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 #include "ice_dcb_lib.h"
  11
  12 #define DRV_VERSION     "0.7.4-k"
  13 #define DRV_SUMMARY     "Intel(R) Ethernet Connection E800 Series Linux Driver"
  14 const char ice_drv_ver[] = DRV_VERSION;
  15 static const char ice_driver_string[] = DRV_SUMMARY;
  16 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
  17
  18 MODULE_AUTHOR("Intel Corporation, <[email protected]>");
  19 MODULE_DESCRIPTION(DRV_SUMMARY);
  20 MODULE_LICENSE("GPL v2");
  21 MODULE_VERSION(DRV_VERSION);
  22
  23 static int debug = -1;
  24 module_param(debug, int, 0644);
  25 #ifndef CONFIG_DYNAMIC_DEBUG
  26 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
  27 #else
  28 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
  29 #endif /* !CONFIG_DYNAMIC_DEBUG */
  30
  31 static struct workqueue_struct *ice_wq;
  32 static const struct net_device_ops ice_netdev_ops;
  33
  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         struct ice_hw *hw;
  65         unsigned int i;
  66         int packets;
  67         u32 v;
  68
  69         ice_for_each_vsi(pf, v)
  70                 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
  71                         vsi = pf->vsi[v];
  72                         break;
  73                 }
  74
  75         if (!vsi || test_bit(__ICE_DOWN, vsi->state))
  76                 return;
  77
  78         if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
  79                 return;
  80
  81         hw = &vsi->back->hw;
  82
  83         for (i = 0; i < vsi->num_txq; i++) {
  84                 struct ice_ring *tx_ring = vsi->tx_rings[i];
  85
  86                 if (tx_ring && tx_ring->desc) {
  87                         /* If packet counter has not changed the queue is
  88                          * likely stalled, so force an interrupt for this
  89                          * queue.
  90                          *
  91                          * prev_pkt would be negative if there was no
  92                          * pending work.
  93                          */
  94                         packets = tx_ring->stats.pkts & INT_MAX;
  95                         if (tx_ring->tx_stats.prev_pkt == packets) {
  96                                 /* Trigger sw interrupt to revive the queue */
  97                                 ice_trigger_sw_intr(hw, tx_ring->q_vector);
  98                                 continue;
  99                         }
 100
 101                         /* Memory barrier between read of packet count and call
 102                          * to ice_get_tx_pending()
 103                          */
 104                         smp_rmb();
 105                         tx_ring->tx_stats.prev_pkt =
 106                             ice_get_tx_pending(tx_ring) ? packets : -1;
 107                 }
 108         }
 109 }
 110
 111 /**
 112  * ice_init_mac_fltr - Set initial MAC filters
 113  * @pf: board private structure
 114  *
 115  * Set initial set of MAC filters for PF VSI; configure filters for permanent
 116  * address and broadcast address. If an error is encountered, netdevice will be
 117  * unregistered.
 118  */
 119 static int ice_init_mac_fltr(struct ice_pf *pf)
 120 {
 121         LIST_HEAD(tmp_add_list);
 122         u8 broadcast[ETH_ALEN];
 123         struct ice_vsi *vsi;
 124         int status;
 125
 126         vsi = ice_find_vsi_by_type(pf, ICE_VSI_PF);
 127         if (!vsi)
 128                 return -EINVAL;
 129
 130         /* To add a MAC filter, first add the MAC to a list and then
 131          * pass the list to ice_add_mac.
 132          */
 133
 134          /* Add a unicast MAC filter so the VSI can get its packets */
 135         status = ice_add_mac_to_list(vsi, &tmp_add_list,
 136                                      vsi->port_info->mac.perm_addr);
 137         if (status)
 138                 goto unregister;
 139
 140         /* VSI needs to receive broadcast traffic, so add the broadcast
 141          * MAC address to the list as well.
 142          */
 143         eth_broadcast_addr(broadcast);
 144         status = ice_add_mac_to_list(vsi, &tmp_add_list, broadcast);
 145         if (status)
 146                 goto free_mac_list;
 147
 148         /* Program MAC filters for entries in tmp_add_list */
 149         status = ice_add_mac(&pf->hw, &tmp_add_list);
 150         if (status)
 151                 status = -ENOMEM;
 152
 153 free_mac_list:
 154         ice_free_fltr_list(&pf->pdev->dev, &tmp_add_list);
 155
 156 unregister:
 157         /* We aren't useful with no MAC filters, so unregister if we
 158          * had an error
 159          */
 160         if (status && vsi->netdev->reg_state == NETREG_REGISTERED) {
 161                 dev_err(&pf->pdev->dev,
 162                         "Could not add MAC filters error %d. Unregistering device\n",
 163                         status);
 164                 unregister_netdev(vsi->netdev);
 165                 free_netdev(vsi->netdev);
 166                 vsi->netdev = NULL;
 167         }
 168
 169         return status;
 170 }
 171
 172 /**
 173  * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
 174  * @netdev: the net device on which the sync is happening
 175  * @addr: MAC address to sync
 176  *
 177  * This is a callback function which is called by the in kernel device sync
 178  * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
 179  * populates the tmp_sync_list, which is later used by ice_add_mac to add the
 180  * MAC filters from the hardware.
 181  */
 182 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
 183 {
 184         struct ice_netdev_priv *np = netdev_priv(netdev);
 185         struct ice_vsi *vsi = np->vsi;
 186
 187         if (ice_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr))
 188                 return -EINVAL;
 189
 190         return 0;
 191 }
 192
 193 /**
 194  * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
 195  * @netdev: the net device on which the unsync is happening
 196  * @addr: MAC address to unsync
 197  *
 198  * This is a callback function which is called by the in kernel device unsync
 199  * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
 200  * populates the tmp_unsync_list, which is later used by ice_remove_mac to
 201  * delete the MAC filters from the hardware.
 202  */
 203 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
 204 {
 205         struct ice_netdev_priv *np = netdev_priv(netdev);
 206         struct ice_vsi *vsi = np->vsi;
 207
 208         if (ice_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr))
 209                 return -EINVAL;
 210
 211         return 0;
 212 }
 213
 214 /**
 215  * ice_vsi_fltr_changed - check if filter state changed
 216  * @vsi: VSI to be checked
 217  *
 218  * returns true if filter state has changed, false otherwise.
 219  */
 220 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
 221 {
 222         return test_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags) ||
 223                test_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags) ||
 224                test_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
 225 }
 226
 227 /**
 228  * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
 229  * @vsi: the VSI being configured
 230  * @promisc_m: mask of promiscuous config bits
 231  * @set_promisc: enable or disable promisc flag request
 232  *
 233  */
 234 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
 235 {
 236         struct ice_hw *hw = &vsi->back->hw;
 237         enum ice_status status = 0;
 238
 239         if (vsi->type != ICE_VSI_PF)
 240                 return 0;
 241
 242         if (vsi->vlan_ena) {
 243                 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
 244                                                   set_promisc);
 245         } else {
 246                 if (set_promisc)
 247                         status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
 248                                                      0);
 249                 else
 250                         status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
 251                                                        0);
 252         }
 253
 254         if (status)
 255                 return -EIO;
 256
 257         return 0;
 258 }
 259
 260 /**
 261  * ice_vsi_sync_fltr - Update the VSI filter list to the HW
 262  * @vsi: ptr to the VSI
 263  *
 264  * Push any outstanding VSI filter changes through the AdminQ.
 265  */
 266 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
 267 {
 268         struct device *dev = &vsi->back->pdev->dev;
 269         struct net_device *netdev = vsi->netdev;
 270         bool promisc_forced_on = false;
 271         struct ice_pf *pf = vsi->back;
 272         struct ice_hw *hw = &pf->hw;
 273         enum ice_status status = 0;
 274         u32 changed_flags = 0;
 275         u8 promisc_m;
 276         int err = 0;
 277
 278         if (!vsi->netdev)
 279                 return -EINVAL;
 280
 281         while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state))
 282                 usleep_range(1000, 2000);
 283
 284         changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
 285         vsi->current_netdev_flags = vsi->netdev->flags;
 286
 287         INIT_LIST_HEAD(&vsi->tmp_sync_list);
 288         INIT_LIST_HEAD(&vsi->tmp_unsync_list);
 289
 290         if (ice_vsi_fltr_changed(vsi)) {
 291                 clear_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
 292                 clear_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
 293                 clear_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
 294
 295                 /* grab the netdev's addr_list_lock */
 296                 netif_addr_lock_bh(netdev);
 297                 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
 298                               ice_add_mac_to_unsync_list);
 299                 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
 300                               ice_add_mac_to_unsync_list);
 301                 /* our temp lists are populated. release lock */
 302                 netif_addr_unlock_bh(netdev);
 303         }
 304
 305         /* Remove MAC addresses in the unsync list */
 306         status = ice_remove_mac(hw, &vsi->tmp_unsync_list);
 307         ice_free_fltr_list(dev, &vsi->tmp_unsync_list);
 308         if (status) {
 309                 netdev_err(netdev, "Failed to delete MAC filters\n");
 310                 /* if we failed because of alloc failures, just bail */
 311                 if (status == ICE_ERR_NO_MEMORY) {
 312                         err = -ENOMEM;
 313                         goto out;
 314                 }
 315         }
 316
 317         /* Add MAC addresses in the sync list */
 318         status = ice_add_mac(hw, &vsi->tmp_sync_list);
 319         ice_free_fltr_list(dev, &vsi->tmp_sync_list);
 320         /* If filter is added successfully or already exists, do not go into
 321          * 'if' condition and report it as error. Instead continue processing
 322          * rest of the function.
 323          */
 324         if (status && status != ICE_ERR_ALREADY_EXISTS) {
 325                 netdev_err(netdev, "Failed to add MAC filters\n");
 326                 /* If there is no more space for new umac filters, VSI
 327                  * should go into promiscuous mode. There should be some
 328                  * space reserved for promiscuous filters.
 329                  */
 330                 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
 331                     !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC,
 332                                       vsi->state)) {
 333                         promisc_forced_on = true;
 334                         netdev_warn(netdev,
 335                                     "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
 336                                     vsi->vsi_num);
 337                 } else {
 338                         err = -EIO;
 339                         goto out;
 340                 }
 341         }
 342         /* check for changes in promiscuous modes */
 343         if (changed_flags & IFF_ALLMULTI) {
 344                 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
 345                         if (vsi->vlan_ena)
 346                                 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
 347                         else
 348                                 promisc_m = ICE_MCAST_PROMISC_BITS;
 349
 350                         err = ice_cfg_promisc(vsi, promisc_m, true);
 351                         if (err) {
 352                                 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
 353                                            vsi->vsi_num);
 354                                 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
 355                                 goto out_promisc;
 356                         }
 357                 } else if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) {
 358                         if (vsi->vlan_ena)
 359                                 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
 360                         else
 361                                 promisc_m = ICE_MCAST_PROMISC_BITS;
 362
 363                         err = ice_cfg_promisc(vsi, promisc_m, false);
 364                         if (err) {
 365                                 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
 366                                            vsi->vsi_num);
 367                                 vsi->current_netdev_flags |= IFF_ALLMULTI;
 368                                 goto out_promisc;
 369                         }
 370                 }
 371         }
 372
 373         if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
 374             test_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags)) {
 375                 clear_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
 376                 if (vsi->current_netdev_flags & IFF_PROMISC) {
 377                         /* Apply Rx filter rule to get traffic from wire */
 378                         status = ice_cfg_dflt_vsi(hw, vsi->idx, true,
 379                                                   ICE_FLTR_RX);
 380                         if (status) {
 381                                 netdev_err(netdev, "Error setting default VSI %i Rx rule\n",
 382                                            vsi->vsi_num);
 383                                 vsi->current_netdev_flags &= ~IFF_PROMISC;
 384                                 err = -EIO;
 385                                 goto out_promisc;
 386                         }
 387                 } else {
 388                         /* Clear Rx filter to remove traffic from wire */
 389                         status = ice_cfg_dflt_vsi(hw, vsi->idx, false,
 390                                                   ICE_FLTR_RX);
 391                         if (status) {
 392                                 netdev_err(netdev, "Error clearing default VSI %i Rx rule\n",
 393                                            vsi->vsi_num);
 394                                 vsi->current_netdev_flags |= IFF_PROMISC;
 395                                 err = -EIO;
 396                                 goto out_promisc;
 397                         }
 398                 }
 399         }
 400         goto exit;
 401
 402 out_promisc:
 403         set_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
 404         goto exit;
 405 out:
 406         /* if something went wrong then set the changed flag so we try again */
 407         set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
 408         set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
 409 exit:
 410         clear_bit(__ICE_CFG_BUSY, vsi->state);
 411         return err;
 412 }
 413
 414 /**
 415  * ice_sync_fltr_subtask - Sync the VSI filter list with HW
 416  * @pf: board private structure
 417  */
 418 static void ice_sync_fltr_subtask(struct ice_pf *pf)
 419 {
 420         int v;
 421
 422         if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
 423                 return;
 424
 425         clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
 426
 427         ice_for_each_vsi(pf, v)
 428                 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
 429                     ice_vsi_sync_fltr(pf->vsi[v])) {
 430                         /* come back and try again later */
 431                         set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
 432                         break;
 433                 }
 434 }
 435
 436 /**
 437  * ice_dis_vsi - pause a VSI
 438  * @vsi: the VSI being paused
 439  * @locked: is the rtnl_lock already held
 440  */
 441 static void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
 442 {
 443         if (test_bit(__ICE_DOWN, vsi->state))
 444                 return;
 445
 446         set_bit(__ICE_NEEDS_RESTART, vsi->state);
 447
 448         if (vsi->type == ICE_VSI_PF && vsi->netdev) {
 449                 if (netif_running(vsi->netdev)) {
 450                         if (!locked) {
 451                                 rtnl_lock();
 452                                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
 453                                 rtnl_unlock();
 454                         } else {
 455                                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
 456                         }
 457                 } else {
 458                         ice_vsi_close(vsi);
 459                 }
 460         }
 461 }
 462
 463 /**
 464  * ice_pf_dis_all_vsi - Pause all VSIs on a PF
 465  * @pf: the PF
 466  * @locked: is the rtnl_lock already held
 467  */
 468 #ifdef CONFIG_DCB
 469 void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
 470 #else
 471 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
 472 #endif /* CONFIG_DCB */
 473 {
 474         int v;
 475
 476         ice_for_each_vsi(pf, v)
 477                 if (pf->vsi[v])
 478                         ice_dis_vsi(pf->vsi[v], locked);
 479 }
 480
 481 /**
 482  * ice_prepare_for_reset - prep for the core to reset
 483  * @pf: board private structure
 484  *
 485  * Inform or close all dependent features in prep for reset.
 486  */
 487 static void
 488 ice_prepare_for_reset(struct ice_pf *pf)
 489 {
 490         struct ice_hw *hw = &pf->hw;
 491
 492         /* already prepared for reset */
 493         if (test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
 494                 return;
 495
 496         /* Notify VFs of impending reset */
 497         if (ice_check_sq_alive(hw, &hw->mailboxq))
 498                 ice_vc_notify_reset(pf);
 499
 500         /* disable the VSIs and their queues that are not already DOWN */
 501         ice_pf_dis_all_vsi(pf, false);
 502
 503         if (hw->port_info)
 504                 ice_sched_clear_port(hw->port_info);
 505
 506         ice_shutdown_all_ctrlq(hw);
 507
 508         set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 509 }
 510
 511 /**
 512  * ice_do_reset - Initiate one of many types of resets
 513  * @pf: board private structure
 514  * @reset_type: reset type requested
 515  * before this function was called.
 516  */
 517 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 518 {
 519         struct device *dev = &pf->pdev->dev;
 520         struct ice_hw *hw = &pf->hw;
 521
 522         dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
 523         WARN_ON(in_interrupt());
 524
 525         ice_prepare_for_reset(pf);
 526
 527         /* trigger the reset */
 528         if (ice_reset(hw, reset_type)) {
 529                 dev_err(dev, "reset %d failed\n", reset_type);
 530                 set_bit(__ICE_RESET_FAILED, pf->state);
 531                 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 532                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 533                 clear_bit(__ICE_PFR_REQ, pf->state);
 534                 clear_bit(__ICE_CORER_REQ, pf->state);
 535                 clear_bit(__ICE_GLOBR_REQ, pf->state);
 536                 return;
 537         }
 538
 539         /* PFR is a bit of a special case because it doesn't result in an OICR
 540          * interrupt. So for PFR, rebuild after the reset and clear the reset-
 541          * associated state bits.
 542          */
 543         if (reset_type == ICE_RESET_PFR) {
 544                 pf->pfr_count++;
 545                 ice_rebuild(pf);
 546                 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 547                 clear_bit(__ICE_PFR_REQ, pf->state);
 548                 ice_reset_all_vfs(pf, true);
 549         }
 550 }
 551
 552 /**
 553  * ice_reset_subtask - Set up for resetting the device and driver
 554  * @pf: board private structure
 555  */
 556 static void ice_reset_subtask(struct ice_pf *pf)
 557 {
 558         enum ice_reset_req reset_type = ICE_RESET_INVAL;
 559
 560         /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
 561          * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
 562          * of reset is pending and sets bits in pf->state indicating the reset
 563          * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set
 564          * prepare for pending reset if not already (for PF software-initiated
 565          * global resets the software should already be prepared for it as
 566          * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
 567          * by firmware or software on other PFs, that bit is not set so prepare
 568          * for the reset now), poll for reset done, rebuild and return.
