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         if (hw->port_info)
 353                 ice_sched_clear_port(hw->port_info);
 354
 355         ice_shutdown_all_ctrlq(hw);
 356
 357         set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 358 }
 359
 360 /**
 361  * ice_do_reset - Initiate one of many types of resets
 362  * @pf: board private structure
 363  * @reset_type: reset type requested
 364  * before this function was called.
 365  */
 366 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 367 {
 368         struct device *dev = &pf->pdev->dev;
 369         struct ice_hw *hw = &pf->hw;
 370
 371         dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
 372         WARN_ON(in_interrupt());
 373
 374         ice_prepare_for_reset(pf);
 375
 376         /* trigger the reset */
 377         if (ice_reset(hw, reset_type)) {
 378                 dev_err(dev, "reset %d failed\n", reset_type);
 379                 set_bit(__ICE_RESET_FAILED, pf->state);
 380                 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 381                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 382                 clear_bit(__ICE_PFR_REQ, pf->state);
 383                 clear_bit(__ICE_CORER_REQ, pf->state);
 384                 clear_bit(__ICE_GLOBR_REQ, pf->state);
 385                 return;
 386         }
 387
 388         /* PFR is a bit of a special case because it doesn't result in an OICR
 389          * interrupt. So for PFR, rebuild after the reset and clear the reset-
 390          * associated state bits.
 391          */
 392         if (reset_type == ICE_RESET_PFR) {
 393                 pf->pfr_count++;
 394                 ice_rebuild(pf);
 395                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 396                 clear_bit(__ICE_PFR_REQ, pf->state);
 397         }
 398 }
 399
 400 /**
 401  * ice_reset_subtask - Set up for resetting the device and driver
 402  * @pf: board private structure
 403  */
 404 static void ice_reset_subtask(struct ice_pf *pf)
 405 {
 406         enum ice_reset_req reset_type = ICE_RESET_INVAL;
 407
 408         /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
 409          * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
 410          * of reset is pending and sets bits in pf->state indicating the reset
 411          * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set
 412          * prepare for pending reset if not already (for PF software-initiated
 413          * global resets the software should already be prepared for it as
 414          * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
 415          * by firmware or software on other PFs, that bit is not set so prepare
 416          * for the reset now), poll for reset done, rebuild and return.
 417          */
 418         if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
 419                 clear_bit(__ICE_GLOBR_RECV, pf->state);
 420                 clear_bit(__ICE_CORER_RECV, pf->state);
 421                 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
 422                         ice_prepare_for_reset(pf);
 423
 424                 /* make sure we are ready to rebuild */
 425                 if (ice_check_reset(&pf->hw)) {
 426                         set_bit(__ICE_RESET_FAILED, pf->state);
 427                 } else {
 428                         /* done with reset. start rebuild */
 429                         pf->hw.reset_ongoing = false;
 430                         ice_rebuild(pf);
 431                         /* clear bit to resume normal operations, but
 432                          * ICE_NEEDS_RESTART bit is set incase rebuild failed
 433                          */
 434                         clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 435                         clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 436                         clear_bit(__ICE_PFR_REQ, pf->state);
 437                         clear_bit(__ICE_CORER_REQ, pf->state);
 438                         clear_bit(__ICE_GLOBR_REQ, pf->state);
 439                 }
 440
 441                 return;
 442         }
 443
 444         /* No pending resets to finish processing. Check for new resets */
 445         if (test_bit(__ICE_PFR_REQ, pf->state))
 446                 reset_type = ICE_RESET_PFR;
 447         if (test_bit(__ICE_CORER_REQ, pf->state))
 448                 reset_type = ICE_RESET_CORER;
 449         if (test_bit(__ICE_GLOBR_REQ, pf->state))
 450                 reset_type = ICE_RESET_GLOBR;
 451         /* If no valid reset type requested just return */
 452         if (reset_type == ICE_RESET_INVAL)
 453                 return;
 454
 455         /* reset if not already down or busy */
 456         if (!test_bit(__ICE_DOWN, pf->state) &&
 457             !test_bit(__ICE_CFG_BUSY, pf->state)) {
 458                 ice_do_reset(pf, reset_type);
 459         }
 460 }
 461
 462 /**
 463  * ice_print_link_msg - print link up or down message
 464  * @vsi: the VSI whose link status is being queried
 465  * @isup: boolean for if the link is now up or down
 466  */
 467 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
 468 {
 469         const char *speed;
 470         const char *fc;
 471
 472         if (vsi->current_isup == isup)
 473                 return;
 474
 475         vsi->current_isup = isup;
 476
 477         if (!isup) {
 478                 netdev_info(vsi->netdev, "NIC Link is Down\n");
 479                 return;
 480         }
 481
 482         switch (vsi->port_info->phy.link_info.link_speed) {
 483         case ICE_AQ_LINK_SPEED_40GB:
 484                 speed = "40 G";
 485                 break;
 486         case ICE_AQ_LINK_SPEED_25GB:
 487                 speed = "25 G";
 488                 break;
 489         case ICE_AQ_LINK_SPEED_20GB:
 490                 speed = "20 G";
 491                 break;
 492         case ICE_AQ_LINK_SPEED_10GB:
 493                 speed = "10 G";
 494                 break;
 495         case ICE_AQ_LINK_SPEED_5GB:
 496                 speed = "5 G";
 497                 break;
 498         case ICE_AQ_LINK_SPEED_2500MB:
 499                 speed = "2.5 G";
 500                 break;
 501         case ICE_AQ_LINK_SPEED_1000MB:
 502                 speed = "1 G";
 503                 break;
 504         case ICE_AQ_LINK_SPEED_100MB:
 505                 speed = "100 M";
 506                 break;
 507         default:
 508                 speed = "Unknown";
 509                 break;
 510         }
 511
 512         switch (vsi->port_info->fc.current_mode) {
 513         case ICE_FC_FULL:
 514                 fc = "RX/TX";
 515                 break;
 516         case ICE_FC_TX_PAUSE:
 517                 fc = "TX";
 518                 break;
 519         case ICE_FC_RX_PAUSE:
 520                 fc = "RX";
 521                 break;
 522         default:
 523                 fc = "Unknown";
 524                 break;
 525         }
 526
 527         netdev_info(vsi->netdev, "NIC Link is up %sbps, Flow Control: %s\n",
 528                     speed, fc);
 529 }
 530
 531 /**
 532  * ice_vsi_link_event - update the vsi's netdev
 533  * @vsi: the vsi on which the link event occurred
 534  * @link_up: whether or not the vsi needs to be set up or down
 535  */
 536 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
 537 {
 538         if (!vsi || test_bit(__ICE_DOWN, vsi->state))
 539                 return;
 540
 541         if (vsi->type == ICE_VSI_PF) {
 542                 if (!vsi->netdev) {
 543                         dev_dbg(&vsi->back->pdev->dev,
 544                                 "vsi->netdev is not initialized!\n");
 545                         return;
 546                 }
 547                 if (link_up) {
 548                         netif_carrier_on(vsi->netdev);
 549                         netif_tx_wake_all_queues(vsi->netdev);
 550                 } else {
 551                         netif_carrier_off(vsi->netdev);
 552                         netif_tx_stop_all_queues(vsi->netdev);
 553                 }
 554         }
 555 }
 556
 557 /**
 558  * ice_link_event - process the link event
 559  * @pf: pf that the link event is associated with
 560  * @pi: port_info for the port that the link event is associated with
 561  *
 562  * Returns -EIO if ice_get_link_status() fails
 563  * Returns 0 on success
 564  */
 565 static int
 566 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi)
 567 {
 568         u8 new_link_speed, old_link_speed;
 569         struct ice_phy_info *phy_info;
 570         bool new_link_same_as_old;
 571         bool new_link, old_link;
 572         u8 lport;
 573         u16 v;
 574
 575         phy_info = &pi->phy;
 576         phy_info->link_info_old = phy_info->link_info;
 577         /* Force ice_get_link_status() to update link info */
 578         phy_info->get_link_info = true;
 579
 580         old_link = (phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
 581         old_link_speed = phy_info->link_info_old.link_speed;
 582
 583         lport = pi->lport;
 584         if (ice_get_link_status(pi, &new_link)) {
 585                 dev_dbg(&pf->pdev->dev,
 586                         "Could not get link status for port %d\n", lport);
 587                 return -EIO;
 588         }
 589
 590         new_link_speed = phy_info->link_info.link_speed;
 591
 592         new_link_same_as_old = (new_link == old_link &&
 593                                 new_link_speed == old_link_speed);
 594
 595         ice_for_each_vsi(pf, v) {
 596                 struct ice_vsi *vsi = pf->vsi[v];
 597
 598                 if (!vsi || !vsi->port_info)
 599                         continue;
 600
 601                 if (new_link_same_as_old &&
 602                     (test_bit(__ICE_DOWN, vsi->state) ||
 603                     new_link == netif_carrier_ok(vsi->netdev)))
 604                         continue;
 605
 606                 if (vsi->port_info->lport == lport) {
 607                         ice_print_link_msg(vsi, new_link);
 608                         ice_vsi_link_event(vsi, new_link);
 609                 }
 610         }
 611
 612         ice_vc_notify_link_state(pf);
 613
 614         return 0;
 615 }
 616
 617 /**
 618  * ice_watchdog_subtask - periodic tasks not using event driven scheduling
 619  * @pf: board private structure
 620  */
 621 static void ice_watchdog_subtask(struct ice_pf *pf)
 622 {
 623         int i;
 624
 625         /* if interface is down do nothing */
 626         if (test_bit(__ICE_DOWN, pf->state) ||
 627             test_bit(__ICE_CFG_BUSY, pf->state))
 628                 return;
 629
 630         /* make sure we don't do these things too often */
 631         if (time_before(jiffies,
 632                         pf->serv_tmr_prev + pf->serv_tmr_period))
 633                 return;
 634
 635         pf->serv_tmr_prev = jiffies;
 636
 637         if (ice_link_event(pf, pf->hw.port_info))
 638                 dev_dbg(&pf->pdev->dev, "ice_link_event failed\n");
 639
 640         /* Update the stats for active netdevs so the network stack
 641          * can look at updated numbers whenever it cares to
 642          */
 643         ice_update_pf_stats(pf);
 644         for (i = 0; i < pf->num_alloc_vsi; i++)
 645                 if (pf->vsi[i] && pf->vsi[i]->netdev)
 646                         ice_update_vsi_stats(pf->vsi[i]);
 647 }
 648
 649 /**
 650  * __ice_clean_ctrlq - helper function to clean controlq rings
 651  * @pf: ptr to struct ice_pf
 652  * @q_type: specific Control queue type
 653  */
 654 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
 655 {
 656         struct ice_rq_event_info event;
 657         struct ice_hw *hw = &pf->hw;
 658         struct ice_ctl_q_info *cq;
 659         u16 pending, i = 0;
 660         const char *qtype;
 661         u32 oldval, val;
 662
 663         /* Do not clean control queue if/when PF reset fails */
 664         if (test_bit(__ICE_RESET_FAILED, pf->state))
 665                 return 0;
 666
 667         switch (q_type) {
 668         case ICE_CTL_Q_ADMIN:
 669                 cq = &hw->adminq;
 670                 qtype = "Admin";
 671                 break;
 672         case ICE_CTL_Q_MAILBOX:
 673                 cq = &hw->mailboxq;
 674                 qtype = "Mailbox";
 675                 break;
 676         default:
 677                 dev_warn(&pf->pdev->dev, "Unknown control queue type 0x%x\n",
 678                          q_type);
 679                 return 0;
 680         }
 681
 682         /* check for error indications - PF_xx_AxQLEN register layout for
 683          * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
 684          */
 685         val = rd32(hw, cq->rq.len);
 686         if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 687                    PF_FW_ARQLEN_ARQCRIT_M)) {
 688                 oldval = val;
 689                 if (val & PF_FW_ARQLEN_ARQVFE_M)
 690                         dev_dbg(&pf->pdev->dev,
 691                                 "%s Receive Queue VF Error detected\n", qtype);
 692                 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
 693                         dev_dbg(&pf->pdev->dev,
 694                                 "%s Receive Queue Overflow Error detected\n",
 695                                 qtype);
 696                 }
 697                 if (val & PF_FW_ARQLEN_ARQCRIT_M)
 698                         dev_dbg(&pf->pdev->dev,
 699                                 "%s Receive Queue Critical Error detected\n",
 700                                 qtype);
 701                 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 702                          PF_FW_ARQLEN_ARQCRIT_M);
 703                 if (oldval != val)
 704                         wr32(hw, cq->rq.len, val);
 705         }
 706
 707         val = rd32(hw, cq->sq.len);
 708         if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 709                    PF_FW_ATQLEN_ATQCRIT_M)) {
 710                 oldval = val;
 711                 if (val & PF_FW_ATQLEN_ATQVFE_M)
 712                         dev_dbg(&pf->pdev->dev,
 713                                 "%s Send Queue VF Error detected\n", qtype);
 714                 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
 715                         dev_dbg(&pf->pdev->dev,
 716                                 "%s Send Queue Overflow Error detected\n",
 717                                 qtype);
 718                 }
 719                 if (val & PF_FW_ATQLEN_ATQCRIT_M)
 720                         dev_dbg(&pf->pdev->dev,
 721                                 "%s Send Queue Critical Error detected\n",
 722                                 qtype);
 723                 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 724                          PF_FW_ATQLEN_ATQCRIT_M);
 725                 if (oldval != val)
 726                         wr32(hw, cq->sq.len, val);
 727         }
 728
 729         event.buf_len = cq->rq_buf_size;
 730         event.msg_buf = devm_kzalloc(&pf->pdev->dev, event.buf_len,
 731                                      GFP_KERNEL);
 732         if (!event.msg_buf)
 733                 return 0;
 734
 735         do {
 736                 enum ice_status ret;
 737                 u16 opcode;
 738
 739                 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
 740                 if (ret == ICE_ERR_AQ_NO_WORK)
 741                         break;
 742                 if (ret) {
 743                         dev_err(&pf->pdev->dev,
 744                                 "%s Receive Queue event error %d\n", qtype,
 745                                 ret);
 746                         break;
 747                 }
 748
 749                 opcode = le16_to_cpu(event.desc.opcode);
 750
 751                 switch (opcode) {
 752                 case ice_mbx_opc_send_msg_to_pf:
 753                         ice_vc_process_vf_msg(pf, &event);
 754                         break;
 755                 case ice_aqc_opc_fw_logging:
 756                         ice_output_fw_log(hw, &event.desc, event.msg_buf);
 757                         break;
 758                 default:
 759                         dev_dbg(&pf->pdev->dev,
 760                                 "%s Receive Queue unknown event 0x%04x ignored\n",
 761                                 qtype, opcode);
 762                         break;
 763                 }
 764         } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
 765
 766         devm_kfree(&pf->pdev->dev, event.msg_buf);
 767
 768         return pending && (i == ICE_DFLT_IRQ_WORK);
 769 }
 770
 771 /**
 772  * ice_ctrlq_pending - check if there is a difference between ntc and ntu
 773  * @hw: pointer to hardware info
 774  * @cq: control queue information
 775  *
 776  * returns true if there are pending messages in a queue, false if there aren't
 777  */
 778 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
 779 {
 780         u16 ntu;
 781
 782         ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
 783         return cq->rq.next_to_clean != ntu;
 784 }
 785
 786 /**
 787  * ice_clean_adminq_subtask - clean the AdminQ rings
 788  * @pf: board private structure
 789  */
 790 static void ice_clean_adminq_subtask(struct ice_pf *pf)
 791 {
 792         struct ice_hw *hw = &pf->hw;
 793
 794         if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
 795                 return;
 796
 797         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
 798                 return;
 799
 800         clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
 801
 802         /* There might be a situation where new messages arrive to a control
 803          * queue between processing the last message and clearing the
 804          * EVENT_PENDING bit. So before exiting, check queue head again (using
 805          * ice_ctrlq_pending) and process new messages if any.
