drivers/net/ethernet/intel/ice/ice_main.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Copyright (c) 2018, Intel Corporation. */
   3
   4 /* Intel(R) Ethernet Connection E800 Series Linux Driver */
   5
   6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
   7
   8 #include "ice.h"
   9 #include "ice_lib.h"
  10
  11 #define DRV_VERSION     "0.7.2-k"
  12 #define DRV_SUMMARY     "Intel(R) Ethernet Connection E800 Series Linux Driver"
  13 const char ice_drv_ver[] = DRV_VERSION;
  14 static const char ice_driver_string[] = DRV_SUMMARY;
  15 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
  16
  17 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
  18 MODULE_DESCRIPTION(DRV_SUMMARY);
  19 MODULE_LICENSE("GPL v2");
  20 MODULE_VERSION(DRV_VERSION);
  21
  22 static int debug = -1;
  23 module_param(debug, int, 0644);
  24 #ifndef CONFIG_DYNAMIC_DEBUG
  25 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
  26 #else
  27 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
  28 #endif /* !CONFIG_DYNAMIC_DEBUG */
  29
  30 static struct workqueue_struct *ice_wq;
  31 static const struct net_device_ops ice_netdev_ops;
  32
  33 static void ice_pf_dis_all_vsi(struct ice_pf *pf);
  34 static void ice_rebuild(struct ice_pf *pf);
  35
  36 static void ice_vsi_release_all(struct ice_pf *pf);
  37 static void ice_update_vsi_stats(struct ice_vsi *vsi);
  38 static void ice_update_pf_stats(struct ice_pf *pf);
  39
  40 /**
  41  * ice_get_tx_pending - returns number of Tx descriptors not processed
  42  * @ring: the ring of descriptors
  43  */
  44 static u32 ice_get_tx_pending(struct ice_ring *ring)
  45 {
  46         u32 head, tail;
  47
  48         head = ring->next_to_clean;
  49         tail = readl(ring->tail);
  50
  51         if (head != tail)
  52                 return (head < tail) ?
  53                         tail - head : (tail + ring->count - head);
  54         return 0;
  55 }
  56
  57 /**
  58  * ice_check_for_hang_subtask - check for and recover hung queues
  59  * @pf: pointer to PF struct
  60  */
  61 static void ice_check_for_hang_subtask(struct ice_pf *pf)
  62 {
  63         struct ice_vsi *vsi = NULL;
  64         unsigned int i;
  65         u32 v, v_idx;
  66         int packets;
  67
  68         ice_for_each_vsi(pf, v)
  69                 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
  70                         vsi = pf->vsi[v];
  71                         break;
  72                 }
  73
  74         if (!vsi || test_bit(__ICE_DOWN, vsi->state))
  75                 return;
  76
  77         if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
  78                 return;
  79
  80         for (i = 0; i < vsi->num_txq; i++) {
  81                 struct ice_ring *tx_ring = vsi->tx_rings[i];
  82
  83                 if (tx_ring && tx_ring->desc) {
  84                         int itr = ICE_ITR_NONE;
  85
  86                         /* If packet counter has not changed the queue is
  87                          * likely stalled, so force an interrupt for this
  88                          * queue.
  89                          *
  90                          * prev_pkt would be negative if there was no
  91                          * pending work.
  92                          */
  93                         packets = tx_ring->stats.pkts & INT_MAX;
  94                         if (tx_ring->tx_stats.prev_pkt == packets) {
  95                                 /* Trigger sw interrupt to revive the queue */
  96                                 v_idx = tx_ring->q_vector->v_idx;
  97                                 wr32(&vsi->back->hw,
  98                                      GLINT_DYN_CTL(vsi->hw_base_vector + v_idx),
  99                                      (itr << GLINT_DYN_CTL_ITR_INDX_S) |
 100                                      GLINT_DYN_CTL_SWINT_TRIG_M |
 101                                      GLINT_DYN_CTL_INTENA_MSK_M);
 102                                 continue;
 103                         }
 104
 105                         /* Memory barrier between read of packet count and call
 106                          * to ice_get_tx_pending()
 107                          */
 108                         smp_rmb();
 109                         tx_ring->tx_stats.prev_pkt =
 110                             ice_get_tx_pending(tx_ring) ? packets : -1;
 111                 }
 112         }
 113 }
 114
 115 /**
 116  * ice_add_mac_to_sync_list - creates list of mac addresses to be synced
 117  * @netdev: the net device on which the sync is happening
 118  * @addr: mac address to sync
 119  *
 120  * This is a callback function which is called by the in kernel device sync
 121  * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
 122  * populates the tmp_sync_list, which is later used by ice_add_mac to add the
 123  * mac filters from the hardware.
 124  */
 125 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
 126 {
 127         struct ice_netdev_priv *np = netdev_priv(netdev);
 128         struct ice_vsi *vsi = np->vsi;
 129
 130         if (ice_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr))
 131                 return -EINVAL;
 132
 133         return 0;
 134 }
 135
 136 /**
 137  * ice_add_mac_to_unsync_list - creates list of mac addresses to be unsynced
 138  * @netdev: the net device on which the unsync is happening
 139  * @addr: mac address to unsync
 140  *
 141  * This is a callback function which is called by the in kernel device unsync
 142  * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
 143  * populates the tmp_unsync_list, which is later used by ice_remove_mac to
 144  * delete the mac filters from the hardware.
 145  */
 146 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
 147 {
 148         struct ice_netdev_priv *np = netdev_priv(netdev);
 149         struct ice_vsi *vsi = np->vsi;
 150
 151         if (ice_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr))
 152                 return -EINVAL;
 153
 154         return 0;
 155 }
 156
 157 /**
 158  * ice_vsi_fltr_changed - check if filter state changed
 159  * @vsi: VSI to be checked
 160  *
 161  * returns true if filter state has changed, false otherwise.
 162  */
 163 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
 164 {
 165         return test_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags) ||
 166                test_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags) ||
 167                test_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
 168 }
 169
 170 /**
 171  * ice_vsi_sync_fltr - Update the VSI filter list to the HW
 172  * @vsi: ptr to the VSI
 173  *
 174  * Push any outstanding VSI filter changes through the AdminQ.
 175  */
 176 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
 177 {
 178         struct device *dev = &vsi->back->pdev->dev;
 179         struct net_device *netdev = vsi->netdev;
 180         bool promisc_forced_on = false;
 181         struct ice_pf *pf = vsi->back;
 182         struct ice_hw *hw = &pf->hw;
 183         enum ice_status status = 0;
 184         u32 changed_flags = 0;
 185         int err = 0;
 186
 187         if (!vsi->netdev)
 188                 return -EINVAL;
 189
 190         while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state))
 191                 usleep_range(1000, 2000);
 192
 193         changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
 194         vsi->current_netdev_flags = vsi->netdev->flags;
 195
 196         INIT_LIST_HEAD(&vsi->tmp_sync_list);
 197         INIT_LIST_HEAD(&vsi->tmp_unsync_list);
 198
 199         if (ice_vsi_fltr_changed(vsi)) {
 200                 clear_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
 201                 clear_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
 202                 clear_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
 203
 204                 /* grab the netdev's addr_list_lock */
 205                 netif_addr_lock_bh(netdev);
 206                 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
 207                               ice_add_mac_to_unsync_list);
 208                 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
 209                               ice_add_mac_to_unsync_list);
 210                 /* our temp lists are populated. release lock */
 211                 netif_addr_unlock_bh(netdev);
 212         }
 213
 214         /* Remove mac addresses in the unsync list */
 215         status = ice_remove_mac(hw, &vsi->tmp_unsync_list);
 216         ice_free_fltr_list(dev, &vsi->tmp_unsync_list);
 217         if (status) {
 218                 netdev_err(netdev, "Failed to delete MAC filters\n");
 219                 /* if we failed because of alloc failures, just bail */
 220                 if (status == ICE_ERR_NO_MEMORY) {
 221                         err = -ENOMEM;
 222                         goto out;
 223                 }
 224         }
 225
 226         /* Add mac addresses in the sync list */
 227         status = ice_add_mac(hw, &vsi->tmp_sync_list);
 228         ice_free_fltr_list(dev, &vsi->tmp_sync_list);
 229         if (status) {
 230                 netdev_err(netdev, "Failed to add MAC filters\n");
 231                 /* If there is no more space for new umac filters, vsi
 232                  * should go into promiscuous mode. There should be some
 233                  * space reserved for promiscuous filters.
 234                  */
 235                 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
 236                     !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC,
 237                                       vsi->state)) {
 238                         promisc_forced_on = true;
 239                         netdev_warn(netdev,
 240                                     "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
 241                                     vsi->vsi_num);
 242                 } else {
 243                         err = -EIO;
 244                         goto out;
 245                 }
 246         }
 247         /* check for changes in promiscuous modes */
 248         if (changed_flags & IFF_ALLMULTI)
 249                 netdev_warn(netdev, "Unsupported configuration\n");
 250
 251         if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
 252             test_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags)) {
 253                 clear_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
 254                 if (vsi->current_netdev_flags & IFF_PROMISC) {
 255                         /* Apply TX filter rule to get traffic from VMs */
 256                         status = ice_cfg_dflt_vsi(hw, vsi->idx, true,
 257                                                   ICE_FLTR_TX);
 258                         if (status) {
 259                                 netdev_err(netdev, "Error setting default VSI %i tx rule\n",
 260                                            vsi->vsi_num);
 261                                 vsi->current_netdev_flags &= ~IFF_PROMISC;
 262                                 err = -EIO;
 263                                 goto out_promisc;
 264                         }
 265                         /* Apply RX filter rule to get traffic from wire */
 266                         status = ice_cfg_dflt_vsi(hw, vsi->idx, true,
 267                                                   ICE_FLTR_RX);
 268                         if (status) {
 269                                 netdev_err(netdev, "Error setting default VSI %i rx rule\n",
 270                                            vsi->vsi_num);
 271                                 vsi->current_netdev_flags &= ~IFF_PROMISC;
 272                                 err = -EIO;
 273                                 goto out_promisc;
 274                         }
 275                 } else {
 276                         /* Clear TX filter rule to stop traffic from VMs */
 277                         status = ice_cfg_dflt_vsi(hw, vsi->idx, false,
 278                                                   ICE_FLTR_TX);
 279                         if (status) {
 280                                 netdev_err(netdev, "Error clearing default VSI %i tx rule\n",
 281                                            vsi->vsi_num);
 282                                 vsi->current_netdev_flags |= IFF_PROMISC;
 283                                 err = -EIO;
 284                                 goto out_promisc;
 285                         }
 286                         /* Clear RX filter to remove traffic from wire */
 287                         status = ice_cfg_dflt_vsi(hw, vsi->idx, false,
 288                                                   ICE_FLTR_RX);
 289                         if (status) {
 290                                 netdev_err(netdev, "Error clearing default VSI %i rx rule\n",
 291                                            vsi->vsi_num);
 292                                 vsi->current_netdev_flags |= IFF_PROMISC;
 293                                 err = -EIO;
 294                                 goto out_promisc;
 295                         }
 296                 }
 297         }
 298         goto exit;
 299
 300 out_promisc:
 301         set_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
 302         goto exit;
 303 out:
 304         /* if something went wrong then set the changed flag so we try again */
 305         set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
 306         set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
 307 exit:
 308         clear_bit(__ICE_CFG_BUSY, vsi->state);
 309         return err;
 310 }
 311
 312 /**
 313  * ice_sync_fltr_subtask - Sync the VSI filter list with HW
 314  * @pf: board private structure
 315  */
 316 static void ice_sync_fltr_subtask(struct ice_pf *pf)
 317 {
 318         int v;
 319
 320         if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
 321                 return;
 322
 323         clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
 324
 325         for (v = 0; v < pf->num_alloc_vsi; v++)
 326                 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
 327                     ice_vsi_sync_fltr(pf->vsi[v])) {
 328                         /* come back and try again later */
 329                         set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
 330                         break;
 331                 }
 332 }
 333
 334 /**
 335  * ice_prepare_for_reset - prep for the core to reset
 336  * @pf: board private structure
 337  *
 338  * Inform or close all dependent features in prep for reset.
 339  */
 340 static void
 341 ice_prepare_for_reset(struct ice_pf *pf)
 342 {
 343         struct ice_hw *hw = &pf->hw;
 344
 345         /* Notify VFs of impending reset */
 346         if (ice_check_sq_alive(hw, &hw->mailboxq))
 347                 ice_vc_notify_reset(pf);
 348
 349         /* disable the VSIs and their queues that are not already DOWN */
 350         ice_pf_dis_all_vsi(pf);
 351
 352         ice_shutdown_all_ctrlq(hw);
 353
 354         set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 355 }
 356
 357 /**
 358  * ice_do_reset - Initiate one of many types of resets
 359  * @pf: board private structure
 360  * @reset_type: reset type requested
 361  * before this function was called.
 362  */
 363 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 364 {
 365         struct device *dev = &pf->pdev->dev;
 366         struct ice_hw *hw = &pf->hw;
 367
 368         dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
 369         WARN_ON(in_interrupt());
 370
 371         ice_prepare_for_reset(pf);
 372
 373         /* trigger the reset */
 374         if (ice_reset(hw, reset_type)) {
 375                 dev_err(dev, "reset %d failed\n", reset_type);
 376                 set_bit(__ICE_RESET_FAILED, pf->state);
 377                 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 378                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 379                 clear_bit(__ICE_PFR_REQ, pf->state);
 380                 clear_bit(__ICE_CORER_REQ, pf->state);
 381                 clear_bit(__ICE_GLOBR_REQ, pf->state);
 382                 return;
 383         }
 384
 385         /* PFR is a bit of a special case because it doesn't result in an OICR
 386          * interrupt. So for PFR, rebuild after the reset and clear the reset-
 387          * associated state bits.
 388          */
 389         if (reset_type == ICE_RESET_PFR) {
 390                 pf->pfr_count++;
 391                 ice_rebuild(pf);
 392                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 393                 clear_bit(__ICE_PFR_REQ, pf->state);
 394         }
 395 }
 396
 397 /**
 398  * ice_reset_subtask - Set up for resetting the device and driver
 399  * @pf: board private structure
 400  */
 401 static void ice_reset_subtask(struct ice_pf *pf)
 402 {
 403         enum ice_reset_req reset_type = ICE_RESET_INVAL;
 404
 405         /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
 406          * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
 407          * of reset is pending and sets bits in pf->state indicating the reset
 408          * type and __ICE_RESET_OICR_RECV.  So, if the latter bit is set
 409          * prepare for pending reset if not already (for PF software-initiated
 410          * global resets the software should already be prepared for it as
 411          * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
 412          * by firmware or software on other PFs, that bit is not set so prepare
 413          * for the reset now), poll for reset done, rebuild and return.
