linux/drivers/net/ethernet/intel/i40e/i40e_main.c
<<
>>
Prefs
   1// SPDX-License-Identifier: GPL-2.0
   2/* Copyright(c) 2013 - 2018 Intel Corporation. */
   3
   4#include <linux/etherdevice.h>
   5#include <linux/of_net.h>
   6#include <linux/pci.h>
   7#include <linux/bpf.h>
   8
   9/* Local includes */
  10#include "i40e.h"
  11#include "i40e_diag.h"
  12#include "i40e_xsk.h"
  13#include <net/udp_tunnel.h>
  14#include <net/xdp_sock.h>
  15/* All i40e tracepoints are defined by the include below, which
  16 * must be included exactly once across the whole kernel with
  17 * CREATE_TRACE_POINTS defined
  18 */
  19#define CREATE_TRACE_POINTS
  20#include "i40e_trace.h"
  21
  22const char i40e_driver_name[] = "i40e";
  23static const char i40e_driver_string[] =
  24                        "Intel(R) Ethernet Connection XL710 Network Driver";
  25
  26#define DRV_KERN "-k"
  27
  28#define DRV_VERSION_MAJOR 2
  29#define DRV_VERSION_MINOR 8
  30#define DRV_VERSION_BUILD 20
  31#define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  32             __stringify(DRV_VERSION_MINOR) "." \
  33             __stringify(DRV_VERSION_BUILD)    DRV_KERN
  34const char i40e_driver_version_str[] = DRV_VERSION;
  35static const char i40e_copyright[] = "Copyright (c) 2013 - 2019 Intel Corporation.";
  36
  37/* a bit of forward declarations */
  38static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  39static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
  40static int i40e_add_vsi(struct i40e_vsi *vsi);
  41static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  42static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  43static int i40e_setup_misc_vector(struct i40e_pf *pf);
  44static void i40e_determine_queue_usage(struct i40e_pf *pf);
  45static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  46static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
  47static int i40e_reset(struct i40e_pf *pf);
  48static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
  49static int i40e_setup_misc_vector_for_recovery_mode(struct i40e_pf *pf);
  50static int i40e_restore_interrupt_scheme(struct i40e_pf *pf);
  51static bool i40e_check_recovery_mode(struct i40e_pf *pf);
  52static int i40e_init_recovery_mode(struct i40e_pf *pf, struct i40e_hw *hw);
  53static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  54static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  55static int i40e_get_capabilities(struct i40e_pf *pf,
  56                                 enum i40e_admin_queue_opc list_type);
  57
  58
  59/* i40e_pci_tbl - PCI Device ID Table
  60 *
  61 * Last entry must be all 0s
  62 *
  63 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  64 *   Class, Class Mask, private data (not used) }
  65 */
  66static const struct pci_device_id i40e_pci_tbl[] = {
  67        {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  68        {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  69        {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  70        {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  71        {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  72        {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  73        {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  74        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  75        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  76        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_BC), 0},
  77        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_SFP), 0},
  78        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_B), 0},
  79        {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
  80        {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
  81        {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  82        {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  83        {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  84        {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
  85        {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  86        {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  87        {PCI_VDEVICE(INTEL, I40E_DEV_ID_X710_N3000), 0},
  88        {PCI_VDEVICE(INTEL, I40E_DEV_ID_XXV710_N3000), 0},
  89        {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
  90        {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
  91        /* required last entry */
  92        {0, }
  93};
  94MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
  95
  96#define I40E_MAX_VF_COUNT 128
  97static int debug = -1;
  98module_param(debug, uint, 0);
  99MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
 100
 101MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 102MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 103MODULE_LICENSE("GPL v2");
 104MODULE_VERSION(DRV_VERSION);
 105
 106static struct workqueue_struct *i40e_wq;
 107
 108/**
 109 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 110 * @hw:   pointer to the HW structure
 111 * @mem:  ptr to mem struct to fill out
 112 * @size: size of memory requested
 113 * @alignment: what to align the allocation to
 114 **/
 115int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 116                            u64 size, u32 alignment)
 117{
 118        struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 119
 120        mem->size = ALIGN(size, alignment);
 121        mem->va = dma_alloc_coherent(&pf->pdev->dev, mem->size, &mem->pa,
 122                                     GFP_KERNEL);
 123        if (!mem->va)
 124                return -ENOMEM;
 125
 126        return 0;
 127}
 128
 129/**
 130 * i40e_free_dma_mem_d - OS specific memory free for shared code
 131 * @hw:   pointer to the HW structure
 132 * @mem:  ptr to mem struct to free
 133 **/
 134int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 135{
 136        struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 137
 138        dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 139        mem->va = NULL;
 140        mem->pa = 0;
 141        mem->size = 0;
 142
 143        return 0;
 144}
 145
 146/**
 147 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 148 * @hw:   pointer to the HW structure
 149 * @mem:  ptr to mem struct to fill out
 150 * @size: size of memory requested
 151 **/
 152int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 153                             u32 size)
 154{
 155        mem->size = size;
 156        mem->va = kzalloc(size, GFP_KERNEL);
 157
 158        if (!mem->va)
 159                return -ENOMEM;
 160
 161        return 0;
 162}
 163
 164/**
 165 * i40e_free_virt_mem_d - OS specific memory free for shared code
 166 * @hw:   pointer to the HW structure
 167 * @mem:  ptr to mem struct to free
 168 **/
 169int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 170{
 171        /* it's ok to kfree a NULL pointer */
 172        kfree(mem->va);
 173        mem->va = NULL;
 174        mem->size = 0;
 175
 176        return 0;
 177}
 178
 179/**
 180 * i40e_get_lump - find a lump of free generic resource
 181 * @pf: board private structure
 182 * @pile: the pile of resource to search
 183 * @needed: the number of items needed
 184 * @id: an owner id to stick on the items assigned
 185 *
 186 * Returns the base item index of the lump, or negative for error
 187 *
 188 * The search_hint trick and lack of advanced fit-finding only work
 189 * because we're highly likely to have all the same size lump requests.
 190 * Linear search time and any fragmentation should be minimal.
 191 **/
 192static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 193                         u16 needed, u16 id)
 194{
 195        int ret = -ENOMEM;
 196        int i, j;
 197
 198        if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 199                dev_info(&pf->pdev->dev,
 200                         "param err: pile=%s needed=%d id=0x%04x\n",
 201                         pile ? "<valid>" : "<null>", needed, id);
 202                return -EINVAL;
 203        }
 204
 205        /* start the linear search with an imperfect hint */
 206        i = pile->search_hint;
 207        while (i < pile->num_entries) {
 208                /* skip already allocated entries */
 209                if (pile->list[i] & I40E_PILE_VALID_BIT) {
 210                        i++;
 211                        continue;
 212                }
 213
 214                /* do we have enough in this lump? */
 215                for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 216                        if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 217                                break;
 218                }
 219
 220                if (j == needed) {
 221                        /* there was enough, so assign it to the requestor */
 222                        for (j = 0; j < needed; j++)
 223                                pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 224                        ret = i;
 225                        pile->search_hint = i + j;
 226                        break;
 227                }
 228
 229                /* not enough, so skip over it and continue looking */
 230                i += j;
 231        }
 232
 233        return ret;
 234}
 235
 236/**
 237 * i40e_put_lump - return a lump of generic resource
 238 * @pile: the pile of resource to search
 239 * @index: the base item index
 240 * @id: the owner id of the items assigned
 241 *
 242 * Returns the count of items in the lump
 243 **/
 244static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 245{
 246        int valid_id = (id | I40E_PILE_VALID_BIT);
 247        int count = 0;
 248        int i;
 249
 250        if (!pile || index >= pile->num_entries)
 251                return -EINVAL;
 252
 253        for (i = index;
 254             i < pile->num_entries && pile->list[i] == valid_id;
 255             i++) {
 256                pile->list[i] = 0;
 257                count++;
 258        }
 259
 260        if (count && index < pile->search_hint)
 261                pile->search_hint = index;
 262
 263        return count;
 264}
 265
 266/**
 267 * i40e_find_vsi_from_id - searches for the vsi with the given id
 268 * @pf: the pf structure to search for the vsi
 269 * @id: id of the vsi it is searching for
 270 **/
 271struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 272{
 273        int i;
 274
 275        for (i = 0; i < pf->num_alloc_vsi; i++)
 276                if (pf->vsi[i] && (pf->vsi[i]->id == id))
 277                        return pf->vsi[i];
 278
 279        return NULL;
 280}
 281
 282/**
 283 * i40e_service_event_schedule - Schedule the service task to wake up
 284 * @pf: board private structure
 285 *
 286 * If not already scheduled, this puts the task into the work queue
 287 **/
 288void i40e_service_event_schedule(struct i40e_pf *pf)
 289{
 290        if ((!test_bit(__I40E_DOWN, pf->state) &&
 291             !test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state)) ||
 292              test_bit(__I40E_RECOVERY_MODE, pf->state))
 293                queue_work(i40e_wq, &pf->service_task);
 294}
 295
 296/**
 297 * i40e_tx_timeout - Respond to a Tx Hang
 298 * @netdev: network interface device structure
 299 *
 300 * If any port has noticed a Tx timeout, it is likely that the whole
 301 * device is munged, not just the one netdev port, so go for the full
 302 * reset.
 303 **/
 304static void i40e_tx_timeout(struct net_device *netdev)
 305{
 306        struct i40e_netdev_priv *np = netdev_priv(netdev);
 307        struct i40e_vsi *vsi = np->vsi;
 308        struct i40e_pf *pf = vsi->back;
 309        struct i40e_ring *tx_ring = NULL;
 310        unsigned int i, hung_queue = 0;
 311        u32 head, val;
 312
 313        pf->tx_timeout_count++;
 314
 315        /* find the stopped queue the same way the stack does */
 316        for (i = 0; i < netdev->num_tx_queues; i++) {
 317                struct netdev_queue *q;
 318                unsigned long trans_start;
 319
 320                q = netdev_get_tx_queue(netdev, i);
 321                trans_start = q->trans_start;
 322                if (netif_xmit_stopped(q) &&
 323                    time_after(jiffies,
 324                               (trans_start + netdev->watchdog_timeo))) {
 325                        hung_queue = i;
 326                        break;
 327                }
 328        }
 329
 330        if (i == netdev->num_tx_queues) {
 331                netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 332        } else {
 333                /* now that we have an index, find the tx_ring struct */
 334                for (i = 0; i < vsi->num_queue_pairs; i++) {
 335                        if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 336                                if (hung_queue ==
 337                                    vsi->tx_rings[i]->queue_index) {
 338                                        tx_ring = vsi->tx_rings[i];
 339                                        break;
 340                                }
 341                        }
 342                }
 343        }
 344
 345        if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 346                pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 347        else if (time_before(jiffies,
 348                      (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 349                return;   /* don't do any new action before the next timeout */
 350
 351        /* don't kick off another recovery if one is already pending */
 352        if (test_and_set_bit(__I40E_TIMEOUT_RECOVERY_PENDING, pf->state))
 353                return;
 354
 355        if (tx_ring) {
 356                head = i40e_get_head(tx_ring);
 357                /* Read interrupt register */
 358                if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 359                        val = rd32(&pf->hw,
 360                             I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 361                                                tx_ring->vsi->base_vector - 1));
 362                else
 363                        val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 364
 365                netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
 366                            vsi->seid, hung_queue, tx_ring->next_to_clean,
 367                            head, tx_ring->next_to_use,
 368                            readl(tx_ring->tail), val);
 369        }
 370
 371        pf->tx_timeout_last_recovery = jiffies;
 372        netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 373                    pf->tx_timeout_recovery_level, hung_queue);
 374
 375        switch (pf->tx_timeout_recovery_level) {
 376        case 1:
 377                set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
 378                break;
 379        case 2:
 380                set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
 381                break;
 382        case 3:
 383                set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
 384                break;
 385        default:
 386                netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 387                break;
 388        }
 389
 390        i40e_service_event_schedule(pf);
 391        pf->tx_timeout_recovery_level++;
 392}
 393
 394/**
 395 * i40e_get_vsi_stats_struct - Get System Network Statistics
 396 * @vsi: the VSI we care about
 397 *
 398 * Returns the address of the device statistics structure.
 399 * The statistics are actually updated from the service task.
 400 **/
 401struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 402{
 403        return &vsi->net_stats;
 404}
 405
 406/**
 407 * i40e_get_netdev_stats_struct_tx - populate stats from a Tx ring
 408 * @ring: Tx ring to get statistics from
 409 * @stats: statistics entry to be updated
 410 **/
 411static void i40e_get_netdev_stats_struct_tx(struct i40e_ring *ring,
 412                                            struct rtnl_link_stats64 *stats)
 413{
 414        u64 bytes, packets;
 415        unsigned int start;
 416
 417        do {
 418                start = u64_stats_fetch_begin_irq(&ring->syncp);
 419                packets = ring->stats.packets;
 420                bytes   = ring->stats.bytes;
 421        } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
 422
 423        stats->tx_packets += packets;
 424        stats->tx_bytes   += bytes;
 425}
 426
 427/**
 428 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 429 * @netdev: network interface device structure
 430 * @stats: data structure to store statistics
 431 *
 432 * Returns the address of the device statistics structure.
 433 * The statistics are actually updated from the service task.
 434 **/
 435static void i40e_get_netdev_stats_struct(struct net_device *netdev,
 436                                  struct rtnl_link_stats64 *stats)
 437{
 438        struct i40e_netdev_priv *np = netdev_priv(netdev);
 439        struct i40e_vsi *vsi = np->vsi;
 440        struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 441        struct i40e_ring *ring;
 442        int i;
 443
 444        if (test_bit(__I40E_VSI_DOWN, vsi->state))
 445                return;
 446
 447        if (!vsi->tx_rings)
 448                return;
 449
 450        rcu_read_lock();
 451        for (i = 0; i < vsi->num_queue_pairs; i++) {
 452                u64 bytes, packets;
 453                unsigned int start;
 454
 455                ring = READ_ONCE(vsi->tx_rings[i]);
 456                if (!ring)
 457                        continue;
 458                i40e_get_netdev_stats_struct_tx(ring, stats);
 459
 460                if (i40e_enabled_xdp_vsi(vsi)) {
 461                        ring++;
 462                        i40e_get_netdev_stats_struct_tx(ring, stats);
 463                }
 464
 465                ring++;
 466                do {
 467                        start   = u64_stats_fetch_begin_irq(&ring->syncp);
 468                        packets = ring->stats.packets;
 469                        bytes   = ring->stats.bytes;
 470                } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
 471
 472                stats->rx_packets += packets;
 473                stats->rx_bytes   += bytes;
 474
 475        }
 476        rcu_read_unlock();
 477
 478        /* following stats updated by i40e_watchdog_subtask() */
 479        stats->multicast        = vsi_stats->multicast;
 480        stats->tx_errors        = vsi_stats->tx_errors;
 481        stats->tx_dropped       = vsi_stats->tx_dropped;
 482        stats->rx_errors        = vsi_stats->rx_errors;
 483        stats->rx_dropped       = vsi_stats->rx_dropped;
 484        stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 485        stats->rx_length_errors = vsi_stats->rx_length_errors;
 486}
 487
 488/**
 489 * i40e_vsi_reset_stats - Resets all stats of the given vsi
 490 * @vsi: the VSI to have its stats reset
 491 **/
 492void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 493{
 494        struct rtnl_link_stats64 *ns;
 495        int i;
 496
 497        if (!vsi)
 498                return;
 499
 500        ns = i40e_get_vsi_stats_struct(vsi);
 501        memset(ns, 0, sizeof(*ns));
 502        memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 503        memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 504        memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 505        if (vsi->rx_rings && vsi->rx_rings[0]) {
 506                for (i = 0; i < vsi->num_queue_pairs; i++) {
 507                        memset(&vsi->rx_rings[i]->stats, 0,
 508                               sizeof(vsi->rx_rings[i]->stats));
 509                        memset(&vsi->rx_rings[i]->rx_stats, 0,
 510                               sizeof(vsi->rx_rings[i]->rx_stats));
 511                        memset(&vsi->tx_rings[i]->stats, 0,
 512                               sizeof(vsi->tx_rings[i]->stats));
 513                        memset(&vsi->tx_rings[i]->tx_stats, 0,
 514                               sizeof(vsi->tx_rings[i]->tx_stats));
 515                }
 516        }
 517        vsi->stat_offsets_loaded = false;
 518}
 519
 520/**
 521 * i40e_pf_reset_stats - Reset all of the stats for the given PF
 522 * @pf: the PF to be reset
 523 **/
 524void i40e_pf_reset_stats(struct i40e_pf *pf)
 525{
 526        int i;
 527
 528        memset(&pf->stats, 0, sizeof(pf->stats));
 529        memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 530        pf->stat_offsets_loaded = false;
 531
 532        for (i = 0; i < I40E_MAX_VEB; i++) {
 533                if (pf->veb[i]) {
 534                        memset(&pf->veb[i]->stats, 0,
 535                               sizeof(pf->veb[i]->stats));
 536                        memset(&pf->veb[i]->stats_offsets, 0,
 537                               sizeof(pf->veb[i]->stats_offsets));
 538                        memset(&pf->veb[i]->tc_stats, 0,
 539                               sizeof(pf->veb[i]->tc_stats));
 540                        memset(&pf->veb[i]->tc_stats_offsets, 0,
 541                               sizeof(pf->veb[i]->tc_stats_offsets));
 542                        pf->veb[i]->stat_offsets_loaded = false;
 543                }
 544        }
 545        pf->hw_csum_rx_error = 0;
 546}
 547
 548/**
 549 * i40e_stat_update48 - read and update a 48 bit stat from the chip
 550 * @hw: ptr to the hardware info
 551 * @hireg: the high 32 bit reg to read
 552 * @loreg: the low 32 bit reg to read
 553 * @offset_loaded: has the initial offset been loaded yet
 554 * @offset: ptr to current offset value
 555 * @stat: ptr to the stat
 556 *
 557 * Since the device stats are not reset at PFReset, they likely will not
 558 * be zeroed when the driver starts.  We'll save the first values read
 559 * and use them as offsets to be subtracted from the raw values in order
 560 * to report stats that count from zero.  In the process, we also manage
 561 * the potential roll-over.
 562 **/
 563static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 564                               bool offset_loaded, u64 *offset, u64 *stat)
 565{
 566        u64 new_data;
 567
 568        if (hw->device_id == I40E_DEV_ID_QEMU) {
 569                new_data = rd32(hw, loreg);
 570                new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 571        } else {
 572                new_data = rd64(hw, loreg);
 573        }
 574        if (!offset_loaded)
 575                *offset = new_data;
 576        if (likely(new_data >= *offset))
 577                *stat = new_data - *offset;
 578        else
 579                *stat = (new_data + BIT_ULL(48)) - *offset;
 580        *stat &= 0xFFFFFFFFFFFFULL;
 581}
 582
 583/**
 584 * i40e_stat_update32 - read and update a 32 bit stat from the chip
 585 * @hw: ptr to the hardware info
 586 * @reg: the hw reg to read
 587 * @offset_loaded: has the initial offset been loaded yet
 588 * @offset: ptr to current offset value
 589 * @stat: ptr to the stat
 590 **/
 591static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 592                               bool offset_loaded, u64 *offset, u64 *stat)
 593{
 594        u32 new_data;
 595
 596        new_data = rd32(hw, reg);
 597        if (!offset_loaded)
 598                *offset = new_data;
 599        if (likely(new_data >= *offset))
 600                *stat = (u32)(new_data - *offset);
 601        else
 602                *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 603}
 604
 605/**
 606 * i40e_stat_update_and_clear32 - read and clear hw reg, update a 32 bit stat
 607 * @hw: ptr to the hardware info
 608 * @reg: the hw reg to read and clear
 609 * @stat: ptr to the stat
 610 **/
 611static void i40e_stat_update_and_clear32(struct i40e_hw *hw, u32 reg, u64 *stat)
 612{
 613        u32 new_data = rd32(hw, reg);
 614
 615        wr32(hw, reg, 1); /* must write a nonzero value to clear register */
 616        *stat += new_data;
 617}
 618
 619/**
 620 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 621 * @vsi: the VSI to be updated
 622 **/
 623void i40e_update_eth_stats(struct i40e_vsi *vsi)
 624{
 625        int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 626        struct i40e_pf *pf = vsi->back;
 627        struct i40e_hw *hw = &pf->hw;
 628        struct i40e_eth_stats *oes;
 629        struct i40e_eth_stats *es;     /* device's eth stats */
 630
 631        es = &vsi->eth_stats;
 632        oes = &vsi->eth_stats_offsets;
 633
 634        /* Gather up the stats that the hw collects */
 635        i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 636                           vsi->stat_offsets_loaded,
 637                           &oes->tx_errors, &es->tx_errors);
 638        i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 639                           vsi->stat_offsets_loaded,
 640                           &oes->rx_discards, &es->rx_discards);
 641        i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 642                           vsi->stat_offsets_loaded,
 643                           &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 644
 645        i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 646                           I40E_GLV_GORCL(stat_idx),
 647                           vsi->stat_offsets_loaded,
 648                           &oes->rx_bytes, &es->rx_bytes);
 649        i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 650                           I40E_GLV_UPRCL(stat_idx),
 651                           vsi->stat_offsets_loaded,
 652                           &oes->rx_unicast, &es->rx_unicast);
 653        i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 654                           I40E_GLV_MPRCL(stat_idx),
 655                           vsi->stat_offsets_loaded,
 656                           &oes->rx_multicast, &es->rx_multicast);
 657        i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 658                           I40E_GLV_BPRCL(stat_idx),
 659                           vsi->stat_offsets_loaded,
 660                           &oes->rx_broadcast, &es->rx_broadcast);
 661
 662        i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 663                           I40E_GLV_GOTCL(stat_idx),
 664                           vsi->stat_offsets_loaded,
 665                           &oes->tx_bytes, &es->tx_bytes);
 666        i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 667                           I40E_GLV_UPTCL(stat_idx),
 668                           vsi->stat_offsets_loaded,
 669                           &oes->tx_unicast, &es->tx_unicast);
 670        i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 671                           I40E_GLV_MPTCL(stat_idx),
 672                           vsi->stat_offsets_loaded,
 673                           &oes->tx_multicast, &es->tx_multicast);
 674        i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 675                           I40E_GLV_BPTCL(stat_idx),
 676                           vsi->stat_offsets_loaded,
 677                           &oes->tx_broadcast, &es->tx_broadcast);
 678        vsi->stat_offsets_loaded = true;
 679}
 680
 681/**
 682 * i40e_update_veb_stats - Update Switch component statistics
 683 * @veb: the VEB being updated
 684 **/
 685void i40e_update_veb_stats(struct i40e_veb *veb)
 686{
 687        struct i40e_pf *pf = veb->pf;
 688        struct i40e_hw *hw = &pf->hw;
 689        struct i40e_eth_stats *oes;
 690        struct i40e_eth_stats *es;     /* device's eth stats */
 691        struct i40e_veb_tc_stats *veb_oes;
 692        struct i40e_veb_tc_stats *veb_es;
 693        int i, idx = 0;
 694
 695        idx = veb->stats_idx;
 696        es = &veb->stats;
 697        oes = &veb->stats_offsets;
 698        veb_es = &veb->tc_stats;
 699        veb_oes = &veb->tc_stats_offsets;
 700
 701        /* Gather up the stats that the hw collects */
 702        i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 703                           veb->stat_offsets_loaded,
 704                           &oes->tx_discards, &es->tx_discards);
 705        if (hw->revision_id > 0)
 706                i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 707                                   veb->stat_offsets_loaded,
 708                                   &oes->rx_unknown_protocol,
 709                                   &es->rx_unknown_protocol);
 710        i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 711                           veb->stat_offsets_loaded,
 712                           &oes->rx_bytes, &es->rx_bytes);
 713        i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 714                           veb->stat_offsets_loaded,
 715                           &oes->rx_unicast, &es->rx_unicast);
 716        i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 717                           veb->stat_offsets_loaded,
 718                           &oes->rx_multicast, &es->rx_multicast);
 719        i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 720                           veb->stat_offsets_loaded,
 721                           &oes->rx_broadcast, &es->rx_broadcast);
 722
 723        i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 724                           veb->stat_offsets_loaded,
 725                           &oes->tx_bytes, &es->tx_bytes);
 726        i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 727                           veb->stat_offsets_loaded,
 728                           &oes->tx_unicast, &es->tx_unicast);
 729        i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 730                           veb->stat_offsets_loaded,
 731                           &oes->tx_multicast, &es->tx_multicast);
 732        i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 733                           veb->stat_offsets_loaded,
 734                           &oes->tx_broadcast, &es->tx_broadcast);
 735        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 736                i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 737                                   I40E_GLVEBTC_RPCL(i, idx),
 738                                   veb->stat_offsets_loaded,
 739                                   &veb_oes->tc_rx_packets[i],
 740                                   &veb_es->tc_rx_packets[i]);
 741                i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 742                                   I40E_GLVEBTC_RBCL(i, idx),
 743                                   veb->stat_offsets_loaded,
 744                                   &veb_oes->tc_rx_bytes[i],
 745                                   &veb_es->tc_rx_bytes[i]);
 746                i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 747                                   I40E_GLVEBTC_TPCL(i, idx),
 748                                   veb->stat_offsets_loaded,
 749                                   &veb_oes->tc_tx_packets[i],
 750                                   &veb_es->tc_tx_packets[i]);
 751                i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 752                                   I40E_GLVEBTC_TBCL(i, idx),
 753                                   veb->stat_offsets_loaded,
 754                                   &veb_oes->tc_tx_bytes[i],
 755                                   &veb_es->tc_tx_bytes[i]);
 756        }
 757        veb->stat_offsets_loaded = true;
 758}
 759
 760/**
 761 * i40e_update_vsi_stats - Update the vsi statistics counters.
 762 * @vsi: the VSI to be updated
 763 *
 764 * There are a few instances where we store the same stat in a
 765 * couple of different structs.  This is partly because we have
 766 * the netdev stats that need to be filled out, which is slightly
 767 * different from the "eth_stats" defined by the chip and used in
 768 * VF communications.  We sort it out here.
 769 **/
 770static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 771{
 772        struct i40e_pf *pf = vsi->back;
 773        struct rtnl_link_stats64 *ons;
 774        struct rtnl_link_stats64 *ns;   /* netdev stats */
 775        struct i40e_eth_stats *oes;
 776        struct i40e_eth_stats *es;     /* device's eth stats */
 777        u32 tx_restart, tx_busy;
 778        struct i40e_ring *p;
 779        u32 rx_page, rx_buf;
 780        u64 bytes, packets;
 781        unsigned int start;
 782        u64 tx_linearize;
 783        u64 tx_force_wb;
 784        u64 rx_p, rx_b;
 785        u64 tx_p, tx_b;
 786        u16 q;
 787
 788        if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
 789            test_bit(__I40E_CONFIG_BUSY, pf->state))
 790                return;
 791
 792        ns = i40e_get_vsi_stats_struct(vsi);
 793        ons = &vsi->net_stats_offsets;
 794        es = &vsi->eth_stats;
 795        oes = &vsi->eth_stats_offsets;
 796
 797        /* Gather up the netdev and vsi stats that the driver collects
 798         * on the fly during packet processing
 799         */
 800        rx_b = rx_p = 0;
 801        tx_b = tx_p = 0;
 802        tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 803        rx_page = 0;
 804        rx_buf = 0;
 805        rcu_read_lock();
 806        for (q = 0; q < vsi->num_queue_pairs; q++) {
 807                /* locate Tx ring */
 808                p = READ_ONCE(vsi->tx_rings[q]);
 809
 810                do {
 811                        start = u64_stats_fetch_begin_irq(&p->syncp);
 812                        packets = p->stats.packets;
 813                        bytes = p->stats.bytes;
 814                } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 815                tx_b += bytes;
 816                tx_p += packets;
 817                tx_restart += p->tx_stats.restart_queue;
 818                tx_busy += p->tx_stats.tx_busy;
 819                tx_linearize += p->tx_stats.tx_linearize;
 820                tx_force_wb += p->tx_stats.tx_force_wb;
 821
 822                /* Rx queue is part of the same block as Tx queue */
 823                p = &p[1];
 824                do {
 825                        start = u64_stats_fetch_begin_irq(&p->syncp);
 826                        packets = p->stats.packets;
 827                        bytes = p->stats.bytes;
 828                } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 829                rx_b += bytes;
 830                rx_p += packets;
 831                rx_buf += p->rx_stats.alloc_buff_failed;
 832                rx_page += p->rx_stats.alloc_page_failed;
 833        }
 834        rcu_read_unlock();
 835        vsi->tx_restart = tx_restart;
 836        vsi->tx_busy = tx_busy;
 837        vsi->tx_linearize = tx_linearize;
 838        vsi->tx_force_wb = tx_force_wb;
 839        vsi->rx_page_failed = rx_page;
 840        vsi->rx_buf_failed = rx_buf;
 841
 842        ns->rx_packets = rx_p;
 843        ns->rx_bytes = rx_b;
 844        ns->tx_packets = tx_p;
 845        ns->tx_bytes = tx_b;
 846
 847        /* update netdev stats from eth stats */
 848        i40e_update_eth_stats(vsi);
 849        ons->tx_errors = oes->tx_errors;
 850        ns->tx_errors = es->tx_errors;
 851        ons->multicast = oes->rx_multicast;
 852        ns->multicast = es->rx_multicast;
 853        ons->rx_dropped = oes->rx_discards;
 854        ns->rx_dropped = es->rx_discards;
 855        ons->tx_dropped = oes->tx_discards;
 856        ns->tx_dropped = es->tx_discards;
 857
 858        /* pull in a couple PF stats if this is the main vsi */
 859        if (vsi == pf->vsi[pf->lan_vsi]) {
 860                ns->rx_crc_errors = pf->stats.crc_errors;
 861                ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 862                ns->rx_length_errors = pf->stats.rx_length_errors;
 863        }
 864}
 865
 866/**
 867 * i40e_update_pf_stats - Update the PF statistics counters.
 868 * @pf: the PF to be updated
 869 **/
 870static void i40e_update_pf_stats(struct i40e_pf *pf)
 871{
 872        struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 873        struct i40e_hw_port_stats *nsd = &pf->stats;
 874        struct i40e_hw *hw = &pf->hw;
 875        u32 val;
 876        int i;
 877
 878        i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 879                           I40E_GLPRT_GORCL(hw->port),
 880                           pf->stat_offsets_loaded,
 881                           &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 882        i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 883                           I40E_GLPRT_GOTCL(hw->port),
 884                           pf->stat_offsets_loaded,
 885                           &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 886        i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 887                           pf->stat_offsets_loaded,
 888                           &osd->eth.rx_discards,
 889                           &nsd->eth.rx_discards);
 890        i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 891                           I40E_GLPRT_UPRCL(hw->port),
 892                           pf->stat_offsets_loaded,
 893                           &osd->eth.rx_unicast,
 894                           &nsd->eth.rx_unicast);
 895        i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 896                           I40E_GLPRT_MPRCL(hw->port),
 897                           pf->stat_offsets_loaded,
 898                           &osd->eth.rx_multicast,
 899                           &nsd->eth.rx_multicast);
 900        i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 901                           I40E_GLPRT_BPRCL(hw->port),
 902                           pf->stat_offsets_loaded,
 903                           &osd->eth.rx_broadcast,
 904                           &nsd->eth.rx_broadcast);
 905        i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 906                           I40E_GLPRT_UPTCL(hw->port),
 907                           pf->stat_offsets_loaded,
 908                           &osd->eth.tx_unicast,
 909                           &nsd->eth.tx_unicast);
 910        i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 911                           I40E_GLPRT_MPTCL(hw->port),
 912                           pf->stat_offsets_loaded,
 913                           &osd->eth.tx_multicast,
 914                           &nsd->eth.tx_multicast);
 915        i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 916                           I40E_GLPRT_BPTCL(hw->port),
 917                           pf->stat_offsets_loaded,
 918                           &osd->eth.tx_broadcast,
 919                           &nsd->eth.tx_broadcast);
 920
 921        i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 922                           pf->stat_offsets_loaded,
 923                           &osd->tx_dropped_link_down,
 924                           &nsd->tx_dropped_link_down);
 925
 926        i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 927                           pf->stat_offsets_loaded,
 928                           &osd->crc_errors, &nsd->crc_errors);
 929
 930        i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 931                           pf->stat_offsets_loaded,
 932                           &osd->illegal_bytes, &nsd->illegal_bytes);
 933
 934        i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 935                           pf->stat_offsets_loaded,
 936                           &osd->mac_local_faults,
 937                           &nsd->mac_local_faults);
 938        i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 939                           pf->stat_offsets_loaded,
 940                           &osd->mac_remote_faults,
 941                           &nsd->mac_remote_faults);
 942
 943        i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 944                           pf->stat_offsets_loaded,
 945                           &osd->rx_length_errors,
 946                           &nsd->rx_length_errors);
 947
 948        i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 949                           pf->stat_offsets_loaded,
 950                           &osd->link_xon_rx, &nsd->link_xon_rx);
 951        i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 952                           pf->stat_offsets_loaded,
 953                           &osd->link_xon_tx, &nsd->link_xon_tx);
 954        i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 955                           pf->stat_offsets_loaded,
 956                           &osd->link_xoff_rx, &nsd->link_xoff_rx);
 957        i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 958                           pf->stat_offsets_loaded,
 959                           &osd->link_xoff_tx, &nsd->link_xoff_tx);
 960
 961        for (i = 0; i < 8; i++) {
 962                i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 963                                   pf->stat_offsets_loaded,
 964                                   &osd->priority_xoff_rx[i],
 965                                   &nsd->priority_xoff_rx[i]);
 966                i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 967                                   pf->stat_offsets_loaded,
 968                                   &osd->priority_xon_rx[i],
 969                                   &nsd->priority_xon_rx[i]);
 970                i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
 971                                   pf->stat_offsets_loaded,
 972                                   &osd->priority_xon_tx[i],
 973                                   &nsd->priority_xon_tx[i]);
 974                i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
 975                                   pf->stat_offsets_loaded,
 976                                   &osd->priority_xoff_tx[i],
 977                                   &nsd->priority_xoff_tx[i]);
 978                i40e_stat_update32(hw,
 979                                   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
 980                                   pf->stat_offsets_loaded,
 981                                   &osd->priority_xon_2_xoff[i],
 982                                   &nsd->priority_xon_2_xoff[i]);
 983        }
 984
 985        i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
 986                           I40E_GLPRT_PRC64L(hw->port),
 987                           pf->stat_offsets_loaded,
 988                           &osd->rx_size_64, &nsd->rx_size_64);
 989        i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
 990                           I40E_GLPRT_PRC127L(hw->port),
 991                           pf->stat_offsets_loaded,
 992                           &osd->rx_size_127, &nsd->rx_size_127);
 993        i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
 994                           I40E_GLPRT_PRC255L(hw->port),
 995                           pf->stat_offsets_loaded,
 996                           &osd->rx_size_255, &nsd->rx_size_255);
 997        i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
 998                           I40E_GLPRT_PRC511L(hw->port),
 999                           pf->stat_offsets_loaded,
1000                           &osd->rx_size_511, &nsd->rx_size_511);
1001        i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1002                           I40E_GLPRT_PRC1023L(hw->port),
1003                           pf->stat_offsets_loaded,
1004                           &osd->rx_size_1023, &nsd->rx_size_1023);
1005        i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1006                           I40E_GLPRT_PRC1522L(hw->port),
1007                           pf->stat_offsets_loaded,
1008                           &osd->rx_size_1522, &nsd->rx_size_1522);
1009        i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1010                           I40E_GLPRT_PRC9522L(hw->port),
1011                           pf->stat_offsets_loaded,
1012                           &osd->rx_size_big, &nsd->rx_size_big);
1013
1014        i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1015                           I40E_GLPRT_PTC64L(hw->port),
1016                           pf->stat_offsets_loaded,
1017                           &osd->tx_size_64, &nsd->tx_size_64);
1018        i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1019                           I40E_GLPRT_PTC127L(hw->port),
1020                           pf->stat_offsets_loaded,
1021                           &osd->tx_size_127, &nsd->tx_size_127);
1022        i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1023                           I40E_GLPRT_PTC255L(hw->port),
1024                           pf->stat_offsets_loaded,
1025                           &osd->tx_size_255, &nsd->tx_size_255);
1026        i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1027                           I40E_GLPRT_PTC511L(hw->port),
1028                           pf->stat_offsets_loaded,
1029                           &osd->tx_size_511, &nsd->tx_size_511);
1030        i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1031                           I40E_GLPRT_PTC1023L(hw->port),
1032                           pf->stat_offsets_loaded,
1033                           &osd->tx_size_1023, &nsd->tx_size_1023);
1034        i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1035                           I40E_GLPRT_PTC1522L(hw->port),
1036                           pf->stat_offsets_loaded,
1037                           &osd->tx_size_1522, &nsd->tx_size_1522);
1038        i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1039                           I40E_GLPRT_PTC9522L(hw->port),
1040                           pf->stat_offsets_loaded,
1041                           &osd->tx_size_big, &nsd->tx_size_big);
1042
1043        i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1044                           pf->stat_offsets_loaded,
1045                           &osd->rx_undersize, &nsd->rx_undersize);
1046        i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1047                           pf->stat_offsets_loaded,
1048                           &osd->rx_fragments, &nsd->rx_fragments);
1049        i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1050                           pf->stat_offsets_loaded,
1051                           &osd->rx_oversize, &nsd->rx_oversize);
1052        i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1053                           pf->stat_offsets_loaded,
1054                           &osd->rx_jabber, &nsd->rx_jabber);
1055
1056        /* FDIR stats */
1057        i40e_stat_update_and_clear32(hw,
1058                        I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(hw->pf_id)),
1059                        &nsd->fd_atr_match);
1060        i40e_stat_update_and_clear32(hw,
1061                        I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(hw->pf_id)),
1062                        &nsd->fd_sb_match);
1063        i40e_stat_update_and_clear32(hw,
1064                        I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(hw->pf_id)),
1065                        &nsd->fd_atr_tunnel_match);
1066
1067        val = rd32(hw, I40E_PRTPM_EEE_STAT);
1068        nsd->tx_lpi_status =
1069                       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1070                        I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1071        nsd->rx_lpi_status =
1072                       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1073                        I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1074        i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1075                           pf->stat_offsets_loaded,
1076                           &osd->tx_lpi_count, &nsd->tx_lpi_count);
1077        i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1078                           pf->stat_offsets_loaded,
1079                           &osd->rx_lpi_count, &nsd->rx_lpi_count);
1080
1081        if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1082            !test_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state))
1083                nsd->fd_sb_status = true;
1084        else
1085                nsd->fd_sb_status = false;
1086
1087        if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1088            !test_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state))
1089                nsd->fd_atr_status = true;
1090        else
1091                nsd->fd_atr_status = false;
1092
1093        pf->stat_offsets_loaded = true;
1094}
1095
1096/**
1097 * i40e_update_stats - Update the various statistics counters.
1098 * @vsi: the VSI to be updated
1099 *
1100 * Update the various stats for this VSI and its related entities.
1101 **/
1102void i40e_update_stats(struct i40e_vsi *vsi)
1103{
1104        struct i40e_pf *pf = vsi->back;
1105
1106        if (vsi == pf->vsi[pf->lan_vsi])
1107                i40e_update_pf_stats(pf);
1108
1109        i40e_update_vsi_stats(vsi);
1110}
1111
1112/**
1113 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1114 * @vsi: the VSI to be searched
1115 * @macaddr: the MAC address
1116 * @vlan: the vlan
1117 *
1118 * Returns ptr to the filter object or NULL
1119 **/
1120static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1121                                                const u8 *macaddr, s16 vlan)
1122{
1123        struct i40e_mac_filter *f;
1124        u64 key;
1125
1126        if (!vsi || !macaddr)
1127                return NULL;
1128
1129        key = i40e_addr_to_hkey(macaddr);
1130        hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1131                if ((ether_addr_equal(macaddr, f->macaddr)) &&
1132                    (vlan == f->vlan))
1133                        return f;
1134        }
1135        return NULL;
1136}
1137
1138/**
1139 * i40e_find_mac - Find a mac addr in the macvlan filters list
1140 * @vsi: the VSI to be searched
1141 * @macaddr: the MAC address we are searching for
1142 *
1143 * Returns the first filter with the provided MAC address or NULL if
1144 * MAC address was not found
1145 **/
1146struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1147{
1148        struct i40e_mac_filter *f;
1149        u64 key;
1150
1151        if (!vsi || !macaddr)
1152                return NULL;
1153
1154        key = i40e_addr_to_hkey(macaddr);
1155        hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1156                if ((ether_addr_equal(macaddr, f->macaddr)))
1157                        return f;
1158        }
1159        return NULL;
1160}
1161
1162/**
1163 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1164 * @vsi: the VSI to be searched
1165 *
1166 * Returns true if VSI is in vlan mode or false otherwise
1167 **/
1168bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1169{
1170        /* If we have a PVID, always operate in VLAN mode */
1171        if (vsi->info.pvid)
1172                return true;
1173
1174        /* We need to operate in VLAN mode whenever we have any filters with
1175         * a VLAN other than I40E_VLAN_ALL. We could check the table each
1176         * time, incurring search cost repeatedly. However, we can notice two
1177         * things:
1178         *
1179         * 1) the only place where we can gain a VLAN filter is in
1180         *    i40e_add_filter.
1181         *
1182         * 2) the only place where filters are actually removed is in
1183         *    i40e_sync_filters_subtask.
1184         *
1185         * Thus, we can simply use a boolean value, has_vlan_filters which we
1186         * will set to true when we add a VLAN filter in i40e_add_filter. Then
1187         * we have to perform the full search after deleting filters in
1188         * i40e_sync_filters_subtask, but we already have to search
1189         * filters here and can perform the check at the same time. This
1190         * results in avoiding embedding a loop for VLAN mode inside another
1191         * loop over all the filters, and should maintain correctness as noted
1192         * above.
1193         */
1194        return vsi->has_vlan_filter;
1195}
1196
1197/**
1198 * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1199 * @vsi: the VSI to configure
1200 * @tmp_add_list: list of filters ready to be added
1201 * @tmp_del_list: list of filters ready to be deleted
1202 * @vlan_filters: the number of active VLAN filters
1203 *
1204 * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1205 * behave as expected. If we have any active VLAN filters remaining or about
1206 * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1207 * so that they only match against untagged traffic. If we no longer have any
1208 * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1209 * so that they match against both tagged and untagged traffic. In this way,
1210 * we ensure that we correctly receive the desired traffic. This ensures that
1211 * when we have an active VLAN we will receive only untagged traffic and
1212 * traffic matching active VLANs. If we have no active VLANs then we will
1213 * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1214 *
1215 * Finally, in a similar fashion, this function also corrects filters when
1216 * there is an active PVID assigned to this VSI.
1217 *
1218 * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1219 *
1220 * This function is only expected to be called from within
1221 * i40e_sync_vsi_filters.
1222 *
1223 * NOTE: This function expects to be called while under the
1224 * mac_filter_hash_lock
1225 */
1226static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1227                                         struct hlist_head *tmp_add_list,
1228                                         struct hlist_head *tmp_del_list,
1229                                         int vlan_filters)
1230{
1231        s16 pvid = le16_to_cpu(vsi->info.pvid);
1232        struct i40e_mac_filter *f, *add_head;
1233        struct i40e_new_mac_filter *new;
1234        struct hlist_node *h;
1235        int bkt, new_vlan;
1236
1237        /* To determine if a particular filter needs to be replaced we
1238         * have the three following conditions:
1239         *
1240         * a) if we have a PVID assigned, then all filters which are
1241         *    not marked as VLAN=PVID must be replaced with filters that
1242         *    are.
1243         * b) otherwise, if we have any active VLANS, all filters
1244         *    which are marked as VLAN=-1 must be replaced with
1245         *    filters marked as VLAN=0
1246         * c) finally, if we do not have any active VLANS, all filters
1247         *    which are marked as VLAN=0 must be replaced with filters
1248         *    marked as VLAN=-1
1249         */
1250
1251        /* Update the filters about to be added in place */
1252        hlist_for_each_entry(new, tmp_add_list, hlist) {
1253                if (pvid && new->f->vlan != pvid)
1254                        new->f->vlan = pvid;
1255                else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1256                        new->f->vlan = 0;
1257                else if (!vlan_filters && new->f->vlan == 0)
1258                        new->f->vlan = I40E_VLAN_ANY;
1259        }
1260
1261        /* Update the remaining active filters */
1262        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1263                /* Combine the checks for whether a filter needs to be changed
1264                 * and then determine the new VLAN inside the if block, in
1265                 * order to avoid duplicating code for adding the new filter
1266                 * then deleting the old filter.
1267                 */
1268                if ((pvid && f->vlan != pvid) ||
1269                    (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1270                    (!vlan_filters && f->vlan == 0)) {
1271                        /* Determine the new vlan we will be adding */
1272                        if (pvid)
1273                                new_vlan = pvid;
1274                        else if (vlan_filters)
1275                                new_vlan = 0;
1276                        else
1277                                new_vlan = I40E_VLAN_ANY;
1278
1279                        /* Create the new filter */
1280                        add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1281                        if (!add_head)
1282                                return -ENOMEM;
1283
1284                        /* Create a temporary i40e_new_mac_filter */
1285                        new = kzalloc(sizeof(*new), GFP_ATOMIC);
1286                        if (!new)
1287                                return -ENOMEM;
1288
1289                        new->f = add_head;
1290                        new->state = add_head->state;
1291
1292                        /* Add the new filter to the tmp list */
1293                        hlist_add_head(&new->hlist, tmp_add_list);
1294
1295                        /* Put the original filter into the delete list */
1296                        f->state = I40E_FILTER_REMOVE;
1297                        hash_del(&f->hlist);
1298                        hlist_add_head(&f->hlist, tmp_del_list);
1299                }
1300        }
1301
1302        vsi->has_vlan_filter = !!vlan_filters;
1303
1304        return 0;
1305}
1306
1307/**
1308 * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1309 * @vsi: the PF Main VSI - inappropriate for any other VSI
1310 * @macaddr: the MAC address
1311 *
1312 * Remove whatever filter the firmware set up so the driver can manage
1313 * its own filtering intelligently.
1314 **/
1315static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1316{
1317        struct i40e_aqc_remove_macvlan_element_data element;
1318        struct i40e_pf *pf = vsi->back;
1319
1320        /* Only appropriate for the PF main VSI */
1321        if (vsi->type != I40E_VSI_MAIN)
1322                return;
1323
1324        memset(&element, 0, sizeof(element));
1325        ether_addr_copy(element.mac_addr, macaddr);
1326        element.vlan_tag = 0;
1327        /* Ignore error returns, some firmware does it this way... */
1328        element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1329        i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1330
1331        memset(&element, 0, sizeof(element));
1332        ether_addr_copy(element.mac_addr, macaddr);
1333        element.vlan_tag = 0;
1334        /* ...and some firmware does it this way. */
1335        element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1336                        I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1337        i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1338}
1339
1340/**
1341 * i40e_add_filter - Add a mac/vlan filter to the VSI
1342 * @vsi: the VSI to be searched
1343 * @macaddr: the MAC address
1344 * @vlan: the vlan
1345 *
1346 * Returns ptr to the filter object or NULL when no memory available.
1347 *
1348 * NOTE: This function is expected to be called with mac_filter_hash_lock
1349 * being held.
1350 **/
1351struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1352                                        const u8 *macaddr, s16 vlan)
1353{
1354        struct i40e_mac_filter *f;
1355        u64 key;
1356
1357        if (!vsi || !macaddr)
1358                return NULL;
1359
1360        f = i40e_find_filter(vsi, macaddr, vlan);
1361        if (!f) {
1362                f = kzalloc(sizeof(*f), GFP_ATOMIC);
1363                if (!f)
1364                        return NULL;
1365
1366                /* Update the boolean indicating if we need to function in
1367                 * VLAN mode.
1368                 */
1369                if (vlan >= 0)
1370                        vsi->has_vlan_filter = true;
1371
1372                ether_addr_copy(f->macaddr, macaddr);
1373                f->vlan = vlan;
1374                f->state = I40E_FILTER_NEW;
1375                INIT_HLIST_NODE(&f->hlist);
1376
1377                key = i40e_addr_to_hkey(macaddr);
1378                hash_add(vsi->mac_filter_hash, &f->hlist, key);
1379
1380                vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1381                set_bit(__I40E_MACVLAN_SYNC_PENDING, vsi->back->state);
1382        }
1383
1384        /* If we're asked to add a filter that has been marked for removal, it
1385         * is safe to simply restore it to active state. __i40e_del_filter
1386         * will have simply deleted any filters which were previously marked
1387         * NEW or FAILED, so if it is currently marked REMOVE it must have
1388         * previously been ACTIVE. Since we haven't yet run the sync filters
1389         * task, just restore this filter to the ACTIVE state so that the
1390         * sync task leaves it in place
1391         */
1392        if (f->state == I40E_FILTER_REMOVE)
1393                f->state = I40E_FILTER_ACTIVE;
1394
1395        return f;
1396}
1397
1398/**
1399 * __i40e_del_filter - Remove a specific filter from the VSI
1400 * @vsi: VSI to remove from
1401 * @f: the filter to remove from the list
1402 *
1403 * This function should be called instead of i40e_del_filter only if you know
1404 * the exact filter you will remove already, such as via i40e_find_filter or
1405 * i40e_find_mac.
1406 *
1407 * NOTE: This function is expected to be called with mac_filter_hash_lock
1408 * being held.
1409 * ANOTHER NOTE: This function MUST be called from within the context of
1410 * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1411 * instead of list_for_each_entry().
1412 **/
1413void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1414{
1415        if (!f)
1416                return;
1417
1418        /* If the filter was never added to firmware then we can just delete it
1419         * directly and we don't want to set the status to remove or else an
1420         * admin queue command will unnecessarily fire.
1421         */
1422        if ((f->state == I40E_FILTER_FAILED) ||
1423            (f->state == I40E_FILTER_NEW)) {
1424                hash_del(&f->hlist);
1425                kfree(f);
1426        } else {
1427                f->state = I40E_FILTER_REMOVE;
1428        }
1429
1430        vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1431        set_bit(__I40E_MACVLAN_SYNC_PENDING, vsi->back->state);
1432}
1433
1434/**
1435 * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1436 * @vsi: the VSI to be searched
1437 * @macaddr: the MAC address
1438 * @vlan: the VLAN
1439 *
1440 * NOTE: This function is expected to be called with mac_filter_hash_lock
1441 * being held.
1442 * ANOTHER NOTE: This function MUST be called from within the context of
1443 * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1444 * instead of list_for_each_entry().
1445 **/
1446void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1447{
1448        struct i40e_mac_filter *f;
1449
1450        if (!vsi || !macaddr)
1451                return;
1452
1453        f = i40e_find_filter(vsi, macaddr, vlan);
1454        __i40e_del_filter(vsi, f);
1455}
1456
1457/**
1458 * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1459 * @vsi: the VSI to be searched
1460 * @macaddr: the mac address to be filtered
1461 *
1462 * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1463 * go through all the macvlan filters and add a macvlan filter for each
1464 * unique vlan that already exists. If a PVID has been assigned, instead only
1465 * add the macaddr to that VLAN.
1466 *
1467 * Returns last filter added on success, else NULL
1468 **/
1469struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1470                                            const u8 *macaddr)
1471{
1472        struct i40e_mac_filter *f, *add = NULL;
1473        struct hlist_node *h;
1474        int bkt;
1475
1476        if (vsi->info.pvid)
1477                return i40e_add_filter(vsi, macaddr,
1478                                       le16_to_cpu(vsi->info.pvid));
1479
1480        if (!i40e_is_vsi_in_vlan(vsi))
1481                return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1482
1483        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1484                if (f->state == I40E_FILTER_REMOVE)
1485                        continue;
1486                add = i40e_add_filter(vsi, macaddr, f->vlan);
1487                if (!add)
1488                        return NULL;
1489        }
1490
1491        return add;
1492}
1493
1494/**
1495 * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1496 * @vsi: the VSI to be searched
1497 * @macaddr: the mac address to be removed
1498 *
1499 * Removes a given MAC address from a VSI regardless of what VLAN it has been
1500 * associated with.
1501 *
1502 * Returns 0 for success, or error
1503 **/
1504int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1505{
1506        struct i40e_mac_filter *f;
1507        struct hlist_node *h;
1508        bool found = false;
1509        int bkt;
1510
1511        lockdep_assert_held(&vsi->mac_filter_hash_lock);
1512        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1513                if (ether_addr_equal(macaddr, f->macaddr)) {
1514                        __i40e_del_filter(vsi, f);
1515                        found = true;
1516                }
1517        }
1518
1519        if (found)
1520                return 0;
1521        else
1522                return -ENOENT;
1523}
1524
1525/**
1526 * i40e_set_mac - NDO callback to set mac address
1527 * @netdev: network interface device structure
1528 * @p: pointer to an address structure
1529 *
1530 * Returns 0 on success, negative on failure
1531 **/
1532static int i40e_set_mac(struct net_device *netdev, void *p)
1533{
1534        struct i40e_netdev_priv *np = netdev_priv(netdev);
1535        struct i40e_vsi *vsi = np->vsi;
1536        struct i40e_pf *pf = vsi->back;
1537        struct i40e_hw *hw = &pf->hw;
1538        struct sockaddr *addr = p;
1539
1540        if (!is_valid_ether_addr(addr->sa_data))
1541                return -EADDRNOTAVAIL;
1542
1543        if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1544                netdev_info(netdev, "already using mac address %pM\n",
1545                            addr->sa_data);
1546                return 0;
1547        }
1548
1549        if (test_bit(__I40E_DOWN, pf->state) ||
1550            test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
1551                return -EADDRNOTAVAIL;
1552
1553        if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1554                netdev_info(netdev, "returning to hw mac address %pM\n",
1555                            hw->mac.addr);
1556        else
1557                netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1558
1559        /* Copy the address first, so that we avoid a possible race with
1560         * .set_rx_mode().
1561         * - Remove old address from MAC filter
1562         * - Copy new address
1563         * - Add new address to MAC filter
1564         */
1565        spin_lock_bh(&vsi->mac_filter_hash_lock);
1566        i40e_del_mac_filter(vsi, netdev->dev_addr);
1567        ether_addr_copy(netdev->dev_addr, addr->sa_data);
1568        i40e_add_mac_filter(vsi, netdev->dev_addr);
1569        spin_unlock_bh(&vsi->mac_filter_hash_lock);
1570
1571        if (vsi->type == I40E_VSI_MAIN) {
1572                i40e_status ret;
1573
1574                ret = i40e_aq_mac_address_write(hw, I40E_AQC_WRITE_TYPE_LAA_WOL,
1575                                                addr->sa_data, NULL);
1576                if (ret)
1577                        netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1578                                    i40e_stat_str(hw, ret),
1579                                    i40e_aq_str(hw, hw->aq.asq_last_status));
1580        }
1581
1582        /* schedule our worker thread which will take care of
1583         * applying the new filter changes
1584         */
1585        i40e_service_event_schedule(pf);
1586        return 0;
1587}
1588
1589/**
1590 * i40e_config_rss_aq - Prepare for RSS using AQ commands
1591 * @vsi: vsi structure
1592 * @seed: RSS hash seed
1593 **/
1594static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
1595                              u8 *lut, u16 lut_size)
1596{
1597        struct i40e_pf *pf = vsi->back;
1598        struct i40e_hw *hw = &pf->hw;
1599        int ret = 0;
1600
1601        if (seed) {
1602                struct i40e_aqc_get_set_rss_key_data *seed_dw =
1603                        (struct i40e_aqc_get_set_rss_key_data *)seed;
1604                ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
1605                if (ret) {
1606                        dev_info(&pf->pdev->dev,
1607                                 "Cannot set RSS key, err %s aq_err %s\n",
1608                                 i40e_stat_str(hw, ret),
1609                                 i40e_aq_str(hw, hw->aq.asq_last_status));
1610                        return ret;
1611                }
1612        }
1613        if (lut) {
1614                bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
1615
1616                ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
1617                if (ret) {
1618                        dev_info(&pf->pdev->dev,
1619                                 "Cannot set RSS lut, err %s aq_err %s\n",
1620                                 i40e_stat_str(hw, ret),
1621                                 i40e_aq_str(hw, hw->aq.asq_last_status));
1622                        return ret;
1623                }
1624        }
1625        return ret;
1626}
1627
1628/**
1629 * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
1630 * @vsi: VSI structure
1631 **/
1632static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
1633{
1634        struct i40e_pf *pf = vsi->back;
1635        u8 seed[I40E_HKEY_ARRAY_SIZE];
1636        u8 *lut;
1637        int ret;
1638
1639        if (!(pf->hw_features & I40E_HW_RSS_AQ_CAPABLE))
1640                return 0;
1641        if (!vsi->rss_size)
1642                vsi->rss_size = min_t(int, pf->alloc_rss_size,
1643                                      vsi->num_queue_pairs);
1644        if (!vsi->rss_size)
1645                return -EINVAL;
1646        lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
1647        if (!lut)
1648                return -ENOMEM;
1649
1650        /* Use the user configured hash keys and lookup table if there is one,
1651         * otherwise use default
1652         */
1653        if (vsi->rss_lut_user)
1654                memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
1655        else
1656                i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
1657        if (vsi->rss_hkey_user)
1658                memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
1659        else
1660                netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
1661        ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
1662        kfree(lut);
1663        return ret;
1664}
1665
1666/**
1667 * i40e_vsi_setup_queue_map_mqprio - Prepares mqprio based tc_config
1668 * @vsi: the VSI being configured,
1669 * @ctxt: VSI context structure
1670 * @enabled_tc: number of traffic classes to enable
1671 *
1672 * Prepares VSI tc_config to have queue configurations based on MQPRIO options.
1673 **/
1674static int i40e_vsi_setup_queue_map_mqprio(struct i40e_vsi *vsi,
1675                                           struct i40e_vsi_context *ctxt,
1676                                           u8 enabled_tc)
1677{
1678        u16 qcount = 0, max_qcount, qmap, sections = 0;
1679        int i, override_q, pow, num_qps, ret;
1680        u8 netdev_tc = 0, offset = 0;
1681
1682        if (vsi->type != I40E_VSI_MAIN)
1683                return -EINVAL;
1684        sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1685        sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1686        vsi->tc_config.numtc = vsi->mqprio_qopt.qopt.num_tc;
1687        vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1688        num_qps = vsi->mqprio_qopt.qopt.count[0];
1689
1690        /* find the next higher power-of-2 of num queue pairs */
1691        pow = ilog2(num_qps);
1692        if (!is_power_of_2(num_qps))
1693                pow++;
1694        qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1695                (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1696
1697        /* Setup queue offset/count for all TCs for given VSI */
1698        max_qcount = vsi->mqprio_qopt.qopt.count[0];
1699        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1700                /* See if the given TC is enabled for the given VSI */
1701                if (vsi->tc_config.enabled_tc & BIT(i)) {
1702                        offset = vsi->mqprio_qopt.qopt.offset[i];
1703                        qcount = vsi->mqprio_qopt.qopt.count[i];
1704                        if (qcount > max_qcount)
1705                                max_qcount = qcount;
1706                        vsi->tc_config.tc_info[i].qoffset = offset;
1707                        vsi->tc_config.tc_info[i].qcount = qcount;
1708                        vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1709                } else {
1710                        /* TC is not enabled so set the offset to
1711                         * default queue and allocate one queue
1712                         * for the given TC.
1713                         */
1714                        vsi->tc_config.tc_info[i].qoffset = 0;
1715                        vsi->tc_config.tc_info[i].qcount = 1;
1716                        vsi->tc_config.tc_info[i].netdev_tc = 0;
1717                }
1718        }
1719
1720        /* Set actual Tx/Rx queue pairs */
1721        vsi->num_queue_pairs = offset + qcount;
1722
1723        /* Setup queue TC[0].qmap for given VSI context */
1724        ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
1725        ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1726        ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1727        ctxt->info.valid_sections |= cpu_to_le16(sections);
1728
1729        /* Reconfigure RSS for main VSI with max queue count */
1730        vsi->rss_size = max_qcount;
1731        ret = i40e_vsi_config_rss(vsi);
1732        if (ret) {
1733                dev_info(&vsi->back->pdev->dev,
1734                         "Failed to reconfig rss for num_queues (%u)\n",
1735                         max_qcount);
1736                return ret;
1737        }
1738        vsi->reconfig_rss = true;
1739        dev_dbg(&vsi->back->pdev->dev,
1740                "Reconfigured rss with num_queues (%u)\n", max_qcount);
1741
1742        /* Find queue count available for channel VSIs and starting offset
1743         * for channel VSIs
1744         */
1745        override_q = vsi->mqprio_qopt.qopt.count[0];
1746        if (override_q && override_q < vsi->num_queue_pairs) {
1747                vsi->cnt_q_avail = vsi->num_queue_pairs - override_q;
1748                vsi->next_base_queue = override_q;
1749        }
1750        return 0;
1751}
1752
1753/**
1754 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1755 * @vsi: the VSI being setup
1756 * @ctxt: VSI context structure
1757 * @enabled_tc: Enabled TCs bitmap
1758 * @is_add: True if called before Add VSI
1759 *
1760 * Setup VSI queue mapping for enabled traffic classes.
1761 **/
1762static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1763                                     struct i40e_vsi_context *ctxt,
1764                                     u8 enabled_tc,
1765                                     bool is_add)
1766{
1767        struct i40e_pf *pf = vsi->back;
1768        u16 sections = 0;
1769        u8 netdev_tc = 0;
1770        u16 numtc = 1;
1771        u16 qcount;
1772        u8 offset;
1773        u16 qmap;
1774        int i;
1775        u16 num_tc_qps = 0;
1776
1777        sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1778        offset = 0;
1779
1780        /* Number of queues per enabled TC */
1781        num_tc_qps = vsi->alloc_queue_pairs;
1782        if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1783                /* Find numtc from enabled TC bitmap */
1784                for (i = 0, numtc = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1785                        if (enabled_tc & BIT(i)) /* TC is enabled */
1786                                numtc++;
1787                }
1788                if (!numtc) {
1789                        dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1790                        numtc = 1;
1791                }
1792                num_tc_qps = num_tc_qps / numtc;
1793                num_tc_qps = min_t(int, num_tc_qps,
1794                                   i40e_pf_get_max_q_per_tc(pf));
1795        }
1796
1797        vsi->tc_config.numtc = numtc;
1798        vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1799
1800        /* Do not allow use more TC queue pairs than MSI-X vectors exist */
1801        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1802                num_tc_qps = min_t(int, num_tc_qps, pf->num_lan_msix);
1803
1804        /* Setup queue offset/count for all TCs for given VSI */
1805        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1806                /* See if the given TC is enabled for the given VSI */
1807                if (vsi->tc_config.enabled_tc & BIT(i)) {
1808                        /* TC is enabled */
1809                        int pow, num_qps;
1810
1811                        switch (vsi->type) {
1812                        case I40E_VSI_MAIN:
1813                                if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED |
1814                                    I40E_FLAG_FD_ATR_ENABLED)) ||
1815                                    vsi->tc_config.enabled_tc != 1) {
1816                                        qcount = min_t(int, pf->alloc_rss_size,
1817                                                       num_tc_qps);
1818                                        break;
1819                                }
1820                                /* fall through */
1821                        case I40E_VSI_FDIR:
1822                        case I40E_VSI_SRIOV:
1823                        case I40E_VSI_VMDQ2:
1824                        default:
1825                                qcount = num_tc_qps;
1826                                WARN_ON(i != 0);
1827                                break;
1828                        }
1829                        vsi->tc_config.tc_info[i].qoffset = offset;
1830                        vsi->tc_config.tc_info[i].qcount = qcount;
1831
1832                        /* find the next higher power-of-2 of num queue pairs */
1833                        num_qps = qcount;
1834                        pow = 0;
1835                        while (num_qps && (BIT_ULL(pow) < qcount)) {
1836                                pow++;
1837                                num_qps >>= 1;
1838                        }
1839
1840                        vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1841                        qmap =
1842                            (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1843                            (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1844
1845                        offset += qcount;
1846                } else {
1847                        /* TC is not enabled so set the offset to
1848                         * default queue and allocate one queue
1849                         * for the given TC.
1850                         */
1851                        vsi->tc_config.tc_info[i].qoffset = 0;
1852                        vsi->tc_config.tc_info[i].qcount = 1;
1853                        vsi->tc_config.tc_info[i].netdev_tc = 0;
1854
1855                        qmap = 0;
1856                }
1857                ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1858        }
1859
1860        /* Set actual Tx/Rx queue pairs */
1861        vsi->num_queue_pairs = offset;
1862        if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1863                if (vsi->req_queue_pairs > 0)
1864                        vsi->num_queue_pairs = vsi->req_queue_pairs;
1865                else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1866                        vsi->num_queue_pairs = pf->num_lan_msix;
1867        }
1868
1869        /* Scheduler section valid can only be set for ADD VSI */
1870        if (is_add) {
1871                sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1872
1873                ctxt->info.up_enable_bits = enabled_tc;
1874        }
1875        if (vsi->type == I40E_VSI_SRIOV) {
1876                ctxt->info.mapping_flags |=
1877                                     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1878                for (i = 0; i < vsi->num_queue_pairs; i++)
1879                        ctxt->info.queue_mapping[i] =
1880                                               cpu_to_le16(vsi->base_queue + i);
1881        } else {
1882                ctxt->info.mapping_flags |=
1883                                        cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1884                ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1885        }
1886        ctxt->info.valid_sections |= cpu_to_le16(sections);
1887}
1888
1889/**
1890 * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1891 * @netdev: the netdevice
1892 * @addr: address to add
1893 *
1894 * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1895 * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1896 */
1897static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1898{
1899        struct i40e_netdev_priv *np = netdev_priv(netdev);
1900        struct i40e_vsi *vsi = np->vsi;
1901
1902        if (i40e_add_mac_filter(vsi, addr))
1903                return 0;
1904        else
1905                return -ENOMEM;
1906}
1907
1908/**
1909 * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1910 * @netdev: the netdevice
1911 * @addr: address to add
1912 *
1913 * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1914 * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1915 */
1916static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1917{
1918        struct i40e_netdev_priv *np = netdev_priv(netdev);
1919        struct i40e_vsi *vsi = np->vsi;
1920
1921        /* Under some circumstances, we might receive a request to delete
1922         * our own device address from our uc list. Because we store the
1923         * device address in the VSI's MAC/VLAN filter list, we need to ignore
1924         * such requests and not delete our device address from this list.
