drivers/net/ethernet/intel/i40e/i40e_main.c

   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
  22 const char i40e_driver_name[] = "i40e";
  23 static 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
  34 const char i40e_driver_version_str[] = DRV_VERSION;
  35 static const char i40e_copyright[] = "Copyright (c) 2013 - 2019 Intel Corporation.";
  36
  37 /* a bit of forward declarations */
  38 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  39 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
  40 static int i40e_add_vsi(struct i40e_vsi *vsi);
  41 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  42 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  43 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  44 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  45 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  46 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
  47 static int i40e_reset(struct i40e_pf *pf);
  48 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
  49 static int i40e_setup_misc_vector_for_recovery_mode(struct i40e_pf *pf);
  50 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf);
  51 static bool i40e_check_recovery_mode(struct i40e_pf *pf);
  52 static int i40e_init_recovery_mode(struct i40e_pf *pf, struct i40e_hw *hw);
  53 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  54 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  55 static 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  */
  66 static 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 };
  94 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
  95
  96 #define I40E_MAX_VF_COUNT 128
  97 static int debug = -1;
  98 module_param(debug, uint, 0);
  99 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
 100
 101 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 102 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 103 MODULE_LICENSE("GPL v2");
 104 MODULE_VERSION(DRV_VERSION);
 105
 106 static 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  **/
 115 int 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  **/
 134 int 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  **/
 152 int 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  **/
 169 int 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  **/
 192 static 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  **/
 244 static 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  **/
 271 struct 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  **/
 288 void 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  **/
 304 static void i40e_tx_timeout(struct net_device *netdev, unsigned int txqueue)
 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;
 311         u32 head, val;
 312
 313         pf->tx_timeout_count++;
 314
 315         /* with txqueue index, find the tx_ring struct */
 316         for (i = 0; i < vsi->num_queue_pairs; i++) {
 317                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 318                         if (txqueue ==
 319                             vsi->tx_rings[i]->queue_index) {
 320                                 tx_ring = vsi->tx_rings[i];
 321                                 break;
 322                         }
 323                 }
 324         }
 325
 326         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 327                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 328         else if (time_before(jiffies,
 329                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 330                 return;   /* don't do any new action before the next timeout */
 331
 332         /* don't kick off another recovery if one is already pending */
 333         if (test_and_set_bit(__I40E_TIMEOUT_RECOVERY_PENDING, pf->state))
 334                 return;
 335
 336         if (tx_ring) {
 337                 head = i40e_get_head(tx_ring);
 338                 /* Read interrupt register */
 339                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 340                         val = rd32(&pf->hw,
 341                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 342                                                 tx_ring->vsi->base_vector - 1));
 343                 else
 344                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 345
 346                 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",
 347                             vsi->seid, txqueue, tx_ring->next_to_clean,
 348                             head, tx_ring->next_to_use,
 349                             readl(tx_ring->tail), val);
 350         }
 351
 352         pf->tx_timeout_last_recovery = jiffies;
 353         netdev_info(netdev, "tx_timeout recovery level %d, txqueue %d\n",
 354                     pf->tx_timeout_recovery_level, txqueue);
 355
 356         switch (pf->tx_timeout_recovery_level) {
 357         case 1:
 358                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
 359                 break;
 360         case 2:
 361                 set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
 362                 break;
 363         case 3:
 364                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
 365                 break;
 366         default:
 367                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 368                 break;
 369         }
 370
 371         i40e_service_event_schedule(pf);
 372         pf->tx_timeout_recovery_level++;
 373 }
 374
 375 /**
 376  * i40e_get_vsi_stats_struct - Get System Network Statistics
 377  * @vsi: the VSI we care about
 378  *
 379  * Returns the address of the device statistics structure.
 380  * The statistics are actually updated from the service task.
 381  **/
 382 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 383 {
 384         return &vsi->net_stats;
 385 }
 386
 387 /**
 388  * i40e_get_netdev_stats_struct_tx - populate stats from a Tx ring
 389  * @ring: Tx ring to get statistics from
 390  * @stats: statistics entry to be updated
 391  **/
 392 static void i40e_get_netdev_stats_struct_tx(struct i40e_ring *ring,
 393                                             struct rtnl_link_stats64 *stats)
 394 {
 395         u64 bytes, packets;
 396         unsigned int start;
 397
 398         do {
 399                 start = u64_stats_fetch_begin_irq(&ring->syncp);
 400                 packets = ring->stats.packets;
 401                 bytes   = ring->stats.bytes;
 402         } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
 403
 404         stats->tx_packets += packets;
 405         stats->tx_bytes   += bytes;
 406 }
 407
 408 /**
 409  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 410  * @netdev: network interface device structure
 411  * @stats: data structure to store statistics
 412  *
 413  * Returns the address of the device statistics structure.
 414  * The statistics are actually updated from the service task.
 415  **/
 416 static void i40e_get_netdev_stats_struct(struct net_device *netdev,
 417                                   struct rtnl_link_stats64 *stats)
 418 {
 419         struct i40e_netdev_priv *np = netdev_priv(netdev);
 420         struct i40e_vsi *vsi = np->vsi;
 421         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 422         struct i40e_ring *ring;
 423         int i;
 424
 425         if (test_bit(__I40E_VSI_DOWN, vsi->state))
 426                 return;
 427
 428         if (!vsi->tx_rings)
 429                 return;
 430
 431         rcu_read_lock();
 432         for (i = 0; i < vsi->num_queue_pairs; i++) {
 433                 u64 bytes, packets;
 434                 unsigned int start;
 435
 436                 ring = READ_ONCE(vsi->tx_rings[i]);
 437                 if (!ring)
 438                         continue;
 439                 i40e_get_netdev_stats_struct_tx(ring, stats);
 440
 441                 if (i40e_enabled_xdp_vsi(vsi)) {
 442                         ring++;
 443                         i40e_get_netdev_stats_struct_tx(ring, stats);
 444                 }
 445
 446                 ring++;
 447                 do {
 448                         start   = u64_stats_fetch_begin_irq(&ring->syncp);
 449                         packets = ring->stats.packets;
 450                         bytes   = ring->stats.bytes;
 451                 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
 452
 453                 stats->rx_packets += packets;
 454                 stats->rx_bytes   += bytes;
 455
 456         }
 457         rcu_read_unlock();
 458
 459         /* following stats updated by i40e_watchdog_subtask() */
 460         stats->multicast        = vsi_stats->multicast;
 461         stats->tx_errors        = vsi_stats->tx_errors;
 462         stats->tx_dropped       = vsi_stats->tx_dropped;
 463         stats->rx_errors        = vsi_stats->rx_errors;
 464         stats->rx_dropped       = vsi_stats->rx_dropped;
 465         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 466         stats->rx_length_errors = vsi_stats->rx_length_errors;
 467 }
 468
 469 /**
 470  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 471  * @vsi: the VSI to have its stats reset
 472  **/
 473 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 474 {
 475         struct rtnl_link_stats64 *ns;
 476         int i;
 477
 478         if (!vsi)
 479                 return;
 480
 481         ns = i40e_get_vsi_stats_struct(vsi);
 482         memset(ns, 0, sizeof(*ns));
 483         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 484         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 485         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 486         if (vsi->rx_rings && vsi->rx_rings[0]) {
 487                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 488                         memset(&vsi->rx_rings[i]->stats, 0,
 489                                sizeof(vsi->rx_rings[i]->stats));
 490                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 491                                sizeof(vsi->rx_rings[i]->rx_stats));
 492                         memset(&vsi->tx_rings[i]->stats, 0,
 493                                sizeof(vsi->tx_rings[i]->stats));
 494                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 495                                sizeof(vsi->tx_rings[i]->tx_stats));
 496                 }
 497         }
 498         vsi->stat_offsets_loaded = false;
 499 }
 500
 501 /**
 502  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 503  * @pf: the PF to be reset
 504  **/
 505 void i40e_pf_reset_stats(struct i40e_pf *pf)
 506 {
 507         int i;
 508
 509         memset(&pf->stats, 0, sizeof(pf->stats));
 510         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 511         pf->stat_offsets_loaded = false;
 512
 513         for (i = 0; i < I40E_MAX_VEB; i++) {
 514                 if (pf->veb[i]) {
 515                         memset(&pf->veb[i]->stats, 0,
 516                                sizeof(pf->veb[i]->stats));
 517                         memset(&pf->veb[i]->stats_offsets, 0,
 518                                sizeof(pf->veb[i]->stats_offsets));
 519                         memset(&pf->veb[i]->tc_stats, 0,
 520                                sizeof(pf->veb[i]->tc_stats));
 521                         memset(&pf->veb[i]->tc_stats_offsets, 0,
 522                                sizeof(pf->veb[i]->tc_stats_offsets));
 523                         pf->veb[i]->stat_offsets_loaded = false;
 524                 }
 525         }
 526         pf->hw_csum_rx_error = 0;
 527 }
 528
 529 /**
 530  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 531  * @hw: ptr to the hardware info
 532  * @hireg: the high 32 bit reg to read
 533  * @loreg: the low 32 bit reg to read
 534  * @offset_loaded: has the initial offset been loaded yet
 535  * @offset: ptr to current offset value
 536  * @stat: ptr to the stat
 537  *
 538  * Since the device stats are not reset at PFReset, they likely will not
 539  * be zeroed when the driver starts.  We'll save the first values read
 540  * and use them as offsets to be subtracted from the raw values in order
 541  * to report stats that count from zero.  In the process, we also manage
 542  * the potential roll-over.
 543  **/
 544 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 545                                bool offset_loaded, u64 *offset, u64 *stat)
 546 {
 547         u64 new_data;
 548
 549         if (hw->device_id == I40E_DEV_ID_QEMU) {
 550                 new_data = rd32(hw, loreg);
 551                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 552         } else {
 553                 new_data = rd64(hw, loreg);
 554         }
 555         if (!offset_loaded)
 556                 *offset = new_data;
 557         if (likely(new_data >= *offset))
 558                 *stat = new_data - *offset;
 559         else
 560                 *stat = (new_data + BIT_ULL(48)) - *offset;
 561         *stat &= 0xFFFFFFFFFFFFULL;
 562 }
 563
 564 /**
 565  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 566  * @hw: ptr to the hardware info
 567  * @reg: the hw reg to read
 568  * @offset_loaded: has the initial offset been loaded yet
 569  * @offset: ptr to current offset value
 570  * @stat: ptr to the stat
 571  **/
 572 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 573                                bool offset_loaded, u64 *offset, u64 *stat)
 574 {
 575         u32 new_data;
 576
 577         new_data = rd32(hw, reg);
 578         if (!offset_loaded)
 579                 *offset = new_data;
 580         if (likely(new_data >= *offset))
 581                 *stat = (u32)(new_data - *offset);
 582         else
 583                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 584 }
 585
 586 /**
 587  * i40e_stat_update_and_clear32 - read and clear hw reg, update a 32 bit stat
 588  * @hw: ptr to the hardware info
 589  * @reg: the hw reg to read and clear
 590  * @stat: ptr to the stat
 591  **/
 592 static void i40e_stat_update_and_clear32(struct i40e_hw *hw, u32 reg, u64 *stat)
 593 {
 594         u32 new_data = rd32(hw, reg);
 595
 596         wr32(hw, reg, 1); /* must write a nonzero value to clear register */
 597         *stat += new_data;
 598 }
 599
 600 /**
 601  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 602  * @vsi: the VSI to be updated
 603  **/
 604 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 605 {
 606         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 607         struct i40e_pf *pf = vsi->back;
 608         struct i40e_hw *hw = &pf->hw;
 609         struct i40e_eth_stats *oes;
 610         struct i40e_eth_stats *es;     /* device's eth stats */
 611
 612         es = &vsi->eth_stats;
 613         oes = &vsi->eth_stats_offsets;
 614
 615         /* Gather up the stats that the hw collects */
 616         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 617                            vsi->stat_offsets_loaded,
 618                            &oes->tx_errors, &es->tx_errors);
 619         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 620                            vsi->stat_offsets_loaded,
 621                            &oes->rx_discards, &es->rx_discards);
 622         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 623                            vsi->stat_offsets_loaded,
 624                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 625
 626         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 627                            I40E_GLV_GORCL(stat_idx),
 628                            vsi->stat_offsets_loaded,
 629                            &oes->rx_bytes, &es->rx_bytes);
 630         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 631                            I40E_GLV_UPRCL(stat_idx),
 632                            vsi->stat_offsets_loaded,
 633                            &oes->rx_unicast, &es->rx_unicast);
 634         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 635                            I40E_GLV_MPRCL(stat_idx),
 636                            vsi->stat_offsets_loaded,
 637                            &oes->rx_multicast, &es->rx_multicast);
 638         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 639                            I40E_GLV_BPRCL(stat_idx),
 640                            vsi->stat_offsets_loaded,
 641                            &oes->rx_broadcast, &es->rx_broadcast);
 642
 643         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 644                            I40E_GLV_GOTCL(stat_idx),
 645                            vsi->stat_offsets_loaded,
 646                            &oes->tx_bytes, &es->tx_bytes);
 647         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 648                            I40E_GLV_UPTCL(stat_idx),
 649                            vsi->stat_offsets_loaded,
 650                            &oes->tx_unicast, &es->tx_unicast);
 651         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 652                            I40E_GLV_MPTCL(stat_idx),
 653                            vsi->stat_offsets_loaded,
 654                            &oes->tx_multicast, &es->tx_multicast);
 655         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 656                            I40E_GLV_BPTCL(stat_idx),
 657                            vsi->stat_offsets_loaded,
 658                            &oes->tx_broadcast, &es->tx_broadcast);
 659         vsi->stat_offsets_loaded = true;
 660 }
 661
 662 /**
 663  * i40e_update_veb_stats - Update Switch component statistics
 664  * @veb: the VEB being updated
 665  **/
 666 void i40e_update_veb_stats(struct i40e_veb *veb)
 667 {
 668         struct i40e_pf *pf = veb->pf;
 669         struct i40e_hw *hw = &pf->hw;
 670         struct i40e_eth_stats *oes;
 671         struct i40e_eth_stats *es;     /* device's eth stats */
 672         struct i40e_veb_tc_stats *veb_oes;
 673         struct i40e_veb_tc_stats *veb_es;
 674         int i, idx = 0;
 675
 676         idx = veb->stats_idx;
 677         es = &veb->stats;
 678         oes = &veb->stats_offsets;
 679         veb_es = &veb->tc_stats;
 680         veb_oes = &veb->tc_stats_offsets;
 681
 682         /* Gather up the stats that the hw collects */
 683         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 684                            veb->stat_offsets_loaded,
 685                            &oes->tx_discards, &es->tx_discards);
 686         if (hw->revision_id > 0)
 687                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 688                                    veb->stat_offsets_loaded,
 689                                    &oes->rx_unknown_protocol,
 690                                    &es->rx_unknown_protocol);
 691         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 692                            veb->stat_offsets_loaded,
 693                            &oes->rx_bytes, &es->rx_bytes);
 694         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 695                            veb->stat_offsets_loaded,
 696                            &oes->rx_unicast, &es->rx_unicast);
 697         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 698                            veb->stat_offsets_loaded,
 699                            &oes->rx_multicast, &es->rx_multicast);
 700         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 701                            veb->stat_offsets_loaded,
 702                            &oes->rx_broadcast, &es->rx_broadcast);
 703
 704         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 705                            veb->stat_offsets_loaded,
 706                            &oes->tx_bytes, &es->tx_bytes);
 707         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 708                            veb->stat_offsets_loaded,
 709                            &oes->tx_unicast, &es->tx_unicast);
 710         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 711                            veb->stat_offsets_loaded,
 712                            &oes->tx_multicast, &es->tx_multicast);
 713         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 714                            veb->stat_offsets_loaded,
 715                            &oes->tx_broadcast, &es->tx_broadcast);
 716         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 717                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 718                                    I40E_GLVEBTC_RPCL(i, idx),
 719                                    veb->stat_offsets_loaded,
 720                                    &veb_oes->tc_rx_packets[i],
 721                                    &veb_es->tc_rx_packets[i]);
 722                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 723                                    I40E_GLVEBTC_RBCL(i, idx),
 724                                    veb->stat_offsets_loaded,
 725                                    &veb_oes->tc_rx_bytes[i],
 726                                    &veb_es->tc_rx_bytes[i]);
 727                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 728                                    I40E_GLVEBTC_TPCL(i, idx),
 729                                    veb->stat_offsets_loaded,
 730                                    &veb_oes->tc_tx_packets[i],
 731                                    &veb_es->tc_tx_packets[i]);
 732                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 733                                    I40E_GLVEBTC_TBCL(i, idx),
 734                                    veb->stat_offsets_loaded,
 735                                    &veb_oes->tc_tx_bytes[i],
 736                                    &veb_es->tc_tx_bytes[i]);
 737         }
 738         veb->stat_offsets_loaded = true;
 739 }
 740
 741 /**
 742  * i40e_update_vsi_stats - Update the vsi statistics counters.
 743  * @vsi: the VSI to be updated
 744  *
 745  * There are a few instances where we store the same stat in a
 746  * couple of different structs.  This is partly because we have
 747  * the netdev stats that need to be filled out, which is slightly
 748  * different from the "eth_stats" defined by the chip and used in
 749  * VF communications.  We sort it out here.
 750  **/
 751 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 752 {
 753         struct i40e_pf *pf = vsi->back;
 754         struct rtnl_link_stats64 *ons;
 755         struct rtnl_link_stats64 *ns;   /* netdev stats */
 756         struct i40e_eth_stats *oes;
 757         struct i40e_eth_stats *es;     /* device's eth stats */
 758         u32 tx_restart, tx_busy;
 759         struct i40e_ring *p;
 760         u32 rx_page, rx_buf;
 761         u64 bytes, packets;
 762         unsigned int start;
 763         u64 tx_linearize;
 764         u64 tx_force_wb;
 765         u64 rx_p, rx_b;
 766         u64 tx_p, tx_b;
 767         u16 q;
 768
 769         if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
 770             test_bit(__I40E_CONFIG_BUSY, pf->state))
 771                 return;
 772
 773         ns = i40e_get_vsi_stats_struct(vsi);
 774         ons = &vsi->net_stats_offsets;
 775         es = &vsi->eth_stats;
 776         oes = &vsi->eth_stats_offsets;
 777
 778         /* Gather up the netdev and vsi stats that the driver collects
 779          * on the fly during packet processing
 780          */
 781         rx_b = rx_p = 0;
 782         tx_b = tx_p = 0;
 783         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 784         rx_page = 0;
 785         rx_buf = 0;
 786         rcu_read_lock();
 787         for (q = 0; q < vsi->num_queue_pairs; q++) {
 788                 /* locate Tx ring */
 789                 p = READ_ONCE(vsi->tx_rings[q]);
 790
 791                 do {
 792                         start = u64_stats_fetch_begin_irq(&p->syncp);
 793                         packets = p->stats.packets;
 794                         bytes = p->stats.bytes;
 795                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 796                 tx_b += bytes;
 797                 tx_p += packets;
 798                 tx_restart += p->tx_stats.restart_queue;
 799                 tx_busy += p->tx_stats.tx_busy;
 800                 tx_linearize += p->tx_stats.tx_linearize;
 801                 tx_force_wb += p->tx_stats.tx_force_wb;
 802
 803                 /* Rx queue is part of the same block as Tx queue */
 804                 p = &p[1];
 805                 do {
 806                         start = u64_stats_fetch_begin_irq(&p->syncp);
 807                         packets = p->stats.packets;
 808                         bytes = p->stats.bytes;
 809                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 810                 rx_b += bytes;
 811                 rx_p += packets;
 812                 rx_buf += p->rx_stats.alloc_buff_failed;
 813                 rx_page += p->rx_stats.alloc_page_failed;
 814         }
 815         rcu_read_unlock();
 816         vsi->tx_restart = tx_restart;
 817         vsi->tx_busy = tx_busy;
 818         vsi->tx_linearize = tx_linearize;
 819         vsi->tx_force_wb = tx_force_wb;
 820         vsi->rx_page_failed = rx_page;
 821         vsi->rx_buf_failed = rx_buf;
 822
 823         ns->rx_packets = rx_p;
 824         ns->rx_bytes = rx_b;
 825         ns->tx_packets = tx_p;
 826         ns->tx_bytes = tx_b;
 827
 828         /* update netdev stats from eth stats */
 829         i40e_update_eth_stats(vsi);
 830         ons->tx_errors = oes->tx_errors;
 831         ns->tx_errors = es->tx_errors;
 832         ons->multicast = oes->rx_multicast;
 833         ns->multicast = es->rx_multicast;
 834         ons->rx_dropped = oes->rx_discards;
 835         ns->rx_dropped = es->rx_discards;
 836         ons->tx_dropped = oes->tx_discards;
 837         ns->tx_dropped = es->tx_discards;
 838
 839         /* pull in a couple PF stats if this is the main vsi */
 840         if (vsi == pf->vsi[pf->lan_vsi]) {
 841                 ns->rx_crc_errors = pf->stats.crc_errors;
 842                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 843                 ns->rx_length_errors = pf->stats.rx_length_errors;
 844         }
 845 }
 846
 847 /**
 848  * i40e_update_pf_stats - Update the PF statistics counters.
 849  * @pf: the PF to be updated
 850  **/
 851 static void i40e_update_pf_stats(struct i40e_pf *pf)
 852 {
 853         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 854         struct i40e_hw_port_stats *nsd = &pf->stats;
 855         struct i40e_hw *hw = &pf->hw;
 856         u32 val;
 857         int i;
 858
 859         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 860                            I40E_GLPRT_GORCL(hw->port),
 861                            pf->stat_offsets_loaded,
 862                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 863         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 864                            I40E_GLPRT_GOTCL(hw->port),
 865                            pf->stat_offsets_loaded,
 866                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 867         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 868                            pf->stat_offsets_loaded,
 869                            &osd->eth.rx_discards,
 870                            &nsd->eth.rx_discards);
 871         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 872                            I40E_GLPRT_UPRCL(hw->port),
 873                            pf->stat_offsets_loaded,
 874                            &osd->eth.rx_unicast,
 875                            &nsd->eth.rx_unicast);
 876         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 877                            I40E_GLPRT_MPRCL(hw->port),
 878                            pf->stat_offsets_loaded,
 879                            &osd->eth.rx_multicast,
 880                            &nsd->eth.rx_multicast);
 881         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 882                            I40E_GLPRT_BPRCL(hw->port),
 883                            pf->stat_offsets_loaded,
 884                            &osd->eth.rx_broadcast,
 885                            &nsd->eth.rx_broadcast);
 886         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 887                            I40E_GLPRT_UPTCL(hw->port),
 888                            pf->stat_offsets_loaded,
 889                            &osd->eth.tx_unicast,
 890                            &nsd->eth.tx_unicast);
 891         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 892                            I40E_GLPRT_MPTCL(hw->port),
 893                            pf->stat_offsets_loaded,
 894                            &osd->eth.tx_multicast,
 895                            &nsd->eth.tx_multicast);
 896         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 897                            I40E_GLPRT_BPTCL(hw->port),
 898                            pf->stat_offsets_loaded,
 899                            &osd->eth.tx_broadcast,
 900                            &nsd->eth.tx_broadcast);
 901
 902         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 903                            pf->stat_offsets_loaded,
 904                            &osd->tx_dropped_link_down,
 905                            &nsd->tx_dropped_link_down);
 906
 907         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 908                            pf->stat_offsets_loaded,
 909                            &osd->crc_errors, &nsd->crc_errors);
 910
 911         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 912                            pf->stat_offsets_loaded,
 913                            &osd->illegal_bytes, &nsd->illegal_bytes);
 914
 915         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 916                            pf->stat_offsets_loaded,
 917                            &osd->mac_local_faults,
 918                            &nsd->mac_local_faults);
 919         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 920                            pf->stat_offsets_loaded,
 921                            &osd->mac_remote_faults,
 922                            &nsd->mac_remote_faults);
 923
 924         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 925                            pf->stat_offsets_loaded,
 926                            &osd->rx_length_errors,
 927                            &nsd->rx_length_errors);
 928
 929         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 930                            pf->stat_offsets_loaded,
 931                            &osd->link_xon_rx, &nsd->link_xon_rx);
 932         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 933                            pf->stat_offsets_loaded,
 934                            &osd->link_xon_tx, &nsd->link_xon_tx);
 935         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 936                            pf->stat_offsets_loaded,
 937                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 938         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 939                            pf->stat_offsets_loaded,
 940                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 941
 942         for (i = 0; i < 8; i++) {
 943                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 944                                    pf->stat_offsets_loaded,
 945                                    &osd->priority_xoff_rx[i],
 946                                    &nsd->priority_xoff_rx[i]);
 947                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 948                                    pf->stat_offsets_loaded,
 949                                    &osd->priority_xon_rx[i],
 950                                    &nsd->priority_xon_rx[i]);
 951                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
 952                                    pf->stat_offsets_loaded,
 953                                    &osd->priority_xon_tx[i],
 954                                    &nsd->priority_xon_tx[i]);
 955                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
 956                                    pf->stat_offsets_loaded,
 957                                    &osd->priority_xoff_tx[i],
 958                                    &nsd->priority_xoff_tx[i]);
 959                 i40e_stat_update32(hw,
 960                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
 961                                    pf->stat_offsets_loaded,
 962                                    &osd->priority_xon_2_xoff[i],
 963                                    &nsd->priority_xon_2_xoff[i]);
 964         }
 965
 966         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
 967                            I40E_GLPRT_PRC64L(hw->port),
 968                            pf->stat_offsets_loaded,
 969                            &osd->rx_size_64, &nsd->rx_size_64);
 970         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
 971                            I40E_GLPRT_PRC127L(hw->port),
 972                            pf->stat_offsets_loaded,
 973                            &osd->rx_size_127, &nsd->rx_size_127);
 974         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
 975                            I40E_GLPRT_PRC255L(hw->port),
 976                            pf->stat_offsets_loaded,
 977                            &osd->rx_size_255, &nsd->rx_size_255);
 978         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
 979                            I40E_GLPRT_PRC511L(hw->port),
 980                            pf->stat_offsets_loaded,
 981                            &osd->rx_size_511, &nsd->rx_size_511);
 982         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
 983                            I40E_GLPRT_PRC1023L(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->rx_size_1023, &nsd->rx_size_1023);
 986         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
 987                            I40E_GLPRT_PRC1522L(hw->port),
 988                            pf->stat_offsets_loaded,
 989                            &osd->rx_size_1522, &nsd->rx_size_1522);
 990         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
 991                            I40E_GLPRT_PRC9522L(hw->port),
 992                            pf->stat_offsets_loaded,
 993                            &osd->rx_size_big, &nsd->rx_size_big);
 994
 995         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
 996                            I40E_GLPRT_PTC64L(hw->port),
 997                            pf->stat_offsets_loaded,
 998                            &osd->tx_size_64, &nsd->tx_size_64);
 999         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1000                            I40E_GLPRT_PTC127L(hw->port),
1001                            pf->stat_offsets_loaded,
1002                            &osd->tx_size_127, &nsd->tx_size_127);
1003         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1004                            I40E_GLPRT_PTC255L(hw->port),
1005                            pf->stat_offsets_loaded,
1006                            &osd->tx_size_255, &nsd->tx_size_255);
1007         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1008                            I40E_GLPRT_PTC511L(hw->port),
1009                            pf->stat_offsets_loaded,
1010                            &osd->tx_size_511, &nsd->tx_size_511);
1011         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1012                            I40E_GLPRT_PTC1023L(hw->port),
1013                            pf->stat_offsets_loaded,
1014                            &osd->tx_size_1023, &nsd->tx_size_1023);
1015         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1016                            I40E_GLPRT_PTC1522L(hw->port),
1017                            pf->stat_offsets_loaded,
1018                            &osd->tx_size_1522, &nsd->tx_size_1522);
1019         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1020                            I40E_GLPRT_PTC9522L(hw->port),
1021                            pf->stat_offsets_loaded,
1022                            &osd->tx_size_big, &nsd->tx_size_big);
1023
1024         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1025                            pf->stat_offsets_loaded,
1026                            &osd->rx_undersize, &nsd->rx_undersize);
1027         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1028                            pf->stat_offsets_loaded,
1029                            &osd->rx_fragments, &nsd->rx_fragments);
1030         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1031                            pf->stat_offsets_loaded,
1032                            &osd->rx_oversize, &nsd->rx_oversize);
1033         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1034                            pf->stat_offsets_loaded,
1035                            &osd->rx_jabber, &nsd->rx_jabber);
1036
1037         /* FDIR stats */
1038         i40e_stat_update_and_clear32(hw,
1039                         I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(hw->pf_id)),
1040                         &nsd->fd_atr_match);
1041         i40e_stat_update_and_clear32(hw,
1042                         I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(hw->pf_id)),
1043                         &nsd->fd_sb_match);
1044         i40e_stat_update_and_clear32(hw,
1045                         I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(hw->pf_id)),
1046                         &nsd->fd_atr_tunnel_match);
1047
1048         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1049         nsd->tx_lpi_status =
1050                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1051                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1052         nsd->rx_lpi_status =
1053                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1054                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1055         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1056                            pf->stat_offsets_loaded,
1057                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1058         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1059                            pf->stat_offsets_loaded,
1060                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1061
1062         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1063             !test_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state))
1064                 nsd->fd_sb_status = true;
1065         else
1066                 nsd->fd_sb_status = false;
1067
1068         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1069             !test_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state))
1070                 nsd->fd_atr_status = true;
1071         else
1072                 nsd->fd_atr_status = false;
1073
1074         pf->stat_offsets_loaded = true;
1075 }
1076
1077 /**
1078  * i40e_update_stats - Update the various statistics counters.
1079  * @vsi: the VSI to be updated
1080  *
1081  * Update the various stats for this VSI and its related entities.
1082  **/
1083 void i40e_update_stats(struct i40e_vsi *vsi)
1084 {
1085         struct i40e_pf *pf = vsi->back;
1086
1087         if (vsi == pf->vsi[pf->lan_vsi])
1088                 i40e_update_pf_stats(pf);
1089
1090         i40e_update_vsi_stats(vsi);
1091 }
1092
1093 /**
1094  * i40e_count_filters - counts VSI mac filters
1095  * @vsi: the VSI to be searched
1096  *
1097  * Returns count of mac filters
1098  **/
1099 int i40e_count_filters(struct i40e_vsi *vsi)
1100 {
1101         struct i40e_mac_filter *f;
1102         struct hlist_node *h;
1103         int bkt;
1104         int cnt = 0;
1105
1106         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
1107                 ++cnt;
1108
1109         return cnt;
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  **/
1120 static 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  **/
1146 struct 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  **/
1168 bool 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  */
1226 static 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  **/
1315 static 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  **/
1351 struct 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  **/
1413 void __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  **/
1446 void 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  **/
1469 struct 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  **/
1504 int 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  **/
1532 static 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  **/
1594 static 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;
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  **/
1632 static 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  **/
1674 static 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  **/
1762 static 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  */
1897 static 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  */
1916 static 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  **/
1938 static 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  **/
1965 static 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  **/
1988 static 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  */
2009 static
2010 struct 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  **/
2030 static int
2031 i40e_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  */
2073 static
2074 void 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  */
2107 static
2108 void 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  **/
2152 static i40e_status
2153 i40e_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  **/
2193 static 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  **/
2259 int 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         }
2562 out:
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
2570 err_no_memory:
2571         /* Restore elements on the temporary add and delete lists */
2572         spin_lock_bh(&vsi->mac_filter_hash_lock);
2573 err_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  **/
2587 static 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  **/
2620 static 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  **/
2635 static 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_dbg(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  **/
2664 int 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  **/
2683 void 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  **/
2717 void 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  **/
2761 int 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  **/
2787 int 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  */
2832 void 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  **/
2849 void 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  **/
2872 static 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  **/
2895 static 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  **/
2914 static 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  **/
2935 static 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  **/
2957 int 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  **/
2989 void 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  **/
3006 static 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  **/
3028 static 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  **/
3055 static 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  **/
3070 static 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  **/
3089 static 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  **/
3111 static 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  **/
3131 static 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  **/
3244 static 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  **/
3369 static 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  **/
3392 static 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  **/
3423 static 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  **/
3459 static 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  **/
3472 static 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  **/
3497 static 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  **/
3515 static 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 >> 1);
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 >> 1);
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  **/
3600 static 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  **/
3638 static 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 >> 1);
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 >> 1);
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  **/
3690 void 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  **/
3703 void 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  **/
3721 static 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  **/
3741 static 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  **/
3758 static 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  **/
3767 static 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
3825 free_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  **/
3840 static 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  **/
3886 static 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  **/
3906 static 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  **/
3928 static 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
4039 enable_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  **/
4058 static 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  **/
4141 static 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  **/
4161 static 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  **/
4197 static 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  **/
4238 static 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  **/
4266 static 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  **/
4299 static 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  **/
4327 static 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  **/
4369 int 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  **/
4393 static 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  **/
4429 static 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  **/
4457 static 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  **/
4494 int 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  **/
4513 static 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  **/
4542 int 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  **/
4559 void 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  **/
4583 void 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  **/
4599 static 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  **/
4724 static 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  **/
4755 static 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  **/
4767 static 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  **/
4789 static 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  **/
4809 static 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  **/
4828 static 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  **/
4847 static 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  **/
4865 static 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  **/
4881 static 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  **/
4896 static 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  **/
4910 static 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  **/
4926 int 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                 }
4954 wait_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  **/
4976 static 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  **/
5000 static 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  **/
5028 static 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  **/
5071 static 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  **/
5090 static 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  **/
5107 static 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  **/
5144 static 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  **/
5172 static 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  **/
5234 static 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  **/
5277 static 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  **/
5332 static 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  **/
5359 static 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);
5478 out:
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  **/
5487 static 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  **/
5515 int 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  **/
5554 static 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  **/
5642 static 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  **/
5661 static 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  **/
5687 static 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  **/
5737 static 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  **/
5796 static 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  **/
5835 static 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
5899 static 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  **/
5934 static 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  **/
5988 static 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  **/
6040 static 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  **/
6075 static 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  **/
6133 int 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  **/
6247 static 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
6290 err_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  **/
6302 int 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
6341 out:
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  **/
6354 static 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  **/
6409 static 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  **/
6435 static 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
6483 out:
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  */
6494 void 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  **/
6612 static 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  **/
6662 static 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  **/
6679 int 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  **/
6695 static 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  **/
6786 void 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 and
6808                          * ndo_xsk_wakeup 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  **/
6823 static 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  **/
6868 static 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  **/
6901 static 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  **/
6927 static 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  */
6951 static 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  */
6974 static 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  */
7019 static 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  */
7093 static 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                         kfree(ch);
7172                         goto err_free;
7173                 }
7174                 ch->parent_vsi = vsi;
7175                 vsi->cnt_q_avail -= ch->num_queue_pairs;
7176                 vsi->macvlan_cnt++;
7177                 list_add_tail(&ch->list, &vsi->macvlan_list);
7178         }
7179
7180         return ret;
7181
7182 err_free:
7183         dev_info(&pf->pdev->dev, "Failed to setup macvlans\n");
7184         i40e_free_macvlan_channels(vsi);
7185
7186         return ret;
7187 }
7188
7189 /**
7190  * i40e_fwd_add - configure macvlans
7191  * @netdev: net device to configure
7192  * @vdev: macvlan netdevice
7193  **/
7194 static void *i40e_fwd_add(struct net_device *netdev, struct net_device *vdev)
7195 {
7196         struct i40e_netdev_priv *np = netdev_priv(netdev);
7197         u16 q_per_macvlan = 0, macvlan_cnt = 0, vectors;
7198         struct i40e_vsi *vsi = np->vsi;
7199         struct i40e_pf *pf = vsi->back;
7200         struct i40e_fwd_adapter *fwd;
7201         int avail_macvlan, ret;
7202
7203         if ((pf->flags & I40E_FLAG_DCB_ENABLED)) {
7204                 netdev_info(netdev, "Macvlans are not supported when DCB is enabled\n");
7205                 return ERR_PTR(-EINVAL);
7206         }
7207         if ((pf->flags & I40E_FLAG_TC_MQPRIO)) {
7208                 netdev_info(netdev, "Macvlans are not supported when HW TC offload is on\n");
7209                 return ERR_PTR(-EINVAL);
7210         }
7211         if (pf->num_lan_msix < I40E_MIN_MACVLAN_VECTORS) {
7212                 netdev_info(netdev, "Not enough vectors available to support macvlans\n");
7213                 return ERR_PTR(-EINVAL);
7214         }
7215
7216         /* The macvlan device has to be a single Q device so that the
7217          * tc_to_txq field can be reused to pick the tx queue.