 806          */
 807         if (ice_ctrlq_pending(hw, &hw->adminq))
 808                 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
 809
 810         ice_flush(hw);
 811 }
 812
 813 /**
 814  * ice_clean_mailboxq_subtask - clean the MailboxQ rings
 815  * @pf: board private structure
 816  */
 817 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
 818 {
 819         struct ice_hw *hw = &pf->hw;
 820
 821         if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
 822                 return;
 823
 824         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
 825                 return;
 826
 827         clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
 828
 829         if (ice_ctrlq_pending(hw, &hw->mailboxq))
 830                 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
 831
 832         ice_flush(hw);
 833 }
 834
 835 /**
 836  * ice_service_task_schedule - schedule the service task to wake up
 837  * @pf: board private structure
 838  *
 839  * If not already scheduled, this puts the task into the work queue.
 840  */
 841 static void ice_service_task_schedule(struct ice_pf *pf)
 842 {
 843         if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
 844             !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
 845             !test_bit(__ICE_NEEDS_RESTART, pf->state))
 846                 queue_work(ice_wq, &pf->serv_task);
 847 }
 848
 849 /**
 850  * ice_service_task_complete - finish up the service task
 851  * @pf: board private structure
 852  */
 853 static void ice_service_task_complete(struct ice_pf *pf)
 854 {
 855         WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
 856
 857         /* force memory (pf->state) to sync before next service task */
 858         smp_mb__before_atomic();
 859         clear_bit(__ICE_SERVICE_SCHED, pf->state);
 860 }
 861
 862 /**
 863  * ice_service_task_stop - stop service task and cancel works
 864  * @pf: board private structure
 865  */
 866 static void ice_service_task_stop(struct ice_pf *pf)
 867 {
 868         set_bit(__ICE_SERVICE_DIS, pf->state);
 869
 870         if (pf->serv_tmr.function)
 871                 del_timer_sync(&pf->serv_tmr);
 872         if (pf->serv_task.func)
 873                 cancel_work_sync(&pf->serv_task);
 874
 875         clear_bit(__ICE_SERVICE_SCHED, pf->state);
 876 }
 877
 878 /**
 879  * ice_service_timer - timer callback to schedule service task
 880  * @t: pointer to timer_list
 881  */
 882 static void ice_service_timer(struct timer_list *t)
 883 {
 884         struct ice_pf *pf = from_timer(pf, t, serv_tmr);
 885
 886         mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
 887         ice_service_task_schedule(pf);
 888 }
 889
 890 /**
 891  * ice_handle_mdd_event - handle malicious driver detect event
 892  * @pf: pointer to the PF structure
 893  *
 894  * Called from service task. OICR interrupt handler indicates MDD event
 895  */
 896 static void ice_handle_mdd_event(struct ice_pf *pf)
 897 {
 898         struct ice_hw *hw = &pf->hw;
 899         bool mdd_detected = false;
 900         u32 reg;
 901         int i;
 902
 903         if (!test_bit(__ICE_MDD_EVENT_PENDING, pf->state))
 904                 return;
 905
 906         /* find what triggered the MDD event */
 907         reg = rd32(hw, GL_MDET_TX_PQM);
 908         if (reg & GL_MDET_TX_PQM_VALID_M) {
 909                 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
 910                                 GL_MDET_TX_PQM_PF_NUM_S;
 911                 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
 912                                 GL_MDET_TX_PQM_VF_NUM_S;
 913                 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
 914                                 GL_MDET_TX_PQM_MAL_TYPE_S;
 915                 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
 916                                 GL_MDET_TX_PQM_QNUM_S);
 917
 918                 if (netif_msg_tx_err(pf))
 919                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
 920                                  event, queue, pf_num, vf_num);
 921                 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
 922                 mdd_detected = true;
 923         }
 924
 925         reg = rd32(hw, GL_MDET_TX_TCLAN);
 926         if (reg & GL_MDET_TX_TCLAN_VALID_M) {
 927                 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
 928                                 GL_MDET_TX_TCLAN_PF_NUM_S;
 929                 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
 930                                 GL_MDET_TX_TCLAN_VF_NUM_S;
 931                 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
 932                                 GL_MDET_TX_TCLAN_MAL_TYPE_S;
 933                 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
 934                                 GL_MDET_TX_TCLAN_QNUM_S);
 935
 936                 if (netif_msg_rx_err(pf))
 937                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
 938                                  event, queue, pf_num, vf_num);
 939                 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
 940                 mdd_detected = true;
 941         }
 942
 943         reg = rd32(hw, GL_MDET_RX);
 944         if (reg & GL_MDET_RX_VALID_M) {
 945                 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
 946                                 GL_MDET_RX_PF_NUM_S;
 947                 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
 948                                 GL_MDET_RX_VF_NUM_S;
 949                 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
 950                                 GL_MDET_RX_MAL_TYPE_S;
 951                 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
 952                                 GL_MDET_RX_QNUM_S);
 953
 954                 if (netif_msg_rx_err(pf))
 955                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
 956                                  event, queue, pf_num, vf_num);
 957                 wr32(hw, GL_MDET_RX, 0xffffffff);
 958                 mdd_detected = true;
 959         }
 960
 961         if (mdd_detected) {
 962                 bool pf_mdd_detected = false;
 963
 964                 reg = rd32(hw, PF_MDET_TX_PQM);
 965                 if (reg & PF_MDET_TX_PQM_VALID_M) {
 966                         wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
 967                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
 968                         pf_mdd_detected = true;
 969                 }
 970
 971                 reg = rd32(hw, PF_MDET_TX_TCLAN);
 972                 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
 973                         wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
 974                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
 975                         pf_mdd_detected = true;
 976                 }
 977
 978                 reg = rd32(hw, PF_MDET_RX);
 979                 if (reg & PF_MDET_RX_VALID_M) {
 980                         wr32(hw, PF_MDET_RX, 0xFFFF);
 981                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
 982                         pf_mdd_detected = true;
 983                 }
 984                 /* Queue belongs to the PF initiate a reset */
 985                 if (pf_mdd_detected) {
 986                         set_bit(__ICE_NEEDS_RESTART, pf->state);
 987                         ice_service_task_schedule(pf);
 988                 }
 989         }
 990
 991         /* see if one of the VFs needs to be reset */
 992         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
 993                 struct ice_vf *vf = &pf->vf[i];
 994
 995                 reg = rd32(hw, VP_MDET_TX_PQM(i));
 996                 if (reg & VP_MDET_TX_PQM_VALID_M) {
 997                         wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
 998                         vf->num_mdd_events++;
 999                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1000                                  i);
1001                 }
1002
1003                 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1004                 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1005                         wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1006                         vf->num_mdd_events++;
1007                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1008                                  i);
1009                 }
1010
1011                 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1012                 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1013                         wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1014                         vf->num_mdd_events++;
1015                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1016                                  i);
1017                 }
1018
1019                 reg = rd32(hw, VP_MDET_RX(i));
1020                 if (reg & VP_MDET_RX_VALID_M) {
1021                         wr32(hw, VP_MDET_RX(i), 0xFFFF);
1022                         vf->num_mdd_events++;
1023                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
1024                                  i);
1025                 }
1026
1027                 if (vf->num_mdd_events > ICE_DFLT_NUM_MDD_EVENTS_ALLOWED) {
1028                         dev_info(&pf->pdev->dev,
1029                                  "Too many MDD events on VF %d, disabled\n", i);
1030                         dev_info(&pf->pdev->dev,
1031                                  "Use PF Control I/F to re-enable the VF\n");
1032                         set_bit(ICE_VF_STATE_DIS, vf->vf_states);
1033                 }
1034         }
1035
1036         /* re-enable MDD interrupt cause */
1037         clear_bit(__ICE_MDD_EVENT_PENDING, pf->state);
1038         reg = rd32(hw, PFINT_OICR_ENA);
1039         reg |= PFINT_OICR_MAL_DETECT_M;
1040         wr32(hw, PFINT_OICR_ENA, reg);
1041         ice_flush(hw);
1042 }
1043
1044 /**
1045  * ice_service_task - manage and run subtasks
1046  * @work: pointer to work_struct contained by the PF struct
1047  */
1048 static void ice_service_task(struct work_struct *work)
1049 {
1050         struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
1051         unsigned long start_time = jiffies;
1052
1053         /* subtasks */
1054
1055         /* process reset requests first */
1056         ice_reset_subtask(pf);
1057
1058         /* bail if a reset/recovery cycle is pending or rebuild failed */
1059         if (ice_is_reset_in_progress(pf->state) ||
1060             test_bit(__ICE_SUSPENDED, pf->state) ||
1061             test_bit(__ICE_NEEDS_RESTART, pf->state)) {
1062                 ice_service_task_complete(pf);
1063                 return;
1064         }
1065
1066         ice_check_for_hang_subtask(pf);
1067         ice_sync_fltr_subtask(pf);
1068         ice_handle_mdd_event(pf);
1069         ice_process_vflr_event(pf);
1070         ice_watchdog_subtask(pf);
1071         ice_clean_adminq_subtask(pf);
1072         ice_clean_mailboxq_subtask(pf);
1073
1074         /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
1075         ice_service_task_complete(pf);
1076
1077         /* If the tasks have taken longer than one service timer period
1078          * or there is more work to be done, reset the service timer to
1079          * schedule the service task now.
1080          */
1081         if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
1082             test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
1083             test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
1084             test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
1085             test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1086                 mod_timer(&pf->serv_tmr, jiffies);
1087 }
1088
1089 /**
1090  * ice_set_ctrlq_len - helper function to set controlq length
1091  * @hw: pointer to the hw instance
1092  */
1093 static void ice_set_ctrlq_len(struct ice_hw *hw)
1094 {
1095         hw->adminq.num_rq_entries = ICE_AQ_LEN;
1096         hw->adminq.num_sq_entries = ICE_AQ_LEN;
1097         hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
1098         hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
1099         hw->mailboxq.num_rq_entries = ICE_MBXQ_LEN;
1100         hw->mailboxq.num_sq_entries = ICE_MBXQ_LEN;
1101         hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1102         hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1103 }
1104
1105 /**
1106  * ice_irq_affinity_notify - Callback for affinity changes
1107  * @notify: context as to what irq was changed
1108  * @mask: the new affinity mask
1109  *
1110  * This is a callback function used by the irq_set_affinity_notifier function
1111  * so that we may register to receive changes to the irq affinity masks.
1112  */
1113 static void ice_irq_affinity_notify(struct irq_affinity_notify *notify,
1114                                     const cpumask_t *mask)
1115 {
1116         struct ice_q_vector *q_vector =
1117                 container_of(notify, struct ice_q_vector, affinity_notify);
1118
1119         cpumask_copy(&q_vector->affinity_mask, mask);
1120 }
1121
1122 /**
1123  * ice_irq_affinity_release - Callback for affinity notifier release
1124  * @ref: internal core kernel usage
1125  *
1126  * This is a callback function used by the irq_set_affinity_notifier function
1127  * to inform the current notification subscriber that they will no longer
1128  * receive notifications.
1129  */
1130 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
1131
1132 /**
1133  * ice_vsi_ena_irq - Enable IRQ for the given VSI
1134  * @vsi: the VSI being configured
1135  */
1136 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
1137 {
1138         struct ice_pf *pf = vsi->back;
1139         struct ice_hw *hw = &pf->hw;
1140
1141         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
1142                 int i;
1143
1144                 for (i = 0; i < vsi->num_q_vectors; i++)
1145                         ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
1146         }
1147
1148         ice_flush(hw);
1149         return 0;
1150 }
1151
1152 /**
1153  * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
1154  * @vsi: the VSI being configured
1155  * @basename: name for the vector
1156  */
1157 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
1158 {
1159         int q_vectors = vsi->num_q_vectors;
1160         struct ice_pf *pf = vsi->back;
1161         int base = vsi->sw_base_vector;
1162         int rx_int_idx = 0;
1163         int tx_int_idx = 0;
1164         int vector, err;
1165         int irq_num;
1166
1167         for (vector = 0; vector < q_vectors; vector++) {
1168                 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
1169
1170                 irq_num = pf->msix_entries[base + vector].vector;
1171
1172                 if (q_vector->tx.ring && q_vector->rx.ring) {
1173                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1174                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
1175                         tx_int_idx++;
1176                 } else if (q_vector->rx.ring) {
1177                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1178                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
1179                 } else if (q_vector->tx.ring) {
1180                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1181                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
1182                 } else {
1183                         /* skip this unused q_vector */
1184                         continue;
1185                 }
1186                 err = devm_request_irq(&pf->pdev->dev,
1187                                        pf->msix_entries[base + vector].vector,
1188                                        vsi->irq_handler, 0, q_vector->name,
1189                                        q_vector);
1190                 if (err) {
1191                         netdev_err(vsi->netdev,
1192                                    "MSIX request_irq failed, error: %d\n", err);
1193                         goto free_q_irqs;
1194                 }
1195
1196                 /* register for affinity change notifications */
1197                 q_vector->affinity_notify.notify = ice_irq_affinity_notify;
1198                 q_vector->affinity_notify.release = ice_irq_affinity_release;
1199                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
1200
1201                 /* assign the mask for this irq */
1202                 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
1203         }
1204
1205         vsi->irqs_ready = true;
1206         return 0;
1207
1208 free_q_irqs:
1209         while (vector) {
1210                 vector--;
1211                 irq_num = pf->msix_entries[base + vector].vector,
1212                 irq_set_affinity_notifier(irq_num, NULL);
1213                 irq_set_affinity_hint(irq_num, NULL);
1214                 devm_free_irq(&pf->pdev->dev, irq_num, &vsi->q_vectors[vector]);
1215         }
1216         return err;
1217 }
1218
1219 /**
1220  * ice_ena_misc_vector - enable the non-queue interrupts
1221  * @pf: board private structure
1222  */
1223 static void ice_ena_misc_vector(struct ice_pf *pf)
1224 {
1225         struct ice_hw *hw = &pf->hw;
1226         u32 val;
1227
1228         /* clear things first */
1229         wr32(hw, PFINT_OICR_ENA, 0);    /* disable all */
1230         rd32(hw, PFINT_OICR);           /* read to clear */
1231
1232         val = (PFINT_OICR_ECC_ERR_M |
1233                PFINT_OICR_MAL_DETECT_M |
1234                PFINT_OICR_GRST_M |
1235                PFINT_OICR_PCI_EXCEPTION_M |
1236                PFINT_OICR_VFLR_M |
1237                PFINT_OICR_HMC_ERR_M |
1238                PFINT_OICR_PE_CRITERR_M);
1239
1240         wr32(hw, PFINT_OICR_ENA, val);
1241
1242         /* SW_ITR_IDX = 0, but don't change INTENA */
1243         wr32(hw, GLINT_DYN_CTL(pf->hw_oicr_idx),
1244              GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
1245 }
1246
1247 /**
1248  * ice_misc_intr - misc interrupt handler
1249  * @irq: interrupt number
1250  * @data: pointer to a q_vector
1251  */
1252 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
1253 {
1254         struct ice_pf *pf = (struct ice_pf *)data;
1255         struct ice_hw *hw = &pf->hw;
1256         irqreturn_t ret = IRQ_NONE;
1257         u32 oicr, ena_mask;
1258
1259         set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1260         set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1261
1262         oicr = rd32(hw, PFINT_OICR);
1263         ena_mask = rd32(hw, PFINT_OICR_ENA);
1264
1265         if (oicr & PFINT_OICR_MAL_DETECT_M) {
1266                 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
1267                 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
1268         }
1269         if (oicr & PFINT_OICR_VFLR_M) {
1270                 ena_mask &= ~PFINT_OICR_VFLR_M;
1271                 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
1272         }
1273
1274         if (oicr & PFINT_OICR_GRST_M) {
1275                 u32 reset;
1276
1277                 /* we have a reset warning */
1278                 ena_mask &= ~PFINT_OICR_GRST_M;
1279                 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
1280                         GLGEN_RSTAT_RESET_TYPE_S;
1281
1282                 if (reset == ICE_RESET_CORER)
1283                         pf->corer_count++;
1284                 else if (reset == ICE_RESET_GLOBR)
1285                         pf->globr_count++;
1286                 else if (reset == ICE_RESET_EMPR)
1287                         pf->empr_count++;
1288                 else
1289                         dev_dbg(&pf->pdev->dev, "Invalid reset type %d\n",
1290                                 reset);
1291
1292                 /* If a reset cycle isn't already in progress, we set a bit in
1293                  * pf->state so that the service task can start a reset/rebuild.