1925         */
1926        if (ether_addr_equal(addr, netdev->dev_addr))
1927                return 0;
1928
1929        i40e_del_mac_filter(vsi, addr);
1930
1931        return 0;
1932}
1933
1934/**
1935 * i40e_set_rx_mode - NDO callback to set the netdev filters
1936 * @netdev: network interface device structure
1937 **/
1938static void i40e_set_rx_mode(struct net_device *netdev)
1939{
1940        struct i40e_netdev_priv *np = netdev_priv(netdev);
1941        struct i40e_vsi *vsi = np->vsi;
1942
1943        spin_lock_bh(&vsi->mac_filter_hash_lock);
1944
1945        __dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1946        __dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1947
1948        spin_unlock_bh(&vsi->mac_filter_hash_lock);
1949
1950        /* check for other flag changes */
1951        if (vsi->current_netdev_flags != vsi->netdev->flags) {
1952                vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1953                set_bit(__I40E_MACVLAN_SYNC_PENDING, vsi->back->state);
1954        }
1955}
1956
1957/**
1958 * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1959 * @vsi: Pointer to VSI struct
1960 * @from: Pointer to list which contains MAC filter entries - changes to
1961 *        those entries needs to be undone.
1962 *
1963 * MAC filter entries from this list were slated for deletion.
1964 **/
1965static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1966                                         struct hlist_head *from)
1967{
1968        struct i40e_mac_filter *f;
1969        struct hlist_node *h;
1970
1971        hlist_for_each_entry_safe(f, h, from, hlist) {
1972                u64 key = i40e_addr_to_hkey(f->macaddr);
1973
1974                /* Move the element back into MAC filter list*/
1975                hlist_del(&f->hlist);
1976                hash_add(vsi->mac_filter_hash, &f->hlist, key);
1977        }
1978}
1979
1980/**
1981 * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1982 * @vsi: Pointer to vsi struct
1983 * @from: Pointer to list which contains MAC filter entries - changes to
1984 *        those entries needs to be undone.
1985 *
1986 * MAC filter entries from this list were slated for addition.
1987 **/
1988static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1989                                         struct hlist_head *from)
1990{
1991        struct i40e_new_mac_filter *new;
1992        struct hlist_node *h;
1993
1994        hlist_for_each_entry_safe(new, h, from, hlist) {
1995                /* We can simply free the wrapper structure */
1996                hlist_del(&new->hlist);
1997                kfree(new);
1998        }
1999}
2000
2001/**
2002 * i40e_next_entry - Get the next non-broadcast filter from a list
2003 * @next: pointer to filter in list
2004 *
2005 * Returns the next non-broadcast filter in the list. Required so that we
2006 * ignore broadcast filters within the list, since these are not handled via
2007 * the normal firmware update path.
2008 */
2009static
2010struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
2011{
2012        hlist_for_each_entry_continue(next, hlist) {
2013                if (!is_broadcast_ether_addr(next->f->macaddr))
2014                        return next;
2015        }
2016
2017        return NULL;
2018}
2019
2020/**
2021 * i40e_update_filter_state - Update filter state based on return data
2022 * from firmware
2023 * @count: Number of filters added
2024 * @add_list: return data from fw
2025 * @add_head: pointer to first filter in current batch
2026 *
2027 * MAC filter entries from list were slated to be added to device. Returns
2028 * number of successful filters. Note that 0 does NOT mean success!
2029 **/
2030static int
2031i40e_update_filter_state(int count,
2032                         struct i40e_aqc_add_macvlan_element_data *add_list,
2033                         struct i40e_new_mac_filter *add_head)
2034{
2035        int retval = 0;
2036        int i;
2037
2038        for (i = 0; i < count; i++) {
2039                /* Always check status of each filter. We don't need to check
2040                 * the firmware return status because we pre-set the filter
2041                 * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
2042                 * request to the adminq. Thus, if it no longer matches then
2043                 * we know the filter is active.
2044                 */
2045                if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
2046                        add_head->state = I40E_FILTER_FAILED;
2047                } else {
2048                        add_head->state = I40E_FILTER_ACTIVE;
2049                        retval++;
2050                }
2051
2052                add_head = i40e_next_filter(add_head);
2053                if (!add_head)
2054                        break;
2055        }
2056
2057        return retval;
2058}
2059
2060/**
2061 * i40e_aqc_del_filters - Request firmware to delete a set of filters
2062 * @vsi: ptr to the VSI
2063 * @vsi_name: name to display in messages
2064 * @list: the list of filters to send to firmware
2065 * @num_del: the number of filters to delete
2066 * @retval: Set to -EIO on failure to delete
2067 *
2068 * Send a request to firmware via AdminQ to delete a set of filters. Uses
2069 * *retval instead of a return value so that success does not force ret_val to
2070 * be set to 0. This ensures that a sequence of calls to this function
2071 * preserve the previous value of *retval on successful delete.
2072 */
2073static
2074void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
2075                          struct i40e_aqc_remove_macvlan_element_data *list,
2076                          int num_del, int *retval)
2077{
2078        struct i40e_hw *hw = &vsi->back->hw;
2079        i40e_status aq_ret;
2080        int aq_err;
2081
2082        aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
2083        aq_err = hw->aq.asq_last_status;
2084
2085        /* Explicitly ignore and do not report when firmware returns ENOENT */
2086        if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
2087                *retval = -EIO;
2088                dev_info(&vsi->back->pdev->dev,
2089                         "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
2090                         vsi_name, i40e_stat_str(hw, aq_ret),
2091                         i40e_aq_str(hw, aq_err));
2092        }
2093}
2094
2095/**
2096 * i40e_aqc_add_filters - Request firmware to add a set of filters
2097 * @vsi: ptr to the VSI
2098 * @vsi_name: name to display in messages
2099 * @list: the list of filters to send to firmware
2100 * @add_head: Position in the add hlist
2101 * @num_add: the number of filters to add
2102 *
2103 * Send a request to firmware via AdminQ to add a chunk of filters. Will set
2104 * __I40E_VSI_OVERFLOW_PROMISC bit in vsi->state if the firmware has run out of
2105 * space for more filters.
2106 */
2107static
2108void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
2109                          struct i40e_aqc_add_macvlan_element_data *list,
2110                          struct i40e_new_mac_filter *add_head,
2111                          int num_add)
2112{
2113        struct i40e_hw *hw = &vsi->back->hw;
2114        int aq_err, fcnt;
2115
2116        i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
2117        aq_err = hw->aq.asq_last_status;
2118        fcnt = i40e_update_filter_state(num_add, list, add_head);
2119
2120        if (fcnt != num_add) {
2121                if (vsi->type == I40E_VSI_MAIN) {
2122                        set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2123                        dev_warn(&vsi->back->pdev->dev,
2124                                 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2125                                 i40e_aq_str(hw, aq_err), vsi_name);
2126                } else if (vsi->type == I40E_VSI_SRIOV ||
2127                           vsi->type == I40E_VSI_VMDQ1 ||
2128                           vsi->type == I40E_VSI_VMDQ2) {
2129                        dev_warn(&vsi->back->pdev->dev,
2130                                 "Error %s adding RX filters on %s, please set promiscuous on manually for %s\n",
2131                                 i40e_aq_str(hw, aq_err), vsi_name, vsi_name);
2132                } else {
2133                        dev_warn(&vsi->back->pdev->dev,
2134                                 "Error %s adding RX filters on %s, incorrect VSI type: %i.\n",
2135                                 i40e_aq_str(hw, aq_err), vsi_name, vsi->type);
2136                }
2137        }
2138}
2139
2140/**
2141 * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
2142 * @vsi: pointer to the VSI
2143 * @vsi_name: the VSI name
2144 * @f: filter data
2145 *
2146 * This function sets or clears the promiscuous broadcast flags for VLAN
2147 * filters in order to properly receive broadcast frames. Assumes that only
2148 * broadcast filters are passed.
2149 *
2150 * Returns status indicating success or failure;
2151 **/
2152static i40e_status
2153i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
2154                          struct i40e_mac_filter *f)
2155{
2156        bool enable = f->state == I40E_FILTER_NEW;
2157        struct i40e_hw *hw = &vsi->back->hw;
2158        i40e_status aq_ret;
2159
2160        if (f->vlan == I40E_VLAN_ANY) {
2161                aq_ret = i40e_aq_set_vsi_broadcast(hw,
2162                                                   vsi->seid,
2163                                                   enable,
2164                                                   NULL);
2165        } else {
2166                aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
2167                                                            vsi->seid,
2168                                                            enable,
2169                                                            f->vlan,
2170                                                            NULL);
2171        }
2172
2173        if (aq_ret) {
2174                set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2175                dev_warn(&vsi->back->pdev->dev,
2176                         "Error %s, forcing overflow promiscuous on %s\n",
2177                         i40e_aq_str(hw, hw->aq.asq_last_status),
2178                         vsi_name);
2179        }
2180
2181        return aq_ret;
2182}
2183
2184/**
2185 * i40e_set_promiscuous - set promiscuous mode
2186 * @pf: board private structure
2187 * @promisc: promisc on or off
2188 *
2189 * There are different ways of setting promiscuous mode on a PF depending on
2190 * what state/environment we're in.  This identifies and sets it appropriately.
2191 * Returns 0 on success.
2192 **/
2193static int i40e_set_promiscuous(struct i40e_pf *pf, bool promisc)
2194{
2195        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
2196        struct i40e_hw *hw = &pf->hw;
2197        i40e_status aq_ret;
2198
2199        if (vsi->type == I40E_VSI_MAIN &&
2200            pf->lan_veb != I40E_NO_VEB &&
2201            !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2202                /* set defport ON for Main VSI instead of true promisc
2203                 * this way we will get all unicast/multicast and VLAN
2204                 * promisc behavior but will not get VF or VMDq traffic
2205                 * replicated on the Main VSI.
2206                 */
2207                if (promisc)
2208                        aq_ret = i40e_aq_set_default_vsi(hw,
2209                                                         vsi->seid,
2210                                                         NULL);
2211                else
2212                        aq_ret = i40e_aq_clear_default_vsi(hw,
2213                                                           vsi->seid,
2214                                                           NULL);
2215                if (aq_ret) {
2216                        dev_info(&pf->pdev->dev,
2217                                 "Set default VSI failed, err %s, aq_err %s\n",
2218                                 i40e_stat_str(hw, aq_ret),
2219                                 i40e_aq_str(hw, hw->aq.asq_last_status));
2220                }
2221        } else {
2222                aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2223                                                  hw,
2224                                                  vsi->seid,
2225                                                  promisc, NULL,
2226                                                  true);
2227                if (aq_ret) {
2228                        dev_info(&pf->pdev->dev,
2229                                 "set unicast promisc failed, err %s, aq_err %s\n",
2230                                 i40e_stat_str(hw, aq_ret),
2231                                 i40e_aq_str(hw, hw->aq.asq_last_status));
2232                }
2233                aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2234                                                  hw,
2235                                                  vsi->seid,
2236                                                  promisc, NULL);
2237                if (aq_ret) {
2238                        dev_info(&pf->pdev->dev,
2239                                 "set multicast promisc failed, err %s, aq_err %s\n",
2240                                 i40e_stat_str(hw, aq_ret),
2241                                 i40e_aq_str(hw, hw->aq.asq_last_status));
2242                }
2243        }
2244
2245        if (!aq_ret)
2246                pf->cur_promisc = promisc;
2247
2248        return aq_ret;
2249}
2250
2251/**
2252 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
2253 * @vsi: ptr to the VSI
2254 *
2255 * Push any outstanding VSI filter changes through the AdminQ.
2256 *
2257 * Returns 0 or error value
2258 **/
2259int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
2260{
2261        struct hlist_head tmp_add_list, tmp_del_list;
2262        struct i40e_mac_filter *f;
2263        struct i40e_new_mac_filter *new, *add_head = NULL;
2264        struct i40e_hw *hw = &vsi->back->hw;
2265        bool old_overflow, new_overflow;
2266        unsigned int failed_filters = 0;
2267        unsigned int vlan_filters = 0;
2268        char vsi_name[16] = "PF";
2269        int filter_list_len = 0;
2270        i40e_status aq_ret = 0;
2271        u32 changed_flags = 0;
2272        struct hlist_node *h;
2273        struct i40e_pf *pf;
2274        int num_add = 0;
2275        int num_del = 0;
2276        int retval = 0;
2277        u16 cmd_flags;
2278        int list_size;
2279        int bkt;
2280
2281        /* empty array typed pointers, kcalloc later */
2282        struct i40e_aqc_add_macvlan_element_data *add_list;
2283        struct i40e_aqc_remove_macvlan_element_data *del_list;
2284
2285        while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2286                usleep_range(1000, 2000);
2287        pf = vsi->back;
2288
2289        old_overflow = test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2290
2291        if (vsi->netdev) {
2292                changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2293                vsi->current_netdev_flags = vsi->netdev->flags;
2294        }
2295
2296        INIT_HLIST_HEAD(&tmp_add_list);
2297        INIT_HLIST_HEAD(&tmp_del_list);
2298
2299        if (vsi->type == I40E_VSI_SRIOV)
2300                snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2301        else if (vsi->type != I40E_VSI_MAIN)
2302                snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2303
2304        if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2305                vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2306
2307                spin_lock_bh(&vsi->mac_filter_hash_lock);
2308                /* Create a list of filters to delete. */
2309                hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2310                        if (f->state == I40E_FILTER_REMOVE) {
2311                                /* Move the element into temporary del_list */
2312                                hash_del(&f->hlist);
2313                                hlist_add_head(&f->hlist, &tmp_del_list);
2314
2315                                /* Avoid counting removed filters */
2316                                continue;
2317                        }
2318                        if (f->state == I40E_FILTER_NEW) {
2319                                /* Create a temporary i40e_new_mac_filter */
2320                                new = kzalloc(sizeof(*new), GFP_ATOMIC);
2321                                if (!new)
2322                                        goto err_no_memory_locked;
2323
2324                                /* Store pointer to the real filter */
2325                                new->f = f;
2326                                new->state = f->state;
2327
2328                                /* Add it to the hash list */
2329                                hlist_add_head(&new->hlist, &tmp_add_list);
2330                        }
2331
2332                        /* Count the number of active (current and new) VLAN
2333                         * filters we have now. Does not count filters which
2334                         * are marked for deletion.
2335                         */
2336                        if (f->vlan > 0)
2337                                vlan_filters++;
2338                }
2339
2340                retval = i40e_correct_mac_vlan_filters(vsi,
2341                                                       &tmp_add_list,
2342                                                       &tmp_del_list,
2343                                                       vlan_filters);
2344                if (retval)
2345                        goto err_no_memory_locked;
2346
2347                spin_unlock_bh(&vsi->mac_filter_hash_lock);
2348        }
2349
2350        /* Now process 'del_list' outside the lock */
2351        if (!hlist_empty(&tmp_del_list)) {
2352                filter_list_len = hw->aq.asq_buf_size /
2353                            sizeof(struct i40e_aqc_remove_macvlan_element_data);
2354                list_size = filter_list_len *
2355                            sizeof(struct i40e_aqc_remove_macvlan_element_data);
2356                del_list = kzalloc(list_size, GFP_ATOMIC);
2357                if (!del_list)
2358                        goto err_no_memory;
2359
2360                hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2361                        cmd_flags = 0;
2362
2363                        /* handle broadcast filters by updating the broadcast
2364                         * promiscuous flag and release filter list.
2365                         */
2366                        if (is_broadcast_ether_addr(f->macaddr)) {
2367                                i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2368
2369                                hlist_del(&f->hlist);
2370                                kfree(f);
2371                                continue;
2372                        }
2373
2374                        /* add to delete list */
2375                        ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2376                        if (f->vlan == I40E_VLAN_ANY) {
2377                                del_list[num_del].vlan_tag = 0;
2378                                cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2379                        } else {
2380                                del_list[num_del].vlan_tag =
2381                                        cpu_to_le16((u16)(f->vlan));
2382                        }
2383
2384                        cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2385                        del_list[num_del].flags = cmd_flags;
2386                        num_del++;
2387
2388                        /* flush a full buffer */
2389                        if (num_del == filter_list_len) {
2390                                i40e_aqc_del_filters(vsi, vsi_name, del_list,
2391                                                     num_del, &retval);
2392                                memset(del_list, 0, list_size);
2393                                num_del = 0;
2394                        }
2395                        /* Release memory for MAC filter entries which were
2396                         * synced up with HW.
2397                         */
2398                        hlist_del(&f->hlist);
2399                        kfree(f);
2400                }
2401
2402                if (num_del) {
2403                        i40e_aqc_del_filters(vsi, vsi_name, del_list,
2404                                             num_del, &retval);
2405                }
2406
2407                kfree(del_list);
2408                del_list = NULL;
2409        }
2410
2411        if (!hlist_empty(&tmp_add_list)) {
2412                /* Do all the adds now. */
2413                filter_list_len = hw->aq.asq_buf_size /
2414                               sizeof(struct i40e_aqc_add_macvlan_element_data);
2415                list_size = filter_list_len *
2416                               sizeof(struct i40e_aqc_add_macvlan_element_data);
2417                add_list = kzalloc(list_size, GFP_ATOMIC);
2418                if (!add_list)
2419                        goto err_no_memory;
2420
2421                num_add = 0;
2422                hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2423                        /* handle broadcast filters by updating the broadcast
2424                         * promiscuous flag instead of adding a MAC filter.
2425                         */
2426                        if (is_broadcast_ether_addr(new->f->macaddr)) {
2427                                if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2428                                                              new->f))
2429                                        new->state = I40E_FILTER_FAILED;
2430                                else
2431                                        new->state = I40E_FILTER_ACTIVE;
2432                                continue;
2433                        }
2434
2435                        /* add to add array */
2436                        if (num_add == 0)
2437                                add_head = new;
2438                        cmd_flags = 0;
2439                        ether_addr_copy(add_list[num_add].mac_addr,
2440                                        new->f->macaddr);
2441                        if (new->f->vlan == I40E_VLAN_ANY) {
2442                                add_list[num_add].vlan_tag = 0;
2443                                cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2444                        } else {
2445                                add_list[num_add].vlan_tag =
2446                                        cpu_to_le16((u16)(new->f->vlan));
2447                        }
2448                        add_list[num_add].queue_number = 0;
2449                        /* set invalid match method for later detection */
2450                        add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2451                        cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2452                        add_list[num_add].flags = cpu_to_le16(cmd_flags);
2453                        num_add++;
2454
2455                        /* flush a full buffer */
2456                        if (num_add == filter_list_len) {
2457                                i40e_aqc_add_filters(vsi, vsi_name, add_list,
2458                                                     add_head, num_add);
2459                                memset(add_list, 0, list_size);
2460                                num_add = 0;
2461                        }
2462                }
2463                if (num_add) {
2464                        i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2465                                             num_add);
2466                }
2467                /* Now move all of the filters from the temp add list back to
2468                 * the VSI's list.
2469                 */
2470                spin_lock_bh(&vsi->mac_filter_hash_lock);
2471                hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2472                        /* Only update the state if we're still NEW */
2473                        if (new->f->state == I40E_FILTER_NEW)
2474                                new->f->state = new->state;
2475                        hlist_del(&new->hlist);
2476                        kfree(new);
2477                }
2478                spin_unlock_bh(&vsi->mac_filter_hash_lock);
2479                kfree(add_list);
2480                add_list = NULL;
2481        }
2482
2483        /* Determine the number of active and failed filters. */
2484        spin_lock_bh(&vsi->mac_filter_hash_lock);
2485        vsi->active_filters = 0;
2486        hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2487                if (f->state == I40E_FILTER_ACTIVE)
2488                        vsi->active_filters++;
2489                else if (f->state == I40E_FILTER_FAILED)
2490                        failed_filters++;
2491        }
2492        spin_unlock_bh(&vsi->mac_filter_hash_lock);
2493
2494        /* Check if we are able to exit overflow promiscuous mode. We can
2495         * safely exit if we didn't just enter, we no longer have any failed
2496         * filters, and we have reduced filters below the threshold value.
2497         */
2498        if (old_overflow && !failed_filters &&
2499            vsi->active_filters < vsi->promisc_threshold) {
2500                dev_info(&pf->pdev->dev,
2501                         "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2502                         vsi_name);
2503                clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2504                vsi->promisc_threshold = 0;
2505        }
2506
2507        /* if the VF is not trusted do not do promisc */
2508        if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2509                clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2510                goto out;
2511        }
2512
2513        new_overflow = test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2514
2515        /* If we are entering overflow promiscuous, we need to calculate a new
2516         * threshold for when we are safe to exit
2517         */
2518        if (!old_overflow && new_overflow)
2519                vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2520
2521        /* check for changes in promiscuous modes */
2522        if (changed_flags & IFF_ALLMULTI) {
2523                bool cur_multipromisc;
2524
2525                cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2526                aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2527                                                               vsi->seid,
2528                                                               cur_multipromisc,
2529                                                               NULL);
2530                if (aq_ret) {
2531                        retval = i40e_aq_rc_to_posix(aq_ret,
2532                                                     hw->aq.asq_last_status);
2533                        dev_info(&pf->pdev->dev,
2534                                 "set multi promisc failed on %s, err %s aq_err %s\n",
2535                                 vsi_name,
2536                                 i40e_stat_str(hw, aq_ret),
2537                                 i40e_aq_str(hw, hw->aq.asq_last_status));
2538                } else {
2539                        dev_info(&pf->pdev->dev, "%s is %s allmulti mode.\n",
2540                                 vsi->netdev->name,
2541                                 cur_multipromisc ? "entering" : "leaving");
2542                }
2543        }
2544
2545        if ((changed_flags & IFF_PROMISC) || old_overflow != new_overflow) {
2546                bool cur_promisc;
2547
2548                cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2549                               new_overflow);
2550                aq_ret = i40e_set_promiscuous(pf, cur_promisc);
2551                if (aq_ret) {
2552                        retval = i40e_aq_rc_to_posix(aq_ret,
2553                                                     hw->aq.asq_last_status);
2554                        dev_info(&pf->pdev->dev,
2555                                 "Setting promiscuous %s failed on %s, err %s aq_err %s\n",
2556                                 cur_promisc ? "on" : "off",
2557                                 vsi_name,
2558                                 i40e_stat_str(hw, aq_ret),
2559                                 i40e_aq_str(hw, hw->aq.asq_last_status));
2560                }
2561        }
2562out:
2563        /* if something went wrong then set the changed flag so we try again */
2564        if (retval)
2565                vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2566
2567        clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2568        return retval;
2569
2570err_no_memory:
2571        /* Restore elements on the temporary add and delete lists */
2572        spin_lock_bh(&vsi->mac_filter_hash_lock);
2573err_no_memory_locked:
2574        i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2575        i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2576        spin_unlock_bh(&vsi->mac_filter_hash_lock);
2577
2578        vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2579        clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2580        return -ENOMEM;
2581}
2582
2583/**
2584 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2585 * @pf: board private structure
2586 **/
2587static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2588{
2589        int v;
2590
2591        if (!pf)
2592                return;
2593        if (!test_and_clear_bit(__I40E_MACVLAN_SYNC_PENDING, pf->state))
2594                return;
2595        if (test_and_set_bit(__I40E_VF_DISABLE, pf->state)) {
2596                set_bit(__I40E_MACVLAN_SYNC_PENDING, pf->state);
2597                return;
2598        }
2599
2600        for (v = 0; v < pf->num_alloc_vsi; v++) {
2601                if (pf->vsi[v] &&
2602                    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2603                        int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2604
2605                        if (ret) {
2606                                /* come back and try again later */
2607                                set_bit(__I40E_MACVLAN_SYNC_PENDING,
2608                                        pf->state);
2609                                break;
2610                        }
2611                }
2612        }
2613        clear_bit(__I40E_VF_DISABLE, pf->state);
2614}
2615
2616/**
2617 * i40e_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2618 * @vsi: the vsi
2619 **/
2620static int i40e_max_xdp_frame_size(struct i40e_vsi *vsi)
2621{
2622        if (PAGE_SIZE >= 8192 || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
2623                return I40E_RXBUFFER_2048;
2624        else
2625                return I40E_RXBUFFER_3072;
2626}
2627
2628/**
2629 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2630 * @netdev: network interface device structure
2631 * @new_mtu: new value for maximum frame size
2632 *
2633 * Returns 0 on success, negative on failure
2634 **/
2635static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2636{
2637        struct i40e_netdev_priv *np = netdev_priv(netdev);
2638        struct i40e_vsi *vsi = np->vsi;
2639        struct i40e_pf *pf = vsi->back;
2640
2641        if (i40e_enabled_xdp_vsi(vsi)) {
2642                int frame_size = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2643
2644                if (frame_size > i40e_max_xdp_frame_size(vsi))
2645                        return -EINVAL;
2646        }
2647
2648        netdev_info(netdev, "changing MTU from %d to %d\n",
2649                    netdev->mtu, new_mtu);
2650        netdev->mtu = new_mtu;
2651        if (netif_running(netdev))
2652                i40e_vsi_reinit_locked(vsi);
2653        set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
2654        set_bit(__I40E_CLIENT_L2_CHANGE, pf->state);
2655        return 0;
2656}
2657
2658/**
2659 * i40e_ioctl - Access the hwtstamp interface
2660 * @netdev: network interface device structure
2661 * @ifr: interface request data
2662 * @cmd: ioctl command
2663 **/
2664int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2665{
2666        struct i40e_netdev_priv *np = netdev_priv(netdev);
2667        struct i40e_pf *pf = np->vsi->back;
2668
2669        switch (cmd) {
2670        case SIOCGHWTSTAMP:
2671                return i40e_ptp_get_ts_config(pf, ifr);
2672        case SIOCSHWTSTAMP:
2673                return i40e_ptp_set_ts_config(pf, ifr);
2674        default:
2675                return -EOPNOTSUPP;
2676        }
2677}
2678
2679/**
2680 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2681 * @vsi: the vsi being adjusted
2682 **/
2683void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2684{
2685        struct i40e_vsi_context ctxt;
2686        i40e_status ret;
2687
2688        /* Don't modify stripping options if a port VLAN is active */
2689        if (vsi->info.pvid)
2690                return;
2691
2692        if ((vsi->info.valid_sections &
2693             cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2694            ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2695                return;  /* already enabled */
2696
2697        vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2698        vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2699                                    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2700
2701        ctxt.seid = vsi->seid;
2702        ctxt.info = vsi->info;
2703        ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2704        if (ret) {
2705                dev_info(&vsi->back->pdev->dev,
2706                         "update vlan stripping failed, err %s aq_err %s\n",
2707                         i40e_stat_str(&vsi->back->hw, ret),
2708                         i40e_aq_str(&vsi->back->hw,
2709                                     vsi->back->hw.aq.asq_last_status));
2710        }
2711}
2712
2713/**
2714 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2715 * @vsi: the vsi being adjusted
2716 **/
2717void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2718{
2719        struct i40e_vsi_context ctxt;
2720        i40e_status ret;
2721
2722        /* Don't modify stripping options if a port VLAN is active */
2723        if (vsi->info.pvid)
2724                return;
2725
2726        if ((vsi->info.valid_sections &
2727             cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2728            ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2729             I40E_AQ_VSI_PVLAN_EMOD_MASK))
2730                return;  /* already disabled */
2731
2732        vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2733        vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2734                                    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2735
2736        ctxt.seid = vsi->seid;
2737        ctxt.info = vsi->info;
2738        ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2739        if (ret) {
2740                dev_info(&vsi->back->pdev->dev,
2741                         "update vlan stripping failed, err %s aq_err %s\n",
2742                         i40e_stat_str(&vsi->back->hw, ret),
2743                         i40e_aq_str(&vsi->back->hw,
2744                                     vsi->back->hw.aq.asq_last_status));
2745        }
2746}
2747
2748/**
2749 * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2750 * @vsi: the vsi being configured
2751 * @vid: vlan id to be added (0 = untagged only , -1 = any)
2752 *
2753 * This is a helper function for adding a new MAC/VLAN filter with the
2754 * specified VLAN for each existing MAC address already in the hash table.
2755 * This function does *not* perform any accounting to update filters based on
2756 * VLAN mode.
2757 *
2758 * NOTE: this function expects to be called while under the
2759 * mac_filter_hash_lock
2760 **/
2761int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2762{
2763        struct i40e_mac_filter *f, *add_f;
2764        struct hlist_node *h;
2765        int bkt;
2766
2767        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2768                if (f->state == I40E_FILTER_REMOVE)
2769                        continue;
2770                add_f = i40e_add_filter(vsi, f->macaddr, vid);
2771                if (!add_f) {
2772                        dev_info(&vsi->back->pdev->dev,
2773                                 "Could not add vlan filter %d for %pM\n",
2774                                 vid, f->macaddr);
2775                        return -ENOMEM;
2776                }
2777        }
2778
2779        return 0;
2780}
2781
2782/**
2783 * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2784 * @vsi: the VSI being configured
2785 * @vid: VLAN id to be added
2786 **/
2787int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2788{
2789        int err;
2790
2791        if (vsi->info.pvid)
2792                return -EINVAL;
2793
2794        /* The network stack will attempt to add VID=0, with the intention to
2795         * receive priority tagged packets with a VLAN of 0. Our HW receives
2796         * these packets by default when configured to receive untagged
2797         * packets, so we don't need to add a filter for this case.
2798         * Additionally, HW interprets adding a VID=0 filter as meaning to
2799         * receive *only* tagged traffic and stops receiving untagged traffic.
2800         * Thus, we do not want to actually add a filter for VID=0
2801         */
2802        if (!vid)
2803                return 0;
2804
2805        /* Locked once because all functions invoked below iterates list*/
2806        spin_lock_bh(&vsi->mac_filter_hash_lock);
2807        err = i40e_add_vlan_all_mac(vsi, vid);
2808        spin_unlock_bh(&vsi->mac_filter_hash_lock);
2809        if (err)
2810                return err;
2811
2812        /* schedule our worker thread which will take care of
2813         * applying the new filter changes
2814         */
2815        i40e_service_event_schedule(vsi->back);
2816        return 0;
2817}
2818
2819/**
2820 * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2821 * @vsi: the vsi being configured
2822 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2823 *
2824 * This function should be used to remove all VLAN filters which match the
2825 * given VID. It does not schedule the service event and does not take the
2826 * mac_filter_hash_lock so it may be combined with other operations under
2827 * a single invocation of the mac_filter_hash_lock.
2828 *
2829 * NOTE: this function expects to be called while under the
2830 * mac_filter_hash_lock
2831 */
2832void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2833{
2834        struct i40e_mac_filter *f;
2835        struct hlist_node *h;
2836        int bkt;
2837
2838        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2839                if (f->vlan == vid)
2840                        __i40e_del_filter(vsi, f);
2841        }
2842}
2843
2844/**
2845 * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2846 * @vsi: the VSI being configured
2847 * @vid: VLAN id to be removed
2848 **/
2849void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2850{
2851        if (!vid || vsi->info.pvid)
2852                return;
2853
2854        spin_lock_bh(&vsi->mac_filter_hash_lock);
2855        i40e_rm_vlan_all_mac(vsi, vid);
2856        spin_unlock_bh(&vsi->mac_filter_hash_lock);
2857
2858        /* schedule our worker thread which will take care of
2859         * applying the new filter changes
2860         */
2861        i40e_service_event_schedule(vsi->back);
2862}
2863
2864/**
2865 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2866 * @netdev: network interface to be adjusted
2867 * @proto: unused protocol value
2868 * @vid: vlan id to be added
2869 *
2870 * net_device_ops implementation for adding vlan ids
2871 **/
2872static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2873                                __always_unused __be16 proto, u16 vid)
2874{
2875        struct i40e_netdev_priv *np = netdev_priv(netdev);
2876        struct i40e_vsi *vsi = np->vsi;
2877        int ret = 0;
2878
2879        if (vid >= VLAN_N_VID)
2880                return -EINVAL;
2881
2882        ret = i40e_vsi_add_vlan(vsi, vid);
2883        if (!ret)
2884                set_bit(vid, vsi->active_vlans);
2885
2886        return ret;
2887}
2888
2889/**
2890 * i40e_vlan_rx_add_vid_up - Add a vlan id filter to HW offload in UP path
2891 * @netdev: network interface to be adjusted
2892 * @proto: unused protocol value
2893 * @vid: vlan id to be added
2894 **/
2895static void i40e_vlan_rx_add_vid_up(struct net_device *netdev,
2896                                    __always_unused __be16 proto, u16 vid)
2897{
2898        struct i40e_netdev_priv *np = netdev_priv(netdev);
2899        struct i40e_vsi *vsi = np->vsi;
2900
2901        if (vid >= VLAN_N_VID)
2902                return;
2903        set_bit(vid, vsi->active_vlans);
2904}
2905
2906/**
2907 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2908 * @netdev: network interface to be adjusted
2909 * @proto: unused protocol value
2910 * @vid: vlan id to be removed
2911 *
2912 * net_device_ops implementation for removing vlan ids
2913 **/
2914static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2915                                 __always_unused __be16 proto, u16 vid)
2916{
2917        struct i40e_netdev_priv *np = netdev_priv(netdev);
2918        struct i40e_vsi *vsi = np->vsi;
2919
2920        /* return code is ignored as there is nothing a user
2921         * can do about failure to remove and a log message was
2922         * already printed from the other function
2923         */
2924        i40e_vsi_kill_vlan(vsi, vid);
2925
2926        clear_bit(vid, vsi->active_vlans);
2927
2928        return 0;
2929}
2930
2931/**
2932 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2933 * @vsi: the vsi being brought back up
2934 **/
2935static void i40e_restore_vlan(struct i40e_vsi *vsi)
2936{
2937        u16 vid;
2938
2939        if (!vsi->netdev)
2940                return;
2941
2942        if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
2943                i40e_vlan_stripping_enable(vsi);
2944        else
2945                i40e_vlan_stripping_disable(vsi);
2946
2947        for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2948                i40e_vlan_rx_add_vid_up(vsi->netdev, htons(ETH_P_8021Q),
2949                                        vid);
2950}
2951
2952/**
2953 * i40e_vsi_add_pvid - Add pvid for the VSI
2954 * @vsi: the vsi being adjusted
2955 * @vid: the vlan id to set as a PVID
2956 **/
2957int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2958{
2959        struct i40e_vsi_context ctxt;
2960        i40e_status ret;
2961
2962        vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2963        vsi->info.pvid = cpu_to_le16(vid);
2964        vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2965                                    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2966                                    I40E_AQ_VSI_PVLAN_EMOD_STR;
2967
2968        ctxt.seid = vsi->seid;
2969        ctxt.info = vsi->info;
2970        ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2971        if (ret) {
2972                dev_info(&vsi->back->pdev->dev,
2973                         "add pvid failed, err %s aq_err %s\n",
2974                         i40e_stat_str(&vsi->back->hw, ret),
2975                         i40e_aq_str(&vsi->back->hw,
2976                                     vsi->back->hw.aq.asq_last_status));
2977                return -ENOENT;
2978        }
2979
2980        return 0;
2981}
2982
2983/**
2984 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2985 * @vsi: the vsi being adjusted
2986 *
2987 * Just use the vlan_rx_register() service to put it back to normal
2988 **/
2989void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2990{
2991        vsi->info.pvid = 0;
2992
2993        i40e_vlan_stripping_disable(vsi);
2994}
2995
2996/**
2997 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2998 * @vsi: ptr to the VSI
2999 *
3000 * If this function returns with an error, then it's possible one or
3001 * more of the rings is populated (while the rest are not).  It is the
3002 * callers duty to clean those orphaned rings.
3003 *
3004 * Return 0 on success, negative on failure
3005 **/
3006static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
3007{
3008        int i, err = 0;
3009
3010        for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3011                err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
3012
3013        if (!i40e_enabled_xdp_vsi(vsi))
3014                return err;
3015
3016        for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3017                err = i40e_setup_tx_descriptors(vsi->xdp_rings[i]);
3018
3019        return err;
3020}
3021
3022/**
3023 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
3024 * @vsi: ptr to the VSI
3025 *
3026 * Free VSI's transmit software resources
3027 **/
3028static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
3029{
3030        int i;
3031
3032        if (vsi->tx_rings) {
3033                for (i = 0; i < vsi->num_queue_pairs; i++)
3034                        if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
3035                                i40e_free_tx_resources(vsi->tx_rings[i]);
3036        }
3037
3038        if (vsi->xdp_rings) {
3039                for (i = 0; i < vsi->num_queue_pairs; i++)
3040                        if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
3041                                i40e_free_tx_resources(vsi->xdp_rings[i]);
3042        }
3043}
3044
3045/**
3046 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
3047 * @vsi: ptr to the VSI
3048 *
3049 * If this function returns with an error, then it's possible one or
3050 * more of the rings is populated (while the rest are not).  It is the
3051 * callers duty to clean those orphaned rings.
3052 *
3053 * Return 0 on success, negative on failure
3054 **/
3055static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
3056{
3057        int i, err = 0;
3058
3059        for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3060                err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
3061        return err;
3062}
3063
3064/**
3065 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
3066 * @vsi: ptr to the VSI
3067 *
3068 * Free all receive software resources
3069 **/
3070static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
3071{
3072        int i;
3073
3074        if (!vsi->rx_rings)
3075                return;
3076
3077        for (i = 0; i < vsi->num_queue_pairs; i++)
3078                if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
3079                        i40e_free_rx_resources(vsi->rx_rings[i]);
3080}
3081
3082/**
3083 * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
3084 * @ring: The Tx ring to configure
3085 *
3086 * This enables/disables XPS for a given Tx descriptor ring
3087 * based on the TCs enabled for the VSI that ring belongs to.
3088 **/
3089static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
3090{
3091        int cpu;
3092
3093        if (!ring->q_vector || !ring->netdev || ring->ch)
3094                return;
3095
3096        /* We only initialize XPS once, so as not to overwrite user settings */
3097        if (test_and_set_bit(__I40E_TX_XPS_INIT_DONE, ring->state))
3098                return;
3099
3100        cpu = cpumask_local_spread(ring->q_vector->v_idx, -1);
3101        netif_set_xps_queue(ring->netdev, get_cpu_mask(cpu),
3102                            ring->queue_index);
3103}
3104
3105/**
3106 * i40e_xsk_umem - Retrieve the AF_XDP ZC if XDP and ZC is enabled
3107 * @ring: The Tx or Rx ring
3108 *
3109 * Returns the UMEM or NULL.
3110 **/
3111static struct xdp_umem *i40e_xsk_umem(struct i40e_ring *ring)
3112{
3113        bool xdp_on = i40e_enabled_xdp_vsi(ring->vsi);
3114        int qid = ring->queue_index;
3115
3116        if (ring_is_xdp(ring))
3117                qid -= ring->vsi->alloc_queue_pairs;
3118
3119        if (!xdp_on || !test_bit(qid, ring->vsi->af_xdp_zc_qps))
3120                return NULL;
3121
3122        return xdp_get_umem_from_qid(ring->vsi->netdev, qid);
3123}
3124
3125/**
3126 * i40e_configure_tx_ring - Configure a transmit ring context and rest
3127 * @ring: The Tx ring to configure
3128 *
3129 * Configure the Tx descriptor ring in the HMC context.
3130 **/
3131static int i40e_configure_tx_ring(struct i40e_ring *ring)
3132{
3133        struct i40e_vsi *vsi = ring->vsi;
3134        u16 pf_q = vsi->base_queue + ring->queue_index;
3135        struct i40e_hw *hw = &vsi->back->hw;
3136        struct i40e_hmc_obj_txq tx_ctx;
3137        i40e_status err = 0;
3138        u32 qtx_ctl = 0;
3139
3140        if (ring_is_xdp(ring))
3141                ring->xsk_umem = i40e_xsk_umem(ring);
3142
3143        /* some ATR related tx ring init */
3144        if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
3145                ring->atr_sample_rate = vsi->back->atr_sample_rate;
3146                ring->atr_count = 0;
3147        } else {
3148                ring->atr_sample_rate = 0;
3149        }
3150
3151        /* configure XPS */
3152        i40e_config_xps_tx_ring(ring);
3153
3154        /* clear the context structure first */
3155        memset(&tx_ctx, 0, sizeof(tx_ctx));
3156
3157        tx_ctx.new_context = 1;
3158        tx_ctx.base = (ring->dma / 128);
3159        tx_ctx.qlen = ring->count;
3160        tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
3161                                               I40E_FLAG_FD_ATR_ENABLED));
3162        tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
3163        /* FDIR VSI tx ring can still use RS bit and writebacks */
3164        if (vsi->type != I40E_VSI_FDIR)
3165                tx_ctx.head_wb_ena = 1;
3166        tx_ctx.head_wb_addr = ring->dma +
3167                              (ring->count * sizeof(struct i40e_tx_desc));
3168
3169        /* As part of VSI creation/update, FW allocates certain
3170         * Tx arbitration queue sets for each TC enabled for
3171         * the VSI. The FW returns the handles to these queue
3172         * sets as part of the response buffer to Add VSI,
3173         * Update VSI, etc. AQ commands. It is expected that
3174         * these queue set handles be associated with the Tx
3175         * queues by the driver as part of the TX queue context
3176         * initialization. This has to be done regardless of
3177         * DCB as by default everything is mapped to TC0.
3178         */
3179
3180        if (ring->ch)
3181                tx_ctx.rdylist =
3182                        le16_to_cpu(ring->ch->info.qs_handle[ring->dcb_tc]);
3183
3184        else
3185                tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
3186
3187        tx_ctx.rdylist_act = 0;
3188
3189        /* clear the context in the HMC */
3190        err = i40e_clear_lan_tx_queue_context(hw, pf_q);
3191        if (err) {
3192                dev_info(&vsi->back->pdev->dev,
3193                         "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
3194                         ring->queue_index, pf_q, err);
3195                return -ENOMEM;
3196        }
3197
3198        /* set the context in the HMC */
3199        err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
3200        if (err) {
3201                dev_info(&vsi->back->pdev->dev,
3202                         "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
3203                         ring->queue_index, pf_q, err);
3204                return -ENOMEM;
3205        }
3206
3207        /* Now associate this queue with this PCI function */
3208        if (ring->ch) {
3209                if (ring->ch->type == I40E_VSI_VMDQ2)
3210                        qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3211                else
3212                        return -EINVAL;
3213
3214                qtx_ctl |= (ring->ch->vsi_number <<
3215                            I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3216                            I40E_QTX_CTL_VFVM_INDX_MASK;
3217        } else {
3218                if (vsi->type == I40E_VSI_VMDQ2) {
3219                        qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
3220                        qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
3221                                    I40E_QTX_CTL_VFVM_INDX_MASK;
3222                } else {
3223                        qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
3224                }
3225        }
3226
3227        qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
3228                    I40E_QTX_CTL_PF_INDX_MASK);
3229        wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
3230        i40e_flush(hw);
3231
3232        /* cache tail off for easier writes later */
3233        ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
3234
3235        return 0;
3236}
3237
3238/**
3239 * i40e_configure_rx_ring - Configure a receive ring context
3240 * @ring: The Rx ring to configure
3241 *
3242 * Configure the Rx descriptor ring in the HMC context.
3243 **/
3244static int i40e_configure_rx_ring(struct i40e_ring *ring)
3245{
3246        struct i40e_vsi *vsi = ring->vsi;
3247        u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
3248        u16 pf_q = vsi->base_queue + ring->queue_index;
3249        struct i40e_hw *hw = &vsi->back->hw;
3250        struct i40e_hmc_obj_rxq rx_ctx;
3251        i40e_status err = 0;
3252        bool ok;
3253        int ret;
3254
3255        bitmap_zero(ring->state, __I40E_RING_STATE_NBITS);
3256
3257        /* clear the context structure first */
3258        memset(&rx_ctx, 0, sizeof(rx_ctx));
3259
3260        if (ring->vsi->type == I40E_VSI_MAIN)
3261                xdp_rxq_info_unreg_mem_model(&ring->xdp_rxq);
3262
3263        ring->xsk_umem = i40e_xsk_umem(ring);
3264        if (ring->xsk_umem) {
3265                ring->rx_buf_len = ring->xsk_umem->chunk_size_nohr -
3266                                   XDP_PACKET_HEADROOM;
3267                /* For AF_XDP ZC, we disallow packets to span on
3268                 * multiple buffers, thus letting us skip that
3269                 * handling in the fast-path.
3270                 */
3271                chain_len = 1;
3272                ring->zca.free = i40e_zca_free;
3273                ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
3274                                                 MEM_TYPE_ZERO_COPY,
3275                                                 &ring->zca);
3276                if (ret)
3277                        return ret;
3278                dev_info(&vsi->back->pdev->dev,
3279                         "Registered XDP mem model MEM_TYPE_ZERO_COPY on Rx ring %d\n",
3280                         ring->queue_index);
3281
3282        } else {
3283                ring->rx_buf_len = vsi->rx_buf_len;
3284                if (ring->vsi->type == I40E_VSI_MAIN) {
3285                        ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
3286                                                         MEM_TYPE_PAGE_SHARED,
3287                                                         NULL);
3288                        if (ret)
3289                                return ret;
3290                }
3291        }
3292
3293        rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3294                                    BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3295
3296        rx_ctx.base = (ring->dma / 128);
3297        rx_ctx.qlen = ring->count;
3298
3299        /* use 32 byte descriptors */
3300        rx_ctx.dsize = 1;
3301
3302        /* descriptor type is always zero
3303         * rx_ctx.dtype = 0;
3304         */
3305        rx_ctx.hsplit_0 = 0;
3306
3307        rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3308        if (hw->revision_id == 0)
3309                rx_ctx.lrxqthresh = 0;
3310        else
3311                rx_ctx.lrxqthresh = 1;
3312        rx_ctx.crcstrip = 1;
3313        rx_ctx.l2tsel = 1;
3314        /* this controls whether VLAN is stripped from inner headers */
3315        rx_ctx.showiv = 0;
3316        /* set the prefena field to 1 because the manual says to */
3317        rx_ctx.prefena = 1;
3318
3319        /* clear the context in the HMC */
3320        err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3321        if (err) {
3322                dev_info(&vsi->back->pdev->dev,
3323                         "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3324                         ring->queue_index, pf_q, err);
3325                return -ENOMEM;
3326        }
3327
3328        /* set the context in the HMC */
3329        err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3330        if (err) {
3331                dev_info(&vsi->back->pdev->dev,
3332                         "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3333                         ring->queue_index, pf_q, err);
3334                return -ENOMEM;
3335        }
3336
3337        /* configure Rx buffer alignment */
3338        if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3339                clear_ring_build_skb_enabled(ring);
3340        else
3341                set_ring_build_skb_enabled(ring);
3342
3343        /* cache tail for quicker writes, and clear the reg before use */
3344        ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3345        writel(0, ring->tail);
3346
3347        ok = ring->xsk_umem ?
3348             i40e_alloc_rx_buffers_zc(ring, I40E_DESC_UNUSED(ring)) :
3349             !i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3350        if (!ok) {
3351                /* Log this in case the user has forgotten to give the kernel
3352                 * any buffers, even later in the application.
3353                 */
3354                dev_info(&vsi->back->pdev->dev,
3355                         "Failed to allocate some buffers on %sRx ring %d (pf_q %d)\n",
3356                         ring->xsk_umem ? "UMEM enabled " : "",
3357                         ring->queue_index, pf_q);
3358        }
3359
3360        return 0;
3361}
3362
3363/**
3364 * i40e_vsi_configure_tx - Configure the VSI for Tx
3365 * @vsi: VSI structure describing this set of rings and resources
3366 *
3367 * Configure the Tx VSI for operation.
3368 **/
3369static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3370{
3371        int err = 0;
3372        u16 i;
3373
3374        for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3375                err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3376
3377        if (err || !i40e_enabled_xdp_vsi(vsi))
3378                return err;
3379
3380        for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3381                err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3382
3383        return err;
3384}
3385
3386/**
3387 * i40e_vsi_configure_rx - Configure the VSI for Rx
3388 * @vsi: the VSI being configured
3389 *
3390 * Configure the Rx VSI for operation.
3391 **/
3392static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3393{
3394        int err = 0;
3395        u16 i;
3396
3397        if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3398                vsi->max_frame = I40E_MAX_RXBUFFER;
3399                vsi->rx_buf_len = I40E_RXBUFFER_2048;
3400#if (PAGE_SIZE < 8192)
3401        } else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3402                   (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3403                vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3404                vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3405#endif
3406        } else {
3407                vsi->max_frame = I40E_MAX_RXBUFFER;
3408                vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3409                                                       I40E_RXBUFFER_2048;
3410        }
3411
3412        /* set up individual rings */
3413        for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3414                err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3415
3416        return err;
3417}
3418
3419/**
3420 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3421 * @vsi: ptr to the VSI
3422 **/
3423static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3424{
3425        struct i40e_ring *tx_ring, *rx_ring;
3426        u16 qoffset, qcount;
3427        int i, n;
3428
3429        if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3430                /* Reset the TC information */
3431                for (i = 0; i < vsi->num_queue_pairs; i++) {
3432                        rx_ring = vsi->rx_rings[i];
3433                        tx_ring = vsi->tx_rings[i];
3434                        rx_ring->dcb_tc = 0;
3435                        tx_ring->dcb_tc = 0;
3436                }
3437                return;
3438        }
3439
3440        for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3441                if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3442                        continue;
3443
3444                qoffset = vsi->tc_config.tc_info[n].qoffset;
3445                qcount = vsi->tc_config.tc_info[n].qcount;
3446                for (i = qoffset; i < (qoffset + qcount); i++) {
3447                        rx_ring = vsi->rx_rings[i];
3448                        tx_ring = vsi->tx_rings[i];
3449                        rx_ring->dcb_tc = n;
3450                        tx_ring->dcb_tc = n;
3451                }
3452        }
3453}
3454
3455/**
3456 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3457 * @vsi: ptr to the VSI
3458 **/
3459static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3460{
3461        if (vsi->netdev)
3462                i40e_set_rx_mode(vsi->netdev);
3463}
3464
3465/**
3466 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3467 * @vsi: Pointer to the targeted VSI
3468 *
3469 * This function replays the hlist on the hw where all the SB Flow Director
3470 * filters were saved.
3471 **/
3472static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3473{
3474        struct i40e_fdir_filter *filter;
3475        struct i40e_pf *pf = vsi->back;
3476        struct hlist_node *node;
3477
3478        if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3479                return;
3480
3481        /* Reset FDir counters as we're replaying all existing filters */
3482        pf->fd_tcp4_filter_cnt = 0;
3483        pf->fd_udp4_filter_cnt = 0;
3484        pf->fd_sctp4_filter_cnt = 0;
3485        pf->fd_ip4_filter_cnt = 0;
3486
3487        hlist_for_each_entry_safe(filter, node,
3488                                  &pf->fdir_filter_list, fdir_node) {
3489                i40e_add_del_fdir(vsi, filter, true);
3490        }
3491}
3492
3493/**
3494 * i40e_vsi_configure - Set up the VSI for action
3495 * @vsi: the VSI being configured
3496 **/
3497static int i40e_vsi_configure(struct i40e_vsi *vsi)
3498{
3499        int err;
3500
3501        i40e_set_vsi_rx_mode(vsi);
3502        i40e_restore_vlan(vsi);
3503        i40e_vsi_config_dcb_rings(vsi);
3504        err = i40e_vsi_configure_tx(vsi);
3505        if (!err)
3506                err = i40e_vsi_configure_rx(vsi);
3507
3508        return err;
3509}
3510
3511/**
3512 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3513 * @vsi: the VSI being configured
3514 **/
3515static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3516{
3517        bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3518        struct i40e_pf *pf = vsi->back;
3519        struct i40e_hw *hw = &pf->hw;
3520        u16 vector;
3521        int i, q;
3522        u32 qp;
3523
3524        /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3525         * and PFINT_LNKLSTn registers, e.g.:
3526         *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3527         */
3528        qp = vsi->base_queue;
3529        vector = vsi->base_vector;
3530        for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3531                struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3532
3533                q_vector->rx.next_update = jiffies + 1;
3534                q_vector->rx.target_itr =
3535                        ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
3536                wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3537                     q_vector->rx.target_itr);
3538                q_vector->rx.current_itr = q_vector->rx.target_itr;
3539
3540                q_vector->tx.next_update = jiffies + 1;
3541                q_vector->tx.target_itr =
3542                        ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
3543                wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3544                     q_vector->tx.target_itr);
3545                q_vector->tx.current_itr = q_vector->tx.target_itr;
3546
3547                wr32(hw, I40E_PFINT_RATEN(vector - 1),
3548                     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3549
3550                /* Linked list for the queuepairs assigned to this vector */
3551                wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3552                for (q = 0; q < q_vector->num_ringpairs; q++) {
3553                        u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3554                        u32 val;
3555
3556                        val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3557                              (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3558                              (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3559                              (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3560                              (I40E_QUEUE_TYPE_TX <<
3561                               I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3562
3563                        wr32(hw, I40E_QINT_RQCTL(qp), val);
3564
3565                        if (has_xdp) {
3566                                val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3567                                      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3568                                      (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3569                                      (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3570                                      (I40E_QUEUE_TYPE_TX <<
3571                                       I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3572
3573                                wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3574                        }
3575
3576                        val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3577                              (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3578                              (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3579                              ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3580                              (I40E_QUEUE_TYPE_RX <<
3581                               I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3582
3583                        /* Terminate the linked list */
3584                        if (q == (q_vector->num_ringpairs - 1))
3585                                val |= (I40E_QUEUE_END_OF_LIST <<
3586                                        I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3587
3588                        wr32(hw, I40E_QINT_TQCTL(qp), val);
3589                        qp++;
3590                }
3591        }
3592
3593        i40e_flush(hw);
3594}
3595
3596/**
3597 * i40e_enable_misc_int_causes - enable the non-queue interrupts
3598 * @pf: pointer to private device data structure
3599 **/
3600static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3601{
3602        struct i40e_hw *hw = &pf->hw;
3603        u32 val;
3604
3605        /* clear things first */
3606        wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3607        rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3608
3609        val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3610              I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3611              I40E_PFINT_ICR0_ENA_GRST_MASK          |
3612              I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3613              I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3614              I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3615              I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3616              I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3617
3618        if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3619                val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3620
3621        if (pf->flags & I40E_FLAG_PTP)
3622                val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3623
3624        wr32(hw, I40E_PFINT_ICR0_ENA, val);
3625
3626        /* SW_ITR_IDX = 0, but don't change INTENA */
3627        wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3628                                        I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3629
3630        /* OTHER_ITR_IDX = 0 */
3631        wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3632}
3633
3634/**
3635 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3636 * @vsi: the VSI being configured
3637 **/
3638static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3639{
3640        u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3641        struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3642        struct i40e_pf *pf = vsi->back;
3643        struct i40e_hw *hw = &pf->hw;
3644        u32 val;
3645
3646        /* set the ITR configuration */
3647        q_vector->rx.next_update = jiffies + 1;
3648        q_vector->rx.target_itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
3649        wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.target_itr);
3650        q_vector->rx.current_itr = q_vector->rx.target_itr;
3651        q_vector->tx.next_update = jiffies + 1;
3652        q_vector->tx.target_itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
3653        wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.target_itr);
3654        q_vector->tx.current_itr = q_vector->tx.target_itr;
3655
3656        i40e_enable_misc_int_causes(pf);
3657
3658        /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3659        wr32(hw, I40E_PFINT_LNKLST0, 0);
3660
3661        /* Associate the queue pair to the vector and enable the queue int */
3662        val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                   |
3663              (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3664              (nextqp      << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3665              (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3666
3667        wr32(hw, I40E_QINT_RQCTL(0), val);
3668
3669        if (i40e_enabled_xdp_vsi(vsi)) {
3670                val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                 |
3671                      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3672                      (I40E_QUEUE_TYPE_TX
3673                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3674
3675                wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3676        }
3677
3678        val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3679              (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3680              (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3681
3682        wr32(hw, I40E_QINT_TQCTL(0), val);
3683        i40e_flush(hw);
3684}
3685
3686/**
3687 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3688 * @pf: board private structure
3689 **/
3690void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3691{
3692        struct i40e_hw *hw = &pf->hw;
3693
3694        wr32(hw, I40E_PFINT_DYN_CTL0,
3695             I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3696        i40e_flush(hw);
3697}
3698
3699/**
3700 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3701 * @pf: board private structure
3702 **/
3703void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3704{
3705        struct i40e_hw *hw = &pf->hw;
3706        u32 val;
3707
3708        val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3709              I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3710              (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3711
3712        wr32(hw, I40E_PFINT_DYN_CTL0, val);
3713        i40e_flush(hw);
3714}
3715
3716/**
3717 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3718 * @irq: interrupt number
3719 * @data: pointer to a q_vector
3720 **/
3721static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3722{
3723        struct i40e_q_vector *q_vector = data;
3724
3725        if (!q_vector->tx.ring && !q_vector->rx.ring)
3726                return IRQ_HANDLED;
3727
3728        napi_schedule_irqoff(&q_vector->napi);
3729
3730        return IRQ_HANDLED;
3731}
3732
3733/**
3734 * i40e_irq_affinity_notify - Callback for affinity changes
3735 * @notify: context as to what irq was changed
3736 * @mask: the new affinity mask
3737 *
3738 * This is a callback function used by the irq_set_affinity_notifier function
3739 * so that we may register to receive changes to the irq affinity masks.
3740 **/
3741static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3742                                     const cpumask_t *mask)
3743{
3744        struct i40e_q_vector *q_vector =
3745                container_of(notify, struct i40e_q_vector, affinity_notify);
3746
3747        cpumask_copy(&q_vector->affinity_mask, mask);
3748}
3749
3750/**
3751 * i40e_irq_affinity_release - Callback for affinity notifier release
3752 * @ref: internal core kernel usage
3753 *
3754 * This is a callback function used by the irq_set_affinity_notifier function
3755 * to inform the current notification subscriber that they will no longer
3756 * receive notifications.
3757 **/
3758static void i40e_irq_affinity_release(struct kref *ref) {}
3759
3760/**
3761 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3762 * @vsi: the VSI being configured
3763 * @basename: name for the vector
3764 *
3765 * Allocates MSI-X vectors and requests interrupts from the kernel.
3766 **/
3767static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3768{
3769        int q_vectors = vsi->num_q_vectors;
3770        struct i40e_pf *pf = vsi->back;
3771        int base = vsi->base_vector;
3772        int rx_int_idx = 0;
3773        int tx_int_idx = 0;
3774        int vector, err;
3775        int irq_num;
3776        int cpu;
3777
3778        for (vector = 0; vector < q_vectors; vector++) {
3779                struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3780
3781                irq_num = pf->msix_entries[base + vector].vector;
3782
3783                if (q_vector->tx.ring && q_vector->rx.ring) {
3784                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3785                                 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3786                        tx_int_idx++;
3787                } else if (q_vector->rx.ring) {
3788                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3789                                 "%s-%s-%d", basename, "rx", rx_int_idx++);
3790                } else if (q_vector->tx.ring) {
3791                        snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3792                                 "%s-%s-%d", basename, "tx", tx_int_idx++);
3793                } else {
3794                        /* skip this unused q_vector */
3795                        continue;
3796                }
3797                err = request_irq(irq_num,
3798                                  vsi->irq_handler,
3799                                  0,
3800                                  q_vector->name,
3801                                  q_vector);
3802                if (err) {
3803                        dev_info(&pf->pdev->dev,
3804                                 "MSIX request_irq failed, error: %d\n", err);
3805                        goto free_queue_irqs;
3806                }
3807
3808                /* register for affinity change notifications */
3809                q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3810                q_vector->affinity_notify.release = i40e_irq_affinity_release;
3811                irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3812                /* Spread affinity hints out across online CPUs.
3813                 *
3814                 * get_cpu_mask returns a static constant mask with
3815                 * a permanent lifetime so it's ok to pass to
3816                 * irq_set_affinity_hint without making a copy.
3817                 */
3818                cpu = cpumask_local_spread(q_vector->v_idx, -1);
3819                irq_set_affinity_hint(irq_num, get_cpu_mask(cpu));
3820        }
3821
3822        vsi->irqs_ready = true;
3823        return 0;
3824
3825free_queue_irqs:
3826        while (vector) {
3827                vector--;
3828                irq_num = pf->msix_entries[base + vector].vector;
3829                irq_set_affinity_notifier(irq_num, NULL);
3830                irq_set_affinity_hint(irq_num, NULL);
3831                free_irq(irq_num, &vsi->q_vectors[vector]);
3832        }
3833        return err;
3834}
3835
3836/**
3837 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3838 * @vsi: the VSI being un-configured
3839 **/
3840static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3841{
3842        struct i40e_pf *pf = vsi->back;
3843        struct i40e_hw *hw = &pf->hw;
3844        int base = vsi->base_vector;
3845        int i;
3846
3847        /* disable interrupt causation from each queue */
3848        for (i = 0; i < vsi->num_queue_pairs; i++) {
3849                u32 val;
3850
3851                val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3852                val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3853                wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3854
3855                val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3856                val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3857                wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3858
3859                if (!i40e_enabled_xdp_vsi(vsi))
3860                        continue;
3861                wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3862        }
3863
3864        /* disable each interrupt */
3865        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3866                for (i = vsi->base_vector;
3867                     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3868                        wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3869
3870                i40e_flush(hw);
3871                for (i = 0; i < vsi->num_q_vectors; i++)
3872                        synchronize_irq(pf->msix_entries[i + base].vector);
3873        } else {
3874                /* Legacy and MSI mode - this stops all interrupt handling */
3875                wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3876                wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3877                i40e_flush(hw);
3878                synchronize_irq(pf->pdev->irq);
3879        }
3880}
3881
3882/**
3883 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3884 * @vsi: the VSI being configured
3885 **/
3886static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3887{
3888        struct i40e_pf *pf = vsi->back;
3889        int i;
3890
3891        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3892                for (i = 0; i < vsi->num_q_vectors; i++)
3893                        i40e_irq_dynamic_enable(vsi, i);
3894        } else {
3895                i40e_irq_dynamic_enable_icr0(pf);
3896        }
3897
3898        i40e_flush(&pf->hw);
3899        return 0;
3900}
3901
3902/**
3903 * i40e_free_misc_vector - Free the vector that handles non-queue events
3904 * @pf: board private structure
3905 **/
3906static void i40e_free_misc_vector(struct i40e_pf *pf)
3907{
3908        /* Disable ICR 0 */
3909        wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3910        i40e_flush(&pf->hw);
3911
3912        if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
3913                synchronize_irq(pf->msix_entries[0].vector);
3914                free_irq(pf->msix_entries[0].vector, pf);
3915                clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
3916        }
3917}
3918
3919/**
3920 * i40e_intr - MSI/Legacy and non-queue interrupt handler
3921 * @irq: interrupt number
3922 * @data: pointer to a q_vector
3923 *
3924 * This is the handler used for all MSI/Legacy interrupts, and deals
3925 * with both queue and non-queue interrupts.  This is also used in
3926 * MSIX mode to handle the non-queue interrupts.
3927 **/
3928static irqreturn_t i40e_intr(int irq, void *data)
3929{
3930        struct i40e_pf *pf = (struct i40e_pf *)data;
3931        struct i40e_hw *hw = &pf->hw;
3932        irqreturn_t ret = IRQ_NONE;
3933        u32 icr0, icr0_remaining;
3934        u32 val, ena_mask;
3935
3936        icr0 = rd32(hw, I40E_PFINT_ICR0);
3937        ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3938
3939        /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3940        if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3941                goto enable_intr;
3942
3943        /* if interrupt but no bits showing, must be SWINT */
3944        if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3945            (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3946                pf->sw_int_count++;
3947
3948        if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3949            (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3950                ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3951                dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3952                set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3953        }
3954
3955        /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3956        if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3957                struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3958                struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3959
3960                /* We do not have a way to disarm Queue causes while leaving
3961                 * interrupt enabled for all other causes, ideally
3962                 * interrupt should be disabled while we are in NAPI but
3963                 * this is not a performance path and napi_schedule()
3964                 * can deal with rescheduling.
3965                 */
3966                if (!test_bit(__I40E_DOWN, pf->state))
3967                        napi_schedule_irqoff(&q_vector->napi);
3968        }
3969
3970        if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3971                ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3972                set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3973                i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3974        }
3975
3976        if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3977                ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3978                set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3979        }
3980
3981        if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3982                ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3983                set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3984        }
3985
3986        if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3987                if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3988                        set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3989                ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3990                val = rd32(hw, I40E_GLGEN_RSTAT);
3991                val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3992                       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3993                if (val == I40E_RESET_CORER) {
3994                        pf->corer_count++;
3995                } else if (val == I40E_RESET_GLOBR) {
3996                        pf->globr_count++;
3997                } else if (val == I40E_RESET_EMPR) {
3998                        pf->empr_count++;
3999                        set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
4000                }
4001        }
4002
4003        if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
4004                icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
4005                dev_info(&pf->pdev->dev, "HMC error interrupt\n");
4006                dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
4007                         rd32(hw, I40E_PFHMC_ERRORINFO),
4008                         rd32(hw, I40E_PFHMC_ERRORDATA));
4009        }
4010
4011        if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
4012                u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
4013
4014                if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
4015                        icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
4016                        i40e_ptp_tx_hwtstamp(pf);
4017                }
4018        }
4019
4020        /* If a critical error is pending we have no choice but to reset the
4021         * device.
4022         * Report and mask out any remaining unexpected interrupts.