7218          */
7219         if (netif_is_multiqueue(vdev))
7220                 return ERR_PTR(-ERANGE);
7221
7222         if (!vsi->macvlan_cnt) {
7223                 /* reserve bit 0 for the pf device */
7224                 set_bit(0, vsi->fwd_bitmask);
7225
7226                 /* Try to reserve as many queues as possible for macvlans. First
7227                  * reserve 3/4th of max vectors, then half, then quarter and
7228                  * calculate Qs per macvlan as you go
7229                  */
7230                 vectors = pf->num_lan_msix;
7231                 if (vectors <= I40E_MAX_MACVLANS && vectors > 64) {
7232                         /* allocate 4 Qs per macvlan and 32 Qs to the PF*/
7233                         q_per_macvlan = 4;
7234                         macvlan_cnt = (vectors - 32) / 4;
7235                 } else if (vectors <= 64 && vectors > 32) {
7236                         /* allocate 2 Qs per macvlan and 16 Qs to the PF*/
7237                         q_per_macvlan = 2;
7238                         macvlan_cnt = (vectors - 16) / 2;
7239                 } else if (vectors <= 32 && vectors > 16) {
7240                         /* allocate 1 Q per macvlan and 16 Qs to the PF*/
7241                         q_per_macvlan = 1;
7242                         macvlan_cnt = vectors - 16;
7243                 } else if (vectors <= 16 && vectors > 8) {
7244                         /* allocate 1 Q per macvlan and 8 Qs to the PF */
7245                         q_per_macvlan = 1;
7246                         macvlan_cnt = vectors - 8;
7247                 } else {
7248                         /* allocate 1 Q per macvlan and 1 Q to the PF */
7249                         q_per_macvlan = 1;
7250                         macvlan_cnt = vectors - 1;
7251                 }
7252
7253                 if (macvlan_cnt == 0)
7254                         return ERR_PTR(-EBUSY);
7255
7256                 /* Quiesce VSI queues */
7257                 i40e_quiesce_vsi(vsi);
7258
7259                 /* sets up the macvlans but does not "enable" them */
7260                 ret = i40e_setup_macvlans(vsi, macvlan_cnt, q_per_macvlan,
7261                                           vdev);
7262                 if (ret)
7263                         return ERR_PTR(ret);
7264
7265                 /* Unquiesce VSI */
7266                 i40e_unquiesce_vsi(vsi);
7267         }
7268         avail_macvlan = find_first_zero_bit(vsi->fwd_bitmask,
7269                                             vsi->macvlan_cnt);
7270         if (avail_macvlan >= I40E_MAX_MACVLANS)
7271                 return ERR_PTR(-EBUSY);
7272
7273         /* create the fwd struct */
7274         fwd = kzalloc(sizeof(*fwd), GFP_KERNEL);
7275         if (!fwd)
7276                 return ERR_PTR(-ENOMEM);
7277
7278         set_bit(avail_macvlan, vsi->fwd_bitmask);
7279         fwd->bit_no = avail_macvlan;
7280         netdev_set_sb_channel(vdev, avail_macvlan);
7281         fwd->netdev = vdev;
7282
7283         if (!netif_running(netdev))
7284                 return fwd;
7285
7286         /* Set fwd ring up */
7287         ret = i40e_fwd_ring_up(vsi, vdev, fwd);
7288         if (ret) {
7289                 /* unbind the queues and drop the subordinate channel config */
7290                 netdev_unbind_sb_channel(netdev, vdev);
7291                 netdev_set_sb_channel(vdev, 0);
7292
7293                 kfree(fwd);
7294                 return ERR_PTR(-EINVAL);
7295         }
7296
7297         return fwd;
7298 }
7299
7300 /**
7301  * i40e_del_all_macvlans - Delete all the mac filters on the channels
7302  * @vsi: the VSI we want to access
7303  */
7304 static void i40e_del_all_macvlans(struct i40e_vsi *vsi)
7305 {
7306         struct i40e_channel *ch, *ch_tmp;
7307         struct i40e_pf *pf = vsi->back;
7308         struct i40e_hw *hw = &pf->hw;
7309         int aq_err, ret = 0;
7310
7311         if (list_empty(&vsi->macvlan_list))
7312                 return;
7313
7314         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7315                 if (i40e_is_channel_macvlan(ch)) {
7316                         ret = i40e_del_macvlan_filter(hw, ch->seid,
7317                                                       i40e_channel_mac(ch),
7318                                                       &aq_err);
7319                         if (!ret) {
7320                                 /* Reset queue contexts */
7321                                 i40e_reset_ch_rings(vsi, ch);
7322                                 clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7323                                 netdev_unbind_sb_channel(vsi->netdev,
7324                                                          ch->fwd->netdev);
7325                                 netdev_set_sb_channel(ch->fwd->netdev, 0);
7326                                 kfree(ch->fwd);
7327                                 ch->fwd = NULL;
7328                         }
7329                 }
7330         }
7331 }
7332
7333 /**
7334  * i40e_fwd_del - delete macvlan interfaces
7335  * @netdev: net device to configure
7336  * @vdev: macvlan netdevice
7337  */
7338 static void i40e_fwd_del(struct net_device *netdev, void *vdev)
7339 {
7340         struct i40e_netdev_priv *np = netdev_priv(netdev);
7341         struct i40e_fwd_adapter *fwd = vdev;
7342         struct i40e_channel *ch, *ch_tmp;
7343         struct i40e_vsi *vsi = np->vsi;
7344         struct i40e_pf *pf = vsi->back;
7345         struct i40e_hw *hw = &pf->hw;
7346         int aq_err, ret = 0;
7347
7348         /* Find the channel associated with the macvlan and del mac filter */
7349         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7350                 if (i40e_is_channel_macvlan(ch) &&
7351                     ether_addr_equal(i40e_channel_mac(ch),
7352                                      fwd->netdev->dev_addr)) {
7353                         ret = i40e_del_macvlan_filter(hw, ch->seid,
7354                                                       i40e_channel_mac(ch),
7355                                                       &aq_err);
7356                         if (!ret) {
7357                                 /* Reset queue contexts */
7358                                 i40e_reset_ch_rings(vsi, ch);
7359                                 clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7360                                 netdev_unbind_sb_channel(netdev, fwd->netdev);
7361                                 netdev_set_sb_channel(fwd->netdev, 0);
7362                                 kfree(ch->fwd);
7363                                 ch->fwd = NULL;
7364                         } else {
7365                                 dev_info(&pf->pdev->dev,
7366                                          "Error deleting mac filter on macvlan err %s, aq_err %s\n",
7367                                           i40e_stat_str(hw, ret),
7368                                           i40e_aq_str(hw, aq_err));
7369                         }
7370                         break;
7371                 }
7372         }
7373 }
7374
7375 /**
7376  * i40e_setup_tc - configure multiple traffic classes
7377  * @netdev: net device to configure
7378  * @type_data: tc offload data
7379  **/
7380 static int i40e_setup_tc(struct net_device *netdev, void *type_data)
7381 {
7382         struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
7383         struct i40e_netdev_priv *np = netdev_priv(netdev);
7384         struct i40e_vsi *vsi = np->vsi;
7385         struct i40e_pf *pf = vsi->back;
7386         u8 enabled_tc = 0, num_tc, hw;
7387         bool need_reset = false;
7388         int old_queue_pairs;
7389         int ret = -EINVAL;
7390         u16 mode;
7391         int i;
7392
7393         old_queue_pairs = vsi->num_queue_pairs;
7394         num_tc = mqprio_qopt->qopt.num_tc;
7395         hw = mqprio_qopt->qopt.hw;
7396         mode = mqprio_qopt->mode;
7397         if (!hw) {
7398                 pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7399                 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
7400                 goto config_tc;
7401         }
7402
7403         /* Check if MFP enabled */
7404         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
7405                 netdev_info(netdev,
7406                             "Configuring TC not supported in MFP mode\n");
7407                 return ret;
7408         }
7409         switch (mode) {
7410         case TC_MQPRIO_MODE_DCB:
7411                 pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7412
7413                 /* Check if DCB enabled to continue */
7414                 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
7415                         netdev_info(netdev,
7416                                     "DCB is not enabled for adapter\n");
7417                         return ret;
7418                 }
7419
7420                 /* Check whether tc count is within enabled limit */
7421                 if (num_tc > i40e_pf_get_num_tc(pf)) {
7422                         netdev_info(netdev,
7423                                     "TC count greater than enabled on link for adapter\n");
7424                         return ret;
7425                 }
7426                 break;
7427         case TC_MQPRIO_MODE_CHANNEL:
7428                 if (pf->flags & I40E_FLAG_DCB_ENABLED) {
7429                         netdev_info(netdev,
7430                                     "Full offload of TC Mqprio options is not supported when DCB is enabled\n");
7431                         return ret;
7432                 }
7433                 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7434                         return ret;
7435                 ret = i40e_validate_mqprio_qopt(vsi, mqprio_qopt);
7436                 if (ret)
7437                         return ret;
7438                 memcpy(&vsi->mqprio_qopt, mqprio_qopt,
7439                        sizeof(*mqprio_qopt));
7440                 pf->flags |= I40E_FLAG_TC_MQPRIO;
7441                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
7442                 break;
7443         default:
7444                 return -EINVAL;
7445         }
7446
7447 config_tc:
7448         /* Generate TC map for number of tc requested */
7449         for (i = 0; i < num_tc; i++)
7450                 enabled_tc |= BIT(i);
7451
7452         /* Requesting same TC configuration as already enabled */
7453         if (enabled_tc == vsi->tc_config.enabled_tc &&
7454             mode != TC_MQPRIO_MODE_CHANNEL)
7455                 return 0;
7456
7457         /* Quiesce VSI queues */
7458         i40e_quiesce_vsi(vsi);
7459
7460         if (!hw && !(pf->flags & I40E_FLAG_TC_MQPRIO))
7461                 i40e_remove_queue_channels(vsi);
7462
7463         /* Configure VSI for enabled TCs */
7464         ret = i40e_vsi_config_tc(vsi, enabled_tc);
7465         if (ret) {
7466                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
7467                             vsi->seid);
7468                 need_reset = true;
7469                 goto exit;
7470         } else {
7471                 dev_info(&vsi->back->pdev->dev,
7472                          "Setup channel (id:%u) utilizing num_queues %d\n",
7473                          vsi->seid, vsi->tc_config.tc_info[0].qcount);
7474         }
7475
7476         if (pf->flags & I40E_FLAG_TC_MQPRIO) {
7477                 if (vsi->mqprio_qopt.max_rate[0]) {
7478                         u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
7479
7480                         do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
7481                         ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
7482                         if (!ret) {
7483                                 u64 credits = max_tx_rate;
7484
7485                                 do_div(credits, I40E_BW_CREDIT_DIVISOR);
7486                                 dev_dbg(&vsi->back->pdev->dev,
7487                                         "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
7488                                         max_tx_rate,
7489                                         credits,
7490                                         vsi->seid);
7491                         } else {
7492                                 need_reset = true;
7493                                 goto exit;
7494                         }
7495                 }
7496                 ret = i40e_configure_queue_channels(vsi);
7497                 if (ret) {
7498                         vsi->num_queue_pairs = old_queue_pairs;
7499                         netdev_info(netdev,
7500                                     "Failed configuring queue channels\n");
7501                         need_reset = true;
7502                         goto exit;
7503                 }
7504         }
7505
7506 exit:
7507         /* Reset the configuration data to defaults, only TC0 is enabled */
7508         if (need_reset) {
7509                 i40e_vsi_set_default_tc_config(vsi);
7510                 need_reset = false;
7511         }
7512
7513         /* Unquiesce VSI */
7514         i40e_unquiesce_vsi(vsi);
7515         return ret;
7516 }
7517
7518 /**
7519  * i40e_set_cld_element - sets cloud filter element data
7520  * @filter: cloud filter rule
7521  * @cld: ptr to cloud filter element data
7522  *
7523  * This is helper function to copy data into cloud filter element
7524  **/
7525 static inline void
7526 i40e_set_cld_element(struct i40e_cloud_filter *filter,
7527                      struct i40e_aqc_cloud_filters_element_data *cld)
7528 {
7529         int i, j;
7530         u32 ipa;
7531
7532         memset(cld, 0, sizeof(*cld));
7533         ether_addr_copy(cld->outer_mac, filter->dst_mac);
7534         ether_addr_copy(cld->inner_mac, filter->src_mac);
7535
7536         if (filter->n_proto != ETH_P_IP && filter->n_proto != ETH_P_IPV6)
7537                 return;
7538
7539         if (filter->n_proto == ETH_P_IPV6) {
7540 #define IPV6_MAX_INDEX  (ARRAY_SIZE(filter->dst_ipv6) - 1)
7541                 for (i = 0, j = 0; i < ARRAY_SIZE(filter->dst_ipv6);
7542                      i++, j += 2) {
7543                         ipa = be32_to_cpu(filter->dst_ipv6[IPV6_MAX_INDEX - i]);
7544                         ipa = cpu_to_le32(ipa);
7545                         memcpy(&cld->ipaddr.raw_v6.data[j], &ipa, sizeof(ipa));
7546                 }
7547         } else {
7548                 ipa = be32_to_cpu(filter->dst_ipv4);
7549                 memcpy(&cld->ipaddr.v4.data, &ipa, sizeof(ipa));
7550         }
7551
7552         cld->inner_vlan = cpu_to_le16(ntohs(filter->vlan_id));
7553
7554         /* tenant_id is not supported by FW now, once the support is enabled
7555          * fill the cld->tenant_id with cpu_to_le32(filter->tenant_id)
7556          */
7557         if (filter->tenant_id)
7558                 return;
7559 }
7560
7561 /**
7562  * i40e_add_del_cloud_filter - Add/del cloud filter
7563  * @vsi: pointer to VSI
7564  * @filter: cloud filter rule
7565  * @add: if true, add, if false, delete
7566  *
7567  * Add or delete a cloud filter for a specific flow spec.
7568  * Returns 0 if the filter were successfully added.
7569  **/
7570 int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
7571                               struct i40e_cloud_filter *filter, bool add)
7572 {
7573         struct i40e_aqc_cloud_filters_element_data cld_filter;
7574         struct i40e_pf *pf = vsi->back;
7575         int ret;
7576         static const u16 flag_table[128] = {
7577                 [I40E_CLOUD_FILTER_FLAGS_OMAC]  =
7578                         I40E_AQC_ADD_CLOUD_FILTER_OMAC,
7579                 [I40E_CLOUD_FILTER_FLAGS_IMAC]  =
7580                         I40E_AQC_ADD_CLOUD_FILTER_IMAC,
7581                 [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN]  =
7582                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN,
7583                 [I40E_CLOUD_FILTER_FLAGS_IMAC_TEN_ID] =
7584                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_TEN_ID,
7585                 [I40E_CLOUD_FILTER_FLAGS_OMAC_TEN_ID_IMAC] =
7586                         I40E_AQC_ADD_CLOUD_FILTER_OMAC_TEN_ID_IMAC,
7587                 [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN_TEN_ID] =
7588                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN_TEN_ID,
7589                 [I40E_CLOUD_FILTER_FLAGS_IIP] =
7590                         I40E_AQC_ADD_CLOUD_FILTER_IIP,
7591         };
7592
7593         if (filter->flags >= ARRAY_SIZE(flag_table))
7594                 return I40E_ERR_CONFIG;
7595
7596         /* copy element needed to add cloud filter from filter */
7597         i40e_set_cld_element(filter, &cld_filter);
7598
7599         if (filter->tunnel_type != I40E_CLOUD_TNL_TYPE_NONE)
7600                 cld_filter.flags = cpu_to_le16(filter->tunnel_type <<
7601                                              I40E_AQC_ADD_CLOUD_TNL_TYPE_SHIFT);
7602
7603         if (filter->n_proto == ETH_P_IPV6)
7604                 cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7605                                                 I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7606         else
7607                 cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7608                                                 I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7609
7610         if (add)
7611                 ret = i40e_aq_add_cloud_filters(&pf->hw, filter->seid,
7612                                                 &cld_filter, 1);
7613         else
7614                 ret = i40e_aq_rem_cloud_filters(&pf->hw, filter->seid,
7615                                                 &cld_filter, 1);
7616         if (ret)
7617                 dev_dbg(&pf->pdev->dev,
7618                         "Failed to %s cloud filter using l4 port %u, err %d aq_err %d\n",
7619                         add ? "add" : "delete", filter->dst_port, ret,
7620                         pf->hw.aq.asq_last_status);
7621         else
7622                 dev_info(&pf->pdev->dev,
7623                          "%s cloud filter for VSI: %d\n",
7624                          add ? "Added" : "Deleted", filter->seid);
7625         return ret;
7626 }
7627
7628 /**
7629  * i40e_add_del_cloud_filter_big_buf - Add/del cloud filter using big_buf
7630  * @vsi: pointer to VSI
7631  * @filter: cloud filter rule
7632  * @add: if true, add, if false, delete
7633  *
7634  * Add or delete a cloud filter for a specific flow spec using big buffer.
7635  * Returns 0 if the filter were successfully added.
7636  **/
7637 int i40e_add_del_cloud_filter_big_buf(struct i40e_vsi *vsi,
7638                                       struct i40e_cloud_filter *filter,
7639                                       bool add)
7640 {
7641         struct i40e_aqc_cloud_filters_element_bb cld_filter;
7642         struct i40e_pf *pf = vsi->back;
7643         int ret;
7644
7645         /* Both (src/dst) valid mac_addr are not supported */
7646         if ((is_valid_ether_addr(filter->dst_mac) &&
7647              is_valid_ether_addr(filter->src_mac)) ||
7648             (is_multicast_ether_addr(filter->dst_mac) &&
7649              is_multicast_ether_addr(filter->src_mac)))
7650                 return -EOPNOTSUPP;
7651
7652         /* Big buffer cloud filter needs 'L4 port' to be non-zero. Also, UDP
7653          * ports are not supported via big buffer now.
7654          */
7655         if (!filter->dst_port || filter->ip_proto == IPPROTO_UDP)
7656                 return -EOPNOTSUPP;
7657
7658         /* adding filter using src_port/src_ip is not supported at this stage */
7659         if (filter->src_port || filter->src_ipv4 ||
7660             !ipv6_addr_any(&filter->ip.v6.src_ip6))
7661                 return -EOPNOTSUPP;
7662
7663         /* copy element needed to add cloud filter from filter */
7664         i40e_set_cld_element(filter, &cld_filter.element);
7665
7666         if (is_valid_ether_addr(filter->dst_mac) ||
7667             is_valid_ether_addr(filter->src_mac) ||
7668             is_multicast_ether_addr(filter->dst_mac) ||
7669             is_multicast_ether_addr(filter->src_mac)) {
7670                 /* MAC + IP : unsupported mode */
7671                 if (filter->dst_ipv4)
7672                         return -EOPNOTSUPP;
7673
7674                 /* since we validated that L4 port must be valid before
7675                  * we get here, start with respective "flags" value
7676                  * and update if vlan is present or not
7677                  */
7678                 cld_filter.element.flags =
7679                         cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_PORT);
7680
7681                 if (filter->vlan_id) {
7682                         cld_filter.element.flags =
7683                         cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_VLAN_PORT);
7684                 }
7685
7686         } else if (filter->dst_ipv4 ||
7687                    !ipv6_addr_any(&filter->ip.v6.dst_ip6)) {
7688                 cld_filter.element.flags =
7689                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_IP_PORT);
7690                 if (filter->n_proto == ETH_P_IPV6)
7691                         cld_filter.element.flags |=
7692                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7693                 else
7694                         cld_filter.element.flags |=
7695                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7696         } else {
7697                 dev_err(&pf->pdev->dev,
7698                         "either mac or ip has to be valid for cloud filter\n");
7699                 return -EINVAL;
7700         }
7701
7702         /* Now copy L4 port in Byte 6..7 in general fields */
7703         cld_filter.general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X16_WORD0] =
7704                                                 be16_to_cpu(filter->dst_port);
7705
7706         if (add) {
7707                 /* Validate current device switch mode, change if necessary */
7708                 ret = i40e_validate_and_set_switch_mode(vsi);
7709                 if (ret) {
7710                         dev_err(&pf->pdev->dev,
7711                                 "failed to set switch mode, ret %d\n",
7712                                 ret);
7713                         return ret;
7714                 }
7715
7716                 ret = i40e_aq_add_cloud_filters_bb(&pf->hw, filter->seid,
7717                                                    &cld_filter, 1);
7718         } else {
7719                 ret = i40e_aq_rem_cloud_filters_bb(&pf->hw, filter->seid,
7720                                                    &cld_filter, 1);
7721         }
7722
7723         if (ret)
7724                 dev_dbg(&pf->pdev->dev,
7725                         "Failed to %s cloud filter(big buffer) err %d aq_err %d\n",
7726                         add ? "add" : "delete", ret, pf->hw.aq.asq_last_status);
7727         else
7728                 dev_info(&pf->pdev->dev,
7729                          "%s cloud filter for VSI: %d, L4 port: %d\n",
7730                          add ? "add" : "delete", filter->seid,
7731                          ntohs(filter->dst_port));
7732         return ret;
7733 }
7734
7735 /**
7736  * i40e_parse_cls_flower - Parse tc flower filters provided by kernel
7737  * @vsi: Pointer to VSI
7738  * @cls_flower: Pointer to struct flow_cls_offload
7739  * @filter: Pointer to cloud filter structure
7740  *
7741  **/
7742 static int i40e_parse_cls_flower(struct i40e_vsi *vsi,
7743                                  struct flow_cls_offload *f,
7744                                  struct i40e_cloud_filter *filter)
7745 {
7746         struct flow_rule *rule = flow_cls_offload_flow_rule(f);
7747         struct flow_dissector *dissector = rule->match.dissector;
7748         u16 n_proto_mask = 0, n_proto_key = 0, addr_type = 0;
7749         struct i40e_pf *pf = vsi->back;
7750         u8 field_flags = 0;
7751
7752         if (dissector->used_keys &
7753             ~(BIT(FLOW_DISSECTOR_KEY_CONTROL) |
7754               BIT(FLOW_DISSECTOR_KEY_BASIC) |
7755               BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
7756               BIT(FLOW_DISSECTOR_KEY_VLAN) |
7757               BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
7758               BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
7759               BIT(FLOW_DISSECTOR_KEY_PORTS) |
7760               BIT(FLOW_DISSECTOR_KEY_ENC_KEYID))) {
7761                 dev_err(&pf->pdev->dev, "Unsupported key used: 0x%x\n",
7762                         dissector->used_keys);
7763                 return -EOPNOTSUPP;
7764         }
7765
7766         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
7767                 struct flow_match_enc_keyid match;
7768
7769                 flow_rule_match_enc_keyid(rule, &match);
7770                 if (match.mask->keyid != 0)
7771                         field_flags |= I40E_CLOUD_FIELD_TEN_ID;
7772
7773                 filter->tenant_id = be32_to_cpu(match.key->keyid);
7774         }
7775
7776         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
7777                 struct flow_match_basic match;
7778
7779                 flow_rule_match_basic(rule, &match);
7780                 n_proto_key = ntohs(match.key->n_proto);
7781                 n_proto_mask = ntohs(match.mask->n_proto);
7782
7783                 if (n_proto_key == ETH_P_ALL) {
7784                         n_proto_key = 0;
7785                         n_proto_mask = 0;
7786                 }
7787                 filter->n_proto = n_proto_key & n_proto_mask;
7788                 filter->ip_proto = match.key->ip_proto;
7789         }
7790
7791         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
7792                 struct flow_match_eth_addrs match;
7793
7794                 flow_rule_match_eth_addrs(rule, &match);
7795
7796                 /* use is_broadcast and is_zero to check for all 0xf or 0 */
7797                 if (!is_zero_ether_addr(match.mask->dst)) {
7798                         if (is_broadcast_ether_addr(match.mask->dst)) {
7799                                 field_flags |= I40E_CLOUD_FIELD_OMAC;
7800                         } else {
7801                                 dev_err(&pf->pdev->dev, "Bad ether dest mask %pM\n",
7802                                         match.mask->dst);
7803                                 return I40E_ERR_CONFIG;
7804                         }
7805                 }
7806
7807                 if (!is_zero_ether_addr(match.mask->src)) {
7808                         if (is_broadcast_ether_addr(match.mask->src)) {
7809                                 field_flags |= I40E_CLOUD_FIELD_IMAC;
7810                         } else {
7811                                 dev_err(&pf->pdev->dev, "Bad ether src mask %pM\n",
7812                                         match.mask->src);
7813                                 return I40E_ERR_CONFIG;
7814                         }
7815                 }
7816                 ether_addr_copy(filter->dst_mac, match.key->dst);
7817                 ether_addr_copy(filter->src_mac, match.key->src);
7818         }
7819
7820         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
7821                 struct flow_match_vlan match;
7822
7823                 flow_rule_match_vlan(rule, &match);
7824                 if (match.mask->vlan_id) {
7825                         if (match.mask->vlan_id == VLAN_VID_MASK) {
7826                                 field_flags |= I40E_CLOUD_FIELD_IVLAN;
7827
7828                         } else {
7829                                 dev_err(&pf->pdev->dev, "Bad vlan mask 0x%04x\n",
7830                                         match.mask->vlan_id);
7831                                 return I40E_ERR_CONFIG;
7832                         }
7833                 }
7834
7835                 filter->vlan_id = cpu_to_be16(match.key->vlan_id);
7836         }
7837
7838         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
7839                 struct flow_match_control match;
7840
7841                 flow_rule_match_control(rule, &match);
7842                 addr_type = match.key->addr_type;
7843         }
7844
7845         if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
7846                 struct flow_match_ipv4_addrs match;
7847
7848                 flow_rule_match_ipv4_addrs(rule, &match);
7849                 if (match.mask->dst) {
7850                         if (match.mask->dst == cpu_to_be32(0xffffffff)) {
7851                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7852                         } else {
7853                                 dev_err(&pf->pdev->dev, "Bad ip dst mask %pI4b\n",
7854                                         &match.mask->dst);
7855                                 return I40E_ERR_CONFIG;
7856                         }
7857                 }
7858
7859                 if (match.mask->src) {
7860                         if (match.mask->src == cpu_to_be32(0xffffffff)) {
7861                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7862                         } else {
7863                                 dev_err(&pf->pdev->dev, "Bad ip src mask %pI4b\n",
7864                                         &match.mask->src);
7865                                 return I40E_ERR_CONFIG;
7866                         }
7867                 }
7868
7869                 if (field_flags & I40E_CLOUD_FIELD_TEN_ID) {
7870                         dev_err(&pf->pdev->dev, "Tenant id not allowed for ip filter\n");
7871                         return I40E_ERR_CONFIG;
7872                 }
7873                 filter->dst_ipv4 = match.key->dst;
7874                 filter->src_ipv4 = match.key->src;
7875         }
7876
7877         if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
7878                 struct flow_match_ipv6_addrs match;
7879
7880                 flow_rule_match_ipv6_addrs(rule, &match);
7881
7882                 /* src and dest IPV6 address should not be LOOPBACK
7883                  * (0:0:0:0:0:0:0:1), which can be represented as ::1
7884                  */
7885                 if (ipv6_addr_loopback(&match.key->dst) ||
7886                     ipv6_addr_loopback(&match.key->src)) {
7887                         dev_err(&pf->pdev->dev,
7888                                 "Bad ipv6, addr is LOOPBACK\n");
7889                         return I40E_ERR_CONFIG;
7890                 }
7891                 if (!ipv6_addr_any(&match.mask->dst) ||
7892                     !ipv6_addr_any(&match.mask->src))
7893                         field_flags |= I40E_CLOUD_FIELD_IIP;
7894
7895                 memcpy(&filter->src_ipv6, &match.key->src.s6_addr32,
7896                        sizeof(filter->src_ipv6));
7897                 memcpy(&filter->dst_ipv6, &match.key->dst.s6_addr32,
7898                        sizeof(filter->dst_ipv6));
7899         }
7900
7901         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS)) {
7902                 struct flow_match_ports match;
7903
7904                 flow_rule_match_ports(rule, &match);
7905                 if (match.mask->src) {
7906                         if (match.mask->src == cpu_to_be16(0xffff)) {
7907                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7908                         } else {
7909                                 dev_err(&pf->pdev->dev, "Bad src port mask 0x%04x\n",
7910                                         be16_to_cpu(match.mask->src));
7911                                 return I40E_ERR_CONFIG;
7912                         }
7913                 }
7914
7915                 if (match.mask->dst) {
7916                         if (match.mask->dst == cpu_to_be16(0xffff)) {
7917                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7918                         } else {
7919                                 dev_err(&pf->pdev->dev, "Bad dst port mask 0x%04x\n",
7920                                         be16_to_cpu(match.mask->dst));
7921                                 return I40E_ERR_CONFIG;
7922                         }
7923                 }
7924
7925                 filter->dst_port = match.key->dst;
7926                 filter->src_port = match.key->src;
7927
7928                 switch (filter->ip_proto) {
7929                 case IPPROTO_TCP:
7930                 case IPPROTO_UDP:
7931                         break;
7932                 default:
7933                         dev_err(&pf->pdev->dev,
7934                                 "Only UDP and TCP transport are supported\n");
7935                         return -EINVAL;
7936                 }
7937         }
7938         filter->flags = field_flags;
7939         return 0;
7940 }
7941
7942 /**
7943  * i40e_handle_tclass: Forward to a traffic class on the device
7944  * @vsi: Pointer to VSI
7945  * @tc: traffic class index on the device
7946  * @filter: Pointer to cloud filter structure
7947  *
7948  **/
7949 static int i40e_handle_tclass(struct i40e_vsi *vsi, u32 tc,
7950                               struct i40e_cloud_filter *filter)
7951 {
7952         struct i40e_channel *ch, *ch_tmp;
7953
7954         /* direct to a traffic class on the same device */
7955         if (tc == 0) {
7956                 filter->seid = vsi->seid;
7957                 return 0;
7958         } else if (vsi->tc_config.enabled_tc & BIT(tc)) {
7959                 if (!filter->dst_port) {
7960                         dev_err(&vsi->back->pdev->dev,
7961                                 "Specify destination port to direct to traffic class that is not default\n");
7962                         return -EINVAL;
7963                 }
7964                 if (list_empty(&vsi->ch_list))
7965                         return -EINVAL;
7966                 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list,
7967                                          list) {
7968                         if (ch->seid == vsi->tc_seid_map[tc])
7969                                 filter->seid = ch->seid;
7970                 }
7971                 return 0;
7972         }
7973         dev_err(&vsi->back->pdev->dev, "TC is not enabled\n");
7974         return -EINVAL;
7975 }
7976
7977 /**
7978  * i40e_configure_clsflower - Configure tc flower filters
7979  * @vsi: Pointer to VSI
7980  * @cls_flower: Pointer to struct flow_cls_offload
7981  *
7982  **/
7983 static int i40e_configure_clsflower(struct i40e_vsi *vsi,
7984                                     struct flow_cls_offload *cls_flower)
7985 {
7986         int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
7987         struct i40e_cloud_filter *filter = NULL;
7988         struct i40e_pf *pf = vsi->back;
7989         int err = 0;
7990
7991         if (tc < 0) {
7992                 dev_err(&vsi->back->pdev->dev, "Invalid traffic class\n");
7993                 return -EOPNOTSUPP;
7994         }
7995
7996         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
7997             test_bit(__I40E_RESET_INTR_RECEIVED, pf->state))
7998                 return -EBUSY;
7999
8000         if (pf->fdir_pf_active_filters ||
8001             (!hlist_empty(&pf->fdir_filter_list))) {
8002                 dev_err(&vsi->back->pdev->dev,
8003                         "Flow Director Sideband filters exists, turn ntuple off to configure cloud filters\n");
8004                 return -EINVAL;
8005         }
8006
8007         if (vsi->back->flags & I40E_FLAG_FD_SB_ENABLED) {
8008                 dev_err(&vsi->back->pdev->dev,
8009                         "Disable Flow Director Sideband, configuring Cloud filters via tc-flower\n");
8010                 vsi->back->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8011                 vsi->back->flags |= I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8012         }
8013
8014         filter = kzalloc(sizeof(*filter), GFP_KERNEL);
8015         if (!filter)
8016                 return -ENOMEM;
8017
8018         filter->cookie = cls_flower->cookie;
8019
8020         err = i40e_parse_cls_flower(vsi, cls_flower, filter);
8021         if (err < 0)
8022                 goto err;
8023
8024         err = i40e_handle_tclass(vsi, tc, filter);
8025         if (err < 0)
8026                 goto err;
8027
8028         /* Add cloud filter */
8029         if (filter->dst_port)
8030                 err = i40e_add_del_cloud_filter_big_buf(vsi, filter, true);
8031         else
8032                 err = i40e_add_del_cloud_filter(vsi, filter, true);
8033
8034         if (err) {
8035                 dev_err(&pf->pdev->dev,
8036                         "Failed to add cloud filter, err %s\n",
8037                         i40e_stat_str(&pf->hw, err));
8038                 goto err;
8039         }
8040
8041         /* add filter to the ordered list */
8042         INIT_HLIST_NODE(&filter->cloud_node);
8043
8044         hlist_add_head(&filter->cloud_node, &pf->cloud_filter_list);
8045
8046         pf->num_cloud_filters++;
8047
8048         return err;
8049 err:
8050         kfree(filter);
8051         return err;
8052 }
8053
8054 /**
8055  * i40e_find_cloud_filter - Find the could filter in the list
8056  * @vsi: Pointer to VSI
8057  * @cookie: filter specific cookie
8058  *
8059  **/
8060 static struct i40e_cloud_filter *i40e_find_cloud_filter(struct i40e_vsi *vsi,
8061                                                         unsigned long *cookie)
8062 {
8063         struct i40e_cloud_filter *filter = NULL;
8064         struct hlist_node *node2;
8065
8066         hlist_for_each_entry_safe(filter, node2,
8067                                   &vsi->back->cloud_filter_list, cloud_node)
8068                 if (!memcmp(cookie, &filter->cookie, sizeof(filter->cookie)))
8069                         return filter;
8070         return NULL;
8071 }
8072
8073 /**
8074  * i40e_delete_clsflower - Remove tc flower filters
8075  * @vsi: Pointer to VSI
8076  * @cls_flower: Pointer to struct flow_cls_offload
8077  *
8078  **/
8079 static int i40e_delete_clsflower(struct i40e_vsi *vsi,
8080                                  struct flow_cls_offload *cls_flower)
8081 {
8082         struct i40e_cloud_filter *filter = NULL;
8083         struct i40e_pf *pf = vsi->back;
8084         int err = 0;
8085
8086         filter = i40e_find_cloud_filter(vsi, &cls_flower->cookie);
8087
8088         if (!filter)
8089                 return -EINVAL;
8090
8091         hash_del(&filter->cloud_node);
8092
8093         if (filter->dst_port)
8094                 err = i40e_add_del_cloud_filter_big_buf(vsi, filter, false);
8095         else
8096                 err = i40e_add_del_cloud_filter(vsi, filter, false);
8097
8098         kfree(filter);
8099         if (err) {
8100                 dev_err(&pf->pdev->dev,
8101                         "Failed to delete cloud filter, err %s\n",
8102                         i40e_stat_str(&pf->hw, err));
8103                 return i40e_aq_rc_to_posix(err, pf->hw.aq.asq_last_status);
8104         }
8105
8106         pf->num_cloud_filters--;
8107         if (!pf->num_cloud_filters)
8108                 if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8109                     !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8110                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8111                         pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8112                         pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8113                 }
8114         return 0;
8115 }
8116
8117 /**
8118  * i40e_setup_tc_cls_flower - flower classifier offloads
8119  * @netdev: net device to configure
8120  * @type_data: offload data
8121  **/
8122 static int i40e_setup_tc_cls_flower(struct i40e_netdev_priv *np,
8123                                     struct flow_cls_offload *cls_flower)
8124 {
8125         struct i40e_vsi *vsi = np->vsi;
8126
8127         switch (cls_flower->command) {
8128         case FLOW_CLS_REPLACE:
8129                 return i40e_configure_clsflower(vsi, cls_flower);
8130         case FLOW_CLS_DESTROY:
8131                 return i40e_delete_clsflower(vsi, cls_flower);
8132         case FLOW_CLS_STATS:
8133                 return -EOPNOTSUPP;
8134         default:
8135                 return -EOPNOTSUPP;
8136         }
8137 }
8138
8139 static int i40e_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
8140                                   void *cb_priv)
8141 {
8142         struct i40e_netdev_priv *np = cb_priv;
8143
8144         if (!tc_cls_can_offload_and_chain0(np->vsi->netdev, type_data))
8145                 return -EOPNOTSUPP;
8146
8147         switch (type) {
8148         case TC_SETUP_CLSFLOWER:
8149                 return i40e_setup_tc_cls_flower(np, type_data);
8150
8151         default:
8152                 return -EOPNOTSUPP;
8153         }
8154 }
8155
8156 static LIST_HEAD(i40e_block_cb_list);
8157
8158 static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8159                            void *type_data)
8160 {
8161         struct i40e_netdev_priv *np = netdev_priv(netdev);
8162
8163         switch (type) {
8164         case TC_SETUP_QDISC_MQPRIO:
8165                 return i40e_setup_tc(netdev, type_data);
8166         case TC_SETUP_BLOCK:
8167                 return flow_block_cb_setup_simple(type_data,
8168                                                   &i40e_block_cb_list,
8169                                                   i40e_setup_tc_block_cb,
8170                                                   np, np, true);
8171         default:
8172                 return -EOPNOTSUPP;
8173         }
8174 }
8175
8176 /**
8177  * i40e_open - Called when a network interface is made active
8178  * @netdev: network interface device structure
8179  *
8180  * The open entry point is called when a network interface is made
8181  * active by the system (IFF_UP).  At this point all resources needed
8182  * for transmit and receive operations are allocated, the interrupt
8183  * handler is registered with the OS, the netdev watchdog subtask is
8184  * enabled, and the stack is notified that the interface is ready.
8185  *
8186  * Returns 0 on success, negative value on failure
8187  **/
8188 int i40e_open(struct net_device *netdev)
8189 {
8190         struct i40e_netdev_priv *np = netdev_priv(netdev);
8191         struct i40e_vsi *vsi = np->vsi;
8192         struct i40e_pf *pf = vsi->back;
8193         int err;
8194
8195         /* disallow open during test or if eeprom is broken */
8196         if (test_bit(__I40E_TESTING, pf->state) ||
8197             test_bit(__I40E_BAD_EEPROM, pf->state))
8198                 return -EBUSY;
8199
8200         netif_carrier_off(netdev);
8201
8202         if (i40e_force_link_state(pf, true))
8203                 return -EAGAIN;
8204
8205         err = i40e_vsi_open(vsi);
8206         if (err)
8207                 return err;
8208
8209         /* configure global TSO hardware offload settings */
8210         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
8211                                                        TCP_FLAG_FIN) >> 16);
8212         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
8213                                                        TCP_FLAG_FIN |
8214                                                        TCP_FLAG_CWR) >> 16);
8215         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
8216
8217         udp_tunnel_get_rx_info(netdev);
8218
8219         return 0;
8220 }
8221
8222 /**
8223  * i40e_vsi_open -
8224  * @vsi: the VSI to open
8225  *
8226  * Finish initialization of the VSI.
8227  *
8228  * Returns 0 on success, negative value on failure
8229  *
8230  * Note: expects to be called while under rtnl_lock()
8231  **/
8232 int i40e_vsi_open(struct i40e_vsi *vsi)
8233 {
8234         struct i40e_pf *pf = vsi->back;
8235         char int_name[I40E_INT_NAME_STR_LEN];
8236         int err;
8237
8238         /* allocate descriptors */
8239         err = i40e_vsi_setup_tx_resources(vsi);
8240         if (err)
8241                 goto err_setup_tx;
8242         err = i40e_vsi_setup_rx_resources(vsi);
8243         if (err)
8244                 goto err_setup_rx;
8245
8246         err = i40e_vsi_configure(vsi);
8247         if (err)
8248                 goto err_setup_rx;
8249
8250         if (vsi->netdev) {
8251                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
8252                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
8253                 err = i40e_vsi_request_irq(vsi, int_name);
8254                 if (err)
8255                         goto err_setup_rx;
8256
8257                 /* Notify the stack of the actual queue counts. */
8258                 err = netif_set_real_num_tx_queues(vsi->netdev,
8259                                                    vsi->num_queue_pairs);
8260                 if (err)
8261                         goto err_set_queues;
8262
8263                 err = netif_set_real_num_rx_queues(vsi->netdev,
8264                                                    vsi->num_queue_pairs);
8265                 if (err)
8266                         goto err_set_queues;
8267
8268         } else if (vsi->type == I40E_VSI_FDIR) {
8269                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
8270                          dev_driver_string(&pf->pdev->dev),
8271                          dev_name(&pf->pdev->dev));
8272                 err = i40e_vsi_request_irq(vsi, int_name);
8273
8274         } else {
8275                 err = -EINVAL;
8276                 goto err_setup_rx;
8277         }
8278
8279         err = i40e_up_complete(vsi);
8280         if (err)
8281                 goto err_up_complete;
8282
8283         return 0;
8284
8285 err_up_complete:
8286         i40e_down(vsi);
8287 err_set_queues:
8288         i40e_vsi_free_irq(vsi);
8289 err_setup_rx:
8290         i40e_vsi_free_rx_resources(vsi);
8291 err_setup_tx:
8292         i40e_vsi_free_tx_resources(vsi);
8293         if (vsi == pf->vsi[pf->lan_vsi])
8294                 i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
8295
8296         return err;
8297 }
8298
8299 /**
8300  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
8301  * @pf: Pointer to PF
8302  *
8303  * This function destroys the hlist where all the Flow Director
8304  * filters were saved.
8305  **/
8306 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
8307 {
8308         struct i40e_fdir_filter *filter;
8309         struct i40e_flex_pit *pit_entry, *tmp;
8310         struct hlist_node *node2;
8311
8312         hlist_for_each_entry_safe(filter, node2,
8313                                   &pf->fdir_filter_list, fdir_node) {
8314                 hlist_del(&filter->fdir_node);
8315                 kfree(filter);
8316         }
8317
8318         list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
8319                 list_del(&pit_entry->list);
8320                 kfree(pit_entry);
8321         }
8322         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
8323
8324         list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
8325                 list_del(&pit_entry->list);
8326                 kfree(pit_entry);
8327         }
8328         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
8329
8330         pf->fdir_pf_active_filters = 0;
8331         pf->fd_tcp4_filter_cnt = 0;
8332         pf->fd_udp4_filter_cnt = 0;
8333         pf->fd_sctp4_filter_cnt = 0;
8334         pf->fd_ip4_filter_cnt = 0;
8335
8336         /* Reprogram the default input set for TCP/IPv4 */
8337         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8338                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8339                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8340
8341         /* Reprogram the default input set for UDP/IPv4 */
8342         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
8343                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8344                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8345
8346         /* Reprogram the default input set for SCTP/IPv4 */
8347         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
8348                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8349                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8350
8351         /* Reprogram the default input set for Other/IPv4 */
8352         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
8353                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8354
8355         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_FRAG_IPV4,
8356                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8357 }
8358
8359 /**
8360  * i40e_cloud_filter_exit - Cleans up the cloud filters
8361  * @pf: Pointer to PF
8362  *
8363  * This function destroys the hlist where all the cloud filters
8364  * were saved.
8365  **/
8366 static void i40e_cloud_filter_exit(struct i40e_pf *pf)
8367 {
8368         struct i40e_cloud_filter *cfilter;
8369         struct hlist_node *node;
8370
8371         hlist_for_each_entry_safe(cfilter, node,
8372                                   &pf->cloud_filter_list, cloud_node) {
8373                 hlist_del(&cfilter->cloud_node);
8374                 kfree(cfilter);
8375         }
8376         pf->num_cloud_filters = 0;
8377
8378         if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8379             !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8380                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8381                 pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8382                 pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8383         }
8384 }
8385
8386 /**
8387  * i40e_close - Disables a network interface
8388  * @netdev: network interface device structure
8389  *
8390  * The close entry point is called when an interface is de-activated
8391  * by the OS.  The hardware is still under the driver's control, but
8392  * this netdev interface is disabled.