1294                  * We also make note of which reset happened so that peer
1295                  * devices/drivers can be informed.
1296                  */
1297                 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
1298                         if (reset == ICE_RESET_CORER)
1299                                 set_bit(__ICE_CORER_RECV, pf->state);
1300                         else if (reset == ICE_RESET_GLOBR)
1301                                 set_bit(__ICE_GLOBR_RECV, pf->state);
1302                         else
1303                                 set_bit(__ICE_EMPR_RECV, pf->state);
1304
1305                         /* There are couple of different bits at play here.
1306                          * hw->reset_ongoing indicates whether the hardware is
1307                          * in reset. This is set to true when a reset interrupt
1308                          * is received and set back to false after the driver
1309                          * has determined that the hardware is out of reset.
1310                          *
1311                          * __ICE_RESET_OICR_RECV in pf->state indicates
1312                          * that a post reset rebuild is required before the
1313                          * driver is operational again. This is set above.
1314                          *
1315                          * As this is the start of the reset/rebuild cycle, set
1316                          * both to indicate that.
1317                          */
1318                         hw->reset_ongoing = true;
1319                 }
1320         }
1321
1322         if (oicr & PFINT_OICR_HMC_ERR_M) {
1323                 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
1324                 dev_dbg(&pf->pdev->dev,
1325                         "HMC Error interrupt - info 0x%x, data 0x%x\n",
1326                         rd32(hw, PFHMC_ERRORINFO),
1327                         rd32(hw, PFHMC_ERRORDATA));
1328         }
1329
1330         /* Report and mask off any remaining unexpected interrupts */
1331         oicr &= ena_mask;
1332         if (oicr) {
1333                 dev_dbg(&pf->pdev->dev, "unhandled interrupt oicr=0x%08x\n",
1334                         oicr);
1335                 /* If a critical error is pending there is no choice but to
1336                  * reset the device.
1337                  */
1338                 if (oicr & (PFINT_OICR_PE_CRITERR_M |
1339                             PFINT_OICR_PCI_EXCEPTION_M |
1340                             PFINT_OICR_ECC_ERR_M)) {
1341                         set_bit(__ICE_PFR_REQ, pf->state);
1342                         ice_service_task_schedule(pf);
1343                 }
1344                 ena_mask &= ~oicr;
1345         }
1346         ret = IRQ_HANDLED;
1347
1348         /* re-enable interrupt causes that are not handled during this pass */
1349         wr32(hw, PFINT_OICR_ENA, ena_mask);
1350         if (!test_bit(__ICE_DOWN, pf->state)) {
1351                 ice_service_task_schedule(pf);
1352                 ice_irq_dynamic_ena(hw, NULL, NULL);
1353         }
1354
1355         return ret;
1356 }
1357
1358 /**
1359  * ice_free_irq_msix_misc - Unroll misc vector setup
1360  * @pf: board private structure
1361  */
1362 static void ice_free_irq_msix_misc(struct ice_pf *pf)
1363 {
1364         /* disable OICR interrupt */
1365         wr32(&pf->hw, PFINT_OICR_ENA, 0);
1366         ice_flush(&pf->hw);
1367
1368         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags) && pf->msix_entries) {
1369                 synchronize_irq(pf->msix_entries[pf->sw_oicr_idx].vector);
1370                 devm_free_irq(&pf->pdev->dev,
1371                               pf->msix_entries[pf->sw_oicr_idx].vector, pf);
1372         }
1373
1374         pf->num_avail_sw_msix += 1;
1375         ice_free_res(pf->sw_irq_tracker, pf->sw_oicr_idx, ICE_RES_MISC_VEC_ID);
1376         pf->num_avail_hw_msix += 1;
1377         ice_free_res(pf->hw_irq_tracker, pf->hw_oicr_idx, ICE_RES_MISC_VEC_ID);
1378 }
1379
1380 /**
1381  * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
1382  * @pf: board private structure
1383  *
1384  * This sets up the handler for MSIX 0, which is used to manage the
1385  * non-queue interrupts, e.g. AdminQ and errors. This is not used
1386  * when in MSI or Legacy interrupt mode.
1387  */
1388 static int ice_req_irq_msix_misc(struct ice_pf *pf)
1389 {
1390         struct ice_hw *hw = &pf->hw;
1391         int oicr_idx, err = 0;
1392         u8 itr_gran;
1393         u32 val;
1394
1395         if (!pf->int_name[0])
1396                 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
1397                          dev_driver_string(&pf->pdev->dev),
1398                          dev_name(&pf->pdev->dev));
1399
1400         /* Do not request IRQ but do enable OICR interrupt since settings are
1401          * lost during reset. Note that this function is called only during
1402          * rebuild path and not while reset is in progress.
1403          */
1404         if (ice_is_reset_in_progress(pf->state))
1405                 goto skip_req_irq;
1406
1407         /* reserve one vector in sw_irq_tracker for misc interrupts */
1408         oicr_idx = ice_get_res(pf, pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1409         if (oicr_idx < 0)
1410                 return oicr_idx;
1411
1412         pf->num_avail_sw_msix -= 1;
1413         pf->sw_oicr_idx = oicr_idx;
1414
1415         /* reserve one vector in hw_irq_tracker for misc interrupts */
1416         oicr_idx = ice_get_res(pf, pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1417         if (oicr_idx < 0) {
1418                 ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1419                 pf->num_avail_sw_msix += 1;
1420                 return oicr_idx;
1421         }
1422         pf->num_avail_hw_msix -= 1;
1423         pf->hw_oicr_idx = oicr_idx;
1424
1425         err = devm_request_irq(&pf->pdev->dev,
1426                                pf->msix_entries[pf->sw_oicr_idx].vector,
1427                                ice_misc_intr, 0, pf->int_name, pf);
1428         if (err) {
1429                 dev_err(&pf->pdev->dev,
1430                         "devm_request_irq for %s failed: %d\n",
1431                         pf->int_name, err);
1432                 ice_free_res(pf->sw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1433                 pf->num_avail_sw_msix += 1;
1434                 ice_free_res(pf->hw_irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1435                 pf->num_avail_hw_msix += 1;
1436                 return err;
1437         }
1438
1439 skip_req_irq:
1440         ice_ena_misc_vector(pf);
1441
1442         val = ((pf->hw_oicr_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
1443                PFINT_OICR_CTL_CAUSE_ENA_M);
1444         wr32(hw, PFINT_OICR_CTL, val);
1445
1446         /* This enables Admin queue Interrupt causes */
1447         val = ((pf->hw_oicr_idx & PFINT_FW_CTL_MSIX_INDX_M) |
1448                PFINT_FW_CTL_CAUSE_ENA_M);
1449         wr32(hw, PFINT_FW_CTL, val);
1450
1451         /* This enables Mailbox queue Interrupt causes */
1452         val = ((pf->hw_oicr_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
1453                PFINT_MBX_CTL_CAUSE_ENA_M);
1454         wr32(hw, PFINT_MBX_CTL, val);
1455
1456         itr_gran = hw->itr_gran;
1457
1458         wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->hw_oicr_idx),
1459              ITR_TO_REG(ICE_ITR_8K, itr_gran));
1460
1461         ice_flush(hw);
1462         ice_irq_dynamic_ena(hw, NULL, NULL);
1463
1464         return 0;
1465 }
1466
1467 /**
1468  * ice_napi_del - Remove NAPI handler for the VSI
1469  * @vsi: VSI for which NAPI handler is to be removed
1470  */
1471 void ice_napi_del(struct ice_vsi *vsi)
1472 {
1473         int v_idx;
1474
1475         if (!vsi->netdev)
1476                 return;
1477
1478         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
1479                 netif_napi_del(&vsi->q_vectors[v_idx]->napi);
1480 }
1481
1482 /**
1483  * ice_napi_add - register NAPI handler for the VSI
1484  * @vsi: VSI for which NAPI handler is to be registered
1485  *
1486  * This function is only called in the driver's load path. Registering the NAPI
1487  * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
1488  * reset/rebuild, etc.)
1489  */
1490 static void ice_napi_add(struct ice_vsi *vsi)
1491 {
1492         int v_idx;
1493
1494         if (!vsi->netdev)
1495                 return;
1496
1497         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
1498                 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
1499                                ice_napi_poll, NAPI_POLL_WEIGHT);
1500 }
1501
1502 /**
1503  * ice_cfg_netdev - Allocate, configure and register a netdev
1504  * @vsi: the VSI associated with the new netdev
1505  *
1506  * Returns 0 on success, negative value on failure
1507  */
1508 static int ice_cfg_netdev(struct ice_vsi *vsi)
1509 {
1510         netdev_features_t csumo_features;
1511         netdev_features_t vlano_features;
1512         netdev_features_t dflt_features;
1513         netdev_features_t tso_features;
1514         struct ice_netdev_priv *np;
1515         struct net_device *netdev;
1516         u8 mac_addr[ETH_ALEN];
1517         int err;
1518
1519         netdev = alloc_etherdev_mqs(sizeof(struct ice_netdev_priv),
1520                                     vsi->alloc_txq, vsi->alloc_rxq);
1521         if (!netdev)
1522                 return -ENOMEM;
1523
1524         vsi->netdev = netdev;
1525         np = netdev_priv(netdev);
1526         np->vsi = vsi;
1527
1528         dflt_features = NETIF_F_SG      |
1529                         NETIF_F_HIGHDMA |
1530                         NETIF_F_RXHASH;
1531
1532         csumo_features = NETIF_F_RXCSUM   |
1533                          NETIF_F_IP_CSUM  |
1534                          NETIF_F_IPV6_CSUM;
1535
1536         vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
1537                          NETIF_F_HW_VLAN_CTAG_TX     |
1538                          NETIF_F_HW_VLAN_CTAG_RX;
1539
1540         tso_features = NETIF_F_TSO;
1541
1542         /* set features that user can change */
1543         netdev->hw_features = dflt_features | csumo_features |
1544                               vlano_features | tso_features;
1545
1546         /* enable features */
1547         netdev->features |= netdev->hw_features;
1548         /* encap and VLAN devices inherit default, csumo and tso features */
1549         netdev->hw_enc_features |= dflt_features | csumo_features |
1550                                    tso_features;
1551         netdev->vlan_features |= dflt_features | csumo_features |
1552                                  tso_features;
1553
1554         if (vsi->type == ICE_VSI_PF) {
1555                 SET_NETDEV_DEV(netdev, &vsi->back->pdev->dev);
1556                 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
1557
1558                 ether_addr_copy(netdev->dev_addr, mac_addr);
1559                 ether_addr_copy(netdev->perm_addr, mac_addr);
1560         }
1561
1562         netdev->priv_flags |= IFF_UNICAST_FLT;
1563
1564         /* assign netdev_ops */
1565         netdev->netdev_ops = &ice_netdev_ops;
1566
1567         /* setup watchdog timeout value to be 5 second */
1568         netdev->watchdog_timeo = 5 * HZ;
1569
1570         ice_set_ethtool_ops(netdev);
1571
1572         netdev->min_mtu = ETH_MIN_MTU;
1573         netdev->max_mtu = ICE_MAX_MTU;
1574
1575         err = register_netdev(vsi->netdev);
1576         if (err)
1577                 return err;
1578
1579         netif_carrier_off(vsi->netdev);
1580
1581         /* make sure transmit queues start off as stopped */
1582         netif_tx_stop_all_queues(vsi->netdev);
1583
1584         return 0;
1585 }
1586
1587 /**
1588  * ice_fill_rss_lut - Fill the RSS lookup table with default values
1589  * @lut: Lookup table
1590  * @rss_table_size: Lookup table size
1591  * @rss_size: Range of queue number for hashing
1592  */
1593 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
1594 {
1595         u16 i;
1596
1597         for (i = 0; i < rss_table_size; i++)
1598                 lut[i] = i % rss_size;
1599 }
1600
1601 /**
1602  * ice_pf_vsi_setup - Set up a PF VSI
1603  * @pf: board private structure
1604  * @pi: pointer to the port_info instance
1605  *
1606  * Returns pointer to the successfully allocated VSI sw struct on success,
1607  * otherwise returns NULL on failure.
1608  */
1609 static struct ice_vsi *
1610 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
1611 {
1612         return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
1613 }
1614
1615 /**
1616  * ice_vlan_rx_add_vid - Add a vlan id filter to HW offload
1617  * @netdev: network interface to be adjusted
1618  * @proto: unused protocol
1619  * @vid: vlan id to be added
1620  *
1621  * net_device_ops implementation for adding vlan ids
1622  */
1623 static int ice_vlan_rx_add_vid(struct net_device *netdev,
1624                                __always_unused __be16 proto, u16 vid)
1625 {
1626         struct ice_netdev_priv *np = netdev_priv(netdev);
1627         struct ice_vsi *vsi = np->vsi;
1628
1629         if (vid >= VLAN_N_VID) {
1630                 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
1631                            vid, VLAN_N_VID);
1632                 return -EINVAL;
1633         }
1634
1635         if (vsi->info.pvid)
1636                 return -EINVAL;
1637
1638         /* Enable VLAN pruning when VLAN 0 is added */
1639         if (unlikely(!vid)) {
1640                 int ret = ice_cfg_vlan_pruning(vsi, true);
1641
1642                 if (ret)
1643                         return ret;
1644         }
1645
1646         /* Add all VLAN ids including 0 to the switch filter. VLAN id 0 is
1647          * needed to continue allowing all untagged packets since VLAN prune
1648          * list is applied to all packets by the switch
1649          */
1650         return ice_vsi_add_vlan(vsi, vid);
1651 }
1652
1653 /**
1654  * ice_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
1655  * @netdev: network interface to be adjusted
1656  * @proto: unused protocol
1657  * @vid: vlan id to be removed
1658  *
1659  * net_device_ops implementation for removing vlan ids
1660  */
1661 static int ice_vlan_rx_kill_vid(struct net_device *netdev,
1662                                 __always_unused __be16 proto, u16 vid)
1663 {
1664         struct ice_netdev_priv *np = netdev_priv(netdev);
1665         struct ice_vsi *vsi = np->vsi;
1666         int status;
1667
1668         if (vsi->info.pvid)
1669                 return -EINVAL;
1670
1671         /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
1672          * information
1673          */
1674         status = ice_vsi_kill_vlan(vsi, vid);
1675         if (status)
1676                 return status;
1677
1678         /* Disable VLAN pruning when VLAN 0 is removed */
1679         if (unlikely(!vid))
1680                 status = ice_cfg_vlan_pruning(vsi, false);
1681
1682         return status;
1683 }
1684
1685 /**
1686  * ice_setup_pf_sw - Setup the HW switch on startup or after reset
1687  * @pf: board private structure
1688  *
1689  * Returns 0 on success, negative value on failure
1690  */
1691 static int ice_setup_pf_sw(struct ice_pf *pf)
1692 {
1693         LIST_HEAD(tmp_add_list);
1694         u8 broadcast[ETH_ALEN];
1695         struct ice_vsi *vsi;
1696         int status = 0;
1697
1698         if (ice_is_reset_in_progress(pf->state))
1699                 return -EBUSY;
1700
1701         vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
1702         if (!vsi) {
1703                 status = -ENOMEM;
1704                 goto unroll_vsi_setup;
1705         }
1706
1707         status = ice_cfg_netdev(vsi);
1708         if (status) {
1709                 status = -ENODEV;
1710                 goto unroll_vsi_setup;
1711         }
1712
1713         /* registering the NAPI handler requires both the queues and
1714          * netdev to be created, which are done in ice_pf_vsi_setup()
1715          * and ice_cfg_netdev() respectively
1716          */
1717         ice_napi_add(vsi);
1718
1719         /* To add a MAC filter, first add the MAC to a list and then
1720          * pass the list to ice_add_mac.
