linux/drivers/net/ethernet/intel/ice/ice_main.c
<<
>>
Prefs
   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 <generated/utsrelease.h>
   9#include "ice.h"
  10#include "ice_base.h"
  11#include "ice_lib.h"
  12#include "ice_fltr.h"
  13#include "ice_dcb_lib.h"
  14#include "ice_dcb_nl.h"
  15#include "ice_devlink.h"
  16/* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
  17 * ice tracepoint functions. This must be done exactly once across the
  18 * ice driver.
  19 */
  20#define CREATE_TRACE_POINTS
  21#include "ice_trace.h"
  22
  23#define DRV_SUMMARY     "Intel(R) Ethernet Connection E800 Series Linux Driver"
  24static const char ice_driver_string[] = DRV_SUMMARY;
  25static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
  26
  27/* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
  28#define ICE_DDP_PKG_PATH        "intel/ice/ddp/"
  29#define ICE_DDP_PKG_FILE        ICE_DDP_PKG_PATH "ice.pkg"
  30
  31MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
  32MODULE_DESCRIPTION(DRV_SUMMARY);
  33MODULE_LICENSE("GPL v2");
  34MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
  35
  36static int debug = -1;
  37module_param(debug, int, 0644);
  38#ifndef CONFIG_DYNAMIC_DEBUG
  39MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
  40#else
  41MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
  42#endif /* !CONFIG_DYNAMIC_DEBUG */
  43
  44static DEFINE_IDA(ice_aux_ida);
  45
  46static struct workqueue_struct *ice_wq;
  47static const struct net_device_ops ice_netdev_safe_mode_ops;
  48static const struct net_device_ops ice_netdev_ops;
  49static int ice_vsi_open(struct ice_vsi *vsi);
  50
  51static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
  52
  53static void ice_vsi_release_all(struct ice_pf *pf);
  54
  55bool netif_is_ice(struct net_device *dev)
  56{
  57        return dev && (dev->netdev_ops == &ice_netdev_ops);
  58}
  59
  60/**
  61 * ice_get_tx_pending - returns number of Tx descriptors not processed
  62 * @ring: the ring of descriptors
  63 */
  64static u16 ice_get_tx_pending(struct ice_ring *ring)
  65{
  66        u16 head, tail;
  67
  68        head = ring->next_to_clean;
  69        tail = ring->next_to_use;
  70
  71        if (head != tail)
  72                return (head < tail) ?
  73                        tail - head : (tail + ring->count - head);
  74        return 0;
  75}
  76
  77/**
  78 * ice_check_for_hang_subtask - check for and recover hung queues
  79 * @pf: pointer to PF struct
  80 */
  81static void ice_check_for_hang_subtask(struct ice_pf *pf)
  82{
  83        struct ice_vsi *vsi = NULL;
  84        struct ice_hw *hw;
  85        unsigned int i;
  86        int packets;
  87        u32 v;
  88
  89        ice_for_each_vsi(pf, v)
  90                if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
  91                        vsi = pf->vsi[v];
  92                        break;
  93                }
  94
  95        if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
  96                return;
  97
  98        if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
  99                return;
 100
 101        hw = &vsi->back->hw;
 102
 103        for (i = 0; i < vsi->num_txq; i++) {
 104                struct ice_ring *tx_ring = vsi->tx_rings[i];
 105
 106                if (tx_ring && tx_ring->desc) {
 107                        /* If packet counter has not changed the queue is
 108                         * likely stalled, so force an interrupt for this
 109                         * queue.
 110                         *
 111                         * prev_pkt would be negative if there was no
 112                         * pending work.
 113                         */
 114                        packets = tx_ring->stats.pkts & INT_MAX;
 115                        if (tx_ring->tx_stats.prev_pkt == packets) {
 116                                /* Trigger sw interrupt to revive the queue */
 117                                ice_trigger_sw_intr(hw, tx_ring->q_vector);
 118                                continue;
 119                        }
 120
 121                        /* Memory barrier between read of packet count and call
 122                         * to ice_get_tx_pending()
 123                         */
 124                        smp_rmb();
 125                        tx_ring->tx_stats.prev_pkt =
 126                            ice_get_tx_pending(tx_ring) ? packets : -1;
 127                }
 128        }
 129}
 130
 131/**
 132 * ice_init_mac_fltr - Set initial MAC filters
 133 * @pf: board private structure
 134 *
 135 * Set initial set of MAC filters for PF VSI; configure filters for permanent
 136 * address and broadcast address. If an error is encountered, netdevice will be
 137 * unregistered.
 138 */
 139static int ice_init_mac_fltr(struct ice_pf *pf)
 140{
 141        enum ice_status status;
 142        struct ice_vsi *vsi;
 143        u8 *perm_addr;
 144
 145        vsi = ice_get_main_vsi(pf);
 146        if (!vsi)
 147                return -EINVAL;
 148
 149        perm_addr = vsi->port_info->mac.perm_addr;
 150        status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
 151        if (status)
 152                return -EIO;
 153
 154        return 0;
 155}
 156
 157/**
 158 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
 159 * @netdev: the net device on which the sync is happening
 160 * @addr: MAC address to sync
 161 *
 162 * This is a callback function which is called by the in kernel device sync
 163 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
 164 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
 165 * MAC filters from the hardware.
 166 */
 167static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
 168{
 169        struct ice_netdev_priv *np = netdev_priv(netdev);
 170        struct ice_vsi *vsi = np->vsi;
 171
 172        if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
 173                                     ICE_FWD_TO_VSI))
 174                return -EINVAL;
 175
 176        return 0;
 177}
 178
 179/**
 180 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
 181 * @netdev: the net device on which the unsync is happening
 182 * @addr: MAC address to unsync
 183 *
 184 * This is a callback function which is called by the in kernel device unsync
 185 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
 186 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
 187 * delete the MAC filters from the hardware.
 188 */
 189static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
 190{
 191        struct ice_netdev_priv *np = netdev_priv(netdev);
 192        struct ice_vsi *vsi = np->vsi;
 193
 194        /* Under some circumstances, we might receive a request to delete our
 195         * own device address from our uc list. Because we store the device
 196         * address in the VSI's MAC filter list, we need to ignore such
 197         * requests and not delete our device address from this list.
 198         */
 199        if (ether_addr_equal(addr, netdev->dev_addr))
 200                return 0;
 201
 202        if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
 203                                     ICE_FWD_TO_VSI))
 204                return -EINVAL;
 205
 206        return 0;
 207}
 208
 209/**
 210 * ice_vsi_fltr_changed - check if filter state changed
 211 * @vsi: VSI to be checked
 212 *
 213 * returns true if filter state has changed, false otherwise.
 214 */
 215static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
 216{
 217        return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
 218               test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
 219               test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
 220}
 221
 222/**
 223 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
 224 * @vsi: the VSI being configured
 225 * @promisc_m: mask of promiscuous config bits
 226 * @set_promisc: enable or disable promisc flag request
 227 *
 228 */
 229static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
 230{
 231        struct ice_hw *hw = &vsi->back->hw;
 232        enum ice_status status = 0;
 233
 234        if (vsi->type != ICE_VSI_PF)
 235                return 0;
 236
 237        if (vsi->num_vlan > 1) {
 238                status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
 239                                                  set_promisc);
 240        } else {
 241                if (set_promisc)
 242                        status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
 243                                                     0);
 244                else
 245                        status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
 246                                                       0);
 247        }
 248
 249        if (status)
 250                return -EIO;
 251
 252        return 0;
 253}
 254
 255/**
 256 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
 257 * @vsi: ptr to the VSI
 258 *
 259 * Push any outstanding VSI filter changes through the AdminQ.
 260 */
 261static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
 262{
 263        struct device *dev = ice_pf_to_dev(vsi->back);
 264        struct net_device *netdev = vsi->netdev;
 265        bool promisc_forced_on = false;
 266        struct ice_pf *pf = vsi->back;
 267        struct ice_hw *hw = &pf->hw;
 268        enum ice_status status = 0;
 269        u32 changed_flags = 0;
 270        u8 promisc_m;
 271        int err = 0;
 272
 273        if (!vsi->netdev)
 274                return -EINVAL;
 275
 276        while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
 277                usleep_range(1000, 2000);
 278
 279        changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
 280        vsi->current_netdev_flags = vsi->netdev->flags;
 281
 282        INIT_LIST_HEAD(&vsi->tmp_sync_list);
 283        INIT_LIST_HEAD(&vsi->tmp_unsync_list);
 284
 285        if (ice_vsi_fltr_changed(vsi)) {
 286                clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
 287                clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
 288                clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
 289
 290                /* grab the netdev's addr_list_lock */
 291                netif_addr_lock_bh(netdev);
 292                __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
 293                              ice_add_mac_to_unsync_list);
 294                __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
 295                              ice_add_mac_to_unsync_list);
 296                /* our temp lists are populated. release lock */
 297                netif_addr_unlock_bh(netdev);
 298        }
 299
 300        /* Remove MAC addresses in the unsync list */
 301        status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
 302        ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
 303        if (status) {
 304                netdev_err(netdev, "Failed to delete MAC filters\n");
 305                /* if we failed because of alloc failures, just bail */
 306                if (status == ICE_ERR_NO_MEMORY) {
 307                        err = -ENOMEM;
 308                        goto out;
 309                }
 310        }
 311
 312        /* Add MAC addresses in the sync list */
 313        status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
 314        ice_fltr_free_list(dev, &vsi->tmp_sync_list);
 315        /* If filter is added successfully or already exists, do not go into
 316         * 'if' condition and report it as error. Instead continue processing
 317         * rest of the function.
 318         */
 319        if (status && status != ICE_ERR_ALREADY_EXISTS) {
 320                netdev_err(netdev, "Failed to add MAC filters\n");
 321                /* If there is no more space for new umac filters, VSI
 322                 * should go into promiscuous mode. There should be some
 323                 * space reserved for promiscuous filters.
 324                 */
 325                if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
 326                    !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
 327                                      vsi->state)) {
 328                        promisc_forced_on = true;
 329                        netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
 330                                    vsi->vsi_num);
 331                } else {
 332                        err = -EIO;
 333                        goto out;
 334                }
 335        }
 336        /* check for changes in promiscuous modes */
 337        if (changed_flags & IFF_ALLMULTI) {
 338                if (vsi->current_netdev_flags & IFF_ALLMULTI) {
 339                        if (vsi->num_vlan > 1)
 340                                promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
 341                        else
 342                                promisc_m = ICE_MCAST_PROMISC_BITS;
 343
 344                        err = ice_cfg_promisc(vsi, promisc_m, true);
 345                        if (err) {
 346                                netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
 347                                           vsi->vsi_num);
 348                                vsi->current_netdev_flags &= ~IFF_ALLMULTI;
 349                                goto out_promisc;
 350                        }
 351                } else {
 352                        /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
 353                        if (vsi->num_vlan > 1)
 354                                promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
 355                        else
 356                                promisc_m = ICE_MCAST_PROMISC_BITS;
 357
 358                        err = ice_cfg_promisc(vsi, promisc_m, false);
 359                        if (err) {
 360                                netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
 361                                           vsi->vsi_num);
 362                                vsi->current_netdev_flags |= IFF_ALLMULTI;
 363                                goto out_promisc;
 364                        }
 365                }
 366        }
 367
 368        if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
 369            test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
 370                clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
 371                if (vsi->current_netdev_flags & IFF_PROMISC) {
 372                        /* Apply Rx filter rule to get traffic from wire */
 373                        if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
 374                                err = ice_set_dflt_vsi(pf->first_sw, vsi);
 375                                if (err && err != -EEXIST) {
 376                                        netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
 377                                                   err, vsi->vsi_num);
 378                                        vsi->current_netdev_flags &=
 379                                                ~IFF_PROMISC;
 380                                        goto out_promisc;
 381                                }
 382                                ice_cfg_vlan_pruning(vsi, false, false);
 383                        }
 384                } else {
 385                        /* Clear Rx filter to remove traffic from wire */
 386                        if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
 387                                err = ice_clear_dflt_vsi(pf->first_sw);
 388                                if (err) {
 389                                        netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
 390                                                   err, vsi->vsi_num);
 391                                        vsi->current_netdev_flags |=
 392                                                IFF_PROMISC;
 393                                        goto out_promisc;
 394                                }
 395                                if (vsi->num_vlan > 1)
 396                                        ice_cfg_vlan_pruning(vsi, true, false);
 397                        }
 398                }
 399        }
 400        goto exit;
 401
 402out_promisc:
 403        set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
 404        goto exit;
 405out:
 406        /* if something went wrong then set the changed flag so we try again */
 407        set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
 408        set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
 409exit:
 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 */
 418static 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_pf_dis_all_vsi - Pause all VSIs on a PF
 438 * @pf: the PF
 439 * @locked: is the rtnl_lock already held
 440 */
 441static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
 442{
 443        int node;
 444        int v;
 445
 446        ice_for_each_vsi(pf, v)
 447                if (pf->vsi[v])
 448                        ice_dis_vsi(pf->vsi[v], locked);
 449
 450        for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
 451                pf->pf_agg_node[node].num_vsis = 0;
 452
 453        for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
 454                pf->vf_agg_node[node].num_vsis = 0;
 455}
 456
 457/**
 458 * ice_prepare_for_reset - prep for the core to reset
 459 * @pf: board private structure
 460 *
 461 * Inform or close all dependent features in prep for reset.
 462 */
 463static void
 464ice_prepare_for_reset(struct ice_pf *pf)
 465{
 466        struct ice_hw *hw = &pf->hw;
 467        unsigned int i;
 468
 469        /* already prepared for reset */
 470        if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
 471                return;
 472
 473        ice_unplug_aux_dev(pf);
 474
 475        /* Notify VFs of impending reset */
 476        if (ice_check_sq_alive(hw, &hw->mailboxq))
 477                ice_vc_notify_reset(pf);
 478
 479        /* Disable VFs until reset is completed */
 480        ice_for_each_vf(pf, i)
 481                ice_set_vf_state_qs_dis(&pf->vf[i]);
 482
 483        /* clear SW filtering DB */
 484        ice_clear_hw_tbls(hw);
 485        /* disable the VSIs and their queues that are not already DOWN */
 486        ice_pf_dis_all_vsi(pf, false);
 487
 488        if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
 489                ice_ptp_release(pf);
 490
 491        if (hw->port_info)
 492                ice_sched_clear_port(hw->port_info);
 493
 494        ice_shutdown_all_ctrlq(hw);
 495
 496        set_bit(ICE_PREPARED_FOR_RESET, pf->state);
 497}
 498
 499/**
 500 * ice_do_reset - Initiate one of many types of resets
 501 * @pf: board private structure
 502 * @reset_type: reset type requested
 503 * before this function was called.