8393  *
8394  * Returns 0, this is not allowed to fail
8395  **/
8396 int i40e_close(struct net_device *netdev)
8397 {
8398         struct i40e_netdev_priv *np = netdev_priv(netdev);
8399         struct i40e_vsi *vsi = np->vsi;
8400
8401         i40e_vsi_close(vsi);
8402
8403         return 0;
8404 }
8405
8406 /**
8407  * i40e_do_reset - Start a PF or Core Reset sequence
8408  * @pf: board private structure
8409  * @reset_flags: which reset is requested
8410  * @lock_acquired: indicates whether or not the lock has been acquired
8411  * before this function was called.
8412  *
8413  * The essential difference in resets is that the PF Reset
8414  * doesn't clear the packet buffers, doesn't reset the PE
8415  * firmware, and doesn't bother the other PFs on the chip.
8416  **/
8417 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
8418 {
8419         u32 val;
8420
8421         WARN_ON(in_interrupt());
8422
8423
8424         /* do the biggest reset indicated */
8425         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
8426
8427                 /* Request a Global Reset
8428                  *
8429                  * This will start the chip's countdown to the actual full
8430                  * chip reset event, and a warning interrupt to be sent
8431                  * to all PFs, including the requestor.  Our handler
8432                  * for the warning interrupt will deal with the shutdown
8433                  * and recovery of the switch setup.
8434                  */
8435                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
8436                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8437                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
8438                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8439
8440         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
8441
8442                 /* Request a Core Reset
8443                  *
8444                  * Same as Global Reset, except does *not* include the MAC/PHY
8445                  */
8446                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
8447                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8448                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
8449                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8450                 i40e_flush(&pf->hw);
8451
8452         } else if (reset_flags & I40E_PF_RESET_FLAG) {
8453
8454                 /* Request a PF Reset
8455                  *
8456                  * Resets only the PF-specific registers
8457                  *
8458                  * This goes directly to the tear-down and rebuild of
8459                  * the switch, since we need to do all the recovery as
8460                  * for the Core Reset.
8461                  */
8462                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
8463                 i40e_handle_reset_warning(pf, lock_acquired);
8464
8465                 dev_info(&pf->pdev->dev,
8466                          pf->flags & I40E_FLAG_DISABLE_FW_LLDP ?
8467                          "FW LLDP is disabled\n" :
8468                          "FW LLDP is enabled\n");
8469
8470         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
8471                 int v;
8472
8473                 /* Find the VSI(s) that requested a re-init */
8474                 dev_info(&pf->pdev->dev,
8475                          "VSI reinit requested\n");
8476                 for (v = 0; v < pf->num_alloc_vsi; v++) {
8477                         struct i40e_vsi *vsi = pf->vsi[v];
8478
8479                         if (vsi != NULL &&
8480                             test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
8481                                                vsi->state))
8482                                 i40e_vsi_reinit_locked(pf->vsi[v]);
8483                 }
8484         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
8485                 int v;
8486
8487                 /* Find the VSI(s) that needs to be brought down */
8488                 dev_info(&pf->pdev->dev, "VSI down requested\n");
8489                 for (v = 0; v < pf->num_alloc_vsi; v++) {
8490                         struct i40e_vsi *vsi = pf->vsi[v];
8491
8492                         if (vsi != NULL &&
8493                             test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
8494                                                vsi->state)) {
8495                                 set_bit(__I40E_VSI_DOWN, vsi->state);
8496                                 i40e_down(vsi);
8497                         }
8498                 }
8499         } else {
8500                 dev_info(&pf->pdev->dev,
8501                          "bad reset request 0x%08x\n", reset_flags);
8502         }
8503 }
8504
8505 #ifdef CONFIG_I40E_DCB
8506 /**
8507  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
8508  * @pf: board private structure
8509  * @old_cfg: current DCB config
8510  * @new_cfg: new DCB config
8511  **/
8512 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
8513                             struct i40e_dcbx_config *old_cfg,
8514                             struct i40e_dcbx_config *new_cfg)
8515 {
8516         bool need_reconfig = false;
8517
8518         /* Check if ETS configuration has changed */
8519         if (memcmp(&new_cfg->etscfg,
8520                    &old_cfg->etscfg,
8521                    sizeof(new_cfg->etscfg))) {
8522                 /* If Priority Table has changed reconfig is needed */
8523                 if (memcmp(&new_cfg->etscfg.prioritytable,
8524                            &old_cfg->etscfg.prioritytable,
8525                            sizeof(new_cfg->etscfg.prioritytable))) {
8526                         need_reconfig = true;
8527                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
8528                 }
8529
8530                 if (memcmp(&new_cfg->etscfg.tcbwtable,
8531                            &old_cfg->etscfg.tcbwtable,
8532                            sizeof(new_cfg->etscfg.tcbwtable)))
8533                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
8534
8535                 if (memcmp(&new_cfg->etscfg.tsatable,
8536                            &old_cfg->etscfg.tsatable,
8537                            sizeof(new_cfg->etscfg.tsatable)))
8538                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
8539         }
8540
8541         /* Check if PFC configuration has changed */
8542         if (memcmp(&new_cfg->pfc,
8543                    &old_cfg->pfc,
8544                    sizeof(new_cfg->pfc))) {
8545                 need_reconfig = true;
8546                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
8547         }
8548
8549         /* Check if APP Table has changed */
8550         if (memcmp(&new_cfg->app,
8551                    &old_cfg->app,
8552                    sizeof(new_cfg->app))) {
8553                 need_reconfig = true;
8554                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
8555         }
8556
8557         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
8558         return need_reconfig;
8559 }
8560
8561 /**
8562  * i40e_handle_lldp_event - Handle LLDP Change MIB event
8563  * @pf: board private structure
8564  * @e: event info posted on ARQ
8565  **/
8566 static int i40e_handle_lldp_event(struct i40e_pf *pf,
8567                                   struct i40e_arq_event_info *e)
8568 {
8569         struct i40e_aqc_lldp_get_mib *mib =
8570                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
8571         struct i40e_hw *hw = &pf->hw;
8572         struct i40e_dcbx_config tmp_dcbx_cfg;
8573         bool need_reconfig = false;
8574         int ret = 0;
8575         u8 type;
8576
8577         /* Not DCB capable or capability disabled */
8578         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
8579                 return ret;
8580
8581         /* Ignore if event is not for Nearest Bridge */
8582         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
8583                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
8584         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
8585         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
8586                 return ret;
8587
8588         /* Check MIB Type and return if event for Remote MIB update */
8589         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
8590         dev_dbg(&pf->pdev->dev,
8591                 "LLDP event mib type %s\n", type ? "remote" : "local");
8592         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
8593                 /* Update the remote cached instance and return */
8594                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
8595                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
8596                                 &hw->remote_dcbx_config);
8597                 goto exit;
8598         }
8599
8600         /* Store the old configuration */
8601         tmp_dcbx_cfg = hw->local_dcbx_config;
8602
8603         /* Reset the old DCBx configuration data */
8604         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
8605         /* Get updated DCBX data from firmware */
8606         ret = i40e_get_dcb_config(&pf->hw);
8607         if (ret) {
8608                 dev_info(&pf->pdev->dev,
8609                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
8610                          i40e_stat_str(&pf->hw, ret),
8611                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8612                 goto exit;
8613         }
8614
8615         /* No change detected in DCBX configs */
8616         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
8617                     sizeof(tmp_dcbx_cfg))) {
8618                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
8619                 goto exit;
8620         }
8621
8622         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
8623                                                &hw->local_dcbx_config);
8624
8625         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
8626
8627         if (!need_reconfig)
8628                 goto exit;
8629
8630         /* Enable DCB tagging only when more than one TC */
8631         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
8632                 pf->flags |= I40E_FLAG_DCB_ENABLED;
8633         else
8634                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8635
8636         set_bit(__I40E_PORT_SUSPENDED, pf->state);
8637         /* Reconfiguration needed quiesce all VSIs */
8638         i40e_pf_quiesce_all_vsi(pf);
8639
8640         /* Changes in configuration update VEB/VSI */
8641         i40e_dcb_reconfigure(pf);
8642
8643         ret = i40e_resume_port_tx(pf);
8644
8645         clear_bit(__I40E_PORT_SUSPENDED, pf->state);
8646         /* In case of error no point in resuming VSIs */
8647         if (ret)
8648                 goto exit;
8649
8650         /* Wait for the PF's queues to be disabled */
8651         ret = i40e_pf_wait_queues_disabled(pf);
8652         if (ret) {
8653                 /* Schedule PF reset to recover */
8654                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8655                 i40e_service_event_schedule(pf);
8656         } else {
8657                 i40e_pf_unquiesce_all_vsi(pf);
8658                 set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
8659                 set_bit(__I40E_CLIENT_L2_CHANGE, pf->state);
8660         }
8661
8662 exit:
8663         return ret;
8664 }
8665 #endif /* CONFIG_I40E_DCB */
8666
8667 /**
8668  * i40e_do_reset_safe - Protected reset path for userland calls.
8669  * @pf: board private structure
8670  * @reset_flags: which reset is requested
8671  *
8672  **/
8673 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
8674 {
8675         rtnl_lock();
8676         i40e_do_reset(pf, reset_flags, true);
8677         rtnl_unlock();
8678 }
8679
8680 /**
8681  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
8682  * @pf: board private structure
8683  * @e: event info posted on ARQ
8684  *
8685  * Handler for LAN Queue Overflow Event generated by the firmware for PF
8686  * and VF queues
8687  **/
8688 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
8689                                            struct i40e_arq_event_info *e)
8690 {
8691         struct i40e_aqc_lan_overflow *data =
8692                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
8693         u32 queue = le32_to_cpu(data->prtdcb_rupto);
8694         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
8695         struct i40e_hw *hw = &pf->hw;
8696         struct i40e_vf *vf;
8697         u16 vf_id;
8698
8699         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
8700                 queue, qtx_ctl);
8701
8702         /* Queue belongs to VF, find the VF and issue VF reset */
8703         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
8704             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
8705                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
8706                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
8707                 vf_id -= hw->func_caps.vf_base_id;
8708                 vf = &pf->vf[vf_id];
8709                 i40e_vc_notify_vf_reset(vf);
8710                 /* Allow VF to process pending reset notification */
8711                 msleep(20);
8712                 i40e_reset_vf(vf, false);
8713         }
8714 }
8715
8716 /**
8717  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
8718  * @pf: board private structure
8719  **/
8720 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
8721 {
8722         u32 val, fcnt_prog;
8723
8724         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8725         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
8726         return fcnt_prog;
8727 }
8728
8729 /**
8730  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
8731  * @pf: board private structure
8732  **/
8733 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
8734 {
8735         u32 val, fcnt_prog;
8736
8737         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8738         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
8739                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
8740                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
8741         return fcnt_prog;
8742 }
8743
8744 /**
8745  * i40e_get_global_fd_count - Get total FD filters programmed on device
8746  * @pf: board private structure
8747  **/
8748 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
8749 {
8750         u32 val, fcnt_prog;
8751
8752         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
8753         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
8754                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
8755                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
8756         return fcnt_prog;
8757 }
8758
8759 /**
8760  * i40e_reenable_fdir_sb - Restore FDir SB capability
8761  * @pf: board private structure
8762  **/
8763 static void i40e_reenable_fdir_sb(struct i40e_pf *pf)
8764 {
8765         if (test_and_clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state))
8766                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8767                     (I40E_DEBUG_FD & pf->hw.debug_mask))
8768                         dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
8769 }
8770
8771 /**
8772  * i40e_reenable_fdir_atr - Restore FDir ATR capability
8773  * @pf: board private structure
8774  **/
8775 static void i40e_reenable_fdir_atr(struct i40e_pf *pf)
8776 {
8777         if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state)) {
8778                 /* ATR uses the same filtering logic as SB rules. It only
8779                  * functions properly if the input set mask is at the default
8780                  * settings. It is safe to restore the default input set
8781                  * because there are no active TCPv4 filter rules.
8782                  */
8783                 i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8784                                         I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8785                                         I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8786
8787                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8788                     (I40E_DEBUG_FD & pf->hw.debug_mask))
8789                         dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
8790         }
8791 }
8792
8793 /**
8794  * i40e_delete_invalid_filter - Delete an invalid FDIR filter
8795  * @pf: board private structure
8796  * @filter: FDir filter to remove
8797  */
8798 static void i40e_delete_invalid_filter(struct i40e_pf *pf,
8799                                        struct i40e_fdir_filter *filter)
8800 {
8801         /* Update counters */
8802         pf->fdir_pf_active_filters--;
8803         pf->fd_inv = 0;
8804
8805         switch (filter->flow_type) {
8806         case TCP_V4_FLOW:
8807                 pf->fd_tcp4_filter_cnt--;
8808                 break;
8809         case UDP_V4_FLOW:
8810                 pf->fd_udp4_filter_cnt--;
8811                 break;
8812         case SCTP_V4_FLOW:
8813                 pf->fd_sctp4_filter_cnt--;
8814                 break;
8815         case IP_USER_FLOW:
8816                 switch (filter->ip4_proto) {
8817                 case IPPROTO_TCP:
8818                         pf->fd_tcp4_filter_cnt--;
8819                         break;
8820                 case IPPROTO_UDP:
8821                         pf->fd_udp4_filter_cnt--;
8822                         break;
8823                 case IPPROTO_SCTP:
8824                         pf->fd_sctp4_filter_cnt--;
8825                         break;
8826                 case IPPROTO_IP:
8827                         pf->fd_ip4_filter_cnt--;
8828                         break;
8829                 }
8830                 break;
8831         }
8832
8833         /* Remove the filter from the list and free memory */
8834         hlist_del(&filter->fdir_node);
8835         kfree(filter);
8836 }
8837
8838 /**
8839  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
8840  * @pf: board private structure
8841  **/
8842 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
8843 {
8844         struct i40e_fdir_filter *filter;
8845         u32 fcnt_prog, fcnt_avail;
8846         struct hlist_node *node;
8847
8848         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8849                 return;
8850
8851         /* Check if we have enough room to re-enable FDir SB capability. */
8852         fcnt_prog = i40e_get_global_fd_count(pf);
8853         fcnt_avail = pf->fdir_pf_filter_count;
8854         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
8855             (pf->fd_add_err == 0) ||
8856             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt))
8857                 i40e_reenable_fdir_sb(pf);
8858
8859         /* We should wait for even more space before re-enabling ATR.
8860          * Additionally, we cannot enable ATR as long as we still have TCP SB
8861          * rules active.
8862          */
8863         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
8864             (pf->fd_tcp4_filter_cnt == 0))
8865                 i40e_reenable_fdir_atr(pf);
8866
8867         /* if hw had a problem adding a filter, delete it */
8868         if (pf->fd_inv > 0) {
8869                 hlist_for_each_entry_safe(filter, node,
8870                                           &pf->fdir_filter_list, fdir_node)
8871                         if (filter->fd_id == pf->fd_inv)
8872                                 i40e_delete_invalid_filter(pf, filter);
8873         }
8874 }
8875
8876 #define I40E_MIN_FD_FLUSH_INTERVAL 10
8877 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
8878 /**
8879  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
8880  * @pf: board private structure
8881  **/
8882 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
8883 {
8884         unsigned long min_flush_time;
8885         int flush_wait_retry = 50;
8886         bool disable_atr = false;
8887         int fd_room;
8888         int reg;
8889
8890         if (!time_after(jiffies, pf->fd_flush_timestamp +
8891                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
8892                 return;
8893
8894         /* If the flush is happening too quick and we have mostly SB rules we
8895          * should not re-enable ATR for some time.
8896          */
8897         min_flush_time = pf->fd_flush_timestamp +
8898                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
8899         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
8900
8901         if (!(time_after(jiffies, min_flush_time)) &&
8902             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
8903                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8904                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
8905                 disable_atr = true;
8906         }
8907
8908         pf->fd_flush_timestamp = jiffies;
8909         set_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8910         /* flush all filters */
8911         wr32(&pf->hw, I40E_PFQF_CTL_1,
8912              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
8913         i40e_flush(&pf->hw);
8914         pf->fd_flush_cnt++;
8915         pf->fd_add_err = 0;
8916         do {
8917                 /* Check FD flush status every 5-6msec */
8918                 usleep_range(5000, 6000);
8919                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
8920                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
8921                         break;
8922         } while (flush_wait_retry--);
8923         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
8924                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
8925         } else {
8926                 /* replay sideband filters */
8927                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
8928                 if (!disable_atr && !pf->fd_tcp4_filter_cnt)
8929                         clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8930                 clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
8931                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8932                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
8933         }
8934 }
8935
8936 /**
8937  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
8938  * @pf: board private structure
8939  **/
8940 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
8941 {
8942         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
8943 }
8944
8945 /* We can see up to 256 filter programming desc in transit if the filters are
8946  * being applied really fast; before we see the first
8947  * filter miss error on Rx queue 0. Accumulating enough error messages before
8948  * reacting will make sure we don't cause flush too often.
8949  */
8950 #define I40E_MAX_FD_PROGRAM_ERROR 256
8951
8952 /**
8953  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
8954  * @pf: board private structure
8955  **/
8956 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
8957 {
8958
8959         /* if interface is down do nothing */
8960         if (test_bit(__I40E_DOWN, pf->state))
8961                 return;
8962
8963         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8964                 i40e_fdir_flush_and_replay(pf);
8965
8966         i40e_fdir_check_and_reenable(pf);
8967
8968 }
8969
8970 /**
8971  * i40e_vsi_link_event - notify VSI of a link event
8972  * @vsi: vsi to be notified
8973  * @link_up: link up or down
8974  **/
8975 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
8976 {
8977         if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
8978                 return;
8979
8980         switch (vsi->type) {
8981         case I40E_VSI_MAIN:
8982                 if (!vsi->netdev || !vsi->netdev_registered)
8983                         break;
8984
8985                 if (link_up) {
8986                         netif_carrier_on(vsi->netdev);
8987                         netif_tx_wake_all_queues(vsi->netdev);
8988                 } else {
8989                         netif_carrier_off(vsi->netdev);
8990                         netif_tx_stop_all_queues(vsi->netdev);
8991                 }
8992                 break;
8993
8994         case I40E_VSI_SRIOV:
8995         case I40E_VSI_VMDQ2:
8996         case I40E_VSI_CTRL:
8997         case I40E_VSI_IWARP:
8998         case I40E_VSI_MIRROR:
8999         default:
9000                 /* there is no notification for other VSIs */
9001                 break;
9002         }
9003 }
9004
9005 /**
9006  * i40e_veb_link_event - notify elements on the veb of a link event
9007  * @veb: veb to be notified
9008  * @link_up: link up or down
9009  **/
9010 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
9011 {
9012         struct i40e_pf *pf;
9013         int i;
9014
9015         if (!veb || !veb->pf)
9016                 return;
9017         pf = veb->pf;
9018
9019         /* depth first... */
9020         for (i = 0; i < I40E_MAX_VEB; i++)
9021                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
9022                         i40e_veb_link_event(pf->veb[i], link_up);
9023
9024         /* ... now the local VSIs */
9025         for (i = 0; i < pf->num_alloc_vsi; i++)
9026                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
9027                         i40e_vsi_link_event(pf->vsi[i], link_up);
9028 }
9029
9030 /**
9031  * i40e_link_event - Update netif_carrier status
9032  * @pf: board private structure
9033  **/
9034 static void i40e_link_event(struct i40e_pf *pf)
9035 {
9036         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9037         u8 new_link_speed, old_link_speed;
9038         i40e_status status;
9039         bool new_link, old_link;
9040
9041         /* set this to force the get_link_status call to refresh state */
9042         pf->hw.phy.get_link_info = true;
9043         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
9044         status = i40e_get_link_status(&pf->hw, &new_link);
9045
9046         /* On success, disable temp link polling */
9047         if (status == I40E_SUCCESS) {
9048                 clear_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9049         } else {
9050                 /* Enable link polling temporarily until i40e_get_link_status
9051                  * returns I40E_SUCCESS
9052                  */
9053                 set_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9054                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
9055                         status);
9056                 return;
9057         }
9058
9059         old_link_speed = pf->hw.phy.link_info_old.link_speed;
9060         new_link_speed = pf->hw.phy.link_info.link_speed;
9061
9062         if (new_link == old_link &&
9063             new_link_speed == old_link_speed &&
9064             (test_bit(__I40E_VSI_DOWN, vsi->state) ||
9065              new_link == netif_carrier_ok(vsi->netdev)))
9066                 return;
9067
9068         i40e_print_link_message(vsi, new_link);
9069
9070         /* Notify the base of the switch tree connected to
9071          * the link.  Floating VEBs are not notified.
9072          */
9073         if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
9074                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
9075         else
9076                 i40e_vsi_link_event(vsi, new_link);
9077
9078         if (pf->vf)
9079                 i40e_vc_notify_link_state(pf);
9080
9081         if (pf->flags & I40E_FLAG_PTP)
9082                 i40e_ptp_set_increment(pf);
9083 }
9084
9085 /**
9086  * i40e_watchdog_subtask - periodic checks not using event driven response
9087  * @pf: board private structure
9088  **/
9089 static void i40e_watchdog_subtask(struct i40e_pf *pf)
9090 {
9091         int i;
9092
9093         /* if interface is down do nothing */
9094         if (test_bit(__I40E_DOWN, pf->state) ||
9095             test_bit(__I40E_CONFIG_BUSY, pf->state))
9096                 return;
9097
9098         /* make sure we don't do these things too often */
9099         if (time_before(jiffies, (pf->service_timer_previous +
9100                                   pf->service_timer_period)))
9101                 return;
9102         pf->service_timer_previous = jiffies;
9103
9104         if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
9105             test_bit(__I40E_TEMP_LINK_POLLING, pf->state))
9106                 i40e_link_event(pf);
9107
9108         /* Update the stats for active netdevs so the network stack
9109          * can look at updated numbers whenever it cares to
9110          */
9111         for (i = 0; i < pf->num_alloc_vsi; i++)
9112                 if (pf->vsi[i] && pf->vsi[i]->netdev)
9113                         i40e_update_stats(pf->vsi[i]);
9114
9115         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
9116                 /* Update the stats for the active switching components */
9117                 for (i = 0; i < I40E_MAX_VEB; i++)
9118                         if (pf->veb[i])
9119                                 i40e_update_veb_stats(pf->veb[i]);
9120         }
9121
9122         i40e_ptp_rx_hang(pf);
9123         i40e_ptp_tx_hang(pf);
9124 }
9125
9126 /**
9127  * i40e_reset_subtask - Set up for resetting the device and driver
9128  * @pf: board private structure
9129  **/
9130 static void i40e_reset_subtask(struct i40e_pf *pf)
9131 {
9132         u32 reset_flags = 0;
9133
9134         if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
9135                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
9136                 clear_bit(__I40E_REINIT_REQUESTED, pf->state);
9137         }
9138         if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
9139                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
9140                 clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
9141         }
9142         if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
9143                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
9144                 clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
9145         }
9146         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
9147                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
9148                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
9149         }
9150         if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
9151                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
9152                 clear_bit(__I40E_DOWN_REQUESTED, pf->state);
9153         }
9154
9155         /* If there's a recovery already waiting, it takes
9156          * precedence before starting a new reset sequence.
9157          */
9158         if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
9159                 i40e_prep_for_reset(pf, false);
9160                 i40e_reset(pf);
9161                 i40e_rebuild(pf, false, false);
9162         }
9163
9164         /* If we're already down or resetting, just bail */
9165         if (reset_flags &&
9166             !test_bit(__I40E_DOWN, pf->state) &&
9167             !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
9168                 i40e_do_reset(pf, reset_flags, false);
9169         }
9170 }
9171
9172 /**
9173  * i40e_handle_link_event - Handle link event
9174  * @pf: board private structure
9175  * @e: event info posted on ARQ
9176  **/
9177 static void i40e_handle_link_event(struct i40e_pf *pf,
9178                                    struct i40e_arq_event_info *e)
9179 {
9180         struct i40e_aqc_get_link_status *status =
9181                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
9182
9183         /* Do a new status request to re-enable LSE reporting
9184          * and load new status information into the hw struct
9185          * This completely ignores any state information
9186          * in the ARQ event info, instead choosing to always
9187          * issue the AQ update link status command.
9188          */
9189         i40e_link_event(pf);
9190
9191         /* Check if module meets thermal requirements */
9192         if (status->phy_type == I40E_PHY_TYPE_NOT_SUPPORTED_HIGH_TEMP) {
9193                 dev_err(&pf->pdev->dev,
9194                         "Rx/Tx is disabled on this device because the module does not meet thermal requirements.\n");
9195                 dev_err(&pf->pdev->dev,
9196                         "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9197         } else {
9198                 /* check for unqualified module, if link is down, suppress
9199                  * the message if link was forced to be down.
9200                  */
9201                 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
9202                     (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
9203                     (!(status->link_info & I40E_AQ_LINK_UP)) &&
9204                     (!(pf->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED))) {
9205                         dev_err(&pf->pdev->dev,
9206                                 "Rx/Tx is disabled on this device because an unsupported SFP module type was detected.\n");
9207                         dev_err(&pf->pdev->dev,
9208                                 "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9209                 }
9210         }
9211 }
9212
9213 /**
9214  * i40e_clean_adminq_subtask - Clean the AdminQ rings
9215  * @pf: board private structure
9216  **/
9217 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
9218 {
9219         struct i40e_arq_event_info event;
9220         struct i40e_hw *hw = &pf->hw;
9221         u16 pending, i = 0;
9222         i40e_status ret;
9223         u16 opcode;
9224         u32 oldval;
9225         u32 val;
9226
9227         /* Do not run clean AQ when PF reset fails */
9228         if (test_bit(__I40E_RESET_FAILED, pf->state))
9229                 return;
9230
9231         /* check for error indications */
9232         val = rd32(&pf->hw, pf->hw.aq.arq.len);
9233         oldval = val;
9234         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
9235                 if (hw->debug_mask & I40E_DEBUG_AQ)
9236                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
9237                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
9238         }
9239         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
9240                 if (hw->debug_mask & I40E_DEBUG_AQ)
9241                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
9242                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
9243                 pf->arq_overflows++;
9244         }
9245         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
9246                 if (hw->debug_mask & I40E_DEBUG_AQ)
9247                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
9248                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
9249         }
9250         if (oldval != val)
9251                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
9252
9253         val = rd32(&pf->hw, pf->hw.aq.asq.len);
9254         oldval = val;
9255         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
9256                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9257                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
9258                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
9259         }
9260         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
9261                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9262                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
9263                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
9264         }
9265         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
9266                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9267                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
9268                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
9269         }
9270         if (oldval != val)
9271                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
9272
9273         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
9274         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
9275         if (!event.msg_buf)
9276                 return;
9277
9278         do {
9279                 ret = i40e_clean_arq_element(hw, &event, &pending);
9280                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
9281                         break;
9282                 else if (ret) {
9283                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
9284                         break;
9285                 }
9286
9287                 opcode = le16_to_cpu(event.desc.opcode);
9288                 switch (opcode) {
9289
9290                 case i40e_aqc_opc_get_link_status:
9291                         i40e_handle_link_event(pf, &event);
9292                         break;
9293                 case i40e_aqc_opc_send_msg_to_pf:
9294                         ret = i40e_vc_process_vf_msg(pf,
9295                                         le16_to_cpu(event.desc.retval),
9296                                         le32_to_cpu(event.desc.cookie_high),
9297                                         le32_to_cpu(event.desc.cookie_low),
9298                                         event.msg_buf,
9299                                         event.msg_len);
9300                         break;
9301                 case i40e_aqc_opc_lldp_update_mib:
9302                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
9303 #ifdef CONFIG_I40E_DCB
9304                         rtnl_lock();
9305                         ret = i40e_handle_lldp_event(pf, &event);
9306                         rtnl_unlock();
9307 #endif /* CONFIG_I40E_DCB */
9308                         break;
9309                 case i40e_aqc_opc_event_lan_overflow:
9310                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
9311                         i40e_handle_lan_overflow_event(pf, &event);
9312                         break;
9313                 case i40e_aqc_opc_send_msg_to_peer:
9314                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
9315                         break;
9316                 case i40e_aqc_opc_nvm_erase:
9317                 case i40e_aqc_opc_nvm_update:
9318                 case i40e_aqc_opc_oem_post_update:
9319                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
9320                                    "ARQ NVM operation 0x%04x completed\n",
9321                                    opcode);
9322                         break;
9323                 default:
9324                         dev_info(&pf->pdev->dev,
9325                                  "ARQ: Unknown event 0x%04x ignored\n",
9326                                  opcode);
9327                         break;
9328                 }
9329         } while (i++ < pf->adminq_work_limit);
9330
9331         if (i < pf->adminq_work_limit)
9332                 clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
9333
9334         /* re-enable Admin queue interrupt cause */
9335         val = rd32(hw, I40E_PFINT_ICR0_ENA);
9336         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
9337         wr32(hw, I40E_PFINT_ICR0_ENA, val);
9338         i40e_flush(hw);
9339
9340         kfree(event.msg_buf);
9341 }
9342
9343 /**
9344  * i40e_verify_eeprom - make sure eeprom is good to use
9345  * @pf: board private structure
9346  **/
9347 static void i40e_verify_eeprom(struct i40e_pf *pf)
9348 {
9349         int err;
9350
9351         err = i40e_diag_eeprom_test(&pf->hw);
9352         if (err) {
9353                 /* retry in case of garbage read */
9354                 err = i40e_diag_eeprom_test(&pf->hw);
9355                 if (err) {
9356                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
9357                                  err);
9358                         set_bit(__I40E_BAD_EEPROM, pf->state);
9359                 }
9360         }
9361
9362         if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
9363                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
9364                 clear_bit(__I40E_BAD_EEPROM, pf->state);
9365         }
9366 }
9367
9368 /**
9369  * i40e_enable_pf_switch_lb
9370  * @pf: pointer to the PF structure
9371  *
9372  * enable switch loop back or die - no point in a return value
9373  **/
9374 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
9375 {
9376         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9377         struct i40e_vsi_context ctxt;
9378         int ret;
9379
9380         ctxt.seid = pf->main_vsi_seid;
9381         ctxt.pf_num = pf->hw.pf_id;
9382         ctxt.vf_num = 0;
9383         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9384         if (ret) {
9385                 dev_info(&pf->pdev->dev,
9386                          "couldn't get PF vsi config, err %s aq_err %s\n",
9387                          i40e_stat_str(&pf->hw, ret),
9388                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9389                 return;
9390         }
9391         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9392         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9393         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9394
9395         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9396         if (ret) {
9397                 dev_info(&pf->pdev->dev,
9398                          "update vsi switch failed, err %s aq_err %s\n",
9399                          i40e_stat_str(&pf->hw, ret),
9400                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9401         }
9402 }
9403
9404 /**
9405  * i40e_disable_pf_switch_lb
9406  * @pf: pointer to the PF structure
9407  *
9408  * disable switch loop back or die - no point in a return value
9409  **/
9410 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
9411 {
9412         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9413         struct i40e_vsi_context ctxt;
9414         int ret;
9415
9416         ctxt.seid = pf->main_vsi_seid;
9417         ctxt.pf_num = pf->hw.pf_id;
9418         ctxt.vf_num = 0;
9419         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9420         if (ret) {
9421                 dev_info(&pf->pdev->dev,
9422                          "couldn't get PF vsi config, err %s aq_err %s\n",
9423                          i40e_stat_str(&pf->hw, ret),
9424                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9425                 return;
9426         }
9427         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9428         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9429         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9430
9431         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9432         if (ret) {
9433                 dev_info(&pf->pdev->dev,
9434                          "update vsi switch failed, err %s aq_err %s\n",
9435                          i40e_stat_str(&pf->hw, ret),
9436                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9437         }
9438 }
9439
9440 /**
9441  * i40e_config_bridge_mode - Configure the HW bridge mode
9442  * @veb: pointer to the bridge instance
9443  *
9444  * Configure the loop back mode for the LAN VSI that is downlink to the
9445  * specified HW bridge instance. It is expected this function is called
9446  * when a new HW bridge is instantiated.
9447  **/
9448 static void i40e_config_bridge_mode(struct i40e_veb *veb)
9449 {
9450         struct i40e_pf *pf = veb->pf;
9451
9452         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
9453                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
9454                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9455         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
9456                 i40e_disable_pf_switch_lb(pf);
9457         else
9458                 i40e_enable_pf_switch_lb(pf);
9459 }
9460
9461 /**
9462  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
9463  * @veb: pointer to the VEB instance
9464  *
9465  * This is a recursive function that first builds the attached VSIs then
9466  * recurses in to build the next layer of VEB.  We track the connections
9467  * through our own index numbers because the seid's from the HW could
9468  * change across the reset.
9469  **/
9470 static int i40e_reconstitute_veb(struct i40e_veb *veb)
9471 {
9472         struct i40e_vsi *ctl_vsi = NULL;
9473         struct i40e_pf *pf = veb->pf;
9474         int v, veb_idx;
9475         int ret;
9476
9477         /* build VSI that owns this VEB, temporarily attached to base VEB */
9478         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
9479                 if (pf->vsi[v] &&
9480                     pf->vsi[v]->veb_idx == veb->idx &&
9481                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
9482                         ctl_vsi = pf->vsi[v];
9483                         break;
9484                 }
9485         }
9486         if (!ctl_vsi) {
9487                 dev_info(&pf->pdev->dev,
9488                          "missing owner VSI for veb_idx %d\n", veb->idx);
9489                 ret = -ENOENT;
9490                 goto end_reconstitute;
9491         }
9492         if (ctl_vsi != pf->vsi[pf->lan_vsi])
9493                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9494         ret = i40e_add_vsi(ctl_vsi);
9495         if (ret) {
9496                 dev_info(&pf->pdev->dev,
9497                          "rebuild of veb_idx %d owner VSI failed: %d\n",
9498                          veb->idx, ret);
9499                 goto end_reconstitute;
9500         }
9501         i40e_vsi_reset_stats(ctl_vsi);
9502
9503         /* create the VEB in the switch and move the VSI onto the VEB */
9504         ret = i40e_add_veb(veb, ctl_vsi);
9505         if (ret)
9506                 goto end_reconstitute;
9507
9508         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
9509                 veb->bridge_mode = BRIDGE_MODE_VEB;
9510         else
9511                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9512         i40e_config_bridge_mode(veb);
9513
9514         /* create the remaining VSIs attached to this VEB */
9515         for (v = 0; v < pf->num_alloc_vsi; v++) {
9516                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
9517                         continue;
9518
9519                 if (pf->vsi[v]->veb_idx == veb->idx) {
9520                         struct i40e_vsi *vsi = pf->vsi[v];
9521
9522                         vsi->uplink_seid = veb->seid;
9523                         ret = i40e_add_vsi(vsi);
9524                         if (ret) {
9525                                 dev_info(&pf->pdev->dev,
9526                                          "rebuild of vsi_idx %d failed: %d\n",
9527                                          v, ret);
9528                                 goto end_reconstitute;
9529                         }
9530                         i40e_vsi_reset_stats(vsi);
9531                 }
9532         }
9533
9534         /* create any VEBs attached to this VEB - RECURSION */
9535         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9536                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
9537                         pf->veb[veb_idx]->uplink_seid = veb->seid;
9538                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
9539                         if (ret)
9540                                 break;
9541                 }
9542         }
9543
9544 end_reconstitute:
9545         return ret;
9546 }
9547
9548 /**
9549  * i40e_get_capabilities - get info about the HW
9550  * @pf: the PF struct
9551  **/
9552 static int i40e_get_capabilities(struct i40e_pf *pf,
9553                                  enum i40e_admin_queue_opc list_type)
9554 {
9555         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
9556         u16 data_size;
9557         int buf_len;
9558         int err;
9559
9560         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
9561         do {
9562                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
9563                 if (!cap_buf)
9564                         return -ENOMEM;
9565
9566                 /* this loads the data into the hw struct for us */
9567                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
9568                                                     &data_size, list_type,
9569                                                     NULL);
9570                 /* data loaded, buffer no longer needed */
9571                 kfree(cap_buf);
9572
9573                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
9574                         /* retry with a larger buffer */
9575                         buf_len = data_size;
9576                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
9577                         dev_info(&pf->pdev->dev,
9578                                  "capability discovery failed, err %s aq_err %s\n",
9579                                  i40e_stat_str(&pf->hw, err),
9580                                  i40e_aq_str(&pf->hw,
9581                                              pf->hw.aq.asq_last_status));
9582                         return -ENODEV;
9583                 }
9584         } while (err);
9585
9586         if (pf->hw.debug_mask & I40E_DEBUG_USER) {
9587                 if (list_type == i40e_aqc_opc_list_func_capabilities) {
9588                         dev_info(&pf->pdev->dev,
9589                                  "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",
9590                                  pf->hw.pf_id, pf->hw.func_caps.num_vfs,
9591                                  pf->hw.func_caps.num_msix_vectors,
9592                                  pf->hw.func_caps.num_msix_vectors_vf,
9593                                  pf->hw.func_caps.fd_filters_guaranteed,
9594                                  pf->hw.func_caps.fd_filters_best_effort,
9595                                  pf->hw.func_caps.num_tx_qp,
9596                                  pf->hw.func_caps.num_vsis);
9597                 } else if (list_type == i40e_aqc_opc_list_dev_capabilities) {
9598                         dev_info(&pf->pdev->dev,
9599                                  "switch_mode=0x%04x, function_valid=0x%08x\n",
9600                                  pf->hw.dev_caps.switch_mode,
9601                                  pf->hw.dev_caps.valid_functions);
9602                         dev_info(&pf->pdev->dev,
9603                                  "SR-IOV=%d, num_vfs for all function=%u\n",
9604                                  pf->hw.dev_caps.sr_iov_1_1,
9605                                  pf->hw.dev_caps.num_vfs);
9606                         dev_info(&pf->pdev->dev,
9607                                  "num_vsis=%u, num_rx:%u, num_tx=%u\n",
9608                                  pf->hw.dev_caps.num_vsis,
9609                                  pf->hw.dev_caps.num_rx_qp,
9610                                  pf->hw.dev_caps.num_tx_qp);
9611                 }
9612         }
9613         if (list_type == i40e_aqc_opc_list_func_capabilities) {
9614 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
9615                        + pf->hw.func_caps.num_vfs)
9616                 if (pf->hw.revision_id == 0 &&
9617                     pf->hw.func_caps.num_vsis < DEF_NUM_VSI) {
9618                         dev_info(&pf->pdev->dev,
9619                                  "got num_vsis %d, setting num_vsis to %d\n",
9620                                  pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
9621                         pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
9622                 }
9623         }
9624         return 0;
9625 }
9626
9627 static int i40e_vsi_clear(struct i40e_vsi *vsi);
9628
9629 /**
9630  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
9631  * @pf: board private structure
9632  **/
9633 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
9634 {
9635         struct i40e_vsi *vsi;
9636
9637         /* quick workaround for an NVM issue that leaves a critical register
9638          * uninitialized
9639          */
9640         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
9641                 static const u32 hkey[] = {
9642                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
9643                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
9644                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
9645                         0x95b3a76d};
9646                 int i;
9647
9648                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
9649                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
9650         }
9651
9652         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
9653                 return;
9654
9655         /* find existing VSI and see if it needs configuring */
9656         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9657
9658         /* create a new VSI if none exists */
9659         if (!vsi) {
9660                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
9661                                      pf->vsi[pf->lan_vsi]->seid, 0);
9662                 if (!vsi) {
9663                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
9664                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9665                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
9666                         return;
9667                 }
9668         }
9669
9670         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
9671 }
9672
9673 /**
9674  * i40e_fdir_teardown - release the Flow Director resources
9675  * @pf: board private structure
9676  **/
9677 static void i40e_fdir_teardown(struct i40e_pf *pf)
9678 {
9679         struct i40e_vsi *vsi;
9680
9681         i40e_fdir_filter_exit(pf);
9682         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9683         if (vsi)
9684                 i40e_vsi_release(vsi);
9685 }
9686
9687 /**
9688  * i40e_rebuild_cloud_filters - Rebuilds cloud filters for VSIs
9689  * @vsi: PF main vsi
9690  * @seid: seid of main or channel VSIs
9691  *
9692  * Rebuilds cloud filters associated with main VSI and channel VSIs if they
9693  * existed before reset
9694  **/
9695 static int i40e_rebuild_cloud_filters(struct i40e_vsi *vsi, u16 seid)
9696 {
9697         struct i40e_cloud_filter *cfilter;
9698         struct i40e_pf *pf = vsi->back;
9699         struct hlist_node *node;
9700         i40e_status ret;
9701
9702         /* Add cloud filters back if they exist */
9703         hlist_for_each_entry_safe(cfilter, node, &pf->cloud_filter_list,
9704                                   cloud_node) {
9705                 if (cfilter->seid != seid)
9706                         continue;
9707
9708                 if (cfilter->dst_port)
9709                         ret = i40e_add_del_cloud_filter_big_buf(vsi, cfilter,
9710                                                                 true);
9711                 else
9712                         ret = i40e_add_del_cloud_filter(vsi, cfilter, true);
9713
9714                 if (ret) {
9715                         dev_dbg(&pf->pdev->dev,
9716                                 "Failed to rebuild cloud filter, err %s aq_err %s\n",
9717                                 i40e_stat_str(&pf->hw, ret),
9718                                 i40e_aq_str(&pf->hw,
9719                                             pf->hw.aq.asq_last_status));
9720                         return ret;
9721                 }
9722         }
9723         return 0;
9724 }
9725
9726 /**
9727  * i40e_rebuild_channels - Rebuilds channel VSIs if they existed before reset
9728  * @vsi: PF main vsi
9729  *
9730  * Rebuilds channel VSIs if they existed before reset
9731  **/
9732 static int i40e_rebuild_channels(struct i40e_vsi *vsi)
9733 {
9734         struct i40e_channel *ch, *ch_tmp;
9735         i40e_status ret;
9736
9737         if (list_empty(&vsi->ch_list))
9738                 return 0;
9739
9740         list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
9741                 if (!ch->initialized)
9742                         break;
9743                 /* Proceed with creation of channel (VMDq2) VSI */
9744                 ret = i40e_add_channel(vsi->back, vsi->uplink_seid, ch);
9745                 if (ret) {
9746                         dev_info(&vsi->back->pdev->dev,
9747                                  "failed to rebuild channels using uplink_seid %u\n",
9748                                  vsi->uplink_seid);
9749                         return ret;
9750                 }
9751                 /* Reconfigure TX queues using QTX_CTL register */
9752                 ret = i40e_channel_config_tx_ring(vsi->back, vsi, ch);
9753                 if (ret) {
9754                         dev_info(&vsi->back->pdev->dev,
9755                                  "failed to configure TX rings for channel %u\n",
9756                                  ch->seid);
9757                         return ret;
9758                 }
9759                 /* update 'next_base_queue' */
9760                 vsi->next_base_queue = vsi->next_base_queue +
9761                                                         ch->num_queue_pairs;
9762                 if (ch->max_tx_rate) {
9763                         u64 credits = ch->max_tx_rate;
9764
9765                         if (i40e_set_bw_limit(vsi, ch->seid,
9766                                               ch->max_tx_rate))
9767                                 return -EINVAL;
9768
9769                         do_div(credits, I40E_BW_CREDIT_DIVISOR);
9770                         dev_dbg(&vsi->back->pdev->dev,
9771                                 "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
9772                                 ch->max_tx_rate,
9773                                 credits,
9774                                 ch->seid);
9775                 }
9776                 ret = i40e_rebuild_cloud_filters(vsi, ch->seid);
9777                 if (ret) {
9778                         dev_dbg(&vsi->back->pdev->dev,
9779                                 "Failed to rebuild cloud filters for channel VSI %u\n",
9780                                 ch->seid);
9781                         return ret;
9782                 }
9783         }
9784         return 0;
9785 }
9786
9787 /**
9788  * i40e_prep_for_reset - prep for the core to reset
9789  * @pf: board private structure
9790  * @lock_acquired: indicates whether or not the lock has been acquired
9791  * before this function was called.