 569          */
 570         if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
 571                 /* Perform the largest reset requested */
 572                 if (test_and_clear_bit(__ICE_CORER_RECV, pf->state))
 573                         reset_type = ICE_RESET_CORER;
 574                 if (test_and_clear_bit(__ICE_GLOBR_RECV, pf->state))
 575                         reset_type = ICE_RESET_GLOBR;
 576                 /* return if no valid reset type requested */
 577                 if (reset_type == ICE_RESET_INVAL)
 578                         return;
 579                 ice_prepare_for_reset(pf);
 580
 581                 /* make sure we are ready to rebuild */
 582                 if (ice_check_reset(&pf->hw)) {
 583                         set_bit(__ICE_RESET_FAILED, pf->state);
 584                 } else {
 585                         /* done with reset. start rebuild */
 586                         pf->hw.reset_ongoing = false;
 587                         ice_rebuild(pf);
 588                         /* clear bit to resume normal operations, but
 589                          * ICE_NEEDS_RESTART bit is set in case rebuild failed
 590                          */
 591                         clear_bit(__ICE_RESET_OICR_RECV, pf->state);
 592                         clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
 593                         clear_bit(__ICE_PFR_REQ, pf->state);
 594                         clear_bit(__ICE_CORER_REQ, pf->state);
 595                         clear_bit(__ICE_GLOBR_REQ, pf->state);
 596                         ice_reset_all_vfs(pf, true);
 597                 }
 598
 599                 return;
 600         }
 601
 602         /* No pending resets to finish processing. Check for new resets */
 603         if (test_bit(__ICE_PFR_REQ, pf->state))
 604                 reset_type = ICE_RESET_PFR;
 605         if (test_bit(__ICE_CORER_REQ, pf->state))
 606                 reset_type = ICE_RESET_CORER;
 607         if (test_bit(__ICE_GLOBR_REQ, pf->state))
 608                 reset_type = ICE_RESET_GLOBR;
 609         /* If no valid reset type requested just return */
 610         if (reset_type == ICE_RESET_INVAL)
 611                 return;
 612
 613         /* reset if not already down or busy */
 614         if (!test_bit(__ICE_DOWN, pf->state) &&
 615             !test_bit(__ICE_CFG_BUSY, pf->state)) {
 616                 ice_do_reset(pf, reset_type);
 617         }
 618 }
 619
 620 /**
 621  * ice_print_link_msg - print link up or down message
 622  * @vsi: the VSI whose link status is being queried
 623  * @isup: boolean for if the link is now up or down
 624  */
 625 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
 626 {
 627         struct ice_aqc_get_phy_caps_data *caps;
 628         enum ice_status status;
 629         const char *fec_req;
 630         const char *speed;
 631         const char *fec;
 632         const char *fc;
 633
 634         if (!vsi)
 635                 return;
 636
 637         if (vsi->current_isup == isup)
 638                 return;
 639
 640         vsi->current_isup = isup;
 641
 642         if (!isup) {
 643                 netdev_info(vsi->netdev, "NIC Link is Down\n");
 644                 return;
 645         }
 646
 647         switch (vsi->port_info->phy.link_info.link_speed) {
 648         case ICE_AQ_LINK_SPEED_100GB:
 649                 speed = "100 G";
 650                 break;
 651         case ICE_AQ_LINK_SPEED_50GB:
 652                 speed = "50 G";
 653                 break;
 654         case ICE_AQ_LINK_SPEED_40GB:
 655                 speed = "40 G";
 656                 break;
 657         case ICE_AQ_LINK_SPEED_25GB:
 658                 speed = "25 G";
 659                 break;
 660         case ICE_AQ_LINK_SPEED_20GB:
 661                 speed = "20 G";
 662                 break;
 663         case ICE_AQ_LINK_SPEED_10GB:
 664                 speed = "10 G";
 665                 break;
 666         case ICE_AQ_LINK_SPEED_5GB:
 667                 speed = "5 G";
 668                 break;
 669         case ICE_AQ_LINK_SPEED_2500MB:
 670                 speed = "2.5 G";
 671                 break;
 672         case ICE_AQ_LINK_SPEED_1000MB:
 673                 speed = "1 G";
 674                 break;
 675         case ICE_AQ_LINK_SPEED_100MB:
 676                 speed = "100 M";
 677                 break;
 678         default:
 679                 speed = "Unknown";
 680                 break;
 681         }
 682
 683         switch (vsi->port_info->fc.current_mode) {
 684         case ICE_FC_FULL:
 685                 fc = "Rx/Tx";
 686                 break;
 687         case ICE_FC_TX_PAUSE:
 688                 fc = "Tx";
 689                 break;
 690         case ICE_FC_RX_PAUSE:
 691                 fc = "Rx";
 692                 break;
 693         case ICE_FC_NONE:
 694                 fc = "None";
 695                 break;
 696         default:
 697                 fc = "Unknown";
 698                 break;
 699         }
 700
 701         /* Get FEC mode based on negotiated link info */
 702         switch (vsi->port_info->phy.link_info.fec_info) {
 703         case ICE_AQ_LINK_25G_RS_528_FEC_EN:
 704                 /* fall through */
 705         case ICE_AQ_LINK_25G_RS_544_FEC_EN:
 706                 fec = "RS-FEC";
 707                 break;
 708         case ICE_AQ_LINK_25G_KR_FEC_EN:
 709                 fec = "FC-FEC/BASE-R";
 710                 break;
 711         default:
 712                 fec = "NONE";
 713                 break;
 714         }
 715
 716         /* Get FEC mode requested based on PHY caps last SW configuration */
 717         caps = devm_kzalloc(&vsi->back->pdev->dev, sizeof(*caps), GFP_KERNEL);
 718         if (!caps) {
 719                 fec_req = "Unknown";
 720                 goto done;
 721         }
 722
 723         status = ice_aq_get_phy_caps(vsi->port_info, false,
 724                                      ICE_AQC_REPORT_SW_CFG, caps, NULL);
 725         if (status)
 726                 netdev_info(vsi->netdev, "Get phy capability failed.\n");
 727
 728         if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
 729             caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
 730                 fec_req = "RS-FEC";
 731         else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
 732                  caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
 733                 fec_req = "FC-FEC/BASE-R";
 734         else
 735                 fec_req = "NONE";
 736
 737         devm_kfree(&vsi->back->pdev->dev, caps);
 738
 739 done:
 740         netdev_info(vsi->netdev, "NIC Link is up %sbps, Requested FEC: %s, FEC: %s, Flow Control: %s\n",
 741                     speed, fec_req, fec, fc);
 742 }
 743
 744 /**
 745  * ice_vsi_link_event - update the VSI's netdev
 746  * @vsi: the VSI on which the link event occurred
 747  * @link_up: whether or not the VSI needs to be set up or down
 748  */
 749 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
 750 {
 751         if (!vsi)
 752                 return;
 753
 754         if (test_bit(__ICE_DOWN, vsi->state) || !vsi->netdev)
 755                 return;
 756
 757         if (vsi->type == ICE_VSI_PF) {
 758                 if (link_up == netif_carrier_ok(vsi->netdev))
 759                         return;
 760
 761                 if (link_up) {
 762                         netif_carrier_on(vsi->netdev);
 763                         netif_tx_wake_all_queues(vsi->netdev);
 764                 } else {
 765                         netif_carrier_off(vsi->netdev);
 766                         netif_tx_stop_all_queues(vsi->netdev);
 767                 }
 768         }
 769 }
 770
 771 /**
 772  * ice_link_event - process the link event
 773  * @pf: PF that the link event is associated with
 774  * @pi: port_info for the port that the link event is associated with
 775  * @link_up: true if the physical link is up and false if it is down
 776  * @link_speed: current link speed received from the link event
 777  *
 778  * Returns 0 on success and negative on failure
 779  */
 780 static int
 781 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
 782                u16 link_speed)
 783 {
 784         struct ice_phy_info *phy_info;
 785         struct ice_vsi *vsi;
 786         u16 old_link_speed;
 787         bool old_link;
 788         int result;
 789
 790         phy_info = &pi->phy;
 791         phy_info->link_info_old = phy_info->link_info;
 792
 793         old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
 794         old_link_speed = phy_info->link_info_old.link_speed;
 795
 796         /* update the link info structures and re-enable link events,
 797          * don't bail on failure due to other book keeping needed
 798          */
 799         result = ice_update_link_info(pi);
 800         if (result)
 801                 dev_dbg(&pf->pdev->dev,
 802                         "Failed to update link status and re-enable link events for port %d\n",
 803                         pi->lport);
 804
 805         /* if the old link up/down and speed is the same as the new */
 806         if (link_up == old_link && link_speed == old_link_speed)
 807                 return result;
 808
 809         vsi = ice_find_vsi_by_type(pf, ICE_VSI_PF);
 810         if (!vsi || !vsi->port_info)
 811                 return -EINVAL;
 812
 813         ice_vsi_link_event(vsi, link_up);
 814         ice_print_link_msg(vsi, link_up);
 815
 816         if (pf->num_alloc_vfs)
 817                 ice_vc_notify_link_state(pf);
 818
 819         return result;
 820 }
 821
 822 /**
 823  * ice_watchdog_subtask - periodic tasks not using event driven scheduling
 824  * @pf: board private structure
 825  */
 826 static void ice_watchdog_subtask(struct ice_pf *pf)
 827 {
 828         int i;
 829
 830         /* if interface is down do nothing */
 831         if (test_bit(__ICE_DOWN, pf->state) ||
 832             test_bit(__ICE_CFG_BUSY, pf->state))
 833                 return;
 834
 835         /* make sure we don't do these things too often */
 836         if (time_before(jiffies,
 837                         pf->serv_tmr_prev + pf->serv_tmr_period))
 838                 return;
 839
 840         pf->serv_tmr_prev = jiffies;
 841
 842         /* Update the stats for active netdevs so the network stack
 843          * can look at updated numbers whenever it cares to
 844          */
 845         ice_update_pf_stats(pf);
 846         ice_for_each_vsi(pf, i)
 847                 if (pf->vsi[i] && pf->vsi[i]->netdev)
 848                         ice_update_vsi_stats(pf->vsi[i]);
 849 }
 850
 851 /**
 852  * ice_init_link_events - enable/initialize link events
 853  * @pi: pointer to the port_info instance
 854  *
 855  * Returns -EIO on failure, 0 on success
 856  */
 857 static int ice_init_link_events(struct ice_port_info *pi)
 858 {
 859         u16 mask;
 860
 861         mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
 862                        ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
 863
 864         if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
 865                 dev_dbg(ice_hw_to_dev(pi->hw),
 866                         "Failed to set link event mask for port %d\n",
 867                         pi->lport);
 868                 return -EIO;
 869         }
 870
 871         if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
 872                 dev_dbg(ice_hw_to_dev(pi->hw),
 873                         "Failed to enable link events for port %d\n",
 874                         pi->lport);
 875                 return -EIO;
 876         }
 877
 878         return 0;
 879 }
 880
 881 /**
 882  * ice_handle_link_event - handle link event via ARQ
 883  * @pf: PF that the link event is associated with
 884  * @event: event structure containing link status info
 885  */
 886 static int
 887 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
 888 {
 889         struct ice_aqc_get_link_status_data *link_data;
 890         struct ice_port_info *port_info;
 891         int status;
 892
 893         link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
 894         port_info = pf->hw.port_info;
 895         if (!port_info)
 896                 return -EINVAL;
 897
 898         status = ice_link_event(pf, port_info,
 899                                 !!(link_data->link_info & ICE_AQ_LINK_UP),
 900                                 le16_to_cpu(link_data->link_speed));
 901         if (status)
 902                 dev_dbg(&pf->pdev->dev,
 903                         "Could not process link event, error %d\n", status);
 904
 905         return status;
 906 }
 907
 908 /**
 909  * __ice_clean_ctrlq - helper function to clean controlq rings
 910  * @pf: ptr to struct ice_pf
 911  * @q_type: specific Control queue type
 912  */
 913 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
 914 {
 915         struct ice_rq_event_info event;
 916         struct ice_hw *hw = &pf->hw;
 917         struct ice_ctl_q_info *cq;
 918         u16 pending, i = 0;
 919         const char *qtype;
 920         u32 oldval, val;
 921
 922         /* Do not clean control queue if/when PF reset fails */
 923         if (test_bit(__ICE_RESET_FAILED, pf->state))
 924                 return 0;
 925
 926         switch (q_type) {
 927         case ICE_CTL_Q_ADMIN:
 928                 cq = &hw->adminq;
 929                 qtype = "Admin";
 930                 break;
 931         case ICE_CTL_Q_MAILBOX:
 932                 cq = &hw->mailboxq;
 933                 qtype = "Mailbox";
 934                 break;
 935         default:
 936                 dev_warn(&pf->pdev->dev, "Unknown control queue type 0x%x\n",
 937                          q_type);
 938                 return 0;
 939         }
 940
 941         /* check for error indications - PF_xx_AxQLEN register layout for
 942          * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
 943          */
 944         val = rd32(hw, cq->rq.len);
 945         if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 946                    PF_FW_ARQLEN_ARQCRIT_M)) {
 947                 oldval = val;
 948                 if (val & PF_FW_ARQLEN_ARQVFE_M)
 949                         dev_dbg(&pf->pdev->dev,
 950                                 "%s Receive Queue VF Error detected\n", qtype);
 951                 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
 952                         dev_dbg(&pf->pdev->dev,
 953                                 "%s Receive Queue Overflow Error detected\n",
 954                                 qtype);
 955                 }
 956                 if (val & PF_FW_ARQLEN_ARQCRIT_M)
 957                         dev_dbg(&pf->pdev->dev,
 958                                 "%s Receive Queue Critical Error detected\n",
 959                                 qtype);
 960                 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
 961                          PF_FW_ARQLEN_ARQCRIT_M);
 962                 if (oldval != val)
 963                         wr32(hw, cq->rq.len, val);
 964         }
 965
 966         val = rd32(hw, cq->sq.len);
 967         if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 968                    PF_FW_ATQLEN_ATQCRIT_M)) {
 969                 oldval = val;
 970                 if (val & PF_FW_ATQLEN_ATQVFE_M)
 971                         dev_dbg(&pf->pdev->dev,
 972                                 "%s Send Queue VF Error detected\n", qtype);
 973                 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
 974                         dev_dbg(&pf->pdev->dev,
 975                                 "%s Send Queue Overflow Error detected\n",
 976                                 qtype);
 977                 }
 978                 if (val & PF_FW_ATQLEN_ATQCRIT_M)
 979                         dev_dbg(&pf->pdev->dev,
 980                                 "%s Send Queue Critical Error detected\n",
 981                                 qtype);
 982                 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
 983                          PF_FW_ATQLEN_ATQCRIT_M);
 984                 if (oldval != val)
 985                         wr32(hw, cq->sq.len, val);
 986         }
 987
 988         event.buf_len = cq->rq_buf_size;
 989         event.msg_buf = devm_kzalloc(&pf->pdev->dev, event.buf_len,
 990                                      GFP_KERNEL);
 991         if (!event.msg_buf)
 992                 return 0;
 993
 994         do {
 995                 enum ice_status ret;
 996                 u16 opcode;
 997
 998                 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
 999                 if (ret == ICE_ERR_AQ_NO_WORK)
1000                         break;
1001                 if (ret) {
1002                         dev_err(&pf->pdev->dev,
1003                                 "%s Receive Queue event error %d\n", qtype,
1004                                 ret);
1005                         break;
1006                 }
1007
1008                 opcode = le16_to_cpu(event.desc.opcode);
1009
1010                 switch (opcode) {
1011                 case ice_aqc_opc_get_link_status:
1012                         if (ice_handle_link_event(pf, &event))
1013                                 dev_err(&pf->pdev->dev,
1014                                         "Could not handle link event\n");
1015                         break;
1016                 case ice_mbx_opc_send_msg_to_pf:
1017                         ice_vc_process_vf_msg(pf, &event);
1018                         break;
1019                 case ice_aqc_opc_fw_logging:
1020                         ice_output_fw_log(hw, &event.desc, event.msg_buf);
1021                         break;
1022                 case ice_aqc_opc_lldp_set_mib_change:
1023                         ice_dcb_process_lldp_set_mib_change(pf, &event);
1024                         break;
1025                 default:
1026                         dev_dbg(&pf->pdev->dev,
1027                                 "%s Receive Queue unknown event 0x%04x ignored\n",
1028                                 qtype, opcode);
1029                         break;
1030                 }
1031         } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1032
1033         devm_kfree(&pf->pdev->dev, event.msg_buf);
1034
1035         return pending && (i == ICE_DFLT_IRQ_WORK);
1036 }
1037
1038 /**
1039  * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1040  * @hw: pointer to hardware info
1041  * @cq: control queue information
1042  *
1043  * returns true if there are pending messages in a queue, false if there aren't
1044  */
1045 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1046 {
1047         u16 ntu;
1048
1049         ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1050         return cq->rq.next_to_clean != ntu;
1051 }
1052
1053 /**
1054  * ice_clean_adminq_subtask - clean the AdminQ rings
1055  * @pf: board private structure
1056  */
1057 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1058 {
1059         struct ice_hw *hw = &pf->hw;
1060
1061         if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1062                 return;
1063
1064         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1065                 return;
1066
1067         clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1068
1069         /* There might be a situation where new messages arrive to a control
1070          * queue between processing the last message and clearing the
1071          * EVENT_PENDING bit. So before exiting, check queue head again (using
1072          * ice_ctrlq_pending) and process new messages if any.
1073          */
1074         if (ice_ctrlq_pending(hw, &hw->adminq))
1075                 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1076
1077         ice_flush(hw);
1078 }
1079
1080 /**
1081  * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1082  * @pf: board private structure
1083  */
1084 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1085 {
1086         struct ice_hw *hw = &pf->hw;
1087
1088         if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1089                 return;
1090
1091         if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1092                 return;
1093
1094         clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1095
1096         if (ice_ctrlq_pending(hw, &hw->mailboxq))
1097                 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1098
1099         ice_flush(hw);
1100 }
1101
1102 /**
1103  * ice_service_task_schedule - schedule the service task to wake up
1104  * @pf: board private structure
1105  *
1106  * If not already scheduled, this puts the task into the work queue.