 504 */
 505static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
 506{
 507        struct device *dev = ice_pf_to_dev(pf);
 508        struct ice_hw *hw = &pf->hw;
 509
 510        dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
 511
 512        ice_prepare_for_reset(pf);
 513
 514        /* trigger the reset */
 515        if (ice_reset(hw, reset_type)) {
 516                dev_err(dev, "reset %d failed\n", reset_type);
 517                set_bit(ICE_RESET_FAILED, pf->state);
 518                clear_bit(ICE_RESET_OICR_RECV, pf->state);
 519                clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
 520                clear_bit(ICE_PFR_REQ, pf->state);
 521                clear_bit(ICE_CORER_REQ, pf->state);
 522                clear_bit(ICE_GLOBR_REQ, pf->state);
 523                wake_up(&pf->reset_wait_queue);
 524                return;
 525        }
 526
 527        /* PFR is a bit of a special case because it doesn't result in an OICR
 528         * interrupt. So for PFR, rebuild after the reset and clear the reset-
 529         * associated state bits.
 530         */
 531        if (reset_type == ICE_RESET_PFR) {
 532                pf->pfr_count++;
 533                ice_rebuild(pf, reset_type);
 534                clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
 535                clear_bit(ICE_PFR_REQ, pf->state);
 536                wake_up(&pf->reset_wait_queue);
 537                ice_reset_all_vfs(pf, true);
 538        }
 539}
 540
 541/**
 542 * ice_reset_subtask - Set up for resetting the device and driver
 543 * @pf: board private structure
 544 */
 545static void ice_reset_subtask(struct ice_pf *pf)
 546{
 547        enum ice_reset_req reset_type = ICE_RESET_INVAL;
 548
 549        /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
 550         * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
 551         * of reset is pending and sets bits in pf->state indicating the reset
 552         * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
 553         * prepare for pending reset if not already (for PF software-initiated
 554         * global resets the software should already be prepared for it as
 555         * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
 556         * by firmware or software on other PFs, that bit is not set so prepare
 557         * for the reset now), poll for reset done, rebuild and return.
 558         */
 559        if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
 560                /* Perform the largest reset requested */
 561                if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
 562                        reset_type = ICE_RESET_CORER;
 563                if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
 564                        reset_type = ICE_RESET_GLOBR;
 565                if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
 566                        reset_type = ICE_RESET_EMPR;
 567                /* return if no valid reset type requested */
 568                if (reset_type == ICE_RESET_INVAL)
 569                        return;
 570                ice_prepare_for_reset(pf);
 571
 572                /* make sure we are ready to rebuild */
 573                if (ice_check_reset(&pf->hw)) {
 574                        set_bit(ICE_RESET_FAILED, pf->state);
 575                } else {
 576                        /* done with reset. start rebuild */
 577                        pf->hw.reset_ongoing = false;
 578                        ice_rebuild(pf, reset_type);
 579                        /* clear bit to resume normal operations, but
 580                         * ICE_NEEDS_RESTART bit is set in case rebuild failed
 581                         */
 582                        clear_bit(ICE_RESET_OICR_RECV, pf->state);
 583                        clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
 584                        clear_bit(ICE_PFR_REQ, pf->state);
 585                        clear_bit(ICE_CORER_REQ, pf->state);
 586                        clear_bit(ICE_GLOBR_REQ, pf->state);
 587                        wake_up(&pf->reset_wait_queue);
 588                        ice_reset_all_vfs(pf, true);
 589                }
 590
 591                return;
 592        }
 593
 594        /* No pending resets to finish processing. Check for new resets */
 595        if (test_bit(ICE_PFR_REQ, pf->state))
 596                reset_type = ICE_RESET_PFR;
 597        if (test_bit(ICE_CORER_REQ, pf->state))
 598                reset_type = ICE_RESET_CORER;
 599        if (test_bit(ICE_GLOBR_REQ, pf->state))
 600                reset_type = ICE_RESET_GLOBR;
 601        /* If no valid reset type requested just return */
 602        if (reset_type == ICE_RESET_INVAL)
 603                return;
 604
 605        /* reset if not already down or busy */
 606        if (!test_bit(ICE_DOWN, pf->state) &&
 607            !test_bit(ICE_CFG_BUSY, pf->state)) {
 608                ice_do_reset(pf, reset_type);
 609        }
 610}
 611
 612/**
 613 * ice_print_topo_conflict - print topology conflict message
 614 * @vsi: the VSI whose topology status is being checked
 615 */
 616static void ice_print_topo_conflict(struct ice_vsi *vsi)
 617{
 618        switch (vsi->port_info->phy.link_info.topo_media_conflict) {
 619        case ICE_AQ_LINK_TOPO_CONFLICT:
 620        case ICE_AQ_LINK_MEDIA_CONFLICT:
 621        case ICE_AQ_LINK_TOPO_UNREACH_PRT:
 622        case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
 623        case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
 624                netdev_info(vsi->netdev, "Potential misconfiguration of the Ethernet port detected. If it was not intended, please use the Intel (R) Ethernet Port Configuration Tool to address the issue.\n");
 625                break;
 626        case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
 627                netdev_info(vsi->netdev, "Rx/Tx is disabled on this device because an unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
 628                break;
 629        default:
 630                break;
 631        }
 632}
 633
 634/**
 635 * ice_print_link_msg - print link up or down message
 636 * @vsi: the VSI whose link status is being queried
 637 * @isup: boolean for if the link is now up or down
 638 */
 639void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
 640{
 641        struct ice_aqc_get_phy_caps_data *caps;
 642        const char *an_advertised;
 643        enum ice_status status;
 644        const char *fec_req;
 645        const char *speed;
 646        const char *fec;
 647        const char *fc;
 648        const char *an;
 649
 650        if (!vsi)
 651                return;
 652
 653        if (vsi->current_isup == isup)
 654                return;
 655
 656        vsi->current_isup = isup;
 657
 658        if (!isup) {
 659                netdev_info(vsi->netdev, "NIC Link is Down\n");
 660                return;
 661        }
 662
 663        switch (vsi->port_info->phy.link_info.link_speed) {
 664        case ICE_AQ_LINK_SPEED_100GB:
 665                speed = "100 G";
 666                break;
 667        case ICE_AQ_LINK_SPEED_50GB:
 668                speed = "50 G";
 669                break;
 670        case ICE_AQ_LINK_SPEED_40GB:
 671                speed = "40 G";
 672                break;
 673        case ICE_AQ_LINK_SPEED_25GB:
 674                speed = "25 G";
 675                break;
 676        case ICE_AQ_LINK_SPEED_20GB:
 677                speed = "20 G";
 678                break;
 679        case ICE_AQ_LINK_SPEED_10GB:
 680                speed = "10 G";
 681                break;
 682        case ICE_AQ_LINK_SPEED_5GB:
 683                speed = "5 G";
 684                break;
 685        case ICE_AQ_LINK_SPEED_2500MB:
 686                speed = "2.5 G";
 687                break;
 688        case ICE_AQ_LINK_SPEED_1000MB:
 689                speed = "1 G";
 690                break;
 691        case ICE_AQ_LINK_SPEED_100MB:
 692                speed = "100 M";
 693                break;
 694        default:
 695                speed = "Unknown ";
 696                break;
 697        }
 698
 699        switch (vsi->port_info->fc.current_mode) {
 700        case ICE_FC_FULL:
 701                fc = "Rx/Tx";
 702                break;
 703        case ICE_FC_TX_PAUSE:
 704                fc = "Tx";
 705                break;
 706        case ICE_FC_RX_PAUSE:
 707                fc = "Rx";
 708                break;
 709        case ICE_FC_NONE:
 710                fc = "None";
 711                break;
 712        default:
 713                fc = "Unknown";
 714                break;
 715        }
 716
 717        /* Get FEC mode based on negotiated link info */
 718        switch (vsi->port_info->phy.link_info.fec_info) {
 719        case ICE_AQ_LINK_25G_RS_528_FEC_EN:
 720        case ICE_AQ_LINK_25G_RS_544_FEC_EN:
 721                fec = "RS-FEC";
 722                break;
 723        case ICE_AQ_LINK_25G_KR_FEC_EN:
 724                fec = "FC-FEC/BASE-R";
 725                break;
 726        default:
 727                fec = "NONE";
 728                break;
 729        }
 730
 731        /* check if autoneg completed, might be false due to not supported */
 732        if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
 733                an = "True";
 734        else
 735                an = "False";
 736
 737        /* Get FEC mode requested based on PHY caps last SW configuration */
 738        caps = kzalloc(sizeof(*caps), GFP_KERNEL);
 739        if (!caps) {
 740                fec_req = "Unknown";
 741                an_advertised = "Unknown";
 742                goto done;
 743        }
 744
 745        status = ice_aq_get_phy_caps(vsi->port_info, false,
 746                                     ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
 747        if (status)
 748                netdev_info(vsi->netdev, "Get phy capability failed.\n");
 749
 750        an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
 751
 752        if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
 753            caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
 754                fec_req = "RS-FEC";
 755        else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
 756                 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
 757                fec_req = "FC-FEC/BASE-R";
 758        else
 759                fec_req = "NONE";
 760
 761        kfree(caps);
 762
 763done:
 764        netdev_info(vsi->netdev, "NIC Link is up %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg Advertised: %s, Autoneg Negotiated: %s, Flow Control: %s\n",
 765                    speed, fec_req, fec, an_advertised, an, fc);
 766        ice_print_topo_conflict(vsi);
 767}
 768
 769/**
 770 * ice_vsi_link_event - update the VSI's netdev
 771 * @vsi: the VSI on which the link event occurred
 772 * @link_up: whether or not the VSI needs to be set up or down
 773 */
 774static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
 775{
 776        if (!vsi)
 777                return;
 778
 779        if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
 780                return;
 781
 782        if (vsi->type == ICE_VSI_PF) {
 783                if (link_up == netif_carrier_ok(vsi->netdev))
 784                        return;
 785
 786                if (link_up) {
 787                        netif_carrier_on(vsi->netdev);
 788                        netif_tx_wake_all_queues(vsi->netdev);
 789                } else {
 790                        netif_carrier_off(vsi->netdev);
 791                        netif_tx_stop_all_queues(vsi->netdev);
 792                }
 793        }
 794}
 795
 796/**
 797 * ice_set_dflt_mib - send a default config MIB to the FW
 798 * @pf: private PF struct
 799 *
 800 * This function sends a default configuration MIB to the FW.
 801 *
 802 * If this function errors out at any point, the driver is still able to
 803 * function.  The main impact is that LFC may not operate as expected.
 804 * Therefore an error state in this function should be treated with a DBG
 805 * message and continue on with driver rebuild/reenable.
 806 */
 807static void ice_set_dflt_mib(struct ice_pf *pf)
 808{
 809        struct device *dev = ice_pf_to_dev(pf);
 810        u8 mib_type, *buf, *lldpmib = NULL;
 811        u16 len, typelen, offset = 0;
 812        struct ice_lldp_org_tlv *tlv;
 813        struct ice_hw *hw = &pf->hw;
 814        u32 ouisubtype;
 815
 816        mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
 817        lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
 818        if (!lldpmib) {
 819                dev_dbg(dev, "%s Failed to allocate MIB memory\n",
 820                        __func__);
 821                return;
 822        }
 823
 824        /* Add ETS CFG TLV */
 825        tlv = (struct ice_lldp_org_tlv *)lldpmib;
 826        typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
 827                   ICE_IEEE_ETS_TLV_LEN);
 828        tlv->typelen = htons(typelen);
 829        ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
 830                      ICE_IEEE_SUBTYPE_ETS_CFG);
 831        tlv->ouisubtype = htonl(ouisubtype);
 832
 833        buf = tlv->tlvinfo;
 834        buf[0] = 0;
 835
 836        /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
 837         * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
 838         * Octets 13 - 20 are TSA values - leave as zeros
 839         */
 840        buf[5] = 0x64;
 841        len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
 842        offset += len + 2;
 843        tlv = (struct ice_lldp_org_tlv *)
 844                ((char *)tlv + sizeof(tlv->typelen) + len);
 845
 846        /* Add ETS REC TLV */
 847        buf = tlv->tlvinfo;
 848        tlv->typelen = htons(typelen);
 849
 850        ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
 851                      ICE_IEEE_SUBTYPE_ETS_REC);
 852        tlv->ouisubtype = htonl(ouisubtype);
 853
 854        /* First octet of buf is reserved
 855         * Octets 1 - 4 map UP to TC - all UPs map to zero
 856         * Octets 5 - 12 are BW values - set TC 0 to 100%.
 857         * Octets 13 - 20 are TSA value - leave as zeros
 858         */
 859        buf[5] = 0x64;
 860        offset += len + 2;
 861        tlv = (struct ice_lldp_org_tlv *)
 862                ((char *)tlv + sizeof(tlv->typelen) + len);
 863
 864        /* Add PFC CFG TLV */
 865        typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
 866                   ICE_IEEE_PFC_TLV_LEN);
 867        tlv->typelen = htons(typelen);
 868
 869        ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
 870                      ICE_IEEE_SUBTYPE_PFC_CFG);
 871        tlv->ouisubtype = htonl(ouisubtype);
 872
 873        /* Octet 1 left as all zeros - PFC disabled */
 874        buf[0] = 0x08;
 875        len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
 876        offset += len + 2;
 877
 878        if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
 879                dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
 880
 881        kfree(lldpmib);
 882}
 883
 884/**
 885 * ice_check_module_power
 886 * @pf: pointer to PF struct
 887 * @link_cfg_err: bitmap from the link info structure
 888 *
 889 * check module power level returned by a previous call to aq_get_link_info
 890 * and print error messages if module power level is not supported
 891 */
 892static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
 893{
 894        /* if module power level is supported, clear the flag */
 895        if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
 896                              ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
 897                clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
 898                return;
 899        }
 900
 901        /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
 902         * above block didn't clear this bit, there's nothing to do
 903         */
 904        if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
 905                return;
 906
 907        if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
 908                dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
 909                set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
 910        } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
 911                dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
 912                set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
 913        }
 914}
 915
 916/**
 917 * ice_link_event - process the link event
 918 * @pf: PF that the link event is associated with
 919 * @pi: port_info for the port that the link event is associated with
 920 * @link_up: true if the physical link is up and false if it is down
 921 * @link_speed: current link speed received from the link event
 922 *
 923 * Returns 0 on success and negative on failure
 924 */
 925static int
 926ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
 927               u16 link_speed)
 928{
 929        struct device *dev = ice_pf_to_dev(pf);
 930        struct ice_phy_info *phy_info;
 931        enum ice_status status;
 932        struct ice_vsi *vsi;
 933        u16 old_link_speed;
 934        bool old_link;
 935
 936        phy_info = &pi->phy;
 937        phy_info->link_info_old = phy_info->link_info;
 938
 939        old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
 940        old_link_speed = phy_info->link_info_old.link_speed;
 941
 942        /* update the link info structures and re-enable link events,
 943         * don't bail on failure due to other book keeping needed
 944         */
 945        status = ice_update_link_info(pi);
 946        if (status)
 947                dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
 948                        pi->lport, ice_stat_str(status),
 949                        ice_aq_str(pi->hw->adminq.sq_last_status));
 950
 951        ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
 952
 953        /* Check if the link state is up after updating link info, and treat
 954         * this event as an UP event since the link is actually UP now.