9792  *
9793  * Close up the VFs and other things in prep for PF Reset.
9794   **/
9795 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
9796 {
9797         struct i40e_hw *hw = &pf->hw;
9798         i40e_status ret = 0;
9799         u32 v;
9800
9801         clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
9802         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
9803                 return;
9804         if (i40e_check_asq_alive(&pf->hw))
9805                 i40e_vc_notify_reset(pf);
9806
9807         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
9808
9809         /* quiesce the VSIs and their queues that are not already DOWN */
9810         /* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
9811         if (!lock_acquired)
9812                 rtnl_lock();
9813         i40e_pf_quiesce_all_vsi(pf);
9814         if (!lock_acquired)
9815                 rtnl_unlock();
9816
9817         for (v = 0; v < pf->num_alloc_vsi; v++) {
9818                 if (pf->vsi[v])
9819                         pf->vsi[v]->seid = 0;
9820         }
9821
9822         i40e_shutdown_adminq(&pf->hw);
9823
9824         /* call shutdown HMC */
9825         if (hw->hmc.hmc_obj) {
9826                 ret = i40e_shutdown_lan_hmc(hw);
9827                 if (ret)
9828                         dev_warn(&pf->pdev->dev,
9829                                  "shutdown_lan_hmc failed: %d\n", ret);
9830         }
9831
9832         /* Save the current PTP time so that we can restore the time after the
9833          * reset completes.
9834          */
9835         i40e_ptp_save_hw_time(pf);
9836 }
9837
9838 /**
9839  * i40e_send_version - update firmware with driver version
9840  * @pf: PF struct
9841  */
9842 static void i40e_send_version(struct i40e_pf *pf)
9843 {
9844         struct i40e_driver_version dv;
9845
9846         dv.major_version = DRV_VERSION_MAJOR;
9847         dv.minor_version = DRV_VERSION_MINOR;
9848         dv.build_version = DRV_VERSION_BUILD;
9849         dv.subbuild_version = 0;
9850         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
9851         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
9852 }
9853
9854 /**
9855  * i40e_get_oem_version - get OEM specific version information
9856  * @hw: pointer to the hardware structure
9857  **/
9858 static void i40e_get_oem_version(struct i40e_hw *hw)
9859 {
9860         u16 block_offset = 0xffff;
9861         u16 block_length = 0;
9862         u16 capabilities = 0;
9863         u16 gen_snap = 0;
9864         u16 release = 0;
9865
9866 #define I40E_SR_NVM_OEM_VERSION_PTR             0x1B
9867 #define I40E_NVM_OEM_LENGTH_OFFSET              0x00
9868 #define I40E_NVM_OEM_CAPABILITIES_OFFSET        0x01
9869 #define I40E_NVM_OEM_GEN_OFFSET                 0x02
9870 #define I40E_NVM_OEM_RELEASE_OFFSET             0x03
9871 #define I40E_NVM_OEM_CAPABILITIES_MASK          0x000F
9872 #define I40E_NVM_OEM_LENGTH                     3
9873
9874         /* Check if pointer to OEM version block is valid. */
9875         i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
9876         if (block_offset == 0xffff)
9877                 return;
9878
9879         /* Check if OEM version block has correct length. */
9880         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
9881                            &block_length);
9882         if (block_length < I40E_NVM_OEM_LENGTH)
9883                 return;
9884
9885         /* Check if OEM version format is as expected. */
9886         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
9887                            &capabilities);
9888         if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
9889                 return;
9890
9891         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
9892                            &gen_snap);
9893         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
9894                            &release);
9895         hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
9896         hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
9897 }
9898
9899 /**
9900  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
9901  * @pf: board private structure
9902  **/
9903 static int i40e_reset(struct i40e_pf *pf)
9904 {
9905         struct i40e_hw *hw = &pf->hw;
9906         i40e_status ret;
9907
9908         ret = i40e_pf_reset(hw);
9909         if (ret) {
9910                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
9911                 set_bit(__I40E_RESET_FAILED, pf->state);
9912                 clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
9913         } else {
9914                 pf->pfr_count++;
9915         }
9916         return ret;
9917 }
9918
9919 /**
9920  * i40e_rebuild - rebuild using a saved config
9921  * @pf: board private structure
9922  * @reinit: if the Main VSI needs to re-initialized.
9923  * @lock_acquired: indicates whether or not the lock has been acquired
9924  * before this function was called.
9925  **/
9926 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
9927 {
9928         int old_recovery_mode_bit = test_bit(__I40E_RECOVERY_MODE, pf->state);
9929         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9930         struct i40e_hw *hw = &pf->hw;
9931         u8 set_fc_aq_fail = 0;
9932         i40e_status ret;
9933         u32 val;
9934         int v;
9935
9936         if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9937             i40e_check_recovery_mode(pf)) {
9938                 i40e_set_ethtool_ops(pf->vsi[pf->lan_vsi]->netdev);
9939         }
9940
9941         if (test_bit(__I40E_DOWN, pf->state) &&
9942             !test_bit(__I40E_RECOVERY_MODE, pf->state) &&
9943             !old_recovery_mode_bit)
9944                 goto clear_recovery;
9945         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
9946
9947         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
9948         ret = i40e_init_adminq(&pf->hw);
9949         if (ret) {
9950                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
9951                          i40e_stat_str(&pf->hw, ret),
9952                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9953                 goto clear_recovery;
9954         }
9955         i40e_get_oem_version(&pf->hw);
9956
9957         if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9958             ((hw->aq.fw_maj_ver == 4 && hw->aq.fw_min_ver <= 33) ||
9959              hw->aq.fw_maj_ver < 4) && hw->mac.type == I40E_MAC_XL710) {
9960                 /* The following delay is necessary for 4.33 firmware and older
9961                  * to recover after EMP reset. 200 ms should suffice but we
9962                  * put here 300 ms to be sure that FW is ready to operate
9963                  * after reset.
9964                  */
9965                 mdelay(300);
9966         }
9967
9968         /* re-verify the eeprom if we just had an EMP reset */
9969         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
9970                 i40e_verify_eeprom(pf);
9971
9972         /* if we are going out of or into recovery mode we have to act
9973          * accordingly with regard to resources initialization
9974          * and deinitialization
9975          */
9976         if (test_bit(__I40E_RECOVERY_MODE, pf->state) ||
9977             old_recovery_mode_bit) {
9978                 if (i40e_get_capabilities(pf,
9979                                           i40e_aqc_opc_list_func_capabilities))
9980                         goto end_unlock;
9981
9982                 if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
9983                         /* we're staying in recovery mode so we'll reinitialize
9984                          * misc vector here
9985                          */
9986                         if (i40e_setup_misc_vector_for_recovery_mode(pf))
9987                                 goto end_unlock;
9988                 } else {
9989                         if (!lock_acquired)
9990                                 rtnl_lock();
9991                         /* we're going out of recovery mode so we'll free
9992                          * the IRQ allocated specifically for recovery mode
9993                          * and restore the interrupt scheme
9994                          */
9995                         free_irq(pf->pdev->irq, pf);
9996                         i40e_clear_interrupt_scheme(pf);
9997                         if (i40e_restore_interrupt_scheme(pf))
9998                                 goto end_unlock;
9999                 }
10000
10001                 /* tell the firmware that we're starting */
10002                 i40e_send_version(pf);
10003
10004                 /* bail out in case recovery mode was detected, as there is
10005                  * no need for further configuration.
10006                  */
10007                 goto end_unlock;
10008         }
10009
10010         i40e_clear_pxe_mode(hw);
10011         ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
10012         if (ret)
10013                 goto end_core_reset;
10014
10015         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10016                                 hw->func_caps.num_rx_qp, 0, 0);
10017         if (ret) {
10018                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
10019                 goto end_core_reset;
10020         }
10021         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10022         if (ret) {
10023                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
10024                 goto end_core_reset;
10025         }
10026
10027         /* Enable FW to write a default DCB config on link-up */
10028         i40e_aq_set_dcb_parameters(hw, true, NULL);
10029
10030 #ifdef CONFIG_I40E_DCB
10031         ret = i40e_init_pf_dcb(pf);
10032         if (ret) {
10033                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
10034                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10035                 /* Continue without DCB enabled */
10036         }
10037 #endif /* CONFIG_I40E_DCB */
10038         /* do basic switch setup */
10039         if (!lock_acquired)
10040                 rtnl_lock();
10041         ret = i40e_setup_pf_switch(pf, reinit);
10042         if (ret)
10043                 goto end_unlock;
10044
10045         /* The driver only wants link up/down and module qualification
10046          * reports from firmware.  Note the negative logic.
10047          */
10048         ret = i40e_aq_set_phy_int_mask(&pf->hw,
10049                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
10050                                          I40E_AQ_EVENT_MEDIA_NA |
10051                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
10052         if (ret)
10053                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10054                          i40e_stat_str(&pf->hw, ret),
10055                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10056
10057         /* make sure our flow control settings are restored */
10058         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
10059         if (ret)
10060                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
10061                         i40e_stat_str(&pf->hw, ret),
10062                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10063
10064         /* Rebuild the VSIs and VEBs that existed before reset.
10065          * They are still in our local switch element arrays, so only
10066          * need to rebuild the switch model in the HW.
10067          *
10068          * If there were VEBs but the reconstitution failed, we'll try
10069          * try to recover minimal use by getting the basic PF VSI working.
10070          */
10071         if (vsi->uplink_seid != pf->mac_seid) {
10072                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
10073                 /* find the one VEB connected to the MAC, and find orphans */
10074                 for (v = 0; v < I40E_MAX_VEB; v++) {
10075                         if (!pf->veb[v])
10076                                 continue;
10077
10078                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
10079                             pf->veb[v]->uplink_seid == 0) {
10080                                 ret = i40e_reconstitute_veb(pf->veb[v]);
10081
10082                                 if (!ret)
10083                                         continue;
10084
10085                                 /* If Main VEB failed, we're in deep doodoo,
10086                                  * so give up rebuilding the switch and set up
10087                                  * for minimal rebuild of PF VSI.
10088                                  * If orphan failed, we'll report the error
10089                                  * but try to keep going.
10090                                  */
10091                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
10092                                         dev_info(&pf->pdev->dev,
10093                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
10094                                                  ret);
10095                                         vsi->uplink_seid = pf->mac_seid;
10096                                         break;
10097                                 } else if (pf->veb[v]->uplink_seid == 0) {
10098                                         dev_info(&pf->pdev->dev,
10099                                                  "rebuild of orphan VEB failed: %d\n",
10100                                                  ret);
10101                                 }
10102                         }
10103                 }
10104         }
10105
10106         if (vsi->uplink_seid == pf->mac_seid) {
10107                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
10108                 /* no VEB, so rebuild only the Main VSI */
10109                 ret = i40e_add_vsi(vsi);
10110                 if (ret) {
10111                         dev_info(&pf->pdev->dev,
10112                                  "rebuild of Main VSI failed: %d\n", ret);
10113                         goto end_unlock;
10114                 }
10115         }
10116
10117         if (vsi->mqprio_qopt.max_rate[0]) {
10118                 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
10119                 u64 credits = 0;
10120
10121                 do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
10122                 ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
10123                 if (ret)
10124                         goto end_unlock;
10125
10126                 credits = max_tx_rate;
10127                 do_div(credits, I40E_BW_CREDIT_DIVISOR);
10128                 dev_dbg(&vsi->back->pdev->dev,
10129                         "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
10130                         max_tx_rate,
10131                         credits,
10132                         vsi->seid);
10133         }
10134
10135         ret = i40e_rebuild_cloud_filters(vsi, vsi->seid);
10136         if (ret)
10137                 goto end_unlock;
10138
10139         /* PF Main VSI is rebuild by now, go ahead and rebuild channel VSIs
10140          * for this main VSI if they exist
10141          */
10142         ret = i40e_rebuild_channels(vsi);
10143         if (ret)
10144                 goto end_unlock;
10145
10146         /* Reconfigure hardware for allowing smaller MSS in the case
10147          * of TSO, so that we avoid the MDD being fired and causing
10148          * a reset in the case of small MSS+TSO.
10149          */
10150 #define I40E_REG_MSS          0x000E64DC
10151 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
10152 #define I40E_64BYTE_MSS       0x400000
10153         val = rd32(hw, I40E_REG_MSS);
10154         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10155                 val &= ~I40E_REG_MSS_MIN_MASK;
10156                 val |= I40E_64BYTE_MSS;
10157                 wr32(hw, I40E_REG_MSS, val);
10158         }
10159
10160         if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
10161                 msleep(75);
10162                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10163                 if (ret)
10164                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10165                                  i40e_stat_str(&pf->hw, ret),
10166                                  i40e_aq_str(&pf->hw,
10167                                              pf->hw.aq.asq_last_status));
10168         }
10169         /* reinit the misc interrupt */
10170         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10171                 ret = i40e_setup_misc_vector(pf);
10172
10173         /* Add a filter to drop all Flow control frames from any VSI from being
10174          * transmitted. By doing so we stop a malicious VF from sending out
10175          * PAUSE or PFC frames and potentially controlling traffic for other
10176          * PF/VF VSIs.
10177          * The FW can still send Flow control frames if enabled.
10178          */
10179         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
10180                                                        pf->main_vsi_seid);
10181
10182         /* restart the VSIs that were rebuilt and running before the reset */
10183         i40e_pf_unquiesce_all_vsi(pf);
10184
10185         /* Release the RTNL lock before we start resetting VFs */
10186         if (!lock_acquired)
10187                 rtnl_unlock();
10188
10189         /* Restore promiscuous settings */
10190         ret = i40e_set_promiscuous(pf, pf->cur_promisc);
10191         if (ret)
10192                 dev_warn(&pf->pdev->dev,
10193                          "Failed to restore promiscuous setting: %s, err %s aq_err %s\n",
10194                          pf->cur_promisc ? "on" : "off",
10195                          i40e_stat_str(&pf->hw, ret),
10196                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10197
10198         i40e_reset_all_vfs(pf, true);
10199
10200         /* tell the firmware that we're starting */
10201         i40e_send_version(pf);
10202
10203         /* We've already released the lock, so don't do it again */
10204         goto end_core_reset;
10205
10206 end_unlock:
10207         if (!lock_acquired)
10208                 rtnl_unlock();
10209 end_core_reset:
10210         clear_bit(__I40E_RESET_FAILED, pf->state);
10211 clear_recovery:
10212         clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
10213         clear_bit(__I40E_TIMEOUT_RECOVERY_PENDING, pf->state);
10214 }
10215
10216 /**
10217  * i40e_reset_and_rebuild - reset and rebuild using a saved config
10218  * @pf: board private structure
10219  * @reinit: if the Main VSI needs to re-initialized.
10220  * @lock_acquired: indicates whether or not the lock has been acquired
10221  * before this function was called.
10222  **/
10223 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
10224                                    bool lock_acquired)
10225 {
10226         int ret;
10227         /* Now we wait for GRST to settle out.
10228          * We don't have to delete the VEBs or VSIs from the hw switch
10229          * because the reset will make them disappear.
10230          */
10231         ret = i40e_reset(pf);
10232         if (!ret)
10233                 i40e_rebuild(pf, reinit, lock_acquired);
10234 }
10235
10236 /**
10237  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
10238  * @pf: board private structure
10239  *
10240  * Close up the VFs and other things in prep for a Core Reset,
10241  * then get ready to rebuild the world.
10242  * @lock_acquired: indicates whether or not the lock has been acquired
10243  * before this function was called.
10244  **/
10245 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
10246 {
10247         i40e_prep_for_reset(pf, lock_acquired);
10248         i40e_reset_and_rebuild(pf, false, lock_acquired);
10249 }
10250
10251 /**
10252  * i40e_handle_mdd_event
10253  * @pf: pointer to the PF structure
10254  *
10255  * Called from the MDD irq handler to identify possibly malicious vfs
10256  **/
10257 static void i40e_handle_mdd_event(struct i40e_pf *pf)
10258 {
10259         struct i40e_hw *hw = &pf->hw;
10260         bool mdd_detected = false;
10261         struct i40e_vf *vf;
10262         u32 reg;
10263         int i;
10264
10265         if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
10266                 return;
10267
10268         /* find what triggered the MDD event */
10269         reg = rd32(hw, I40E_GL_MDET_TX);
10270         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
10271                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
10272                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
10273                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
10274                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
10275                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
10276                                 I40E_GL_MDET_TX_EVENT_SHIFT;
10277                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
10278                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
10279                                 pf->hw.func_caps.base_queue;
10280                 if (netif_msg_tx_err(pf))
10281                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
10282                                  event, queue, pf_num, vf_num);
10283                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
10284                 mdd_detected = true;
10285         }
10286         reg = rd32(hw, I40E_GL_MDET_RX);
10287         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
10288                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
10289                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
10290                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
10291                                 I40E_GL_MDET_RX_EVENT_SHIFT;
10292                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
10293                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
10294                                 pf->hw.func_caps.base_queue;
10295                 if (netif_msg_rx_err(pf))
10296                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
10297                                  event, queue, func);
10298                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
10299                 mdd_detected = true;
10300         }
10301
10302         if (mdd_detected) {
10303                 reg = rd32(hw, I40E_PF_MDET_TX);
10304                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
10305                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
10306                         dev_dbg(&pf->pdev->dev, "TX driver issue detected on PF\n");
10307                 }
10308                 reg = rd32(hw, I40E_PF_MDET_RX);
10309                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
10310                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
10311                         dev_dbg(&pf->pdev->dev, "RX driver issue detected on PF\n");
10312                 }
10313         }
10314
10315         /* see if one of the VFs needs its hand slapped */
10316         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
10317                 vf = &(pf->vf[i]);
10318                 reg = rd32(hw, I40E_VP_MDET_TX(i));
10319                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
10320                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
10321                         vf->num_mdd_events++;
10322                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
10323                                  i);
10324                         dev_info(&pf->pdev->dev,
10325                                  "Use PF Control I/F to re-enable the VF\n");
10326                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10327                 }
10328
10329                 reg = rd32(hw, I40E_VP_MDET_RX(i));
10330                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
10331                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
10332                         vf->num_mdd_events++;
10333                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
10334                                  i);
10335                         dev_info(&pf->pdev->dev,
10336                                  "Use PF Control I/F to re-enable the VF\n");
10337                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10338                 }
10339         }
10340
10341         /* re-enable mdd interrupt cause */
10342         clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
10343         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
10344         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
10345         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
10346         i40e_flush(hw);
10347 }
10348
10349 static const char *i40e_tunnel_name(u8 type)
10350 {
10351         switch (type) {
10352         case UDP_TUNNEL_TYPE_VXLAN:
10353                 return "vxlan";
10354         case UDP_TUNNEL_TYPE_GENEVE:
10355                 return "geneve";
10356         default:
10357                 return "unknown";
10358         }
10359 }
10360
10361 /**
10362  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
10363  * @pf: board private structure
10364  **/
10365 static void i40e_sync_udp_filters(struct i40e_pf *pf)
10366 {
10367         int i;
10368
10369         /* loop through and set pending bit for all active UDP filters */
10370         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10371                 if (pf->udp_ports[i].port)
10372                         pf->pending_udp_bitmap |= BIT_ULL(i);
10373         }
10374
10375         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
10376 }
10377
10378 /**
10379  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
10380  * @pf: board private structure
10381  **/
10382 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
10383 {
10384         struct i40e_hw *hw = &pf->hw;
10385         u8 filter_index, type;
10386         u16 port;
10387         int i;
10388
10389         if (!test_and_clear_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state))
10390                 return;
10391
10392         /* acquire RTNL to maintain state of flags and port requests */
10393         rtnl_lock();
10394
10395         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10396                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
10397                         struct i40e_udp_port_config *udp_port;
10398                         i40e_status ret = 0;
10399
10400                         udp_port = &pf->udp_ports[i];
10401                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
10402
10403                         port = READ_ONCE(udp_port->port);
10404                         type = READ_ONCE(udp_port->type);
10405                         filter_index = READ_ONCE(udp_port->filter_index);
10406
10407                         /* release RTNL while we wait on AQ command */
10408                         rtnl_unlock();
10409
10410                         if (port)
10411                                 ret = i40e_aq_add_udp_tunnel(hw, port,
10412                                                              type,
10413                                                              &filter_index,
10414                                                              NULL);
10415                         else if (filter_index != I40E_UDP_PORT_INDEX_UNUSED)
10416                                 ret = i40e_aq_del_udp_tunnel(hw, filter_index,
10417                                                              NULL);
10418
10419                         /* reacquire RTNL so we can update filter_index */
10420                         rtnl_lock();
10421
10422                         if (ret) {
10423                                 dev_info(&pf->pdev->dev,
10424                                          "%s %s port %d, index %d failed, err %s aq_err %s\n",
10425                                          i40e_tunnel_name(type),
10426                                          port ? "add" : "delete",
10427                                          port,
10428                                          filter_index,
10429                                          i40e_stat_str(&pf->hw, ret),
10430                                          i40e_aq_str(&pf->hw,
10431                                                      pf->hw.aq.asq_last_status));
10432                                 if (port) {
10433                                         /* failed to add, just reset port,
10434                                          * drop pending bit for any deletion
10435                                          */
10436                                         udp_port->port = 0;
10437                                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
10438                                 }
10439                         } else if (port) {
10440                                 /* record filter index on success */
10441                                 udp_port->filter_index = filter_index;
10442                         }
10443                 }
10444         }
10445
10446         rtnl_unlock();
10447 }
10448
10449 /**
10450  * i40e_service_task - Run the driver's async subtasks
10451  * @work: pointer to work_struct containing our data
10452  **/
10453 static void i40e_service_task(struct work_struct *work)
10454 {
10455         struct i40e_pf *pf = container_of(work,
10456                                           struct i40e_pf,
10457                                           service_task);
10458         unsigned long start_time = jiffies;
10459
10460         /* don't bother with service tasks if a reset is in progress */
10461         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
10462             test_bit(__I40E_SUSPENDED, pf->state))
10463                 return;
10464
10465         if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
10466                 return;
10467
10468         if (!test_bit(__I40E_RECOVERY_MODE, pf->state)) {
10469                 i40e_detect_recover_hung(pf->vsi[pf->lan_vsi]);
10470                 i40e_sync_filters_subtask(pf);
10471                 i40e_reset_subtask(pf);
10472                 i40e_handle_mdd_event(pf);
10473                 i40e_vc_process_vflr_event(pf);
10474                 i40e_watchdog_subtask(pf);
10475                 i40e_fdir_reinit_subtask(pf);
10476                 if (test_and_clear_bit(__I40E_CLIENT_RESET, pf->state)) {
10477                         /* Client subtask will reopen next time through. */
10478                         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi],
10479                                                            true);
10480                 } else {
10481                         i40e_client_subtask(pf);
10482                         if (test_and_clear_bit(__I40E_CLIENT_L2_CHANGE,
10483                                                pf->state))
10484                                 i40e_notify_client_of_l2_param_changes(
10485                                                                 pf->vsi[pf->lan_vsi]);
10486                 }
10487                 i40e_sync_filters_subtask(pf);
10488                 i40e_sync_udp_filters_subtask(pf);
10489         } else {
10490                 i40e_reset_subtask(pf);
10491         }
10492
10493         i40e_clean_adminq_subtask(pf);
10494
10495         /* flush memory to make sure state is correct before next watchdog */
10496         smp_mb__before_atomic();
10497         clear_bit(__I40E_SERVICE_SCHED, pf->state);
10498
10499         /* If the tasks have taken longer than one timer cycle or there
10500          * is more work to be done, reschedule the service task now
10501          * rather than wait for the timer to tick again.
10502          */
10503         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
10504             test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)             ||
10505             test_bit(__I40E_MDD_EVENT_PENDING, pf->state)                ||
10506             test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
10507                 i40e_service_event_schedule(pf);
10508 }
10509
10510 /**
10511  * i40e_service_timer - timer callback
10512  * @data: pointer to PF struct
10513  **/
10514 static void i40e_service_timer(struct timer_list *t)
10515 {
10516         struct i40e_pf *pf = from_timer(pf, t, service_timer);
10517
10518         mod_timer(&pf->service_timer,
10519                   round_jiffies(jiffies + pf->service_timer_period));
10520         i40e_service_event_schedule(pf);
10521 }
10522
10523 /**
10524  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
10525  * @vsi: the VSI being configured
10526  **/
10527 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
10528 {
10529         struct i40e_pf *pf = vsi->back;
10530
10531         switch (vsi->type) {
10532         case I40E_VSI_MAIN:
10533                 vsi->alloc_queue_pairs = pf->num_lan_qps;
10534                 if (!vsi->num_tx_desc)
10535                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10536                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10537                 if (!vsi->num_rx_desc)
10538                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10539                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10540                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10541                         vsi->num_q_vectors = pf->num_lan_msix;
10542                 else
10543                         vsi->num_q_vectors = 1;
10544
10545                 break;
10546
10547         case I40E_VSI_FDIR:
10548                 vsi->alloc_queue_pairs = 1;
10549                 vsi->num_tx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10550                                          I40E_REQ_DESCRIPTOR_MULTIPLE);
10551                 vsi->num_rx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10552                                          I40E_REQ_DESCRIPTOR_MULTIPLE);
10553                 vsi->num_q_vectors = pf->num_fdsb_msix;
10554                 break;
10555
10556         case I40E_VSI_VMDQ2:
10557                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
10558                 if (!vsi->num_tx_desc)
10559                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10560                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10561                 if (!vsi->num_rx_desc)
10562                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10563                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10564                 vsi->num_q_vectors = pf->num_vmdq_msix;
10565                 break;
10566
10567         case I40E_VSI_SRIOV:
10568                 vsi->alloc_queue_pairs = pf->num_vf_qps;
10569                 if (!vsi->num_tx_desc)
10570                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10571                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10572                 if (!vsi->num_rx_desc)
10573                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10574                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10575                 break;
10576
10577         default:
10578                 WARN_ON(1);
10579                 return -ENODATA;
10580         }
10581
10582         return 0;
10583 }
10584
10585 /**
10586  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
10587  * @vsi: VSI pointer
10588  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
10589  *
10590  * On error: returns error code (negative)
10591  * On success: returns 0
10592  **/
10593 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
10594 {
10595         struct i40e_ring **next_rings;
10596         int size;
10597         int ret = 0;
10598
10599         /* allocate memory for both Tx, XDP Tx and Rx ring pointers */
10600         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
10601                (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
10602         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
10603         if (!vsi->tx_rings)
10604                 return -ENOMEM;
10605         next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
10606         if (i40e_enabled_xdp_vsi(vsi)) {
10607                 vsi->xdp_rings = next_rings;
10608                 next_rings += vsi->alloc_queue_pairs;
10609         }
10610         vsi->rx_rings = next_rings;
10611
10612         if (alloc_qvectors) {
10613                 /* allocate memory for q_vector pointers */
10614                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
10615                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
10616                 if (!vsi->q_vectors) {
10617                         ret = -ENOMEM;
10618                         goto err_vectors;
10619                 }
10620         }
10621         return ret;
10622
10623 err_vectors:
10624         kfree(vsi->tx_rings);
10625         return ret;
10626 }
10627
10628 /**
10629  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
10630  * @pf: board private structure
10631  * @type: type of VSI
10632  *
10633  * On error: returns error code (negative)
10634  * On success: returns vsi index in PF (positive)
10635  **/
10636 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
10637 {
10638         int ret = -ENODEV;
10639         struct i40e_vsi *vsi;
10640         int vsi_idx;
10641         int i;
10642
10643         /* Need to protect the allocation of the VSIs at the PF level */
10644         mutex_lock(&pf->switch_mutex);
10645
10646         /* VSI list may be fragmented if VSI creation/destruction has
10647          * been happening.  We can afford to do a quick scan to look
10648          * for any free VSIs in the list.
10649          *
10650          * find next empty vsi slot, looping back around if necessary
10651          */
10652         i = pf->next_vsi;
10653         while (i < pf->num_alloc_vsi && pf->vsi[i])
10654                 i++;
10655         if (i >= pf->num_alloc_vsi) {
10656                 i = 0;
10657                 while (i < pf->next_vsi && pf->vsi[i])
10658                         i++;
10659         }
10660
10661         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
10662                 vsi_idx = i;             /* Found one! */
10663         } else {
10664                 ret = -ENODEV;
10665                 goto unlock_pf;  /* out of VSI slots! */
10666         }
10667         pf->next_vsi = ++i;
10668
10669         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
10670         if (!vsi) {
10671                 ret = -ENOMEM;
10672                 goto unlock_pf;
10673         }
10674         vsi->type = type;
10675         vsi->back = pf;
10676         set_bit(__I40E_VSI_DOWN, vsi->state);
10677         vsi->flags = 0;
10678         vsi->idx = vsi_idx;
10679         vsi->int_rate_limit = 0;
10680         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
10681                                 pf->rss_table_size : 64;
10682         vsi->netdev_registered = false;
10683         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
10684         hash_init(vsi->mac_filter_hash);
10685         vsi->irqs_ready = false;
10686
10687         if (type == I40E_VSI_MAIN) {
10688                 vsi->af_xdp_zc_qps = bitmap_zalloc(pf->num_lan_qps, GFP_KERNEL);
10689                 if (!vsi->af_xdp_zc_qps)
10690                         goto err_rings;
10691         }
10692
10693         ret = i40e_set_num_rings_in_vsi(vsi);
10694         if (ret)
10695                 goto err_rings;
10696
10697         ret = i40e_vsi_alloc_arrays(vsi, true);
10698         if (ret)
10699                 goto err_rings;
10700
10701         /* Setup default MSIX irq handler for VSI */
10702         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
10703
10704         /* Initialize VSI lock */
10705         spin_lock_init(&vsi->mac_filter_hash_lock);
10706         pf->vsi[vsi_idx] = vsi;
10707         ret = vsi_idx;
10708         goto unlock_pf;
10709
10710 err_rings:
10711         bitmap_free(vsi->af_xdp_zc_qps);
10712         pf->next_vsi = i - 1;
10713         kfree(vsi);
10714 unlock_pf:
10715         mutex_unlock(&pf->switch_mutex);
10716         return ret;
10717 }
10718
10719 /**
10720  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
10721  * @vsi: VSI pointer
10722  * @free_qvectors: a bool to specify if q_vectors need to be freed.
10723  *
10724  * On error: returns error code (negative)
10725  * On success: returns 0
10726  **/
10727 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
10728 {
10729         /* free the ring and vector containers */
10730         if (free_qvectors) {
10731                 kfree(vsi->q_vectors);
10732                 vsi->q_vectors = NULL;
10733         }
10734         kfree(vsi->tx_rings);
10735         vsi->tx_rings = NULL;
10736         vsi->rx_rings = NULL;
10737         vsi->xdp_rings = NULL;
10738 }
10739
10740 /**
10741  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
10742  * and lookup table
10743  * @vsi: Pointer to VSI structure
10744  */
10745 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
10746 {
10747         if (!vsi)
10748                 return;
10749
10750         kfree(vsi->rss_hkey_user);
10751         vsi->rss_hkey_user = NULL;
10752
10753         kfree(vsi->rss_lut_user);
10754         vsi->rss_lut_user = NULL;
10755 }
10756
10757 /**
10758  * i40e_vsi_clear - Deallocate the VSI provided
10759  * @vsi: the VSI being un-configured
10760  **/
10761 static int i40e_vsi_clear(struct i40e_vsi *vsi)
10762 {
10763         struct i40e_pf *pf;
10764
10765         if (!vsi)
10766                 return 0;
10767
10768         if (!vsi->back)
10769                 goto free_vsi;
10770         pf = vsi->back;
10771
10772         mutex_lock(&pf->switch_mutex);
10773         if (!pf->vsi[vsi->idx]) {
10774                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](type %d)\n",
10775                         vsi->idx, vsi->idx, vsi->type);
10776                 goto unlock_vsi;
10777         }
10778
10779         if (pf->vsi[vsi->idx] != vsi) {
10780                 dev_err(&pf->pdev->dev,
10781                         "pf->vsi[%d](type %d) != vsi[%d](type %d): no free!\n",
10782                         pf->vsi[vsi->idx]->idx,
10783                         pf->vsi[vsi->idx]->type,
10784                         vsi->idx, vsi->type);
10785                 goto unlock_vsi;
10786         }
10787
10788         /* updates the PF for this cleared vsi */
10789         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
10790         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
10791
10792         bitmap_free(vsi->af_xdp_zc_qps);
10793         i40e_vsi_free_arrays(vsi, true);
10794         i40e_clear_rss_config_user(vsi);
10795
10796         pf->vsi[vsi->idx] = NULL;
10797         if (vsi->idx < pf->next_vsi)
10798                 pf->next_vsi = vsi->idx;
10799
10800 unlock_vsi:
10801         mutex_unlock(&pf->switch_mutex);
10802 free_vsi:
10803         kfree(vsi);
10804
10805         return 0;
10806 }
10807
10808 /**
10809  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
10810  * @vsi: the VSI being cleaned
10811  **/
10812 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
10813 {
10814         int i;
10815
10816         if (vsi->tx_rings && vsi->tx_rings[0]) {
10817                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10818                         kfree_rcu(vsi->tx_rings[i], rcu);
10819                         vsi->tx_rings[i] = NULL;
10820                         vsi->rx_rings[i] = NULL;
10821                         if (vsi->xdp_rings)
10822                                 vsi->xdp_rings[i] = NULL;
10823                 }
10824         }
10825 }
10826
10827 /**
10828  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
10829  * @vsi: the VSI being configured
10830  **/
10831 static int i40e_alloc_rings(struct i40e_vsi *vsi)
10832 {
10833         int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
10834         struct i40e_pf *pf = vsi->back;
10835         struct i40e_ring *ring;
10836
10837         /* Set basic values in the rings to be used later during open() */
10838         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10839                 /* allocate space for both Tx and Rx in one shot */
10840                 ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
10841                 if (!ring)
10842                         goto err_out;
10843
10844                 ring->queue_index = i;
10845                 ring->reg_idx = vsi->base_queue + i;
10846                 ring->ring_active = false;
10847                 ring->vsi = vsi;
10848                 ring->netdev = vsi->netdev;
10849                 ring->dev = &pf->pdev->dev;
10850                 ring->count = vsi->num_tx_desc;
10851                 ring->size = 0;
10852                 ring->dcb_tc = 0;
10853                 if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10854                         ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10855                 ring->itr_setting = pf->tx_itr_default;
10856                 vsi->tx_rings[i] = ring++;
10857
10858                 if (!i40e_enabled_xdp_vsi(vsi))
10859                         goto setup_rx;
10860
10861                 ring->queue_index = vsi->alloc_queue_pairs + i;
10862                 ring->reg_idx = vsi->base_queue + ring->queue_index;
10863                 ring->ring_active = false;
10864                 ring->vsi = vsi;
10865                 ring->netdev = NULL;
10866                 ring->dev = &pf->pdev->dev;
10867                 ring->count = vsi->num_tx_desc;
10868                 ring->size = 0;
10869                 ring->dcb_tc = 0;
10870                 if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10871                         ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10872                 set_ring_xdp(ring);
10873                 ring->itr_setting = pf->tx_itr_default;
10874                 vsi->xdp_rings[i] = ring++;
10875
10876 setup_rx:
10877                 ring->queue_index = i;
10878                 ring->reg_idx = vsi->base_queue + i;
10879                 ring->ring_active = false;
10880                 ring->vsi = vsi;
10881                 ring->netdev = vsi->netdev;
10882                 ring->dev = &pf->pdev->dev;
10883                 ring->count = vsi->num_rx_desc;
10884                 ring->size = 0;
10885                 ring->dcb_tc = 0;
10886                 ring->itr_setting = pf->rx_itr_default;
10887                 vsi->rx_rings[i] = ring;
10888         }
10889
10890         return 0;
10891
10892 err_out:
10893         i40e_vsi_clear_rings(vsi);
10894         return -ENOMEM;
10895 }
10896
10897 /**
10898  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
10899  * @pf: board private structure
10900  * @vectors: the number of MSI-X vectors to request
10901  *
10902  * Returns the number of vectors reserved, or error
10903  **/
10904 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
10905 {
10906         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
10907                                         I40E_MIN_MSIX, vectors);
10908         if (vectors < 0) {
10909                 dev_info(&pf->pdev->dev,
10910                          "MSI-X vector reservation failed: %d\n", vectors);
10911                 vectors = 0;
10912         }
10913
10914         return vectors;
10915 }
10916
10917 /**
10918  * i40e_init_msix - Setup the MSIX capability
10919  * @pf: board private structure
10920  *
10921  * Work with the OS to set up the MSIX vectors needed.