1107  */
1108 static void ice_service_task_schedule(struct ice_pf *pf)
1109 {
1110         if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
1111             !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
1112             !test_bit(__ICE_NEEDS_RESTART, pf->state))
1113                 queue_work(ice_wq, &pf->serv_task);
1114 }
1115
1116 /**
1117  * ice_service_task_complete - finish up the service task
1118  * @pf: board private structure
1119  */
1120 static void ice_service_task_complete(struct ice_pf *pf)
1121 {
1122         WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
1123
1124         /* force memory (pf->state) to sync before next service task */
1125         smp_mb__before_atomic();
1126         clear_bit(__ICE_SERVICE_SCHED, pf->state);
1127 }
1128
1129 /**
1130  * ice_service_task_stop - stop service task and cancel works
1131  * @pf: board private structure
1132  */
1133 static void ice_service_task_stop(struct ice_pf *pf)
1134 {
1135         set_bit(__ICE_SERVICE_DIS, pf->state);
1136
1137         if (pf->serv_tmr.function)
1138                 del_timer_sync(&pf->serv_tmr);
1139         if (pf->serv_task.func)
1140                 cancel_work_sync(&pf->serv_task);
1141
1142         clear_bit(__ICE_SERVICE_SCHED, pf->state);
1143 }
1144
1145 /**
1146  * ice_service_task_restart - restart service task and schedule works
1147  * @pf: board private structure
1148  *
1149  * This function is needed for suspend and resume works (e.g WoL scenario)
1150  */
1151 static void ice_service_task_restart(struct ice_pf *pf)
1152 {
1153         clear_bit(__ICE_SERVICE_DIS, pf->state);
1154         ice_service_task_schedule(pf);
1155 }
1156
1157 /**
1158  * ice_service_timer - timer callback to schedule service task
1159  * @t: pointer to timer_list
1160  */
1161 static void ice_service_timer(struct timer_list *t)
1162 {
1163         struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1164
1165         mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1166         ice_service_task_schedule(pf);
1167 }
1168
1169 /**
1170  * ice_handle_mdd_event - handle malicious driver detect event
1171  * @pf: pointer to the PF structure
1172  *
1173  * Called from service task. OICR interrupt handler indicates MDD event
1174  */
1175 static void ice_handle_mdd_event(struct ice_pf *pf)
1176 {
1177         struct ice_hw *hw = &pf->hw;
1178         bool mdd_detected = false;
1179         u32 reg;
1180         int i;
1181
1182         if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state))
1183                 return;
1184
1185         /* find what triggered the MDD event */
1186         reg = rd32(hw, GL_MDET_TX_PQM);
1187         if (reg & GL_MDET_TX_PQM_VALID_M) {
1188                 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1189                                 GL_MDET_TX_PQM_PF_NUM_S;
1190                 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1191                                 GL_MDET_TX_PQM_VF_NUM_S;
1192                 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1193                                 GL_MDET_TX_PQM_MAL_TYPE_S;
1194                 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1195                                 GL_MDET_TX_PQM_QNUM_S);
1196
1197                 if (netif_msg_tx_err(pf))
1198                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1199                                  event, queue, pf_num, vf_num);
1200                 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1201                 mdd_detected = true;
1202         }
1203
1204         reg = rd32(hw, GL_MDET_TX_TCLAN);
1205         if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1206                 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1207                                 GL_MDET_TX_TCLAN_PF_NUM_S;
1208                 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1209                                 GL_MDET_TX_TCLAN_VF_NUM_S;
1210                 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1211                                 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1212                 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1213                                 GL_MDET_TX_TCLAN_QNUM_S);
1214
1215                 if (netif_msg_rx_err(pf))
1216                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1217                                  event, queue, pf_num, vf_num);
1218                 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1219                 mdd_detected = true;
1220         }
1221
1222         reg = rd32(hw, GL_MDET_RX);
1223         if (reg & GL_MDET_RX_VALID_M) {
1224                 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1225                                 GL_MDET_RX_PF_NUM_S;
1226                 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1227                                 GL_MDET_RX_VF_NUM_S;
1228                 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1229                                 GL_MDET_RX_MAL_TYPE_S;
1230                 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1231                                 GL_MDET_RX_QNUM_S);
1232
1233                 if (netif_msg_rx_err(pf))
1234                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1235                                  event, queue, pf_num, vf_num);
1236                 wr32(hw, GL_MDET_RX, 0xffffffff);
1237                 mdd_detected = true;
1238         }
1239
1240         if (mdd_detected) {
1241                 bool pf_mdd_detected = false;
1242
1243                 reg = rd32(hw, PF_MDET_TX_PQM);
1244                 if (reg & PF_MDET_TX_PQM_VALID_M) {
1245                         wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1246                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
1247                         pf_mdd_detected = true;
1248                 }
1249
1250                 reg = rd32(hw, PF_MDET_TX_TCLAN);
1251                 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1252                         wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1253                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
1254                         pf_mdd_detected = true;
1255                 }
1256
1257                 reg = rd32(hw, PF_MDET_RX);
1258                 if (reg & PF_MDET_RX_VALID_M) {
1259                         wr32(hw, PF_MDET_RX, 0xFFFF);
1260                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
1261                         pf_mdd_detected = true;
1262                 }
1263                 /* Queue belongs to the PF initiate a reset */
1264                 if (pf_mdd_detected) {
1265                         set_bit(__ICE_NEEDS_RESTART, pf->state);
1266                         ice_service_task_schedule(pf);
1267                 }
1268         }
1269
1270         /* check to see if one of the VFs caused the MDD */
1271         for (i = 0; i < pf->num_alloc_vfs; i++) {
1272                 struct ice_vf *vf = &pf->vf[i];
1273
1274                 bool vf_mdd_detected = false;
1275
1276                 reg = rd32(hw, VP_MDET_TX_PQM(i));
1277                 if (reg & VP_MDET_TX_PQM_VALID_M) {
1278                         wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1279                         vf_mdd_detected = true;
1280                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1281                                  i);
1282                 }
1283
1284                 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1285                 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1286                         wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1287                         vf_mdd_detected = true;
1288                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1289                                  i);
1290                 }
1291
1292                 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1293                 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1294                         wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1295                         vf_mdd_detected = true;
1296                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
1297                                  i);
1298                 }
1299
1300                 reg = rd32(hw, VP_MDET_RX(i));
1301                 if (reg & VP_MDET_RX_VALID_M) {
1302                         wr32(hw, VP_MDET_RX(i), 0xFFFF);
1303                         vf_mdd_detected = true;
1304                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
1305                                  i);
1306                 }
1307
1308                 if (vf_mdd_detected) {
1309                         vf->num_mdd_events++;
1310                         if (vf->num_mdd_events > 1)
1311                                 dev_info(&pf->pdev->dev, "VF %d has had %llu MDD events since last boot\n",
1312                                          i, vf->num_mdd_events);
1313                 }
1314         }
1315 }
1316
1317 /**
1318  * ice_service_task - manage and run subtasks
1319  * @work: pointer to work_struct contained by the PF struct
1320  */
1321 static void ice_service_task(struct work_struct *work)
1322 {
1323         struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
1324         unsigned long start_time = jiffies;
1325
1326         /* subtasks */
1327
1328         /* process reset requests first */
1329         ice_reset_subtask(pf);
1330
1331         /* bail if a reset/recovery cycle is pending or rebuild failed */
1332         if (ice_is_reset_in_progress(pf->state) ||
1333             test_bit(__ICE_SUSPENDED, pf->state) ||
1334             test_bit(__ICE_NEEDS_RESTART, pf->state)) {
1335                 ice_service_task_complete(pf);
1336                 return;
1337         }
1338
1339         ice_check_for_hang_subtask(pf);
1340         ice_sync_fltr_subtask(pf);
1341         ice_handle_mdd_event(pf);
1342         ice_process_vflr_event(pf);
1343         ice_watchdog_subtask(pf);
1344         ice_clean_adminq_subtask(pf);
1345         ice_clean_mailboxq_subtask(pf);
1346
1347         /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
1348         ice_service_task_complete(pf);
1349
1350         /* If the tasks have taken longer than one service timer period
1351          * or there is more work to be done, reset the service timer to
1352          * schedule the service task now.
1353          */
1354         if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
1355             test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
1356             test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
1357             test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
1358             test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1359                 mod_timer(&pf->serv_tmr, jiffies);
1360 }
1361
1362 /**
1363  * ice_set_ctrlq_len - helper function to set controlq length
1364  * @hw: pointer to the HW instance
1365  */
1366 static void ice_set_ctrlq_len(struct ice_hw *hw)
1367 {
1368         hw->adminq.num_rq_entries = ICE_AQ_LEN;
1369         hw->adminq.num_sq_entries = ICE_AQ_LEN;
1370         hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
1371         hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
1372         hw->mailboxq.num_rq_entries = ICE_MBXQ_LEN;
1373         hw->mailboxq.num_sq_entries = ICE_MBXQ_LEN;
1374         hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1375         hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
1376 }
1377
1378 /**
1379  * ice_irq_affinity_notify - Callback for affinity changes
1380  * @notify: context as to what irq was changed
1381  * @mask: the new affinity mask
1382  *
1383  * This is a callback function used by the irq_set_affinity_notifier function
1384  * so that we may register to receive changes to the irq affinity masks.
1385  */
1386 static void
1387 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
1388                         const cpumask_t *mask)
1389 {
1390         struct ice_q_vector *q_vector =
1391                 container_of(notify, struct ice_q_vector, affinity_notify);
1392
1393         cpumask_copy(&q_vector->affinity_mask, mask);
1394 }
1395
1396 /**
1397  * ice_irq_affinity_release - Callback for affinity notifier release
1398  * @ref: internal core kernel usage
1399  *
1400  * This is a callback function used by the irq_set_affinity_notifier function
1401  * to inform the current notification subscriber that they will no longer
1402  * receive notifications.
1403  */
1404 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
1405
1406 /**
1407  * ice_vsi_ena_irq - Enable IRQ for the given VSI
1408  * @vsi: the VSI being configured
1409  */
1410 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
1411 {
1412         struct ice_pf *pf = vsi->back;
1413         struct ice_hw *hw = &pf->hw;
1414
1415         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
1416                 int i;
1417
1418                 ice_for_each_q_vector(vsi, i)
1419                         ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
1420         }
1421
1422         ice_flush(hw);
1423         return 0;
1424 }
1425
1426 /**
1427  * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
1428  * @vsi: the VSI being configured
1429  * @basename: name for the vector
1430  */
1431 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
1432 {
1433         int q_vectors = vsi->num_q_vectors;
1434         struct ice_pf *pf = vsi->back;
1435         int base = vsi->base_vector;
1436         int rx_int_idx = 0;
1437         int tx_int_idx = 0;
1438         int vector, err;
1439         int irq_num;
1440
1441         for (vector = 0; vector < q_vectors; vector++) {
1442                 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
1443
1444                 irq_num = pf->msix_entries[base + vector].vector;
1445
1446                 if (q_vector->tx.ring && q_vector->rx.ring) {
1447                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1448                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
1449                         tx_int_idx++;
1450                 } else if (q_vector->rx.ring) {
1451                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1452                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
1453                 } else if (q_vector->tx.ring) {
1454                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
1455                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
1456                 } else {
1457                         /* skip this unused q_vector */
1458                         continue;
1459                 }
1460                 err = devm_request_irq(&pf->pdev->dev, irq_num,
1461                                        vsi->irq_handler, 0,
1462                                        q_vector->name, q_vector);
1463                 if (err) {
1464                         netdev_err(vsi->netdev,
1465                                    "MSIX request_irq failed, error: %d\n", err);
1466                         goto free_q_irqs;
1467                 }
1468
1469                 /* register for affinity change notifications */
1470                 q_vector->affinity_notify.notify = ice_irq_affinity_notify;
1471                 q_vector->affinity_notify.release = ice_irq_affinity_release;
1472                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
1473
1474                 /* assign the mask for this irq */
1475                 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
1476         }
1477
1478         vsi->irqs_ready = true;
1479         return 0;
1480
1481 free_q_irqs:
1482         while (vector) {
1483                 vector--;
1484                 irq_num = pf->msix_entries[base + vector].vector,
1485                 irq_set_affinity_notifier(irq_num, NULL);
1486                 irq_set_affinity_hint(irq_num, NULL);
1487                 devm_free_irq(&pf->pdev->dev, irq_num, &vsi->q_vectors[vector]);
1488         }
1489         return err;
1490 }
1491
1492 /**
1493  * ice_ena_misc_vector - enable the non-queue interrupts
1494  * @pf: board private structure
1495  */
1496 static void ice_ena_misc_vector(struct ice_pf *pf)
1497 {
1498         struct ice_hw *hw = &pf->hw;
1499         u32 val;
1500
1501         /* clear things first */
1502         wr32(hw, PFINT_OICR_ENA, 0);    /* disable all */
1503         rd32(hw, PFINT_OICR);           /* read to clear */
1504
1505         val = (PFINT_OICR_ECC_ERR_M |
1506                PFINT_OICR_MAL_DETECT_M |
1507                PFINT_OICR_GRST_M |
1508                PFINT_OICR_PCI_EXCEPTION_M |
1509                PFINT_OICR_VFLR_M |
1510                PFINT_OICR_HMC_ERR_M |
1511                PFINT_OICR_PE_CRITERR_M);
1512
1513         wr32(hw, PFINT_OICR_ENA, val);
1514
1515         /* SW_ITR_IDX = 0, but don't change INTENA */
1516         wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
1517              GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
1518 }
1519
1520 /**
1521  * ice_misc_intr - misc interrupt handler
1522  * @irq: interrupt number
1523  * @data: pointer to a q_vector
1524  */
1525 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
1526 {
1527         struct ice_pf *pf = (struct ice_pf *)data;
1528         struct ice_hw *hw = &pf->hw;
1529         irqreturn_t ret = IRQ_NONE;
1530         u32 oicr, ena_mask;
1531
1532         set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1533         set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1534
1535         oicr = rd32(hw, PFINT_OICR);
1536         ena_mask = rd32(hw, PFINT_OICR_ENA);
1537
1538         if (oicr & PFINT_OICR_SWINT_M) {
1539                 ena_mask &= ~PFINT_OICR_SWINT_M;
1540                 pf->sw_int_count++;
1541         }
1542
1543         if (oicr & PFINT_OICR_MAL_DETECT_M) {
1544                 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
1545                 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
1546         }
1547         if (oicr & PFINT_OICR_VFLR_M) {
1548                 ena_mask &= ~PFINT_OICR_VFLR_M;
1549                 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
1550         }
1551
1552         if (oicr & PFINT_OICR_GRST_M) {
1553                 u32 reset;
1554
1555                 /* we have a reset warning */
1556                 ena_mask &= ~PFINT_OICR_GRST_M;
1557                 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
1558                         GLGEN_RSTAT_RESET_TYPE_S;
1559
1560                 if (reset == ICE_RESET_CORER)
1561                         pf->corer_count++;
1562                 else if (reset == ICE_RESET_GLOBR)
1563                         pf->globr_count++;
1564                 else if (reset == ICE_RESET_EMPR)
1565                         pf->empr_count++;
1566                 else
1567                         dev_dbg(&pf->pdev->dev, "Invalid reset type %d\n",
1568                                 reset);
1569
1570                 /* If a reset cycle isn't already in progress, we set a bit in
1571                  * pf->state so that the service task can start a reset/rebuild.
1572                  * We also make note of which reset happened so that peer
1573                  * devices/drivers can be informed.
1574                  */
1575                 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
1576                         if (reset == ICE_RESET_CORER)
1577                                 set_bit(__ICE_CORER_RECV, pf->state);
1578                         else if (reset == ICE_RESET_GLOBR)
1579                                 set_bit(__ICE_GLOBR_RECV, pf->state);
1580                         else
1581                                 set_bit(__ICE_EMPR_RECV, pf->state);
1582
1583                         /* There are couple of different bits at play here.
1584                          * hw->reset_ongoing indicates whether the hardware is
1585                          * in reset. This is set to true when a reset interrupt
1586                          * is received and set back to false after the driver
1587                          * has determined that the hardware is out of reset.
1588                          *
1589                          * __ICE_RESET_OICR_RECV in pf->state indicates
1590                          * that a post reset rebuild is required before the
1591                          * driver is operational again. This is set above.
1592                          *
1593                          * As this is the start of the reset/rebuild cycle, set
1594                          * both to indicate that.
1595                          */
1596                         hw->reset_ongoing = true;
1597                 }
1598         }
1599
1600         if (oicr & PFINT_OICR_HMC_ERR_M) {
1601                 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
1602                 dev_dbg(&pf->pdev->dev,
1603                         "HMC Error interrupt - info 0x%x, data 0x%x\n",
1604                         rd32(hw, PFHMC_ERRORINFO),
1605                         rd32(hw, PFHMC_ERRORDATA));
1606         }
1607
1608         /* Report any remaining unexpected interrupts */
1609         oicr &= ena_mask;
1610         if (oicr) {
1611                 dev_dbg(&pf->pdev->dev, "unhandled interrupt oicr=0x%08x\n",
1612                         oicr);
1613                 /* If a critical error is pending there is no choice but to
1614                  * reset the device.
1615                  */
1616                 if (oicr & (PFINT_OICR_PE_CRITERR_M |
1617                             PFINT_OICR_PCI_EXCEPTION_M |
1618                             PFINT_OICR_ECC_ERR_M)) {
1619                         set_bit(__ICE_PFR_REQ, pf->state);
1620                         ice_service_task_schedule(pf);
1621                 }
1622         }
1623         ret = IRQ_HANDLED;
1624
1625         if (!test_bit(__ICE_DOWN, pf->state)) {
1626                 ice_service_task_schedule(pf);
1627                 ice_irq_dynamic_ena(hw, NULL, NULL);
1628         }
1629
1630         return ret;
1631 }
1632
1633 /**
1634  * ice_dis_ctrlq_interrupts - disable control queue interrupts
1635  * @hw: pointer to HW structure
1636  */
1637 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
1638 {
1639         /* disable Admin queue Interrupt causes */
1640         wr32(hw, PFINT_FW_CTL,
1641              rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
1642
1643         /* disable Mailbox queue Interrupt causes */
1644         wr32(hw, PFINT_MBX_CTL,
1645              rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
1646
1647         /* disable Control queue Interrupt causes */
1648         wr32(hw, PFINT_OICR_CTL,
1649              rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
1650
1651         ice_flush(hw);
1652 }
1653
1654 /**
1655  * ice_free_irq_msix_misc - Unroll misc vector setup
1656  * @pf: board private structure
1657  */
1658 static void ice_free_irq_msix_misc(struct ice_pf *pf)
1659 {
1660         struct ice_hw *hw = &pf->hw;
1661
1662         ice_dis_ctrlq_interrupts(hw);
1663
1664         /* disable OICR interrupt */
1665         wr32(hw, PFINT_OICR_ENA, 0);
1666         ice_flush(hw);
1667
1668         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags) && pf->msix_entries) {
1669                 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
1670                 devm_free_irq(&pf->pdev->dev,
1671                               pf->msix_entries[pf->oicr_idx].vector, pf);
1672         }
1673
1674         pf->num_avail_sw_msix += 1;
1675         ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
1676 }
1677
1678 /**
1679  * ice_ena_ctrlq_interrupts - enable control queue interrupts
1680  * @hw: pointer to HW structure
1681  * @reg_idx: HW vector index to associate the control queue interrupts with
1682  */
1683 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
1684 {
1685         u32 val;
1686
1687         val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
1688                PFINT_OICR_CTL_CAUSE_ENA_M);
1689         wr32(hw, PFINT_OICR_CTL, val);
1690
1691         /* enable Admin queue Interrupt causes */
1692         val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
1693                PFINT_FW_CTL_CAUSE_ENA_M);
1694         wr32(hw, PFINT_FW_CTL, val);
1695
1696         /* enable Mailbox queue Interrupt causes */
1697         val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
1698                PFINT_MBX_CTL_CAUSE_ENA_M);
1699         wr32(hw, PFINT_MBX_CTL, val);
1700
1701         ice_flush(hw);
1702 }
1703
1704 /**
1705  * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
1706  * @pf: board private structure
1707  *
1708  * This sets up the handler for MSIX 0, which is used to manage the
1709  * non-queue interrupts, e.g. AdminQ and errors. This is not used
1710  * when in MSI or Legacy interrupt mode.