 414          */
 415         if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
 416                 clear_bit(__ICE_GLOBR_RECV, pf->state);
 417                 clear_bit(__ICE_CORER_RECV, pf->state);
 418                 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
 419                         ice_prepare_for_reset(pf);
 420
 421                 /* make sure we are ready to rebuild */
 422                 if (ice_check_reset(&pf->hw)) {
 423                         set_bit(__ICE_RESET_FAILED, pf->state);
 424                 } else {
 425                         /* done with reset. start rebuild */
 426                         pf->hw.reset_ongoing = false;
 427                         ice_rebuild(pf);
 428                         /* clear bit to resume normal operations, but
 429                          * ICE_NEEDS_RESTART bit is set incase rebuild failed
 430                          */
 431                         clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 432                         clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 433                         clear_bit(__ICE_PFR_REQ, pf->state);
 434                         clear_bit(__ICE_CORER_REQ, pf->state);
 435                         clear_bit(__ICE_GLOBR_REQ, pf->state);
 436                 }
 437
 438                 return;
 439         }
 440
 441         /* No pending resets to finish processing. Check for new resets */
 442         if (test_bit(__ICE_PFR_REQ, pf->state))
 443                 reset_type = ICE_RESET_PFR;
 444         if (test_bit(__ICE_CORER_REQ, pf->state))
 445                 reset_type = ICE_RESET_CORER;
 446         if (test_bit(__ICE_GLOBR_REQ, pf->state))
 447                 reset_type = ICE_RESET_GLOBR;
 448         /* If no valid reset type requested just return */
 449         if (reset_type == ICE_RESET_INVAL)
 450                 return;
 451
 452         /* reset if not already down or busy */
 453         if (!test_bit(__ICE_DOWN, pf->state) &&
 454             !test_bit(__ICE_CFG_BUSY, pf->state)) {
 455                 ice_do_reset(pf, reset_type);
 456         }
 457 }
 458
 459 /**
 460  * ice_print_link_msg - print link up or down message
 461  * @vsi: the VSI whose link status is being queried
 462  * @isup: boolean for if the link is now up or down
 463  */
 464 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
 465 {
 466         const char *speed;
 467         const char *fc;
 468
 469         if (vsi->current_isup == isup)
 470                 return;
 471
 472         vsi->current_isup = isup;
 473
 474         if (!isup) {
 475                 netdev_info(vsi->netdev, "NIC Link is Down\n");
 476                 return;
 477         }
 478
 479         switch (vsi->port_info->phy.link_info.link_speed) {
 480         case ICE_AQ_LINK_SPEED_40GB:
 481                 speed = "40 G";
 482                 break;
 483         case ICE_AQ_LINK_SPEED_25GB:
 484                 speed = "25 G";
 485                 break;
 486         case ICE_AQ_LINK_SPEED_20GB:
 487                 speed = "20 G";
 488                 break;
 489         case ICE_AQ_LINK_SPEED_10GB:
 490                 speed = "10 G";
 491                 break;
 492         case ICE_AQ_LINK_SPEED_5GB:
 493                 speed = "5 G";
 494                 break;
 495         case ICE_AQ_LINK_SPEED_2500MB:
 496                 speed = "2.5 G";
 497                 break;
 498         case ICE_AQ_LINK_SPEED_1000MB:
 499                 speed = "1 G";
 500                 break;
 501         case ICE_AQ_LINK_SPEED_100MB:
 502                 speed = "100 M";
 503                 break;
 504         default:
 505                 speed = "Unknown";
 506                 break;
 507         }
 508
 509         switch (vsi->port_info->fc.current_mode) {
 510         case ICE_FC_FULL:
 511                 fc = "RX/TX";
 512                 break;
 513         case ICE_FC_TX_PAUSE:
 514                 fc = "TX";
 515                 break;
 516         case ICE_FC_RX_PAUSE:
 517                 fc = "RX";
 518                 break;
 519         default:
 520                 fc = "Unknown";
 521                 break;
 522         }
 523
 524         netdev_info(vsi->netdev, "NIC Link is up %sbps, Flow Control: %s\n",
 525                     speed, fc);
 526 }
 527
 528 /**
 529  * ice_vsi_link_event - update the vsi's netdev
 530  * @vsi: the vsi on which the link event occurred
 531  * @link_up: whether or not the vsi needs to be set up or down
 532  */
 533 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
 534 {
 535         if (!vsi || test_bit(__ICE_DOWN, vsi->state))
 536                 return;
 537
 538         if (vsi->type == ICE_VSI_PF) {
 539                 if (!vsi->netdev) {
 540                         dev_dbg(&vsi->back->pdev->dev,
 541                                 "vsi->netdev is not initialized!\n");
 542                         return;
 543                 }
 544                 if (link_up) {
 545                         netif_carrier_on(vsi->netdev);
 546                         netif_tx_wake_all_queues(vsi->netdev);
 547                 } else {
 548                         netif_carrier_off(vsi->netdev);
 549                         netif_tx_stop_all_queues(vsi->netdev);
 550                 }
 551         }
 552 }
 553
 554 /**
 555  * ice_link_event - process the link event
 556  * @pf: pf that the link event is associated with
 557  * @pi: port_info for the port that the link event is associated with
 558  *
 559  * Returns -EIO if ice_get_link_status() fails
 560  * Returns 0 on success
 561  */
 562 static int
 563 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi)
 564 {
 565         u8 new_link_speed, old_link_speed;
 566         struct ice_phy_info *phy_info;
 567         bool new_link_same_as_old;
 568         bool new_link, old_link;
 569         u8 lport;
 570         u16 v;
 571
 572         phy_info = &pi->phy;
 573         phy_info->link_info_old = phy_info->link_info;
 574         /* Force ice_get_link_status() to update link info */
 575         phy_info->get_link_info = true;
 576
 577         old_link = (phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
 578         old_link_speed = phy_info->link_info_old.link_speed;
 579
 580         lport = pi->lport;
 581         if (ice_get_link_status(pi, &new_link)) {
 582                 dev_dbg(&pf->pdev->dev,
 583                         "Could not get link status for port %d\n", lport);
 584                 return -EIO;
 585         }
 586
 587         new_link_speed = phy_info->link_info.link_speed;
 588
 589         new_link_same_as_old = (new_link == old_link &&
 590                                 new_link_speed == old_link_speed);
 591
 592         ice_for_each_vsi(pf, v) {
 593                 struct ice_vsi *vsi = pf->vsi[v];
 594
 595                 if (!vsi || !vsi->port_info)
 596                         continue;
 597
 598                 if (new_link_same_as_old &&
 599                     (test_bit(__ICE_DOWN, vsi->state) ||
 600                     new_link == netif_carrier_ok(vsi->netdev)))
 601                         continue;
 602
 603                 if (vsi->port_info->lport == lport) {
 604                         ice_print_link_msg(vsi, new_link);
 605                         ice_vsi_link_event(vsi, new_link);
 606                 }
 607         }
 608
 609         ice_vc_notify_link_state(pf);
 610
 611         return 0;
 612 }
 613
 614 /**
 615  * ice_watchdog_subtask - periodic tasks not using event driven scheduling
 616  * @pf: board private structure
 617  */
 618 static void ice_watchdog_subtask(struct ice_pf *pf)
 619 {
 620         int i;
 621
 622         /* if interface is down do nothing */
 623         if (test_bit(__ICE_DOWN, pf->state) ||
 624             test_bit(__ICE_CFG_BUSY, pf->state))
 625                 return;
 626
 627         /* make sure we don't do these things too often */
 628         if (time_before(jiffies,
 629                         pf->serv_tmr_prev + pf->serv_tmr_period))
 630                 return;
 631
 632         pf->serv_tmr_prev = jiffies;
 633
 634         if (ice_link_event(pf, pf->hw.port_info))
 635                 dev_dbg(&pf->pdev->dev, "ice_link_event failed\n");
 636
 637         /* Update the stats for active netdevs so the network stack
 638          * can look at updated numbers whenever it cares to
 639          */
 640         ice_update_pf_stats(pf);
 641         for (i = 0; i < pf->num_alloc_vsi; i++)
 642                 if (pf->vsi[i] && pf->vsi[i]->netdev)
 643                         ice_update_vsi_stats(pf->vsi[i]);
 644 }
 645
 646 /**
 647  * __ice_clean_ctrlq - helper function to clean controlq rings
 648  * @pf: ptr to struct ice_pf
 649  * @q_type: specific Control queue type
 650  */
 651 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
 652 {
 653         struct ice_rq_event_info event;
 654         struct ice_hw *hw = &pf->hw;
 655         struct ice_ctl_q_info *cq;
 656         u16 pending, i = 0;
 657         const char *qtype;
 658         u32 oldval, val;
 659
 660         /* Do not clean control queue if/when PF reset fails */
 661         if (test_bit(__ICE_RESET_FAILED, pf->state))
 662                 return 0;
 663
 664         switch (q_type) {
 665         case ICE_CTL_Q_ADMIN:
 666                 cq = &hw->adminq;
 667                 qtype = "Admin";
 668                 break;
 669         case ICE_CTL_Q_MAILBOX:
 670                 cq = &hw->mailboxq;
 671                 qtype = "Mailbox";
 672                 break;
 673         default:
 674                 dev_warn(&pf->pdev->dev, "Unknown control queue type 0x%x\n",
 675                          q_type);
 676                 return 0;
 677         }
 678
 679         /* check for error indications - PF_xx_AxQLEN register layout for
 680          * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
 681          */
 682         val = rd32(hw, cq->rq.len);
 683         if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 684                    PF_FW_ARQLEN_ARQCRIT_M)) {
 685                 oldval = val;
 686                 if (val & PF_FW_ARQLEN_ARQVFE_M)
 687                         dev_dbg(&pf->pdev->dev,
 688                                 "%s Receive Queue VF Error detected\n", qtype);
 689                 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
 690                         dev_dbg(&pf->pdev->dev,
 691                                 "%s Receive Queue Overflow Error detected\n",
 692                                 qtype);
 693                 }
 694                 if (val & PF_FW_ARQLEN_ARQCRIT_M)
 695                         dev_dbg(&pf->pdev->dev,
 696                                 "%s Receive Queue Critical Error detected\n",
 697                                 qtype);
 698                 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 699                          PF_FW_ARQLEN_ARQCRIT_M);
 700                 if (oldval != val)
 701                         wr32(hw, cq->rq.len, val);
 702         }
 703
 704         val = rd32(hw, cq->sq.len);
 705         if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 706                    PF_FW_ATQLEN_ATQCRIT_M)) {
 707                 oldval = val;
 708                 if (val & PF_FW_ATQLEN_ATQVFE_M)
 709                         dev_dbg(&pf->pdev->dev,
 710                                 "%s Send Queue VF Error detected\n", qtype);
 711                 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
 712                         dev_dbg(&pf->pdev->dev,
 713                                 "%s Send Queue Overflow Error detected\n",
 714                                 qtype);
 715                 }
 716                 if (val & PF_FW_ATQLEN_ATQCRIT_M)
 717                         dev_dbg(&pf->pdev->dev,
 718                                 "%s Send Queue Critical Error detected\n",
 719                                 qtype);
 720                 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 721                          PF_FW_ATQLEN_ATQCRIT_M);
 722                 if (oldval != val)
 723                         wr32(hw, cq->sq.len, val);
 724         }
 725
 726         event.buf_len = cq->rq_buf_size;
 727         event.msg_buf = devm_kzalloc(&pf->pdev->dev, event.buf_len,
 728                                      GFP_KERNEL);
 729         if (!event.msg_buf)
 730                 return 0;
 731
 732         do {
 733                 enum ice_status ret;
 734                 u16 opcode;
 735
 736                 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
 737                 if (ret == ICE_ERR_AQ_NO_WORK)
 738                         break;
 739                 if (ret) {
 740                         dev_err(&pf->pdev->dev,
 741                                 "%s Receive Queue event error %d\n", qtype,
 742                                 ret);
 743                         break;
 744                 }
 745
 746                 opcode = le16_to_cpu(event.desc.opcode);
 747
 748                 switch (opcode) {
 749                 case ice_mbx_opc_send_msg_to_pf:
 750                         ice_vc_process_vf_msg(pf, &event);
 751                         break;
 752                 case ice_aqc_opc_fw_logging:
 753                         ice_output_fw_log(hw, &event.desc, event.msg_buf);
 754                         break;
 755                 default:
 756                         dev_dbg(&pf->pdev->dev,
 757                                 "%s Receive Queue unknown event 0x%04x ignored\n",
 758                                 qtype, opcode);
 759                         break;
 760                 }
 761         } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
 762
 763         devm_kfree(&pf->pdev->dev, event.msg_buf);
 764
 765         return pending && (i == ICE_DFLT_IRQ_WORK);
 766 }
 767
 768 /**
 769  * ice_ctrlq_pending - check if there is a difference between ntc and ntu
 770  * @hw: pointer to hardware info
 771  * @cq: control queue information
 772  *
 773  * returns true if there are pending messages in a queue, false if there aren't
 774  */
 775 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
 776 {
 777         u16 ntu;
 778
 779         ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
 780         return cq->rq.next_to_clean != ntu;
 781 }
 782
 783 /**
 784  * ice_clean_adminq_subtask - clean the AdminQ rings
 785  * @pf: board private structure
 786  */
 787 static void ice_clean_adminq_subtask(struct ice_pf *pf)
 788 {
 789         struct ice_hw *hw = &pf->hw;
 790
 791         if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
 792                 return;
 793
 794         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
 795                 return;
 796
 797         clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
 798
 799         /* There might be a situation where new messages arrive to a control
 800          * queue between processing the last message and clearing the
 801          * EVENT_PENDING bit. So before exiting, check queue head again (using
 802          * ice_ctrlq_pending) and process new messages if any.
 803          */
 804         if (ice_ctrlq_pending(hw, &hw->adminq))
 805                 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
 806
 807         ice_flush(hw);
 808 }
 809
 810 /**
 811  * ice_clean_mailboxq_subtask - clean the MailboxQ rings
 812  * @pf: board private structure
 813  */
 814 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
 815 {
 816         struct ice_hw *hw = &pf->hw;
 817
 818         if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
 819                 return;
 820
 821         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
 822                 return;
 823
 824         clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
 825
 826         if (ice_ctrlq_pending(hw, &hw->mailboxq))
 827                 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
 828
 829         ice_flush(hw);
 830 }
 831
 832 /**
 833  * ice_service_task_schedule - schedule the service task to wake up
 834  * @pf: board private structure
 835  *
 836  * If not already scheduled, this puts the task into the work queue.
 837  */
 838 static void ice_service_task_schedule(struct ice_pf *pf)
 839 {
 840         if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
 841             !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
 842             !test_bit(__ICE_NEEDS_RESTART, pf->state))
 843                 queue_work(ice_wq, &pf->serv_task);
 844 }
 845
 846 /**
 847  * ice_service_task_complete - finish up the service task
 848  * @pf: board private structure
 849  */
 850 static void ice_service_task_complete(struct ice_pf *pf)
 851 {
 852         WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
 853
 854         /* force memory (pf->state) to sync before next service task */
 855         smp_mb__before_atomic();
 856         clear_bit(__ICE_SERVICE_SCHED, pf->state);
 857 }
 858
 859 /**
 860  * ice_service_task_stop - stop service task and cancel works
 861  * @pf: board private structure
 862  */
 863 static void ice_service_task_stop(struct ice_pf *pf)
 864 {
 865         set_bit(__ICE_SERVICE_DIS, pf->state);
 866
 867         if (pf->serv_tmr.function)
 868                 del_timer_sync(&pf->serv_tmr);
 869         if (pf->serv_task.func)
 870                 cancel_work_sync(&pf->serv_task);
 871
 872         clear_bit(__ICE_SERVICE_SCHED, pf->state);
 873 }
 874
 875 /**
 876  * ice_service_timer - timer callback to schedule service task
 877  * @t: pointer to timer_list
 878  */
 879 static void ice_service_timer(struct timer_list *t)
 880 {
 881         struct ice_pf *pf = from_timer(pf, t, serv_tmr);
 882
 883         mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
 884         ice_service_task_schedule(pf);
 885 }
 886
 887 /**
 888  * ice_handle_mdd_event - handle malicious driver detect event
 889  * @pf: pointer to the PF structure
 890  *
 891  * Called from service task. OICR interrupt handler indicates MDD event
 892  */
 893 static void ice_handle_mdd_event(struct ice_pf *pf)
 894 {
 895         struct ice_hw *hw = &pf->hw;
 896         bool mdd_detected = false;
 897         u32 reg;
 898         int i;
 899
 900         if (!test_bit(__ICE_MDD_EVENT_PENDING, pf->state))
 901                 return;
 902
 903         /* find what triggered the MDD event */
 904         reg = rd32(hw, GL_MDET_TX_PQM);
 905         if (reg & GL_MDET_TX_PQM_VALID_M) {
 906                 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
 907                                 GL_MDET_TX_PQM_PF_NUM_S;
 908                 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
 909                                 GL_MDET_TX_PQM_VF_NUM_S;
 910                 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
 911                                 GL_MDET_TX_PQM_MAL_TYPE_S;
 912                 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
 913                                 GL_MDET_TX_PQM_QNUM_S);
 914
 915                 if (netif_msg_tx_err(pf))
 916                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
 917                                  event, queue, pf_num, vf_num);
 918                 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
 919                 mdd_detected = true;
 920         }
 921
 922         reg = rd32(hw, GL_MDET_TX_TCLAN);
 923         if (reg & GL_MDET_TX_TCLAN_VALID_M) {
 924                 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
 925                                 GL_MDET_TX_TCLAN_PF_NUM_S;
 926                 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
 927                                 GL_MDET_TX_TCLAN_VF_NUM_S;
 928                 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
 929                                 GL_MDET_TX_TCLAN_MAL_TYPE_S;
 930                 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
 931                                 GL_MDET_TX_TCLAN_QNUM_S);
 932
 933                 if (netif_msg_rx_err(pf))
 934                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
 935                                  event, queue, pf_num, vf_num);
 936                 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
 937                 mdd_detected = true;
 938         }
 939
 940         reg = rd32(hw, GL_MDET_RX);
 941         if (reg & GL_MDET_RX_VALID_M) {
 942                 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
 943                                 GL_MDET_RX_PF_NUM_S;
 944                 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
 945                                 GL_MDET_RX_VF_NUM_S;
 946                 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
 947                                 GL_MDET_RX_MAL_TYPE_S;
 948                 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
 949                                 GL_MDET_RX_QNUM_S);
 950
 951                 if (netif_msg_rx_err(pf))
 952                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
 953                                  event, queue, pf_num, vf_num);
 954                 wr32(hw, GL_MDET_RX, 0xffffffff);
 955                 mdd_detected = true;
 956         }
 957
 958         if (mdd_detected) {
 959                 bool pf_mdd_detected = false;
 960
 961                 reg = rd32(hw, PF_MDET_TX_PQM);
 962                 if (reg & PF_MDET_TX_PQM_VALID_M) {
 963                         wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
 964                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
 965                         pf_mdd_detected = true;
 966                 }
 967
 968                 reg = rd32(hw, PF_MDET_TX_TCLAN);
 969                 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
 970                         wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
 971                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
 972                         pf_mdd_detected = true;
 973                 }
 974
 975                 reg = rd32(hw, PF_MDET_RX);
 976                 if (reg & PF_MDET_RX_VALID_M) {
 977                         wr32(hw, PF_MDET_RX, 0xFFFF);
 978                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
 979                         pf_mdd_detected = true;
 980                 }
 981                 /* Queue belongs to the PF initiate a reset */
 982                 if (pf_mdd_detected) {
 983                         set_bit(__ICE_NEEDS_RESTART, pf->state);
 984                         ice_service_task_schedule(pf);
 985                 }
 986         }
 987
 988         /* see if one of the VFs needs to be reset */
 989         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
 990                 struct ice_vf *vf = &pf->vf[i];
 991
 992                 reg = rd32(hw, VP_MDET_TX_PQM(i));
 993                 if (reg & VP_MDET_TX_PQM_VALID_M) {
 994                         wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
 995                         vf->num_mdd_events++;
 996                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
 997                                  i);
 998                 }
 999
1000                 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1001                 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1002                         wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1003                         vf->num_mdd_events++;
1004                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1005                                  i);
1006                 }
1007
1008                 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1009                 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1010                         wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1011                         vf->num_mdd_events++;
1012                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1013                                  i);
1014                 }
1015
1016                 reg = rd32(hw, VP_MDET_RX(i));
1017                 if (reg & VP_MDET_RX_VALID_M) {
1018                         wr32(hw, VP_MDET_RX(i), 0xFFFF);
1019                         vf->num_mdd_events++;
1020                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
1021                                  i);
1022                 }
1023
1024                 if (vf->num_mdd_events > ICE_DFLT_NUM_MDD_EVENTS_ALLOWED) {
1025                         dev_info(&pf->pdev->dev,
1026                                  "Too many MDD events on VF %d, disabled\n", i);
1027                         dev_info(&pf->pdev->dev,
1028                                  "Use PF Control I/F to re-enable the VF\n");
1029                         set_bit(ICE_VF_STATE_DIS, vf->vf_states);
1030                 }
1031         }
1032
1033         /* re-enable MDD interrupt cause */
1034         clear_bit(__ICE_MDD_EVENT_PENDING, pf->state);
1035         reg = rd32(hw, PFINT_OICR_ENA);
1036         reg |= PFINT_OICR_MAL_DETECT_M;
1037         wr32(hw, PFINT_OICR_ENA, reg);
1038         ice_flush(hw);
1039 }
1040
1041 /**
1042  * ice_service_task - manage and run subtasks
1043  * @work: pointer to work_struct contained by the PF struct
1044  */
1045 static void ice_service_task(struct work_struct *work)
1046 {
1047         struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
1048         unsigned long start_time = jiffies;
1049
1050         /* subtasks */
1051
1052         /* process reset requests first */
1053         ice_reset_subtask(pf);
1054
1055         /* bail if a reset/recovery cycle is pending or rebuild failed */
1056         if (ice_is_reset_in_progress(pf->state) ||
1057             test_bit(__ICE_SUSPENDED, pf->state) ||
1058             test_bit(__ICE_NEEDS_RESTART, pf->state)) {
1059                 ice_service_task_complete(pf);
1060                 return;
1061         }
1062
1063         ice_check_for_hang_subtask(pf);
1064         ice_sync_fltr_subtask(pf);
1065         ice_handle_mdd_event(pf);
1066         ice_process_vflr_event(pf);
1067         ice_watchdog_subtask(pf);
1068         ice_clean_adminq_subtask(pf);
1069         ice_clean_mailboxq_subtask(pf);
1070
1071         /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
1072         ice_service_task_complete(pf);
1073
1074         /* If the tasks have taken longer than one service timer period
1075          * or there is more work to be done, reset the service timer to
1076          * schedule the service task now.
