_vector);
8028
8029        if (icr & E1000_ICR_DRSTA)
8030                schedule_work(&adapter->reset_task);
8031
8032        if (icr & E1000_ICR_DOUTSYNC) {
8033                /* HW is reporting DMA is out of sync */
8034                adapter->stats.doosync++;
8035        }
8036
8037        if (icr & (E1000_ICR_RXSEQ | E1000_ICR_LSC)) {
8038                hw->mac.get_link_status = 1;
8039                /* guard against interrupt when we're going down */
8040                if (!test_bit(__IGB_DOWN, &adapter->state))
8041                        mod_timer(&adapter->watchdog_timer, jiffies + 1);
8042        }
8043
8044        if (icr & E1000_ICR_TS)
8045                igb_tsync_interrupt(adapter);
8046
8047        napi_schedule(&q_vector->napi);
8048
8049        return IRQ_HANDLED;
8050}
8051
8052static void igb_ring_irq_enable(struct igb_q_vector *q_vector)
8053{
8054        struct igb_adapter *adapter = q_vector->adapter;
8055        struct e1000_hw *hw = &adapter->hw;
8056
8057        if ((q_vector->rx.ring && (adapter->rx_itr_setting & 3)) ||
8058            (!q_vector->rx.ring && (adapter->tx_itr_setting & 3))) {
8059                if ((adapter->num_q_vectors == 1) && !adapter->vf_data)
8060                        igb_set_itr(q_vector);
8061                else
8062                        igb_update_ring_itr(q_vector);
8063        }
8064
8065        if (!test_bit(__IGB_DOWN, &adapter->state)) {
8066                if (adapter->flags & IGB_FLAG_HAS_MSIX)
8067                        wr32(E1000_EIMS, q_vector->eims_value);
8068                else
8069                        igb_irq_enable(adapter);
8070        }
8071}
8072
8073/**
8074 *  igb_poll - NAPI Rx polling callback
8075 *  @napi: napi polling structure
8076 *  @budget: count of how many packets we should handle
8077 **/
8078static int igb_poll(struct napi_struct *napi, int budget)
8079{
8080        struct igb_q_vector *q_vector = container_of(napi,
8081                                                     struct igb_q_vector,
8082                                                     napi);
8083        bool clean_complete = true;
8084        int work_done = 0;
8085
8086#ifdef CONFIG_IGB_DCA
8087        if (q_vector->adapter->flags & IGB_FLAG_DCA_ENABLED)
8088                igb_update_dca(q_vector);
8089#endif
8090        if (q_vector->tx.ring)
8091                clean_complete = igb_clean_tx_irq(q_vector, budget);
8092
8093        if (q_vector->rx.ring) {
8094                int cleaned = igb_clean_rx_irq(q_vector, budget);
8095
8096                work_done += cleaned;
8097                if (cleaned >= budget)
8098                        clean_complete = false;
8099        }
8100
8101        /* If all work not completed, return budget and keep polling */
8102        if (!clean_complete)
8103                return budget;
8104
8105        /* Exit the polling mode, but don't re-enable interrupts if stack might
8106         * poll us due to busy-polling
8107         */
8108        if (likely(napi_complete_done(napi, work_done)))
8109                igb_ring_irq_enable(q_vector);
8110
8111        return work_done;
8112}
8113
8114/**
8115 *  igb_clean_tx_irq - Reclaim resources after transmit completes
8116 *  @q_vector: pointer to q_vector containing needed info
8117 *  @napi_budget: Used to determine if we are in netpoll
8118 *
8119 *  returns true if ring is completely cleaned
8120 **/
8121static bool igb_clean_tx_irq(struct igb_q_vector *q_vector, int napi_budget)
8122{
8123        struct igb_adapter *adapter = q_vector->adapter;
8124        struct igb_ring *tx_ring = q_vector->tx.ring;
8125        struct igb_tx_buffer *tx_buffer;
8126        union e1000_adv_tx_desc *tx_desc;
8127        unsigned int total_bytes = 0, total_packets = 0;
8128        unsigned int budget = q_vector->tx.work_limit;
8129        unsigned int i = tx_ring->next_to_clean;
8130
8131        if (test_bit(__IGB_DOWN, &adapter->state))
8132                return true;
8133
8134        tx_buffer = &tx_ring->tx_buffer_info[i];
8135        tx_desc = IGB_TX_DESC(tx_ring, i);
8136        i -= tx_ring->count;
8137
8138        do {
8139                union e1000_adv_tx_desc *eop_desc = tx_buffer->next_to_watch;
8140
8141                /* if next_to_watch is not set then there is no work pending */
8142                if (!eop_desc)
8143                        break;
8144
8145                /* prevent any other reads prior to eop_desc */
8146                smp_rmb();
8147
8148                /* if DD is not set pending work has not been completed */
8149                if (!(eop_desc->wb.status & cpu_to_le32(E1000_TXD_STAT_DD)))
8150                        break;
8151
8152                /* clear next_to_watch to prevent false hangs */
8153                tx_buffer->next_to_watch = NULL;
8154
8155                /* update the statistics for this packet */
8156                total_bytes += tx_buffer->bytecount;
8157                total_packets += tx_buffer->gso_segs;
8158
8159                /* free the skb */
8160                if (tx_buffer->type == IGB_TYPE_SKB)
8161                        napi_consume_skb(tx_buffer->skb, napi_budget);
8162                else
8163                        xdp_return_frame(tx_buffer->xdpf);
8164
8165                /* unmap skb header data */
8166                dma_unmap_single(tx_ring->dev,
8167                                 dma_unmap_addr(tx_buffer, dma),
8168                                 dma_unmap_len(tx_buffer, len),
8169                                 DMA_TO_DEVICE);
8170
8171                /* clear tx_buffer data */
8172                dma_unmap_len_set(tx_buffer, len, 0);
8173
8174                /* clear last DMA location and unmap remaining buffers */
8175                while (tx_desc != eop_desc) {
8176                        tx_buffer++;
8177                        tx_desc++;
8178                        i++;
8179                        if (unlikely(!i)) {
8180                                i -= tx_ring->count;
8181                                tx_buffer = tx_ring->tx_buffer_info;
8182                                tx_desc = IGB_TX_DESC(tx_ring, 0);
8183                        }
8184
8185                        /* unmap any remaining paged data */
8186                        if (dma_unmap_len(tx_buffer, len)) {
8187                                dma_unmap_page(tx_ring->dev,
8188                                               dma_unmap_addr(tx_buffer, dma),
8189                                               dma_unmap_len(tx_buffer, len),
8190                                               DMA_TO_DEVICE);
8191                                dma_unmap_len_set(tx_buffer, len, 0);
8192                        }
8193                }
8194
8195                /* move us one more past the eop_desc for start of next pkt */
8196                tx_buffer++;
8197                tx_desc++;
8198                i++;
8199                if (unlikely(!i)) {
8200                        i -= tx_ring->count;
8201                        tx_buffer = tx_ring->tx_buffer_info;
8202                        tx_desc = IGB_TX_DESC(tx_ring, 0);
8203                }
8204
8205                /* issue prefetch for next Tx descriptor */
8206                prefetch(tx_desc);
8207
8208                /* update budget accounting */
8209                budget--;
8210        } while (likely(budget));
8211
8212        netdev_tx_completed_queue(txring_txq(tx_ring),
8213                                  total_packets, total_bytes);
8214        i += tx_ring->count;
8215        tx_ring->next_to_clean = i;
8216        u64_stats_update_begin(&tx_ring->tx_syncp);
8217        tx_ring->tx_stats.bytes += total_bytes;
8218        tx_ring->tx_stats.packets += total_packets;
8219        u64_stats_update_end(&tx_ring->tx_syncp);
8220        q_vector->tx.total_bytes += total_bytes;
8221        q_vector->tx.total_packets += total_packets;
8222
8223        if (test_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags)) {
8224                struct e1000_hw *hw = &adapter->hw;
8225
8226                /* Detect a transmit hang in hardware, this serializes the
8227                 * check with the clearing of time_stamp and movement of i
8228                 */
8229                clear_bit(IGB_RING_FLAG_TX_DETECT_HANG, &tx_ring->flags);
8230                if (tx_buffer->next_to_watch &&
8231                    time_after(jiffies, tx_buffer->time_stamp +
8232                               (adapter->tx_timeout_factor * HZ)) &&
8233                    !(rd32(E1000_STATUS) & E1000_STATUS_TXOFF)) {
8234
8235                        /* detected Tx unit hang */
8236                        dev_err(tx_ring->dev,
8237                                "Detected Tx Unit Hang\n"
8238                                "  Tx Queue             <%d>\n"
8239                                "  TDH                  <%x>\n"
8240                                "  TDT                  <%x>\n"
8241                                "  next_to_use          <%x>\n"
8242                                "  next_to_clean        <%x>\n"
8243                                "buffer_info[next_to_clean]\n"
8244                                "  time_stamp           <%lx>\n"
8245                                "  next_to_watch        <%p>\n"
8246                                "  jiffies              <%lx>\n"
8247                                "  desc.status          <%x>\n",
8248                                tx_ring->queue_index,
8249                                rd32(E1000_TDH(tx_ring->reg_idx)),
8250                                readl(tx_ring->tail),
8251                                tx_ring->next_to_use,
8252                                tx_ring->next_to_clean,
8253                                tx_buffer->time_stamp,
8254                                tx_buffer->next_to_watch,
8255                                jiffies,
8256                                tx_buffer->next_to_watch->wb.status);
8257                        netif_stop_subqueue(tx_ring->netdev,
8258                                            tx_ring->queue_index);
8259
8260                        /* we are about to reset, no point in enabling stuff */
8261                        return true;
8262                }
8263        }
8264
8265#define TX_WAKE_THRESHOLD (DESC_NEEDED * 2)
8266        if (unlikely(total_packets &&
8267            netif_carrier_ok(tx_ring->netdev) &&
8268            igb_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD)) {
8269                /* Make sure that anybody stopping the queue after this
8270                 * sees the new next_to_clean.