4023         */
4024        icr0_remaining = icr0 & ena_mask;
4025        if (icr0_remaining) {
4026                dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
4027                         icr0_remaining);
4028                if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
4029                    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
4030                    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
4031                        dev_info(&pf->pdev->dev, "device will be reset\n");
4032                        set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
4033                        i40e_service_event_schedule(pf);
4034                }
4035                ena_mask &= ~icr0_remaining;
4036        }
4037        ret = IRQ_HANDLED;
4038
4039enable_intr:
4040        /* re-enable interrupt causes */
4041        wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
4042        if (!test_bit(__I40E_DOWN, pf->state) ||
4043            test_bit(__I40E_RECOVERY_MODE, pf->state)) {
4044                i40e_service_event_schedule(pf);
4045                i40e_irq_dynamic_enable_icr0(pf);
4046        }
4047
4048        return ret;
4049}
4050
4051/**
4052 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
4053 * @tx_ring:  tx ring to clean
4054 * @budget:   how many cleans we're allowed
4055 *
4056 * Returns true if there's any budget left (e.g. the clean is finished)
4057 **/
4058static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
4059{
4060        struct i40e_vsi *vsi = tx_ring->vsi;
4061        u16 i = tx_ring->next_to_clean;
4062        struct i40e_tx_buffer *tx_buf;
4063        struct i40e_tx_desc *tx_desc;
4064
4065        tx_buf = &tx_ring->tx_bi[i];
4066        tx_desc = I40E_TX_DESC(tx_ring, i);
4067        i -= tx_ring->count;
4068
4069        do {
4070                struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
4071
4072                /* if next_to_watch is not set then there is no work pending */
4073                if (!eop_desc)
4074                        break;
4075
4076                /* prevent any other reads prior to eop_desc */
4077                smp_rmb();
4078
4079                /* if the descriptor isn't done, no work yet to do */
4080                if (!(eop_desc->cmd_type_offset_bsz &
4081                      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
4082                        break;
4083
4084                /* clear next_to_watch to prevent false hangs */
4085                tx_buf->next_to_watch = NULL;
4086
4087                tx_desc->buffer_addr = 0;
4088                tx_desc->cmd_type_offset_bsz = 0;
4089                /* move past filter desc */
4090                tx_buf++;
4091                tx_desc++;
4092                i++;
4093                if (unlikely(!i)) {
4094                        i -= tx_ring->count;
4095                        tx_buf = tx_ring->tx_bi;
4096                        tx_desc = I40E_TX_DESC(tx_ring, 0);
4097                }
4098                /* unmap skb header data */
4099                dma_unmap_single(tx_ring->dev,
4100                                 dma_unmap_addr(tx_buf, dma),
4101                                 dma_unmap_len(tx_buf, len),
4102                                 DMA_TO_DEVICE);
4103                if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
4104                        kfree(tx_buf->raw_buf);
4105
4106                tx_buf->raw_buf = NULL;
4107                tx_buf->tx_flags = 0;
4108                tx_buf->next_to_watch = NULL;
4109                dma_unmap_len_set(tx_buf, len, 0);
4110                tx_desc->buffer_addr = 0;
4111                tx_desc->cmd_type_offset_bsz = 0;
4112
4113                /* move us past the eop_desc for start of next FD desc */
4114                tx_buf++;
4115                tx_desc++;
4116                i++;
4117                if (unlikely(!i)) {
4118                        i -= tx_ring->count;
4119                        tx_buf = tx_ring->tx_bi;
4120                        tx_desc = I40E_TX_DESC(tx_ring, 0);
4121                }
4122
4123                /* update budget accounting */
4124                budget--;
4125        } while (likely(budget));
4126
4127        i += tx_ring->count;
4128        tx_ring->next_to_clean = i;
4129
4130        if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
4131                i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
4132
4133        return budget > 0;
4134}
4135
4136/**
4137 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
4138 * @irq: interrupt number
4139 * @data: pointer to a q_vector
4140 **/
4141static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
4142{
4143        struct i40e_q_vector *q_vector = data;
4144        struct i40e_vsi *vsi;
4145
4146        if (!q_vector->tx.ring)
4147                return IRQ_HANDLED;
4148
4149        vsi = q_vector->tx.ring->vsi;
4150        i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
4151
4152        return IRQ_HANDLED;
4153}
4154
4155/**
4156 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
4157 * @vsi: the VSI being configured
4158 * @v_idx: vector index
4159 * @qp_idx: queue pair index
4160 **/
4161static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
4162{
4163        struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4164        struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
4165        struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
4166
4167        tx_ring->q_vector = q_vector;
4168        tx_ring->next = q_vector->tx.ring;
4169        q_vector->tx.ring = tx_ring;
4170        q_vector->tx.count++;
4171
4172        /* Place XDP Tx ring in the same q_vector ring list as regular Tx */
4173        if (i40e_enabled_xdp_vsi(vsi)) {
4174                struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
4175
4176                xdp_ring->q_vector = q_vector;
4177                xdp_ring->next = q_vector->tx.ring;
4178                q_vector->tx.ring = xdp_ring;
4179                q_vector->tx.count++;
4180        }
4181
4182        rx_ring->q_vector = q_vector;
4183        rx_ring->next = q_vector->rx.ring;
4184        q_vector->rx.ring = rx_ring;
4185        q_vector->rx.count++;
4186}
4187
4188/**
4189 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
4190 * @vsi: the VSI being configured
4191 *
4192 * This function maps descriptor rings to the queue-specific vectors
4193 * we were allotted through the MSI-X enabling code.  Ideally, we'd have
4194 * one vector per queue pair, but on a constrained vector budget, we
4195 * group the queue pairs as "efficiently" as possible.
4196 **/
4197static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
4198{
4199        int qp_remaining = vsi->num_queue_pairs;
4200        int q_vectors = vsi->num_q_vectors;
4201        int num_ringpairs;
4202        int v_start = 0;
4203        int qp_idx = 0;
4204
4205        /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
4206         * group them so there are multiple queues per vector.
4207         * It is also important to go through all the vectors available to be
4208         * sure that if we don't use all the vectors, that the remaining vectors
4209         * are cleared. This is especially important when decreasing the
4210         * number of queues in use.
4211         */
4212        for (; v_start < q_vectors; v_start++) {
4213                struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
4214
4215                num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
4216
4217                q_vector->num_ringpairs = num_ringpairs;
4218                q_vector->reg_idx = q_vector->v_idx + vsi->base_vector - 1;
4219
4220                q_vector->rx.count = 0;
4221                q_vector->tx.count = 0;
4222                q_vector->rx.ring = NULL;
4223                q_vector->tx.ring = NULL;
4224
4225                while (num_ringpairs--) {
4226                        i40e_map_vector_to_qp(vsi, v_start, qp_idx);
4227                        qp_idx++;
4228                        qp_remaining--;
4229                }
4230        }
4231}
4232
4233/**
4234 * i40e_vsi_request_irq - Request IRQ from the OS
4235 * @vsi: the VSI being configured
4236 * @basename: name for the vector
4237 **/
4238static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
4239{
4240        struct i40e_pf *pf = vsi->back;
4241        int err;
4242
4243        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4244                err = i40e_vsi_request_irq_msix(vsi, basename);
4245        else if (pf->flags & I40E_FLAG_MSI_ENABLED)
4246                err = request_irq(pf->pdev->irq, i40e_intr, 0,
4247                                  pf->int_name, pf);
4248        else
4249                err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
4250                                  pf->int_name, pf);
4251
4252        if (err)
4253                dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
4254
4255        return err;
4256}
4257
4258#ifdef CONFIG_NET_POLL_CONTROLLER
4259/**
4260 * i40e_netpoll - A Polling 'interrupt' handler
4261 * @netdev: network interface device structure
4262 *
4263 * This is used by netconsole to send skbs without having to re-enable
4264 * interrupts.  It's not called while the normal interrupt routine is executing.
4265 **/
4266static void i40e_netpoll(struct net_device *netdev)
4267{
4268        struct i40e_netdev_priv *np = netdev_priv(netdev);
4269        struct i40e_vsi *vsi = np->vsi;
4270        struct i40e_pf *pf = vsi->back;
4271        int i;
4272
4273        /* if interface is down do nothing */
4274        if (test_bit(__I40E_VSI_DOWN, vsi->state))
4275                return;
4276
4277        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4278                for (i = 0; i < vsi->num_q_vectors; i++)
4279                        i40e_msix_clean_rings(0, vsi->q_vectors[i]);
4280        } else {
4281                i40e_intr(pf->pdev->irq, netdev);
4282        }
4283}
4284#endif
4285
4286#define I40E_QTX_ENA_WAIT_COUNT 50
4287
4288/**
4289 * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
4290 * @pf: the PF being configured
4291 * @pf_q: the PF queue
4292 * @enable: enable or disable state of the queue
4293 *
4294 * This routine will wait for the given Tx queue of the PF to reach the
4295 * enabled or disabled state.
4296 * Returns -ETIMEDOUT in case of failing to reach the requested state after
4297 * multiple retries; else will return 0 in case of success.
4298 **/
4299static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4300{
4301        int i;
4302        u32 tx_reg;
4303
4304        for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4305                tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4306                if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4307                        break;
4308
4309                usleep_range(10, 20);
4310        }
4311        if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4312                return -ETIMEDOUT;
4313
4314        return 0;
4315}
4316
4317/**
4318 * i40e_control_tx_q - Start or stop a particular Tx queue
4319 * @pf: the PF structure
4320 * @pf_q: the PF queue to configure
4321 * @enable: start or stop the queue
4322 *
4323 * This function enables or disables a single queue. Note that any delay
4324 * required after the operation is expected to be handled by the caller of
4325 * this function.
4326 **/
4327static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4328{
4329        struct i40e_hw *hw = &pf->hw;
4330        u32 tx_reg;
4331        int i;
4332
4333        /* warn the TX unit of coming changes */
4334        i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4335        if (!enable)
4336                usleep_range(10, 20);
4337
4338        for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4339                tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4340                if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4341                    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4342                        break;
4343                usleep_range(1000, 2000);
4344        }
4345
4346        /* Skip if the queue is already in the requested state */
4347        if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4348                return;
4349
4350        /* turn on/off the queue */
4351        if (enable) {
4352                wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4353                tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4354        } else {
4355                tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4356        }
4357
4358        wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4359}
4360
4361/**
4362 * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4363 * @seid: VSI SEID
4364 * @pf: the PF structure
4365 * @pf_q: the PF queue to configure
4366 * @is_xdp: true if the queue is used for XDP
4367 * @enable: start or stop the queue
4368 **/
4369int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4370                           bool is_xdp, bool enable)
4371{
4372        int ret;
4373
4374        i40e_control_tx_q(pf, pf_q, enable);
4375
4376        /* wait for the change to finish */
4377        ret = i40e_pf_txq_wait(pf, pf_q, enable);
4378        if (ret) {
4379                dev_info(&pf->pdev->dev,
4380                         "VSI seid %d %sTx ring %d %sable timeout\n",
4381                         seid, (is_xdp ? "XDP " : ""), pf_q,
4382                         (enable ? "en" : "dis"));
4383        }
4384
4385        return ret;
4386}
4387
4388/**
4389 * i40e_vsi_control_tx - Start or stop a VSI's rings
4390 * @vsi: the VSI being configured
4391 * @enable: start or stop the rings
4392 **/
4393static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4394{
4395        struct i40e_pf *pf = vsi->back;
4396        int i, pf_q, ret = 0;
4397
4398        pf_q = vsi->base_queue;
4399        for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4400                ret = i40e_control_wait_tx_q(vsi->seid, pf,
4401                                             pf_q,
4402                                             false /*is xdp*/, enable);
4403                if (ret)
4404                        break;
4405
4406                if (!i40e_enabled_xdp_vsi(vsi))
4407                        continue;
4408
4409                ret = i40e_control_wait_tx_q(vsi->seid, pf,
4410                                             pf_q + vsi->alloc_queue_pairs,
4411                                             true /*is xdp*/, enable);
4412                if (ret)
4413                        break;
4414        }
4415        return ret;
4416}
4417
4418/**
4419 * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4420 * @pf: the PF being configured
4421 * @pf_q: the PF queue
4422 * @enable: enable or disable state of the queue
4423 *
4424 * This routine will wait for the given Rx queue of the PF to reach the
4425 * enabled or disabled state.
4426 * Returns -ETIMEDOUT in case of failing to reach the requested state after
4427 * multiple retries; else will return 0 in case of success.
4428 **/
4429static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4430{
4431        int i;
4432        u32 rx_reg;
4433
4434        for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4435                rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4436                if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4437                        break;
4438
4439                usleep_range(10, 20);
4440        }
4441        if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4442                return -ETIMEDOUT;
4443
4444        return 0;
4445}
4446
4447/**
4448 * i40e_control_rx_q - Start or stop a particular Rx queue
4449 * @pf: the PF structure
4450 * @pf_q: the PF queue to configure
4451 * @enable: start or stop the queue
4452 *
4453 * This function enables or disables a single queue. Note that
4454 * any delay required after the operation is expected to be
4455 * handled by the caller of this function.
4456 **/
4457static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4458{
4459        struct i40e_hw *hw = &pf->hw;
4460        u32 rx_reg;
4461        int i;
4462
4463        for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4464                rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4465                if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4466                    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4467                        break;
4468                usleep_range(1000, 2000);
4469        }
4470
4471        /* Skip if the queue is already in the requested state */
4472        if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4473                return;
4474
4475        /* turn on/off the queue */
4476        if (enable)
4477                rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4478        else
4479                rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4480
4481        wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4482}
4483
4484/**
4485 * i40e_control_wait_rx_q
4486 * @pf: the PF structure
4487 * @pf_q: queue being configured
4488 * @enable: start or stop the rings
4489 *
4490 * This function enables or disables a single queue along with waiting
4491 * for the change to finish. The caller of this function should handle
4492 * the delays needed in the case of disabling queues.
4493 **/
4494int i40e_control_wait_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4495{
4496        int ret = 0;
4497
4498        i40e_control_rx_q(pf, pf_q, enable);
4499
4500        /* wait for the change to finish */
4501        ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4502        if (ret)
4503                return ret;
4504
4505        return ret;
4506}
4507
4508/**
4509 * i40e_vsi_control_rx - Start or stop a VSI's rings
4510 * @vsi: the VSI being configured
4511 * @enable: start or stop the rings
4512 **/
4513static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4514{
4515        struct i40e_pf *pf = vsi->back;
4516        int i, pf_q, ret = 0;
4517
4518        pf_q = vsi->base_queue;
4519        for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4520                ret = i40e_control_wait_rx_q(pf, pf_q, enable);
4521                if (ret) {
4522                        dev_info(&pf->pdev->dev,
4523                                 "VSI seid %d Rx ring %d %sable timeout\n",
4524                                 vsi->seid, pf_q, (enable ? "en" : "dis"));
4525                        break;
4526                }
4527        }
4528
4529        /* Due to HW errata, on Rx disable only, the register can indicate done
4530         * before it really is. Needs 50ms to be sure
4531         */
4532        if (!enable)
4533                mdelay(50);
4534
4535        return ret;
4536}
4537
4538/**
4539 * i40e_vsi_start_rings - Start a VSI's rings
4540 * @vsi: the VSI being configured
4541 **/
4542int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4543{
4544        int ret = 0;
4545
4546        /* do rx first for enable and last for disable */
4547        ret = i40e_vsi_control_rx(vsi, true);
4548        if (ret)
4549                return ret;
4550        ret = i40e_vsi_control_tx(vsi, true);
4551
4552        return ret;
4553}
4554
4555/**
4556 * i40e_vsi_stop_rings - Stop a VSI's rings
4557 * @vsi: the VSI being configured
4558 **/
4559void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4560{
4561        /* When port TX is suspended, don't wait */
4562        if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4563                return i40e_vsi_stop_rings_no_wait(vsi);
4564
4565        /* do rx first for enable and last for disable
4566         * Ignore return value, we need to shutdown whatever we can
4567         */
4568        i40e_vsi_control_tx(vsi, false);
4569        i40e_vsi_control_rx(vsi, false);
4570}
4571
4572/**
4573 * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4574 * @vsi: the VSI being shutdown
4575 *
4576 * This function stops all the rings for a VSI but does not delay to verify
4577 * that rings have been disabled. It is expected that the caller is shutting
4578 * down multiple VSIs at once and will delay together for all the VSIs after
4579 * initiating the shutdown. This is particularly useful for shutting down lots
4580 * of VFs together. Otherwise, a large delay can be incurred while configuring
4581 * each VSI in serial.
4582 **/
4583void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4584{
4585        struct i40e_pf *pf = vsi->back;
4586        int i, pf_q;
4587
4588        pf_q = vsi->base_queue;
4589        for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4590                i40e_control_tx_q(pf, pf_q, false);
4591                i40e_control_rx_q(pf, pf_q, false);
4592        }
4593}
4594
4595/**
4596 * i40e_vsi_free_irq - Free the irq association with the OS
4597 * @vsi: the VSI being configured
4598 **/
4599static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4600{
4601        struct i40e_pf *pf = vsi->back;
4602        struct i40e_hw *hw = &pf->hw;
4603        int base = vsi->base_vector;
4604        u32 val, qp;
4605        int i;
4606
4607        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4608                if (!vsi->q_vectors)
4609                        return;
4610
4611                if (!vsi->irqs_ready)
4612                        return;
4613
4614                vsi->irqs_ready = false;
4615                for (i = 0; i < vsi->num_q_vectors; i++) {
4616                        int irq_num;
4617                        u16 vector;
4618
4619                        vector = i + base;
4620                        irq_num = pf->msix_entries[vector].vector;
4621
4622                        /* free only the irqs that were actually requested */
4623                        if (!vsi->q_vectors[i] ||
4624                            !vsi->q_vectors[i]->num_ringpairs)
4625                                continue;
4626
4627                        /* clear the affinity notifier in the IRQ descriptor */
4628                        irq_set_affinity_notifier(irq_num, NULL);
4629                        /* remove our suggested affinity mask for this IRQ */
4630                        irq_set_affinity_hint(irq_num, NULL);
4631                        synchronize_irq(irq_num);
4632                        free_irq(irq_num, vsi->q_vectors[i]);
4633
4634                        /* Tear down the interrupt queue link list
4635                         *
4636                         * We know that they come in pairs and always
4637                         * the Rx first, then the Tx.  To clear the
4638                         * link list, stick the EOL value into the
4639                         * next_q field of the registers.
4640                         */
4641                        val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4642                        qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4643                                >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4644                        val |= I40E_QUEUE_END_OF_LIST
4645                                << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4646                        wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4647
4648                        while (qp != I40E_QUEUE_END_OF_LIST) {
4649                                u32 next;
4650
4651                                val = rd32(hw, I40E_QINT_RQCTL(qp));
4652
4653                                val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4654                                         I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4655                                         I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4656                                         I40E_QINT_RQCTL_INTEVENT_MASK);
4657
4658                                val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4659                                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4660
4661                                wr32(hw, I40E_QINT_RQCTL(qp), val);
4662
4663                                val = rd32(hw, I40E_QINT_TQCTL(qp));
4664
4665                                next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4666                                        >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4667
4668                                val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4669                                         I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4670                                         I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4671                                         I40E_QINT_TQCTL_INTEVENT_MASK);
4672
4673                                val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4674                                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4675
4676                                wr32(hw, I40E_QINT_TQCTL(qp), val);
4677                                qp = next;
4678                        }
4679                }
4680        } else {
4681                free_irq(pf->pdev->irq, pf);
4682
4683                val = rd32(hw, I40E_PFINT_LNKLST0);
4684                qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4685                        >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4686                val |= I40E_QUEUE_END_OF_LIST
4687                        << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4688                wr32(hw, I40E_PFINT_LNKLST0, val);
4689
4690                val = rd32(hw, I40E_QINT_RQCTL(qp));
4691                val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4692                         I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4693                         I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4694                         I40E_QINT_RQCTL_INTEVENT_MASK);
4695
4696                val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4697                        I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4698
4699                wr32(hw, I40E_QINT_RQCTL(qp), val);
4700
4701                val = rd32(hw, I40E_QINT_TQCTL(qp));
4702
4703                val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4704                         I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4705                         I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4706                         I40E_QINT_TQCTL_INTEVENT_MASK);
4707
4708                val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4709                        I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4710
4711                wr32(hw, I40E_QINT_TQCTL(qp), val);
4712        }
4713}
4714
4715/**
4716 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4717 * @vsi: the VSI being configured
4718 * @v_idx: Index of vector to be freed
4719 *
4720 * This function frees the memory allocated to the q_vector.  In addition if
4721 * NAPI is enabled it will delete any references to the NAPI struct prior
4722 * to freeing the q_vector.
4723 **/
4724static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4725{
4726        struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4727        struct i40e_ring *ring;
4728
4729        if (!q_vector)
4730                return;
4731
4732        /* disassociate q_vector from rings */
4733        i40e_for_each_ring(ring, q_vector->tx)
4734                ring->q_vector = NULL;
4735
4736        i40e_for_each_ring(ring, q_vector->rx)
4737                ring->q_vector = NULL;
4738
4739        /* only VSI w/ an associated netdev is set up w/ NAPI */
4740        if (vsi->netdev)
4741                netif_napi_del(&q_vector->napi);
4742
4743        vsi->q_vectors[v_idx] = NULL;
4744
4745        kfree_rcu(q_vector, rcu);
4746}
4747
4748/**
4749 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4750 * @vsi: the VSI being un-configured
4751 *
4752 * This frees the memory allocated to the q_vectors and
4753 * deletes references to the NAPI struct.
4754 **/
4755static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4756{
4757        int v_idx;
4758
4759        for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4760                i40e_free_q_vector(vsi, v_idx);
4761}
4762
4763/**
4764 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4765 * @pf: board private structure
4766 **/
4767static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4768{
4769        /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4770        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4771                pci_disable_msix(pf->pdev);
4772                kfree(pf->msix_entries);
4773                pf->msix_entries = NULL;
4774                kfree(pf->irq_pile);
4775                pf->irq_pile = NULL;
4776        } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4777                pci_disable_msi(pf->pdev);
4778        }
4779        pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4780}
4781
4782/**
4783 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4784 * @pf: board private structure
4785 *
4786 * We go through and clear interrupt specific resources and reset the structure
4787 * to pre-load conditions
4788 **/
4789static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4790{
4791        int i;
4792
4793        i40e_free_misc_vector(pf);
4794
4795        i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4796                      I40E_IWARP_IRQ_PILE_ID);
4797
4798        i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4799        for (i = 0; i < pf->num_alloc_vsi; i++)
4800                if (pf->vsi[i])
4801                        i40e_vsi_free_q_vectors(pf->vsi[i]);
4802        i40e_reset_interrupt_capability(pf);
4803}
4804
4805/**
4806 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4807 * @vsi: the VSI being configured
4808 **/
4809static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4810{
4811        int q_idx;
4812
4813        if (!vsi->netdev)
4814                return;
4815
4816        for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4817                struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4818
4819                if (q_vector->rx.ring || q_vector->tx.ring)
4820                        napi_enable(&q_vector->napi);
4821        }
4822}
4823
4824/**
4825 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4826 * @vsi: the VSI being configured
4827 **/
4828static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4829{
4830        int q_idx;
4831
4832        if (!vsi->netdev)
4833                return;
4834
4835        for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4836                struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4837
4838                if (q_vector->rx.ring || q_vector->tx.ring)
4839                        napi_disable(&q_vector->napi);
4840        }
4841}
4842
4843/**
4844 * i40e_vsi_close - Shut down a VSI
4845 * @vsi: the vsi to be quelled
4846 **/
4847static void i40e_vsi_close(struct i40e_vsi *vsi)
4848{
4849        struct i40e_pf *pf = vsi->back;
4850        if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4851                i40e_down(vsi);
4852        i40e_vsi_free_irq(vsi);
4853        i40e_vsi_free_tx_resources(vsi);
4854        i40e_vsi_free_rx_resources(vsi);
4855        vsi->current_netdev_flags = 0;
4856        set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
4857        if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4858                set_bit(__I40E_CLIENT_RESET, pf->state);
4859}
4860
4861/**
4862 * i40e_quiesce_vsi - Pause a given VSI
4863 * @vsi: the VSI being paused
4864 **/
4865static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4866{
4867        if (test_bit(__I40E_VSI_DOWN, vsi->state))
4868                return;
4869
4870        set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4871        if (vsi->netdev && netif_running(vsi->netdev))
4872                vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4873        else
4874                i40e_vsi_close(vsi);
4875}
4876
4877/**
4878 * i40e_unquiesce_vsi - Resume a given VSI
4879 * @vsi: the VSI being resumed
4880 **/
4881static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4882{
4883        if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4884                return;
4885
4886        if (vsi->netdev && netif_running(vsi->netdev))
4887                vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4888        else
4889                i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4890}
4891
4892/**
4893 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4894 * @pf: the PF
4895 **/
4896static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4897{
4898        int v;
4899
4900        for (v = 0; v < pf->num_alloc_vsi; v++) {
4901                if (pf->vsi[v])
4902                        i40e_quiesce_vsi(pf->vsi[v]);
4903        }
4904}
4905
4906/**
4907 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4908 * @pf: the PF
4909 **/
4910static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4911{
4912        int v;
4913
4914        for (v = 0; v < pf->num_alloc_vsi; v++) {
4915                if (pf->vsi[v])
4916                        i40e_unquiesce_vsi(pf->vsi[v]);
4917        }
4918}
4919
4920/**
4921 * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4922 * @vsi: the VSI being configured
4923 *
4924 * Wait until all queues on a given VSI have been disabled.
4925 **/
4926int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4927{
4928        struct i40e_pf *pf = vsi->back;
4929        int i, pf_q, ret;
4930
4931        pf_q = vsi->base_queue;
4932        for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4933                /* Check and wait for the Tx queue */
4934                ret = i40e_pf_txq_wait(pf, pf_q, false);
4935                if (ret) {
4936                        dev_info(&pf->pdev->dev,
4937                                 "VSI seid %d Tx ring %d disable timeout\n",
4938                                 vsi->seid, pf_q);
4939                        return ret;
4940                }
4941
4942                if (!i40e_enabled_xdp_vsi(vsi))
4943                        goto wait_rx;
4944
4945                /* Check and wait for the XDP Tx queue */
4946                ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4947                                       false);
4948                if (ret) {
4949                        dev_info(&pf->pdev->dev,
4950                                 "VSI seid %d XDP Tx ring %d disable timeout\n",
4951                                 vsi->seid, pf_q);
4952                        return ret;
4953                }
4954wait_rx:
4955                /* Check and wait for the Rx queue */
4956                ret = i40e_pf_rxq_wait(pf, pf_q, false);
4957                if (ret) {
4958                        dev_info(&pf->pdev->dev,
4959                                 "VSI seid %d Rx ring %d disable timeout\n",
4960                                 vsi->seid, pf_q);
4961                        return ret;
4962                }
4963        }
4964
4965        return 0;
4966}
4967
4968#ifdef CONFIG_I40E_DCB
4969/**
4970 * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4971 * @pf: the PF
4972 *
4973 * This function waits for the queues to be in disabled state for all the
4974 * VSIs that are managed by this PF.
4975 **/
4976static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4977{
4978        int v, ret = 0;
4979
4980        for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4981                if (pf->vsi[v]) {
4982                        ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4983                        if (ret)
4984                                break;
4985                }
4986        }
4987
4988        return ret;
4989}
4990
4991#endif
4992
4993/**
4994 * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4995 * @pf: pointer to PF
4996 *
4997 * Get TC map for ISCSI PF type that will include iSCSI TC
4998 * and LAN TC.
4999 **/
5000static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
5001{
5002        struct i40e_dcb_app_priority_table app;
5003        struct i40e_hw *hw = &pf->hw;
5004        u8 enabled_tc = 1; /* TC0 is always enabled */
5005        u8 tc, i;
5006        /* Get the iSCSI APP TLV */
5007        struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5008
5009        for (i = 0; i < dcbcfg->numapps; i++) {
5010                app = dcbcfg->app[i];
5011                if (app.selector == I40E_APP_SEL_TCPIP &&
5012                    app.protocolid == I40E_APP_PROTOID_ISCSI) {
5013                        tc = dcbcfg->etscfg.prioritytable[app.priority];
5014                        enabled_tc |= BIT(tc);
5015                        break;
5016                }
5017        }
5018
5019        return enabled_tc;
5020}
5021
5022/**
5023 * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
5024 * @dcbcfg: the corresponding DCBx configuration structure
5025 *
5026 * Return the number of TCs from given DCBx configuration
5027 **/
5028static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
5029{
5030        int i, tc_unused = 0;
5031        u8 num_tc = 0;
5032        u8 ret = 0;
5033
5034        /* Scan the ETS Config Priority Table to find
5035         * traffic class enabled for a given priority
5036         * and create a bitmask of enabled TCs
5037         */
5038        for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
5039                num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
5040
5041        /* Now scan the bitmask to check for
5042         * contiguous TCs starting with TC0
5043         */
5044        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5045                if (num_tc & BIT(i)) {
5046                        if (!tc_unused) {
5047                                ret++;
5048                        } else {
5049                                pr_err("Non-contiguous TC - Disabling DCB\n");
5050                                return 1;
5051                        }
5052                } else {
5053                        tc_unused = 1;
5054                }
5055        }
5056
5057        /* There is always at least TC0 */
5058        if (!ret)
5059                ret = 1;
5060
5061        return ret;
5062}
5063
5064/**
5065 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
5066 * @dcbcfg: the corresponding DCBx configuration structure
5067 *
5068 * Query the current DCB configuration and return the number of
5069 * traffic classes enabled from the given DCBX config
5070 **/
5071static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
5072{
5073        u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
5074        u8 enabled_tc = 1;
5075        u8 i;
5076
5077        for (i = 0; i < num_tc; i++)
5078                enabled_tc |= BIT(i);
5079
5080        return enabled_tc;
5081}
5082
5083/**
5084 * i40e_mqprio_get_enabled_tc - Get enabled traffic classes
5085 * @pf: PF being queried
5086 *
5087 * Query the current MQPRIO configuration and return the number of
5088 * traffic classes enabled.
5089 **/
5090static u8 i40e_mqprio_get_enabled_tc(struct i40e_pf *pf)
5091{
5092        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5093        u8 num_tc = vsi->mqprio_qopt.qopt.num_tc;
5094        u8 enabled_tc = 1, i;
5095
5096        for (i = 1; i < num_tc; i++)
5097                enabled_tc |= BIT(i);
5098        return enabled_tc;
5099}
5100
5101/**
5102 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
5103 * @pf: PF being queried
5104 *
5105 * Return number of traffic classes enabled for the given PF
5106 **/
5107static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
5108{
5109        struct i40e_hw *hw = &pf->hw;
5110        u8 i, enabled_tc = 1;
5111        u8 num_tc = 0;
5112        struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5113
5114        if (pf->flags & I40E_FLAG_TC_MQPRIO)
5115                return pf->vsi[pf->lan_vsi]->mqprio_qopt.qopt.num_tc;
5116
5117        /* If neither MQPRIO nor DCB is enabled, then always use single TC */
5118        if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5119                return 1;
5120
5121        /* SFP mode will be enabled for all TCs on port */
5122        if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5123                return i40e_dcb_get_num_tc(dcbcfg);
5124
5125        /* MFP mode return count of enabled TCs for this PF */
5126        if (pf->hw.func_caps.iscsi)
5127                enabled_tc =  i40e_get_iscsi_tc_map(pf);
5128        else
5129                return 1; /* Only TC0 */
5130
5131        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5132                if (enabled_tc & BIT(i))
5133                        num_tc++;
5134        }
5135        return num_tc;
5136}
5137
5138/**
5139 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
5140 * @pf: PF being queried
5141 *
5142 * Return a bitmap for enabled traffic classes for this PF.
5143 **/
5144static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
5145{
5146        if (pf->flags & I40E_FLAG_TC_MQPRIO)
5147                return i40e_mqprio_get_enabled_tc(pf);
5148
5149        /* If neither MQPRIO nor DCB is enabled for this PF then just return
5150         * default TC
5151         */
5152        if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5153                return I40E_DEFAULT_TRAFFIC_CLASS;
5154
5155        /* SFP mode we want PF to be enabled for all TCs */
5156        if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5157                return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
5158
5159        /* MFP enabled and iSCSI PF type */
5160        if (pf->hw.func_caps.iscsi)
5161                return i40e_get_iscsi_tc_map(pf);
5162        else
5163                return I40E_DEFAULT_TRAFFIC_CLASS;
5164}
5165
5166/**
5167 * i40e_vsi_get_bw_info - Query VSI BW Information
5168 * @vsi: the VSI being queried
5169 *
5170 * Returns 0 on success, negative value on failure
5171 **/
5172static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
5173{
5174        struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
5175        struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5176        struct i40e_pf *pf = vsi->back;
5177        struct i40e_hw *hw = &pf->hw;
5178        i40e_status ret;
5179        u32 tc_bw_max;
5180        int i;
5181
5182        /* Get the VSI level BW configuration */
5183        ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
5184        if (ret) {
5185                dev_info(&pf->pdev->dev,
5186                         "couldn't get PF vsi bw config, err %s aq_err %s\n",
5187                         i40e_stat_str(&pf->hw, ret),
5188                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5189                return -EINVAL;
5190        }
5191
5192        /* Get the VSI level BW configuration per TC */
5193        ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
5194                                               NULL);
5195        if (ret) {
5196                dev_info(&pf->pdev->dev,
5197                         "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
5198                         i40e_stat_str(&pf->hw, ret),
5199                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5200                return -EINVAL;
5201        }
5202
5203        if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
5204                dev_info(&pf->pdev->dev,
5205                         "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
5206                         bw_config.tc_valid_bits,
5207                         bw_ets_config.tc_valid_bits);
5208                /* Still continuing */
5209        }
5210
5211        vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
5212        vsi->bw_max_quanta = bw_config.max_bw;
5213        tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
5214                    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
5215        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5216                vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
5217                vsi->bw_ets_limit_credits[i] =
5218                                        le16_to_cpu(bw_ets_config.credits[i]);
5219                /* 3 bits out of 4 for each TC */
5220                vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
5221        }
5222
5223        return 0;
5224}
5225
5226/**
5227 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
5228 * @vsi: the VSI being configured
5229 * @enabled_tc: TC bitmap
5230 * @bw_share: BW shared credits per TC
5231 *
5232 * Returns 0 on success, negative value on failure
5233 **/
5234static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
5235                                       u8 *bw_share)
5236{
5237        struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5238        struct i40e_pf *pf = vsi->back;
5239        i40e_status ret;
5240        int i;
5241
5242        /* There is no need to reset BW when mqprio mode is on.  */
5243        if (pf->flags & I40E_FLAG_TC_MQPRIO)
5244                return 0;
5245        if (!vsi->mqprio_qopt.qopt.hw && !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5246                ret = i40e_set_bw_limit(vsi, vsi->seid, 0);
5247                if (ret)
5248                        dev_info(&pf->pdev->dev,
5249                                 "Failed to reset tx rate for vsi->seid %u\n",
5250                                 vsi->seid);
5251                return ret;
5252        }
5253        bw_data.tc_valid_bits = enabled_tc;
5254        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5255                bw_data.tc_bw_credits[i] = bw_share[i];
5256
5257        ret = i40e_aq_config_vsi_tc_bw(&pf->hw, vsi->seid, &bw_data, NULL);
5258        if (ret) {
5259                dev_info(&pf->pdev->dev,
5260                         "AQ command Config VSI BW allocation per TC failed = %d\n",
5261                         pf->hw.aq.asq_last_status);
5262                return -EINVAL;
5263        }
5264
5265        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5266                vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5267
5268        return 0;
5269}
5270
5271/**
5272 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5273 * @vsi: the VSI being configured
5274 * @enabled_tc: TC map to be enabled
5275 *
5276 **/
5277static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5278{
5279        struct net_device *netdev = vsi->netdev;
5280        struct i40e_pf *pf = vsi->back;
5281        struct i40e_hw *hw = &pf->hw;
5282        u8 netdev_tc = 0;
5283        int i;
5284        struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5285
5286        if (!netdev)
5287                return;
5288
5289        if (!enabled_tc) {
5290                netdev_reset_tc(netdev);
5291                return;
5292        }
5293
5294        /* Set up actual enabled TCs on the VSI */
5295        if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5296                return;
5297
5298        /* set per TC queues for the VSI */
5299        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5300                /* Only set TC queues for enabled tcs
5301                 *
5302                 * e.g. For a VSI that has TC0 and TC3 enabled the
5303                 * enabled_tc bitmap would be 0x00001001; the driver
5304                 * will set the numtc for netdev as 2 that will be
5305                 * referenced by the netdev layer as TC 0 and 1.
5306                 */
5307                if (vsi->tc_config.enabled_tc & BIT(i))
5308                        netdev_set_tc_queue(netdev,
5309                                        vsi->tc_config.tc_info[i].netdev_tc,
5310                                        vsi->tc_config.tc_info[i].qcount,
5311                                        vsi->tc_config.tc_info[i].qoffset);
5312        }
5313
5314        if (pf->flags & I40E_FLAG_TC_MQPRIO)
5315                return;
5316
5317        /* Assign UP2TC map for the VSI */
5318        for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5319                /* Get the actual TC# for the UP */
5320                u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5321                /* Get the mapped netdev TC# for the UP */
5322                netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5323                netdev_set_prio_tc_map(netdev, i, netdev_tc);
5324        }
5325}
5326
5327/**
5328 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5329 * @vsi: the VSI being configured
5330 * @ctxt: the ctxt buffer returned from AQ VSI update param command
5331 **/
5332static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5333                                      struct i40e_vsi_context *ctxt)
5334{
5335        /* copy just the sections touched not the entire info
5336         * since not all sections are valid as returned by
5337         * update vsi params
5338         */
5339        vsi->info.mapping_flags = ctxt->info.mapping_flags;
5340        memcpy(&vsi->info.queue_mapping,
5341               &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5342        memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5343               sizeof(vsi->info.tc_mapping));
5344}
5345
5346/**
5347 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5348 * @vsi: VSI to be configured
5349 * @enabled_tc: TC bitmap
5350 *
5351 * This configures a particular VSI for TCs that are mapped to the
5352 * given TC bitmap. It uses default bandwidth share for TCs across
5353 * VSIs to configure TC for a particular VSI.
5354 *
5355 * NOTE:
5356 * It is expected that the VSI queues have been quisced before calling
5357 * this function.
5358 **/
5359static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5360{
5361        u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5362        struct i40e_pf *pf = vsi->back;
5363        struct i40e_hw *hw = &pf->hw;
5364        struct i40e_vsi_context ctxt;
5365        int ret = 0;
5366        int i;
5367
5368        /* Check if enabled_tc is same as existing or new TCs */
5369        if (vsi->tc_config.enabled_tc == enabled_tc &&
5370            vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
5371                return ret;
5372
5373        /* Enable ETS TCs with equal BW Share for now across all VSIs */
5374        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5375                if (enabled_tc & BIT(i))
5376                        bw_share[i] = 1;
5377        }
5378
5379        ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5380        if (ret) {
5381                struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5382
5383                dev_info(&pf->pdev->dev,
5384                         "Failed configuring TC map %d for VSI %d\n",
5385                         enabled_tc, vsi->seid);
5386                ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid,
5387                                                  &bw_config, NULL);
5388                if (ret) {
5389                        dev_info(&pf->pdev->dev,
5390                                 "Failed querying vsi bw info, err %s aq_err %s\n",
5391                                 i40e_stat_str(hw, ret),
5392                                 i40e_aq_str(hw, hw->aq.asq_last_status));
5393                        goto out;
5394                }
5395                if ((bw_config.tc_valid_bits & enabled_tc) != enabled_tc) {
5396                        u8 valid_tc = bw_config.tc_valid_bits & enabled_tc;
5397
5398                        if (!valid_tc)
5399                                valid_tc = bw_config.tc_valid_bits;
5400                        /* Always enable TC0, no matter what */
5401                        valid_tc |= 1;
5402                        dev_info(&pf->pdev->dev,
5403                                 "Requested tc 0x%x, but FW reports 0x%x as valid. Attempting to use 0x%x.\n",
5404                                 enabled_tc, bw_config.tc_valid_bits, valid_tc);
5405                        enabled_tc = valid_tc;
5406                }
5407
5408                ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5409                if (ret) {
5410                        dev_err(&pf->pdev->dev,
5411                                "Unable to  configure TC map %d for VSI %d\n",
5412                                enabled_tc, vsi->seid);
5413                        goto out;
5414                }
5415        }
5416
5417        /* Update Queue Pairs Mapping for currently enabled UPs */
5418        ctxt.seid = vsi->seid;
5419        ctxt.pf_num = vsi->back->hw.pf_id;
5420        ctxt.vf_num = 0;
5421        ctxt.uplink_seid = vsi->uplink_seid;
5422        ctxt.info = vsi->info;
5423        if (vsi->back->flags & I40E_FLAG_TC_MQPRIO) {
5424                ret = i40e_vsi_setup_queue_map_mqprio(vsi, &ctxt, enabled_tc);
5425                if (ret)
5426                        goto out;
5427        } else {
5428                i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5429        }
5430
5431        /* On destroying the qdisc, reset vsi->rss_size, as number of enabled
5432         * queues changed.
5433         */
5434        if (!vsi->mqprio_qopt.qopt.hw && vsi->reconfig_rss) {
5435                vsi->rss_size = min_t(int, vsi->back->alloc_rss_size,
5436                                      vsi->num_queue_pairs);
5437                ret = i40e_vsi_config_rss(vsi);
5438                if (ret) {
5439                        dev_info(&vsi->back->pdev->dev,
5440                                 "Failed to reconfig rss for num_queues\n");
5441                        return ret;
5442                }
5443                vsi->reconfig_rss = false;
5444        }
5445        if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5446                ctxt.info.valid_sections |=
5447                                cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5448                ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5449        }
5450
5451        /* Update the VSI after updating the VSI queue-mapping
5452         * information
5453         */
5454        ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
5455        if (ret) {
5456                dev_info(&pf->pdev->dev,
5457                         "Update vsi tc config failed, err %s aq_err %s\n",
5458                         i40e_stat_str(hw, ret),
5459                         i40e_aq_str(hw, hw->aq.asq_last_status));
5460                goto out;
5461        }
5462        /* update the local VSI info with updated queue map */
5463        i40e_vsi_update_queue_map(vsi, &ctxt);
5464        vsi->info.valid_sections = 0;
5465
5466        /* Update current VSI BW information */
5467        ret = i40e_vsi_get_bw_info(vsi);
5468        if (ret) {
5469                dev_info(&pf->pdev->dev,
5470                         "Failed updating vsi bw info, err %s aq_err %s\n",
5471                         i40e_stat_str(hw, ret),
5472                         i40e_aq_str(hw, hw->aq.asq_last_status));
5473                goto out;
5474        }
5475
5476        /* Update the netdev TC setup */
5477        i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5478out:
5479        return ret;
5480}
5481
5482/**
5483 * i40e_get_link_speed - Returns link speed for the interface
5484 * @vsi: VSI to be configured
5485 *
5486 **/
5487static int i40e_get_link_speed(struct i40e_vsi *vsi)
5488{
5489        struct i40e_pf *pf = vsi->back;
5490
5491        switch (pf->hw.phy.link_info.link_speed) {
5492        case I40E_LINK_SPEED_40GB:
5493                return 40000;
5494        case I40E_LINK_SPEED_25GB:
5495                return 25000;
5496        case I40E_LINK_SPEED_20GB:
5497                return 20000;
5498        case I40E_LINK_SPEED_10GB:
5499                return 10000;
5500        case I40E_LINK_SPEED_1GB:
5501                return 1000;
5502        default:
5503                return -EINVAL;
5504        }
5505}
5506
5507/**
5508 * i40e_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
5509 * @vsi: VSI to be configured
5510 * @seid: seid of the channel/VSI
5511 * @max_tx_rate: max TX rate to be configured as BW limit
5512 *
5513 * Helper function to set BW limit for a given VSI
5514 **/
5515int i40e_set_bw_limit(struct i40e_vsi *vsi, u16 seid, u64 max_tx_rate)
5516{
5517        struct i40e_pf *pf = vsi->back;
5518        u64 credits = 0;
5519        int speed = 0;
5520        int ret = 0;
5521
5522        speed = i40e_get_link_speed(vsi);
5523        if (max_tx_rate > speed) {
5524                dev_err(&pf->pdev->dev,
5525                        "Invalid max tx rate %llu specified for VSI seid %d.",
5526                        max_tx_rate, seid);
5527                return -EINVAL;
5528        }
5529        if (max_tx_rate && max_tx_rate < 50) {
5530                dev_warn(&pf->pdev->dev,
5531                         "Setting max tx rate to minimum usable value of 50Mbps.\n");
5532                max_tx_rate = 50;
5533        }
5534
5535        /* Tx rate credits are in values of 50Mbps, 0 is disabled */
5536        credits = max_tx_rate;
5537        do_div(credits, I40E_BW_CREDIT_DIVISOR);
5538        ret = i40e_aq_config_vsi_bw_limit(&pf->hw, seid, credits,
5539                                          I40E_MAX_BW_INACTIVE_ACCUM, NULL);
5540        if (ret)
5541                dev_err(&pf->pdev->dev,
5542                        "Failed set tx rate (%llu Mbps) for vsi->seid %u, err %s aq_err %s\n",
5543                        max_tx_rate, seid, i40e_stat_str(&pf->hw, ret),
5544                        i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5545        return ret;
5546}
5547
5548/**
5549 * i40e_remove_queue_channels - Remove queue channels for the TCs
5550 * @vsi: VSI to be configured
5551 *
5552 * Remove queue channels for the TCs
5553 **/
5554static void i40e_remove_queue_channels(struct i40e_vsi *vsi)
5555{
5556        enum i40e_admin_queue_err last_aq_status;
5557        struct i40e_cloud_filter *cfilter;
5558        struct i40e_channel *ch, *ch_tmp;
5559        struct i40e_pf *pf = vsi->back;
5560        struct hlist_node *node;
5561        int ret, i;
5562
5563        /* Reset rss size that was stored when reconfiguring rss for
5564         * channel VSIs with non-power-of-2 queue count.
5565         */
5566        vsi->current_rss_size = 0;
5567
5568        /* perform cleanup for channels if they exist */
5569        if (list_empty(&vsi->ch_list))
5570                return;
5571
5572        list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5573                struct i40e_vsi *p_vsi;
5574
5575                list_del(&ch->list);
5576                p_vsi = ch->parent_vsi;
5577                if (!p_vsi || !ch->initialized) {
5578                        kfree(ch);
5579                        continue;
5580                }
5581                /* Reset queue contexts */
5582                for (i = 0; i < ch->num_queue_pairs; i++) {
5583                        struct i40e_ring *tx_ring, *rx_ring;
5584                        u16 pf_q;
5585
5586                        pf_q = ch->base_queue + i;
5587                        tx_ring = vsi->tx_rings[pf_q];
5588                        tx_ring->ch = NULL;
5589
5590                        rx_ring = vsi->rx_rings[pf_q];
5591                        rx_ring->ch = NULL;
5592                }
5593
5594                /* Reset BW configured for this VSI via mqprio */
5595                ret = i40e_set_bw_limit(vsi, ch->seid, 0);
5596                if (ret)
5597                        dev_info(&vsi->back->pdev->dev,
5598                                 "Failed to reset tx rate for ch->seid %u\n",
5599                                 ch->seid);
5600
5601                /* delete cloud filters associated with this channel */
5602                hlist_for_each_entry_safe(cfilter, node,
5603                                          &pf->cloud_filter_list, cloud_node) {
5604                        if (cfilter->seid != ch->seid)
5605                                continue;
5606
5607                        hash_del(&cfilter->cloud_node);
5608                        if (cfilter->dst_port)
5609                                ret = i40e_add_del_cloud_filter_big_buf(vsi,
5610                                                                        cfilter,
5611                                                                        false);
5612                        else
5613                                ret = i40e_add_del_cloud_filter(vsi, cfilter,
5614                                                                false);
5615                        last_aq_status = pf->hw.aq.asq_last_status;
5616                        if (ret)
5617                                dev_info(&pf->pdev->dev,
5618                                         "Failed to delete cloud filter, err %s aq_err %s\n",
5619                                         i40e_stat_str(&pf->hw, ret),
5620                                         i40e_aq_str(&pf->hw, last_aq_status));
5621                        kfree(cfilter);
5622                }
5623
5624                /* delete VSI from FW */
5625                ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
5626                                             NULL);
5627                if (ret)
5628                        dev_err(&vsi->back->pdev->dev,
5629                                "unable to remove channel (%d) for parent VSI(%d)\n",
5630                                ch->seid, p_vsi->seid);
5631                kfree(ch);
5632        }
5633        INIT_LIST_HEAD(&vsi->ch_list);
5634}
5635
5636/**
5637 * i40e_is_any_channel - channel exist or not
5638 * @vsi: ptr to VSI to which channels are associated with
5639 *
5640 * Returns true or false if channel(s) exist for associated VSI or not
5641 **/
5642static bool i40e_is_any_channel(struct i40e_vsi *vsi)
5643{
5644        struct i40e_channel *ch, *ch_tmp;
5645
5646        list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5647                if (ch->initialized)
5648                        return true;
5649        }
5650
5651        return false;
5652}
5653
5654/**
5655 * i40e_get_max_queues_for_channel
5656 * @vsi: ptr to VSI to which channels are associated with
5657 *
5658 * Helper function which returns max value among the queue counts set on the
5659 * channels/TCs created.
5660 **/
5661static int i40e_get_max_queues_for_channel(struct i40e_vsi *vsi)
5662{
5663        struct i40e_channel *ch, *ch_tmp;
5664        int max = 0;
5665
5666        list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5667                if (!ch->initialized)
5668                        continue;
5669                if (ch->num_queue_pairs > max)
5670                        max = ch->num_queue_pairs;
5671        }
5672
5673        return max;
5674}
5675
5676/**
5677 * i40e_validate_num_queues - validate num_queues w.r.t channel
5678 * @pf: ptr to PF device
5679 * @num_queues: number of queues
5680 * @vsi: the parent VSI
5681 * @reconfig_rss: indicates should the RSS be reconfigured or not
5682 *
5683 * This function validates number of queues in the context of new channel
5684 * which is being established and determines if RSS should be reconfigured
5685 * or not for parent VSI.
5686 **/
5687static int i40e_validate_num_queues(struct i40e_pf *pf, int num_queues,
5688                                    struct i40e_vsi *vsi, bool *reconfig_rss)
5689{
5690        int max_ch_queues;
5691
5692        if (!reconfig_rss)
5693                return -EINVAL;
5694
5695        *reconfig_rss = false;
5696        if (vsi->current_rss_size) {
5697                if (num_queues > vsi->current_rss_size) {
5698                        dev_dbg(&pf->pdev->dev,
5699                                "Error: num_queues (%d) > vsi's current_size(%d)\n",
5700                                num_queues, vsi->current_rss_size);
5701                        return -EINVAL;
5702                } else if ((num_queues < vsi->current_rss_size) &&
5703                           (!is_power_of_2(num_queues))) {
5704                        dev_dbg(&pf->pdev->dev,
5705                                "Error: num_queues (%d) < vsi's current_size(%d), but not power of 2\n",
5706                                num_queues, vsi->current_rss_size);
5707                        return -EINVAL;
5708                }
5709        }
5710
5711        if (!is_power_of_2(num_queues)) {
5712                /* Find the max num_queues configured for channel if channel
5713                 * exist.
5714                 * if channel exist, then enforce 'num_queues' to be more than
5715                 * max ever queues configured for channel.
5716                 */
5717                max_ch_queues = i40e_get_max_queues_for_channel(vsi);
5718                if (num_queues < max_ch_queues) {
5719                        dev_dbg(&pf->pdev->dev,
5720                                "Error: num_queues (%d) < max queues configured for channel(%d)\n",
5721                                num_queues, max_ch_queues);
5722                        return -EINVAL;
5723                }
5724                *reconfig_rss = true;
5725        }
5726
5727        return 0;
5728}
5729
5730/**
5731 * i40e_vsi_reconfig_rss - reconfig RSS based on specified rss_size
5732 * @vsi: the VSI being setup
5733 * @rss_size: size of RSS, accordingly LUT gets reprogrammed
5734 *
5735 * This function reconfigures RSS by reprogramming LUTs using 'rss_size'
5736 **/
5737static int i40e_vsi_reconfig_rss(struct i40e_vsi *vsi, u16 rss_size)
5738{
5739        struct i40e_pf *pf = vsi->back;
5740        u8 seed[I40E_HKEY_ARRAY_SIZE];
5741        struct i40e_hw *hw = &pf->hw;
5742        int local_rss_size;
5743        u8 *lut;
5744        int ret;
5745
5746        if (!vsi->rss_size)
5747                return -EINVAL;
5748
5749        if (rss_size > vsi->rss_size)
5750                return -EINVAL;
5751
5752        local_rss_size = min_t(int, vsi->rss_size, rss_size);
5753        lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
5754        if (!lut)
5755                return -ENOMEM;
5756
5757        /* Ignoring user configured lut if there is one */
5758        i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, local_rss_size);
5759
5760        /* Use user configured hash key if there is one, otherwise
5761         * use default.
5762         */
5763        if (vsi->rss_hkey_user)
5764                memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
5765        else
5766                netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
5767
5768        ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
5769        if (ret) {
5770                dev_info(&pf->pdev->dev,
5771                         "Cannot set RSS lut, err %s aq_err %s\n",
5772                         i40e_stat_str(hw, ret),
5773                         i40e_aq_str(hw, hw->aq.asq_last_status));
5774                kfree(lut);
5775                return ret;
5776        }
5777        kfree(lut);
5778
5779        /* Do the update w.r.t. storing rss_size */
5780        if (!vsi->orig_rss_size)
5781                vsi->orig_rss_size = vsi->rss_size;
5782        vsi->current_rss_size = local_rss_size;
5783
5784        return ret;
5785}
5786
5787/**
5788 * i40e_channel_setup_queue_map - Setup a channel queue map
5789 * @pf: ptr to PF device
5790 * @vsi: the VSI being setup
5791 * @ctxt: VSI context structure
5792 * @ch: ptr to channel structure
5793 *
5794 * Setup queue map for a specific channel
5795 **/
5796static void i40e_channel_setup_queue_map(struct i40e_pf *pf,
5797                                         struct i40e_vsi_context *ctxt,
5798                                         struct i40e_channel *ch)
5799{
5800        u16 qcount, qmap, sections = 0;
5801        u8 offset = 0;
5802        int pow;
5803
5804        sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
5805        sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
5806
5807        qcount = min_t(int, ch->num_queue_pairs, pf->num_lan_msix);
5808        ch->num_queue_pairs = qcount;
5809
5810        /* find the next higher power-of-2 of num queue pairs */
5811        pow = ilog2(qcount);
5812        if (!is_power_of_2(qcount))
5813                pow++;
5814
5815        qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
5816                (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
5817
5818        /* Setup queue TC[0].qmap for given VSI context */
5819        ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
5820
5821        ctxt->info.up_enable_bits = 0x1; /* TC0 enabled */
5822        ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
5823        ctxt->info.queue_mapping[0] = cpu_to_le16(ch->base_queue);
5824        ctxt->info.valid_sections |= cpu_to_le16(sections);
5825}
5826
5827/**
5828 * i40e_add_channel - add a channel by adding VSI
5829 * @pf: ptr to PF device
5830 * @uplink_seid: underlying HW switching element (VEB) ID
5831 * @ch: ptr to channel structure
5832 *
5833 * Add a channel (VSI) using add_vsi and queue_map
5834 **/
5835static int i40e_add_channel(struct i40e_pf *pf, u16 uplink_seid,
5836                            struct i40e_channel *ch)
5837{
5838        struct i40e_hw *hw = &pf->hw;
5839        struct i40e_vsi_context ctxt;
5840        u8 enabled_tc = 0x1; /* TC0 enabled */
5841        int ret;
5842
5843        if (ch->type != I40E_VSI_VMDQ2) {
5844                dev_info(&pf->pdev->dev,
5845                         "add new vsi failed, ch->type %d\n", ch->type);
5846                return -EINVAL;
5847        }
5848
5849        memset(&ctxt, 0, sizeof(ctxt));
5850        ctxt.pf_num = hw->pf_id;
5851        ctxt.vf_num = 0;
5852        ctxt.uplink_seid = uplink_seid;
5853        ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
5854        if (ch->type == I40E_VSI_VMDQ2)
5855                ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5856
5857        if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED) {
5858                ctxt.info.valid_sections |=
5859                     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5860                ctxt.info.switch_id =
5861                   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5862        }
5863
5864        /* Set queue map for a given VSI context */
5865        i40e_channel_setup_queue_map(pf, &ctxt, ch);
5866
5867        /* Now time to create VSI */
5868        ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5869        if (ret) {
5870                dev_info(&pf->pdev->dev,
5871                         "add new vsi failed, err %s aq_err %s\n",
5872                         i40e_stat_str(&pf->hw, ret),
5873                         i40e_aq_str(&pf->hw,
5874                                     pf->hw.aq.asq_last_status));
5875                return -ENOENT;
5876        }
5877
5878        /* Success, update channel, set enabled_tc only if the channel
5879         * is not a macvlan
5880         */
5881        ch->enabled_tc = !i40e_is_channel_macvlan(ch) && enabled_tc;
5882        ch->seid = ctxt.seid;
5883        ch->vsi_number = ctxt.vsi_number;
5884        ch->stat_counter_idx = cpu_to_le16(ctxt.info.stat_counter_idx);
5885
5886        /* copy just the sections touched not the entire info
5887         * since not all sections are valid as returned by
5888         * update vsi params
5889         */
5890        ch->info.mapping_flags = ctxt.info.mapping_flags;
5891        memcpy(&ch->info.queue_mapping,
5892               &ctxt.info.queue_mapping, sizeof(ctxt.info.queue_mapping));
5893        memcpy(&ch->info.tc_mapping, ctxt.info.tc_mapping,
5894               sizeof(ctxt.info.tc_mapping));
5895
5896        return 0;
5897}
5898
5899static int i40e_channel_config_bw(struct i40e_vsi *vsi, struct i40e_channel *ch,
5900                                  u8 *bw_share)
5901{
5902        struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5903        i40e_status ret;
5904        int i;
5905
5906        bw_data.tc_valid_bits = ch->enabled_tc;
5907        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5908                bw_data.tc_bw_credits[i] = bw_share[i];
5909
5910        ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, ch->seid,
5911                                       &bw_data, NULL);
5912        if (ret) {
5913                dev_info(&vsi->back->pdev->dev,
5914                         "Config VSI BW allocation per TC failed, aq_err: %d for new_vsi->seid %u\n",
5915                         vsi->back->hw.aq.asq_last_status, ch->seid);
5916                return -EINVAL;
5917        }
5918
5919        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5920                ch->info.qs_handle[i] = bw_data.qs_handles[i];
5921
5922        return 0;
5923}
5924
5925/**
5926 * i40e_channel_config_tx_ring - config TX ring associated with new channel
5927 * @pf: ptr to PF device
5928 * @vsi: the VSI being setup
5929 * @ch: ptr to channel structure
5930 *
5931 * Configure TX rings associated with channel (VSI) since queues are being
5932 * from parent VSI.
5933 **/
5934static int i40e_channel_config_tx_ring(struct i40e_pf *pf,
5935                                       struct i40e_vsi *vsi,
5936                                       struct i40e_channel *ch)
5937{
5938        i40e_status ret;
5939        int i;
5940        u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5941
5942        /* Enable ETS TCs with equal BW Share for now across all VSIs */
5943        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5944                if (ch->enabled_tc & BIT(i))
5945                        bw_share[i] = 1;
5946        }
5947
5948        /* configure BW for new VSI */
5949        ret = i40e_channel_config_bw(vsi, ch, bw_share);
5950        if (ret) {
5951                dev_info(&vsi->back->pdev->dev,
5952                         "Failed configuring TC map %d for channel (seid %u)\n",
5953                         ch->enabled_tc, ch->seid);
5954                return ret;
5955        }
5956
5957        for (i = 0; i < ch->num_queue_pairs; i++) {
5958                struct i40e_ring *tx_ring, *rx_ring;
5959                u16 pf_q;
5960
5961                pf_q = ch->base_queue + i;
5962
5963                /* Get to TX ring ptr of main VSI, for re-setup TX queue
5964                 * context
5965                 */
5966                tx_ring = vsi->tx_rings[pf_q];
5967                tx_ring->ch = ch;
5968
5969                /* Get the RX ring ptr */
5970                rx_ring = vsi->rx_rings[pf_q];
5971                rx_ring->ch = ch;
5972        }
5973
5974        return 0;
5975}
5976
5977/**
5978 * i40e_setup_hw_channel - setup new channel
5979 * @pf: ptr to PF device
5980 * @vsi: the VSI being setup
5981 * @ch: ptr to channel structure
5982 * @uplink_seid: underlying HW switching element (VEB) ID
5983 * @type: type of channel to be created (VMDq2/VF)
5984 *
5985 * Setup new channel (VSI) based on specified type (VMDq2/VF)
5986 * and configures TX rings accordingly
5987 **/
5988static inline int i40e_setup_hw_channel(struct i40e_pf *pf,
5989                                        struct i40e_vsi *vsi,
5990                                        struct i40e_channel *ch,
5991                                        u16 uplink_seid, u8 type)
5992{
5993        int ret;
5994
5995        ch->initialized = false;
5996        ch->base_queue = vsi->next_base_queue;
5997        ch->type = type;
5998
5999        /* Proceed with creation of channel (VMDq2) VSI */
6000        ret = i40e_add_channel(pf, uplink_seid, ch);
6001        if (ret) {
6002                dev_info(&pf->pdev->dev,
6003                         "failed to add_channel using uplink_seid %u\n",
6004                         uplink_seid);
6005                return ret;
6006        }
6007
6008        /* Mark the successful creation of channel */
6009        ch->initialized = true;
6010
6011        /* Reconfigure TX queues using QTX_CTL register */
6012        ret = i40e_channel_config_tx_ring(pf, vsi, ch);
6013        if (ret) {
6014                dev_info(&pf->pdev->dev,
6015                         "failed to configure TX rings for channel %u\n",
6016                         ch->seid);
6017                return ret;
6018        }
6019
6020        /* update 'next_base_queue' */
6021        vsi->next_base_queue = vsi->next_base_queue + ch->num_queue_pairs;
6022        dev_dbg(&pf->pdev->dev,
6023                "Added channel: vsi_seid %u, vsi_number %u, stat_counter_idx %u, num_queue_pairs %u, pf->next_base_queue %d\n",
6024                ch->seid, ch->vsi_number, ch->stat_counter_idx,
6025                ch->num_queue_pairs,
6026                vsi->next_base_queue);
6027        return ret;
6028}
6029
6030/**
6031 * i40e_setup_channel - setup new channel using uplink element
6032 * @pf: ptr to PF device
6033 * @type: type of channel to be created (VMDq2/VF)
6034 * @uplink_seid: underlying HW switching element (VEB) ID
6035 * @ch: ptr to channel structure
6036 *
6037 * Setup new channel (VSI) based on specified type (VMDq2/VF)
6038 * and uplink switching element (uplink_seid)
6039 **/
6040static bool i40e_setup_channel(struct i40e_pf *pf, struct i40e_vsi *vsi,
6041                               struct i40e_channel *ch)
6042{
6043        u8 vsi_type;
6044        u16 seid;
6045        int ret;
6046
6047        if (vsi->type == I40E_VSI_MAIN) {
6048                vsi_type = I40E_VSI_VMDQ2;
6049        } else {
6050                dev_err(&pf->pdev->dev, "unsupported parent vsi type(%d)\n",
6051                        vsi->type);
6052                return false;
6053        }
6054
6055        /* underlying switching element */
6056        seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6057
6058        /* create channel (VSI), configure TX rings */
6059        ret = i40e_setup_hw_channel(pf, vsi, ch, seid, vsi_type);
6060        if (ret) {
6061                dev_err(&pf->pdev->dev, "failed to setup hw_channel\n");
6062                return false;
6063        }
6064
6065        return ch->initialized ? true : false;
6066}
6067
6068/**
6069 * i40e_validate_and_set_switch_mode - sets up switch mode correctly
6070 * @vsi: ptr to VSI which has PF backing
6071 *
6072 * Sets up switch mode correctly if it needs to be changed and perform
6073 * what are allowed modes.
6074 **/
6075static int i40e_validate_and_set_switch_mode(struct i40e_vsi *vsi)
6076{
6077        u8 mode;
6078        struct i40e_pf *pf = vsi->back;
6079        struct i40e_hw *hw = &pf->hw;
6080        int ret;
6081
6082        ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_dev_capabilities);
6083        if (ret)
6084                return -EINVAL;
6085
6086        if (hw->dev_caps.switch_mode) {
6087                /* if switch mode is set, support mode2 (non-tunneled for
6088                 * cloud filter) for now
6089                 */
6090                u32 switch_mode = hw->dev_caps.switch_mode &
6091                                  I40E_SWITCH_MODE_MASK;
6092                if (switch_mode >= I40E_CLOUD_FILTER_MODE1) {
6093                        if (switch_mode == I40E_CLOUD_FILTER_MODE2)
6094                                return 0;
6095                        dev_err(&pf->pdev->dev,
6096                                "Invalid switch_mode (%d), only non-tunneled mode for cloud filter is supported\n",
6097                                hw->dev_caps.switch_mode);
6098                        return -EINVAL;
6099                }
6100        }
6101
6102        /* Set Bit 7 to be valid */
6103        mode = I40E_AQ_SET_SWITCH_BIT7_VALID;
6104
6105        /* Set L4type for TCP support */
6106        mode |= I40E_AQ_SET_SWITCH_L4_TYPE_TCP;
6107
6108        /* Set cloud filter mode */
6109        mode |= I40E_AQ_SET_SWITCH_MODE_NON_TUNNEL;
6110
6111        /* Prep mode field for set_switch_config */
6112        ret = i40e_aq_set_switch_config(hw, pf->last_sw_conf_flags,
6113                                        pf->last_sw_conf_valid_flags,
6114                                        mode, NULL);
6115        if (ret && hw->aq.asq_last_status != I40E_AQ_RC_ESRCH)
6116                dev_err(&pf->pdev->dev,
6117                        "couldn't set switch config bits, err %s aq_err %s\n",
6118                        i40e_stat_str(hw, ret),
6119                        i40e_aq_str(hw,
6120                                    hw->aq.asq_last_status));
6121
6122        return ret;
6123}
6124
6125/**
6126 * i40e_create_queue_channel - function to create channel
6127 * @vsi: VSI to be configured
6128 * @ch: ptr to channel (it contains channel specific params)
6129 *
6130 * This function creates channel (VSI) using num_queues specified by user,
6131 * reconfigs RSS if needed.
6132 **/
6133int i40e_create_queue_channel(struct i40e_vsi *vsi,
6134                              struct i40e_channel *ch)
6135{
6136        struct i40e_pf *pf = vsi->back;
6137        bool reconfig_rss;
6138        int err;
6139
6140        if (!ch)
6141                return -EINVAL;
6142
6143        if (!ch->num_queue_pairs) {
6144                dev_err(&pf->pdev->dev, "Invalid num_queues requested: %d\n",
6145                        ch->num_queue_pairs);
6146                return -EINVAL;
6147        }
6148
6149        /* validate user requested num_queues for channel */
6150        err = i40e_validate_num_queues(pf, ch->num_queue_pairs, vsi,
6151                                       &reconfig_rss);
6152        if (err) {
6153                dev_info(&pf->pdev->dev, "Failed to validate num_queues (%d)\n",
6154                         ch->num_queue_pairs);
6155                return -EINVAL;
6156        }
6157
6158        /* By default we are in VEPA mode, if this is the first VF/VMDq
6159         * VSI to be added switch to VEB mode.
6160         */
6161        if ((!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) ||
6162            (!i40e_is_any_channel(vsi))) {
6163                if (!is_power_of_2(vsi->tc_config.tc_info[0].qcount)) {
6164                        dev_dbg(&pf->pdev->dev,
6165                                "Failed to create channel. Override queues (%u) not power of 2\n",
6166                                vsi->tc_config.tc_info[0].qcount);
6167                        return -EINVAL;
6168                }
6169
6170                if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
6171                        pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
6172
6173                        if (vsi->type == I40E_VSI_MAIN) {
6174                                if (pf->flags & I40E_FLAG_TC_MQPRIO)
6175                                        i40e_do_reset(pf, I40E_PF_RESET_FLAG,
6176                                                      true);
6177                                else
6178                                        i40e_do_reset_safe(pf,
6179                                                           I40E_PF_RESET_FLAG);
6180                        }
6181                }
6182                /* now onwards for main VSI, number of queues will be value
6183                 * of TC0's queue count
6184                 */
6185        }
6186
6187        /* By this time, vsi->cnt_q_avail shall be set to non-zero and
6188         * it should be more than num_queues
6189         */
6190        if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_queue_pairs) {
6191                dev_dbg(&pf->pdev->dev,
6192                        "Error: cnt_q_avail (%u) less than num_queues %d\n",
6193                        vsi->cnt_q_avail, ch->num_queue_pairs);
6194                return -EINVAL;
6195        }
6196
6197        /* reconfig_rss only if vsi type is MAIN_VSI */
6198        if (reconfig_rss && (vsi->type == I40E_VSI_MAIN)) {
6199                err = i40e_vsi_reconfig_rss(vsi, ch->num_queue_pairs);
6200                if (err) {
6201                        dev_info(&pf->pdev->dev,
6202                                 "Error: unable to reconfig rss for num_queues (%u)\n",
6203                                 ch->num_queue_pairs);
6204                        return -EINVAL;
6205                }
6206        }
6207
6208        if (!i40e_setup_channel(pf, vsi, ch)) {
6209                dev_info(&pf->pdev->dev, "Failed to setup channel\n");
6210                return -EINVAL;
6211        }
6212
6213        dev_info(&pf->pdev->dev,
6214                 "Setup channel (id:%u) utilizing num_queues %d\n",
6215                 ch->seid, ch->num_queue_pairs);
6216
6217        /* configure VSI for BW limit */
6218        if (ch->max_tx_rate) {
6219                u64 credits = ch->max_tx_rate;
6220
6221                if (i40e_set_bw_limit(vsi, ch->seid, ch->max_tx_rate))
6222                        return -EINVAL;
6223
6224                do_div(credits, I40E_BW_CREDIT_DIVISOR);
6225                dev_dbg(&pf->pdev->dev,
6226                        "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6227                        ch->max_tx_rate,
6228                        credits,
6229                        ch->seid);
6230        }
6231
6232        /* in case of VF, this will be main SRIOV VSI */
6233        ch->parent_vsi = vsi;
6234
6235        /* and update main_vsi's count for queue_available to use */
6236        vsi->cnt_q_avail -= ch->num_queue_pairs;
6237
6238        return 0;
6239}
6240
6241/**
6242 * i40e_configure_queue_channels - Add queue channel for the given TCs
6243 * @vsi: VSI to be configured
6244 *
6245 * Configures queue channel mapping to the given TCs
6246 **/
6247static int i40e_configure_queue_channels(struct i40e_vsi *vsi)
6248{
6249        struct i40e_channel *ch;
6250        u64 max_rate = 0;
6251        int ret = 0, i;
6252
6253        /* Create app vsi with the TCs. Main VSI with TC0 is already set up */
6254        vsi->tc_seid_map[0] = vsi->seid;
6255        for (i = 1; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6256                if (vsi->tc_config.enabled_tc & BIT(i)) {
6257                        ch = kzalloc(sizeof(*ch), GFP_KERNEL);
6258                        if (!ch) {
6259                                ret = -ENOMEM;
6260                                goto err_free;
6261                        }
6262
6263                        INIT_LIST_HEAD(&ch->list);
6264                        ch->num_queue_pairs =
6265                                vsi->tc_config.tc_info[i].qcount;
6266                        ch->base_queue =
6267                                vsi->tc_config.tc_info[i].qoffset;
6268
6269                        /* Bandwidth limit through tc interface is in bytes/s,
6270                         * change to Mbit/s
6271                         */
6272                        max_rate = vsi->mqprio_qopt.max_rate[i];
6273                        do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6274                        ch->max_tx_rate = max_rate;
6275
6276                        list_add_tail(&ch->list, &vsi->ch_list);
6277
6278                        ret = i40e_create_queue_channel(vsi, ch);
6279                        if (ret) {
6280                                dev_err(&vsi->back->pdev->dev,
6281                                        "Failed creating queue channel with TC%d: queues %d\n",
6282                                        i, ch->num_queue_pairs);
6283                                goto err_free;
6284                        }
6285                        vsi->tc_seid_map[i] = ch->seid;
6286                }
6287        }
6288        return ret;
6289
6290err_free:
6291        i40e_remove_queue_channels(vsi);
6292        return ret;
6293}
6294
6295/**
6296 * i40e_veb_config_tc - Configure TCs for given VEB
6297 * @veb: given VEB
6298 * @enabled_tc: TC bitmap
6299 *
6300 * Configures given TC bitmap for VEB (switching) element
6301 **/
6302int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
6303{
6304        struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
6305        struct i40e_pf *pf = veb->pf;
6306        int ret = 0;
6307        int i;
6308
6309        /* No TCs or already enabled TCs just return */
6310        if (!enabled_tc || veb->enabled_tc == enabled_tc)
6311                return ret;
6312
6313        bw_data.tc_valid_bits = enabled_tc;
6314        /* bw_data.absolute_credits is not set (relative) */
6315
6316        /* Enable ETS TCs with equal BW Share for now */
6317        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6318                if (enabled_tc & BIT(i))
6319                        bw_data.tc_bw_share_credits[i] = 1;
6320        }
6321
6322        ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
6323                                                   &bw_data, NULL);
6324        if (ret) {
6325                dev_info(&pf->pdev->dev,
6326                         "VEB bw config failed, err %s aq_err %s\n",
6327                         i40e_stat_str(&pf->hw, ret),
6328                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6329                goto out;
6330        }
6331
6332        /* Update the BW information */
6333        ret = i40e_veb_get_bw_info(veb);
6334        if (ret) {
6335                dev_info(&pf->pdev->dev,
6336                         "Failed getting veb bw config, err %s aq_err %s\n",
6337                         i40e_stat_str(&pf->hw, ret),
6338                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6339        }
6340
6341out:
6342        return ret;
6343}
6344
6345#ifdef CONFIG_I40E_DCB
6346/**
6347 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
6348 * @pf: PF struct
6349 *
6350 * Reconfigure VEB/VSIs on a given PF; it is assumed that
6351 * the caller would've quiesce all the VSIs before calling
6352 * this function
6353 **/
6354static void i40e_dcb_reconfigure(struct i40e_pf *pf)
6355{
6356        u8 tc_map = 0;
6357        int ret;
6358        u8 v;
6359
6360        /* Enable the TCs available on PF to all VEBs */
6361        tc_map = i40e_pf_get_tc_map(pf);
6362        for (v = 0; v < I40E_MAX_VEB; v++) {
6363                if (!pf->veb[v])
6364                        continue;
6365                ret = i40e_veb_config_tc(pf->veb[v], tc_map);
6366                if (ret) {
6367                        dev_info(&pf->pdev->dev,
6368                                 "Failed configuring TC for VEB seid=%d\n",
6369                                 pf->veb[v]->seid);
6370                        /* Will try to configure as many components */
6371                }
6372        }
6373
6374        /* Update each VSI */
6375        for (v = 0; v < pf->num_alloc_vsi; v++) {
6376                if (!pf->vsi[v])
6377                        continue;
6378
6379                /* - Enable all TCs for the LAN VSI
6380                 * - For all others keep them at TC0 for now
6381                 */
6382                if (v == pf->lan_vsi)
6383                        tc_map = i40e_pf_get_tc_map(pf);
6384                else
6385                        tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
6386
6387                ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
6388                if (ret) {
6389                        dev_info(&pf->pdev->dev,
6390                                 "Failed configuring TC for VSI seid=%d\n",
6391                                 pf->vsi[v]->seid);
6392                        /* Will try to configure as many components */
6393                } else {
6394                        /* Re-configure VSI vectors based on updated TC map */
6395                        i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
6396                        if (pf->vsi[v]->netdev)
6397                                i40e_dcbnl_set_all(pf->vsi[v]);
6398                }
6399        }
6400}
6401
6402/**
6403 * i40e_resume_port_tx - Resume port Tx
6404 * @pf: PF struct
6405 *
6406 * Resume a port's Tx and issue a PF reset in case of failure to
6407 * resume.
6408 **/
6409static int i40e_resume_port_tx(struct i40e_pf *pf)
6410{
6411        struct i40e_hw *hw = &pf->hw;
6412        int ret;
6413
6414        ret = i40e_aq_resume_port_tx(hw, NULL);
6415        if (ret) {
6416                dev_info(&pf->pdev->dev,
6417                         "Resume Port Tx failed, err %s aq_err %s\n",
6418                          i40e_stat_str(&pf->hw, ret),
6419                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6420                /* Schedule PF reset to recover */
6421                set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6422                i40e_service_event_schedule(pf);
6423        }
6424
6425        return ret;
6426}
6427
6428/**
6429 * i40e_init_pf_dcb - Initialize DCB configuration
6430 * @pf: PF being configured
6431 *
6432 * Query the current DCB configuration and cache it
6433 * in the hardware structure
6434 **/
6435static int i40e_init_pf_dcb(struct i40e_pf *pf)
6436{
6437        struct i40e_hw *hw = &pf->hw;
6438        int err = 0;
6439
6440        /* Do not enable DCB for SW1 and SW2 images even if the FW is capable
6441         * Also do not enable DCBx if FW LLDP agent is disabled
6442         */
6443        if ((pf->hw_features & I40E_HW_NO_DCB_SUPPORT) ||
6444            (pf->flags & I40E_FLAG_DISABLE_FW_LLDP)) {
6445                dev_info(&pf->pdev->dev, "DCB is not supported or FW LLDP is disabled\n");
6446                err = I40E_NOT_SUPPORTED;
6447                goto out;
6448        }
6449
6450        err = i40e_init_dcb(hw, true);
6451        if (!err) {
6452                /* Device/Function is not DCBX capable */
6453                if ((!hw->func_caps.dcb) ||
6454                    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
6455                        dev_info(&pf->pdev->dev,
6456                                 "DCBX offload is not supported or is disabled for this PF.\n");
6457                } else {
6458                        /* When status is not DISABLED then DCBX in FW */
6459                        pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
6460                                       DCB_CAP_DCBX_VER_IEEE;
6461
6462                        pf->flags |= I40E_FLAG_DCB_CAPABLE;
6463                        /* Enable DCB tagging only when more than one TC
6464                         * or explicitly disable if only one TC
6465                         */
6466                        if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6467                                pf->flags |= I40E_FLAG_DCB_ENABLED;
6468                        else
6469                                pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6470                        dev_dbg(&pf->pdev->dev,
6471                                "DCBX offload is supported for this PF.\n");
6472                }
6473        } else if (pf->hw.aq.asq_last_status == I40E_AQ_RC_EPERM) {
6474                dev_info(&pf->pdev->dev, "FW LLDP disabled for this PF.\n");
6475                pf->flags |= I40E_FLAG_DISABLE_FW_LLDP;
6476        } else {
6477                dev_info(&pf->pdev->dev,
6478                         "Query for DCB configuration failed, err %s aq_err %s\n",
6479                         i40e_stat_str(&pf->hw, err),
6480                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6481        }
6482
6483out:
6484        return err;
6485}
6486#endif /* CONFIG_I40E_DCB */
6487#define SPEED_SIZE 14
6488#define FC_SIZE 8
6489/**
6490 * i40e_print_link_message - print link up or down
6491 * @vsi: the VSI for which link needs a message
6492 * @isup: true of link is up, false otherwise
6493 */
6494void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
6495{
6496        enum i40e_aq_link_speed new_speed;
6497        struct i40e_pf *pf = vsi->back;
6498        char *speed = "Unknown";
6499        char *fc = "Unknown";
6500        char *fec = "";
6501        char *req_fec = "";
6502        char *an = "";
6503
6504        if (isup)
6505                new_speed = pf->hw.phy.link_info.link_speed;
6506        else
6507                new_speed = I40E_LINK_SPEED_UNKNOWN;
6508
6509        if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
6510                return;
6511        vsi->current_isup = isup;
6512        vsi->current_speed = new_speed;
6513        if (!isup) {
6514                netdev_info(vsi->netdev, "NIC Link is Down\n");
6515                return;
6516        }
6517
6518        /* Warn user if link speed on NPAR enabled partition is not at
6519         * least 10GB
6520         */
6521        if (pf->hw.func_caps.npar_enable &&
6522            (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
6523             pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
6524                netdev_warn(vsi->netdev,
6525                            "The partition detected link speed that is less than 10Gbps\n");
6526
6527        switch (pf->hw.phy.link_info.link_speed) {
6528        case I40E_LINK_SPEED_40GB:
6529                speed = "40 G";
6530                break;
6531        case I40E_LINK_SPEED_20GB:
6532                speed = "20 G";
6533                break;
6534        case I40E_LINK_SPEED_25GB:
6535                speed = "25 G";
6536                break;
6537        case I40E_LINK_SPEED_10GB:
6538                speed = "10 G";
6539                break;
6540        case I40E_LINK_SPEED_5GB:
6541                speed = "5 G";
6542                break;
6543        case I40E_LINK_SPEED_2_5GB:
6544                speed = "2.5 G";
6545                break;
6546        case I40E_LINK_SPEED_1GB:
6547                speed = "1000 M";
6548                break;
6549        case I40E_LINK_SPEED_100MB:
6550                speed = "100 M";
6551                break;
6552        default:
6553                break;
6554        }
6555
6556        switch (pf->hw.fc.current_mode) {
6557        case I40E_FC_FULL:
6558                fc = "RX/TX";
6559                break;
6560        case I40E_FC_TX_PAUSE:
6561                fc = "TX";
6562                break;
6563        case I40E_FC_RX_PAUSE:
6564                fc = "RX";
6565                break;
6566        default:
6567                fc = "None";
6568                break;
6569        }
6570
6571        if (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
6572                req_fec = "None";
6573                fec = "None";
6574                an = "False";
6575
6576                if (pf->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
6577                        an = "True";
6578
6579                if (pf->hw.phy.link_info.fec_info &
6580                    I40E_AQ_CONFIG_FEC_KR_ENA)
6581                        fec = "CL74 FC-FEC/BASE-R";
6582                else if (pf->hw.phy.link_info.fec_info &
6583                         I40E_AQ_CONFIG_FEC_RS_ENA)
6584                        fec = "CL108 RS-FEC";
6585
6586                /* 'CL108 RS-FEC' should be displayed when RS is requested, or
6587                 * both RS and FC are requested
6588                 */
6589                if (vsi->back->hw.phy.link_info.req_fec_info &
6590                    (I40E_AQ_REQUEST_FEC_KR | I40E_AQ_REQUEST_FEC_RS)) {
6591                        if (vsi->back->hw.phy.link_info.req_fec_info &
6592                            I40E_AQ_REQUEST_FEC_RS)
6593                                req_fec = "CL108 RS-FEC";
6594                        else
6595                                req_fec = "CL74 FC-FEC/BASE-R";
6596                }
6597                netdev_info(vsi->netdev,
6598                            "NIC Link is Up, %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg: %s, Flow Control: %s\n",
6599                            speed, req_fec, fec, an, fc);
6600        } else {
6601                netdev_info(vsi->netdev,
6602                            "NIC Link is Up, %sbps Full Duplex, Flow Control: %s\n",
6603                            speed, fc);
6604        }
6605
6606}
6607
6608/**
6609 * i40e_up_complete - Finish the last steps of bringing up a connection
6610 * @vsi: the VSI being configured
6611 **/
6612static int i40e_up_complete(struct i40e_vsi *vsi)
6613{
6614        struct i40e_pf *pf = vsi->back;
6615        int err;
6616
6617        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6618                i40e_vsi_configure_msix(vsi);
6619        else
6620                i40e_configure_msi_and_legacy(vsi);
6621
6622        /* start rings */
6623        err = i40e_vsi_start_rings(vsi);
6624        if (err)
6625                return err;
6626
6627        clear_bit(__I40E_VSI_DOWN, vsi->state);
6628        i40e_napi_enable_all(vsi);
6629        i40e_vsi_enable_irq(vsi);
6630
6631        if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
6632            (vsi->netdev)) {
6633                i40e_print_link_message(vsi, true);
6634                netif_tx_start_all_queues(vsi->netdev);
6635                netif_carrier_on(vsi->netdev);
6636        }
6637
6638        /* replay FDIR SB filters */
6639        if (vsi->type == I40E_VSI_FDIR) {
6640                /* reset fd counters */
6641                pf->fd_add_err = 0;
6642                pf->fd_atr_cnt = 0;
6643                i40e_fdir_filter_restore(vsi);
6644        }
6645
6646        /* On the next run of the service_task, notify any clients of the new
6647         * opened netdev
6648         */
6649        set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
6650        i40e_service_event_schedule(pf);
6651
6652        return 0;
6653}
6654
6655/**
6656 * i40e_vsi_reinit_locked - Reset the VSI
6657 * @vsi: the VSI being configured
6658 *
6659 * Rebuild the ring structs after some configuration
6660 * has changed, e.g. MTU size.
6661 **/
6662static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
6663{
6664        struct i40e_pf *pf = vsi->back;
6665
6666        WARN_ON(in_interrupt());
6667        while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
6668                usleep_range(1000, 2000);
6669        i40e_down(vsi);
6670
6671        i40e_up(vsi);
6672        clear_bit(__I40E_CONFIG_BUSY, pf->state);
6673}
6674
6675/**
6676 * i40e_up - Bring the connection back up after being down
6677 * @vsi: the VSI being configured
6678 **/
6679int i40e_up(struct i40e_vsi *vsi)
6680{
6681        int err;
6682
6683        err = i40e_vsi_configure(vsi);
6684        if (!err)
6685                err = i40e_up_complete(vsi);
6686
6687        return err;
6688}
6689
6690/**
6691 * i40e_force_link_state - Force the link status
6692 * @pf: board private structure
6693 * @is_up: whether the link state should be forced up or down
6694 **/
6695static i40e_status i40e_force_link_state(struct i40e_pf *pf, bool is_up)
6696{
6697        struct i40e_aq_get_phy_abilities_resp abilities;
6698        struct i40e_aq_set_phy_config config = {0};
6699        struct i40e_hw *hw = &pf->hw;
6700        i40e_status err;
6701        u64 mask;
6702        u8 speed;
6703
6704        /* Card might've been put in an unstable state by other drivers
6705         * and applications, which causes incorrect speed values being
6706         * set on startup. In order to clear speed registers, we call
6707         * get_phy_capabilities twice, once to get initial state of
6708         * available speeds, and once to get current PHY config.
6709         */
6710        err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities,
6711                                           NULL);
6712        if (err) {
6713                dev_err(&pf->pdev->dev,
6714                        "failed to get phy cap., ret =  %s last_status =  %s\n",
6715                        i40e_stat_str(hw, err),
6716                        i40e_aq_str(hw, hw->aq.asq_last_status));
6717                return err;
6718        }
6719        speed = abilities.link_speed;
6720
6721        /* Get the current phy config */
6722        err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities,
6723                                           NULL);
6724        if (err) {
6725                dev_err(&pf->pdev->dev,
6726                        "failed to get phy cap., ret =  %s last_status =  %s\n",
6727                        i40e_stat_str(hw, err),
6728                        i40e_aq_str(hw, hw->aq.asq_last_status));
6729                return err;
6730        }
6731
6732        /* If link needs to go up, but was not forced to go down,
6733         * and its speed values are OK, no need for a flap
6734         */
6735        if (is_up && abilities.phy_type != 0 && abilities.link_speed != 0)
6736                return I40E_SUCCESS;
6737
6738        /* To force link we need to set bits for all supported PHY types,
6739         * but there are now more than 32, so we need to split the bitmap
6740         * across two fields.
6741         */
6742        mask = I40E_PHY_TYPES_BITMASK;
6743        config.phy_type = is_up ? cpu_to_le32((u32)(mask & 0xffffffff)) : 0;
6744        config.phy_type_ext = is_up ? (u8)((mask >> 32) & 0xff) : 0;
6745        /* Copy the old settings, except of phy_type */
6746        config.abilities = abilities.abilities;
6747        if (abilities.link_speed != 0)
6748                config.link_speed = abilities.link_speed;
6749        else
6750                config.link_speed = speed;
6751        config.eee_capability = abilities.eee_capability;
6752        config.eeer = abilities.eeer_val;
6753        config.low_power_ctrl = abilities.d3_lpan;
6754        config.fec_config = abilities.fec_cfg_curr_mod_ext_info &
6755                            I40E_AQ_PHY_FEC_CONFIG_MASK;
6756        err = i40e_aq_set_phy_config(hw, &config, NULL);
6757
6758        if (err) {
6759                dev_err(&pf->pdev->dev,
6760                        "set phy config ret =  %s last_status =  %s\n",
6761                        i40e_stat_str(&pf->hw, err),
6762                        i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6763                return err;
6764        }
6765
6766        /* Update the link info */
6767        err = i40e_update_link_info(hw);
6768        if (err) {
6769                /* Wait a little bit (on 40G cards it sometimes takes a really
6770                 * long time for link to come back from the atomic reset)
6771                 * and try once more
6772                 */
6773                msleep(1000);
6774                i40e_update_link_info(hw);
6775        }
6776
6777        i40e_aq_set_link_restart_an(hw, true, NULL);
6778
6779        return I40E_SUCCESS;
6780}
6781
6782/**
6783 * i40e_down - Shutdown the connection processing
6784 * @vsi: the VSI being stopped
6785 **/
6786void i40e_down(struct i40e_vsi *vsi)
6787{
6788        int i;
6789
6790        /* It is assumed that the caller of this function
6791         * sets the vsi->state __I40E_VSI_DOWN bit.
6792         */
6793        if (vsi->netdev) {
6794                netif_carrier_off(vsi->netdev);
6795                netif_tx_disable(vsi->netdev);
6796        }
6797        i40e_vsi_disable_irq(vsi);
6798        i40e_vsi_stop_rings(vsi);
6799        if (vsi->type == I40E_VSI_MAIN &&
6800            vsi->back->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED)
6801                i40e_force_link_state(vsi->back, false);
6802        i40e_napi_disable_all(vsi);
6803
6804        for (i = 0; i < vsi->num_queue_pairs; i++) {
6805                i40e_clean_tx_ring(vsi->tx_rings[i]);
6806                if (i40e_enabled_xdp_vsi(vsi)) {
6807                        /* Make sure that in-progress ndo_xdp_xmit
6808                         * calls are completed.
6809                         */
6810                        synchronize_rcu();
6811                        i40e_clean_tx_ring(vsi->xdp_rings[i]);
6812                }
6813                i40e_clean_rx_ring(vsi->rx_rings[i]);
6814        }
6815
6816}
6817
6818/**
6819 * i40e_validate_mqprio_qopt- validate queue mapping info
6820 * @vsi: the VSI being configured
6821 * @mqprio_qopt: queue parametrs
6822 **/
6823static int i40e_validate_mqprio_qopt(struct i40e_vsi *vsi,
6824                                     struct tc_mqprio_qopt_offload *mqprio_qopt)
6825{
6826        u64 sum_max_rate = 0;
6827        u64 max_rate = 0;
6828        int i;
6829
6830        if (mqprio_qopt->qopt.offset[0] != 0 ||
6831            mqprio_qopt->qopt.num_tc < 1 ||
6832            mqprio_qopt->qopt.num_tc > I40E_MAX_TRAFFIC_CLASS)
6833                return -EINVAL;
6834        for (i = 0; ; i++) {
6835                if (!mqprio_qopt->qopt.count[i])
6836                        return -EINVAL;
6837                if (mqprio_qopt->min_rate[i]) {
6838                        dev_err(&vsi->back->pdev->dev,
6839                                "Invalid min tx rate (greater than 0) specified\n");
6840                        return -EINVAL;
6841                }
6842                max_rate = mqprio_qopt->max_rate[i];
6843                do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6844                sum_max_rate += max_rate;
6845
6846                if (i >= mqprio_qopt->qopt.num_tc - 1)
6847                        break;
6848                if (mqprio_qopt->qopt.offset[i + 1] !=
6849                    (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
6850                        return -EINVAL;
6851        }
6852        if (vsi->num_queue_pairs <
6853            (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) {
6854                return -EINVAL;
6855        }
6856        if (sum_max_rate > i40e_get_link_speed(vsi)) {
6857                dev_err(&vsi->back->pdev->dev,
6858                        "Invalid max tx rate specified\n");
6859                return -EINVAL;
6860        }
6861        return 0;
6862}
6863
6864/**
6865 * i40e_vsi_set_default_tc_config - set default values for tc configuration
6866 * @vsi: the VSI being configured
6867 **/
6868static void i40e_vsi_set_default_tc_config(struct i40e_vsi *vsi)
6869{
6870        u16 qcount;
6871        int i;
6872
6873        /* Only TC0 is enabled */
6874        vsi->tc_config.numtc = 1;
6875        vsi->tc_config.enabled_tc = 1;
6876        qcount = min_t(int, vsi->alloc_queue_pairs,
6877                       i40e_pf_get_max_q_per_tc(vsi->back));
6878        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6879                /* For the TC that is not enabled set the offset to to default
6880                 * queue and allocate one queue for the given TC.
6881                 */
6882                vsi->tc_config.tc_info[i].qoffset = 0;
6883                if (i == 0)
6884                        vsi->tc_config.tc_info[i].qcount = qcount;
6885                else
6886                        vsi->tc_config.tc_info[i].qcount = 1;
6887                vsi->tc_config.tc_info[i].netdev_tc = 0;
6888        }
6889}
6890
6891/**
6892 * i40e_del_macvlan_filter
6893 * @hw: pointer to the HW structure
6894 * @seid: seid of the channel VSI
6895 * @macaddr: the mac address to apply as a filter
6896 * @aq_err: store the admin Q error
6897 *
6898 * This function deletes a mac filter on the channel VSI which serves as the
6899 * macvlan. Returns 0 on success.
6900 **/
6901static i40e_status i40e_del_macvlan_filter(struct i40e_hw *hw, u16 seid,
6902                                           const u8 *macaddr, int *aq_err)
6903{
6904        struct i40e_aqc_remove_macvlan_element_data element;
6905        i40e_status status;
6906
6907        memset(&element, 0, sizeof(element));
6908        ether_addr_copy(element.mac_addr, macaddr);
6909        element.vlan_tag = 0;
6910        element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
6911        status = i40e_aq_remove_macvlan(hw, seid, &element, 1, NULL);
6912        *aq_err = hw->aq.asq_last_status;
6913
6914        return status;
6915}
6916
6917/**
6918 * i40e_add_macvlan_filter
6919 * @hw: pointer to the HW structure
6920 * @seid: seid of the channel VSI
6921 * @macaddr: the mac address to apply as a filter
6922 * @aq_err: store the admin Q error
6923 *
6924 * This function adds a mac filter on the channel VSI which serves as the
6925 * macvlan. Returns 0 on success.
6926 **/
6927static i40e_status i40e_add_macvlan_filter(struct i40e_hw *hw, u16 seid,
6928                                           const u8 *macaddr, int *aq_err)
6929{
6930        struct i40e_aqc_add_macvlan_element_data element;
6931        i40e_status status;
6932        u16 cmd_flags = 0;
6933
6934        ether_addr_copy(element.mac_addr, macaddr);
6935        element.vlan_tag = 0;
6936        element.queue_number = 0;
6937        element.match_method = I40E_AQC_MM_ERR_NO_RES;
6938        cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
6939        element.flags = cpu_to_le16(cmd_flags);
6940        status = i40e_aq_add_macvlan(hw, seid, &element, 1, NULL);
6941        *aq_err = hw->aq.asq_last_status;
6942
6943        return status;
6944}
6945
6946/**
6947 * i40e_reset_ch_rings - Reset the queue contexts in a channel
6948 * @vsi: the VSI we want to access
6949 * @ch: the channel we want to access
6950 */
6951static void i40e_reset_ch_rings(struct i40e_vsi *vsi, struct i40e_channel *ch)
6952{
6953        struct i40e_ring *tx_ring, *rx_ring;
6954        u16 pf_q;
6955        int i;
6956
6957        for (i = 0; i < ch->num_queue_pairs; i++) {
6958                pf_q = ch->base_queue + i;
6959                tx_ring = vsi->tx_rings[pf_q];
6960                tx_ring->ch = NULL;
6961                rx_ring = vsi->rx_rings[pf_q];
6962                rx_ring->ch = NULL;
6963        }
6964}
6965
6966/**
6967 * i40e_free_macvlan_channels
6968 * @vsi: the VSI we want to access
6969 *
6970 * This function frees the Qs of the channel VSI from
6971 * the stack and also deletes the channel VSIs which
6972 * serve as macvlans.
6973 */
6974static void i40e_free_macvlan_channels(struct i40e_vsi *vsi)
6975{
6976        struct i40e_channel *ch, *ch_tmp;
6977        int ret;
6978
6979        if (list_empty(&vsi->macvlan_list))
6980                return;
6981
6982        list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
6983                struct i40e_vsi *parent_vsi;
6984
6985                if (i40e_is_channel_macvlan(ch)) {
6986                        i40e_reset_ch_rings(vsi, ch);
6987                        clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
6988                        netdev_unbind_sb_channel(vsi->netdev, ch->fwd->netdev);
6989                        netdev_set_sb_channel(ch->fwd->netdev, 0);
6990                        kfree(ch->fwd);
6991                        ch->fwd = NULL;
6992                }
6993
6994                list_del(&ch->list);
6995                parent_vsi = ch->parent_vsi;
6996                if (!parent_vsi || !ch->initialized) {
6997                        kfree(ch);
6998                        continue;
6999                }
7000
7001                /* remove the VSI */
7002                ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
7003                                             NULL);
7004                if (ret)
7005                        dev_err(&vsi->back->pdev->dev,
7006                                "unable to remove channel (%d) for parent VSI(%d)\n",
7007                                ch->seid, parent_vsi->seid);
7008                kfree(ch);
7009        }
7010        vsi->macvlan_cnt = 0;
7011}
7012
7013/**
7014 * i40e_fwd_ring_up - bring the macvlan device up
7015 * @vsi: the VSI we want to access
7016 * @vdev: macvlan netdevice
7017 * @fwd: the private fwd structure
7018 */
7019static int i40e_fwd_ring_up(struct i40e_vsi *vsi, struct net_device *vdev,
7020                            struct i40e_fwd_adapter *fwd)
7021{
7022        int ret = 0, num_tc = 1,  i, aq_err;
7023        struct i40e_channel *ch, *ch_tmp;
7024        struct i40e_pf *pf = vsi->back;
7025        struct i40e_hw *hw = &pf->hw;
7026
7027        if (list_empty(&vsi->macvlan_list))
7028                return -EINVAL;
7029
7030        /* Go through the list and find an available channel */
7031        list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7032                if (!i40e_is_channel_macvlan(ch)) {
7033                        ch->fwd = fwd;
7034                        /* record configuration for macvlan interface in vdev */
7035                        for (i = 0; i < num_tc; i++)
7036                                netdev_bind_sb_channel_queue(vsi->netdev, vdev,
7037                                                             i,
7038                                                             ch->num_queue_pairs,
7039                                                             ch->base_queue);
7040                        for (i = 0; i < ch->num_queue_pairs; i++) {
7041                                struct i40e_ring *tx_ring, *rx_ring;
7042                                u16 pf_q;
7043
7044                                pf_q = ch->base_queue + i;
7045
7046                                /* Get to TX ring ptr */
7047                                tx_ring = vsi->tx_rings[pf_q];
7048                                tx_ring->ch = ch;
7049
7050                                /* Get the RX ring ptr */
7051                                rx_ring = vsi->rx_rings[pf_q];
7052                                rx_ring->ch = ch;
7053                        }
7054                        break;
7055                }
7056        }
7057
7058        /* Guarantee all rings are updated before we update the
7059         * MAC address filter.
7060         */
7061        wmb();
7062
7063        /* Add a mac filter */
7064        ret = i40e_add_macvlan_filter(hw, ch->seid, vdev->dev_addr, &aq_err);
7065        if (ret) {
7066                /* if we cannot add the MAC rule then disable the offload */
7067                macvlan_release_l2fw_offload(vdev);
7068                for (i = 0; i < ch->num_queue_pairs; i++) {
7069                        struct i40e_ring *rx_ring;
7070                        u16 pf_q;
7071
7072                        pf_q = ch->base_queue + i;
7073                        rx_ring = vsi->rx_rings[pf_q];
7074                        rx_ring->netdev = NULL;
7075                }
7076                dev_info(&pf->pdev->dev,
7077                         "Error adding mac filter on macvlan err %s, aq_err %s\n",
7078                          i40e_stat_str(hw, ret),
7079                          i40e_aq_str(hw, aq_err));
7080                netdev_err(vdev, "L2fwd offload disabled to L2 filter error\n");
7081        }
7082
7083        return ret;
7084}
7085
7086/**
7087 * i40e_setup_macvlans - create the channels which will be macvlans
7088 * @vsi: the VSI we want to access
7089 * @macvlan_cnt: no. of macvlans to be setup
7090 * @qcnt: no. of Qs per macvlan
7091 * @vdev: macvlan netdevice
7092 */
7093static int i40e_setup_macvlans(struct i40e_vsi *vsi, u16 macvlan_cnt, u16 qcnt,
7094                               struct net_device *vdev)
7095{
7096        struct i40e_pf *pf = vsi->back;
7097        struct i40e_hw *hw = &pf->hw;
7098        struct i40e_vsi_context ctxt;
7099        u16 sections, qmap, num_qps;
7100        struct i40e_channel *ch;
7101        int i, pow, ret = 0;
7102        u8 offset = 0;
7103
7104        if (vsi->type != I40E_VSI_MAIN || !macvlan_cnt)
7105                return -EINVAL;
7106
7107        num_qps = vsi->num_queue_pairs - (macvlan_cnt * qcnt);
7108
7109        /* find the next higher power-of-2 of num queue pairs */
7110        pow = fls(roundup_pow_of_two(num_qps) - 1);
7111
7112        qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
7113                (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
7114
7115        /* Setup context bits for the main VSI */
7116        sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
7117        sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
7118        memset(&ctxt, 0, sizeof(ctxt));
7119        ctxt.seid = vsi->seid;
7120        ctxt.pf_num = vsi->back->hw.pf_id;
7121        ctxt.vf_num = 0;
7122        ctxt.uplink_seid = vsi->uplink_seid;
7123        ctxt.info = vsi->info;
7124        ctxt.info.tc_mapping[0] = cpu_to_le16(qmap);
7125        ctxt.info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
7126        ctxt.info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
7127        ctxt.info.valid_sections |= cpu_to_le16(sections);
7128
7129        /* Reconfigure RSS for main VSI with new max queue count */
7130        vsi->rss_size = max_t(u16, num_qps, qcnt);
7131        ret = i40e_vsi_config_rss(vsi);
7132        if (ret) {
7133                dev_info(&pf->pdev->dev,
7134                         "Failed to reconfig RSS for num_queues (%u)\n",
7135                         vsi->rss_size);
7136                return ret;
7137        }
7138        vsi->reconfig_rss = true;
7139        dev_dbg(&vsi->back->pdev->dev,
7140                "Reconfigured RSS with num_queues (%u)\n", vsi->rss_size);
7141        vsi->next_base_queue = num_qps;
7142        vsi->cnt_q_avail = vsi->num_queue_pairs - num_qps;
7143
7144        /* Update the VSI after updating the VSI queue-mapping
7145         * information
7146         */
7147        ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
7148        if (ret) {
7149                dev_info(&pf->pdev->dev,
7150                         "Update vsi tc config failed, err %s aq_err %s\n",
7151                         i40e_stat_str(hw, ret),
7152                         i40e_aq_str(hw, hw->aq.asq_last_status));
7153                return ret;
7154        }
7155        /* update the local VSI info with updated queue map */
7156        i40e_vsi_update_queue_map(vsi, &ctxt);
7157        vsi->info.valid_sections = 0;
7158
7159        /* Create channels for macvlans */
7160        INIT_LIST_HEAD(&vsi->macvlan_list);
7161        for (i = 0; i < macvlan_cnt; i++) {
7162                ch = kzalloc(sizeof(*ch), GFP_KERNEL);
7163                if (!ch) {
7164                        ret = -ENOMEM;
7165                        goto err_free;
7166                }
7167                INIT_LIST_HEAD(&ch->list);
7168                ch->num_queue_pairs = qcnt;
7169                if (!i40e_setup_channel(pf, vsi, ch)) {
7170                        ret = -EINVAL;
7171                        goto err_free;
7172                }
7173                ch->parent_vsi = vsi;
7174                vsi->cnt_q_avail -= ch->num_queue_pairs;
7175                vsi->macvlan_cnt++;
7176                list_add_tail(&ch->list, &vsi->macvlan_list);
7177        }
7178
7179        return ret;
7180
7181err_free:
7182        dev_info(&pf->pdev->dev, "Failed to setup macvlans\n");
7183        i40e_free_macvlan_channels(vsi);
7184
7185        return ret;
7186}
7187
7188/**
7189 * i40e_fwd_add - configure macvlans
7190 * @netdev: net device to configure
7191 * @vdev: macvlan netdevice
7192 **/
7193static void *i40e_fwd_add(struct net_device *netdev, struct net_device *vdev)
7194{
7195        struct i40e_netdev_priv *np = netdev_priv(netdev);
7196        u16 q_per_macvlan = 0, macvlan_cnt = 0, vectors;
7197        struct i40e_vsi *vsi = np->vsi;
7198        struct i40e_pf *pf = vsi->back;
7199        struct i40e_fwd_adapter *fwd;
7200        int avail_macvlan, ret;
7201
7202        if ((pf->flags & I40E_FLAG_DCB_ENABLED)) {
7203                netdev_info(netdev, "Macvlans are not supported when DCB is enabled\n");
7204                return ERR_PTR(-EINVAL);
7205        }
7206        if ((pf->flags & I40E_FLAG_TC_MQPRIO)) {
7207                netdev_info(netdev, "Macvlans are not supported when HW TC offload is on\n");
7208                return ERR_PTR(-EINVAL);
7209        }
7210        if (pf->num_lan_msix < I40E_MIN_MACVLAN_VECTORS) {
7211                netdev_info(netdev, "Not enough vectors available to support macvlans\n");
7212                return ERR_PTR(-EINVAL);
7213        }
7214
7215        /* The macvlan device has to be a single Q device so that the
7216         * tc_to_txq field can be reused to pick the tx queue.
7217         */
7218        if (netif_is_multiqueue(vdev))
7219                return ERR_PTR(-ERANGE);
7220
7221        if (!vsi->macvlan_cnt) {
7222                /* reserve bit 0 for the pf device */
7223                set_bit(0, vsi->fwd_bitmask);
7224
7225                /* Try to reserve as many queues as possible for macvlans. First
7226                 * reserve 3/4th of max vectors, then half, then quarter and
7227                 * calculate Qs per macvlan as you go
7228                 */
7229                vectors = pf->num_lan_msix;
7230                if (vectors <= I40E_MAX_MACVLANS && vectors > 64) {
7231                        /* allocate 4 Qs per macvlan and 32 Qs to the PF*/
7232                        q_per_macvlan = 4;
7233                        macvlan_cnt = (vectors - 32) / 4;
7234                } else if (vectors <= 64 && vectors > 32) {
7235                        /* allocate 2 Qs per macvlan and 16 Qs to the PF*/
7236                        q_per_macvlan = 2;
7237                        macvlan_cnt = (vectors - 16) / 2;
7238                } else if (vectors <= 32 && vectors > 16) {
7239                        /* allocate 1 Q per macvlan and 16 Qs to the PF*/
7240                        q_per_macvlan = 1;
7241                        macvlan_cnt = vectors - 16;
7242                } else if (vectors <= 16 && vectors > 8) {
7243                        /* allocate 1 Q per macvlan and 8 Qs to the PF */
7244                        q_per_macvlan = 1;
7245                        macvlan_cnt = vectors - 8;
7246                } else {
7247                        /* allocate 1 Q per macvlan and 1 Q to the PF */
7248                        q_per_macvlan = 1;
7249                        macvlan_cnt = vectors - 1;
7250                }
7251
7252                if (macvlan_cnt == 0)
7253                        return ERR_PTR(-EBUSY);
7254
7255                /* Quiesce VSI queues */
7256                i40e_quiesce_vsi(vsi);
7257
7258                /* sets up the macvlans but does not "enable" them */
7259                ret = i40e_setup_macvlans(vsi, macvlan_cnt, q_per_macvlan,
7260                                          vdev);
7261                if (ret)
7262                        return ERR_PTR(ret);
7263
7264                /* Unquiesce VSI */
7265                i40e_unquiesce_vsi(vsi);
7266        }
7267        avail_macvlan = find_first_zero_bit(vsi->fwd_bitmask,
7268                                            vsi->macvlan_cnt);
7269        if (avail_macvlan >= I40E_MAX_MACVLANS)
7270                return ERR_PTR(-EBUSY);
7271
7272        /* create the fwd struct */
7273        fwd = kzalloc(sizeof(*fwd), GFP_KERNEL);
7274        if (!fwd)
7275                return ERR_PTR(-ENOMEM);
7276
7277        set_bit(avail_macvlan, vsi->fwd_bitmask);
7278        fwd->bit_no = avail_macvlan;
7279        netdev_set_sb_channel(vdev, avail_macvlan);
7280        fwd->netdev = vdev;
7281
7282        if (!netif_running(netdev))
7283                return fwd;
7284
7285        /* Set fwd ring up */
7286        ret = i40e_fwd_ring_up(vsi, vdev, fwd);
7287        if (ret) {
7288                /* unbind the queues and drop the subordinate channel config */
7289                netdev_unbind_sb_channel(netdev, vdev);
7290                netdev_set_sb_channel(vdev, 0);
7291
7292                kfree(fwd);
7293                return ERR_PTR(-EINVAL);
7294        }
7295
7296        return fwd;
7297}
7298
7299/**
7300 * i40e_del_all_macvlans - Delete all the mac filters on the channels
7301 * @vsi: the VSI we want to access
7302 */
7303static void i40e_del_all_macvlans(struct i40e_vsi *vsi)
7304{
7305        struct i40e_channel *ch, *ch_tmp;
7306        struct i40e_pf *pf = vsi->back;
7307        struct i40e_hw *hw = &pf->hw;
7308        int aq_err, ret = 0;
7309
7310        if (list_empty(&vsi->macvlan_list))
7311                return;
7312
7313        list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7314                if (i40e_is_channel_macvlan(ch)) {
7315                        ret = i40e_del_macvlan_filter(hw, ch->seid,
7316                                                      i40e_channel_mac(ch),
7317                                                      &aq_err);
7318                        if (!ret) {
7319                                /* Reset queue contexts */
7320                                i40e_reset_ch_rings(vsi, ch);
7321                                clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7322                                netdev_unbind_sb_channel(vsi->netdev,
7323                                                         ch->fwd->netdev);
7324                                netdev_set_sb_channel(ch->fwd->netdev, 0);
7325                                kfree(ch->fwd);
7326                                ch->fwd = NULL;
7327                        }
7328                }
7329        }
7330}
7331
7332/**
7333 * i40e_fwd_del - delete macvlan interfaces
7334 * @netdev: net device to configure
7335 * @vdev: macvlan netdevice
7336 */
7337static void i40e_fwd_del(struct net_device *netdev, void *vdev)
7338{
7339        struct i40e_netdev_priv *np = netdev_priv(netdev);
7340        struct i40e_fwd_adapter *fwd = vdev;
7341        struct i40e_channel *ch, *ch_tmp;
7342        struct i40e_vsi *vsi = np->vsi;
7343        struct i40e_pf *pf = vsi->back;
7344        struct i40e_hw *hw = &pf->hw;
7345        int aq_err, ret = 0;
7346
7347        /* Find the channel associated with the macvlan and del mac filter */
7348        list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7349                if (i40e_is_channel_macvlan(ch) &&
7350                    ether_addr_equal(i40e_channel_mac(ch),
7351                                     fwd->netdev->dev_addr)) {
7352                        ret = i40e_del_macvlan_filter(hw, ch->seid,
7353                                                      i40e_channel_mac(ch),
7354                                                      &aq_err);
7355                        if (!ret) {
7356                                /* Reset queue contexts */
7357                                i40e_reset_ch_rings(vsi, ch);
7358                                clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7359                                netdev_unbind_sb_channel(netdev, fwd->netdev);
7360                                netdev_set_sb_channel(fwd->netdev, 0);
7361                                kfree(ch->fwd);
7362                                ch->fwd = NULL;
7363                        } else {
7364                                dev_info(&pf->pdev->dev,
7365                                         "Error deleting mac filter on macvlan err %s, aq_err %s\n",
7366                                          i40e_stat_str(hw, ret),
7367                                          i40e_aq_str(hw, aq_err));
7368                        }
7369                        break;
7370                }
7371        }
7372}
7373
7374/**
7375 * i40e_setup_tc - configure multiple traffic classes
7376 * @netdev: net device to configure
7377 * @type_data: tc offload data
7378 **/
7379static int i40e_setup_tc(struct net_device *netdev, void *type_data)
7380{
7381        struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
7382        struct i40e_netdev_priv *np = netdev_priv(netdev);
7383        struct i40e_vsi *vsi = np->vsi;
7384        struct i40e_pf *pf = vsi->back;
7385        u8 enabled_tc = 0, num_tc, hw;
7386        bool need_reset = false;
7387        int old_queue_pairs;
7388        int ret = -EINVAL;
7389        u16 mode;
7390        int i;
7391
7392        old_queue_pairs = vsi->num_queue_pairs;
7393        num_tc = mqprio_qopt->qopt.num_tc;
7394        hw = mqprio_qopt->qopt.hw;
7395        mode = mqprio_qopt->mode;
7396        if (!hw) {
7397                pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7398                memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
7399                goto config_tc;
7400        }
7401
7402        /* Check if MFP enabled */
7403        if (pf->flags & I40E_FLAG_MFP_ENABLED) {
7404                netdev_info(netdev,
7405                            "Configuring TC not supported in MFP mode\n");
7406                return ret;
7407        }
7408        switch (mode) {
7409        case TC_MQPRIO_MODE_DCB:
7410                pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7411
7412                /* Check if DCB enabled to continue */
7413                if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
7414                        netdev_info(netdev,
7415                                    "DCB is not enabled for adapter\n");
7416                        return ret;
7417                }
7418
7419                /* Check whether tc count is within enabled limit */
7420                if (num_tc > i40e_pf_get_num_tc(pf)) {
7421                        netdev_info(netdev,
7422                                    "TC count greater than enabled on link for adapter\n");
7423                        return ret;
7424                }
7425                break;
7426        case TC_MQPRIO_MODE_CHANNEL:
7427                if (pf->flags & I40E_FLAG_DCB_ENABLED) {
7428                        netdev_info(netdev,
7429                                    "Full offload of TC Mqprio options is not supported when DCB is enabled\n");
7430                        return ret;
7431                }
7432                if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7433                        return ret;
7434                ret = i40e_validate_mqprio_qopt(vsi, mqprio_qopt);
7435                if (ret)
7436                        return ret;
7437                memcpy(&vsi->mqprio_qopt, mqprio_qopt,
7438                       sizeof(*mqprio_qopt));
7439                pf->flags |= I40E_FLAG_TC_MQPRIO;
7440                pf->flags &= ~I40E_FLAG_DCB_ENABLED;
7441                break;
7442        default:
7443                return -EINVAL;
7444        }
7445
7446config_tc:
7447        /* Generate TC map for number of tc requested */
7448        for (i = 0; i < num_tc; i++)
7449                enabled_tc |= BIT(i);
7450
7451        /* Requesting same TC configuration as already enabled */
7452        if (enabled_tc == vsi->tc_config.enabled_tc &&
7453            mode != TC_MQPRIO_MODE_CHANNEL)
7454                return 0;
7455
7456        /* Quiesce VSI queues */
7457        i40e_quiesce_vsi(vsi);
7458
7459        if (!hw && !(pf->flags & I40E_FLAG_TC_MQPRIO))
7460                i40e_remove_queue_channels(vsi);
7461
7462        /* Configure VSI for enabled TCs */
7463        ret = i40e_vsi_config_tc(vsi, enabled_tc);
7464        if (ret) {
7465                netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
7466                            vsi->seid);
7467                need_reset = true;
7468                goto exit;
7469        } else {
7470                dev_info(&vsi->back->pdev->dev,
7471                         "Setup channel (id:%u) utilizing num_queues %d\n",
7472                         vsi->seid, vsi->tc_config.tc_info[0].qcount);
7473        }
7474
7475        if (pf->flags & I40E_FLAG_TC_MQPRIO) {
7476                if (vsi->mqprio_qopt.max_rate[0]) {
7477                        u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
7478
7479                        do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
7480                        ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
7481                        if (!ret) {
7482                                u64 credits = max_tx_rate;
7483
7484                                do_div(credits, I40E_BW_CREDIT_DIVISOR);
7485                                dev_dbg(&vsi->back->pdev->dev,
7486                                        "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
7487                                        max_tx_rate,
7488                                        credits,
7489                                        vsi->seid);
7490                        } else {
7491                                need_reset = true;
7492                                goto exit;
7493                        }
7494                }
7495                ret = i40e_configure_queue_channels(vsi);
7496                if (ret) {
7497                        vsi->num_queue_pairs = old_queue_pairs;
7498                        netdev_info(netdev,
7499                                    "Failed configuring queue channels\n");
7500                        need_reset = true;
7501                        goto exit;
7502                }
7503        }
7504
7505exit:
7506        /* Reset the configuration data to defaults, only TC0 is enabled */
7507        if (need_reset) {
7508                i40e_vsi_set_default_tc_config(vsi);
7509                need_reset = false;
7510        }
7511
7512        /* Unquiesce VSI */
7513        i40e_unquiesce_vsi(vsi);
7514        return ret;
7515}
7516
7517/**
7518 * i40e_set_cld_element - sets cloud filter element data
7519 * @filter: cloud filter rule
7520 * @cld: ptr to cloud filter element data
7521 *
7522 * This is helper function to copy data into cloud filter element
7523 **/
7524static inline void
7525i40e_set_cld_element(struct i40e_cloud_filter *filter,
7526                     struct i40e_aqc_cloud_filters_element_data *cld)
7527{
7528        int i, j;
7529        u32 ipa;
7530
7531        memset(cld, 0, sizeof(*cld));
7532        ether_addr_copy(cld->outer_mac, filter->dst_mac);
7533        ether_addr_copy(cld->inner_mac, filter->src_mac);
7534
7535        if (filter->n_proto != ETH_P_IP && filter->n_proto != ETH_P_IPV6)
7536                return;
7537
7538        if (filter->n_proto == ETH_P_IPV6) {
7539#define IPV6_MAX_INDEX  (ARRAY_SIZE(filter->dst_ipv6) - 1)
7540                for (i = 0, j = 0; i < ARRAY_SIZE(filter->dst_ipv6);
7541                     i++, j += 2) {
7542                        ipa = be32_to_cpu(filter->dst_ipv6[IPV6_MAX_INDEX - i]);
7543                        ipa = cpu_to_le32(ipa);
7544                        memcpy(&cld->ipaddr.raw_v6.data[j], &ipa, sizeof(ipa));
7545                }
7546        } else {
7547                ipa = be32_to_cpu(filter->dst_ipv4);
7548                memcpy(&cld->ipaddr.v4.data, &ipa, sizeof(ipa));
7549        }
7550
7551        cld->inner_vlan = cpu_to_le16(ntohs(filter->vlan_id));
7552
7553        /* tenant_id is not supported by FW now, once the support is enabled
7554         * fill the cld->tenant_id with cpu_to_le32(filter->tenant_id)
7555         */
7556        if (filter->tenant_id)
7557                return;
7558}
7559
7560/**
7561 * i40e_add_del_cloud_filter - Add/del cloud filter
7562 * @vsi: pointer to VSI
7563 * @filter: cloud filter rule
7564 * @add: if true, add, if false, delete
7565 *
7566 * Add or delete a cloud filter for a specific flow spec.
7567 * Returns 0 if the filter were successfully added.
7568 **/
7569int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
7570                              struct i40e_cloud_filter *filter, bool add)
7571{
7572        struct i40e_aqc_cloud_filters_element_data cld_filter;
7573        struct i40e_pf *pf = vsi->back;
7574        int ret;
7575        static const u16 flag_table[128] = {
7576                [I40E_CLOUD_FILTER_FLAGS_OMAC]  =
7577                        I40E_AQC_ADD_CLOUD_FILTER_OMAC,
7578                [I40E_CLOUD_FILTER_FLAGS_IMAC]  =
7579                        I40E_AQC_ADD_CLOUD_FILTER_IMAC,
7580                [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN]  =
7581                        I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN,
7582                [I40E_CLOUD_FILTER_FLAGS_IMAC_TEN_ID] =
7583                        I40E_AQC_ADD_CLOUD_FILTER_IMAC_TEN_ID,
7584                [I40E_CLOUD_FILTER_FLAGS_OMAC_TEN_ID_IMAC] =
7585                        I40E_AQC_ADD_CLOUD_FILTER_OMAC_TEN_ID_IMAC,
7586                [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN_TEN_ID] =
7587                        I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN_TEN_ID,
7588                [I40E_CLOUD_FILTER_FLAGS_IIP] =
7589                        I40E_AQC_ADD_CLOUD_FILTER_IIP,
7590        };
7591
7592        if (filter->flags >= ARRAY_SIZE(flag_table))
7593                return I40E_ERR_CONFIG;
7594
7595        /* copy element needed to add cloud filter from filter */
7596        i40e_set_cld_element(filter, &cld_filter);
7597
7598        if (filter->tunnel_type != I40E_CLOUD_TNL_TYPE_NONE)
7599                cld_filter.flags = cpu_to_le16(filter->tunnel_type <<
7600                                             I40E_AQC_ADD_CLOUD_TNL_TYPE_SHIFT);
7601
7602        if (filter->n_proto == ETH_P_IPV6)
7603                cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7604                                                I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7605        else
7606                cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7607                                                I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7608
7609        if (add)
7610                ret = i40e_aq_add_cloud_filters(&pf->hw, filter->seid,
7611                                                &cld_filter, 1);
7612        else
7613                ret = i40e_aq_rem_cloud_filters(&pf->hw, filter->seid,
7614                                                &cld_filter, 1);
7615        if (ret)
7616                dev_dbg(&pf->pdev->dev,
7617                        "Failed to %s cloud filter using l4 port %u, err %d aq_err %d\n",
7618                        add ? "add" : "delete", filter->dst_port, ret,
7619                        pf->hw.aq.asq_last_status);
7620        else
7621                dev_info(&pf->pdev->dev,
7622                         "%s cloud filter for VSI: %d\n",
7623                         add ? "Added" : "Deleted", filter->seid);
7624        return ret;
7625}
7626
7627/**
7628 * i40e_add_del_cloud_filter_big_buf - Add/del cloud filter using big_buf
7629 * @vsi: pointer to VSI
7630 * @filter: cloud filter rule
7631 * @add: if true, add, if false, delete
7632 *
7633 * Add or delete a cloud filter for a specific flow spec using big buffer.
7634 * Returns 0 if the filter were successfully added.
7635 **/
7636int i40e_add_del_cloud_filter_big_buf(struct i40e_vsi *vsi,
7637                                      struct i40e_cloud_filter *filter,
7638                                      bool add)
7639{
7640        struct i40e_aqc_cloud_filters_element_bb cld_filter;
7641        struct i40e_pf *pf = vsi->back;
7642        int ret;
7643
7644        /* Both (src/dst) valid mac_addr are not supported */
7645        if ((is_valid_ether_addr(filter->dst_mac) &&
7646             is_valid_ether_addr(filter->src_mac)) ||
7647            (is_multicast_ether_addr(filter->dst_mac) &&
7648             is_multicast_ether_addr(filter->src_mac)))
7649                return -EOPNOTSUPP;
7650
7651        /* Big buffer cloud filter needs 'L4 port' to be non-zero. Also, UDP
7652         * ports are not supported via big buffer now.
7653         */
7654        if (!filter->dst_port || filter->ip_proto == IPPROTO_UDP)
7655                return -EOPNOTSUPP;
7656
7657        /* adding filter using src_port/src_ip is not supported at this stage */
7658        if (filter->src_port || filter->src_ipv4 ||
7659            !ipv6_addr_any(&filter->ip.v6.src_ip6))
7660                return -EOPNOTSUPP;
7661
7662        /* copy element needed to add cloud filter from filter */
7663        i40e_set_cld_element(filter, &cld_filter.element);
7664
7665        if (is_valid_ether_addr(filter->dst_mac) ||
7666            is_valid_ether_addr(filter->src_mac) ||
7667            is_multicast_ether_addr(filter->dst_mac) ||
7668            is_multicast_ether_addr(filter->src_mac)) {
7669                /* MAC + IP : unsupported mode */
7670                if (filter->dst_ipv4)
7671                        return -EOPNOTSUPP;
7672
7673                /* since we validated that L4 port must be valid before
7674                 * we get here, start with respective "flags" value
7675                 * and update if vlan is present or not
7676                 */
7677                cld_filter.element.flags =
7678                        cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_PORT);
7679
7680                if (filter->vlan_id) {
7681                        cld_filter.element.flags =
7682                        cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_VLAN_PORT);
7683                }
7684
7685        } else if (filter->dst_ipv4 ||
7686                   !ipv6_addr_any(&filter->ip.v6.dst_ip6)) {
7687                cld_filter.element.flags =
7688                                cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_IP_PORT);
7689                if (filter->n_proto == ETH_P_IPV6)
7690                        cld_filter.element.flags |=
7691                                cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7692                else
7693                        cld_filter.element.flags |=
7694                                cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7695        } else {
7696                dev_err(&pf->pdev->dev,
7697                        "either mac or ip has to be valid for cloud filter\n");
7698                return -EINVAL;
7699        }
7700
7701        /* Now copy L4 port in Byte 6..7 in general fields */
7702        cld_filter.general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X16_WORD0] =
7703                                                be16_to_cpu(filter->dst_port);
7704
7705        if (add) {
7706                /* Validate current device switch mode, change if necessary */
7707                ret = i40e_validate_and_set_switch_mode(vsi);
7708                if (ret) {
7709                        dev_err(&pf->pdev->dev,
7710                                "failed to set switch mode, ret %d\n",
7711                                ret);
7712                        return ret;
7713                }
7714
7715                ret = i40e_aq_add_cloud_filters_bb(&pf->hw, filter->seid,
7716                                                   &cld_filter, 1);
7717        } else {
7718                ret = i40e_aq_rem_cloud_filters_bb(&pf->hw, filter->seid,
7719                                                   &cld_filter, 1);
7720        }
7721
7722        if (ret)
7723                dev_dbg(&pf->pdev->dev,
7724                        "Failed to %s cloud filter(big buffer) err %d aq_err %d\n",
7725                        add ? "add" : "delete", ret, pf->hw.aq.asq_last_status);
7726        else
7727                dev_info(&pf->pdev->dev,
7728                         "%s cloud filter for VSI: %d, L4 port: %d\n",
7729                         add ? "add" : "delete", filter->seid,
7730                         ntohs(filter->dst_port));
7731        return ret;
7732}
7733
7734/**
7735 * i40e_parse_cls_flower - Parse tc flower filters provided by kernel
7736 * @vsi: Pointer to VSI
7737 * @cls_flower: Pointer to struct flow_cls_offload
7738 * @filter: Pointer to cloud filter structure
7739 *
7740 **/
7741static int i40e_parse_cls_flower(struct i40e_vsi *vsi,
7742                                 struct flow_cls_offload *f,
7743                                 struct i40e_cloud_filter *filter)
7744{
7745        struct flow_rule *rule = flow_cls_offload_flow_rule(f);
7746        struct flow_dissector *dissector = rule->match.dissector;
7747        u16 n_proto_mask = 0, n_proto_key = 0, addr_type = 0;
7748        struct i40e_pf *pf = vsi->back;
7749        u8 field_flags = 0;
7750
7751        if (dissector->used_keys &
7752            ~(BIT(FLOW_DISSECTOR_KEY_CONTROL) |
7753              BIT(FLOW_DISSECTOR_KEY_BASIC) |
7754              BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
7755              BIT(FLOW_DISSECTOR_KEY_VLAN) |
7756              BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
7757              BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
7758              BIT(FLOW_DISSECTOR_KEY_PORTS) |
7759              BIT(FLOW_DISSECTOR_KEY_ENC_KEYID))) {
7760                dev_err(&pf->pdev->dev, "Unsupported key used: 0x%x\n",
7761                        dissector->used_keys);
7762                return -EOPNOTSUPP;
7763        }
7764
7765        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
7766                struct flow_match_enc_keyid match;
7767
7768                flow_rule_match_enc_keyid(rule, &match);
7769                if (match.mask->keyid != 0)
7770                        field_flags |= I40E_CLOUD_FIELD_TEN_ID;
7771
7772                filter->tenant_id = be32_to_cpu(match.key->keyid);
7773        }
7774
7775        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
7776                struct flow_match_basic match;
7777
7778                flow_rule_match_basic(rule, &match);
7779                n_proto_key = ntohs(match.key->n_proto);
7780                n_proto_mask = ntohs(match.mask->n_proto);
7781
7782                if (n_proto_key == ETH_P_ALL) {
7783                        n_proto_key = 0;
7784                        n_proto_mask = 0;
7785                }
7786                filter->n_proto = n_proto_key & n_proto_mask;
7787                filter->ip_proto = match.key->ip_proto;
7788        }
7789
7790        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
7791                struct flow_match_eth_addrs match;
7792
7793                flow_rule_match_eth_addrs(rule, &match);
7794
7795                /* use is_broadcast and is_zero to check for all 0xf or 0 */
7796                if (!is_zero_ether_addr(match.mask->dst)) {
7797                        if (is_broadcast_ether_addr(match.mask->dst)) {
7798                                field_flags |= I40E_CLOUD_FIELD_OMAC;
7799                        } else {
7800                                dev_err(&pf->pdev->dev, "Bad ether dest mask %pM\n",
7801                                        match.mask->dst);
7802                                return I40E_ERR_CONFIG;
7803                        }
7804                }
7805
7806                if (!is_zero_ether_addr(match.mask->src)) {
7807                        if (is_broadcast_ether_addr(match.mask->src)) {
7808                                field_flags |= I40E_CLOUD_FIELD_IMAC;
7809                        } else {
7810                                dev_err(&pf->pdev->dev, "Bad ether src mask %pM\n",
7811                                        match.mask->src);
7812                                return I40E_ERR_CONFIG;
7813                        }
7814                }
7815                ether_addr_copy(filter->dst_mac, match.key->dst);
7816                ether_addr_copy(filter->src_mac, match.key->src);
7817        }
7818
7819        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
7820                struct flow_match_vlan match;
7821
7822                flow_rule_match_vlan(rule, &match);
7823                if (match.mask->vlan_id) {
7824                        if (match.mask->vlan_id == VLAN_VID_MASK) {
7825                                field_flags |= I40E_CLOUD_FIELD_IVLAN;
7826
7827                        } else {
7828                                dev_err(&pf->pdev->dev, "Bad vlan mask 0x%04x\n",
7829                                        match.mask->vlan_id);
7830                                return I40E_ERR_CONFIG;
7831                        }
7832                }
7833
7834                filter->vlan_id = cpu_to_be16(match.key->vlan_id);
7835        }
7836
7837        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
7838                struct flow_match_control match;
7839
7840                flow_rule_match_control(rule, &match);
7841                addr_type = match.key->addr_type;
7842        }
7843
7844        if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
7845                struct flow_match_ipv4_addrs match;
7846
7847                flow_rule_match_ipv4_addrs(rule, &match);
7848                if (match.mask->dst) {
7849                        if (match.mask->dst == cpu_to_be32(0xffffffff)) {
7850                                field_flags |= I40E_CLOUD_FIELD_IIP;
7851                        } else {
7852                                dev_err(&pf->pdev->dev, "Bad ip dst mask %pI4b\n",
7853                                        &match.mask->dst);
7854                                return I40E_ERR_CONFIG;
7855                        }
7856                }
7857
7858                if (match.mask->src) {
7859                        if (match.mask->src == cpu_to_be32(0xffffffff)) {
7860                                field_flags |= I40E_CLOUD_FIELD_IIP;
7861                        } else {
7862                                dev_err(&pf->pdev->dev, "Bad ip src mask %pI4b\n",
7863                                        &match.mask->src);
7864                                return I40E_ERR_CONFIG;
7865                        }
7866                }
7867
7868                if (field_flags & I40E_CLOUD_FIELD_TEN_ID) {
7869                        dev_err(&pf->pdev->dev, "Tenant id not allowed for ip filter\n");
7870                        return I40E_ERR_CONFIG;
7871                }
7872                filter->dst_ipv4 = match.key->dst;
7873                filter->src_ipv4 = match.key->src;
7874        }
7875
7876        if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
7877                struct flow_match_ipv6_addrs match;
7878
7879                flow_rule_match_ipv6_addrs(rule, &match);
7880
7881                /* src and dest IPV6 address should not be LOOPBACK
7882                 * (0:0:0:0:0:0:0:1), which can be represented as ::1
7883                 */
7884                if (ipv6_addr_loopback(&match.key->dst) ||
7885                    ipv6_addr_loopback(&match.key->src)) {
7886                        dev_err(&pf->pdev->dev,
7887                                "Bad ipv6, addr is LOOPBACK\n");
7888                        return I40E_ERR_CONFIG;
7889                }
7890                if (!ipv6_addr_any(&match.mask->dst) ||
7891                    !ipv6_addr_any(&match.mask->src))
7892                        field_flags |= I40E_CLOUD_FIELD_IIP;
7893
7894                memcpy(&filter->src_ipv6, &match.key->src.s6_addr32,
7895                       sizeof(filter->src_ipv6));
7896                memcpy(&filter->dst_ipv6, &match.key->dst.s6_addr32,
7897                       sizeof(filter->dst_ipv6));
7898        }
7899
7900        if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS)) {
7901                struct flow_match_ports match;
7902
7903                flow_rule_match_ports(rule, &match);
7904                if (match.mask->src) {
7905                        if (match.mask->src == cpu_to_be16(0xffff)) {
7906                                field_flags |= I40E_CLOUD_FIELD_IIP;
7907                        } else {
7908                                dev_err(&pf->pdev->dev, "Bad src port mask 0x%04x\n",
7909                                        be16_to_cpu(match.mask->src));
7910                                return I40E_ERR_CONFIG;
7911                        }
7912                }
7913
7914                if (match.mask->dst) {
7915                        if (match.mask->dst == cpu_to_be16(0xffff)) {
7916                                field_flags |= I40E_CLOUD_FIELD_IIP;
7917                        } else {
7918                                dev_err(&pf->pdev->dev, "Bad dst port mask 0x%04x\n",
7919                                        be16_to_cpu(match.mask->dst));
7920                                return I40E_ERR_CONFIG;
7921                        }
7922                }
7923
7924                filter->dst_port = match.key->dst;
7925                filter->src_port = match.key->src;
7926
7927                switch (filter->ip_proto) {
7928                case IPPROTO_TCP:
7929                case IPPROTO_UDP:
7930                        break;
7931                default:
7932                        dev_err(&pf->pdev->dev,
7933                                "Only UDP and TCP transport are supported\n");
7934                        return -EINVAL;
7935                }
7936        }
7937        filter->flags = field_flags;
7938        return 0;
7939}
7940
7941/**
7942 * i40e_handle_tclass: Forward to a traffic class on the device
7943 * @vsi: Pointer to VSI
7944 * @tc: traffic class index on the device
7945 * @filter: Pointer to cloud filter structure
7946 *
7947 **/
7948static int i40e_handle_tclass(struct i40e_vsi *vsi, u32 tc,
7949                              struct i40e_cloud_filter *filter)
7950{
7951        struct i40e_channel *ch, *ch_tmp;
7952
7953        /* direct to a traffic class on the same device */
7954        if (tc == 0) {
7955                filter->seid = vsi->seid;
7956                return 0;
7957        } else if (vsi->tc_config.enabled_tc & BIT(tc)) {
7958                if (!filter->dst_port) {
7959                        dev_err(&vsi->back->pdev->dev,
7960                                "Specify destination port to direct to traffic class that is not default\n");
7961                        return -EINVAL;
7962                }
7963                if (list_empty(&vsi->ch_list))
7964                        return -EINVAL;
7965                list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list,
7966                                         list) {
7967                        if (ch->seid == vsi->tc_seid_map[tc])
7968                                filter->seid = ch->seid;
7969                }
7970                return 0;
7971        }
7972        dev_err(&vsi->back->pdev->dev, "TC is not enabled\n");
7973        return -EINVAL;
7974}
7975
7976/**
7977 * i40e_configure_clsflower - Configure tc flower filters
7978 * @vsi: Pointer to VSI
7979 * @cls_flower: Pointer to struct flow_cls_offload
7980 *
7981 **/
7982static int i40e_configure_clsflower(struct i40e_vsi *vsi,
7983                                    struct flow_cls_offload *cls_flower)
7984{
7985        int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
7986        struct i40e_cloud_filter *filter = NULL;
7987        struct i40e_pf *pf = vsi->back;
7988        int err = 0;
7989
7990        if (tc < 0) {
7991                dev_err(&vsi->back->pdev->dev, "Invalid traffic class\n");
7992                return -EOPNOTSUPP;
7993        }
7994
7995        if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
7996            test_bit(__I40E_RESET_INTR_RECEIVED, pf->state))
7997                return -EBUSY;
7998
7999        if (pf->fdir_pf_active_filters ||
8000            (!hlist_empty(&pf->fdir_filter_list))) {
8001                dev_err(&vsi->back->pdev->dev,
8002                        "Flow Director Sideband filters exists, turn ntuple off to configure cloud filters\n");
8003                return -EINVAL;
8004        }
8005
8006        if (vsi->back->flags & I40E_FLAG_FD_SB_ENABLED) {
8007                dev_err(&vsi->back->pdev->dev,
8008                        "Disable Flow Director Sideband, configuring Cloud filters via tc-flower\n");
8009                vsi->back->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8010                vsi->back->flags |= I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8011        }
8012
8013        filter = kzalloc(sizeof(*filter), GFP_KERNEL);
8014        if (!filter)
8015                return -ENOMEM;
8016
8017        filter->cookie = cls_flower->cookie;
8018
8019        err = i40e_parse_cls_flower(vsi, cls_flower, filter);
8020        if (err < 0)
8021                goto err;
8022
8023        err = i40e_handle_tclass(vsi, tc, filter);
8024        if (err < 0)
8025                goto err;
8026
8027        /* Add cloud filter */
8028        if (filter->dst_port)
8029                err = i40e_add_del_cloud_filter_big_buf(vsi, filter, true);
8030        else
8031                err = i40e_add_del_cloud_filter(vsi, filter, true);
8032
8033        if (err) {
8034                dev_err(&pf->pdev->dev,
8035                        "Failed to add cloud filter, err %s\n",
8036                        i40e_stat_str(&pf->hw, err));
8037                goto err;
8038        }
8039
8040        /* add filter to the ordered list */
8041        INIT_HLIST_NODE(&filter->cloud_node);
8042
8043        hlist_add_head(&filter->cloud_node, &pf->cloud_filter_list);
8044
8045        pf->num_cloud_filters++;
8046
8047        return err;
8048err:
8049        kfree(filter);
8050        return err;
8051}
8052
8053/**
8054 * i40e_find_cloud_filter - Find the could filter in the list
8055 * @vsi: Pointer to VSI
8056 * @cookie: filter specific cookie
8057 *
8058 **/
8059static struct i40e_cloud_filter *i40e_find_cloud_filter(struct i40e_vsi *vsi,
8060                                                        unsigned long *cookie)
8061{
8062        struct i40e_cloud_filter *filter = NULL;
8063        struct hlist_node *node2;
8064
8065        hlist_for_each_entry_safe(filter, node2,
8066                                  &vsi->back->cloud_filter_list, cloud_node)
8067                if (!memcmp(cookie, &filter->cookie, sizeof(filter->cookie)))
8068                        return filter;
8069        return NULL;
8070}
8071
8072/**
8073 * i40e_delete_clsflower - Remove tc flower filters
8074 * @vsi: Pointer to VSI
8075 * @cls_flower: Pointer to struct flow_cls_offload
8076 *
8077 **/
8078static int i40e_delete_clsflower(struct i40e_vsi *vsi,
8079                                 struct flow_cls_offload *cls_flower)
8080{
8081        struct i40e_cloud_filter *filter = NULL;
8082        struct i40e_pf *pf = vsi->back;
8083        int err = 0;
8084
8085        filter = i40e_find_cloud_filter(vsi, &cls_flower->cookie);
8086
8087        if (!filter)
8088                return -EINVAL;
8089
8090        hash_del(&filter->cloud_node);
8091
8092        if (filter->dst_port)
8093                err = i40e_add_del_cloud_filter_big_buf(vsi, filter, false);
8094        else
8095                err = i40e_add_del_cloud_filter(vsi, filter, false);
8096
8097        kfree(filter);
8098        if (err) {
8099                dev_err(&pf->pdev->dev,
8100                        "Failed to delete cloud filter, err %s\n",
8101                        i40e_stat_str(&pf->hw, err));
8102                return i40e_aq_rc_to_posix(err, pf->hw.aq.asq_last_status);
8103        }
8104
8105        pf->num_cloud_filters--;
8106        if (!pf->num_cloud_filters)
8107                if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8108                    !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8109                        pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8110                        pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8111                        pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8112                }
8113        return 0;
8114}
8115
8116/**
8117 * i40e_setup_tc_cls_flower - flower classifier offloads
8118 * @netdev: net device to configure
8119 * @type_data: offload data
8120 **/
8121static int i40e_setup_tc_cls_flower(struct i40e_netdev_priv *np,
8122                                    struct flow_cls_offload *cls_flower)
8123{
8124        struct i40e_vsi *vsi = np->vsi;
8125
8126        switch (cls_flower->command) {
8127        case FLOW_CLS_REPLACE:
8128                return i40e_configure_clsflower(vsi, cls_flower);
8129        case FLOW_CLS_DESTROY:
8130                return i40e_delete_clsflower(vsi, cls_flower);
8131        case FLOW_CLS_STATS:
8132                return -EOPNOTSUPP;
8133        default:
8134                return -EOPNOTSUPP;
8135        }
8136}
8137
8138static int i40e_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
8139                                  void *cb_priv)
8140{
8141        struct i40e_netdev_priv *np = cb_priv;
8142
8143        if (!tc_cls_can_offload_and_chain0(np->vsi->netdev, type_data))
8144                return -EOPNOTSUPP;
8145
8146        switch (type) {
8147        case TC_SETUP_CLSFLOWER:
8148                return i40e_setup_tc_cls_flower(np, type_data);
8149
8150        default:
8151                return -EOPNOTSUPP;
8152        }
8153}
8154
8155static LIST_HEAD(i40e_block_cb_list);
8156
8157static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8158                           void *type_data)
8159{
8160        struct i40e_netdev_priv *np = netdev_priv(netdev);
8161
8162        switch (type) {
8163        case TC_SETUP_QDISC_MQPRIO:
8164                return i40e_setup_tc(netdev, type_data);
8165        case TC_SETUP_BLOCK:
8166                return flow_block_cb_setup_simple(type_data,
8167                                                  &i40e_block_cb_list,
8168                                                  i40e_setup_tc_block_cb,
8169                                                  np, np, true);
8170        default:
8171                return -EOPNOTSUPP;
8172        }
8173}
8174
8175/**
8176 * i40e_open - Called when a network interface is made active
8177 * @netdev: network interface device structure
8178 *
8179 * The open entry point is called when a network interface is made
8180 * active by the system (IFF_UP).  At this point all resources needed
8181 * for transmit and receive operations are allocated, the interrupt
8182 * handler is registered with the OS, the netdev watchdog subtask is
8183 * enabled, and the stack is notified that the interface is ready.
8184 *
8185 * Returns 0 on success, negative value on failure
8186 **/
8187int i40e_open(struct net_device *netdev)
8188{
8189        struct i40e_netdev_priv *np = netdev_priv(netdev);
8190        struct i40e_vsi *vsi = np->vsi;
8191        struct i40e_pf *pf = vsi->back;
8192        int err;
8193
8194        /* disallow open during test or if eeprom is broken */
8195        if (test_bit(__I40E_TESTING, pf->state) ||
8196            test_bit(__I40E_BAD_EEPROM, pf->state))
8197                return -EBUSY;
8198
8199        netif_carrier_off(netdev);
8200
8201        if (i40e_force_link_state(pf, true))
8202                return -EAGAIN;
8203
8204        err = i40e_vsi_open(vsi);
8205        if (err)
8206                return err;
8207
8208        /* configure global TSO hardware offload settings */
8209        wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
8210                                                       TCP_FLAG_FIN) >> 16);
8211        wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
8212                                                       TCP_FLAG_FIN |
8213                                                       TCP_FLAG_CWR) >> 16);
8214        wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
8215
8216        udp_tunnel_get_rx_info(netdev);
8217
8218        return 0;
8219}
8220
8221/**
8222 * i40e_vsi_open -
8223 * @vsi: the VSI to open
8224 *
8225 * Finish initialization of the VSI.
8226 *
8227 * Returns 0 on success, negative value on failure
8228 *
8229 * Note: expects to be called while under rtnl_lock()
8230 **/
8231int i40e_vsi_open(struct i40e_vsi *vsi)
8232{
8233        struct i40e_pf *pf = vsi->back;
8234        char int_name[I40E_INT_NAME_STR_LEN];
8235        int err;
8236
8237        /* allocate descriptors */
8238        err = i40e_vsi_setup_tx_resources(vsi);
8239        if (err)
8240                goto err_setup_tx;
8241        err = i40e_vsi_setup_rx_resources(vsi);
8242        if (err)
8243                goto err_setup_rx;
8244
8245        err = i40e_vsi_configure(vsi);
8246        if (err)
8247                goto err_setup_rx;
8248
8249        if (vsi->netdev) {
8250                snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
8251                         dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
8252                err = i40e_vsi_request_irq(vsi, int_name);
8253                if (err)
8254                        goto err_setup_rx;
8255
8256                /* Notify the stack of the actual queue counts. */
8257                err = netif_set_real_num_tx_queues(vsi->netdev,
8258                                                   vsi->num_queue_pairs);
8259                if (err)
8260                        goto err_set_queues;
8261
8262                err = netif_set_real_num_rx_queues(vsi->netdev,
8263                                                   vsi->num_queue_pairs);
8264                if (err)
8265                        goto err_set_queues;
8266
8267        } else if (vsi->type == I40E_VSI_FDIR) {
8268                snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
8269                         dev_driver_string(&pf->pdev->dev),
8270                         dev_name(&pf->pdev->dev));
8271                err = i40e_vsi_request_irq(vsi, int_name);
8272
8273        } else {
8274                err = -EINVAL;
8275                goto err_setup_rx;
8276        }
8277
8278        err = i40e_up_complete(vsi);
8279        if (err)
8280                goto err_up_complete;
8281
8282        return 0;
8283
8284err_up_complete:
8285        i40e_down(vsi);
8286err_set_queues:
8287        i40e_vsi_free_irq(vsi);
8288err_setup_rx:
8289        i40e_vsi_free_rx_resources(vsi);
8290err_setup_tx:
8291        i40e_vsi_free_tx_resources(vsi);
8292        if (vsi == pf->vsi[pf->lan_vsi])
8293                i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
8294
8295        return err;
8296}
8297
8298/**
8299 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
8300 * @pf: Pointer to PF
8301 *
8302 * This function destroys the hlist where all the Flow Director
8303 * filters were saved.
8304 **/
8305static void i40e_fdir_filter_exit(struct i40e_pf *pf)
8306{
8307        struct i40e_fdir_filter *filter;
8308        struct i40e_flex_pit *pit_entry, *tmp;
8309        struct hlist_node *node2;
8310
8311        hlist_for_each_entry_safe(filter, node2,
8312                                  &pf->fdir_filter_list, fdir_node) {
8313                hlist_del(&filter->fdir_node);
8314                kfree(filter);
8315        }
8316
8317        list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
8318                list_del(&pit_entry->list);
8319                kfree(pit_entry);
8320        }
8321        INIT_LIST_HEAD(&pf->l3_flex_pit_list);
8322
8323        list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
8324                list_del(&pit_entry->list);
8325                kfree(pit_entry);
8326        }
8327        INIT_LIST_HEAD(&pf->l4_flex_pit_list);
8328
8329        pf->fdir_pf_active_filters = 0;
8330        pf->fd_tcp4_filter_cnt = 0;
8331        pf->fd_udp4_filter_cnt = 0;
8332        pf->fd_sctp4_filter_cnt = 0;
8333        pf->fd_ip4_filter_cnt = 0;
8334
8335        /* Reprogram the default input set for TCP/IPv4 */
8336        i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8337                                I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8338                                I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8339
8340        /* Reprogram the default input set for UDP/IPv4 */
8341        i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
8342                                I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8343                                I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8344
8345        /* Reprogram the default input set for SCTP/IPv4 */
8346        i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
8347                                I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8348                                I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8349
8350        /* Reprogram the default input set for Other/IPv4 */
8351        i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
8352                                I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8353
8354        i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_FRAG_IPV4,
8355                                I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8356}
8357
8358/**
8359 * i40e_cloud_filter_exit - Cleans up the cloud filters
8360 * @pf: Pointer to PF
8361 *
8362 * This function destroys the hlist where all the cloud filters
8363 * were saved.
8364 **/
8365static void i40e_cloud_filter_exit(struct i40e_pf *pf)
8366{
8367        struct i40e_cloud_filter *cfilter;
8368        struct hlist_node *node;
8369
8370        hlist_for_each_entry_safe(cfilter, node,
8371                                  &pf->cloud_filter_list, cloud_node) {
8372                hlist_del(&cfilter->cloud_node);
8373                kfree(cfilter);
8374        }
8375        pf->num_cloud_filters = 0;
8376
8377        if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8378            !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8379                pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8380                pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8381                pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8382        }
8383}
8384
8385/**
8386 * i40e_close - Disables a network interface
8387 * @netdev: network interface device structure
8388 *
8389 * The close entry point is called when an interface is de-activated
8390 * by the OS.  The hardware is still under the driver's control, but
8391 * this netdev interface is disabled.
8392 *
8393 * Returns 0, this is not allowed to fail
8394 **/
8395int i40e_close(struct net_device *netdev)
8396{
8397        struct i40e_netdev_priv *np = netdev_priv(netdev);
8398        struct i40e_vsi *vsi = np->vsi;
8399
8400        i40e_vsi_close(vsi);
8401
8402        return 0;
8403}
8404
8405/**
8406 * i40e_do_reset - Start a PF or Core Reset sequence
8407 * @pf: board private structure
8408 * @reset_flags: which reset is requested
8409 * @lock_acquired: indicates whether or not the lock has been acquired
8410 * before this function was called.
8411 *
8412 * The essential difference in resets is that the PF Reset
8413 * doesn't clear the packet buffers, doesn't reset the PE
8414 * firmware, and doesn't bother the other PFs on the chip.
8415 **/
8416void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
8417{
8418        u32 val;
8419
8420        WARN_ON(in_interrupt());
8421
8422
8423        /* do the biggest reset indicated */
8424        if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
8425
8426                /* Request a Global Reset
8427                 *
8428                 * This will start the chip's countdown to the actual full
8429                 * chip reset event, and a warning interrupt to be sent
8430                 * to all PFs, including the requestor.  Our handler
8431                 * for the warning interrupt will deal with the shutdown
8432                 * and recovery of the switch setup.
8433                 */
8434                dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
8435                val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8436                val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
8437                wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8438
8439        } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
8440
8441                /* Request a Core Reset
8442                 *
8443                 * Same as Global Reset, except does *not* include the MAC/PHY
8444                 */
8445                dev_dbg(&pf->pdev->dev, "CoreR requested\n");
8446                val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8447                val |= I40E_GLGEN_RTRIG_CORER_MASK;
8448                wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8449                i40e_flush(&pf->hw);
8450
8451        } else if (reset_flags & I40E_PF_RESET_FLAG) {
8452
8453                /* Request a PF Reset
8454                 *
8455                 * Resets only the PF-specific registers
8456                 *
8457                 * This goes directly to the tear-down and rebuild of
8458                 * the switch, since we need to do all the recovery as
8459                 * for the Core Reset.
8460                 */
8461                dev_dbg(&pf->pdev->dev, "PFR requested\n");
8462                i40e_handle_reset_warning(pf, lock_acquired);
8463
8464                dev_info(&pf->pdev->dev,
8465                         pf->flags & I40E_FLAG_DISABLE_FW_LLDP ?
8466                         "FW LLDP is disabled\n" :
8467                         "FW LLDP is enabled\n");
8468
8469        } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
8470                int v;
8471
8472                /* Find the VSI(s) that requested a re-init */
8473                dev_info(&pf->pdev->dev,
8474                         "VSI reinit requested\n");
8475                for (v = 0; v < pf->num_alloc_vsi; v++) {
8476                        struct i40e_vsi *vsi = pf->vsi[v];
8477
8478                        if (vsi != NULL &&
8479                            test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
8480                                               vsi->state))
8481                                i40e_vsi_reinit_locked(pf->vsi[v]);
8482                }
8483        } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
8484                int v;
8485
8486                /* Find the VSI(s) that needs to be brought down */
8487                dev_info(&pf->pdev->dev, "VSI down requested\n");
8488                for (v = 0; v < pf->num_alloc_vsi; v++) {
8489                        struct i40e_vsi *vsi = pf->vsi[v];
8490
8491                        if (vsi != NULL &&
8492                            test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
8493                                               vsi->state)) {
8494                                set_bit(__I40E_VSI_DOWN, vsi->state);
8495                                i40e_down(vsi);
8496                        }
8497                }
8498        } else {
8499                dev_info(&pf->pdev->dev,
8500                         "bad reset request 0x%08x\n", reset_flags);
8501        }
8502}
8503
8504#ifdef CONFIG_I40E_DCB
8505/**
8506 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
8507 * @pf: board private structure
8508 * @old_cfg: current DCB config
8509 * @new_cfg: new DCB config
8510 **/
8511bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
8512                            struct i40e_dcbx_config *old_cfg,
8513                            struct i40e_dcbx_config *new_cfg)
8514{
8515        bool need_reconfig = false;
8516
8517        /* Check if ETS configuration has changed */
8518        if (memcmp(&new_cfg->etscfg,
8519                   &old_cfg->etscfg,
8520                   sizeof(new_cfg->etscfg))) {
8521                /* If Priority Table has changed reconfig is needed */
8522                if (memcmp(&new_cfg->etscfg.prioritytable,
8523                           &old_cfg->etscfg.prioritytable,
8524                           sizeof(new_cfg->etscfg.prioritytable))) {
8525                        need_reconfig = true;
8526                        dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
8527                }
8528
8529                if (memcmp(&new_cfg->etscfg.tcbwtable,
8530                           &old_cfg->etscfg.tcbwtable,
8531                           sizeof(new_cfg->etscfg.tcbwtable)))
8532                        dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
8533
8534                if (memcmp(&new_cfg->etscfg.tsatable,
8535                           &old_cfg->etscfg.tsatable,
8536                           sizeof(new_cfg->etscfg.tsatable)))
8537                        dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
8538        }
8539
8540        /* Check if PFC configuration has changed */
8541        if (memcmp(&new_cfg->pfc,
8542                   &old_cfg->pfc,
8543                   sizeof(new_cfg->pfc))) {
8544                need_reconfig = true;
8545                dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
8546        }
8547
8548        /* Check if APP Table has changed */
8549        if (memcmp(&new_cfg->app,
8550                   &old_cfg->app,
8551                   sizeof(new_cfg->app))) {
8552                need_reconfig = true;
8553                dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
8554        }
8555
8556        dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
8557        return need_reconfig;
8558}
8559
8560/**
8561 * i40e_handle_lldp_event - Handle LLDP Change MIB event
8562 * @pf: board private structure
8563 * @e: event info posted on ARQ
8564 **/
8565static int i40e_handle_lldp_event(struct i40e_pf *pf,
8566                                  struct i40e_arq_event_info *e)
8567{
8568        struct i40e_aqc_lldp_get_mib *mib =
8569                (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
8570        struct i40e_hw *hw = &pf->hw;
8571        struct i40e_dcbx_config tmp_dcbx_cfg;
8572        bool need_reconfig = false;
8573        int ret = 0;
8574        u8 type;
8575
8576        /* Not DCB capable or capability disabled */
8577        if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
8578                return ret;
8579
8580        /* Ignore if event is not for Nearest Bridge */
8581        type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
8582                & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
8583        dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
8584        if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
8585                return ret;
8586
8587        /* Check MIB Type and return if event for Remote MIB update */
8588        type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
8589        dev_dbg(&pf->pdev->dev,
8590                "LLDP event mib type %s\n", type ? "remote" : "local");
8591        if (type == I40E_AQ_LLDP_MIB_REMOTE) {
8592                /* Update the remote cached instance and return */
8593                ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
8594                                I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
8595                                &hw->remote_dcbx_config);
8596                goto exit;
8597        }
8598
8599        /* Store the old configuration */
8600        tmp_dcbx_cfg = hw->local_dcbx_config;
8601
8602        /* Reset the old DCBx configuration data */
8603        memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
8604        /* Get updated DCBX data from firmware */
8605        ret = i40e_get_dcb_config(&pf->hw);
8606        if (ret) {
8607                dev_info(&pf->pdev->dev,
8608                         "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
8609                         i40e_stat_str(&pf->hw, ret),
8610                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8611                goto exit;
8612        }
8613
8614        /* No change detected in DCBX configs */
8615        if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
8616                    sizeof(tmp_dcbx_cfg))) {
8617                dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
8618                goto exit;
8619        }
8620
8621        need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
8622                                               &hw->local_dcbx_config);
8623
8624        i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
8625
8626        if (!need_reconfig)
8627                goto exit;
8628
8629        /* Enable DCB tagging only when more than one TC */
8630        if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
8631                pf->flags |= I40E_FLAG_DCB_ENABLED;
8632        else
8633                pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8634
8635        set_bit(__I40E_PORT_SUSPENDED, pf->state);
8636        /* Reconfiguration needed quiesce all VSIs */
8637        i40e_pf_quiesce_all_vsi(pf);
8638
8639        /* Changes in configuration update VEB/VSI */
8640        i40e_dcb_reconfigure(pf);
8641
8642        ret = i40e_resume_port_tx(pf);
8643
8644        clear_bit(__I40E_PORT_SUSPENDED, pf->state);
8645        /* In case of error no point in resuming VSIs */
8646        if (ret)
8647                goto exit;
8648
8649        /* Wait for the PF's queues to be disabled */
8650        ret = i40e_pf_wait_queues_disabled(pf);
8651        if (ret) {
8652                /* Schedule PF reset to recover */
8653                set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8654                i40e_service_event_schedule(pf);
8655        } else {
8656                i40e_pf_unquiesce_all_vsi(pf);
8657                set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
8658                set_bit(__I40E_CLIENT_L2_CHANGE, pf->state);
8659        }
8660
8661exit:
8662        return ret;
8663}
8664#endif /* CONFIG_I40E_DCB */
8665
8666/**
8667 * i40e_do_reset_safe - Protected reset path for userland calls.
8668 * @pf: board private structure
8669 * @reset_flags: which reset is requested
8670 *
8671 **/
8672void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
8673{
8674        rtnl_lock();
8675        i40e_do_reset(pf, reset_flags, true);
8676        rtnl_unlock();
8677}
8678
8679/**
8680 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
8681 * @pf: board private structure
8682 * @e: event info posted on ARQ
8683 *
8684 * Handler for LAN Queue Overflow Event generated by the firmware for PF
8685 * and VF queues
8686 **/
8687static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
8688                                           struct i40e_arq_event_info *e)
8689{
8690        struct i40e_aqc_lan_overflow *data =
8691                (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
8692        u32 queue = le32_to_cpu(data->prtdcb_rupto);
8693        u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
8694        struct i40e_hw *hw = &pf->hw;
8695        struct i40e_vf *vf;
8696        u16 vf_id;
8697
8698        dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
8699                queue, qtx_ctl);
8700
8701        /* Queue belongs to VF, find the VF and issue VF reset */
8702        if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
8703            >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
8704                vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
8705                         >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
8706                vf_id -= hw->func_caps.vf_base_id;
8707                vf = &pf->vf[vf_id];
8708                i40e_vc_notify_vf_reset(vf);
8709                /* Allow VF to process pending reset notification */
8710                msleep(20);
8711                i40e_reset_vf(vf, false);
8712        }
8713}
8714
8715/**
8716 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
8717 * @pf: board private structure
8718 **/
8719u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
8720{
8721        u32 val, fcnt_prog;
8722
8723        val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8724        fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
8725        return fcnt_prog;
8726}
8727
8728/**
8729 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
8730 * @pf: board private structure
8731 **/
8732u32 i40e_get_current_fd_count(struct i40e_pf *pf)
8733{
8734        u32 val, fcnt_prog;
8735
8736        val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8737        fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
8738                    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
8739                      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
8740        return fcnt_prog;
8741}
8742
8743/**
8744 * i40e_get_global_fd_count - Get total FD filters programmed on device
8745 * @pf: board private structure
8746 **/
8747u32 i40e_get_global_fd_count(struct i40e_pf *pf)
8748{
8749        u32 val, fcnt_prog;
8750
8751        val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
8752        fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
8753                    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
8754                     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
8755        return fcnt_prog;
8756}
8757
8758/**
8759 * i40e_reenable_fdir_sb - Restore FDir SB capability
8760 * @pf: board private structure
8761 **/
8762static void i40e_reenable_fdir_sb(struct i40e_pf *pf)
8763{
8764        if (test_and_clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state))
8765                if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8766                    (I40E_DEBUG_FD & pf->hw.debug_mask))
8767                        dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
8768}
8769
8770/**
8771 * i40e_reenable_fdir_atr - Restore FDir ATR capability
8772 * @pf: board private structure
8773 **/
8774static void i40e_reenable_fdir_atr(struct i40e_pf *pf)
8775{
8776        if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state)) {
8777                /* ATR uses the same filtering logic as SB rules. It only
8778                 * functions properly if the input set mask is at the default
8779                 * settings. It is safe to restore the default input set
8780                 * because there are no active TCPv4 filter rules.
8781                 */
8782                i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8783                                        I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8784                                        I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8785
8786                if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8787                    (I40E_DEBUG_FD & pf->hw.debug_mask))
8788                        dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
8789        }
8790}
8791
8792/**
8793 * i40e_delete_invalid_filter - Delete an invalid FDIR filter
8794 * @pf: board private structure
8795 * @filter: FDir filter to remove
8796 */
8797static void i40e_delete_invalid_filter(struct i40e_pf *pf,
8798                                       struct i40e_fdir_filter *filter)
8799{
8800        /* Update counters */
8801        pf->fdir_pf_active_filters--;
8802        pf->fd_inv = 0;
8803
8804        switch (filter->flow_type) {
8805        case TCP_V4_FLOW:
8806                pf->fd_tcp4_filter_cnt--;
8807                break;
8808        case UDP_V4_FLOW:
8809                pf->fd_udp4_filter_cnt--;
8810                break;
8811        case SCTP_V4_FLOW:
8812                pf->fd_sctp4_filter_cnt--;
8813                break;
8814        case IP_USER_FLOW:
8815                switch (filter->ip4_proto) {
8816                case IPPROTO_TCP:
8817                        pf->fd_tcp4_filter_cnt--;
8818                        break;
8819                case IPPROTO_UDP:
8820                        pf->fd_udp4_filter_cnt--;
8821                        break;
8822                case IPPROTO_SCTP:
8823                        pf->fd_sctp4_filter_cnt--;
8824                        break;
8825                case IPPROTO_IP:
8826                        pf->fd_ip4_filter_cnt--;
8827                        break;
8828                }
8829                break;
8830        }
8831
8832        /* Remove the filter from the list and free memory */
8833        hlist_del(&filter->fdir_node);
8834        kfree(filter);
8835}
8836
8837/**
8838 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
8839 * @pf: board private structure
8840 **/
8841void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
8842{
8843        struct i40e_fdir_filter *filter;
8844        u32 fcnt_prog, fcnt_avail;
8845        struct hlist_node *node;
8846
8847        if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8848                return;
8849
8850        /* Check if we have enough room to re-enable FDir SB capability. */
8851        fcnt_prog = i40e_get_global_fd_count(pf);
8852        fcnt_avail = pf->fdir_pf_filter_count;
8853        if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
8854            (pf->fd_add_err == 0) ||
8855            (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt))
8856                i40e_reenable_fdir_sb(pf);
8857
8858        /* We should wait for even more space before re-enabling ATR.
8859         * Additionally, we cannot enable ATR as long as we still have TCP SB
8860         * rules active.
8861         */
8862        if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
8863            (pf->fd_tcp4_filter_cnt == 0))
8864                i40e_reenable_fdir_atr(pf);
8865
8866        /* if hw had a problem adding a filter, delete it */
8867        if (pf->fd_inv > 0) {
8868                hlist_for_each_entry_safe(filter, node,
8869                                          &pf->fdir_filter_list, fdir_node)
8870                        if (filter->fd_id == pf->fd_inv)
8871                                i40e_delete_invalid_filter(pf, filter);
8872        }
8873}
8874
8875#define I40E_MIN_FD_FLUSH_INTERVAL 10
8876#define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
8877/**
8878 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
8879 * @pf: board private structure
8880 **/
8881static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
8882{
8883        unsigned long min_flush_time;
8884        int flush_wait_retry = 50;
8885        bool disable_atr = false;
8886        int fd_room;
8887        int reg;
8888
8889        if (!time_after(jiffies, pf->fd_flush_timestamp +
8890                                 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
8891                return;
8892
8893        /* If the flush is happening too quick and we have mostly SB rules we
8894         * should not re-enable ATR for some time.
8895         */
8896        min_flush_time = pf->fd_flush_timestamp +
8897                         (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
8898        fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
8899
8900        if (!(time_after(jiffies, min_flush_time)) &&
8901            (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
8902                if (I40E_DEBUG_FD & pf->hw.debug_mask)
8903                        dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
8904                disable_atr = true;
8905        }
8906
8907        pf->fd_flush_timestamp = jiffies;
8908        set_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8909        /* flush all filters */
8910        wr32(&pf->hw, I40E_PFQF_CTL_1,
8911             I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
8912        i40e_flush(&pf->hw);
8913        pf->fd_flush_cnt++;
8914        pf->fd_add_err = 0;
8915        do {
8916                /* Check FD flush status every 5-6msec */
8917                usleep_range(5000, 6000);
8918                reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
8919                if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
8920                        break;
8921        } while (flush_wait_retry--);
8922        if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
8923                dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
8924        } else {
8925                /* replay sideband filters */
8926                i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
8927                if (!disable_atr && !pf->fd_tcp4_filter_cnt)
8928                        clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8929                clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
8930                if (I40E_DEBUG_FD & pf->hw.debug_mask)
8931                        dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
8932        }
8933}
8934
8935/**
8936 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
8937 * @pf: board private structure
8938 **/
8939u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
8940{
8941        return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
8942}
8943
8944/* We can see up to 256 filter programming desc in transit if the filters are
8945 * being applied really fast; before we see the first
8946 * filter miss error on Rx queue 0. Accumulating enough error messages before
8947 * reacting will make sure we don't cause flush too often.
8948 */
8949#define I40E_MAX_FD_PROGRAM_ERROR 256
8950
8951/**
8952 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
8953 * @pf: board private structure
8954 **/
8955static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
8956{
8957
8958        /* if interface is down do nothing */
8959        if (test_bit(__I40E_DOWN, pf->state))
8960                return;
8961
8962        if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8963                i40e_fdir_flush_and_replay(pf);
8964
8965        i40e_fdir_check_and_reenable(pf);
8966
8967}
8968
8969/**
8970 * i40e_vsi_link_event - notify VSI of a link event
8971 * @vsi: vsi to be notified
8972 * @link_up: link up or down
8973 **/
8974static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
8975{
8976        if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
8977                return;
8978
8979        switch (vsi->type) {
8980        case I40E_VSI_MAIN:
8981                if (!vsi->netdev || !vsi->netdev_registered)
8982                        break;
8983
8984                if (link_up) {
8985                        netif_carrier_on(vsi->netdev);
8986                        netif_tx_wake_all_queues(vsi->netdev);
8987                } else {
8988                        netif_carrier_off(vsi->netdev);
8989                        netif_tx_stop_all_queues(vsi->netdev);
8990                }
8991                break;
8992
8993        case I40E_VSI_SRIOV:
8994        case I40E_VSI_VMDQ2:
8995        case I40E_VSI_CTRL:
8996        case I40E_VSI_IWARP:
8997        case I40E_VSI_MIRROR:
8998        default:
8999                /* there is no notification for other VSIs */
9000                break;
9001        }
9002}
9003
9004/**
9005 * i40e_veb_link_event - notify elements on the veb of a link event
9006 * @veb: veb to be notified
9007 * @link_up: link up or down
9008 **/
9009static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
9010{
9011        struct i40e_pf *pf;
9012        int i;
9013
9014        if (!veb || !veb->pf)
9015                return;
9016        pf = veb->pf;
9017
9018        /* depth first... */
9019        for (i = 0; i < I40E_MAX_VEB; i++)
9020                if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
9021                        i40e_veb_link_event(pf->veb[i], link_up);
9022
9023        /* ... now the local VSIs */
9024        for (i = 0; i < pf->num_alloc_vsi; i++)
9025                if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
9026                        i40e_vsi_link_event(pf->vsi[i], link_up);
9027}
9028
9029/**
9030 * i40e_link_event - Update netif_carrier status
9031 * @pf: board private structure
9032 **/
9033static void i40e_link_event(struct i40e_pf *pf)
9034{
9035        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9036        u8 new_link_speed, old_link_speed;
9037        i40e_status status;
9038        bool new_link, old_link;
9039
9040        /* set this to force the get_link_status call to refresh state */
9041        pf->hw.phy.get_link_info = true;
9042        old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
9043        status = i40e_get_link_status(&pf->hw, &new_link);
9044
9045        /* On success, disable temp link polling */
9046        if (status == I40E_SUCCESS) {
9047                clear_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9048        } else {
9049                /* Enable link polling temporarily until i40e_get_link_status
9050                 * returns I40E_SUCCESS
9051                 */
9052                set_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9053                dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
9054                        status);
9055                return;
9056        }
9057
9058        old_link_speed = pf->hw.phy.link_info_old.link_speed;
9059        new_link_speed = pf->hw.phy.link_info.link_speed;
9060
9061        if (new_link == old_link &&
9062            new_link_speed == old_link_speed &&
9063            (test_bit(__I40E_VSI_DOWN, vsi->state) ||
9064             new_link == netif_carrier_ok(vsi->netdev)))
9065                return;
9066
9067        i40e_print_link_message(vsi, new_link);
9068
9069        /* Notify the base of the switch tree connected to
9070         * the link.  Floating VEBs are not notified.
9071         */
9072        if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
9073                i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
9074        else
9075                i40e_vsi_link_event(vsi, new_link);
9076
9077        if (pf->vf)
9078                i40e_vc_notify_link_state(pf);
9079
9080        if (pf->flags & I40E_FLAG_PTP)
9081                i40e_ptp_set_increment(pf);
9082}
9083
9084/**
9085 * i40e_watchdog_subtask - periodic checks not using event driven response
9086 * @pf: board private structure
9087 **/
9088static void i40e_watchdog_subtask(struct i40e_pf *pf)
9089{
9090        int i;
9091
9092        /* if interface is down do nothing */
9093        if (test_bit(__I40E_DOWN, pf->state) ||
9094            test_bit(__I40E_CONFIG_BUSY, pf->state))
9095                return;
9096
9097        /* make sure we don't do these things too often */
9098        if (time_before(jiffies, (pf->service_timer_previous +
9099                                  pf->service_timer_period)))
9100                return;
9101        pf->service_timer_previous = jiffies;
9102
9103        if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
9104            test_bit(__I40E_TEMP_LINK_POLLING, pf->state))
9105                i40e_link_event(pf);
9106
9107        /* Update the stats for active netdevs so the network stack
9108         * can look at updated numbers whenever it cares to
9109         */
9110        for (i = 0; i < pf->num_alloc_vsi; i++)
9111                if (pf->vsi[i] && pf->vsi[i]->netdev)
9112                        i40e_update_stats(pf->vsi[i]);
9113
9114        if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
9115                /* Update the stats for the active switching components */
9116                for (i = 0; i < I40E_MAX_VEB; i++)
9117                        if (pf->veb[i])
9118                                i40e_update_veb_stats(pf->veb[i]);
9119        }
9120
9121        i40e_ptp_rx_hang(pf);
9122        i40e_ptp_tx_hang(pf);
9123}
9124
9125/**
9126 * i40e_reset_subtask - Set up for resetting the device and driver
9127 * @pf: board private structure
9128 **/
9129static void i40e_reset_subtask(struct i40e_pf *pf)
9130{
9131        u32 reset_flags = 0;
9132
9133        if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
9134                reset_flags |= BIT(__I40E_REINIT_REQUESTED);
9135                clear_bit(__I40E_REINIT_REQUESTED, pf->state);
9136        }
9137        if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
9138                reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
9139                clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
9140        }
9141        if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
9142                reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
9143                clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
9144        }
9145        if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
9146                reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
9147                clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
9148        }
9149        if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
9150                reset_flags |= BIT(__I40E_DOWN_REQUESTED);
9151                clear_bit(__I40E_DOWN_REQUESTED, pf->state);
9152        }
9153
9154        /* If there's a recovery already waiting, it takes
9155         * precedence before starting a new reset sequence.
9156         */
9157        if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
9158                i40e_prep_for_reset(pf, false);
9159                i40e_reset(pf);
9160                i40e_rebuild(pf, false, false);
9161        }
9162
9163        /* If we're already down or resetting, just bail */
9164        if (reset_flags &&
9165            !test_bit(__I40E_DOWN, pf->state) &&
9166            !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
9167                i40e_do_reset(pf, reset_flags, false);
9168        }
9169}
9170
9171/**
9172 * i40e_handle_link_event - Handle link event
9173 * @pf: board private structure
9174 * @e: event info posted on ARQ
9175 **/
9176static void i40e_handle_link_event(struct i40e_pf *pf,
9177                                   struct i40e_arq_event_info *e)
9178{
9179        struct i40e_aqc_get_link_status *status =
9180                (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
9181
9182        /* Do a new status request to re-enable LSE reporting
9183         * and load new status information into the hw struct
9184         * This completely ignores any state information
9185         * in the ARQ event info, instead choosing to always
9186         * issue the AQ update link status command.
9187         */
9188        i40e_link_event(pf);
9189
9190        /* Check if module meets thermal requirements */
9191        if (status->phy_type == I40E_PHY_TYPE_NOT_SUPPORTED_HIGH_TEMP) {
9192                dev_err(&pf->pdev->dev,
9193                        "Rx/Tx is disabled on this device because the module does not meet thermal requirements.\n");
9194                dev_err(&pf->pdev->dev,
9195                        "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9196        } else {
9197                /* check for unqualified module, if link is down, suppress
9198                 * the message if link was forced to be down.
9199                 */
9200                if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
9201                    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
9202                    (!(status->link_info & I40E_AQ_LINK_UP)) &&
9203                    (!(pf->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED))) {
9204                        dev_err(&pf->pdev->dev,
9205                                "Rx/Tx is disabled on this device because an unsupported SFP module type was detected.\n");
9206                        dev_err(&pf->pdev->dev,
9207                                "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9208                }
9209        }
9210}
9211
9212/**
9213 * i40e_clean_adminq_subtask - Clean the AdminQ rings
9214 * @pf: board private structure
9215 **/
9216static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
9217{
9218        struct i40e_arq_event_info event;
9219        struct i40e_hw *hw = &pf->hw;
9220        u16 pending, i = 0;
9221        i40e_status ret;
9222        u16 opcode;
9223        u32 oldval;
9224        u32 val;
9225
9226        /* Do not run clean AQ when PF reset fails */
9227        if (test_bit(__I40E_RESET_FAILED, pf->state))
9228                return;
9229
9230        /* check for error indications */
9231        val = rd32(&pf->hw, pf->hw.aq.arq.len);
9232        oldval = val;
9233        if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
9234                if (hw->debug_mask & I40E_DEBUG_AQ)
9235                        dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
9236                val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
9237        }
9238        if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
9239                if (hw->debug_mask & I40E_DEBUG_AQ)
9240                        dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
9241                val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
9242                pf->arq_overflows++;
9243        }
9244        if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
9245                if (hw->debug_mask & I40E_DEBUG_AQ)
9246                        dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
9247                val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
9248        }
9249        if (oldval != val)
9250                wr32(&pf->hw, pf->hw.aq.arq.len, val);
9251
9252        val = rd32(&pf->hw, pf->hw.aq.asq.len);
9253        oldval = val;
9254        if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
9255                if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9256                        dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
9257                val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
9258        }
9259        if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
9260                if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9261                        dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
9262                val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
9263        }
9264        if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
9265                if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9266                        dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
9267                val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
9268        }
9269        if (oldval != val)
9270                wr32(&pf->hw, pf->hw.aq.asq.len, val);
9271
9272        event.buf_len = I40E_MAX_AQ_BUF_SIZE;
9273        event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
9274        if (!event.msg_buf)
9275                return;
9276
9277        do {
9278                ret = i40e_clean_arq_element(hw, &event, &pending);
9279                if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
9280                        break;
9281                else if (ret) {
9282                        dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
9283                        break;
9284                }
9285
9286                opcode = le16_to_cpu(event.desc.opcode);
9287                switch (opcode) {
9288
9289                case i40e_aqc_opc_get_link_status:
9290                        i40e_handle_link_event(pf, &event);
9291                        break;
9292                case i40e_aqc_opc_send_msg_to_pf:
9293                        ret = i40e_vc_process_vf_msg(pf,
9294                                        le16_to_cpu(event.desc.retval),
9295                                        le32_to_cpu(event.desc.cookie_high),
9296                                        le32_to_cpu(event.desc.cookie_low),
9297                                        event.msg_buf,
9298                                        event.msg_len);
9299                        break;
9300                case i40e_aqc_opc_lldp_update_mib:
9301                        dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
9302#ifdef CONFIG_I40E_DCB
9303                        rtnl_lock();
9304                        ret = i40e_handle_lldp_event(pf, &event);
9305                        rtnl_unlock();
9306#endif /* CONFIG_I40E_DCB */
9307                        break;
9308                case i40e_aqc_opc_event_lan_overflow:
9309                        dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
9310                        i40e_handle_lan_overflow_event(pf, &event);
9311                        break;
9312                case i40e_aqc_opc_send_msg_to_peer:
9313                        dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
9314                        break;
9315                case i40e_aqc_opc_nvm_erase:
9316                case i40e_aqc_opc_nvm_update:
9317                case i40e_aqc_opc_oem_post_update:
9318                        i40e_debug(&pf->hw, I40E_DEBUG_NVM,
9319                                   "ARQ NVM operation 0x%04x completed\n",
9320                                   opcode);
9321                        break;
9322                default:
9323                        dev_info(&pf->pdev->dev,
9324                                 "ARQ: Unknown event 0x%04x ignored\n",
9325                                 opcode);
9326                        break;
9327                }
9328        } while (i++ < pf->adminq_work_limit);
9329
9330        if (i < pf->adminq_work_limit)
9331                clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
9332
9333        /* re-enable Admin queue interrupt cause */
9334        val = rd32(hw, I40E_PFINT_ICR0_ENA);
9335        val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
9336        wr32(hw, I40E_PFINT_ICR0_ENA, val);
9337        i40e_flush(hw);
9338
9339        kfree(event.msg_buf);
9340}
9341
9342/**
9343 * i40e_verify_eeprom - make sure eeprom is good to use
9344 * @pf: board private structure
9345 **/
9346static void i40e_verify_eeprom(struct i40e_pf *pf)
9347{
9348        int err;
9349
9350        err = i40e_diag_eeprom_test(&pf->hw);
9351        if (err) {
9352                /* retry in case of garbage read */
9353                err = i40e_diag_eeprom_test(&pf->hw);
9354                if (err) {
9355                        dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
9356                                 err);
9357                        set_bit(__I40E_BAD_EEPROM, pf->state);
9358                }
9359        }
9360
9361        if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
9362                dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
9363                clear_bit(__I40E_BAD_EEPROM, pf->state);
9364        }
9365}
9366
9367/**
9368 * i40e_enable_pf_switch_lb
9369 * @pf: pointer to the PF structure
9370 *
9371 * enable switch loop back or die - no point in a return value
9372 **/
9373static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
9374{
9375        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9376        struct i40e_vsi_context ctxt;
9377        int ret;
9378
9379        ctxt.seid = pf->main_vsi_seid;
9380        ctxt.pf_num = pf->hw.pf_id;
9381        ctxt.vf_num = 0;
9382        ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9383        if (ret) {
9384                dev_info(&pf->pdev->dev,
9385                         "couldn't get PF vsi config, err %s aq_err %s\n",
9386                         i40e_stat_str(&pf->hw, ret),
9387                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9388                return;
9389        }
9390        ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9391        ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9392        ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9393
9394        ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9395        if (ret) {
9396                dev_info(&pf->pdev->dev,
9397                         "update vsi switch failed, err %s aq_err %s\n",
9398                         i40e_stat_str(&pf->hw, ret),
9399                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9400        }
9401}
9402
9403/**
9404 * i40e_disable_pf_switch_lb
9405 * @pf: pointer to the PF structure
9406 *
9407 * disable switch loop back or die - no point in a return value
9408 **/
9409static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
9410{
9411        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9412        struct i40e_vsi_context ctxt;
9413        int ret;
9414
9415        ctxt.seid = pf->main_vsi_seid;
9416        ctxt.pf_num = pf->hw.pf_id;
9417        ctxt.vf_num = 0;
9418        ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9419        if (ret) {
9420                dev_info(&pf->pdev->dev,
9421                         "couldn't get PF vsi config, err %s aq_err %s\n",
9422                         i40e_stat_str(&pf->hw, ret),
9423                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9424                return;
9425        }
9426        ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9427        ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9428        ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9429
9430        ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9431        if (ret) {
9432                dev_info(&pf->pdev->dev,
9433                         "update vsi switch failed, err %s aq_err %s\n",
9434                         i40e_stat_str(&pf->hw, ret),
9435                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9436        }
9437}
9438
9439/**
9440 * i40e_config_bridge_mode - Configure the HW bridge mode
9441 * @veb: pointer to the bridge instance
9442 *
9443 * Configure the loop back mode for the LAN VSI that is downlink to the
9444 * specified HW bridge instance. It is expected this function is called
9445 * when a new HW bridge is instantiated.
9446 **/
9447static void i40e_config_bridge_mode(struct i40e_veb *veb)
9448{
9449        struct i40e_pf *pf = veb->pf;
9450
9451        if (pf->hw.debug_mask & I40E_DEBUG_LAN)
9452                dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
9453                         veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9454        if (veb->bridge_mode & BRIDGE_MODE_VEPA)
9455                i40e_disable_pf_switch_lb(pf);
9456        else
9457                i40e_enable_pf_switch_lb(pf);
9458}
9459
9460/**
9461 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
9462 * @veb: pointer to the VEB instance
9463 *
9464 * This is a recursive function that first builds the attached VSIs then
9465 * recurses in to build the next layer of VEB.  We track the connections
9466 * through our own index numbers because the seid's from the HW could
9467 * change across the reset.
9468 **/
9469static int i40e_reconstitute_veb(struct i40e_veb *veb)
9470{
9471        struct i40e_vsi *ctl_vsi = NULL;
9472        struct i40e_pf *pf = veb->pf;
9473        int v, veb_idx;
9474        int ret;
9475
9476        /* build VSI that owns this VEB, temporarily attached to base VEB */
9477        for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
9478                if (pf->vsi[v] &&
9479                    pf->vsi[v]->veb_idx == veb->idx &&
9480                    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
9481                        ctl_vsi = pf->vsi[v];
9482                        break;
9483                }
9484        }
9485        if (!ctl_vsi) {
9486                dev_info(&pf->pdev->dev,
9487                         "missing owner VSI for veb_idx %d\n", veb->idx);
9488                ret = -ENOENT;
9489                goto end_reconstitute;
9490        }
9491        if (ctl_vsi != pf->vsi[pf->lan_vsi])
9492                ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9493        ret = i40e_add_vsi(ctl_vsi);
9494        if (ret) {
9495                dev_info(&pf->pdev->dev,
9496                         "rebuild of veb_idx %d owner VSI failed: %d\n",
9497                         veb->idx, ret);
9498                goto end_reconstitute;
9499        }
9500        i40e_vsi_reset_stats(ctl_vsi);
9501
9502        /* create the VEB in the switch and move the VSI onto the VEB */
9503        ret = i40e_add_veb(veb, ctl_vsi);
9504        if (ret)
9505                goto end_reconstitute;
9506
9507        if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
9508                veb->bridge_mode = BRIDGE_MODE_VEB;
9509        else
9510                veb->bridge_mode = BRIDGE_MODE_VEPA;
9511        i40e_config_bridge_mode(veb);
9512
9513        /* create the remaining VSIs attached to this VEB */
9514        for (v = 0; v < pf->num_alloc_vsi; v++) {
9515                if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
9516                        continue;
9517
9518                if (pf->vsi[v]->veb_idx == veb->idx) {
9519                        struct i40e_vsi *vsi = pf->vsi[v];
9520
9521                        vsi->uplink_seid = veb->seid;
9522                        ret = i40e_add_vsi(vsi);
9523                        if (ret) {
9524                                dev_info(&pf->pdev->dev,
9525                                         "rebuild of vsi_idx %d failed: %d\n",
9526                                         v, ret);
9527                                goto end_reconstitute;
9528                        }
9529                        i40e_vsi_reset_stats(vsi);
9530                }
9531        }
9532
9533        /* create any VEBs attached to this VEB - RECURSION */
9534        for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9535                if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
9536                        pf->veb[veb_idx]->uplink_seid = veb->seid;
9537                        ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
9538                        if (ret)
9539                                break;
9540                }
9541        }
9542
9543end_reconstitute:
9544        return ret;
9545}
9546
9547/**
9548 * i40e_get_capabilities - get info about the HW
9549 * @pf: the PF struct
9550 **/
9551static int i40e_get_capabilities(struct i40e_pf *pf,
9552                                 enum i40e_admin_queue_opc list_type)
9553{
9554        struct i40e_aqc_list_capabilities_element_resp *cap_buf;
9555        u16 data_size;
9556        int buf_len;
9557        int err;
9558
9559        buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
9560        do {
9561                cap_buf = kzalloc(buf_len, GFP_KERNEL);
9562                if (!cap_buf)
9563                        return -ENOMEM;
9564
9565                /* this loads the data into the hw struct for us */
9566                err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
9567                                                    &data_size, list_type,
9568                                                    NULL);
9569                /* data loaded, buffer no longer needed */
9570                kfree(cap_buf);
9571
9572                if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
9573                        /* retry with a larger buffer */
9574                        buf_len = data_size;
9575                } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
9576                        dev_info(&pf->pdev->dev,
9577                                 "capability discovery failed, err %s aq_err %s\n",
9578                                 i40e_stat_str(&pf->hw, err),
9579                                 i40e_aq_str(&pf->hw,
9580                                             pf->hw.aq.asq_last_status));
9581                        return -ENODEV;
9582                }
9583        } while (err);
9584
9585        if (pf->hw.debug_mask & I40E_DEBUG_USER) {
9586                if (list_type == i40e_aqc_opc_list_func_capabilities) {
9587                        dev_info(&pf->pdev->dev,
9588                                 "pf=%d, num_vfs=%d, msix_pf=%d, msix_vf=%d, fd_g=%d, fd_b=%d, pf_max_q=%d num_vsi=%d\n",
9589                                 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
9590                                 pf->hw.func_caps.num_msix_vectors,
9591                                 pf->hw.func_caps.num_msix_vectors_vf,
9592                                 pf->hw.func_caps.fd_filters_guaranteed,
9593                                 pf->hw.func_caps.fd_filters_best_effort,
9594                                 pf->hw.func_caps.num_tx_qp,
9595                                 pf->hw.func_caps.num_vsis);
9596                } else if (list_type == i40e_aqc_opc_list_dev_capabilities) {
9597                        dev_info(&pf->pdev->dev,
9598                                 "switch_mode=0x%04x, function_valid=0x%08x\n",
9599                                 pf->hw.dev_caps.switch_mode,
9600                                 pf->hw.dev_caps.valid_functions);
9601                        dev_info(&pf->pdev->dev,
9602                                 "SR-IOV=%d, num_vfs for all function=%u\n",
9603                                 pf->hw.dev_caps.sr_iov_1_1,
9604                                 pf->hw.dev_caps.num_vfs);
9605                        dev_info(&pf->pdev->dev,
9606                                 "num_vsis=%u, num_rx:%u, num_tx=%u\n",
9607                                 pf->hw.dev_caps.num_vsis,
9608                                 pf->hw.dev_caps.num_rx_qp,
9609                                 pf->hw.dev_caps.num_tx_qp);
9610                }
9611        }
9612        if (list_type == i40e_aqc_opc_list_func_capabilities) {
9613#define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
9614                       + pf->hw.func_caps.num_vfs)
9615                if (pf->hw.revision_id == 0 &&
9616                    pf->hw.func_caps.num_vsis < DEF_NUM_VSI) {
9617                        dev_info(&pf->pdev->dev,
9618                                 "got num_vsis %d, setting num_vsis to %d\n",
9619                                 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
9620                        pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
9621                }
9622        }
9623        return 0;
9624}
9625
9626static int i40e_vsi_clear(struct i40e_vsi *vsi);
9627
9628/**
9629 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
9630 * @pf: board private structure
9631 **/
9632static void i40e_fdir_sb_setup(struct i40e_pf *pf)
9633{
9634        struct i40e_vsi *vsi;
9635
9636        /* quick workaround for an NVM issue that leaves a critical register
9637         * uninitialized
9638         */
9639        if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
9640                static const u32 hkey[] = {
9641                        0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
9642                        0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
9643                        0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
9644                        0x95b3a76d};
9645                int i;
9646
9647                for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
9648                        wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
9649        }
9650
9651        if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
9652                return;
9653
9654        /* find existing VSI and see if it needs configuring */
9655        vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9656
9657        /* create a new VSI if none exists */
9658        if (!vsi) {
9659                vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
9660                                     pf->vsi[pf->lan_vsi]->seid, 0);
9661                if (!vsi) {
9662                        dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
9663                        pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9664                        pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
9665                        return;
9666                }
9667        }
9668
9669        i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
9670}
9671
9672/**
9673 * i40e_fdir_teardown - release the Flow Director resources
9674 * @pf: board private structure
9675 **/
9676static void i40e_fdir_teardown(struct i40e_pf *pf)
9677{
9678        struct i40e_vsi *vsi;
9679
9680        i40e_fdir_filter_exit(pf);
9681        vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9682        if (vsi)
9683                i40e_vsi_release(vsi);
9684}
9685
9686/**
9687 * i40e_rebuild_cloud_filters - Rebuilds cloud filters for VSIs
9688 * @vsi: PF main vsi
9689 * @seid: seid of main or channel VSIs
9690 *
9691 * Rebuilds cloud filters associated with main VSI and channel VSIs if they
9692 * existed before reset
9693 **/
9694static int i40e_rebuild_cloud_filters(struct i40e_vsi *vsi, u16 seid)
9695{
9696        struct i40e_cloud_filter *cfilter;
9697        struct i40e_pf *pf = vsi->back;
9698        struct hlist_node *node;
9699        i40e_status ret;
9700
9701        /* Add cloud filters back if they exist */
9702        hlist_for_each_entry_safe(cfilter, node, &pf->cloud_filter_list,
9703                                  cloud_node) {
9704                if (cfilter->seid != seid)
9705                        continue;
9706
9707                if (cfilter->dst_port)
9708                        ret = i40e_add_del_cloud_filter_big_buf(vsi, cfilter,
9709                                                                true);
9710                else
9711                        ret = i40e_add_del_cloud_filter(vsi, cfilter, true);
9712
9713                if (ret) {
9714                        dev_dbg(&pf->pdev->dev,
9715                                "Failed to rebuild cloud filter, err %s aq_err %s\n",
9716                                i40e_stat_str(&pf->hw, ret),
9717                                i40e_aq_str(&pf->hw,
9718                                            pf->hw.aq.asq_last_status));
9719                        return ret;
9720                }
9721        }
9722        return 0;
9723}
9724
9725/**
9726 * i40e_rebuild_channels - Rebuilds channel VSIs if they existed before reset
9727 * @vsi: PF main vsi
9728 *
9729 * Rebuilds channel VSIs if they existed before reset
9730 **/
9731static int i40e_rebuild_channels(struct i40e_vsi *vsi)
9732{
9733        struct i40e_channel *ch, *ch_tmp;
9734        i40e_status ret;
9735
9736        if (list_empty(&vsi->ch_list))
9737                return 0;
9738
9739        list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
9740                if (!ch->initialized)
9741                        break;
9742                /* Proceed with creation of channel (VMDq2) VSI */
9743                ret = i40e_add_channel(vsi->back, vsi->uplink_seid, ch);
9744                if (ret) {
9745                        dev_info(&vsi->back->pdev->dev,
9746                                 "failed to rebuild channels using uplink_seid %u\n",
9747                                 vsi->uplink_seid);
9748                        return ret;
9749                }
9750                /* Reconfigure TX queues using QTX_CTL register */
9751                ret = i40e_channel_config_tx_ring(vsi->back, vsi, ch);
9752                if (ret) {
9753                        dev_info(&vsi->back->pdev->dev,
9754                                 "failed to configure TX rings for channel %u\n",
9755                                 ch->seid);
9756                        return ret;
9757                }
9758                /* update 'next_base_queue' */
9759                vsi->next_base_queue = vsi->next_base_queue +
9760                                                        ch->num_queue_pairs;
9761                if (ch->max_tx_rate) {
9762                        u64 credits = ch->max_tx_rate;
9763
9764                        if (i40e_set_bw_limit(vsi, ch->seid,
9765                                              ch->max_tx_rate))
9766                                return -EINVAL;
9767
9768                        do_div(credits, I40E_BW_CREDIT_DIVISOR);
9769                        dev_dbg(&vsi->back->pdev->dev,
9770                                "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
9771                                ch->max_tx_rate,
9772                                credits,
9773                                ch->seid);
9774                }
9775                ret = i40e_rebuild_cloud_filters(vsi, ch->seid);
9776                if (ret) {
9777                        dev_dbg(&vsi->back->pdev->dev,
9778                                "Failed to rebuild cloud filters for channel VSI %u\n",
9779                                ch->seid);
9780                        return ret;
9781                }
9782        }
9783        return 0;
9784}
9785
9786/**
9787 * i40e_prep_for_reset - prep for the core to reset
9788 * @pf: board private structure
9789 * @lock_acquired: indicates whether or not the lock has been acquired
9790 * before this function was called.
9791 *
9792 * Close up the VFs and other things in prep for PF Reset.
9793  **/
9794static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
9795{
9796        struct i40e_hw *hw = &pf->hw;
9797        i40e_status ret = 0;
9798        u32 v;
9799
9800        clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
9801        if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
9802                return;
9803        if (i40e_check_asq_alive(&pf->hw))
9804                i40e_vc_notify_reset(pf);
9805
9806        dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
9807
9808        /* quiesce the VSIs and their queues that are not already DOWN */
9809        /* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
9810        if (!lock_acquired)
9811                rtnl_lock();
9812        i40e_pf_quiesce_all_vsi(pf);
9813        if (!lock_acquired)
9814                rtnl_unlock();
9815
9816        for (v = 0; v < pf->num_alloc_vsi; v++) {
9817                if (pf->vsi[v])
9818                        pf->vsi[v]->seid = 0;
9819        }
9820
9821        i40e_shutdown_adminq(&pf->hw);
9822
9823        /* call shutdown HMC */
9824        if (hw->hmc.hmc_obj) {
9825                ret = i40e_shutdown_lan_hmc(hw);
9826                if (ret)
9827                        dev_warn(&pf->pdev->dev,
9828                                 "shutdown_lan_hmc failed: %d\n", ret);
9829        }
9830
9831        /* Save the current PTP time so that we can restore the time after the
9832         * reset completes.
9833         */
9834        i40e_ptp_save_hw_time(pf);
9835}
9836
9837/**
9838 * i40e_send_version - update firmware with driver version
9839 * @pf: PF struct
9840 */
9841static void i40e_send_version(struct i40e_pf *pf)
9842{
9843        struct i40e_driver_version dv;
9844
9845        dv.major_version = DRV_VERSION_MAJOR;
9846        dv.minor_version = DRV_VERSION_MINOR;
9847        dv.build_version = DRV_VERSION_BUILD;
9848        dv.subbuild_version = 0;
9849        strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
9850        i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
9851}
9852
9853/**
9854 * i40e_get_oem_version - get OEM specific version information
9855 * @hw: pointer to the hardware structure
9856 **/
9857static void i40e_get_oem_version(struct i40e_hw *hw)
9858{
9859        u16 block_offset = 0xffff;
9860        u16 block_length = 0;
9861        u16 capabilities = 0;
9862        u16 gen_snap = 0;
9863        u16 release = 0;
9864
9865#define I40E_SR_NVM_OEM_VERSION_PTR             0x1B
9866#define I40E_NVM_OEM_LENGTH_OFFSET              0x00
9867#define I40E_NVM_OEM_CAPABILITIES_OFFSET        0x01
9868#define I40E_NVM_OEM_GEN_OFFSET                 0x02
9869#define I40E_NVM_OEM_RELEASE_OFFSET             0x03
9870#define I40E_NVM_OEM_CAPABILITIES_MASK          0x000F
9871#define I40E_NVM_OEM_LENGTH                     3
9872
9873        /* Check if pointer to OEM version block is valid. */
9874        i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
9875        if (block_offset == 0xffff)
9876                return;
9877
9878        /* Check if OEM version block has correct length. */
9879        i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
9880                           &block_length);
9881        if (block_length < I40E_NVM_OEM_LENGTH)
9882                return;
9883
9884        /* Check if OEM version format is as expected. */
9885        i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
9886                           &capabilities);
9887        if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
9888                return;
9889
9890        i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
9891                           &gen_snap);
9892        i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
9893                           &release);
9894        hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
9895        hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
9896}
9897
9898/**
9899 * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
9900 * @pf: board private structure
9901 **/
9902static int i40e_reset(struct i40e_pf *pf)
9903{
9904        struct i40e_hw *hw = &pf->hw;
9905        i40e_status ret;
9906
9907        ret = i40e_pf_reset(hw);
9908        if (ret) {
9909                dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
9910                set_bit(__I40E_RESET_FAILED, pf->state);
9911                clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
9912        } else {
9913                pf->pfr_count++;
9914        }
9915        return ret;
9916}
9917
9918/**
9919 * i40e_rebuild - rebuild using a saved config
9920 * @pf: board private structure
9921 * @reinit: if the Main VSI needs to re-initialized.
9922 * @lock_acquired: indicates whether or not the lock has been acquired
9923 * before this function was called.
9924 **/
9925static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
9926{
9927        int old_recovery_mode_bit = test_bit(__I40E_RECOVERY_MODE, pf->state);
9928        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9929        struct i40e_hw *hw = &pf->hw;
9930        u8 set_fc_aq_fail = 0;
9931        i40e_status ret;
9932        u32 val;
9933        int v;
9934
9935        if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9936            i40e_check_recovery_mode(pf)) {
9937                i40e_set_ethtool_ops(pf->vsi[pf->lan_vsi]->netdev);
9938        }
9939
9940        if (test_bit(__I40E_DOWN, pf->state) &&
9941            !test_bit(__I40E_RECOVERY_MODE, pf->state) &&
9942            !old_recovery_mode_bit)
9943                goto clear_recovery;
9944        dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
9945
9946        /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
9947        ret = i40e_init_adminq(&pf->hw);
9948        if (ret) {
9949                dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
9950                         i40e_stat_str(&pf->hw, ret),
9951                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9952                goto clear_recovery;
9953        }
9954        i40e_get_oem_version(&pf->hw);
9955
9956        if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9957            ((hw->aq.fw_maj_ver == 4 && hw->aq.fw_min_ver <= 33) ||
9958             hw->aq.fw_maj_ver < 4) && hw->mac.type == I40E_MAC_XL710) {
9959                /* The following delay is necessary for 4.33 firmware and older
9960                 * to recover after EMP reset. 200 ms should suffice but we
9961                 * put here 300 ms to be sure that FW is ready to operate
9962                 * after reset.
9963                 */
9964                mdelay(300);
9965        }
9966
9967        /* re-verify the eeprom if we just had an EMP reset */
9968        if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
9969                i40e_verify_eeprom(pf);
9970
9971        /* if we are going out of or into recovery mode we have to act
9972         * accordingly with regard to resources initialization
9973         * and deinitialization
9974         */
9975        if (test_bit(__I40E_RECOVERY_MODE, pf->state) ||
9976            old_recovery_mode_bit) {
9977                if (i40e_get_capabilities(pf,
9978                                          i40e_aqc_opc_list_func_capabilities))
9979                        goto end_unlock;
9980
9981                if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
9982                        /* we're staying in recovery mode so we'll reinitialize
9983                         * misc vector here
9984                         */
9985                        if (i40e_setup_misc_vector_for_recovery_mode(pf))
9986                                goto end_unlock;
9987                } else {
9988                        if (!lock_acquired)
9989                                rtnl_lock();
9990                        /* we're going out of recovery mode so we'll free
9991                         * the IRQ allocated specifically for recovery mode
9992                         * and restore the interrupt scheme
9993                         */
9994                        free_irq(pf->pdev->irq, pf);
9995                        i40e_clear_interrupt_scheme(pf);
9996                        if (i40e_restore_interrupt_scheme(pf))
9997                                goto end_unlock;
9998                }
9999
10000                /* tell the firmware that we're starting */
10001                i40e_send_version(pf);
10002
10003                /* bail out in case recovery mode was detected, as there is
10004                 * no need for further configuration.
10005                 */
10006                goto end_unlock;
10007        }
10008
10009        i40e_clear_pxe_mode(hw);
10010        ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
10011        if (ret)
10012                goto end_core_reset;
10013
10014        ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10015                                hw->func_caps.num_rx_qp, 0, 0);
10016        if (ret) {
10017                dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
10018                goto end_core_reset;
10019        }
10020        ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10021        if (ret) {
10022                dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
10023                goto end_core_reset;
10024        }
10025
10026        /* Enable FW to write a default DCB config on link-up */
10027        i40e_aq_set_dcb_parameters(hw, true, NULL);
10028
10029#ifdef CONFIG_I40E_DCB
10030        ret = i40e_init_pf_dcb(pf);
10031        if (ret) {
10032                dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
10033                pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10034                /* Continue without DCB enabled */
10035        }
10036#endif /* CONFIG_I40E_DCB */
10037        /* do basic switch setup */
10038        if (!lock_acquired)
10039                rtnl_lock();
10040        ret = i40e_setup_pf_switch(pf, reinit);
10041        if (ret)
10042                goto end_unlock;
10043
10044        /* The driver only wants link up/down and module qualification
10045         * reports from firmware.  Note the negative logic.
10046         */
10047        ret = i40e_aq_set_phy_int_mask(&pf->hw,
10048                                       ~(I40E_AQ_EVENT_LINK_UPDOWN |
10049                                         I40E_AQ_EVENT_MEDIA_NA |
10050                                         I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
10051        if (ret)
10052                dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10053                         i40e_stat_str(&pf->hw, ret),
10054                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10055
10056        /* make sure our flow control settings are restored */
10057        ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
10058        if (ret)
10059                dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
10060                        i40e_stat_str(&pf->hw, ret),
10061                        i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10062
10063        /* Rebuild the VSIs and VEBs that existed before reset.
10064         * They are still in our local switch element arrays, so only
10065         * need to rebuild the switch model in the HW.
10066         *
10067         * If there were VEBs but the reconstitution failed, we'll try
10068         * try to recover minimal use by getting the basic PF VSI working.
10069         */
10070        if (vsi->uplink_seid != pf->mac_seid) {
10071                dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
10072                /* find the one VEB connected to the MAC, and find orphans */
10073                for (v = 0; v < I40E_MAX_VEB; v++) {
10074                        if (!pf->veb[v])
10075                                continue;
10076
10077                        if (pf->veb[v]->uplink_seid == pf->mac_seid ||
10078                            pf->veb[v]->uplink_seid == 0) {
10079                                ret = i40e_reconstitute_veb(pf->veb[v]);
10080
10081                                if (!ret)
10082                                        continue;
10083
10084                                /* If Main VEB failed, we're in deep doodoo,
10085                                 * so give up rebuilding the switch and set up
10086                                 * for minimal rebuild of PF VSI.
10087                                 * If orphan failed, we'll report the error
10088                                 * but try to keep going.
10089                                 */
10090                                if (pf->veb[v]->uplink_seid == pf->mac_seid) {
10091                                        dev_info(&pf->pdev->dev,
10092                                                 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
10093                                                 ret);
10094                                        vsi->uplink_seid = pf->mac_seid;
10095                                        break;
10096                                } else if (pf->veb[v]->uplink_seid == 0) {
10097                                        dev_info(&pf->pdev->dev,
10098                                                 "rebuild of orphan VEB failed: %d\n",
10099                                                 ret);
10100                                }
10101                        }
10102                }
10103        }
10104
10105        if (vsi->uplink_seid == pf->mac_seid) {
10106                dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
10107                /* no VEB, so rebuild only the Main VSI */
10108                ret = i40e_add_vsi(vsi);
10109                if (ret) {
10110                        dev_info(&pf->pdev->dev,
10111                                 "rebuild of Main VSI failed: %d\n", ret);
10112                        goto end_unlock;
10113                }
10114        }
10115
10116        if (vsi->mqprio_qopt.max_rate[0]) {
10117                u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
10118                u64 credits = 0;
10119
10120                do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
10121                ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
10122                if (ret)
10123                        goto end_unlock;
10124
10125                credits = max_tx_rate;
10126                do_div(credits, I40E_BW_CREDIT_DIVISOR);
10127                dev_dbg(&vsi->back->pdev->dev,
10128                        "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
10129                        max_tx_rate,
10130                        credits,
10131                        vsi->seid);
10132        }
10133
10134        ret = i40e_rebuild_cloud_filters(vsi, vsi->seid);
10135        if (ret)
10136                goto end_unlock;
10137
10138        /* PF Main VSI is rebuild by now, go ahead and rebuild channel VSIs
10139         * for this main VSI if they exist
10140         */
10141        ret = i40e_rebuild_channels(vsi);
10142        if (ret)
10143                goto end_unlock;
10144
10145        /* Reconfigure hardware for allowing smaller MSS in the case
10146         * of TSO, so that we avoid the MDD being fired and causing
10147         * a reset in the case of small MSS+TSO.
10148         */
10149#define I40E_REG_MSS          0x000E64DC
10150#define I40E_REG_MSS_MIN_MASK 0x3FF0000
10151#define I40E_64BYTE_MSS       0x400000
10152        val = rd32(hw, I40E_REG_MSS);
10153        if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10154                val &= ~I40E_REG_MSS_MIN_MASK;
10155                val |= I40E_64BYTE_MSS;
10156                wr32(hw, I40E_REG_MSS, val);
10157        }
10158
10159        if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
10160                msleep(75);
10161                ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10162                if (ret)
10163                        dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10164                                 i40e_stat_str(&pf->hw, ret),
10165                                 i40e_aq_str(&pf->hw,
10166                                             pf->hw.aq.asq_last_status));
10167        }
10168        /* reinit the misc interrupt */
10169        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10170                ret = i40e_setup_misc_vector(pf);
10171
10172        /* Add a filter to drop all Flow control frames from any VSI from being
10173         * transmitted. By doing so we stop a malicious VF from sending out
10174         * PAUSE or PFC frames and potentially controlling traffic for other
10175         * PF/VF VSIs.
10176         * The FW can still send Flow control frames if enabled.
10177         */
10178        i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
10179                                                       pf->main_vsi_seid);
10180
10181        /* restart the VSIs that were rebuilt and running before the reset */
10182        i40e_pf_unquiesce_all_vsi(pf);
10183
10184        /* Release the RTNL lock before we start resetting VFs */
10185        if (!lock_acquired)
10186                rtnl_unlock();
10187
10188        /* Restore promiscuous settings */
10189        ret = i40e_set_promiscuous(pf, pf->cur_promisc);
10190        if (ret)
10191                dev_warn(&pf->pdev->dev,
10192                         "Failed to restore promiscuous setting: %s, err %s aq_err %s\n",
10193                         pf->cur_promisc ? "on" : "off",
10194                         i40e_stat_str(&pf->hw, ret),
10195                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10196
10197        i40e_reset_all_vfs(pf, true);
10198
10199        /* tell the firmware that we're starting */
10200        i40e_send_version(pf);
10201
10202        /* We've already released the lock, so don't do it again */
10203        goto end_core_reset;
10204
10205end_unlock:
10206        if (!lock_acquired)
10207                rtnl_unlock();
10208end_core_reset:
10209        clear_bit(__I40E_RESET_FAILED, pf->state);
10210clear_recovery:
10211        clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
10212        clear_bit(__I40E_TIMEOUT_RECOVERY_PENDING, pf->state);
10213}
10214
10215/**
10216 * i40e_reset_and_rebuild - reset and rebuild using a saved config
10217 * @pf: board private structure
10218 * @reinit: if the Main VSI needs to re-initialized.
10219 * @lock_acquired: indicates whether or not the lock has been acquired
10220 * before this function was called.
10221 **/
10222static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
10223                                   bool lock_acquired)
10224{
10225        int ret;
10226        /* Now we wait for GRST to settle out.
10227         * We don't have to delete the VEBs or VSIs from the hw switch
10228         * because the reset will make them disappear.
10229         */
10230        ret = i40e_reset(pf);
10231        if (!ret)
10232                i40e_rebuild(pf, reinit, lock_acquired);
10233}
10234
10235/**
10236 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
10237 * @pf: board private structure
10238 *
10239 * Close up the VFs and other things in prep for a Core Reset,
10240 * then get ready to rebuild the world.
10241 * @lock_acquired: indicates whether or not the lock has been acquired
10242 * before this function was called.
10243 **/
10244static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
10245{
10246        i40e_prep_for_reset(pf, lock_acquired);
10247        i40e_reset_and_rebuild(pf, false, lock_acquired);
10248}
10249
10250/**
10251 * i40e_handle_mdd_event
10252 * @pf: pointer to the PF structure
10253 *
10254 * Called from the MDD irq handler to identify possibly malicious vfs
10255 **/
10256static void i40e_handle_mdd_event(struct i40e_pf *pf)
10257{
10258        struct i40e_hw *hw = &pf->hw;
10259        bool mdd_detected = false;
10260        struct i40e_vf *vf;
10261        u32 reg;
10262        int i;
10263
10264        if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
10265                return;
10266
10267        /* find what triggered the MDD event */
10268        reg = rd32(hw, I40E_GL_MDET_TX);
10269        if (reg & I40E_GL_MDET_TX_VALID_MASK) {
10270                u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
10271                                I40E_GL_MDET_TX_PF_NUM_SHIFT;
10272                u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
10273                                I40E_GL_MDET_TX_VF_NUM_SHIFT;
10274                u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
10275                                I40E_GL_MDET_TX_EVENT_SHIFT;
10276                u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
10277                                I40E_GL_MDET_TX_QUEUE_SHIFT) -
10278                                pf->hw.func_caps.base_queue;
10279                if (netif_msg_tx_err(pf))
10280                        dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
10281                                 event, queue, pf_num, vf_num);
10282                wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
10283                mdd_detected = true;
10284        }
10285        reg = rd32(hw, I40E_GL_MDET_RX);
10286        if (reg & I40E_GL_MDET_RX_VALID_MASK) {
10287                u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
10288                                I40E_GL_MDET_RX_FUNCTION_SHIFT;
10289                u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
10290                                I40E_GL_MDET_RX_EVENT_SHIFT;
10291                u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
10292                                I40E_GL_MDET_RX_QUEUE_SHIFT) -
10293                                pf->hw.func_caps.base_queue;
10294                if (netif_msg_rx_err(pf))
10295                        dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
10296                                 event, queue, func);
10297                wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
10298                mdd_detected = true;
10299        }
10300
10301        if (mdd_detected) {
10302                reg = rd32(hw, I40E_PF_MDET_TX);
10303                if (reg & I40E_PF_MDET_TX_VALID_MASK) {
10304                        wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
10305                        dev_dbg(&pf->pdev->dev, "TX driver issue detected on PF\n");
10306                }
10307                reg = rd32(hw, I40E_PF_MDET_RX);
10308                if (reg & I40E_PF_MDET_RX_VALID_MASK) {
10309                        wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
10310                        dev_dbg(&pf->pdev->dev, "RX driver issue detected on PF\n");
10311                }
10312        }
10313
10314        /* see if one of the VFs needs its hand slapped */
10315        for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
10316                vf = &(pf->vf[i]);
10317                reg = rd32(hw, I40E_VP_MDET_TX(i));
10318                if (reg & I40E_VP_MDET_TX_VALID_MASK) {
10319                        wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
10320                        vf->num_mdd_events++;
10321                        dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
10322                                 i);
10323                        dev_info(&pf->pdev->dev,
10324                                 "Use PF Control I/F to re-enable the VF\n");
10325                        set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10326                }
10327
10328                reg = rd32(hw, I40E_VP_MDET_RX(i));
10329                if (reg & I40E_VP_MDET_RX_VALID_MASK) {
10330                        wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
10331                        vf->num_mdd_events++;
10332                        dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
10333                                 i);
10334                        dev_info(&pf->pdev->dev,
10335                                 "Use PF Control I/F to re-enable the VF\n");
10336                        set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10337                }
10338        }
10339
10340        /* re-enable mdd interrupt cause */
10341        clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
10342        reg = rd32(hw, I40E_PFINT_ICR0_ENA);
10343        reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
10344        wr32(hw, I40E_PFINT_ICR0_ENA, reg);
10345        i40e_flush(hw);
10346}
10347
10348static const char *i40e_tunnel_name(u8 type)
10349{
10350        switch (type) {
10351        case UDP_TUNNEL_TYPE_VXLAN:
10352                return "vxlan";
10353        case UDP_TUNNEL_TYPE_GENEVE:
10354                return "geneve";
10355        default:
10356                return "unknown";
10357        }
10358}
10359
10360/**
10361 * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
10362 * @pf: board private structure
10363 **/
10364static void i40e_sync_udp_filters(struct i40e_pf *pf)
10365{
10366        int i;
10367
10368        /* loop through and set pending bit for all active UDP filters */
10369        for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10370                if (pf->udp_ports[i].port)
10371                        pf->pending_udp_bitmap |= BIT_ULL(i);
10372        }
10373
10374        set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
10375}
10376
10377/**
10378 * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
10379 * @pf: board private structure
10380 **/
10381static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
10382{
10383        struct i40e_hw *hw = &pf->hw;
10384        u8 filter_index, type;
10385        u16 port;
10386        int i;
10387
10388        if (!test_and_clear_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state))
10389                return;
10390
10391        /* acquire RTNL to maintain state of flags and port requests */
10392        rtnl_lock();
10393
10394        for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10395                if (pf->pending_udp_bitmap & BIT_ULL(i)) {
10396                        struct i40e_udp_port_config *udp_port;
10397                        i40e_status ret = 0;
10398
10399                        udp_port = &pf->udp_ports[i];
10400                        pf->pending_udp_bitmap &= ~BIT_ULL(i);
10401
10402                        port = READ_ONCE(udp_port->port);
10403                        type = READ_ONCE(udp_port->type);
10404                        filter_index = READ_ONCE(udp_port->filter_index);
10405
10406                        /* release RTNL while we wait on AQ command */
10407                        rtnl_unlock();
10408
10409                        if (port)
10410                                ret = i40e_aq_add_udp_tunnel(hw, port,
10411                                                             type,
10412                                                             &filter_index,
10413                                                             NULL);
10414                        else if (filter_index != I40E_UDP_PORT_INDEX_UNUSED)
10415                                ret = i40e_aq_del_udp_tunnel(hw, filter_index,
10416                                                             NULL);
10417
10418                        /* reacquire RTNL so we can update filter_index */
10419                        rtnl_lock();
10420
10421                        if (ret) {
10422                                dev_info(&pf->pdev->dev,
10423                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
10424                                         i40e_tunnel_name(type),
10425                                         port ? "add" : "delete",
10426                                         port,
10427                                         filter_index,
10428                                         i40e_stat_str(&pf->hw, ret),
10429                                         i40e_aq_str(&pf->hw,
10430                                                     pf->hw.aq.asq_last_status));
10431                                if (port) {
10432                                        /* failed to add, just reset port,
10433                                         * drop pending bit for any deletion
10434                                         */
10435                                        udp_port->port = 0;
10436                                        pf->pending_udp_bitmap &= ~BIT_ULL(i);
10437                                }
10438                        } else if (port) {
10439                                /* record filter index on success */
10440                                udp_port->filter_index = filter_index;
10441                        }
10442                }
10443        }
10444
10445        rtnl_unlock();
10446}
10447
10448/**
10449 * i40e_service_task - Run the driver's async subtasks
10450 * @work: pointer to work_struct containing our data
10451 **/
10452static void i40e_service_task(struct work_struct *work)
10453{
10454        struct i40e_pf *pf = container_of(work,
10455                                          struct i40e_pf,
10456                                          service_task);
10457        unsigned long start_time = jiffies;
10458
10459        /* don't bother with service tasks if a reset is in progress */
10460        if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
10461            test_bit(__I40E_SUSPENDED, pf->state))
10462                return;
10463
10464        if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
10465                return;
10466
10467        if (!test_bit(__I40E_RECOVERY_MODE, pf->state)) {
10468                i40e_detect_recover_hung(pf->vsi[pf->lan_vsi]);
10469                i40e_sync_filters_subtask(pf);
10470                i40e_reset_subtask(pf);
10471                i40e_handle_mdd_event(pf);
10472                i40e_vc_process_vflr_event(pf);
10473                i40e_watchdog_subtask(pf);
10474                i40e_fdir_reinit_subtask(pf);
10475                if (test_and_clear_bit(__I40E_CLIENT_RESET, pf->state)) {
10476                        /* Client subtask will reopen next time through. */
10477                        i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi],
10478                                                           true);
10479                } else {
10480                        i40e_client_subtask(pf);
10481                        if (test_and_clear_bit(__I40E_CLIENT_L2_CHANGE,
10482                                               pf->state))
10483                                i40e_notify_client_of_l2_param_changes(
10484                                                                pf->vsi[pf->lan_vsi]);
10485                }
10486                i40e_sync_filters_subtask(pf);
10487                i40e_sync_udp_filters_subtask(pf);
10488        } else {
10489                i40e_reset_subtask(pf);
10490        }
10491
10492        i40e_clean_adminq_subtask(pf);
10493
10494        /* flush memory to make sure state is correct before next watchdog */
10495        smp_mb__before_atomic();
10496        clear_bit(__I40E_SERVICE_SCHED, pf->state);
10497
10498        /* If the tasks have taken longer than one timer cycle or there
10499         * is more work to be done, reschedule the service task now
10500         * rather than wait for the timer to tick again.
10501         */
10502        if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
10503            test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)             ||
10504            test_bit(__I40E_MDD_EVENT_PENDING, pf->state)                ||
10505            test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
10506                i40e_service_event_schedule(pf);
10507}
10508
10509/**
10510 * i40e_service_timer - timer callback
10511 * @data: pointer to PF struct
10512 **/
10513static void i40e_service_timer(struct timer_list *t)
10514{
10515        struct i40e_pf *pf = from_timer(pf, t, service_timer);
10516
10517        mod_timer(&pf->service_timer,
10518                  round_jiffies(jiffies + pf->service_timer_period));
10519        i40e_service_event_schedule(pf);
10520}
10521
10522/**
10523 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
10524 * @vsi: the VSI being configured
10525 **/
10526static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
10527{
10528        struct i40e_pf *pf = vsi->back;
10529
10530        switch (vsi->type) {
10531        case I40E_VSI_MAIN:
10532                vsi->alloc_queue_pairs = pf->num_lan_qps;
10533                if (!vsi->num_tx_desc)
10534                        vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10535                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10536                if (!vsi->num_rx_desc)
10537                        vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10538                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10539                if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10540                        vsi->num_q_vectors = pf->num_lan_msix;
10541                else
10542                        vsi->num_q_vectors = 1;
10543
10544                break;
10545
10546        case I40E_VSI_FDIR:
10547                vsi->alloc_queue_pairs = 1;
10548                vsi->num_tx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10549                                         I40E_REQ_DESCRIPTOR_MULTIPLE);
10550                vsi->num_rx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10551                                         I40E_REQ_DESCRIPTOR_MULTIPLE);
10552                vsi->num_q_vectors = pf->num_fdsb_msix;
10553                break;
10554
10555        case I40E_VSI_VMDQ2:
10556                vsi->alloc_queue_pairs = pf->num_vmdq_qps;
10557                if (!vsi->num_tx_desc)
10558                        vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10559                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10560                if (!vsi->num_rx_desc)
10561                        vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10562                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10563                vsi->num_q_vectors = pf->num_vmdq_msix;
10564                break;
10565
10566        case I40E_VSI_SRIOV:
10567                vsi->alloc_queue_pairs = pf->num_vf_qps;
10568                if (!vsi->num_tx_desc)
10569                        vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10570                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10571                if (!vsi->num_rx_desc)
10572                        vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10573                                                 I40E_REQ_DESCRIPTOR_MULTIPLE);
10574                break;
10575
10576        default:
10577                WARN_ON(1);
10578                return -ENODATA;
10579        }
10580
10581        return 0;
10582}
10583
10584/**
10585 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
10586 * @vsi: VSI pointer
10587 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
10588 *
10589 * On error: returns error code (negative)
10590 * On success: returns 0
10591 **/
10592static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
10593{
10594        struct i40e_ring **next_rings;
10595        int size;
10596        int ret = 0;
10597
10598        /* allocate memory for both Tx, XDP Tx and Rx ring pointers */
10599        size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
10600               (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
10601        vsi->tx_rings = kzalloc(size, GFP_KERNEL);
10602        if (!vsi->tx_rings)
10603                return -ENOMEM;
10604        next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
10605        if (i40e_enabled_xdp_vsi(vsi)) {
10606                vsi->xdp_rings = next_rings;
10607                next_rings += vsi->alloc_queue_pairs;
10608        }
10609        vsi->rx_rings = next_rings;
10610
10611        if (alloc_qvectors) {
10612                /* allocate memory for q_vector pointers */
10613                size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
10614                vsi->q_vectors = kzalloc(size, GFP_KERNEL);
10615                if (!vsi->q_vectors) {
10616                        ret = -ENOMEM;
10617                        goto err_vectors;
10618                }
10619        }
10620        return ret;
10621
10622err_vectors:
10623        kfree(vsi->tx_rings);
10624        return ret;
10625}
10626
10627/**
10628 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
10629 * @pf: board private structure
10630 * @type: type of VSI
10631 *
10632 * On error: returns error code (negative)
10633 * On success: returns vsi index in PF (positive)
10634 **/
10635static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
10636{
10637        int ret = -ENODEV;
10638        struct i40e_vsi *vsi;
10639        int vsi_idx;
10640        int i;
10641
10642        /* Need to protect the allocation of the VSIs at the PF level */
10643        mutex_lock(&pf->switch_mutex);
10644
10645        /* VSI list may be fragmented if VSI creation/destruction has
10646         * been happening.  We can afford to do a quick scan to look
10647         * for any free VSIs in the list.
10648         *
10649         * find next empty vsi slot, looping back around if necessary
10650         */
10651        i = pf->next_vsi;
10652        while (i < pf->num_alloc_vsi && pf->vsi[i])
10653                i++;
10654        if (i >= pf->num_alloc_vsi) {
10655                i = 0;
10656                while (i < pf->next_vsi && pf->vsi[i])
10657                        i++;
10658        }
10659
10660        if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
10661                vsi_idx = i;             /* Found one! */
10662        } else {
10663                ret = -ENODEV;
10664                goto unlock_pf;  /* out of VSI slots! */
10665        }
10666        pf->next_vsi = ++i;
10667
10668        vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
10669        if (!vsi) {
10670                ret = -ENOMEM;
10671                goto unlock_pf;
10672        }
10673        vsi->type = type;
10674        vsi->back = pf;
10675        set_bit(__I40E_VSI_DOWN, vsi->state);
10676        vsi->flags = 0;
10677        vsi->idx = vsi_idx;
10678        vsi->int_rate_limit = 0;
10679        vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
10680                                pf->rss_table_size : 64;
10681        vsi->netdev_registered = false;
10682        vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
10683        hash_init(vsi->mac_filter_hash);
10684        vsi->irqs_ready = false;
10685
10686        if (type == I40E_VSI_MAIN) {
10687                vsi->af_xdp_zc_qps = bitmap_zalloc(pf->num_lan_qps, GFP_KERNEL);
10688                if (!vsi->af_xdp_zc_qps)
10689                        goto err_rings;
10690        }
10691
10692        ret = i40e_set_num_rings_in_vsi(vsi);
10693        if (ret)
10694                goto err_rings;
10695
10696        ret = i40e_vsi_alloc_arrays(vsi, true);
10697        if (ret)
10698                goto err_rings;
10699
10700        /* Setup default MSIX irq handler for VSI */
10701        i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
10702
10703        /* Initialize VSI lock */
10704        spin_lock_init(&vsi->mac_filter_hash_lock);
10705        pf->vsi[vsi_idx] = vsi;
10706        ret = vsi_idx;
10707        goto unlock_pf;
10708
10709err_rings:
10710        bitmap_free(vsi->af_xdp_zc_qps);
10711        pf->next_vsi = i - 1;
10712        kfree(vsi);
10713unlock_pf:
10714        mutex_unlock(&pf->switch_mutex);
10715        return ret;
10716}
10717
10718/**
10719 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
10720 * @vsi: VSI pointer
10721 * @free_qvectors: a bool to specify if q_vectors need to be freed.
10722 *
10723 * On error: returns error code (negative)
10724 * On success: returns 0
10725 **/
10726static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
10727{
10728        /* free the ring and vector containers */
10729        if (free_qvectors) {
10730                kfree(vsi->q_vectors);
10731                vsi->q_vectors = NULL;
10732        }
10733        kfree(vsi->tx_rings);
10734        vsi->tx_rings = NULL;
10735        vsi->rx_rings = NULL;
10736        vsi->xdp_rings = NULL;
10737}
10738
10739/**
10740 * i40e_clear_rss_config_user - clear the user configured RSS hash keys
10741 * and lookup table
10742 * @vsi: Pointer to VSI structure
10743 */
10744static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
10745{
10746        if (!vsi)
10747                return;
10748
10749        kfree(vsi->rss_hkey_user);
10750        vsi->rss_hkey_user = NULL;
10751
10752        kfree(vsi->rss_lut_user);
10753        vsi->rss_lut_user = NULL;
10754}
10755
10756/**
10757 * i40e_vsi_clear - Deallocate the VSI provided
10758 * @vsi: the VSI being un-configured
10759 **/
10760static int i40e_vsi_clear(struct i40e_vsi *vsi)
10761{
10762        struct i40e_pf *pf;
10763
10764        if (!vsi)
10765                return 0;
10766
10767        if (!vsi->back)
10768                goto free_vsi;
10769        pf = vsi->back;
10770
10771        mutex_lock(&pf->switch_mutex);
10772        if (!pf->vsi[vsi->idx]) {
10773                dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](type %d)\n",
10774                        vsi->idx, vsi->idx, vsi->type);
10775                goto unlock_vsi;
10776        }
10777
10778        if (pf->vsi[vsi->idx] != vsi) {
10779                dev_err(&pf->pdev->dev,
10780                        "pf->vsi[%d](type %d) != vsi[%d](type %d): no free!\n",
10781                        pf->vsi[vsi->idx]->idx,
10782                        pf->vsi[vsi->idx]->type,
10783                        vsi->idx, vsi->type);
10784                goto unlock_vsi;
10785        }
10786
10787        /* updates the PF for this cleared vsi */
10788        i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
10789        i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
10790
10791        bitmap_free(vsi->af_xdp_zc_qps);
10792        i40e_vsi_free_arrays(vsi, true);
10793        i40e_clear_rss_config_user(vsi);
10794
10795        pf->vsi[vsi->idx] = NULL;
10796        if (vsi->idx < pf->next_vsi)
10797                pf->next_vsi = vsi->idx;
10798
10799unlock_vsi:
10800        mutex_unlock(&pf->switch_mutex);
10801free_vsi:
10802        kfree(vsi);
10803
10804        return 0;
10805}
10806
10807/**
10808 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
10809 * @vsi: the VSI being cleaned
10810 **/
10811static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
10812{
10813        int i;
10814
10815        if (vsi->tx_rings && vsi->tx_rings[0]) {
10816                for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10817                        kfree_rcu(vsi->tx_rings[i], rcu);
10818                        vsi->tx_rings[i] = NULL;
10819                        vsi->rx_rings[i] = NULL;
10820                        if (vsi->xdp_rings)
10821                                vsi->xdp_rings[i] = NULL;
10822                }
10823        }
10824}
10825
10826/**
10827 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
10828 * @vsi: the VSI being configured
10829 **/
10830static int i40e_alloc_rings(struct i40e_vsi *vsi)
10831{
10832        int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
10833        struct i40e_pf *pf = vsi->back;
10834        struct i40e_ring *ring;
10835
10836        /* Set basic values in the rings to be used later during open() */
10837        for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10838                /* allocate space for both Tx and Rx in one shot */
10839                ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
10840                if (!ring)
10841                        goto err_out;
10842
10843                ring->queue_index = i;
10844                ring->reg_idx = vsi->base_queue + i;
10845                ring->ring_active = false;
10846                ring->vsi = vsi;
10847                ring->netdev = vsi->netdev;
10848                ring->dev = &pf->pdev->dev;
10849                ring->count = vsi->num_tx_desc;
10850                ring->size = 0;
10851                ring->dcb_tc = 0;
10852                if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10853                        ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10854                ring->itr_setting = pf->tx_itr_default;
10855                vsi->tx_rings[i] = ring++;
10856
10857                if (!i40e_enabled_xdp_vsi(vsi))
10858                        goto setup_rx;
10859
10860                ring->queue_index = vsi->alloc_queue_pairs + i;
10861                ring->reg_idx = vsi->base_queue + ring->queue_index;
10862                ring->ring_active = false;
10863                ring->vsi = vsi;
10864                ring->netdev = NULL;
10865                ring->dev = &pf->pdev->dev;
10866                ring->count = vsi->num_tx_desc;
10867                ring->size = 0;
10868                ring->dcb_tc = 0;
10869                if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10870                        ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10871                set_ring_xdp(ring);
10872                ring->itr_setting = pf->tx_itr_default;
10873                vsi->xdp_rings[i] = ring++;
10874
10875setup_rx:
10876                ring->queue_index = i;
10877                ring->reg_idx = vsi->base_queue + i;
10878                ring->ring_active = false;
10879                ring->vsi = vsi;
10880                ring->netdev = vsi->netdev;
10881                ring->dev = &pf->pdev->dev;
10882                ring->count = vsi->num_rx_desc;
10883                ring->size = 0;
10884                ring->dcb_tc = 0;
10885                ring->itr_setting = pf->rx_itr_default;
10886                vsi->rx_rings[i] = ring;
10887        }
10888
10889        return 0;
10890
10891err_out:
10892        i40e_vsi_clear_rings(vsi);
10893        return -ENOMEM;
10894}
10895
10896/**
10897 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
10898 * @pf: board private structure
10899 * @vectors: the number of MSI-X vectors to request
10900 *
10901 * Returns the number of vectors reserved, or error
10902 **/
10903static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
10904{
10905        vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
10906                                        I40E_MIN_MSIX, vectors);
10907        if (vectors < 0) {
10908                dev_info(&pf->pdev->dev,
10909                         "MSI-X vector reservation failed: %d\n", vectors);
10910                vectors = 0;
10911        }
10912
10913        return vectors;
10914}
10915
10916/**
10917 * i40e_init_msix - Setup the MSIX capability
10918 * @pf: board private structure
10919 *
10920 * Work with the OS to set up the MSIX vectors needed.
10921 *
10922 * Returns the number of vectors reserved or negative on failure
10923 **/
10924static int i40e_init_msix(struct i40e_pf *pf)
10925{
10926        struct i40e_hw *hw = &pf->hw;
10927        int cpus, extra_vectors;
10928        int vectors_left;
10929        int v_budget, i;
10930        int v_actual;
10931        int iwarp_requested = 0;
10932
10933        if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10934                return -ENODEV;
10935
10936        /* The number of vectors we'll request will be comprised of:
10937         *   - Add 1 for "other" cause for Admin Queue events, etc.
10938         *   - The number of LAN queue pairs
10939         *      - Queues being used for RSS.
10940         *              We don't need as many as max_rss_size vectors.
10941         *              use rss_size instead in the calculation since that
10942         *              is governed by number of cpus in the system.
10943         *      - assumes symmetric Tx/Rx pairing
10944         *   - The number of VMDq pairs
10945         *   - The CPU count within the NUMA node if iWARP is enabled
10946         * Once we count this up, try the request.
10947         *
10948         * If we can't get what we want, we'll simplify to nearly nothing
10949         * and try again.  If that still fails, we punt.
10950         */
10951        vectors_left = hw->func_caps.num_msix_vectors;
10952        v_budget = 0;
10953
10954        /* reserve one vector for miscellaneous handler */
10955        if (vectors_left) {
10956                v_budget++;
10957                vectors_left--;
10958        }
10959
10960        /* reserve some vectors for the main PF traffic queues. Initially we
10961         * only reserve at most 50% of the available vectors, in the case that
10962         * the number of online CPUs is large. This ensures that we can enable
10963         * extra features as well. Once we've enabled the other features, we
10964         * will use any remaining vectors to reach as close as we can to the
10965         * number of online CPUs.
10966         */
10967        cpus = num_online_cpus();
10968        pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
10969        vectors_left -= pf->num_lan_msix;
10970
10971        /* reserve one vector for sideband flow director */
10972        if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10973                if (vectors_left) {
10974                        pf->num_fdsb_msix = 1;
10975                        v_budget++;
10976                        vectors_left--;
10977                } else {
10978                        pf->num_fdsb_msix = 0;
10979                }
10980        }
10981
10982        /* can we reserve enough for iWARP? */
10983        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10984                iwarp_requested = pf->num_iwarp_msix;
10985
10986                if (!vectors_left)
10987                        pf->num_iwarp_msix = 0;
10988                else if (vectors_left < pf->num_iwarp_msix)
10989                        pf->num_iwarp_msix = 1;
10990                v_budget += pf->num_iwarp_msix;
10991                vectors_left -= pf->num_iwarp_msix;
10992        }
10993
10994        /* any vectors left over go for VMDq support */
10995        if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
10996                if (!vectors_left) {
10997                        pf->num_vmdq_msix = 0;
10998                        pf->num_vmdq_qps = 0;
10999                } else {
11000                        int vmdq_vecs_wanted =
11001                                pf->num_vmdq_vsis * pf->num_vmdq_qps;
11002                        int vmdq_vecs =
11003                                min_t(int, vectors_left, vmdq_vecs_wanted);
11004
11005                        /* if we're short on vectors for what's desired, we limit
11006                         * the queues per vmdq.  If this is still more than are
11007                         * available, the user will need to change the number of
11008                         * queues/vectors used by the PF later with the ethtool
11009                         * channels command
11010                         */
11011                        if (vectors_left < vmdq_vecs_wanted) {
11012                                pf->num_vmdq_qps = 1;
11013                                vmdq_vecs_wanted = pf->num_vmdq_vsis;
11014                                vmdq_vecs = min_t(int,
11015                                                  vectors_left,
11016                                                  vmdq_vecs_wanted);
11017                        }
11018                        pf->num_vmdq_msix = pf->num_vmdq_qps;
11019
11020                        v_budget += vmdq_vecs;
11021                        vectors_left -= vmdq_vecs;
11022                }
11023        }
11024
11025        /* On systems with a large number of SMP cores, we previously limited
11026         * the number of vectors for num_lan_msix to be at most 50% of the
11027         * available vectors, to allow for other features. Now, we add back
11028         * the remaining vectors. However, we ensure that the total
11029         * num_lan_msix will not exceed num_online_cpus(). To do this, we
11030         * calculate the number of vectors we can add without going over the
11031         * cap of CPUs. For systems with a small number of CPUs this will be
11032         * zero.
11033         */
11034        extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
11035        pf->num_lan_msix += extra_vectors;
11036        vectors_left -= extra_vectors;
11037
11038        WARN(vectors_left < 0,
11039             "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
11040
11041        v_budget += pf->num_lan_msix;
11042        pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
11043                                   GFP_KERNEL);
11044        if (!pf->msix_entries)
11045                return -ENOMEM;
11046
11047        for (i = 0; i < v_budget; i++)
11048                pf->msix_entries[i].entry = i;
11049        v_actual = i40e_reserve_msix_vectors(pf, v_budget);
11050
11051        if (v_actual < I40E_MIN_MSIX) {
11052                pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
11053                kfree(pf->msix_entries);
11054                pf->msix_entries = NULL;
11055                pci_disable_msix(pf->pdev);
11056                return -ENODEV;
11057
11058        } else if (v_actual == I40E_MIN_MSIX) {
11059                /* Adjust for minimal MSIX use */
11060                pf->num_vmdq_vsis = 0;
11061                pf->num_vmdq_qps = 0;
11062                pf->num_lan_qps = 1;
11063                pf->num_lan_msix = 1;
11064
11065        } else if (v_actual != v_budget) {
11066                /* If we have limited resources, we will start with no vectors
11067                 * for the special features and then allocate vectors to some
11068                 * of these features based on the policy and at the end disable
11069                 * the features that did not get any vectors.
11070                 */
11071                int vec;
11072
11073                dev_info(&pf->pdev->dev,
11074                         "MSI-X vector limit reached with %d, wanted %d, attempting to redistribute vectors\n",
11075                         v_actual, v_budget);
11076                /* reserve the misc vector */
11077                vec = v_actual - 1;
11078
11079                /* Scale vector usage down */
11080                pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
11081                pf->num_vmdq_vsis = 1;
11082                pf->num_vmdq_qps = 1;
11083
11084                /* partition out the remaining vectors */
11085                switch (vec) {
11086                case 2:
11087                        pf->num_lan_msix = 1;
11088                        break;
11089                case 3:
11090                        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11091                                pf->num_lan_msix = 1;
11092                                pf->num_iwarp_msix = 1;
11093                        } else {
11094                                pf->num_lan_msix = 2;
11095                        }
11096                        break;
11097                default:
11098                        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11099                                pf->num_iwarp_msix = min_t(int, (vec / 3),
11100                                                 iwarp_requested);
11101                                pf->num_vmdq_vsis = min_t(int, (vec / 3),
11102                                                  I40E_DEFAULT_NUM_VMDQ_VSI);
11103                        } else {
11104                                pf->num_vmdq_vsis = min_t(int, (vec / 2),
11105                                                  I40E_DEFAULT_NUM_VMDQ_VSI);
11106                        }
11107                        if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11108                                pf->num_fdsb_msix = 1;
11109                                vec--;
11110                        }
11111                        pf->num_lan_msix = min_t(int,
11112                               (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
11113                                                              pf->num_lan_msix);
11114                        pf->num_lan_qps = pf->num_lan_msix;
11115                        break;
11116                }
11117        }
11118
11119        if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
11120            (pf->num_fdsb_msix == 0)) {
11121                dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
11122                pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
11123                pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11124        }
11125        if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
11126            (pf->num_vmdq_msix == 0)) {
11127                dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
11128                pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
11129        }
11130
11131        if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
11132            (pf->num_iwarp_msix == 0)) {
11133                dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
11134                pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11135        }
11136        i40e_debug(&pf->hw, I40E_DEBUG_INIT,
11137                   "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
11138                   pf->num_lan_msix,
11139                   pf->num_vmdq_msix * pf->num_vmdq_vsis,
11140                   pf->num_fdsb_msix,
11141                   pf->num_iwarp_msix);
11142
11143        return v_actual;
11144}
11145
11146/**
11147 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
11148 * @vsi: the VSI being configured
11149 * @v_idx: index of the vector in the vsi struct
11150 * @cpu: cpu to be used on affinity_mask
11151 *
11152 * We allocate one q_vector.  If allocation fails we return -ENOMEM.
11153 **/
11154static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
11155{
11156        struct i40e_q_vector *q_vector;
11157
11158        /* allocate q_vector */
11159        q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
11160        if (!q_vector)
11161                return -ENOMEM;
11162
11163        q_vector->vsi = vsi;
11164        q_vector->v_idx = v_idx;
11165        cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
11166
11167        if (vsi->netdev)
11168                netif_napi_add(vsi->netdev, &q_vector->napi,
11169                               i40e_napi_poll, NAPI_POLL_WEIGHT);
11170
11171        /* tie q_vector and vsi together */
11172        vsi->q_vectors[v_idx] = q_vector;
11173
11174        return 0;
11175}
11176
11177/**
11178 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
11179 * @vsi: the VSI being configured
11180 *
11181 * We allocate one q_vector per queue interrupt.  If allocation fails we
11182 * return -ENOMEM.
11183 **/
11184static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
11185{
11186        struct i40e_pf *pf = vsi->back;
11187        int err, v_idx, num_q_vectors, current_cpu;
11188
11189        /* if not MSIX, give the one vector only to the LAN VSI */
11190        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
11191                num_q_vectors = vsi->num_q_vectors;
11192        else if (vsi == pf->vsi[pf->lan_vsi])
11193                num_q_vectors = 1;
11194        else
11195                return -EINVAL;
11196
11197        current_cpu = cpumask_first(cpu_online_mask);
11198
11199        for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
11200                err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
11201                if (err)
11202                        goto err_out;
11203                current_cpu = cpumask_next(current_cpu, cpu_online_mask);
11204                if (unlikely(current_cpu >= nr_cpu_ids))
11205                        current_cpu = cpumask_first(cpu_online_mask);
11206        }
11207
11208        return 0;
11209
11210err_out:
11211        while (v_idx--)
11212                i40e_free_q_vector(vsi, v_idx);
11213
11214        return err;
11215}
11216
11217/**
11218 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
11219 * @pf: board private structure to initialize
11220 **/
11221static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
11222{
11223        int vectors = 0;
11224        ssize_t size;
11225
11226        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11227                vectors = i40e_init_msix(pf);
11228                if (vectors < 0) {
11229                        pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
11230                                       I40E_FLAG_IWARP_ENABLED  |
11231                                       I40E_FLAG_RSS_ENABLED    |
11232                                       I40E_FLAG_DCB_CAPABLE    |
11233                                       I40E_FLAG_DCB_ENABLED    |
11234                                       I40E_FLAG_SRIOV_ENABLED  |
11235                                       I40E_FLAG_FD_SB_ENABLED  |
11236                                       I40E_FLAG_FD_ATR_ENABLED |
11237                                       I40E_FLAG_VMDQ_ENABLED);
11238                        pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11239
11240                        /* rework the queue expectations without MSIX */
11241                        i40e_determine_queue_usage(pf);
11242                }
11243        }
11244
11245        if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11246            (pf->flags & I40E_FLAG_MSI_ENABLED)) {
11247                dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
11248                vectors = pci_enable_msi(pf->pdev);
11249                if (vectors < 0) {
11250                        dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
11251                                 vectors);
11252                        pf->flags &= ~I40E_FLAG_MSI_ENABLED;
11253                }
11254                vectors = 1;  /* one MSI or Legacy vector */
11255        }
11256
11257        if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
11258                dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
11259
11260        /* set up vector assignment tracking */
11261        size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
11262        pf->irq_pile = kzalloc(size, GFP_KERNEL);
11263        if (!pf->irq_pile)
11264                return -ENOMEM;
11265
11266        pf->irq_pile->num_entries = vectors;
11267        pf->irq_pile->search_hint = 0;
11268
11269        /* track first vector for misc interrupts, ignore return */
11270        (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
11271
11272        return 0;
11273}
11274
11275/**
11276 * i40e_restore_interrupt_scheme - Restore the interrupt scheme
11277 * @pf: private board data structure
11278 *
11279 * Restore the interrupt scheme that was cleared when we suspended the
11280 * device. This should be called during resume to re-allocate the q_vectors
11281 * and reacquire IRQs.
11282 */
11283static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
11284{
11285        int err, i;
11286
11287        /* We cleared the MSI and MSI-X flags when disabling the old interrupt
11288         * scheme. We need to re-enabled them here in order to attempt to
11289         * re-acquire the MSI or MSI-X vectors
11290         */
11291        pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
11292
11293        err = i40e_init_interrupt_scheme(pf);
11294        if (err)
11295                return err;
11296
11297        /* Now that we've re-acquired IRQs, we need to remap the vectors and
11298         * rings together again.
11299         */
11300        for (i = 0; i < pf->num_alloc_vsi; i++) {
11301                if (pf->vsi[i]) {
11302                        err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
11303                        if (err)
11304                                goto err_unwind;
11305                        i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
11306                }
11307        }
11308
11309        err = i40e_setup_misc_vector(pf);
11310        if (err)
11311                goto err_unwind;
11312
11313        if (pf->flags & I40E_FLAG_IWARP_ENABLED)
11314                i40e_client_update_msix_info(pf);
11315
11316        return 0;
11317
11318err_unwind:
11319        while (i--) {
11320                if (pf->vsi[i])
11321                        i40e_vsi_free_q_vectors(pf->vsi[i]);
11322        }
11323
11324        return err;
11325}
11326
11327/**
11328 * i40e_setup_misc_vector_for_recovery_mode - Setup the misc vector to handle
11329 * non queue events in recovery mode
11330 * @pf: board private structure
11331 *
11332 * This sets up the handler for MSIX 0 or MSI/legacy, which is used to manage
11333 * the non-queue interrupts, e.g. AdminQ and errors in recovery mode.
11334 * This is handled differently than in recovery mode since no Tx/Rx resources
11335 * are being allocated.
11336 **/
11337static int i40e_setup_misc_vector_for_recovery_mode(struct i40e_pf *pf)
11338{
11339        int err;
11340
11341        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11342                err = i40e_setup_misc_vector(pf);
11343
11344                if (err) {
11345                        dev_info(&pf->pdev->dev,
11346                                 "MSI-X misc vector request failed, error %d\n",
11347                                 err);
11348                        return err;
11349                }
11350        } else {
11351                u32 flags = pf->flags & I40E_FLAG_MSI_ENABLED ? 0 : IRQF_SHARED;
11352
11353                err = request_irq(pf->pdev->irq, i40e_intr, flags,
11354                                  pf->int_name, pf);
11355
11356                if (err) {
11357                        dev_info(&pf->pdev->dev,
11358                                 "MSI/legacy misc vector request failed, error %d\n",
11359                                 err);
11360                        return err;
11361                }
11362                i40e_enable_misc_int_causes(pf);
11363                i40e_irq_dynamic_enable_icr0(pf);
11364        }
11365
11366        return 0;
11367}
11368
11369/**
11370 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
11371 * @pf: board private structure
11372 *
11373 * This sets up the handler for MSIX 0, which is used to manage the
11374 * non-queue interrupts, e.g. AdminQ and errors.  This is not used
11375 * when in MSI or Legacy interrupt mode.
11376 **/
11377static int i40e_setup_misc_vector(struct i40e_pf *pf)
11378{
11379        struct i40e_hw *hw = &pf->hw;
11380        int err = 0;
11381
11382        /* Only request the IRQ once, the first time through. */
11383        if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
11384                err = request_irq(pf->msix_entries[0].vector,
11385                                  i40e_intr, 0, pf->int_name, pf);
11386                if (err) {
11387                        clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
11388                        dev_info(&pf->pdev->dev,
11389                                 "request_irq for %s failed: %d\n",
11390                                 pf->int_name, err);
11391                        return -EFAULT;
11392                }
11393        }
11394
11395        i40e_enable_misc_int_causes(pf);
11396
11397        /* associate no queues to the misc vector */
11398        wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
11399        wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
11400
11401        i40e_flush(hw);
11402
11403        i40e_irq_dynamic_enable_icr0(pf);
11404
11405        return err;
11406}
11407
11408/**
11409 * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
11410 * @vsi: Pointer to vsi structure
11411 * @seed: Buffter to store the hash keys
11412 * @lut: Buffer to store the lookup table entries
11413 * @lut_size: Size of buffer to store the lookup table entries
11414 *
11415 * Return 0 on success, negative on failure
11416 */
11417static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
11418                           u8 *lut, u16 lut_size)
11419{
11420        struct i40e_pf *pf = vsi->back;
11421        struct i40e_hw *hw = &pf->hw;
11422        int ret = 0;
11423
11424        if (seed) {
11425                ret = i40e_aq_get_rss_key(hw, vsi->id,
11426                        (struct i40e_aqc_get_set_rss_key_data *)seed);
11427                if (ret) {
11428                        dev_info(&pf->pdev->dev,
11429                                 "Cannot get RSS key, err %s aq_err %s\n",
11430                                 i40e_stat_str(&pf->hw, ret),
11431                                 i40e_aq_str(&pf->hw,
11432                                             pf->hw.aq.asq_last_status));
11433                        return ret;
11434                }
11435        }
11436
11437        if (lut) {
11438                bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
11439
11440                ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
11441                if (ret) {
11442                        dev_info(&pf->pdev->dev,
11443                                 "Cannot get RSS lut, err %s aq_err %s\n",
11444                                 i40e_stat_str(&pf->hw, ret),
11445                                 i40e_aq_str(&pf->hw,
11446                                             pf->hw.aq.asq_last_status));
11447                        return ret;
11448                }
11449        }
11450
11451        return ret;
11452}
11453
11454/**
11455 * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
11456 * @vsi: Pointer to vsi structure
11457 * @seed: RSS hash seed
11458 * @lut: Lookup table
11459 * @lut_size: Lookup table size
11460 *
11461 * Returns 0 on success, negative on failure
11462 **/
11463static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
11464                               const u8 *lut, u16 lut_size)
11465{
11466        struct i40e_pf *pf = vsi->back;
11467        struct i40e_hw *hw = &pf->hw;
11468        u16 vf_id = vsi->vf_id;
11469        u8 i;
11470
11471        /* Fill out hash function seed */
11472        if (seed) {
11473                u32 *seed_dw = (u32 *)seed;
11474
11475                if (vsi->type == I40E_VSI_MAIN) {
11476                        for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11477                                wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
11478                } else if (vsi->type == I40E_VSI_SRIOV) {
11479                        for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
11480                                wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
11481                } else {
11482                        dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
11483                }
11484        }
11485
11486        if (lut) {
11487                u32 *lut_dw = (u32 *)lut;
11488
11489                if (vsi->type == I40E_VSI_MAIN) {
11490                        if (lut_size != I40E_HLUT_ARRAY_SIZE)
11491                                return -EINVAL;
11492                        for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11493                                wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
11494                } else if (vsi->type == I40E_VSI_SRIOV) {
11495                        if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
11496                                return -EINVAL;
11497                        for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
11498                                wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
11499                } else {
11500                        dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
11501                }
11502        }
11503        i40e_flush(hw);
11504
11505        return 0;
11506}
11507
11508/**
11509 * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
11510 * @vsi: Pointer to VSI structure
11511 * @seed: Buffer to store the keys
11512 * @lut: Buffer to store the lookup table entries
11513 * @lut_size: Size of buffer to store the lookup table entries
11514 *
11515 * Returns 0 on success, negative on failure
11516 */
11517static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
11518                            u8 *lut, u16 lut_size)
11519{
11520        struct i40e_pf *pf = vsi->back;
11521        struct i40e_hw *hw = &pf->hw;
11522        u16 i;
11523
11524        if (seed) {
11525                u32 *seed_dw = (u32 *)seed;
11526
11527                for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11528                        seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
11529        }
11530        if (lut) {
11531                u32 *lut_dw = (u32 *)lut;
11532
11533                if (lut_size != I40E_HLUT_ARRAY_SIZE)
11534                        return -EINVAL;
11535                for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11536                        lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
11537        }
11538
11539        return 0;
11540}
11541
11542/**
11543 * i40e_config_rss - Configure RSS keys and lut
11544 * @vsi: Pointer to VSI structure
11545 * @seed: RSS hash seed
11546 * @lut: Lookup table
11547 * @lut_size: Lookup table size
11548 *
11549 * Returns 0 on success, negative on failure
11550 */
11551int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11552{
11553        struct i40e_pf *pf = vsi->back;
11554
11555        if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11556                return i40e_config_rss_aq(vsi, seed, lut, lut_size);
11557        else
11558                return i40e_config_rss_reg(vsi, seed, lut, lut_size);
11559}
11560
11561/**
11562 * i40e_get_rss - Get RSS keys and lut
11563 * @vsi: Pointer to VSI structure
11564 * @seed: Buffer to store the keys
11565 * @lut: Buffer to store the lookup table entries
11566 * @lut_size: Size of buffer to store the lookup table entries
11567 *
11568 * Returns 0 on success, negative on failure
11569 */
11570int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11571{
11572        struct i40e_pf *pf = vsi->back;
11573
11574        if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11575                return i40e_get_rss_aq(vsi, seed, lut, lut_size);
11576        else
11577                return i40e_get_rss_reg(vsi, seed, lut, lut_size);
11578}
11579
11580/**
11581 * i40e_fill_rss_lut - Fill the RSS lookup table with default values
11582 * @pf: Pointer to board private structure
11583 * @lut: Lookup table
11584 * @rss_table_size: Lookup table size
11585 * @rss_size: Range of queue number for hashing
11586 */
11587void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
11588                       u16 rss_table_size, u16 rss_size)
11589{
11590        u16 i;
11591
11592        for (i = 0; i < rss_table_size; i++)
11593                lut[i] = i % rss_size;
11594}
11595
11596/**
11597 * i40e_pf_config_rss - Prepare for RSS if used
11598 * @pf: board private structure
11599 **/
11600static int i40e_pf_config_rss(struct i40e_pf *pf)
11601{
11602        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11603        u8 seed[I40E_HKEY_ARRAY_SIZE];
11604        u8 *lut;
11605        struct i40e_hw *hw = &pf->hw;
11606        u32 reg_val;
11607        u64 hena;
11608        int ret;
11609
11610        /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
11611        hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
11612                ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
11613        hena |= i40e_pf_get_default_rss_hena(pf);
11614
11615        i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
11616        i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
11617
11618        /* Determine the RSS table size based on the hardware capabilities */
11619        reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
11620        reg_val = (pf->rss_table_size == 512) ?
11621                        (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
11622                        (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
11623        i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
11624
11625        /* Determine the RSS size of the VSI */
11626        if (!vsi->rss_size) {
11627                u16 qcount;
11628                /* If the firmware does something weird during VSI init, we
11629                 * could end up with zero TCs. Check for that to avoid
11630                 * divide-by-zero. It probably won't pass traffic, but it also
11631                 * won't panic.
11632                 */
11633                qcount = vsi->num_queue_pairs /
11634                         (vsi->tc_config.numtc ? vsi->tc_config.numtc : 1);
11635                vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11636        }
11637        if (!vsi->rss_size)
11638                return -EINVAL;
11639
11640        lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
11641        if (!lut)
11642                return -ENOMEM;
11643
11644        /* Use user configured lut if there is one, otherwise use default */
11645        if (vsi->rss_lut_user)
11646                memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
11647        else
11648                i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
11649
11650        /* Use user configured hash key if there is one, otherwise
11651         * use default.
11652         */
11653        if (vsi->rss_hkey_user)
11654                memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
11655        else
11656                netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
11657        ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
11658        kfree(lut);
11659
11660        return ret;
11661}
11662
11663/**
11664 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
11665 * @pf: board private structure
11666 * @queue_count: the requested queue count for rss.
11667 *
11668 * returns 0 if rss is not enabled, if enabled returns the final rss queue
11669 * count which may be different from the requested queue count.
11670 * Note: expects to be called while under rtnl_lock()
11671 **/
11672int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
11673{
11674        struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11675        int new_rss_size;
11676
11677        if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
11678                return 0;
11679
11680        queue_count = min_t(int, queue_count, num_online_cpus());
11681        new_rss_size = min_t(int, queue_count, pf->rss_size_max);
11682
11683        if (queue_count != vsi->num_queue_pairs) {
11684                u16 qcount;
11685
11686                vsi->req_queue_pairs = queue_count;
11687                i40e_prep_for_reset(pf, true);
11688
11689                pf->alloc_rss_size = new_rss_size;
11690
11691                i40e_reset_and_rebuild(pf, true, true);
11692
11693                /* Discard the user configured hash keys and lut, if less
11694                 * queues are enabled.
11695                 */
11696                if (queue_count < vsi->rss_size) {
11697                        i40e_clear_rss_config_user(vsi);
11698                        dev_dbg(&pf->pdev->dev,
11699                                "discard user configured hash keys and lut\n");
11700                }
11701
11702                /* Reset vsi->rss_size, as number of enabled queues changed */
11703                qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
11704                vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11705
11706                i40e_pf_config_rss(pf);
11707        }
11708        dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
11709                 vsi->req_queue_pairs, pf->rss_size_max);
11710        return pf->alloc_rss_size;
11711}
11712
11713/**
11714 * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
11715 * @pf: board private structure
11716 **/
11717i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
11718{
11719        i40e_status status;
11720        bool min_valid, max_valid;
11721        u32 max_bw, min_bw;
11722
11723        status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
11724                                           &min_valid, &max_valid);
11725
11726        if (!status) {
11727                if (min_valid)
11728                        pf->min_bw = min_bw;
11729                if (max_valid)
11730                        pf->max_bw = max_bw;
11731        }
11732
11733        return status;
11734}
11735
11736/**
11737 * i40e_set_partition_bw_setting - Set BW settings for this PF partition
11738 * @pf: board private structure
11739 **/
11740i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
11741{
11742        struct i40e_aqc_configure_partition_bw_data bw_data;
11743        i40e_status status;
11744
11745        /* Set the valid bit for this PF */
11746        bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
11747        bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
11748        bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
11749
11750        /* Set the new bandwidths */
11751        status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
11752
11753        return status;
11754}
11755
11756/**
11757 * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
11758 * @pf: board private structure
11759 **/
11760i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
11761{
11762        /* Commit temporary BW setting to permanent NVM image */
11763        enum i40e_admin_queue_err last_aq_status;
11764        i40e_status ret;
11765        u16 nvm_word;
11766
11767        if (pf->hw.partition_id != 1) {
11768                dev_info(&pf->pdev->dev,
11769                         "Commit BW only works on partition 1! This is partition %d",
11770                         pf->hw.partition_id);
11771                ret = I40E_NOT_SUPPORTED;
11772                goto bw_commit_out;
11773        }
11774
11775        /* Acquire NVM for read access */
11776        ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
11777        last_aq_status = pf->hw.aq.asq_last_status;
11778        if (ret) {
11779                dev_info(&pf->pdev->dev,
11780                         "Cannot acquire NVM for read access, err %s aq_err %s\n",
11781                         i40e_stat_str(&pf->hw, ret),
11782                         i40e_aq_str(&pf->hw, last_aq_status));
11783                goto bw_commit_out;
11784        }
11785
11786        /* Read word 0x10 of NVM - SW compatibility word 1 */
11787        ret = i40e_aq_read_nvm(&pf->hw,
11788                               I40E_SR_NVM_CONTROL_WORD,
11789                               0x10, sizeof(nvm_word), &nvm_word,
11790                               false, NULL);
11791        /* Save off last admin queue command status before releasing
11792         * the NVM
11793         */
11794        last_aq_status = pf->hw.aq.asq_last_status;
11795        i40e_release_nvm(&pf->hw);
11796        if (ret) {
11797                dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
11798                         i40e_stat_str(&pf->hw, ret),
11799                         i40e_aq_str(&pf->hw, last_aq_status));
11800                goto bw_commit_out;
11801        }
11802
11803        /* Wait a bit for NVM release to complete */
11804        msleep(50);
11805
11806        /* Acquire NVM for write access */
11807        ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
11808        last_aq_status = pf->hw.aq.asq_last_status;
11809        if (ret) {
11810                dev_info(&pf->pdev->dev,
11811                         "Cannot acquire NVM for write access, err %s aq_err %s\n",
11812                         i40e_stat_str(&pf->hw, ret),
11813                         i40e_aq_str(&pf->hw, last_aq_status));
11814                goto bw_commit_out;
11815        }
11816        /* Write it back out unchanged to initiate update NVM,
11817         * which will force a write of the shadow (alt) RAM to
11818         * the NVM - thus storing the bandwidth values permanently.
11819         */
11820        ret = i40e_aq_update_nvm(&pf->hw,
11821                                 I40E_SR_NVM_CONTROL_WORD,
11822                                 0x10, sizeof(nvm_word),
11823                                 &nvm_word, true, 0, NULL);
11824        /* Save off last admin queue command status before releasing
11825         * the NVM
11826         */
11827        last_aq_status = pf->hw.aq.asq_last_status;
11828        i40e_release_nvm(&pf->hw);
11829        if (ret)
11830                dev_info(&pf->pdev->dev,
11831                         "BW settings NOT SAVED, err %s aq_err %s\n",
11832                         i40e_stat_str(&pf->hw, ret),
11833                         i40e_aq_str(&pf->hw, last_aq_status));
11834bw_commit_out:
11835
11836        return ret;
11837}
11838
11839/**
11840 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
11841 * @pf: board private structure to initialize
11842 *
11843 * i40e_sw_init initializes the Adapter private data structure.
11844 * Fields are initialized based on PCI device information and
11845 * OS network device settings (MTU size).
11846 **/
11847static int i40e_sw_init(struct i40e_pf *pf)
11848{
11849        int err = 0;
11850        int size;
11851
11852        /* Set default capability flags */
11853        pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
11854                    I40E_FLAG_MSI_ENABLED     |
11855                    I40E_FLAG_MSIX_ENABLED;
11856
11857        /* Set default ITR */
11858        pf->rx_itr_default = I40E_ITR_RX_DEF;
11859        pf->tx_itr_default = I40E_ITR_TX_DEF;
11860
11861        /* Depending on PF configurations, it is possible that the RSS
11862         * maximum might end up larger than the available queues
11863         */
11864        pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
11865        pf->alloc_rss_size = 1;
11866        pf->rss_table_size = pf->hw.func_caps.rss_table_size;
11867        pf->rss_size_max = min_t(int, pf->rss_size_max,
11868                                 pf->hw.func_caps.num_tx_qp);
11869        if (pf->hw.func_caps.rss) {
11870                pf->flags |= I40E_FLAG_RSS_ENABLED;
11871                pf->alloc_rss_size = min_t(int, pf->rss_size_max,
11872                                           num_online_cpus());
11873        }
11874
11875        /* MFP mode enabled */
11876        if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
11877                pf->flags |= I40E_FLAG_MFP_ENABLED;
11878                dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
11879                if (i40e_get_partition_bw_setting(pf)) {
11880                        dev_warn(&pf->pdev->dev,
11881                                 "Could not get partition bw settings\n");
11882                } else {
11883                        dev_info(&pf->pdev->dev,
11884                                 "Partition BW Min = %8.8x, Max = %8.8x\n",
11885                                 pf->min_bw, pf->max_bw);
11886
11887                        /* nudge the Tx scheduler */
11888                        i40e_set_partition_bw_setting(pf);
11889                }
11890        }
11891
11892        if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
11893            (pf->hw.func_caps.fd_filters_best_effort > 0)) {
11894                pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
11895                pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
11896                if (pf->flags & I40E_FLAG_MFP_ENABLED &&
11897                    pf->hw.num_partitions > 1)
11898                        dev_info(&pf->pdev->dev,
11899                                 "Flow Director Sideband mode Disabled in MFP mode\n");
11900                else
11901                        pf->flags |= I40E_FLAG_FD_SB_ENABLED;
11902                pf->fdir_pf_filter_count =
11903                                 pf->hw.func_caps.fd_filters_guaranteed;
11904                pf->hw.fdir_shared_filter_count =
11905                                 pf->hw.func_caps.fd_filters_best_effort;
11906        }
11907
11908        if (pf->hw.mac.type == I40E_MAC_X722) {
11909                pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
11910                                    I40E_HW_128_QP_RSS_CAPABLE |
11911                                    I40E_HW_ATR_EVICT_CAPABLE |
11912                                    I40E_HW_WB_ON_ITR_CAPABLE |
11913                                    I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
11914                                    I40E_HW_NO_PCI_LINK_CHECK |
11915                                    I40E_HW_USE_SET_LLDP_MIB |
11916                                    I40E_HW_GENEVE_OFFLOAD_CAPABLE |
11917                                    I40E_HW_PTP_L4_CAPABLE |
11918                                    I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
11919                                    I40E_HW_OUTER_UDP_CSUM_CAPABLE);
11920
11921#define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
11922                if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
11923                    I40E_FDEVICT_PCTYPE_DEFAULT) {
11924                        dev_warn(&pf->pdev->dev,
11925                                 "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
11926                        pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
11927                }
11928        } else if ((pf->hw.aq.api_maj_ver > 1) ||
11929                   ((pf->hw.aq.api_maj_ver == 1) &&
11930                    (pf->hw.aq.api_min_ver > 4))) {
11931                /* Supported in FW API version higher than 1.4 */
11932                pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
11933        }
11934
11935        /* Enable HW ATR eviction if possible */
11936        if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
11937                pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
11938
11939        if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11940            (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
11941            (pf->hw.aq.fw_maj_ver < 4))) {
11942                pf->hw_features |= I40E_HW_RESTART_AUTONEG;
11943                /* No DCB support  for FW < v4.33 */
11944                pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
11945        }
11946
11947        /* Disable FW LLDP if FW < v4.3 */
11948        if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11949            (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
11950            (pf->hw.aq.fw_maj_ver < 4)))
11951                pf->hw_features |= I40E_HW_STOP_FW_LLDP;
11952
11953        /* Use the FW Set LLDP MIB API if FW > v4.40 */
11954        if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11955            (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
11956            (pf->hw.aq.fw_maj_ver >= 5)))
11957                pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
11958
11959        /* Enable PTP L4 if FW > v6.0 */
11960        if (pf->hw.mac.type == I40E_MAC_XL710 &&
11961            pf->hw.aq.fw_maj_ver >= 6)
11962                pf->hw_features |= I40E_HW_PTP_L4_CAPABLE;
11963
11964        if (pf->hw.func_caps.vmdq && num_online_cpus() != 1) {
11965                pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
11966                pf->flags |= I40E_FLAG_VMDQ_ENABLED;
11967                pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
11968        }
11969
11970        if (pf->hw.func_caps.iwarp && num_online_cpus() != 1) {
11971                pf->flags |= I40E_FLAG_IWARP_ENABLED;
11972                /* IWARP needs one extra vector for CQP just like MISC.*/
11973                pf->num_iwarp_msix = (int)num_online_cpus() + 1;
11974        }
11975        /* Stopping FW LLDP engine is supported on XL710 and X722
11976         * starting from FW versions determined in i40e_init_adminq.
11977         * Stopping the FW LLDP engine is not supported on XL710
11978         * if NPAR is functioning so unset this hw flag in this case.
11979         */
11980        if (pf->hw.mac.type == I40E_MAC_XL710 &&
11981            pf->hw.func_caps.npar_enable &&
11982            (pf->hw.flags & I40E_HW_FLAG_FW_LLDP_STOPPABLE))
11983                pf->hw.flags &= ~I40E_HW_FLAG_FW_LLDP_STOPPABLE;
11984
11985#ifdef CONFIG_PCI_IOV
11986        if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
11987                pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
11988                pf->flags |= I40E_FLAG_SRIOV_ENABLED;
11989                pf->num_req_vfs = min_t(int,
11990                                        pf->hw.func_caps.num_vfs,
11991                                        I40E_MAX_VF_COUNT);
11992        }
11993#endif /* CONFIG_PCI_IOV */
11994        pf->eeprom_version = 0xDEAD;
11995        pf->lan_veb = I40E_NO_VEB;
11996        pf->lan_vsi = I40E_NO_VSI;
11997
11998        /* By default FW has this off for performance reasons */
11999        pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
12000
12001        /* set up queue assignment tracking */
12002        size = sizeof(struct i40e_lump_tracking)
12003                + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
12004        pf->qp_pile = kzalloc(size, GFP_KERNEL);
12005        if (!pf->qp_pile) {
12006                err = -ENOMEM;
12007                goto sw_init_done;
12008        }
12009        pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
12010        pf->qp_pile->search_hint = 0;
12011
12012        pf->tx_timeout_recovery_level = 1;
12013
12014        mutex_init(&pf->switch_mutex);
12015
12016sw_init_done:
12017        return err;
12018}
12019
12020/**
12021 * i40e_set_ntuple - set the ntuple feature flag and take action
12022 * @pf: board private structure to initialize
12023 * @features: the feature set that the stack is suggesting
12024 *
12025 * returns a bool to indicate if reset needs to happen
12026 **/
12027bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
12028{
12029        bool need_reset = false;
12030
12031        /* Check if Flow Director n-tuple support was enabled or disabled.  If
12032         * the state changed, we need to reset.
12033         */
12034        if (features & NETIF_F_NTUPLE) {
12035                /* Enable filters and mark for reset */
12036                if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
12037                        need_reset = true;
12038                /* enable FD_SB only if there is MSI-X vector and no cloud
12039                 * filters exist
12040                 */
12041                if (pf->num_fdsb_msix > 0 && !pf->num_cloud_filters) {
12042                        pf->flags |= I40E_FLAG_FD_SB_ENABLED;
12043                        pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
12044                }
12045        } else {
12046                /* turn off filters, mark for reset and clear SW filter list */
12047                if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
12048                        need_reset = true;
12049                        i40e_fdir_filter_exit(pf);
12050                }
12051                pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
12052                clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state);
12053                pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
12054
12055                /* reset fd counters */
12056                pf->fd_add_err = 0;
12057                pf->fd_atr_cnt = 0;
12058                /* if ATR was auto disabled it can be re-enabled. */
12059                if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state))
12060                        if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
12061                            (I40E_DEBUG_FD & pf->hw.debug_mask))
12062                                dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
12063        }
12064        return need_reset;
12065}
12066
12067/**
12068 * i40e_clear_rss_lut - clear the rx hash lookup table
12069 * @vsi: the VSI being configured
12070 **/
12071static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
12072{
12073        struct i40e_pf *pf = vsi->back;
12074        struct i40e_hw *hw = &pf->hw;
12075        u16 vf_id = vsi->vf_id;
12076        u8 i;
12077
12078        if (vsi->type == I40E_VSI_MAIN) {
12079                for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
12080                        wr32(hw, I40E_PFQF_HLUT(i), 0);
12081        } else if (vsi->type == I40E_VSI_SRIOV) {
12082                for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
12083                        i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
12084        } else {
12085                dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
12086        }
12087}
12088
12089/**
12090 * i40e_set_features - set the netdev feature flags
12091 * @netdev: ptr to the netdev being adjusted
12092 * @features: the feature set that the stack is suggesting
12093 * Note: expects to be called while under rtnl_lock()
12094 **/
12095static int i40e_set_features(struct net_device *netdev,
12096                             netdev_features_t features)
12097{
12098        struct i40e_netdev_priv *np = netdev_priv(netdev);
12099        struct i40e_vsi *vsi = np->vsi;
12100        struct i40e_pf *pf = vsi->back;
12101        bool need_reset;
12102
12103        if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
12104                i40e_pf_config_rss(pf);
12105        else if (!(features & NETIF_F_RXHASH) &&
12106                 netdev->features & NETIF_F_RXHASH)
12107                i40e_clear_rss_lut(vsi);
12108
12109        if (features & NETIF_F_HW_VLAN_CTAG_RX)
12110                i40e_vlan_stripping_enable(vsi);
12111        else
12112                i40e_vlan_stripping_disable(vsi);
12113
12114        if (!(features & NETIF_F_HW_TC) && pf->num_cloud_filters) {
12115                dev_err(&pf->pdev->dev,
12116                        "Offloaded tc filters active, can't turn hw_tc_offload off");
12117                return -EINVAL;
12118        }
12119
12120        if (!(features & NETIF_F_HW_L2FW_DOFFLOAD) && vsi->macvlan_cnt)
12121                i40e_del_all_macvlans(vsi);
12122
12123        need_reset = i40e_set_ntuple(pf, features);
12124
12125        if (need_reset)
12126                i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12127
12128        return 0;
12129}
12130
12131/**
12132 * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
12133 * @pf: board private structure
12134 * @port: The UDP port to look up
12135 *
12136 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
12137 **/
12138static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
12139{
12140        u8 i;
12141
12142        for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
12143                /* Do not report ports with pending deletions as
12144                 * being available.
12145                 */
12146                if (!port && (pf->pending_udp_bitmap & BIT_ULL(i)))
12147                        continue;
12148                if (pf->udp_ports[i].port == port)
12149                        return i;
12150        }
12151
12152        return i;
12153}
12154
12155/**
12156 * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
12157 * @netdev: This physical port's netdev
12158 * @ti: Tunnel endpoint information
12159 **/
12160static void i40e_udp_tunnel_add(struct net_device *netdev,
12161                                struct udp_tunnel_info *ti)
12162{
12163        struct i40e_netdev_priv *np = netdev_priv(netdev);
12164        struct i40e_vsi *vsi = np->vsi;
12165        struct i40e_pf *pf = vsi->back;
12166        u16 port = ntohs(ti->port);
12167        u8 next_idx;
12168        u8 idx;
12169
12170        idx = i40e_get_udp_port_idx(pf, port);
12171
12172        /* Check if port already exists */
12173        if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12174                netdev_info(netdev, "port %d already offloaded\n", port);
12175                return;
12176        }
12177
12178        /* Now check if there is space to add the new port */
12179        next_idx = i40e_get_udp_port_idx(pf, 0);
12180
12181        if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12182                netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
12183                            port);
12184                return;
12185        }
12186
12187        switch (ti->type) {
12188        case UDP_TUNNEL_TYPE_VXLAN:
12189                pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
12190                break;
12191        case UDP_TUNNEL_TYPE_GENEVE:
12192                if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
12193                        return;
12194                pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
12195                break;
12196        default:
12197                return;
12198        }
12199
12200        /* New port: add it and mark its index in the bitmap */
12201        pf->udp_ports[next_idx].port = port;
12202        pf->udp_ports[next_idx].filter_index = I40E_UDP_PORT_INDEX_UNUSED;
12203        pf->pending_udp_bitmap |= BIT_ULL(next_idx);
12204        set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12205}
12206
12207/**
12208 * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
12209 * @netdev: This physical port's netdev
12210 * @ti: Tunnel endpoint information
12211 **/
12212static void i40e_udp_tunnel_del(struct net_device *netdev,
12213                                struct udp_tunnel_info *ti)
12214{
12215        struct i40e_netdev_priv *np = netdev_priv(netdev);
12216        struct i40e_vsi *vsi = np->vsi;
12217        struct i40e_pf *pf = vsi->back;
12218        u16 port = ntohs(ti->port);
12219        u8 idx;
12220
12221        idx = i40e_get_udp_port_idx(pf, port);
12222
12223        /* Check if port already exists */
12224        if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
12225                goto not_found;
12226
12227        switch (ti->type) {
12228        case UDP_TUNNEL_TYPE_VXLAN:
12229                if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
12230                        goto not_found;
12231                break;
12232        case UDP_TUNNEL_TYPE_GENEVE:
12233                if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
12234                        goto not_found;
12235                break;
12236        default:
12237                goto not_found;
12238        }
12239
12240        /* if port exists, set it to 0 (mark for deletion)
12241         * and make it pending
12242         */
12243        pf->udp_ports[idx].port = 0;
12244
12245        /* Toggle pending bit instead of setting it. This way if we are
12246         * deleting a port that has yet to be added we just clear the pending
12247         * bit and don't have to worry about it.
12248         */
12249        pf->pending_udp_bitmap ^= BIT_ULL(idx);
12250        set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12251
12252        return;
12253not_found:
12254        netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
12255                    port);
12256}
12257
12258static int i40e_get_phys_port_id(struct net_device *netdev,
12259                                 struct netdev_phys_item_id *ppid)
12260{
12261        struct i40e_netdev_priv *np = netdev_priv(netdev);
12262        struct i40e_pf *pf = np->vsi->back;
12263        struct i40e_hw *hw = &pf->hw;
12264
12265        if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
12266                return -EOPNOTSUPP;
12267
12268        ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
12269        memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
12270
12271        return 0;
12272}
12273
12274/**
12275 * i40e_ndo_fdb_add - add an entry to the hardware database
12276 * @ndm: the input from the stack
12277 * @tb: pointer to array of nladdr (unused)
12278 * @dev: the net device pointer
12279 * @addr: the MAC address entry being added
12280 * @vid: VLAN ID
12281 * @flags: instructions from stack about fdb operation
12282 */
12283static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
12284                            struct net_device *dev,
12285                            const unsigned char *addr, u16 vid,
12286                            u16 flags,
12287                            struct netlink_ext_ack *extack)
12288{
12289        struct i40e_netdev_priv *np = netdev_priv(dev);
12290        struct i40e_pf *pf = np->vsi->back;
12291        int err = 0;
12292
12293        if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
12294                return -EOPNOTSUPP;
12295
12296        if (vid) {
12297                pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
12298                return -EINVAL;
12299        }
12300
12301        /* Hardware does not support aging addresses so if a
12302         * ndm_state is given only allow permanent addresses
12303         */
12304        if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
12305                netdev_info(dev, "FDB only supports static addresses\n");
12306                return -EINVAL;
12307        }
12308
12309        if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
12310                err = dev_uc_add_excl(dev, addr);
12311        else if (is_multicast_ether_addr(addr))
12312                err = dev_mc_add_excl(dev, addr);
12313        else
12314                err = -EINVAL;
12315
12316        /* Only return duplicate errors if NLM_F_EXCL is set */
12317        if (err == -EEXIST && !(flags & NLM_F_EXCL))
12318                err = 0;
12319
12320        return err;
12321}
12322
12323/**
12324 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
12325 * @dev: the netdev being configured
12326 * @nlh: RTNL message
12327 * @flags: bridge flags
12328 * @extack: netlink extended ack
12329 *
12330 * Inserts a new hardware bridge if not already created and
12331 * enables the bridging mode requested (VEB or VEPA). If the
12332 * hardware bridge has already been inserted and the request
12333 * is to change the mode then that requires a PF reset to
12334 * allow rebuild of the components with required hardware
12335 * bridge mode enabled.
12336 *
12337 * Note: expects to be called while under rtnl_lock()
12338 **/
12339static int i40e_ndo_bridge_setlink(struct net_device *dev,
12340                                   struct nlmsghdr *nlh,
12341                                   u16 flags,
12342                                   struct netlink_ext_ack *extack)
12343{
12344        struct i40e_netdev_priv *np = netdev_priv(dev);
12345        struct i40e_vsi *vsi = np->vsi;
12346        struct i40e_pf *pf = vsi->back;
12347        struct i40e_veb *veb = NULL;
12348        struct nlattr *attr, *br_spec;
12349        int i, rem;
12350
12351        /* Only for PF VSI for now */
12352        if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12353                return -EOPNOTSUPP;
12354
12355        /* Find the HW bridge for PF VSI */
12356        for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12357                if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12358                        veb = pf->veb[i];
12359        }
12360
12361        br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
12362
12363        nla_for_each_nested(attr, br_spec, rem) {
12364                __u16 mode;
12365
12366                if (nla_type(attr) != IFLA_BRIDGE_MODE)
12367                        continue;
12368
12369                mode = nla_get_u16(attr);
12370                if ((mode != BRIDGE_MODE_VEPA) &&
12371                    (mode != BRIDGE_MODE_VEB))
12372                        return -EINVAL;
12373
12374                /* Insert a new HW bridge */
12375                if (!veb) {
12376                        veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
12377                                             vsi->tc_config.enabled_tc);
12378                        if (veb) {
12379                                veb->bridge_mode = mode;
12380                                i40e_config_bridge_mode(veb);
12381                        } else {
12382                                /* No Bridge HW offload available */
12383                                return -ENOENT;
12384                        }
12385                        break;
12386                } else if (mode != veb->bridge_mode) {
12387                        /* Existing HW bridge but different mode needs reset */
12388                        veb->bridge_mode = mode;
12389                        /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
12390                        if (mode == BRIDGE_MODE_VEB)
12391                                pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
12392                        else
12393                                pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
12394                        i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12395                        break;
12396                }
12397        }
12398
12399        return 0;
12400}
12401
12402/**
12403 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
12404 * @skb: skb buff
12405 * @pid: process id
12406 * @seq: RTNL message seq #
12407 * @dev: the netdev being configured
12408 * @filter_mask: unused
12409 * @nlflags: netlink flags passed in
12410 *
12411 * Return the mode in which the hardware bridge is operating in
12412 * i.e VEB or VEPA.
12413 **/
12414static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
12415                                   struct net_device *dev,
12416                                   u32 __always_unused filter_mask,
12417                                   int nlflags)
12418{
12419        struct i40e_netdev_priv *np = netdev_priv(dev);
12420        struct i40e_vsi *vsi = np->vsi;
12421        struct i40e_pf *pf = vsi->back;
12422        struct i40e_veb *veb = NULL;
12423        int i;
12424
12425        /* Only for PF VSI for now */
12426        if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12427                return -EOPNOTSUPP;
12428
12429        /* Find the HW bridge for the PF VSI */
12430        for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12431                if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12432                        veb = pf->veb[i];
12433        }
12434
12435        if (!veb)
12436                return 0;
12437
12438        return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
12439                                       0, 0, nlflags, filter_mask, NULL);
12440}
12441
12442/**
12443 * i40e_features_check - Validate encapsulated packet conforms to limits
12444 * @skb: skb buff
12445 * @dev: This physical port's netdev
12446 * @features: Offload features that the stack believes apply
12447 **/
12448static netdev_features_t i40e_features_check(struct sk_buff *skb,
12449                                             struct net_device *dev,
12450                                             netdev_features_t features)
12451{
12452        size_t len;
12453
12454        /* No point in doing any of this if neither checksum nor GSO are
12455         * being requested for this frame.  We can rule out both by just
12456         * checking for CHECKSUM_PARTIAL
12457         */
12458        if (skb->ip_summed != CHECKSUM_PARTIAL)
12459                return features;
12460
12461        /* We cannot support GSO if the MSS is going to be less than
12462         * 64 bytes.  If it is then we need to drop support for GSO.
12463         */
12464        if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
12465                features &= ~NETIF_F_GSO_MASK;
12466
12467        /* MACLEN can support at most 63 words */
12468        len = skb_network_header(skb) - skb->data;
12469        if (len & ~(63 * 2))
12470                goto out_err;
12471
12472        /* IPLEN and EIPLEN can support at most 127 dwords */
12473        len = skb_transport_header(skb) - skb_network_header(skb);
12474        if (len & ~(127 * 4))
12475                goto out_err;
12476
12477        if (skb->encapsulation) {
12478                /* L4TUNLEN can support 127 words */
12479                len = skb_inner_network_header(skb) - skb_transport_header(skb);
12480                if (len & ~(127 * 2))
12481                        goto out_err;
12482
12483                /* IPLEN can support at most 127 dwords */
12484                len = skb_inner_transport_header(skb) -
12485                      skb_inner_network_header(skb);
12486                if (len & ~(127 * 4))
12487                        goto out_err;
12488        }
12489
12490        /* No need to validate L4LEN as TCP is the only protocol with a
12491         * a flexible value and we support all possible values supported
12492         * by TCP, which is at most 15 dwords
12493         */
12494
12495        return features;
12496out_err:
12497        return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
12498}
12499
12500/**
12501 * i40e_xdp_setup - add/remove an XDP program
12502 * @vsi: VSI to changed
12503 * @prog: XDP program
12504 **/
12505static int i40e_xdp_setup(struct i40e_vsi *vsi,
12506                          struct bpf_prog *prog)
12507{
12508        int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
12509        struct i40e_pf *pf = vsi->back;
12510        struct bpf_prog *old_prog;
12511        bool need_reset;
12512        int i;
12513
12514        /* Don't allow frames that span over multiple buffers */
12515        if (frame_size > vsi->rx_buf_len)
12516                return -EINVAL;
12517
12518        if (!i40e_enabled_xdp_vsi(vsi) && !prog)
12519                return 0;
12520
12521        /* When turning XDP on->off/off->on we reset and rebuild the rings. */
12522        need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
12523
12524        if (need_reset)
12525                i40e_prep_for_reset(pf, true);
12526
12527        old_prog = xchg(&vsi->xdp_prog, prog);
12528
12529        if (need_reset)
12530                i40e_reset_and_rebuild(pf, true, true);
12531
12532        for (i = 0; i < vsi->num_queue_pairs; i++)
12533                WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
12534
12535        if (old_prog)
12536                bpf_prog_put(old_prog);
12537
12538        /* Kick start the NAPI context if there is an AF_XDP socket open
12539         * on that queue id. This so that receiving will start.
12540         */
12541        if (need_reset && prog)
12542                for (i = 0; i < vsi->num_queue_pairs; i++)
12543                        if (vsi->xdp_rings[i]->xsk_umem)
12544                                (void)i40e_xsk_wakeup(vsi->netdev, i,
12545                                                      XDP_WAKEUP_RX);
12546
12547        return 0;
12548}
12549
12550/**
12551 * i40e_enter_busy_conf - Enters busy config state
12552 * @vsi: vsi
12553 *
12554 * Returns 0 on success, <0 for failure.
12555 **/
12556static int i40e_enter_busy_conf(struct i40e_vsi *vsi)
12557{
12558        struct i40e_pf *pf = vsi->back;
12559        int timeout = 50;
12560
12561        while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state)) {
12562                timeout--;
12563                if (!timeout)
12564                        return -EBUSY;
12565                usleep_range(1000, 2000);
12566        }
12567
12568        return 0;
12569}
12570
12571/**
12572 * i40e_exit_busy_conf - Exits busy config state
12573 * @vsi: vsi
12574 **/
12575static void i40e_exit_busy_conf(struct i40e_vsi *vsi)
12576{
12577        struct i40e_pf *pf = vsi->back;
12578
12579        clear_bit(__I40E_CONFIG_BUSY, pf->state);
12580}
12581
12582/**
12583 * i40e_queue_pair_reset_stats - Resets all statistics for a queue pair
12584 * @vsi: vsi
12585 * @queue_pair: queue pair
12586 **/
12587static void i40e_queue_pair_reset_stats(struct i40e_vsi *vsi, int queue_pair)
12588{
12589        memset(&vsi->rx_rings[queue_pair]->rx_stats, 0,
12590               sizeof(vsi->rx_rings[queue_pair]->rx_stats));
12591        memset(&vsi->tx_rings[queue_pair]->stats, 0,
12592               sizeof(vsi->tx_rings[queue_pair]->stats));
12593        if (i40e_enabled_xdp_vsi(vsi)) {
12594                memset(&vsi->xdp_rings[queue_pair]->stats, 0,
12595                       sizeof(vsi->xdp_rings[queue_pair]->stats));
12596        }
12597}
12598
12599/**
12600 * i40e_queue_pair_clean_rings - Cleans all the rings of a queue pair
12601 * @vsi: vsi
12602 * @queue_pair: queue pair
12603 **/
12604static void i40e_queue_pair_clean_rings(struct i40e_vsi *vsi, int queue_pair)
12605{
12606        i40e_clean_tx_ring(vsi->tx_rings[queue_pair]);
12607        if (i40e_enabled_xdp_vsi(vsi)) {
12608                /* Make sure that in-progress ndo_xdp_xmit calls are
12609                 * completed.
12610                 */
12611                synchronize_rcu();
12612                i40e_clean_tx_ring(vsi->xdp_rings[queue_pair]);
12613        }
12614        i40e_clean_rx_ring(vsi->rx_rings[queue_pair]);
12615}
12616
12617/**
12618 * i40e_queue_pair_toggle_napi - Enables/disables NAPI for a queue pair
12619 * @vsi: vsi
12620 * @queue_pair: queue pair
12621 * @enable: true for enable, false for disable
12622 **/
12623static void i40e_queue_pair_toggle_napi(struct i40e_vsi *vsi, int queue_pair,
12624                                        bool enable)
12625{
12626        struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12627        struct i40e_q_vector *q_vector = rxr->q_vector;
12628
12629        if (!vsi->netdev)
12630                return;
12631
12632        /* All rings in a qp belong to the same qvector. */
12633        if (q_vector->rx.ring || q_vector->tx.ring) {
12634                if (enable)
12635                        napi_enable(&q_vector->napi);
12636                else
12637                        napi_disable(&q_vector->napi);
12638        }
12639}
12640
12641/**
12642 * i40e_queue_pair_toggle_rings - Enables/disables all rings for a queue pair
12643 * @vsi: vsi
12644 * @queue_pair: queue pair
12645 * @enable: true for enable, false for disable
12646 *
12647 * Returns 0 on success, <0 on failure.
12648 **/
12649static int i40e_queue_pair_toggle_rings(struct i40e_vsi *vsi, int queue_pair,
12650                                        bool enable)
12651{
12652        struct i40e_pf *pf = vsi->back;
12653        int pf_q, ret = 0;
12654
12655        pf_q = vsi->base_queue + queue_pair;
12656        ret = i40e_control_wait_tx_q(vsi->seid, pf, pf_q,
12657                                     false /*is xdp*/, enable);
12658        if (ret) {
12659                dev_info(&pf->pdev->dev,
12660                         "VSI seid %d Tx ring %d %sable timeout\n",
12661                         vsi->seid, pf_q, (enable ? "en" : "dis"));
12662                return ret;
12663        }
12664
12665        i40e_control_rx_q(pf, pf_q, enable);
12666        ret = i40e_pf_rxq_wait(pf, pf_q, enable);
12667        if (ret) {
12668                dev_info(&pf->pdev->dev,
12669                         "VSI seid %d Rx ring %d %sable timeout\n",
12670                         vsi->seid, pf_q, (enable ? "en" : "dis"));
12671                return ret;
12672        }
12673
12674        /* Due to HW errata, on Rx disable only, the register can
12675         * indicate done before it really is. Needs 50ms to be sure
12676         */
12677        if (!enable)
12678                mdelay(50);
12679
12680        if (!i40e_enabled_xdp_vsi(vsi))
12681                return ret;
12682
12683        ret = i40e_control_wait_tx_q(vsi->seid, pf,
12684                                     pf_q + vsi->alloc_queue_pairs,
12685                                     true /*is xdp*/, enable);
12686        if (ret) {
12687                dev_info(&pf->pdev->dev,
12688                         "VSI seid %d XDP Tx ring %d %sable timeout\n",
12689                         vsi->seid, pf_q, (enable ? "en" : "dis"));
12690        }
12691
12692        return ret;
12693}
12694
12695/**
12696 * i40e_queue_pair_enable_irq - Enables interrupts for a queue pair
12697 * @vsi: vsi
12698 * @queue_pair: queue_pair
12699 **/
12700static void i40e_queue_pair_enable_irq(struct i40e_vsi *vsi, int queue_pair)
12701{
12702        struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12703        struct i40e_pf *pf = vsi->back;
12704        struct i40e_hw *hw = &pf->hw;
12705
12706        /* All rings in a qp belong to the same qvector. */
12707        if (pf->flags & I40E_FLAG_MSIX_ENABLED)
12708                i40e_irq_dynamic_enable(vsi, rxr->q_vector->v_idx);
12709        else
12710                i40e_irq_dynamic_enable_icr0(pf);
12711
12712        i40e_flush(hw);
12713}
12714
12715/**
12716 * i40e_queue_pair_disable_irq - Disables interrupts for a queue pair
12717 * @vsi: vsi
12718 * @queue_pair: queue_pair
12719 **/
12720static void i40e_queue_pair_disable_irq(struct i40e_vsi *vsi, int queue_pair)
12721{
12722        struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12723        struct i40e_pf *pf = vsi->back;
12724        struct i40e_hw *hw = &pf->hw;
12725
12726        /* For simplicity, instead of removing the qp interrupt causes
12727         * from the interrupt linked list, we simply disable the interrupt, and
12728         * leave the list intact.
12729         *
12730         * All rings in a qp belong to the same qvector.
12731         */
12732        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
12733                u32 intpf = vsi->base_vector + rxr->q_vector->v_idx;
12734
12735                wr32(hw, I40E_PFINT_DYN_CTLN(intpf - 1), 0);
12736                i40e_flush(hw);
12737                synchronize_irq(pf->msix_entries[intpf].vector);
12738        } else {
12739                /* Legacy and MSI mode - this stops all interrupt handling */
12740                wr32(hw, I40E_PFINT_ICR0_ENA, 0);
12741                wr32(hw, I40E_PFINT_DYN_CTL0, 0);
12742                i40e_flush(hw);
12743                synchronize_irq(pf->pdev->irq);
12744        }
12745}
12746
12747/**
12748 * i40e_queue_pair_disable - Disables a queue pair
12749 * @vsi: vsi
12750 * @queue_pair: queue pair
12751 *
12752 * Returns 0 on success, <0 on failure.
12753 **/
12754int i40e_queue_pair_disable(struct i40e_vsi *vsi, int queue_pair)
12755{
12756        int err;
12757
12758        err = i40e_enter_busy_conf(vsi);
12759        if (err)
12760                return err;
12761
12762        i40e_queue_pair_disable_irq(vsi, queue_pair);
12763        err = i40e_queue_pair_toggle_rings(vsi, queue_pair, false /* off */);
12764        i40e_queue_pair_toggle_napi(vsi, queue_pair, false /* off */);
12765        i40e_queue_pair_clean_rings(vsi, queue_pair);
12766        i40e_queue_pair_reset_stats(vsi, queue_pair);
12767
12768        return err;
12769}
12770
12771/**
12772 * i40e_queue_pair_enable - Enables a queue pair
12773 * @vsi: vsi
12774 * @queue_pair: queue pair
12775 *
12776 * Returns 0 on success, <0 on failure.
12777 **/
12778int i40e_queue_pair_enable(struct i40e_vsi *vsi, int queue_pair)
12779{
12780        int err;
12781
12782        err = i40e_configure_tx_ring(vsi->tx_rings[queue_pair]);
12783        if (err)
12784                return err;
12785
12786        if (i40e_enabled_xdp_vsi(vsi)) {
12787                err = i40e_configure_tx_ring(vsi->xdp_rings[queue_pair]);
12788                if (err)
12789                        return err;
12790        }
12791
12792        err = i40e_configure_rx_ring(vsi->rx_rings[queue_pair]);
12793        if (err)
12794                return err;
12795
12796        err = i40e_queue_pair_toggle_rings(vsi, queue_pair, true /* on */);
12797        i40e_queue_pair_toggle_napi(vsi, queue_pair, true /* on */);
12798        i40e_queue_pair_enable_irq(vsi, queue_pair);
12799
12800        i40e_exit_busy_conf(vsi);
12801
12802        return err;
12803}
12804
12805/**
12806 * i40e_xdp - implements ndo_bpf for i40e
12807 * @dev: netdevice
12808 * @xdp: XDP command
12809 **/
12810static int i40e_xdp(struct net_device *dev,
12811                    struct netdev_bpf *xdp)
12812{
12813        struct i40e_netdev_priv *np = netdev_priv(dev);
12814        struct i40e_vsi *vsi = np->vsi;
12815
12816        if (vsi->type != I40E_VSI_MAIN)
12817                return -EINVAL;
12818
12819        switch (xdp->command) {
12820        case XDP_SETUP_PROG:
12821                return i40e_xdp_setup(vsi, xdp->prog);
12822        case XDP_QUERY_PROG:
12823                xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
12824                return 0;
12825        case XDP_SETUP_XSK_UMEM:
12826                return i40e_xsk_umem_setup(vsi, xdp->xsk.umem,
12827                                           xdp->xsk.queue_id);
12828        default:
12829                return -EINVAL;
12830        }
12831}
12832
12833static const struct net_device_ops i40e_netdev_ops = {
12834        .ndo_open               = i40e_open,
12835        .ndo_stop               = i40e_close,
12836        .ndo_start_xmit         = i40e_lan_xmit_frame,
12837        .ndo_get_stats64        = i40e_get_netdev_stats_struct,
12838        .ndo_set_rx_mode        = i40e_set_rx_mode,
12839        .ndo_validate_addr      = eth_validate_addr,
12840        .ndo_set_mac_address    = i40e_set_mac,
12841        .ndo_change_mtu         = i40e_change_mtu,
12842        .ndo_do_ioctl           = i40e_ioctl,
12843        .ndo_tx_timeout         = i40e_tx_timeout,
12844        .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
12845        .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
12846#ifdef CONFIG_NET_POLL_CONTROLLER
12847        .ndo_poll_controller    = i40e_netpoll,
12848#endif
12849        .ndo_setup_tc           = __i40e_setup_tc,
12850        .ndo_set_features       = i40e_set_features,
12851        .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
12852        .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
12853        .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
12854        .ndo_get_vf_config      = i40e_ndo_get_vf_config,
12855        .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
12856        .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
12857        .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
12858        .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
12859        .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
12860        .ndo_get_phys_port_id   = i40e_get_phys_port_id,
12861        .ndo_fdb_add            = i40e_ndo_fdb_add,
12862        .ndo_features_check     = i40e_features_check,
12863        .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
12864        .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
12865        .ndo_bpf                = i40e_xdp,
12866        .ndo_xdp_xmit           = i40e_xdp_xmit,
12867        .ndo_xsk_wakeup         = i40e_xsk_wakeup,
12868        .ndo_dfwd_add_station   = i40e_fwd_add,
12869        .ndo_dfwd_del_station   = i40e_fwd_del,
12870};
12871
12872/**
12873 * i40e_config_netdev - Setup the netdev flags
12874 * @vsi: the VSI being configured
12875 *
12876 * Returns 0 on success, negative value on failure
12877 **/
12878static int i40e_config_netdev(struct i40e_vsi *vsi)
12879{
12880        struct i40e_pf *pf = vsi->back;
12881        struct i40e_hw *hw = &pf->hw;
12882        struct i40e_netdev_priv *np;
12883        struct net_device *netdev;
12884        u8 broadcast[ETH_ALEN];
12885        u8 mac_addr[ETH_ALEN];
12886        int etherdev_size;
12887        netdev_features_t hw_enc_features;
12888        netdev_features_t hw_features;
12889
12890        etherdev_size = sizeof(struct i40e_netdev_priv);
12891        netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
12892        if (!netdev)
12893                return -ENOMEM;
12894
12895        vsi->netdev = netdev;
12896        np = netdev_priv(netdev);
12897        np->vsi = vsi;
12898
12899        hw_enc_features = NETIF_F_SG                    |
12900                          NETIF_F_IP_CSUM               |
12901                          NETIF_F_IPV6_CSUM             |
12902                          NETIF_F_HIGHDMA               |
12903                          NETIF_F_SOFT_FEATURES         |
12904                          NETIF_F_TSO                   |
12905                          NETIF_F_TSO_ECN               |
12906                          NETIF_F_TSO6                  |
12907                          NETIF_F_GSO_GRE               |
12908                          NETIF_F_GSO_GRE_CSUM          |
12909                          NETIF_F_GSO_PARTIAL           |
12910                          NETIF_F_GSO_IPXIP4            |
12911                          NETIF_F_GSO_IPXIP6            |
12912                          NETIF_F_GSO_UDP_TUNNEL        |
12913                          NETIF_F_GSO_UDP_TUNNEL_CSUM   |
12914                          NETIF_F_SCTP_CRC              |
12915                          NETIF_F_RXHASH                |
12916                          NETIF_F_RXCSUM                |
12917                          0;
12918
12919        if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
12920                netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
12921
12922        netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
12923
12924        netdev->hw_enc_features |= hw_enc_features;
12925
12926        /* record features VLANs can make use of */
12927        netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
12928
12929        /* enable macvlan offloads */
12930        netdev->hw_features |= NETIF_F_HW_L2FW_DOFFLOAD;
12931
12932        hw_features = hw_enc_features           |
12933                      NETIF_F_HW_VLAN_CTAG_TX   |
12934                      NETIF_F_HW_VLAN_CTAG_RX;
12935
12936        if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
12937                hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
12938
12939        netdev->hw_features |= hw_features;
12940
12941        netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
12942        netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
12943
12944        if (vsi->type == I40E_VSI_MAIN) {
12945                SET_NETDEV_DEV(netdev, &pf->pdev->dev);
12946                ether_addr_copy(mac_addr, hw->mac.perm_addr);
12947                /* The following steps are necessary for two reasons. First,
12948                 * some older NVM configurations load a default MAC-VLAN
12949                 * filter that will accept any tagged packet, and we want to
12950                 * replace this with a normal filter. Additionally, it is
12951                 * possible our MAC address was provided by the platform using
12952                 * Open Firmware or similar.
12953                 *
12954                 * Thus, we need to remove the default filter and install one
12955                 * specific to the MAC address.
12956                 */
12957                i40e_rm_default_mac_filter(vsi, mac_addr);
12958                spin_lock_bh(&vsi->mac_filter_hash_lock);
12959                i40e_add_mac_filter(vsi, mac_addr);
12960                spin_unlock_bh(&vsi->mac_filter_hash_lock);
12961        } else {
12962                /* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
12963                 * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
12964                 * the end, which is 4 bytes long, so force truncation of the
12965                 * original name by IFNAMSIZ - 4
12966                 */
12967                snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
12968                         IFNAMSIZ - 4,
12969                         pf->vsi[pf->lan_vsi]->netdev->name);
12970                eth_random_addr(mac_addr);
12971
12972                spin_lock_bh(&vsi->mac_filter_hash_lock);
12973                i40e_add_mac_filter(vsi, mac_addr);
12974                spin_unlock_bh(&vsi->mac_filter_hash_lock);
12975        }
12976
12977        /* Add the broadcast filter so that we initially will receive
12978         * broadcast packets. Note that when a new VLAN is first added the
12979         * driver will convert all filters marked I40E_VLAN_ANY into VLAN
12980         * specific filters as part of transitioning into "vlan" operation.
12981         * When more VLANs are added, the driver will copy each existing MAC
12982         * filter and add it for the new VLAN.
12983         *
12984         * Broadcast filters are handled specially by
12985         * i40e_sync_filters_subtask, as the driver must to set the broadcast
12986         * promiscuous bit instead of adding this directly as a MAC/VLAN
12987         * filter. The subtask will update the correct broadcast promiscuous
12988         * bits as VLANs become active or inactive.
12989         */
12990        eth_broadcast_addr(broadcast);
12991        spin_lock_bh(&vsi->mac_filter_hash_lock);
12992        i40e_add_mac_filter(vsi, broadcast);
12993        spin_unlock_bh(&vsi->mac_filter_hash_lock);
12994
12995        ether_addr_copy(netdev->dev_addr, mac_addr);
12996        ether_addr_copy(netdev->perm_addr, mac_addr);
12997
12998        /* i40iw_net_event() reads 16 bytes from neigh->primary_key */
12999        netdev->neigh_priv_len = sizeof(u32) * 4;
13000
13001        netdev->priv_flags |= IFF_UNICAST_FLT;
13002        netdev->priv_flags |= IFF_SUPP_NOFCS;
13003        /* Setup netdev TC information */
13004        i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
13005
13006        netdev->netdev_ops = &i40e_netdev_ops;
13007        netdev->watchdog_timeo = 5 * HZ;
13008        i40e_set_ethtool_ops(netdev);
13009
13010        /* MTU range: 68 - 9706 */
13011        netdev->min_mtu = ETH_MIN_MTU;
13012        netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
13013
13014        return 0;
13015}
13016
13017/**
13018 * i40e_vsi_delete - Delete a VSI from the switch
13019 * @vsi: the VSI being removed
13020 *
13021 * Returns 0 on success, negative value on failure
13022 **/
13023static void i40e_vsi_delete(struct i40e_vsi *vsi)
13024{
13025        /* remove default VSI is not allowed */
13026        if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
13027                return;
13028
13029        i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
13030}
13031
13032/**
13033 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
13034 * @vsi: the VSI being queried
13035 *
13036 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
13037 **/
13038int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
13039{
13040        struct i40e_veb *veb;
13041        struct i40e_pf *pf = vsi->back;
13042
13043        /* Uplink is not a bridge so default to VEB */
13044        if (vsi->veb_idx >= I40E_MAX_VEB)
13045                return 1;
13046
13047        veb = pf->veb[vsi->veb_idx];
13048        if (!veb) {
13049                dev_info(&pf->pdev->dev,
13050                         "There is no veb associated with the bridge\n");
13051                return -ENOENT;
13052        }
13053
13054        /* Uplink is a bridge in VEPA mode */
13055        if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
13056                return 0;
13057        } else {
13058                /* Uplink is a bridge in VEB mode */
13059                return 1;
13060        }
13061
13062        /* VEPA is now default bridge, so return 0 */
13063        return 0;
13064}
13065
13066/**
13067 * i40e_add_vsi - Add a VSI to the switch
13068 * @vsi: the VSI being configured
13069 *
13070 * This initializes a VSI context depending on the VSI type to be added and
13071 * passes it down to the add_vsi aq command.
13072 **/
13073static int i40e_add_vsi(struct i40e_vsi *vsi)
13074{
13075        int ret = -ENODEV;
13076        struct i40e_pf *pf = vsi->back;
13077        struct i40e_hw *hw = &pf->hw;
13078        struct i40e_vsi_context ctxt;
13079        struct i40e_mac_filter *f;
13080        struct hlist_node *h;
13081        int bkt;
13082
13083        u8 enabled_tc = 0x1; /* TC0 enabled */
13084        int f_count = 0;
13085
13086        memset(&ctxt, 0, sizeof(ctxt));
13087        switch (vsi->type) {
13088        case I40E_VSI_MAIN:
13089                /* The PF's main VSI is already setup as part of the
13090                 * device initialization, so we'll not bother with
13091                 * the add_vsi call, but we will retrieve the current
13092                 * VSI context.
13093                 */
13094                ctxt.seid = pf->main_vsi_seid;
13095                ctxt.pf_num = pf->hw.pf_id;
13096                ctxt.vf_num = 0;
13097                ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
13098                ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13099                if (ret) {
13100                        dev_info(&pf->pdev->dev,
13101                                 "couldn't get PF vsi config, err %s aq_err %s\n",
13102                                 i40e_stat_str(&pf->hw, ret),
13103                                 i40e_aq_str(&pf->hw,
13104                                             pf->hw.aq.asq_last_status));
13105                        return -ENOENT;
13106                }
13107                vsi->info = ctxt.info;
13108                vsi->info.valid_sections = 0;
13109
13110                vsi->seid = ctxt.seid;
13111                vsi->id = ctxt.vsi_number;
13112
13113                enabled_tc = i40e_pf_get_tc_map(pf);
13114
13115                /* Source pruning is enabled by default, so the flag is
13116                 * negative logic - if it's set, we need to fiddle with
13117                 * the VSI to disable source pruning.
13118                 */
13119                if (pf->flags & I40E_FLAG_SOURCE_PRUNING_DISABLED) {
13120                        memset(&ctxt, 0, sizeof(ctxt));
13121                        ctxt.seid = pf->main_vsi_seid;
13122                        ctxt.pf_num = pf->hw.pf_id;
13123                        ctxt.vf_num = 0;
13124                        ctxt.info.valid_sections |=
13125                                     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13126                        ctxt.info.switch_id =
13127                                   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
13128                        ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13129                        if (ret) {
13130                                dev_info(&pf->pdev->dev,
13131                                         "update vsi failed, err %s aq_err %s\n",
13132                                         i40e_stat_str(&pf->hw, ret),
13133                                         i40e_aq_str(&pf->hw,
13134                                                     pf->hw.aq.asq_last_status));
13135                                ret = -ENOENT;
13136                                goto err;
13137                        }
13138                }
13139
13140                /* MFP mode setup queue map and update VSI */
13141                if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
13142                    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
13143                        memset(&ctxt, 0, sizeof(ctxt));
13144                        ctxt.seid = pf->main_vsi_seid;
13145                        ctxt.pf_num = pf->hw.pf_id;
13146                        ctxt.vf_num = 0;
13147                        i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
13148                        ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13149                        if (ret) {
13150                                dev_info(&pf->pdev->dev,
13151                                         "update vsi failed, err %s aq_err %s\n",
13152                                         i40e_stat_str(&pf->hw, ret),
13153                                         i40e_aq_str(&pf->hw,
13154                                                    pf->hw.aq.asq_last_status));
13155                                ret = -ENOENT;
13156                                goto err;
13157                        }
13158                        /* update the local VSI info queue map */
13159                        i40e_vsi_update_queue_map(vsi, &ctxt);
13160                        vsi->info.valid_sections = 0;
13161                } else {
13162                        /* Default/Main VSI is only enabled for TC0
13163                         * reconfigure it to enable all TCs that are
13164                         * available on the port in SFP mode.
13165                         * For MFP case the iSCSI PF would use this
13166                         * flow to enable LAN+iSCSI TC.
13167                         */
13168                        ret = i40e_vsi_config_tc(vsi, enabled_tc);
13169                        if (ret) {
13170                                /* Single TC condition is not fatal,
13171                                 * message and continue
13172                                 */
13173                                dev_info(&pf->pdev->dev,
13174                                         "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
13175                                         enabled_tc,
13176                                         i40e_stat_str(&pf->hw, ret),
13177                                         i40e_aq_str(&pf->hw,
13178                                                    pf->hw.aq.asq_last_status));
13179                        }
13180                }
13181                break;
13182
13183        case I40E_VSI_FDIR:
13184                ctxt.pf_num = hw->pf_id;
13185                ctxt.vf_num = 0;
13186                ctxt.uplink_seid = vsi->uplink_seid;
13187                ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13188                ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13189                if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
13190                    (i40e_is_vsi_uplink_mode_veb(vsi))) {
13191                        ctxt.info.valid_sections |=
13192                             cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13193                        ctxt.info.switch_id =
13194                           cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13195                }
13196                i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13197                break;
13198
13199        case I40E_VSI_VMDQ2:
13200                ctxt.pf_num = hw->pf_id;
13201                ctxt.vf_num = 0;
13202                ctxt.uplink_seid = vsi->uplink_seid;
13203                ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13204                ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
13205
13206                /* This VSI is connected to VEB so the switch_id
13207                 * should be set to zero by default.
13208                 */
13209                if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13210                        ctxt.info.valid_sections |=
13211                                cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13212                        ctxt.info.switch_id =
13213                                cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13214                }
13215
13216                /* Setup the VSI tx/rx queue map for TC0 only for now */
13217                i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13218                break;
13219
13220        case I40E_VSI_SRIOV:
13221                ctxt.pf_num = hw->pf_id;
13222                ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
13223                ctxt.uplink_seid = vsi->uplink_seid;
13224                ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13225                ctxt.flags = I40E_AQ_VSI_TYPE_VF;
13226
13227                /* This VSI is connected to VEB so the switch_id
13228                 * should be set to zero by default.
13229                 */
13230                if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13231                        ctxt.info.valid_sections |=
13232                                cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13233                        ctxt.info.switch_id =
13234                                cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13235                }
13236
13237                if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
13238                        ctxt.info.valid_sections |=
13239                                cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
13240                        ctxt.info.queueing_opt_flags |=
13241                                (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
13242                                 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
13243                }
13244
13245                ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
13246                ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
13247                if (pf->vf[vsi->vf_id].spoofchk) {
13248                        ctxt.info.valid_sections |=
13249                                cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
13250                        ctxt.info.sec_flags |=
13251                                (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
13252                                 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
13253                }
13254                /* Setup the VSI tx/rx queue map for TC0 only for now */
13255                i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13256                break;
13257
13258        case I40E_VSI_IWARP:
13259                /* send down message to iWARP */
13260                break;
13261
13262        default:
13263                return -ENODEV;
13264        }
13265
13266        if (vsi->type != I40E_VSI_MAIN) {
13267                ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
13268                if (ret) {
13269                        dev_info(&vsi->back->pdev->dev,
13270                                 "add vsi failed, err %s aq_err %s\n",
13271                                 i40e_stat_str(&pf->hw, ret),
13272                                 i40e_aq_str(&pf->hw,
13273                                             pf->hw.aq.asq_last_status));
13274                        ret = -ENOENT;
13275                        goto err;
13276                }
13277                vsi->info = ctxt.info;
13278                vsi->info.valid_sections = 0;
13279                vsi->seid = ctxt.seid;
13280                vsi->id = ctxt.vsi_number;
13281        }
13282
13283        vsi->active_filters = 0;
13284        clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
13285        spin_lock_bh(&vsi->mac_filter_hash_lock);
13286        /* If macvlan filters already exist, force them to get loaded */
13287        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
13288                f->state = I40E_FILTER_NEW;
13289                f_count++;
13290        }
13291        spin_unlock_bh(&vsi->mac_filter_hash_lock);
13292
13293        if (f_count) {
13294                vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
13295                set_bit(__I40E_MACVLAN_SYNC_PENDING, pf->state);
13296        }
13297
13298        /* Update VSI BW information */
13299        ret = i40e_vsi_get_bw_info(vsi);
13300        if (ret) {
13301                dev_info(&pf->pdev->dev,
13302                         "couldn't get vsi bw info, err %s aq_err %s\n",
13303                         i40e_stat_str(&pf->hw, ret),
13304                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13305                /* VSI is already added so not tearing that up */
13306                ret = 0;
13307        }
13308
13309err:
13310        return ret;
13311}
13312
13313/**
13314 * i40e_vsi_release - Delete a VSI and free its resources
13315 * @vsi: the VSI being removed
13316 *
13317 * Returns 0 on success or < 0 on error
13318 **/
13319int i40e_vsi_release(struct i40e_vsi *vsi)
13320{
13321        struct i40e_mac_filter *f;
13322        struct hlist_node *h;
13323        struct i40e_veb *veb = NULL;
13324        struct i40e_pf *pf;
13325        u16 uplink_seid;
13326        int i, n, bkt;
13327
13328        pf = vsi->back;
13329
13330        /* release of a VEB-owner or last VSI is not allowed */
13331        if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
13332                dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
13333                         vsi->seid, vsi->uplink_seid);
13334                return -ENODEV;
13335        }
13336        if (vsi == pf->vsi[pf->lan_vsi] &&
13337            !test_bit(__I40E_DOWN, pf->state)) {
13338                dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
13339                return -ENODEV;
13340        }
13341
13342        uplink_seid = vsi->uplink_seid;
13343        if (vsi->type != I40E_VSI_SRIOV) {
13344                if (vsi->netdev_registered) {
13345                        vsi->netdev_registered = false;
13346                        if (vsi->netdev) {
13347                                /* results in a call to i40e_close() */
13348                                unregister_netdev(vsi->netdev);
13349                        }
13350                } else {
13351                        i40e_vsi_close(vsi);
13352                }
13353                i40e_vsi_disable_irq(vsi);
13354        }
13355
13356        spin_lock_bh(&vsi->mac_filter_hash_lock);
13357
13358        /* clear the sync flag on all filters */
13359        if (vsi->netdev) {
13360                __dev_uc_unsync(vsi->netdev, NULL);
13361                __dev_mc_unsync(vsi->netdev, NULL);
13362        }
13363
13364        /* make sure any remaining filters are marked for deletion */
13365        hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
13366                __i40e_del_filter(vsi, f);
13367
13368        spin_unlock_bh(&vsi->mac_filter_hash_lock);
13369
13370        i40e_sync_vsi_filters(vsi);
13371
13372        i40e_vsi_delete(vsi);
13373        i40e_vsi_free_q_vectors(vsi);
13374        if (vsi->netdev) {
13375                free_netdev(vsi->netdev);
13376                vsi->netdev = NULL;
13377        }
13378        i40e_vsi_clear_rings(vsi);
13379        i40e_vsi_clear(vsi);
13380
13381        /* If this was the last thing on the VEB, except for the
13382         * controlling VSI, remove the VEB, which puts the controlling
13383         * VSI onto the next level down in the switch.
13384         *
13385         * Well, okay, there's one more exception here: don't remove
13386         * the orphan VEBs yet.  We'll wait for an explicit remove request
13387         * from up the network stack.
13388         */
13389        for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
13390                if (pf->vsi[i] &&
13391                    pf->vsi[i]->uplink_seid == uplink_seid &&
13392                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13393                        n++;      /* count the VSIs */
13394                }
13395        }
13396        for (i = 0; i < I40E_MAX_VEB; i++) {
13397                if (!pf->veb[i])
13398                        continue;
13399                if (pf->veb[i]->uplink_seid == uplink_seid)
13400                        n++;     /* count the VEBs */
13401                if (pf->veb[i]->seid == uplink_seid)
13402                        veb = pf->veb[i];
13403        }
13404        if (n == 0 && veb && veb->uplink_seid != 0)
13405                i40e_veb_release(veb);
13406
13407        return 0;
13408}
13409
13410/**
13411 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
13412 * @vsi: ptr to the VSI
13413 *
13414 * This should only be called after i40e_vsi_mem_alloc() which allocates the
13415 * corresponding SW VSI structure and initializes num_queue_pairs for the
13416 * newly allocated VSI.
13417 *
13418 * Returns 0 on success or negative on failure
13419 **/
13420static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
13421{
13422        int ret = -ENOENT;
13423        struct i40e_pf *pf = vsi->back;
13424
13425        if (vsi->q_vectors[0]) {
13426                dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
13427                         vsi->seid);
13428                return -EEXIST;
13429        }
13430
13431        if (vsi->base_vector) {
13432                dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
13433                         vsi->seid, vsi->base_vector);
13434                return -EEXIST;
13435        }
13436
13437        ret = i40e_vsi_alloc_q_vectors(vsi);
13438        if (ret) {
13439                dev_info(&pf->pdev->dev,
13440                         "failed to allocate %d q_vector for VSI %d, ret=%d\n",
13441                         vsi->num_q_vectors, vsi->seid, ret);
13442                vsi->num_q_vectors = 0;
13443                goto vector_setup_out;
13444        }
13445
13446        /* In Legacy mode, we do not have to get any other vector since we
13447         * piggyback on the misc/ICR0 for queue interrupts.
13448        */
13449        if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
13450                return ret;
13451        if (vsi->num_q_vectors)
13452                vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
13453                                                 vsi->num_q_vectors, vsi->idx);
13454        if (vsi->base_vector < 0) {
13455                dev_info(&pf->pdev->dev,
13456                         "failed to get tracking for %d vectors for VSI %d, err=%d\n",
13457                         vsi->num_q_vectors, vsi->seid, vsi->base_vector);
13458                i40e_vsi_free_q_vectors(vsi);
13459                ret = -ENOENT;
13460                goto vector_setup_out;
13461        }
13462
13463vector_setup_out:
13464        return ret;
13465}
13466
13467/**
13468 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
13469 * @vsi: pointer to the vsi.
13470 *
13471 * This re-allocates a vsi's queue resources.
13472 *
13473 * Returns pointer to the successfully allocated and configured VSI sw struct
13474 * on success, otherwise returns NULL on failure.
13475 **/
13476static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
13477{
13478        u16 alloc_queue_pairs;
13479        struct i40e_pf *pf;
13480        u8 enabled_tc;
13481        int ret;
13482
13483        if (!vsi)
13484                return NULL;
13485
13486        pf = vsi->back;
13487
13488        i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
13489        i40e_vsi_clear_rings(vsi);
13490
13491        i40e_vsi_free_arrays(vsi, false);
13492        i40e_set_num_rings_in_vsi(vsi);
13493        ret = i40e_vsi_alloc_arrays(vsi, false);
13494        if (ret)
13495                goto err_vsi;
13496
13497        alloc_queue_pairs = vsi->alloc_queue_pairs *
13498                            (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13499
13500        ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13501        if (ret < 0) {
13502                dev_info(&pf->pdev->dev,
13503                         "failed to get tracking for %d queues for VSI %d err %d\n",
13504                         alloc_queue_pairs, vsi->seid, ret);
13505                goto err_vsi;
13506        }
13507        vsi->base_queue = ret;
13508
13509        /* Update the FW view of the VSI. Force a reset of TC and queue
13510         * layout configurations.
13511         */
13512        enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
13513        pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
13514        pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
13515        i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
13516        if (vsi->type == I40E_VSI_MAIN)
13517                i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
13518
13519        /* assign it some queues */
13520        ret = i40e_alloc_rings(vsi);
13521        if (ret)
13522                goto err_rings;
13523
13524        /* map all of the rings to the q_vectors */
13525        i40e_vsi_map_rings_to_vectors(vsi);
13526        return vsi;
13527
13528err_rings:
13529        i40e_vsi_free_q_vectors(vsi);
13530        if (vsi->netdev_registered) {
13531                vsi->netdev_registered = false;
13532                unregister_netdev(vsi->netdev);
13533                free_netdev(vsi->netdev);
13534                vsi->netdev = NULL;
13535        }
13536        i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13537err_vsi:
13538        i40e_vsi_clear(vsi);
13539        return NULL;
13540}
13541
13542/**
13543 * i40e_vsi_setup - Set up a VSI by a given type
13544 * @pf: board private structure
13545 * @type: VSI type
13546 * @uplink_seid: the switch element to link to
13547 * @param1: usage depends upon VSI type. For VF types, indicates VF id
13548 *
13549 * This allocates the sw VSI structure and its queue resources, then add a VSI
13550 * to the identified VEB.
13551 *
13552 * Returns pointer to the successfully allocated and configure VSI sw struct on
13553 * success, otherwise returns NULL on failure.
13554 **/
13555struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
13556                                u16 uplink_seid, u32 param1)
13557{
13558        struct i40e_vsi *vsi = NULL;
13559        struct i40e_veb *veb = NULL;
13560        u16 alloc_queue_pairs;
13561        int ret, i;
13562        int v_idx;
13563
13564        /* The requested uplink_seid must be either
13565         *     - the PF's port seid
13566         *              no VEB is needed because this is the PF
13567         *              or this is a Flow Director special case VSI
13568         *     - seid of an existing VEB
13569         *     - seid of a VSI that owns an existing VEB
13570         *     - seid of a VSI that doesn't own a VEB
13571         *              a new VEB is created and the VSI becomes the owner
13572         *     - seid of the PF VSI, which is what creates the first VEB
13573         *              this is a special case of the previous
13574         *
13575         * Find which uplink_seid we were given and create a new VEB if needed
13576         */
13577        for (i = 0; i < I40E_MAX_VEB; i++) {
13578                if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
13579                        veb = pf->veb[i];
13580                        break;
13581                }
13582        }
13583
13584        if (!veb && uplink_seid != pf->mac_seid) {
13585
13586                for (i = 0; i < pf->num_alloc_vsi; i++) {
13587                        if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
13588                                vsi = pf->vsi[i];
13589                                break;
13590                        }
13591                }
13592                if (!vsi) {
13593                        dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
13594                                 uplink_seid);
13595                        return NULL;
13596                }
13597
13598                if (vsi->uplink_seid == pf->mac_seid)
13599                        veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
13600                                             vsi->tc_config.enabled_tc);
13601                else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
13602                        veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
13603                                             vsi->tc_config.enabled_tc);
13604                if (veb) {
13605                        if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
13606                                dev_info(&vsi->back->pdev->dev,
13607                                         "New VSI creation error, uplink seid of LAN VSI expected.\n");
13608                                return NULL;
13609                        }
13610                        /* We come up by default in VEPA mode if SRIOV is not
13611                         * already enabled, in which case we can't force VEPA
13612                         * mode.
13613                         */
13614                        if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
13615                                veb->bridge_mode = BRIDGE_MODE_VEPA;
13616                                pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
13617                        }
13618                        i40e_config_bridge_mode(veb);
13619                }
13620                for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
13621                        if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
13622                                veb = pf->veb[i];
13623                }
13624                if (!veb) {
13625                        dev_info(&pf->pdev->dev, "couldn't add VEB\n");
13626                        return NULL;
13627                }
13628
13629                vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13630                uplink_seid = veb->seid;
13631        }
13632
13633        /* get vsi sw struct */
13634        v_idx = i40e_vsi_mem_alloc(pf, type);
13635        if (v_idx < 0)
13636                goto err_alloc;
13637        vsi = pf->vsi[v_idx];
13638        if (!vsi)
13639                goto err_alloc;
13640        vsi->type = type;
13641        vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
13642
13643        if (type == I40E_VSI_MAIN)
13644                pf->lan_vsi = v_idx;
13645        else if (type == I40E_VSI_SRIOV)
13646                vsi->vf_id = param1;
13647        /* assign it some queues */
13648        alloc_queue_pairs = vsi->alloc_queue_pairs *
13649                            (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13650
13651        ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13652        if (ret < 0) {
13653                dev_info(&pf->pdev->dev,
13654                         "failed to get tracking for %d queues for VSI %d err=%d\n",
13655                         alloc_queue_pairs, vsi->seid, ret);
13656                goto err_vsi;
13657        }
13658        vsi->base_queue = ret;
13659
13660        /* get a VSI from the hardware */
13661        vsi->uplink_seid = uplink_seid;
13662        ret = i40e_add_vsi(vsi);
13663        if (ret)
13664                goto err_vsi;
13665
13666        switch (vsi->type) {
13667        /* setup the netdev if needed */
13668        case I40E_VSI_MAIN:
13669        case I40E_VSI_VMDQ2:
13670                ret = i40e_config_netdev(vsi);
13671                if (ret)
13672                        goto err_netdev;
13673                ret = register_netdev(vsi->netdev);
13674                if (ret)
13675                        goto err_netdev;
13676                vsi->netdev_registered = true;
13677                netif_carrier_off(vsi->netdev);
13678#ifdef CONFIG_I40E_DCB
13679                /* Setup DCB netlink interface */
13680                i40e_dcbnl_setup(vsi);
13681#endif /* CONFIG_I40E_DCB */
13682                /* fall through */
13683
13684        case I40E_VSI_FDIR:
13685                /* set up vectors and rings if needed */
13686                ret = i40e_vsi_setup_vectors(vsi);
13687                if (ret)
13688                        goto err_msix;
13689
13690                ret = i40e_alloc_rings(vsi);
13691                if (ret)
13692                        goto err_rings;
13693
13694                /* map all of the rings to the q_vectors */
13695                i40e_vsi_map_rings_to_vectors(vsi);
13696
13697                i40e_vsi_reset_stats(vsi);
13698                break;
13699
13700        default:
13701                /* no netdev or rings for the other VSI types */
13702                break;
13703        }
13704
13705        if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
13706            (vsi->type == I40E_VSI_VMDQ2)) {
13707                ret = i40e_vsi_config_rss(vsi);
13708        }
13709        return vsi;
13710
13711err_rings:
13712        i40e_vsi_free_q_vectors(vsi);
13713err_msix:
13714        if (vsi->netdev_registered) {
13715                vsi->netdev_registered = false;
13716                unregister_netdev(vsi->netdev);
13717                free_netdev(vsi->netdev);
13718                vsi->netdev = NULL;
13719        }
13720err_netdev:
13721        i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13722err_vsi:
13723        i40e_vsi_clear(vsi);
13724err_alloc:
13725        return NULL;
13726}
13727
13728/**
13729 * i40e_veb_get_bw_info - Query VEB BW information
13730 * @veb: the veb to query
13731 *
13732 * Query the Tx scheduler BW configuration data for given VEB
13733 **/
13734static int i40e_veb_get_bw_info(struct i40e_veb *veb)
13735{
13736        struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
13737        struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
13738        struct i40e_pf *pf = veb->pf;
13739        struct i40e_hw *hw = &pf->hw;
13740        u32 tc_bw_max;
13741        int ret = 0;
13742        int i;
13743
13744        ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
13745                                                  &bw_data, NULL);
13746        if (ret) {
13747                dev_info(&pf->pdev->dev,
13748                         "query veb bw config failed, err %s aq_err %s\n",
13749                         i40e_stat_str(&pf->hw, ret),
13750                         i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13751                goto out;
13752        }
13753
13754        ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
13755                                                   &ets_data, NULL);
13756        if (ret) {
13757                dev_info(&pf->pdev->dev,
13758                         "query veb bw ets config failed, err %s aq_err %s\n",
13759                         i40e_stat_str(&pf->hw, ret),
13760                         i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13761                goto out;
13762        }
13763
13764        veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
13765        veb->bw_max_quanta = ets_data.tc_bw_max;
13766        veb->is_abs_credits = bw_data.absolute_credits_enable;
13767        veb->enabled_tc = ets_data.tc_valid_bits;
13768        tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
13769                    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
13770        for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
13771                veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
13772                veb->bw_tc_limit_credits[i] =
13773                                        le16_to_cpu(bw_data.tc_bw_limits[i]);
13774                veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
13775        }
13776
13777out:
13778        return ret;
13779}
13780
13781/**
13782 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
13783 * @pf: board private structure
13784 *
13785 * On error: returns error code (negative)
13786 * On success: returns vsi index in PF (positive)
13787 **/
13788static int i40e_veb_mem_alloc(struct i40e_pf *pf)
13789{
13790        int ret = -ENOENT;
13791        struct i40e_veb *veb;
13792        int i;
13793
13794        /* Need to protect the allocation of switch elements at the PF level */
13795        mutex_lock(&pf->switch_mutex);
13796
13797        /* VEB list may be fragmented if VEB creation/destruction has
13798         * been happening.  We can afford to do a quick scan to look
13799         * for any free slots in the list.
13800         *
13801         * find next empty veb slot, looping back around if necessary
13802         */
13803        i = 0;
13804        while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
13805                i++;
13806        if (i >= I40E_MAX_VEB) {
13807                ret = -ENOMEM;
13808                goto err_alloc_veb;  /* out of VEB slots! */
13809        }
13810
13811        veb = kzalloc(sizeof(*veb), GFP_KERNEL);
13812        if (!veb) {
13813                ret = -ENOMEM;
13814                goto err_alloc_veb;
13815        }
13816        veb->pf = pf;
13817        veb->idx = i;
13818        veb->enabled_tc = 1;
13819
13820        pf->veb[i] = veb;
13821        ret = i;
13822err_alloc_veb:
13823        mutex_unlock(&pf->switch_mutex);
13824        return ret;
13825}
13826
13827/**
13828 * i40e_switch_branch_release - Delete a branch of the switch tree
13829 * @branch: where to start deleting
13830 *
13831 * This uses recursion to find the tips of the branch to be
13832 * removed, deleting until we get back to and can delete this VEB.
13833 **/
13834static void i40e_switch_branch_release(struct i40e_veb *branch)
13835{
13836        struct i40e_pf *pf = branch->pf;
13837        u16 branch_seid = branch->seid;
13838        u16 veb_idx = branch->idx;
13839        int i;
13840
13841        /* release any VEBs on this VEB - RECURSION */
13842        for (i = 0; i < I40E_MAX_VEB; i++) {
13843                if (!pf->veb[i])
13844                        continue;
13845                if (pf->veb[i]->uplink_seid == branch->seid)
13846                        i40e_switch_branch_release(pf->veb[i]);
13847        }
13848
13849        /* Release the VSIs on this VEB, but not the owner VSI.
13850         *
13851         * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
13852         *       the VEB itself, so don't use (*branch) after this loop.