 955         */
 956        if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
 957                link_up = true;
 958
 959        vsi = ice_get_main_vsi(pf);
 960        if (!vsi || !vsi->port_info)
 961                return -EINVAL;
 962
 963        /* turn off PHY if media was removed */
 964        if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
 965            !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
 966                set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
 967                ice_set_link(vsi, false);
 968        }
 969
 970        /* if the old link up/down and speed is the same as the new */
 971        if (link_up == old_link && link_speed == old_link_speed)
 972                return 0;
 973
 974        if (ice_is_dcb_active(pf)) {
 975                if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
 976                        ice_dcb_rebuild(pf);
 977        } else {
 978                if (link_up)
 979                        ice_set_dflt_mib(pf);
 980        }
 981        ice_vsi_link_event(vsi, link_up);
 982        ice_print_link_msg(vsi, link_up);
 983
 984        ice_vc_notify_link_state(pf);
 985
 986        return 0;
 987}
 988
 989/**
 990 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
 991 * @pf: board private structure
 992 */
 993static void ice_watchdog_subtask(struct ice_pf *pf)
 994{
 995        int i;
 996
 997        /* if interface is down do nothing */
 998        if (test_bit(ICE_DOWN, pf->state) ||
 999            test_bit(ICE_CFG_BUSY, pf->state))
1000                return;
1001
1002        /* make sure we don't do these things too often */
1003        if (time_before(jiffies,
1004                        pf->serv_tmr_prev + pf->serv_tmr_period))
1005                return;
1006
1007        pf->serv_tmr_prev = jiffies;
1008
1009        /* Update the stats for active netdevs so the network stack
1010         * can look at updated numbers whenever it cares to
1011         */
1012        ice_update_pf_stats(pf);
1013        ice_for_each_vsi(pf, i)
1014                if (pf->vsi[i] && pf->vsi[i]->netdev)
1015                        ice_update_vsi_stats(pf->vsi[i]);
1016}
1017
1018/**
1019 * ice_init_link_events - enable/initialize link events
1020 * @pi: pointer to the port_info instance
1021 *
1022 * Returns -EIO on failure, 0 on success
1023 */
1024static int ice_init_link_events(struct ice_port_info *pi)
1025{
1026        u16 mask;
1027
1028        mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1029                       ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
1030
1031        if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1032                dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1033                        pi->lport);
1034                return -EIO;
1035        }
1036
1037        if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1038                dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1039                        pi->lport);
1040                return -EIO;
1041        }
1042
1043        return 0;
1044}
1045
1046/**
1047 * ice_handle_link_event - handle link event via ARQ
1048 * @pf: PF that the link event is associated with
1049 * @event: event structure containing link status info
1050 */
1051static int
1052ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1053{
1054        struct ice_aqc_get_link_status_data *link_data;
1055        struct ice_port_info *port_info;
1056        int status;
1057
1058        link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1059        port_info = pf->hw.port_info;
1060        if (!port_info)
1061                return -EINVAL;
1062
1063        status = ice_link_event(pf, port_info,
1064                                !!(link_data->link_info & ICE_AQ_LINK_UP),
1065                                le16_to_cpu(link_data->link_speed));
1066        if (status)
1067                dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1068                        status);
1069
1070        return status;
1071}
1072
1073enum ice_aq_task_state {
1074        ICE_AQ_TASK_WAITING = 0,
1075        ICE_AQ_TASK_COMPLETE,
1076        ICE_AQ_TASK_CANCELED,
1077};
1078
1079struct ice_aq_task {
1080        struct hlist_node entry;
1081
1082        u16 opcode;
1083        struct ice_rq_event_info *event;
1084        enum ice_aq_task_state state;
1085};
1086
1087/**
1088 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1089 * @pf: pointer to the PF private structure
1090 * @opcode: the opcode to wait for
1091 * @timeout: how long to wait, in jiffies
1092 * @event: storage for the event info
1093 *
1094 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1095 * current thread will be put to sleep until the specified event occurs or
1096 * until the given timeout is reached.
1097 *
1098 * To obtain only the descriptor contents, pass an event without an allocated
1099 * msg_buf. If the complete data buffer is desired, allocate the
1100 * event->msg_buf with enough space ahead of time.
1101 *
1102 * Returns: zero on success, or a negative error code on failure.
1103 */
1104int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1105                          struct ice_rq_event_info *event)
1106{
1107        struct device *dev = ice_pf_to_dev(pf);
1108        struct ice_aq_task *task;
1109        unsigned long start;
1110        long ret;
1111        int err;
1112
1113        task = kzalloc(sizeof(*task), GFP_KERNEL);
1114        if (!task)
1115                return -ENOMEM;
1116
1117        INIT_HLIST_NODE(&task->entry);
1118        task->opcode = opcode;
1119        task->event = event;
1120        task->state = ICE_AQ_TASK_WAITING;
1121
1122        spin_lock_bh(&pf->aq_wait_lock);
1123        hlist_add_head(&task->entry, &pf->aq_wait_list);
1124        spin_unlock_bh(&pf->aq_wait_lock);
1125
1126        start = jiffies;
1127
1128        ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1129                                               timeout);
1130        switch (task->state) {
1131        case ICE_AQ_TASK_WAITING:
1132                err = ret < 0 ? ret : -ETIMEDOUT;
1133                break;
1134        case ICE_AQ_TASK_CANCELED:
1135                err = ret < 0 ? ret : -ECANCELED;
1136                break;
1137        case ICE_AQ_TASK_COMPLETE:
1138                err = ret < 0 ? ret : 0;
1139                break;
1140        default:
1141                WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1142                err = -EINVAL;
1143                break;
1144        }
1145
1146        dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1147                jiffies_to_msecs(jiffies - start),
1148                jiffies_to_msecs(timeout),
1149                opcode);
1150
1151        spin_lock_bh(&pf->aq_wait_lock);
1152        hlist_del(&task->entry);
1153        spin_unlock_bh(&pf->aq_wait_lock);
1154        kfree(task);
1155
1156        return err;
1157}
1158
1159/**
1160 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1161 * @pf: pointer to the PF private structure
1162 * @opcode: the opcode of the event
1163 * @event: the event to check
1164 *
1165 * Loops over the current list of pending threads waiting for an AdminQ event.
1166 * For each matching task, copy the contents of the event into the task
1167 * structure and wake up the thread.
1168 *
1169 * If multiple threads wait for the same opcode, they will all be woken up.
1170 *
1171 * Note that event->msg_buf will only be duplicated if the event has a buffer
1172 * with enough space already allocated. Otherwise, only the descriptor and
1173 * message length will be copied.
1174 *
1175 * Returns: true if an event was found, false otherwise
1176 */
1177static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1178                                struct ice_rq_event_info *event)
1179{
1180        struct ice_aq_task *task;
1181        bool found = false;
1182
1183        spin_lock_bh(&pf->aq_wait_lock);
1184        hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1185                if (task->state || task->opcode != opcode)
1186                        continue;
1187
1188                memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1189                task->event->msg_len = event->msg_len;
1190
1191                /* Only copy the data buffer if a destination was set */
1192                if (task->event->msg_buf &&
1193                    task->event->buf_len > event->buf_len) {
1194                        memcpy(task->event->msg_buf, event->msg_buf,
1195                               event->buf_len);
1196                        task->event->buf_len = event->buf_len;
1197                }
1198
1199                task->state = ICE_AQ_TASK_COMPLETE;
1200                found = true;
1201        }
1202        spin_unlock_bh(&pf->aq_wait_lock);
1203
1204        if (found)
1205                wake_up(&pf->aq_wait_queue);
1206}
1207
1208/**
1209 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1210 * @pf: the PF private structure
1211 *
1212 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1213 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1214 */
1215static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1216{
1217        struct ice_aq_task *task;
1218
1219        spin_lock_bh(&pf->aq_wait_lock);
1220        hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1221                task->state = ICE_AQ_TASK_CANCELED;
1222        spin_unlock_bh(&pf->aq_wait_lock);
1223
1224        wake_up(&pf->aq_wait_queue);
1225}
1226
1227/**
1228 * __ice_clean_ctrlq - helper function to clean controlq rings
1229 * @pf: ptr to struct ice_pf
1230 * @q_type: specific Control queue type
1231 */
1232static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1233{
1234        struct device *dev = ice_pf_to_dev(pf);
1235        struct ice_rq_event_info event;
1236        struct ice_hw *hw = &pf->hw;
1237        struct ice_ctl_q_info *cq;
1238        u16 pending, i = 0;
1239        const char *qtype;
1240        u32 oldval, val;
1241
1242        /* Do not clean control queue if/when PF reset fails */
1243        if (test_bit(ICE_RESET_FAILED, pf->state))
1244                return 0;
1245
1246        switch (q_type) {
1247        case ICE_CTL_Q_ADMIN:
1248                cq = &hw->adminq;
1249                qtype = "Admin";
1250                break;
1251        case ICE_CTL_Q_SB:
1252                cq = &hw->sbq;
1253                qtype = "Sideband";
1254                break;
1255        case ICE_CTL_Q_MAILBOX:
1256                cq = &hw->mailboxq;
1257                qtype = "Mailbox";
1258                /* we are going to try to detect a malicious VF, so set the
1259                 * state to begin detection
1260                 */
1261                hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1262                break;
1263        default:
1264                dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1265                return 0;
1266        }
1267
1268        /* check for error indications - PF_xx_AxQLEN register layout for
1269         * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1270         */
1271        val = rd32(hw, cq->rq.len);
1272        if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1273                   PF_FW_ARQLEN_ARQCRIT_M)) {
1274                oldval = val;
1275                if (val & PF_FW_ARQLEN_ARQVFE_M)
1276                        dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1277                                qtype);
1278                if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1279                        dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1280                                qtype);
1281                }
1282                if (val & PF_FW_ARQLEN_ARQCRIT_M)
1283                        dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1284                                qtype);
1285                val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1286                         PF_FW_ARQLEN_ARQCRIT_M);
1287                if (oldval != val)
1288                        wr32(hw, cq->rq.len, val);
1289        }
1290
1291        val = rd32(hw, cq->sq.len);
1292        if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1293                   PF_FW_ATQLEN_ATQCRIT_M)) {
1294                oldval = val;
1295                if (val & PF_FW_ATQLEN_ATQVFE_M)
1296                        dev_dbg(dev, "%s Send Queue VF Error detected\n",
1297                                qtype);
1298                if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1299                        dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1300                                qtype);
1301                }
1302                if (val & PF_FW_ATQLEN_ATQCRIT_M)
1303                        dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1304                                qtype);
1305                val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1306                         PF_FW_ATQLEN_ATQCRIT_M);
1307                if (oldval != val)
1308                        wr32(hw, cq->sq.len, val);
1309        }
1310
1311        event.buf_len = cq->rq_buf_size;
1312        event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1313        if (!event.msg_buf)
1314                return 0;
1315
1316        do {
1317                enum ice_status ret;
1318                u16 opcode;
1319
1320                ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1321                if (ret == ICE_ERR_AQ_NO_WORK)
1322                        break;
1323                if (ret) {
1324                        dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1325                                ice_stat_str(ret));
1326                        break;
1327                }
1328
1329                opcode = le16_to_cpu(event.desc.opcode);
1330
1331                /* Notify any thread that might be waiting for this event */
1332                ice_aq_check_events(pf, opcode, &event);
1333
1334                switch (opcode) {
1335                case ice_aqc_opc_get_link_status:
1336                        if (ice_handle_link_event(pf, &event))
1337                                dev_err(dev, "Could not handle link event\n");
1338                        break;
1339                case ice_aqc_opc_event_lan_overflow:
1340                        ice_vf_lan_overflow_event(pf, &event);
1341                        break;
1342                case ice_mbx_opc_send_msg_to_pf:
1343                        if (!ice_is_malicious_vf(pf, &event, i, pending))
1344                                ice_vc_process_vf_msg(pf, &event);
1345                        break;
1346                case ice_aqc_opc_fw_logging:
1347                        ice_output_fw_log(hw, &event.desc, event.msg_buf);
1348                        break;
1349                case ice_aqc_opc_lldp_set_mib_change:
1350                        ice_dcb_process_lldp_set_mib_change(pf, &event);
1351                        break;
1352                default:
1353                        dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1354                                qtype, opcode);
1355                        break;
1356                }
1357        } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1358
1359        kfree(event.msg_buf);
1360
1361        return pending && (i == ICE_DFLT_IRQ_WORK);
1362}
1363
1364/**
1365 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1366 * @hw: pointer to hardware info
1367 * @cq: control queue information
1368 *
1369 * returns true if there are pending messages in a queue, false if there aren't
1370 */
1371static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1372{
1373        u16 ntu;
1374
1375        ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1376        return cq->rq.next_to_clean != ntu;
1377}
1378
1379/**
1380 * ice_clean_adminq_subtask - clean the AdminQ rings
1381 * @pf: board private structure
1382 */
1383static void ice_clean_adminq_subtask(struct ice_pf *pf)
1384{
1385        struct ice_hw *hw = &pf->hw;
1386
1387        if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1388                return;
1389
1390        if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1391                return;
1392
1393        clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1394
1395        /* There might be a situation where new messages arrive to a control
1396         * queue between processing the last message and clearing the
1397         * EVENT_PENDING bit. So before exiting, check queue head again (using
1398         * ice_ctrlq_pending) and process new messages if any.
1399         */
1400        if (ice_ctrlq_pending(hw, &hw->adminq))
1401                __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1402
1403        ice_flush(hw);
1404}
1405
1406/**
1407 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1408 * @pf: board private structure
1409 */
1410static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1411{
1412        struct ice_hw *hw = &pf->hw;
1413
1414        if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1415                return;
1416
1417        if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1418                return;
1419
1420        clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1421
1422        if (ice_ctrlq_pending(hw, &hw->mailboxq))
1423                __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1424
1425        ice_flush(hw);
1426}
1427
1428/**
1429 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1430 * @pf: board private structure
1431 */
1432static void ice_clean_sbq_subtask(struct ice_pf *pf)
1433{
1434        struct ice_hw *hw = &pf->hw;
1435
1436        /* Nothing to do here if sideband queue is not supported */
1437        if (!ice_is_sbq_supported(hw)) {
1438                clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1439                return;
1440        }
1441
1442        if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1443                return;
1444
1445        if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1446                return;
1447
1448        clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1449
1450        if (ice_ctrlq_pending(hw, &hw->sbq))
1451                __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1452
1453        ice_flush(hw);
1454}
1455
1456/**
1457 * ice_service_task_schedule - schedule the service task to wake up
1458 * @pf: board private structure
1459 *
1460 * If not already scheduled, this puts the task into the work queue.
1461 */
1462void ice_service_task_schedule(struct ice_pf *pf)
1463{
1464        if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1465            !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1466            !test_bit(ICE_NEEDS_RESTART, pf->state))
1467                queue_work(ice_wq, &pf->serv_task);
1468}
1469
1470/**
1471 * ice_service_task_complete - finish up the service task
1472 * @pf: board private structure
1473 */
1474static void ice_service_task_complete(struct ice_pf *pf)
1475{
1476        WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1477
1478        /* force memory (pf->state) to sync before next service task */
1479        smp_mb__before_atomic();
1480        clear_bit(ICE_SERVICE_SCHED, pf->state);
1481}
1482
1483/**
1484 * ice_service_task_stop - stop service task and cancel works
1485 * @pf: board private structure
1486 *
1487 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1488 * 1 otherwise.