8271                 */
8272                smp_mb();
8273                if (__netif_subqueue_stopped(tx_ring->netdev,
8274                                             tx_ring->queue_index) &&
8275                    !(test_bit(__IGB_DOWN, &adapter->state))) {
8276                        netif_wake_subqueue(tx_ring->netdev,
8277                                            tx_ring->queue_index);
8278
8279                        u64_stats_update_begin(&tx_ring->tx_syncp);
8280                        tx_ring->tx_stats.restart_queue++;
8281                        u64_stats_update_end(&tx_ring->tx_syncp);
8282                }
8283        }
8284
8285        return !!budget;
8286}
8287
8288/**
8289 *  igb_reuse_rx_page - page flip buffer and store it back on the ring
8290 *  @rx_ring: rx descriptor ring to store buffers on
8291 *  @old_buff: donor buffer to have page reused
8292 *
8293 *  Synchronizes page for reuse by the adapter
8294 **/
8295static void igb_reuse_rx_page(struct igb_ring *rx_ring,
8296                              struct igb_rx_buffer *old_buff)
8297{
8298        struct igb_rx_buffer *new_buff;
8299        u16 nta = rx_ring->next_to_alloc;
8300
8301        new_buff = &rx_ring->rx_buffer_info[nta];
8302
8303        /* update, and store next to alloc */
8304        nta++;
8305        rx_ring->next_to_alloc = (nta < rx_ring->count) ? nta : 0;
8306
8307        /* Transfer page from old buffer to new buffer.
8308         * Move each member individually to avoid possible store
8309         * forwarding stalls.
8310         */
8311        new_buff->dma           = old_buff->dma;
8312        new_buff->page          = old_buff->page;
8313        new_buff->page_offset   = old_buff->page_offset;
8314        new_buff->pagecnt_bias  = old_buff->pagecnt_bias;
8315}
8316
8317static bool igb_can_reuse_rx_page(struct igb_rx_buffer *rx_buffer,
8318                                  int rx_buf_pgcnt)
8319{
8320        unsigned int pagecnt_bias = rx_buffer->pagecnt_bias;
8321        struct page *page = rx_buffer->page;
8322
8323        /* avoid re-using remote and pfmemalloc pages */
8324        if (!dev_page_is_reusable(page))
8325                return false;
8326
8327#if (PAGE_SIZE < 8192)
8328        /* if we are only owner of page we can reuse it */
8329        if (unlikely((rx_buf_pgcnt - pagecnt_bias) > 1))
8330                return false;
8331#else
8332#define IGB_LAST_OFFSET \
8333        (SKB_WITH_OVERHEAD(PAGE_SIZE) - IGB_RXBUFFER_2048)
8334
8335        if (rx_buffer->page_offset > IGB_LAST_OFFSET)
8336                return false;
8337#endif
8338
8339        /* If we have drained the page fragment pool we need to update
8340         * the pagecnt_bias and page count so that we fully restock the
8341         * number of references the driver holds.
8342         */
8343        if (unlikely(pagecnt_bias == 1)) {
8344                page_ref_add(page, USHRT_MAX - 1);
8345                rx_buffer->pagecnt_bias = USHRT_MAX;
8346        }
8347
8348        return true;
8349}
8350
8351/**
8352 *  igb_add_rx_frag - Add contents of Rx buffer to sk_buff
8353 *  @rx_ring: rx descriptor ring to transact packets on
8354 *  @rx_buffer: buffer containing page to add
8355 *  @skb: sk_buff to place the data into
8356 *  @size: size of buffer to be added
8357 *
8358 *  This function will add the data contained in rx_buffer->page to the skb.
8359 **/
8360static void igb_add_rx_frag(struct igb_ring *rx_ring,
8361                            struct igb_rx_buffer *rx_buffer,
8362                            struct sk_buff *skb,
8363                            unsigned int size)
8364{
8365#if (PAGE_SIZE < 8192)
8366        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8367#else
8368        unsigned int truesize = ring_uses_build_skb(rx_ring) ?
8369                                SKB_DATA_ALIGN(IGB_SKB_PAD + size) :
8370                                SKB_DATA_ALIGN(size);
8371#endif
8372        skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, rx_buffer->page,
8373                        rx_buffer->page_offset, size, truesize);
8374#if (PAGE_SIZE < 8192)
8375        rx_buffer->page_offset ^= truesize;
8376#else
8377        rx_buffer->page_offset += truesize;
8378#endif
8379}
8380
8381static struct sk_buff *igb_construct_skb(struct igb_ring *rx_ring,
8382                                         struct igb_rx_buffer *rx_buffer,
8383                                         struct xdp_buff *xdp,
8384                                         ktime_t timestamp)
8385{
8386#if (PAGE_SIZE < 8192)
8387        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8388#else
8389        unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
8390                                               xdp->data_hard_start);
8391#endif
8392        unsigned int size = xdp->data_end - xdp->data;
8393        unsigned int headlen;
8394        struct sk_buff *skb;
8395
8396        /* prefetch first cache line of first page */
8397        net_prefetch(xdp->data);
8398
8399        /* allocate a skb to store the frags */
8400        skb = napi_alloc_skb(&rx_ring->q_vector->napi, IGB_RX_HDR_LEN);
8401        if (unlikely(!skb))
8402                return NULL;
8403
8404        if (timestamp)
8405                skb_hwtstamps(skb)->hwtstamp = timestamp;
8406
8407        /* Determine available headroom for copy */
8408        headlen = size;
8409        if (headlen > IGB_RX_HDR_LEN)
8410                headlen = eth_get_headlen(skb->dev, xdp->data, IGB_RX_HDR_LEN);
8411
8412        /* align pull length to size of long to optimize memcpy performance */
8413        memcpy(__skb_put(skb, headlen), xdp->data, ALIGN(headlen, sizeof(long)));
8414
8415        /* update all of the pointers */
8416        size -= headlen;
8417        if (size) {
8418                skb_add_rx_frag(skb, 0, rx_buffer->page,
8419                                (xdp->data + headlen) - page_address(rx_buffer->page),
8420                                size, truesize);
8421#if (PAGE_SIZE < 8192)
8422                rx_buffer->page_offset ^= truesize;
8423#else
8424                rx_buffer->page_offset += truesize;
8425#endif
8426        } else {
8427                rx_buffer->pagecnt_bias++;
8428        }
8429
8430        return skb;
8431}
8432
8433static struct sk_buff *igb_build_skb(struct igb_ring *rx_ring,
8434                                     struct igb_rx_buffer *rx_buffer,
8435                                     struct xdp_buff *xdp,
8436                                     ktime_t timestamp)
8437{
8438#if (PAGE_SIZE < 8192)
8439        unsigned int truesize = igb_rx_pg_size(rx_ring) / 2;
8440#else
8441        unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
8442                                SKB_DATA_ALIGN(xdp->data_end -
8443                                               xdp->data_hard_start);
8444#endif
8445        unsigned int metasize = xdp->data - xdp->data_meta;
8446        struct sk_buff *skb;
8447
8448        /* prefetch first cache line of first page */
8449        net_prefetch(xdp->data_meta);
8450
8451        /* build an skb around the page buffer */
8452        skb = napi_build_skb(xdp->data_hard_start, truesize);
8453        if (unlikely(!skb))
8454                return NULL;
8455
8456        /* update pointers within the skb to store the data */
8457        skb_reserve(skb, xdp->data - xdp->data_hard_start);
8458        __skb_put(skb, xdp->data_end - xdp->data);
8459
8460        if (metasize)
8461                skb_metadata_set(skb, metasize);
8462
8463        if (timestamp)
8464                skb_hwtstamps(skb)->hwtstamp = timestamp;
8465
8466        /* update buffer offset */
8467#if (PAGE_SIZE < 8192)
8468        rx_buffer->page_offset ^= truesize;
8469#else
8470        rx_buffer->page_offset += truesize;
8471#endif
8472
8473        return skb;
8474}
8475
8476static struct sk_buff *igb_run_xdp(struct igb_adapter *adapter,
8477                                   struct igb_ring *rx_ring,
8478                                   struct xdp_buff *xdp)
8479{
8480        int err, result = IGB_XDP_PASS;
8481        struct bpf_prog *xdp_prog;
8482        u32 act;
8483
8484        xdp_prog = READ_ONCE(rx_ring->xdp_prog);
8485
8486        if (!xdp_prog)
8487                goto xdp_out;
8488
8489        prefetchw(xdp->data_hard_start); /* xdp_frame write */
8490
8491        act = bpf_prog_run_xdp(xdp_prog, xdp);
8492        switch (act) {
8493        case XDP_PASS:
8494                break;
8495        case XDP_TX:
8496                result = igb_xdp_xmit_back(adapter, xdp);
8497                if (result == IGB_XDP_CONSUMED)
8498                        goto out_failure;
8499                break;
8500        case XDP_REDIRECT:
8501                err = xdp_do_redirect(adapter->netdev, xdp, xdp_prog);
8502                if (err)
8503                        goto out_failure;
8504                result = IGB_XDP_REDIR;
8505                break;
8506        default:
8507                bpf_warn_invalid_xdp_action(adapter->netdev, xdp_prog, act);
8508                fallthrough;
8509        case XDP_ABORTED:
8510out_failure:
8511                trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
8512                fallthrough;
8513        case XDP_DROP:
8514                result = IGB_XDP_CONSUMED;
8515                break;
8516        }
8517xdp_out:
8518        return ERR_PTR(-result);
8519}
8520
8521static unsigned int igb_rx_frame_truesize(struct igb_ring *rx_ring,
8522                                          unsigned int size)
8523{
8524        unsigned int truesize;
8525
8526#if (PAGE_SIZE < 8192)
8527        truesize = igb_rx_pg_size(rx_ring) / 2; /* Must be power-of-2 */
8528#else
8529        truesize = ring_uses_build_skb(rx_ring) ?