1077          */
1078         if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
1079             test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
1080             test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
1081             test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
1082             test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1083                 mod_timer(&pf->serv_tmr, jiffies);
1084 }
1085
1086 /**
1087  * ice_set_ctrlq_len - helper function to set controlq length
1088  * @hw: pointer to the hw instance
1089  */
1090 static void ice_set_ctrlq_len(struct ice_hw *hw)
1091 {
1092         hw->adminq.num_rq_entries = ICE_AQ_LEN;
1093         hw->adminq.num_sq_entries = ICE_AQ_LEN;
1094         hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
1095         hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
1096         hw->mailboxq.num_rq_entries = ICE_MBXQ_LEN;
1097         hw->mailboxq.num_sq_entries = ICE_MBXQ_LEN;
1098         hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1099         hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1100 }
1101
1102 /**
1103  * ice_irq_affinity_notify - Callback for affinity changes
1104  * @notify: context as to what irq was changed
1105  * @mask: the new affinity mask
1106  *
1107  * This is a callback function used by the irq_set_affinity_notifier function
1108  * so that we may register to receive changes to the irq affinity masks.
1109  */
1110 static void ice_irq_affinity_notify(struct irq_affinity_notify *notify,
1111                                     const cpumask_t *mask)
1112 {
1113         struct ice_q_vector *q_vector =
1114                 container_of(notify, struct ice_q_vector, affinity_notify);
1115
1116         cpumask_copy(&q_vector->affinity_mask, mask);
1117 }
1118
1119 /**
1120  * ice_irq_affinity_release - Callback for affinity notifier release
1121  * @ref: internal core kernel usage
1122  *
1123  * This is a callback function used by the irq_set_affinity_notifier function
1124  * to inform the current notification subscriber that they will no longer
1125  * receive notifications.
1126  */
1127 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
1128
1129 /**
1130  * ice_vsi_ena_irq - Enable IRQ for the given VSI
1131  * @vsi: the VSI being configured
1132  */
1133 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
1134 {
1135         struct ice_pf *pf = vsi->back;
1136         struct ice_hw *hw = &pf->hw;
1137
1138         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
1139                 int i;
1140
1141                 for (i = 0; i < vsi->num_q_vectors; i++)
1142                         ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
1143         }
1144
1145         ice_flush(hw);
1146         return 0;
1147 }
1148
1149 /**
1150  * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
1151  * @vsi: the VSI being configured
1152  * @basename: name for the vector
1153  */
1154 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
1155 {
1156         int q_vectors = vsi->num_q_vectors;
1157         struct ice_pf *pf = vsi->back;
1158         int base = vsi->sw_base_vector;
1159         int rx_int_idx = 0;
1160         int tx_int_idx = 0;
1161         int vector, err;
1162         int irq_num;
1163
1164         for (vector = 0; vector < q_vectors; vector++) {
1165                 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
1166
1167                 irq_num = pf->msix_entries[base + vector].vector;
1168
1169                 if (q_vector->tx.ring && q_vector->rx.ring) {
1170                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1171                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
1172                         tx_int_idx++;
1173                 } else if (q_vector->rx.ring) {
1174                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1175                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
1176                 } else if (q_vector->tx.ring) {
1177                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1178                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
1179                 } else {
1180                         /* skip this unused q_vector */
1181                         continue;
1182                 }
1183                 err = devm_request_irq(&pf->pdev->dev,
1184                                        pf->msix_entries[base + vector].vector,
1185                                        vsi->irq_handler, 0, q_vector->name,
1186                                        q_vector);
1187                 if (err) {
1188                         netdev_err(vsi->netdev,
1189                                    "MSIX request_irq failed, error: %d\n", err);
1190                         goto free_q_irqs;
1191                 }
1192
1193                 /* register for affinity change notifications */
1194                 q_vector->affinity_notify.notify = ice_irq_affinity_notify;
1195                 q_vector->affinity_notify.release = ice_irq_affinity_release;
1196                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
1197
1198                 /* assign the mask for this irq */
1199                 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
1200         }
1201
1202         vsi->irqs_ready = true;
1203         return 0;
1204
1205 free_q_irqs:
1206         while (vector) {
1207                 vector--;
1208                 irq_num = pf->msix_entries[base + vector].vector,
1209                 irq_set_affinity_notifier(irq_num, NULL);
1210                 irq_set_affinity_hint(irq_num, NULL);
1211                 devm_free_irq(&pf->pdev->dev, irq_num, &vsi->q_vectors[vector]);
1212         }
1213         return err;
1214 }
1215
1216 /**
1217  * ice_ena_misc_vector - enable the non-queue interrupts
1218  * @pf: board private structure
1219  */
1220 static void ice_ena_misc_vector(struct ice_pf *pf)
1221 {
1222         struct ice_hw *hw = &pf->hw;
1223         u32 val;
1224
1225         /* clear things first */
1226         wr32(hw, PFINT_OICR_ENA, 0);    /* disable all */
1227         rd32(hw, PFINT_OICR);           /* read to clear */
1228
1229         val = (PFINT_OICR_ECC_ERR_M |
1230                PFINT_OICR_MAL_DETECT_M |
1231                PFINT_OICR_GRST_M |
1232                PFINT_OICR_PCI_EXCEPTION_M |
1233                PFINT_OICR_VFLR_M |
1234                PFINT_OICR_HMC_ERR_M |
1235                PFINT_OICR_PE_CRITERR_M);
1236
1237         wr32(hw, PFINT_OICR_ENA, val);
1238
1239         /* SW_ITR_IDX = 0, but don't change INTENA */
1240         wr32(hw, GLINT_DYN_CTL(pf->hw_oicr_idx),
1241              GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
1242 }
1243
1244 /**
1245  * ice_misc_intr - misc interrupt handler
1246  * @irq: interrupt number
1247  * @data: pointer to a q_vector
1248  */
1249 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
1250 {
1251         struct ice_pf *pf = (struct ice_pf *)data;
1252         struct ice_hw *hw = &pf->hw;
1253         irqreturn_t ret = IRQ_NONE;
1254         u32 oicr, ena_mask;
1255
1256         set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1257         set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1258
1259         oicr = rd32(hw, PFINT_OICR);
1260         ena_mask = rd32(hw, PFINT_OICR_ENA);
1261
1262         if (oicr & PFINT_OICR_MAL_DETECT_M) {
1263                 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
1264                 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
1265         }
1266         if (oicr & PFINT_OICR_VFLR_M) {
1267                 ena_mask &= ~PFINT_OICR_VFLR_M;
1268                 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
1269         }
1270
1271         if (oicr & PFINT_OICR_GRST_M) {
1272                 u32 reset;
1273
1274                 /* we have a reset warning */
1275                 ena_mask &= ~PFINT_OICR_GRST_M;
1276                 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
1277                         GLGEN_RSTAT_RESET_TYPE_S;
1278
1279                 if (reset == ICE_RESET_CORER)
1280                         pf->corer_count++;
1281                 else if (reset == ICE_RESET_GLOBR)
1282                         pf->globr_count++;
1283                 else if (reset == ICE_RESET_EMPR)
1284                         pf->empr_count++;
1285                 else
1286                         dev_dbg(&pf->pdev->dev, "Invalid reset type %d\n",
1287                                 reset);
1288
1289                 /* If a reset cycle isn't already in progress, we set a bit in
1290                  * pf->state so that the service task can start a reset/rebuild.
1291                  * We also make note of which reset happened so that peer
1292                  * devices/drivers can be informed.
1293                  */
1294                 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
1295                         if (reset == ICE_RESET_CORER)
1296                                 set_bit(__ICE_CORER_RECV, pf->state);
1297                         else if (reset == ICE_RESET_GLOBR)
1298                                 set_bit(__ICE_GLOBR_RECV, pf->state);
1299                         else
1300                                 set_bit(__ICE_EMPR_RECV, pf->state);
1301
1302                         /* There are couple of different bits at play here.
1303                          * hw->reset_ongoing indicates whether the hardware is
1304                          * in reset. This is set to true when a reset interrupt
1305                          * is received and set back to false after the driver
1306                          * has determined that the hardware is out of reset.
1307                          *
1308                          * __ICE_RESET_OICR_RECV in pf->state indicates
1309                          * that a post reset rebuild is required before the
1310                          * driver is operational again. This is set above.
1311                          *
1312                          * As this is the start of the reset/rebuild cycle, set
1313                          * both to indicate that.
1314                          */
1315                         hw->reset_ongoing = true;
1316                 }
1317         }
1318
1319         if (oicr & PFINT_OICR_HMC_ERR_M) {
1320                 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
1321                 dev_dbg(&pf->pdev->dev,
1322                         "HMC Error interrupt - info 0x%x, data 0x%x\n",
1323                         rd32(hw, PFHMC_ERRORINFO),
1324                         rd32(hw, PFHMC_ERRORDATA));
1325         }
1326
1327         /* Report and mask off any remaining unexpected interrupts */
1328         oicr &= ena_mask;
1329         if (oicr) {
1330                 dev_dbg(&pf->pdev->dev, "unhandled interrupt oicr=0x%08x\n",
1331                         oicr);
1332                 /* If a critical error is pending there is no choice but to
1333                  * reset the device.
1334                  */
1335                 if (oicr & (PFINT_OICR_PE_CRITERR_M |
1336                             PFINT_OICR_PCI_EXCEPTION_M |
1337                             PFINT_OICR_ECC_ERR_M)) {
1338                         set_bit(__ICE_PFR_REQ, pf->state);
1339                         ice_service_task_schedule(pf);
1340                 }
1341                 ena_mask &= ~oicr;
1342         }
1343         ret = IRQ_HANDLED;
1344
1345         /* re-enable interrupt causes that are not handled during this pass */
1346         wr32(hw, PFINT_OICR_ENA, ena_mask);
1347         if (!test_bit(__ICE_DOWN, pf->state)) {
1348                 ice_service_task_schedule(pf);
1349                 ice_irq_dynamic_ena(hw, NULL, NULL);
1350         }
1351
1352         return ret;
1353 }
1354
1355 /**
1356  * ice_free_irq_msix_misc - Unroll misc vector setup
1357  * @pf: board private structure
1358  */
1359 static void ice_free_irq_msix_misc(struct ice_pf *pf)
1360 {
1361         /* disable OICR interrupt */
1362         wr32(&pf->hw, PFINT_OICR_ENA, 0);
1363         ice_flush(&pf->hw);
1364
1365         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags) && pf->msix_entries) {
1366                 synchronize_irq(pf->msix_entries[pf->sw_oicr_idx].vector);
1367                 devm_free_irq(&pf->pdev->dev,
1368                               pf->msix_entries[pf->sw_oicr_idx].vector, pf);
1369         }
1370
1371         pf->num_avail_sw_msix += 1;
1372         ice_free_res(pf->sw_irq_tracker, pf->sw_oicr_idx, ICE_RES_MISC_VEC_ID);
1373         pf->num_avail_hw_msix += 1;
1374         ice_free_res(pf->hw_irq_tracker, pf->hw_oicr_idx, ICE_RES_MISC_VEC_ID);
1375 }
1376
1377 /**
1378  * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
1379  * @pf: board private structure
1380  *
1381  * This sets up the handler for MSIX 0, which is used to manage the
1382  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
1383  * when in MSI or Legacy interrupt mode.
1384  */
1385 static int ice_req_irq_msix_misc(struct ice_pf *pf)
1386 {
1387         struct ice_hw *hw = &pf->hw;
1388         int oicr_idx, err = 0;
1389         u8 itr_gran;
1390         u32 val;
1391
1392         if (!pf->int_name[0])
1393                 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
1394                          dev_driver_string(&pf->pdev->dev),
1395                          dev_name(&pf->pdev->dev));
1396
1397         /* Do not request IRQ but do enable OICR interrupt since settings are
1398          * lost during reset. Note that this function is called only during
1399          * rebuild path and not while reset is in progress.
1400          */
1401         if (ice_is_reset_in_progress(pf->state))
1402                 goto skip_req_irq;
1403
1404         /* reserve one vector in sw_irq_tracker for misc interrupts */
1405         oicr_idx = ice_get_res(pf, pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1406         if (oicr_idx < 0)
1407                 return oicr_idx;
1408
1409         pf->num_avail_sw_msix -= 1;
1410         pf->sw_oicr_idx = oicr_idx;
1411
1412         /* reserve one vector in hw_irq_tracker for misc interrupts */
1413         oicr_idx = ice_get_res(pf, pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1414         if (oicr_idx < 0) {
1415                 ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1416                 pf->num_avail_sw_msix += 1;
1417                 return oicr_idx;
1418         }
1419         pf->num_avail_hw_msix -= 1;
1420         pf->hw_oicr_idx = oicr_idx;
1421
1422         err = devm_request_irq(&pf->pdev->dev,
1423                                pf->msix_entries[pf->sw_oicr_idx].vector,
1424                                ice_misc_intr, 0, pf->int_name, pf);
1425         if (err) {
1426                 dev_err(&pf->pdev->dev,
1427                         "devm_request_irq for %s failed: %d\n",
1428                         pf->int_name, err);
1429                 ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1430                 pf->num_avail_sw_msix += 1;
1431                 ice_free_res(pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1432                 pf->num_avail_hw_msix += 1;
1433                 return err;
1434         }
1435
1436 skip_req_irq:
1437         ice_ena_misc_vector(pf);
1438
1439         val = ((pf->hw_oicr_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
1440                PFINT_OICR_CTL_CAUSE_ENA_M);
1441         wr32(hw, PFINT_OICR_CTL, val);
1442
1443         /* This enables Admin queue Interrupt causes */
1444         val = ((pf->hw_oicr_idx & PFINT_FW_CTL_MSIX_INDX_M) |
1445                PFINT_FW_CTL_CAUSE_ENA_M);
1446         wr32(hw, PFINT_FW_CTL, val);
1447
1448         /* This enables Mailbox queue Interrupt causes */
1449         val = ((pf->hw_oicr_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
1450                PFINT_MBX_CTL_CAUSE_ENA_M);
1451         wr32(hw, PFINT_MBX_CTL, val);
1452
1453         itr_gran = hw->itr_gran;
1454
1455         wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->hw_oicr_idx),
1456              ITR_TO_REG(ICE_ITR_8K, itr_gran));
1457
1458         ice_flush(hw);
1459         ice_irq_dynamic_ena(hw, NULL, NULL);
1460
1461         return 0;
1462 }
1463
1464 /**
1465  * ice_napi_del - Remove NAPI handler for the VSI
1466  * @vsi: VSI for which NAPI handler is to be removed
1467  */
1468 static void ice_napi_del(struct ice_vsi *vsi)
1469 {
1470         int v_idx;
1471
1472         if (!vsi->netdev)
1473                 return;
1474
1475         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
1476                 netif_napi_del(&vsi->q_vectors[v_idx]->napi);
1477 }
1478
1479 /**
1480  * ice_napi_add - register NAPI handler for the VSI
1481  * @vsi: VSI for which NAPI handler is to be registered
1482  *
1483  * This function is only called in the driver's load path. Registering the NAPI
1484  * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
1485  * reset/rebuild, etc.)