1489 */
1490static int ice_service_task_stop(struct ice_pf *pf)
1491{
1492        int ret;
1493
1494        ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1495
1496        if (pf->serv_tmr.function)
1497                del_timer_sync(&pf->serv_tmr);
1498        if (pf->serv_task.func)
1499                cancel_work_sync(&pf->serv_task);
1500
1501        clear_bit(ICE_SERVICE_SCHED, pf->state);
1502        return ret;
1503}
1504
1505/**
1506 * ice_service_task_restart - restart service task and schedule works
1507 * @pf: board private structure
1508 *
1509 * This function is needed for suspend and resume works (e.g WoL scenario)
1510 */
1511static void ice_service_task_restart(struct ice_pf *pf)
1512{
1513        clear_bit(ICE_SERVICE_DIS, pf->state);
1514        ice_service_task_schedule(pf);
1515}
1516
1517/**
1518 * ice_service_timer - timer callback to schedule service task
1519 * @t: pointer to timer_list
1520 */
1521static void ice_service_timer(struct timer_list *t)
1522{
1523        struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1524
1525        mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1526        ice_service_task_schedule(pf);
1527}
1528
1529/**
1530 * ice_handle_mdd_event - handle malicious driver detect event
1531 * @pf: pointer to the PF structure
1532 *
1533 * Called from service task. OICR interrupt handler indicates MDD event.
1534 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1535 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1536 * disable the queue, the PF can be configured to reset the VF using ethtool
1537 * private flag mdd-auto-reset-vf.
1538 */
1539static void ice_handle_mdd_event(struct ice_pf *pf)
1540{
1541        struct device *dev = ice_pf_to_dev(pf);
1542        struct ice_hw *hw = &pf->hw;
1543        unsigned int i;
1544        u32 reg;
1545
1546        if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1547                /* Since the VF MDD event logging is rate limited, check if
1548                 * there are pending MDD events.
1549                 */
1550                ice_print_vfs_mdd_events(pf);
1551                return;
1552        }
1553
1554        /* find what triggered an MDD event */
1555        reg = rd32(hw, GL_MDET_TX_PQM);
1556        if (reg & GL_MDET_TX_PQM_VALID_M) {
1557                u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1558                                GL_MDET_TX_PQM_PF_NUM_S;
1559                u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1560                                GL_MDET_TX_PQM_VF_NUM_S;
1561                u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1562                                GL_MDET_TX_PQM_MAL_TYPE_S;
1563                u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1564                                GL_MDET_TX_PQM_QNUM_S);
1565
1566                if (netif_msg_tx_err(pf))
1567                        dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1568                                 event, queue, pf_num, vf_num);
1569                wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1570        }
1571
1572        reg = rd32(hw, GL_MDET_TX_TCLAN);
1573        if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1574                u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1575                                GL_MDET_TX_TCLAN_PF_NUM_S;
1576                u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1577                                GL_MDET_TX_TCLAN_VF_NUM_S;
1578                u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1579                                GL_MDET_TX_TCLAN_MAL_TYPE_S;
1580                u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1581                                GL_MDET_TX_TCLAN_QNUM_S);
1582
1583                if (netif_msg_tx_err(pf))
1584                        dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1585                                 event, queue, pf_num, vf_num);
1586                wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1587        }
1588
1589        reg = rd32(hw, GL_MDET_RX);
1590        if (reg & GL_MDET_RX_VALID_M) {
1591                u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1592                                GL_MDET_RX_PF_NUM_S;
1593                u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1594                                GL_MDET_RX_VF_NUM_S;
1595                u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1596                                GL_MDET_RX_MAL_TYPE_S;
1597                u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1598                                GL_MDET_RX_QNUM_S);
1599
1600                if (netif_msg_rx_err(pf))
1601                        dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1602                                 event, queue, pf_num, vf_num);
1603                wr32(hw, GL_MDET_RX, 0xffffffff);
1604        }
1605
1606        /* check to see if this PF caused an MDD event */
1607        reg = rd32(hw, PF_MDET_TX_PQM);
1608        if (reg & PF_MDET_TX_PQM_VALID_M) {
1609                wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1610                if (netif_msg_tx_err(pf))
1611                        dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1612        }
1613
1614        reg = rd32(hw, PF_MDET_TX_TCLAN);
1615        if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1616                wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1617                if (netif_msg_tx_err(pf))
1618                        dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1619        }
1620
1621        reg = rd32(hw, PF_MDET_RX);
1622        if (reg & PF_MDET_RX_VALID_M) {
1623                wr32(hw, PF_MDET_RX, 0xFFFF);
1624                if (netif_msg_rx_err(pf))
1625                        dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1626        }
1627
1628        /* Check to see if one of the VFs caused an MDD event, and then
1629         * increment counters and set print pending
1630         */
1631        ice_for_each_vf(pf, i) {
1632                struct ice_vf *vf = &pf->vf[i];
1633
1634                reg = rd32(hw, VP_MDET_TX_PQM(i));
1635                if (reg & VP_MDET_TX_PQM_VALID_M) {
1636                        wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1637                        vf->mdd_tx_events.count++;
1638                        set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1639                        if (netif_msg_tx_err(pf))
1640                                dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1641                                         i);
1642                }
1643
1644                reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1645                if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1646                        wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1647                        vf->mdd_tx_events.count++;
1648                        set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1649                        if (netif_msg_tx_err(pf))
1650                                dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1651                                         i);
1652                }
1653
1654                reg = rd32(hw, VP_MDET_TX_TDPU(i));
1655                if (reg & VP_MDET_TX_TDPU_VALID_M) {
1656                        wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1657                        vf->mdd_tx_events.count++;
1658                        set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1659                        if (netif_msg_tx_err(pf))
1660                                dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1661                                         i);
1662                }
1663
1664                reg = rd32(hw, VP_MDET_RX(i));
1665                if (reg & VP_MDET_RX_VALID_M) {
1666                        wr32(hw, VP_MDET_RX(i), 0xFFFF);
1667                        vf->mdd_rx_events.count++;
1668                        set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1669                        if (netif_msg_rx_err(pf))
1670                                dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1671                                         i);
1672
1673                        /* Since the queue is disabled on VF Rx MDD events, the
1674                         * PF can be configured to reset the VF through ethtool
1675                         * private flag mdd-auto-reset-vf.
1676                         */
1677                        if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1678                                /* VF MDD event counters will be cleared by
1679                                 * reset, so print the event prior to reset.
1680                                 */
1681                                ice_print_vf_rx_mdd_event(vf);
1682                                ice_reset_vf(&pf->vf[i], false);
1683                        }
1684                }
1685        }
1686
1687        ice_print_vfs_mdd_events(pf);
1688}
1689
1690/**
1691 * ice_force_phys_link_state - Force the physical link state
1692 * @vsi: VSI to force the physical link state to up/down
1693 * @link_up: true/false indicates to set the physical link to up/down
1694 *
1695 * Force the physical link state by getting the current PHY capabilities from
1696 * hardware and setting the PHY config based on the determined capabilities. If
1697 * link changes a link event will be triggered because both the Enable Automatic
1698 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1699 *
1700 * Returns 0 on success, negative on failure
1701 */
1702static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1703{
1704        struct ice_aqc_get_phy_caps_data *pcaps;
1705        struct ice_aqc_set_phy_cfg_data *cfg;
1706        struct ice_port_info *pi;
1707        struct device *dev;
1708        int retcode;
1709
1710        if (!vsi || !vsi->port_info || !vsi->back)
1711                return -EINVAL;
1712        if (vsi->type != ICE_VSI_PF)
1713                return 0;
1714
1715        dev = ice_pf_to_dev(vsi->back);
1716
1717        pi = vsi->port_info;
1718
1719        pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1720        if (!pcaps)
1721                return -ENOMEM;
1722
1723        retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1724                                      NULL);
1725        if (retcode) {
1726                dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1727                        vsi->vsi_num, retcode);
1728                retcode = -EIO;
1729                goto out;
1730        }
1731
1732        /* No change in link */
1733        if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1734            link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1735                goto out;
1736
1737        /* Use the current user PHY configuration. The current user PHY
1738         * configuration is initialized during probe from PHY capabilities
1739         * software mode, and updated on set PHY configuration.
1740         */
1741        cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1742        if (!cfg) {
1743                retcode = -ENOMEM;
1744                goto out;
1745        }
1746
1747        cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1748        if (link_up)
1749                cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1750        else
1751                cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1752
1753        retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1754        if (retcode) {
1755                dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1756                        vsi->vsi_num, retcode);
1757                retcode = -EIO;
1758        }
1759
1760        kfree(cfg);
1761out:
1762        kfree(pcaps);
1763        return retcode;
1764}
1765
1766/**
1767 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1768 * @pi: port info structure
1769 *
1770 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1771 */
1772static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1773{
1774        struct ice_aqc_get_phy_caps_data *pcaps;
1775        struct ice_pf *pf = pi->hw->back;
1776        enum ice_status status;
1777        int err = 0;
1778
1779        pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1780        if (!pcaps)
1781                return -ENOMEM;
1782
1783        status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1784                                     NULL);
1785
1786        if (status) {
1787                dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1788                err = -EIO;
1789                goto out;
1790        }
1791
1792        pf->nvm_phy_type_hi = pcaps->phy_type_high;
1793        pf->nvm_phy_type_lo = pcaps->phy_type_low;
1794
1795out:
1796        kfree(pcaps);
1797        return err;
1798}
1799
1800/**
1801 * ice_init_link_dflt_override - Initialize link default override
1802 * @pi: port info structure
1803 *
1804 * Initialize link default override and PHY total port shutdown during probe
1805 */
1806static void ice_init_link_dflt_override(struct ice_port_info *pi)
1807{
1808        struct ice_link_default_override_tlv *ldo;
1809        struct ice_pf *pf = pi->hw->back;
1810
1811        ldo = &pf->link_dflt_override;
1812        if (ice_get_link_default_override(ldo, pi))
1813                return;
1814
1815        if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1816                return;
1817
1818        /* Enable Total Port Shutdown (override/replace link-down-on-close
1819         * ethtool private flag) for ports with Port Disable bit set.
1820         */
1821        set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1822        set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1823}
1824
1825/**
1826 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1827 * @pi: port info structure
1828 *
1829 * If default override is enabled, initialize the user PHY cfg speed and FEC
1830 * settings using the default override mask from the NVM.
1831 *
1832 * The PHY should only be configured with the default override settings the
1833 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1834 * is used to indicate that the user PHY cfg default override is initialized
1835 * and the PHY has not been configured with the default override settings. The
1836 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1837 * configured.
1838 *
1839 * This function should be called only if the FW doesn't support default
1840 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1841 */
1842static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1843{
1844        struct ice_link_default_override_tlv *ldo;
1845        struct ice_aqc_set_phy_cfg_data *cfg;
1846        struct ice_phy_info *phy = &pi->phy;
1847        struct ice_pf *pf = pi->hw->back;
1848
1849        ldo = &pf->link_dflt_override;
1850
1851        /* If link default override is enabled, use to mask NVM PHY capabilities
1852         * for speed and FEC default configuration.
1853         */
1854        cfg = &phy->curr_user_phy_cfg;
1855
1856        if (ldo->phy_type_low || ldo->phy_type_high) {
1857                cfg->phy_type_low = pf->nvm_phy_type_lo &
1858                                    cpu_to_le64(ldo->phy_type_low);
1859                cfg->phy_type_high = pf->nvm_phy_type_hi &
1860                                     cpu_to_le64(ldo->phy_type_high);
1861        }
1862        cfg->link_fec_opt = ldo->fec_options;
1863        phy->curr_user_fec_req = ICE_FEC_AUTO;
1864
1865        set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1866}
1867
1868/**
1869 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1870 * @pi: port info structure
1871 *
1872 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1873 * mode to default. The PHY defaults are from get PHY capabilities topology
1874 * with media so call when media is first available. An error is returned if
1875 * called when media is not available. The PHY initialization completed state is
1876 * set here.
1877 *
1878 * These configurations are used when setting PHY
1879 * configuration. The user PHY configuration is updated on set PHY
1880 * configuration. Returns 0 on success, negative on failure
1881 */
1882static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1883{
1884        struct ice_aqc_get_phy_caps_data *pcaps;
1885        struct ice_phy_info *phy = &pi->phy;
1886        struct ice_pf *pf = pi->hw->back;
1887        enum ice_status status;
1888        int err = 0;
1889
1890        if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1891                return -EIO;
1892
1893        pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1894        if (!pcaps)
1895                return -ENOMEM;
1896
1897        if (ice_fw_supports_report_dflt_cfg(pi->hw))
1898                status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1899                                             pcaps, NULL);
1900        else
1901                status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1902                                             pcaps, NULL);
1903        if (status) {
1904                dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1905                err = -EIO;
1906                goto err_out;
1907        }
1908
1909        ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1910
1911        /* check if lenient mode is supported and enabled */
1912        if (ice_fw_supports_link_override(pi->hw) &&
1913            !(pcaps->module_compliance_enforcement &
1914              ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1915                set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1916
1917                /* if the FW supports default PHY configuration mode, then the driver
1918                 * does not have to apply link override settings. If not,
1919                 * initialize user PHY configuration with link override values
1920                 */
1921                if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1922                    (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1923                        ice_init_phy_cfg_dflt_override(pi);
1924                        goto out;
1925                }
1926        }
1927
1928        /* if link default override is not enabled, set user flow control and
1929         * FEC settings based on what get_phy_caps returned
1930         */
1931        phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1932                                                      pcaps->link_fec_options);
1933        phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1934
1935out:
1936        phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1937        set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
1938err_out:
1939        kfree(pcaps);
1940        return err;
1941}
1942
1943/**
1944 * ice_configure_phy - configure PHY
1945 * @vsi: VSI of PHY
1946 *
1947 * Set the PHY configuration. If the current PHY configuration is the same as
1948 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1949 * configure the based get PHY capabilities for topology with media.
1950 */
1951static int ice_configure_phy(struct ice_vsi *vsi)
1952{
1953        struct device *dev = ice_pf_to_dev(vsi->back);
1954        struct ice_port_info *pi = vsi->port_info;
1955        struct ice_aqc_get_phy_caps_data *pcaps;
1956        struct ice_aqc_set_phy_cfg_data *cfg;
1957        struct ice_phy_info *phy = &pi->phy;
1958        struct ice_pf *pf = vsi->back;
1959        enum ice_status status;
1960        int err = 0;
1961
1962        /* Ensure we have media as we cannot configure a medialess port */
1963        if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1964                return -EPERM;
1965
1966        ice_print_topo_conflict(vsi);
1967
1968        if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1969                return -EPERM;
1970
1971        if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1972                return ice_force_phys_link_state(vsi, true);
1973
1974        pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1975        if (!pcaps)
1976                return -ENOMEM;
1977
1978        /* Get current PHY config */
1979        status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1980                                     NULL);
1981        if (status) {
1982                dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1983                        vsi->vsi_num, ice_stat_str(status));
1984                err = -EIO;
1985                goto done;
1986        }
1987
1988        /* If PHY enable link is configured and configuration has not changed,
1989         * there's nothing to do
1990         */
1991        if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1992            ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1993                goto done;
1994
1995        /* Use PHY topology as baseline for configuration */
1996        memset(pcaps, 0, sizeof(*pcaps));
1997        if (ice_fw_supports_report_dflt_cfg(pi->hw))
1998                status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1999                                             pcaps, NULL);
2000        else
2001                status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2002                                             pcaps, NULL);
2003        if (status) {
2004                dev_err(dev, "Failed to get PHY caps, VSI %d error %s\n",
2005                        vsi->vsi_num, ice_stat_str(status));
2006                err = -EIO;
2007                goto done;
2008        }
2009
2010        cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2011        if (!cfg) {
2012                err = -ENOMEM;
2013                goto done;
2014        }
2015
2016        ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2017
2018        /* Speed - If default override pending, use curr_user_phy_cfg set in
2019         * ice_init_phy_user_cfg_ldo.