8530                SKB_DATA_ALIGN(IGB_SKB_PAD + size) +
8531                SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) :
8532                SKB_DATA_ALIGN(size);
8533#endif
8534        return truesize;
8535}
8536
8537static void igb_rx_buffer_flip(struct igb_ring *rx_ring,
8538                               struct igb_rx_buffer *rx_buffer,
8539                               unsigned int size)
8540{
8541        unsigned int truesize = igb_rx_frame_truesize(rx_ring, size);
8542#if (PAGE_SIZE < 8192)
8543        rx_buffer->page_offset ^= truesize;
8544#else
8545        rx_buffer->page_offset += truesize;
8546#endif
8547}
8548
8549static inline void igb_rx_checksum(struct igb_ring *ring,
8550                                   union e1000_adv_rx_desc *rx_desc,
8551                                   struct sk_buff *skb)
8552{
8553        skb_checksum_none_assert(skb);
8554
8555        /* Ignore Checksum bit is set */
8556        if (igb_test_staterr(rx_desc, E1000_RXD_STAT_IXSM))
8557                return;
8558
8559        /* Rx checksum disabled via ethtool */
8560        if (!(ring->netdev->features & NETIF_F_RXCSUM))
8561                return;
8562
8563        /* TCP/UDP checksum error bit is set */
8564        if (igb_test_staterr(rx_desc,
8565                             E1000_RXDEXT_STATERR_TCPE |
8566                             E1000_RXDEXT_STATERR_IPE)) {
8567                /* work around errata with sctp packets where the TCPE aka
8568                 * L4E bit is set incorrectly on 64 byte (60 byte w/o crc)
8569                 * packets, (aka let the stack check the crc32c)
8570                 */
8571                if (!((skb->len == 60) &&
8572                      test_bit(IGB_RING_FLAG_RX_SCTP_CSUM, &ring->flags))) {
8573                        u64_stats_update_begin(&ring->rx_syncp);
8574                        ring->rx_stats.csum_err++;
8575                        u64_stats_update_end(&ring->rx_syncp);
8576                }
8577                /* let the stack verify checksum errors */
8578                return;
8579        }
8580        /* It must be a TCP or UDP packet with a valid checksum */
8581        if (igb_test_staterr(rx_desc, E1000_RXD_STAT_TCPCS |
8582                                      E1000_RXD_STAT_UDPCS))
8583                skb->ip_summed = CHECKSUM_UNNECESSARY;
8584
8585        dev_dbg(ring->dev, "cksum success: bits %08X\n",
8586                le32_to_cpu(rx_desc->wb.upper.status_error));
8587}
8588
8589static inline void igb_rx_hash(struct igb_ring *ring,
8590                               union e1000_adv_rx_desc *rx_desc,
8591                               struct sk_buff *skb)
8592{
8593        if (ring->netdev->features & NETIF_F_RXHASH)
8594                skb_set_hash(skb,
8595                             le32_to_cpu(rx_desc->wb.lower.hi_dword.rss),
8596                             PKT_HASH_TYPE_L3);
8597}
8598
8599/**
8600 *  igb_is_non_eop - process handling of non-EOP buffers
8601 *  @rx_ring: Rx ring being processed
8602 *  @rx_desc: Rx descriptor for current buffer
8603 *
8604 *  This function updates next to clean.  If the buffer is an EOP buffer
8605 *  this function exits returning false, otherwise it will place the
8606 *  sk_buff in the next buffer to be chained and return true indicating
8607 *  that this is in fact a non-EOP buffer.
8608 **/
8609static bool igb_is_non_eop(struct igb_ring *rx_ring,
8610                           union e1000_adv_rx_desc *rx_desc)
8611{
8612        u32 ntc = rx_ring->next_to_clean + 1;
8613
8614        /* fetch, update, and store next to clean */
8615        ntc = (ntc < rx_ring->count) ? ntc : 0;
8616        rx_ring->next_to_clean = ntc;
8617
8618        prefetch(IGB_RX_DESC(rx_ring, ntc));
8619
8620        if (likely(igb_test_staterr(rx_desc, E1000_RXD_STAT_EOP)))
8621                return false;
8622
8623        return true;
8624}
8625
8626/**
8627 *  igb_cleanup_headers - Correct corrupted or empty headers
8628 *  @rx_ring: rx descriptor ring packet is being transacted on
8629 *  @rx_desc: pointer to the EOP Rx descriptor
8630 *  @skb: pointer to current skb being fixed
8631 *
8632 *  Address the case where we are pulling data in on pages only
8633 *  and as such no data is present in the skb header.
8634 *
8635 *  In addition if skb is not at least 60 bytes we need to pad it so that
8636 *  it is large enough to qualify as a valid Ethernet frame.
8637 *
8638 *  Returns true if an error was encountered and skb was freed.
8639 **/
8640static bool igb_cleanup_headers(struct igb_ring *rx_ring,
8641                                union e1000_adv_rx_desc *rx_desc,
8642                                struct sk_buff *skb)
8643{
8644        /* XDP packets use error pointer so abort at this point */
8645        if (IS_ERR(skb))
8646                return true;
8647
8648        if (unlikely((igb_test_staterr(rx_desc,
8649                                       E1000_RXDEXT_ERR_FRAME_ERR_MASK)))) {
8650                struct net_device *netdev = rx_ring->netdev;
8651                if (!(netdev->features & NETIF_F_RXALL)) {
8652                        dev_kfree_skb_any(skb);
8653                        return true;
8654                }
8655        }
8656
8657        /* if eth_skb_pad returns an error the skb was freed */
8658        if (eth_skb_pad(skb))
8659                return true;
8660
8661        return false;
8662}
8663
8664/**
8665 *  igb_process_skb_fields - Populate skb header fields from Rx descriptor
8666 *  @rx_ring: rx descriptor ring packet is being transacted on
8667 *  @rx_desc: pointer to the EOP Rx descriptor
8668 *  @skb: pointer to current skb being populated
8669 *
8670 *  This function checks the ring, descriptor, and packet information in
8671 *  order to populate the hash, checksum, VLAN, timestamp, protocol, and
8672 *  other fields within the skb.
8673 **/
8674static void igb_process_skb_fields(struct igb_ring *rx_ring,
8675                                   union e1000_adv_rx_desc *rx_desc,
8676                                   struct sk_buff *skb)
8677{
8678        struct net_device *dev = rx_ring->netdev;
8679
8680        igb_rx_hash(rx_ring, rx_desc, skb);
8681
8682        igb_rx_checksum(rx_ring, rx_desc, skb);
8683
8684        if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TS) &&
8685            !igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP))
8686                igb_ptp_rx_rgtstamp(rx_ring->q_vector, skb);
8687
8688        if ((dev->features & NETIF_F_HW_VLAN_CTAG_RX) &&
8689            igb_test_staterr(rx_desc, E1000_RXD_STAT_VP)) {
8690                u16 vid;
8691
8692                if (igb_test_staterr(rx_desc, E1000_RXDEXT_STATERR_LB) &&
8693                    test_bit(IGB_RING_FLAG_RX_LB_VLAN_BSWAP, &rx_ring->flags))
8694                        vid = be16_to_cpu((__force __be16)rx_desc->wb.upper.vlan);
8695                else
8696                        vid = le16_to_cpu(rx_desc->wb.upper.vlan);
8697
8698                __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vid);
8699        }
8700
8701        skb_record_rx_queue(skb, rx_ring->queue_index);
8702
8703        skb->protocol = eth_type_trans(skb, rx_ring->netdev);
8704}
8705
8706static unsigned int igb_rx_offset(struct igb_ring *rx_ring)
8707{
8708        return ring_uses_build_skb(rx_ring) ? IGB_SKB_PAD : 0;
8709}
8710
8711static struct igb_rx_buffer *igb_get_rx_buffer(struct igb_ring *rx_ring,
8712                                               const unsigned int size, int *rx_buf_pgcnt)
8713{
8714        struct igb_rx_buffer *rx_buffer;
8715
8716        rx_buffer = &rx_ring->rx_buffer_info[rx_ring->next_to_clean];
8717        *rx_buf_pgcnt =
8718#if (PAGE_SIZE < 8192)
8719                page_count(rx_buffer->page);
8720#else
8721                0;
8722#endif
8723        prefetchw(rx_buffer->page);
8724
8725        /* we are reusing so sync this buffer for CPU use */
8726        dma_sync_single_range_for_cpu(rx_ring->dev,
8727                                      rx_buffer->dma,
8728                                      rx_buffer->page_offset,
8729                                      size,
8730                                      DMA_FROM_DEVICE);
8731
8732        rx_buffer->pagecnt_bias--;
8733
8734        return rx_buffer;
8735}
8736
8737static void igb_put_rx_buffer(struct igb_ring *rx_ring,
8738                              struct igb_rx_buffer *rx_buffer, int rx_buf_pgcnt)
8739{
8740        if (igb_can_reuse_rx_page(rx_buffer, rx_buf_pgcnt)) {
8741                /* hand second half of page back to the ring */
8742                igb_reuse_rx_page(rx_ring, rx_buffer);
8743        } else {
8744                /* We are not reusing the buffer so unmap it and free
8745                 * any references we are holding to it
8746                 */
8747                dma_unmap_page_attrs(rx_ring->dev, rx_buffer->dma,
8748                                     igb_rx_pg_size(rx_ring), DMA_FROM_DEVICE,
8749                                     IGB_RX_DMA_ATTR);
8750                __page_frag_cache_drain(rx_buffer->page,
8751                                        rx_buffer->pagecnt_bias);
8752        }
8753
8754        /* clear contents of rx_buffer */
8755        rx_buffer->page = NULL;
8756}
8757
8758static int igb_clean_rx_irq(struct igb_q_vector *q_vector, const int budget)
8759{
8760        struct igb_adapter *adapter = q_vector->adapter;
8761        struct igb_ring *rx_ring = q_vector->rx.ring;
8762        struct sk_buff *skb = rx_ring->skb;
8763        unsigned int total_bytes = 0, total_packets = 0;
8764        u16 cleaned_count = igb_desc_unused(rx_ring);
8765        unsigned int xdp_xmit = 0;
8766        struct xdp_buff xdp;
8767        u32 frame_sz = 0;
8768        int rx_buf_pgcnt;
8769