13853         */
13854        for (i = 0; i < pf->num_alloc_vsi; i++) {
13855                if (!pf->vsi[i])
13856                        continue;
13857                if (pf->vsi[i]->uplink_seid == branch_seid &&
13858                   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13859                        i40e_vsi_release(pf->vsi[i]);
13860                }
13861        }
13862
13863        /* There's one corner case where the VEB might not have been
13864         * removed, so double check it here and remove it if needed.
13865         * This case happens if the veb was created from the debugfs
13866         * commands and no VSIs were added to it.
13867         */
13868        if (pf->veb[veb_idx])
13869                i40e_veb_release(pf->veb[veb_idx]);
13870}
13871
13872/**
13873 * i40e_veb_clear - remove veb struct
13874 * @veb: the veb to remove
13875 **/
13876static void i40e_veb_clear(struct i40e_veb *veb)
13877{
13878        if (!veb)
13879                return;
13880
13881        if (veb->pf) {
13882                struct i40e_pf *pf = veb->pf;
13883
13884                mutex_lock(&pf->switch_mutex);
13885                if (pf->veb[veb->idx] == veb)
13886                        pf->veb[veb->idx] = NULL;
13887                mutex_unlock(&pf->switch_mutex);
13888        }
13889
13890        kfree(veb);
13891}
13892
13893/**
13894 * i40e_veb_release - Delete a VEB and free its resources
13895 * @veb: the VEB being removed
13896 **/
13897void i40e_veb_release(struct i40e_veb *veb)
13898{
13899        struct i40e_vsi *vsi = NULL;
13900        struct i40e_pf *pf;
13901        int i, n = 0;
13902
13903        pf = veb->pf;
13904
13905        /* find the remaining VSI and check for extras */
13906        for (i = 0; i < pf->num_alloc_vsi; i++) {
13907                if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
13908                        n++;
13909                        vsi = pf->vsi[i];
13910                }
13911        }
13912        if (n != 1) {
13913                dev_info(&pf->pdev->dev,
13914                         "can't remove VEB %d with %d VSIs left\n",
13915                         veb->seid, n);
13916                return;
13917        }
13918
13919        /* move the remaining VSI to uplink veb */
13920        vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
13921        if (veb->uplink_seid) {
13922                vsi->uplink_seid = veb->uplink_seid;
13923                if (veb->uplink_seid == pf->mac_seid)
13924                        vsi->veb_idx = I40E_NO_VEB;
13925                else
13926                        vsi->veb_idx = veb->veb_idx;
13927        } else {
13928                /* floating VEB */
13929                vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
13930                vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
13931        }
13932
13933        i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13934        i40e_veb_clear(veb);
13935}
13936
13937/**
13938 * i40e_add_veb - create the VEB in the switch
13939 * @veb: the VEB to be instantiated
13940 * @vsi: the controlling VSI
13941 **/
13942static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
13943{
13944        struct i40e_pf *pf = veb->pf;
13945        bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
13946        int ret;
13947
13948        ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
13949                              veb->enabled_tc, false,
13950                              &veb->seid, enable_stats, NULL);
13951
13952        /* get a VEB from the hardware */
13953        if (ret) {
13954                dev_info(&pf->pdev->dev,
13955                         "couldn't add VEB, err %s aq_err %s\n",
13956                         i40e_stat_str(&pf->hw, ret),
13957                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13958                return -EPERM;
13959        }
13960
13961        /* get statistics counter */
13962        ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
13963                                         &veb->stats_idx, NULL, NULL, NULL);
13964        if (ret) {
13965                dev_info(&pf->pdev->dev,
13966                         "couldn't get VEB statistics idx, err %s aq_err %s\n",
13967                         i40e_stat_str(&pf->hw, ret),
13968                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13969                return -EPERM;
13970        }
13971        ret = i40e_veb_get_bw_info(veb);
13972        if (ret) {
13973                dev_info(&pf->pdev->dev,
13974                         "couldn't get VEB bw info, err %s aq_err %s\n",
13975                         i40e_stat_str(&pf->hw, ret),
13976                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13977                i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13978                return -ENOENT;
13979        }
13980
13981        vsi->uplink_seid = veb->seid;
13982        vsi->veb_idx = veb->idx;
13983        vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13984
13985        return 0;
13986}
13987
13988/**
13989 * i40e_veb_setup - Set up a VEB
13990 * @pf: board private structure
13991 * @flags: VEB setup flags
13992 * @uplink_seid: the switch element to link to
13993 * @vsi_seid: the initial VSI seid
13994 * @enabled_tc: Enabled TC bit-map
13995 *
13996 * This allocates the sw VEB structure and links it into the switch
13997 * It is possible and legal for this to be a duplicate of an already
13998 * existing VEB.  It is also possible for both uplink and vsi seids
13999 * to be zero, in order to create a floating VEB.
14000 *
14001 * Returns pointer to the successfully allocated VEB sw struct on
14002 * success, otherwise returns NULL on failure.
14003 **/
14004struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
14005                                u16 uplink_seid, u16 vsi_seid,
14006                                u8 enabled_tc)
14007{
14008        struct i40e_veb *veb, *uplink_veb = NULL;
14009        int vsi_idx, veb_idx;
14010        int ret;
14011
14012        /* if one seid is 0, the other must be 0 to create a floating relay */
14013        if ((uplink_seid == 0 || vsi_seid == 0) &&
14014            (uplink_seid + vsi_seid != 0)) {
14015                dev_info(&pf->pdev->dev,
14016                         "one, not both seid's are 0: uplink=%d vsi=%d\n",
14017                         uplink_seid, vsi_seid);
14018                return NULL;
14019        }
14020
14021        /* make sure there is such a vsi and uplink */
14022        for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
14023                if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
14024                        break;
14025        if (vsi_idx == pf->num_alloc_vsi && vsi_seid != 0) {
14026                dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
14027                         vsi_seid);
14028                return NULL;
14029        }
14030
14031        if (uplink_seid && uplink_seid != pf->mac_seid) {
14032                for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
14033                        if (pf->veb[veb_idx] &&
14034                            pf->veb[veb_idx]->seid == uplink_seid) {
14035                                uplink_veb = pf->veb[veb_idx];
14036                                break;
14037                        }
14038                }
14039                if (!uplink_veb) {
14040                        dev_info(&pf->pdev->dev,
14041                                 "uplink seid %d not found\n", uplink_seid);
14042                        return NULL;
14043                }
14044        }
14045
14046        /* get veb sw struct */
14047        veb_idx = i40e_veb_mem_alloc(pf);
14048        if (veb_idx < 0)
14049                goto err_alloc;
14050        veb = pf->veb[veb_idx];
14051        veb->flags = flags;
14052        veb->uplink_seid = uplink_seid;
14053        veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
14054        veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
14055
14056        /* create the VEB in the switch */
14057        ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
14058        if (ret)
14059                goto err_veb;
14060        if (vsi_idx == pf->lan_vsi)
14061                pf->lan_veb = veb->idx;
14062
14063        return veb;
14064
14065err_veb:
14066        i40e_veb_clear(veb);
14067err_alloc:
14068        return NULL;
14069}
14070
14071/**
14072 * i40e_setup_pf_switch_element - set PF vars based on switch type
14073 * @pf: board private structure
14074 * @ele: element we are building info from
14075 * @num_reported: total number of elements
14076 * @printconfig: should we print the contents
14077 *
14078 * helper function to assist in extracting a few useful SEID values.
14079 **/
14080static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
14081                                struct i40e_aqc_switch_config_element_resp *ele,
14082                                u16 num_reported, bool printconfig)
14083{
14084        u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
14085        u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
14086        u8 element_type = ele->element_type;
14087        u16 seid = le16_to_cpu(ele->seid);
14088
14089        if (printconfig)
14090                dev_info(&pf->pdev->dev,
14091                         "type=%d seid=%d uplink=%d downlink=%d\n",
14092                         element_type, seid, uplink_seid, downlink_seid);
14093
14094        switch (element_type) {
14095        case I40E_SWITCH_ELEMENT_TYPE_MAC:
14096                pf->mac_seid = seid;
14097                break;
14098        case I40E_SWITCH_ELEMENT_TYPE_VEB:
14099                /* Main VEB? */
14100                if (uplink_seid != pf->mac_seid)
14101                        break;
14102                if (pf->lan_veb >= I40E_MAX_VEB) {
14103                        int v;
14104
14105                        /* find existing or else empty VEB */
14106                        for (v = 0; v < I40E_MAX_VEB; v++) {
14107                                if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
14108                                        pf->lan_veb = v;
14109                                        break;
14110                                }
14111                        }
14112                        if (pf->lan_veb >= I40E_MAX_VEB) {
14113                                v = i40e_veb_mem_alloc(pf);
14114                                if (v < 0)
14115                                        break;
14116                                pf->lan_veb = v;
14117                        }
14118                }
14119                if (pf->lan_veb >= I40E_MAX_VEB)
14120                        break;
14121
14122                pf->veb[pf->lan_veb]->seid = seid;
14123                pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
14124                pf->veb[pf->lan_veb]->pf = pf;
14125                pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
14126                break;
14127        case I40E_SWITCH_ELEMENT_TYPE_VSI:
14128                if (num_reported != 1)
14129                        break;
14130                /* This is immediately after a reset so we can assume this is
14131                 * the PF's VSI
14132                 */
14133                pf->mac_seid = uplink_seid;
14134                pf->pf_seid = downlink_seid;
14135                pf->main_vsi_seid = seid;
14136                if (printconfig)
14137                        dev_info(&pf->pdev->dev,
14138                                 "pf_seid=%d main_vsi_seid=%d\n",
14139                                 pf->pf_seid, pf->main_vsi_seid);
14140                break;
14141        case I40E_SWITCH_ELEMENT_TYPE_PF:
14142        case I40E_SWITCH_ELEMENT_TYPE_VF:
14143        case I40E_SWITCH_ELEMENT_TYPE_EMP:
14144        case I40E_SWITCH_ELEMENT_TYPE_BMC:
14145        case I40E_SWITCH_ELEMENT_TYPE_PE:
14146        case I40E_SWITCH_ELEMENT_TYPE_PA:
14147                /* ignore these for now */
14148                break;
14149        default:
14150                dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
14151                         element_type, seid);
14152                break;
14153        }
14154}
14155
14156/**
14157 * i40e_fetch_switch_configuration - Get switch config from firmware
14158 * @pf: board private structure
14159 * @printconfig: should we print the contents
14160 *
14161 * Get the current switch configuration from the device and
14162 * extract a few useful SEID values.
14163 **/
14164int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
14165{
14166        struct i40e_aqc_get_switch_config_resp *sw_config;
14167        u16 next_seid = 0;
14168        int ret = 0;
14169        u8 *aq_buf;
14170        int i;
14171
14172        aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
14173        if (!aq_buf)
14174                return -ENOMEM;
14175
14176        sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
14177        do {
14178                u16 num_reported, num_total;
14179
14180                ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
14181                                                I40E_AQ_LARGE_BUF,
14182                                                &next_seid, NULL);
14183                if (ret) {
14184                        dev_info(&pf->pdev->dev,
14185                                 "get switch config failed err %s aq_err %s\n",
14186                                 i40e_stat_str(&pf->hw, ret),
14187                                 i40e_aq_str(&pf->hw,
14188                                             pf->hw.aq.asq_last_status));
14189                        kfree(aq_buf);
14190                        return -ENOENT;
14191                }
14192
14193                num_reported = le16_to_cpu(sw_config->header.num_reported);
14194                num_total = le16_to_cpu(sw_config->header.num_total);
14195
14196                if (printconfig)
14197                        dev_info(&pf->pdev->dev,
14198                                 "header: %d reported %d total\n",
14199                                 num_reported, num_total);
14200
14201                for (i = 0; i < num_reported; i++) {
14202                        struct i40e_aqc_switch_config_element_resp *ele =
14203                                &sw_config->element[i];
14204
14205                        i40e_setup_pf_switch_element(pf, ele, num_reported,
14206                                                     printconfig);
14207                }
14208        } while (next_seid != 0);
14209
14210        kfree(aq_buf);
14211        return ret;
14212}
14213
14214/**
14215 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
14216 * @pf: board private structure
14217 * @reinit: if the Main VSI needs to re-initialized.
14218 *
14219 * Returns 0 on success, negative value on failure
14220 **/
14221static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
14222{
14223        u16 flags = 0;
14224        int ret;
14225
14226        /* find out what's out there already */
14227        ret = i40e_fetch_switch_configuration(pf, false);
14228        if (ret) {
14229                dev_info(&pf->pdev->dev,
14230                         "couldn't fetch switch config, err %s aq_err %s\n",
14231                         i40e_stat_str(&pf->hw, ret),
14232                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
14233                return ret;
14234        }
14235        i40e_pf_reset_stats(pf);
14236
14237        /* set the switch config bit for the whole device to
14238         * support limited promisc or true promisc
14239         * when user requests promisc. The default is limited
14240         * promisc.
14241        */
14242
14243        if ((pf->hw.pf_id == 0) &&
14244            !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT)) {
14245                flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14246                pf->last_sw_conf_flags = flags;
14247        }
14248
14249        if (pf->hw.pf_id == 0) {
14250                u16 valid_flags;
14251
14252                valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14253                ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags, 0,
14254                                                NULL);
14255                if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
14256                        dev_info(&pf->pdev->dev,
14257                                 "couldn't set switch config bits, err %s aq_err %s\n",
14258                                 i40e_stat_str(&pf->hw, ret),
14259                                 i40e_aq_str(&pf->hw,
14260                                             pf->hw.aq.asq_last_status));
14261                        /* not a fatal problem, just keep going */
14262                }
14263                pf->last_sw_conf_valid_flags = valid_flags;
14264        }
14265
14266        /* first time setup */
14267        if (pf->lan_vsi == I40E_NO_VSI || reinit) {
14268                struct i40e_vsi *vsi = NULL;
14269                u16 uplink_seid;
14270
14271                /* Set up the PF VSI associated with the PF's main VSI
14272                 * that is already in the HW switch
14273                 */
14274                if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
14275                        uplink_seid = pf->veb[pf->lan_veb]->seid;
14276                else
14277                        uplink_seid = pf->mac_seid;
14278                if (pf->lan_vsi == I40E_NO_VSI)
14279                        vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
14280                else if (reinit)
14281                        vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
14282                if (!vsi) {
14283                        dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
14284                        i40e_cloud_filter_exit(pf);
14285                        i40e_fdir_teardown(pf);
14286                        return -EAGAIN;
14287                }
14288        } else {
14289                /* force a reset of TC and queue layout configurations */
14290                u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
14291
14292                pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
14293                pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
14294                i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
14295        }
14296        i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
14297
14298        i40e_fdir_sb_setup(pf);
14299
14300        /* Setup static PF queue filter control settings */
14301        ret = i40e_setup_pf_filter_control(pf);
14302        if (ret) {
14303                dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
14304                         ret);
14305                /* Failure here should not stop continuing other steps */
14306        }
14307
14308        /* enable RSS in the HW, even for only one queue, as the stack can use
14309         * the hash
14310         */
14311        if ((pf->flags & I40E_FLAG_RSS_ENABLED))
14312                i40e_pf_config_rss(pf);
14313
14314        /* fill in link information and enable LSE reporting */
14315        i40e_link_event(pf);
14316
14317        /* Initialize user-specific link properties */
14318        pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
14319                                  I40E_AQ_AN_COMPLETED) ? true : false);
14320
14321        i40e_ptp_init(pf);
14322
14323        /* repopulate tunnel port filters */
14324        i40e_sync_udp_filters(pf);
14325
14326        return ret;
14327}
14328
14329/**
14330 * i40e_determine_queue_usage - Work out queue distribution
14331 * @pf: board private structure
14332 **/
14333static void i40e_determine_queue_usage(struct i40e_pf *pf)
14334{
14335        int queues_left;
14336        int q_max;
14337
14338        pf->num_lan_qps = 0;
14339
14340        /* Find the max queues to be put into basic use.  We'll always be
14341         * using TC0, whether or not DCB is running, and TC0 will get the
14342         * big RSS set.
14343         */
14344        queues_left = pf->hw.func_caps.num_tx_qp;
14345
14346        if ((queues_left == 1) ||
14347            !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
14348                /* one qp for PF, no queues for anything else */
14349                queues_left = 0;
14350                pf->alloc_rss_size = pf->num_lan_qps = 1;
14351
14352                /* make sure all the fancies are disabled */
14353                pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14354                               I40E_FLAG_IWARP_ENABLED  |
14355                               I40E_FLAG_FD_SB_ENABLED  |
14356                               I40E_FLAG_FD_ATR_ENABLED |
14357                               I40E_FLAG_DCB_CAPABLE    |
14358                               I40E_FLAG_DCB_ENABLED    |
14359                               I40E_FLAG_SRIOV_ENABLED  |
14360                               I40E_FLAG_VMDQ_ENABLED);
14361                pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14362        } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
14363                                  I40E_FLAG_FD_SB_ENABLED |
14364                                  I40E_FLAG_FD_ATR_ENABLED |
14365                                  I40E_FLAG_DCB_CAPABLE))) {
14366                /* one qp for PF */
14367                pf->alloc_rss_size = pf->num_lan_qps = 1;
14368                queues_left -= pf->num_lan_qps;
14369
14370                pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14371                               I40E_FLAG_IWARP_ENABLED  |
14372                               I40E_FLAG_FD_SB_ENABLED  |
14373                               I40E_FLAG_FD_ATR_ENABLED |
14374                               I40E_FLAG_DCB_ENABLED    |
14375                               I40E_FLAG_VMDQ_ENABLED);
14376                pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14377        } else {
14378                /* Not enough queues for all TCs */
14379                if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
14380                    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
14381                        pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
14382                                        I40E_FLAG_DCB_ENABLED);
14383                        dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
14384                }
14385
14386                /* limit lan qps to the smaller of qps, cpus or msix */
14387                q_max = max_t(int, pf->rss_size_max, num_online_cpus());
14388                q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
14389                q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
14390                pf->num_lan_qps = q_max;
14391
14392                queues_left -= pf->num_lan_qps;
14393        }
14394
14395        if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14396                if (queues_left > 1) {
14397                        queues_left -= 1; /* save 1 queue for FD */
14398                } else {
14399                        pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
14400                        pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14401                        dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
14402                }
14403        }
14404
14405        if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
14406            pf->num_vf_qps && pf->num_req_vfs && queues_left) {
14407                pf->num_req_vfs = min_t(int, pf->num_req_vfs,
14408                                        (queues_left / pf->num_vf_qps));
14409                queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
14410        }
14411
14412        if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
14413            pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
14414                pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
14415                                          (queues_left / pf->num_vmdq_qps));
14416                queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
14417        }
14418
14419        pf->queues_left = queues_left;
14420        dev_dbg(&pf->pdev->dev,
14421                "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
14422                pf->hw.func_caps.num_tx_qp,
14423                !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
14424                pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
14425                pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
14426                queues_left);
14427}
14428
14429/**
14430 * i40e_setup_pf_filter_control - Setup PF static filter control
14431 * @pf: PF to be setup
14432 *
14433 * i40e_setup_pf_filter_control sets up a PF's initial filter control
14434 * settings. If PE/FCoE are enabled then it will also set the per PF
14435 * based filter sizes required for them. It also enables Flow director,
14436 * ethertype and macvlan type filter settings for the pf.
14437 *
14438 * Returns 0 on success, negative on failure
14439 **/
14440static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
14441{
14442        struct i40e_filter_control_settings *settings = &pf->filter_settings;
14443
14444        settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
14445
14446        /* Flow Director is enabled */
14447        if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
14448                settings->enable_fdir = true;
14449
14450        /* Ethtype and MACVLAN filters enabled for PF */
14451        settings->enable_ethtype = true;
14452        settings->enable_macvlan = true;
14453
14454        if (i40e_set_filter_control(&pf->hw, settings))
14455                return -ENOENT;
14456
14457        return 0;
14458}
14459
14460#define INFO_STRING_LEN 255
14461#define REMAIN(__x) (INFO_STRING_LEN - (__x))
14462static void i40e_print_features(struct i40e_pf *pf)
14463{
14464        struct i40e_hw *hw = &pf->hw;
14465        char *buf;
14466        int i;
14467
14468        buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
14469        if (!buf)
14470                return;
14471
14472        i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
14473#ifdef CONFIG_PCI_IOV
14474        i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
14475#endif
14476        i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
14477                      pf->hw.func_caps.num_vsis,
14478                      pf->vsi[pf->lan_vsi]->num_queue_pairs);
14479        if (pf->flags & I40E_FLAG_RSS_ENABLED)
14480                i += snprintf(&buf[i], REMAIN(i), " RSS");
14481        if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
14482                i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
14483        if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14484                i += snprintf(&buf[i], REMAIN(i), " FD_SB");
14485                i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
14486        }
14487        if (pf->flags & I40E_FLAG_DCB_CAPABLE)
14488                i += snprintf(&buf[i], REMAIN(i), " DCB");
14489        i += snprintf(&buf[i], REMAIN(i), " VxLAN");
14490        i += snprintf(&buf[i], REMAIN(i), " Geneve");
14491        if (pf->flags & I40E_FLAG_PTP)
14492                i += snprintf(&buf[i], REMAIN(i), " PTP");
14493        if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
14494                i += snprintf(&buf[i], REMAIN(i), " VEB");
14495        else
14496                i += snprintf(&buf[i], REMAIN(i), " VEPA");
14497
14498        dev_info(&pf->pdev->dev, "%s\n", buf);
14499        kfree(buf);
14500        WARN_ON(i > INFO_STRING_LEN);
14501}
14502
14503/**
14504 * i40e_get_platform_mac_addr - get platform-specific MAC address
14505 * @pdev: PCI device information struct
14506 * @pf: board private structure
14507 *
14508 * Look up the MAC address for the device. First we'll try
14509 * eth_platform_get_mac_address, which will check Open Firmware, or arch
14510 * specific fallback. Otherwise, we'll default to the stored value in
14511 * firmware.
14512 **/
14513static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
14514{
14515        if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
14516                i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
14517}
14518
14519/**
14520 * i40e_set_fec_in_flags - helper function for setting FEC options in flags
14521 * @fec_cfg: FEC option to set in flags
14522 * @flags: ptr to flags in which we set FEC option
14523 **/
14524void i40e_set_fec_in_flags(u8 fec_cfg, u32 *flags)
14525{
14526        if (fec_cfg & I40E_AQ_SET_FEC_AUTO)
14527                *flags |= I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC;
14528        if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_RS) ||
14529            (fec_cfg & I40E_AQ_SET_FEC_ABILITY_RS)) {
14530                *flags |= I40E_FLAG_RS_FEC;
14531                *flags &= ~I40E_FLAG_BASE_R_FEC;
14532        }
14533        if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_KR) ||
14534            (fec_cfg & I40E_AQ_SET_FEC_ABILITY_KR)) {
14535                *flags |= I40E_FLAG_BASE_R_FEC;
14536                *flags &= ~I40E_FLAG_RS_FEC;
14537        }
14538        if (fec_cfg == 0)
14539                *flags &= ~(I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC);
14540}
14541
14542/**
14543 * i40e_check_recovery_mode - check if we are running transition firmware
14544 * @pf: board private structure
14545 *
14546 * Check registers indicating the firmware runs in recovery mode. Sets the
14547 * appropriate driver state.
14548 *
14549 * Returns true if the recovery mode was detected, false otherwise
14550 **/
14551static bool i40e_check_recovery_mode(struct i40e_pf *pf)
14552{
14553        u32 val = rd32(&pf->hw, I40E_GL_FWSTS) & I40E_GL_FWSTS_FWS1B_MASK;
14554        bool is_recovery_mode = false;
14555
14556        if (pf->hw.mac.type == I40E_MAC_XL710)
14557                is_recovery_mode =
14558                val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14559                val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK ||
14560                val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_TRANSITION_MASK ||
14561                val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_NVM_MASK;
14562        if (pf->hw.mac.type == I40E_MAC_X722)
14563                is_recovery_mode =
14564                val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14565                val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK;
14566        if (is_recovery_mode) {
14567                dev_notice(&pf->pdev->dev, "Firmware recovery mode detected. Limiting functionality.\n");
14568                dev_notice(&pf->pdev->dev, "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n");
14569                set_bit(__I40E_RECOVERY_MODE, pf->state);
14570
14571                return true;
14572        }
14573        if (test_and_clear_bit(__I40E_RECOVERY_MODE, pf->state))
14574                dev_info(&pf->pdev->dev, "Reinitializing in normal mode with full functionality.\n");
14575
14576        return false;
14577}
14578
14579/**
14580 * i40e_pf_loop_reset - perform reset in a loop.
14581 * @pf: board private structure
14582 *
14583 * This function is useful when a NIC is about to enter recovery mode.
14584 * When a NIC's internal data structures are corrupted the NIC's
14585 * firmware is going to enter recovery mode.
14586 * Right after a POR it takes about 7 minutes for firmware to enter
14587 * recovery mode. Until that time a NIC is in some kind of intermediate
14588 * state. After that time period the NIC almost surely enters
14589 * recovery mode. The only way for a driver to detect intermediate
14590 * state is to issue a series of pf-resets and check a return value.
14591 * If a PF reset returns success then the firmware could be in recovery
14592 * mode so the caller of this code needs to check for recovery mode
14593 * if this function returns success. There is a little chance that
14594 * firmware will hang in intermediate state forever.
14595 * Since waiting 7 minutes is quite a lot of time this function waits
14596 * 10 seconds and then gives up by returning an error.
14597 *
14598 * Return 0 on success, negative on failure.
14599 **/
14600static i40e_status i40e_pf_loop_reset(struct i40e_pf *pf)
14601{
14602        const unsigned short MAX_CNT = 1000;
14603        const unsigned short MSECS = 10;
14604        struct i40e_hw *hw = &pf->hw;
14605        i40e_status ret;
14606        int cnt;
14607
14608        for (cnt = 0; cnt < MAX_CNT; ++cnt) {
14609                ret = i40e_pf_reset(hw);
14610                if (!ret)
14611                        break;
14612                msleep(MSECS);
14613        }
14614
14615        if (cnt == MAX_CNT) {
14616                dev_info(&pf->pdev->dev, "PF reset failed: %d\n", ret);
14617                return ret;
14618        }
14619
14620        pf->pfr_count++;
14621        return ret;
14622}
14623
14624/**
14625 * i40e_init_recovery_mode - initialize subsystems needed in recovery mode
14626 * @pf: board private structure
14627 * @hw: ptr to the hardware info
14628 *
14629 * This function does a minimal setup of all subsystems needed for running
14630 * recovery mode.
14631 *
14632 * Returns 0 on success, negative on failure
14633 **/
14634static int i40e_init_recovery_mode(struct i40e_pf *pf, struct i40e_hw *hw)
14635{
14636        struct i40e_vsi *vsi;
14637        int err;
14638        int v_idx;
14639
14640        pci_save_state(pf->pdev);
14641
14642        /* set up periodic task facility */
14643        timer_setup(&pf->service_timer, i40e_service_timer, 0);
14644        pf->service_timer_period = HZ;
14645
14646        INIT_WORK(&pf->service_task, i40e_service_task);
14647        clear_bit(__I40E_SERVICE_SCHED, pf->state);
14648
14649        err = i40e_init_interrupt_scheme(pf);
14650        if (err)
14651                goto err_switch_setup;
14652
14653        /* The number of VSIs reported by the FW is the minimum guaranteed
14654         * to us; HW supports far more and we share the remaining pool with
14655         * the other PFs. We allocate space for more than the guarantee with
14656         * the understanding that we might not get them all later.
14657         */
14658        if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
14659                pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
14660        else
14661                pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
14662
14663        /* Set up the vsi struct and our local tracking of the MAIN PF vsi. */
14664        pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
14665                          GFP_KERNEL);
14666        if (!pf->vsi) {
14667                err = -ENOMEM;
14668                goto err_switch_setup;
14669        }
14670
14671        /* We allocate one VSI which is needed as absolute minimum
14672         * in order to register the netdev
14673         */
14674        v_idx = i40e_vsi_mem_alloc(pf, I40E_VSI_MAIN);
14675        if (v_idx < 0)
14676                goto err_switch_setup;
14677        pf->lan_vsi = v_idx;
14678        vsi = pf->vsi[v_idx];
14679        if (!vsi)
14680                goto err_switch_setup;
14681        vsi->alloc_queue_pairs = 1;
14682        err = i40e_config_netdev(vsi);
14683        if (err)
14684                goto err_switch_setup;
14685        err = register_netdev(vsi->netdev);
14686        if (err)
14687                goto err_switch_setup;
14688        vsi->netdev_registered = true;
14689        i40e_dbg_pf_init(pf);
14690
14691        err = i40e_setup_misc_vector_for_recovery_mode(pf);
14692        if (err)
14693                goto err_switch_setup;
14694
14695        /* tell the firmware that we're starting */
14696        i40e_send_version(pf);
14697
14698        /* since everything's happy, start the service_task timer */
14699        mod_timer(&pf->service_timer,
14700                  round_jiffies(jiffies + pf->service_timer_period));
14701
14702        return 0;
14703
14704err_switch_setup:
14705        i40e_reset_interrupt_capability(pf);
14706        del_timer_sync(&pf->service_timer);
14707        i40e_shutdown_adminq(hw);
14708        iounmap(hw->hw_addr);
14709        pci_disable_pcie_error_reporting(pf->pdev);
14710        pci_release_mem_regions(pf->pdev);
14711        pci_disable_device(pf->pdev);
14712        kfree(pf);
14713
14714        return err;
14715}
14716
14717/**
14718 * i40e_probe - Device initialization routine
14719 * @pdev: PCI device information struct
14720 * @ent: entry in i40e_pci_tbl
14721 *
14722 * i40e_probe initializes a PF identified by a pci_dev structure.
14723 * The OS initialization, configuring of the PF private structure,
14724 * and a hardware reset occur.
14725 *
14726 * Returns 0 on success, negative on failure
14727 **/
14728static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
14729{
14730        struct i40e_aq_get_phy_abilities_resp abilities;
14731        struct i40e_pf *pf;
14732        struct i40e_hw *hw;
14733        static u16 pfs_found;
14734        u16 wol_nvm_bits;
14735        u16 link_status;
14736        int err;
14737        u32 val;
14738        u32 i;
14739        u8 set_fc_aq_fail;
14740
14741        err = pci_enable_device_mem(pdev);
14742        if (err)
14743                return err;
14744
14745        /* set up for high or low dma */
14746        err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
14747        if (err) {
14748                err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
14749                if (err) {
14750                        dev_err(&pdev->dev,
14751                                "DMA configuration failed: 0x%x\n", err);
14752                        goto err_dma;
14753                }
14754        }
14755
14756        /* set up pci connections */
14757        err = pci_request_mem_regions(pdev, i40e_driver_name);
14758        if (err) {
14759                dev_info(&pdev->dev,
14760                         "pci_request_selected_regions failed %d\n", err);
14761                goto err_pci_reg;
14762        }
14763
14764        pci_enable_pcie_error_reporting(pdev);
14765        pci_set_master(pdev);
14766
14767        /* Now that we have a PCI connection, we need to do the
14768         * low level device setup.  This is primarily setting up
14769         * the Admin Queue structures and then querying for the
14770         * device's current profile information.
14771         */
14772        pf = kzalloc(sizeof(*pf), GFP_KERNEL);
14773        if (!pf) {
14774                err = -ENOMEM;
14775                goto err_pf_alloc;
14776        }
14777        pf->next_vsi = 0;
14778        pf->pdev = pdev;
14779        set_bit(__I40E_DOWN, pf->state);
14780
14781        hw = &pf->hw;
14782        hw->back = pf;
14783
14784        pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
14785                                I40E_MAX_CSR_SPACE);
14786        /* We believe that the highest register to read is
14787         * I40E_GLGEN_STAT_CLEAR, so we check if the BAR size
14788         * is not less than that before mapping to prevent a
14789         * kernel panic.
14790         */
14791        if (pf->ioremap_len < I40E_GLGEN_STAT_CLEAR) {
14792                dev_err(&pdev->dev, "Cannot map registers, bar size 0x%X too small, aborting\n",
14793                        pf->ioremap_len);
14794                err = -ENOMEM;
14795                goto err_ioremap;
14796        }
14797        hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
14798        if (!hw->hw_addr) {
14799                err = -EIO;
14800                dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
14801                         (unsigned int)pci_resource_start(pdev, 0),
14802                         pf->ioremap_len, err);
14803                goto err_ioremap;
14804        }
14805        hw->vendor_id = pdev->vendor;
14806        hw->device_id = pdev->device;
14807        pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
14808        hw->subsystem_vendor_id = pdev->subsystem_vendor;
14809        hw->subsystem_device_id = pdev->subsystem_device;
14810        hw->bus.device = PCI_SLOT(pdev->devfn);
14811        hw->bus.func = PCI_FUNC(pdev->devfn);
14812        hw->bus.bus_id = pdev->bus->number;
14813        pf->instance = pfs_found;
14814
14815        /* Select something other than the 802.1ad ethertype for the
14816         * switch to use internally and drop on ingress.
14817         */
14818        hw->switch_tag = 0xffff;
14819        hw->first_tag = ETH_P_8021AD;
14820        hw->second_tag = ETH_P_8021Q;
14821
14822        INIT_LIST_HEAD(&pf->l3_flex_pit_list);
14823        INIT_LIST_HEAD(&pf->l4_flex_pit_list);
14824        INIT_LIST_HEAD(&pf->ddp_old_prof);
14825
14826        /* set up the locks for the AQ, do this only once in probe
14827         * and destroy them only once in remove
14828         */
14829        mutex_init(&hw->aq.asq_mutex);
14830        mutex_init(&hw->aq.arq_mutex);
14831
14832        pf->msg_enable = netif_msg_init(debug,
14833                                        NETIF_MSG_DRV |
14834                                        NETIF_MSG_PROBE |
14835                                        NETIF_MSG_LINK);
14836        if (debug < -1)
14837                pf->hw.debug_mask = debug;
14838
14839        /* do a special CORER for clearing PXE mode once at init */
14840        if (hw->revision_id == 0 &&
14841            (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
14842                wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
14843                i40e_flush(hw);
14844                msleep(200);
14845                pf->corer_count++;
14846
14847                i40e_clear_pxe_mode(hw);
14848        }
14849
14850        /* Reset here to make sure all is clean and to define PF 'n' */
14851        i40e_clear_hw(hw);
14852
14853        err = i40e_set_mac_type(hw);
14854        if (err) {
14855                dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14856                         err);
14857                goto err_pf_reset;
14858        }
14859
14860        err = i40e_pf_loop_reset(pf);
14861        if (err) {
14862                dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
14863                goto err_pf_reset;
14864        }
14865
14866        i40e_check_recovery_mode(pf);
14867
14868        hw->aq.num_arq_entries = I40E_AQ_LEN;
14869        hw->aq.num_asq_entries = I40E_AQ_LEN;
14870        hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14871        hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14872        pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
14873
14874        snprintf(pf->int_name, sizeof(pf->int_name) - 1,
14875                 "%s-%s:misc",
14876                 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
14877
14878        err = i40e_init_shared_code(hw);
14879        if (err) {
14880                dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14881                         err);
14882                goto err_pf_reset;
14883        }
14884
14885        /* set up a default setting for link flow control */
14886        pf->hw.fc.requested_mode = I40E_FC_NONE;
14887
14888        err = i40e_init_adminq(hw);
14889        if (err) {
14890                if (err == I40E_ERR_FIRMWARE_API_VERSION)
14891                        dev_info(&pdev->dev,
14892                                 "The driver for the device stopped because the NVM image v%u.%u is newer than expected v%u.%u. You must install the most recent version of the network driver.\n",
14893                                 hw->aq.api_maj_ver,
14894                                 hw->aq.api_min_ver,
14895                                 I40E_FW_API_VERSION_MAJOR,
14896                                 I40E_FW_MINOR_VERSION(hw));
14897                else
14898                        dev_info(&pdev->dev,
14899                                 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
14900
14901                goto err_pf_reset;
14902        }
14903        i40e_get_oem_version(hw);
14904
14905        /* provide nvm, fw, api versions, vendor:device id, subsys vendor:device id */
14906        dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s [%04x:%04x] [%04x:%04x]\n",
14907                 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
14908                 hw->aq.api_maj_ver, hw->aq.api_min_ver,
14909                 i40e_nvm_version_str(hw), hw->vendor_id, hw->device_id,
14910                 hw->subsystem_vendor_id, hw->subsystem_device_id);
14911
14912        if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
14913            hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
14914                dev_info(&pdev->dev,
14915                         "The driver for the device detected a newer version of the NVM image v%u.%u than expected v%u.%u. Please install the most recent version of the network driver.\n",
14916                         hw->aq.api_maj_ver,
14917                         hw->aq.api_min_ver,
14918                         I40E_FW_API_VERSION_MAJOR,
14919                         I40E_FW_MINOR_VERSION(hw));
14920        else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
14921                dev_info(&pdev->dev,
14922                         "The driver for the device detected an older version of the NVM image v%u.%u than expected v%u.%u. Please update the NVM image.\n",
14923                         hw->aq.api_maj_ver,
14924                         hw->aq.api_min_ver,
14925                         I40E_FW_API_VERSION_MAJOR,
14926                         I40E_FW_MINOR_VERSION(hw));
14927
14928        i40e_verify_eeprom(pf);
14929
14930        /* Rev 0 hardware was never productized */
14931        if (hw->revision_id < 1)
14932                dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
14933
14934        i40e_clear_pxe_mode(hw);
14935
14936        err = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
14937        if (err)
14938                goto err_adminq_setup;
14939
14940        err = i40e_sw_init(pf);
14941        if (err) {
14942                dev_info(&pdev->dev, "sw_init failed: %d\n", err);
14943                goto err_sw_init;
14944        }
14945
14946        if (test_bit(__I40E_RECOVERY_MODE, pf->state))
14947                return i40e_init_recovery_mode(pf, hw);
14948
14949        err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
14950                                hw->func_caps.num_rx_qp, 0, 0);
14951        if (err) {
14952                dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
14953                goto err_init_lan_hmc;
14954        }
14955
14956        err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
14957        if (err) {
14958                dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
14959                err = -ENOENT;
14960                goto err_configure_lan_hmc;
14961        }
14962
14963        /* Disable LLDP for NICs that have firmware versions lower than v4.3.
14964         * Ignore error return codes because if it was already disabled via
14965         * hardware settings this will fail
14966         */
14967        if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
14968                dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
14969                i40e_aq_stop_lldp(hw, true, false, NULL);
14970        }
14971
14972        /* allow a platform config to override the HW addr */
14973        i40e_get_platform_mac_addr(pdev, pf);
14974
14975        if (!is_valid_ether_addr(hw->mac.addr)) {
14976                dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
14977                err = -EIO;
14978                goto err_mac_addr;
14979        }
14980        dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
14981        ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
14982        i40e_get_port_mac_addr(hw, hw->mac.port_addr);
14983        if (is_valid_ether_addr(hw->mac.port_addr))
14984                pf->hw_features |= I40E_HW_PORT_ID_VALID;
14985
14986        pci_set_drvdata(pdev, pf);
14987        pci_save_state(pdev);
14988
14989        dev_info(&pdev->dev,
14990                 (pf->flags & I40E_FLAG_DISABLE_FW_LLDP) ?
14991                        "FW LLDP is disabled\n" :
14992                        "FW LLDP is enabled\n");
14993
14994        /* Enable FW to write default DCB config on link-up */
14995        i40e_aq_set_dcb_parameters(hw, true, NULL);
14996
14997#ifdef CONFIG_I40E_DCB
14998        err = i40e_init_pf_dcb(pf);
14999        if (err) {
15000                dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
15001                pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
15002                /* Continue without DCB enabled */
15003        }
15004#endif /* CONFIG_I40E_DCB */
15005
15006        /* set up periodic task facility */
15007        timer_setup(&pf->service_timer, i40e_service_timer, 0);
15008        pf->service_timer_period = HZ;
15009
15010        INIT_WORK(&pf->service_task, i40e_service_task);
15011        clear_bit(__I40E_SERVICE_SCHED, pf->state);
15012
15013        /* NVM bit on means WoL disabled for the port */
15014        i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
15015        if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
15016                pf->wol_en = false;
15017        else
15018                pf->wol_en = true;
15019        device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
15020
15021        /* set up the main switch operations */
15022        i40e_determine_queue_usage(pf);
15023        err = i40e_init_interrupt_scheme(pf);
15024        if (err)
15025                goto err_switch_setup;
15026
15027        /* The number of VSIs reported by the FW is the minimum guaranteed
15028         * to us; HW supports far more and we share the remaining pool with
15029         * the other PFs. We allocate space for more than the guarantee with
15030         * the understanding that we might not get them all later.
15031         */
15032        if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
15033                pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
15034        else
15035                pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
15036
15037        /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
15038        pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
15039                          GFP_KERNEL);
15040        if (!pf->vsi) {
15041                err = -ENOMEM;
15042                goto err_switch_setup;
15043        }
15044
15045#ifdef CONFIG_PCI_IOV
15046        /* prep for VF support */
15047        if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15048            (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15049            !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15050                if (pci_num_vf(pdev))
15051                        pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
15052        }
15053#endif
15054        err = i40e_setup_pf_switch(pf, false);
15055        if (err) {
15056                dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
15057                goto err_vsis;
15058        }
15059        INIT_LIST_HEAD(&pf->vsi[pf->lan_vsi]->ch_list);
15060
15061        /* Make sure flow control is set according to current settings */
15062        err = i40e_set_fc(hw, &set_fc_aq_fail, true);
15063        if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
15064                dev_dbg(&pf->pdev->dev,
15065                        "Set fc with err %s aq_err %s on get_phy_cap\n",
15066                        i40e_stat_str(hw, err),
15067                        i40e_aq_str(hw, hw->aq.asq_last_status));
15068        if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
15069                dev_dbg(&pf->pdev->dev,
15070                        "Set fc with err %s aq_err %s on set_phy_config\n",
15071                        i40e_stat_str(hw, err),
15072                        i40e_aq_str(hw, hw->aq.asq_last_status));
15073        if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
15074                dev_dbg(&pf->pdev->dev,
15075                        "Set fc with err %s aq_err %s on get_link_info\n",
15076                        i40e_stat_str(hw, err),
15077                        i40e_aq_str(hw, hw->aq.asq_last_status));
15078
15079        /* if FDIR VSI was set up, start it now */
15080        for (i = 0; i < pf->num_alloc_vsi; i++) {
15081                if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
15082                        i40e_vsi_open(pf->vsi[i]);
15083                        break;
15084                }
15085        }
15086
15087        /* The driver only wants link up/down and module qualification
15088         * reports from firmware.  Note the negative logic.
15089         */
15090        err = i40e_aq_set_phy_int_mask(&pf->hw,
15091                                       ~(I40E_AQ_EVENT_LINK_UPDOWN |
15092                                         I40E_AQ_EVENT_MEDIA_NA |
15093                                         I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
15094        if (err)
15095                dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
15096                         i40e_stat_str(&pf->hw, err),
15097                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15098
15099        /* Reconfigure hardware for allowing smaller MSS in the case
15100         * of TSO, so that we avoid the MDD being fired and causing
15101         * a reset in the case of small MSS+TSO.
15102         */
15103        val = rd32(hw, I40E_REG_MSS);
15104        if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
15105                val &= ~I40E_REG_MSS_MIN_MASK;
15106                val |= I40E_64BYTE_MSS;
15107                wr32(hw, I40E_REG_MSS, val);
15108        }
15109
15110        if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
15111                msleep(75);
15112                err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
15113                if (err)
15114                        dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
15115                                 i40e_stat_str(&pf->hw, err),
15116                                 i40e_aq_str(&pf->hw,
15117                                             pf->hw.aq.asq_last_status));
15118        }
15119        /* The main driver is (mostly) up and happy. We need to set this state
15120         * before setting up the misc vector or we get a race and the vector
15121         * ends up disabled forever.
15122         */
15123        clear_bit(__I40E_DOWN, pf->state);
15124
15125        /* In case of MSIX we are going to setup the misc vector right here
15126         * to handle admin queue events etc. In case of legacy and MSI
15127         * the misc functionality and queue processing is combined in
15128         * the same vector and that gets setup at open.
15129         */
15130        if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
15131                err = i40e_setup_misc_vector(pf);
15132                if (err) {
15133                        dev_info(&pdev->dev,
15134                                 "setup of misc vector failed: %d\n", err);
15135                        goto err_vsis;
15136                }
15137        }
15138
15139#ifdef CONFIG_PCI_IOV
15140        /* prep for VF support */
15141        if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15142            (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15143            !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15144                /* disable link interrupts for VFs */
15145                val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
15146                val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
15147                wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
15148                i40e_flush(hw);
15149
15150                if (pci_num_vf(pdev)) {
15151                        dev_info(&pdev->dev,
15152                                 "Active VFs found, allocating resources.\n");
15153                        err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
15154                        if (err)
15155                                dev_info(&pdev->dev,
15156                                         "Error %d allocating resources for existing VFs\n",
15157                                         err);
15158                }
15159        }
15160#endif /* CONFIG_PCI_IOV */
15161
15162        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15163                pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
15164                                                      pf->num_iwarp_msix,
15165                                                      I40E_IWARP_IRQ_PILE_ID);
15166                if (pf->iwarp_base_vector < 0) {
15167                        dev_info(&pdev->dev,
15168                                 "failed to get tracking for %d vectors for IWARP err=%d\n",
15169                                 pf->num_iwarp_msix, pf->iwarp_base_vector);
15170                        pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
15171                }
15172        }
15173
15174        i40e_dbg_pf_init(pf);
15175
15176        /* tell the firmware that we're starting */
15177        i40e_send_version(pf);
15178
15179        /* since everything's happy, start the service_task timer */
15180        mod_timer(&pf->service_timer,
15181                  round_jiffies(jiffies + pf->service_timer_period));
15182
15183        /* add this PF to client device list and launch a client service task */
15184        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15185                err = i40e_lan_add_device(pf);
15186                if (err)
15187                        dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
15188                                 err);
15189        }
15190
15191#define PCI_SPEED_SIZE 8
15192#define PCI_WIDTH_SIZE 8
15193        /* Devices on the IOSF bus do not have this information
15194         * and will report PCI Gen 1 x 1 by default so don't bother
15195         * checking them.
15196         */
15197        if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
15198                char speed[PCI_SPEED_SIZE] = "Unknown";
15199                char width[PCI_WIDTH_SIZE] = "Unknown";
15200
15201                /* Get the negotiated link width and speed from PCI config
15202                 * space
15203                 */
15204                pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
15205                                          &link_status);
15206
15207                i40e_set_pci_config_data(hw, link_status);
15208
15209                switch (hw->bus.speed) {
15210                case i40e_bus_speed_8000:
15211                        strlcpy(speed, "8.0", PCI_SPEED_SIZE); break;
15212                case i40e_bus_speed_5000:
15213                        strlcpy(speed, "5.0", PCI_SPEED_SIZE); break;
15214                case i40e_bus_speed_2500:
15215                        strlcpy(speed, "2.5", PCI_SPEED_SIZE); break;
15216                default:
15217                        break;
15218                }
15219                switch (hw->bus.width) {
15220                case i40e_bus_width_pcie_x8:
15221                        strlcpy(width, "8", PCI_WIDTH_SIZE); break;
15222                case i40e_bus_width_pcie_x4:
15223                        strlcpy(width, "4", PCI_WIDTH_SIZE); break;
15224                case i40e_bus_width_pcie_x2:
15225                        strlcpy(width, "2", PCI_WIDTH_SIZE); break;
15226                case i40e_bus_width_pcie_x1:
15227                        strlcpy(width, "1", PCI_WIDTH_SIZE); break;
15228                default:
15229                        break;
15230                }
15231
15232                dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
15233                         speed, width);
15234
15235                if (hw->bus.width < i40e_bus_width_pcie_x8 ||
15236                    hw->bus.speed < i40e_bus_speed_8000) {
15237                        dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
15238                        dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
15239                }
15240        }
15241
15242        /* get the requested speeds from the fw */
15243        err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
15244        if (err)
15245                dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
15246                        i40e_stat_str(&pf->hw, err),
15247                        i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15248        pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
15249
15250        /* set the FEC config due to the board capabilities */
15251        i40e_set_fec_in_flags(abilities.fec_cfg_curr_mod_ext_info, &pf->flags);
15252
15253        /* get the supported phy types from the fw */
15254        err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
15255        if (err)
15256                dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
15257                        i40e_stat_str(&pf->hw, err),
15258                        i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15259
15260        /* Add a filter to drop all Flow control frames from any VSI from being
15261         * transmitted. By doing so we stop a malicious VF from sending out
15262         * PAUSE or PFC frames and potentially controlling traffic for other
15263         * PF/VF VSIs.
15264         * The FW can still send Flow control frames if enabled.
15265         */
15266        i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
15267                                                       pf->main_vsi_seid);
15268
15269        if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
15270                (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
15271                pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
15272        if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
15273                pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
15274        /* print a string summarizing features */
15275        i40e_print_features(pf);
15276
15277        return 0;
15278
15279        /* Unwind what we've done if something failed in the setup */
15280err_vsis:
15281        set_bit(__I40E_DOWN, pf->state);
15282        i40e_clear_interrupt_scheme(pf);
15283        kfree(pf->vsi);
15284err_switch_setup:
15285        i40e_reset_interrupt_capability(pf);
15286        del_timer_sync(&pf->service_timer);
15287err_mac_addr:
15288err_configure_lan_hmc:
15289        (void)i40e_shutdown_lan_hmc(hw);
15290err_init_lan_hmc:
15291        kfree(pf->qp_pile);
15292err_sw_init:
15293err_adminq_setup:
15294err_pf_reset:
15295        iounmap(hw->hw_addr);
15296err_ioremap:
15297        kfree(pf);
15298err_pf_alloc:
15299        pci_disable_pcie_error_reporting(pdev);
15300        pci_release_mem_regions(pdev);
15301err_pci_reg:
15302err_dma:
15303        pci_disable_device(pdev);
15304        return err;
15305}
15306
15307/**
15308 * i40e_remove - Device removal routine
15309 * @pdev: PCI device information struct
15310 *
15311 * i40e_remove is called by the PCI subsystem to alert the driver
15312 * that is should release a PCI device.  This could be caused by a
15313 * Hot-Plug event, or because the driver is going to be removed from
15314 * memory.
15315 **/
15316static void i40e_remove(struct pci_dev *pdev)
15317{
15318        struct i40e_pf *pf = pci_get_drvdata(pdev);
15319        struct i40e_hw *hw = &pf->hw;
15320        i40e_status ret_code;
15321        int i;
15322
15323        i40e_dbg_pf_exit(pf);
15324
15325        i40e_ptp_stop(pf);
15326
15327        /* Disable RSS in hw */
15328        i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
15329        i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
15330
15331        /* no more scheduling of any task */
15332        set_bit(__I40E_SUSPENDED, pf->state);
15333        set_bit(__I40E_DOWN, pf->state);
15334        if (pf->service_timer.function)
15335                del_timer_sync(&pf->service_timer);
15336        if (pf->service_task.func)
15337                cancel_work_sync(&pf->service_task);
15338
15339        if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
15340                struct i40e_vsi *vsi = pf->vsi[0];
15341
15342                /* We know that we have allocated only one vsi for this PF,
15343                 * it was just for registering netdevice, so the interface
15344                 * could be visible in the 'ifconfig' output
15345                 */
15346                unregister_netdev(vsi->netdev);
15347                free_netdev(vsi->netdev);
15348
15349                goto unmap;
15350        }
15351
15352        /* Client close must be called explicitly here because the timer
15353         * has been stopped.
15354         */
15355        i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15356
15357        if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
15358                i40e_free_vfs(pf);
15359                pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
15360        }
15361
15362        i40e_fdir_teardown(pf);
15363
15364        /* If there is a switch structure or any orphans, remove them.
15365         * This will leave only the PF's VSI remaining.
15366         */
15367        for (i = 0; i < I40E_MAX_VEB; i++) {
15368                if (!pf->veb[i])
15369                        continue;
15370
15371                if (pf->veb[i]->uplink_seid == pf->mac_seid ||
15372                    pf->veb[i]->uplink_seid == 0)
15373                        i40e_switch_branch_release(pf->veb[i]);
15374        }
15375
15376        /* Now we can shutdown the PF's VSI, just before we kill
15377         * adminq and hmc.
15378         */
15379        if (pf->vsi[pf->lan_vsi])
15380                i40e_vsi_release(pf->vsi[pf->lan_vsi]);
15381
15382        i40e_cloud_filter_exit(pf);
15383
15384        /* remove attached clients */
15385        if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15386                ret_code = i40e_lan_del_device(pf);
15387                if (ret_code)
15388                        dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
15389                                 ret_code);
15390        }
15391
15392        /* shutdown and destroy the HMC */
15393        if (hw->hmc.hmc_obj) {
15394                ret_code = i40e_shutdown_lan_hmc(hw);
15395                if (ret_code)
15396                        dev_warn(&pdev->dev,
15397                                 "Failed to destroy the HMC resources: %d\n",
15398                                 ret_code);
15399        }
15400
15401unmap:
15402        /* Free MSI/legacy interrupt 0 when in recovery mode. */
15403        if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15404            !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15405                free_irq(pf->pdev->irq, pf);
15406
15407        /* shutdown the adminq */
15408        i40e_shutdown_adminq(hw);
15409
15410        /* destroy the locks only once, here */
15411        mutex_destroy(&hw->aq.arq_mutex);
15412        mutex_destroy(&hw->aq.asq_mutex);
15413
15414        /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
15415        rtnl_lock();
15416        i40e_clear_interrupt_scheme(pf);
15417        for (i = 0; i < pf->num_alloc_vsi; i++) {
15418                if (pf->vsi[i]) {
15419                        if (!test_bit(__I40E_RECOVERY_MODE, pf->state))
15420                                i40e_vsi_clear_rings(pf->vsi[i]);
15421                        i40e_vsi_clear(pf->vsi[i]);
15422                        pf->vsi[i] = NULL;
15423                }
15424        }
15425        rtnl_unlock();
15426
15427        for (i = 0; i < I40E_MAX_VEB; i++) {
15428                kfree(pf->veb[i]);
15429                pf->veb[i] = NULL;
15430        }
15431
15432        kfree(pf->qp_pile);
15433        kfree(pf->vsi);
15434
15435        iounmap(hw->hw_addr);
15436        kfree(pf);
15437        pci_release_mem_regions(pdev);
15438
15439        pci_disable_pcie_error_reporting(pdev);
15440        pci_disable_device(pdev);
15441}
15442
15443/**
15444 * i40e_pci_error_detected - warning that something funky happened in PCI land
15445 * @pdev: PCI device information struct
15446 * @error: the type of PCI error
15447 *
15448 * Called to warn that something happened and the error handling steps
15449 * are in progress.  Allows the driver to quiesce things, be ready for
15450 * remediation.
15451 **/
15452static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
15453                                                enum pci_channel_state error)
15454{
15455        struct i40e_pf *pf = pci_get_drvdata(pdev);
15456
15457        dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
15458
15459        if (!pf) {
15460                dev_info(&pdev->dev,
15461                         "Cannot recover - error happened during device probe\n");
15462                return PCI_ERS_RESULT_DISCONNECT;
15463        }
15464
15465        /* shutdown all operations */
15466        if (!test_bit(__I40E_SUSPENDED, pf->state))
15467                i40e_prep_for_reset(pf, false);
15468
15469        /* Request a slot reset */
15470        return PCI_ERS_RESULT_NEED_RESET;
15471}
15472
15473/**
15474 * i40e_pci_error_slot_reset - a PCI slot reset just happened
15475 * @pdev: PCI device information struct
15476 *
15477 * Called to find if the driver can work with the device now that
15478 * the pci slot has been reset.  If a basic connection seems good
15479 * (registers are readable and have sane content) then return a
15480 * happy little PCI_ERS_RESULT_xxx.
15481 **/
15482static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
15483{
15484        struct i40e_pf *pf = pci_get_drvdata(pdev);
15485        pci_ers_result_t result;
15486        u32 reg;
15487
15488        dev_dbg(&pdev->dev, "%s\n", __func__);
15489        if (pci_enable_device_mem(pdev)) {
15490                dev_info(&pdev->dev,
15491                         "Cannot re-enable PCI device after reset.\n");
15492                result = PCI_ERS_RESULT_DISCONNECT;
15493        } else {
15494                pci_set_master(pdev);
15495                pci_restore_state(pdev);
15496                pci_save_state(pdev);
15497                pci_wake_from_d3(pdev, false);
15498
15499                reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
15500                if (reg == 0)
15501                        result = PCI_ERS_RESULT_RECOVERED;
15502                else
15503                        result = PCI_ERS_RESULT_DISCONNECT;
15504        }
15505
15506        return result;
15507}
15508
15509/**
15510 * i40e_pci_error_reset_prepare - prepare device driver for pci reset
15511 * @pdev: PCI device information struct
15512 */
15513static void i40e_pci_error_reset_prepare(struct pci_dev *pdev)
15514{
15515        struct i40e_pf *pf = pci_get_drvdata(pdev);
15516
15517        i40e_prep_for_reset(pf, false);
15518}
15519
15520/**
15521 * i40e_pci_error_reset_done - pci reset done, device driver reset can begin
15522 * @pdev: PCI device information struct
15523 */
15524static void i40e_pci_error_reset_done(struct pci_dev *pdev)
15525{
15526        struct i40e_pf *pf = pci_get_drvdata(pdev);
15527
15528        i40e_reset_and_rebuild(pf, false, false);
15529}
15530
15531/**
15532 * i40e_pci_error_resume - restart operations after PCI error recovery
15533 * @pdev: PCI device information struct
15534 *
15535 * Called to allow the driver to bring things back up after PCI error
15536 * and/or reset recovery has finished.
15537 **/
15538static void i40e_pci_error_resume(struct pci_dev *pdev)
15539{
15540        struct i40e_pf *pf = pci_get_drvdata(pdev);
15541
15542        dev_dbg(&pdev->dev, "%s\n", __func__);
15543        if (test_bit(__I40E_SUSPENDED, pf->state))
15544                return;
15545
15546        i40e_handle_reset_warning(pf, false);
15547}
15548
15549/**
15550 * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
15551 * using the mac_address_write admin q function
15552 * @pf: pointer to i40e_pf struct
15553 **/
15554static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
15555{
15556        struct i40e_hw *hw = &pf->hw;
15557        i40e_status ret;
15558        u8 mac_addr[6];
15559        u16 flags = 0;
15560
15561        /* Get current MAC address in case it's an LAA */
15562        if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
15563                ether_addr_copy(mac_addr,
15564                                pf->vsi[pf->lan_vsi]->netdev->dev_addr);
15565        } else {
15566                dev_err(&pf->pdev->dev,
15567                        "Failed to retrieve MAC address; using default\n");
15568                ether_addr_copy(mac_addr, hw->mac.addr);
15569        }
15570
15571        /* The FW expects the mac address write cmd to first be called with
15572         * one of these flags before calling it again with the multicast
15573         * enable flags.
15574         */
15575        flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
15576
15577        if (hw->func_caps.flex10_enable && hw->partition_id != 1)
15578                flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
15579
15580        ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15581        if (ret) {
15582                dev_err(&pf->pdev->dev,
15583                        "Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
15584                return;
15585        }
15586
15587        flags = I40E_AQC_MC_MAG_EN
15588                        | I40E_AQC_WOL_PRESERVE_ON_PFR
15589                        | I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
15590        ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15591        if (ret)
15592                dev_err(&pf->pdev->dev,
15593                        "Failed to enable Multicast Magic Packet wake up\n");
15594}
15595
15596/**
15597 * i40e_shutdown - PCI callback for shutting down
15598 * @pdev: PCI device information struct
15599 **/
15600static void i40e_shutdown(struct pci_dev *pdev)
15601{
15602        struct i40e_pf *pf = pci_get_drvdata(pdev);
15603        struct i40e_hw *hw = &pf->hw;
15604
15605        set_bit(__I40E_SUSPENDED, pf->state);
15606        set_bit(__I40E_DOWN, pf->state);
15607
15608        del_timer_sync(&pf->service_timer);
15609        cancel_work_sync(&pf->service_task);
15610        i40e_cloud_filter_exit(pf);
15611        i40e_fdir_teardown(pf);
15612
15613        /* Client close must be called explicitly here because the timer
15614         * has been stopped.
15615         */
15616        i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15617
15618        if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15619                i40e_enable_mc_magic_wake(pf);
15620
15621        i40e_prep_for_reset(pf, false);
15622
15623        wr32(hw, I40E_PFPM_APM,
15624             (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15625        wr32(hw, I40E_PFPM_WUFC,
15626             (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15627
15628        /* Free MSI/legacy interrupt 0 when in recovery mode. */
15629        if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15630            !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15631                free_irq(pf->pdev->irq, pf);
15632
15633        /* Since we're going to destroy queues during the
15634         * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15635         * whole section
15636         */
15637        rtnl_lock();
15638        i40e_clear_interrupt_scheme(pf);
15639        rtnl_unlock();
15640
15641        if (system_state == SYSTEM_POWER_OFF) {
15642                pci_wake_from_d3(pdev, pf->wol_en);
15643                pci_set_power_state(pdev, PCI_D3hot);
15644        }
15645}
15646
15647/**
15648 * i40e_suspend - PM callback for moving to D3
15649 * @dev: generic device information structure
15650 **/
15651static int __maybe_unused i40e_suspend(struct device *dev)
15652{
15653        struct i40e_pf *pf = dev_get_drvdata(dev);
15654        struct i40e_hw *hw = &pf->hw;
15655
15656        /* If we're already suspended, then there is nothing to do */
15657        if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
15658                return 0;
15659
15660        set_bit(__I40E_DOWN, pf->state);
15661
15662        /* Ensure service task will not be running */
15663        del_timer_sync(&pf->service_timer);
15664        cancel_work_sync(&pf->service_task);
15665
15666        /* Client close must be called explicitly here because the timer
15667         * has been stopped.
15668         */
15669        i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15670
15671        if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15672                i40e_enable_mc_magic_wake(pf);
15673
15674        /* Since we're going to destroy queues during the
15675         * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15676         * whole section
15677         */
15678        rtnl_lock();
15679
15680        i40e_prep_for_reset(pf, true);
15681
15682        wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15683        wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15684
15685        /* Clear the interrupt scheme and release our IRQs so that the system
15686         * can safely hibernate even when there are a large number of CPUs.
15687         * Otherwise hibernation might fail when mapping all the vectors back
15688         * to CPU0.
15689         */
15690        i40e_clear_interrupt_scheme(pf);
15691
15692        rtnl_unlock();
15693
15694        return 0;
15695}
15696
15697/**
15698 * i40e_resume - PM callback for waking up from D3
15699 * @dev: generic device information structure
15700 **/
15701static int __maybe_unused i40e_resume(struct device *dev)
15702{
15703        struct i40e_pf *pf = dev_get_drvdata(dev);
15704        int err;
15705
15706        /* If we're not suspended, then there is nothing to do */
15707        if (!test_bit(__I40E_SUSPENDED, pf->state))
15708                return 0;
15709
15710        /* We need to hold the RTNL lock prior to restoring interrupt schemes,
15711         * since we're going to be restoring queues
15712         */
15713        rtnl_lock();
15714
15715        /* We cleared the interrupt scheme when we suspended, so we need to
15716         * restore it now to resume device functionality.
15717         */
15718        err = i40e_restore_interrupt_scheme(pf);
15719        if (err) {
15720                dev_err(dev, "Cannot restore interrupt scheme: %d\n",
15721                        err);
15722        }
15723
15724        clear_bit(__I40E_DOWN, pf->state);
15725        i40e_reset_and_rebuild(pf, false, true);
15726
15727        rtnl_unlock();
15728
15729        /* Clear suspended state last after everything is recovered */
15730        clear_bit(__I40E_SUSPENDED, pf->state);
15731
15732        /* Restart the service task */
15733        mod_timer(&pf->service_timer,
15734                  round_jiffies(jiffies + pf->service_timer_period));
15735
15736        return 0;
15737}
15738
15739static const struct pci_error_handlers i40e_err_handler = {
15740        .error_detected = i40e_pci_error_detected,
15741        .slot_reset = i40e_pci_error_slot_reset,
15742        .reset_prepare = i40e_pci_error_reset_prepare,
15743        .reset_done = i40e_pci_error_reset_done,
15744        .resume = i40e_pci_error_resume,
15745};
15746
15747static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
15748
15749static struct pci_driver i40e_driver = {
15750        .name     = i40e_driver_name,
15751        .id_table = i40e_pci_tbl,
15752        .probe    = i40e_probe,
15753        .remove   = i40e_remove,
15754        .driver   = {
15755                .pm = &i40e_pm_ops,
15756        },
15757        .shutdown = i40e_shutdown,
15758        .err_handler = &i40e_err_handler,
15759        .sriov_configure = i40e_pci_sriov_configure,
15760};
15761
15762/**
15763 * i40e_init_module - Driver registration routine
15764 *
15765 * i40e_init_module is the first routine called when the driver is
15766 * loaded. All it does is register with the PCI subsystem.
15767 **/
15768static int __init i40e_init_module(void)
15769{
15770        pr_info("%s: %s - version %s\n", i40e_driver_name,
15771                i40e_driver_string, i40e_driver_version_str);
15772        pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
15773
15774        /* There is no need to throttle the number of active tasks because
15775         * each device limits its own task using a state bit for scheduling
15776         * the service task, and the device tasks do not interfere with each
15777         * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
15778         * since we need to be able to guarantee forward progress even under
15779         * memory pressure.
15780         */
15781        i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
15782        if (!i40e_wq) {
15783                pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
15784                return -ENOMEM;
15785        }
15786
15787        i40e_dbg_init();
15788        return pci_register_driver(&i40e_driver);
15789}
15790module_init(i40e_init_module);
15791
15792/**
15793 * i40e_exit_module - Driver exit cleanup routine
15794 *
15795 * i40e_exit_module is called just before the driver is removed
15796 * from memory.
15797 **/
15798static void __exit i40e_exit_module(void)
15799{
15800        pci_unregister_driver(&i40e_driver);
15801        destroy_workqueue(i40e_wq);
15802        i40e_dbg_exit();
15803}
15804module_exit(i40e_exit_module);
15805