10922  *
10923  * Returns the number of vectors reserved or negative on failure
10924  **/
10925 static int i40e_init_msix(struct i40e_pf *pf)
10926 {
10927         struct i40e_hw *hw = &pf->hw;
10928         int cpus, extra_vectors;
10929         int vectors_left;
10930         int v_budget, i;
10931         int v_actual;
10932         int iwarp_requested = 0;
10933
10934         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10935                 return -ENODEV;
10936
10937         /* The number of vectors we'll request will be comprised of:
10938          *   - Add 1 for "other" cause for Admin Queue events, etc.
10939          *   - The number of LAN queue pairs
10940          *      - Queues being used for RSS.
10941          *              We don't need as many as max_rss_size vectors.
10942          *              use rss_size instead in the calculation since that
10943          *              is governed by number of cpus in the system.
10944          *      - assumes symmetric Tx/Rx pairing
10945          *   - The number of VMDq pairs
10946          *   - The CPU count within the NUMA node if iWARP is enabled
10947          * Once we count this up, try the request.
10948          *
10949          * If we can't get what we want, we'll simplify to nearly nothing
10950          * and try again.  If that still fails, we punt.
10951          */
10952         vectors_left = hw->func_caps.num_msix_vectors;
10953         v_budget = 0;
10954
10955         /* reserve one vector for miscellaneous handler */
10956         if (vectors_left) {
10957                 v_budget++;
10958                 vectors_left--;
10959         }
10960
10961         /* reserve some vectors for the main PF traffic queues. Initially we
10962          * only reserve at most 50% of the available vectors, in the case that
10963          * the number of online CPUs is large. This ensures that we can enable
10964          * extra features as well. Once we've enabled the other features, we
10965          * will use any remaining vectors to reach as close as we can to the
10966          * number of online CPUs.
10967          */
10968         cpus = num_online_cpus();
10969         pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
10970         vectors_left -= pf->num_lan_msix;
10971
10972         /* reserve one vector for sideband flow director */
10973         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10974                 if (vectors_left) {
10975                         pf->num_fdsb_msix = 1;
10976                         v_budget++;
10977                         vectors_left--;
10978                 } else {
10979                         pf->num_fdsb_msix = 0;
10980                 }
10981         }
10982
10983         /* can we reserve enough for iWARP? */
10984         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10985                 iwarp_requested = pf->num_iwarp_msix;
10986
10987                 if (!vectors_left)
10988                         pf->num_iwarp_msix = 0;
10989                 else if (vectors_left < pf->num_iwarp_msix)
10990                         pf->num_iwarp_msix = 1;
10991                 v_budget += pf->num_iwarp_msix;
10992                 vectors_left -= pf->num_iwarp_msix;
10993         }
10994
10995         /* any vectors left over go for VMDq support */
10996         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
10997                 if (!vectors_left) {
10998                         pf->num_vmdq_msix = 0;
10999                         pf->num_vmdq_qps = 0;
11000                 } else {
11001                         int vmdq_vecs_wanted =
11002                                 pf->num_vmdq_vsis * pf->num_vmdq_qps;
11003                         int vmdq_vecs =
11004                                 min_t(int, vectors_left, vmdq_vecs_wanted);
11005
11006                         /* if we're short on vectors for what's desired, we limit
11007                          * the queues per vmdq.  If this is still more than are
11008                          * available, the user will need to change the number of
11009                          * queues/vectors used by the PF later with the ethtool
11010                          * channels command
11011                          */
11012                         if (vectors_left < vmdq_vecs_wanted) {
11013                                 pf->num_vmdq_qps = 1;
11014                                 vmdq_vecs_wanted = pf->num_vmdq_vsis;
11015                                 vmdq_vecs = min_t(int,
11016                                                   vectors_left,
11017                                                   vmdq_vecs_wanted);
11018                         }
11019                         pf->num_vmdq_msix = pf->num_vmdq_qps;
11020
11021                         v_budget += vmdq_vecs;
11022                         vectors_left -= vmdq_vecs;
11023                 }
11024         }
11025
11026         /* On systems with a large number of SMP cores, we previously limited
11027          * the number of vectors for num_lan_msix to be at most 50% of the
11028          * available vectors, to allow for other features. Now, we add back
11029          * the remaining vectors. However, we ensure that the total
11030          * num_lan_msix will not exceed num_online_cpus(). To do this, we
11031          * calculate the number of vectors we can add without going over the
11032          * cap of CPUs. For systems with a small number of CPUs this will be
11033          * zero.
11034          */
11035         extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
11036         pf->num_lan_msix += extra_vectors;
11037         vectors_left -= extra_vectors;
11038
11039         WARN(vectors_left < 0,
11040              "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
11041
11042         v_budget += pf->num_lan_msix;
11043         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
11044                                    GFP_KERNEL);
11045         if (!pf->msix_entries)
11046                 return -ENOMEM;
11047
11048         for (i = 0; i < v_budget; i++)
11049                 pf->msix_entries[i].entry = i;
11050         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
11051
11052         if (v_actual < I40E_MIN_MSIX) {
11053                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
11054                 kfree(pf->msix_entries);
11055                 pf->msix_entries = NULL;
11056                 pci_disable_msix(pf->pdev);
11057                 return -ENODEV;
11058
11059         } else if (v_actual == I40E_MIN_MSIX) {
11060                 /* Adjust for minimal MSIX use */
11061                 pf->num_vmdq_vsis = 0;
11062                 pf->num_vmdq_qps = 0;
11063                 pf->num_lan_qps = 1;
11064                 pf->num_lan_msix = 1;
11065
11066         } else if (v_actual != v_budget) {
11067                 /* If we have limited resources, we will start with no vectors
11068                  * for the special features and then allocate vectors to some
11069                  * of these features based on the policy and at the end disable
11070                  * the features that did not get any vectors.
11071                  */
11072                 int vec;
11073
11074                 dev_info(&pf->pdev->dev,
11075                          "MSI-X vector limit reached with %d, wanted %d, attempting to redistribute vectors\n",
11076                          v_actual, v_budget);
11077                 /* reserve the misc vector */
11078                 vec = v_actual - 1;
11079
11080                 /* Scale vector usage down */
11081                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
11082                 pf->num_vmdq_vsis = 1;
11083                 pf->num_vmdq_qps = 1;
11084
11085                 /* partition out the remaining vectors */
11086                 switch (vec) {
11087                 case 2:
11088                         pf->num_lan_msix = 1;
11089                         break;
11090                 case 3:
11091                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11092                                 pf->num_lan_msix = 1;
11093                                 pf->num_iwarp_msix = 1;
11094                         } else {
11095                                 pf->num_lan_msix = 2;
11096                         }
11097                         break;
11098                 default:
11099                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11100                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
11101                                                  iwarp_requested);
11102                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
11103                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
11104                         } else {
11105                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
11106                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
11107                         }
11108                         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11109                                 pf->num_fdsb_msix = 1;
11110                                 vec--;
11111                         }
11112                         pf->num_lan_msix = min_t(int,
11113                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
11114                                                               pf->num_lan_msix);
11115                         pf->num_lan_qps = pf->num_lan_msix;
11116                         break;
11117                 }
11118         }
11119
11120         if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
11121             (pf->num_fdsb_msix == 0)) {
11122                 dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
11123                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
11124                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11125         }
11126         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
11127             (pf->num_vmdq_msix == 0)) {
11128                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
11129                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
11130         }
11131
11132         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
11133             (pf->num_iwarp_msix == 0)) {
11134                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
11135                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11136         }
11137         i40e_debug(&pf->hw, I40E_DEBUG_INIT,
11138                    "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
11139                    pf->num_lan_msix,
11140                    pf->num_vmdq_msix * pf->num_vmdq_vsis,
11141                    pf->num_fdsb_msix,
11142                    pf->num_iwarp_msix);
11143
11144         return v_actual;
11145 }
11146
11147 /**
11148  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
11149  * @vsi: the VSI being configured
11150  * @v_idx: index of the vector in the vsi struct
11151  * @cpu: cpu to be used on affinity_mask
11152  *
11153  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
11154  **/
11155 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
11156 {
11157         struct i40e_q_vector *q_vector;
11158
11159         /* allocate q_vector */
11160         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
11161         if (!q_vector)
11162                 return -ENOMEM;
11163
11164         q_vector->vsi = vsi;
11165         q_vector->v_idx = v_idx;
11166         cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
11167
11168         if (vsi->netdev)
11169                 netif_napi_add(vsi->netdev, &q_vector->napi,
11170                                i40e_napi_poll, NAPI_POLL_WEIGHT);
11171
11172         /* tie q_vector and vsi together */
11173         vsi->q_vectors[v_idx] = q_vector;
11174
11175         return 0;
11176 }
11177
11178 /**
11179  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
11180  * @vsi: the VSI being configured
11181  *
11182  * We allocate one q_vector per queue interrupt.  If allocation fails we
11183  * return -ENOMEM.
11184  **/
11185 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
11186 {
11187         struct i40e_pf *pf = vsi->back;
11188         int err, v_idx, num_q_vectors, current_cpu;
11189
11190         /* if not MSIX, give the one vector only to the LAN VSI */
11191         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
11192                 num_q_vectors = vsi->num_q_vectors;
11193         else if (vsi == pf->vsi[pf->lan_vsi])
11194                 num_q_vectors = 1;
11195         else
11196                 return -EINVAL;
11197
11198         current_cpu = cpumask_first(cpu_online_mask);
11199
11200         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
11201                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
11202                 if (err)
11203                         goto err_out;
11204                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
11205                 if (unlikely(current_cpu >= nr_cpu_ids))
11206                         current_cpu = cpumask_first(cpu_online_mask);
11207         }
11208
11209         return 0;
11210
11211 err_out:
11212         while (v_idx--)
11213                 i40e_free_q_vector(vsi, v_idx);
11214
11215         return err;
11216 }
11217
11218 /**
11219  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
11220  * @pf: board private structure to initialize
11221  **/
11222 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
11223 {
11224         int vectors = 0;
11225         ssize_t size;
11226
11227         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11228                 vectors = i40e_init_msix(pf);
11229                 if (vectors < 0) {
11230                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
11231                                        I40E_FLAG_IWARP_ENABLED  |
11232                                        I40E_FLAG_RSS_ENABLED    |
11233                                        I40E_FLAG_DCB_CAPABLE    |
11234                                        I40E_FLAG_DCB_ENABLED    |
11235                                        I40E_FLAG_SRIOV_ENABLED  |
11236                                        I40E_FLAG_FD_SB_ENABLED  |
11237                                        I40E_FLAG_FD_ATR_ENABLED |
11238                                        I40E_FLAG_VMDQ_ENABLED);
11239                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11240
11241                         /* rework the queue expectations without MSIX */
11242                         i40e_determine_queue_usage(pf);
11243                 }
11244         }
11245
11246         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11247             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
11248                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
11249                 vectors = pci_enable_msi(pf->pdev);
11250                 if (vectors < 0) {
11251                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
11252                                  vectors);
11253                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
11254                 }
11255                 vectors = 1;  /* one MSI or Legacy vector */
11256         }
11257
11258         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
11259                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
11260
11261         /* set up vector assignment tracking */
11262         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
11263         pf->irq_pile = kzalloc(size, GFP_KERNEL);
11264         if (!pf->irq_pile)
11265                 return -ENOMEM;
11266
11267         pf->irq_pile->num_entries = vectors;
11268         pf->irq_pile->search_hint = 0;
11269
11270         /* track first vector for misc interrupts, ignore return */
11271         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
11272
11273         return 0;
11274 }
11275
11276 /**
11277  * i40e_restore_interrupt_scheme - Restore the interrupt scheme
11278  * @pf: private board data structure
11279  *
11280  * Restore the interrupt scheme that was cleared when we suspended the
11281  * device. This should be called during resume to re-allocate the q_vectors
11282  * and reacquire IRQs.
11283  */
11284 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
11285 {
11286         int err, i;
11287
11288         /* We cleared the MSI and MSI-X flags when disabling the old interrupt
11289          * scheme. We need to re-enabled them here in order to attempt to
11290          * re-acquire the MSI or MSI-X vectors
11291          */
11292         pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
11293
11294         err = i40e_init_interrupt_scheme(pf);
11295         if (err)
11296                 return err;
11297
11298         /* Now that we've re-acquired IRQs, we need to remap the vectors and
11299          * rings together again.
11300          */
11301         for (i = 0; i < pf->num_alloc_vsi; i++) {
11302                 if (pf->vsi[i]) {
11303                         err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
11304                         if (err)
11305                                 goto err_unwind;
11306                         i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
11307                 }
11308         }
11309
11310         err = i40e_setup_misc_vector(pf);
11311         if (err)
11312                 goto err_unwind;
11313
11314         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
11315                 i40e_client_update_msix_info(pf);
11316
11317         return 0;
11318
11319 err_unwind:
11320         while (i--) {
11321                 if (pf->vsi[i])
11322                         i40e_vsi_free_q_vectors(pf->vsi[i]);
11323         }
11324
11325         return err;
11326 }
11327
11328 /**
11329  * i40e_setup_misc_vector_for_recovery_mode - Setup the misc vector to handle
11330  * non queue events in recovery mode
11331  * @pf: board private structure
11332  *
11333  * This sets up the handler for MSIX 0 or MSI/legacy, which is used to manage
11334  * the non-queue interrupts, e.g. AdminQ and errors in recovery mode.
11335  * This is handled differently than in recovery mode since no Tx/Rx resources
11336  * are being allocated.
11337  **/
11338 static int i40e_setup_misc_vector_for_recovery_mode(struct i40e_pf *pf)
11339 {
11340         int err;
11341
11342         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11343                 err = i40e_setup_misc_vector(pf);
11344
11345                 if (err) {
11346                         dev_info(&pf->pdev->dev,
11347                                  "MSI-X misc vector request failed, error %d\n",
11348                                  err);
11349                         return err;
11350                 }
11351         } else {
11352                 u32 flags = pf->flags & I40E_FLAG_MSI_ENABLED ? 0 : IRQF_SHARED;
11353
11354                 err = request_irq(pf->pdev->irq, i40e_intr, flags,
11355                                   pf->int_name, pf);
11356
11357                 if (err) {
11358                         dev_info(&pf->pdev->dev,
11359                                  "MSI/legacy misc vector request failed, error %d\n",
11360                                  err);
11361                         return err;
11362                 }
11363                 i40e_enable_misc_int_causes(pf);
11364                 i40e_irq_dynamic_enable_icr0(pf);
11365         }
11366
11367         return 0;
11368 }
11369
11370 /**
11371  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
11372  * @pf: board private structure
11373  *
11374  * This sets up the handler for MSIX 0, which is used to manage the
11375  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
11376  * when in MSI or Legacy interrupt mode.
11377  **/
11378 static int i40e_setup_misc_vector(struct i40e_pf *pf)
11379 {
11380         struct i40e_hw *hw = &pf->hw;
11381         int err = 0;
11382
11383         /* Only request the IRQ once, the first time through. */
11384         if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
11385                 err = request_irq(pf->msix_entries[0].vector,
11386                                   i40e_intr, 0, pf->int_name, pf);
11387                 if (err) {
11388                         clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
11389                         dev_info(&pf->pdev->dev,
11390                                  "request_irq for %s failed: %d\n",
11391                                  pf->int_name, err);
11392                         return -EFAULT;
11393                 }
11394         }
11395
11396         i40e_enable_misc_int_causes(pf);
11397
11398         /* associate no queues to the misc vector */
11399         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
11400         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K >> 1);
11401
11402         i40e_flush(hw);
11403
11404         i40e_irq_dynamic_enable_icr0(pf);
11405
11406         return err;
11407 }
11408
11409 /**
11410  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
11411  * @vsi: Pointer to vsi structure
11412  * @seed: Buffter to store the hash keys
11413  * @lut: Buffer to store the lookup table entries
11414  * @lut_size: Size of buffer to store the lookup table entries
11415  *
11416  * Return 0 on success, negative on failure
11417  */
11418 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
11419                            u8 *lut, u16 lut_size)
11420 {
11421         struct i40e_pf *pf = vsi->back;
11422         struct i40e_hw *hw = &pf->hw;
11423         int ret = 0;
11424
11425         if (seed) {
11426                 ret = i40e_aq_get_rss_key(hw, vsi->id,
11427                         (struct i40e_aqc_get_set_rss_key_data *)seed);
11428                 if (ret) {
11429                         dev_info(&pf->pdev->dev,
11430                                  "Cannot get RSS key, err %s aq_err %s\n",
11431                                  i40e_stat_str(&pf->hw, ret),
11432                                  i40e_aq_str(&pf->hw,
11433                                              pf->hw.aq.asq_last_status));
11434                         return ret;
11435                 }
11436         }
11437
11438         if (lut) {
11439                 bool pf_lut = vsi->type == I40E_VSI_MAIN;
11440
11441                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
11442                 if (ret) {
11443                         dev_info(&pf->pdev->dev,
11444                                  "Cannot get RSS lut, err %s aq_err %s\n",
11445                                  i40e_stat_str(&pf->hw, ret),
11446                                  i40e_aq_str(&pf->hw,
11447                                              pf->hw.aq.asq_last_status));
11448                         return ret;
11449                 }
11450         }
11451
11452         return ret;
11453 }
11454
11455 /**
11456  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
11457  * @vsi: Pointer to vsi structure
11458  * @seed: RSS hash seed
11459  * @lut: Lookup table
11460  * @lut_size: Lookup table size
11461  *
11462  * Returns 0 on success, negative on failure
11463  **/
11464 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
11465                                const u8 *lut, u16 lut_size)
11466 {
11467         struct i40e_pf *pf = vsi->back;
11468         struct i40e_hw *hw = &pf->hw;
11469         u16 vf_id = vsi->vf_id;
11470         u8 i;
11471
11472         /* Fill out hash function seed */
11473         if (seed) {
11474                 u32 *seed_dw = (u32 *)seed;
11475
11476                 if (vsi->type == I40E_VSI_MAIN) {
11477                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11478                                 wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
11479                 } else if (vsi->type == I40E_VSI_SRIOV) {
11480                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
11481                                 wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
11482                 } else {
11483                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
11484                 }
11485         }
11486
11487         if (lut) {
11488                 u32 *lut_dw = (u32 *)lut;
11489
11490                 if (vsi->type == I40E_VSI_MAIN) {
11491                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
11492                                 return -EINVAL;
11493                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11494                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
11495                 } else if (vsi->type == I40E_VSI_SRIOV) {
11496                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
11497                                 return -EINVAL;
11498                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
11499                                 wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
11500                 } else {
11501                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
11502                 }
11503         }
11504         i40e_flush(hw);
11505
11506         return 0;
11507 }
11508
11509 /**
11510  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
11511  * @vsi: Pointer to VSI structure
11512  * @seed: Buffer to store the keys
11513  * @lut: Buffer to store the lookup table entries
11514  * @lut_size: Size of buffer to store the lookup table entries
11515  *
11516  * Returns 0 on success, negative on failure
11517  */
11518 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
11519                             u8 *lut, u16 lut_size)
11520 {
11521         struct i40e_pf *pf = vsi->back;
11522         struct i40e_hw *hw = &pf->hw;
11523         u16 i;
11524
11525         if (seed) {
11526                 u32 *seed_dw = (u32 *)seed;
11527
11528                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11529                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
11530         }
11531         if (lut) {
11532                 u32 *lut_dw = (u32 *)lut;
11533
11534                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
11535                         return -EINVAL;
11536                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11537                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
11538         }
11539
11540         return 0;
11541 }
11542
11543 /**
11544  * i40e_config_rss - Configure RSS keys and lut
11545  * @vsi: Pointer to VSI structure
11546  * @seed: RSS hash seed
11547  * @lut: Lookup table
11548  * @lut_size: Lookup table size
11549  *
11550  * Returns 0 on success, negative on failure
11551  */
11552 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11553 {
11554         struct i40e_pf *pf = vsi->back;
11555
11556         if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11557                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
11558         else
11559                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
11560 }
11561
11562 /**
11563  * i40e_get_rss - Get RSS keys and lut
11564  * @vsi: Pointer to VSI structure
11565  * @seed: Buffer to store the keys
11566  * @lut: Buffer to store the lookup table entries
11567  * @lut_size: Size of buffer to store the lookup table entries
11568  *
11569  * Returns 0 on success, negative on failure
11570  */
11571 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11572 {
11573         struct i40e_pf *pf = vsi->back;
11574
11575         if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11576                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
11577         else
11578                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
11579 }
11580
11581 /**
11582  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
11583  * @pf: Pointer to board private structure
11584  * @lut: Lookup table
11585  * @rss_table_size: Lookup table size
11586  * @rss_size: Range of queue number for hashing
11587  */
11588 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
11589                        u16 rss_table_size, u16 rss_size)
11590 {
11591         u16 i;
11592
11593         for (i = 0; i < rss_table_size; i++)
11594                 lut[i] = i % rss_size;
11595 }
11596
11597 /**
11598  * i40e_pf_config_rss - Prepare for RSS if used
11599  * @pf: board private structure
11600  **/
11601 static int i40e_pf_config_rss(struct i40e_pf *pf)
11602 {
11603         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11604         u8 seed[I40E_HKEY_ARRAY_SIZE];
11605         u8 *lut;
11606         struct i40e_hw *hw = &pf->hw;
11607         u32 reg_val;
11608         u64 hena;
11609         int ret;
11610
11611         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
11612         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
11613                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
11614         hena |= i40e_pf_get_default_rss_hena(pf);
11615
11616         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
11617         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
11618
11619         /* Determine the RSS table size based on the hardware capabilities */
11620         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
11621         reg_val = (pf->rss_table_size == 512) ?
11622                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
11623                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
11624         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
11625
11626         /* Determine the RSS size of the VSI */
11627         if (!vsi->rss_size) {
11628                 u16 qcount;
11629                 /* If the firmware does something weird during VSI init, we
11630                  * could end up with zero TCs. Check for that to avoid
11631                  * divide-by-zero. It probably won't pass traffic, but it also
11632                  * won't panic.
11633                  */
11634                 qcount = vsi->num_queue_pairs /
11635                          (vsi->tc_config.numtc ? vsi->tc_config.numtc : 1);
11636                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11637         }
11638         if (!vsi->rss_size)
11639                 return -EINVAL;
11640
11641         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
11642         if (!lut)
11643                 return -ENOMEM;
11644
11645         /* Use user configured lut if there is one, otherwise use default */
11646         if (vsi->rss_lut_user)
11647                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
11648         else
11649                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
11650
11651         /* Use user configured hash key if there is one, otherwise
11652          * use default.
11653          */
11654         if (vsi->rss_hkey_user)
11655                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
11656         else
11657                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
11658         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
11659         kfree(lut);
11660
11661         return ret;
11662 }
11663
11664 /**
11665  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
11666  * @pf: board private structure
11667  * @queue_count: the requested queue count for rss.
11668  *
11669  * returns 0 if rss is not enabled, if enabled returns the final rss queue
11670  * count which may be different from the requested queue count.
11671  * Note: expects to be called while under rtnl_lock()
11672  **/
11673 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
11674 {
11675         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11676         int new_rss_size;
11677
11678         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
11679                 return 0;
11680
11681         queue_count = min_t(int, queue_count, num_online_cpus());
11682         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
11683
11684         if (queue_count != vsi->num_queue_pairs) {
11685                 u16 qcount;
11686
11687                 vsi->req_queue_pairs = queue_count;
11688                 i40e_prep_for_reset(pf, true);
11689
11690                 pf->alloc_rss_size = new_rss_size;
11691
11692                 i40e_reset_and_rebuild(pf, true, true);
11693
11694                 /* Discard the user configured hash keys and lut, if less
11695                  * queues are enabled.
11696                  */
11697                 if (queue_count < vsi->rss_size) {
11698                         i40e_clear_rss_config_user(vsi);
11699                         dev_dbg(&pf->pdev->dev,
11700                                 "discard user configured hash keys and lut\n");
11701                 }
11702
11703                 /* Reset vsi->rss_size, as number of enabled queues changed */
11704                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
11705                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11706
11707                 i40e_pf_config_rss(pf);
11708         }
11709         dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
11710                  vsi->req_queue_pairs, pf->rss_size_max);
11711         return pf->alloc_rss_size;
11712 }
11713
11714 /**
11715  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
11716  * @pf: board private structure
11717  **/
11718 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
11719 {
11720         i40e_status status;
11721         bool min_valid, max_valid;
11722         u32 max_bw, min_bw;
11723
11724         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
11725                                            &min_valid, &max_valid);
11726
11727         if (!status) {
11728                 if (min_valid)
11729                         pf->min_bw = min_bw;
11730                 if (max_valid)
11731                         pf->max_bw = max_bw;
11732         }
11733
11734         return status;
11735 }
11736
11737 /**
11738  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
11739  * @pf: board private structure
11740  **/
11741 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
11742 {
11743         struct i40e_aqc_configure_partition_bw_data bw_data;
11744         i40e_status status;
11745
11746         /* Set the valid bit for this PF */
11747         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
11748         bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
11749         bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
11750
11751         /* Set the new bandwidths */
11752         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
11753
11754         return status;
11755 }
11756
11757 /**
11758  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
11759  * @pf: board private structure
11760  **/
11761 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
11762 {
11763         /* Commit temporary BW setting to permanent NVM image */
11764         enum i40e_admin_queue_err last_aq_status;
11765         i40e_status ret;
11766         u16 nvm_word;
11767
11768         if (pf->hw.partition_id != 1) {
11769                 dev_info(&pf->pdev->dev,
11770                          "Commit BW only works on partition 1! This is partition %d",
11771                          pf->hw.partition_id);
11772                 ret = I40E_NOT_SUPPORTED;
11773                 goto bw_commit_out;
11774         }
11775
11776         /* Acquire NVM for read access */
11777         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
11778         last_aq_status = pf->hw.aq.asq_last_status;
11779         if (ret) {
11780                 dev_info(&pf->pdev->dev,
11781                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
11782                          i40e_stat_str(&pf->hw, ret),
11783                          i40e_aq_str(&pf->hw, last_aq_status));
11784                 goto bw_commit_out;
11785         }
11786
11787         /* Read word 0x10 of NVM - SW compatibility word 1 */
11788         ret = i40e_aq_read_nvm(&pf->hw,
11789                                I40E_SR_NVM_CONTROL_WORD,
11790                                0x10, sizeof(nvm_word), &nvm_word,
11791                                false, NULL);
11792         /* Save off last admin queue command status before releasing
11793          * the NVM
11794          */
11795         last_aq_status = pf->hw.aq.asq_last_status;
11796         i40e_release_nvm(&pf->hw);
11797         if (ret) {
11798                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
11799                          i40e_stat_str(&pf->hw, ret),
11800                          i40e_aq_str(&pf->hw, last_aq_status));
11801                 goto bw_commit_out;
11802         }
11803
11804         /* Wait a bit for NVM release to complete */
11805         msleep(50);
11806
11807         /* Acquire NVM for write access */
11808         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
11809         last_aq_status = pf->hw.aq.asq_last_status;
11810         if (ret) {
11811                 dev_info(&pf->pdev->dev,
11812                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
11813                          i40e_stat_str(&pf->hw, ret),
11814                          i40e_aq_str(&pf->hw, last_aq_status));
11815                 goto bw_commit_out;
11816         }
11817         /* Write it back out unchanged to initiate update NVM,
11818          * which will force a write of the shadow (alt) RAM to
11819          * the NVM - thus storing the bandwidth values permanently.
11820          */
11821         ret = i40e_aq_update_nvm(&pf->hw,
11822                                  I40E_SR_NVM_CONTROL_WORD,
11823                                  0x10, sizeof(nvm_word),
11824                                  &nvm_word, true, 0, NULL);
11825         /* Save off last admin queue command status before releasing
11826          * the NVM
11827          */
11828         last_aq_status = pf->hw.aq.asq_last_status;
11829         i40e_release_nvm(&pf->hw);
11830         if (ret)
11831                 dev_info(&pf->pdev->dev,
11832                          "BW settings NOT SAVED, err %s aq_err %s\n",
11833                          i40e_stat_str(&pf->hw, ret),
11834                          i40e_aq_str(&pf->hw, last_aq_status));
11835 bw_commit_out:
11836
11837         return ret;
11838 }
11839
11840 /**
11841  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
11842  * @pf: board private structure to initialize
11843  *
11844  * i40e_sw_init initializes the Adapter private data structure.
11845  * Fields are initialized based on PCI device information and
11846  * OS network device settings (MTU size).
11847  **/
11848 static int i40e_sw_init(struct i40e_pf *pf)
11849 {
11850         int err = 0;
11851         int size;
11852
11853         /* Set default capability flags */
11854         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
11855                     I40E_FLAG_MSI_ENABLED     |
11856                     I40E_FLAG_MSIX_ENABLED;
11857
11858         /* Set default ITR */
11859         pf->rx_itr_default = I40E_ITR_RX_DEF;
11860         pf->tx_itr_default = I40E_ITR_TX_DEF;
11861
11862         /* Depending on PF configurations, it is possible that the RSS
11863          * maximum might end up larger than the available queues
11864          */
11865         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
11866         pf->alloc_rss_size = 1;
11867         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
11868         pf->rss_size_max = min_t(int, pf->rss_size_max,
11869                                  pf->hw.func_caps.num_tx_qp);
11870         if (pf->hw.func_caps.rss) {
11871                 pf->flags |= I40E_FLAG_RSS_ENABLED;
11872                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
11873                                            num_online_cpus());
11874         }
11875
11876         /* MFP mode enabled */
11877         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
11878                 pf->flags |= I40E_FLAG_MFP_ENABLED;
11879                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
11880                 if (i40e_get_partition_bw_setting(pf)) {
11881                         dev_warn(&pf->pdev->dev,
11882                                  "Could not get partition bw settings\n");
11883                 } else {
11884                         dev_info(&pf->pdev->dev,
11885                                  "Partition BW Min = %8.8x, Max = %8.8x\n",
11886                                  pf->min_bw, pf->max_bw);
11887
11888                         /* nudge the Tx scheduler */
11889                         i40e_set_partition_bw_setting(pf);
11890                 }
11891         }
11892
11893         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
11894             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
11895                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
11896                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
11897                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
11898                     pf->hw.num_partitions > 1)
11899                         dev_info(&pf->pdev->dev,
11900                                  "Flow Director Sideband mode Disabled in MFP mode\n");
11901                 else
11902                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
11903                 pf->fdir_pf_filter_count =
11904                                  pf->hw.func_caps.fd_filters_guaranteed;
11905                 pf->hw.fdir_shared_filter_count =
11906                                  pf->hw.func_caps.fd_filters_best_effort;
11907         }
11908
11909         if (pf->hw.mac.type == I40E_MAC_X722) {
11910                 pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
11911                                     I40E_HW_128_QP_RSS_CAPABLE |
11912                                     I40E_HW_ATR_EVICT_CAPABLE |
11913                                     I40E_HW_WB_ON_ITR_CAPABLE |
11914                                     I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
11915                                     I40E_HW_NO_PCI_LINK_CHECK |
11916                                     I40E_HW_USE_SET_LLDP_MIB |
11917                                     I40E_HW_GENEVE_OFFLOAD_CAPABLE |
11918                                     I40E_HW_PTP_L4_CAPABLE |
11919                                     I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
11920                                     I40E_HW_OUTER_UDP_CSUM_CAPABLE);
11921
11922 #define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
11923                 if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
11924                     I40E_FDEVICT_PCTYPE_DEFAULT) {
11925                         dev_warn(&pf->pdev->dev,
11926                                  "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
11927                         pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
11928                 }
11929         } else if ((pf->hw.aq.api_maj_ver > 1) ||
11930                    ((pf->hw.aq.api_maj_ver == 1) &&
11931                     (pf->hw.aq.api_min_ver > 4))) {
11932                 /* Supported in FW API version higher than 1.4 */
11933                 pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
11934         }
11935
11936         /* Enable HW ATR eviction if possible */
11937         if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
11938                 pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
11939
11940         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11941             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
11942             (pf->hw.aq.fw_maj_ver < 4))) {
11943                 pf->hw_features |= I40E_HW_RESTART_AUTONEG;
11944                 /* No DCB support  for FW < v4.33 */
11945                 pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
11946         }
11947
11948         /* Disable FW LLDP if FW < v4.3 */
11949         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11950             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
11951             (pf->hw.aq.fw_maj_ver < 4)))
11952                 pf->hw_features |= I40E_HW_STOP_FW_LLDP;
11953
11954         /* Use the FW Set LLDP MIB API if FW > v4.40 */
11955         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11956             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
11957             (pf->hw.aq.fw_maj_ver >= 5)))
11958                 pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
11959
11960         /* Enable PTP L4 if FW > v6.0 */
11961         if (pf->hw.mac.type == I40E_MAC_XL710 &&
11962             pf->hw.aq.fw_maj_ver >= 6)
11963                 pf->hw_features |= I40E_HW_PTP_L4_CAPABLE;
11964
11965         if (pf->hw.func_caps.vmdq && num_online_cpus() != 1) {
11966                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
11967                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
11968                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
11969         }
11970
11971         if (pf->hw.func_caps.iwarp && num_online_cpus() != 1) {
11972                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
11973                 /* IWARP needs one extra vector for CQP just like MISC.*/
11974                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
11975         }
11976         /* Stopping FW LLDP engine is supported on XL710 and X722
11977          * starting from FW versions determined in i40e_init_adminq.
11978          * Stopping the FW LLDP engine is not supported on XL710
11979          * if NPAR is functioning so unset this hw flag in this case.
11980          */
11981         if (pf->hw.mac.type == I40E_MAC_XL710 &&
11982             pf->hw.func_caps.npar_enable &&
11983             (pf->hw.flags & I40E_HW_FLAG_FW_LLDP_STOPPABLE))
11984                 pf->hw.flags &= ~I40E_HW_FLAG_FW_LLDP_STOPPABLE;
11985
11986 #ifdef CONFIG_PCI_IOV
11987         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
11988                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
11989                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
11990                 pf->num_req_vfs = min_t(int,
11991                                         pf->hw.func_caps.num_vfs,
11992                                         I40E_MAX_VF_COUNT);
11993         }
11994 #endif /* CONFIG_PCI_IOV */
11995         pf->eeprom_version = 0xDEAD;
11996         pf->lan_veb = I40E_NO_VEB;
11997         pf->lan_vsi = I40E_NO_VSI;
11998
11999         /* By default FW has this off for performance reasons */
12000         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
12001
12002         /* set up queue assignment tracking */
12003         size = sizeof(struct i40e_lump_tracking)
12004                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
12005         pf->qp_pile = kzalloc(size, GFP_KERNEL);
12006         if (!pf->qp_pile) {
12007                 err = -ENOMEM;
12008                 goto sw_init_done;
12009         }
12010         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
12011         pf->qp_pile->search_hint = 0;
12012
12013         pf->tx_timeout_recovery_level = 1;
12014
12015         mutex_init(&pf->switch_mutex);
12016
12017 sw_init_done:
12018         return err;
12019 }
12020
12021 /**
12022  * i40e_set_ntuple - set the ntuple feature flag and take action
12023  * @pf: board private structure to initialize
12024  * @features: the feature set that the stack is suggesting
12025  *
12026  * returns a bool to indicate if reset needs to happen
12027  **/
12028 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
12029 {
12030         bool need_reset = false;
12031
12032         /* Check if Flow Director n-tuple support was enabled or disabled.  If
12033          * the state changed, we need to reset.
12034          */
12035         if (features & NETIF_F_NTUPLE) {
12036                 /* Enable filters and mark for reset */
12037                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
12038                         need_reset = true;
12039                 /* enable FD_SB only if there is MSI-X vector and no cloud
12040                  * filters exist
12041                  */
12042                 if (pf->num_fdsb_msix > 0 && !pf->num_cloud_filters) {
12043                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
12044                         pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
12045                 }
12046         } else {
12047                 /* turn off filters, mark for reset and clear SW filter list */
12048                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
12049                         need_reset = true;
12050                         i40e_fdir_filter_exit(pf);
12051                 }
12052                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
12053                 clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state);
12054                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
12055
12056                 /* reset fd counters */
12057                 pf->fd_add_err = 0;
12058                 pf->fd_atr_cnt = 0;
12059                 /* if ATR was auto disabled it can be re-enabled. */
12060                 if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state))
12061                         if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
12062                             (I40E_DEBUG_FD & pf->hw.debug_mask))
12063                                 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
12064         }
12065         return need_reset;
12066 }
12067
12068 /**
12069  * i40e_clear_rss_lut - clear the rx hash lookup table
12070  * @vsi: the VSI being configured
12071  **/
12072 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
12073 {
12074         struct i40e_pf *pf = vsi->back;
12075         struct i40e_hw *hw = &pf->hw;
12076         u16 vf_id = vsi->vf_id;
12077         u8 i;
12078
12079         if (vsi->type == I40E_VSI_MAIN) {
12080                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
12081                         wr32(hw, I40E_PFQF_HLUT(i), 0);
12082         } else if (vsi->type == I40E_VSI_SRIOV) {
12083                 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
12084                         i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
12085         } else {
12086                 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
12087         }
12088 }
12089
12090 /**
12091  * i40e_set_features - set the netdev feature flags
12092  * @netdev: ptr to the netdev being adjusted
12093  * @features: the feature set that the stack is suggesting
12094  * Note: expects to be called while under rtnl_lock()
12095  **/
12096 static int i40e_set_features(struct net_device *netdev,
12097                              netdev_features_t features)
12098 {
12099         struct i40e_netdev_priv *np = netdev_priv(netdev);
12100         struct i40e_vsi *vsi = np->vsi;
12101         struct i40e_pf *pf = vsi->back;
12102         bool need_reset;
12103
12104         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
12105                 i40e_pf_config_rss(pf);
12106         else if (!(features & NETIF_F_RXHASH) &&
12107                  netdev->features & NETIF_F_RXHASH)
12108                 i40e_clear_rss_lut(vsi);
12109
12110         if (features & NETIF_F_HW_VLAN_CTAG_RX)
12111                 i40e_vlan_stripping_enable(vsi);
12112         else
12113                 i40e_vlan_stripping_disable(vsi);
12114
12115         if (!(features & NETIF_F_HW_TC) && pf->num_cloud_filters) {
12116                 dev_err(&pf->pdev->dev,
12117                         "Offloaded tc filters active, can't turn hw_tc_offload off");
12118                 return -EINVAL;
12119         }
12120
12121         if (!(features & NETIF_F_HW_L2FW_DOFFLOAD) && vsi->macvlan_cnt)
12122                 i40e_del_all_macvlans(vsi);
12123
12124         need_reset = i40e_set_ntuple(pf, features);
12125
12126         if (need_reset)
12127                 i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12128
12129         return 0;
12130 }
12131
12132 /**
12133  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
12134  * @pf: board private structure
12135  * @port: The UDP port to look up
12136  *
12137  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
12138  **/
12139 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
12140 {
12141         u8 i;
12142
12143         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
12144                 /* Do not report ports with pending deletions as
12145                  * being available.