8770        /* Frame size depend on rx_ring setup when PAGE_SIZE=4K */
8771#if (PAGE_SIZE < 8192)
8772        frame_sz = igb_rx_frame_truesize(rx_ring, 0);
8773#endif
8774        xdp_init_buff(&xdp, frame_sz, &rx_ring->xdp_rxq);
8775
8776        while (likely(total_packets < budget)) {
8777                union e1000_adv_rx_desc *rx_desc;
8778                struct igb_rx_buffer *rx_buffer;
8779                ktime_t timestamp = 0;
8780                int pkt_offset = 0;
8781                unsigned int size;
8782                void *pktbuf;
8783
8784                /* return some buffers to hardware, one at a time is too slow */
8785                if (cleaned_count >= IGB_RX_BUFFER_WRITE) {
8786                        igb_alloc_rx_buffers(rx_ring, cleaned_count);
8787                        cleaned_count = 0;
8788                }
8789
8790                rx_desc = IGB_RX_DESC(rx_ring, rx_ring->next_to_clean);
8791                size = le16_to_cpu(rx_desc->wb.upper.length);
8792                if (!size)
8793                        break;
8794
8795                /* This memory barrier is needed to keep us from reading
8796                 * any other fields out of the rx_desc until we know the
8797                 * descriptor has been written back
8798                 */
8799                dma_rmb();
8800
8801                rx_buffer = igb_get_rx_buffer(rx_ring, size, &rx_buf_pgcnt);
8802                pktbuf = page_address(rx_buffer->page) + rx_buffer->page_offset;
8803
8804                /* pull rx packet timestamp if available and valid */
8805                if (igb_test_staterr(rx_desc, E1000_RXDADV_STAT_TSIP)) {
8806                        int ts_hdr_len;
8807
8808                        ts_hdr_len = igb_ptp_rx_pktstamp(rx_ring->q_vector,
8809                                                         pktbuf, &timestamp);
8810
8811                        pkt_offset += ts_hdr_len;
8812                        size -= ts_hdr_len;
8813                }
8814
8815                /* retrieve a buffer from the ring */
8816                if (!skb) {
8817                        unsigned char *hard_start = pktbuf - igb_rx_offset(rx_ring);
8818                        unsigned int offset = pkt_offset + igb_rx_offset(rx_ring);
8819
8820                        xdp_prepare_buff(&xdp, hard_start, offset, size, true);
8821#if (PAGE_SIZE > 4096)
8822                        /* At larger PAGE_SIZE, frame_sz depend on len size */
8823                        xdp.frame_sz = igb_rx_frame_truesize(rx_ring, size);
8824#endif
8825                        skb = igb_run_xdp(adapter, rx_ring, &xdp);
8826                }
8827
8828                if (IS_ERR(skb)) {
8829                        unsigned int xdp_res = -PTR_ERR(skb);
8830
8831                        if (xdp_res & (IGB_XDP_TX | IGB_XDP_REDIR)) {
8832                                xdp_xmit |= xdp_res;
8833                                igb_rx_buffer_flip(rx_ring, rx_buffer, size);
8834                        } else {
8835                                rx_buffer->pagecnt_bias++;
8836                        }
8837                        total_packets++;
8838                        total_bytes += size;
8839                } else if (skb)
8840                        igb_add_rx_frag(rx_ring, rx_buffer, skb, size);
8841                else if (ring_uses_build_skb(rx_ring))
8842                        skb = igb_build_skb(rx_ring, rx_buffer, &xdp,
8843                                            timestamp);
8844                else
8845                        skb = igb_construct_skb(rx_ring, rx_buffer,
8846                                                &xdp, timestamp);
8847
8848                /* exit if we failed to retrieve a buffer */
8849                if (!skb) {
8850                        rx_ring->rx_stats.alloc_failed++;
8851                        rx_buffer->pagecnt_bias++;
8852                        break;
8853                }
8854
8855                igb_put_rx_buffer(rx_ring, rx_buffer, rx_buf_pgcnt);
8856                cleaned_count++;
8857
8858                /* fetch next buffer in frame if non-eop */
8859                if (igb_is_non_eop(rx_ring, rx_desc))
8860                        continue;
8861
8862                /* verify the packet layout is correct */
8863                if (igb_cleanup_headers(rx_ring, rx_desc, skb)) {
8864                        skb = NULL;
8865                        continue;
8866                }
8867
8868                /* probably a little skewed due to removing CRC */
8869                total_bytes += skb->len;
8870
8871                /* populate checksum, timestamp, VLAN, and protocol */
8872                igb_process_skb_fields(rx_ring, rx_desc, skb);
8873
8874                napi_gro_receive(&q_vector->napi, skb);
8875
8876                /* reset skb pointer */
8877                skb = NULL;
8878
8879                /* update budget accounting */
8880                total_packets++;
8881        }
8882
8883        /* place incomplete frames back on ring for completion */
8884        rx_ring->skb = skb;
8885
8886        if (xdp_xmit & IGB_XDP_REDIR)
8887                xdp_do_flush();
8888
8889        if (xdp_xmit & IGB_XDP_TX) {
8890                struct igb_ring *tx_ring = igb_xdp_tx_queue_mapping(adapter);
8891
8892                igb_xdp_ring_update_tail(tx_ring);
8893        }
8894
8895        u64_stats_update_begin(&rx_ring->rx_syncp);
8896        rx_ring->rx_stats.packets += total_packets;
8897        rx_ring->rx_stats.bytes += total_bytes;
8898        u64_stats_update_end(&rx_ring->rx_syncp);
8899        q_vector->rx.total_packets += total_packets;
8900        q_vector->rx.total_bytes += total_bytes;
8901
8902        if (cleaned_count)
8903                igb_alloc_rx_buffers(rx_ring, cleaned_count);
8904
8905        return total_packets;
8906}
8907
8908static bool igb_alloc_mapped_page(struct igb_ring *rx_ring,
8909                                  struct igb_rx_buffer *bi)
8910{
8911        struct page *page = bi->page;
8912        dma_addr_t dma;
8913
8914        /* since we are recycling buffers we should seldom need to alloc */
8915        if (likely(page))
8916                return true;
8917
8918        /* alloc new page for storage */
8919        page = dev_alloc_pages(igb_rx_pg_order(rx_ring));
8920        if (unlikely(!page)) {
8921                rx_ring->rx_stats.alloc_failed++;
8922                return false;
8923        }
8924
8925        /* map page for use */
8926        dma = dma_map_page_attrs(rx_ring->dev, page, 0,
8927                                 igb_rx_pg_size(rx_ring),
8928                                 DMA_FROM_DEVICE,
8929                                 IGB_RX_DMA_ATTR);
8930
8931        /* if mapping failed free memory back to system since
8932         * there isn't much point in holding memory we can't use
8933         */
8934        if (dma_mapping_error(rx_ring->dev, dma)) {
8935                __free_pages(page, igb_rx_pg_order(rx_ring));
8936
8937                rx_ring->rx_stats.alloc_failed++;
8938                return false;
8939        }
8940
8941        bi->dma = dma;
8942        bi->page = page;
8943        bi->page_offset = igb_rx_offset(rx_ring);
8944        page_ref_add(page, USHRT_MAX - 1);
8945        bi->pagecnt_bias = USHRT_MAX;
8946
8947        return true;
8948}
8949
8950/**
8951 *  igb_alloc_rx_buffers - Replace used receive buffers
8952 *  @rx_ring: rx descriptor ring to allocate new receive buffers
8953 *  @cleaned_count: count of buffers to allocate
8954 **/
8955void igb_alloc_rx_buffers(struct igb_ring *rx_ring, u16 cleaned_count)
8956{
8957        union e1000_adv_rx_desc *rx_desc;
8958        struct igb_rx_buffer *bi;
8959        u16 i = rx_ring->next_to_use;
8960        u16 bufsz;
8961
8962        /* nothing to do */
8963        if (!cleaned_count)
8964                return;
8965
8966        rx_desc = IGB_RX_DESC(rx_ring, i);
8967        bi = &rx_ring->rx_buffer_info[i];
8968        i -= rx_ring->count;
8969
8970        bufsz = igb_rx_bufsz(rx_ring);
8971
8972        do {
8973                if (!igb_alloc_mapped_page(rx_ring, bi))
8974                        break;
8975
8976                /* sync the buffer for use by the device */
8977                dma_sync_single_range_for_device(rx_ring->dev, bi->dma,
8978                                                 bi->page_offset, bufsz,
8979                                                 DMA_FROM_DEVICE);
8980
8981                /* Refresh the desc even if buffer_addrs didn't change
8982                 * because each write-back erases this info.
8983                 */
8984                rx_desc->read.pkt_addr = cpu_to_le64(bi->dma + bi->page_offset);
8985
8986                rx_desc++;
8987                bi++;
8988                i++;
8989                if (unlikely(!i)) {
8990                        rx_desc = IGB_RX_DESC(rx_ring, 0);
8991                        bi = rx_ring->rx_buffer_info;
8992                        i -= rx_ring->count;
8993                }
8994
8995                /* clear the length for the next_to_use descriptor */
8996                rx_desc->wb.upper.length = 0;
8997
8998                cleaned_count--;
8999        } while (cleaned_count);
9000
9001        i += rx_ring->count;
9002
9003        if (rx_ring->next_to_use != i) {
9004                /* record the next descriptor to use */
9005                rx_ring->next_to_use = i;
9006
9007                /* update next to alloc since we have filled the ring */
9008                rx_ring->next_to_alloc = i;
9009
9010                /* Force memory writes to complete before letting h/w
9011                 * know there are new descriptors to fetch.  (Only
9012                 * applicable for weak-ordered memory model archs,
9013                 * such as IA-64).