1721          */
1722
1723          /* Add a unicast MAC filter so the VSI can get its packets */
1724         status = ice_add_mac_to_list(vsi, &tmp_add_list,
1725                                      vsi->port_info->mac.perm_addr);
1726         if (status)
1727                 goto unroll_napi_add;
1728
1729         /* VSI needs to receive broadcast traffic, so add the broadcast
1730          * MAC address to the list as well.
1731          */
1732         eth_broadcast_addr(broadcast);
1733         status = ice_add_mac_to_list(vsi, &tmp_add_list, broadcast);
1734         if (status)
1735                 goto free_mac_list;
1736
1737         /* program MAC filters for entries in tmp_add_list */
1738         status = ice_add_mac(&pf->hw, &tmp_add_list);
1739         if (status) {
1740                 dev_err(&pf->pdev->dev, "Could not add MAC filters\n");
1741                 status = -ENOMEM;
1742                 goto free_mac_list;
1743         }
1744
1745         ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list);
1746         return status;
1747
1748 free_mac_list:
1749         ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list);
1750
1751 unroll_napi_add:
1752         if (vsi) {
1753                 ice_napi_del(vsi);
1754                 if (vsi->netdev) {
1755                         if (vsi->netdev->reg_state == NETREG_REGISTERED)
1756                                 unregister_netdev(vsi->netdev);
1757                         free_netdev(vsi->netdev);
1758                         vsi->netdev = NULL;
1759                 }
1760         }
1761
1762 unroll_vsi_setup:
1763         if (vsi) {
1764                 ice_vsi_free_q_vectors(vsi);
1765                 ice_vsi_delete(vsi);
1766                 ice_vsi_put_qs(vsi);
1767                 pf->q_left_tx += vsi->alloc_txq;
1768                 pf->q_left_rx += vsi->alloc_rxq;
1769                 ice_vsi_clear(vsi);
1770         }
1771         return status;
1772 }
1773
1774 /**
1775  * ice_determine_q_usage - Calculate queue distribution
1776  * @pf: board private structure
1777  *
1778  * Return -ENOMEM if we don't get enough queues for all ports
1779  */
1780 static void ice_determine_q_usage(struct ice_pf *pf)
1781 {
1782         u16 q_left_tx, q_left_rx;
1783
1784         q_left_tx = pf->hw.func_caps.common_cap.num_txq;
1785         q_left_rx = pf->hw.func_caps.common_cap.num_rxq;
1786
1787         pf->num_lan_tx = min_t(int, q_left_tx, num_online_cpus());
1788
1789         /* only 1 Rx queue unless RSS is enabled */
1790         if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags))
1791                 pf->num_lan_rx = 1;
1792         else
1793                 pf->num_lan_rx = min_t(int, q_left_rx, num_online_cpus());
1794
1795         pf->q_left_tx = q_left_tx - pf->num_lan_tx;
1796         pf->q_left_rx = q_left_rx - pf->num_lan_rx;
1797 }
1798
1799 /**
1800  * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
1801  * @pf: board private structure to initialize
1802  */
1803 static void ice_deinit_pf(struct ice_pf *pf)
1804 {
1805         ice_service_task_stop(pf);
1806         mutex_destroy(&pf->sw_mutex);
1807         mutex_destroy(&pf->avail_q_mutex);
1808 }
1809
1810 /**
1811  * ice_init_pf - Initialize general software structures (struct ice_pf)
1812  * @pf: board private structure to initialize
1813  */
1814 static void ice_init_pf(struct ice_pf *pf)
1815 {
1816         bitmap_zero(pf->flags, ICE_PF_FLAGS_NBITS);
1817         set_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1818 #ifdef CONFIG_PCI_IOV
1819         if (pf->hw.func_caps.common_cap.sr_iov_1_1) {
1820                 struct ice_hw *hw = &pf->hw;
1821
1822                 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
1823                 pf->num_vfs_supported = min_t(int, hw->func_caps.num_allocd_vfs,
1824                                               ICE_MAX_VF_COUNT);
1825         }
1826 #endif /* CONFIG_PCI_IOV */
1827
1828         mutex_init(&pf->sw_mutex);
1829         mutex_init(&pf->avail_q_mutex);
1830
1831         /* Clear avail_[t|r]x_qs bitmaps (set all to avail) */
1832         mutex_lock(&pf->avail_q_mutex);
1833         bitmap_zero(pf->avail_txqs, ICE_MAX_TXQS);
1834         bitmap_zero(pf->avail_rxqs, ICE_MAX_RXQS);
1835         mutex_unlock(&pf->avail_q_mutex);
1836
1837         if (pf->hw.func_caps.common_cap.rss_table_size)
1838                 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
1839
1840         /* setup service timer and periodic service task */
1841         timer_setup(&pf->serv_tmr, ice_service_timer, 0);
1842         pf->serv_tmr_period = HZ;
1843         INIT_WORK(&pf->serv_task, ice_service_task);
1844         clear_bit(__ICE_SERVICE_SCHED, pf->state);
1845 }
1846
1847 /**
1848  * ice_ena_msix_range - Request a range of MSIX vectors from the OS
1849  * @pf: board private structure
1850  *
1851  * compute the number of MSIX vectors required (v_budget) and request from
1852  * the OS. Return the number of vectors reserved or negative on failure
1853  */
1854 static int ice_ena_msix_range(struct ice_pf *pf)
1855 {
1856         int v_left, v_actual, v_budget = 0;
1857         int needed, err, i;
1858
1859         v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
1860
1861         /* reserve one vector for miscellaneous handler */
1862         needed = 1;
1863         v_budget += needed;
1864         v_left -= needed;
1865
1866         /* reserve vectors for LAN traffic */
1867         pf->num_lan_msix = min_t(int, num_online_cpus(), v_left);
1868         v_budget += pf->num_lan_msix;
1869         v_left -= pf->num_lan_msix;
1870
1871         pf->msix_entries = devm_kcalloc(&pf->pdev->dev, v_budget,
1872                                         sizeof(struct msix_entry), GFP_KERNEL);
1873
1874         if (!pf->msix_entries) {
1875                 err = -ENOMEM;
1876                 goto exit_err;
1877         }
1878
1879         for (i = 0; i < v_budget; i++)
1880                 pf->msix_entries[i].entry = i;
1881
1882         /* actually reserve the vectors */
1883         v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
1884                                          ICE_MIN_MSIX, v_budget);
1885
1886         if (v_actual < 0) {
1887                 dev_err(&pf->pdev->dev, "unable to reserve MSI-X vectors\n");
1888                 err = v_actual;
1889                 goto msix_err;
1890         }
1891
1892         if (v_actual < v_budget) {
1893                 dev_warn(&pf->pdev->dev,
1894                          "not enough vectors. requested = %d, obtained = %d\n",
1895                          v_budget, v_actual);
1896                 if (v_actual >= (pf->num_lan_msix + 1)) {
1897                         pf->num_avail_sw_msix = v_actual -
1898                                                 (pf->num_lan_msix + 1);
1899                 } else if (v_actual >= 2) {
1900                         pf->num_lan_msix = 1;
1901                         pf->num_avail_sw_msix = v_actual - 2;
1902                 } else {
1903                         pci_disable_msix(pf->pdev);
1904                         err = -ERANGE;
1905                         goto msix_err;
1906                 }
1907         }
1908
1909         return v_actual;
1910
1911 msix_err:
1912         devm_kfree(&pf->pdev->dev, pf->msix_entries);
1913         goto exit_err;
1914
1915 exit_err:
1916         pf->num_lan_msix = 0;
1917         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1918         return err;
1919 }
1920
1921 /**
1922  * ice_dis_msix - Disable MSI-X interrupt setup in OS
1923  * @pf: board private structure
1924  */
1925 static void ice_dis_msix(struct ice_pf *pf)
1926 {
1927         pci_disable_msix(pf->pdev);
1928         devm_kfree(&pf->pdev->dev, pf->msix_entries);
1929         pf->msix_entries = NULL;
1930         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
1931 }
1932
1933 /**
1934  * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
1935  * @pf: board private structure
1936  */
1937 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
1938 {
1939         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
1940                 ice_dis_msix(pf);
1941
1942         if (pf->sw_irq_tracker) {
1943                 devm_kfree(&pf->pdev->dev, pf->sw_irq_tracker);
1944                 pf->sw_irq_tracker = NULL;
1945         }
1946
1947         if (pf->hw_irq_tracker) {
1948                 devm_kfree(&pf->pdev->dev, pf->hw_irq_tracker);
1949                 pf->hw_irq_tracker = NULL;
1950         }
1951 }
1952
1953 /**
1954  * ice_init_interrupt_scheme - Determine proper interrupt scheme
1955  * @pf: board private structure to initialize
1956  */
1957 static int ice_init_interrupt_scheme(struct ice_pf *pf)
1958 {
1959         int vectors = 0, hw_vectors = 0;
1960         ssize_t size;
1961
1962         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
1963                 vectors = ice_ena_msix_range(pf);
1964         else
1965                 return -ENODEV;
1966
1967         if (vectors < 0)
1968                 return vectors;
1969
1970         /* set up vector assignment tracking */
1971         size = sizeof(struct ice_res_tracker) + (sizeof(u16) * vectors);
1972
1973         pf->sw_irq_tracker = devm_kzalloc(&pf->pdev->dev, size, GFP_KERNEL);
1974         if (!pf->sw_irq_tracker) {
1975                 ice_dis_msix(pf);
1976                 return -ENOMEM;
1977         }
1978
1979         /* populate SW interrupts pool with number of OS granted IRQs. */
1980         pf->num_avail_sw_msix = vectors;
1981         pf->sw_irq_tracker->num_entries = vectors;
1982
1983         /* set up HW vector assignment tracking */
1984         hw_vectors = pf->hw.func_caps.common_cap.num_msix_vectors;
1985         size = sizeof(struct ice_res_tracker) + (sizeof(u16) * hw_vectors);
1986
1987         pf->hw_irq_tracker = devm_kzalloc(&pf->pdev->dev, size, GFP_KERNEL);
1988         if (!pf->hw_irq_tracker) {
1989                 ice_clear_interrupt_scheme(pf);
1990                 return -ENOMEM;
1991         }
1992
1993         /* populate HW interrupts pool with number of HW supported irqs. */
1994         pf->num_avail_hw_msix = hw_vectors;
1995         pf->hw_irq_tracker->num_entries = hw_vectors;
1996
1997         return 0;
1998 }
1999
2000 /**
2001  * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
2002  * @pf: pointer to the PF structure
2003  *
2004  * There is no error returned here because the driver should be able to handle
2005  * 128 Byte cache lines, so we only print a warning in case issues are seen,
2006  * specifically with Tx.
2007  */
2008 static void ice_verify_cacheline_size(struct ice_pf *pf)
2009 {
2010         if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
2011                 dev_warn(&pf->pdev->dev,
2012                          "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
2013                          ICE_CACHE_LINE_BYTES);
2014 }
2015
2016 /**
2017  * ice_probe - Device initialization routine
2018  * @pdev: PCI device information struct
2019  * @ent: entry in ice_pci_tbl
2020  *
2021  * Returns 0 on success, negative on failure
2022  */
2023 static int ice_probe(struct pci_dev *pdev,
2024                      const struct pci_device_id __always_unused *ent)
2025 {
2026         struct ice_pf *pf;
2027         struct ice_hw *hw;
2028         int err;
2029
2030         /* this driver uses devres, see Documentation/driver-model/devres.txt */
2031         err = pcim_enable_device(pdev);
2032         if (err)
2033                 return err;
2034
2035         err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
2036         if (err) {
2037                 dev_err(&pdev->dev, "BAR0 I/O map error %d\n", err);
2038                 return err;
2039         }
2040
2041         pf = devm_kzalloc(&pdev->dev, sizeof(*pf), GFP_KERNEL);
2042         if (!pf)
2043                 return -ENOMEM;
2044
2045         /* set up for high or low dma */
2046         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
2047         if (err)
2048                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
2049         if (err) {
2050                 dev_err(&pdev->dev, "DMA configuration failed: 0x%x\n", err);
2051                 return err;
2052         }
2053
2054         pci_enable_pcie_error_reporting(pdev);
2055         pci_set_master(pdev);
2056
2057         pf->pdev = pdev;
2058         pci_set_drvdata(pdev, pf);
2059         set_bit(__ICE_DOWN, pf->state);
2060         /* Disable service task until DOWN bit is cleared */
2061         set_bit(__ICE_SERVICE_DIS, pf->state);
2062
2063         hw = &pf->hw;
2064         hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
2065         hw->back = pf;
2066         hw->vendor_id = pdev->vendor;
2067         hw->device_id = pdev->device;
2068         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
2069         hw->subsystem_vendor_id = pdev->subsystem_vendor;
2070         hw->subsystem_device_id = pdev->subsystem_device;
2071         hw->bus.device = PCI_SLOT(pdev->devfn);
2072         hw->bus.func = PCI_FUNC(pdev->devfn);
2073         ice_set_ctrlq_len(hw);
2074
2075         pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
2076
2077 #ifndef CONFIG_DYNAMIC_DEBUG
2078         if (debug < -1)
2079                 hw->debug_mask = debug;
2080 #endif
2081
2082         err = ice_init_hw(hw);
2083         if (err) {
2084                 dev_err(&pdev->dev, "ice_init_hw failed: %d\n", err);
2085                 err = -EIO;
2086                 goto err_exit_unroll;
2087         }
2088
2089         dev_info(&pdev->dev, "firmware %d.%d.%05d api %d.%d\n",
2090                  hw->fw_maj_ver, hw->fw_min_ver, hw->fw_build,
2091                  hw->api_maj_ver, hw->api_min_ver);
2092
2093         ice_init_pf(pf);
2094
2095         ice_determine_q_usage(pf);
2096
2097         pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
2098         if (!pf->num_alloc_vsi) {
2099                 err = -EIO;
2100                 goto err_init_pf_unroll;
2101         }
2102
2103         pf->vsi = devm_kcalloc(&pdev->dev, pf->num_alloc_vsi,
2104                                sizeof(struct ice_vsi *), GFP_KERNEL);
2105         if (!pf->vsi) {
2106                 err = -ENOMEM;
2107                 goto err_init_pf_unroll;
2108         }
2109
2110         err = ice_init_interrupt_scheme(pf);
2111         if (err) {
2112                 dev_err(&pdev->dev,
2113                         "ice_init_interrupt_scheme failed: %d\n", err);
2114                 err = -EIO;
2115                 goto err_init_interrupt_unroll;
2116         }
2117
2118         /* Driver is mostly up */
2119         clear_bit(__ICE_DOWN, pf->state);
2120
2121         /* In case of MSIX we are going to setup the misc vector right here
2122          * to handle admin queue events etc. In case of legacy and MSI
2123          * the misc functionality and queue processing is combined in
2124          * the same vector and that gets setup at open.
2125          */
2126         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
2127                 err = ice_req_irq_msix_misc(pf);
2128                 if (err) {
2129                         dev_err(&pdev->dev,
2130                                 "setup of misc vector failed: %d\n", err);
2131                         goto err_init_interrupt_unroll;
2132                 }
2133         }
2134
2135         /* create switch struct for the switch element created by FW on boot */
2136         pf->first_sw = devm_kzalloc(&pdev->dev, sizeof(struct ice_sw),
2137                                     GFP_KERNEL);
2138         if (!pf->first_sw) {
2139                 err = -ENOMEM;
2140                 goto err_msix_misc_unroll;
2141         }
2142
2143         if (hw->evb_veb)
2144                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
2145         else
2146                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
2147
2148         pf->first_sw->pf = pf;
2149
2150         /* record the sw_id available for later use */
2151         pf->first_sw->sw_id = hw->port_info->sw_id;
2152
2153         err = ice_setup_pf_sw(pf);
2154         if (err) {
2155                 dev_err(&pdev->dev,
2156                         "probe failed due to setup pf switch:%d\n", err);
2157                 goto err_alloc_sw_unroll;
2158         }
2159
2160         clear_bit(__ICE_SERVICE_DIS, pf->state);
2161
2162         /* since everything is good, start the service timer */
2163         mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
2164
2165         ice_verify_cacheline_size(pf);
2166
2167         return 0;
2168
2169 err_alloc_sw_unroll:
2170         set_bit(__ICE_SERVICE_DIS, pf->state);
2171         set_bit(__ICE_DOWN, pf->state);
2172         devm_kfree(&pf->pdev->dev, pf->first_sw);
2173 err_msix_misc_unroll:
2174         ice_free_irq_msix_misc(pf);
2175 err_init_interrupt_unroll:
2176         ice_clear_interrupt_scheme(pf);
2177         devm_kfree(&pdev->dev, pf->vsi);
2178 err_init_pf_unroll:
2179         ice_deinit_pf(pf);
2180         ice_deinit_hw(hw);
2181 err_exit_unroll:
2182         pci_disable_pcie_error_reporting(pdev);
2183         return err;
2184 }
2185
2186 /**
2187  * ice_remove - Device removal routine
2188  * @pdev: PCI device information struct
2189  */
2190 static void ice_remove(struct pci_dev *pdev)
2191 {
2192         struct ice_pf *pf = pci_get_drvdata(pdev);
2193         int i;
2194
2195         if (!pf)
2196                 return;
2197
2198         for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
2199                 if (!ice_is_reset_in_progress(pf->state))
2200                         break;
2201                 msleep(100);
2202         }
2203
2204         set_bit(__ICE_DOWN, pf->state);
2205         ice_service_task_stop(pf);
2206
2207         if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags))
2208                 ice_free_vfs(pf);
2209         ice_vsi_release_all(pf);
2210         ice_free_irq_msix_misc(pf);
2211         ice_for_each_vsi(pf, i) {
2212                 if (!pf->vsi[i])
2213                         continue;
2214                 ice_vsi_free_q_vectors(pf->vsi[i]);
2215         }
2216         ice_clear_interrupt_scheme(pf);
2217         ice_deinit_pf(pf);
2218         ice_deinit_hw(&pf->hw);
2219         pci_disable_pcie_error_reporting(pdev);
2220 }
2221
2222 /* ice_pci_tbl - PCI Device ID Table
2223  *
2224  * Wildcard entries (PCI_ANY_ID) should come last
2225  * Last entry must be all 0s
2226  *
2227  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
2228  *   Class, Class Mask, private data (not used) }
2229  */
2230 static const struct pci_device_id ice_pci_tbl[] = {
2231         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
2232         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
2233         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
2234         /* required last entry */
2235         { 0, }
2236 };
2237 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
2238
2239 static struct pci_driver ice_driver = {
2240         .name = KBUILD_MODNAME,
2241         .id_table = ice_pci_tbl,
2242         .probe = ice_probe,
2243         .remove = ice_remove,
2244         .sriov_configure = ice_sriov_configure,
2245 };
2246
2247 /**
2248  * ice_module_init - Driver registration routine
2249  *
2250  * ice_module_init is the first routine called when the driver is
2251  * loaded. All it does is register with the PCI subsystem.
2252  */
2253 static int __init ice_module_init(void)
2254 {
2255         int status;
2256
2257         pr_info("%s - version %s\n", ice_driver_string, ice_drv_ver);
2258         pr_info("%s\n", ice_copyright);
2259
2260         ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
2261         if (!ice_wq) {
2262                 pr_err("Failed to create workqueue\n");
2263                 return -ENOMEM;
2264         }
2265
2266         status = pci_register_driver(&ice_driver);
2267         if (status) {
2268                 pr_err("failed to register pci driver, err %d\n", status);
2269                 destroy_workqueue(ice_wq);
2270         }
2271
2272         return status;
2273 }
2274 module_init(ice_module_init);
2275
2276 /**
2277  * ice_module_exit - Driver exit cleanup routine
2278  *
2279  * ice_module_exit is called just before the driver is removed
2280  * from memory.
2281  */
2282 static void __exit ice_module_exit(void)
2283 {
2284         pci_unregister_driver(&ice_driver);
2285         destroy_workqueue(ice_wq);
2286         pr_info("module unloaded\n");
2287 }
2288 module_exit(ice_module_exit);
2289
2290 /**
2291  * ice_set_mac_address - NDO callback to set mac address
2292  * @netdev: network interface device structure
2293  * @pi: pointer to an address structure
2294  *
2295  * Returns 0 on success, negative on failure
2296  */
2297 static int ice_set_mac_address(struct net_device *netdev, void *pi)
2298 {
2299         struct ice_netdev_priv *np = netdev_priv(netdev);
2300         struct ice_vsi *vsi = np->vsi;
2301         struct ice_pf *pf = vsi->back;
2302         struct ice_hw *hw = &pf->hw;
2303         struct sockaddr *addr = pi;
2304         enum ice_status status;
2305         LIST_HEAD(a_mac_list);
2306         LIST_HEAD(r_mac_list);
2307         u8 flags = 0;
2308         int err;
2309         u8 *mac;
2310
2311         mac = (u8 *)addr->sa_data;
2312
2313         if (!is_valid_ether_addr(mac))
2314                 return -EADDRNOTAVAIL;
2315
2316         if (ether_addr_equal(netdev->dev_addr, mac)) {
2317                 netdev_warn(netdev, "already using mac %pM\n", mac);
2318                 return 0;
2319         }
2320
2321         if (test_bit(__ICE_DOWN, pf->state) ||
2322             ice_is_reset_in_progress(pf->state)) {
2323                 netdev_err(netdev, "can't set mac %pM. device not ready\n",
2324                            mac);
2325                 return -EBUSY;
2326         }
2327
2328         /* When we change the mac address we also have to change the mac address
2329          * based filter rules that were created previously for the old mac
2330          * address. So first, we remove the old filter rule using ice_remove_mac
2331          * and then create a new filter rule using ice_add_mac. Note that for
2332          * both these operations, we first need to form a "list" of mac
2333          * addresses (even though in this case, we have only 1 mac address to be
2334          * added/removed) and this done using ice_add_mac_to_list. Depending on
2335          * the ensuing operation this "list" of mac addresses is either to be
2336          * added or removed from the filter.
2337          */
2338         err = ice_add_mac_to_list(vsi, &r_mac_list, netdev->dev_addr);
2339         if (err) {
2340                 err = -EADDRNOTAVAIL;
2341                 goto free_lists;
2342         }
2343
2344         status = ice_remove_mac(hw, &r_mac_list);
2345         if (status) {
2346                 err = -EADDRNOTAVAIL;
2347                 goto free_lists;
2348         }
2349
2350         err = ice_add_mac_to_list(vsi, &a_mac_list, mac);
2351         if (err) {
2352                 err = -EADDRNOTAVAIL;
2353                 goto free_lists;
2354         }
2355
2356         status = ice_add_mac(hw, &a_mac_list);
2357         if (status) {
2358                 err = -EADDRNOTAVAIL;
2359                 goto free_lists;
2360         }
2361
2362 free_lists:
2363         /* free list entries */
2364         ice_free_fltr_list(&pf->pdev->dev, &r_mac_list);
2365         ice_free_fltr_list(&pf->pdev->dev, &a_mac_list);
2366
2367         if (err) {
2368                 netdev_err(netdev, "can't set mac %pM. filter update failed\n",
2369                            mac);
2370                 return err;
2371         }
2372
2373         /* change the netdev's mac address */
2374         memcpy(netdev->dev_addr, mac, netdev->addr_len);
2375         netdev_dbg(vsi->netdev, "updated mac address to %pM\n",
2376                    netdev->dev_addr);
2377
2378         /* write new mac address to the firmware */
2379         flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
2380         status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
2381         if (status) {
2382                 netdev_err(netdev, "can't set mac %pM. write to firmware failed.\n",
2383                            mac);
2384         }
2385         return 0;
2386 }
2387
2388 /**
2389  * ice_set_rx_mode - NDO callback to set the netdev filters
2390  * @netdev: network interface device structure
2391  */
2392 static void ice_set_rx_mode(struct net_device *netdev)
2393 {
2394         struct ice_netdev_priv *np = netdev_priv(netdev);
2395         struct ice_vsi *vsi = np->vsi;
2396
2397         if (!vsi)
2398                 return;
2399
2400         /* Set the flags to synchronize filters
2401          * ndo_set_rx_mode may be triggered even without a change in netdev
2402          * flags
2403          */
2404         set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
2405         set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
2406         set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
2407
2408         /* schedule our worker thread which will take care of
2409          * applying the new filter changes
2410          */
2411         ice_service_task_schedule(vsi->back);
2412 }
2413
2414 /**
2415  * ice_fdb_add - add an entry to the hardware database
2416  * @ndm: the input from the stack
2417  * @tb: pointer to array of nladdr (unused)
2418  * @dev: the net device pointer
2419  * @addr: the MAC address entry being added
2420  * @vid: VLAN id
2421  * @flags: instructions from stack about fdb operation
2422  */
2423 static int ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
2424                        struct net_device *dev, const unsigned char *addr,
2425                        u16 vid, u16 flags)
2426 {
2427         int err;
2428
2429         if (vid) {
2430                 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
2431                 return -EINVAL;
2432         }
2433         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2434                 netdev_err(dev, "FDB only supports static addresses\n");
2435                 return -EINVAL;
2436         }
2437
2438         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2439                 err = dev_uc_add_excl(dev, addr);
2440         else if (is_multicast_ether_addr(addr))
2441                 err = dev_mc_add_excl(dev, addr);
2442         else
2443                 err = -EINVAL;
2444
2445         /* Only return duplicate errors if NLM_F_EXCL is set */
2446         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2447                 err = 0;
2448
2449         return err;
2450 }
2451
2452 /**
2453  * ice_fdb_del - delete an entry from the hardware database
2454  * @ndm: the input from the stack
2455  * @tb: pointer to array of nladdr (unused)
2456  * @dev: the net device pointer
2457  * @addr: the MAC address entry being added
2458  * @vid: VLAN id
2459  */
2460 static int ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
2461                        struct net_device *dev, const unsigned char *addr,
2462                        __always_unused u16 vid)
2463 {
2464         int err;
2465
2466         if (ndm->ndm_state & NUD_PERMANENT) {
2467                 netdev_err(dev, "FDB only supports static addresses\n");
2468                 return -EINVAL;
2469         }
2470
2471         if (is_unicast_ether_addr(addr))
2472                 err = dev_uc_del(dev, addr);
2473         else if (is_multicast_ether_addr(addr))
2474                 err = dev_mc_del(dev, addr);
2475         else
2476                 err = -EINVAL;
2477
2478         return err;
2479 }
2480
2481 /**
2482  * ice_set_features - set the netdev feature flags
2483  * @netdev: ptr to the netdev being adjusted
2484  * @features: the feature set that the stack is suggesting
2485  */
2486 static int ice_set_features(struct net_device *netdev,
2487                             netdev_features_t features)
2488 {
2489         struct ice_netdev_priv *np = netdev_priv(netdev);
2490         struct ice_vsi *vsi = np->vsi;
2491         int ret = 0;
2492
2493         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
2494                 ret = ice_vsi_manage_rss_lut(vsi, true);
2495         else if (!(features & NETIF_F_RXHASH) &&
2496                  netdev->features & NETIF_F_RXHASH)
2497                 ret = ice_vsi_manage_rss_lut(vsi, false);
2498
2499         if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
2500             !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2501                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2502         else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
2503                  (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2504                 ret = ice_vsi_manage_vlan_stripping(vsi, false);
2505         else if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
2506                  !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2507                 ret = ice_vsi_manage_vlan_insertion(vsi);
2508         else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
2509                  (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2510                 ret = ice_vsi_manage_vlan_insertion(vsi);
2511
2512         return ret;
2513 }
2514
2515 /**
2516  * ice_vsi_vlan_setup - Setup vlan offload properties on a VSI
2517  * @vsi: VSI to setup vlan properties for
2518  */
2519 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
2520 {
2521         int ret = 0;
2522
2523         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
2524                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2525         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
2526                 ret = ice_vsi_manage_vlan_insertion(vsi);
2527
2528         return ret;
2529 }
2530
2531 /**
2532  * ice_vsi_cfg - Setup the VSI
2533  * @vsi: the VSI being configured
2534  *
2535  * Return 0 on success and negative value on error
2536  */
2537 static int ice_vsi_cfg(struct ice_vsi *vsi)
2538 {
2539         int err;
2540
2541         if (vsi->netdev) {
2542                 ice_set_rx_mode(vsi->netdev);
2543
2544                 err = ice_vsi_vlan_setup(vsi);
2545
2546                 if (err)
2547                         return err;
2548         }
2549         err = ice_vsi_cfg_txqs(vsi);
2550         if (!err)
2551                 err = ice_vsi_cfg_rxqs(vsi);
2552
2553         return err;
2554 }
2555
2556 /**
2557  * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
2558  * @vsi: the VSI being configured
2559  */
2560 static void ice_napi_enable_all(struct ice_vsi *vsi)
2561 {
2562         int q_idx;
2563
2564         if (!vsi->netdev)
2565                 return;
2566
2567         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
2568                 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
2569
2570                 if (q_vector->rx.ring || q_vector->tx.ring)
2571                         napi_enable(&q_vector->napi);
2572         }
2573 }
2574
2575 /**
2576  * ice_up_complete - Finish the last steps of bringing up a connection
2577  * @vsi: The VSI being configured
2578  *
2579  * Return 0 on success and negative value on error
2580  */
2581 static int ice_up_complete(struct ice_vsi *vsi)
2582 {
2583         struct ice_pf *pf = vsi->back;
2584         int err;
2585
2586         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
2587                 ice_vsi_cfg_msix(vsi);
2588         else
2589                 return -ENOTSUPP;
2590
2591         /* Enable only Rx rings, Tx rings were enabled by the FW when the
2592          * Tx queue group list was configured and the context bits were
2593          * programmed using ice_vsi_cfg_txqs
2594          */
2595         err = ice_vsi_start_rx_rings(vsi);
2596         if (err)
2597                 return err;
2598
2599         clear_bit(__ICE_DOWN, vsi->state);
2600         ice_napi_enable_all(vsi);
2601         ice_vsi_ena_irq(vsi);
2602
2603         if (vsi->port_info &&
2604             (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
2605             vsi->netdev) {
2606                 ice_print_link_msg(vsi, true);
2607                 netif_tx_start_all_queues(vsi->netdev);
2608                 netif_carrier_on(vsi->netdev);
2609         }
2610
2611         ice_service_task_schedule(pf);
2612
2613         return err;
2614 }
2615
2616 /**
2617  * ice_up - Bring the connection back up after being down
2618  * @vsi: VSI being configured
2619  */
2620 int ice_up(struct ice_vsi *vsi)
2621 {
2622         int err;
2623
2624         err = ice_vsi_cfg(vsi);
2625         if (!err)
2626                 err = ice_up_complete(vsi);
2627
2628         return err;
2629 }
2630
2631 /**
2632  * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
2633  * @ring: Tx or Rx ring to read stats from
2634  * @pkts: packets stats counter
2635  * @bytes: bytes stats counter
2636  *
2637  * This function fetches stats from the ring considering the atomic operations
2638  * that needs to be performed to read u64 values in 32 bit machine.