12146                  */
12147                 if (!port && (pf->pending_udp_bitmap & BIT_ULL(i)))
12148                         continue;
12149                 if (pf->udp_ports[i].port == port)
12150                         return i;
12151         }
12152
12153         return i;
12154 }
12155
12156 /**
12157  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
12158  * @netdev: This physical port's netdev
12159  * @ti: Tunnel endpoint information
12160  **/
12161 static void i40e_udp_tunnel_add(struct net_device *netdev,
12162                                 struct udp_tunnel_info *ti)
12163 {
12164         struct i40e_netdev_priv *np = netdev_priv(netdev);
12165         struct i40e_vsi *vsi = np->vsi;
12166         struct i40e_pf *pf = vsi->back;
12167         u16 port = ntohs(ti->port);
12168         u8 next_idx;
12169         u8 idx;
12170
12171         idx = i40e_get_udp_port_idx(pf, port);
12172
12173         /* Check if port already exists */
12174         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12175                 netdev_info(netdev, "port %d already offloaded\n", port);
12176                 return;
12177         }
12178
12179         /* Now check if there is space to add the new port */
12180         next_idx = i40e_get_udp_port_idx(pf, 0);
12181
12182         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12183                 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
12184                             port);
12185                 return;
12186         }
12187
12188         switch (ti->type) {
12189         case UDP_TUNNEL_TYPE_VXLAN:
12190                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
12191                 break;
12192         case UDP_TUNNEL_TYPE_GENEVE:
12193                 if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
12194                         return;
12195                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
12196                 break;
12197         default:
12198                 return;
12199         }
12200
12201         /* New port: add it and mark its index in the bitmap */
12202         pf->udp_ports[next_idx].port = port;
12203         pf->udp_ports[next_idx].filter_index = I40E_UDP_PORT_INDEX_UNUSED;
12204         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
12205         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12206 }
12207
12208 /**
12209  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
12210  * @netdev: This physical port's netdev
12211  * @ti: Tunnel endpoint information
12212  **/
12213 static void i40e_udp_tunnel_del(struct net_device *netdev,
12214                                 struct udp_tunnel_info *ti)
12215 {
12216         struct i40e_netdev_priv *np = netdev_priv(netdev);
12217         struct i40e_vsi *vsi = np->vsi;
12218         struct i40e_pf *pf = vsi->back;
12219         u16 port = ntohs(ti->port);
12220         u8 idx;
12221
12222         idx = i40e_get_udp_port_idx(pf, port);
12223
12224         /* Check if port already exists */
12225         if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
12226                 goto not_found;
12227
12228         switch (ti->type) {
12229         case UDP_TUNNEL_TYPE_VXLAN:
12230                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
12231                         goto not_found;
12232                 break;
12233         case UDP_TUNNEL_TYPE_GENEVE:
12234                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
12235                         goto not_found;
12236                 break;
12237         default:
12238                 goto not_found;
12239         }
12240
12241         /* if port exists, set it to 0 (mark for deletion)
12242          * and make it pending
12243          */
12244         pf->udp_ports[idx].port = 0;
12245
12246         /* Toggle pending bit instead of setting it. This way if we are
12247          * deleting a port that has yet to be added we just clear the pending
12248          * bit and don't have to worry about it.
12249          */
12250         pf->pending_udp_bitmap ^= BIT_ULL(idx);
12251         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12252
12253         return;
12254 not_found:
12255         netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
12256                     port);
12257 }
12258
12259 static int i40e_get_phys_port_id(struct net_device *netdev,
12260                                  struct netdev_phys_item_id *ppid)
12261 {
12262         struct i40e_netdev_priv *np = netdev_priv(netdev);
12263         struct i40e_pf *pf = np->vsi->back;
12264         struct i40e_hw *hw = &pf->hw;
12265
12266         if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
12267                 return -EOPNOTSUPP;
12268
12269         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
12270         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
12271
12272         return 0;
12273 }
12274
12275 /**
12276  * i40e_ndo_fdb_add - add an entry to the hardware database
12277  * @ndm: the input from the stack
12278  * @tb: pointer to array of nladdr (unused)
12279  * @dev: the net device pointer
12280  * @addr: the MAC address entry being added
12281  * @vid: VLAN ID
12282  * @flags: instructions from stack about fdb operation
12283  */
12284 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
12285                             struct net_device *dev,
12286                             const unsigned char *addr, u16 vid,
12287                             u16 flags,
12288                             struct netlink_ext_ack *extack)
12289 {
12290         struct i40e_netdev_priv *np = netdev_priv(dev);
12291         struct i40e_pf *pf = np->vsi->back;
12292         int err = 0;
12293
12294         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
12295                 return -EOPNOTSUPP;
12296
12297         if (vid) {
12298                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
12299                 return -EINVAL;
12300         }
12301
12302         /* Hardware does not support aging addresses so if a
12303          * ndm_state is given only allow permanent addresses
12304          */
12305         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
12306                 netdev_info(dev, "FDB only supports static addresses\n");
12307                 return -EINVAL;
12308         }
12309
12310         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
12311                 err = dev_uc_add_excl(dev, addr);
12312         else if (is_multicast_ether_addr(addr))
12313                 err = dev_mc_add_excl(dev, addr);
12314         else
12315                 err = -EINVAL;
12316
12317         /* Only return duplicate errors if NLM_F_EXCL is set */
12318         if (err == -EEXIST && !(flags & NLM_F_EXCL))
12319                 err = 0;
12320
12321         return err;
12322 }
12323
12324 /**
12325  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
12326  * @dev: the netdev being configured
12327  * @nlh: RTNL message
12328  * @flags: bridge flags
12329  * @extack: netlink extended ack
12330  *
12331  * Inserts a new hardware bridge if not already created and
12332  * enables the bridging mode requested (VEB or VEPA). If the
12333  * hardware bridge has already been inserted and the request
12334  * is to change the mode then that requires a PF reset to
12335  * allow rebuild of the components with required hardware
12336  * bridge mode enabled.
12337  *
12338  * Note: expects to be called while under rtnl_lock()
12339  **/
12340 static int i40e_ndo_bridge_setlink(struct net_device *dev,
12341                                    struct nlmsghdr *nlh,
12342                                    u16 flags,
12343                                    struct netlink_ext_ack *extack)
12344 {
12345         struct i40e_netdev_priv *np = netdev_priv(dev);
12346         struct i40e_vsi *vsi = np->vsi;
12347         struct i40e_pf *pf = vsi->back;
12348         struct i40e_veb *veb = NULL;
12349         struct nlattr *attr, *br_spec;
12350         int i, rem;
12351
12352         /* Only for PF VSI for now */
12353         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12354                 return -EOPNOTSUPP;
12355
12356         /* Find the HW bridge for PF VSI */
12357         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12358                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12359                         veb = pf->veb[i];
12360         }
12361
12362         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
12363
12364         nla_for_each_nested(attr, br_spec, rem) {
12365                 __u16 mode;
12366
12367                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
12368                         continue;
12369
12370                 mode = nla_get_u16(attr);
12371                 if ((mode != BRIDGE_MODE_VEPA) &&
12372                     (mode != BRIDGE_MODE_VEB))
12373                         return -EINVAL;
12374
12375                 /* Insert a new HW bridge */
12376                 if (!veb) {
12377                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
12378                                              vsi->tc_config.enabled_tc);
12379                         if (veb) {
12380                                 veb->bridge_mode = mode;
12381                                 i40e_config_bridge_mode(veb);
12382                         } else {
12383                                 /* No Bridge HW offload available */
12384                                 return -ENOENT;
12385                         }
12386                         break;
12387                 } else if (mode != veb->bridge_mode) {
12388                         /* Existing HW bridge but different mode needs reset */
12389                         veb->bridge_mode = mode;
12390                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
12391                         if (mode == BRIDGE_MODE_VEB)
12392                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
12393                         else
12394                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
12395                         i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12396                         break;
12397                 }
12398         }
12399
12400         return 0;
12401 }
12402
12403 /**
12404  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
12405  * @skb: skb buff
12406  * @pid: process id
12407  * @seq: RTNL message seq #
12408  * @dev: the netdev being configured
12409  * @filter_mask: unused
12410  * @nlflags: netlink flags passed in
12411  *
12412  * Return the mode in which the hardware bridge is operating in
12413  * i.e VEB or VEPA.
12414  **/
12415 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
12416                                    struct net_device *dev,
12417                                    u32 __always_unused filter_mask,
12418                                    int nlflags)
12419 {
12420         struct i40e_netdev_priv *np = netdev_priv(dev);
12421         struct i40e_vsi *vsi = np->vsi;
12422         struct i40e_pf *pf = vsi->back;
12423         struct i40e_veb *veb = NULL;
12424         int i;
12425
12426         /* Only for PF VSI for now */
12427         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12428                 return -EOPNOTSUPP;
12429
12430         /* Find the HW bridge for the PF VSI */
12431         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12432                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12433                         veb = pf->veb[i];
12434         }
12435
12436         if (!veb)
12437                 return 0;
12438
12439         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
12440                                        0, 0, nlflags, filter_mask, NULL);
12441 }
12442
12443 /**
12444  * i40e_features_check - Validate encapsulated packet conforms to limits
12445  * @skb: skb buff
12446  * @dev: This physical port's netdev
12447  * @features: Offload features that the stack believes apply
12448  **/
12449 static netdev_features_t i40e_features_check(struct sk_buff *skb,
12450                                              struct net_device *dev,
12451                                              netdev_features_t features)
12452 {
12453         size_t len;
12454
12455         /* No point in doing any of this if neither checksum nor GSO are
12456          * being requested for this frame.  We can rule out both by just
12457          * checking for CHECKSUM_PARTIAL
12458          */
12459         if (skb->ip_summed != CHECKSUM_PARTIAL)
12460                 return features;
12461
12462         /* We cannot support GSO if the MSS is going to be less than
12463          * 64 bytes.  If it is then we need to drop support for GSO.
12464          */
12465         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
12466                 features &= ~NETIF_F_GSO_MASK;
12467
12468         /* MACLEN can support at most 63 words */
12469         len = skb_network_header(skb) - skb->data;
12470         if (len & ~(63 * 2))
12471                 goto out_err;
12472
12473         /* IPLEN and EIPLEN can support at most 127 dwords */
12474         len = skb_transport_header(skb) - skb_network_header(skb);
12475         if (len & ~(127 * 4))
12476                 goto out_err;
12477
12478         if (skb->encapsulation) {
12479                 /* L4TUNLEN can support 127 words */
12480                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
12481                 if (len & ~(127 * 2))
12482                         goto out_err;
12483
12484                 /* IPLEN can support at most 127 dwords */
12485                 len = skb_inner_transport_header(skb) -
12486                       skb_inner_network_header(skb);
12487                 if (len & ~(127 * 4))
12488                         goto out_err;
12489         }
12490
12491         /* No need to validate L4LEN as TCP is the only protocol with a
12492          * a flexible value and we support all possible values supported
12493          * by TCP, which is at most 15 dwords
12494          */
12495
12496         return features;
12497 out_err:
12498         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
12499 }
12500
12501 /**
12502  * i40e_xdp_setup - add/remove an XDP program
12503  * @vsi: VSI to changed
12504  * @prog: XDP program
12505  **/
12506 static int i40e_xdp_setup(struct i40e_vsi *vsi,
12507                           struct bpf_prog *prog)
12508 {
12509         int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
12510         struct i40e_pf *pf = vsi->back;
12511         struct bpf_prog *old_prog;
12512         bool need_reset;
12513         int i;
12514
12515         /* Don't allow frames that span over multiple buffers */
12516         if (frame_size > vsi->rx_buf_len)
12517                 return -EINVAL;
12518
12519         if (!i40e_enabled_xdp_vsi(vsi) && !prog)
12520                 return 0;
12521
12522         /* When turning XDP on->off/off->on we reset and rebuild the rings. */
12523         need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
12524
12525         if (need_reset)
12526                 i40e_prep_for_reset(pf, true);
12527
12528         old_prog = xchg(&vsi->xdp_prog, prog);
12529
12530         if (need_reset) {
12531                 if (!prog)
12532                         /* Wait until ndo_xsk_wakeup completes. */
12533                         synchronize_rcu();
12534                 i40e_reset_and_rebuild(pf, true, true);
12535         }
12536
12537         for (i = 0; i < vsi->num_queue_pairs; i++)
12538                 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
12539
12540         if (old_prog)
12541                 bpf_prog_put(old_prog);
12542
12543         /* Kick start the NAPI context if there is an AF_XDP socket open
12544          * on that queue id. This so that receiving will start.
12545          */
12546         if (need_reset && prog)
12547                 for (i = 0; i < vsi->num_queue_pairs; i++)
12548                         if (vsi->xdp_rings[i]->xsk_umem)
12549                                 (void)i40e_xsk_wakeup(vsi->netdev, i,
12550                                                       XDP_WAKEUP_RX);
12551
12552         return 0;
12553 }
12554
12555 /**
12556  * i40e_enter_busy_conf - Enters busy config state
12557  * @vsi: vsi
12558  *
12559  * Returns 0 on success, <0 for failure.
12560  **/
12561 static int i40e_enter_busy_conf(struct i40e_vsi *vsi)
12562 {
12563         struct i40e_pf *pf = vsi->back;
12564         int timeout = 50;
12565
12566         while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state)) {
12567                 timeout--;
12568                 if (!timeout)
12569                         return -EBUSY;
12570                 usleep_range(1000, 2000);
12571         }
12572
12573         return 0;
12574 }
12575
12576 /**
12577  * i40e_exit_busy_conf - Exits busy config state
12578  * @vsi: vsi
12579  **/
12580 static void i40e_exit_busy_conf(struct i40e_vsi *vsi)
12581 {
12582         struct i40e_pf *pf = vsi->back;
12583
12584         clear_bit(__I40E_CONFIG_BUSY, pf->state);
12585 }
12586
12587 /**
12588  * i40e_queue_pair_reset_stats - Resets all statistics for a queue pair
12589  * @vsi: vsi
12590  * @queue_pair: queue pair
12591  **/
12592 static void i40e_queue_pair_reset_stats(struct i40e_vsi *vsi, int queue_pair)
12593 {
12594         memset(&vsi->rx_rings[queue_pair]->rx_stats, 0,
12595                sizeof(vsi->rx_rings[queue_pair]->rx_stats));
12596         memset(&vsi->tx_rings[queue_pair]->stats, 0,
12597                sizeof(vsi->tx_rings[queue_pair]->stats));
12598         if (i40e_enabled_xdp_vsi(vsi)) {
12599                 memset(&vsi->xdp_rings[queue_pair]->stats, 0,
12600                        sizeof(vsi->xdp_rings[queue_pair]->stats));
12601         }
12602 }
12603
12604 /**
12605  * i40e_queue_pair_clean_rings - Cleans all the rings of a queue pair
12606  * @vsi: vsi
12607  * @queue_pair: queue pair
12608  **/
12609 static void i40e_queue_pair_clean_rings(struct i40e_vsi *vsi, int queue_pair)
12610 {
12611         i40e_clean_tx_ring(vsi->tx_rings[queue_pair]);
12612         if (i40e_enabled_xdp_vsi(vsi)) {
12613                 /* Make sure that in-progress ndo_xdp_xmit calls are
12614                  * completed.
12615                  */
12616                 synchronize_rcu();
12617                 i40e_clean_tx_ring(vsi->xdp_rings[queue_pair]);
12618         }
12619         i40e_clean_rx_ring(vsi->rx_rings[queue_pair]);
12620 }
12621
12622 /**
12623  * i40e_queue_pair_toggle_napi - Enables/disables NAPI for a queue pair
12624  * @vsi: vsi
12625  * @queue_pair: queue pair
12626  * @enable: true for enable, false for disable
12627  **/
12628 static void i40e_queue_pair_toggle_napi(struct i40e_vsi *vsi, int queue_pair,
12629                                         bool enable)
12630 {
12631         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12632         struct i40e_q_vector *q_vector = rxr->q_vector;
12633
12634         if (!vsi->netdev)
12635                 return;
12636
12637         /* All rings in a qp belong to the same qvector. */
12638         if (q_vector->rx.ring || q_vector->tx.ring) {
12639                 if (enable)
12640                         napi_enable(&q_vector->napi);
12641                 else
12642                         napi_disable(&q_vector->napi);
12643         }
12644 }
12645
12646 /**
12647  * i40e_queue_pair_toggle_rings - Enables/disables all rings for a queue pair
12648  * @vsi: vsi
12649  * @queue_pair: queue pair
12650  * @enable: true for enable, false for disable
12651  *
12652  * Returns 0 on success, <0 on failure.
12653  **/
12654 static int i40e_queue_pair_toggle_rings(struct i40e_vsi *vsi, int queue_pair,
12655                                         bool enable)
12656 {
12657         struct i40e_pf *pf = vsi->back;
12658         int pf_q, ret = 0;
12659
12660         pf_q = vsi->base_queue + queue_pair;
12661         ret = i40e_control_wait_tx_q(vsi->seid, pf, pf_q,
12662                                      false /*is xdp*/, enable);
12663         if (ret) {
12664                 dev_info(&pf->pdev->dev,
12665                          "VSI seid %d Tx ring %d %sable timeout\n",
12666                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12667                 return ret;
12668         }
12669
12670         i40e_control_rx_q(pf, pf_q, enable);
12671         ret = i40e_pf_rxq_wait(pf, pf_q, enable);
12672         if (ret) {
12673                 dev_info(&pf->pdev->dev,
12674                          "VSI seid %d Rx ring %d %sable timeout\n",
12675                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12676                 return ret;
12677         }
12678
12679         /* Due to HW errata, on Rx disable only, the register can
12680          * indicate done before it really is. Needs 50ms to be sure
12681          */
12682         if (!enable)
12683                 mdelay(50);
12684
12685         if (!i40e_enabled_xdp_vsi(vsi))
12686                 return ret;
12687
12688         ret = i40e_control_wait_tx_q(vsi->seid, pf,
12689                                      pf_q + vsi->alloc_queue_pairs,
12690                                      true /*is xdp*/, enable);
12691         if (ret) {
12692                 dev_info(&pf->pdev->dev,
12693                          "VSI seid %d XDP Tx ring %d %sable timeout\n",
12694                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12695         }
12696
12697         return ret;
12698 }
12699
12700 /**
12701  * i40e_queue_pair_enable_irq - Enables interrupts for a queue pair
12702  * @vsi: vsi
12703  * @queue_pair: queue_pair
12704  **/
12705 static void i40e_queue_pair_enable_irq(struct i40e_vsi *vsi, int queue_pair)
12706 {
12707         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12708         struct i40e_pf *pf = vsi->back;
12709         struct i40e_hw *hw = &pf->hw;
12710
12711         /* All rings in a qp belong to the same qvector. */
12712         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
12713                 i40e_irq_dynamic_enable(vsi, rxr->q_vector->v_idx);
12714         else
12715                 i40e_irq_dynamic_enable_icr0(pf);
12716
12717         i40e_flush(hw);
12718 }
12719
12720 /**
12721  * i40e_queue_pair_disable_irq - Disables interrupts for a queue pair
12722  * @vsi: vsi
12723  * @queue_pair: queue_pair
12724  **/
12725 static void i40e_queue_pair_disable_irq(struct i40e_vsi *vsi, int queue_pair)
12726 {
12727         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12728         struct i40e_pf *pf = vsi->back;
12729         struct i40e_hw *hw = &pf->hw;
12730
12731         /* For simplicity, instead of removing the qp interrupt causes
12732          * from the interrupt linked list, we simply disable the interrupt, and
12733          * leave the list intact.
12734          *
12735          * All rings in a qp belong to the same qvector.
12736          */
12737         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
12738                 u32 intpf = vsi->base_vector + rxr->q_vector->v_idx;
12739
12740                 wr32(hw, I40E_PFINT_DYN_CTLN(intpf - 1), 0);
12741                 i40e_flush(hw);
12742                 synchronize_irq(pf->msix_entries[intpf].vector);
12743         } else {
12744                 /* Legacy and MSI mode - this stops all interrupt handling */
12745                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
12746                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
12747                 i40e_flush(hw);
12748                 synchronize_irq(pf->pdev->irq);
12749         }
12750 }
12751
12752 /**
12753  * i40e_queue_pair_disable - Disables a queue pair
12754  * @vsi: vsi
12755  * @queue_pair: queue pair
12756  *
12757  * Returns 0 on success, <0 on failure.
12758  **/
12759 int i40e_queue_pair_disable(struct i40e_vsi *vsi, int queue_pair)
12760 {
12761         int err;
12762
12763         err = i40e_enter_busy_conf(vsi);
12764         if (err)
12765                 return err;
12766
12767         i40e_queue_pair_disable_irq(vsi, queue_pair);
12768         err = i40e_queue_pair_toggle_rings(vsi, queue_pair, false /* off */);
12769         i40e_queue_pair_toggle_napi(vsi, queue_pair, false /* off */);
12770         i40e_queue_pair_clean_rings(vsi, queue_pair);
12771         i40e_queue_pair_reset_stats(vsi, queue_pair);
12772
12773         return err;
12774 }
12775
12776 /**
12777  * i40e_queue_pair_enable - Enables a queue pair
12778  * @vsi: vsi
12779  * @queue_pair: queue pair
12780  *
12781  * Returns 0 on success, <0 on failure.
12782  **/
12783 int i40e_queue_pair_enable(struct i40e_vsi *vsi, int queue_pair)
12784 {
12785         int err;
12786
12787         err = i40e_configure_tx_ring(vsi->tx_rings[queue_pair]);
12788         if (err)
12789                 return err;
12790
12791         if (i40e_enabled_xdp_vsi(vsi)) {
12792                 err = i40e_configure_tx_ring(vsi->xdp_rings[queue_pair]);
12793                 if (err)
12794                         return err;
12795         }
12796
12797         err = i40e_configure_rx_ring(vsi->rx_rings[queue_pair]);
12798         if (err)
12799                 return err;
12800
12801         err = i40e_queue_pair_toggle_rings(vsi, queue_pair, true /* on */);
12802         i40e_queue_pair_toggle_napi(vsi, queue_pair, true /* on */);
12803         i40e_queue_pair_enable_irq(vsi, queue_pair);
12804
12805         i40e_exit_busy_conf(vsi);
12806
12807         return err;
12808 }
12809
12810 /**
12811  * i40e_xdp - implements ndo_bpf for i40e
12812  * @dev: netdevice
12813  * @xdp: XDP command
12814  **/
12815 static int i40e_xdp(struct net_device *dev,
12816                     struct netdev_bpf *xdp)
12817 {
12818         struct i40e_netdev_priv *np = netdev_priv(dev);
12819         struct i40e_vsi *vsi = np->vsi;
12820
12821         if (vsi->type != I40E_VSI_MAIN)
12822                 return -EINVAL;
12823
12824         switch (xdp->command) {
12825         case XDP_SETUP_PROG:
12826                 return i40e_xdp_setup(vsi, xdp->prog);
12827         case XDP_QUERY_PROG:
12828                 xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
12829                 return 0;
12830         case XDP_SETUP_XSK_UMEM:
12831                 return i40e_xsk_umem_setup(vsi, xdp->xsk.umem,
12832                                            xdp->xsk.queue_id);
12833         default:
12834                 return -EINVAL;
12835         }
12836 }
12837
12838 static const struct net_device_ops i40e_netdev_ops = {
12839         .ndo_open               = i40e_open,
12840         .ndo_stop               = i40e_close,
12841         .ndo_start_xmit         = i40e_lan_xmit_frame,
12842         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
12843         .ndo_set_rx_mode        = i40e_set_rx_mode,
12844         .ndo_validate_addr      = eth_validate_addr,
12845         .ndo_set_mac_address    = i40e_set_mac,
12846         .ndo_change_mtu         = i40e_change_mtu,
12847         .ndo_do_ioctl           = i40e_ioctl,
12848         .ndo_tx_timeout         = i40e_tx_timeout,
12849         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
12850         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
12851 #ifdef CONFIG_NET_POLL_CONTROLLER
12852         .ndo_poll_controller    = i40e_netpoll,
12853 #endif
12854         .ndo_setup_tc           = __i40e_setup_tc,
12855         .ndo_set_features       = i40e_set_features,
12856         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
12857         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
12858         .ndo_get_vf_stats       = i40e_get_vf_stats,
12859         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
12860         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
12861         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
12862         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
12863         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
12864         .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
12865         .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
12866         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
12867         .ndo_fdb_add            = i40e_ndo_fdb_add,
12868         .ndo_features_check     = i40e_features_check,
12869         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
12870         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
12871         .ndo_bpf                = i40e_xdp,
12872         .ndo_xdp_xmit           = i40e_xdp_xmit,
12873         .ndo_xsk_wakeup         = i40e_xsk_wakeup,
12874         .ndo_dfwd_add_station   = i40e_fwd_add,
12875         .ndo_dfwd_del_station   = i40e_fwd_del,
12876 };
12877
12878 /**
12879  * i40e_config_netdev - Setup the netdev flags
12880  * @vsi: the VSI being configured
12881  *
12882  * Returns 0 on success, negative value on failure
12883  **/
12884 static int i40e_config_netdev(struct i40e_vsi *vsi)
12885 {
12886         struct i40e_pf *pf = vsi->back;
12887         struct i40e_hw *hw = &pf->hw;
12888         struct i40e_netdev_priv *np;
12889         struct net_device *netdev;
12890         u8 broadcast[ETH_ALEN];
12891         u8 mac_addr[ETH_ALEN];
12892         int etherdev_size;
12893         netdev_features_t hw_enc_features;
12894         netdev_features_t hw_features;
12895
12896         etherdev_size = sizeof(struct i40e_netdev_priv);
12897         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
12898         if (!netdev)
12899                 return -ENOMEM;
12900
12901         vsi->netdev = netdev;
12902         np = netdev_priv(netdev);
12903         np->vsi = vsi;
12904
12905         hw_enc_features = NETIF_F_SG                    |
12906                           NETIF_F_IP_CSUM               |
12907                           NETIF_F_IPV6_CSUM             |
12908                           NETIF_F_HIGHDMA               |
12909                           NETIF_F_SOFT_FEATURES         |
12910                           NETIF_F_TSO                   |
12911                           NETIF_F_TSO_ECN               |
12912                           NETIF_F_TSO6                  |
12913                           NETIF_F_GSO_GRE               |
12914                           NETIF_F_GSO_GRE_CSUM          |
12915                           NETIF_F_GSO_PARTIAL           |
12916                           NETIF_F_GSO_IPXIP4            |
12917                           NETIF_F_GSO_IPXIP6            |
12918                           NETIF_F_GSO_UDP_TUNNEL        |
12919                           NETIF_F_GSO_UDP_TUNNEL_CSUM   |
12920                           NETIF_F_GSO_UDP_L4            |
12921                           NETIF_F_SCTP_CRC              |
12922                           NETIF_F_RXHASH                |
12923                           NETIF_F_RXCSUM                |
12924                           0;
12925
12926         if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
12927                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
12928
12929         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
12930
12931         netdev->hw_enc_features |= hw_enc_features;
12932
12933         /* record features VLANs can make use of */
12934         netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
12935
12936         /* enable macvlan offloads */
12937         netdev->hw_features |= NETIF_F_HW_L2FW_DOFFLOAD;
12938
12939         hw_features = hw_enc_features           |
12940                       NETIF_F_HW_VLAN_CTAG_TX   |
12941                       NETIF_F_HW_VLAN_CTAG_RX;
12942
12943         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
12944                 hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
12945
12946         netdev->hw_features |= hw_features;
12947
12948         netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
12949         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
12950
12951         if (vsi->type == I40E_VSI_MAIN) {
12952                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
12953                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
12954                 /* The following steps are necessary for two reasons. First,
12955                  * some older NVM configurations load a default MAC-VLAN
12956                  * filter that will accept any tagged packet, and we want to
12957                  * replace this with a normal filter. Additionally, it is
12958                  * possible our MAC address was provided by the platform using
12959                  * Open Firmware or similar.
12960                  *
12961                  * Thus, we need to remove the default filter and install one
12962                  * specific to the MAC address.
12963                  */
12964                 i40e_rm_default_mac_filter(vsi, mac_addr);
12965                 spin_lock_bh(&vsi->mac_filter_hash_lock);
12966                 i40e_add_mac_filter(vsi, mac_addr);
12967                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
12968         } else {
12969                 /* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
12970                  * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
12971                  * the end, which is 4 bytes long, so force truncation of the
12972                  * original name by IFNAMSIZ - 4
12973                  */
12974                 snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
12975                          IFNAMSIZ - 4,
12976                          pf->vsi[pf->lan_vsi]->netdev->name);
12977                 eth_random_addr(mac_addr);
12978
12979                 spin_lock_bh(&vsi->mac_filter_hash_lock);
12980                 i40e_add_mac_filter(vsi, mac_addr);
12981                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
12982         }
12983
12984         /* Add the broadcast filter so that we initially will receive
12985          * broadcast packets. Note that when a new VLAN is first added the
12986          * driver will convert all filters marked I40E_VLAN_ANY into VLAN
12987          * specific filters as part of transitioning into "vlan" operation.
12988          * When more VLANs are added, the driver will copy each existing MAC
12989          * filter and add it for the new VLAN.
12990          *
12991          * Broadcast filters are handled specially by
12992          * i40e_sync_filters_subtask, as the driver must to set the broadcast
12993          * promiscuous bit instead of adding this directly as a MAC/VLAN
12994          * filter. The subtask will update the correct broadcast promiscuous
12995          * bits as VLANs become active or inactive.
12996          */
12997         eth_broadcast_addr(broadcast);
12998         spin_lock_bh(&vsi->mac_filter_hash_lock);
12999         i40e_add_mac_filter(vsi, broadcast);
13000         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13001
13002         ether_addr_copy(netdev->dev_addr, mac_addr);
13003         ether_addr_copy(netdev->perm_addr, mac_addr);
13004
13005         /* i40iw_net_event() reads 16 bytes from neigh->primary_key */
13006         netdev->neigh_priv_len = sizeof(u32) * 4;
13007
13008         netdev->priv_flags |= IFF_UNICAST_FLT;
13009         netdev->priv_flags |= IFF_SUPP_NOFCS;
13010         /* Setup netdev TC information */
13011         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
13012
13013         netdev->netdev_ops = &i40e_netdev_ops;
13014         netdev->watchdog_timeo = 5 * HZ;
13015         i40e_set_ethtool_ops(netdev);
13016
13017         /* MTU range: 68 - 9706 */
13018         netdev->min_mtu = ETH_MIN_MTU;
13019         netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
13020
13021         return 0;
13022 }
13023
13024 /**
13025  * i40e_vsi_delete - Delete a VSI from the switch
13026  * @vsi: the VSI being removed
13027  *
13028  * Returns 0 on success, negative value on failure
13029  **/
13030 static void i40e_vsi_delete(struct i40e_vsi *vsi)
13031 {
13032         /* remove default VSI is not allowed */
13033         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
13034                 return;
13035
13036         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
13037 }
13038
13039 /**
13040  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
13041  * @vsi: the VSI being queried
13042  *
13043  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
13044  **/
13045 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
13046 {
13047         struct i40e_veb *veb;
13048         struct i40e_pf *pf = vsi->back;
13049
13050         /* Uplink is not a bridge so default to VEB */
13051         if (vsi->veb_idx >= I40E_MAX_VEB)
13052                 return 1;
13053
13054         veb = pf->veb[vsi->veb_idx];
13055         if (!veb) {
13056                 dev_info(&pf->pdev->dev,
13057                          "There is no veb associated with the bridge\n");
13058                 return -ENOENT;
13059         }
13060
13061         /* Uplink is a bridge in VEPA mode */
13062         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
13063                 return 0;
13064         } else {
13065                 /* Uplink is a bridge in VEB mode */
13066                 return 1;
13067         }
13068
13069         /* VEPA is now default bridge, so return 0 */
13070         return 0;
13071 }
13072
13073 /**
13074  * i40e_add_vsi - Add a VSI to the switch
13075  * @vsi: the VSI being configured
13076  *
13077  * This initializes a VSI context depending on the VSI type to be added and
13078  * passes it down to the add_vsi aq command.
13079  **/
13080 static int i40e_add_vsi(struct i40e_vsi *vsi)
13081 {
13082         int ret = -ENODEV;
13083         struct i40e_pf *pf = vsi->back;
13084         struct i40e_hw *hw = &pf->hw;
13085         struct i40e_vsi_context ctxt;
13086         struct i40e_mac_filter *f;
13087         struct hlist_node *h;
13088         int bkt;
13089
13090         u8 enabled_tc = 0x1; /* TC0 enabled */
13091         int f_count = 0;
13092
13093         memset(&ctxt, 0, sizeof(ctxt));
13094         switch (vsi->type) {
13095         case I40E_VSI_MAIN:
13096                 /* The PF's main VSI is already setup as part of the
13097                  * device initialization, so we'll not bother with
13098                  * the add_vsi call, but we will retrieve the current
13099                  * VSI context.
13100                  */
13101                 ctxt.seid = pf->main_vsi_seid;
13102                 ctxt.pf_num = pf->hw.pf_id;
13103                 ctxt.vf_num = 0;
13104                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
13105                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13106                 if (ret) {
13107                         dev_info(&pf->pdev->dev,
13108                                  "couldn't get PF vsi config, err %s aq_err %s\n",
13109                                  i40e_stat_str(&pf->hw, ret),
13110                                  i40e_aq_str(&pf->hw,
13111                                              pf->hw.aq.asq_last_status));
13112                         return -ENOENT;
13113                 }
13114                 vsi->info = ctxt.info;
13115                 vsi->info.valid_sections = 0;
13116
13117                 vsi->seid = ctxt.seid;
13118                 vsi->id = ctxt.vsi_number;
13119
13120                 enabled_tc = i40e_pf_get_tc_map(pf);
13121
13122                 /* Source pruning is enabled by default, so the flag is
13123                  * negative logic - if it's set, we need to fiddle with
13124                  * the VSI to disable source pruning.
13125                  */
13126                 if (pf->flags & I40E_FLAG_SOURCE_PRUNING_DISABLED) {
13127                         memset(&ctxt, 0, sizeof(ctxt));
13128                         ctxt.seid = pf->main_vsi_seid;
13129                         ctxt.pf_num = pf->hw.pf_id;
13130                         ctxt.vf_num = 0;
13131                         ctxt.info.valid_sections |=
13132                                      cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13133                         ctxt.info.switch_id =
13134                                    cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
13135                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13136                         if (ret) {
13137                                 dev_info(&pf->pdev->dev,
13138                                          "update vsi failed, err %s aq_err %s\n",
13139                                          i40e_stat_str(&pf->hw, ret),
13140                                          i40e_aq_str(&pf->hw,
13141                                                      pf->hw.aq.asq_last_status));
13142                                 ret = -ENOENT;
13143                                 goto err;
13144                         }
13145                 }
13146
13147                 /* MFP mode setup queue map and update VSI */
13148                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
13149                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
13150                         memset(&ctxt, 0, sizeof(ctxt));
13151                         ctxt.seid = pf->main_vsi_seid;
13152                         ctxt.pf_num = pf->hw.pf_id;
13153                         ctxt.vf_num = 0;
13154                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
13155                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13156                         if (ret) {
13157                                 dev_info(&pf->pdev->dev,
13158                                          "update vsi failed, err %s aq_err %s\n",
13159                                          i40e_stat_str(&pf->hw, ret),
13160                                          i40e_aq_str(&pf->hw,
13161                                                     pf->hw.aq.asq_last_status));
13162                                 ret = -ENOENT;
13163                                 goto err;
13164                         }
13165                         /* update the local VSI info queue map */
13166                         i40e_vsi_update_queue_map(vsi, &ctxt);
13167                         vsi->info.valid_sections = 0;
13168                 } else {
13169                         /* Default/Main VSI is only enabled for TC0
13170                          * reconfigure it to enable all TCs that are
13171                          * available on the port in SFP mode.
13172                          * For MFP case the iSCSI PF would use this
13173                          * flow to enable LAN+iSCSI TC.
13174                          */
13175                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
13176                         if (ret) {
13177                                 /* Single TC condition is not fatal,
13178                                  * message and continue
13179                                  */
13180                                 dev_info(&pf->pdev->dev,
13181                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
13182                                          enabled_tc,
13183                                          i40e_stat_str(&pf->hw, ret),
13184                                          i40e_aq_str(&pf->hw,
13185                                                     pf->hw.aq.asq_last_status));
13186                         }
13187                 }
13188                 break;
13189
13190         case I40E_VSI_FDIR:
13191                 ctxt.pf_num = hw->pf_id;
13192                 ctxt.vf_num = 0;
13193                 ctxt.uplink_seid = vsi->uplink_seid;
13194                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13195                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13196                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
13197                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
13198                         ctxt.info.valid_sections |=
13199                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13200                         ctxt.info.switch_id =
13201                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13202                 }
13203                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13204                 break;
13205
13206         case I40E_VSI_VMDQ2:
13207                 ctxt.pf_num = hw->pf_id;
13208                 ctxt.vf_num = 0;
13209                 ctxt.uplink_seid = vsi->uplink_seid;
13210                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13211                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
13212
13213                 /* This VSI is connected to VEB so the switch_id
13214                  * should be set to zero by default.
13215                  */
13216                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13217                         ctxt.info.valid_sections |=
13218                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13219                         ctxt.info.switch_id =
13220                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13221                 }
13222
13223                 /* Setup the VSI tx/rx queue map for TC0 only for now */
13224                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13225                 break;
13226
13227         case I40E_VSI_SRIOV:
13228                 ctxt.pf_num = hw->pf_id;
13229                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
13230                 ctxt.uplink_seid = vsi->uplink_seid;
13231                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13232                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
13233
13234                 /* This VSI is connected to VEB so the switch_id
13235                  * should be set to zero by default.
13236                  */
13237                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13238                         ctxt.info.valid_sections |=
13239                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13240                         ctxt.info.switch_id =
13241                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13242                 }
13243
13244                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
13245                         ctxt.info.valid_sections |=
13246                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
13247                         ctxt.info.queueing_opt_flags |=
13248                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
13249                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
13250                 }
13251
13252                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
13253                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
13254                 if (pf->vf[vsi->vf_id].spoofchk) {
13255                         ctxt.info.valid_sections |=
13256                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
13257                         ctxt.info.sec_flags |=
13258                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
13259                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
13260                 }
13261                 /* Setup the VSI tx/rx queue map for TC0 only for now */
13262                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13263                 break;
13264
13265         case I40E_VSI_IWARP:
13266                 /* send down message to iWARP */
13267                 break;
13268
13269         default:
13270                 return -ENODEV;
13271         }
13272
13273         if (vsi->type != I40E_VSI_MAIN) {
13274                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
13275                 if (ret) {
13276                         dev_info(&vsi->back->pdev->dev,
13277                                  "add vsi failed, err %s aq_err %s\n",
13278                                  i40e_stat_str(&pf->hw, ret),
13279                                  i40e_aq_str(&pf->hw,
13280                                              pf->hw.aq.asq_last_status));
13281                         ret = -ENOENT;
13282                         goto err;
13283                 }
13284                 vsi->info = ctxt.info;
13285                 vsi->info.valid_sections = 0;
13286                 vsi->seid = ctxt.seid;
13287                 vsi->id = ctxt.vsi_number;
13288         }
13289
13290         vsi->active_filters = 0;
13291         clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
13292         spin_lock_bh(&vsi->mac_filter_hash_lock);
13293         /* If macvlan filters already exist, force them to get loaded */
13294         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
13295                 f->state = I40E_FILTER_NEW;
13296                 f_count++;
13297         }
13298         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13299
13300         if (f_count) {
13301                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
13302                 set_bit(__I40E_MACVLAN_SYNC_PENDING, pf->state);
13303         }
13304
13305         /* Update VSI BW information */
13306         ret = i40e_vsi_get_bw_info(vsi);
13307         if (ret) {
13308                 dev_info(&pf->pdev->dev,
13309                          "couldn't get vsi bw info, err %s aq_err %s\n",
13310                          i40e_stat_str(&pf->hw, ret),
13311                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13312                 /* VSI is already added so not tearing that up */
13313                 ret = 0;
13314         }
13315
13316 err:
13317         return ret;
13318 }
13319
13320 /**
13321  * i40e_vsi_release - Delete a VSI and free its resources
13322  * @vsi: the VSI being removed
13323  *
13324  * Returns 0 on success or < 0 on error
13325  **/
13326 int i40e_vsi_release(struct i40e_vsi *vsi)
13327 {
13328         struct i40e_mac_filter *f;
13329         struct hlist_node *h;
13330         struct i40e_veb *veb = NULL;
13331         struct i40e_pf *pf;
13332         u16 uplink_seid;
13333         int i, n, bkt;
13334
13335         pf = vsi->back;
13336
13337         /* release of a VEB-owner or last VSI is not allowed */
13338         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
13339                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
13340                          vsi->seid, vsi->uplink_seid);
13341                 return -ENODEV;
13342         }
13343         if (vsi == pf->vsi[pf->lan_vsi] &&
13344             !test_bit(__I40E_DOWN, pf->state)) {
13345                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
13346                 return -ENODEV;
13347         }
13348
13349         uplink_seid = vsi->uplink_seid;
13350         if (vsi->type != I40E_VSI_SRIOV) {
13351                 if (vsi->netdev_registered) {
13352                         vsi->netdev_registered = false;
13353                         if (vsi->netdev) {
13354                                 /* results in a call to i40e_close() */
13355                                 unregister_netdev(vsi->netdev);
13356                         }
13357                 } else {
13358                         i40e_vsi_close(vsi);
13359                 }
13360                 i40e_vsi_disable_irq(vsi);
13361         }
13362
13363         spin_lock_bh(&vsi->mac_filter_hash_lock);
13364
13365         /* clear the sync flag on all filters */
13366         if (vsi->netdev) {
13367                 __dev_uc_unsync(vsi->netdev, NULL);
13368                 __dev_mc_unsync(vsi->netdev, NULL);
13369         }
13370
13371         /* make sure any remaining filters are marked for deletion */
13372         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
13373                 __i40e_del_filter(vsi, f);
13374
13375         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13376
13377         i40e_sync_vsi_filters(vsi);
13378
13379         i40e_vsi_delete(vsi);
13380         i40e_vsi_free_q_vectors(vsi);
13381         if (vsi->netdev) {
13382                 free_netdev(vsi->netdev);
13383                 vsi->netdev = NULL;
13384         }
13385         i40e_vsi_clear_rings(vsi);
13386         i40e_vsi_clear(vsi);
13387
13388         /* If this was the last thing on the VEB, except for the
13389          * controlling VSI, remove the VEB, which puts the controlling
13390          * VSI onto the next level down in the switch.