2020         */
2021        if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2022                               vsi->back->state)) {
2023                cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2024                cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2025        } else {
2026                u64 phy_low = 0, phy_high = 0;
2027
2028                ice_update_phy_type(&phy_low, &phy_high,
2029                                    pi->phy.curr_user_speed_req);
2030                cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2031                cfg->phy_type_high = pcaps->phy_type_high &
2032                                     cpu_to_le64(phy_high);
2033        }
2034
2035        /* Can't provide what was requested; use PHY capabilities */
2036        if (!cfg->phy_type_low && !cfg->phy_type_high) {
2037                cfg->phy_type_low = pcaps->phy_type_low;
2038                cfg->phy_type_high = pcaps->phy_type_high;
2039        }
2040
2041        /* FEC */
2042        ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2043
2044        /* Can't provide what was requested; use PHY capabilities */
2045        if (cfg->link_fec_opt !=
2046            (cfg->link_fec_opt & pcaps->link_fec_options)) {
2047                cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2048                cfg->link_fec_opt = pcaps->link_fec_options;
2049        }
2050
2051        /* Flow Control - always supported; no need to check against
2052         * capabilities
2053         */
2054        ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2055
2056        /* Enable link and link update */
2057        cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2058
2059        status = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2060        if (status) {
2061                dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
2062                        vsi->vsi_num, ice_stat_str(status));
2063                err = -EIO;
2064        }
2065
2066        kfree(cfg);
2067done:
2068        kfree(pcaps);
2069        return err;
2070}
2071
2072/**
2073 * ice_check_media_subtask - Check for media
2074 * @pf: pointer to PF struct
2075 *
2076 * If media is available, then initialize PHY user configuration if it is not
2077 * been, and configure the PHY if the interface is up.
2078 */
2079static void ice_check_media_subtask(struct ice_pf *pf)
2080{
2081        struct ice_port_info *pi;
2082        struct ice_vsi *vsi;
2083        int err;
2084
2085        /* No need to check for media if it's already present */
2086        if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2087                return;
2088
2089        vsi = ice_get_main_vsi(pf);
2090        if (!vsi)
2091                return;
2092
2093        /* Refresh link info and check if media is present */
2094        pi = vsi->port_info;
2095        err = ice_update_link_info(pi);
2096        if (err)
2097                return;
2098
2099        ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
2100
2101        if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2102                if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2103                        ice_init_phy_user_cfg(pi);
2104
2105                /* PHY settings are reset on media insertion, reconfigure
2106                 * PHY to preserve settings.
2107                 */
2108                if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2109                    test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2110                        return;
2111
2112                err = ice_configure_phy(vsi);
2113                if (!err)
2114                        clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2115
2116                /* A Link Status Event will be generated; the event handler
2117                 * will complete bringing the interface up
2118                 */
2119        }
2120}
2121
2122/**
2123 * ice_service_task - manage and run subtasks
2124 * @work: pointer to work_struct contained by the PF struct
2125 */
2126static void ice_service_task(struct work_struct *work)
2127{
2128        struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2129        unsigned long start_time = jiffies;
2130
2131        /* subtasks */
2132
2133        /* process reset requests first */
2134        ice_reset_subtask(pf);
2135
2136        /* bail if a reset/recovery cycle is pending or rebuild failed */
2137        if (ice_is_reset_in_progress(pf->state) ||
2138            test_bit(ICE_SUSPENDED, pf->state) ||
2139            test_bit(ICE_NEEDS_RESTART, pf->state)) {
2140                ice_service_task_complete(pf);
2141                return;
2142        }
2143
2144        ice_clean_adminq_subtask(pf);
2145        ice_check_media_subtask(pf);
2146        ice_check_for_hang_subtask(pf);
2147        ice_sync_fltr_subtask(pf);
2148        ice_handle_mdd_event(pf);
2149        ice_watchdog_subtask(pf);
2150
2151        if (ice_is_safe_mode(pf)) {
2152                ice_service_task_complete(pf);
2153                return;
2154        }
2155
2156        ice_process_vflr_event(pf);
2157        ice_clean_mailboxq_subtask(pf);
2158        ice_clean_sbq_subtask(pf);
2159        ice_sync_arfs_fltrs(pf);
2160        ice_flush_fdir_ctx(pf);
2161
2162        /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2163        ice_service_task_complete(pf);
2164
2165        /* If the tasks have taken longer than one service timer period
2166         * or there is more work to be done, reset the service timer to
2167         * schedule the service task now.
2168         */
2169        if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2170            test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2171            test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2172            test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2173            test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2174            test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2175            test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2176                mod_timer(&pf->serv_tmr, jiffies);
2177}
2178
2179/**
2180 * ice_set_ctrlq_len - helper function to set controlq length
2181 * @hw: pointer to the HW instance
2182 */
2183static void ice_set_ctrlq_len(struct ice_hw *hw)
2184{
2185        hw->adminq.num_rq_entries = ICE_AQ_LEN;
2186        hw->adminq.num_sq_entries = ICE_AQ_LEN;
2187        hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2188        hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2189        hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2190        hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2191        hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2192        hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2193        hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2194        hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2195        hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2196        hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2197}
2198
2199/**
2200 * ice_schedule_reset - schedule a reset
2201 * @pf: board private structure
2202 * @reset: reset being requested
2203 */
2204int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2205{
2206        struct device *dev = ice_pf_to_dev(pf);
2207
2208        /* bail out if earlier reset has failed */
2209        if (test_bit(ICE_RESET_FAILED, pf->state)) {
2210                dev_dbg(dev, "earlier reset has failed\n");
2211                return -EIO;
2212        }
2213        /* bail if reset/recovery already in progress */
2214        if (ice_is_reset_in_progress(pf->state)) {
2215                dev_dbg(dev, "Reset already in progress\n");
2216                return -EBUSY;
2217        }
2218
2219        ice_unplug_aux_dev(pf);
2220
2221        switch (reset) {
2222        case ICE_RESET_PFR:
2223                set_bit(ICE_PFR_REQ, pf->state);
2224                break;
2225        case ICE_RESET_CORER:
2226                set_bit(ICE_CORER_REQ, pf->state);
2227                break;
2228        case ICE_RESET_GLOBR:
2229                set_bit(ICE_GLOBR_REQ, pf->state);
2230                break;
2231        default:
2232                return -EINVAL;
2233        }
2234
2235        ice_service_task_schedule(pf);
2236        return 0;
2237}
2238
2239/**
2240 * ice_irq_affinity_notify - Callback for affinity changes
2241 * @notify: context as to what irq was changed
2242 * @mask: the new affinity mask
2243 *
2244 * This is a callback function used by the irq_set_affinity_notifier function
2245 * so that we may register to receive changes to the irq affinity masks.
2246 */
2247static void
2248ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2249                        const cpumask_t *mask)
2250{
2251        struct ice_q_vector *q_vector =
2252                container_of(notify, struct ice_q_vector, affinity_notify);
2253
2254        cpumask_copy(&q_vector->affinity_mask, mask);
2255}
2256
2257/**
2258 * ice_irq_affinity_release - Callback for affinity notifier release
2259 * @ref: internal core kernel usage
2260 *
2261 * This is a callback function used by the irq_set_affinity_notifier function
2262 * to inform the current notification subscriber that they will no longer
2263 * receive notifications.
2264 */
2265static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2266
2267/**
2268 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2269 * @vsi: the VSI being configured
2270 */
2271static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2272{
2273        struct ice_hw *hw = &vsi->back->hw;
2274        int i;
2275
2276        ice_for_each_q_vector(vsi, i)
2277                ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2278
2279        ice_flush(hw);
2280        return 0;
2281}
2282
2283/**
2284 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2285 * @vsi: the VSI being configured
2286 * @basename: name for the vector
2287 */
2288static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2289{
2290        int q_vectors = vsi->num_q_vectors;
2291        struct ice_pf *pf = vsi->back;
2292        int base = vsi->base_vector;
2293        struct device *dev;
2294        int rx_int_idx = 0;
2295        int tx_int_idx = 0;
2296        int vector, err;
2297        int irq_num;
2298
2299        dev = ice_pf_to_dev(pf);
2300        for (vector = 0; vector < q_vectors; vector++) {
2301                struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2302
2303                irq_num = pf->msix_entries[base + vector].vector;
2304
2305                if (q_vector->tx.ring && q_vector->rx.ring) {
2306                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2307                                 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2308                        tx_int_idx++;
2309                } else if (q_vector->rx.ring) {
2310                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2311                                 "%s-%s-%d", basename, "rx", rx_int_idx++);
2312                } else if (q_vector->tx.ring) {
2313                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2314                                 "%s-%s-%d", basename, "tx", tx_int_idx++);
2315                } else {
2316                        /* skip this unused q_vector */
2317                        continue;
2318                }
2319                if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2320                        err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2321                                               IRQF_SHARED, q_vector->name,
2322                                               q_vector);
2323                else
2324                        err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2325                                               0, q_vector->name, q_vector);
2326                if (err) {
2327                        netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2328                                   err);
2329                        goto free_q_irqs;
2330                }
2331
2332                /* register for affinity change notifications */
2333                if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2334                        struct irq_affinity_notify *affinity_notify;
2335
2336                        affinity_notify = &q_vector->affinity_notify;
2337                        affinity_notify->notify = ice_irq_affinity_notify;
2338                        affinity_notify->release = ice_irq_affinity_release;
2339                        irq_set_affinity_notifier(irq_num, affinity_notify);
2340                }
2341
2342                /* assign the mask for this irq */
2343                irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2344        }
2345
2346        vsi->irqs_ready = true;
2347        return 0;
2348
2349free_q_irqs:
2350        while (vector) {
2351                vector--;
2352                irq_num = pf->msix_entries[base + vector].vector;
2353                if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2354                        irq_set_affinity_notifier(irq_num, NULL);
2355                irq_set_affinity_hint(irq_num, NULL);
2356                devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2357        }
2358        return err;
2359}
2360
2361/**
2362 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2363 * @vsi: VSI to setup Tx rings used by XDP
2364 *
2365 * Return 0 on success and negative value on error
2366 */
2367static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2368{
2369        struct device *dev = ice_pf_to_dev(vsi->back);
2370        int i;
2371
2372        for (i = 0; i < vsi->num_xdp_txq; i++) {
2373                u16 xdp_q_idx = vsi->alloc_txq + i;
2374                struct ice_ring *xdp_ring;
2375
2376                xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2377
2378                if (!xdp_ring)
2379                        goto free_xdp_rings;
2380
2381                xdp_ring->q_index = xdp_q_idx;
2382                xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2383                xdp_ring->ring_active = false;
2384                xdp_ring->vsi = vsi;
2385                xdp_ring->netdev = NULL;
2386                xdp_ring->dev = dev;
2387                xdp_ring->count = vsi->num_tx_desc;
2388                WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2389                if (ice_setup_tx_ring(xdp_ring))
2390                        goto free_xdp_rings;
2391                ice_set_ring_xdp(xdp_ring);
2392                xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2393        }
2394
2395        return 0;
2396
2397free_xdp_rings:
2398        for (; i >= 0; i--)
2399                if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2400                        ice_free_tx_ring(vsi->xdp_rings[i]);
2401        return -ENOMEM;
2402}
2403
2404/**
2405 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2406 * @vsi: VSI to set the bpf prog on
2407 * @prog: the bpf prog pointer
2408 */
2409static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2410{
2411        struct bpf_prog *old_prog;
2412        int i;
2413
2414        old_prog = xchg(&vsi->xdp_prog, prog);
2415        if (old_prog)
2416                bpf_prog_put(old_prog);
2417
2418        ice_for_each_rxq(vsi, i)
2419                WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2420}
2421
2422/**
2423 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2424 * @vsi: VSI to bring up Tx rings used by XDP
2425 * @prog: bpf program that will be assigned to VSI
2426 *
2427 * Return 0 on success and negative value on error
2428 */
2429int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2430{
2431        u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2432        int xdp_rings_rem = vsi->num_xdp_txq;
2433        struct ice_pf *pf = vsi->back;
2434        struct ice_qs_cfg xdp_qs_cfg = {
2435                .qs_mutex = &pf->avail_q_mutex,
2436                .pf_map = pf->avail_txqs,
2437                .pf_map_size = pf->max_pf_txqs,
2438                .q_count = vsi->num_xdp_txq,
2439                .scatter_count = ICE_MAX_SCATTER_TXQS,
2440                .vsi_map = vsi->txq_map,
2441                .vsi_map_offset = vsi->alloc_txq,
2442                .mapping_mode = ICE_VSI_MAP_CONTIG
2443        };
2444        enum ice_status status;
2445        struct device *dev;
2446        int i, v_idx;
2447
2448        dev = ice_pf_to_dev(pf);
2449        vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2450                                      sizeof(*vsi->xdp_rings), GFP_KERNEL);
2451        if (!vsi->xdp_rings)
2452                return -ENOMEM;
2453
2454        vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2455        if (__ice_vsi_get_qs(&xdp_qs_cfg))
2456                goto err_map_xdp;
2457
2458        if (ice_xdp_alloc_setup_rings(vsi))
2459                goto clear_xdp_rings;
2460
2461        /* follow the logic from ice_vsi_map_rings_to_vectors */
2462        ice_for_each_q_vector(vsi, v_idx) {
2463                struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2464                int xdp_rings_per_v, q_id, q_base;
2465
2466                xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2467                                               vsi->num_q_vectors - v_idx);
2468                q_base = vsi->num_xdp_txq - xdp_rings_rem;
2469
2470                for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2471                        struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2472
2473                        xdp_ring->q_vector = q_vector;
2474                        xdp_ring->next = q_vector->tx.ring;
2475                        q_vector->tx.ring = xdp_ring;
2476                }
2477                xdp_rings_rem -= xdp_rings_per_v;
2478        }
2479
2480        /* omit the scheduler update if in reset path; XDP queues will be
2481         * taken into account at the end of ice_vsi_rebuild, where
2482         * ice_cfg_vsi_lan is being called
2483         */
2484        if (ice_is_reset_in_progress(pf->state))
2485                return 0;
2486
2487        /* tell the Tx scheduler that right now we have
2488         * additional queues
2489         */
2490        for (i = 0; i < vsi->tc_cfg.numtc; i++)
2491                max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2492
2493        status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2494                                 max_txqs);