9014                 */
9015                dma_wmb();
9016                writel(i, rx_ring->tail);
9017        }
9018}
9019
9020/**
9021 * igb_mii_ioctl -
9022 * @netdev: pointer to netdev struct
9023 * @ifr: interface structure
9024 * @cmd: ioctl command to execute
9025 **/
9026static int igb_mii_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9027{
9028        struct igb_adapter *adapter = netdev_priv(netdev);
9029        struct mii_ioctl_data *data = if_mii(ifr);
9030
9031        if (adapter->hw.phy.media_type != e1000_media_type_copper)
9032                return -EOPNOTSUPP;
9033
9034        switch (cmd) {
9035        case SIOCGMIIPHY:
9036                data->phy_id = adapter->hw.phy.addr;
9037                break;
9038        case SIOCGMIIREG:
9039                if (igb_read_phy_reg(&adapter->hw, data->reg_num & 0x1F,
9040                                     &data->val_out))
9041                        return -EIO;
9042                break;
9043        case SIOCSMIIREG:
9044        default:
9045                return -EOPNOTSUPP;
9046        }
9047        return 0;
9048}
9049
9050/**
9051 * igb_ioctl -
9052 * @netdev: pointer to netdev struct
9053 * @ifr: interface structure
9054 * @cmd: ioctl command to execute
9055 **/
9056static int igb_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
9057{
9058        switch (cmd) {
9059        case SIOCGMIIPHY:
9060        case SIOCGMIIREG:
9061        case SIOCSMIIREG:
9062                return igb_mii_ioctl(netdev, ifr, cmd);
9063        case SIOCGHWTSTAMP:
9064                return igb_ptp_get_ts_config(netdev, ifr);
9065        case SIOCSHWTSTAMP:
9066                return igb_ptp_set_ts_config(netdev, ifr);
9067        default:
9068                return -EOPNOTSUPP;
9069        }
9070}
9071
9072void igb_read_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9073{
9074        struct igb_adapter *adapter = hw->back;
9075
9076        pci_read_config_word(adapter->pdev, reg, value);
9077}
9078
9079void igb_write_pci_cfg(struct e1000_hw *hw, u32 reg, u16 *value)
9080{
9081        struct igb_adapter *adapter = hw->back;
9082
9083        pci_write_config_word(adapter->pdev, reg, *value);
9084}
9085
9086s32 igb_read_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9087{
9088        struct igb_adapter *adapter = hw->back;
9089
9090        if (pcie_capability_read_word(adapter->pdev, reg, value))
9091                return -E1000_ERR_CONFIG;
9092
9093        return 0;
9094}
9095
9096s32 igb_write_pcie_cap_reg(struct e1000_hw *hw, u32 reg, u16 *value)
9097{
9098        struct igb_adapter *adapter = hw->back;
9099
9100        if (pcie_capability_write_word(adapter->pdev, reg, *value))
9101                return -E1000_ERR_CONFIG;
9102
9103        return 0;
9104}
9105
9106static void igb_vlan_mode(struct net_device *netdev, netdev_features_t features)
9107{
9108        struct igb_adapter *adapter = netdev_priv(netdev);
9109        struct e1000_hw *hw = &adapter->hw;
9110        u32 ctrl, rctl;
9111        bool enable = !!(features & NETIF_F_HW_VLAN_CTAG_RX);
9112
9113        if (enable) {
9114                /* enable VLAN tag insert/strip */
9115                ctrl = rd32(E1000_CTRL);
9116                ctrl |= E1000_CTRL_VME;
9117                wr32(E1000_CTRL, ctrl);
9118
9119                /* Disable CFI check */
9120                rctl = rd32(E1000_RCTL);
9121                rctl &= ~E1000_RCTL_CFIEN;
9122                wr32(E1000_RCTL, rctl);
9123        } else {
9124                /* disable VLAN tag insert/strip */
9125                ctrl = rd32(E1000_CTRL);
9126                ctrl &= ~E1000_CTRL_VME;
9127                wr32(E1000_CTRL, ctrl);
9128        }
9129
9130        igb_set_vf_vlan_strip(adapter, adapter->vfs_allocated_count, enable);
9131}
9132
9133static int igb_vlan_rx_add_vid(struct net_device *netdev,
9134                               __be16 proto, u16 vid)
9135{
9136        struct igb_adapter *adapter = netdev_priv(netdev);
9137        struct e1000_hw *hw = &adapter->hw;
9138        int pf_id = adapter->vfs_allocated_count;
9139
9140        /* add the filter since PF can receive vlans w/o entry in vlvf */
9141        if (!vid || !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9142                igb_vfta_set(hw, vid, pf_id, true, !!vid);
9143
9144        set_bit(vid, adapter->active_vlans);
9145
9146        return 0;
9147}
9148
9149static int igb_vlan_rx_kill_vid(struct net_device *netdev,
9150                                __be16 proto, u16 vid)
9151{
9152        struct igb_adapter *adapter = netdev_priv(netdev);
9153        int pf_id = adapter->vfs_allocated_count;
9154        struct e1000_hw *hw = &adapter->hw;
9155
9156        /* remove VID from filter table */
9157        if (vid && !(adapter->flags & IGB_FLAG_VLAN_PROMISC))
9158                igb_vfta_set(hw, vid, pf_id, false, true);
9159
9160        clear_bit(vid, adapter->active_vlans);
9161
9162        return 0;
9163}
9164
9165static void igb_restore_vlan(struct igb_adapter *adapter)
9166{
9167        u16 vid = 1;
9168
9169        igb_vlan_mode(adapter->netdev, adapter->netdev->features);
9170        igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
9171
9172        for_each_set_bit_from(vid, adapter->active_vlans, VLAN_N_VID)
9173                igb_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
9174}
9175
9176int igb_set_spd_dplx(struct igb_adapter *adapter, u32 spd, u8 dplx)
9177{
9178        struct pci_dev *pdev = adapter->pdev;
9179        struct e1000_mac_info *mac = &adapter->hw.mac;
9180
9181        mac->autoneg = 0;
9182
9183        /* Make sure dplx is at most 1 bit and lsb of speed is not set
9184         * for the switch() below to work
9185         */
9186        if ((spd & 1) || (dplx & ~1))
9187                goto err_inval;
9188
9189        /* Fiber NIC's only allow 1000 gbps Full duplex
9190         * and 100Mbps Full duplex for 100baseFx sfp
9191         */
9192        if (adapter->hw.phy.media_type == e1000_media_type_internal_serdes) {
9193                switch (spd + dplx) {
9194                case SPEED_10 + DUPLEX_HALF:
9195                case SPEED_10 + DUPLEX_FULL:
9196                case SPEED_100 + DUPLEX_HALF:
9197                        goto err_inval;
9198                default:
9199                        break;
9200                }
9201        }
9202
9203        switch (spd + dplx) {
9204        case SPEED_10 + DUPLEX_HALF:
9205                mac->forced_speed_duplex = ADVERTISE_10_HALF;
9206                break;
9207        case SPEED_10 + DUPLEX_FULL:
9208                mac->forced_speed_duplex = ADVERTISE_10_FULL;
9209                break;
9210        case SPEED_100 + DUPLEX_HALF:
9211                mac->forced_speed_duplex = ADVERTISE_100_HALF;
9212                break;
9213        case SPEED_100 + DUPLEX_FULL:
9214                mac->forced_speed_duplex = ADVERTISE_100_FULL;
9215                break;
9216        case SPEED_1000 + DUPLEX_FULL:
9217                mac->autoneg = 1;
9218                adapter->hw.phy.autoneg_advertised = ADVERTISE_1000_FULL;
9219                break;
9220        case SPEED_1000 + DUPLEX_HALF: /* not supported */
9221        default:
9222                goto err_inval;
9223        }
9224
9225        /* clear MDI, MDI(-X) override is only allowed when autoneg enabled */
9226        adapter->hw.phy.mdix = AUTO_ALL_MODES;
9227
9228        return 0;
9229
9230err_inval:
9231        dev_err(&pdev->dev, "Unsupported Speed/Duplex configuration\n");
9232        return -EINVAL;
9233}
9234
9235static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
9236                          bool runtime)
9237{
9238        struct net_device *netdev = pci_get_drvdata(pdev);
9239        struct igb_adapter *adapter = netdev_priv(netdev);
9240        struct e1000_hw *hw = &adapter->hw;
9241        u32 ctrl, rctl, status;
9242        u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
9243        bool wake;
9244
9245        rtnl_lock();
9246        netif_device_detach(netdev);
9247
9248        if (netif_running(netdev))
9249                __igb_close(netdev, true);
9250
9251        igb_ptp_suspend(adapter);
9252
9253        igb_clear_interrupt_scheme(adapter);
9254        rtnl_unlock();
9255
9256        status = rd32(E1000_STATUS);
9257        if (status & E1000_STATUS_LU)
9258                wufc &= ~E1000_WUFC_LNKC;
9259
9260        if (wufc) {
9261                igb_setup_rctl(adapter);
9262                igb_set_rx_mode(netdev);
9263
9264                /* turn on all-multi mode if wake on multicast is enabled */
9265                if (wufc & E1000_WUFC_MC) {
9266                        rctl = rd32(E1000_RCTL);
9267                        rctl |= E1000_RCTL_MPE;
9268                        wr32(E1000_RCTL, rctl);
9269                }
9270
9271                ctrl = rd32(E1000_CTRL);
9272                ctrl |= E1000_CTRL_ADVD3WUC;
9273                wr32(E1000_CTRL, ctrl);
9274
9275                /* Allow time for pending master requests to run */
9276                igb_disable_pcie_master(hw);
9277
9278                wr32(E1000_WUC, E1000_WUC_PME_EN);
9279                wr32(E1000_WUFC, wufc);
9280        } else {
9281                wr32(E1000_WUC, 0);
9282                wr32(E1000_WUFC, 0);
9283        }
9284
9285        wake = wufc || adapter->en_mng_pt;
9286        if (!wake)
9287                igb_power_down_link(adapter);
9288        else
9289                igb_power_up_link(adapter);
9290
9291        if (enable_wake)
9292                *enable_wake = wake;
9293
9294        /* Release control of h/w to f/w.  If f/w is AMT enabled, this
9295         * would have already happened in close and is redundant.