1486  */
1487 static void ice_napi_add(struct ice_vsi *vsi)
1488 {
1489         int v_idx;
1490
1491         if (!vsi->netdev)
1492                 return;
1493
1494         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
1495                 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
1496                                ice_napi_poll, NAPI_POLL_WEIGHT);
1497 }
1498
1499 /**
1500  * ice_cfg_netdev - Allocate, configure and register a netdev
1501  * @vsi: the VSI associated with the new netdev
1502  *
1503  * Returns 0 on success, negative value on failure
1504  */
1505 static int ice_cfg_netdev(struct ice_vsi *vsi)
1506 {
1507         netdev_features_t csumo_features;
1508         netdev_features_t vlano_features;
1509         netdev_features_t dflt_features;
1510         netdev_features_t tso_features;
1511         struct ice_netdev_priv *np;
1512         struct net_device *netdev;
1513         u8 mac_addr[ETH_ALEN];
1514         int err;
1515
1516         netdev = alloc_etherdev_mqs(sizeof(struct ice_netdev_priv),
1517                                     vsi->alloc_txq, vsi->alloc_rxq);
1518         if (!netdev)
1519                 return -ENOMEM;
1520
1521         vsi->netdev = netdev;
1522         np = netdev_priv(netdev);
1523         np->vsi = vsi;
1524
1525         dflt_features = NETIF_F_SG      |
1526                         NETIF_F_HIGHDMA |
1527                         NETIF_F_RXHASH;
1528
1529         csumo_features = NETIF_F_RXCSUM   |
1530                          NETIF_F_IP_CSUM  |
1531                          NETIF_F_IPV6_CSUM;
1532
1533         vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
1534                          NETIF_F_HW_VLAN_CTAG_TX     |
1535                          NETIF_F_HW_VLAN_CTAG_RX;
1536
1537         tso_features = NETIF_F_TSO;
1538
1539         /* set features that user can change */
1540         netdev->hw_features = dflt_features | csumo_features |
1541                               vlano_features | tso_features;
1542
1543         /* enable features */
1544         netdev->features |= netdev->hw_features;
1545         /* encap and VLAN devices inherit default, csumo and tso features */
1546         netdev->hw_enc_features |= dflt_features | csumo_features |
1547                                    tso_features;
1548         netdev->vlan_features |= dflt_features | csumo_features |
1549                                  tso_features;
1550
1551         if (vsi->type == ICE_VSI_PF) {
1552                 SET_NETDEV_DEV(netdev, &vsi->back->pdev->dev);
1553                 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
1554
1555                 ether_addr_copy(netdev->dev_addr, mac_addr);
1556                 ether_addr_copy(netdev->perm_addr, mac_addr);
1557         }
1558
1559         netdev->priv_flags |= IFF_UNICAST_FLT;
1560
1561         /* assign netdev_ops */
1562         netdev->netdev_ops = &ice_netdev_ops;
1563
1564         /* setup watchdog timeout value to be 5 second */
1565         netdev->watchdog_timeo = 5 * HZ;
1566
1567         ice_set_ethtool_ops(netdev);
1568
1569         netdev->min_mtu = ETH_MIN_MTU;
1570         netdev->max_mtu = ICE_MAX_MTU;
1571
1572         err = register_netdev(vsi->netdev);
1573         if (err)
1574                 return err;
1575
1576         netif_carrier_off(vsi->netdev);
1577
1578         /* make sure transmit queues start off as stopped */
1579         netif_tx_stop_all_queues(vsi->netdev);
1580
1581         return 0;
1582 }
1583
1584 /**
1585  * ice_fill_rss_lut - Fill the RSS lookup table with default values
1586  * @lut: Lookup table
1587  * @rss_table_size: Lookup table size
1588  * @rss_size: Range of queue number for hashing
1589  */
1590 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
1591 {
1592         u16 i;
1593
1594         for (i = 0; i < rss_table_size; i++)
1595                 lut[i] = i % rss_size;
1596 }
1597
1598 /**
1599  * ice_pf_vsi_setup - Set up a PF VSI
1600  * @pf: board private structure
1601  * @pi: pointer to the port_info instance
1602  *
1603  * Returns pointer to the successfully allocated VSI sw struct on success,
1604  * otherwise returns NULL on failure.
1605  */
1606 static struct ice_vsi *
1607 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
1608 {
1609         return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
1610 }
1611
1612 /**
1613  * ice_vlan_rx_add_vid - Add a vlan id filter to HW offload
1614  * @netdev: network interface to be adjusted
1615  * @proto: unused protocol
1616  * @vid: vlan id to be added
1617  *
1618  * net_device_ops implementation for adding vlan ids
1619  */
1620 static int ice_vlan_rx_add_vid(struct net_device *netdev,
1621                                __always_unused __be16 proto, u16 vid)
1622 {
1623         struct ice_netdev_priv *np = netdev_priv(netdev);
1624         struct ice_vsi *vsi = np->vsi;
1625         int ret;
1626
1627         if (vid >= VLAN_N_VID) {
1628                 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
1629                            vid, VLAN_N_VID);
1630                 return -EINVAL;
1631         }
1632
1633         if (vsi->info.pvid)
1634                 return -EINVAL;
1635
1636         /* Enable VLAN pruning when VLAN 0 is added */
1637         if (unlikely(!vid)) {
1638                 ret = ice_cfg_vlan_pruning(vsi, true);
1639                 if (ret)
1640                         return ret;
1641         }
1642
1643         /* Add all VLAN ids including 0 to the switch filter. VLAN id 0 is
1644          * needed to continue allowing all untagged packets since VLAN prune
1645          * list is applied to all packets by the switch
1646          */
1647         ret = ice_vsi_add_vlan(vsi, vid);
1648
1649         if (!ret)
1650                 set_bit(vid, vsi->active_vlans);
1651
1652         return ret;
1653 }
1654
1655 /**
1656  * ice_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
1657  * @netdev: network interface to be adjusted
1658  * @proto: unused protocol
1659  * @vid: vlan id to be removed
1660  *
1661  * net_device_ops implementation for removing vlan ids
1662  */
1663 static int ice_vlan_rx_kill_vid(struct net_device *netdev,
1664                                 __always_unused __be16 proto, u16 vid)
1665 {
1666         struct ice_netdev_priv *np = netdev_priv(netdev);
1667         struct ice_vsi *vsi = np->vsi;
1668         int status;
1669
1670         if (vsi->info.pvid)
1671                 return -EINVAL;
1672
1673         /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
1674          * information
1675          */
1676         status = ice_vsi_kill_vlan(vsi, vid);
1677         if (status)
1678                 return status;
1679
1680         clear_bit(vid, vsi->active_vlans);
1681
1682         /* Disable VLAN pruning when VLAN 0 is removed */
1683         if (unlikely(!vid))
1684                 status = ice_cfg_vlan_pruning(vsi, false);
1685
1686         return status;
1687 }
1688
1689 /**
1690  * ice_setup_pf_sw - Setup the HW switch on startup or after reset
1691  * @pf: board private structure
1692  *
1693  * Returns 0 on success, negative value on failure
1694  */
1695 static int ice_setup_pf_sw(struct ice_pf *pf)
1696 {
1697         LIST_HEAD(tmp_add_list);
1698         u8 broadcast[ETH_ALEN];
1699         struct ice_vsi *vsi;
1700         int status = 0;
1701
1702         if (ice_is_reset_in_progress(pf->state))
1703                 return -EBUSY;
1704
1705         vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
1706         if (!vsi) {
1707                 status = -ENOMEM;
1708                 goto unroll_vsi_setup;
1709         }
1710
1711         status = ice_cfg_netdev(vsi);
1712         if (status) {
1713                 status = -ENODEV;
1714                 goto unroll_vsi_setup;
1715         }
1716
1717         /* registering the NAPI handler requires both the queues and
1718          * netdev to be created, which are done in ice_pf_vsi_setup()
1719          * and ice_cfg_netdev() respectively
1720          */
1721         ice_napi_add(vsi);
1722
1723         /* To add a MAC filter, first add the MAC to a list and then
1724          * pass the list to ice_add_mac.
1725          */
1726
1727          /* Add a unicast MAC filter so the VSI can get its packets */
1728         status = ice_add_mac_to_list(vsi, &tmp_add_list,
1729                                      vsi->port_info->mac.perm_addr);
1730         if (status)
1731                 goto unroll_napi_add;
1732
1733         /* VSI needs to receive broadcast traffic, so add the broadcast
1734          * MAC address to the list as well.
1735          */
1736         eth_broadcast_addr(broadcast);
1737         status = ice_add_mac_to_list(vsi, &tmp_add_list, broadcast);
1738         if (status)
1739                 goto free_mac_list;
1740
1741         /* program MAC filters for entries in tmp_add_list */
1742         status = ice_add_mac(&pf->hw, &tmp_add_list);
1743         if (status) {
1744                 dev_err(&pf->pdev->dev, "Could not add MAC filters\n");
1745                 status = -ENOMEM;
1746                 goto free_mac_list;
1747         }
1748
1749         ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list);
1750         return status;
1751
1752 free_mac_list:
1753         ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list);
1754
1755 unroll_napi_add:
1756         if (vsi) {
1757                 ice_napi_del(vsi);
1758                 if (vsi->netdev) {
1759                         if (vsi->netdev->reg_state == NETREG_REGISTERED)
1760                                 unregister_netdev(vsi->netdev);
1761                         free_netdev(vsi->netdev);
1762                         vsi->netdev = NULL;
1763                 }
1764         }
1765
1766 unroll_vsi_setup:
1767         if (vsi) {
1768                 ice_vsi_free_q_vectors(vsi);
1769                 ice_vsi_delete(vsi);
1770                 ice_vsi_put_qs(vsi);
1771                 pf->q_left_tx += vsi->alloc_txq;
1772                 pf->q_left_rx += vsi->alloc_rxq;
1773                 ice_vsi_clear(vsi);
1774         }
1775         return status;
1776 }
1777
1778 /**
1779  * ice_determine_q_usage - Calculate queue distribution
1780  * @pf: board private structure
1781  *
1782  * Return -ENOMEM if we don't get enough queues for all ports
1783  */
1784 static void ice_determine_q_usage(struct ice_pf *pf)
1785 {
1786         u16 q_left_tx, q_left_rx;
1787
1788         q_left_tx = pf->hw.func_caps.common_cap.num_txq;
1789         q_left_rx = pf->hw.func_caps.common_cap.num_rxq;
1790
1791         pf->num_lan_tx = min_t(int, q_left_tx, num_online_cpus());
1792
1793         /* only 1 rx queue unless RSS is enabled */
1794         if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags))
1795                 pf->num_lan_rx = 1;
1796         else
1797                 pf->num_lan_rx = min_t(int, q_left_rx, num_online_cpus());
1798
1799         pf->q_left_tx = q_left_tx - pf->num_lan_tx;
1800         pf->q_left_rx = q_left_rx - pf->num_lan_rx;
1801 }
1802
1803 /**
1804  * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
1805  * @pf: board private structure to initialize
1806  */
1807 static void ice_deinit_pf(struct ice_pf *pf)
1808 {
1809         ice_service_task_stop(pf);
1810         mutex_destroy(&pf->sw_mutex);
1811         mutex_destroy(&pf->avail_q_mutex);
1812 }
1813
1814 /**
1815  * ice_init_pf - Initialize general software structures (struct ice_pf)
1816  * @pf: board private structure to initialize
1817  */
1818 static void ice_init_pf(struct ice_pf *pf)
1819 {
1820         bitmap_zero(pf->flags, ICE_PF_FLAGS_NBITS);
1821         set_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1822 #ifdef CONFIG_PCI_IOV
1823         if (pf->hw.func_caps.common_cap.sr_iov_1_1) {
1824                 struct ice_hw *hw = &pf->hw;
1825
1826                 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
1827                 pf->num_vfs_supported = min_t(int, hw->func_caps.num_allocd_vfs,
1828                                               ICE_MAX_VF_COUNT);
1829         }
1830 #endif /* CONFIG_PCI_IOV */
1831
1832         mutex_init(&pf->sw_mutex);
1833         mutex_init(&pf->avail_q_mutex);
1834
1835         /* Clear avail_[t|r]x_qs bitmaps (set all to avail) */
1836         mutex_lock(&pf->avail_q_mutex);
1837         bitmap_zero(pf->avail_txqs, ICE_MAX_TXQS);
1838         bitmap_zero(pf->avail_rxqs, ICE_MAX_RXQS);
1839         mutex_unlock(&pf->avail_q_mutex);
1840
1841         if (pf->hw.func_caps.common_cap.rss_table_size)
1842                 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
1843
1844         /* setup service timer and periodic service task */
1845         timer_setup(&pf->serv_tmr, ice_service_timer, 0);
1846         pf->serv_tmr_period = HZ;
1847         INIT_WORK(&pf->serv_task, ice_service_task);
1848         clear_bit(__ICE_SERVICE_SCHED, pf->state);
1849 }
1850
1851 /**
1852  * ice_ena_msix_range - Request a range of MSIX vectors from the OS
1853  * @pf: board private structure
1854  *
1855  * compute the number of MSIX vectors required (v_budget) and request from
1856  * the OS. Return the number of vectors reserved or negative on failure
1857  */
1858 static int ice_ena_msix_range(struct ice_pf *pf)
1859 {
1860         int v_left, v_actual, v_budget = 0;
1861         int needed, err, i;
1862
1863         v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
1864
1865         /* reserve one vector for miscellaneous handler */
1866         needed = 1;
1867         v_budget += needed;
1868         v_left -= needed;
1869
1870         /* reserve vectors for LAN traffic */
1871         pf->num_lan_msix = min_t(int, num_online_cpus(), v_left);
1872         v_budget += pf->num_lan_msix;
1873         v_left -= pf->num_lan_msix;
1874
1875         pf->msix_entries = devm_kcalloc(&pf->pdev->dev, v_budget,
1876                                         sizeof(struct msix_entry), GFP_KERNEL);
1877
1878         if (!pf->msix_entries) {
1879                 err = -ENOMEM;
1880                 goto exit_err;
1881         }
1882
1883         for (i = 0; i < v_budget; i++)
1884                 pf->msix_entries[i].entry = i;
1885
1886         /* actually reserve the vectors */
1887         v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
1888                                          ICE_MIN_MSIX, v_budget);
1889
1890         if (v_actual < 0) {
1891                 dev_err(&pf->pdev->dev, "unable to reserve MSI-X vectors\n");
1892                 err = v_actual;
1893                 goto msix_err;
1894         }
1895
1896         if (v_actual < v_budget) {
1897                 dev_warn(&pf->pdev->dev,
1898                          "not enough vectors. requested = %d, obtained = %d\n",
1899                          v_budget, v_actual);
1900                 if (v_actual >= (pf->num_lan_msix + 1)) {
1901                         pf->num_avail_sw_msix = v_actual -
1902                                                 (pf->num_lan_msix + 1);