2639  */
2640 static void ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts,
2641                                          u64 *bytes)
2642 {
2643         unsigned int start;
2644         *pkts = 0;
2645         *bytes = 0;
2646
2647         if (!ring)
2648                 return;
2649         do {
2650                 start = u64_stats_fetch_begin_irq(&ring->syncp);
2651                 *pkts = ring->stats.pkts;
2652                 *bytes = ring->stats.bytes;
2653         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
2654 }
2655
2656 /**
2657  * ice_update_vsi_ring_stats - Update VSI stats counters
2658  * @vsi: the VSI to be updated
2659  */
2660 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
2661 {
2662         struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
2663         struct ice_ring *ring;
2664         u64 pkts, bytes;
2665         int i;
2666
2667         /* reset netdev stats */
2668         vsi_stats->tx_packets = 0;
2669         vsi_stats->tx_bytes = 0;
2670         vsi_stats->rx_packets = 0;
2671         vsi_stats->rx_bytes = 0;
2672
2673         /* reset non-netdev (extended) stats */
2674         vsi->tx_restart = 0;
2675         vsi->tx_busy = 0;
2676         vsi->tx_linearize = 0;
2677         vsi->rx_buf_failed = 0;
2678         vsi->rx_page_failed = 0;
2679
2680         rcu_read_lock();
2681
2682         /* update Tx rings counters */
2683         ice_for_each_txq(vsi, i) {
2684                 ring = READ_ONCE(vsi->tx_rings[i]);
2685                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
2686                 vsi_stats->tx_packets += pkts;
2687                 vsi_stats->tx_bytes += bytes;
2688                 vsi->tx_restart += ring->tx_stats.restart_q;
2689                 vsi->tx_busy += ring->tx_stats.tx_busy;
2690                 vsi->tx_linearize += ring->tx_stats.tx_linearize;
2691         }
2692
2693         /* update Rx rings counters */
2694         ice_for_each_rxq(vsi, i) {
2695                 ring = READ_ONCE(vsi->rx_rings[i]);
2696                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
2697                 vsi_stats->rx_packets += pkts;
2698                 vsi_stats->rx_bytes += bytes;
2699                 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
2700                 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
2701         }
2702
2703         rcu_read_unlock();
2704 }
2705
2706 /**
2707  * ice_update_vsi_stats - Update VSI stats counters
2708  * @vsi: the VSI to be updated
2709  */
2710 static void ice_update_vsi_stats(struct ice_vsi *vsi)
2711 {
2712         struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
2713         struct ice_eth_stats *cur_es = &vsi->eth_stats;
2714         struct ice_pf *pf = vsi->back;
2715
2716         if (test_bit(__ICE_DOWN, vsi->state) ||
2717             test_bit(__ICE_CFG_BUSY, pf->state))
2718                 return;
2719
2720         /* get stats as recorded by Tx/Rx rings */
2721         ice_update_vsi_ring_stats(vsi);
2722
2723         /* get VSI stats as recorded by the hardware */
2724         ice_update_eth_stats(vsi);
2725
2726         cur_ns->tx_errors = cur_es->tx_errors;
2727         cur_ns->rx_dropped = cur_es->rx_discards;
2728         cur_ns->tx_dropped = cur_es->tx_discards;
2729         cur_ns->multicast = cur_es->rx_multicast;
2730
2731         /* update some more netdev stats if this is main VSI */
2732         if (vsi->type == ICE_VSI_PF) {
2733                 cur_ns->rx_crc_errors = pf->stats.crc_errors;
2734                 cur_ns->rx_errors = pf->stats.crc_errors +
2735                                     pf->stats.illegal_bytes;
2736                 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
2737         }
2738 }
2739
2740 /**
2741  * ice_update_pf_stats - Update PF port stats counters
2742  * @pf: PF whose stats needs to be updated
2743  */
2744 static void ice_update_pf_stats(struct ice_pf *pf)
2745 {
2746         struct ice_hw_port_stats *prev_ps, *cur_ps;
2747         struct ice_hw *hw = &pf->hw;
2748         u8 pf_id;
2749
2750         prev_ps = &pf->stats_prev;
2751         cur_ps = &pf->stats;
2752         pf_id = hw->pf_id;
2753
2754         ice_stat_update40(hw, GLPRT_GORCH(pf_id), GLPRT_GORCL(pf_id),
2755                           pf->stat_prev_loaded, &prev_ps->eth.rx_bytes,
2756                           &cur_ps->eth.rx_bytes);
2757
2758         ice_stat_update40(hw, GLPRT_UPRCH(pf_id), GLPRT_UPRCL(pf_id),
2759                           pf->stat_prev_loaded, &prev_ps->eth.rx_unicast,
2760                           &cur_ps->eth.rx_unicast);
2761
2762         ice_stat_update40(hw, GLPRT_MPRCH(pf_id), GLPRT_MPRCL(pf_id),
2763                           pf->stat_prev_loaded, &prev_ps->eth.rx_multicast,
2764                           &cur_ps->eth.rx_multicast);
2765
2766         ice_stat_update40(hw, GLPRT_BPRCH(pf_id), GLPRT_BPRCL(pf_id),
2767                           pf->stat_prev_loaded, &prev_ps->eth.rx_broadcast,
2768                           &cur_ps->eth.rx_broadcast);
2769
2770         ice_stat_update40(hw, GLPRT_GOTCH(pf_id), GLPRT_GOTCL(pf_id),
2771                           pf->stat_prev_loaded, &prev_ps->eth.tx_bytes,
2772                           &cur_ps->eth.tx_bytes);
2773
2774         ice_stat_update40(hw, GLPRT_UPTCH(pf_id), GLPRT_UPTCL(pf_id),
2775                           pf->stat_prev_loaded, &prev_ps->eth.tx_unicast,
2776                           &cur_ps->eth.tx_unicast);
2777
2778         ice_stat_update40(hw, GLPRT_MPTCH(pf_id), GLPRT_MPTCL(pf_id),
2779                           pf->stat_prev_loaded, &prev_ps->eth.tx_multicast,
2780                           &cur_ps->eth.tx_multicast);
2781
2782         ice_stat_update40(hw, GLPRT_BPTCH(pf_id), GLPRT_BPTCL(pf_id),
2783                           pf->stat_prev_loaded, &prev_ps->eth.tx_broadcast,
2784                           &cur_ps->eth.tx_broadcast);
2785
2786         ice_stat_update32(hw, GLPRT_TDOLD(pf_id), pf->stat_prev_loaded,
2787                           &prev_ps->tx_dropped_link_down,
2788                           &cur_ps->tx_dropped_link_down);
2789
2790         ice_stat_update40(hw, GLPRT_PRC64H(pf_id), GLPRT_PRC64L(pf_id),
2791                           pf->stat_prev_loaded, &prev_ps->rx_size_64,
2792                           &cur_ps->rx_size_64);
2793
2794         ice_stat_update40(hw, GLPRT_PRC127H(pf_id), GLPRT_PRC127L(pf_id),
2795                           pf->stat_prev_loaded, &prev_ps->rx_size_127,
2796                           &cur_ps->rx_size_127);
2797
2798         ice_stat_update40(hw, GLPRT_PRC255H(pf_id), GLPRT_PRC255L(pf_id),
2799                           pf->stat_prev_loaded, &prev_ps->rx_size_255,
2800                           &cur_ps->rx_size_255);
2801
2802         ice_stat_update40(hw, GLPRT_PRC511H(pf_id), GLPRT_PRC511L(pf_id),
2803                           pf->stat_prev_loaded, &prev_ps->rx_size_511,
2804                           &cur_ps->rx_size_511);
2805
2806         ice_stat_update40(hw, GLPRT_PRC1023H(pf_id),
2807                           GLPRT_PRC1023L(pf_id), pf->stat_prev_loaded,
2808                           &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
2809
2810         ice_stat_update40(hw, GLPRT_PRC1522H(pf_id),
2811                           GLPRT_PRC1522L(pf_id), pf->stat_prev_loaded,
2812                           &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
2813
2814         ice_stat_update40(hw, GLPRT_PRC9522H(pf_id),
2815                           GLPRT_PRC9522L(pf_id), pf->stat_prev_loaded,
2816                           &prev_ps->rx_size_big, &cur_ps->rx_size_big);
2817
2818         ice_stat_update40(hw, GLPRT_PTC64H(pf_id), GLPRT_PTC64L(pf_id),
2819                           pf->stat_prev_loaded, &prev_ps->tx_size_64,
2820                           &cur_ps->tx_size_64);
2821
2822         ice_stat_update40(hw, GLPRT_PTC127H(pf_id), GLPRT_PTC127L(pf_id),
2823                           pf->stat_prev_loaded, &prev_ps->tx_size_127,
2824                           &cur_ps->tx_size_127);
2825
2826         ice_stat_update40(hw, GLPRT_PTC255H(pf_id), GLPRT_PTC255L(pf_id),
2827                           pf->stat_prev_loaded, &prev_ps->tx_size_255,
2828                           &cur_ps->tx_size_255);
2829
2830         ice_stat_update40(hw, GLPRT_PTC511H(pf_id), GLPRT_PTC511L(pf_id),
2831                           pf->stat_prev_loaded, &prev_ps->tx_size_511,
2832                           &cur_ps->tx_size_511);
2833
2834         ice_stat_update40(hw, GLPRT_PTC1023H(pf_id),
2835                           GLPRT_PTC1023L(pf_id), pf->stat_prev_loaded,
2836                           &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
2837
2838         ice_stat_update40(hw, GLPRT_PTC1522H(pf_id),
2839                           GLPRT_PTC1522L(pf_id), pf->stat_prev_loaded,
2840                           &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
2841
2842         ice_stat_update40(hw, GLPRT_PTC9522H(pf_id),
2843                           GLPRT_PTC9522L(pf_id), pf->stat_prev_loaded,
2844                           &prev_ps->tx_size_big, &cur_ps->tx_size_big);
2845
2846         ice_stat_update32(hw, GLPRT_LXONRXC(pf_id), pf->stat_prev_loaded,
2847                           &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
2848
2849         ice_stat_update32(hw, GLPRT_LXOFFRXC(pf_id), pf->stat_prev_loaded,
2850                           &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
2851
2852         ice_stat_update32(hw, GLPRT_LXONTXC(pf_id), pf->stat_prev_loaded,
2853                           &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
2854
2855         ice_stat_update32(hw, GLPRT_LXOFFTXC(pf_id), pf->stat_prev_loaded,
2856                           &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
2857
2858         ice_stat_update32(hw, GLPRT_CRCERRS(pf_id), pf->stat_prev_loaded,
2859                           &prev_ps->crc_errors, &cur_ps->crc_errors);
2860
2861         ice_stat_update32(hw, GLPRT_ILLERRC(pf_id), pf->stat_prev_loaded,
2862                           &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
2863
2864         ice_stat_update32(hw, GLPRT_MLFC(pf_id), pf->stat_prev_loaded,
2865                           &prev_ps->mac_local_faults,
2866                           &cur_ps->mac_local_faults);
2867
2868         ice_stat_update32(hw, GLPRT_MRFC(pf_id), pf->stat_prev_loaded,
2869                           &prev_ps->mac_remote_faults,
2870                           &cur_ps->mac_remote_faults);
2871
2872         ice_stat_update32(hw, GLPRT_RLEC(pf_id), pf->stat_prev_loaded,
2873                           &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
2874
2875         ice_stat_update32(hw, GLPRT_RUC(pf_id), pf->stat_prev_loaded,
2876                           &prev_ps->rx_undersize, &cur_ps->rx_undersize);
2877
2878         ice_stat_update32(hw, GLPRT_RFC(pf_id), pf->stat_prev_loaded,
2879                           &prev_ps->rx_fragments, &cur_ps->rx_fragments);
2880
2881         ice_stat_update32(hw, GLPRT_ROC(pf_id), pf->stat_prev_loaded,
2882                           &prev_ps->rx_oversize, &cur_ps->rx_oversize);
2883
2884         ice_stat_update32(hw, GLPRT_RJC(pf_id), pf->stat_prev_loaded,
2885                           &prev_ps->rx_jabber, &cur_ps->rx_jabber);
2886
2887         pf->stat_prev_loaded = true;
2888 }
2889
2890 /**
2891  * ice_get_stats64 - get statistics for network device structure
2892  * @netdev: network interface device structure
2893  * @stats: main device statistics structure
2894  */
2895 static
2896 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
2897 {
2898         struct ice_netdev_priv *np = netdev_priv(netdev);
2899         struct rtnl_link_stats64 *vsi_stats;
2900         struct ice_vsi *vsi = np->vsi;
2901
2902         vsi_stats = &vsi->net_stats;
2903
2904         if (test_bit(__ICE_DOWN, vsi->state) || !vsi->num_txq || !vsi->num_rxq)
2905                 return;
2906         /* netdev packet/byte stats come from ring counter. These are obtained
2907          * by summing up ring counters (done by ice_update_vsi_ring_stats).
2908          */
2909         ice_update_vsi_ring_stats(vsi);
2910         stats->tx_packets = vsi_stats->tx_packets;
2911         stats->tx_bytes = vsi_stats->tx_bytes;
2912         stats->rx_packets = vsi_stats->rx_packets;
2913         stats->rx_bytes = vsi_stats->rx_bytes;
2914
2915         /* The rest of the stats can be read from the hardware but instead we
2916          * just return values that the watchdog task has already obtained from
2917          * the hardware.