9296         */
9297        igb_release_hw_control(adapter);
9298
9299        pci_disable_device(pdev);
9300
9301        return 0;
9302}
9303
9304static void igb_deliver_wake_packet(struct net_device *netdev)
9305{
9306        struct igb_adapter *adapter = netdev_priv(netdev);
9307        struct e1000_hw *hw = &adapter->hw;
9308        struct sk_buff *skb;
9309        u32 wupl;
9310
9311        wupl = rd32(E1000_WUPL) & E1000_WUPL_MASK;
9312
9313        /* WUPM stores only the first 128 bytes of the wake packet.
9314         * Read the packet only if we have the whole thing.
9315         */
9316        if ((wupl == 0) || (wupl > E1000_WUPM_BYTES))
9317                return;
9318
9319        skb = netdev_alloc_skb_ip_align(netdev, E1000_WUPM_BYTES);
9320        if (!skb)
9321                return;
9322
9323        skb_put(skb, wupl);
9324
9325        /* Ensure reads are 32-bit aligned */
9326        wupl = roundup(wupl, 4);
9327
9328        memcpy_fromio(skb->data, hw->hw_addr + E1000_WUPM_REG(0), wupl);
9329
9330        skb->protocol = eth_type_trans(skb, netdev);
9331        netif_rx(skb);
9332}
9333
9334static int __maybe_unused igb_suspend(struct device *dev)
9335{
9336        return __igb_shutdown(to_pci_dev(dev), NULL, 0);
9337}
9338
9339static int __maybe_unused __igb_resume(struct device *dev, bool rpm)
9340{
9341        struct pci_dev *pdev = to_pci_dev(dev);
9342        struct net_device *netdev = pci_get_drvdata(pdev);
9343        struct igb_adapter *adapter = netdev_priv(netdev);
9344        struct e1000_hw *hw = &adapter->hw;
9345        u32 err, val;
9346
9347        pci_set_power_state(pdev, PCI_D0);
9348        pci_restore_state(pdev);
9349        pci_save_state(pdev);
9350
9351        if (!pci_device_is_present(pdev))
9352                return -ENODEV;
9353        err = pci_enable_device_mem(pdev);
9354        if (err) {
9355                dev_err(&pdev->dev,
9356                        "igb: Cannot enable PCI device from suspend\n");
9357                return err;
9358        }
9359        pci_set_master(pdev);
9360
9361        pci_enable_wake(pdev, PCI_D3hot, 0);
9362        pci_enable_wake(pdev, PCI_D3cold, 0);
9363
9364        if (igb_init_interrupt_scheme(adapter, true)) {
9365                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9366                return -ENOMEM;
9367        }
9368
9369        igb_reset(adapter);
9370
9371        /* let the f/w know that the h/w is now under the control of the
9372         * driver.
9373         */
9374        igb_get_hw_control(adapter);
9375
9376        val = rd32(E1000_WUS);
9377        if (val & WAKE_PKT_WUS)
9378                igb_deliver_wake_packet(netdev);
9379
9380        wr32(E1000_WUS, ~0);
9381
9382        if (!rpm)
9383                rtnl_lock();
9384        if (!err && netif_running(netdev))
9385                err = __igb_open(netdev, true);
9386
9387        if (!err)
9388                netif_device_attach(netdev);
9389        if (!rpm)
9390                rtnl_unlock();
9391
9392        return err;
9393}
9394
9395static int __maybe_unused igb_resume(struct device *dev)
9396{
9397        return __igb_resume(dev, false);
9398}
9399
9400static int __maybe_unused igb_runtime_idle(struct device *dev)
9401{
9402        struct net_device *netdev = dev_get_drvdata(dev);
9403        struct igb_adapter *adapter = netdev_priv(netdev);
9404
9405        if (!igb_has_link(adapter))
9406                pm_schedule_suspend(dev, MSEC_PER_SEC * 5);
9407
9408        return -EBUSY;
9409}
9410
9411static int __maybe_unused igb_runtime_suspend(struct device *dev)
9412{
9413        return __igb_shutdown(to_pci_dev(dev), NULL, 1);
9414}
9415
9416static int __maybe_unused igb_runtime_resume(struct device *dev)
9417{
9418        return __igb_resume(dev, true);
9419}
9420
9421static void igb_shutdown(struct pci_dev *pdev)
9422{
9423        bool wake;
9424
9425        __igb_shutdown(pdev, &wake, 0);
9426
9427        if (system_state == SYSTEM_POWER_OFF) {
9428                pci_wake_from_d3(pdev, wake);
9429                pci_set_power_state(pdev, PCI_D3hot);
9430        }
9431}
9432
9433#ifdef CONFIG_PCI_IOV
9434static int igb_sriov_reinit(struct pci_dev *dev)
9435{
9436        struct net_device *netdev = pci_get_drvdata(dev);
9437        struct igb_adapter *adapter = netdev_priv(netdev);
9438        struct pci_dev *pdev = adapter->pdev;
9439
9440        rtnl_lock();
9441
9442        if (netif_running(netdev))
9443                igb_close(netdev);
9444        else
9445                igb_reset(adapter);
9446
9447        igb_clear_interrupt_scheme(adapter);
9448
9449        igb_init_queue_configuration(adapter);
9450
9451        if (igb_init_interrupt_scheme(adapter, true)) {
9452                rtnl_unlock();
9453                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
9454                return -ENOMEM;
9455        }
9456
9457        if (netif_running(netdev))
9458                igb_open(netdev);
9459
9460        rtnl_unlock();
9461
9462        return 0;
9463}
9464
9465static int igb_pci_disable_sriov(struct pci_dev *dev)
9466{
9467        int err = igb_disable_sriov(dev);
9468
9469        if (!err)
9470                err = igb_sriov_reinit(dev);
9471
9472        return err;
9473}
9474
9475static int igb_pci_enable_sriov(struct pci_dev *dev, int num_vfs)
9476{
9477        int err = igb_enable_sriov(dev, num_vfs);
9478
9479        if (err)
9480                goto out;
9481
9482        err = igb_sriov_reinit(dev);
9483        if (!err)
9484                return num_vfs;
9485
9486out:
9487        return err;
9488}
9489
9490#endif
9491static int igb_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
9492{
9493#ifdef CONFIG_PCI_IOV
9494        if (num_vfs == 0)
9495                return igb_pci_disable_sriov(dev);
9496        else
9497                return igb_pci_enable_sriov(dev, num_vfs);
9498#endif
9499        return 0;
9500}
9501
9502/**
9503 *  igb_io_error_detected - called when PCI error is detected
9504 *  @pdev: Pointer to PCI device
9505 *  @state: The current pci connection state
9506 *
9507 *  This function is called after a PCI bus error affecting
9508 *  this device has been detected.
9509 **/
9510static pci_ers_result_t igb_io_error_detected(struct pci_dev *pdev,
9511                                              pci_channel_state_t state)
9512{
9513        struct net_device *netdev = pci_get_drvdata(pdev);
9514        struct igb_adapter *adapter = netdev_priv(netdev);
9515
9516        netif_device_detach(netdev);
9517
9518        if (state == pci_channel_io_perm_failure)
9519                return PCI_ERS_RESULT_DISCONNECT;
9520
9521        if (netif_running(netdev))
9522                igb_down(adapter);
9523        pci_disable_device(pdev);
9524
9525        /* Request a slot slot reset. */
9526        return PCI_ERS_RESULT_NEED_RESET;
9527}
9528
9529/**
9530 *  igb_io_slot_reset - called after the pci bus has been reset.
9531 *  @pdev: Pointer to PCI device
9532 *
9533 *  Restart the card from scratch, as if from a cold-boot. Implementation
9534 *  resembles the first-half of the __igb_resume routine.
9535 **/
9536static pci_ers_result_t igb_io_slot_reset(struct pci_dev *pdev)
9537{
9538        struct net_device *netdev = pci_get_drvdata(pdev);
9539        struct igb_adapter *adapter = netdev_priv(netdev);
9540        struct e1000_hw *hw = &adapter->hw;
9541        pci_ers_result_t result;
9542
9543        if (pci_enable_device_mem(pdev)) {
9544                dev_err(&pdev->dev,
9545                        "Cannot re-enable PCI device after reset.\n");
9546                result = PCI_ERS_RESULT_DISCONNECT;
9547        } else {
9548                pci_set_master(pdev);
9549                pci_restore_state(pdev);
9550                pci_save_state(pdev);
9551
9552                pci_enable_wake(pdev, PCI_D3hot, 0);
9553                pci_enable_wake(pdev, PCI_D3cold, 0);
9554
9555                /* In case of PCI error, adapter lose its HW address
9556                 * so we should re-assign it here.
9557                 */
9558                hw->hw_addr = adapter->io_addr;
9559
9560                igb_reset(adapter);
9561                wr32(E1000_WUS, ~0);
9562                result = PCI_ERS_RESULT_RECOVERED;
9563        }
9564
9565        return result;
9566}
9567
9568/**
9569 *  igb_io_resume - called when traffic can start flowing again.
9570 *  @pdev: Pointer to PCI device
9571 *
9572 *  This callback is called when the error recovery driver tells us that
9573 *  its OK to resume normal operation. Implementation resembles the
9574 *  second-half of the __igb_resume routine.
9575 */
9576static void igb_io_resume(struct pci_dev *pdev)
9577{
9578        struct net_device *netdev = pci_get_drvdata(pdev);
9579        struct igb_adapter *adapter = netdev_priv(netdev);
9580
9581        if (netif_running(netdev)) {
9582                if (igb_up(adapter)) {
9583                        dev_err(&pdev->dev, "igb_up failed after reset\n");
9584                        return;
9585                }
9586        }
9587
9588        netif_device_attach(netdev);
9589
9590        /* let the f/w know that the h/w is now under the control of the
9591         * driver.