1711  */
1712 static int ice_req_irq_msix_misc(struct ice_pf *pf)
1713 {
1714         struct ice_hw *hw = &pf->hw;
1715         int oicr_idx, err = 0;
1716
1717         if (!pf->int_name[0])
1718                 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
1719                          dev_driver_string(&pf->pdev->dev),
1720                          dev_name(&pf->pdev->dev));
1721
1722         /* Do not request IRQ but do enable OICR interrupt since settings are
1723          * lost during reset. Note that this function is called only during
1724          * rebuild path and not while reset is in progress.
1725          */
1726         if (ice_is_reset_in_progress(pf->state))
1727                 goto skip_req_irq;
1728
1729         /* reserve one vector in irq_tracker for misc interrupts */
1730         oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1731         if (oicr_idx < 0)
1732                 return oicr_idx;
1733
1734         pf->num_avail_sw_msix -= 1;
1735         pf->oicr_idx = oicr_idx;
1736
1737         err = devm_request_irq(&pf->pdev->dev,
1738                                pf->msix_entries[pf->oicr_idx].vector,
1739                                ice_misc_intr, 0, pf->int_name, pf);
1740         if (err) {
1741                 dev_err(&pf->pdev->dev,
1742                         "devm_request_irq for %s failed: %d\n",
1743                         pf->int_name, err);
1744                 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
1745                 pf->num_avail_sw_msix += 1;
1746                 return err;
1747         }
1748
1749 skip_req_irq:
1750         ice_ena_misc_vector(pf);
1751
1752         ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
1753         wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
1754              ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
1755
1756         ice_flush(hw);
1757         ice_irq_dynamic_ena(hw, NULL, NULL);
1758
1759         return 0;
1760 }
1761
1762 /**
1763  * ice_napi_add - register NAPI handler for the VSI
1764  * @vsi: VSI for which NAPI handler is to be registered
1765  *
1766  * This function is only called in the driver's load path. Registering the NAPI
1767  * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
1768  * reset/rebuild, etc.)
1769  */
1770 static void ice_napi_add(struct ice_vsi *vsi)
1771 {
1772         int v_idx;
1773
1774         if (!vsi->netdev)
1775                 return;
1776
1777         ice_for_each_q_vector(vsi, v_idx)
1778                 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
1779                                ice_napi_poll, NAPI_POLL_WEIGHT);
1780 }
1781
1782 /**
1783  * ice_cfg_netdev - Allocate, configure and register a netdev
1784  * @vsi: the VSI associated with the new netdev
1785  *
1786  * Returns 0 on success, negative value on failure
1787  */
1788 static int ice_cfg_netdev(struct ice_vsi *vsi)
1789 {
1790         netdev_features_t csumo_features;
1791         netdev_features_t vlano_features;
1792         netdev_features_t dflt_features;
1793         netdev_features_t tso_features;
1794         struct ice_netdev_priv *np;
1795         struct net_device *netdev;
1796         u8 mac_addr[ETH_ALEN];
1797         int err;
1798
1799         netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
1800                                     vsi->alloc_rxq);
1801         if (!netdev)
1802                 return -ENOMEM;
1803
1804         vsi->netdev = netdev;
1805         np = netdev_priv(netdev);
1806         np->vsi = vsi;
1807
1808         dflt_features = NETIF_F_SG      |
1809                         NETIF_F_HIGHDMA |
1810                         NETIF_F_RXHASH;
1811
1812         csumo_features = NETIF_F_RXCSUM   |
1813                          NETIF_F_IP_CSUM  |
1814                          NETIF_F_SCTP_CRC |
1815                          NETIF_F_IPV6_CSUM;
1816
1817         vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
1818                          NETIF_F_HW_VLAN_CTAG_TX     |
1819                          NETIF_F_HW_VLAN_CTAG_RX;
1820
1821         tso_features = NETIF_F_TSO;
1822
1823         /* set features that user can change */
1824         netdev->hw_features = dflt_features | csumo_features |
1825                               vlano_features | tso_features;
1826
1827         /* enable features */
1828         netdev->features |= netdev->hw_features;
1829         /* encap and VLAN devices inherit default, csumo and tso features */
1830         netdev->hw_enc_features |= dflt_features | csumo_features |
1831                                    tso_features;
1832         netdev->vlan_features |= dflt_features | csumo_features |
1833                                  tso_features;
1834
1835         if (vsi->type == ICE_VSI_PF) {
1836                 SET_NETDEV_DEV(netdev, &vsi->back->pdev->dev);
1837                 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
1838
1839                 ether_addr_copy(netdev->dev_addr, mac_addr);
1840                 ether_addr_copy(netdev->perm_addr, mac_addr);
1841         }
1842
1843         netdev->priv_flags |= IFF_UNICAST_FLT;
1844
1845         /* assign netdev_ops */
1846         netdev->netdev_ops = &ice_netdev_ops;
1847
1848         /* setup watchdog timeout value to be 5 second */
1849         netdev->watchdog_timeo = 5 * HZ;
1850
1851         ice_set_ethtool_ops(netdev);
1852
1853         netdev->min_mtu = ETH_MIN_MTU;
1854         netdev->max_mtu = ICE_MAX_MTU;
1855
1856         err = register_netdev(vsi->netdev);
1857         if (err)
1858                 return err;
1859
1860         netif_carrier_off(vsi->netdev);
1861
1862         /* make sure transmit queues start off as stopped */
1863         netif_tx_stop_all_queues(vsi->netdev);
1864
1865         return 0;
1866 }
1867
1868 /**
1869  * ice_fill_rss_lut - Fill the RSS lookup table with default values
1870  * @lut: Lookup table
1871  * @rss_table_size: Lookup table size
1872  * @rss_size: Range of queue number for hashing
1873  */
1874 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
1875 {
1876         u16 i;
1877
1878         for (i = 0; i < rss_table_size; i++)
1879                 lut[i] = i % rss_size;
1880 }
1881
1882 /**
1883  * ice_pf_vsi_setup - Set up a PF VSI
1884  * @pf: board private structure
1885  * @pi: pointer to the port_info instance
1886  *
1887  * Returns pointer to the successfully allocated VSI software struct
1888  * on success, otherwise returns NULL on failure.
1889  */
1890 static struct ice_vsi *
1891 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
1892 {
1893         return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
1894 }
1895
1896 /**
1897  * ice_lb_vsi_setup - Set up a loopback VSI
1898  * @pf: board private structure
1899  * @pi: pointer to the port_info instance
1900  *
1901  * Returns pointer to the successfully allocated VSI software struct
1902  * on success, otherwise returns NULL on failure.
1903  */
1904 struct ice_vsi *
1905 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
1906 {
1907         return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
1908 }
1909
1910 /**
1911  * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
1912  * @netdev: network interface to be adjusted
1913  * @proto: unused protocol
1914  * @vid: VLAN ID to be added
1915  *
1916  * net_device_ops implementation for adding VLAN IDs
1917  */
1918 static int
1919 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
1920                     u16 vid)
1921 {
1922         struct ice_netdev_priv *np = netdev_priv(netdev);
1923         struct ice_vsi *vsi = np->vsi;
1924         int ret;
1925
1926         if (vid >= VLAN_N_VID) {
1927                 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
1928                            vid, VLAN_N_VID);
1929                 return -EINVAL;
1930         }
1931
1932         if (vsi->info.pvid)
1933                 return -EINVAL;
1934
1935         /* Enable VLAN pruning when VLAN 0 is added */
1936         if (unlikely(!vid)) {
1937                 ret = ice_cfg_vlan_pruning(vsi, true, false);
1938                 if (ret)
1939                         return ret;
1940         }
1941
1942         /* Add all VLAN IDs including 0 to the switch filter. VLAN ID 0 is
1943          * needed to continue allowing all untagged packets since VLAN prune
1944          * list is applied to all packets by the switch
1945          */
1946         ret = ice_vsi_add_vlan(vsi, vid);
1947         if (!ret) {
1948                 vsi->vlan_ena = true;
1949                 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
1950         }
1951
1952         return ret;
1953 }
1954
1955 /**
1956  * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
1957  * @netdev: network interface to be adjusted
1958  * @proto: unused protocol
1959  * @vid: VLAN ID to be removed
1960  *
1961  * net_device_ops implementation for removing VLAN IDs
1962  */
1963 static int
1964 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
1965                      u16 vid)
1966 {
1967         struct ice_netdev_priv *np = netdev_priv(netdev);
1968         struct ice_vsi *vsi = np->vsi;
1969         int ret;
1970
1971         if (vsi->info.pvid)
1972                 return -EINVAL;
1973
1974         /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
1975          * information
1976          */
1977         ret = ice_vsi_kill_vlan(vsi, vid);
1978         if (ret)
1979                 return ret;
1980
1981         /* Disable VLAN pruning when VLAN 0 is removed */
1982         if (unlikely(!vid))
1983                 ret = ice_cfg_vlan_pruning(vsi, false, false);
1984
1985         vsi->vlan_ena = false;
1986         set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
1987         return ret;
1988 }
1989
1990 /**
1991  * ice_setup_pf_sw - Setup the HW switch on startup or after reset
1992  * @pf: board private structure
1993  *
1994  * Returns 0 on success, negative value on failure
1995  */
1996 static int ice_setup_pf_sw(struct ice_pf *pf)
1997 {
1998         struct ice_vsi *vsi;
1999         int status = 0;
2000
2001         if (ice_is_reset_in_progress(pf->state))
2002                 return -EBUSY;
2003
2004         vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
2005         if (!vsi) {
2006                 status = -ENOMEM;
2007                 goto unroll_vsi_setup;
2008         }
2009
2010         status = ice_cfg_netdev(vsi);
2011         if (status) {
2012                 status = -ENODEV;
2013                 goto unroll_vsi_setup;
2014         }
2015
2016         /* registering the NAPI handler requires both the queues and
2017          * netdev to be created, which are done in ice_pf_vsi_setup()
2018          * and ice_cfg_netdev() respectively
2019          */
2020         ice_napi_add(vsi);
2021
2022         status = ice_init_mac_fltr(pf);
2023         if (status)
2024                 goto unroll_napi_add;
2025
2026         return status;
2027
2028 unroll_napi_add:
2029         if (vsi) {
2030                 ice_napi_del(vsi);
2031                 if (vsi->netdev) {
2032                         if (vsi->netdev->reg_state == NETREG_REGISTERED)
2033                                 unregister_netdev(vsi->netdev);
2034                         free_netdev(vsi->netdev);
2035                         vsi->netdev = NULL;
2036                 }
2037         }
2038
2039 unroll_vsi_setup:
2040         if (vsi) {
2041                 ice_vsi_free_q_vectors(vsi);
2042                 ice_vsi_delete(vsi);
2043                 ice_vsi_put_qs(vsi);
2044                 pf->q_left_tx += vsi->alloc_txq;
2045                 pf->q_left_rx += vsi->alloc_rxq;
2046                 ice_vsi_clear(vsi);
2047         }
2048         return status;
2049 }
2050
2051 /**
2052  * ice_determine_q_usage - Calculate queue distribution
2053  * @pf: board private structure
2054  *
2055  * Return -ENOMEM if we don't get enough queues for all ports
2056  */
2057 static void ice_determine_q_usage(struct ice_pf *pf)
2058 {
2059         u16 q_left_tx, q_left_rx;
2060
2061         q_left_tx = pf->hw.func_caps.common_cap.num_txq;
2062         q_left_rx = pf->hw.func_caps.common_cap.num_rxq;
2063
2064         pf->num_lan_tx = min_t(int, q_left_tx, num_online_cpus());
2065
2066         /* only 1 Rx queue unless RSS is enabled */
2067         if (!test_bit(ICE_FLAG_RSS_ENA, pf->flags))
2068                 pf->num_lan_rx = 1;
2069         else
2070                 pf->num_lan_rx = min_t(int, q_left_rx, num_online_cpus());
2071
2072         pf->q_left_tx = q_left_tx - pf->num_lan_tx;
2073         pf->q_left_rx = q_left_rx - pf->num_lan_rx;
2074 }
2075
2076 /**
2077  * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
2078  * @pf: board private structure to initialize
2079  */
2080 static void ice_deinit_pf(struct ice_pf *pf)
2081 {
2082         ice_service_task_stop(pf);
2083         mutex_destroy(&pf->sw_mutex);
2084         mutex_destroy(&pf->avail_q_mutex);
2085 }
2086
2087 /**
2088  * ice_init_pf - Initialize general software structures (struct ice_pf)
2089  * @pf: board private structure to initialize
2090  */
2091 static void ice_init_pf(struct ice_pf *pf)
2092 {
2093         bitmap_zero(pf->flags, ICE_PF_FLAGS_NBITS);
2094         set_bit(ICE_FLAG_MSIX_ENA, pf->flags);
2095 #ifdef CONFIG_PCI_IOV
2096         if (pf->hw.func_caps.common_cap.sr_iov_1_1) {
2097                 struct ice_hw *hw = &pf->hw;
2098
2099                 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
2100                 pf->num_vfs_supported = min_t(int, hw->func_caps.num_allocd_vfs,
2101                                               ICE_MAX_VF_COUNT);
2102         }
2103 #endif /* CONFIG_PCI_IOV */
2104
2105         mutex_init(&pf->sw_mutex);
2106         mutex_init(&pf->avail_q_mutex);
2107
2108         /* Clear avail_[t|r]x_qs bitmaps (set all to avail) */
2109         mutex_lock(&pf->avail_q_mutex);
2110         bitmap_zero(pf->avail_txqs, ICE_MAX_TXQS);
2111         bitmap_zero(pf->avail_rxqs, ICE_MAX_RXQS);
2112         mutex_unlock(&pf->avail_q_mutex);
2113
2114         if (pf->hw.func_caps.common_cap.rss_table_size)
2115                 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
2116
2117         /* setup service timer and periodic service task */
2118         timer_setup(&pf->serv_tmr, ice_service_timer, 0);
2119         pf->serv_tmr_period = HZ;
2120         INIT_WORK(&pf->serv_task, ice_service_task);
2121         clear_bit(__ICE_SERVICE_SCHED, pf->state);
2122 }
2123
2124 /**
2125  * ice_ena_msix_range - Request a range of MSIX vectors from the OS
2126  * @pf: board private structure
2127  *
2128  * compute the number of MSIX vectors required (v_budget) and request from
2129  * the OS. Return the number of vectors reserved or negative on failure
2130  */
2131 static int ice_ena_msix_range(struct ice_pf *pf)
2132 {
2133         int v_left, v_actual, v_budget = 0;
2134         int needed, err, i;
2135
2136         v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
2137
2138         /* reserve one vector for miscellaneous handler */
2139         needed = 1;
2140         v_budget += needed;
2141         v_left -= needed;
2142
2143         /* reserve vectors for LAN traffic */
2144         pf->num_lan_msix = min_t(int, num_online_cpus(), v_left);
2145         v_budget += pf->num_lan_msix;
2146         v_left -= pf->num_lan_msix;
2147
2148         pf->msix_entries = devm_kcalloc(&pf->pdev->dev, v_budget,
2149                                         sizeof(*pf->msix_entries), GFP_KERNEL);
2150
2151         if (!pf->msix_entries) {
2152                 err = -ENOMEM;
2153                 goto exit_err;
2154         }
2155
2156         for (i = 0; i < v_budget; i++)
2157                 pf->msix_entries[i].entry = i;
2158
2159         /* actually reserve the vectors */
2160         v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
2161                                          ICE_MIN_MSIX, v_budget);
2162
2163         if (v_actual < 0) {
2164                 dev_err(&pf->pdev->dev, "unable to reserve MSI-X vectors\n");
2165                 err = v_actual;
2166                 goto msix_err;
2167         }
2168
2169         if (v_actual < v_budget) {
2170                 dev_warn(&pf->pdev->dev,
2171                          "not enough vectors. requested = %d, obtained = %d\n",
2172                          v_budget, v_actual);
2173                 if (v_actual >= (pf->num_lan_msix + 1)) {
2174                         pf->num_avail_sw_msix = v_actual -
2175                                                 (pf->num_lan_msix + 1);
2176                 } else if (v_actual >= 2) {
2177                         pf->num_lan_msix = 1;
2178                         pf->num_avail_sw_msix = v_actual - 2;
2179                 } else {
2180                         pci_disable_msix(pf->pdev);
2181                         err = -ERANGE;
2182                         goto msix_err;
2183                 }
2184         }
2185
2186         return v_actual;
2187
2188 msix_err:
2189         devm_kfree(&pf->pdev->dev, pf->msix_entries);
2190         goto exit_err;
2191
2192 exit_err:
2193         pf->num_lan_msix = 0;
2194         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
2195         return err;
2196 }
2197
2198 /**
2199  * ice_dis_msix - Disable MSI-X interrupt setup in OS
2200  * @pf: board private structure
2201  */
2202 static void ice_dis_msix(struct ice_pf *pf)
2203 {
2204         pci_disable_msix(pf->pdev);
2205         devm_kfree(&pf->pdev->dev, pf->msix_entries);
2206         pf->msix_entries = NULL;
2207         clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
2208 }
2209
2210 /**
2211  * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
2212  * @pf: board private structure
2213  */
2214 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
2215 {
2216         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
2217                 ice_dis_msix(pf);
2218
2219         if (pf->irq_tracker) {
2220                 devm_kfree(&pf->pdev->dev, pf->irq_tracker);
2221                 pf->irq_tracker = NULL;
2222         }
2223 }
2224
2225 /**
2226  * ice_init_interrupt_scheme - Determine proper interrupt scheme
2227  * @pf: board private structure to initialize
2228  */
2229 static int ice_init_interrupt_scheme(struct ice_pf *pf)
2230 {
2231         int vectors;
2232
2233         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
2234                 vectors = ice_ena_msix_range(pf);
2235         else
2236                 return -ENODEV;
2237
2238         if (vectors < 0)
2239                 return vectors;
2240
2241         /* set up vector assignment tracking */
2242         pf->irq_tracker =
2243                 devm_kzalloc(&pf->pdev->dev, sizeof(*pf->irq_tracker) +
2244                              (sizeof(u16) * vectors), GFP_KERNEL);
2245         if (!pf->irq_tracker) {
2246                 ice_dis_msix(pf);
2247                 return -ENOMEM;
2248         }
2249
2250         /* populate SW interrupts pool with number of OS granted IRQs. */
2251         pf->num_avail_sw_msix = vectors;
2252         pf->irq_tracker->num_entries = vectors;
2253         pf->irq_tracker->end = pf->irq_tracker->num_entries;
2254
2255         return 0;
2256 }
2257
2258 /**
2259  * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
2260  * @pf: pointer to the PF structure
2261  *
2262  * There is no error returned here because the driver should be able to handle
2263  * 128 Byte cache lines, so we only print a warning in case issues are seen,
2264  * specifically with Tx.
2265  */
2266 static void ice_verify_cacheline_size(struct ice_pf *pf)
2267 {
2268         if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
2269                 dev_warn(&pf->pdev->dev,
2270                          "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
2271                          ICE_CACHE_LINE_BYTES);
2272 }
2273
2274 /**
2275  * ice_probe - Device initialization routine
2276  * @pdev: PCI device information struct
2277  * @ent: entry in ice_pci_tbl
2278  *
2279  * Returns 0 on success, negative on failure
2280  */
2281 static int
2282 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
2283 {
2284         struct device *dev = &pdev->dev;
2285         struct ice_pf *pf;
2286         struct ice_hw *hw;
2287         int err;
2288
2289         /* this driver uses devres, see Documentation/driver-model/devres.rst */
2290         err = pcim_enable_device(pdev);
2291         if (err)
2292                 return err;
2293
2294         err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
2295         if (err) {
2296                 dev_err(dev, "BAR0 I/O map error %d\n", err);
2297                 return err;
2298         }
2299
2300         pf = devm_kzalloc(dev, sizeof(*pf), GFP_KERNEL);
2301         if (!pf)
2302                 return -ENOMEM;
2303
2304         /* set up for high or low DMA */
2305         err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
2306         if (err)
2307                 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
2308         if (err) {
2309                 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
2310                 return err;
2311         }
2312
2313         pci_enable_pcie_error_reporting(pdev);
2314         pci_set_master(pdev);
2315
2316         pf->pdev = pdev;
2317         pci_set_drvdata(pdev, pf);
2318         set_bit(__ICE_DOWN, pf->state);
2319         /* Disable service task until DOWN bit is cleared */
2320         set_bit(__ICE_SERVICE_DIS, pf->state);
2321
2322         hw = &pf->hw;
2323         hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
2324         hw->back = pf;
2325         hw->vendor_id = pdev->vendor;
2326         hw->device_id = pdev->device;
2327         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
2328         hw->subsystem_vendor_id = pdev->subsystem_vendor;
2329         hw->subsystem_device_id = pdev->subsystem_device;
2330         hw->bus.device = PCI_SLOT(pdev->devfn);
2331         hw->bus.func = PCI_FUNC(pdev->devfn);
2332         ice_set_ctrlq_len(hw);
2333
2334         pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
2335
2336 #ifndef CONFIG_DYNAMIC_DEBUG
2337         if (debug < -1)
2338                 hw->debug_mask = debug;
2339 #endif
2340
2341         err = ice_init_hw(hw);
2342         if (err) {
2343                 dev_err(dev, "ice_init_hw failed: %d\n", err);
2344                 err = -EIO;
2345                 goto err_exit_unroll;
2346         }
2347
2348         dev_info(dev, "firmware %d.%d.%05d api %d.%d\n",
2349                  hw->fw_maj_ver, hw->fw_min_ver, hw->fw_build,
2350                  hw->api_maj_ver, hw->api_min_ver);
2351
2352         ice_init_pf(pf);
2353
2354         err = ice_init_pf_dcb(pf, false);
2355         if (err) {
2356                 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
2357                 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
2358
2359                 /* do not fail overall init if DCB init fails */
2360                 err = 0;
2361         }
2362
2363         ice_determine_q_usage(pf);
2364
2365         pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
2366         if (!pf->num_alloc_vsi) {
2367                 err = -EIO;
2368                 goto err_init_pf_unroll;
2369         }
2370
2371         pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
2372                                GFP_KERNEL);
2373         if (!pf->vsi) {
2374                 err = -ENOMEM;
2375                 goto err_init_pf_unroll;
2376         }
2377
2378         err = ice_init_interrupt_scheme(pf);
2379         if (err) {
2380                 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
2381                 err = -EIO;
2382                 goto err_init_interrupt_unroll;
2383         }
2384
2385         /* Driver is mostly up */
2386         clear_bit(__ICE_DOWN, pf->state);
2387
2388         /* In case of MSIX we are going to setup the misc vector right here
2389          * to handle admin queue events etc. In case of legacy and MSI
2390          * the misc functionality and queue processing is combined in
2391          * the same vector and that gets setup at open.
2392          */
2393         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
2394                 err = ice_req_irq_msix_misc(pf);
2395                 if (err) {
2396                         dev_err(dev, "setup of misc vector failed: %d\n", err);
2397                         goto err_init_interrupt_unroll;
2398                 }
2399         }
2400
2401         /* create switch struct for the switch element created by FW on boot */
2402         pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
2403         if (!pf->first_sw) {
2404                 err = -ENOMEM;
2405                 goto err_msix_misc_unroll;
2406         }
2407
2408         if (hw->evb_veb)
2409                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
2410         else
2411                 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
2412
2413         pf->first_sw->pf = pf;
2414
2415         /* record the sw_id available for later use */
2416         pf->first_sw->sw_id = hw->port_info->sw_id;
2417
2418         err = ice_setup_pf_sw(pf);
2419         if (err) {
2420                 dev_err(dev, "probe failed due to setup PF switch:%d\n", err);
2421                 goto err_alloc_sw_unroll;
2422         }
2423
2424         clear_bit(__ICE_SERVICE_DIS, pf->state);
2425
2426         /* since everything is good, start the service timer */
2427         mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
2428
2429         err = ice_init_link_events(pf->hw.port_info);
2430         if (err) {
2431                 dev_err(dev, "ice_init_link_events failed: %d\n", err);
2432                 goto err_alloc_sw_unroll;
2433         }
2434
2435         ice_verify_cacheline_size(pf);
2436
2437         return 0;
2438
2439 err_alloc_sw_unroll:
2440         set_bit(__ICE_SERVICE_DIS, pf->state);
2441         set_bit(__ICE_DOWN, pf->state);
2442         devm_kfree(&pf->pdev->dev, pf->first_sw);
2443 err_msix_misc_unroll:
2444         ice_free_irq_msix_misc(pf);
2445 err_init_interrupt_unroll:
2446         ice_clear_interrupt_scheme(pf);
2447         devm_kfree(dev, pf->vsi);
2448 err_init_pf_unroll:
2449         ice_deinit_pf(pf);
2450         ice_deinit_hw(hw);
2451 err_exit_unroll:
2452         pci_disable_pcie_error_reporting(pdev);
2453         return err;
2454 }
2455
2456 /**
2457  * ice_remove - Device removal routine
2458  * @pdev: PCI device information struct
2459  */
2460 static void ice_remove(struct pci_dev *pdev)
2461 {
2462         struct ice_pf *pf = pci_get_drvdata(pdev);
2463         int i;
2464
2465         if (!pf)
2466                 return;
2467
2468         for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
2469                 if (!ice_is_reset_in_progress(pf->state))
2470                         break;
2471                 msleep(100);
2472         }
2473
2474         set_bit(__ICE_DOWN, pf->state);
2475         ice_service_task_stop(pf);
2476
2477         if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags))
2478                 ice_free_vfs(pf);
2479         ice_vsi_release_all(pf);
2480         ice_free_irq_msix_misc(pf);
2481         ice_for_each_vsi(pf, i) {
2482                 if (!pf->vsi[i])
2483                         continue;
2484                 ice_vsi_free_q_vectors(pf->vsi[i]);
2485         }
2486         ice_clear_interrupt_scheme(pf);
2487         ice_deinit_pf(pf);
2488         ice_deinit_hw(&pf->hw);
2489         pci_disable_pcie_error_reporting(pdev);
2490 }
2491
2492 /**
2493  * ice_pci_err_detected - warning that PCI error has been detected
2494  * @pdev: PCI device information struct
2495  * @err: the type of PCI error
2496  *
2497  * Called to warn that something happened on the PCI bus and the error handling
2498  * is in progress.  Allows the driver to gracefully prepare/handle PCI errors.
2499  */
2500 static pci_ers_result_t
2501 ice_pci_err_detected(struct pci_dev *pdev, enum pci_channel_state err)
2502 {
2503         struct ice_pf *pf = pci_get_drvdata(pdev);
2504
2505         if (!pf) {
2506                 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
2507                         __func__, err);
2508                 return PCI_ERS_RESULT_DISCONNECT;
2509         }
2510
2511         if (!test_bit(__ICE_SUSPENDED, pf->state)) {
2512                 ice_service_task_stop(pf);
2513
2514                 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
2515                         set_bit(__ICE_PFR_REQ, pf->state);
2516                         ice_prepare_for_reset(pf);
2517                 }
2518         }
2519
2520         return PCI_ERS_RESULT_NEED_RESET;
2521 }
2522
2523 /**
2524  * ice_pci_err_slot_reset - a PCI slot reset has just happened
2525  * @pdev: PCI device information struct
2526  *
2527  * Called to determine if the driver can recover from the PCI slot reset by
2528  * using a register read to determine if the device is recoverable.
2529  */
2530 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
2531 {
2532         struct ice_pf *pf = pci_get_drvdata(pdev);
2533         pci_ers_result_t result;
2534         int err;
2535         u32 reg;
2536
2537         err = pci_enable_device_mem(pdev);
2538         if (err) {
2539                 dev_err(&pdev->dev,
2540                         "Cannot re-enable PCI device after reset, error %d\n",
2541                         err);
2542                 result = PCI_ERS_RESULT_DISCONNECT;
2543         } else {
2544                 pci_set_master(pdev);
2545                 pci_restore_state(pdev);
2546                 pci_save_state(pdev);
2547                 pci_wake_from_d3(pdev, false);
2548
2549                 /* Check for life */
2550                 reg = rd32(&pf->hw, GLGEN_RTRIG);
2551                 if (!reg)
2552                         result = PCI_ERS_RESULT_RECOVERED;
2553                 else
2554                         result = PCI_ERS_RESULT_DISCONNECT;
2555         }
2556
2557         err = pci_cleanup_aer_uncorrect_error_status(pdev);
2558         if (err)
2559                 dev_dbg(&pdev->dev,
2560                         "pci_cleanup_aer_uncorrect_error_status failed, error %d\n",
2561                         err);
2562                 /* non-fatal, continue */
2563
2564         return result;
2565 }
2566
2567 /**
2568  * ice_pci_err_resume - restart operations after PCI error recovery
2569  * @pdev: PCI device information struct
2570  *
2571  * Called to allow the driver to bring things back up after PCI error and/or
2572  * reset recovery have finished
2573  */
2574 static void ice_pci_err_resume(struct pci_dev *pdev)
2575 {
2576         struct ice_pf *pf = pci_get_drvdata(pdev);
2577
2578         if (!pf) {
2579                 dev_err(&pdev->dev,
2580                         "%s failed, device is unrecoverable\n", __func__);
2581                 return;
2582         }
2583
2584         if (test_bit(__ICE_SUSPENDED, pf->state)) {
2585                 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
2586                         __func__);
2587                 return;
2588         }
2589
2590         ice_do_reset(pf, ICE_RESET_PFR);
2591         ice_service_task_restart(pf);
2592         mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
2593 }
2594
2595 /**
2596  * ice_pci_err_reset_prepare - prepare device driver for PCI reset
2597  * @pdev: PCI device information struct
2598  */
2599 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
2600 {
2601         struct ice_pf *pf = pci_get_drvdata(pdev);
2602
2603         if (!test_bit(__ICE_SUSPENDED, pf->state)) {
2604                 ice_service_task_stop(pf);
2605
2606                 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
2607                         set_bit(__ICE_PFR_REQ, pf->state);
2608                         ice_prepare_for_reset(pf);
2609                 }
2610         }
2611 }
2612
2613 /**
2614  * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
2615  * @pdev: PCI device information struct
2616  */
2617 static void ice_pci_err_reset_done(struct pci_dev *pdev)
2618 {
2619         ice_pci_err_resume(pdev);
2620 }
2621
2622 /* ice_pci_tbl - PCI Device ID Table
2623  *
2624  * Wildcard entries (PCI_ANY_ID) should come last
2625  * Last entry must be all 0s
2626  *
2627  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
2628  *   Class, Class Mask, private data (not used) }
2629  */
2630 static const struct pci_device_id ice_pci_tbl[] = {
2631         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
2632         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
2633         { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
2634         /* required last entry */
2635         { 0, }
2636 };
2637 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
2638
2639 static const struct pci_error_handlers ice_pci_err_handler = {
2640         .error_detected = ice_pci_err_detected,
2641         .slot_reset = ice_pci_err_slot_reset,
2642         .reset_prepare = ice_pci_err_reset_prepare,
2643         .reset_done = ice_pci_err_reset_done,
2644         .resume = ice_pci_err_resume
2645 };
2646
2647 static struct pci_driver ice_driver = {
2648         .name = KBUILD_MODNAME,
2649         .id_table = ice_pci_tbl,
2650         .probe = ice_probe,
2651         .remove = ice_remove,
2652         .sriov_configure = ice_sriov_configure,
2653         .err_handler = &ice_pci_err_handler
2654 };
2655
2656 /**
2657  * ice_module_init - Driver registration routine
2658  *
2659  * ice_module_init is the first routine called when the driver is
2660  * loaded. All it does is register with the PCI subsystem.
2661  */
2662 static int __init ice_module_init(void)
2663 {
2664         int status;
2665
2666         pr_info("%s - version %s\n", ice_driver_string, ice_drv_ver);
2667         pr_info("%s\n", ice_copyright);
2668
2669         ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
2670         if (!ice_wq) {
2671                 pr_err("Failed to create workqueue\n");
2672                 return -ENOMEM;
2673         }
2674
2675         status = pci_register_driver(&ice_driver);
2676         if (status) {
2677                 pr_err("failed to register PCI driver, err %d\n", status);
2678                 destroy_workqueue(ice_wq);
2679         }
2680
2681         return status;
2682 }
2683 module_init(ice_module_init);
2684
2685 /**
2686  * ice_module_exit - Driver exit cleanup routine
2687  *
2688  * ice_module_exit is called just before the driver is removed
2689  * from memory.
2690  */
2691 static void __exit ice_module_exit(void)
2692 {
2693         pci_unregister_driver(&ice_driver);
2694         destroy_workqueue(ice_wq);
2695         pr_info("module unloaded\n");
2696 }
2697 module_exit(ice_module_exit);
2698
2699 /**
2700  * ice_set_mac_address - NDO callback to set MAC address
2701  * @netdev: network interface device structure
2702  * @pi: pointer to an address structure
2703  *
2704  * Returns 0 on success, negative on failure
2705  */
2706 static int ice_set_mac_address(struct net_device *netdev, void *pi)
2707 {
2708         struct ice_netdev_priv *np = netdev_priv(netdev);
2709         struct ice_vsi *vsi = np->vsi;
2710         struct ice_pf *pf = vsi->back;
2711         struct ice_hw *hw = &pf->hw;
2712         struct sockaddr *addr = pi;
2713         enum ice_status status;
2714         LIST_HEAD(a_mac_list);
2715         LIST_HEAD(r_mac_list);
2716         u8 flags = 0;
2717         int err;
2718         u8 *mac;
2719
2720         mac = (u8 *)addr->sa_data;
2721
2722         if (!is_valid_ether_addr(mac))
2723                 return -EADDRNOTAVAIL;
2724
2725         if (ether_addr_equal(netdev->dev_addr, mac)) {
2726                 netdev_warn(netdev, "already using mac %pM\n", mac);
2727                 return 0;
2728         }
2729
2730         if (test_bit(__ICE_DOWN, pf->state) ||
2731             ice_is_reset_in_progress(pf->state)) {
2732                 netdev_err(netdev, "can't set mac %pM. device not ready\n",
2733                            mac);
2734                 return -EBUSY;
2735         }
2736
2737         /* When we change the MAC address we also have to change the MAC address
2738          * based filter rules that were created previously for the old MAC
2739          * address. So first, we remove the old filter rule using ice_remove_mac
2740          * and then create a new filter rule using ice_add_mac. Note that for
2741          * both these operations, we first need to form a "list" of MAC
2742          * addresses (even though in this case, we have only 1 MAC address to be
2743          * added/removed) and this done using ice_add_mac_to_list. Depending on
2744          * the ensuing operation this "list" of MAC addresses is either to be
2745          * added or removed from the filter.
2746          */
2747         err = ice_add_mac_to_list(vsi, &r_mac_list, netdev->dev_addr);
2748         if (err) {
2749                 err = -EADDRNOTAVAIL;
2750                 goto free_lists;
2751         }
2752
2753         status = ice_remove_mac(hw, &r_mac_list);
2754         if (status) {
2755                 err = -EADDRNOTAVAIL;
2756                 goto free_lists;
2757         }
2758
2759         err = ice_add_mac_to_list(vsi, &a_mac_list, mac);
2760         if (err) {
2761                 err = -EADDRNOTAVAIL;
2762                 goto free_lists;
2763         }
2764
2765         status = ice_add_mac(hw, &a_mac_list);
2766         if (status) {
2767                 err = -EADDRNOTAVAIL;
2768                 goto free_lists;
2769         }
2770
2771 free_lists:
2772         /* free list entries */
2773         ice_free_fltr_list(&pf->pdev->dev, &r_mac_list);
2774         ice_free_fltr_list(&pf->pdev->dev, &a_mac_list);
2775
2776         if (err) {
2777                 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
2778                            mac);
2779                 return err;
2780         }
2781
2782         /* change the netdev's MAC address */
2783         memcpy(netdev->dev_addr, mac, netdev->addr_len);
2784         netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
2785                    netdev->dev_addr);
2786
2787         /* write new MAC address to the firmware */
2788         flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
2789         status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
2790         if (status) {
2791                 netdev_err(netdev, "can't set MAC %pM. write to firmware failed.\n",
2792                            mac);
2793         }
2794         return 0;
2795 }
2796
2797 /**
2798  * ice_set_rx_mode - NDO callback to set the netdev filters
2799  * @netdev: network interface device structure
2800  */
2801 static void ice_set_rx_mode(struct net_device *netdev)
2802 {
2803         struct ice_netdev_priv *np = netdev_priv(netdev);
2804         struct ice_vsi *vsi = np->vsi;
2805
2806         if (!vsi)
2807                 return;
2808
2809         /* Set the flags to synchronize filters
2810          * ndo_set_rx_mode may be triggered even without a change in netdev
2811          * flags
2812          */
2813         set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
2814         set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
2815         set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
2816
2817         /* schedule our worker thread which will take care of
2818          * applying the new filter changes
2819          */
2820         ice_service_task_schedule(vsi->back);
2821 }
2822
2823 /**
2824  * ice_fdb_add - add an entry to the hardware database
2825  * @ndm: the input from the stack
2826  * @tb: pointer to array of nladdr (unused)
2827  * @dev: the net device pointer
2828  * @addr: the MAC address entry being added
2829  * @vid: VLAN ID
2830  * @flags: instructions from stack about fdb operation
2831  * @extack: netlink extended ack
2832  */
2833 static int
2834 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
2835             struct net_device *dev, const unsigned char *addr, u16 vid,
2836             u16 flags, struct netlink_ext_ack __always_unused *extack)
2837 {
2838         int err;
2839
2840         if (vid) {
2841                 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
2842                 return -EINVAL;
2843         }
2844         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
2845                 netdev_err(dev, "FDB only supports static addresses\n");
2846                 return -EINVAL;
2847         }
2848
2849         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
2850                 err = dev_uc_add_excl(dev, addr);
2851         else if (is_multicast_ether_addr(addr))
2852                 err = dev_mc_add_excl(dev, addr);
2853         else
2854                 err = -EINVAL;
2855
2856         /* Only return duplicate errors if NLM_F_EXCL is set */
2857         if (err == -EEXIST && !(flags & NLM_F_EXCL))
2858                 err = 0;
2859
2860         return err;
2861 }
2862
2863 /**
2864  * ice_fdb_del - delete an entry from the hardware database
2865  * @ndm: the input from the stack
2866  * @tb: pointer to array of nladdr (unused)
2867  * @dev: the net device pointer
2868  * @addr: the MAC address entry being added
2869  * @vid: VLAN ID
2870  */
2871 static int
2872 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
2873             struct net_device *dev, const unsigned char *addr,
2874             __always_unused u16 vid)
2875 {
2876         int err;
2877
2878         if (ndm->ndm_state & NUD_PERMANENT) {
2879                 netdev_err(dev, "FDB only supports static addresses\n");
2880                 return -EINVAL;
2881         }
2882
2883         if (is_unicast_ether_addr(addr))
2884                 err = dev_uc_del(dev, addr);
2885         else if (is_multicast_ether_addr(addr))
2886                 err = dev_mc_del(dev, addr);
2887         else
2888                 err = -EINVAL;
2889
2890         return err;
2891 }
2892
2893 /**
2894  * ice_set_features - set the netdev feature flags
2895  * @netdev: ptr to the netdev being adjusted
2896  * @features: the feature set that the stack is suggesting
2897  */
2898 static int
2899 ice_set_features(struct net_device *netdev, netdev_features_t features)
2900 {
2901         struct ice_netdev_priv *np = netdev_priv(netdev);
2902         struct ice_vsi *vsi = np->vsi;
2903         int ret = 0;
2904
2905         /* Multiple features can be changed in one call so keep features in
2906          * separate if/else statements to guarantee each feature is checked
2907          */
2908         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
2909                 ret = ice_vsi_manage_rss_lut(vsi, true);
2910         else if (!(features & NETIF_F_RXHASH) &&
2911                  netdev->features & NETIF_F_RXHASH)
2912                 ret = ice_vsi_manage_rss_lut(vsi, false);
2913
2914         if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
2915             !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2916                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2917         else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
2918                  (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
2919                 ret = ice_vsi_manage_vlan_stripping(vsi, false);
2920
2921         if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
2922             !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2923                 ret = ice_vsi_manage_vlan_insertion(vsi);
2924         else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
2925                  (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
2926                 ret = ice_vsi_manage_vlan_insertion(vsi);
2927
2928         if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
2929             !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
2930                 ret = ice_cfg_vlan_pruning(vsi, true, false);
2931         else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
2932                  (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
2933                 ret = ice_cfg_vlan_pruning(vsi, false, false);
2934
2935         return ret;
2936 }
2937
2938 /**
2939  * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
2940  * @vsi: VSI to setup VLAN properties for
2941  */
2942 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
2943 {
2944         int ret = 0;
2945
2946         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
2947                 ret = ice_vsi_manage_vlan_stripping(vsi, true);
2948         if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
2949                 ret = ice_vsi_manage_vlan_insertion(vsi);
2950
2951         return ret;
2952 }
2953
2954 /**
2955  * ice_vsi_cfg - Setup the VSI
2956  * @vsi: the VSI being configured
2957  *
2958  * Return 0 on success and negative value on error
2959  */
2960 int ice_vsi_cfg(struct ice_vsi *vsi)
2961 {
2962         int err;
2963
2964         if (vsi->netdev) {
2965                 ice_set_rx_mode(vsi->netdev);
2966
2967                 err = ice_vsi_vlan_setup(vsi);
2968
2969                 if (err)
2970                         return err;
2971         }
2972         ice_vsi_cfg_dcb_rings(vsi);
2973
2974         err = ice_vsi_cfg_lan_txqs(vsi);
2975         if (!err)
2976                 err = ice_vsi_cfg_rxqs(vsi);
2977
2978         return err;
2979 }
2980
2981 /**
2982  * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
2983  * @vsi: the VSI being configured
2984  */
2985 static void ice_napi_enable_all(struct ice_vsi *vsi)
2986 {
2987         int q_idx;
2988
2989         if (!vsi->netdev)
2990                 return;
2991
2992         ice_for_each_q_vector(vsi, q_idx) {
2993                 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
2994
2995                 if (q_vector->rx.ring || q_vector->tx.ring)
2996                         napi_enable(&q_vector->napi);
2997         }
2998 }
2999
3000 /**
3001  * ice_up_complete - Finish the last steps of bringing up a connection
3002  * @vsi: The VSI being configured
3003  *
3004  * Return 0 on success and negative value on error
3005  */
3006 static int ice_up_complete(struct ice_vsi *vsi)
3007 {
3008         struct ice_pf *pf = vsi->back;
3009         int err;
3010
3011         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3012                 ice_vsi_cfg_msix(vsi);
3013         else
3014                 return -ENOTSUPP;
3015
3016         /* Enable only Rx rings, Tx rings were enabled by the FW when the
3017          * Tx queue group list was configured and the context bits were
3018          * programmed using ice_vsi_cfg_txqs
3019          */
3020         err = ice_vsi_start_rx_rings(vsi);
3021         if (err)
3022                 return err;
3023
3024         clear_bit(__ICE_DOWN, vsi->state);
3025         ice_napi_enable_all(vsi);
3026         ice_vsi_ena_irq(vsi);
3027
3028         if (vsi->port_info &&
3029             (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
3030             vsi->netdev) {
3031                 ice_print_link_msg(vsi, true);
3032                 netif_tx_start_all_queues(vsi->netdev);
3033                 netif_carrier_on(vsi->netdev);
3034         }
3035
3036         ice_service_task_schedule(pf);
3037
3038         return 0;
3039 }
3040
3041 /**
3042  * ice_up - Bring the connection back up after being down
3043  * @vsi: VSI being configured
3044  */
3045 int ice_up(struct ice_vsi *vsi)
3046 {
3047         int err;
3048
3049         err = ice_vsi_cfg(vsi);
3050         if (!err)
3051                 err = ice_up_complete(vsi);
3052
3053         return err;
3054 }
3055
3056 /**
3057  * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
3058  * @ring: Tx or Rx ring to read stats from
3059  * @pkts: packets stats counter
3060  * @bytes: bytes stats counter
3061  *
3062  * This function fetches stats from the ring considering the atomic operations
3063  * that needs to be performed to read u64 values in 32 bit machine.
3064  */
3065 static void
3066 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
3067 {
3068         unsigned int start;
3069         *pkts = 0;
3070         *bytes = 0;
3071
3072         if (!ring)
3073                 return;
3074         do {
3075                 start = u64_stats_fetch_begin_irq(&ring->syncp);
3076                 *pkts = ring->stats.pkts;
3077                 *bytes = ring->stats.bytes;
3078         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
3079 }
3080
3081 /**
3082  * ice_update_vsi_ring_stats - Update VSI stats counters
3083  * @vsi: the VSI to be updated
3084  */
3085 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
3086 {
3087         struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
3088         struct ice_ring *ring;
3089         u64 pkts, bytes;
3090         int i;
3091
3092         /* reset netdev stats */
3093         vsi_stats->tx_packets = 0;
3094         vsi_stats->tx_bytes = 0;
3095         vsi_stats->rx_packets = 0;
3096         vsi_stats->rx_bytes = 0;
3097
3098         /* reset non-netdev (extended) stats */
3099         vsi->tx_restart = 0;
3100         vsi->tx_busy = 0;
3101         vsi->tx_linearize = 0;
3102         vsi->rx_buf_failed = 0;
3103         vsi->rx_page_failed = 0;
3104
3105         rcu_read_lock();
3106
3107         /* update Tx rings counters */
3108         ice_for_each_txq(vsi, i) {
3109                 ring = READ_ONCE(vsi->tx_rings[i]);
3110                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
3111                 vsi_stats->tx_packets += pkts;
3112                 vsi_stats->tx_bytes += bytes;
3113                 vsi->tx_restart += ring->tx_stats.restart_q;
3114                 vsi->tx_busy += ring->tx_stats.tx_busy;
3115                 vsi->tx_linearize += ring->tx_stats.tx_linearize;
3116         }
3117
3118         /* update Rx rings counters */
3119         ice_for_each_rxq(vsi, i) {
3120                 ring = READ_ONCE(vsi->rx_rings[i]);
3121                 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
3122                 vsi_stats->rx_packets += pkts;
3123                 vsi_stats->rx_bytes += bytes;
3124                 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
3125                 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
3126         }
3127
3128         rcu_read_unlock();
3129 }
3130
3131 /**
3132  * ice_update_vsi_stats - Update VSI stats counters
3133  * @vsi: the VSI to be updated
3134  */
3135 static void ice_update_vsi_stats(struct ice_vsi *vsi)
3136 {
3137         struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
3138         struct ice_eth_stats *cur_es = &vsi->eth_stats;
3139         struct ice_pf *pf = vsi->back;
3140
3141         if (test_bit(__ICE_DOWN, vsi->state) ||
3142             test_bit(__ICE_CFG_BUSY, pf->state))
3143                 return;
3144
3145         /* get stats as recorded by Tx/Rx rings */
3146         ice_update_vsi_ring_stats(vsi);
3147
3148         /* get VSI stats as recorded by the hardware */
3149         ice_update_eth_stats(vsi);
3150
3151         cur_ns->tx_errors = cur_es->tx_errors;
3152         cur_ns->rx_dropped = cur_es->rx_discards;
3153         cur_ns->tx_dropped = cur_es->tx_discards;
3154         cur_ns->multicast = cur_es->rx_multicast;
3155
3156         /* update some more netdev stats if this is main VSI */
3157         if (vsi->type == ICE_VSI_PF) {
3158                 cur_ns->rx_crc_errors = pf->stats.crc_errors;
3159                 cur_ns->rx_errors = pf->stats.crc_errors +
3160                                     pf->stats.illegal_bytes;
3161                 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
3162         }
3163 }
3164
3165 /**
3166  * ice_update_pf_stats - Update PF port stats counters
3167  * @pf: PF whose stats needs to be updated
3168  */
3169 static void ice_update_pf_stats(struct ice_pf *pf)
3170 {
3171         struct ice_hw_port_stats *prev_ps, *cur_ps;
3172         struct ice_hw *hw = &pf->hw;
3173         u8 pf_id;
3174
3175         prev_ps = &pf->stats_prev;
3176         cur_ps = &pf->stats;
3177         pf_id = hw->pf_id;
3178
3179         ice_stat_update40(hw, GLPRT_GORCH(pf_id), GLPRT_GORCL(pf_id),
3180                           pf->stat_prev_loaded, &prev_ps->eth.rx_bytes,
3181                           &cur_ps->eth.rx_bytes);
3182
3183         ice_stat_update40(hw, GLPRT_UPRCH(pf_id), GLPRT_UPRCL(pf_id),
3184                           pf->stat_prev_loaded, &prev_ps->eth.rx_unicast,
3185                           &cur_ps->eth.rx_unicast);
3186
3187         ice_stat_update40(hw, GLPRT_MPRCH(pf_id), GLPRT_MPRCL(pf_id),
3188                           pf->stat_prev_loaded, &prev_ps->eth.rx_multicast,
3189                           &cur_ps->eth.rx_multicast);
3190
3191         ice_stat_update40(hw, GLPRT_BPRCH(pf_id), GLPRT_BPRCL(pf_id),
3192                           pf->stat_prev_loaded, &prev_ps->eth.rx_broadcast,
3193                           &cur_ps->eth.rx_broadcast);
3194
3195         ice_stat_update40(hw, GLPRT_GOTCH(pf_id), GLPRT_GOTCL(pf_id),
3196                           pf->stat_prev_loaded, &prev_ps->eth.tx_bytes,
3197                           &cur_ps->eth.tx_bytes);
3198
3199         ice_stat_update40(hw, GLPRT_UPTCH(pf_id), GLPRT_UPTCL(pf_id),
3200                           pf->stat_prev_loaded, &prev_ps->eth.tx_unicast,
3201                           &cur_ps->eth.tx_unicast);
3202
3203         ice_stat_update40(hw, GLPRT_MPTCH(pf_id), GLPRT_MPTCL(pf_id),
3204                           pf->stat_prev_loaded, &prev_ps->eth.tx_multicast,
3205                           &cur_ps->eth.tx_multicast);
3206
3207         ice_stat_update40(hw, GLPRT_BPTCH(pf_id), GLPRT_BPTCL(pf_id),
3208                           pf->stat_prev_loaded, &prev_ps->eth.tx_broadcast,
3209                           &cur_ps->eth.tx_broadcast);
3210
3211         ice_stat_update32(hw, GLPRT_TDOLD(pf_id), pf->stat_prev_loaded,
3212                           &prev_ps->tx_dropped_link_down,
3213                           &cur_ps->tx_dropped_link_down);
3214
3215         ice_stat_update40(hw, GLPRT_PRC64H(pf_id), GLPRT_PRC64L(pf_id),
3216                           pf->stat_prev_loaded, &prev_ps->rx_size_64,
3217                           &cur_ps->rx_size_64);
3218
3219         ice_stat_update40(hw, GLPRT_PRC127H(pf_id), GLPRT_PRC127L(pf_id),
3220                           pf->stat_prev_loaded, &prev_ps->rx_size_127,
3221                           &cur_ps->rx_size_127);
3222
3223         ice_stat_update40(hw, GLPRT_PRC255H(pf_id), GLPRT_PRC255L(pf_id),
3224                           pf->stat_prev_loaded, &prev_ps->rx_size_255,
3225                           &cur_ps->rx_size_255);
3226
3227         ice_stat_update40(hw, GLPRT_PRC511H(pf_id), GLPRT_PRC511L(pf_id),
3228                           pf->stat_prev_loaded, &prev_ps->rx_size_511,
3229                           &cur_ps->rx_size_511);
3230
3231         ice_stat_update40(hw, GLPRT_PRC1023H(pf_id),
3232                           GLPRT_PRC1023L(pf_id), pf->stat_prev_loaded,
3233                           &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
3234
3235         ice_stat_update40(hw, GLPRT_PRC1522H(pf_id),
3236                           GLPRT_PRC1522L(pf_id), pf->stat_prev_loaded,
3237                           &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
3238
3239         ice_stat_update40(hw, GLPRT_PRC9522H(pf_id),
3240                           GLPRT_PRC9522L(pf_id), pf->stat_prev_loaded,
3241                           &prev_ps->rx_size_big, &cur_ps->rx_size_big);
3242
3243         ice_stat_update40(hw, GLPRT_PTC64H(pf_id), GLPRT_PTC64L(pf_id),
3244                           pf->stat_prev_loaded, &prev_ps->tx_size_64,
3245                           &cur_ps->tx_size_64);
3246
3247         ice_stat_update40(hw, GLPRT_PTC127H(pf_id), GLPRT_PTC127L(pf_id),
3248                           pf->stat_prev_loaded, &prev_ps->tx_size_127,
3249                           &cur_ps->tx_size_127);
3250
3251         ice_stat_update40(hw, GLPRT_PTC255H(pf_id), GLPRT_PTC255L(pf_id),
3252                           pf->stat_prev_loaded, &prev_ps->tx_size_255,
3253                           &cur_ps->tx_size_255);
3254
3255         ice_stat_update40(hw, GLPRT_PTC511H(pf_id), GLPRT_PTC511L(pf_id),
3256                           pf->stat_prev_loaded, &prev_ps->tx_size_511,
3257                           &cur_ps->tx_size_511);
3258
3259         ice_stat_update40(hw, GLPRT_PTC1023H(pf_id),
3260                           GLPRT_PTC1023L(pf_id), pf->stat_prev_loaded,
3261                           &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
3262
3263         ice_stat_update40(hw, GLPRT_PTC1522H(pf_id),
3264                           GLPRT_PTC1522L(pf_id), pf->stat_prev_loaded,
3265                           &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
3266
3267         ice_stat_update40(hw, GLPRT_PTC9522H(pf_id),
3268                           GLPRT_PTC9522L(pf_id), pf->stat_prev_loaded,
3269                           &prev_ps->tx_size_big, &cur_ps->tx_size_big);
3270
3271         ice_stat_update32(hw, GLPRT_LXONRXC(pf_id), pf->stat_prev_loaded,
3272                           &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
3273
3274         ice_stat_update32(hw, GLPRT_LXOFFRXC(pf_id), pf->stat_prev_loaded,
3275                           &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
3276
3277         ice_stat_update32(hw, GLPRT_LXONTXC(pf_id), pf->stat_prev_loaded,
3278                           &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
3279
3280         ice_stat_update32(hw, GLPRT_LXOFFTXC(pf_id), pf->stat_prev_loaded,
3281                           &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
3282
3283         ice_update_dcb_stats(pf);
3284
3285         ice_stat_update32(hw, GLPRT_CRCERRS(pf_id), pf->stat_prev_loaded,
3286                           &prev_ps->crc_errors, &cur_ps->crc_errors);
3287
3288         ice_stat_update32(hw, GLPRT_ILLERRC(pf_id), pf->stat_prev_loaded,
3289                           &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
3290
3291         ice_stat_update32(hw, GLPRT_MLFC(pf_id), pf->stat_prev_loaded,
3292                           &prev_ps->mac_local_faults,
3293                           &cur_ps->mac_local_faults);
3294
3295         ice_stat_update32(hw, GLPRT_MRFC(pf_id), pf->stat_prev_loaded,
3296                           &prev_ps->mac_remote_faults,
3297                           &cur_ps->mac_remote_faults);
3298
3299         ice_stat_update32(hw, GLPRT_RLEC(pf_id), pf->stat_prev_loaded,
3300                           &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
3301
3302         ice_stat_update32(hw, GLPRT_RUC(pf_id), pf->stat_prev_loaded,
3303                           &prev_ps->rx_undersize, &cur_ps->rx_undersize);
3304
3305         ice_stat_update32(hw, GLPRT_RFC(pf_id), pf->stat_prev_loaded,
3306                           &prev_ps->rx_fragments, &cur_ps->rx_fragments);
3307
3308         ice_stat_update32(hw, GLPRT_ROC(pf_id), pf->stat_prev_loaded,
3309                           &prev_ps->rx_oversize, &cur_ps->rx_oversize);
3310
3311         ice_stat_update32(hw, GLPRT_RJC(pf_id), pf->stat_prev_loaded,
3312                           &prev_ps->rx_jabber, &cur_ps->rx_jabber);
3313
3314         pf->stat_prev_loaded = true;
3315 }
3316
3317 /**
3318  * ice_get_stats64 - get statistics for network device structure
3319  * @netdev: network interface device structure
3320  * @stats: main device statistics structure
3321  */
3322 static
3323 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
3324 {
3325         struct ice_netdev_priv *np = netdev_priv(netdev);
3326         struct rtnl_link_stats64 *vsi_stats;
3327         struct ice_vsi *vsi = np->vsi;
3328
3329         vsi_stats = &vsi->net_stats;
3330
3331         if (test_bit(__ICE_DOWN, vsi->state) || !vsi->num_txq || !vsi->num_rxq)
3332                 return;
3333         /* netdev packet/byte stats come from ring counter. These are obtained
3334          * by summing up ring counters (done by ice_update_vsi_ring_stats).