2495        if (status) {
2496                dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2497                        ice_stat_str(status));
2498                goto clear_xdp_rings;
2499        }
2500        ice_vsi_assign_bpf_prog(vsi, prog);
2501
2502        return 0;
2503clear_xdp_rings:
2504        for (i = 0; i < vsi->num_xdp_txq; i++)
2505                if (vsi->xdp_rings[i]) {
2506                        kfree_rcu(vsi->xdp_rings[i], rcu);
2507                        vsi->xdp_rings[i] = NULL;
2508                }
2509
2510err_map_xdp:
2511        mutex_lock(&pf->avail_q_mutex);
2512        for (i = 0; i < vsi->num_xdp_txq; i++) {
2513                clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2514                vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2515        }
2516        mutex_unlock(&pf->avail_q_mutex);
2517
2518        devm_kfree(dev, vsi->xdp_rings);
2519        return -ENOMEM;
2520}
2521
2522/**
2523 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2524 * @vsi: VSI to remove XDP rings
2525 *
2526 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2527 * resources
2528 */
2529int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2530{
2531        u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2532        struct ice_pf *pf = vsi->back;
2533        int i, v_idx;
2534
2535        /* q_vectors are freed in reset path so there's no point in detaching
2536         * rings; in case of rebuild being triggered not from reset bits
2537         * in pf->state won't be set, so additionally check first q_vector
2538         * against NULL
2539         */
2540        if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2541                goto free_qmap;
2542
2543        ice_for_each_q_vector(vsi, v_idx) {
2544                struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2545                struct ice_ring *ring;
2546
2547                ice_for_each_ring(ring, q_vector->tx)
2548                        if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2549                                break;
2550
2551                /* restore the value of last node prior to XDP setup */
2552                q_vector->tx.ring = ring;
2553        }
2554
2555free_qmap:
2556        mutex_lock(&pf->avail_q_mutex);
2557        for (i = 0; i < vsi->num_xdp_txq; i++) {
2558                clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2559                vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2560        }
2561        mutex_unlock(&pf->avail_q_mutex);
2562
2563        for (i = 0; i < vsi->num_xdp_txq; i++)
2564                if (vsi->xdp_rings[i]) {
2565                        if (vsi->xdp_rings[i]->desc)
2566                                ice_free_tx_ring(vsi->xdp_rings[i]);
2567                        kfree_rcu(vsi->xdp_rings[i], rcu);
2568                        vsi->xdp_rings[i] = NULL;
2569                }
2570
2571        devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2572        vsi->xdp_rings = NULL;
2573
2574        if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2575                return 0;
2576
2577        ice_vsi_assign_bpf_prog(vsi, NULL);
2578
2579        /* notify Tx scheduler that we destroyed XDP queues and bring
2580         * back the old number of child nodes
2581         */
2582        for (i = 0; i < vsi->tc_cfg.numtc; i++)
2583                max_txqs[i] = vsi->num_txq;
2584
2585        /* change number of XDP Tx queues to 0 */
2586        vsi->num_xdp_txq = 0;
2587
2588        return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2589                               max_txqs);
2590}
2591
2592/**
2593 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2594 * @vsi: VSI to schedule napi on
2595 */
2596static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2597{
2598        int i;
2599
2600        ice_for_each_rxq(vsi, i) {
2601                struct ice_ring *rx_ring = vsi->rx_rings[i];
2602
2603                if (rx_ring->xsk_pool)
2604                        napi_schedule(&rx_ring->q_vector->napi);
2605        }
2606}
2607
2608/**
2609 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2610 * @vsi: VSI to setup XDP for
2611 * @prog: XDP program
2612 * @extack: netlink extended ack
2613 */
2614static int
2615ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2616                   struct netlink_ext_ack *extack)
2617{
2618        int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2619        bool if_running = netif_running(vsi->netdev);
2620        int ret = 0, xdp_ring_err = 0;
2621
2622        if (frame_size > vsi->rx_buf_len) {
2623                NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2624                return -EOPNOTSUPP;
2625        }
2626
2627        /* need to stop netdev while setting up the program for Rx rings */
2628        if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2629                ret = ice_down(vsi);
2630                if (ret) {
2631                        NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2632                        return ret;
2633                }
2634        }
2635
2636        if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2637                vsi->num_xdp_txq = vsi->alloc_rxq;
2638                xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2639                if (xdp_ring_err)
2640                        NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2641        } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2642                xdp_ring_err = ice_destroy_xdp_rings(vsi);
2643                if (xdp_ring_err)
2644                        NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2645        } else {
2646                ice_vsi_assign_bpf_prog(vsi, prog);
2647        }
2648
2649        if (if_running)
2650                ret = ice_up(vsi);
2651
2652        if (!ret && prog)
2653                ice_vsi_rx_napi_schedule(vsi);
2654
2655        return (ret || xdp_ring_err) ? -ENOMEM : 0;
2656}
2657
2658/**
2659 * ice_xdp_safe_mode - XDP handler for safe mode
2660 * @dev: netdevice
2661 * @xdp: XDP command
2662 */
2663static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2664                             struct netdev_bpf *xdp)
2665{
2666        NL_SET_ERR_MSG_MOD(xdp->extack,
2667                           "Please provide working DDP firmware package in order to use XDP\n"
2668                           "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2669        return -EOPNOTSUPP;
2670}
2671
2672/**
2673 * ice_xdp - implements XDP handler
2674 * @dev: netdevice
2675 * @xdp: XDP command
2676 */
2677static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2678{
2679        struct ice_netdev_priv *np = netdev_priv(dev);
2680        struct ice_vsi *vsi = np->vsi;
2681
2682        if (vsi->type != ICE_VSI_PF) {
2683                NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2684                return -EINVAL;
2685        }
2686
2687        switch (xdp->command) {
2688        case XDP_SETUP_PROG:
2689                return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2690        case XDP_SETUP_XSK_POOL:
2691                return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2692                                          xdp->xsk.queue_id);
2693        default:
2694                return -EINVAL;
2695        }
2696}
2697
2698/**
2699 * ice_ena_misc_vector - enable the non-queue interrupts
2700 * @pf: board private structure
2701 */
2702static void ice_ena_misc_vector(struct ice_pf *pf)
2703{
2704        struct ice_hw *hw = &pf->hw;
2705        u32 val;
2706
2707        /* Disable anti-spoof detection interrupt to prevent spurious event
2708         * interrupts during a function reset. Anti-spoof functionally is
2709         * still supported.
2710         */
2711        val = rd32(hw, GL_MDCK_TX_TDPU);
2712        val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2713        wr32(hw, GL_MDCK_TX_TDPU, val);
2714
2715        /* clear things first */
2716        wr32(hw, PFINT_OICR_ENA, 0);    /* disable all */
2717        rd32(hw, PFINT_OICR);           /* read to clear */
2718
2719        val = (PFINT_OICR_ECC_ERR_M |
2720               PFINT_OICR_MAL_DETECT_M |
2721               PFINT_OICR_GRST_M |
2722               PFINT_OICR_PCI_EXCEPTION_M |
2723               PFINT_OICR_VFLR_M |
2724               PFINT_OICR_HMC_ERR_M |
2725               PFINT_OICR_PE_PUSH_M |
2726               PFINT_OICR_PE_CRITERR_M);
2727
2728        wr32(hw, PFINT_OICR_ENA, val);
2729
2730        /* SW_ITR_IDX = 0, but don't change INTENA */
2731        wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2732             GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2733}
2734
2735/**
2736 * ice_misc_intr - misc interrupt handler
2737 * @irq: interrupt number
2738 * @data: pointer to a q_vector
2739 */
2740static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2741{
2742        struct ice_pf *pf = (struct ice_pf *)data;
2743        struct ice_hw *hw = &pf->hw;
2744        irqreturn_t ret = IRQ_NONE;
2745        struct device *dev;
2746        u32 oicr, ena_mask;
2747
2748        dev = ice_pf_to_dev(pf);
2749        set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2750        set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2751        set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
2752
2753        oicr = rd32(hw, PFINT_OICR);
2754        ena_mask = rd32(hw, PFINT_OICR_ENA);
2755
2756        if (oicr & PFINT_OICR_SWINT_M) {
2757                ena_mask &= ~PFINT_OICR_SWINT_M;
2758                pf->sw_int_count++;
2759        }
2760
2761        if (oicr & PFINT_OICR_MAL_DETECT_M) {
2762                ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2763                set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2764        }
2765        if (oicr & PFINT_OICR_VFLR_M) {
2766                /* disable any further VFLR event notifications */
2767                if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2768                        u32 reg = rd32(hw, PFINT_OICR_ENA);
2769
2770                        reg &= ~PFINT_OICR_VFLR_M;
2771                        wr32(hw, PFINT_OICR_ENA, reg);
2772                } else {
2773                        ena_mask &= ~PFINT_OICR_VFLR_M;
2774                        set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2775                }
2776        }
2777
2778        if (oicr & PFINT_OICR_GRST_M) {
2779                u32 reset;
2780
2781                /* we have a reset warning */
2782                ena_mask &= ~PFINT_OICR_GRST_M;
2783                reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2784                        GLGEN_RSTAT_RESET_TYPE_S;
2785
2786                if (reset == ICE_RESET_CORER)
2787                        pf->corer_count++;
2788                else if (reset == ICE_RESET_GLOBR)
2789                        pf->globr_count++;
2790                else if (reset == ICE_RESET_EMPR)
2791                        pf->empr_count++;
2792                else
2793                        dev_dbg(dev, "Invalid reset type %d\n", reset);
2794
2795                /* If a reset cycle isn't already in progress, we set a bit in
2796                 * pf->state so that the service task can start a reset/rebuild.
2797                 */
2798                if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
2799                        if (reset == ICE_RESET_CORER)
2800                                set_bit(ICE_CORER_RECV, pf->state);
2801                        else if (reset == ICE_RESET_GLOBR)
2802                                set_bit(ICE_GLOBR_RECV, pf->state);
2803                        else
2804                                set_bit(ICE_EMPR_RECV, pf->state);
2805
2806                        /* There are couple of different bits at play here.
2807                         * hw->reset_ongoing indicates whether the hardware is
2808                         * in reset. This is set to true when a reset interrupt
2809                         * is received and set back to false after the driver
2810                         * has determined that the hardware is out of reset.
2811                         *
2812                         * ICE_RESET_OICR_RECV in pf->state indicates
2813                         * that a post reset rebuild is required before the
2814                         * driver is operational again. This is set above.
2815                         *
2816                         * As this is the start of the reset/rebuild cycle, set
2817                         * both to indicate that.
2818                         */
2819                        hw->reset_ongoing = true;
2820                }
2821        }
2822
2823        if (oicr & PFINT_OICR_TSYN_TX_M) {
2824                ena_mask &= ~PFINT_OICR_TSYN_TX_M;
2825                ice_ptp_process_ts(pf);
2826        }
2827
2828        if (oicr & PFINT_OICR_TSYN_EVNT_M) {
2829                u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
2830                u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
2831
2832                /* Save EVENTs from GTSYN register */
2833                pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
2834                                                     GLTSYN_STAT_EVENT1_M |
2835                                                     GLTSYN_STAT_EVENT2_M);
2836                ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
2837                kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
2838        }
2839
2840#define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
2841        if (oicr & ICE_AUX_CRIT_ERR) {
2842                struct iidc_event *event;
2843
2844                ena_mask &= ~ICE_AUX_CRIT_ERR;
2845                event = kzalloc(sizeof(*event), GFP_KERNEL);
2846                if (event) {
2847                        set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2848                        /* report the entire OICR value to AUX driver */
2849                        event->reg = oicr;
2850                        ice_send_event_to_aux(pf, event);
2851                        kfree(event);
2852                }
2853        }
2854
2855        /* Report any remaining unexpected interrupts */
2856        oicr &= ena_mask;
2857        if (oicr) {
2858                dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2859                /* If a critical error is pending there is no choice but to
2860                 * reset the device.
2861                 */
2862                if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
2863                            PFINT_OICR_ECC_ERR_M)) {
2864                        set_bit(ICE_PFR_REQ, pf->state);
2865                        ice_service_task_schedule(pf);
2866                }
2867        }
2868        ret = IRQ_HANDLED;
2869
2870        ice_service_task_schedule(pf);
2871        ice_irq_dynamic_ena(hw, NULL, NULL);
2872
2873        return ret;
2874}
2875
2876/**
2877 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2878 * @hw: pointer to HW structure
2879 */
2880static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2881{
2882        /* disable Admin queue Interrupt causes */
2883        wr32(hw, PFINT_FW_CTL,
2884             rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2885
2886        /* disable Mailbox queue Interrupt causes */
2887        wr32(hw, PFINT_MBX_CTL,
2888             rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2889
2890        wr32(hw, PFINT_SB_CTL,
2891             rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
2892
2893        /* disable Control queue Interrupt causes */
2894        wr32(hw, PFINT_OICR_CTL,
2895             rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2896
2897        ice_flush(hw);
2898}
2899
2900/**
2901 * ice_free_irq_msix_misc - Unroll misc vector setup
2902 * @pf: board private structure
2903 */
2904static void ice_free_irq_msix_misc(struct ice_pf *pf)
2905{
2906        struct ice_hw *hw = &pf->hw;
2907
2908        ice_dis_ctrlq_interrupts(hw);
2909
2910        /* disable OICR interrupt */
2911        wr32(hw, PFINT_OICR_ENA, 0);
2912        ice_flush(hw);
2913
2914        if (pf->msix_entries) {
2915                synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2916                devm_free_irq(ice_pf_to_dev(pf),
2917                              pf->msix_entries[pf->oicr_idx].vector, pf);
2918        }
2919
2920        pf->num_avail_sw_msix += 1;
2921        ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2922}
2923
2924/**
2925 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2926 * @hw: pointer to HW structure
2927 * @reg_idx: HW vector index to associate the control queue interrupts with
2928 */
2929static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2930{
2931        u32 val;
2932
2933        val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2934               PFINT_OICR_CTL_CAUSE_ENA_M);
2935        wr32(hw, PFINT_OICR_CTL, val);
2936
2937        /* enable Admin queue Interrupt causes */
2938        val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2939               PFINT_FW_CTL_CAUSE_ENA_M);
2940        wr32(hw, PFINT_FW_CTL, val);
2941
2942        /* enable Mailbox queue Interrupt causes */
2943        val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2944               PFINT_MBX_CTL_CAUSE_ENA_M);
2945        wr32(hw, PFINT_MBX_CTL, val);
2946
2947        /* This enables Sideband queue Interrupt causes */
2948        val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
2949               PFINT_SB_CTL_CAUSE_ENA_M);
2950        wr32(hw, PFINT_SB_CTL, val);
2951
2952        ice_flush(hw);
2953}
2954
2955/**
2956 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2957 * @pf: board private structure
2958 *
2959 * This sets up the handler for MSIX 0, which is used to manage the
2960 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2961 * when in MSI or Legacy interrupt mode.
2962 */
2963static int ice_req_irq_msix_misc(struct ice_pf *pf)
2964{
2965        struct device *dev = ice_pf_to_dev(pf);
2966        struct ice_hw *hw = &pf->hw;
2967        int oicr_idx, err = 0;
2968
2969        if (!pf->int_name[0])
2970                snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2971                         dev_driver_string(dev), dev_name(dev));
2972
2973        /* Do not request IRQ but do enable OICR interrupt since settings are
2974         * lost during reset. Note that this function is called only during
2975         * rebuild path and not while reset is in progress.