13391          *
13392          * Well, okay, there's one more exception here: don't remove
13393          * the orphan VEBs yet.  We'll wait for an explicit remove request
13394          * from up the network stack.
13395          */
13396         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
13397                 if (pf->vsi[i] &&
13398                     pf->vsi[i]->uplink_seid == uplink_seid &&
13399                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13400                         n++;      /* count the VSIs */
13401                 }
13402         }
13403         for (i = 0; i < I40E_MAX_VEB; i++) {
13404                 if (!pf->veb[i])
13405                         continue;
13406                 if (pf->veb[i]->uplink_seid == uplink_seid)
13407                         n++;     /* count the VEBs */
13408                 if (pf->veb[i]->seid == uplink_seid)
13409                         veb = pf->veb[i];
13410         }
13411         if (n == 0 && veb && veb->uplink_seid != 0)
13412                 i40e_veb_release(veb);
13413
13414         return 0;
13415 }
13416
13417 /**
13418  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
13419  * @vsi: ptr to the VSI
13420  *
13421  * This should only be called after i40e_vsi_mem_alloc() which allocates the
13422  * corresponding SW VSI structure and initializes num_queue_pairs for the
13423  * newly allocated VSI.
13424  *
13425  * Returns 0 on success or negative on failure
13426  **/
13427 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
13428 {
13429         int ret = -ENOENT;
13430         struct i40e_pf *pf = vsi->back;
13431
13432         if (vsi->q_vectors[0]) {
13433                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
13434                          vsi->seid);
13435                 return -EEXIST;
13436         }
13437
13438         if (vsi->base_vector) {
13439                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
13440                          vsi->seid, vsi->base_vector);
13441                 return -EEXIST;
13442         }
13443
13444         ret = i40e_vsi_alloc_q_vectors(vsi);
13445         if (ret) {
13446                 dev_info(&pf->pdev->dev,
13447                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
13448                          vsi->num_q_vectors, vsi->seid, ret);
13449                 vsi->num_q_vectors = 0;
13450                 goto vector_setup_out;
13451         }
13452
13453         /* In Legacy mode, we do not have to get any other vector since we
13454          * piggyback on the misc/ICR0 for queue interrupts.
13455         */
13456         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
13457                 return ret;
13458         if (vsi->num_q_vectors)
13459                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
13460                                                  vsi->num_q_vectors, vsi->idx);
13461         if (vsi->base_vector < 0) {
13462                 dev_info(&pf->pdev->dev,
13463                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
13464                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
13465                 i40e_vsi_free_q_vectors(vsi);
13466                 ret = -ENOENT;
13467                 goto vector_setup_out;
13468         }
13469
13470 vector_setup_out:
13471         return ret;
13472 }
13473
13474 /**
13475  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
13476  * @vsi: pointer to the vsi.
13477  *
13478  * This re-allocates a vsi's queue resources.
13479  *
13480  * Returns pointer to the successfully allocated and configured VSI sw struct
13481  * on success, otherwise returns NULL on failure.
13482  **/
13483 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
13484 {
13485         u16 alloc_queue_pairs;
13486         struct i40e_pf *pf;
13487         u8 enabled_tc;
13488         int ret;
13489
13490         if (!vsi)
13491                 return NULL;
13492
13493         pf = vsi->back;
13494
13495         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
13496         i40e_vsi_clear_rings(vsi);
13497
13498         i40e_vsi_free_arrays(vsi, false);
13499         i40e_set_num_rings_in_vsi(vsi);
13500         ret = i40e_vsi_alloc_arrays(vsi, false);
13501         if (ret)
13502                 goto err_vsi;
13503
13504         alloc_queue_pairs = vsi->alloc_queue_pairs *
13505                             (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13506
13507         ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13508         if (ret < 0) {
13509                 dev_info(&pf->pdev->dev,
13510                          "failed to get tracking for %d queues for VSI %d err %d\n",
13511                          alloc_queue_pairs, vsi->seid, ret);
13512                 goto err_vsi;
13513         }
13514         vsi->base_queue = ret;
13515
13516         /* Update the FW view of the VSI. Force a reset of TC and queue
13517          * layout configurations.
13518          */
13519         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
13520         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
13521         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
13522         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
13523         if (vsi->type == I40E_VSI_MAIN)
13524                 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
13525
13526         /* assign it some queues */
13527         ret = i40e_alloc_rings(vsi);
13528         if (ret)
13529                 goto err_rings;
13530
13531         /* map all of the rings to the q_vectors */
13532         i40e_vsi_map_rings_to_vectors(vsi);
13533         return vsi;
13534
13535 err_rings:
13536         i40e_vsi_free_q_vectors(vsi);
13537         if (vsi->netdev_registered) {
13538                 vsi->netdev_registered = false;
13539                 unregister_netdev(vsi->netdev);
13540                 free_netdev(vsi->netdev);
13541                 vsi->netdev = NULL;
13542         }
13543         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13544 err_vsi:
13545         i40e_vsi_clear(vsi);
13546         return NULL;
13547 }
13548
13549 /**
13550  * i40e_vsi_setup - Set up a VSI by a given type
13551  * @pf: board private structure
13552  * @type: VSI type
13553  * @uplink_seid: the switch element to link to
13554  * @param1: usage depends upon VSI type. For VF types, indicates VF id
13555  *
13556  * This allocates the sw VSI structure and its queue resources, then add a VSI
13557  * to the identified VEB.
13558  *
13559  * Returns pointer to the successfully allocated and configure VSI sw struct on
13560  * success, otherwise returns NULL on failure.
13561  **/
13562 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
13563                                 u16 uplink_seid, u32 param1)
13564 {
13565         struct i40e_vsi *vsi = NULL;
13566         struct i40e_veb *veb = NULL;
13567         u16 alloc_queue_pairs;
13568         int ret, i;
13569         int v_idx;
13570
13571         /* The requested uplink_seid must be either
13572          *     - the PF's port seid
13573          *              no VEB is needed because this is the PF
13574          *              or this is a Flow Director special case VSI
13575          *     - seid of an existing VEB
13576          *     - seid of a VSI that owns an existing VEB
13577          *     - seid of a VSI that doesn't own a VEB
13578          *              a new VEB is created and the VSI becomes the owner
13579          *     - seid of the PF VSI, which is what creates the first VEB
13580          *              this is a special case of the previous
13581          *
13582          * Find which uplink_seid we were given and create a new VEB if needed
13583          */
13584         for (i = 0; i < I40E_MAX_VEB; i++) {
13585                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
13586                         veb = pf->veb[i];
13587                         break;
13588                 }
13589         }
13590
13591         if (!veb && uplink_seid != pf->mac_seid) {
13592
13593                 for (i = 0; i < pf->num_alloc_vsi; i++) {
13594                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
13595                                 vsi = pf->vsi[i];
13596                                 break;
13597                         }
13598                 }
13599                 if (!vsi) {
13600                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
13601                                  uplink_seid);
13602                         return NULL;
13603                 }
13604
13605                 if (vsi->uplink_seid == pf->mac_seid)
13606                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
13607                                              vsi->tc_config.enabled_tc);
13608                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
13609                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
13610                                              vsi->tc_config.enabled_tc);
13611                 if (veb) {
13612                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
13613                                 dev_info(&vsi->back->pdev->dev,
13614                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
13615                                 return NULL;
13616                         }
13617                         /* We come up by default in VEPA mode if SRIOV is not
13618                          * already enabled, in which case we can't force VEPA
13619                          * mode.
13620                          */
13621                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
13622                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
13623                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
13624                         }
13625                         i40e_config_bridge_mode(veb);
13626                 }
13627                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
13628                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
13629                                 veb = pf->veb[i];
13630                 }
13631                 if (!veb) {
13632                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
13633                         return NULL;
13634                 }
13635
13636                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13637                 uplink_seid = veb->seid;
13638         }
13639
13640         /* get vsi sw struct */
13641         v_idx = i40e_vsi_mem_alloc(pf, type);
13642         if (v_idx < 0)
13643                 goto err_alloc;
13644         vsi = pf->vsi[v_idx];
13645         if (!vsi)
13646                 goto err_alloc;
13647         vsi->type = type;
13648         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
13649
13650         if (type == I40E_VSI_MAIN)
13651                 pf->lan_vsi = v_idx;
13652         else if (type == I40E_VSI_SRIOV)
13653                 vsi->vf_id = param1;
13654         /* assign it some queues */
13655         alloc_queue_pairs = vsi->alloc_queue_pairs *
13656                             (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13657
13658         ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13659         if (ret < 0) {
13660                 dev_info(&pf->pdev->dev,
13661                          "failed to get tracking for %d queues for VSI %d err=%d\n",
13662                          alloc_queue_pairs, vsi->seid, ret);
13663                 goto err_vsi;
13664         }
13665         vsi->base_queue = ret;
13666
13667         /* get a VSI from the hardware */
13668         vsi->uplink_seid = uplink_seid;
13669         ret = i40e_add_vsi(vsi);
13670         if (ret)
13671                 goto err_vsi;
13672
13673         switch (vsi->type) {
13674         /* setup the netdev if needed */
13675         case I40E_VSI_MAIN:
13676         case I40E_VSI_VMDQ2:
13677                 ret = i40e_config_netdev(vsi);
13678                 if (ret)
13679                         goto err_netdev;
13680                 ret = register_netdev(vsi->netdev);
13681                 if (ret)
13682                         goto err_netdev;
13683                 vsi->netdev_registered = true;
13684                 netif_carrier_off(vsi->netdev);
13685 #ifdef CONFIG_I40E_DCB
13686                 /* Setup DCB netlink interface */
13687                 i40e_dcbnl_setup(vsi);
13688 #endif /* CONFIG_I40E_DCB */
13689                 /* fall through */
13690
13691         case I40E_VSI_FDIR:
13692                 /* set up vectors and rings if needed */
13693                 ret = i40e_vsi_setup_vectors(vsi);
13694                 if (ret)
13695                         goto err_msix;
13696
13697                 ret = i40e_alloc_rings(vsi);
13698                 if (ret)
13699                         goto err_rings;
13700
13701                 /* map all of the rings to the q_vectors */
13702                 i40e_vsi_map_rings_to_vectors(vsi);
13703
13704                 i40e_vsi_reset_stats(vsi);
13705                 break;
13706
13707         default:
13708                 /* no netdev or rings for the other VSI types */
13709                 break;
13710         }
13711
13712         if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
13713             (vsi->type == I40E_VSI_VMDQ2)) {
13714                 ret = i40e_vsi_config_rss(vsi);
13715         }
13716         return vsi;
13717
13718 err_rings:
13719         i40e_vsi_free_q_vectors(vsi);
13720 err_msix:
13721         if (vsi->netdev_registered) {
13722                 vsi->netdev_registered = false;
13723                 unregister_netdev(vsi->netdev);
13724                 free_netdev(vsi->netdev);
13725                 vsi->netdev = NULL;
13726         }
13727 err_netdev:
13728         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13729 err_vsi:
13730         i40e_vsi_clear(vsi);
13731 err_alloc:
13732         return NULL;
13733 }
13734
13735 /**
13736  * i40e_veb_get_bw_info - Query VEB BW information
13737  * @veb: the veb to query
13738  *
13739  * Query the Tx scheduler BW configuration data for given VEB
13740  **/
13741 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
13742 {
13743         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
13744         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
13745         struct i40e_pf *pf = veb->pf;
13746         struct i40e_hw *hw = &pf->hw;
13747         u32 tc_bw_max;
13748         int ret = 0;
13749         int i;
13750
13751         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
13752                                                   &bw_data, NULL);
13753         if (ret) {
13754                 dev_info(&pf->pdev->dev,
13755                          "query veb bw config failed, err %s aq_err %s\n",
13756                          i40e_stat_str(&pf->hw, ret),
13757                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13758                 goto out;
13759         }
13760
13761         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
13762                                                    &ets_data, NULL);
13763         if (ret) {
13764                 dev_info(&pf->pdev->dev,
13765                          "query veb bw ets config failed, err %s aq_err %s\n",
13766                          i40e_stat_str(&pf->hw, ret),
13767                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13768                 goto out;
13769         }
13770
13771         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
13772         veb->bw_max_quanta = ets_data.tc_bw_max;
13773         veb->is_abs_credits = bw_data.absolute_credits_enable;
13774         veb->enabled_tc = ets_data.tc_valid_bits;
13775         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
13776                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
13777         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
13778                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
13779                 veb->bw_tc_limit_credits[i] =
13780                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
13781                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
13782         }
13783
13784 out:
13785         return ret;
13786 }
13787
13788 /**
13789  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
13790  * @pf: board private structure
13791  *
13792  * On error: returns error code (negative)
13793  * On success: returns vsi index in PF (positive)
13794  **/
13795 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
13796 {
13797         int ret = -ENOENT;
13798         struct i40e_veb *veb;
13799         int i;
13800
13801         /* Need to protect the allocation of switch elements at the PF level */
13802         mutex_lock(&pf->switch_mutex);
13803
13804         /* VEB list may be fragmented if VEB creation/destruction has
13805          * been happening.  We can afford to do a quick scan to look
13806          * for any free slots in the list.
13807          *
13808          * find next empty veb slot, looping back around if necessary
13809          */
13810         i = 0;
13811         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
13812                 i++;
13813         if (i >= I40E_MAX_VEB) {
13814                 ret = -ENOMEM;
13815                 goto err_alloc_veb;  /* out of VEB slots! */
13816         }
13817
13818         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
13819         if (!veb) {
13820                 ret = -ENOMEM;
13821                 goto err_alloc_veb;
13822         }
13823         veb->pf = pf;
13824         veb->idx = i;
13825         veb->enabled_tc = 1;
13826
13827         pf->veb[i] = veb;
13828         ret = i;
13829 err_alloc_veb:
13830         mutex_unlock(&pf->switch_mutex);
13831         return ret;
13832 }
13833
13834 /**
13835  * i40e_switch_branch_release - Delete a branch of the switch tree
13836  * @branch: where to start deleting
13837  *
13838  * This uses recursion to find the tips of the branch to be
13839  * removed, deleting until we get back to and can delete this VEB.
13840  **/
13841 static void i40e_switch_branch_release(struct i40e_veb *branch)
13842 {
13843         struct i40e_pf *pf = branch->pf;
13844         u16 branch_seid = branch->seid;
13845         u16 veb_idx = branch->idx;
13846         int i;
13847
13848         /* release any VEBs on this VEB - RECURSION */
13849         for (i = 0; i < I40E_MAX_VEB; i++) {
13850                 if (!pf->veb[i])
13851                         continue;
13852                 if (pf->veb[i]->uplink_seid == branch->seid)
13853                         i40e_switch_branch_release(pf->veb[i]);
13854         }
13855
13856         /* Release the VSIs on this VEB, but not the owner VSI.
13857          *
13858          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
13859          *       the VEB itself, so don't use (*branch) after this loop.
13860          */
13861         for (i = 0; i < pf->num_alloc_vsi; i++) {
13862                 if (!pf->vsi[i])
13863                         continue;
13864                 if (pf->vsi[i]->uplink_seid == branch_seid &&
13865                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13866                         i40e_vsi_release(pf->vsi[i]);
13867                 }
13868         }
13869
13870         /* There's one corner case where the VEB might not have been
13871          * removed, so double check it here and remove it if needed.
13872          * This case happens if the veb was created from the debugfs
13873          * commands and no VSIs were added to it.
13874          */
13875         if (pf->veb[veb_idx])
13876                 i40e_veb_release(pf->veb[veb_idx]);
13877 }
13878
13879 /**
13880  * i40e_veb_clear - remove veb struct
13881  * @veb: the veb to remove
13882  **/
13883 static void i40e_veb_clear(struct i40e_veb *veb)
13884 {
13885         if (!veb)
13886                 return;
13887
13888         if (veb->pf) {
13889                 struct i40e_pf *pf = veb->pf;
13890
13891                 mutex_lock(&pf->switch_mutex);
13892                 if (pf->veb[veb->idx] == veb)
13893                         pf->veb[veb->idx] = NULL;
13894                 mutex_unlock(&pf->switch_mutex);
13895         }
13896
13897         kfree(veb);
13898 }
13899
13900 /**
13901  * i40e_veb_release - Delete a VEB and free its resources
13902  * @veb: the VEB being removed
13903  **/
13904 void i40e_veb_release(struct i40e_veb *veb)
13905 {
13906         struct i40e_vsi *vsi = NULL;
13907         struct i40e_pf *pf;
13908         int i, n = 0;
13909
13910         pf = veb->pf;
13911
13912         /* find the remaining VSI and check for extras */
13913         for (i = 0; i < pf->num_alloc_vsi; i++) {
13914                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
13915                         n++;
13916                         vsi = pf->vsi[i];
13917                 }
13918         }
13919         if (n != 1) {
13920                 dev_info(&pf->pdev->dev,
13921                          "can't remove VEB %d with %d VSIs left\n",
13922                          veb->seid, n);
13923                 return;
13924         }
13925
13926         /* move the remaining VSI to uplink veb */
13927         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
13928         if (veb->uplink_seid) {
13929                 vsi->uplink_seid = veb->uplink_seid;
13930                 if (veb->uplink_seid == pf->mac_seid)
13931                         vsi->veb_idx = I40E_NO_VEB;
13932                 else
13933                         vsi->veb_idx = veb->veb_idx;
13934         } else {
13935                 /* floating VEB */
13936                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
13937                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
13938         }
13939
13940         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13941         i40e_veb_clear(veb);
13942 }
13943
13944 /**
13945  * i40e_add_veb - create the VEB in the switch
13946  * @veb: the VEB to be instantiated
13947  * @vsi: the controlling VSI
13948  **/
13949 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
13950 {
13951         struct i40e_pf *pf = veb->pf;
13952         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
13953         int ret;
13954
13955         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
13956                               veb->enabled_tc, false,
13957                               &veb->seid, enable_stats, NULL);
13958
13959         /* get a VEB from the hardware */
13960         if (ret) {
13961                 dev_info(&pf->pdev->dev,
13962                          "couldn't add VEB, err %s aq_err %s\n",
13963                          i40e_stat_str(&pf->hw, ret),
13964                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13965                 return -EPERM;
13966         }
13967
13968         /* get statistics counter */
13969         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
13970                                          &veb->stats_idx, NULL, NULL, NULL);
13971         if (ret) {
13972                 dev_info(&pf->pdev->dev,
13973                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
13974                          i40e_stat_str(&pf->hw, ret),
13975                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13976                 return -EPERM;
13977         }
13978         ret = i40e_veb_get_bw_info(veb);
13979         if (ret) {
13980                 dev_info(&pf->pdev->dev,
13981                          "couldn't get VEB bw info, err %s aq_err %s\n",
13982                          i40e_stat_str(&pf->hw, ret),
13983                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13984                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13985                 return -ENOENT;
13986         }
13987
13988         vsi->uplink_seid = veb->seid;
13989         vsi->veb_idx = veb->idx;
13990         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13991
13992         return 0;
13993 }
13994
13995 /**
13996  * i40e_veb_setup - Set up a VEB
13997  * @pf: board private structure
13998  * @flags: VEB setup flags
13999  * @uplink_seid: the switch element to link to
14000  * @vsi_seid: the initial VSI seid
14001  * @enabled_tc: Enabled TC bit-map
14002  *
14003  * This allocates the sw VEB structure and links it into the switch
14004  * It is possible and legal for this to be a duplicate of an already
14005  * existing VEB.  It is also possible for both uplink and vsi seids
14006  * to be zero, in order to create a floating VEB.
14007  *
14008  * Returns pointer to the successfully allocated VEB sw struct on
14009  * success, otherwise returns NULL on failure.
14010  **/
14011 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
14012                                 u16 uplink_seid, u16 vsi_seid,
14013                                 u8 enabled_tc)
14014 {
14015         struct i40e_veb *veb, *uplink_veb = NULL;
14016         int vsi_idx, veb_idx;
14017         int ret;
14018
14019         /* if one seid is 0, the other must be 0 to create a floating relay */
14020         if ((uplink_seid == 0 || vsi_seid == 0) &&
14021             (uplink_seid + vsi_seid != 0)) {
14022                 dev_info(&pf->pdev->dev,
14023                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
14024                          uplink_seid, vsi_seid);
14025                 return NULL;
14026         }
14027
14028         /* make sure there is such a vsi and uplink */
14029         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
14030                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
14031                         break;
14032         if (vsi_idx == pf->num_alloc_vsi && vsi_seid != 0) {
14033                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
14034                          vsi_seid);
14035                 return NULL;
14036         }
14037
14038         if (uplink_seid && uplink_seid != pf->mac_seid) {
14039                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
14040                         if (pf->veb[veb_idx] &&
14041                             pf->veb[veb_idx]->seid == uplink_seid) {
14042                                 uplink_veb = pf->veb[veb_idx];
14043                                 break;
14044                         }
14045                 }
14046                 if (!uplink_veb) {
14047                         dev_info(&pf->pdev->dev,
14048                                  "uplink seid %d not found\n", uplink_seid);
14049                         return NULL;
14050                 }
14051         }
14052
14053         /* get veb sw struct */
14054         veb_idx = i40e_veb_mem_alloc(pf);
14055         if (veb_idx < 0)
14056                 goto err_alloc;
14057         veb = pf->veb[veb_idx];
14058         veb->flags = flags;
14059         veb->uplink_seid = uplink_seid;
14060         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
14061         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
14062
14063         /* create the VEB in the switch */
14064         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
14065         if (ret)
14066                 goto err_veb;
14067         if (vsi_idx == pf->lan_vsi)
14068                 pf->lan_veb = veb->idx;
14069
14070         return veb;
14071
14072 err_veb:
14073         i40e_veb_clear(veb);
14074 err_alloc:
14075         return NULL;
14076 }
14077
14078 /**
14079  * i40e_setup_pf_switch_element - set PF vars based on switch type
14080  * @pf: board private structure
14081  * @ele: element we are building info from
14082  * @num_reported: total number of elements
14083  * @printconfig: should we print the contents
14084  *
14085  * helper function to assist in extracting a few useful SEID values.
14086  **/
14087 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
14088                                 struct i40e_aqc_switch_config_element_resp *ele,
14089                                 u16 num_reported, bool printconfig)
14090 {
14091         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
14092         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
14093         u8 element_type = ele->element_type;
14094         u16 seid = le16_to_cpu(ele->seid);
14095
14096         if (printconfig)
14097                 dev_info(&pf->pdev->dev,
14098                          "type=%d seid=%d uplink=%d downlink=%d\n",
14099                          element_type, seid, uplink_seid, downlink_seid);
14100
14101         switch (element_type) {
14102         case I40E_SWITCH_ELEMENT_TYPE_MAC:
14103                 pf->mac_seid = seid;
14104                 break;
14105         case I40E_SWITCH_ELEMENT_TYPE_VEB:
14106                 /* Main VEB? */
14107                 if (uplink_seid != pf->mac_seid)
14108                         break;
14109                 if (pf->lan_veb >= I40E_MAX_VEB) {
14110                         int v;
14111
14112                         /* find existing or else empty VEB */
14113                         for (v = 0; v < I40E_MAX_VEB; v++) {
14114                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
14115                                         pf->lan_veb = v;
14116                                         break;
14117                                 }
14118                         }
14119                         if (pf->lan_veb >= I40E_MAX_VEB) {
14120                                 v = i40e_veb_mem_alloc(pf);
14121                                 if (v < 0)
14122                                         break;
14123                                 pf->lan_veb = v;
14124                         }
14125                 }
14126                 if (pf->lan_veb >= I40E_MAX_VEB)
14127                         break;
14128
14129                 pf->veb[pf->lan_veb]->seid = seid;
14130                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
14131                 pf->veb[pf->lan_veb]->pf = pf;
14132                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
14133                 break;
14134         case I40E_SWITCH_ELEMENT_TYPE_VSI:
14135                 if (num_reported != 1)
14136                         break;
14137                 /* This is immediately after a reset so we can assume this is
14138                  * the PF's VSI
14139                  */
14140                 pf->mac_seid = uplink_seid;
14141                 pf->pf_seid = downlink_seid;
14142                 pf->main_vsi_seid = seid;
14143                 if (printconfig)
14144                         dev_info(&pf->pdev->dev,
14145                                  "pf_seid=%d main_vsi_seid=%d\n",
14146                                  pf->pf_seid, pf->main_vsi_seid);
14147                 break;
14148         case I40E_SWITCH_ELEMENT_TYPE_PF:
14149         case I40E_SWITCH_ELEMENT_TYPE_VF:
14150         case I40E_SWITCH_ELEMENT_TYPE_EMP:
14151         case I40E_SWITCH_ELEMENT_TYPE_BMC:
14152         case I40E_SWITCH_ELEMENT_TYPE_PE:
14153         case I40E_SWITCH_ELEMENT_TYPE_PA:
14154                 /* ignore these for now */
14155                 break;
14156         default:
14157                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
14158                          element_type, seid);
14159                 break;
14160         }
14161 }
14162
14163 /**
14164  * i40e_fetch_switch_configuration - Get switch config from firmware
14165  * @pf: board private structure
14166  * @printconfig: should we print the contents
14167  *
14168  * Get the current switch configuration from the device and
14169  * extract a few useful SEID values.
14170  **/
14171 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
14172 {
14173         struct i40e_aqc_get_switch_config_resp *sw_config;
14174         u16 next_seid = 0;
14175         int ret = 0;
14176         u8 *aq_buf;
14177         int i;
14178
14179         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
14180         if (!aq_buf)
14181                 return -ENOMEM;
14182
14183         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
14184         do {
14185                 u16 num_reported, num_total;
14186
14187                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
14188                                                 I40E_AQ_LARGE_BUF,
14189                                                 &next_seid, NULL);
14190                 if (ret) {
14191                         dev_info(&pf->pdev->dev,
14192                                  "get switch config failed err %s aq_err %s\n",
14193                                  i40e_stat_str(&pf->hw, ret),
14194                                  i40e_aq_str(&pf->hw,
14195                                              pf->hw.aq.asq_last_status));
14196                         kfree(aq_buf);
14197                         return -ENOENT;
14198                 }
14199
14200                 num_reported = le16_to_cpu(sw_config->header.num_reported);
14201                 num_total = le16_to_cpu(sw_config->header.num_total);
14202
14203                 if (printconfig)
14204                         dev_info(&pf->pdev->dev,
14205                                  "header: %d reported %d total\n",
14206                                  num_reported, num_total);
14207
14208                 for (i = 0; i < num_reported; i++) {
14209                         struct i40e_aqc_switch_config_element_resp *ele =
14210                                 &sw_config->element[i];
14211
14212                         i40e_setup_pf_switch_element(pf, ele, num_reported,
14213                                                      printconfig);
14214                 }
14215         } while (next_seid != 0);
14216
14217         kfree(aq_buf);
14218         return ret;
14219 }
14220
14221 /**
14222  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
14223  * @pf: board private structure
14224  * @reinit: if the Main VSI needs to re-initialized.
14225  *
14226  * Returns 0 on success, negative value on failure
14227  **/
14228 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
14229 {
14230         u16 flags = 0;
14231         int ret;
14232
14233         /* find out what's out there already */
14234         ret = i40e_fetch_switch_configuration(pf, false);
14235         if (ret) {
14236                 dev_info(&pf->pdev->dev,
14237                          "couldn't fetch switch config, err %s aq_err %s\n",
14238                          i40e_stat_str(&pf->hw, ret),
14239                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
14240                 return ret;
14241         }
14242         i40e_pf_reset_stats(pf);
14243
14244         /* set the switch config bit for the whole device to
14245          * support limited promisc or true promisc
14246          * when user requests promisc. The default is limited
14247          * promisc.
14248         */
14249
14250         if ((pf->hw.pf_id == 0) &&
14251             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT)) {
14252                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14253                 pf->last_sw_conf_flags = flags;
14254         }
14255
14256         if (pf->hw.pf_id == 0) {
14257                 u16 valid_flags;
14258
14259                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14260                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags, 0,
14261                                                 NULL);
14262                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
14263                         dev_info(&pf->pdev->dev,
14264                                  "couldn't set switch config bits, err %s aq_err %s\n",
14265                                  i40e_stat_str(&pf->hw, ret),
14266                                  i40e_aq_str(&pf->hw,
14267                                              pf->hw.aq.asq_last_status));
14268                         /* not a fatal problem, just keep going */
14269                 }
14270                 pf->last_sw_conf_valid_flags = valid_flags;
14271         }
14272
14273         /* first time setup */
14274         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
14275                 struct i40e_vsi *vsi = NULL;
14276                 u16 uplink_seid;
14277
14278                 /* Set up the PF VSI associated with the PF's main VSI
14279                  * that is already in the HW switch
14280                  */
14281                 if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
14282                         uplink_seid = pf->veb[pf->lan_veb]->seid;
14283                 else
14284                         uplink_seid = pf->mac_seid;
14285                 if (pf->lan_vsi == I40E_NO_VSI)
14286                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
14287                 else if (reinit)
14288                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
14289                 if (!vsi) {
14290                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
14291                         i40e_cloud_filter_exit(pf);
14292                         i40e_fdir_teardown(pf);
14293                         return -EAGAIN;
14294                 }
14295         } else {
14296                 /* force a reset of TC and queue layout configurations */
14297                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
14298
14299                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
14300                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
14301                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
14302         }
14303         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
14304
14305         i40e_fdir_sb_setup(pf);
14306
14307         /* Setup static PF queue filter control settings */
14308         ret = i40e_setup_pf_filter_control(pf);
14309         if (ret) {
14310                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
14311                          ret);
14312                 /* Failure here should not stop continuing other steps */
14313         }
14314
14315         /* enable RSS in the HW, even for only one queue, as the stack can use
14316          * the hash
14317          */
14318         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
14319                 i40e_pf_config_rss(pf);
14320
14321         /* fill in link information and enable LSE reporting */
14322         i40e_link_event(pf);
14323
14324         /* Initialize user-specific link properties */
14325         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
14326                                   I40E_AQ_AN_COMPLETED) ? true : false);
14327
14328         i40e_ptp_init(pf);
14329
14330         /* repopulate tunnel port filters */
14331         i40e_sync_udp_filters(pf);
14332
14333         return ret;
14334 }
14335
14336 /**
14337  * i40e_determine_queue_usage - Work out queue distribution
14338  * @pf: board private structure
14339  **/
14340 static void i40e_determine_queue_usage(struct i40e_pf *pf)
14341 {
14342         int queues_left;
14343         int q_max;
14344
14345         pf->num_lan_qps = 0;
14346
14347         /* Find the max queues to be put into basic use.  We'll always be
14348          * using TC0, whether or not DCB is running, and TC0 will get the
14349          * big RSS set.
14350          */
14351         queues_left = pf->hw.func_caps.num_tx_qp;
14352
14353         if ((queues_left == 1) ||
14354             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
14355                 /* one qp for PF, no queues for anything else */
14356                 queues_left = 0;
14357                 pf->alloc_rss_size = pf->num_lan_qps = 1;
14358
14359                 /* make sure all the fancies are disabled */
14360                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14361                                I40E_FLAG_IWARP_ENABLED  |
14362                                I40E_FLAG_FD_SB_ENABLED  |
14363                                I40E_FLAG_FD_ATR_ENABLED |
14364                                I40E_FLAG_DCB_CAPABLE    |
14365                                I40E_FLAG_DCB_ENABLED    |
14366                                I40E_FLAG_SRIOV_ENABLED  |
14367                                I40E_FLAG_VMDQ_ENABLED);
14368                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14369         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
14370                                   I40E_FLAG_FD_SB_ENABLED |
14371                                   I40E_FLAG_FD_ATR_ENABLED |
14372                                   I40E_FLAG_DCB_CAPABLE))) {
14373                 /* one qp for PF */
14374                 pf->alloc_rss_size = pf->num_lan_qps = 1;
14375                 queues_left -= pf->num_lan_qps;
14376
14377                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14378                                I40E_FLAG_IWARP_ENABLED  |
14379                                I40E_FLAG_FD_SB_ENABLED  |
14380                                I40E_FLAG_FD_ATR_ENABLED |
14381                                I40E_FLAG_DCB_ENABLED    |
14382                                I40E_FLAG_VMDQ_ENABLED);
14383                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14384         } else {
14385                 /* Not enough queues for all TCs */
14386                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
14387                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
14388                         pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
14389                                         I40E_FLAG_DCB_ENABLED);
14390                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
14391                 }
14392
14393                 /* limit lan qps to the smaller of qps, cpus or msix */
14394                 q_max = max_t(int, pf->rss_size_max, num_online_cpus());
14395                 q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
14396                 q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
14397                 pf->num_lan_qps = q_max;
14398
14399                 queues_left -= pf->num_lan_qps;
14400         }
14401
14402         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14403                 if (queues_left > 1) {
14404                         queues_left -= 1; /* save 1 queue for FD */
14405                 } else {
14406                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
14407                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14408                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
14409                 }
14410         }
14411
14412         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
14413             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
14414                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
14415                                         (queues_left / pf->num_vf_qps));
14416                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
14417         }
14418
14419         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
14420             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
14421                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
14422                                           (queues_left / pf->num_vmdq_qps));
14423                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
14424         }
14425
14426         pf->queues_left = queues_left;
14427         dev_dbg(&pf->pdev->dev,
14428                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
14429                 pf->hw.func_caps.num_tx_qp,
14430                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
14431                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
14432                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
14433                 queues_left);
14434 }
14435
14436 /**
14437  * i40e_setup_pf_filter_control - Setup PF static filter control
14438  * @pf: PF to be setup
14439  *
14440  * i40e_setup_pf_filter_control sets up a PF's initial filter control
14441  * settings. If PE/FCoE are enabled then it will also set the per PF
14442  * based filter sizes required for them. It also enables Flow director,
14443  * ethertype and macvlan type filter settings for the pf.
14444  *
14445  * Returns 0 on success, negative on failure
14446  **/
14447 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
14448 {
14449         struct i40e_filter_control_settings *settings = &pf->filter_settings;
14450
14451         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
14452
14453         /* Flow Director is enabled */
14454         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
14455                 settings->enable_fdir = true;
14456
14457         /* Ethtype and MACVLAN filters enabled for PF */
14458         settings->enable_ethtype = true;
14459         settings->enable_macvlan = true;
14460
14461         if (i40e_set_filter_control(&pf->hw, settings))
14462                 return -ENOENT;
14463
14464         return 0;
14465 }
14466
14467 #define INFO_STRING_LEN 255
14468 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
14469 static void i40e_print_features(struct i40e_pf *pf)
14470 {
14471         struct i40e_hw *hw = &pf->hw;
14472         char *buf;
14473         int i;
14474
14475         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
14476         if (!buf)
14477                 return;
14478
14479         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
14480 #ifdef CONFIG_PCI_IOV
14481         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
14482 #endif
14483         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
14484                       pf->hw.func_caps.num_vsis,
14485                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
14486         if (pf->flags & I40E_FLAG_RSS_ENABLED)
14487                 i += snprintf(&buf[i], REMAIN(i), " RSS");
14488         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
14489                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
14490         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14491                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
14492                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
14493         }
14494         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
14495                 i += snprintf(&buf[i], REMAIN(i), " DCB");
14496         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
14497         i += snprintf(&buf[i], REMAIN(i), " Geneve");
14498         if (pf->flags & I40E_FLAG_PTP)
14499                 i += snprintf(&buf[i], REMAIN(i), " PTP");
14500         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
14501                 i += snprintf(&buf[i], REMAIN(i), " VEB");
14502         else
14503                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
14504
14505         dev_info(&pf->pdev->dev, "%s\n", buf);
14506         kfree(buf);
14507         WARN_ON(i > INFO_STRING_LEN);
14508 }
14509
14510 /**
14511  * i40e_get_platform_mac_addr - get platform-specific MAC address
14512  * @pdev: PCI device information struct
14513  * @pf: board private structure
14514  *
14515  * Look up the MAC address for the device. First we'll try
14516  * eth_platform_get_mac_address, which will check Open Firmware, or arch
14517  * specific fallback. Otherwise, we'll default to the stored value in
14518  * firmware.
14519  **/
14520 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
14521 {
14522         if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
14523                 i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
14524 }
14525
14526 /**
14527  * i40e_set_fec_in_flags - helper function for setting FEC options in flags
14528  * @fec_cfg: FEC option to set in flags
14529  * @flags: ptr to flags in which we set FEC option
14530  **/
14531 void i40e_set_fec_in_flags(u8 fec_cfg, u32 *flags)
14532 {
14533         if (fec_cfg & I40E_AQ_SET_FEC_AUTO)
14534                 *flags |= I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC;
14535         if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_RS) ||
14536             (fec_cfg & I40E_AQ_SET_FEC_ABILITY_RS)) {
14537                 *flags |= I40E_FLAG_RS_FEC;
14538                 *flags &= ~I40E_FLAG_BASE_R_FEC;
14539         }
14540         if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_KR) ||
14541             (fec_cfg & I40E_AQ_SET_FEC_ABILITY_KR)) {
14542                 *flags |= I40E_FLAG_BASE_R_FEC;
14543                 *flags &= ~I40E_FLAG_RS_FEC;
14544         }
14545         if (fec_cfg == 0)
14546                 *flags &= ~(I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC);
14547 }
14548
14549 /**
14550  * i40e_check_recovery_mode - check if we are running transition firmware
14551  * @pf: board private structure
14552  *
14553  * Check registers indicating the firmware runs in recovery mode. Sets the
14554  * appropriate driver state.
14555  *
14556  * Returns true if the recovery mode was detected, false otherwise
14557  **/
14558 static bool i40e_check_recovery_mode(struct i40e_pf *pf)
14559 {
14560         u32 val = rd32(&pf->hw, I40E_GL_FWSTS) & I40E_GL_FWSTS_FWS1B_MASK;
14561         bool is_recovery_mode = false;
14562
14563         if (pf->hw.mac.type == I40E_MAC_XL710)
14564                 is_recovery_mode =
14565                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14566                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK ||
14567                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_TRANSITION_MASK ||
14568                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_NVM_MASK;
14569         if (pf->hw.mac.type == I40E_MAC_X722)
14570                 is_recovery_mode =
14571                 val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14572                 val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK;
14573         if (is_recovery_mode) {
14574                 dev_notice(&pf->pdev->dev, "Firmware recovery mode detected. Limiting functionality.\n");
14575                 dev_notice(&pf->pdev->dev, "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n");
14576                 set_bit(__I40E_RECOVERY_MODE, pf->state);
14577
14578                 return true;
14579         }
14580         if (test_and_clear_bit(__I40E_RECOVERY_MODE, pf->state))
14581                 dev_info(&pf->pdev->dev, "Reinitializing in normal mode with full functionality.\n");
14582
14583         return false;
14584 }
14585
14586 /**
14587  * i40e_pf_loop_reset - perform reset in a loop.
14588  * @pf: board private structure
14589  *
14590  * This function is useful when a NIC is about to enter recovery mode.
14591  * When a NIC's internal data structures are corrupted the NIC's
14592  * firmware is going to enter recovery mode.