3335          */
3336         ice_update_vsi_ring_stats(vsi);
3337         stats->tx_packets = vsi_stats->tx_packets;
3338         stats->tx_bytes = vsi_stats->tx_bytes;
3339         stats->rx_packets = vsi_stats->rx_packets;
3340         stats->rx_bytes = vsi_stats->rx_bytes;
3341
3342         /* The rest of the stats can be read from the hardware but instead we
3343          * just return values that the watchdog task has already obtained from
3344          * the hardware.
3345          */
3346         stats->multicast = vsi_stats->multicast;
3347         stats->tx_errors = vsi_stats->tx_errors;
3348         stats->tx_dropped = vsi_stats->tx_dropped;
3349         stats->rx_errors = vsi_stats->rx_errors;
3350         stats->rx_dropped = vsi_stats->rx_dropped;
3351         stats->rx_crc_errors = vsi_stats->rx_crc_errors;
3352         stats->rx_length_errors = vsi_stats->rx_length_errors;
3353 }
3354
3355 /**
3356  * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3357  * @vsi: VSI having NAPI disabled
3358  */
3359 static void ice_napi_disable_all(struct ice_vsi *vsi)
3360 {
3361         int q_idx;
3362
3363         if (!vsi->netdev)
3364                 return;
3365
3366         ice_for_each_q_vector(vsi, q_idx) {
3367                 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
3368
3369                 if (q_vector->rx.ring || q_vector->tx.ring)
3370                         napi_disable(&q_vector->napi);
3371         }
3372 }
3373
3374 /**
3375  * ice_force_phys_link_state - Force the physical link state
3376  * @vsi: VSI to force the physical link state to up/down
3377  * @link_up: true/false indicates to set the physical link to up/down
3378  *
3379  * Force the physical link state by getting the current PHY capabilities from
3380  * hardware and setting the PHY config based on the determined capabilities. If
3381  * link changes a link event will be triggered because both the Enable Automatic
3382  * Link Update and LESM Enable bits are set when setting the PHY capabilities.
3383  *
3384  * Returns 0 on success, negative on failure
3385  */
3386 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
3387 {
3388         struct ice_aqc_get_phy_caps_data *pcaps;
3389         struct ice_aqc_set_phy_cfg_data *cfg;
3390         struct ice_port_info *pi;
3391         struct device *dev;
3392         int retcode;
3393
3394         if (!vsi || !vsi->port_info || !vsi->back)
3395                 return -EINVAL;
3396         if (vsi->type != ICE_VSI_PF)
3397                 return 0;
3398
3399         dev = &vsi->back->pdev->dev;
3400
3401         pi = vsi->port_info;
3402
3403         pcaps = devm_kzalloc(dev, sizeof(*pcaps), GFP_KERNEL);
3404         if (!pcaps)
3405                 return -ENOMEM;
3406
3407         retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
3408                                       NULL);
3409         if (retcode) {
3410                 dev_err(dev,
3411                         "Failed to get phy capabilities, VSI %d error %d\n",
3412                         vsi->vsi_num, retcode);
3413                 retcode = -EIO;
3414                 goto out;
3415         }
3416
3417         /* No change in link */
3418         if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
3419             link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
3420                 goto out;
3421
3422         cfg = devm_kzalloc(dev, sizeof(*cfg), GFP_KERNEL);
3423         if (!cfg) {
3424                 retcode = -ENOMEM;
3425                 goto out;
3426         }
3427
3428         cfg->phy_type_low = pcaps->phy_type_low;
3429         cfg->phy_type_high = pcaps->phy_type_high;
3430         cfg->caps = pcaps->caps | ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
3431         cfg->low_power_ctrl = pcaps->low_power_ctrl;
3432         cfg->eee_cap = pcaps->eee_cap;
3433         cfg->eeer_value = pcaps->eeer_value;
3434         cfg->link_fec_opt = pcaps->link_fec_options;
3435         if (link_up)
3436                 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
3437         else
3438                 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
3439
3440         retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi->lport, cfg, NULL);
3441         if (retcode) {
3442                 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
3443                         vsi->vsi_num, retcode);
3444                 retcode = -EIO;
3445         }
3446
3447         devm_kfree(dev, cfg);
3448 out:
3449         devm_kfree(dev, pcaps);
3450         return retcode;
3451 }
3452
3453 /**
3454  * ice_down - Shutdown the connection
3455  * @vsi: The VSI being stopped
3456  */
3457 int ice_down(struct ice_vsi *vsi)
3458 {
3459         int i, tx_err, rx_err, link_err = 0;
3460
3461         /* Caller of this function is expected to set the
3462          * vsi->state __ICE_DOWN bit
3463          */
3464         if (vsi->netdev) {
3465                 netif_carrier_off(vsi->netdev);
3466                 netif_tx_disable(vsi->netdev);
3467         }
3468
3469         ice_vsi_dis_irq(vsi);
3470
3471         tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
3472         if (tx_err)
3473                 netdev_err(vsi->netdev,
3474                            "Failed stop Tx rings, VSI %d error %d\n",
3475                            vsi->vsi_num, tx_err);
3476
3477         rx_err = ice_vsi_stop_rx_rings(vsi);
3478         if (rx_err)
3479                 netdev_err(vsi->netdev,
3480                            "Failed stop Rx rings, VSI %d error %d\n",
3481                            vsi->vsi_num, rx_err);
3482
3483         ice_napi_disable_all(vsi);
3484
3485         if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
3486                 link_err = ice_force_phys_link_state(vsi, false);
3487                 if (link_err)
3488                         netdev_err(vsi->netdev,
3489                                    "Failed to set physical link down, VSI %d error %d\n",
3490                                    vsi->vsi_num, link_err);
3491         }
3492
3493         ice_for_each_txq(vsi, i)
3494                 ice_clean_tx_ring(vsi->tx_rings[i]);
3495
3496         ice_for_each_rxq(vsi, i)
3497                 ice_clean_rx_ring(vsi->rx_rings[i]);
3498
3499         if (tx_err || rx_err || link_err) {
3500                 netdev_err(vsi->netdev,
3501                            "Failed to close VSI 0x%04X on switch 0x%04X\n",
3502                            vsi->vsi_num, vsi->vsw->sw_id);
3503                 return -EIO;
3504         }
3505
3506         return 0;
3507 }
3508
3509 /**
3510  * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
3511  * @vsi: VSI having resources allocated
3512  *
3513  * Return 0 on success, negative on failure
3514  */
3515 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
3516 {
3517         int i, err = 0;
3518
3519         if (!vsi->num_txq) {
3520                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Tx queues\n",
3521                         vsi->vsi_num);
3522                 return -EINVAL;
3523         }
3524
3525         ice_for_each_txq(vsi, i) {
3526                 vsi->tx_rings[i]->netdev = vsi->netdev;
3527                 err = ice_setup_tx_ring(vsi->tx_rings[i]);
3528                 if (err)
3529                         break;
3530         }
3531
3532         return err;
3533 }
3534
3535 /**
3536  * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
3537  * @vsi: VSI having resources allocated
3538  *
3539  * Return 0 on success, negative on failure
3540  */
3541 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
3542 {
3543         int i, err = 0;
3544
3545         if (!vsi->num_rxq) {
3546                 dev_err(&vsi->back->pdev->dev, "VSI %d has 0 Rx queues\n",
3547                         vsi->vsi_num);
3548                 return -EINVAL;
3549         }
3550
3551         ice_for_each_rxq(vsi, i) {
3552                 vsi->rx_rings[i]->netdev = vsi->netdev;
3553                 err = ice_setup_rx_ring(vsi->rx_rings[i]);
3554                 if (err)
3555                         break;
3556         }
3557
3558         return err;
3559 }
3560
3561 /**
3562  * ice_vsi_req_irq - Request IRQ from the OS
3563  * @vsi: The VSI IRQ is being requested for
3564  * @basename: name for the vector
3565  *
3566  * Return 0 on success and a negative value on error
3567  */
3568 static int ice_vsi_req_irq(struct ice_vsi *vsi, char *basename)
3569 {
3570         struct ice_pf *pf = vsi->back;
3571         int err = -EINVAL;
3572
3573         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
3574                 err = ice_vsi_req_irq_msix(vsi, basename);
3575
3576         return err;
3577 }
3578
3579 /**
3580  * ice_vsi_open - Called when a network interface is made active
3581  * @vsi: the VSI to open
3582  *
3583  * Initialization of the VSI
3584  *
3585  * Returns 0 on success, negative value on error
3586  */
3587 static int ice_vsi_open(struct ice_vsi *vsi)
3588 {
3589         char int_name[ICE_INT_NAME_STR_LEN];
3590         struct ice_pf *pf = vsi->back;
3591         int err;
3592
3593         /* allocate descriptors */
3594         err = ice_vsi_setup_tx_rings(vsi);
3595         if (err)
3596                 goto err_setup_tx;
3597
3598         err = ice_vsi_setup_rx_rings(vsi);
3599         if (err)
3600                 goto err_setup_rx;
3601
3602         err = ice_vsi_cfg(vsi);
3603         if (err)
3604                 goto err_setup_rx;
3605
3606         snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3607                  dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
3608         err = ice_vsi_req_irq(vsi, int_name);
3609         if (err)
3610                 goto err_setup_rx;
3611
3612         /* Notify the stack of the actual queue counts. */
3613         err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
3614         if (err)
3615                 goto err_set_qs;
3616
3617         err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
3618         if (err)
3619                 goto err_set_qs;
3620
3621         err = ice_up_complete(vsi);
3622         if (err)
3623                 goto err_up_complete;
3624
3625         return 0;
3626
3627 err_up_complete:
3628         ice_down(vsi);
3629 err_set_qs:
3630         ice_vsi_free_irq(vsi);
3631 err_setup_rx:
3632         ice_vsi_free_rx_rings(vsi);
3633 err_setup_tx:
3634         ice_vsi_free_tx_rings(vsi);
3635
3636         return err;
3637 }
3638
3639 /**
3640  * ice_vsi_release_all - Delete all VSIs
3641  * @pf: PF from which all VSIs are being removed
3642  */
3643 static void ice_vsi_release_all(struct ice_pf *pf)
3644 {
3645         int err, i;
3646
3647         if (!pf->vsi)
3648                 return;
3649
3650         ice_for_each_vsi(pf, i) {
3651                 if (!pf->vsi[i])
3652                         continue;
3653
3654                 err = ice_vsi_release(pf->vsi[i]);
3655                 if (err)
3656                         dev_dbg(&pf->pdev->dev,
3657                                 "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
3658                                 i, err, pf->vsi[i]->vsi_num);
3659         }
3660 }
3661
3662 /**
3663  * ice_ena_vsi - resume a VSI
3664  * @vsi: the VSI being resume
3665  * @locked: is the rtnl_lock already held
3666  */
3667 static int ice_ena_vsi(struct ice_vsi *vsi, bool locked)
3668 {
3669         int err = 0;
3670
3671         if (!test_bit(__ICE_NEEDS_RESTART, vsi->state))
3672                 return err;
3673
3674         clear_bit(__ICE_NEEDS_RESTART, vsi->state);
3675
3676         if (vsi->netdev && vsi->type == ICE_VSI_PF) {
3677                 struct net_device *netd = vsi->netdev;
3678
3679                 if (netif_running(vsi->netdev)) {
3680                         if (locked) {
3681                                 err = netd->netdev_ops->ndo_open(netd);
3682                         } else {
3683                                 rtnl_lock();
3684                                 err = netd->netdev_ops->ndo_open(netd);
3685                                 rtnl_unlock();
3686                         }
3687                 } else {
3688                         err = ice_vsi_open(vsi);
3689                 }
3690         }
3691
3692         return err;
3693 }
3694
3695 /**
3696  * ice_pf_ena_all_vsi - Resume all VSIs on a PF
3697  * @pf: the PF
3698  * @locked: is the rtnl_lock already held
3699  */
3700 #ifdef CONFIG_DCB
3701 int ice_pf_ena_all_vsi(struct ice_pf *pf, bool locked)
3702 #else
3703 static int ice_pf_ena_all_vsi(struct ice_pf *pf, bool locked)
3704 #endif /* CONFIG_DCB */
3705 {
3706         int v;
3707
3708         ice_for_each_vsi(pf, v)
3709                 if (pf->vsi[v])
3710                         if (ice_ena_vsi(pf->vsi[v], locked))
3711                                 return -EIO;
3712
3713         return 0;
3714 }
3715
3716 /**
3717  * ice_vsi_rebuild_all - rebuild all VSIs in PF
3718  * @pf: the PF
3719  */
3720 static int ice_vsi_rebuild_all(struct ice_pf *pf)
3721 {
3722         int i;
3723
3724         /* loop through pf->vsi array and reinit the VSI if found */
3725         ice_for_each_vsi(pf, i) {
3726                 int err;
3727
3728                 if (!pf->vsi[i])
3729                         continue;
3730
3731                 err = ice_vsi_rebuild(pf->vsi[i]);
3732                 if (err) {
3733                         dev_err(&pf->pdev->dev,
3734                                 "VSI at index %d rebuild failed\n",
3735                                 pf->vsi[i]->idx);
3736                         return err;
3737                 }
3738
3739                 dev_info(&pf->pdev->dev,
3740                          "VSI at index %d rebuilt. vsi_num = 0x%x\n",
3741                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3742         }
3743
3744         return 0;
3745 }
3746
3747 /**
3748  * ice_vsi_replay_all - replay all VSIs configuration in the PF
3749  * @pf: the PF
3750  */
3751 static int ice_vsi_replay_all(struct ice_pf *pf)
3752 {
3753         struct ice_hw *hw = &pf->hw;
3754         enum ice_status ret;
3755         int i;
3756
3757         /* loop through pf->vsi array and replay the VSI if found */
3758         ice_for_each_vsi(pf, i) {
3759                 if (!pf->vsi[i])
3760                         continue;
3761
3762                 ret = ice_replay_vsi(hw, pf->vsi[i]->idx);
3763                 if (ret) {
3764                         dev_err(&pf->pdev->dev,
3765                                 "VSI at index %d replay failed %d\n",
3766                                 pf->vsi[i]->idx, ret);
3767                         return -EIO;
3768                 }
3769
3770                 /* Re-map HW VSI number, using VSI handle that has been
3771                  * previously validated in ice_replay_vsi() call above
3772                  */
3773                 pf->vsi[i]->vsi_num = ice_get_hw_vsi_num(hw, pf->vsi[i]->idx);
3774
3775                 dev_info(&pf->pdev->dev,
3776                          "VSI at index %d filter replayed successfully - vsi_num %i\n",
3777                          pf->vsi[i]->idx, pf->vsi[i]->vsi_num);
3778         }
3779
3780         /* Clean up replay filter after successful re-configuration */
3781         ice_replay_post(hw);
3782         return 0;
3783 }
3784
3785 /**
3786  * ice_rebuild - rebuild after reset
3787  * @pf: PF to rebuild
3788  */
3789 static void ice_rebuild(struct ice_pf *pf)
3790 {
3791         struct device *dev = &pf->pdev->dev;
3792         struct ice_hw *hw = &pf->hw;
3793         enum ice_status ret;
3794         int err, i;
3795
3796         if (test_bit(__ICE_DOWN, pf->state))
3797                 goto clear_recovery;
3798
3799         dev_dbg(dev, "rebuilding PF\n");
3800
3801         ret = ice_init_all_ctrlq(hw);
3802         if (ret) {
3803                 dev_err(dev, "control queues init failed %d\n", ret);
3804                 goto err_init_ctrlq;
3805         }
3806
3807         ret = ice_clear_pf_cfg(hw);
3808         if (ret) {
3809                 dev_err(dev, "clear PF configuration failed %d\n", ret);
3810                 goto err_init_ctrlq;
3811         }
3812
3813         ice_clear_pxe_mode(hw);
3814
3815         ret = ice_get_caps(hw);
3816         if (ret) {
3817                 dev_err(dev, "ice_get_caps failed %d\n", ret);
3818                 goto err_init_ctrlq;
3819         }
3820
3821         err = ice_sched_init_port(hw->port_info);
3822         if (err)
3823                 goto err_sched_init_port;
3824
3825         ice_dcb_rebuild(pf);
3826
3827         err = ice_vsi_rebuild_all(pf);
3828         if (err) {
3829                 dev_err(dev, "ice_vsi_rebuild_all failed\n");
3830                 goto err_vsi_rebuild;
3831         }
3832
3833         err = ice_update_link_info(hw->port_info);
3834         if (err)
3835                 dev_err(&pf->pdev->dev, "Get link status error %d\n", err);
3836
3837         /* Replay all VSIs Configuration, including filters after reset */
3838         if (ice_vsi_replay_all(pf)) {
3839                 dev_err(&pf->pdev->dev,
3840                         "error replaying VSI configurations with switch filter rules\n");
3841                 goto err_vsi_rebuild;
3842         }
3843
3844         /* start misc vector */
3845         if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags)) {
3846                 err = ice_req_irq_msix_misc(pf);
3847                 if (err) {
3848                         dev_err(dev, "misc vector setup failed: %d\n", err);
3849                         goto err_vsi_rebuild;
3850                 }
3851         }
3852
3853         /* restart the VSIs that were rebuilt and running before the reset */
3854         err = ice_pf_ena_all_vsi(pf, false);
3855         if (err) {
3856                 dev_err(&pf->pdev->dev, "error enabling VSIs\n");
3857                 /* no need to disable VSIs in tear down path in ice_rebuild()
3858                  * since its already taken care in ice_vsi_open()
3859                  */
3860                 goto err_vsi_rebuild;
3861         }
3862
3863         ice_for_each_vsi(pf, i) {
3864                 bool link_up;
3865
3866                 if (!pf->vsi[i] || pf->vsi[i]->type != ICE_VSI_PF)
3867                         continue;
3868                 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
3869                 if (link_up) {
3870                         netif_carrier_on(pf->vsi[i]->netdev);
3871                         netif_tx_wake_all_queues(pf->vsi[i]->netdev);
3872                 } else {
3873                         netif_carrier_off(pf->vsi[i]->netdev);
3874                         netif_tx_stop_all_queues(pf->vsi[i]->netdev);
3875                 }
3876         }
3877
3878         /* if we get here, reset flow is successful */
3879         clear_bit(__ICE_RESET_FAILED, pf->state);
3880         return;
3881
3882 err_vsi_rebuild:
3883         ice_vsi_release_all(pf);
3884 err_sched_init_port:
3885         ice_sched_cleanup_all(hw);
3886 err_init_ctrlq:
3887         ice_shutdown_all_ctrlq(hw);
3888         set_bit(__ICE_RESET_FAILED, pf->state);
3889 clear_recovery:
3890         /* set this bit in PF state to control service task scheduling */
3891         set_bit(__ICE_NEEDS_RESTART, pf->state);
3892         dev_err(dev, "Rebuild failed, unload and reload driver\n");
3893 }
3894
3895 /**
3896  * ice_change_mtu - NDO callback to change the MTU
3897  * @netdev: network interface device structure
3898  * @new_mtu: new value for maximum frame size
3899  *
3900  * Returns 0 on success, negative on failure
3901  */
3902 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
3903 {
3904         struct ice_netdev_priv *np = netdev_priv(netdev);
3905         struct ice_vsi *vsi = np->vsi;
3906         struct ice_pf *pf = vsi->back;
3907         u8 count = 0;
3908
3909         if (new_mtu == netdev->mtu) {
3910                 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