2976         */
2977        if (ice_is_reset_in_progress(pf->state))
2978                goto skip_req_irq;
2979
2980        /* reserve one vector in irq_tracker for misc interrupts */
2981        oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2982        if (oicr_idx < 0)
2983                return oicr_idx;
2984
2985        pf->num_avail_sw_msix -= 1;
2986        pf->oicr_idx = (u16)oicr_idx;
2987
2988        err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2989                               ice_misc_intr, 0, pf->int_name, pf);
2990        if (err) {
2991                dev_err(dev, "devm_request_irq for %s failed: %d\n",
2992                        pf->int_name, err);
2993                ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2994                pf->num_avail_sw_msix += 1;
2995                return err;
2996        }
2997
2998skip_req_irq:
2999        ice_ena_misc_vector(pf);
3000
3001        ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3002        wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3003             ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3004
3005        ice_flush(hw);
3006        ice_irq_dynamic_ena(hw, NULL, NULL);
3007
3008        return 0;
3009}
3010
3011/**
3012 * ice_napi_add - register NAPI handler for the VSI
3013 * @vsi: VSI for which NAPI handler is to be registered
3014 *
3015 * This function is only called in the driver's load path. Registering the NAPI
3016 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3017 * reset/rebuild, etc.)
3018 */
3019static void ice_napi_add(struct ice_vsi *vsi)
3020{
3021        int v_idx;
3022
3023        if (!vsi->netdev)
3024                return;
3025
3026        ice_for_each_q_vector(vsi, v_idx)
3027                netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3028                               ice_napi_poll, NAPI_POLL_WEIGHT);
3029}
3030
3031/**
3032 * ice_set_ops - set netdev and ethtools ops for the given netdev
3033 * @netdev: netdev instance
3034 */
3035static void ice_set_ops(struct net_device *netdev)
3036{
3037        struct ice_pf *pf = ice_netdev_to_pf(netdev);
3038
3039        if (ice_is_safe_mode(pf)) {
3040                netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3041                ice_set_ethtool_safe_mode_ops(netdev);
3042                return;
3043        }
3044
3045        netdev->netdev_ops = &ice_netdev_ops;
3046        netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3047        ice_set_ethtool_ops(netdev);
3048}
3049
3050/**
3051 * ice_set_netdev_features - set features for the given netdev
3052 * @netdev: netdev instance
3053 */
3054static void ice_set_netdev_features(struct net_device *netdev)
3055{
3056        struct ice_pf *pf = ice_netdev_to_pf(netdev);
3057        netdev_features_t csumo_features;
3058        netdev_features_t vlano_features;
3059        netdev_features_t dflt_features;
3060        netdev_features_t tso_features;
3061
3062        if (ice_is_safe_mode(pf)) {
3063                /* safe mode */
3064                netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3065                netdev->hw_features = netdev->features;
3066                return;
3067        }
3068
3069        dflt_features = NETIF_F_SG      |
3070                        NETIF_F_HIGHDMA |
3071                        NETIF_F_NTUPLE  |
3072                        NETIF_F_RXHASH;
3073
3074        csumo_features = NETIF_F_RXCSUM   |
3075                         NETIF_F_IP_CSUM  |
3076                         NETIF_F_SCTP_CRC |
3077                         NETIF_F_IPV6_CSUM;
3078
3079        vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3080                         NETIF_F_HW_VLAN_CTAG_TX     |
3081                         NETIF_F_HW_VLAN_CTAG_RX;
3082
3083        tso_features = NETIF_F_TSO                      |
3084                       NETIF_F_TSO_ECN                  |
3085                       NETIF_F_TSO6                     |
3086                       NETIF_F_GSO_GRE                  |
3087                       NETIF_F_GSO_UDP_TUNNEL           |
3088                       NETIF_F_GSO_GRE_CSUM             |
3089                       NETIF_F_GSO_UDP_TUNNEL_CSUM      |
3090                       NETIF_F_GSO_PARTIAL              |
3091                       NETIF_F_GSO_IPXIP4               |
3092                       NETIF_F_GSO_IPXIP6               |
3093                       NETIF_F_GSO_UDP_L4;
3094
3095        netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3096                                        NETIF_F_GSO_GRE_CSUM;
3097        /* set features that user can change */
3098        netdev->hw_features = dflt_features | csumo_features |
3099                              vlano_features | tso_features;
3100
3101        /* add support for HW_CSUM on packets with MPLS header */
3102        netdev->mpls_features =  NETIF_F_HW_CSUM;
3103
3104        /* enable features */
3105        netdev->features |= netdev->hw_features;
3106        /* encap and VLAN devices inherit default, csumo and tso features */
3107        netdev->hw_enc_features |= dflt_features | csumo_features |
3108                                   tso_features;
3109        netdev->vlan_features |= dflt_features | csumo_features |
3110                                 tso_features;
3111}
3112
3113/**
3114 * ice_cfg_netdev - Allocate, configure and register a netdev
3115 * @vsi: the VSI associated with the new netdev
3116 *
3117 * Returns 0 on success, negative value on failure
3118 */
3119static int ice_cfg_netdev(struct ice_vsi *vsi)
3120{
3121        struct ice_netdev_priv *np;
3122        struct net_device *netdev;
3123        u8 mac_addr[ETH_ALEN];
3124
3125        netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3126                                    vsi->alloc_rxq);
3127        if (!netdev)
3128                return -ENOMEM;
3129
3130        set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3131        vsi->netdev = netdev;
3132        np = netdev_priv(netdev);
3133        np->vsi = vsi;
3134
3135        ice_set_netdev_features(netdev);
3136
3137        ice_set_ops(netdev);
3138
3139        if (vsi->type == ICE_VSI_PF) {
3140                SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3141                ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3142                ether_addr_copy(netdev->dev_addr, mac_addr);
3143                ether_addr_copy(netdev->perm_addr, mac_addr);
3144        }
3145
3146        netdev->priv_flags |= IFF_UNICAST_FLT;
3147
3148        /* Setup netdev TC information */
3149        ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3150
3151        /* setup watchdog timeout value to be 5 second */
3152        netdev->watchdog_timeo = 5 * HZ;
3153
3154        netdev->min_mtu = ETH_MIN_MTU;
3155        netdev->max_mtu = ICE_MAX_MTU;
3156
3157        return 0;
3158}
3159
3160/**
3161 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3162 * @lut: Lookup table
3163 * @rss_table_size: Lookup table size
3164 * @rss_size: Range of queue number for hashing
3165 */
3166void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3167{
3168        u16 i;
3169
3170        for (i = 0; i < rss_table_size; i++)
3171                lut[i] = i % rss_size;
3172}
3173
3174/**
3175 * ice_pf_vsi_setup - Set up a PF VSI
3176 * @pf: board private structure
3177 * @pi: pointer to the port_info instance
3178 *
3179 * Returns pointer to the successfully allocated VSI software struct
3180 * on success, otherwise returns NULL on failure.
3181 */
3182static struct ice_vsi *
3183ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3184{
3185        return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3186}
3187
3188/**
3189 * ice_ctrl_vsi_setup - Set up a control VSI
3190 * @pf: board private structure
3191 * @pi: pointer to the port_info instance
3192 *
3193 * Returns pointer to the successfully allocated VSI software struct
3194 * on success, otherwise returns NULL on failure.
3195 */
3196static struct ice_vsi *
3197ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3198{
3199        return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3200}
3201
3202/**
3203 * ice_lb_vsi_setup - Set up a loopback VSI
3204 * @pf: board private structure
3205 * @pi: pointer to the port_info instance
3206 *
3207 * Returns pointer to the successfully allocated VSI software struct
3208 * on success, otherwise returns NULL on failure.
3209 */
3210struct ice_vsi *
3211ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3212{
3213        return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3214}
3215
3216/**
3217 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3218 * @netdev: network interface to be adjusted
3219 * @proto: unused protocol
3220 * @vid: VLAN ID to be added
3221 *
3222 * net_device_ops implementation for adding VLAN IDs
3223 */
3224static int
3225ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3226                    u16 vid)
3227{
3228        struct ice_netdev_priv *np = netdev_priv(netdev);
3229        struct ice_vsi *vsi = np->vsi;
3230        int ret;
3231
3232        /* VLAN 0 is added by default during load/reset */
3233        if (!vid)
3234                return 0;
3235
3236        /* Enable VLAN pruning when a VLAN other than 0 is added */
3237        if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3238                ret = ice_cfg_vlan_pruning(vsi, true, false);
3239                if (ret)
3240                        return ret;
3241        }
3242
3243        /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3244         * packets aren't pruned by the device's internal switch on Rx
3245         */
3246        ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3247        if (!ret)
3248                set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3249
3250        return ret;
3251}
3252
3253/**
3254 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3255 * @netdev: network interface to be adjusted
3256 * @proto: unused protocol
3257 * @vid: VLAN ID to be removed
3258 *
3259 * net_device_ops implementation for removing VLAN IDs
3260 */
3261static int
3262ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3263                     u16 vid)
3264{
3265        struct ice_netdev_priv *np = netdev_priv(netdev);
3266        struct ice_vsi *vsi = np->vsi;
3267        int ret;
3268
3269        /* don't allow removal of VLAN 0 */
3270        if (!vid)
3271                return 0;
3272
3273        /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3274         * information
3275         */
3276        ret = ice_vsi_kill_vlan(vsi, vid);
3277        if (ret)
3278                return ret;
3279
3280        /* Disable pruning when VLAN 0 is the only VLAN rule */
3281        if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3282                ret = ice_cfg_vlan_pruning(vsi, false, false);
3283
3284        set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3285        return ret;
3286}
3287
3288/**
3289 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3290 * @pf: board private structure
3291 *
3292 * Returns 0 on success, negative value on failure
3293 */
3294static int ice_setup_pf_sw(struct ice_pf *pf)
3295{
3296        struct ice_vsi *vsi;
3297        int status = 0;
3298
3299        if (ice_is_reset_in_progress(pf->state))
3300                return -EBUSY;
3301
3302        vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3303        if (!vsi)
3304                return -ENOMEM;
3305
3306        status = ice_cfg_netdev(vsi);
3307        if (status) {
3308                status = -ENODEV;
3309                goto unroll_vsi_setup;
3310        }
3311        /* netdev has to be configured before setting frame size */
3312        ice_vsi_cfg_frame_size(vsi);
3313
3314        /* Setup DCB netlink interface */
3315        ice_dcbnl_setup(vsi);
3316
3317        /* registering the NAPI handler requires both the queues and
3318         * netdev to be created, which are done in ice_pf_vsi_setup()
3319         * and ice_cfg_netdev() respectively
3320         */
3321        ice_napi_add(vsi);
3322
3323        status = ice_set_cpu_rx_rmap(vsi);
3324        if (status) {
3325                dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3326                        vsi->vsi_num, status);
3327                status = -EINVAL;
3328                goto unroll_napi_add;
3329        }
3330        status = ice_init_mac_fltr(pf);
3331        if (status)
3332                goto free_cpu_rx_map;
3333
3334        return status;
3335
3336free_cpu_rx_map:
3337        ice_free_cpu_rx_rmap(vsi);
3338
3339unroll_napi_add:
3340        if (vsi) {
3341                ice_napi_del(vsi);
3342                if (vsi->netdev) {
3343                        clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3344                        free_netdev(vsi->netdev);
3345                        vsi->netdev = NULL;
3346                }
3347        }
3348
3349unroll_vsi_setup:
3350        ice_vsi_release(vsi);
3351        return status;
3352}
3353
3354/**
3355 * ice_get_avail_q_count - Get count of queues in use
3356 * @pf_qmap: bitmap to get queue use count from
3357 * @lock: pointer to a mutex that protects access to pf_qmap
3358 * @size: size of the bitmap
3359 */
3360static u16
3361ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3362{
3363        unsigned long bit;
3364        u16 count = 0;
3365
3366        mutex_lock(lock);
3367        for_each_clear_bit(bit, pf_qmap, size)
3368                count++;
3369        mutex_unlock(lock);
3370
3371        return count;
3372}
3373
3374/**
3375 * ice_get_avail_txq_count - Get count of Tx queues in use
3376 * @pf: pointer to an ice_pf instance
3377 */
3378u16 ice_get_avail_txq_count(struct ice_pf *pf)
3379{
3380        return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3381                                     pf->max_pf_txqs);
3382}
3383
3384/**
3385 * ice_get_avail_rxq_count - Get count of Rx queues in use
3386 * @pf: pointer to an ice_pf instance
3387 */
3388u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3389{
3390        return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3391                                     pf->max_pf_rxqs);
3392}
3393
3394/**
3395 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3396 * @pf: board private structure to initialize
3397 */
3398static void ice_deinit_pf(struct ice_pf *pf)
3399{
3400        ice_service_task_stop(pf);
3401        mutex_destroy(&pf->sw_mutex);
3402        mutex_destroy(&pf->tc_mutex);
3403        mutex_destroy(&pf->avail_q_mutex);
3404
3405        if (pf->avail_txqs) {
3406                bitmap_free(pf->avail_txqs);
3407                pf->avail_txqs = NULL;
3408        }
3409
3410        if (pf->avail_rxqs) {
3411                bitmap_free(pf->avail_rxqs);
3412                pf->avail_rxqs = NULL;
3413        }
3414
3415        if (pf->ptp.clock)
3416                ptp_clock_unregister(pf->ptp.clock);
3417}
3418
3419/**
3420 * ice_set_pf_caps - set PFs capability flags
3421 * @pf: pointer to the PF instance
3422 */
3423static void ice_set_pf_caps(struct ice_pf *pf)
3424{
3425        struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3426
3427        clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3428        clear_bit(ICE_FLAG_AUX_ENA, pf->flags);
3429        if (func_caps->common_cap.rdma) {
3430                set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3431                set_bit(ICE_FLAG_AUX_ENA, pf->flags);
3432        }
3433        clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3434        if (func_caps->common_cap.dcb)
3435                set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3436        clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3437        if (func_caps->common_cap.sr_iov_1_1) {
3438                set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3439                pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3440                                              ICE_MAX_VF_COUNT);
3441        }
3442        clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3443        if (func_caps->common_cap.rss_table_size)
3444                set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3445
3446        clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3447        if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3448                u16 unused;
3449
3450                /* ctrl_vsi_idx will be set to a valid value when flow director
3451                 * is setup by ice_init_fdir
3452                 */
3453                pf->ctrl_vsi_idx = ICE_NO_VSI;
3454                set_bit(ICE_FLAG_FD_ENA, pf->flags);
3455                /* force guaranteed filter pool for PF */
3456                ice_alloc_fd_guar_item(&pf->hw, &unused,
3457                                       func_caps->fd_fltr_guar);
3458                /* force shared filter pool for PF */
3459                ice_alloc_fd_shrd_item(&pf->hw, &unused,
3460                                       func_caps->fd_fltr_best_effort);
3461        }
3462
3463        clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3464        if (func_caps->common_cap.ieee_1588)
3465                set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3466
3467        pf->max_pf_txqs = func_caps->common_cap.num_txq;
3468        pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3469}
3470
3471/**
3472 * ice_init_pf - Initialize general software structures (struct ice_pf)
3473 * @pf: board private structure to initialize
3474 */
3475static int ice_init_pf(struct ice_pf *pf)
3476{
3477        ice_set_pf_caps(pf);
3478
3479        mutex_init(&pf->sw_mutex);
3480        mutex_init(&pf->tc_mutex);
3481
3482        INIT_HLIST_HEAD(&pf->aq_wait_list);
3483        spin_lock_init(&pf->aq_wait_lock);
3484        init_waitqueue_head(&pf->aq_wait_queue);
3485
3486        init_waitqueue_head(&pf->reset_wait_queue);
3487
3488        /* setup service timer and periodic service task */
3489        timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3490        pf->serv_tmr_period = HZ;
3491        INIT_WORK(&pf->serv_task, ice_service_task);
3492        clear_bit(ICE_SERVICE_SCHED, pf->state);
3493
3494        mutex_init(&pf->avail_q_mutex);
3495        pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3496        if (!pf->avail_txqs)
3497                return -ENOMEM;
3498
3499        pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3500        if (!pf->avail_rxqs) {
3501                devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3502                pf->avail_txqs = NULL;
3503                return -ENOMEM;
3504        }
3505
3506        return 0;
3507}
3508
3509/**