1903                 } else if (v_actual >= 2) {
1904                         pf->num_lan_msix = 1;
1905                         pf->num_avail_sw_msix = v_actual - 2;
1906                 } else {
1907                         pci_disable_msix(pf->pdev);
1908                         err = -ERANGE;
1909                         goto msix_err;
1910                 }
1911         }
1912
1913         return v_actual;
1914
1915 msix_err:
1916         devm_kfree(&pf->pdev->dev, pf->msix_entries);
1917         goto exit_err;
1918
1919 exit_err:
1920         pf->num_lan_msix = 0;
1921         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1922         return err;
1923 }
1924
1925 /**
1926  * ice_dis_msix - Disable MSI-X interrupt setup in OS
1927  * @pf: board private structure
1928  */
1929 static void ice_dis_msix(struct ice_pf *pf)
1930 {
1931         pci_disable_msix(pf->pdev);
1932         devm_kfree(&pf->pdev->dev, pf->msix_entries);
1933         pf->msix_entries = NULL;
1934         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1935 }
1936
1937 /**
1938  * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
1939  * @pf: board private structure
1940  */
1941 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
1942 {
1943         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
1944                 ice_dis_msix(pf);
1945
1946         if (pf->sw_irq_tracker) {
1947                 devm_kfree(&pf->pdev->dev, pf->sw_irq_tracker);
1948                 pf->sw_irq_tracker = NULL;
1949         }
1950
1951         if (pf->hw_irq_tracker) {
1952                 devm_kfree(&pf->pdev->dev, pf->hw_irq_tracker);
1953                 pf->hw_irq_tracker = NULL;
1954         }
1955 }
1956
1957 /**
1958  * ice_init_interrupt_scheme - Determine proper interrupt scheme
1959  * @pf: board private structure to initialize
1960  */
1961 static int ice_init_interrupt_scheme(struct ice_pf *pf)
1962 {
1963         int vectors = 0, hw_vectors = 0;
1964         ssize_t size;
1965
1966         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
1967                 vectors = ice_ena_msix_range(pf);
1968         else
1969                 return -ENODEV;
1970
1971         if (vectors < 0)
1972                 return vectors;
1973
1974         /* set up vector assignment tracking */
1975         size = sizeof(struct ice_res_tracker) + (sizeof(u16) * vectors);
1976
1977         pf->sw_irq_tracker = devm_kzalloc(&pf->pdev->dev, size, GFP_KERNEL);
1978         if (!pf->sw_irq_tracker) {
1979                 ice_dis_msix(pf);
1980                 return -ENOMEM;
1981         }
1982
1983         /* populate SW interrupts pool with number of OS granted IRQs. */
1984         pf->num_avail_sw_msix = vectors;
1985         pf->sw_irq_tracker->num_entries = vectors;
1986
1987         /* set up HW vector assignment tracking */
1988         hw_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
1989         size = sizeof(struct ice_res_tracker) + (sizeof(u16) * hw_vectors);
1990
1991         pf->hw_irq_tracker = devm_kzalloc(&pf->pdev->dev, size, GFP_KERNEL);
1992         if (!pf->hw_irq_tracker) {
1993                 ice_clear_interrupt_scheme(pf);
1994                 return -ENOMEM;
1995         }
1996
1997         /* populate HW interrupts pool with number of HW supported irqs. */
1998         pf->num_avail_hw_msix = hw_vectors;
1999         pf->hw_irq_tracker->num_entries = hw_vectors;
2000
2001         return 0;
2002 }
2003
2004 /**
2005  * ice_probe - Device initialization routine
2006  * @pdev: PCI device information struct
2007  * @ent: entry in ice_pci_tbl
2008  *
2009  * Returns 0 on success, negative on failure
2010  */
2011 static int ice_probe(struct pci_dev *pdev,
2012                      const struct pci_device_id __always_unused *ent)
2013 {
2014         struct ice_pf *pf;
2015         struct ice_hw *hw;
2016         int err;
2017
2018         /* this driver uses devres, see Documentation/driver-model/devres.txt */
2019         err = pcim_enable_device(pdev);
2020         if (err)
2021                 return err;
2022
2023         err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
2024         if (err) {
2025                 dev_err(&pdev->dev, "BAR0 I/O map error %d\n", err);
2026                 return err;
2027         }
2028
2029         pf = devm_kzalloc(&pdev->dev, sizeof(*pf), GFP_KERNEL);
2030         if (!pf)
2031                 return -ENOMEM;
2032
2033         /* set up for high or low dma */
2034         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2035         if (err)
2036                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2037         if (err) {
2038                 dev_err(&pdev->dev, "DMA configuration failed: 0x%x\n", err);
2039                 return err;
2040         }
2041
2042         pci_enable_pcie_error_reporting(pdev);
2043         pci_set_master(pdev);
2044
2045         pf->pdev = pdev;
2046         pci_set_drvdata(pdev, pf);
2047         set_bit(__ICE_DOWN, pf->state);
2048         /* Disable service task until DOWN bit is cleared */
2049         set_bit(__ICE_SERVICE_DIS, pf->state);
2050
2051         hw = &pf->hw;
2052         hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
2053         hw->back = pf;
2054         hw->vendor_id = pdev->vendor;
2055         hw->device_id = pdev->device;
2056         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
2057         hw->subsystem_vendor_id = pdev->subsystem_vendor;
2058         hw->subsystem_device_id = pdev->subsystem_device;
2059         hw->bus.device = PCI_SLOT(pdev->devfn);
2060         hw->bus.func = PCI_FUNC(pdev->devfn);
2061         ice_set_ctrlq_len(hw);
2062
2063         pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
2064
2065 #ifndef CONFIG_DYNAMIC_DEBUG
2066         if (debug < -1)
2067                 hw->debug_mask = debug;
2068 #endif
2069
2070         err = ice_init_hw(hw);
2071         if (err) {
2072                 dev_err(&pdev->dev, "ice_init_hw failed: %d\n", err);
2073                 err = -EIO;
2074                 goto err_exit_unroll;
2075         }
2076
2077         dev_info(&pdev->dev, "firmware %d.%d.%05d api %d.%d\n",
2078                  hw->fw_maj_ver, hw->fw_min_ver, hw->fw_build,
2079                  hw->api_maj_ver, hw->api_min_ver);
2080
2081         ice_init_pf(pf);
2082
2083         ice_determine_q_usage(pf);
2084
2085         pf->num_alloc_vsi = min_t(u16, ICE_MAX_VSI_ALLOC,
2086                                   hw->func_caps.guaranteed_num_vsi);
2087         if (!pf->num_alloc_vsi) {
2088                 err = -EIO;
2089                 goto err_init_pf_unroll;
2090         }
2091
2092         pf->vsi = devm_kcalloc(&pdev->dev, pf->num_alloc_vsi,
2093                                sizeof(struct ice_vsi *), GFP_KERNEL);
2094         if (!pf->vsi) {
2095                 err = -ENOMEM;
2096                 goto err_init_pf_unroll;
2097         }
2098
2099         err = ice_init_interrupt_scheme(pf);
2100         if (err) {
2101                 dev_err(&pdev->dev,
2102                         "ice_init_interrupt_scheme failed: %d\n", err);
2103                 err = -EIO;
2104                 goto err_init_interrupt_unroll;
2105         }
2106
2107         /* Driver is mostly up */
2108         clear_bit(__ICE_DOWN, pf->state);
2109
2110         /* In case of MSIX we are going to setup the misc vector right here
2111          * to handle admin queue events etc. In case of legacy and MSI
2112          * the misc functionality and queue processing is combined in
2113          * the same vector and that gets setup at open.
2114          */
2115         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
2116                 err = ice_req_irq_msix_misc(pf);
2117                 if (err) {
2118                         dev_err(&pdev->dev,
2119                                 "setup of misc vector failed: %d\n", err);
2120                         goto err_init_interrupt_unroll;
2121                 }
2122         }
2123
2124         /* create switch struct for the switch element created by FW on boot */
2125         pf->first_sw = devm_kzalloc(&pdev->dev, sizeof(struct ice_sw),
2126                                     GFP_KERNEL);
2127         if (!pf->first_sw) {
2128                 err = -ENOMEM;
2129                 goto err_msix_misc_unroll;
2130         }
2131
2132         if (hw->evb_veb)
2133                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
2134         else
2135                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
2136
2137         pf->first_sw->pf = pf;
2138
2139         /* record the sw_id available for later use */
2140         pf->first_sw->sw_id = hw->port_info->sw_id;
2141
2142         err = ice_setup_pf_sw(pf);
2143         if (err) {
2144                 dev_err(&pdev->dev,
2145                         "probe failed due to setup pf switch:%d\n", err);
2146                 goto err_alloc_sw_unroll;
2147         }
2148
2149         clear_bit(__ICE_SERVICE_DIS, pf->state);
2150
2151         /* since everything is good, start the service timer */
2152         mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
2153
2154         return 0;
2155
2156 err_alloc_sw_unroll:
2157         set_bit(__ICE_SERVICE_DIS, pf->state);
2158         set_bit(__ICE_DOWN, pf->state);
2159         devm_kfree(&pf->pdev->dev, pf->first_sw);
2160 err_msix_misc_unroll:
2161         ice_free_irq_msix_misc(pf);
2162 err_init_interrupt_unroll:
2163         ice_clear_interrupt_scheme(pf);
2164         devm_kfree(&pdev->dev, pf->vsi);
2165 err_init_pf_unroll:
2166         ice_deinit_pf(pf);
2167         ice_deinit_hw(hw);
2168 err_exit_unroll:
2169         pci_disable_pcie_error_reporting(pdev);
2170         return err;
2171 }
2172
2173 /**
2174  * ice_remove - Device removal routine
2175  * @pdev: PCI device information struct
2176  */
2177 static void ice_remove(struct pci_dev *pdev)
2178 {
2179         struct ice_pf *pf = pci_get_drvdata(pdev);
2180         int i;
2181
2182         if (!pf)
2183                 return;
2184
2185         set_bit(__ICE_DOWN, pf->state);
2186         ice_service_task_stop(pf);
2187
2188         if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags))
2189                 ice_free_vfs(pf);
2190         ice_vsi_release_all(pf);
2191         ice_free_irq_msix_misc(pf);
2192         ice_for_each_vsi(pf, i) {
2193                 if (!pf->vsi[i])
2194                         continue;
2195                 ice_vsi_free_q_vectors(pf->vsi[i]);
2196         }
2197         ice_clear_interrupt_scheme(pf);
2198         ice_deinit_pf(pf);
2199         ice_deinit_hw(&pf->hw);
2200         pci_disable_pcie_error_reporting(pdev);
2201 }
2202
2203 /* ice_pci_tbl - PCI Device ID Table
2204  *
2205  * Wildcard entries (PCI_ANY_ID) should come last
2206  * Last entry must be all 0s
2207  *
2208  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
2209  *   Class, Class Mask, private data (not used) }
2210  */
2211 static const struct pci_device_id ice_pci_tbl[] = {
2212         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
2213         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
2214         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
2215         /* required last entry */
2216         { 0, }
2217 };
2218 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
2219
2220 static struct pci_driver ice_driver = {
2221         .name = KBUILD_MODNAME,
2222         .id_table = ice_pci_tbl,
2223         .probe = ice_probe,
2224         .remove = ice_remove,
2225         .sriov_configure = ice_sriov_configure,
2226 };
2227
2228 /**
2229  * ice_module_init - Driver registration routine
2230  *
2231  * ice_module_init is the first routine called when the driver is
2232  * loaded. All it does is register with the PCI subsystem.
2233  */
2234 static int __init ice_module_init(void)
2235 {
2236         int status;
2237
2238         pr_info("%s - version %s\n", ice_driver_string, ice_drv_ver);
2239         pr_info("%s\n", ice_copyright);
2240
2241         ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
2242         if (!ice_wq) {
2243                 pr_err("Failed to create workqueue\n");
2244                 return -ENOMEM;
2245         }
2246
2247         status = pci_register_driver(&ice_driver);
2248         if (status) {
2249                 pr_err("failed to register pci driver, err %d\n", status);
2250                 destroy_workqueue(ice_wq);
2251         }
2252
2253         return status;
2254 }
2255 module_init(ice_module_init);
2256
2257 /**
2258  * ice_module_exit - Driver exit cleanup routine
2259  *
2260  * ice_module_exit is called just before the driver is removed
2261  * from memory.
2262  */
2263 static void __exit ice_module_exit(void)
2264 {
2265         pci_unregister_driver(&ice_driver);
2266         destroy_workqueue(ice_wq);
2267         pr_info("module unloaded\n");
2268 }
2269 module_exit(ice_module_exit);
2270
2271 /**
2272  * ice_set_mac_address - NDO callback to set mac address
2273  * @netdev: network interface device structure
2274  * @pi: pointer to an address structure
2275  *
2276  * Returns 0 on success, negative on failure
2277  */
2278 static int ice_set_mac_address(struct net_device *netdev, void *pi)
2279 {
2280         struct ice_netdev_priv *np = netdev_priv(netdev);
2281         struct ice_vsi *vsi = np->vsi;
2282         struct ice_pf *pf = vsi->back;
2283         struct ice_hw *hw = &pf->hw;
2284         struct sockaddr *addr = pi;
2285         enum ice_status status;
2286         LIST_HEAD(a_mac_list);
2287         LIST_HEAD(r_mac_list);
2288         u8 flags = 0;
2289         int err;
2290         u8 *mac;
2291
2292         mac = (u8 *)addr->sa_data;
2293
2294         if (!is_valid_ether_addr(mac))
2295                 return -EADDRNOTAVAIL;
2296
2297         if (ether_addr_equal(netdev->dev_addr, mac)) {
2298                 netdev_warn(netdev, "already using mac %pM\n", mac);
2299                 return 0;
2300         }
2301
2302         if (test_bit(__ICE_DOWN, pf->state) ||
2303             ice_is_reset_in_progress(pf->state)) {
2304                 netdev_err(netdev, "can't set mac %pM. device not ready\n",
2305                            mac);
2306                 return -EBUSY;
2307         }
2308
2309         /* When we change the mac address we also have to change the mac address
2310          * based filter rules that were created previously for the old mac
2311          * address. So first, we remove the old filter rule using ice_remove_mac
2312          * and then create a new filter rule using ice_add_mac. Note that for
2313          * both these operations, we first need to form a "list" of mac
2314          * addresses (even though in this case, we have only 1 mac address to be
2315          * added/removed) and this done using ice_add_mac_to_list. Depending on
2316          * the ensuing operation this "list" of mac addresses is either to be
2317          * added or removed from the filter.
2318          */
2319         err = ice_add_mac_to_list(vsi, &r_mac_list, netdev->dev_addr);
2320         if (err) {
2321                 err = -EADDRNOTAVAIL;
2322                 goto free_lists;
2323         }
2324
2325         status = ice_remove_mac(hw, &r_mac_list);
2326         if (status) {
2327                 err = -EADDRNOTAVAIL;
2328                 goto free_lists;
2329         }
2330
2331         err = ice_add_mac_to_list(vsi, &a_mac_list, mac);
2332         if (err) {
2333                 err = -EADDRNOTAVAIL;
2334                 goto free_lists;
2335         }
2336
2337         status = ice_add_mac(hw, &a_mac_list);
2338         if (status) {
2339                 err = -EADDRNOTAVAIL;
2340                 goto free_lists;
2341         }
2342
2343 free_lists:
2344         /* free list entries */
2345         ice_free_fltr_list(&pf->pdev->dev, &r_mac_list);
2346         ice_free_fltr_list(&pf->pdev->dev, &a_mac_list);
2347
2348         if (err) {
2349                 netdev_err(netdev, "can't set mac %pM. filter update failed\n",
2350                            mac);
2351                 return err;
2352         }
2353
2354         /* change the netdev's mac address */
2355         memcpy(netdev->dev_addr, mac, netdev->addr_len);
2356         netdev_dbg(vsi->netdev, "updated mac address to %pM\n",
2357                    netdev->dev_addr);
2358
2359         /* write new mac address to the firmware */
2360         flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
2361         status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
2362         if (status) {
2363                 netdev_err(netdev, "can't set mac %pM. write to firmware failed.\n",
2364                            mac);
2365         }
2366         return 0;
2367 }
2368
2369 /**
2370  * ice_set_rx_mode - NDO callback to set the netdev filters
2371  * @netdev: network interface device structure
2372  */
2373 static void ice_set_rx_mode(struct net_device *netdev)
2374 {
2375         struct ice_netdev_priv *np = netdev_priv(netdev);
2376         struct ice_vsi *vsi = np->vsi;
2377
2378         if (!vsi)
2379                 return;
2380
2381         /* Set the flags to synchronize filters
2382          * ndo_set_rx_mode may be triggered even without a change in netdev
2383          * flags
2384          */
2385         set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
2386         set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
2387         set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
2388
2389         /* schedule our worker thread which will take care of
2390          * applying the new filter changes
2391          */
2392         ice_service_task_schedule(vsi->back);
2393 }
2394
2395 /**
2396  * ice_fdb_add - add an entry to the hardware database
2397  * @ndm: the input from the stack
2398  * @tb: pointer to array of nladdr (unused)
2399  * @dev: the net device pointer
2400  * @addr: the MAC address entry being added
2401  * @vid: VLAN id
2402  * @flags: instructions from stack about fdb operation
2403  */
2404 static int ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
2405                        struct net_device *dev, const unsigned char *addr,
2406                        u16 vid, u16 flags)
2407 {
2408         int err;
2409
2410         if (vid) {
2411                 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
2412                 return -EINVAL;
2413         }
2414         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2415                 netdev_err(dev, "FDB only supports static addresses\n");
2416                 return -EINVAL;
2417         }
2418
2419         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2420                 err = dev_uc_add_excl(dev, addr);
2421         else if (is_multicast_ether_addr(addr))
2422                 err = dev_mc_add_excl(dev, addr);
2423         else
2424                 err = -EINVAL;
2425
2426         /* Only return duplicate errors if NLM_F_EXCL is set */
2427         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2428                 err = 0;
2429
2430         return err;
2431 }
2432
2433 /**
2434  * ice_fdb_del - delete an entry from the hardware database
2435  * @ndm: the input from the stack
2436  * @tb: pointer to array of nladdr (unused)
2437  * @dev: the net device pointer
2438  * @addr: the MAC address entry being added
2439  * @vid: VLAN id
2440  */
2441 static int ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
2442                        struct net_device *dev, const unsigned char *addr,
2443                        __always_unused u16 vid)
2444 {
2445         int err;
2446
2447         if (ndm->ndm_state & NUD_PERMANENT) {
2448                 netdev_err(dev, "FDB only supports static addresses\n");
2449                 return -EINVAL;
2450         }
2451
2452         if (is_unicast_ether_addr(addr))
2453                 err = dev_uc_del(dev, addr);
2454         else if (is_multicast_ether_addr(addr))
2455                 err = dev_mc_del(dev, addr);
2456         else
2457                 err = -EINVAL;
2458
2459         return err;
2460 }
2461
2462 /**
2463  * ice_set_features - set the netdev feature flags
2464  * @netdev: ptr to the netdev being adjusted
2465  * @features: the feature set that the stack is suggesting
2466  */
2467 static int ice_set_features(struct net_device *netdev,
2468                             netdev_features_t features)
2469 {
2470         struct ice_netdev_priv *np = netdev_priv(netdev);
2471         struct ice_vsi *vsi = np->vsi;
2472         int ret = 0;
2473
2474         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
2475                 ret = ice_vsi_manage_rss_lut(vsi, true);
2476         else if (!(features & NETIF_F_RXHASH) &&
2477                  netdev->features & NETIF_F_RXHASH)
2478                 ret = ice_vsi_manage_rss_lut(vsi, false);
2479
2480         if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
2481             !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2482                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2483         else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
2484                  (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2485                 ret = ice_vsi_manage_vlan_stripping(vsi, false);
2486         else if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
2487                  !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2488                 ret = ice_vsi_manage_vlan_insertion(vsi);
2489         else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
2490                  (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2491                 ret = ice_vsi_manage_vlan_insertion(vsi);
2492
2493         return ret;
2494 }
2495
2496 /**
2497  * ice_vsi_vlan_setup - Setup vlan offload properties on a VSI
2498  * @vsi: VSI to setup vlan properties for
2499  */
2500 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
2501 {
2502         int ret = 0;
2503
2504         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
2505                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2506         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
2507                 ret = ice_vsi_manage_vlan_insertion(vsi);
2508
2509         return ret;
2510 }
2511
2512 /**
2513  * ice_restore_vlan - Reinstate VLANs when vsi/netdev comes back up
2514  * @vsi: the VSI being brought back up
2515  */
2516 static int ice_restore_vlan(struct ice_vsi *vsi)
2517 {
2518         int err;
2519         u16 vid;
2520
2521         if (!vsi->netdev)
2522                 return -EINVAL;
2523
2524         err = ice_vsi_vlan_setup(vsi);
2525         if (err)
2526                 return err;
2527
2528         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID) {
2529                 err = ice_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q), vid);
2530                 if (err)
2531                         break;
2532         }
2533
2534         return err;
2535 }
2536
2537 /**
2538  * ice_vsi_cfg - Setup the VSI
2539  * @vsi: the VSI being configured
2540  *
2541  * Return 0 on success and negative value on error
2542  */
2543 static int ice_vsi_cfg(struct ice_vsi *vsi)
2544 {
2545         int err;
2546
2547         if (vsi->netdev) {
2548                 ice_set_rx_mode(vsi->netdev);
2549                 err = ice_restore_vlan(vsi);
2550                 if (err)
2551                         return err;
2552         }
2553
2554         err = ice_vsi_cfg_txqs(vsi);
2555         if (!err)
2556                 err = ice_vsi_cfg_rxqs(vsi);
2557
2558         return err;
2559 }
2560
2561 /**
2562  * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
2563  * @vsi: the VSI being configured
2564  */
2565 static void ice_napi_enable_all(struct ice_vsi *vsi)
2566 {
2567         int q_idx;
2568
2569         if (!vsi->netdev)
2570                 return;
2571
2572         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
2573                 napi_enable(&vsi->q_vectors[q_idx]->napi);
2574 }
2575
2576 /**
2577  * ice_up_complete - Finish the last steps of bringing up a connection
2578  * @vsi: The VSI being configured
2579  *
2580  * Return 0 on success and negative value on error
2581  */
2582 static int ice_up_complete(struct ice_vsi *vsi)
2583 {
2584         struct ice_pf *pf = vsi->back;
2585         int err;
2586
2587         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
2588                 ice_vsi_cfg_msix(vsi);
2589         else
2590                 return -ENOTSUPP;
2591
2592         /* Enable only Rx rings, Tx rings were enabled by the FW when the
2593          * Tx queue group list was configured and the context bits were
2594          * programmed using ice_vsi_cfg_txqs
2595          */
2596         err = ice_vsi_start_rx_rings(vsi);
2597         if (err)
2598                 return err;
2599
2600         clear_bit(__ICE_DOWN, vsi->state);
2601         ice_napi_enable_all(vsi);
2602         ice_vsi_ena_irq(vsi);
2603
2604         if (vsi->port_info &&
2605             (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
2606             vsi->netdev) {
2607                 ice_print_link_msg(vsi, true);
2608                 netif_tx_start_all_queues(vsi->netdev);
2609                 netif_carrier_on(vsi->netdev);
2610         }
2611
2612         ice_service_task_schedule(pf);
2613
2614         return err;
2615 }
2616
2617 /**
2618  * ice_up - Bring the connection back up after being down
2619  * @vsi: VSI being configured
2620  */
2621 int ice_up(struct ice_vsi *vsi)
2622 {
2623         int err;
2624
2625         err = ice_vsi_cfg(vsi);
2626         if (!err)
2627                 err = ice_up_complete(vsi);
2628
2629         return err;
2630 }
2631
2632 /**
2633  * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
2634  * @ring: Tx or Rx ring to read stats from
2635  * @pkts: packets stats counter
2636  * @bytes: bytes stats counter
2637  *
2638  * This function fetches stats from the ring considering the atomic operations
2639  * that needs to be performed to read u64 values in 32 bit machine.
2640  */
2641 static void ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts,
2642                                          u64 *bytes)
2643 {
2644         unsigned int start;
2645         *pkts = 0;
2646         *bytes = 0;
2647
2648         if (!ring)
2649                 return;
2650         do {
2651                 start = u64_stats_fetch_begin_irq(&ring->syncp);
2652                 *pkts = ring->stats.pkts;
2653                 *bytes = ring->stats.bytes;
2654         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
2655 }
2656
2657 /**
2658  * ice_update_vsi_ring_stats - Update VSI stats counters
2659  * @vsi: the VSI to be updated
2660  */
2661 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
2662 {
2663         struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
2664         struct ice_ring *ring;
2665         u64 pkts, bytes;
2666         int i;
2667
2668         /* reset netdev stats */
2669         vsi_stats->tx_packets = 0;
2670         vsi_stats->tx_bytes = 0;
2671         vsi_stats->rx_packets = 0;
2672         vsi_stats->rx_bytes = 0;
2673
2674         /* reset non-netdev (extended) stats */
2675         vsi->tx_restart = 0;
2676         vsi->tx_busy = 0;
2677         vsi->tx_linearize = 0;
2678         vsi->rx_buf_failed = 0;
2679         vsi->rx_page_failed = 0;
2680
2681         rcu_read_lock();
2682
2683         /* update Tx rings counters */
2684         ice_for_each_txq(vsi, i) {
2685                 ring = READ_ONCE(vsi->tx_rings[i]);
2686                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
2687                 vsi_stats->tx_packets += pkts;
2688                 vsi_stats->tx_bytes += bytes;
2689                 vsi->tx_restart += ring->tx_stats.restart_q;
2690                 vsi->tx_busy += ring->tx_stats.tx_busy;
2691                 vsi->tx_linearize += ring->tx_stats.tx_linearize;
2692         }
2693
2694         /* update Rx rings counters */
2695         ice_for_each_rxq(vsi, i) {
2696                 ring = READ_ONCE(vsi->rx_rings[i]);
2697                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
2698                 vsi_stats->rx_packets += pkts;
2699                 vsi_stats->rx_bytes += bytes;
2700                 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
2701                 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
2702         }
2703
2704         rcu_read_unlock();
2705 }
2706
2707 /**
2708  * ice_update_vsi_stats - Update VSI stats counters
2709  * @vsi: the VSI to be updated
2710  */
2711 static void ice_update_vsi_stats(struct ice_vsi *vsi)
2712 {
2713         struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
2714         struct ice_eth_stats *cur_es = &vsi->eth_stats;
2715         struct ice_pf *pf = vsi->back;
2716
2717         if (test_bit(__ICE_DOWN, vsi->state) ||
2718             test_bit(__ICE_CFG_BUSY, pf->state))
2719                 return;
2720
2721         /* get stats as recorded by Tx/Rx rings */
2722         ice_update_vsi_ring_stats(vsi);
2723
2724         /* get VSI stats as recorded by the hardware */
2725         ice_update_eth_stats(vsi);
2726
2727         cur_ns->tx_errors = cur_es->tx_errors;
2728         cur_ns->rx_dropped = cur_es->rx_discards;
2729         cur_ns->tx_dropped = cur_es->tx_discards;
2730         cur_ns->multicast = cur_es->rx_multicast;
2731
2732         /* update some more netdev stats if this is main VSI */
2733         if (vsi->type == ICE_VSI_PF) {
2734                 cur_ns->rx_crc_errors = pf->stats.crc_errors;
2735                 cur_ns->rx_errors = pf->stats.crc_errors +
2736                                     pf->stats.illegal_bytes;
2737                 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
2738         }
2739 }
2740
2741 /**
2742  * ice_update_pf_stats - Update PF port stats counters
2743  * @pf: PF whose stats needs to be updated
2744  */
2745 static void ice_update_pf_stats(struct ice_pf *pf)
2746 {
2747         struct ice_hw_port_stats *prev_ps, *cur_ps;
2748         struct ice_hw *hw = &pf->hw;
2749         u8 pf_id;
2750
2751         prev_ps = &pf->stats_prev;
2752         cur_ps = &pf->stats;
2753         pf_id = hw->pf_id;
2754
2755         ice_stat_update40(hw, GLPRT_GORCH(pf_id), GLPRT_GORCL(pf_id),
2756                           pf->stat_prev_loaded, &prev_ps->eth.rx_bytes,
2757                           &cur_ps->eth.rx_bytes);
2758
2759         ice_stat_update40(hw, GLPRT_UPRCH(pf_id), GLPRT_UPRCL(pf_id),
2760                           pf->stat_prev_loaded, &prev_ps->eth.rx_unicast,
2761                           &cur_ps->eth.rx_unicast);
2762
2763         ice_stat_update40(hw, GLPRT_MPRCH(pf_id), GLPRT_MPRCL(pf_id),
2764                           pf->stat_prev_loaded, &prev_ps->eth.rx_multicast,
2765                           &cur_ps->eth.rx_multicast);
2766
2767         ice_stat_update40(hw, GLPRT_BPRCH(pf_id), GLPRT_BPRCL(pf_id),
2768                           pf->stat_prev_loaded, &prev_ps->eth.rx_broadcast,
2769                           &cur_ps->eth.rx_broadcast);
2770
2771         ice_stat_update40(hw, GLPRT_GOTCH(pf_id), GLPRT_GOTCL(pf_id),
2772                           pf->stat_prev_loaded, &prev_ps->eth.tx_bytes,
2773                           &cur_ps->eth.tx_bytes);
2774
2775         ice_stat_update40(hw, GLPRT_UPTCH(pf_id), GLPRT_UPTCL(pf_id),
2776                           pf->stat_prev_loaded, &prev_ps->eth.tx_unicast,
2777                           &cur_ps->eth.tx_unicast);
2778
2779         ice_stat_update40(hw, GLPRT_MPTCH(pf_id), GLPRT_MPTCL(pf_id),
2780                           pf->stat_prev_loaded, &prev_ps->eth.tx_multicast,
2781                           &cur_ps->eth.tx_multicast);
2782
2783         ice_stat_update40(hw, GLPRT_BPTCH(pf_id), GLPRT_BPTCL(pf_id),
2784                           pf->stat_prev_loaded, &prev_ps->eth.tx_broadcast,
2785                           &cur_ps->eth.tx_broadcast);
2786
2787         ice_stat_update32(hw, GLPRT_TDOLD(pf_id), pf->stat_prev_loaded,
2788                           &prev_ps->tx_dropped_link_down,
2789                           &cur_ps->tx_dropped_link_down);
2790
2791         ice_stat_update40(hw, GLPRT_PRC64H(pf_id), GLPRT_PRC64L(pf_id),
2792                           pf->stat_prev_loaded, &prev_ps->rx_size_64,
2793                           &cur_ps->rx_size_64);
2794
2795         ice_stat_update40(hw, GLPRT_PRC127H(pf_id), GLPRT_PRC127L(pf_id),
2796                           pf->stat_prev_loaded, &prev_ps->rx_size_127,
2797                           &cur_ps->rx_size_127);
2798
2799         ice_stat_update40(hw, GLPRT_PRC255H(pf_id), GLPRT_PRC255L(pf_id),
2800                           pf->stat_prev_loaded, &prev_ps->rx_size_255,
2801                           &cur_ps->rx_size_255);
2802
2803         ice_stat_update40(hw, GLPRT_PRC511H(pf_id), GLPRT_PRC511L(pf_id),
2804                           pf->stat_prev_loaded, &prev_ps->rx_size_511,
2805                           &cur_ps->rx_size_511);
2806
2807         ice_stat_update40(hw, GLPRT_PRC1023H(pf_id),
2808                           GLPRT_PRC1023L(pf_id), pf->stat_prev_loaded,
2809                           &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
2810
2811         ice_stat_update40(hw, GLPRT_PRC1522H(pf_id),
2812                           GLPRT_PRC1522L(pf_id), pf->stat_prev_loaded,
2813                           &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
2814
2815         ice_stat_update40(hw, GLPRT_PRC9522H(pf_id),
2816                           GLPRT_PRC9522L(pf_id), pf->stat_prev_loaded,
2817                           &prev_ps->rx_size_big, &cur_ps->rx_size_big);
2818
2819         ice_stat_update40(hw, GLPRT_PTC64H(pf_id), GLPRT_PTC64L(pf_id),
2820                           pf->stat_prev_loaded, &prev_ps->tx_size_64,
2821                           &cur_ps->tx_size_64);
2822
2823         ice_stat_update40(hw, GLPRT_PTC127H(pf_id), GLPRT_PTC127L(pf_id),
2824                           pf->stat_prev_loaded, &prev_ps->tx_size_127,
2825                           &cur_ps->tx_size_127);
2826
2827         ice_stat_update40(hw, GLPRT_PTC255H(pf_id), GLPRT_PTC255L(pf_id),
2828                           pf->stat_prev_loaded, &prev_ps->tx_size_255,
2829                           &cur_ps->tx_size_255);
2830
2831         ice_stat_update40(hw, GLPRT_PTC511H(pf_id), GLPRT_PTC511L(pf_id),
2832                           pf->stat_prev_loaded, &prev_ps->tx_size_511,
2833                           &cur_ps->tx_size_511);
2834
2835         ice_stat_update40(hw, GLPRT_PTC1023H(pf_id),
2836                           GLPRT_PTC1023L(pf_id), pf->stat_prev_loaded,
2837                           &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
2838
2839         ice_stat_update40(hw, GLPRT_PTC1522H(pf_id),
2840                           GLPRT_PTC1522L(pf_id), pf->stat_prev_loaded,
2841                           &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
2842
2843         ice_stat_update40(hw, GLPRT_PTC9522H(pf_id),
2844                           GLPRT_PTC9522L(pf_id), pf->stat_prev_loaded,
2845                           &prev_ps->tx_size_big, &cur_ps->tx_size_big);
2846
2847         ice_stat_update32(hw, GLPRT_LXONRXC(pf_id), pf->stat_prev_loaded,
2848                           &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
2849
2850         ice_stat_update32(hw, GLPRT_LXOFFRXC(pf_id), pf->stat_prev_loaded,
2851                           &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
2852
2853         ice_stat_update32(hw, GLPRT_LXONTXC(pf_id), pf->stat_prev_loaded,
2854                           &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
2855
2856         ice_stat_update32(hw, GLPRT_LXOFFTXC(pf_id), pf->stat_prev_loaded,
2857                           &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
2858
2859         ice_stat_update32(hw, GLPRT_CRCERRS(pf_id), pf->stat_prev_loaded,
2860                           &prev_ps->crc_errors, &cur_ps->crc_errors);
2861
2862         ice_stat_update32(hw, GLPRT_ILLERRC(pf_id), pf->stat_prev_loaded,
2863                           &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
2864
2865         ice_stat_update32(hw, GLPRT_MLFC(pf_id), pf->stat_prev_loaded,
2866                           &prev_ps->mac_local_faults,
2867                           &cur_ps->mac_local_faults);
2868
2869         ice_stat_update32(hw, GLPRT_MRFC(pf_id), pf->stat_prev_loaded,
2870                           &prev_ps->mac_remote_faults,
2871                           &cur_ps->mac_remote_faults);
2872
2873         ice_stat_update32(hw, GLPRT_RLEC(pf_id), pf->stat_prev_loaded,
2874                           &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
2875
2876         ice_stat_update32(hw, GLPRT_RUC(pf_id), pf->stat_prev_loaded,
2877                           &prev_ps->rx_undersize, &cur_ps->rx_undersize);
2878
2879         ice_stat_update32(hw, GLPRT_RFC(pf_id), pf->stat_prev_loaded,
2880                           &prev_ps->rx_fragments, &cur_ps->rx_fragments);
2881
2882         ice_stat_update32(hw, GLPRT_ROC(pf_id), pf->stat_prev_loaded,
2883                           &prev_ps->rx_oversize, &cur_ps->rx_oversize);
2884
2885         ice_stat_update32(hw, GLPRT_RJC(pf_id), pf->stat_prev_loaded,
2886                           &prev_ps->rx_jabber, &cur_ps->rx_jabber);
2887
2888         pf->stat_prev_loaded = true;
2889 }
2890
2891 /**
2892  * ice_get_stats64 - get statistics for network device structure
2893  * @netdev: network interface device structure
2894  * @stats: main device statistics structure
2895  */
2896 static
2897 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
2898 {
2899         struct ice_netdev_priv *np = netdev_priv(netdev);
2900         struct rtnl_link_stats64 *vsi_stats;
2901         struct ice_vsi *vsi = np->vsi;
2902
2903         vsi_stats = &vsi->net_stats;
2904
2905         if (test_bit(__ICE_DOWN, vsi->state) || !vsi->num_txq || !vsi->num_rxq)
2906                 return;
2907         /* netdev packet/byte stats come from ring counter. These are obtained
2908          * by summing up ring counters (done by ice_update_vsi_ring_stats).
2909          */
2910         ice_update_vsi_ring_stats(vsi);
2911         stats->tx_packets = vsi_stats->tx_packets;
2912         stats->tx_bytes = vsi_stats->tx_bytes;
2913         stats->rx_packets = vsi_stats->rx_packets;
2914         stats->rx_bytes = vsi_stats->rx_bytes;
2915
2916         /* The rest of the stats can be read from the hardware but instead we
2917          * just return values that the watchdog task has already obtained from
2918          * the hardware.
2919          */
2920         stats->multicast = vsi_stats->multicast;
2921         stats->tx_errors = vsi_stats->tx_errors;
2922         stats->tx_dropped = vsi_stats->tx_dropped;
2923         stats->rx_errors = vsi_stats->rx_errors;
2924         stats->rx_dropped = vsi_stats->rx_dropped;
2925         stats->rx_crc_errors = vsi_stats->rx_crc_errors;
2926         stats->rx_length_errors = vsi_stats->rx_length_errors;
2927 }
2928
2929 /**
2930  * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
2931  * @vsi: VSI having NAPI disabled
2932  */
2933 static void ice_napi_disable_all(struct ice_vsi *vsi)
2934 {
2935         int q_idx;
2936
2937         if (!vsi->netdev)
2938                 return;
2939
2940         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
2941                 napi_disable(&vsi->q_vectors[q_idx]->napi);
2942 }
2943
2944 /**
2945  * ice_down - Shutdown the connection
2946  * @vsi: The VSI being stopped
2947  */
2948 int ice_down(struct ice_vsi *vsi)
2949 {
2950         int i, tx_err, rx_err;
2951
2952         /* Caller of this function is expected to set the
2953          * vsi->state __ICE_DOWN bit
2954          */
2955         if (vsi->netdev) {
2956                 netif_carrier_off(vsi->netdev);
2957                 netif_tx_disable(vsi->netdev);
2958         }
2959
2960         ice_vsi_dis_irq(vsi);
2961         tx_err = ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0);
2962         if (tx_err)
2963                 netdev_err(vsi->netdev,
2964                            "Failed stop Tx rings, VSI %d error %d\n",
2965                            vsi->vsi_num, tx_err);
2966
2967         rx_err = ice_vsi_stop_rx_rings(vsi);
2968         if (rx_err)
2969                 netdev_err(vsi->netdev,
2970                            "Failed stop Rx rings, VSI %d error %d\n",
2971                            vsi->vsi_num, rx_err);
2972
2973         ice_napi_disable_all(vsi);
2974
2975         ice_for_each_txq(vsi, i)
2976                 ice_clean_tx_ring(vsi->tx_rings[i]);
2977
2978         ice_for_each_rxq(vsi, i)
2979                 ice_clean_rx_ring(vsi->rx_rings[i]);
2980
2981         if (tx_err || rx_err) {
2982                 netdev_err(vsi->netdev,
2983                            "Failed to close VSI 0x%04X on switch 0x%04X\n",
2984                            vsi->vsi_num, vsi->vsw->sw_id);
2985                 return -EIO;
2986         }
2987
2988         return 0;
2989 }
2990
2991 /**
2992  * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
2993  * @vsi: VSI having resources allocated
2994  *
2995  * Return 0 on success, negative on failure
2996  */
2997 static int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
2998 {
2999         int i, err = 0;
3000
3001         if (!vsi->num_txq) {
3002                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Tx queues\n",
3003                         vsi->vsi_num);
3004                 return -EINVAL;
3005         }
3006
3007         ice_for_each_txq(vsi, i) {
3008                 vsi->tx_rings[i]->netdev = vsi->netdev;
3009                 err = ice_setup_tx_ring(vsi->tx_rings[i]);
3010                 if (err)
3011                         break;
3012         }
3013
3014         return err;
3015 }
3016
3017 /**
3018  * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
3019  * @vsi: VSI having resources allocated
3020  *
3021  * Return 0 on success, negative on failure
3022  */
3023 static int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
3024 {
3025         int i, err = 0;
3026
3027         if (!vsi->num_rxq) {
3028                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Rx queues\n",
3029                         vsi->vsi_num);
3030                 return -EINVAL;
3031         }
3032
3033         ice_for_each_rxq(vsi, i) {
3034                 vsi->rx_rings[i]->netdev = vsi->netdev;
3035                 err = ice_setup_rx_ring(vsi->rx_rings[i]);
3036                 if (err)
3037                         break;
3038         }
3039
3040         return err;
3041 }
3042
3043 /**
3044  * ice_vsi_req_irq - Request IRQ from the OS
3045  * @vsi: The VSI IRQ is being requested for
3046  * @basename: name for the vector
3047  *
3048  * Return 0 on success and a negative value on error
3049  */
3050 static int ice_vsi_req_irq(struct ice_vsi *vsi, char *basename)
3051 {
3052         struct ice_pf *pf = vsi->back;
3053         int err = -EINVAL;
3054
3055         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3056                 err = ice_vsi_req_irq_msix(vsi, basename);
3057
3058         return err;
3059 }
3060
3061 /**
3062  * ice_vsi_open - Called when a network interface is made active
3063  * @vsi: the VSI to open
3064  *
3065  * Initialization of the VSI
3066  *
3067  * Returns 0 on success, negative value on error
3068  */
3069 static int ice_vsi_open(struct ice_vsi *vsi)
3070 {
3071         char int_name[ICE_INT_NAME_STR_LEN];
3072         struct ice_pf *pf = vsi->back;
3073         int err;
3074
3075         /* allocate descriptors */
3076         err = ice_vsi_setup_tx_rings(vsi);
3077         if (err)
3078                 goto err_setup_tx;
3079
3080         err = ice_vsi_setup_rx_rings(vsi);
3081         if (err)
3082                 goto err_setup_rx;
3083
3084         err = ice_vsi_cfg(vsi);
3085         if (err)
3086                 goto err_setup_rx;
3087
3088         snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3089                  dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
3090         err = ice_vsi_req_irq(vsi, int_name);
3091         if (err)
3092                 goto err_setup_rx;
3093
3094         /* Notify the stack of the actual queue counts. */
3095         err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
3096         if (err)
3097                 goto err_set_qs;
3098
3099         err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
3100         if (err)
3101                 goto err_set_qs;
3102
3103         err = ice_up_complete(vsi);
3104         if (err)
3105                 goto err_up_complete;
3106
3107         return 0;
3108
3109 err_up_complete:
3110         ice_down(vsi);
3111 err_set_qs:
3112         ice_vsi_free_irq(vsi);
3113 err_setup_rx:
3114         ice_vsi_free_rx_rings(vsi);
3115 err_setup_tx:
3116         ice_vsi_free_tx_rings(vsi);
3117
3118         return err;
3119 }
3120
3121 /**
3122  * ice_vsi_release_all - Delete all VSIs
3123  * @pf: PF from which all VSIs are being removed
3124  */
3125 static void ice_vsi_release_all(struct ice_pf *pf)
3126 {
3127         int err, i;
3128
3129         if (!pf->vsi)
3130                 return;
3131
3132         for (i = 0; i < pf->num_alloc_vsi; i++) {
3133                 if (!pf->vsi[i])
3134                         continue;
3135
3136                 err = ice_vsi_release(pf->vsi[i]);
3137                 if (err)
3138                         dev_dbg(&pf->pdev->dev,
3139                                 "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
3140                                 i, err, pf->vsi[i]->vsi_num);
3141         }
3142 }
3143
3144 /**
3145  * ice_dis_vsi - pause a VSI
3146  * @vsi: the VSI being paused
3147  */
3148 static void ice_dis_vsi(struct ice_vsi *vsi)
3149 {
3150         if (test_bit(__ICE_DOWN, vsi->state))
3151                 return;
3152
3153         set_bit(__ICE_NEEDS_RESTART, vsi->state);
3154
3155         if (vsi->type == ICE_VSI_PF && vsi->netdev) {
3156                 if (netif_running(vsi->netdev)) {
3157                         rtnl_lock();
3158                         vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3159                         rtnl_unlock();
3160                 } else {
3161                         ice_vsi_close(vsi);
3162                 }
3163         }
3164 }
3165
3166 /**
3167  * ice_ena_vsi - resume a VSI
3168  * @vsi: the VSI being resume
3169  */
3170 static int ice_ena_vsi(struct ice_vsi *vsi)
3171 {
3172         int err = 0;
3173
3174         if (test_and_clear_bit(__ICE_NEEDS_RESTART, vsi->state) &&
3175             vsi->netdev) {
3176                 if (netif_running(vsi->netdev)) {
3177                         rtnl_lock();
3178                         err = vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3179                         rtnl_unlock();
3180                 } else {
3181                         err = ice_vsi_open(vsi);
3182                 }
3183         }
3184
3185         return err;
3186 }
3187
3188 /**
3189  * ice_pf_dis_all_vsi - Pause all VSIs on a PF
3190  * @pf: the PF
3191  */
3192 static void ice_pf_dis_all_vsi(struct ice_pf *pf)
3193 {
3194         int v;
3195
3196         ice_for_each_vsi(pf, v)
3197                 if (pf->vsi[v])
3198                         ice_dis_vsi(pf->vsi[v]);
3199 }
3200
3201 /**
3202  * ice_pf_ena_all_vsi - Resume all VSIs on a PF
3203  * @pf: the PF
3204  */
3205 static int ice_pf_ena_all_vsi(struct ice_pf *pf)
3206 {
3207         int v;
3208
3209         ice_for_each_vsi(pf, v)
3210                 if (pf->vsi[v])
3211                         if (ice_ena_vsi(pf->vsi[v]))
3212                                 return -EIO;
3213
3214         return 0;
3215 }
3216
3217 /**
3218  * ice_vsi_rebuild_all - rebuild all VSIs in pf
3219  * @pf: the PF
3220  */
3221 static int ice_vsi_rebuild_all(struct ice_pf *pf)
3222 {
3223         int i;
3224
3225         /* loop through pf->vsi array and reinit the VSI if found */
3226         for (i = 0; i < pf->num_alloc_vsi; i++) {
3227                 int err;
3228
3229                 if (!pf->vsi[i])
3230                         continue;
3231
3232                 /* VF VSI rebuild isn't supported yet */
3233                 if (pf->vsi[i]->type == ICE_VSI_VF)
3234                         continue;
3235
3236                 err = ice_vsi_rebuild(pf->vsi[i]);
3237                 if (err) {
3238                         dev_err(&pf->pdev->dev,
3239                                 "VSI at index %d rebuild failed\n",
3240                                 pf->vsi[i]->idx);
3241                         return err;
3242                 }
3243
3244                 dev_info(&pf->pdev->dev,
3245                          "VSI at index %d rebuilt. vsi_num = 0x%x\n",
3246                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3247         }
3248
3249         return 0;
3250 }
3251
3252 /**
3253  * ice_vsi_replay_all - replay all VSIs configuration in the PF
3254  * @pf: the PF
3255  */
3256 static int ice_vsi_replay_all(struct ice_pf *pf)
3257 {
3258         struct ice_hw *hw = &pf->hw;
3259         enum ice_status ret;
3260         int i;
3261
3262         /* loop through pf->vsi array and replay the VSI if found */
3263         for (i = 0; i < pf->num_alloc_vsi; i++) {
3264                 if (!pf->vsi[i])
3265                         continue;
3266
3267                 ret = ice_replay_vsi(hw, pf->vsi[i]->idx);
3268                 if (ret) {
3269                         dev_err(&pf->pdev->dev,
3270                                 "VSI at index %d replay failed %d\n",
3271                                 pf->vsi[i]->idx, ret);
3272                         return -EIO;
3273                 }
3274
3275                 /* Re-map HW VSI number, using VSI handle that has been
3276                  * previously validated in ice_replay_vsi() call above
3277                  */
3278                 pf->vsi[i]->vsi_num = ice_get_hw_vsi_num(hw, pf->vsi[i]->idx);
3279
3280                 dev_info(&pf->pdev->dev,
3281                          "VSI at index %d filter replayed successfully - vsi_num %i\n",
3282                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3283         }
3284
3285         /* Clean up replay filter after successful re-configuration */
3286         ice_replay_post(hw);
3287         return 0;
3288 }
3289
3290 /**
3291  * ice_rebuild - rebuild after reset
3292  * @pf: pf to rebuild
3293  */
3294 static void ice_rebuild(struct ice_pf *pf)
3295 {
3296         struct device *dev = &pf->pdev->dev;
3297         struct ice_hw *hw = &pf->hw;
3298         enum ice_status ret;
3299         int err;
3300
3301         if (test_bit(__ICE_DOWN, pf->state))
3302                 goto clear_recovery;
3303
3304         dev_dbg(dev, "rebuilding pf\n");
3305
3306         ret = ice_init_all_ctrlq(hw);
3307         if (ret) {
3308                 dev_err(dev, "control queues init failed %d\n", ret);
3309                 goto err_init_ctrlq;
3310         }
3311
3312         ret = ice_clear_pf_cfg(hw);
3313         if (ret) {
3314                 dev_err(dev, "clear PF configuration failed %d\n", ret);
3315                 goto err_init_ctrlq;
3316         }
3317
3318         ice_clear_pxe_mode(hw);
3319
3320         ret = ice_get_caps(hw);
3321         if (ret) {
3322                 dev_err(dev, "ice_get_caps failed %d\n", ret);
3323                 goto err_init_ctrlq;
3324         }
3325
3326         err = ice_sched_init_port(hw->port_info);
3327         if (err)
3328                 goto err_sched_init_port;
3329
3330         /* reset search_hint of irq_trackers to 0 since interrupts are
3331          * reclaimed and could be allocated from beginning during VSI rebuild
3332          */
3333         pf->sw_irq_tracker->search_hint = 0;
3334         pf->hw_irq_tracker->search_hint = 0;
3335
3336         err = ice_vsi_rebuild_all(pf);
3337         if (err) {
3338                 dev_err(dev, "ice_vsi_rebuild_all failed\n");
3339                 goto err_vsi_rebuild;
3340         }
3341
3342         err = ice_update_link_info(hw->port_info);
3343         if (err)
3344                 dev_err(&pf->pdev->dev, "Get link status error %d\n", err);
3345
3346         /* Replay all VSIs Configuration, including filters after reset */
3347         if (ice_vsi_replay_all(pf)) {
3348                 dev_err(&pf->pdev->dev,
3349                         "error replaying VSI configurations with switch filter rules\n");
3350                 goto err_vsi_rebuild;
3351         }
3352
3353         /* start misc vector */
3354         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
3355                 err = ice_req_irq_msix_misc(pf);
3356                 if (err) {
3357                         dev_err(dev, "misc vector setup failed: %d\n", err);
3358                         goto err_vsi_rebuild;
3359                 }
3360         }
3361
3362         /* restart the VSIs that were rebuilt and running before the reset */
3363         err = ice_pf_ena_all_vsi(pf);
3364         if (err) {
3365                 dev_err(&pf->pdev->dev, "error enabling VSIs\n");
3366                 /* no need to disable VSIs in tear down path in ice_rebuild()
3367                  * since its already taken care in ice_vsi_open()
3368                  */
3369                 goto err_vsi_rebuild;
3370         }
3371
3372         ice_reset_all_vfs(pf, true);
3373         /* if we get here, reset flow is successful */
3374         clear_bit(__ICE_RESET_FAILED, pf->state);
3375         return;
3376
3377 err_vsi_rebuild:
3378         ice_vsi_release_all(pf);
3379 err_sched_init_port:
3380         ice_sched_cleanup_all(hw);
3381 err_init_ctrlq:
3382         ice_shutdown_all_ctrlq(hw);
3383         set_bit(__ICE_RESET_FAILED, pf->state);
3384 clear_recovery:
3385         /* set this bit in PF state to control service task scheduling */
3386         set_bit(__ICE_NEEDS_RESTART, pf->state);
3387         dev_err(dev, "Rebuild failed, unload and reload driver\n");
3388 }
3389
3390 /**
3391  * ice_change_mtu - NDO callback to change the MTU
3392  * @netdev: network interface device structure
3393  * @new_mtu: new value for maximum frame size
3394  *
3395  * Returns 0 on success, negative on failure
3396  */
3397 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
3398 {
3399         struct ice_netdev_priv *np = netdev_priv(netdev);
3400         struct ice_vsi *vsi = np->vsi;
3401         struct ice_pf *pf = vsi->back;
3402         u8 count = 0;
3403
3404         if (new_mtu == netdev->mtu) {
3405                 netdev_warn(netdev, "mtu is already %u\n", netdev->mtu);
3406                 return 0;
3407         }
3408
3409         if (new_mtu < netdev->min_mtu) {
3410                 netdev_err(netdev, "new mtu invalid. min_mtu is %d\n",
3411                            netdev->min_mtu);
3412                 return -EINVAL;
3413         } else if (new_mtu > netdev->max_mtu) {
3414                 netdev_err(netdev, "new mtu invalid. max_mtu is %d\n",
3415                            netdev->min_mtu);
3416                 return -EINVAL;
3417         }
3418         /* if a reset is in progress, wait for some time for it to complete */
3419         do {
3420                 if (ice_is_reset_in_progress(pf->state)) {
3421                         count++;
3422                         usleep_range(1000, 2000);
3423                 } else {
3424                         break;
3425                 }
3426
3427         } while (count < 100);
3428
3429         if (count == 100) {
3430                 netdev_err(netdev, "can't change mtu. Device is busy\n");
3431                 return -EBUSY;
3432         }
3433
3434         netdev->mtu = new_mtu;
3435
3436         /* if VSI is up, bring it down and then back up */
3437         if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
3438                 int err;
3439
3440                 err = ice_down(vsi);
3441                 if (err) {
3442                         netdev_err(netdev, "change mtu if_up err %d\n", err);
3443                         return err;
3444                 }
3445
3446                 err = ice_up(vsi);
3447                 if (err) {
3448                         netdev_err(netdev, "change mtu if_up err %d\n", err);
3449                         return err;
3450                 }
3451         }
3452
3453         netdev_dbg(netdev, "changed mtu to %d\n", new_mtu);
3454         return 0;
3455 }
3456
3457 /**
3458  * ice_set_rss - Set RSS keys and lut
3459  * @vsi: Pointer to VSI structure
3460  * @seed: RSS hash seed
3461  * @lut: Lookup table
3462  * @lut_size: Lookup table size
3463  *
3464  * Returns 0 on success, negative on failure
3465  */
3466 int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
3467 {
3468         struct ice_pf *pf = vsi->back;
3469         struct ice_hw *hw = &pf->hw;
3470         enum ice_status status;
3471
3472         if (seed) {
3473                 struct ice_aqc_get_set_rss_keys *buf =
3474                                   (struct ice_aqc_get_set_rss_keys *)seed;
3475
3476                 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
3477
3478                 if (status) {
3479                         dev_err(&pf->pdev->dev,
3480                                 "Cannot set RSS key, err %d aq_err %d\n",
3481                                 status, hw->adminq.rq_last_status);
3482                         return -EIO;
3483                 }
3484         }
3485
3486         if (lut) {
3487                 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
3488                                             lut, lut_size);
3489                 if (status) {
3490                         dev_err(&pf->pdev->dev,
3491                                 "Cannot set RSS lut, err %d aq_err %d\n",
3492                                 status, hw->adminq.rq_last_status);
3493                         return -EIO;
3494                 }
3495         }
3496
3497         return 0;
3498 }
3499
3500 /**
3501  * ice_get_rss - Get RSS keys and lut
3502  * @vsi: Pointer to VSI structure
3503  * @seed: Buffer to store the keys
3504  * @lut: Buffer to store the lookup table entries
3505  * @lut_size: Size of buffer to store the lookup table entries
3506  *
3507  * Returns 0 on success, negative on failure
3508  */
3509 int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
3510 {
3511         struct ice_pf *pf = vsi->back;
3512         struct ice_hw *hw = &pf->hw;
3513         enum ice_status status;
3514
3515         if (seed) {
3516                 struct ice_aqc_get_set_rss_keys *buf =
3517                                   (struct ice_aqc_get_set_rss_keys *)seed;
3518
3519                 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
3520                 if (status) {
3521                         dev_err(&pf->pdev->dev,
3522                                 "Cannot get RSS key, err %d aq_err %d\n",
3523                                 status, hw->adminq.rq_last_status);
3524                         return -EIO;
3525                 }
3526         }
3527
3528         if (lut) {
3529                 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
3530                                             lut, lut_size);
3531                 if (status) {
3532                         dev_err(&pf->pdev->dev,
3533                                 "Cannot get RSS lut, err %d aq_err %d\n",
3534                                 status, hw->adminq.rq_last_status);
3535                         return -EIO;
3536                 }
3537         }
3538
3539         return 0;
3540 }
3541
3542 /**
3543  * ice_bridge_getlink - Get the hardware bridge mode
3544  * @skb: skb buff
3545  * @pid: process id
3546  * @seq: RTNL message seq
3547  * @dev: the netdev being configured
3548  * @filter_mask: filter mask passed in
3549  * @nlflags: netlink flags passed in
3550  *
3551  * Return the bridge mode (VEB/VEPA)
3552  */
3553 static int
3554 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3555                    struct net_device *dev, u32 filter_mask, int nlflags)
3556 {
3557         struct ice_netdev_priv *np = netdev_priv(dev);
3558         struct ice_vsi *vsi = np->vsi;
3559         struct ice_pf *pf = vsi->back;
3560         u16 bmode;
3561
3562         bmode = pf->first_sw->bridge_mode;
3563
3564         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
3565                                        filter_mask, NULL);
3566 }
3567
3568 /**
3569  * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
3570  * @vsi: Pointer to VSI structure
3571  * @bmode: Hardware bridge mode (VEB/VEPA)
3572  *
3573  * Returns 0 on success, negative on failure
3574  */
3575 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
3576 {
3577         struct device *dev = &vsi->back->pdev->dev;
3578         struct ice_aqc_vsi_props *vsi_props;
3579         struct ice_hw *hw = &vsi->back->hw;
3580         struct ice_vsi_ctx ctxt = { 0 };
3581         enum ice_status status;
3582
3583         vsi_props = &vsi->info;
3584         ctxt.info = vsi->info;
3585
3586         if (bmode == BRIDGE_MODE_VEB)
3587                 /* change from VEPA to VEB mode */
3588                 ctxt.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
3589         else
3590                 /* change from VEB to VEPA mode */
3591                 ctxt.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
3592         ctxt.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
3593
3594         status = ice_update_vsi(hw, vsi->idx, &ctxt, NULL);
3595         if (status) {
3596                 dev_err(dev, "update VSI for bridge mode failed, bmode = %d err %d aq_err %d\n",
3597                         bmode, status, hw->adminq.sq_last_status);
3598                 return -EIO;
3599         }
3600         /* Update sw flags for book keeping */
3601         vsi_props->sw_flags = ctxt.info.sw_flags;
3602
3603         return 0;
3604 }
3605
3606 /**
3607  * ice_bridge_setlink - Set the hardware bridge mode
3608  * @dev: the netdev being configured
3609  * @nlh: RTNL message
3610  * @flags: bridge setlink flags
3611  *
3612  * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
3613  * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
3614  * not already set for all VSIs connected to this switch. And also update the
3615  * unicast switch filter rules for the corresponding switch of the netdev.
3616  */
3617 static int
3618 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
3619                    u16 __always_unused flags)
3620 {
3621         struct ice_netdev_priv *np = netdev_priv(dev);
3622         struct ice_pf *pf = np->vsi->back;
3623         struct nlattr *attr, *br_spec;
3624         struct ice_hw *hw = &pf->hw;
3625         enum ice_status status;
3626         struct ice_sw *pf_sw;
3627         int rem, v, err = 0;
3628
3629         pf_sw = pf->first_sw;
3630         /* find the attribute in the netlink message */
3631         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3632
3633         nla_for_each_nested(attr, br_spec, rem) {
3634                 __u16 mode;
3635
3636                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
3637                         continue;
3638                 mode = nla_get_u16(attr);
3639                 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
3640                         return -EINVAL;
3641                 /* Continue  if bridge mode is not being flipped */
3642                 if (mode == pf_sw->bridge_mode)
3643                         continue;
3644                 /* Iterates through the PF VSI list and update the loopback
3645                  * mode of the VSI
3646                  */
3647                 ice_for_each_vsi(pf, v) {
3648                         if (!pf->vsi[v])
3649                                 continue;
3650                         err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
3651                         if (err)
3652                                 return err;
3653                 }
3654
3655                 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
3656                 /* Update the unicast switch filter rules for the corresponding
3657                  * switch of the netdev
3658                  */
3659                 status = ice_update_sw_rule_bridge_mode(hw);
3660                 if (status) {
3661                         netdev_err(dev, "update SW_RULE for bridge mode failed,  = %d err %d aq_err %d\n",
3662                                    mode, status, hw->adminq.sq_last_status);
3663                         /* revert hw->evb_veb */
3664                         hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
3665                         return -EIO;
3666                 }
3667
3668                 pf_sw->bridge_mode = mode;
3669         }
3670
3671         return 0;
3672 }
3673
3674 /**
3675  * ice_tx_timeout - Respond to a Tx Hang
3676  * @netdev: network interface device structure
3677  */
3678 static void ice_tx_timeout(struct net_device *netdev)
3679 {
3680         struct ice_netdev_priv *np = netdev_priv(netdev);
3681         struct ice_ring *tx_ring = NULL;
3682         struct ice_vsi *vsi = np->vsi;
3683         struct ice_pf *pf = vsi->back;
3684         u32 head, val = 0, i;
3685         int hung_queue = -1;
3686
3687         pf->tx_timeout_count++;
3688
3689         /* find the stopped queue the same way the stack does */
3690         for (i = 0; i < netdev->num_tx_queues; i++) {
3691                 struct netdev_queue *q;
3692                 unsigned long trans_start;
3693
3694                 q = netdev_get_tx_queue(netdev, i);
3695                 trans_start = q->trans_start;
3696                 if (netif_xmit_stopped(q) &&
3697                     time_after(jiffies,
3698                                (trans_start + netdev->watchdog_timeo))) {
3699                         hung_queue = i;
3700                         break;
3701                 }
3702         }
3703
3704         if (i == netdev->num_tx_queues) {
3705                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
3706         } else {
3707                 /* now that we have an index, find the tx_ring struct */
3708                 for (i = 0; i < vsi->num_txq; i++) {
3709                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
3710                                 if (hung_queue ==
3711                                     vsi->tx_rings[i]->q_index) {
3712                                         tx_ring = vsi->tx_rings[i];
3713                                         break;
3714                                 }
3715                         }
3716                 }
3717         }
3718
3719         /* Reset recovery level if enough time has elapsed after last timeout.
3720          * Also ensure no new reset action happens before next timeout period.
3721          */
3722         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
3723                 pf->tx_timeout_recovery_level = 1;
3724         else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
3725                                        netdev->watchdog_timeo)))
3726                 return;
3727
3728         if (tx_ring) {
3729                 head = tx_ring->next_to_clean;
3730                 /* Read interrupt register */
3731                 if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3732                         val = rd32(&pf->hw,
3733                                    GLINT_DYN_CTL(tx_ring->q_vector->v_idx +
3734                                         tx_ring->vsi->hw_base_vector));
3735
3736                 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
3737                             vsi->vsi_num, hung_queue, tx_ring->next_to_clean,
3738                             head, tx_ring->next_to_use,
3739                             readl(tx_ring->tail), val);
3740         }
3741
3742         pf->tx_timeout_last_recovery = jiffies;
3743         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
3744                     pf->tx_timeout_recovery_level, hung_queue);
3745
3746         switch (pf->tx_timeout_recovery_level) {
3747         case 1:
3748                 set_bit(__ICE_PFR_REQ, pf->state);
3749                 break;
3750         case 2:
3751                 set_bit(__ICE_CORER_REQ, pf->state);
3752                 break;
3753         case 3:
3754                 set_bit(__ICE_GLOBR_REQ, pf->state);
3755                 break;
3756         default:
3757                 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
3758                 set_bit(__ICE_DOWN, pf->state);
3759                 set_bit(__ICE_NEEDS_RESTART, vsi->state);
3760                 set_bit(__ICE_SERVICE_DIS, pf->state);
3761                 break;
3762         }
3763
3764         ice_service_task_schedule(pf);
3765         pf->tx_timeout_recovery_level++;
3766 }
3767
3768 /**
3769  * ice_open - Called when a network interface becomes active
3770  * @netdev: network interface device structure
3771  *
3772  * The open entry point is called when a network interface is made
3773  * active by the system (IFF_UP).  At this point all resources needed
3774  * for transmit and receive operations are allocated, the interrupt
3775  * handler is registered with the OS, the netdev watchdog is enabled,
3776  * and the stack is notified that the interface is ready.
3777  *
3778  * Returns 0 on success, negative value on failure
3779  */
3780 static int ice_open(struct net_device *netdev)
3781 {
3782         struct ice_netdev_priv *np = netdev_priv(netdev);
3783         struct ice_vsi *vsi = np->vsi;
3784         int err;
3785
3786         if (test_bit(__ICE_NEEDS_RESTART, vsi->back->state)) {
3787                 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
3788                 return -EIO;
3789         }
3790
3791         netif_carrier_off(netdev);
3792
3793         err = ice_vsi_open(vsi);
3794
3795         if (err)
3796                 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
3797                            vsi->vsi_num, vsi->vsw->sw_id);
3798         return err;
3799 }
3800
3801 /**
3802  * ice_stop - Disables a network interface
3803  * @netdev: network interface device structure
3804  *
3805  * The stop entry point is called when an interface is de-activated by the OS,
3806  * and the netdevice enters the DOWN state.  The hardware is still under the
3807  * driver's control, but the netdev interface is disabled.
3808  *
3809  * Returns success only - not allowed to fail
3810  */
3811 static int ice_stop(struct net_device *netdev)
3812 {
3813         struct ice_netdev_priv *np = netdev_priv(netdev);
3814         struct ice_vsi *vsi = np->vsi;
3815
3816         ice_vsi_close(vsi);
3817
3818         return 0;
3819 }
3820
3821 /**
3822  * ice_features_check - Validate encapsulated packet conforms to limits
3823  * @skb: skb buffer
3824  * @netdev: This port's netdev
3825  * @features: Offload features that the stack believes apply
3826  */
3827 static netdev_features_t
3828 ice_features_check(struct sk_buff *skb,
3829                    struct net_device __always_unused *netdev,
3830                    netdev_features_t features)
3831 {
3832         size_t len;
3833
3834         /* No point in doing any of this if neither checksum nor GSO are
3835          * being requested for this frame.  We can rule out both by just
3836          * checking for CHECKSUM_PARTIAL
3837          */
3838         if (skb->ip_summed != CHECKSUM_PARTIAL)
3839                 return features;
3840
3841         /* We cannot support GSO if the MSS is going to be less than
3842          * 64 bytes.  If it is then we need to drop support for GSO.
3843          */
3844         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
3845                 features &= ~NETIF_F_GSO_MASK;
3846
3847         len = skb_network_header(skb) - skb->data;
3848         if (len & ~(ICE_TXD_MACLEN_MAX))
3849                 goto out_rm_features;
3850
3851         len = skb_transport_header(skb) - skb_network_header(skb);
3852         if (len & ~(ICE_TXD_IPLEN_MAX))
3853                 goto out_rm_features;
3854
3855         if (skb->encapsulation) {
3856                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
3857                 if (len & ~(ICE_TXD_L4LEN_MAX))
3858                         goto out_rm_features;
3859
3860                 len = skb_inner_transport_header(skb) -
3861                       skb_inner_network_header(skb);
3862                 if (len & ~(ICE_TXD_IPLEN_MAX))
3863                         goto out_rm_features;
3864         }
3865
3866         return features;
3867 out_rm_features:
3868         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
3869 }
3870
3871 static const struct net_device_ops ice_netdev_ops = {
3872         .ndo_open = ice_open,
3873         .ndo_stop = ice_stop,
3874         .ndo_start_xmit = ice_start_xmit,
3875         .ndo_features_check = ice_features_check,
3876         .ndo_set_rx_mode = ice_set_rx_mode,
3877         .ndo_set_mac_address = ice_set_mac_address,
3878         .ndo_validate_addr = eth_validate_addr,
3879         .ndo_change_mtu = ice_change_mtu,
3880         .ndo_get_stats64 = ice_get_stats64,
3881         .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
3882         .ndo_set_vf_mac = ice_set_vf_mac,
3883         .ndo_get_vf_config = ice_get_vf_cfg,
3884         .ndo_set_vf_trust = ice_set_vf_trust,
3885         .ndo_set_vf_vlan = ice_set_vf_port_vlan,
3886         .ndo_set_vf_link_state = ice_set_vf_link_state,
3887         .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
3888         .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
3889         .ndo_set_features = ice_set_features,
3890         .ndo_bridge_getlink = ice_bridge_getlink,
3891         .ndo_bridge_setlink = ice_bridge_setlink,
3892         .ndo_fdb_add = ice_fdb_add,
3893         .ndo_fdb_del = ice_fdb_del,
3894         .ndo_tx_timeout = ice_tx_timeout,
3895 };