2918          */
2919         stats->multicast = vsi_stats->multicast;
2920         stats->tx_errors = vsi_stats->tx_errors;
2921         stats->tx_dropped = vsi_stats->tx_dropped;
2922         stats->rx_errors = vsi_stats->rx_errors;
2923         stats->rx_dropped = vsi_stats->rx_dropped;
2924         stats->rx_crc_errors = vsi_stats->rx_crc_errors;
2925         stats->rx_length_errors = vsi_stats->rx_length_errors;
2926 }
2927
2928 /**
2929  * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
2930  * @vsi: VSI having NAPI disabled
2931  */
2932 static void ice_napi_disable_all(struct ice_vsi *vsi)
2933 {
2934         int q_idx;
2935
2936         if (!vsi->netdev)
2937                 return;
2938
2939         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
2940                 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
2941
2942                 if (q_vector->rx.ring || q_vector->tx.ring)
2943                         napi_disable(&q_vector->napi);
2944         }
2945 }
2946
2947 /**
2948  * ice_down - Shutdown the connection
2949  * @vsi: The VSI being stopped
2950  */
2951 int ice_down(struct ice_vsi *vsi)
2952 {
2953         int i, tx_err, rx_err;
2954
2955         /* Caller of this function is expected to set the
2956          * vsi->state __ICE_DOWN bit
2957          */
2958         if (vsi->netdev) {
2959                 netif_carrier_off(vsi->netdev);
2960                 netif_tx_disable(vsi->netdev);
2961         }
2962
2963         ice_vsi_dis_irq(vsi);
2964         tx_err = ice_vsi_stop_tx_rings(vsi, ICE_NO_RESET, 0);
2965         if (tx_err)
2966                 netdev_err(vsi->netdev,
2967                            "Failed stop Tx rings, VSI %d error %d\n",
2968                            vsi->vsi_num, tx_err);
2969
2970         rx_err = ice_vsi_stop_rx_rings(vsi);
2971         if (rx_err)
2972                 netdev_err(vsi->netdev,
2973                            "Failed stop Rx rings, VSI %d error %d\n",
2974                            vsi->vsi_num, rx_err);
2975
2976         ice_napi_disable_all(vsi);
2977
2978         ice_for_each_txq(vsi, i)
2979                 ice_clean_tx_ring(vsi->tx_rings[i]);
2980
2981         ice_for_each_rxq(vsi, i)
2982                 ice_clean_rx_ring(vsi->rx_rings[i]);
2983
2984         if (tx_err || rx_err) {
2985                 netdev_err(vsi->netdev,
2986                            "Failed to close VSI 0x%04X on switch 0x%04X\n",
2987                            vsi->vsi_num, vsi->vsw->sw_id);
2988                 return -EIO;
2989         }
2990
2991         return 0;
2992 }
2993
2994 /**
2995  * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
2996  * @vsi: VSI having resources allocated
2997  *
2998  * Return 0 on success, negative on failure
2999  */
3000 static int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
3001 {
3002         int i, err = 0;
3003
3004         if (!vsi->num_txq) {
3005                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Tx queues\n",
3006                         vsi->vsi_num);
3007                 return -EINVAL;
3008         }
3009
3010         ice_for_each_txq(vsi, i) {
3011                 vsi->tx_rings[i]->netdev = vsi->netdev;
3012                 err = ice_setup_tx_ring(vsi->tx_rings[i]);
3013                 if (err)
3014                         break;
3015         }
3016
3017         return err;
3018 }
3019
3020 /**
3021  * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
3022  * @vsi: VSI having resources allocated
3023  *
3024  * Return 0 on success, negative on failure
3025  */
3026 static int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
3027 {
3028         int i, err = 0;
3029
3030         if (!vsi->num_rxq) {
3031                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Rx queues\n",
3032                         vsi->vsi_num);
3033                 return -EINVAL;
3034         }
3035
3036         ice_for_each_rxq(vsi, i) {
3037                 vsi->rx_rings[i]->netdev = vsi->netdev;
3038                 err = ice_setup_rx_ring(vsi->rx_rings[i]);
3039                 if (err)
3040                         break;
3041         }
3042
3043         return err;
3044 }
3045
3046 /**
3047  * ice_vsi_req_irq - Request IRQ from the OS
3048  * @vsi: The VSI IRQ is being requested for
3049  * @basename: name for the vector
3050  *
3051  * Return 0 on success and a negative value on error
3052  */
3053 static int ice_vsi_req_irq(struct ice_vsi *vsi, char *basename)
3054 {
3055         struct ice_pf *pf = vsi->back;
3056         int err = -EINVAL;
3057
3058         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3059                 err = ice_vsi_req_irq_msix(vsi, basename);
3060
3061         return err;
3062 }
3063
3064 /**
3065  * ice_vsi_open - Called when a network interface is made active
3066  * @vsi: the VSI to open
3067  *
3068  * Initialization of the VSI
3069  *
3070  * Returns 0 on success, negative value on error
3071  */
3072 static int ice_vsi_open(struct ice_vsi *vsi)
3073 {
3074         char int_name[ICE_INT_NAME_STR_LEN];
3075         struct ice_pf *pf = vsi->back;
3076         int err;
3077
3078         /* allocate descriptors */
3079         err = ice_vsi_setup_tx_rings(vsi);
3080         if (err)
3081                 goto err_setup_tx;
3082
3083         err = ice_vsi_setup_rx_rings(vsi);
3084         if (err)
3085                 goto err_setup_rx;
3086
3087         err = ice_vsi_cfg(vsi);
3088         if (err)
3089                 goto err_setup_rx;
3090
3091         snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3092                  dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
3093         err = ice_vsi_req_irq(vsi, int_name);
3094         if (err)
3095                 goto err_setup_rx;
3096
3097         /* Notify the stack of the actual queue counts. */
3098         err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
3099         if (err)
3100                 goto err_set_qs;
3101
3102         err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
3103         if (err)
3104                 goto err_set_qs;
3105
3106         err = ice_up_complete(vsi);
3107         if (err)
3108                 goto err_up_complete;
3109
3110         return 0;
3111
3112 err_up_complete:
3113         ice_down(vsi);
3114 err_set_qs:
3115         ice_vsi_free_irq(vsi);
3116 err_setup_rx:
3117         ice_vsi_free_rx_rings(vsi);
3118 err_setup_tx:
3119         ice_vsi_free_tx_rings(vsi);
3120
3121         return err;
3122 }
3123
3124 /**
3125  * ice_vsi_release_all - Delete all VSIs
3126  * @pf: PF from which all VSIs are being removed
3127  */
3128 static void ice_vsi_release_all(struct ice_pf *pf)
3129 {
3130         int err, i;
3131
3132         if (!pf->vsi)
3133                 return;
3134
3135         for (i = 0; i < pf->num_alloc_vsi; i++) {
3136                 if (!pf->vsi[i])
3137                         continue;
3138
3139                 err = ice_vsi_release(pf->vsi[i]);
3140                 if (err)
3141                         dev_dbg(&pf->pdev->dev,
3142                                 "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
3143                                 i, err, pf->vsi[i]->vsi_num);
3144         }
3145 }
3146
3147 /**
3148  * ice_dis_vsi - pause a VSI
3149  * @vsi: the VSI being paused
3150  * @locked: is the rtnl_lock already held
3151  */
3152 static void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
3153 {
3154         if (test_bit(__ICE_DOWN, vsi->state))
3155                 return;
3156
3157         set_bit(__ICE_NEEDS_RESTART, vsi->state);
3158
3159         if (vsi->type == ICE_VSI_PF && vsi->netdev) {
3160                 if (netif_running(vsi->netdev)) {
3161                         if (!locked) {
3162                                 rtnl_lock();
3163                                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3164                                 rtnl_unlock();
3165                         } else {
3166                                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3167                         }
3168                 } else {
3169                         ice_vsi_close(vsi);
3170                 }
3171         }
3172 }
3173
3174 /**
3175  * ice_ena_vsi - resume a VSI
3176  * @vsi: the VSI being resume
3177  */
3178 static int ice_ena_vsi(struct ice_vsi *vsi)
3179 {
3180         int err = 0;
3181
3182         if (test_and_clear_bit(__ICE_NEEDS_RESTART, vsi->state) &&
3183             vsi->netdev) {
3184                 if (netif_running(vsi->netdev)) {
3185                         rtnl_lock();
3186                         err = vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3187                         rtnl_unlock();
3188                 } else {
3189                         err = ice_vsi_open(vsi);
3190                 }
3191         }
3192
3193         return err;
3194 }
3195
3196 /**
3197  * ice_pf_dis_all_vsi - Pause all VSIs on a PF
3198  * @pf: the PF
3199  */
3200 static void ice_pf_dis_all_vsi(struct ice_pf *pf)
3201 {
3202         int v;
3203
3204         ice_for_each_vsi(pf, v)
3205                 if (pf->vsi[v])
3206                         ice_dis_vsi(pf->vsi[v], false);
3207 }
3208
3209 /**
3210  * ice_pf_ena_all_vsi - Resume all VSIs on a PF
3211  * @pf: the PF
3212  */
3213 static int ice_pf_ena_all_vsi(struct ice_pf *pf)
3214 {
3215         int v;
3216
3217         ice_for_each_vsi(pf, v)
3218                 if (pf->vsi[v])
3219                         if (ice_ena_vsi(pf->vsi[v]))
3220                                 return -EIO;
3221
3222         return 0;
3223 }
3224
3225 /**
3226  * ice_vsi_rebuild_all - rebuild all VSIs in pf
3227  * @pf: the PF
3228  */
3229 static int ice_vsi_rebuild_all(struct ice_pf *pf)
3230 {
3231         int i;
3232
3233         /* loop through pf->vsi array and reinit the VSI if found */
3234         for (i = 0; i < pf->num_alloc_vsi; i++) {
3235                 int err;
3236
3237                 if (!pf->vsi[i])
3238                         continue;
3239
3240                 /* VF VSI rebuild isn't supported yet */
3241                 if (pf->vsi[i]->type == ICE_VSI_VF)
3242                         continue;
3243
3244                 err = ice_vsi_rebuild(pf->vsi[i]);
3245                 if (err) {
3246                         dev_err(&pf->pdev->dev,
3247                                 "VSI at index %d rebuild failed\n",
3248                                 pf->vsi[i]->idx);
3249                         return err;
3250                 }
3251
3252                 dev_info(&pf->pdev->dev,
3253                          "VSI at index %d rebuilt. vsi_num = 0x%x\n",
3254                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3255         }
3256
3257         return 0;
3258 }
3259
3260 /**
3261  * ice_vsi_replay_all - replay all VSIs configuration in the PF
3262  * @pf: the PF
3263  */
3264 static int ice_vsi_replay_all(struct ice_pf *pf)
3265 {
3266         struct ice_hw *hw = &pf->hw;
3267         enum ice_status ret;
3268         int i;
3269
3270         /* loop through pf->vsi array and replay the VSI if found */
3271         for (i = 0; i < pf->num_alloc_vsi; i++) {
3272                 if (!pf->vsi[i])
3273                         continue;
3274
3275                 ret = ice_replay_vsi(hw, pf->vsi[i]->idx);
3276                 if (ret) {
3277                         dev_err(&pf->pdev->dev,
3278                                 "VSI at index %d replay failed %d\n",
3279                                 pf->vsi[i]->idx, ret);
3280                         return -EIO;
3281                 }
3282
3283                 /* Re-map HW VSI number, using VSI handle that has been
3284                  * previously validated in ice_replay_vsi() call above
3285                  */
3286                 pf->vsi[i]->vsi_num = ice_get_hw_vsi_num(hw, pf->vsi[i]->idx);
3287
3288                 dev_info(&pf->pdev->dev,
3289                          "VSI at index %d filter replayed successfully - vsi_num %i\n",
3290                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3291         }
3292
3293         /* Clean up replay filter after successful re-configuration */
3294         ice_replay_post(hw);
3295         return 0;
3296 }
3297
3298 /**
3299  * ice_rebuild - rebuild after reset
3300  * @pf: pf to rebuild
3301  */
3302 static void ice_rebuild(struct ice_pf *pf)
3303 {
3304         struct device *dev = &pf->pdev->dev;
3305         struct ice_hw *hw = &pf->hw;
3306         enum ice_status ret;
3307         int err, i;
3308
3309         if (test_bit(__ICE_DOWN, pf->state))
3310                 goto clear_recovery;
3311
3312         dev_dbg(dev, "rebuilding pf\n");
3313
3314         ret = ice_init_all_ctrlq(hw);
3315         if (ret) {
3316                 dev_err(dev, "control queues init failed %d\n", ret);
3317                 goto err_init_ctrlq;
3318         }
3319
3320         ret = ice_clear_pf_cfg(hw);
3321         if (ret) {
3322                 dev_err(dev, "clear PF configuration failed %d\n", ret);
3323                 goto err_init_ctrlq;
3324         }
3325
3326         ice_clear_pxe_mode(hw);
3327
3328         ret = ice_get_caps(hw);
3329         if (ret) {
3330                 dev_err(dev, "ice_get_caps failed %d\n", ret);
3331                 goto err_init_ctrlq;
3332         }
3333
3334         err = ice_sched_init_port(hw->port_info);
3335         if (err)
3336                 goto err_sched_init_port;
3337
3338         /* reset search_hint of irq_trackers to 0 since interrupts are
3339          * reclaimed and could be allocated from beginning during VSI rebuild
3340          */
3341         pf->sw_irq_tracker->search_hint = 0;
3342         pf->hw_irq_tracker->search_hint = 0;
3343
3344         err = ice_vsi_rebuild_all(pf);
3345         if (err) {
3346                 dev_err(dev, "ice_vsi_rebuild_all failed\n");
3347                 goto err_vsi_rebuild;
3348         }
3349
3350         err = ice_update_link_info(hw->port_info);
3351         if (err)
3352                 dev_err(&pf->pdev->dev, "Get link status error %d\n", err);
3353
3354         /* Replay all VSIs Configuration, including filters after reset */
3355         if (ice_vsi_replay_all(pf)) {
3356                 dev_err(&pf->pdev->dev,
3357                         "error replaying VSI configurations with switch filter rules\n");
3358                 goto err_vsi_rebuild;
3359         }
3360
3361         /* start misc vector */
3362         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
3363                 err = ice_req_irq_msix_misc(pf);
3364                 if (err) {
3365                         dev_err(dev, "misc vector setup failed: %d\n", err);
3366                         goto err_vsi_rebuild;
3367                 }
3368         }
3369
3370         /* restart the VSIs that were rebuilt and running before the reset */
3371         err = ice_pf_ena_all_vsi(pf);
3372         if (err) {
3373                 dev_err(&pf->pdev->dev, "error enabling VSIs\n");
3374                 /* no need to disable VSIs in tear down path in ice_rebuild()
3375                  * since its already taken care in ice_vsi_open()
3376                  */
3377                 goto err_vsi_rebuild;
3378         }
3379
3380         ice_reset_all_vfs(pf, true);
3381
3382         for (i = 0; i < pf->num_alloc_vsi; i++) {
3383                 bool link_up;
3384
3385                 if (!pf->vsi[i] || pf->vsi[i]->type != ICE_VSI_PF)
3386                         continue;
3387                 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
3388                 if (link_up) {
3389                         netif_carrier_on(pf->vsi[i]->netdev);
3390                         netif_tx_wake_all_queues(pf->vsi[i]->netdev);
3391                 } else {
3392                         netif_carrier_off(pf->vsi[i]->netdev);
3393                         netif_tx_stop_all_queues(pf->vsi[i]->netdev);
3394                 }
3395         }
3396
3397         /* if we get here, reset flow is successful */
3398         clear_bit(__ICE_RESET_FAILED, pf->state);
3399         return;
3400
3401 err_vsi_rebuild:
3402         ice_vsi_release_all(pf);
3403 err_sched_init_port:
3404         ice_sched_cleanup_all(hw);
3405 err_init_ctrlq:
3406         ice_shutdown_all_ctrlq(hw);
3407         set_bit(__ICE_RESET_FAILED, pf->state);
3408 clear_recovery:
3409         /* set this bit in PF state to control service task scheduling */
3410         set_bit(__ICE_NEEDS_RESTART, pf->state);
3411         dev_err(dev, "Rebuild failed, unload and reload driver\n");
3412 }
3413
3414 /**
3415  * ice_change_mtu - NDO callback to change the MTU
3416  * @netdev: network interface device structure
3417  * @new_mtu: new value for maximum frame size
3418  *
3419  * Returns 0 on success, negative on failure
3420  */
3421 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
3422 {
3423         struct ice_netdev_priv *np = netdev_priv(netdev);
3424         struct ice_vsi *vsi = np->vsi;
3425         struct ice_pf *pf = vsi->back;
3426         u8 count = 0;
3427
3428         if (new_mtu == netdev->mtu) {
3429                 netdev_warn(netdev, "mtu is already %u\n", netdev->mtu);
3430                 return 0;
3431         }
3432
3433         if (new_mtu < netdev->min_mtu) {
3434                 netdev_err(netdev, "new mtu invalid. min_mtu is %d\n",
3435                            netdev->min_mtu);
3436                 return -EINVAL;
3437         } else if (new_mtu > netdev->max_mtu) {
3438                 netdev_err(netdev, "new mtu invalid. max_mtu is %d\n",
3439                            netdev->min_mtu);
3440                 return -EINVAL;
3441         }
3442         /* if a reset is in progress, wait for some time for it to complete */
3443         do {
3444                 if (ice_is_reset_in_progress(pf->state)) {
3445                         count++;
3446                         usleep_range(1000, 2000);
3447                 } else {
3448                         break;
3449                 }
3450
3451         } while (count < 100);
3452
3453         if (count == 100) {
3454                 netdev_err(netdev, "can't change mtu. Device is busy\n");
3455                 return -EBUSY;
3456         }
3457
3458         netdev->mtu = new_mtu;
3459
3460         /* if VSI is up, bring it down and then back up */