3510 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3511 * @pf: board private structure
3512 *
3513 * compute the number of MSIX vectors required (v_budget) and request from
3514 * the OS. Return the number of vectors reserved or negative on failure
3515 */
3516static int ice_ena_msix_range(struct ice_pf *pf)
3517{
3518        int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3519        struct device *dev = ice_pf_to_dev(pf);
3520        int needed, err, i;
3521
3522        v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3523        num_cpus = num_online_cpus();
3524
3525        /* reserve for LAN miscellaneous handler */
3526        needed = ICE_MIN_LAN_OICR_MSIX;
3527        if (v_left < needed)
3528                goto no_hw_vecs_left_err;
3529        v_budget += needed;
3530        v_left -= needed;
3531
3532        /* reserve for flow director */
3533        if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3534                needed = ICE_FDIR_MSIX;
3535                if (v_left < needed)
3536                        goto no_hw_vecs_left_err;
3537                v_budget += needed;
3538                v_left -= needed;
3539        }
3540
3541        /* total used for non-traffic vectors */
3542        v_other = v_budget;
3543
3544        /* reserve vectors for LAN traffic */
3545        needed = num_cpus;
3546        if (v_left < needed)
3547                goto no_hw_vecs_left_err;
3548        pf->num_lan_msix = needed;
3549        v_budget += needed;
3550        v_left -= needed;
3551
3552        /* reserve vectors for RDMA auxiliary driver */
3553        if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3554                needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3555                if (v_left < needed)
3556                        goto no_hw_vecs_left_err;
3557                pf->num_rdma_msix = needed;
3558                v_budget += needed;
3559                v_left -= needed;
3560        }
3561
3562        pf->msix_entries = devm_kcalloc(dev, v_budget,
3563                                        sizeof(*pf->msix_entries), GFP_KERNEL);
3564        if (!pf->msix_entries) {
3565                err = -ENOMEM;
3566                goto exit_err;
3567        }
3568
3569        for (i = 0; i < v_budget; i++)
3570                pf->msix_entries[i].entry = i;
3571
3572        /* actually reserve the vectors */
3573        v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3574                                         ICE_MIN_MSIX, v_budget);
3575        if (v_actual < 0) {
3576                dev_err(dev, "unable to reserve MSI-X vectors\n");
3577                err = v_actual;
3578                goto msix_err;
3579        }
3580
3581        if (v_actual < v_budget) {
3582                dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3583                         v_budget, v_actual);
3584
3585                if (v_actual < ICE_MIN_MSIX) {
3586                        /* error if we can't get minimum vectors */
3587                        pci_disable_msix(pf->pdev);
3588                        err = -ERANGE;
3589                        goto msix_err;
3590                } else {
3591                        int v_remain = v_actual - v_other;
3592                        int v_rdma = 0, v_min_rdma = 0;
3593
3594                        if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3595                                /* Need at least 1 interrupt in addition to
3596                                 * AEQ MSIX
3597                                 */
3598                                v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3599                                v_min_rdma = ICE_MIN_RDMA_MSIX;
3600                        }
3601
3602                        if (v_actual == ICE_MIN_MSIX ||
3603                            v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
3604                                dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
3605                                clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3606
3607                                pf->num_rdma_msix = 0;
3608                                pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3609                        } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3610                                   (v_remain - v_rdma < v_rdma)) {
3611                                /* Support minimum RDMA and give remaining
3612                                 * vectors to LAN MSIX
3613                                 */
3614                                pf->num_rdma_msix = v_min_rdma;
3615                                pf->num_lan_msix = v_remain - v_min_rdma;
3616                        } else {
3617                                /* Split remaining MSIX with RDMA after
3618                                 * accounting for AEQ MSIX
3619                                 */
3620                                pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3621                                                    ICE_RDMA_NUM_AEQ_MSIX;
3622                                pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3623                        }
3624
3625                        dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3626                                   pf->num_lan_msix);
3627
3628                        if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
3629                                dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
3630                                           pf->num_rdma_msix);
3631                }
3632        }
3633
3634        return v_actual;
3635
3636msix_err:
3637        devm_kfree(dev, pf->msix_entries);
3638        goto exit_err;
3639
3640no_hw_vecs_left_err:
3641        dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3642                needed, v_left);
3643        err = -ERANGE;
3644exit_err:
3645        pf->num_rdma_msix = 0;
3646        pf->num_lan_msix = 0;
3647        return err;
3648}
3649
3650/**
3651 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3652 * @pf: board private structure
3653 */
3654static void ice_dis_msix(struct ice_pf *pf)
3655{
3656        pci_disable_msix(pf->pdev);
3657        devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3658        pf->msix_entries = NULL;
3659}
3660
3661/**
3662 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3663 * @pf: board private structure
3664 */
3665static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3666{
3667        ice_dis_msix(pf);
3668
3669        if (pf->irq_tracker) {
3670                devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3671                pf->irq_tracker = NULL;
3672        }
3673}
3674
3675/**
3676 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3677 * @pf: board private structure to initialize
3678 */
3679static int ice_init_interrupt_scheme(struct ice_pf *pf)
3680{
3681        int vectors;
3682
3683        vectors = ice_ena_msix_range(pf);
3684
3685        if (vectors < 0)
3686                return vectors;
3687
3688        /* set up vector assignment tracking */
3689        pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3690                                       struct_size(pf->irq_tracker, list, vectors),
3691                                       GFP_KERNEL);
3692        if (!pf->irq_tracker) {
3693                ice_dis_msix(pf);
3694                return -ENOMEM;
3695        }
3696
3697        /* populate SW interrupts pool with number of OS granted IRQs. */
3698        pf->num_avail_sw_msix = (u16)vectors;
3699        pf->irq_tracker->num_entries = (u16)vectors;
3700        pf->irq_tracker->end = pf->irq_tracker->num_entries;
3701
3702        return 0;
3703}
3704
3705/**
3706 * ice_is_wol_supported - check if WoL is supported
3707 * @hw: pointer to hardware info
3708 *
3709 * Check if WoL is supported based on the HW configuration.
3710 * Returns true if NVM supports and enables WoL for this port, false otherwise
3711 */
3712bool ice_is_wol_supported(struct ice_hw *hw)
3713{
3714        u16 wol_ctrl;
3715
3716        /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3717         * word) indicates WoL is not supported on the corresponding PF ID.
3718         */
3719        if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3720                return false;
3721
3722        return !(BIT(hw->port_info->lport) & wol_ctrl);
3723}
3724
3725/**
3726 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3727 * @vsi: VSI being changed
3728 * @new_rx: new number of Rx queues
3729 * @new_tx: new number of Tx queues
3730 *
3731 * Only change the number of queues if new_tx, or new_rx is non-0.
3732 *
3733 * Returns 0 on success.
3734 */
3735int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3736{
3737        struct ice_pf *pf = vsi->back;
3738        int err = 0, timeout = 50;
3739
3740        if (!new_rx && !new_tx)
3741                return -EINVAL;
3742
3743        while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3744                timeout--;
3745                if (!timeout)
3746                        return -EBUSY;
3747                usleep_range(1000, 2000);
3748        }
3749
3750        if (new_tx)
3751                vsi->req_txq = (u16)new_tx;
3752        if (new_rx)
3753                vsi->req_rxq = (u16)new_rx;
3754
3755        /* set for the next time the netdev is started */
3756        if (!netif_running(vsi->netdev)) {
3757                ice_vsi_rebuild(vsi, false);
3758                dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3759                goto done;
3760        }
3761
3762        ice_vsi_close(vsi);
3763        ice_vsi_rebuild(vsi, false);
3764        ice_pf_dcb_recfg(pf);
3765        ice_vsi_open(vsi);
3766done:
3767        clear_bit(ICE_CFG_BUSY, pf->state);
3768        return err;
3769}
3770
3771/**
3772 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3773 * @pf: PF to configure
3774 *
3775 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3776 * VSI can still Tx/Rx VLAN tagged packets.
3777 */
3778static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3779{
3780        struct ice_vsi *vsi = ice_get_main_vsi(pf);
3781        struct ice_vsi_ctx *ctxt;
3782        enum ice_status status;
3783        struct ice_hw *hw;
3784
3785        if (!vsi)
3786                return;
3787
3788        ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3789        if (!ctxt)
3790                return;
3791
3792        hw = &pf->hw;
3793        ctxt->info = vsi->info;
3794
3795        ctxt->info.valid_sections =
3796                cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3797                            ICE_AQ_VSI_PROP_SECURITY_VALID |
3798                            ICE_AQ_VSI_PROP_SW_VALID);
3799
3800        /* disable VLAN anti-spoof */
3801        ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3802                                  ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3803
3804        /* disable VLAN pruning and keep all other settings */
3805        ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3806
3807        /* allow all VLANs on Tx and don't strip on Rx */
3808        ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3809                ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3810
3811        status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3812        if (status) {
3813                dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3814                        ice_stat_str(status),
3815                        ice_aq_str(hw->adminq.sq_last_status));
3816        } else {
3817                vsi->info.sec_flags = ctxt->info.sec_flags;
3818                vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3819                vsi->info.vlan_flags = ctxt->info.vlan_flags;
3820        }
3821
3822        kfree(ctxt);
3823}
3824
3825/**
3826 * ice_log_pkg_init - log result of DDP package load
3827 * @hw: pointer to hardware info
3828 * @status: status of package load
3829 */
3830static void
3831ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3832{
3833        struct ice_pf *pf = (struct ice_pf *)hw->back;
3834        struct device *dev = ice_pf_to_dev(pf);
3835
3836        switch (*status) {
3837        case ICE_SUCCESS:
3838                /* The package download AdminQ command returned success because
3839                 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3840                 * already a package loaded on the device.
3841                 */
3842                if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3843                    hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3844                    hw->pkg_ver.update == hw->active_pkg_ver.update &&
3845                    hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3846                    !memcmp(hw->pkg_name, hw->active_pkg_name,
3847                            sizeof(hw->pkg_name))) {
3848                        if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3849                                dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3850                                         hw->active_pkg_name,
3851                                         hw->active_pkg_ver.major,
3852                                         hw->active_pkg_ver.minor,
3853                                         hw->active_pkg_ver.update,
3854                                         hw->active_pkg_ver.draft);
3855                        else
3856                                dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3857                                         hw->active_pkg_name,
3858                                         hw->active_pkg_ver.major,
3859                                         hw->active_pkg_ver.minor,
3860                                         hw->active_pkg_ver.update,
3861                                         hw->active_pkg_ver.draft);
3862                } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3863                           hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3864                        dev_err(dev, "The device has a DDP package that is not supported by the driver.  The device has package '%s' version %d.%d.x.x.  The driver requires version %d.%d.x.x.  Entering Safe Mode.\n",
3865                                hw->active_pkg_name,
3866                                hw->active_pkg_ver.major,
3867                                hw->active_pkg_ver.minor,
3868                                ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3869                        *status = ICE_ERR_NOT_SUPPORTED;
3870                } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3871                           hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3872                        dev_info(dev, "The driver could not load the DDP package file because a compatible DDP package is already present on the device.  The device has package '%s' version %d.%d.%d.%d.  The package file found by the driver: '%s' version %d.%d.%d.%d.\n",
3873                                 hw->active_pkg_name,
3874                                 hw->active_pkg_ver.major,
3875                                 hw->active_pkg_ver.minor,
3876                                 hw->active_pkg_ver.update,
3877                                 hw->active_pkg_ver.draft,
3878                                 hw->pkg_name,
3879                                 hw->pkg_ver.major,
3880                                 hw->pkg_ver.minor,
3881                                 hw->pkg_ver.update,
3882                                 hw->pkg_ver.draft);
3883                } else {
3884                        dev_err(dev, "An unknown error occurred when loading the DDP package, please reboot the system.  If the problem persists, update the NVM.  Entering Safe Mode.\n");
3885                        *status = ICE_ERR_NOT_SUPPORTED;
3886                }
3887                break;
3888        case ICE_ERR_FW_DDP_MISMATCH:
3889                dev_err(dev, "The firmware loaded on the device is not compatible with the DDP package.  Please update the device's NVM.  Entering safe mode.\n");
3890                break;
3891        case ICE_ERR_BUF_TOO_SHORT:
3892        case ICE_ERR_CFG:
3893                dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3894                break;
3895        case ICE_ERR_NOT_SUPPORTED:
3896                /* Package File version not supported */
3897                if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3898                    (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3899                     hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3900                        dev_err(dev, "The DDP package file version is higher than the driver supports.  Please use an updated driver.  Entering Safe Mode.\n");
3901                else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3902                         (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3903                          hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3904                        dev_err(dev, "The DDP package file version is lower than the driver supports.  The driver requires version %d.%d.x.x.  Please use an updated DDP Package file.  Entering Safe Mode.\n",
3905                                ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3906                break;
3907        case ICE_ERR_AQ_ERROR:
3908                switch (hw->pkg_dwnld_status) {
3909                case ICE_AQ_RC_ENOSEC:
3910                case ICE_AQ_RC_EBADSIG:
3911                        dev_err(dev, "The DDP package could not be loaded because its signature is not valid.  Please use a valid DDP Package.  Entering Safe Mode.\n");
3912                        return;
3913                case ICE_AQ_RC_ESVN:
3914                        dev_err(dev, "The DDP Package could not be loaded because its security revision is too low.  Please use an updated DDP Package.  Entering Safe Mode.\n");
3915                        return;
3916                case ICE_AQ_RC_EBADMAN:
3917                case ICE_AQ_RC_EBADBUF:
3918                        dev_err(dev, "An error occurred on the device while loading the DDP package.  The device will be reset.\n");
3919                        /* poll for reset to complete */
3920                        if (ice_check_reset(hw))
3921                                dev_err(dev, "Error resetting device. Please reload the driver\n");
3922                        return;
3923                default:
3924                        break;
3925                }
3926                fallthrough;
3927        default: