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_drv.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         kfree(ring->rx_bi);
3264         ring->xsk_umem = i40e_xsk_umem(ring);
3265         if (ring->xsk_umem) {
3266                 ret = i40e_alloc_rx_bi_zc(ring);
3267                 if (ret)
3268                         return ret;
3269                 ring->rx_buf_len = xsk_umem_get_rx_frame_size(ring->xsk_umem);
3270                 /* For AF_XDP ZC, we disallow packets to span on
3271                  * multiple buffers, thus letting us skip that
3272                  * handling in the fast-path.
3273                  */
3274                 chain_len = 1;
3275                 ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
3276                                                  MEM_TYPE_XSK_BUFF_POOL,
3277                                                  NULL);
3278                 if (ret)
3279                         return ret;
3280                 dev_info(&vsi->back->pdev->dev,
3281                          "Registered XDP mem model MEM_TYPE_XSK_BUFF_POOL on Rx ring %d\n",
3282                          ring->queue_index);
3283
3284         } else {
3285                 ret = i40e_alloc_rx_bi(ring);
3286                 if (ret)
3287                         return ret;
3288                 ring->rx_buf_len = vsi->rx_buf_len;
3289                 if (ring->vsi->type == I40E_VSI_MAIN) {
3290                         ret = xdp_rxq_info_reg_mem_model(&ring->xdp_rxq,
3291                                                          MEM_TYPE_PAGE_SHARED,
3292                                                          NULL);
3293                         if (ret)
3294                                 return ret;
3295                 }
3296         }
3297
3298         rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3299                                     BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3300
3301         rx_ctx.base = (ring->dma / 128);
3302         rx_ctx.qlen = ring->count;
3303
3304         /* use 32 byte descriptors */
3305         rx_ctx.dsize = 1;
3306
3307         /* descriptor type is always zero
3308          * rx_ctx.dtype = 0;
3309          */
3310         rx_ctx.hsplit_0 = 0;
3311
3312         rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3313         if (hw->revision_id == 0)
3314                 rx_ctx.lrxqthresh = 0;
3315         else
3316                 rx_ctx.lrxqthresh = 1;
3317         rx_ctx.crcstrip = 1;
3318         rx_ctx.l2tsel = 1;
3319         /* this controls whether VLAN is stripped from inner headers */
3320         rx_ctx.showiv = 0;
3321         /* set the prefena field to 1 because the manual says to */
3322         rx_ctx.prefena = 1;
3323
3324         /* clear the context in the HMC */
3325         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3326         if (err) {
3327                 dev_info(&vsi->back->pdev->dev,
3328                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3329                          ring->queue_index, pf_q, err);
3330                 return -ENOMEM;
3331         }
3332
3333         /* set the context in the HMC */
3334         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3335         if (err) {
3336                 dev_info(&vsi->back->pdev->dev,
3337                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3338                          ring->queue_index, pf_q, err);
3339                 return -ENOMEM;
3340         }
3341
3342         /* configure Rx buffer alignment */
3343         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3344                 clear_ring_build_skb_enabled(ring);
3345         else
3346                 set_ring_build_skb_enabled(ring);
3347
3348         /* cache tail for quicker writes, and clear the reg before use */
3349         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3350         writel(0, ring->tail);
3351
3352         if (ring->xsk_umem) {
3353                 xsk_buff_set_rxq_info(ring->xsk_umem, &ring->xdp_rxq);
3354                 ok = i40e_alloc_rx_buffers_zc(ring, I40E_DESC_UNUSED(ring));
3355         } else {
3356                 ok = !i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3357         }
3358         if (!ok) {
3359                 /* Log this in case the user has forgotten to give the kernel
3360                  * any buffers, even later in the application.
3361                  */
3362                 dev_info(&vsi->back->pdev->dev,
3363                          "Failed to allocate some buffers on %sRx ring %d (pf_q %d)\n",
3364                          ring->xsk_umem ? "UMEM enabled " : "",
3365                          ring->queue_index, pf_q);
3366         }
3367
3368         return 0;
3369 }
3370
3371 /**
3372  * i40e_vsi_configure_tx - Configure the VSI for Tx
3373  * @vsi: VSI structure describing this set of rings and resources
3374  *
3375  * Configure the Tx VSI for operation.
3376  **/
3377 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3378 {
3379         int err = 0;
3380         u16 i;
3381
3382         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3383                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3384
3385         if (err || !i40e_enabled_xdp_vsi(vsi))
3386                 return err;
3387
3388         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3389                 err = i40e_configure_tx_ring(vsi->xdp_rings[i]);
3390
3391         return err;
3392 }
3393
3394 /**
3395  * i40e_vsi_configure_rx - Configure the VSI for Rx
3396  * @vsi: the VSI being configured
3397  *
3398  * Configure the Rx VSI for operation.
3399  **/
3400 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3401 {
3402         int err = 0;
3403         u16 i;
3404
3405         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3406                 vsi->max_frame = I40E_MAX_RXBUFFER;
3407                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
3408 #if (PAGE_SIZE < 8192)
3409         } else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3410                    (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3411                 vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3412                 vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3413 #endif
3414         } else {
3415                 vsi->max_frame = I40E_MAX_RXBUFFER;
3416                 vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3417                                                        I40E_RXBUFFER_2048;
3418         }
3419
3420         /* set up individual rings */
3421         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3422                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3423
3424         return err;
3425 }
3426
3427 /**
3428  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3429  * @vsi: ptr to the VSI
3430  **/
3431 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3432 {
3433         struct i40e_ring *tx_ring, *rx_ring;
3434         u16 qoffset, qcount;
3435         int i, n;
3436
3437         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3438                 /* Reset the TC information */
3439                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3440                         rx_ring = vsi->rx_rings[i];
3441                         tx_ring = vsi->tx_rings[i];
3442                         rx_ring->dcb_tc = 0;
3443                         tx_ring->dcb_tc = 0;
3444                 }
3445                 return;
3446         }
3447
3448         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3449                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3450                         continue;
3451
3452                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3453                 qcount = vsi->tc_config.tc_info[n].qcount;
3454                 for (i = qoffset; i < (qoffset + qcount); i++) {
3455                         rx_ring = vsi->rx_rings[i];
3456                         tx_ring = vsi->tx_rings[i];
3457                         rx_ring->dcb_tc = n;
3458                         tx_ring->dcb_tc = n;
3459                 }
3460         }
3461 }
3462
3463 /**
3464  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3465  * @vsi: ptr to the VSI
3466  **/
3467 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3468 {
3469         if (vsi->netdev)
3470                 i40e_set_rx_mode(vsi->netdev);
3471 }
3472
3473 /**
3474  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3475  * @vsi: Pointer to the targeted VSI
3476  *
3477  * This function replays the hlist on the hw where all the SB Flow Director
3478  * filters were saved.
3479  **/
3480 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3481 {
3482         struct i40e_fdir_filter *filter;
3483         struct i40e_pf *pf = vsi->back;
3484         struct hlist_node *node;
3485
3486         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3487                 return;
3488
3489         /* Reset FDir counters as we're replaying all existing filters */
3490         pf->fd_tcp4_filter_cnt = 0;
3491         pf->fd_udp4_filter_cnt = 0;
3492         pf->fd_sctp4_filter_cnt = 0;
3493         pf->fd_ip4_filter_cnt = 0;
3494
3495         hlist_for_each_entry_safe(filter, node,
3496                                   &pf->fdir_filter_list, fdir_node) {
3497                 i40e_add_del_fdir(vsi, filter, true);
3498         }
3499 }
3500
3501 /**
3502  * i40e_vsi_configure - Set up the VSI for action
3503  * @vsi: the VSI being configured
3504  **/
3505 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3506 {
3507         int err;
3508
3509         i40e_set_vsi_rx_mode(vsi);
3510         i40e_restore_vlan(vsi);
3511         i40e_vsi_config_dcb_rings(vsi);
3512         err = i40e_vsi_configure_tx(vsi);
3513         if (!err)
3514                 err = i40e_vsi_configure_rx(vsi);
3515
3516         return err;
3517 }
3518
3519 /**
3520  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3521  * @vsi: the VSI being configured
3522  **/
3523 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3524 {
3525         bool has_xdp = i40e_enabled_xdp_vsi(vsi);
3526         struct i40e_pf *pf = vsi->back;
3527         struct i40e_hw *hw = &pf->hw;
3528         u16 vector;
3529         int i, q;
3530         u32 qp;
3531
3532         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3533          * and PFINT_LNKLSTn registers, e.g.:
3534          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3535          */
3536         qp = vsi->base_queue;
3537         vector = vsi->base_vector;
3538         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3539                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3540
3541                 q_vector->rx.next_update = jiffies + 1;
3542                 q_vector->rx.target_itr =
3543                         ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
3544                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3545                      q_vector->rx.target_itr >> 1);
3546                 q_vector->rx.current_itr = q_vector->rx.target_itr;
3547
3548                 q_vector->tx.next_update = jiffies + 1;
3549                 q_vector->tx.target_itr =
3550                         ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
3551                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3552                      q_vector->tx.target_itr >> 1);
3553                 q_vector->tx.current_itr = q_vector->tx.target_itr;
3554
3555                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3556                      i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3557
3558                 /* Linked list for the queuepairs assigned to this vector */
3559                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3560                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3561                         u32 nextqp = has_xdp ? qp + vsi->alloc_queue_pairs : qp;
3562                         u32 val;
3563
3564                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3565                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3566                               (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3567                               (nextqp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT) |
3568                               (I40E_QUEUE_TYPE_TX <<
3569                                I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3570
3571                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3572
3573                         if (has_xdp) {
3574                                 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3575                                       (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3576                                       (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3577                                       (qp << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3578                                       (I40E_QUEUE_TYPE_TX <<
3579                                        I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3580
3581                                 wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3582                         }
3583
3584                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3585                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3586                               (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3587                               ((qp + 1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT) |
3588                               (I40E_QUEUE_TYPE_RX <<
3589                                I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3590
3591                         /* Terminate the linked list */
3592                         if (q == (q_vector->num_ringpairs - 1))
3593                                 val |= (I40E_QUEUE_END_OF_LIST <<
3594                                         I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3595
3596                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3597                         qp++;
3598                 }
3599         }
3600
3601         i40e_flush(hw);
3602 }
3603
3604 /**
3605  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3606  * @pf: pointer to private device data structure
3607  **/
3608 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3609 {
3610         struct i40e_hw *hw = &pf->hw;
3611         u32 val;
3612
3613         /* clear things first */
3614         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3615         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3616
3617         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3618               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3619               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3620               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3621               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3622               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3623               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3624               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3625
3626         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3627                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3628
3629         if (pf->flags & I40E_FLAG_PTP)
3630                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3631
3632         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3633
3634         /* SW_ITR_IDX = 0, but don't change INTENA */
3635         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3636                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3637
3638         /* OTHER_ITR_IDX = 0 */
3639         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3640 }
3641
3642 /**
3643  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3644  * @vsi: the VSI being configured
3645  **/
3646 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3647 {
3648         u32 nextqp = i40e_enabled_xdp_vsi(vsi) ? vsi->alloc_queue_pairs : 0;
3649         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3650         struct i40e_pf *pf = vsi->back;
3651         struct i40e_hw *hw = &pf->hw;
3652         u32 val;
3653
3654         /* set the ITR configuration */
3655         q_vector->rx.next_update = jiffies + 1;
3656         q_vector->rx.target_itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
3657         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.target_itr >> 1);
3658         q_vector->rx.current_itr = q_vector->rx.target_itr;
3659         q_vector->tx.next_update = jiffies + 1;
3660         q_vector->tx.target_itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
3661         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.target_itr >> 1);
3662         q_vector->tx.current_itr = q_vector->tx.target_itr;
3663
3664         i40e_enable_misc_int_causes(pf);
3665
3666         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3667         wr32(hw, I40E_PFINT_LNKLST0, 0);
3668
3669         /* Associate the queue pair to the vector and enable the queue int */
3670         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                   |
3671               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3672               (nextqp      << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3673               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3674
3675         wr32(hw, I40E_QINT_RQCTL(0), val);
3676
3677         if (i40e_enabled_xdp_vsi(vsi)) {
3678                 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                 |
3679                       (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)|
3680                       (I40E_QUEUE_TYPE_TX
3681                        << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3682
3683                 wr32(hw, I40E_QINT_TQCTL(nextqp), val);
3684         }
3685
3686         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3687               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3688               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3689
3690         wr32(hw, I40E_QINT_TQCTL(0), val);
3691         i40e_flush(hw);
3692 }
3693
3694 /**
3695  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3696  * @pf: board private structure
3697  **/
3698 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3699 {
3700         struct i40e_hw *hw = &pf->hw;
3701
3702         wr32(hw, I40E_PFINT_DYN_CTL0,
3703              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3704         i40e_flush(hw);
3705 }
3706
3707 /**
3708  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3709  * @pf: board private structure
3710  **/
3711 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3712 {
3713         struct i40e_hw *hw = &pf->hw;
3714         u32 val;
3715
3716         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3717               I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3718               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3719
3720         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3721         i40e_flush(hw);
3722 }
3723
3724 /**
3725  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3726  * @irq: interrupt number
3727  * @data: pointer to a q_vector
3728  **/
3729 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3730 {
3731         struct i40e_q_vector *q_vector = data;
3732
3733         if (!q_vector->tx.ring && !q_vector->rx.ring)
3734                 return IRQ_HANDLED;
3735
3736         napi_schedule_irqoff(&q_vector->napi);
3737
3738         return IRQ_HANDLED;
3739 }
3740
3741 /**
3742  * i40e_irq_affinity_notify - Callback for affinity changes
3743  * @notify: context as to what irq was changed
3744  * @mask: the new affinity mask
3745  *
3746  * This is a callback function used by the irq_set_affinity_notifier function
3747  * so that we may register to receive changes to the irq affinity masks.
3748  **/
3749 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3750                                      const cpumask_t *mask)
3751 {
3752         struct i40e_q_vector *q_vector =
3753                 container_of(notify, struct i40e_q_vector, affinity_notify);
3754
3755         cpumask_copy(&q_vector->affinity_mask, mask);
3756 }
3757
3758 /**
3759  * i40e_irq_affinity_release - Callback for affinity notifier release
3760  * @ref: internal core kernel usage
3761  *
3762  * This is a callback function used by the irq_set_affinity_notifier function
3763  * to inform the current notification subscriber that they will no longer
3764  * receive notifications.
3765  **/
3766 static void i40e_irq_affinity_release(struct kref *ref) {}
3767
3768 /**
3769  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3770  * @vsi: the VSI being configured
3771  * @basename: name for the vector
3772  *
3773  * Allocates MSI-X vectors and requests interrupts from the kernel.
3774  **/
3775 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3776 {
3777         int q_vectors = vsi->num_q_vectors;
3778         struct i40e_pf *pf = vsi->back;
3779         int base = vsi->base_vector;
3780         int rx_int_idx = 0;
3781         int tx_int_idx = 0;
3782         int vector, err;
3783         int irq_num;
3784         int cpu;
3785
3786         for (vector = 0; vector < q_vectors; vector++) {
3787                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3788
3789                 irq_num = pf->msix_entries[base + vector].vector;
3790
3791                 if (q_vector->tx.ring && q_vector->rx.ring) {
3792                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3793                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3794                         tx_int_idx++;
3795                 } else if (q_vector->rx.ring) {
3796                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3797                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3798                 } else if (q_vector->tx.ring) {
3799                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3800                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3801                 } else {
3802                         /* skip this unused q_vector */
3803                         continue;
3804                 }
3805                 err = request_irq(irq_num,
3806                                   vsi->irq_handler,
3807                                   0,
3808                                   q_vector->name,
3809                                   q_vector);
3810                 if (err) {
3811                         dev_info(&pf->pdev->dev,
3812                                  "MSIX request_irq failed, error: %d\n", err);
3813                         goto free_queue_irqs;
3814                 }
3815
3816                 /* register for affinity change notifications */
3817                 q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3818                 q_vector->affinity_notify.release = i40e_irq_affinity_release;
3819                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3820                 /* Spread affinity hints out across online CPUs.
3821                  *
3822                  * get_cpu_mask returns a static constant mask with
3823                  * a permanent lifetime so it's ok to pass to
3824                  * irq_set_affinity_hint without making a copy.
3825                  */
3826                 cpu = cpumask_local_spread(q_vector->v_idx, -1);
3827                 irq_set_affinity_hint(irq_num, get_cpu_mask(cpu));
3828         }
3829
3830         vsi->irqs_ready = true;
3831         return 0;
3832
3833 free_queue_irqs:
3834         while (vector) {
3835                 vector--;
3836                 irq_num = pf->msix_entries[base + vector].vector;
3837                 irq_set_affinity_notifier(irq_num, NULL);
3838                 irq_set_affinity_hint(irq_num, NULL);
3839                 free_irq(irq_num, &vsi->q_vectors[vector]);
3840         }
3841         return err;
3842 }
3843
3844 /**
3845  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3846  * @vsi: the VSI being un-configured
3847  **/
3848 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3849 {
3850         struct i40e_pf *pf = vsi->back;
3851         struct i40e_hw *hw = &pf->hw;
3852         int base = vsi->base_vector;
3853         int i;
3854
3855         /* disable interrupt causation from each queue */
3856         for (i = 0; i < vsi->num_queue_pairs; i++) {
3857                 u32 val;
3858
3859                 val = rd32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx));
3860                 val &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3861                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), val);
3862
3863                 val = rd32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx));
3864                 val &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3865                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), val);
3866
3867                 if (!i40e_enabled_xdp_vsi(vsi))
3868                         continue;
3869                 wr32(hw, I40E_QINT_TQCTL(vsi->xdp_rings[i]->reg_idx), 0);
3870         }
3871
3872         /* disable each interrupt */
3873         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3874                 for (i = vsi->base_vector;
3875                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3876                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3877
3878                 i40e_flush(hw);
3879                 for (i = 0; i < vsi->num_q_vectors; i++)
3880                         synchronize_irq(pf->msix_entries[i + base].vector);
3881         } else {
3882                 /* Legacy and MSI mode - this stops all interrupt handling */
3883                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3884                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3885                 i40e_flush(hw);
3886                 synchronize_irq(pf->pdev->irq);
3887         }
3888 }
3889
3890 /**
3891  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3892  * @vsi: the VSI being configured
3893  **/
3894 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3895 {
3896         struct i40e_pf *pf = vsi->back;
3897         int i;
3898
3899         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3900                 for (i = 0; i < vsi->num_q_vectors; i++)
3901                         i40e_irq_dynamic_enable(vsi, i);
3902         } else {
3903                 i40e_irq_dynamic_enable_icr0(pf);
3904         }
3905
3906         i40e_flush(&pf->hw);
3907         return 0;
3908 }
3909
3910 /**
3911  * i40e_free_misc_vector - Free the vector that handles non-queue events
3912  * @pf: board private structure
3913  **/
3914 static void i40e_free_misc_vector(struct i40e_pf *pf)
3915 {
3916         /* Disable ICR 0 */
3917         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3918         i40e_flush(&pf->hw);
3919
3920         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
3921                 synchronize_irq(pf->msix_entries[0].vector);
3922                 free_irq(pf->msix_entries[0].vector, pf);
3923                 clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
3924         }
3925 }
3926
3927 /**
3928  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3929  * @irq: interrupt number
3930  * @data: pointer to a q_vector
3931  *
3932  * This is the handler used for all MSI/Legacy interrupts, and deals
3933  * with both queue and non-queue interrupts.  This is also used in
3934  * MSIX mode to handle the non-queue interrupts.
3935  **/
3936 static irqreturn_t i40e_intr(int irq, void *data)
3937 {
3938         struct i40e_pf *pf = (struct i40e_pf *)data;
3939         struct i40e_hw *hw = &pf->hw;
3940         irqreturn_t ret = IRQ_NONE;
3941         u32 icr0, icr0_remaining;
3942         u32 val, ena_mask;
3943
3944         icr0 = rd32(hw, I40E_PFINT_ICR0);
3945         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3946
3947         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3948         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3949                 goto enable_intr;
3950
3951         /* if interrupt but no bits showing, must be SWINT */
3952         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3953             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3954                 pf->sw_int_count++;
3955
3956         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3957             (icr0 & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3958                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3959                 dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3960                 set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
3961         }
3962
3963         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3964         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3965                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3966                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3967
3968                 /* We do not have a way to disarm Queue causes while leaving
3969                  * interrupt enabled for all other causes, ideally
3970                  * interrupt should be disabled while we are in NAPI but
3971                  * this is not a performance path and napi_schedule()
3972                  * can deal with rescheduling.
3973                  */
3974                 if (!test_bit(__I40E_DOWN, pf->state))
3975                         napi_schedule_irqoff(&q_vector->napi);
3976         }
3977
3978         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3979                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3980                 set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3981                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3982         }
3983
3984         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3985                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3986                 set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3987         }
3988
3989         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3990                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3991                 set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3992         }
3993
3994         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3995                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3996                         set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3997                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3998                 val = rd32(hw, I40E_GLGEN_RSTAT);
3999                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
4000                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
4001                 if (val == I40E_RESET_CORER) {
4002                         pf->corer_count++;
4003                 } else if (val == I40E_RESET_GLOBR) {
4004                         pf->globr_count++;
4005                 } else if (val == I40E_RESET_EMPR) {
4006                         pf->empr_count++;
4007                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
4008                 }
4009         }
4010
4011         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
4012                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
4013                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
4014                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
4015                          rd32(hw, I40E_PFHMC_ERRORINFO),
4016                          rd32(hw, I40E_PFHMC_ERRORDATA));
4017         }
4018
4019         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
4020                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
4021
4022                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
4023                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
4024                         i40e_ptp_tx_hwtstamp(pf);
4025                 }
4026         }
4027
4028         /* If a critical error is pending we have no choice but to reset the
4029          * device.
4030          * Report and mask out any remaining unexpected interrupts.
4031          */
4032         icr0_remaining = icr0 & ena_mask;
4033         if (icr0_remaining) {
4034                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
4035                          icr0_remaining);
4036                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
4037                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
4038                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
4039                         dev_info(&pf->pdev->dev, "device will be reset\n");
4040                         set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
4041                         i40e_service_event_schedule(pf);
4042                 }
4043                 ena_mask &= ~icr0_remaining;
4044         }
4045         ret = IRQ_HANDLED;
4046
4047 enable_intr:
4048         /* re-enable interrupt causes */
4049         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
4050         if (!test_bit(__I40E_DOWN, pf->state) ||
4051             test_bit(__I40E_RECOVERY_MODE, pf->state)) {
4052                 i40e_service_event_schedule(pf);
4053                 i40e_irq_dynamic_enable_icr0(pf);
4054         }
4055
4056         return ret;
4057 }
4058
4059 /**
4060  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
4061  * @tx_ring:  tx ring to clean
4062  * @budget:   how many cleans we're allowed
4063  *
4064  * Returns true if there's any budget left (e.g. the clean is finished)
4065  **/
4066 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
4067 {
4068         struct i40e_vsi *vsi = tx_ring->vsi;
4069         u16 i = tx_ring->next_to_clean;
4070         struct i40e_tx_buffer *tx_buf;
4071         struct i40e_tx_desc *tx_desc;
4072
4073         tx_buf = &tx_ring->tx_bi[i];
4074         tx_desc = I40E_TX_DESC(tx_ring, i);
4075         i -= tx_ring->count;
4076
4077         do {
4078                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
4079
4080                 /* if next_to_watch is not set then there is no work pending */
4081                 if (!eop_desc)
4082                         break;
4083
4084                 /* prevent any other reads prior to eop_desc */
4085                 smp_rmb();
4086
4087                 /* if the descriptor isn't done, no work yet to do */
4088                 if (!(eop_desc->cmd_type_offset_bsz &
4089                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
4090                         break;
4091
4092                 /* clear next_to_watch to prevent false hangs */
4093                 tx_buf->next_to_watch = NULL;
4094
4095                 tx_desc->buffer_addr = 0;
4096                 tx_desc->cmd_type_offset_bsz = 0;
4097                 /* move past filter desc */
4098                 tx_buf++;
4099                 tx_desc++;
4100                 i++;
4101                 if (unlikely(!i)) {
4102                         i -= tx_ring->count;
4103                         tx_buf = tx_ring->tx_bi;
4104                         tx_desc = I40E_TX_DESC(tx_ring, 0);
4105                 }
4106                 /* unmap skb header data */
4107                 dma_unmap_single(tx_ring->dev,
4108                                  dma_unmap_addr(tx_buf, dma),
4109                                  dma_unmap_len(tx_buf, len),
4110                                  DMA_TO_DEVICE);
4111                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
4112                         kfree(tx_buf->raw_buf);
4113
4114                 tx_buf->raw_buf = NULL;
4115                 tx_buf->tx_flags = 0;
4116                 tx_buf->next_to_watch = NULL;
4117                 dma_unmap_len_set(tx_buf, len, 0);
4118                 tx_desc->buffer_addr = 0;
4119                 tx_desc->cmd_type_offset_bsz = 0;
4120
4121                 /* move us past the eop_desc for start of next FD desc */
4122                 tx_buf++;
4123                 tx_desc++;
4124                 i++;
4125                 if (unlikely(!i)) {
4126                         i -= tx_ring->count;
4127                         tx_buf = tx_ring->tx_bi;
4128                         tx_desc = I40E_TX_DESC(tx_ring, 0);
4129                 }
4130
4131                 /* update budget accounting */
4132                 budget--;
4133         } while (likely(budget));
4134
4135         i += tx_ring->count;
4136         tx_ring->next_to_clean = i;
4137
4138         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
4139                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
4140
4141         return budget > 0;
4142 }
4143
4144 /**
4145  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
4146  * @irq: interrupt number
4147  * @data: pointer to a q_vector
4148  **/
4149 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
4150 {
4151         struct i40e_q_vector *q_vector = data;
4152         struct i40e_vsi *vsi;
4153
4154         if (!q_vector->tx.ring)
4155                 return IRQ_HANDLED;
4156
4157         vsi = q_vector->tx.ring->vsi;
4158         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
4159
4160         return IRQ_HANDLED;
4161 }
4162
4163 /**
4164  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
4165  * @vsi: the VSI being configured
4166  * @v_idx: vector index
4167  * @qp_idx: queue pair index
4168  **/
4169 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
4170 {
4171         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4172         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
4173         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
4174
4175         tx_ring->q_vector = q_vector;
4176         tx_ring->next = q_vector->tx.ring;
4177         q_vector->tx.ring = tx_ring;
4178         q_vector->tx.count++;
4179
4180         /* Place XDP Tx ring in the same q_vector ring list as regular Tx */
4181         if (i40e_enabled_xdp_vsi(vsi)) {
4182                 struct i40e_ring *xdp_ring = vsi->xdp_rings[qp_idx];
4183
4184                 xdp_ring->q_vector = q_vector;
4185                 xdp_ring->next = q_vector->tx.ring;
4186                 q_vector->tx.ring = xdp_ring;
4187                 q_vector->tx.count++;
4188         }
4189
4190         rx_ring->q_vector = q_vector;
4191         rx_ring->next = q_vector->rx.ring;
4192         q_vector->rx.ring = rx_ring;
4193         q_vector->rx.count++;
4194 }
4195
4196 /**
4197  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
4198  * @vsi: the VSI being configured
4199  *
4200  * This function maps descriptor rings to the queue-specific vectors
4201  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
4202  * one vector per queue pair, but on a constrained vector budget, we
4203  * group the queue pairs as "efficiently" as possible.
4204  **/
4205 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
4206 {
4207         int qp_remaining = vsi->num_queue_pairs;
4208         int q_vectors = vsi->num_q_vectors;
4209         int num_ringpairs;
4210         int v_start = 0;
4211         int qp_idx = 0;
4212
4213         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
4214          * group them so there are multiple queues per vector.
4215          * It is also important to go through all the vectors available to be
4216          * sure that if we don't use all the vectors, that the remaining vectors
4217          * are cleared. This is especially important when decreasing the
4218          * number of queues in use.
4219          */
4220         for (; v_start < q_vectors; v_start++) {
4221                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
4222
4223                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
4224
4225                 q_vector->num_ringpairs = num_ringpairs;
4226                 q_vector->reg_idx = q_vector->v_idx + vsi->base_vector - 1;
4227
4228                 q_vector->rx.count = 0;
4229                 q_vector->tx.count = 0;
4230                 q_vector->rx.ring = NULL;
4231                 q_vector->tx.ring = NULL;
4232
4233                 while (num_ringpairs--) {
4234                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
4235                         qp_idx++;
4236                         qp_remaining--;
4237                 }
4238         }
4239 }
4240
4241 /**
4242  * i40e_vsi_request_irq - Request IRQ from the OS
4243  * @vsi: the VSI being configured
4244  * @basename: name for the vector
4245  **/
4246 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
4247 {
4248         struct i40e_pf *pf = vsi->back;
4249         int err;
4250
4251         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4252                 err = i40e_vsi_request_irq_msix(vsi, basename);
4253         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
4254                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
4255                                   pf->int_name, pf);
4256         else
4257                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
4258                                   pf->int_name, pf);
4259
4260         if (err)
4261                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
4262
4263         return err;
4264 }
4265
4266 #ifdef CONFIG_NET_POLL_CONTROLLER
4267 /**
4268  * i40e_netpoll - A Polling 'interrupt' handler
4269  * @netdev: network interface device structure
4270  *
4271  * This is used by netconsole to send skbs without having to re-enable
4272  * interrupts.  It's not called while the normal interrupt routine is executing.
4273  **/
4274 static void i40e_netpoll(struct net_device *netdev)
4275 {
4276         struct i40e_netdev_priv *np = netdev_priv(netdev);
4277         struct i40e_vsi *vsi = np->vsi;
4278         struct i40e_pf *pf = vsi->back;
4279         int i;
4280
4281         /* if interface is down do nothing */
4282         if (test_bit(__I40E_VSI_DOWN, vsi->state))
4283                 return;
4284
4285         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4286                 for (i = 0; i < vsi->num_q_vectors; i++)
4287                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
4288         } else {
4289                 i40e_intr(pf->pdev->irq, netdev);
4290         }
4291 }
4292 #endif
4293
4294 #define I40E_QTX_ENA_WAIT_COUNT 50
4295
4296 /**
4297  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
4298  * @pf: the PF being configured
4299  * @pf_q: the PF queue
4300  * @enable: enable or disable state of the queue
4301  *
4302  * This routine will wait for the given Tx queue of the PF to reach the
4303  * enabled or disabled state.
4304  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4305  * multiple retries; else will return 0 in case of success.
4306  **/
4307 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4308 {
4309         int i;
4310         u32 tx_reg;
4311
4312         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4313                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
4314                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4315                         break;
4316
4317                 usleep_range(10, 20);
4318         }
4319         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4320                 return -ETIMEDOUT;
4321
4322         return 0;
4323 }
4324
4325 /**
4326  * i40e_control_tx_q - Start or stop a particular Tx queue
4327  * @pf: the PF structure
4328  * @pf_q: the PF queue to configure
4329  * @enable: start or stop the queue
4330  *
4331  * This function enables or disables a single queue. Note that any delay
4332  * required after the operation is expected to be handled by the caller of
4333  * this function.
4334  **/
4335 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
4336 {
4337         struct i40e_hw *hw = &pf->hw;
4338         u32 tx_reg;
4339         int i;
4340
4341         /* warn the TX unit of coming changes */
4342         i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
4343         if (!enable)
4344                 usleep_range(10, 20);
4345
4346         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4347                 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
4348                 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
4349                     ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
4350                         break;
4351                 usleep_range(1000, 2000);
4352         }
4353
4354         /* Skip if the queue is already in the requested state */
4355         if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
4356                 return;
4357
4358         /* turn on/off the queue */
4359         if (enable) {
4360                 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
4361                 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
4362         } else {
4363                 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
4364         }
4365
4366         wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
4367 }
4368
4369 /**
4370  * i40e_control_wait_tx_q - Start/stop Tx queue and wait for completion
4371  * @seid: VSI SEID
4372  * @pf: the PF structure
4373  * @pf_q: the PF queue to configure
4374  * @is_xdp: true if the queue is used for XDP
4375  * @enable: start or stop the queue
4376  **/
4377 int i40e_control_wait_tx_q(int seid, struct i40e_pf *pf, int pf_q,
4378                            bool is_xdp, bool enable)
4379 {
4380         int ret;
4381
4382         i40e_control_tx_q(pf, pf_q, enable);
4383
4384         /* wait for the change to finish */
4385         ret = i40e_pf_txq_wait(pf, pf_q, enable);
4386         if (ret) {
4387                 dev_info(&pf->pdev->dev,
4388                          "VSI seid %d %sTx ring %d %sable timeout\n",
4389                          seid, (is_xdp ? "XDP " : ""), pf_q,
4390                          (enable ? "en" : "dis"));
4391         }
4392
4393         return ret;
4394 }
4395
4396 /**
4397  * i40e_vsi_control_tx - Start or stop a VSI's rings
4398  * @vsi: the VSI being configured
4399  * @enable: start or stop the rings
4400  **/
4401 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4402 {
4403         struct i40e_pf *pf = vsi->back;
4404         int i, pf_q, ret = 0;
4405
4406         pf_q = vsi->base_queue;
4407         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4408                 ret = i40e_control_wait_tx_q(vsi->seid, pf,
4409                                              pf_q,
4410                                              false /*is xdp*/, enable);
4411                 if (ret)
4412                         break;
4413
4414                 if (!i40e_enabled_xdp_vsi(vsi))
4415                         continue;
4416
4417                 ret = i40e_control_wait_tx_q(vsi->seid, pf,
4418                                              pf_q + vsi->alloc_queue_pairs,
4419                                              true /*is xdp*/, enable);
4420                 if (ret)
4421                         break;
4422         }
4423         return ret;
4424 }
4425
4426 /**
4427  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4428  * @pf: the PF being configured
4429  * @pf_q: the PF queue
4430  * @enable: enable or disable state of the queue
4431  *
4432  * This routine will wait for the given Rx queue of the PF to reach the
4433  * enabled or disabled state.
4434  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4435  * multiple retries; else will return 0 in case of success.
4436  **/
4437 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4438 {
4439         int i;
4440         u32 rx_reg;
4441
4442         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4443                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4444                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4445                         break;
4446
4447                 usleep_range(10, 20);
4448         }
4449         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4450                 return -ETIMEDOUT;
4451
4452         return 0;
4453 }
4454
4455 /**
4456  * i40e_control_rx_q - Start or stop a particular Rx queue
4457  * @pf: the PF structure
4458  * @pf_q: the PF queue to configure
4459  * @enable: start or stop the queue
4460  *
4461  * This function enables or disables a single queue. Note that
4462  * any delay required after the operation is expected to be
4463  * handled by the caller of this function.
4464  **/
4465 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4466 {
4467         struct i40e_hw *hw = &pf->hw;
4468         u32 rx_reg;
4469         int i;
4470
4471         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4472                 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4473                 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4474                     ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4475                         break;
4476                 usleep_range(1000, 2000);
4477         }
4478
4479         /* Skip if the queue is already in the requested state */
4480         if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4481                 return;
4482
4483         /* turn on/off the queue */
4484         if (enable)
4485                 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4486         else
4487                 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4488
4489         wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4490 }
4491
4492 /**
4493  * i40e_control_wait_rx_q
4494  * @pf: the PF structure
4495  * @pf_q: queue being configured
4496  * @enable: start or stop the rings
4497  *
4498  * This function enables or disables a single queue along with waiting
4499  * for the change to finish. The caller of this function should handle
4500  * the delays needed in the case of disabling queues.
4501  **/
4502 int i40e_control_wait_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4503 {
4504         int ret = 0;
4505
4506         i40e_control_rx_q(pf, pf_q, enable);
4507
4508         /* wait for the change to finish */
4509         ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4510         if (ret)
4511                 return ret;
4512
4513         return ret;
4514 }
4515
4516 /**
4517  * i40e_vsi_control_rx - Start or stop a VSI's rings
4518  * @vsi: the VSI being configured
4519  * @enable: start or stop the rings
4520  **/
4521 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4522 {
4523         struct i40e_pf *pf = vsi->back;
4524         int i, pf_q, ret = 0;
4525
4526         pf_q = vsi->base_queue;
4527         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4528                 ret = i40e_control_wait_rx_q(pf, pf_q, enable);
4529                 if (ret) {
4530                         dev_info(&pf->pdev->dev,
4531                                  "VSI seid %d Rx ring %d %sable timeout\n",
4532                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
4533                         break;
4534                 }
4535         }
4536
4537         /* Due to HW errata, on Rx disable only, the register can indicate done
4538          * before it really is. Needs 50ms to be sure
4539          */
4540         if (!enable)
4541                 mdelay(50);
4542
4543         return ret;
4544 }
4545
4546 /**
4547  * i40e_vsi_start_rings - Start a VSI's rings
4548  * @vsi: the VSI being configured
4549  **/
4550 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4551 {
4552         int ret = 0;
4553
4554         /* do rx first for enable and last for disable */
4555         ret = i40e_vsi_control_rx(vsi, true);
4556         if (ret)
4557                 return ret;
4558         ret = i40e_vsi_control_tx(vsi, true);
4559
4560         return ret;
4561 }
4562
4563 /**
4564  * i40e_vsi_stop_rings - Stop a VSI's rings
4565  * @vsi: the VSI being configured
4566  **/
4567 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4568 {
4569         /* When port TX is suspended, don't wait */
4570         if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4571                 return i40e_vsi_stop_rings_no_wait(vsi);
4572
4573         /* do rx first for enable and last for disable
4574          * Ignore return value, we need to shutdown whatever we can
4575          */
4576         i40e_vsi_control_tx(vsi, false);
4577         i40e_vsi_control_rx(vsi, false);
4578 }
4579
4580 /**
4581  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4582  * @vsi: the VSI being shutdown
4583  *
4584  * This function stops all the rings for a VSI but does not delay to verify
4585  * that rings have been disabled. It is expected that the caller is shutting
4586  * down multiple VSIs at once and will delay together for all the VSIs after
4587  * initiating the shutdown. This is particularly useful for shutting down lots
4588  * of VFs together. Otherwise, a large delay can be incurred while configuring
4589  * each VSI in serial.
4590  **/
4591 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4592 {
4593         struct i40e_pf *pf = vsi->back;
4594         int i, pf_q;
4595
4596         pf_q = vsi->base_queue;
4597         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4598                 i40e_control_tx_q(pf, pf_q, false);
4599                 i40e_control_rx_q(pf, pf_q, false);
4600         }
4601 }
4602
4603 /**
4604  * i40e_vsi_free_irq - Free the irq association with the OS
4605  * @vsi: the VSI being configured
4606  **/
4607 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4608 {
4609         struct i40e_pf *pf = vsi->back;
4610         struct i40e_hw *hw = &pf->hw;
4611         int base = vsi->base_vector;
4612         u32 val, qp;
4613         int i;
4614
4615         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4616                 if (!vsi->q_vectors)
4617                         return;
4618
4619                 if (!vsi->irqs_ready)
4620                         return;
4621
4622                 vsi->irqs_ready = false;
4623                 for (i = 0; i < vsi->num_q_vectors; i++) {
4624                         int irq_num;
4625                         u16 vector;
4626
4627                         vector = i + base;
4628                         irq_num = pf->msix_entries[vector].vector;
4629
4630                         /* free only the irqs that were actually requested */
4631                         if (!vsi->q_vectors[i] ||
4632                             !vsi->q_vectors[i]->num_ringpairs)
4633                                 continue;
4634
4635                         /* clear the affinity notifier in the IRQ descriptor */
4636                         irq_set_affinity_notifier(irq_num, NULL);
4637                         /* remove our suggested affinity mask for this IRQ */
4638                         irq_set_affinity_hint(irq_num, NULL);
4639                         synchronize_irq(irq_num);
4640                         free_irq(irq_num, vsi->q_vectors[i]);
4641
4642                         /* Tear down the interrupt queue link list
4643                          *
4644                          * We know that they come in pairs and always
4645                          * the Rx first, then the Tx.  To clear the
4646                          * link list, stick the EOL value into the
4647                          * next_q field of the registers.
4648                          */
4649                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4650                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4651                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4652                         val |= I40E_QUEUE_END_OF_LIST
4653                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4654                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4655
4656                         while (qp != I40E_QUEUE_END_OF_LIST) {
4657                                 u32 next;
4658
4659                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4660
4661                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4662                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4663                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4664                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4665
4666                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4667                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4668
4669                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4670
4671                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4672
4673                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4674                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4675
4676                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4677                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4678                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4679                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4680
4681                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4682                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4683
4684                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4685                                 qp = next;
4686                         }
4687                 }
4688         } else {
4689                 free_irq(pf->pdev->irq, pf);
4690
4691                 val = rd32(hw, I40E_PFINT_LNKLST0);
4692                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4693                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4694                 val |= I40E_QUEUE_END_OF_LIST
4695                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4696                 wr32(hw, I40E_PFINT_LNKLST0, val);
4697
4698                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4699                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4700                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4701                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4702                          I40E_QINT_RQCTL_INTEVENT_MASK);
4703
4704                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4705                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4706
4707                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4708
4709                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4710
4711                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4712                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4713                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4714                          I40E_QINT_TQCTL_INTEVENT_MASK);
4715
4716                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4717                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4718
4719                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4720         }
4721 }
4722
4723 /**
4724  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4725  * @vsi: the VSI being configured
4726  * @v_idx: Index of vector to be freed
4727  *
4728  * This function frees the memory allocated to the q_vector.  In addition if
4729  * NAPI is enabled it will delete any references to the NAPI struct prior
4730  * to freeing the q_vector.
4731  **/
4732 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4733 {
4734         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4735         struct i40e_ring *ring;
4736
4737         if (!q_vector)
4738                 return;
4739
4740         /* disassociate q_vector from rings */
4741         i40e_for_each_ring(ring, q_vector->tx)
4742                 ring->q_vector = NULL;
4743
4744         i40e_for_each_ring(ring, q_vector->rx)
4745                 ring->q_vector = NULL;
4746
4747         /* only VSI w/ an associated netdev is set up w/ NAPI */
4748         if (vsi->netdev)
4749                 netif_napi_del(&q_vector->napi);
4750
4751         vsi->q_vectors[v_idx] = NULL;
4752
4753         kfree_rcu(q_vector, rcu);
4754 }
4755
4756 /**
4757  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4758  * @vsi: the VSI being un-configured
4759  *
4760  * This frees the memory allocated to the q_vectors and
4761  * deletes references to the NAPI struct.
4762  **/
4763 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4764 {
4765         int v_idx;
4766
4767         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4768                 i40e_free_q_vector(vsi, v_idx);
4769 }
4770
4771 /**
4772  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4773  * @pf: board private structure
4774  **/
4775 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4776 {
4777         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4778         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4779                 pci_disable_msix(pf->pdev);
4780                 kfree(pf->msix_entries);
4781                 pf->msix_entries = NULL;
4782                 kfree(pf->irq_pile);
4783                 pf->irq_pile = NULL;
4784         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4785                 pci_disable_msi(pf->pdev);
4786         }
4787         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4788 }
4789
4790 /**
4791  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4792  * @pf: board private structure
4793  *
4794  * We go through and clear interrupt specific resources and reset the structure
4795  * to pre-load conditions
4796  **/
4797 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4798 {
4799         int i;
4800
4801         i40e_free_misc_vector(pf);
4802
4803         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4804                       I40E_IWARP_IRQ_PILE_ID);
4805
4806         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4807         for (i = 0; i < pf->num_alloc_vsi; i++)
4808                 if (pf->vsi[i])
4809                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4810         i40e_reset_interrupt_capability(pf);
4811 }
4812
4813 /**
4814  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4815  * @vsi: the VSI being configured
4816  **/
4817 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4818 {
4819         int q_idx;
4820
4821         if (!vsi->netdev)
4822                 return;
4823
4824         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4825                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4826
4827                 if (q_vector->rx.ring || q_vector->tx.ring)
4828                         napi_enable(&q_vector->napi);
4829         }
4830 }
4831
4832 /**
4833  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4834  * @vsi: the VSI being configured
4835  **/
4836 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4837 {
4838         int q_idx;
4839
4840         if (!vsi->netdev)
4841                 return;
4842
4843         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4844                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4845
4846                 if (q_vector->rx.ring || q_vector->tx.ring)
4847                         napi_disable(&q_vector->napi);
4848         }
4849 }
4850
4851 /**
4852  * i40e_vsi_close - Shut down a VSI
4853  * @vsi: the vsi to be quelled
4854  **/
4855 static void i40e_vsi_close(struct i40e_vsi *vsi)
4856 {
4857         struct i40e_pf *pf = vsi->back;
4858         if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4859                 i40e_down(vsi);
4860         i40e_vsi_free_irq(vsi);
4861         i40e_vsi_free_tx_resources(vsi);
4862         i40e_vsi_free_rx_resources(vsi);
4863         vsi->current_netdev_flags = 0;
4864         set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
4865         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4866                 set_bit(__I40E_CLIENT_RESET, pf->state);
4867 }
4868
4869 /**
4870  * i40e_quiesce_vsi - Pause a given VSI
4871  * @vsi: the VSI being paused
4872  **/
4873 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4874 {
4875         if (test_bit(__I40E_VSI_DOWN, vsi->state))
4876                 return;
4877
4878         set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4879         if (vsi->netdev && netif_running(vsi->netdev))
4880                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4881         else
4882                 i40e_vsi_close(vsi);
4883 }
4884
4885 /**
4886  * i40e_unquiesce_vsi - Resume a given VSI
4887  * @vsi: the VSI being resumed
4888  **/
4889 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4890 {
4891         if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4892                 return;
4893
4894         if (vsi->netdev && netif_running(vsi->netdev))
4895                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4896         else
4897                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4898 }
4899
4900 /**
4901  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4902  * @pf: the PF
4903  **/
4904 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4905 {
4906         int v;
4907
4908         for (v = 0; v < pf->num_alloc_vsi; v++) {
4909                 if (pf->vsi[v])
4910                         i40e_quiesce_vsi(pf->vsi[v]);
4911         }
4912 }
4913
4914 /**
4915  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4916  * @pf: the PF
4917  **/
4918 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4919 {
4920         int v;
4921
4922         for (v = 0; v < pf->num_alloc_vsi; v++) {
4923                 if (pf->vsi[v])
4924                         i40e_unquiesce_vsi(pf->vsi[v]);
4925         }
4926 }
4927
4928 /**
4929  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4930  * @vsi: the VSI being configured
4931  *
4932  * Wait until all queues on a given VSI have been disabled.
4933  **/
4934 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4935 {
4936         struct i40e_pf *pf = vsi->back;
4937         int i, pf_q, ret;
4938
4939         pf_q = vsi->base_queue;
4940         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4941                 /* Check and wait for the Tx queue */
4942                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4943                 if (ret) {
4944                         dev_info(&pf->pdev->dev,
4945                                  "VSI seid %d Tx ring %d disable timeout\n",
4946                                  vsi->seid, pf_q);
4947                         return ret;
4948                 }
4949
4950                 if (!i40e_enabled_xdp_vsi(vsi))
4951                         goto wait_rx;
4952
4953                 /* Check and wait for the XDP Tx queue */
4954                 ret = i40e_pf_txq_wait(pf, pf_q + vsi->alloc_queue_pairs,
4955                                        false);
4956                 if (ret) {
4957                         dev_info(&pf->pdev->dev,
4958                                  "VSI seid %d XDP Tx ring %d disable timeout\n",
4959                                  vsi->seid, pf_q);
4960                         return ret;
4961                 }
4962 wait_rx:
4963                 /* Check and wait for the Rx queue */
4964                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4965                 if (ret) {
4966                         dev_info(&pf->pdev->dev,
4967                                  "VSI seid %d Rx ring %d disable timeout\n",
4968                                  vsi->seid, pf_q);
4969                         return ret;
4970                 }
4971         }
4972
4973         return 0;
4974 }
4975
4976 #ifdef CONFIG_I40E_DCB
4977 /**
4978  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4979  * @pf: the PF
4980  *
4981  * This function waits for the queues to be in disabled state for all the
4982  * VSIs that are managed by this PF.
4983  **/
4984 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4985 {
4986         int v, ret = 0;
4987
4988         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4989                 if (pf->vsi[v]) {
4990                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4991                         if (ret)
4992                                 break;
4993                 }
4994         }
4995
4996         return ret;
4997 }
4998
4999 #endif
5000
5001 /**
5002  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
5003  * @pf: pointer to PF
5004  *
5005  * Get TC map for ISCSI PF type that will include iSCSI TC
5006  * and LAN TC.
5007  **/
5008 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
5009 {
5010         struct i40e_dcb_app_priority_table app;
5011         struct i40e_hw *hw = &pf->hw;
5012         u8 enabled_tc = 1; /* TC0 is always enabled */
5013         u8 tc, i;
5014         /* Get the iSCSI APP TLV */
5015         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5016
5017         for (i = 0; i < dcbcfg->numapps; i++) {
5018                 app = dcbcfg->app[i];
5019                 if (app.selector == I40E_APP_SEL_TCPIP &&
5020                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
5021                         tc = dcbcfg->etscfg.prioritytable[app.priority];
5022                         enabled_tc |= BIT(tc);
5023                         break;
5024                 }
5025         }
5026
5027         return enabled_tc;
5028 }
5029
5030 /**
5031  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
5032  * @dcbcfg: the corresponding DCBx configuration structure
5033  *
5034  * Return the number of TCs from given DCBx configuration
5035  **/
5036 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
5037 {
5038         int i, tc_unused = 0;
5039         u8 num_tc = 0;
5040         u8 ret = 0;
5041
5042         /* Scan the ETS Config Priority Table to find
5043          * traffic class enabled for a given priority
5044          * and create a bitmask of enabled TCs
5045          */
5046         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
5047                 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
5048
5049         /* Now scan the bitmask to check for
5050          * contiguous TCs starting with TC0
5051          */
5052         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5053                 if (num_tc & BIT(i)) {
5054                         if (!tc_unused) {
5055                                 ret++;
5056                         } else {
5057                                 pr_err("Non-contiguous TC - Disabling DCB\n");
5058                                 return 1;
5059                         }
5060                 } else {
5061                         tc_unused = 1;
5062                 }
5063         }
5064
5065         /* There is always at least TC0 */
5066         if (!ret)
5067                 ret = 1;
5068
5069         return ret;
5070 }
5071
5072 /**
5073  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
5074  * @dcbcfg: the corresponding DCBx configuration structure
5075  *
5076  * Query the current DCB configuration and return the number of
5077  * traffic classes enabled from the given DCBX config
5078  **/
5079 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
5080 {
5081         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
5082         u8 enabled_tc = 1;
5083         u8 i;
5084
5085         for (i = 0; i < num_tc; i++)
5086                 enabled_tc |= BIT(i);
5087
5088         return enabled_tc;
5089 }
5090
5091 /**
5092  * i40e_mqprio_get_enabled_tc - Get enabled traffic classes
5093  * @pf: PF being queried
5094  *
5095  * Query the current MQPRIO configuration and return the number of
5096  * traffic classes enabled.
5097  **/
5098 static u8 i40e_mqprio_get_enabled_tc(struct i40e_pf *pf)
5099 {
5100         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5101         u8 num_tc = vsi->mqprio_qopt.qopt.num_tc;
5102         u8 enabled_tc = 1, i;
5103
5104         for (i = 1; i < num_tc; i++)
5105                 enabled_tc |= BIT(i);
5106         return enabled_tc;
5107 }
5108
5109 /**
5110  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
5111  * @pf: PF being queried
5112  *
5113  * Return number of traffic classes enabled for the given PF
5114  **/
5115 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
5116 {
5117         struct i40e_hw *hw = &pf->hw;
5118         u8 i, enabled_tc = 1;
5119         u8 num_tc = 0;
5120         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5121
5122         if (pf->flags & I40E_FLAG_TC_MQPRIO)
5123                 return pf->vsi[pf->lan_vsi]->mqprio_qopt.qopt.num_tc;
5124
5125         /* If neither MQPRIO nor DCB is enabled, then always use single TC */
5126         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5127                 return 1;
5128
5129         /* SFP mode will be enabled for all TCs on port */
5130         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5131                 return i40e_dcb_get_num_tc(dcbcfg);
5132
5133         /* MFP mode return count of enabled TCs for this PF */
5134         if (pf->hw.func_caps.iscsi)
5135                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
5136         else
5137                 return 1; /* Only TC0 */
5138
5139         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5140                 if (enabled_tc & BIT(i))
5141                         num_tc++;
5142         }
5143         return num_tc;
5144 }
5145
5146 /**
5147  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
5148  * @pf: PF being queried
5149  *
5150  * Return a bitmap for enabled traffic classes for this PF.
5151  **/
5152 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
5153 {
5154         if (pf->flags & I40E_FLAG_TC_MQPRIO)
5155                 return i40e_mqprio_get_enabled_tc(pf);
5156
5157         /* If neither MQPRIO nor DCB is enabled for this PF then just return
5158          * default TC
5159          */
5160         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5161                 return I40E_DEFAULT_TRAFFIC_CLASS;
5162
5163         /* SFP mode we want PF to be enabled for all TCs */
5164         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
5165                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
5166
5167         /* MFP enabled and iSCSI PF type */
5168         if (pf->hw.func_caps.iscsi)
5169                 return i40e_get_iscsi_tc_map(pf);
5170         else
5171                 return I40E_DEFAULT_TRAFFIC_CLASS;
5172 }
5173
5174 /**
5175  * i40e_vsi_get_bw_info - Query VSI BW Information
5176  * @vsi: the VSI being queried
5177  *
5178  * Returns 0 on success, negative value on failure
5179  **/
5180 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
5181 {
5182         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
5183         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5184         struct i40e_pf *pf = vsi->back;
5185         struct i40e_hw *hw = &pf->hw;
5186         i40e_status ret;
5187         u32 tc_bw_max;
5188         int i;
5189
5190         /* Get the VSI level BW configuration */
5191         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
5192         if (ret) {
5193                 dev_info(&pf->pdev->dev,
5194                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
5195                          i40e_stat_str(&pf->hw, ret),
5196                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5197                 return -EINVAL;
5198         }
5199
5200         /* Get the VSI level BW configuration per TC */
5201         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
5202                                                NULL);
5203         if (ret) {
5204                 dev_info(&pf->pdev->dev,
5205                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
5206                          i40e_stat_str(&pf->hw, ret),
5207                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5208                 return -EINVAL;
5209         }
5210
5211         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
5212                 dev_info(&pf->pdev->dev,
5213                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
5214                          bw_config.tc_valid_bits,
5215                          bw_ets_config.tc_valid_bits);
5216                 /* Still continuing */
5217         }
5218
5219         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
5220         vsi->bw_max_quanta = bw_config.max_bw;
5221         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
5222                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
5223         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5224                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
5225                 vsi->bw_ets_limit_credits[i] =
5226                                         le16_to_cpu(bw_ets_config.credits[i]);
5227                 /* 3 bits out of 4 for each TC */
5228                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
5229         }
5230
5231         return 0;
5232 }
5233
5234 /**
5235  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
5236  * @vsi: the VSI being configured
5237  * @enabled_tc: TC bitmap
5238  * @bw_share: BW shared credits per TC
5239  *
5240  * Returns 0 on success, negative value on failure
5241  **/
5242 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
5243                                        u8 *bw_share)
5244 {
5245         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5246         struct i40e_pf *pf = vsi->back;
5247         i40e_status ret;
5248         int i;
5249
5250         /* There is no need to reset BW when mqprio mode is on.  */
5251         if (pf->flags & I40E_FLAG_TC_MQPRIO)
5252                 return 0;
5253         if (!vsi->mqprio_qopt.qopt.hw && !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5254                 ret = i40e_set_bw_limit(vsi, vsi->seid, 0);
5255                 if (ret)
5256                         dev_info(&pf->pdev->dev,
5257                                  "Failed to reset tx rate for vsi->seid %u\n",
5258                                  vsi->seid);
5259                 return ret;
5260         }
5261         bw_data.tc_valid_bits = enabled_tc;
5262         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5263                 bw_data.tc_bw_credits[i] = bw_share[i];
5264
5265         ret = i40e_aq_config_vsi_tc_bw(&pf->hw, vsi->seid, &bw_data, NULL);
5266         if (ret) {
5267                 dev_info(&pf->pdev->dev,
5268                          "AQ command Config VSI BW allocation per TC failed = %d\n",
5269                          pf->hw.aq.asq_last_status);
5270                 return -EINVAL;
5271         }
5272
5273         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5274                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
5275
5276         return 0;
5277 }
5278
5279 /**
5280  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
5281  * @vsi: the VSI being configured
5282  * @enabled_tc: TC map to be enabled
5283  *
5284  **/
5285 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5286 {
5287         struct net_device *netdev = vsi->netdev;
5288         struct i40e_pf *pf = vsi->back;
5289         struct i40e_hw *hw = &pf->hw;
5290         u8 netdev_tc = 0;
5291         int i;
5292         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
5293
5294         if (!netdev)
5295                 return;
5296
5297         if (!enabled_tc) {
5298                 netdev_reset_tc(netdev);
5299                 return;
5300         }
5301
5302         /* Set up actual enabled TCs on the VSI */
5303         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
5304                 return;
5305
5306         /* set per TC queues for the VSI */
5307         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5308                 /* Only set TC queues for enabled tcs
5309                  *
5310                  * e.g. For a VSI that has TC0 and TC3 enabled the
5311                  * enabled_tc bitmap would be 0x00001001; the driver
5312                  * will set the numtc for netdev as 2 that will be
5313                  * referenced by the netdev layer as TC 0 and 1.