3461         if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
3462                 int err;
3463
3464                 err = ice_down(vsi);
3465                 if (err) {
3466                         netdev_err(netdev, "change mtu if_up err %d\n", err);
3467                         return err;
3468                 }
3469
3470                 err = ice_up(vsi);
3471                 if (err) {
3472                         netdev_err(netdev, "change mtu if_up err %d\n", err);
3473                         return err;
3474                 }
3475         }
3476
3477         netdev_dbg(netdev, "changed mtu to %d\n", new_mtu);
3478         return 0;
3479 }
3480
3481 /**
3482  * ice_set_rss - Set RSS keys and lut
3483  * @vsi: Pointer to VSI structure
3484  * @seed: RSS hash seed
3485  * @lut: Lookup table
3486  * @lut_size: Lookup table size
3487  *
3488  * Returns 0 on success, negative on failure
3489  */
3490 int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
3491 {
3492         struct ice_pf *pf = vsi->back;
3493         struct ice_hw *hw = &pf->hw;
3494         enum ice_status status;
3495
3496         if (seed) {
3497                 struct ice_aqc_get_set_rss_keys *buf =
3498                                   (struct ice_aqc_get_set_rss_keys *)seed;
3499
3500                 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
3501
3502                 if (status) {
3503                         dev_err(&pf->pdev->dev,
3504                                 "Cannot set RSS key, err %d aq_err %d\n",
3505                                 status, hw->adminq.rq_last_status);
3506                         return -EIO;
3507                 }
3508         }
3509
3510         if (lut) {
3511                 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
3512                                             lut, lut_size);
3513                 if (status) {
3514                         dev_err(&pf->pdev->dev,
3515                                 "Cannot set RSS lut, err %d aq_err %d\n",
3516                                 status, hw->adminq.rq_last_status);
3517                         return -EIO;
3518                 }
3519         }
3520
3521         return 0;
3522 }
3523
3524 /**
3525  * ice_get_rss - Get RSS keys and lut
3526  * @vsi: Pointer to VSI structure
3527  * @seed: Buffer to store the keys
3528  * @lut: Buffer to store the lookup table entries
3529  * @lut_size: Size of buffer to store the lookup table entries
3530  *
3531  * Returns 0 on success, negative on failure
3532  */
3533 int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
3534 {
3535         struct ice_pf *pf = vsi->back;
3536         struct ice_hw *hw = &pf->hw;
3537         enum ice_status status;
3538
3539         if (seed) {
3540                 struct ice_aqc_get_set_rss_keys *buf =
3541                                   (struct ice_aqc_get_set_rss_keys *)seed;
3542
3543                 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
3544                 if (status) {
3545                         dev_err(&pf->pdev->dev,
3546                                 "Cannot get RSS key, err %d aq_err %d\n",
3547                                 status, hw->adminq.rq_last_status);
3548                         return -EIO;
3549                 }
3550         }
3551
3552         if (lut) {
3553                 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
3554                                             lut, lut_size);
3555                 if (status) {
3556                         dev_err(&pf->pdev->dev,
3557                                 "Cannot get RSS lut, err %d aq_err %d\n",
3558                                 status, hw->adminq.rq_last_status);
3559                         return -EIO;
3560                 }
3561         }
3562
3563         return 0;
3564 }
3565
3566 /**
3567  * ice_bridge_getlink - Get the hardware bridge mode
3568  * @skb: skb buff
3569  * @pid: process id
3570  * @seq: RTNL message seq
3571  * @dev: the netdev being configured
3572  * @filter_mask: filter mask passed in
3573  * @nlflags: netlink flags passed in
3574  *
3575  * Return the bridge mode (VEB/VEPA)
3576  */
3577 static int
3578 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
3579                    struct net_device *dev, u32 filter_mask, int nlflags)
3580 {
3581         struct ice_netdev_priv *np = netdev_priv(dev);
3582         struct ice_vsi *vsi = np->vsi;
3583         struct ice_pf *pf = vsi->back;
3584         u16 bmode;
3585
3586         bmode = pf->first_sw->bridge_mode;
3587
3588         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
3589                                        filter_mask, NULL);
3590 }
3591
3592 /**
3593  * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
3594  * @vsi: Pointer to VSI structure
3595  * @bmode: Hardware bridge mode (VEB/VEPA)
3596  *
3597  * Returns 0 on success, negative on failure
3598  */
3599 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
3600 {
3601         struct device *dev = &vsi->back->pdev->dev;
3602         struct ice_aqc_vsi_props *vsi_props;
3603         struct ice_hw *hw = &vsi->back->hw;
3604         struct ice_vsi_ctx ctxt = { 0 };
3605         enum ice_status status;
3606
3607         vsi_props = &vsi->info;
3608         ctxt.info = vsi->info;
3609
3610         if (bmode == BRIDGE_MODE_VEB)
3611                 /* change from VEPA to VEB mode */
3612                 ctxt.info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
3613         else
3614                 /* change from VEB to VEPA mode */
3615                 ctxt.info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
3616         ctxt.info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
3617
3618         status = ice_update_vsi(hw, vsi->idx, &ctxt, NULL);
3619         if (status) {
3620                 dev_err(dev, "update VSI for bridge mode failed, bmode = %d err %d aq_err %d\n",
3621                         bmode, status, hw->adminq.sq_last_status);
3622                 return -EIO;
3623         }
3624         /* Update sw flags for book keeping */
3625         vsi_props->sw_flags = ctxt.info.sw_flags;
3626
3627         return 0;
3628 }
3629
3630 /**
3631  * ice_bridge_setlink - Set the hardware bridge mode
3632  * @dev: the netdev being configured
3633  * @nlh: RTNL message
3634  * @flags: bridge setlink flags
3635  * @extack: netlink extended ack
3636  *
3637  * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
3638  * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
3639  * not already set for all VSIs connected to this switch. And also update the
3640  * unicast switch filter rules for the corresponding switch of the netdev.
3641  */
3642 static int
3643 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
3644                    u16 __always_unused flags, struct netlink_ext_ack *extack)
3645 {
3646         struct ice_netdev_priv *np = netdev_priv(dev);
3647         struct ice_pf *pf = np->vsi->back;
3648         struct nlattr *attr, *br_spec;
3649         struct ice_hw *hw = &pf->hw;
3650         enum ice_status status;
3651         struct ice_sw *pf_sw;
3652         int rem, v, err = 0;
3653
3654         pf_sw = pf->first_sw;
3655         /* find the attribute in the netlink message */
3656         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
3657
3658         nla_for_each_nested(attr, br_spec, rem) {
3659                 __u16 mode;
3660
3661                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
3662                         continue;
3663                 mode = nla_get_u16(attr);
3664                 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
3665                         return -EINVAL;
3666                 /* Continue  if bridge mode is not being flipped */
3667                 if (mode == pf_sw->bridge_mode)
3668                         continue;
3669                 /* Iterates through the PF VSI list and update the loopback
3670                  * mode of the VSI
3671                  */
3672                 ice_for_each_vsi(pf, v) {
3673                         if (!pf->vsi[v])
3674                                 continue;
3675                         err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
3676                         if (err)
3677                                 return err;
3678                 }
3679
3680                 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
3681                 /* Update the unicast switch filter rules for the corresponding
3682                  * switch of the netdev
3683                  */
3684                 status = ice_update_sw_rule_bridge_mode(hw);
3685                 if (status) {
3686                         netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %d\n",
3687                                    mode, status, hw->adminq.sq_last_status);
3688                         /* revert hw->evb_veb */
3689                         hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
3690                         return -EIO;
3691                 }
3692
3693                 pf_sw->bridge_mode = mode;
3694         }
3695
3696         return 0;
3697 }
3698
3699 /**
3700  * ice_tx_timeout - Respond to a Tx Hang
3701  * @netdev: network interface device structure
3702  */
3703 static void ice_tx_timeout(struct net_device *netdev)
3704 {
3705         struct ice_netdev_priv *np = netdev_priv(netdev);
3706         struct ice_ring *tx_ring = NULL;
3707         struct ice_vsi *vsi = np->vsi;
3708         struct ice_pf *pf = vsi->back;
3709         int hung_queue = -1;
3710         u32 i;
3711
3712         pf->tx_timeout_count++;
3713
3714         /* find the stopped queue the same way dev_watchdog() does */
3715         for (i = 0; i < netdev->num_tx_queues; i++) {
3716                 unsigned long trans_start;
3717                 struct netdev_queue *q;
3718
3719                 q = netdev_get_tx_queue(netdev, i);
3720                 trans_start = q->trans_start;
3721                 if (netif_xmit_stopped(q) &&
3722                     time_after(jiffies,
3723                                trans_start + netdev->watchdog_timeo)) {
3724                         hung_queue = i;
3725                         break;
3726                 }
3727         }
3728
3729         if (i == netdev->num_tx_queues)
3730                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
3731         else
3732                 /* now that we have an index, find the tx_ring struct */
3733                 for (i = 0; i < vsi->num_txq; i++)
3734                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
3735                                 if (hung_queue == vsi->tx_rings[i]->q_index) {
3736                                         tx_ring = vsi->tx_rings[i];
3737                                         break;
3738                                 }
3739
3740         /* Reset recovery level if enough time has elapsed after last timeout.
3741          * Also ensure no new reset action happens before next timeout period.
3742          */
3743         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
3744                 pf->tx_timeout_recovery_level = 1;
3745         else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
3746                                        netdev->watchdog_timeo)))
3747                 return;
3748
3749         if (tx_ring) {
3750                 struct ice_hw *hw = &pf->hw;
3751                 u32 head, val = 0;
3752
3753                 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[hung_queue])) &
3754                         QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
3755                 /* Read interrupt register */
3756                 if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3757                         val = rd32(hw,
3758                                    GLINT_DYN_CTL(tx_ring->q_vector->v_idx +
3759                                                  tx_ring->vsi->hw_base_vector));
3760
3761                 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %d, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
3762                             vsi->vsi_num, hung_queue, tx_ring->next_to_clean,
3763                             head, tx_ring->next_to_use, val);
3764         }
3765
3766         pf->tx_timeout_last_recovery = jiffies;
3767         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
3768                     pf->tx_timeout_recovery_level, hung_queue);
3769
3770         switch (pf->tx_timeout_recovery_level) {
3771         case 1:
3772                 set_bit(__ICE_PFR_REQ, pf->state);
3773                 break;
3774         case 2:
3775                 set_bit(__ICE_CORER_REQ, pf->state);
3776                 break;
3777         case 3:
3778                 set_bit(__ICE_GLOBR_REQ, pf->state);
3779                 break;
3780         default:
3781                 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
3782                 set_bit(__ICE_DOWN, pf->state);
3783                 set_bit(__ICE_NEEDS_RESTART, vsi->state);
3784                 set_bit(__ICE_SERVICE_DIS, pf->state);
3785                 break;
3786         }
3787
3788         ice_service_task_schedule(pf);
3789         pf->tx_timeout_recovery_level++;
3790 }
3791
3792 /**
3793  * ice_open - Called when a network interface becomes active
3794  * @netdev: network interface device structure
3795  *
3796  * The open entry point is called when a network interface is made
3797  * active by the system (IFF_UP). At this point all resources needed
3798  * for transmit and receive operations are allocated, the interrupt
3799  * handler is registered with the OS, the netdev watchdog is enabled,
3800  * and the stack is notified that the interface is ready.
3801  *
3802  * Returns 0 on success, negative value on failure
3803  */
3804 static int ice_open(struct net_device *netdev)
3805 {
3806         struct ice_netdev_priv *np = netdev_priv(netdev);
3807         struct ice_vsi *vsi = np->vsi;
3808         int err;
3809
3810         if (test_bit(__ICE_NEEDS_RESTART, vsi->back->state)) {
3811                 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
3812                 return -EIO;
3813         }
3814
3815         netif_carrier_off(netdev);
3816
3817         err = ice_vsi_open(vsi);
3818
3819         if (err)
3820                 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
3821                            vsi->vsi_num, vsi->vsw->sw_id);
3822         return err;
3823 }
3824
3825 /**
3826  * ice_stop - Disables a network interface
3827  * @netdev: network interface device structure
3828  *
3829  * The stop entry point is called when an interface is de-activated by the OS,
3830  * and the netdevice enters the DOWN state. The hardware is still under the
3831  * driver's control, but the netdev interface is disabled.
3832  *
3833  * Returns success only - not allowed to fail
3834  */
3835 static int ice_stop(struct net_device *netdev)
3836 {
3837         struct ice_netdev_priv *np = netdev_priv(netdev);
3838         struct ice_vsi *vsi = np->vsi;
3839
3840         ice_vsi_close(vsi);
3841
3842         return 0;
3843 }
3844
3845 /**
3846  * ice_features_check - Validate encapsulated packet conforms to limits
3847  * @skb: skb buffer
3848  * @netdev: This port's netdev
3849  * @features: Offload features that the stack believes apply
3850  */
3851 static netdev_features_t
3852 ice_features_check(struct sk_buff *skb,
3853                    struct net_device __always_unused *netdev,
3854                    netdev_features_t features)
3855 {
3856         size_t len;
3857
3858         /* No point in doing any of this if neither checksum nor GSO are
3859          * being requested for this frame. We can rule out both by just
3860          * checking for CHECKSUM_PARTIAL
3861          */
3862         if (skb->ip_summed != CHECKSUM_PARTIAL)
3863                 return features;
3864
3865         /* We cannot support GSO if the MSS is going to be less than
3866          * 64 bytes. If it is then we need to drop support for GSO.
3867          */
3868         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
3869                 features &= ~NETIF_F_GSO_MASK;
3870
3871         len = skb_network_header(skb) - skb->data;
3872         if (len & ~(ICE_TXD_MACLEN_MAX))
3873                 goto out_rm_features;
3874
3875         len = skb_transport_header(skb) - skb_network_header(skb);
3876         if (len & ~(ICE_TXD_IPLEN_MAX))
3877                 goto out_rm_features;
3878
3879         if (skb->encapsulation) {
3880                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
3881                 if (len & ~(ICE_TXD_L4LEN_MAX))
3882                         goto out_rm_features;
3883
3884                 len = skb_inner_transport_header(skb) -
3885                       skb_inner_network_header(skb);
3886                 if (len & ~(ICE_TXD_IPLEN_MAX))
3887                         goto out_rm_features;
3888         }
3889
3890         return features;
3891 out_rm_features:
3892         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
3893 }
3894
3895 static const struct net_device_ops ice_netdev_ops = {
3896         .ndo_open = ice_open,
3897         .ndo_stop = ice_stop,
3898         .ndo_start_xmit = ice_start_xmit,
3899         .ndo_features_check = ice_features_check,
3900         .ndo_set_rx_mode = ice_set_rx_mode,
3901         .ndo_set_mac_address = ice_set_mac_address,
3902         .ndo_validate_addr = eth_validate_addr,
3903         .ndo_change_mtu = ice_change_mtu,
3904         .ndo_get_stats64 = ice_get_stats64,
3905         .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
3906         .ndo_set_vf_mac = ice_set_vf_mac,
3907         .ndo_get_vf_config = ice_get_vf_cfg,
3908         .ndo_set_vf_trust = ice_set_vf_trust,
3909         .ndo_set_vf_vlan = ice_set_vf_port_vlan,
3910         .ndo_set_vf_link_state = ice_set_vf_link_state,
3911         .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
3912         .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
3913         .ndo_set_features = ice_set_features,
3914         .ndo_bridge_getlink = ice_bridge_getlink,
3915         .ndo_bridge_setlink = ice_bridge_setlink,
3916         .ndo_fdb_add = ice_fdb_add,
3917         .ndo_fdb_del = ice_fdb_del,
3918         .ndo_tx_timeout = ice_tx_timeout,
3919 };