14593  * Right after a POR it takes about 7 minutes for firmware to enter
14594  * recovery mode. Until that time a NIC is in some kind of intermediate
14595  * state. After that time period the NIC almost surely enters
14596  * recovery mode. The only way for a driver to detect intermediate
14597  * state is to issue a series of pf-resets and check a return value.
14598  * If a PF reset returns success then the firmware could be in recovery
14599  * mode so the caller of this code needs to check for recovery mode
14600  * if this function returns success. There is a little chance that
14601  * firmware will hang in intermediate state forever.
14602  * Since waiting 7 minutes is quite a lot of time this function waits
14603  * 10 seconds and then gives up by returning an error.
14604  *
14605  * Return 0 on success, negative on failure.
14606  **/
14607 static i40e_status i40e_pf_loop_reset(struct i40e_pf *pf)
14608 {
14609         const unsigned short MAX_CNT = 1000;
14610         const unsigned short MSECS = 10;
14611         struct i40e_hw *hw = &pf->hw;
14612         i40e_status ret;
14613         int cnt;
14614
14615         for (cnt = 0; cnt < MAX_CNT; ++cnt) {
14616                 ret = i40e_pf_reset(hw);
14617                 if (!ret)
14618                         break;
14619                 msleep(MSECS);
14620         }
14621
14622         if (cnt == MAX_CNT) {
14623                 dev_info(&pf->pdev->dev, "PF reset failed: %d\n", ret);
14624                 return ret;
14625         }
14626
14627         pf->pfr_count++;
14628         return ret;
14629 }
14630
14631 /**
14632  * i40e_init_recovery_mode - initialize subsystems needed in recovery mode
14633  * @pf: board private structure
14634  * @hw: ptr to the hardware info
14635  *
14636  * This function does a minimal setup of all subsystems needed for running
14637  * recovery mode.
14638  *
14639  * Returns 0 on success, negative on failure
14640  **/
14641 static int i40e_init_recovery_mode(struct i40e_pf *pf, struct i40e_hw *hw)
14642 {
14643         struct i40e_vsi *vsi;
14644         int err;
14645         int v_idx;
14646
14647         pci_save_state(pf->pdev);
14648
14649         /* set up periodic task facility */
14650         timer_setup(&pf->service_timer, i40e_service_timer, 0);
14651         pf->service_timer_period = HZ;
14652
14653         INIT_WORK(&pf->service_task, i40e_service_task);
14654         clear_bit(__I40E_SERVICE_SCHED, pf->state);
14655
14656         err = i40e_init_interrupt_scheme(pf);
14657         if (err)
14658                 goto err_switch_setup;
14659
14660         /* The number of VSIs reported by the FW is the minimum guaranteed
14661          * to us; HW supports far more and we share the remaining pool with
14662          * the other PFs. We allocate space for more than the guarantee with
14663          * the understanding that we might not get them all later.
14664          */
14665         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
14666                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
14667         else
14668                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
14669
14670         /* Set up the vsi struct and our local tracking of the MAIN PF vsi. */
14671         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
14672                           GFP_KERNEL);
14673         if (!pf->vsi) {
14674                 err = -ENOMEM;
14675                 goto err_switch_setup;
14676         }
14677
14678         /* We allocate one VSI which is needed as absolute minimum
14679          * in order to register the netdev
14680          */
14681         v_idx = i40e_vsi_mem_alloc(pf, I40E_VSI_MAIN);
14682         if (v_idx < 0)
14683                 goto err_switch_setup;
14684         pf->lan_vsi = v_idx;
14685         vsi = pf->vsi[v_idx];
14686         if (!vsi)
14687                 goto err_switch_setup;
14688         vsi->alloc_queue_pairs = 1;
14689         err = i40e_config_netdev(vsi);
14690         if (err)
14691                 goto err_switch_setup;
14692         err = register_netdev(vsi->netdev);
14693         if (err)
14694                 goto err_switch_setup;
14695         vsi->netdev_registered = true;
14696         i40e_dbg_pf_init(pf);
14697
14698         err = i40e_setup_misc_vector_for_recovery_mode(pf);
14699         if (err)
14700                 goto err_switch_setup;
14701
14702         /* tell the firmware that we're starting */
14703         i40e_send_version(pf);
14704
14705         /* since everything's happy, start the service_task timer */
14706         mod_timer(&pf->service_timer,
14707                   round_jiffies(jiffies + pf->service_timer_period));
14708
14709         return 0;
14710
14711 err_switch_setup:
14712         i40e_reset_interrupt_capability(pf);
14713         del_timer_sync(&pf->service_timer);
14714         i40e_shutdown_adminq(hw);
14715         iounmap(hw->hw_addr);
14716         pci_disable_pcie_error_reporting(pf->pdev);
14717         pci_release_mem_regions(pf->pdev);
14718         pci_disable_device(pf->pdev);
14719         kfree(pf);
14720
14721         return err;
14722 }
14723
14724 /**
14725  * i40e_probe - Device initialization routine
14726  * @pdev: PCI device information struct
14727  * @ent: entry in i40e_pci_tbl
14728  *
14729  * i40e_probe initializes a PF identified by a pci_dev structure.
14730  * The OS initialization, configuring of the PF private structure,
14731  * and a hardware reset occur.
14732  *
14733  * Returns 0 on success, negative on failure
14734  **/
14735 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
14736 {
14737         struct i40e_aq_get_phy_abilities_resp abilities;
14738         struct i40e_pf *pf;
14739         struct i40e_hw *hw;
14740         static u16 pfs_found;
14741         u16 wol_nvm_bits;
14742         u16 link_status;
14743         int err;
14744         u32 val;
14745         u32 i;
14746         u8 set_fc_aq_fail;
14747
14748         err = pci_enable_device_mem(pdev);
14749         if (err)
14750                 return err;
14751
14752         /* set up for high or low dma */
14753         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
14754         if (err) {
14755                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
14756                 if (err) {
14757                         dev_err(&pdev->dev,
14758                                 "DMA configuration failed: 0x%x\n", err);
14759                         goto err_dma;
14760                 }
14761         }
14762
14763         /* set up pci connections */
14764         err = pci_request_mem_regions(pdev, i40e_driver_name);
14765         if (err) {
14766                 dev_info(&pdev->dev,
14767                          "pci_request_selected_regions failed %d\n", err);
14768                 goto err_pci_reg;
14769         }
14770
14771         pci_enable_pcie_error_reporting(pdev);
14772         pci_set_master(pdev);
14773
14774         /* Now that we have a PCI connection, we need to do the
14775          * low level device setup.  This is primarily setting up
14776          * the Admin Queue structures and then querying for the
14777          * device's current profile information.
14778          */
14779         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
14780         if (!pf) {
14781                 err = -ENOMEM;
14782                 goto err_pf_alloc;
14783         }
14784         pf->next_vsi = 0;
14785         pf->pdev = pdev;
14786         set_bit(__I40E_DOWN, pf->state);
14787
14788         hw = &pf->hw;
14789         hw->back = pf;
14790
14791         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
14792                                 I40E_MAX_CSR_SPACE);
14793         /* We believe that the highest register to read is
14794          * I40E_GLGEN_STAT_CLEAR, so we check if the BAR size
14795          * is not less than that before mapping to prevent a
14796          * kernel panic.
14797          */
14798         if (pf->ioremap_len < I40E_GLGEN_STAT_CLEAR) {
14799                 dev_err(&pdev->dev, "Cannot map registers, bar size 0x%X too small, aborting\n",
14800                         pf->ioremap_len);
14801                 err = -ENOMEM;
14802                 goto err_ioremap;
14803         }
14804         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
14805         if (!hw->hw_addr) {
14806                 err = -EIO;
14807                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
14808                          (unsigned int)pci_resource_start(pdev, 0),
14809                          pf->ioremap_len, err);
14810                 goto err_ioremap;
14811         }
14812         hw->vendor_id = pdev->vendor;
14813         hw->device_id = pdev->device;
14814         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
14815         hw->subsystem_vendor_id = pdev->subsystem_vendor;
14816         hw->subsystem_device_id = pdev->subsystem_device;
14817         hw->bus.device = PCI_SLOT(pdev->devfn);
14818         hw->bus.func = PCI_FUNC(pdev->devfn);
14819         hw->bus.bus_id = pdev->bus->number;
14820         pf->instance = pfs_found;
14821
14822         /* Select something other than the 802.1ad ethertype for the
14823          * switch to use internally and drop on ingress.
14824          */
14825         hw->switch_tag = 0xffff;
14826         hw->first_tag = ETH_P_8021AD;
14827         hw->second_tag = ETH_P_8021Q;
14828
14829         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
14830         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
14831         INIT_LIST_HEAD(&pf->ddp_old_prof);
14832
14833         /* set up the locks for the AQ, do this only once in probe
14834          * and destroy them only once in remove
14835          */
14836         mutex_init(&hw->aq.asq_mutex);
14837         mutex_init(&hw->aq.arq_mutex);
14838
14839         pf->msg_enable = netif_msg_init(debug,
14840                                         NETIF_MSG_DRV |
14841                                         NETIF_MSG_PROBE |
14842                                         NETIF_MSG_LINK);
14843         if (debug < -1)
14844                 pf->hw.debug_mask = debug;
14845
14846         /* do a special CORER for clearing PXE mode once at init */
14847         if (hw->revision_id == 0 &&
14848             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
14849                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
14850                 i40e_flush(hw);
14851                 msleep(200);
14852                 pf->corer_count++;
14853
14854                 i40e_clear_pxe_mode(hw);
14855         }
14856
14857         /* Reset here to make sure all is clean and to define PF 'n' */
14858         i40e_clear_hw(hw);
14859
14860         err = i40e_set_mac_type(hw);
14861         if (err) {
14862                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14863                          err);
14864                 goto err_pf_reset;
14865         }
14866
14867         err = i40e_pf_loop_reset(pf);
14868         if (err) {
14869                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
14870                 goto err_pf_reset;
14871         }
14872
14873         i40e_check_recovery_mode(pf);
14874
14875         hw->aq.num_arq_entries = I40E_AQ_LEN;
14876         hw->aq.num_asq_entries = I40E_AQ_LEN;
14877         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14878         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14879         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
14880
14881         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
14882                  "%s-%s:misc",
14883                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
14884
14885         err = i40e_init_shared_code(hw);
14886         if (err) {
14887                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14888                          err);
14889                 goto err_pf_reset;
14890         }
14891
14892         /* set up a default setting for link flow control */
14893         pf->hw.fc.requested_mode = I40E_FC_NONE;
14894
14895         err = i40e_init_adminq(hw);
14896         if (err) {
14897                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
14898                         dev_info(&pdev->dev,
14899                                  "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",
14900                                  hw->aq.api_maj_ver,
14901                                  hw->aq.api_min_ver,
14902                                  I40E_FW_API_VERSION_MAJOR,
14903                                  I40E_FW_MINOR_VERSION(hw));
14904                 else
14905                         dev_info(&pdev->dev,
14906                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
14907
14908                 goto err_pf_reset;
14909         }
14910         i40e_get_oem_version(hw);
14911
14912         /* provide nvm, fw, api versions, vendor:device id, subsys vendor:device id */
14913         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s [%04x:%04x] [%04x:%04x]\n",
14914                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
14915                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
14916                  i40e_nvm_version_str(hw), hw->vendor_id, hw->device_id,
14917                  hw->subsystem_vendor_id, hw->subsystem_device_id);
14918
14919         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
14920             hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
14921                 dev_info(&pdev->dev,
14922                          "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",
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         else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
14928                 dev_info(&pdev->dev,
14929                          "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",
14930                          hw->aq.api_maj_ver,
14931                          hw->aq.api_min_ver,
14932                          I40E_FW_API_VERSION_MAJOR,
14933                          I40E_FW_MINOR_VERSION(hw));
14934
14935         i40e_verify_eeprom(pf);
14936
14937         /* Rev 0 hardware was never productized */
14938         if (hw->revision_id < 1)
14939                 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");
14940
14941         i40e_clear_pxe_mode(hw);
14942
14943         err = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
14944         if (err)
14945                 goto err_adminq_setup;
14946
14947         err = i40e_sw_init(pf);
14948         if (err) {
14949                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
14950                 goto err_sw_init;
14951         }
14952
14953         if (test_bit(__I40E_RECOVERY_MODE, pf->state))
14954                 return i40e_init_recovery_mode(pf, hw);
14955
14956         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
14957                                 hw->func_caps.num_rx_qp, 0, 0);
14958         if (err) {
14959                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
14960                 goto err_init_lan_hmc;
14961         }
14962
14963         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
14964         if (err) {
14965                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
14966                 err = -ENOENT;
14967                 goto err_configure_lan_hmc;
14968         }
14969
14970         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
14971          * Ignore error return codes because if it was already disabled via
14972          * hardware settings this will fail
14973          */
14974         if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
14975                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
14976                 i40e_aq_stop_lldp(hw, true, false, NULL);
14977         }
14978
14979         /* allow a platform config to override the HW addr */
14980         i40e_get_platform_mac_addr(pdev, pf);
14981
14982         if (!is_valid_ether_addr(hw->mac.addr)) {
14983                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
14984                 err = -EIO;
14985                 goto err_mac_addr;
14986         }
14987         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
14988         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
14989         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
14990         if (is_valid_ether_addr(hw->mac.port_addr))
14991                 pf->hw_features |= I40E_HW_PORT_ID_VALID;
14992
14993         pci_set_drvdata(pdev, pf);
14994         pci_save_state(pdev);
14995
14996         dev_info(&pdev->dev,
14997                  (pf->flags & I40E_FLAG_DISABLE_FW_LLDP) ?
14998                         "FW LLDP is disabled\n" :
14999                         "FW LLDP is enabled\n");
15000
15001         /* Enable FW to write default DCB config on link-up */
15002         i40e_aq_set_dcb_parameters(hw, true, NULL);
15003
15004 #ifdef CONFIG_I40E_DCB
15005         err = i40e_init_pf_dcb(pf);
15006         if (err) {
15007                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
15008                 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
15009                 /* Continue without DCB enabled */
15010         }
15011 #endif /* CONFIG_I40E_DCB */
15012
15013         /* set up periodic task facility */
15014         timer_setup(&pf->service_timer, i40e_service_timer, 0);
15015         pf->service_timer_period = HZ;
15016
15017         INIT_WORK(&pf->service_task, i40e_service_task);
15018         clear_bit(__I40E_SERVICE_SCHED, pf->state);
15019
15020         /* NVM bit on means WoL disabled for the port */
15021         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
15022         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
15023                 pf->wol_en = false;
15024         else
15025                 pf->wol_en = true;
15026         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
15027
15028         /* set up the main switch operations */
15029         i40e_determine_queue_usage(pf);
15030         err = i40e_init_interrupt_scheme(pf);
15031         if (err)
15032                 goto err_switch_setup;
15033
15034         /* The number of VSIs reported by the FW is the minimum guaranteed
15035          * to us; HW supports far more and we share the remaining pool with
15036          * the other PFs. We allocate space for more than the guarantee with
15037          * the understanding that we might not get them all later.
15038          */
15039         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
15040                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
15041         else
15042                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
15043
15044         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
15045         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
15046                           GFP_KERNEL);
15047         if (!pf->vsi) {
15048                 err = -ENOMEM;
15049                 goto err_switch_setup;
15050         }
15051
15052 #ifdef CONFIG_PCI_IOV
15053         /* prep for VF support */
15054         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15055             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15056             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15057                 if (pci_num_vf(pdev))
15058                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
15059         }
15060 #endif
15061         err = i40e_setup_pf_switch(pf, false);
15062         if (err) {
15063                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
15064                 goto err_vsis;
15065         }
15066         INIT_LIST_HEAD(&pf->vsi[pf->lan_vsi]->ch_list);
15067
15068         /* Make sure flow control is set according to current settings */
15069         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
15070         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
15071                 dev_dbg(&pf->pdev->dev,
15072                         "Set fc with err %s aq_err %s on get_phy_cap\n",
15073                         i40e_stat_str(hw, err),
15074                         i40e_aq_str(hw, hw->aq.asq_last_status));
15075         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
15076                 dev_dbg(&pf->pdev->dev,
15077                         "Set fc with err %s aq_err %s on set_phy_config\n",
15078                         i40e_stat_str(hw, err),
15079                         i40e_aq_str(hw, hw->aq.asq_last_status));
15080         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
15081                 dev_dbg(&pf->pdev->dev,
15082                         "Set fc with err %s aq_err %s on get_link_info\n",
15083                         i40e_stat_str(hw, err),
15084                         i40e_aq_str(hw, hw->aq.asq_last_status));
15085
15086         /* if FDIR VSI was set up, start it now */
15087         for (i = 0; i < pf->num_alloc_vsi; i++) {
15088                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
15089                         i40e_vsi_open(pf->vsi[i]);
15090                         break;
15091                 }
15092         }
15093
15094         /* The driver only wants link up/down and module qualification
15095          * reports from firmware.  Note the negative logic.
15096          */
15097         err = i40e_aq_set_phy_int_mask(&pf->hw,
15098                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
15099                                          I40E_AQ_EVENT_MEDIA_NA |
15100                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
15101         if (err)
15102                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
15103                          i40e_stat_str(&pf->hw, err),
15104                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15105
15106         /* Reconfigure hardware for allowing smaller MSS in the case
15107          * of TSO, so that we avoid the MDD being fired and causing
15108          * a reset in the case of small MSS+TSO.
15109          */
15110         val = rd32(hw, I40E_REG_MSS);
15111         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
15112                 val &= ~I40E_REG_MSS_MIN_MASK;
15113                 val |= I40E_64BYTE_MSS;
15114                 wr32(hw, I40E_REG_MSS, val);
15115         }
15116
15117         if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
15118                 msleep(75);
15119                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
15120                 if (err)
15121                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
15122                                  i40e_stat_str(&pf->hw, err),
15123                                  i40e_aq_str(&pf->hw,
15124                                              pf->hw.aq.asq_last_status));
15125         }
15126         /* The main driver is (mostly) up and happy. We need to set this state
15127          * before setting up the misc vector or we get a race and the vector
15128          * ends up disabled forever.
15129          */
15130         clear_bit(__I40E_DOWN, pf->state);
15131
15132         /* In case of MSIX we are going to setup the misc vector right here
15133          * to handle admin queue events etc. In case of legacy and MSI
15134          * the misc functionality and queue processing is combined in
15135          * the same vector and that gets setup at open.
15136          */
15137         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
15138                 err = i40e_setup_misc_vector(pf);
15139                 if (err) {
15140                         dev_info(&pdev->dev,
15141                                  "setup of misc vector failed: %d\n", err);
15142                         goto err_vsis;
15143                 }
15144         }
15145
15146 #ifdef CONFIG_PCI_IOV
15147         /* prep for VF support */
15148         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15149             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15150             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15151                 /* disable link interrupts for VFs */
15152                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
15153                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
15154                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
15155                 i40e_flush(hw);
15156
15157                 if (pci_num_vf(pdev)) {
15158                         dev_info(&pdev->dev,
15159                                  "Active VFs found, allocating resources.\n");
15160                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
15161                         if (err)
15162                                 dev_info(&pdev->dev,
15163                                          "Error %d allocating resources for existing VFs\n",
15164                                          err);
15165                 }
15166         }
15167 #endif /* CONFIG_PCI_IOV */
15168
15169         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15170                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
15171                                                       pf->num_iwarp_msix,
15172                                                       I40E_IWARP_IRQ_PILE_ID);
15173                 if (pf->iwarp_base_vector < 0) {
15174                         dev_info(&pdev->dev,
15175                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
15176                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
15177                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
15178                 }
15179         }
15180
15181         i40e_dbg_pf_init(pf);
15182
15183         /* tell the firmware that we're starting */
15184         i40e_send_version(pf);
15185
15186         /* since everything's happy, start the service_task timer */
15187         mod_timer(&pf->service_timer,
15188                   round_jiffies(jiffies + pf->service_timer_period));
15189
15190         /* add this PF to client device list and launch a client service task */
15191         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15192                 err = i40e_lan_add_device(pf);
15193                 if (err)
15194                         dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
15195                                  err);
15196         }
15197
15198 #define PCI_SPEED_SIZE 8
15199 #define PCI_WIDTH_SIZE 8
15200         /* Devices on the IOSF bus do not have this information
15201          * and will report PCI Gen 1 x 1 by default so don't bother
15202          * checking them.
15203          */
15204         if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
15205                 char speed[PCI_SPEED_SIZE] = "Unknown";
15206                 char width[PCI_WIDTH_SIZE] = "Unknown";
15207
15208                 /* Get the negotiated link width and speed from PCI config
15209                  * space
15210                  */
15211                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
15212                                           &link_status);
15213
15214                 i40e_set_pci_config_data(hw, link_status);
15215
15216                 switch (hw->bus.speed) {
15217                 case i40e_bus_speed_8000:
15218                         strlcpy(speed, "8.0", PCI_SPEED_SIZE); break;
15219                 case i40e_bus_speed_5000:
15220                         strlcpy(speed, "5.0", PCI_SPEED_SIZE); break;
15221                 case i40e_bus_speed_2500:
15222                         strlcpy(speed, "2.5", PCI_SPEED_SIZE); break;
15223                 default:
15224                         break;
15225                 }
15226                 switch (hw->bus.width) {
15227                 case i40e_bus_width_pcie_x8:
15228                         strlcpy(width, "8", PCI_WIDTH_SIZE); break;
15229                 case i40e_bus_width_pcie_x4:
15230                         strlcpy(width, "4", PCI_WIDTH_SIZE); break;
15231                 case i40e_bus_width_pcie_x2:
15232                         strlcpy(width, "2", PCI_WIDTH_SIZE); break;
15233                 case i40e_bus_width_pcie_x1:
15234                         strlcpy(width, "1", PCI_WIDTH_SIZE); break;
15235                 default:
15236                         break;
15237                 }
15238
15239                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
15240                          speed, width);
15241
15242                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
15243                     hw->bus.speed < i40e_bus_speed_8000) {
15244                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
15245                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
15246                 }
15247         }
15248
15249         /* get the requested speeds from the fw */
15250         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
15251         if (err)
15252                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
15253                         i40e_stat_str(&pf->hw, err),
15254                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15255         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
15256
15257         /* set the FEC config due to the board capabilities */
15258         i40e_set_fec_in_flags(abilities.fec_cfg_curr_mod_ext_info, &pf->flags);
15259
15260         /* get the supported phy types from the fw */
15261         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
15262         if (err)
15263                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
15264                         i40e_stat_str(&pf->hw, err),
15265                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15266
15267         /* Add a filter to drop all Flow control frames from any VSI from being
15268          * transmitted. By doing so we stop a malicious VF from sending out
15269          * PAUSE or PFC frames and potentially controlling traffic for other
15270          * PF/VF VSIs.
15271          * The FW can still send Flow control frames if enabled.
15272          */
15273         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
15274                                                        pf->main_vsi_seid);
15275
15276         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
15277                 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
15278                 pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
15279         if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
15280                 pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
15281         /* print a string summarizing features */
15282         i40e_print_features(pf);
15283
15284         return 0;
15285
15286         /* Unwind what we've done if something failed in the setup */
15287 err_vsis:
15288         set_bit(__I40E_DOWN, pf->state);
15289         i40e_clear_interrupt_scheme(pf);
15290         kfree(pf->vsi);
15291 err_switch_setup:
15292         i40e_reset_interrupt_capability(pf);
15293         del_timer_sync(&pf->service_timer);
15294 err_mac_addr:
15295 err_configure_lan_hmc:
15296         (void)i40e_shutdown_lan_hmc(hw);
15297 err_init_lan_hmc:
15298         kfree(pf->qp_pile);
15299 err_sw_init:
15300 err_adminq_setup:
15301 err_pf_reset:
15302         iounmap(hw->hw_addr);
15303 err_ioremap:
15304         kfree(pf);
15305 err_pf_alloc:
15306         pci_disable_pcie_error_reporting(pdev);
15307         pci_release_mem_regions(pdev);
15308 err_pci_reg:
15309 err_dma:
15310         pci_disable_device(pdev);
15311         return err;
15312 }
15313
15314 /**
15315  * i40e_remove - Device removal routine
15316  * @pdev: PCI device information struct
15317  *
15318  * i40e_remove is called by the PCI subsystem to alert the driver
15319  * that is should release a PCI device.  This could be caused by a
15320  * Hot-Plug event, or because the driver is going to be removed from
15321  * memory.
15322  **/
15323 static void i40e_remove(struct pci_dev *pdev)
15324 {
15325         struct i40e_pf *pf = pci_get_drvdata(pdev);
15326         struct i40e_hw *hw = &pf->hw;
15327         i40e_status ret_code;
15328         int i;
15329
15330         i40e_dbg_pf_exit(pf);
15331
15332         i40e_ptp_stop(pf);
15333
15334         /* Disable RSS in hw */
15335         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
15336         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
15337
15338         /* no more scheduling of any task */
15339         set_bit(__I40E_SUSPENDED, pf->state);
15340         set_bit(__I40E_DOWN, pf->state);
15341         if (pf->service_timer.function)
15342                 del_timer_sync(&pf->service_timer);
15343         if (pf->service_task.func)
15344                 cancel_work_sync(&pf->service_task);
15345
15346         if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
15347                 struct i40e_vsi *vsi = pf->vsi[0];
15348
15349                 /* We know that we have allocated only one vsi for this PF,
15350                  * it was just for registering netdevice, so the interface
15351                  * could be visible in the 'ifconfig' output
15352                  */
15353                 unregister_netdev(vsi->netdev);
15354                 free_netdev(vsi->netdev);
15355
15356                 goto unmap;
15357         }
15358
15359         /* Client close must be called explicitly here because the timer
15360          * has been stopped.
15361          */
15362         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15363
15364         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
15365                 i40e_free_vfs(pf);
15366                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
15367         }
15368
15369         i40e_fdir_teardown(pf);
15370
15371         /* If there is a switch structure or any orphans, remove them.
15372          * This will leave only the PF's VSI remaining.
15373          */
15374         for (i = 0; i < I40E_MAX_VEB; i++) {
15375                 if (!pf->veb[i])
15376                         continue;
15377
15378                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
15379                     pf->veb[i]->uplink_seid == 0)
15380                         i40e_switch_branch_release(pf->veb[i]);
15381         }
15382
15383         /* Now we can shutdown the PF's VSI, just before we kill
15384          * adminq and hmc.
15385          */
15386         if (pf->vsi[pf->lan_vsi])
15387                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
15388
15389         i40e_cloud_filter_exit(pf);
15390
15391         /* remove attached clients */
15392         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15393                 ret_code = i40e_lan_del_device(pf);
15394                 if (ret_code)
15395                         dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
15396                                  ret_code);
15397         }
15398
15399         /* shutdown and destroy the HMC */
15400         if (hw->hmc.hmc_obj) {
15401                 ret_code = i40e_shutdown_lan_hmc(hw);
15402                 if (ret_code)
15403                         dev_warn(&pdev->dev,
15404                                  "Failed to destroy the HMC resources: %d\n",
15405                                  ret_code);
15406         }
15407
15408 unmap:
15409         /* Free MSI/legacy interrupt 0 when in recovery mode. */
15410         if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15411             !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15412                 free_irq(pf->pdev->irq, pf);
15413
15414         /* shutdown the adminq */
15415         i40e_shutdown_adminq(hw);
15416
15417         /* destroy the locks only once, here */
15418         mutex_destroy(&hw->aq.arq_mutex);
15419         mutex_destroy(&hw->aq.asq_mutex);
15420
15421         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
15422         rtnl_lock();
15423         i40e_clear_interrupt_scheme(pf);
15424         for (i = 0; i < pf->num_alloc_vsi; i++) {
15425                 if (pf->vsi[i]) {
15426                         if (!test_bit(__I40E_RECOVERY_MODE, pf->state))
15427                                 i40e_vsi_clear_rings(pf->vsi[i]);
15428                         i40e_vsi_clear(pf->vsi[i]);
15429                         pf->vsi[i] = NULL;
15430                 }
15431         }
15432         rtnl_unlock();
15433
15434         for (i = 0; i < I40E_MAX_VEB; i++) {
15435                 kfree(pf->veb[i]);
15436                 pf->veb[i] = NULL;
15437         }
15438
15439         kfree(pf->qp_pile);
15440         kfree(pf->vsi);
15441
15442         iounmap(hw->hw_addr);
15443         kfree(pf);
15444         pci_release_mem_regions(pdev);
15445
15446         pci_disable_pcie_error_reporting(pdev);
15447         pci_disable_device(pdev);
15448 }
15449
15450 /**
15451  * i40e_pci_error_detected - warning that something funky happened in PCI land
15452  * @pdev: PCI device information struct
15453  * @error: the type of PCI error
15454  *
15455  * Called to warn that something happened and the error handling steps
15456  * are in progress.  Allows the driver to quiesce things, be ready for
15457  * remediation.
15458  **/
15459 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
15460                                                 enum pci_channel_state error)
15461 {
15462         struct i40e_pf *pf = pci_get_drvdata(pdev);
15463
15464         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
15465
15466         if (!pf) {
15467                 dev_info(&pdev->dev,
15468                          "Cannot recover - error happened during device probe\n");
15469                 return PCI_ERS_RESULT_DISCONNECT;
15470         }
15471
15472         /* shutdown all operations */
15473         if (!test_bit(__I40E_SUSPENDED, pf->state))
15474                 i40e_prep_for_reset(pf, false);
15475
15476         /* Request a slot reset */
15477         return PCI_ERS_RESULT_NEED_RESET;
15478 }
15479
15480 /**
15481  * i40e_pci_error_slot_reset - a PCI slot reset just happened
15482  * @pdev: PCI device information struct
15483  *
15484  * Called to find if the driver can work with the device now that
15485  * the pci slot has been reset.  If a basic connection seems good
15486  * (registers are readable and have sane content) then return a
15487  * happy little PCI_ERS_RESULT_xxx.
15488  **/
15489 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
15490 {
15491         struct i40e_pf *pf = pci_get_drvdata(pdev);
15492         pci_ers_result_t result;
15493         u32 reg;
15494
15495         dev_dbg(&pdev->dev, "%s\n", __func__);
15496         if (pci_enable_device_mem(pdev)) {
15497                 dev_info(&pdev->dev,
15498                          "Cannot re-enable PCI device after reset.\n");
15499                 result = PCI_ERS_RESULT_DISCONNECT;
15500         } else {
15501                 pci_set_master(pdev);
15502                 pci_restore_state(pdev);
15503                 pci_save_state(pdev);
15504                 pci_wake_from_d3(pdev, false);
15505
15506                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
15507                 if (reg == 0)
15508                         result = PCI_ERS_RESULT_RECOVERED;
15509                 else
15510                         result = PCI_ERS_RESULT_DISCONNECT;
15511         }
15512
15513         return result;
15514 }
15515
15516 /**
15517  * i40e_pci_error_reset_prepare - prepare device driver for pci reset
15518  * @pdev: PCI device information struct
15519  */
15520 static void i40e_pci_error_reset_prepare(struct pci_dev *pdev)
15521 {
15522         struct i40e_pf *pf = pci_get_drvdata(pdev);
15523
15524         i40e_prep_for_reset(pf, false);
15525 }
15526
15527 /**
15528  * i40e_pci_error_reset_done - pci reset done, device driver reset can begin
15529  * @pdev: PCI device information struct
15530  */
15531 static void i40e_pci_error_reset_done(struct pci_dev *pdev)
15532 {
15533         struct i40e_pf *pf = pci_get_drvdata(pdev);
15534
15535         i40e_reset_and_rebuild(pf, false, false);
15536 }
15537
15538 /**
15539  * i40e_pci_error_resume - restart operations after PCI error recovery
15540  * @pdev: PCI device information struct
15541  *
15542  * Called to allow the driver to bring things back up after PCI error
15543  * and/or reset recovery has finished.
15544  **/
15545 static void i40e_pci_error_resume(struct pci_dev *pdev)
15546 {
15547         struct i40e_pf *pf = pci_get_drvdata(pdev);
15548
15549         dev_dbg(&pdev->dev, "%s\n", __func__);
15550         if (test_bit(__I40E_SUSPENDED, pf->state))
15551                 return;
15552
15553         i40e_handle_reset_warning(pf, false);
15554 }
15555
15556 /**
15557  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
15558  * using the mac_address_write admin q function
15559  * @pf: pointer to i40e_pf struct
15560  **/
15561 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
15562 {
15563         struct i40e_hw *hw = &pf->hw;
15564         i40e_status ret;
15565         u8 mac_addr[6];
15566         u16 flags = 0;
15567
15568         /* Get current MAC address in case it's an LAA */
15569         if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
15570                 ether_addr_copy(mac_addr,
15571                                 pf->vsi[pf->lan_vsi]->netdev->dev_addr);
15572         } else {
15573                 dev_err(&pf->pdev->dev,
15574                         "Failed to retrieve MAC address; using default\n");
15575                 ether_addr_copy(mac_addr, hw->mac.addr);
15576         }
15577
15578         /* The FW expects the mac address write cmd to first be called with
15579          * one of these flags before calling it again with the multicast
15580          * enable flags.
15581          */
15582         flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
15583
15584         if (hw->func_caps.flex10_enable && hw->partition_id != 1)
15585                 flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
15586
15587         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15588         if (ret) {
15589                 dev_err(&pf->pdev->dev,
15590                         "Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
15591                 return;
15592         }
15593
15594         flags = I40E_AQC_MC_MAG_EN
15595                         | I40E_AQC_WOL_PRESERVE_ON_PFR
15596                         | I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
15597         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15598         if (ret)
15599                 dev_err(&pf->pdev->dev,
15600                         "Failed to enable Multicast Magic Packet wake up\n");
15601 }
15602
15603 /**
15604  * i40e_shutdown - PCI callback for shutting down
15605  * @pdev: PCI device information struct
15606  **/
15607 static void i40e_shutdown(struct pci_dev *pdev)
15608 {
15609         struct i40e_pf *pf = pci_get_drvdata(pdev);
15610         struct i40e_hw *hw = &pf->hw;
15611
15612         set_bit(__I40E_SUSPENDED, pf->state);
15613         set_bit(__I40E_DOWN, pf->state);
15614
15615         del_timer_sync(&pf->service_timer);
15616         cancel_work_sync(&pf->service_task);
15617         i40e_cloud_filter_exit(pf);
15618         i40e_fdir_teardown(pf);
15619
15620         /* Client close must be called explicitly here because the timer
15621          * has been stopped.
15622          */
15623         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15624
15625         if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15626                 i40e_enable_mc_magic_wake(pf);
15627
15628         i40e_prep_for_reset(pf, false);
15629
15630         wr32(hw, I40E_PFPM_APM,
15631              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15632         wr32(hw, I40E_PFPM_WUFC,
15633              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15634
15635         /* Free MSI/legacy interrupt 0 when in recovery mode. */
15636         if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15637             !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15638                 free_irq(pf->pdev->irq, pf);
15639
15640         /* Since we're going to destroy queues during the
15641          * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15642          * whole section
15643          */
15644         rtnl_lock();
15645         i40e_clear_interrupt_scheme(pf);
15646         rtnl_unlock();
15647
15648         if (system_state == SYSTEM_POWER_OFF) {
15649                 pci_wake_from_d3(pdev, pf->wol_en);
15650                 pci_set_power_state(pdev, PCI_D3hot);
15651         }
15652 }
15653
15654 /**
15655  * i40e_suspend - PM callback for moving to D3
15656  * @dev: generic device information structure
15657  **/
15658 static int __maybe_unused i40e_suspend(struct device *dev)
15659 {
15660         struct i40e_pf *pf = dev_get_drvdata(dev);
15661         struct i40e_hw *hw = &pf->hw;
15662
15663         /* If we're already suspended, then there is nothing to do */
15664         if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
15665                 return 0;
15666
15667         set_bit(__I40E_DOWN, pf->state);
15668
15669         /* Ensure service task will not be running */
15670         del_timer_sync(&pf->service_timer);
15671         cancel_work_sync(&pf->service_task);
15672
15673         /* Client close must be called explicitly here because the timer
15674          * has been stopped.
15675          */
15676         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15677
15678         if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15679                 i40e_enable_mc_magic_wake(pf);
15680
15681         /* Since we're going to destroy queues during the
15682          * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15683          * whole section
15684          */
15685         rtnl_lock();
15686
15687         i40e_prep_for_reset(pf, true);
15688
15689         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15690         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15691
15692         /* Clear the interrupt scheme and release our IRQs so that the system
15693          * can safely hibernate even when there are a large number of CPUs.
15694          * Otherwise hibernation might fail when mapping all the vectors back
15695          * to CPU0.
15696          */
15697         i40e_clear_interrupt_scheme(pf);
15698
15699         rtnl_unlock();
15700
15701         return 0;
15702 }
15703
15704 /**
15705  * i40e_resume - PM callback for waking up from D3
15706  * @dev: generic device information structure
15707  **/
15708 static int __maybe_unused i40e_resume(struct device *dev)
15709 {
15710         struct i40e_pf *pf = dev_get_drvdata(dev);
15711         int err;
15712
15713         /* If we're not suspended, then there is nothing to do */
15714         if (!test_bit(__I40E_SUSPENDED, pf->state))
15715                 return 0;
15716
15717         /* We need to hold the RTNL lock prior to restoring interrupt schemes,
15718          * since we're going to be restoring queues
15719          */
15720         rtnl_lock();
15721
15722         /* We cleared the interrupt scheme when we suspended, so we need to
15723          * restore it now to resume device functionality.
15724          */
15725         err = i40e_restore_interrupt_scheme(pf);
15726         if (err) {
15727                 dev_err(dev, "Cannot restore interrupt scheme: %d\n",
15728                         err);
15729         }
15730
15731         clear_bit(__I40E_DOWN, pf->state);
15732         i40e_reset_and_rebuild(pf, false, true);
15733
15734         rtnl_unlock();
15735
15736         /* Clear suspended state last after everything is recovered */
15737         clear_bit(__I40E_SUSPENDED, pf->state);
15738
15739         /* Restart the service task */
15740         mod_timer(&pf->service_timer,
15741                   round_jiffies(jiffies + pf->service_timer_period));
15742
15743         return 0;
15744 }
15745
15746 static const struct pci_error_handlers i40e_err_handler = {
15747         .error_detected = i40e_pci_error_detected,
15748         .slot_reset = i40e_pci_error_slot_reset,
15749         .reset_prepare = i40e_pci_error_reset_prepare,
15750         .reset_done = i40e_pci_error_reset_done,
15751         .resume = i40e_pci_error_resume,
15752 };
15753
15754 static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
15755
15756 static struct pci_driver i40e_driver = {
15757         .name     = i40e_driver_name,
15758         .id_table = i40e_pci_tbl,
15759         .probe    = i40e_probe,
15760         .remove   = i40e_remove,
15761         .driver   = {
15762                 .pm = &i40e_pm_ops,
15763         },
15764         .shutdown = i40e_shutdown,
15765         .err_handler = &i40e_err_handler,
15766         .sriov_configure = i40e_pci_sriov_configure,
15767 };
15768
15769 /**
15770  * i40e_init_module - Driver registration routine
15771  *
15772  * i40e_init_module is the first routine called when the driver is
15773  * loaded. All it does is register with the PCI subsystem.
15774  **/
15775 static int __init i40e_init_module(void)
15776 {
15777         pr_info("%s: %s - version %s\n", i40e_driver_name,
15778                 i40e_driver_string, i40e_driver_version_str);
15779         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
15780
15781         /* There is no need to throttle the number of active tasks because
15782          * each device limits its own task using a state bit for scheduling
15783          * the service task, and the device tasks do not interfere with each
15784          * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
15785          * since we need to be able to guarantee forward progress even under
15786          * memory pressure.
15787          */
15788         i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
15789         if (!i40e_wq) {
15790                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
15791                 return -ENOMEM;
15792         }
15793
15794         i40e_dbg_init();
15795         return pci_register_driver(&i40e_driver);
15796 }
15797 module_init(i40e_init_module);
15798
15799 /**
15800  * i40e_exit_module - Driver exit cleanup routine
15801  *
15802  * i40e_exit_module is called just before the driver is removed
15803  * from memory.
15804  **/
15805 static void __exit i40e_exit_module(void)
15806 {
15807         pci_unregister_driver(&i40e_driver);
15808         destroy_workqueue(i40e_wq);
15809         i40e_dbg_exit();
15810 }
15811 module_exit(i40e_exit_module);