5314                  */
5315                 if (vsi->tc_config.enabled_tc & BIT(i))
5316                         netdev_set_tc_queue(netdev,
5317                                         vsi->tc_config.tc_info[i].netdev_tc,
5318                                         vsi->tc_config.tc_info[i].qcount,
5319                                         vsi->tc_config.tc_info[i].qoffset);
5320         }
5321
5322         if (pf->flags & I40E_FLAG_TC_MQPRIO)
5323                 return;
5324
5325         /* Assign UP2TC map for the VSI */
5326         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
5327                 /* Get the actual TC# for the UP */
5328                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
5329                 /* Get the mapped netdev TC# for the UP */
5330                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
5331                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
5332         }
5333 }
5334
5335 /**
5336  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
5337  * @vsi: the VSI being configured
5338  * @ctxt: the ctxt buffer returned from AQ VSI update param command
5339  **/
5340 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
5341                                       struct i40e_vsi_context *ctxt)
5342 {
5343         /* copy just the sections touched not the entire info
5344          * since not all sections are valid as returned by
5345          * update vsi params
5346          */
5347         vsi->info.mapping_flags = ctxt->info.mapping_flags;
5348         memcpy(&vsi->info.queue_mapping,
5349                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
5350         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
5351                sizeof(vsi->info.tc_mapping));
5352 }
5353
5354 /**
5355  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
5356  * @vsi: VSI to be configured
5357  * @enabled_tc: TC bitmap
5358  *
5359  * This configures a particular VSI for TCs that are mapped to the
5360  * given TC bitmap. It uses default bandwidth share for TCs across
5361  * VSIs to configure TC for a particular VSI.
5362  *
5363  * NOTE:
5364  * It is expected that the VSI queues have been quisced before calling
5365  * this function.
5366  **/
5367 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
5368 {
5369         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5370         struct i40e_pf *pf = vsi->back;
5371         struct i40e_hw *hw = &pf->hw;
5372         struct i40e_vsi_context ctxt;
5373         int ret = 0;
5374         int i;
5375
5376         /* Check if enabled_tc is same as existing or new TCs */
5377         if (vsi->tc_config.enabled_tc == enabled_tc &&
5378             vsi->mqprio_qopt.mode != TC_MQPRIO_MODE_CHANNEL)
5379                 return ret;
5380
5381         /* Enable ETS TCs with equal BW Share for now across all VSIs */
5382         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5383                 if (enabled_tc & BIT(i))
5384                         bw_share[i] = 1;
5385         }
5386
5387         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5388         if (ret) {
5389                 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
5390
5391                 dev_info(&pf->pdev->dev,
5392                          "Failed configuring TC map %d for VSI %d\n",
5393                          enabled_tc, vsi->seid);
5394                 ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid,
5395                                                   &bw_config, NULL);
5396                 if (ret) {
5397                         dev_info(&pf->pdev->dev,
5398                                  "Failed querying vsi bw info, err %s aq_err %s\n",
5399                                  i40e_stat_str(hw, ret),
5400                                  i40e_aq_str(hw, hw->aq.asq_last_status));
5401                         goto out;
5402                 }
5403                 if ((bw_config.tc_valid_bits & enabled_tc) != enabled_tc) {
5404                         u8 valid_tc = bw_config.tc_valid_bits & enabled_tc;
5405
5406                         if (!valid_tc)
5407                                 valid_tc = bw_config.tc_valid_bits;
5408                         /* Always enable TC0, no matter what */
5409                         valid_tc |= 1;
5410                         dev_info(&pf->pdev->dev,
5411                                  "Requested tc 0x%x, but FW reports 0x%x as valid. Attempting to use 0x%x.\n",
5412                                  enabled_tc, bw_config.tc_valid_bits, valid_tc);
5413                         enabled_tc = valid_tc;
5414                 }
5415
5416                 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5417                 if (ret) {
5418                         dev_err(&pf->pdev->dev,
5419                                 "Unable to  configure TC map %d for VSI %d\n",
5420                                 enabled_tc, vsi->seid);
5421                         goto out;
5422                 }
5423         }
5424
5425         /* Update Queue Pairs Mapping for currently enabled UPs */
5426         ctxt.seid = vsi->seid;
5427         ctxt.pf_num = vsi->back->hw.pf_id;
5428         ctxt.vf_num = 0;
5429         ctxt.uplink_seid = vsi->uplink_seid;
5430         ctxt.info = vsi->info;
5431         if (vsi->back->flags & I40E_FLAG_TC_MQPRIO) {
5432                 ret = i40e_vsi_setup_queue_map_mqprio(vsi, &ctxt, enabled_tc);
5433                 if (ret)
5434                         goto out;
5435         } else {
5436                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5437         }
5438
5439         /* On destroying the qdisc, reset vsi->rss_size, as number of enabled
5440          * queues changed.