3911                 return 0;
3912         }
3913
3914         if (new_mtu < netdev->min_mtu) {
3915                 netdev_err(netdev, "new MTU invalid. min_mtu is %d\n",
3916                            netdev->min_mtu);
3917                 return -EINVAL;
3918         } else if (new_mtu > netdev->max_mtu) {
3919                 netdev_err(netdev, "new MTU invalid. max_mtu is %d\n",
3920                            netdev->min_mtu);
3921                 return -EINVAL;
3922         }
3923         /* if a reset is in progress, wait for some time for it to complete */
3924         do {
3925                 if (ice_is_reset_in_progress(pf->state)) {
3926                         count++;
3927                         usleep_range(1000, 2000);
3928                 } else {
3929                         break;
3930                 }
3931
3932         } while (count < 100);
3933
3934         if (count == 100) {
3935                 netdev_err(netdev, "can't change MTU. Device is busy\n");
3936                 return -EBUSY;
3937         }
3938
3939         netdev->mtu = new_mtu;
3940
3941         /* if VSI is up, bring it down and then back up */
3942         if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
3943                 int err;
3944
3945                 err = ice_down(vsi);
3946                 if (err) {
3947                         netdev_err(netdev, "change MTU if_up err %d\n", err);
3948                         return err;
3949                 }
3950
3951                 err = ice_up(vsi);
3952                 if (err) {
3953                         netdev_err(netdev, "change MTU if_up err %d\n", err);
3954                         return err;
3955                 }
3956         }
3957
3958         netdev_info(netdev, "changed MTU to %d\n", new_mtu);
3959         return 0;
3960 }
3961
3962 /**
3963  * ice_set_rss - Set RSS keys and lut
3964  * @vsi: Pointer to VSI structure
3965  * @seed: RSS hash seed
3966  * @lut: Lookup table
3967  * @lut_size: Lookup table size
3968  *
3969  * Returns 0 on success, negative on failure
3970  */
3971 int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
3972 {
3973         struct ice_pf *pf = vsi->back;
3974         struct ice_hw *hw = &pf->hw;
3975         enum ice_status status;
3976
3977         if (seed) {
3978                 struct ice_aqc_get_set_rss_keys *buf =
3979                                   (struct ice_aqc_get_set_rss_keys *)seed;
3980
3981                 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
3982
3983                 if (status) {
3984                         dev_err(&pf->pdev->dev,
3985                                 "Cannot set RSS key, err %d aq_err %d\n",
3986                                 status, hw->adminq.rq_last_status);
3987                         return -EIO;
3988                 }
3989         }
3990
3991         if (lut) {
3992                 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
3993                                             lut, lut_size);
3994                 if (status) {
3995                         dev_err(&pf->pdev->dev,
3996                                 "Cannot set RSS lut, err %d aq_err %d\n",
3997                                 status, hw->adminq.rq_last_status);
3998                         return -EIO;
3999                 }
4000         }
4001
4002         return 0;
4003 }
4004
4005 /**
4006  * ice_get_rss - Get RSS keys and lut
4007  * @vsi: Pointer to VSI structure
4008  * @seed: Buffer to store the keys
4009  * @lut: Buffer to store the lookup table entries
4010  * @lut_size: Size of buffer to store the lookup table entries
4011  *
4012  * Returns 0 on success, negative on failure
4013  */
4014 int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
4015 {
4016         struct ice_pf *pf = vsi->back;
4017         struct ice_hw *hw = &pf->hw;
4018         enum ice_status status;
4019
4020         if (seed) {
4021                 struct ice_aqc_get_set_rss_keys *buf =
4022                                   (struct ice_aqc_get_set_rss_keys *)seed;
4023
4024                 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
4025                 if (status) {
4026                         dev_err(&pf->pdev->dev,
4027                                 "Cannot get RSS key, err %d aq_err %d\n",
4028                                 status, hw->adminq.rq_last_status);
4029                         return -EIO;
4030                 }
4031         }
4032
4033         if (lut) {
4034                 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
4035                                             lut, lut_size);
4036                 if (status) {
4037                         dev_err(&pf->pdev->dev,
4038                                 "Cannot get RSS lut, err %d aq_err %d\n",
4039                                 status, hw->adminq.rq_last_status);
4040                         return -EIO;
4041                 }
4042         }
4043
4044         return 0;
4045 }
4046
4047 /**
4048  * ice_bridge_getlink - Get the hardware bridge mode
4049  * @skb: skb buff
4050  * @pid: process ID
4051  * @seq: RTNL message seq
4052  * @dev: the netdev being configured
4053  * @filter_mask: filter mask passed in
4054  * @nlflags: netlink flags passed in
4055  *
4056  * Return the bridge mode (VEB/VEPA)
4057  */
4058 static int
4059 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4060                    struct net_device *dev, u32 filter_mask, int nlflags)
4061 {
4062         struct ice_netdev_priv *np = netdev_priv(dev);
4063         struct ice_vsi *vsi = np->vsi;
4064         struct ice_pf *pf = vsi->back;
4065         u16 bmode;
4066
4067         bmode = pf->first_sw->bridge_mode;
4068
4069         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
4070                                        filter_mask, NULL);
4071 }
4072
4073 /**
4074  * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
4075  * @vsi: Pointer to VSI structure
4076  * @bmode: Hardware bridge mode (VEB/VEPA)
4077  *
4078  * Returns 0 on success, negative on failure
4079  */
4080 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
4081 {
4082         struct device *dev = &vsi->back->pdev->dev;
4083         struct ice_aqc_vsi_props *vsi_props;
4084         struct ice_hw *hw = &vsi->back->hw;
4085         struct ice_vsi_ctx *ctxt;
4086         enum ice_status status;
4087         int ret = 0;
4088
4089         vsi_props = &vsi->info;
4090
4091         ctxt = devm_kzalloc(dev, sizeof(*ctxt), GFP_KERNEL);
4092         if (!ctxt)
4093                 return -ENOMEM;
4094
4095         ctxt->info = vsi->info;
4096
4097         if (bmode == BRIDGE_MODE_VEB)
4098                 /* change from VEPA to VEB mode */
4099                 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
4100         else
4101                 /* change from VEB to VEPA mode */
4102                 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
4103         ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
4104
4105         status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4106         if (status) {
4107                 dev_err(dev, "update VSI for bridge mode failed, bmode = %d err %d aq_err %d\n",
4108                         bmode, status, hw->adminq.sq_last_status);
4109                 ret = -EIO;
4110                 goto out;
4111         }
4112         /* Update sw flags for book keeping */
4113         vsi_props->sw_flags = ctxt->info.sw_flags;
4114
4115 out:
4116         devm_kfree(dev, ctxt);
4117         return ret;
4118 }
4119
4120 /**
4121  * ice_bridge_setlink - Set the hardware bridge mode
4122  * @dev: the netdev being configured
4123  * @nlh: RTNL message
4124  * @flags: bridge setlink flags
4125  * @extack: netlink extended ack
4126  *
4127  * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
4128  * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
4129  * not already set for all VSIs connected to this switch. And also update the
4130  * unicast switch filter rules for the corresponding switch of the netdev.
4131  */
4132 static int
4133 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
4134                    u16 __always_unused flags,
4135                    struct netlink_ext_ack __always_unused *extack)
4136 {
4137         struct ice_netdev_priv *np = netdev_priv(dev);
4138         struct ice_pf *pf = np->vsi->back;
4139         struct nlattr *attr, *br_spec;
4140         struct ice_hw *hw = &pf->hw;
4141         enum ice_status status;
4142         struct ice_sw *pf_sw;
4143         int rem, v, err = 0;
4144
4145         pf_sw = pf->first_sw;
4146         /* find the attribute in the netlink message */
4147         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4148
4149         nla_for_each_nested(attr, br_spec, rem) {
4150                 __u16 mode;
4151
4152                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
4153                         continue;
4154                 mode = nla_get_u16(attr);
4155                 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
4156                         return -EINVAL;
4157                 /* Continue  if bridge mode is not being flipped */
4158                 if (mode == pf_sw->bridge_mode)
4159                         continue;
4160                 /* Iterates through the PF VSI list and update the loopback
4161                  * mode of the VSI
4162                  */
4163                 ice_for_each_vsi(pf, v) {
4164                         if (!pf->vsi[v])
4165                                 continue;
4166                         err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
4167                         if (err)
4168                                 return err;
4169                 }
4170
4171                 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
4172                 /* Update the unicast switch filter rules for the corresponding
4173                  * switch of the netdev
4174                  */
4175                 status = ice_update_sw_rule_bridge_mode(hw);
4176                 if (status) {
4177                         netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %d\n",
4178                                    mode, status, hw->adminq.sq_last_status);
4179                         /* revert hw->evb_veb */
4180                         hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
4181                         return -EIO;
4182                 }
4183
4184                 pf_sw->bridge_mode = mode;
4185         }
4186
4187         return 0;
4188 }
4189
4190 /**
4191  * ice_tx_timeout - Respond to a Tx Hang
4192  * @netdev: network interface device structure
4193  */
4194 static void ice_tx_timeout(struct net_device *netdev)
4195 {
4196         struct ice_netdev_priv *np = netdev_priv(netdev);
4197         struct ice_ring *tx_ring = NULL;
4198         struct ice_vsi *vsi = np->vsi;
4199         struct ice_pf *pf = vsi->back;
4200         int hung_queue = -1;
4201         u32 i;
4202
4203         pf->tx_timeout_count++;
4204
4205         /* find the stopped queue the same way dev_watchdog() does */
4206         for (i = 0; i < netdev->num_tx_queues; i++) {
4207                 unsigned long trans_start;
4208                 struct netdev_queue *q;
4209
4210                 q = netdev_get_tx_queue(netdev, i);
4211                 trans_start = q->trans_start;
4212                 if (netif_xmit_stopped(q) &&
4213                     time_after(jiffies,
4214                                trans_start + netdev->watchdog_timeo)) {
4215                         hung_queue = i;
4216                         break;
4217                 }
4218         }
4219
4220         if (i == netdev->num_tx_queues)
4221                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
4222         else
4223                 /* now that we have an index, find the tx_ring struct */
4224                 for (i = 0; i < vsi->num_txq; i++)
4225                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
4226                                 if (hung_queue == vsi->tx_rings[i]->q_index) {
4227                                         tx_ring = vsi->tx_rings[i];
4228                                         break;
4229                                 }
4230
4231         /* Reset recovery level if enough time has elapsed after last timeout.
4232          * Also ensure no new reset action happens before next timeout period.
4233          */
4234         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
4235                 pf->tx_timeout_recovery_level = 1;
4236         else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
4237                                        netdev->watchdog_timeo)))
4238                 return;
4239
4240         if (tx_ring) {
4241                 struct ice_hw *hw = &pf->hw;
4242                 u32 head, val = 0;
4243
4244                 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[hung_queue])) &
4245                         QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
4246                 /* Read interrupt register */
4247                 if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
4248                         val = rd32(hw,
4249                                    GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
4250
4251                 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %d, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
4252                             vsi->vsi_num, hung_queue, tx_ring->next_to_clean,
4253                             head, tx_ring->next_to_use, val);
4254         }
4255
4256         pf->tx_timeout_last_recovery = jiffies;
4257         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
4258                     pf->tx_timeout_recovery_level, hung_queue);
4259
4260         switch (pf->tx_timeout_recovery_level) {
4261         case 1:
4262                 set_bit(__ICE_PFR_REQ, pf->state);
4263                 break;
4264         case 2:
4265                 set_bit(__ICE_CORER_REQ, pf->state);
4266                 break;
4267         case 3:
4268                 set_bit(__ICE_GLOBR_REQ, pf->state);
4269                 break;
4270         default:
4271                 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
4272                 set_bit(__ICE_DOWN, pf->state);
4273                 set_bit(__ICE_NEEDS_RESTART, vsi->state);
4274                 set_bit(__ICE_SERVICE_DIS, pf->state);
4275                 break;
4276         }
4277
4278         ice_service_task_schedule(pf);
4279         pf->tx_timeout_recovery_level++;
4280 }
4281
4282 /**
4283  * ice_open - Called when a network interface becomes active
4284  * @netdev: network interface device structure
4285  *
4286  * The open entry point is called when a network interface is made
4287  * active by the system (IFF_UP). At this point all resources needed
4288  * for transmit and receive operations are allocated, the interrupt
4289  * handler is registered with the OS, the netdev watchdog is enabled,
4290  * and the stack is notified that the interface is ready.
4291  *
4292  * Returns 0 on success, negative value on failure
4293  */
4294 int ice_open(struct net_device *netdev)
4295 {
4296         struct ice_netdev_priv *np = netdev_priv(netdev);
4297         struct ice_vsi *vsi = np->vsi;
4298         int err;
4299
4300         if (test_bit(__ICE_NEEDS_RESTART, vsi->back->state)) {
4301                 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
4302                 return -EIO;
4303         }
4304
4305         netif_carrier_off(netdev);
4306
4307         err = ice_force_phys_link_state(vsi, true);
4308         if (err) {
4309                 netdev_err(netdev,
4310                            "Failed to set physical link up, error %d\n", err);
4311                 return err;
4312         }
4313
4314         err = ice_vsi_open(vsi);
4315         if (err)
4316                 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
4317                            vsi->vsi_num, vsi->vsw->sw_id);
4318         return err;
4319 }
4320
4321 /**
4322  * ice_stop - Disables a network interface
4323  * @netdev: network interface device structure
4324  *
4325  * The stop entry point is called when an interface is de-activated by the OS,
4326  * and the netdevice enters the DOWN state. The hardware is still under the
4327  * driver's control, but the netdev interface is disabled.
4328  *
4329  * Returns success only - not allowed to fail
4330  */
4331 int ice_stop(struct net_device *netdev)
4332 {
4333         struct ice_netdev_priv *np = netdev_priv(netdev);
4334         struct ice_vsi *vsi = np->vsi;
4335
4336         ice_vsi_close(vsi);
4337
4338         return 0;
4339 }
4340
4341 /**
4342  * ice_features_check - Validate encapsulated packet conforms to limits
4343  * @skb: skb buffer
4344  * @netdev: This port's netdev
4345  * @features: Offload features that the stack believes apply
4346  */
4347 static netdev_features_t
4348 ice_features_check(struct sk_buff *skb,
4349                    struct net_device __always_unused *netdev,
4350                    netdev_features_t features)
4351 {
4352         size_t len;
4353
4354         /* No point in doing any of this if neither checksum nor GSO are
4355          * being requested for this frame. We can rule out both by just
4356          * checking for CHECKSUM_PARTIAL
4357          */
4358         if (skb->ip_summed != CHECKSUM_PARTIAL)
4359                 return features;
4360
4361         /* We cannot support GSO if the MSS is going to be less than
4362          * 64 bytes. If it is then we need to drop support for GSO.
4363          */
4364         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
4365                 features &= ~NETIF_F_GSO_MASK;
4366
4367         len = skb_network_header(skb) - skb->data;
4368         if (len & ~(ICE_TXD_MACLEN_MAX))
4369                 goto out_rm_features;
4370
4371         len = skb_transport_header(skb) - skb_network_header(skb);
4372         if (len & ~(ICE_TXD_IPLEN_MAX))
4373                 goto out_rm_features;
4374
4375         if (skb->encapsulation) {
4376                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
4377                 if (len & ~(ICE_TXD_L4LEN_MAX))
4378                         goto out_rm_features;
4379
4380                 len = skb_inner_transport_header(skb) -
4381                       skb_inner_network_header(skb);
4382                 if (len & ~(ICE_TXD_IPLEN_MAX))
4383                         goto out_rm_features;
4384         }
4385
4386         return features;
4387 out_rm_features:
4388         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
4389 }
4390
4391 static const struct net_device_ops ice_netdev_ops = {
4392         .ndo_open = ice_open,
4393         .ndo_stop = ice_stop,
4394         .ndo_start_xmit = ice_start_xmit,
4395         .ndo_features_check = ice_features_check,
4396         .ndo_set_rx_mode = ice_set_rx_mode,
4397         .ndo_set_mac_address = ice_set_mac_address,
4398         .ndo_validate_addr = eth_validate_addr,
4399         .ndo_change_mtu = ice_change_mtu,
4400         .ndo_get_stats64 = ice_get_stats64,
4401         .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
4402         .ndo_set_vf_mac = ice_set_vf_mac,
4403         .ndo_get_vf_config = ice_get_vf_cfg,
4404         .ndo_set_vf_trust = ice_set_vf_trust,
4405         .ndo_set_vf_vlan = ice_set_vf_port_vlan,
4406         .ndo_set_vf_link_state = ice_set_vf_link_state,
4407         .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
4408         .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
4409         .ndo_set_features = ice_set_features,
4410         .ndo_bridge_getlink = ice_bridge_getlink,
4411         .ndo_bridge_setlink = ice_bridge_setlink,
4412         .ndo_fdb_add = ice_fdb_add,
4413         .ndo_fdb_del = ice_fdb_del,
4414         .ndo_tx_timeout = ice_tx_timeout,
4415 };