9592         */
9593        igb_get_hw_control(adapter);
9594}
9595
9596/**
9597 *  igb_rar_set_index - Sync RAL[index] and RAH[index] registers with MAC table
9598 *  @adapter: Pointer to adapter structure
9599 *  @index: Index of the RAR entry which need to be synced with MAC table
9600 **/
9601static void igb_rar_set_index(struct igb_adapter *adapter, u32 index)
9602{
9603        struct e1000_hw *hw = &adapter->hw;
9604        u32 rar_low, rar_high;
9605        u8 *addr = adapter->mac_table[index].addr;
9606
9607        /* HW expects these to be in network order when they are plugged
9608         * into the registers which are little endian.  In order to guarantee
9609         * that ordering we need to do an leXX_to_cpup here in order to be
9610         * ready for the byteswap that occurs with writel
9611         */
9612        rar_low = le32_to_cpup((__le32 *)(addr));
9613        rar_high = le16_to_cpup((__le16 *)(addr + 4));
9614
9615        /* Indicate to hardware the Address is Valid. */
9616        if (adapter->mac_table[index].state & IGB_MAC_STATE_IN_USE) {
9617                if (is_valid_ether_addr(addr))
9618                        rar_high |= E1000_RAH_AV;
9619
9620                if (adapter->mac_table[index].state & IGB_MAC_STATE_SRC_ADDR)
9621                        rar_high |= E1000_RAH_ASEL_SRC_ADDR;
9622
9623                switch (hw->mac.type) {
9624                case e1000_82575:
9625                case e1000_i210:
9626                        if (adapter->mac_table[index].state &
9627                            IGB_MAC_STATE_QUEUE_STEERING)
9628                                rar_high |= E1000_RAH_QSEL_ENABLE;
9629
9630                        rar_high |= E1000_RAH_POOL_1 *
9631                                    adapter->mac_table[index].queue;
9632                        break;
9633                default:
9634                        rar_high |= E1000_RAH_POOL_1 <<
9635                                    adapter->mac_table[index].queue;
9636                        break;
9637                }
9638        }
9639
9640        wr32(E1000_RAL(index), rar_low);
9641        wrfl();
9642        wr32(E1000_RAH(index), rar_high);
9643        wrfl();
9644}
9645
9646static int igb_set_vf_mac(struct igb_adapter *adapter,
9647                          int vf, unsigned char *mac_addr)
9648{
9649        struct e1000_hw *hw = &adapter->hw;
9650        /* VF MAC addresses start at end of receive addresses and moves
9651         * towards the first, as a result a collision should not be possible
9652         */
9653        int rar_entry = hw->mac.rar_entry_count - (vf + 1);
9654        unsigned char *vf_mac_addr = adapter->vf_data[vf].vf_mac_addresses;
9655
9656        ether_addr_copy(vf_mac_addr, mac_addr);
9657        ether_addr_copy(adapter->mac_table[rar_entry].addr, mac_addr);
9658        adapter->mac_table[rar_entry].queue = vf;
9659        adapter->mac_table[rar_entry].state |= IGB_MAC_STATE_IN_USE;
9660        igb_rar_set_index(adapter, rar_entry);
9661
9662        return 0;
9663}
9664
9665static int igb_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
9666{
9667        struct igb_adapter *adapter = netdev_priv(netdev);
9668
9669        if (vf >= adapter->vfs_allocated_count)
9670                return -EINVAL;
9671
9672        /* Setting the VF MAC to 0 reverts the IGB_VF_FLAG_PF_SET_MAC
9673         * flag and allows to overwrite the MAC via VF netdev.  This
9674         * is necessary to allow libvirt a way to restore the original
9675         * MAC after unbinding vfio-pci and reloading igbvf after shutting
9676         * down a VM.
9677         */
9678        if (is_zero_ether_addr(mac)) {
9679                adapter->vf_data[vf].flags &= ~IGB_VF_FLAG_PF_SET_MAC;
9680                dev_info(&adapter->pdev->dev,
9681                         "remove administratively set MAC on VF %d\n",
9682                         vf);
9683        } else if (is_valid_ether_addr(mac)) {
9684                adapter->vf_data[vf].flags |= IGB_VF_FLAG_PF_SET_MAC;
9685                dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n",
9686                         mac, vf);
9687                dev_info(&adapter->pdev->dev,
9688                         "Reload the VF driver to make this change effective.");
9689                /* Generate additional warning if PF is down */
9690                if (test_bit(__IGB_DOWN, &adapter->state)) {
9691                        dev_warn(&adapter->pdev->dev,
9692                                 "The VF MAC address has been set, but the PF device is not up.\n");
9693                        dev_warn(&adapter->pdev->dev,
9694                                 "Bring the PF device up before attempting to use the VF device.\n");
9695                }
9696        } else {
9697                return -EINVAL;
9698        }
9699        return igb_set_vf_mac(adapter, vf, mac);
9700}
9701
9702static int igb_link_mbps(int internal_link_speed)
9703{
9704        switch (internal_link_speed) {
9705        case SPEED_100:
9706                return 100;
9707        case SPEED_1000:
9708                return 1000;
9709        default:
9710                return 0;
9711        }
9712}
9713
9714static void igb_set_vf_rate_limit(struct e1000_hw *hw, int vf, int tx_rate,
9715                                  int link_speed)
9716{
9717        int rf_dec, rf_int;
9718        u32 bcnrc_val;
9719
9720        if (tx_rate != 0) {
9721                /* Calculate the rate factor values to set */
9722                rf_int = link_speed / tx_rate;
9723                rf_dec = (link_speed - (rf_int * tx_rate));
9724                rf_dec = (rf_dec * BIT(E1000_RTTBCNRC_RF_INT_SHIFT)) /
9725                         tx_rate;
9726
9727                bcnrc_val = E1000_RTTBCNRC_RS_ENA;
9728                bcnrc_val |= ((rf_int << E1000_RTTBCNRC_RF_INT_SHIFT) &
9729                              E1000_RTTBCNRC_RF_INT_MASK);
9730                bcnrc_val |= (rf_dec & E1000_RTTBCNRC_RF_DEC_MASK);
9731        } else {
9732                bcnrc_val = 0;
9733        }
9734
9735        wr32(E1000_RTTDQSEL, vf); /* vf X uses queue X */
9736        /* Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
9737         * register. MMW_SIZE=0x014 if 9728-byte jumbo is supported.
9738         */
9739        wr32(E1000_RTTBCNRM, 0x14);
9740        wr32(E1000_RTTBCNRC, bcnrc_val);
9741}
9742
9743static void igb_check_vf_rate_limit(struct igb_adapter *adapter)
9744{
9745        int actual_link_speed, i;
9746        bool reset_rate = false;
9747
9748        /* VF TX rate limit was not set or not supported */
9749        if ((adapter->vf_rate_link_speed == 0) ||
9750            (adapter->hw.mac.type != e1000_82576))
9751                return;
9752
9753        actual_link_speed = igb_link_mbps(adapter->link_speed);
9754        if (actual_link_speed != adapter->vf_rate_link_speed) {
9755                reset_rate = true;
9756                adapter->vf_rate_link_speed = 0;
9757                dev_info(&adapter->pdev->dev,
9758                         "Link speed has been changed. VF Transmit rate is disabled\n");
9759        }
9760
9761        for (i = 0; i < adapter->vfs_allocated_count; i++) {
9762                if (reset_rate)
9763                        adapter->vf_data[i].tx_rate = 0;
9764
9765                igb_set_vf_rate_limit(&adapter->hw, i,
9766                                      adapter->vf_data[i].tx_rate,
9767                                      actual_link_speed);
9768        }
9769}
9770
9771static int igb_ndo_set_vf_bw(struct net_device *netdev, int vf,
9772                             int min_tx_rate, int max_tx_rate)
9773{
9774        struct igb_adapter *adapter = netdev_priv(netdev);
9775        struct e1000_hw *hw = &adapter->hw;
9776        int actual_link_speed;
9777
9778        if (hw->mac.type != e1000_82576)
9779                return -EOPNOTSUPP;
9780
9781        if (min_tx_rate)
9782                return -EINVAL;
9783
9784        actual_link_speed = igb_link_mbps(adapter->link_speed);
9785        if ((vf >= adapter->vfs_allocated_count) ||
9786            (!(rd32(E1000_STATUS) & E1000_STATUS_LU)) ||
9787            (max_tx_rate < 0) ||
9788            (max_tx_rate > actual_link_speed))
9789                return -EINVAL;
9790
9791        adapter->vf_rate_link_speed = actual_link_speed;
9792        adapter->vf_data[vf].tx_rate = (u16)max_tx_rate;
9793        igb_set_vf_rate_limit(hw, vf, max_tx_rate, actual_link_speed);
9794
9795        return 0;
9796}
9797
9798static int igb_ndo_set_vf_spoofchk(struct net_device *netdev, int vf,
9799                                   bool setting)
9800{
9801        struct igb_adapter *adapter = netdev_priv(netdev);
9802        struct e1000_hw *hw = &adapter->hw;
9803        u32 reg_val, reg_offset;
9804
9805        if (!adapter->vfs_allocated_count)
9806                return -EOPNOTSUPP;
9807
9808        if (vf >= adapter->vfs_allocated_count)
9809                return -EINVAL;
9810
9811        reg_offset = (hw->mac.type == e1000_82576) ? E1000_DTXSWC : E1000_TXSWC;
9812        reg_val = rd32(reg_offset);
9813        if (setting)
9814                reg_val |= (BIT(vf) |
9815                            BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9816        else
9817                reg_val &= ~(BIT(vf) |
9818                             BIT(vf + E1000_DTXSWC_VLAN_SPOOF_SHIFT));
9819        wr32(reg_offset, reg_val);
9820
9821        adapter->vf_data[vf].spoofchk_enabled = setting;
9822        return 0;
9823}
9824
9825static int igb_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
9826{
9827        struct igb_adapter *adapter = netdev_priv(netdev);
9828
9829        if (vf >= adapter->vfs_allocated_count)
9830                return -EINVAL;
9831        if (adapter->vf_data[vf].trusted == setting)
9832                return 0;
9833
9834        adapter->vf_data[vf].trusted = setting;
9835
9836        dev_info(&adapter->pdev->dev, "VF %u is %strusted\n",
9837                 vf, setting ? "" : "not ");
9838        return 0;
9839}
9840
9841static int igb_ndo_get_vf_config(struct net_device *netdev,
9842                                 int vf, struct ifla_vf_info *ivi)
9843{
9844        struct igb_adapter *adapter = netdev_priv(netdev);
9845        if (vf >= adapter->vfs_allocated_count)
9846                return -EINVAL;
9847        ivi->vf = vf;
9848        memcpy(&ivi->mac, adapter->vf_data[vf].vf_mac_addresses, ETH_ALEN);
9849        ivi->max_tx_rate = adapter->vf_data[vf].tx_rate;
9850        ivi->min_tx_rate = 0;
9851        ivi->vlan = adapter->vf_data[vf].pf_vlan;
9852        ivi->qos = adapter->vf_data[vf].pf_qos;
9853        ivi->spoofchk = adapter->vf_data[vf].spoofchk_enabled;
9854        ivi->trusted = adapter->vf_data[vf].trusted;
9855        return 0;
9856}
9857
9858static void igb_vmm_control(struct igb_adapter *adapter)
9859{
9860        struct e1000_hw *hw = &adapter->hw;
9861        u32 reg;
9862
9863        switch (hw->mac.type) {
9864        case e1000_82575:
9865        case e1000_i210:
9866        case e1000_i211:
9867        case e1000_i354:
9868        default:
9869                /* replication is not supported for 82575 */
9870                return;
9871        case e1000_82576:
9872                /* notify HW that the MAC is adding vlan tags */
9873                reg = rd32(E1000_DTXCTL);
9874                reg |= E1000_DTXCTL_VLAN_ADDED;
9875                wr32(E1000_DTXCTL, reg);
9876                fallthrough;
9877        case e1000_82580:
9878                /* enable replication vlan tag stripping */
9879                reg = rd32(E1000_RPLOLR);
9880                reg |= E1000_RPLOLR_STRVLAN;
9881                wr32(E1000_RPLOLR, reg);
9882                fallthrough;
9883        case e1000_i350:
9884                /* none of the above registers are supported by i350 */
9885                break;
9886        }
9887
9888        if (adapter->vfs_allocated_count) {
9889                igb_vmdq_set_loopback_pf(hw, true);
9890                igb_vmdq_set_replication_pf(hw, true);
9891                igb_vmdq_set_anti_spoofing_pf(hw, true,
9892                                              adapter->vfs_allocated_count);
9893        } else {
9894                igb_vmdq_set_loopback_pf(hw, false);
9895                igb_vmdq_set_replication_pf(hw, false);
9896        }
9897}
9898
9899static void igb_init_dmac(struct igb_adapter *adapter, u32 pba)
9900{
9901        struct e1000_hw *hw = &adapter->hw;
9902        u32 dmac_thr;
9903        u16 hwm;
9904        u32 reg;
9905
9906        if (hw->mac.type > e1000_82580) {
9907                if (adapter->flags & IGB_FLAG_DMAC) {
9908                        /* force threshold to 0. */
9909                        wr32(E1000_DMCTXTH, 0);
9910
9911                        /* DMA Coalescing high water mark needs to be greater
9912                         * than the Rx threshold. Set hwm to PBA - max frame
9913                         * size in 16B units, capping it at PBA - 6KB.
9914                         */
9915                        hwm = 64 * (pba - 6);
9916                        reg = rd32(E1000_FCRTC);
9917                        reg &= ~E1000_FCRTC_RTH_COAL_MASK;
9918                        reg |= ((hwm << E1000_FCRTC_RTH_COAL_SHIFT)
9919                                & E1000_FCRTC_RTH_COAL_MASK);
9920                        wr32(E1000_FCRTC, reg);
9921
9922                        /* Set the DMA Coalescing Rx threshold to PBA - 2 * max
9923                         * frame size, capping it at PBA - 10KB.
9924                         */
9925                        dmac_thr = pba - 10;
9926                        reg = rd32(E1000_DMACR);
9927                        reg &= ~E1000_DMACR_DMACTHR_MASK;
9928                        reg |= ((dmac_thr << E1000_DMACR_DMACTHR_SHIFT)
9929                                & E1000_DMACR_DMACTHR_MASK);
9930
9931                        /* transition to L0x or L1 if available..*/
9932                        reg |= (E1000_DMACR_DMAC_EN | E1000_DMACR_DMAC_LX_MASK);
9933
9934                        /* watchdog timer= +-1000 usec in 32usec intervals */
9935                        reg |= (1000 >> 5);
9936
9937                        /* Disable BMC-to-OS Watchdog Enable */
9938                        if (hw->mac.type != e1000_i354)
9939                                reg &= ~E1000_DMACR_DC_BMC2OSW_EN;
9940                        wr32(E1000_DMACR, reg);
9941
9942                        /* no lower threshold to disable
9943                         * coalescing(smart fifb)-UTRESH=0
9944                         */
9945                        wr32(E1000_DMCRTRH, 0);
9946
9947                        reg = (IGB_DMCTLX_DCFLUSH_DIS | 0x4);
9948
9949                        wr32(E1000_DMCTLX, reg);
9950
9951                        /* free space in tx packet buffer to wake from
9952                         * DMA coal
9953                         */
9954                        wr32(E1000_DMCTXTH, (IGB_MIN_TXPBSIZE -
9955                             (IGB_TX_BUF_4096 + adapter->max_frame_size)) >> 6);
9956                }
9957
9958                if (hw->mac.type >= e1000_i210 ||
9959                    (adapter->flags & IGB_FLAG_DMAC)) {
9960                        reg = rd32(E1000_PCIEMISC);
9961                        reg |= E1000_PCIEMISC_LX_DECISION;
9962                        wr32(E1000_PCIEMISC, reg);
9963                } /* endif adapter->dmac is not disabled */
9964        } else if (hw->mac.type == e1000_82580) {
9965                u32 reg = rd32(E1000_PCIEMISC);
9966
9967                wr32(E1000_PCIEMISC, reg & ~E1000_PCIEMISC_LX_DECISION);
9968                wr32(E1000_DMACR, 0);
9969        }
9970}
9971
9972/**
9973 *  igb_read_i2c_byte - Reads 8 bit word over I2C
9974 *  @hw: pointer to hardware structure
9975 *  @byte_offset: byte offset to read
9976 *  @dev_addr: device address
9977 *  @data: value read
9978 *
9979 *  Performs byte read operation over I2C interface at
9980 *  a specified device address.
9981 **/
9982s32 igb_read_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
9983                      u8 dev_addr, u8 *data)
9984{
9985        struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
9986        struct i2c_client *this_client = adapter->i2c_client;
9987        s32 status;
9988        u16 swfw_mask = 0;
9989
9990        if (!this_client)
9991                return E1000_ERR_I2C;
9992
9993        swfw_mask = E1000_SWFW_PHY0_SM;
9994
9995        if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
9996                return E1000_ERR_SWFW_SYNC;
9997
9998        status = i2c_smbus_read_byte_data(this_client, byte_offset);
9999        hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10000
10001        if (status < 0)
10002                return E1000_ERR_I2C;
10003        else {
10004                *data = status;
10005                return 0;
10006        }
10007}
10008
10009/**
10010 *  igb_write_i2c_byte - Writes 8 bit word over I2C
10011 *  @hw: pointer to hardware structure
10012 *  @byte_offset: byte offset to write
10013 *  @dev_addr: device address
10014 *  @data: value to write
10015 *
10016 *  Performs byte write operation over I2C interface at
10017 *  a specified device address.
10018 **/
10019s32 igb_write_i2c_byte(struct e1000_hw *hw, u8 byte_offset,
10020                       u8 dev_addr, u8 data)
10021{
10022        struct igb_adapter *adapter = container_of(hw, struct igb_adapter, hw);
10023        struct i2c_client *this_client = adapter->i2c_client;
10024        s32 status;
10025        u16 swfw_mask = E1000_SWFW_PHY0_SM;
10026
10027        if (!this_client)
10028                return E1000_ERR_I2C;
10029
10030        if (hw->mac.ops.acquire_swfw_sync(hw, swfw_mask))
10031                return E1000_ERR_SWFW_SYNC;
10032        status = i2c_smbus_write_byte_data(this_client, byte_offset, data);
10033        hw->mac.ops.release_swfw_sync(hw, swfw_mask);
10034
10035        if (status)
10036                return E1000_ERR_I2C;
10037        else
10038                return 0;
10039
10040}
10041
10042int igb_reinit_queues(struct igb_adapter *adapter)
10043{
10044        struct net_device *netdev = adapter->netdev;
10045        struct pci_dev *pdev = adapter->pdev;
10046        int err = 0;
10047
10048        if (netif_running(netdev))
10049                igb_close(netdev);
10050
10051        igb_reset_interrupt_capability(adapter);
10052
10053        if (igb_init_interrupt_scheme(adapter, true)) {
10054                dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
10055                return -ENOMEM;
10056        }
10057
10058        if (netif_running(netdev))
10059                err = igb_open(netdev);
10060
10061        return err;
10062}
10063
10064static void igb_nfc_filter_exit(struct igb_adapter *adapter)
10065{
10066        struct igb_nfc_filter *rule;
10067
10068        spin_lock(&adapter->nfc_lock);
10069
10070        hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10071                igb_erase_filter(adapter, rule);
10072
10073        hlist_for_each_entry(rule, &adapter->cls_flower_list, nfc_node)
10074                igb_erase_filter(adapter, rule);
10075
10076        spin_unlock(&adapter->nfc_lock);
10077}
10078
10079static void igb_nfc_filter_restore(struct igb_adapter *adapter)
10080{
10081        struct igb_nfc_filter *rule;
10082
10083        spin_lock(&adapter->nfc_lock);
10084
10085        hlist_for_each_entry(rule, &adapter->nfc_filter_list, nfc_node)
10086                igb_add_filter(adapter, rule);
10087
10088        spin_unlock(&adapter->nfc_lock);
10089}
10090/* igb_main.c */
10091