5441          */
5442         if (!vsi->mqprio_qopt.qopt.hw && vsi->reconfig_rss) {
5443                 vsi->rss_size = min_t(int, vsi->back->alloc_rss_size,
5444                                       vsi->num_queue_pairs);
5445                 ret = i40e_vsi_config_rss(vsi);
5446                 if (ret) {
5447                         dev_info(&vsi->back->pdev->dev,
5448                                  "Failed to reconfig rss for num_queues\n");
5449                         return ret;
5450                 }
5451                 vsi->reconfig_rss = false;
5452         }
5453         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5454                 ctxt.info.valid_sections |=
5455                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5456                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5457         }
5458
5459         /* Update the VSI after updating the VSI queue-mapping
5460          * information
5461          */
5462         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
5463         if (ret) {
5464                 dev_info(&pf->pdev->dev,
5465                          "Update vsi tc config failed, err %s aq_err %s\n",
5466                          i40e_stat_str(hw, ret),
5467                          i40e_aq_str(hw, hw->aq.asq_last_status));
5468                 goto out;
5469         }
5470         /* update the local VSI info with updated queue map */
5471         i40e_vsi_update_queue_map(vsi, &ctxt);
5472         vsi->info.valid_sections = 0;
5473
5474         /* Update current VSI BW information */
5475         ret = i40e_vsi_get_bw_info(vsi);
5476         if (ret) {
5477                 dev_info(&pf->pdev->dev,
5478                          "Failed updating vsi bw info, err %s aq_err %s\n",
5479                          i40e_stat_str(hw, ret),
5480                          i40e_aq_str(hw, hw->aq.asq_last_status));
5481                 goto out;
5482         }
5483
5484         /* Update the netdev TC setup */
5485         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5486 out:
5487         return ret;
5488 }
5489
5490 /**
5491  * i40e_get_link_speed - Returns link speed for the interface
5492  * @vsi: VSI to be configured
5493  *
5494  **/
5495 static int i40e_get_link_speed(struct i40e_vsi *vsi)
5496 {
5497         struct i40e_pf *pf = vsi->back;
5498
5499         switch (pf->hw.phy.link_info.link_speed) {
5500         case I40E_LINK_SPEED_40GB:
5501                 return 40000;
5502         case I40E_LINK_SPEED_25GB:
5503                 return 25000;
5504         case I40E_LINK_SPEED_20GB:
5505                 return 20000;
5506         case I40E_LINK_SPEED_10GB:
5507                 return 10000;
5508         case I40E_LINK_SPEED_1GB:
5509                 return 1000;
5510         default:
5511                 return -EINVAL;
5512         }
5513 }
5514
5515 /**
5516  * i40e_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
5517  * @vsi: VSI to be configured
5518  * @seid: seid of the channel/VSI
5519  * @max_tx_rate: max TX rate to be configured as BW limit
5520  *
5521  * Helper function to set BW limit for a given VSI
5522  **/
5523 int i40e_set_bw_limit(struct i40e_vsi *vsi, u16 seid, u64 max_tx_rate)
5524 {
5525         struct i40e_pf *pf = vsi->back;
5526         u64 credits = 0;
5527         int speed = 0;
5528         int ret = 0;
5529
5530         speed = i40e_get_link_speed(vsi);
5531         if (max_tx_rate > speed) {
5532                 dev_err(&pf->pdev->dev,
5533                         "Invalid max tx rate %llu specified for VSI seid %d.",
5534                         max_tx_rate, seid);
5535                 return -EINVAL;
5536         }
5537         if (max_tx_rate && max_tx_rate < 50) {
5538                 dev_warn(&pf->pdev->dev,
5539                          "Setting max tx rate to minimum usable value of 50Mbps.\n");
5540                 max_tx_rate = 50;
5541         }
5542
5543         /* Tx rate credits are in values of 50Mbps, 0 is disabled */
5544         credits = max_tx_rate;
5545         do_div(credits, I40E_BW_CREDIT_DIVISOR);
5546         ret = i40e_aq_config_vsi_bw_limit(&pf->hw, seid, credits,
5547                                           I40E_MAX_BW_INACTIVE_ACCUM, NULL);
5548         if (ret)
5549                 dev_err(&pf->pdev->dev,
5550                         "Failed set tx rate (%llu Mbps) for vsi->seid %u, err %s aq_err %s\n",
5551                         max_tx_rate, seid, i40e_stat_str(&pf->hw, ret),
5552                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5553         return ret;
5554 }
5555
5556 /**
5557  * i40e_remove_queue_channels - Remove queue channels for the TCs
5558  * @vsi: VSI to be configured
5559  *
5560  * Remove queue channels for the TCs
5561  **/
5562 static void i40e_remove_queue_channels(struct i40e_vsi *vsi)
5563 {
5564         enum i40e_admin_queue_err last_aq_status;
5565         struct i40e_cloud_filter *cfilter;
5566         struct i40e_channel *ch, *ch_tmp;
5567         struct i40e_pf *pf = vsi->back;
5568         struct hlist_node *node;
5569         int ret, i;
5570
5571         /* Reset rss size that was stored when reconfiguring rss for
5572          * channel VSIs with non-power-of-2 queue count.
5573          */
5574         vsi->current_rss_size = 0;
5575
5576         /* perform cleanup for channels if they exist */
5577         if (list_empty(&vsi->ch_list))
5578                 return;
5579
5580         list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5581                 struct i40e_vsi *p_vsi;
5582
5583                 list_del(&ch->list);
5584                 p_vsi = ch->parent_vsi;
5585                 if (!p_vsi || !ch->initialized) {
5586                         kfree(ch);
5587                         continue;
5588                 }
5589                 /* Reset queue contexts */
5590                 for (i = 0; i < ch->num_queue_pairs; i++) {
5591                         struct i40e_ring *tx_ring, *rx_ring;
5592                         u16 pf_q;
5593
5594                         pf_q = ch->base_queue + i;
5595                         tx_ring = vsi->tx_rings[pf_q];
5596                         tx_ring->ch = NULL;
5597
5598                         rx_ring = vsi->rx_rings[pf_q];
5599                         rx_ring->ch = NULL;
5600                 }
5601
5602                 /* Reset BW configured for this VSI via mqprio */
5603                 ret = i40e_set_bw_limit(vsi, ch->seid, 0);
5604                 if (ret)
5605                         dev_info(&vsi->back->pdev->dev,
5606                                  "Failed to reset tx rate for ch->seid %u\n",
5607                                  ch->seid);
5608
5609                 /* delete cloud filters associated with this channel */
5610                 hlist_for_each_entry_safe(cfilter, node,
5611                                           &pf->cloud_filter_list, cloud_node) {
5612                         if (cfilter->seid != ch->seid)
5613                                 continue;
5614
5615                         hash_del(&cfilter->cloud_node);
5616                         if (cfilter->dst_port)
5617                                 ret = i40e_add_del_cloud_filter_big_buf(vsi,
5618                                                                         cfilter,
5619                                                                         false);
5620                         else
5621                                 ret = i40e_add_del_cloud_filter(vsi, cfilter,
5622                                                                 false);
5623                         last_aq_status = pf->hw.aq.asq_last_status;
5624                         if (ret)
5625                                 dev_info(&pf->pdev->dev,
5626                                          "Failed to delete cloud filter, err %s aq_err %s\n",
5627                                          i40e_stat_str(&pf->hw, ret),
5628                                          i40e_aq_str(&pf->hw, last_aq_status));
5629                         kfree(cfilter);
5630                 }
5631
5632                 /* delete VSI from FW */
5633                 ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
5634                                              NULL);
5635                 if (ret)
5636                         dev_err(&vsi->back->pdev->dev,
5637                                 "unable to remove channel (%d) for parent VSI(%d)\n",
5638                                 ch->seid, p_vsi->seid);
5639                 kfree(ch);
5640         }
5641         INIT_LIST_HEAD(&vsi->ch_list);
5642 }
5643
5644 /**
5645  * i40e_is_any_channel - channel exist or not
5646  * @vsi: ptr to VSI to which channels are associated with
5647  *
5648  * Returns true or false if channel(s) exist for associated VSI or not
5649  **/
5650 static bool i40e_is_any_channel(struct i40e_vsi *vsi)
5651 {
5652         struct i40e_channel *ch, *ch_tmp;
5653
5654         list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5655                 if (ch->initialized)
5656                         return true;
5657         }
5658
5659         return false;
5660 }
5661
5662 /**
5663  * i40e_get_max_queues_for_channel
5664  * @vsi: ptr to VSI to which channels are associated with
5665  *
5666  * Helper function which returns max value among the queue counts set on the
5667  * channels/TCs created.
5668  **/
5669 static int i40e_get_max_queues_for_channel(struct i40e_vsi *vsi)
5670 {
5671         struct i40e_channel *ch, *ch_tmp;
5672         int max = 0;
5673
5674         list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
5675                 if (!ch->initialized)
5676                         continue;
5677                 if (ch->num_queue_pairs > max)
5678                         max = ch->num_queue_pairs;
5679         }
5680
5681         return max;
5682 }
5683
5684 /**
5685  * i40e_validate_num_queues - validate num_queues w.r.t channel
5686  * @pf: ptr to PF device
5687  * @num_queues: number of queues
5688  * @vsi: the parent VSI
5689  * @reconfig_rss: indicates should the RSS be reconfigured or not
5690  *
5691  * This function validates number of queues in the context of new channel
5692  * which is being established and determines if RSS should be reconfigured
5693  * or not for parent VSI.
5694  **/
5695 static int i40e_validate_num_queues(struct i40e_pf *pf, int num_queues,
5696                                     struct i40e_vsi *vsi, bool *reconfig_rss)
5697 {
5698         int max_ch_queues;
5699
5700         if (!reconfig_rss)
5701                 return -EINVAL;
5702
5703         *reconfig_rss = false;
5704         if (vsi->current_rss_size) {
5705                 if (num_queues > vsi->current_rss_size) {
5706                         dev_dbg(&pf->pdev->dev,
5707                                 "Error: num_queues (%d) > vsi's current_size(%d)\n",
5708                                 num_queues, vsi->current_rss_size);
5709                         return -EINVAL;
5710                 } else if ((num_queues < vsi->current_rss_size) &&
5711                            (!is_power_of_2(num_queues))) {
5712                         dev_dbg(&pf->pdev->dev,
5713                                 "Error: num_queues (%d) < vsi's current_size(%d), but not power of 2\n",
5714                                 num_queues, vsi->current_rss_size);
5715                         return -EINVAL;
5716                 }
5717         }
5718
5719         if (!is_power_of_2(num_queues)) {
5720                 /* Find the max num_queues configured for channel if channel
5721                  * exist.
5722                  * if channel exist, then enforce 'num_queues' to be more than
5723                  * max ever queues configured for channel.
5724                  */
5725                 max_ch_queues = i40e_get_max_queues_for_channel(vsi);
5726                 if (num_queues < max_ch_queues) {
5727                         dev_dbg(&pf->pdev->dev,
5728                                 "Error: num_queues (%d) < max queues configured for channel(%d)\n",
5729                                 num_queues, max_ch_queues);
5730                         return -EINVAL;
5731                 }
5732                 *reconfig_rss = true;
5733         }
5734
5735         return 0;
5736 }
5737
5738 /**
5739  * i40e_vsi_reconfig_rss - reconfig RSS based on specified rss_size
5740  * @vsi: the VSI being setup
5741  * @rss_size: size of RSS, accordingly LUT gets reprogrammed
5742  *
5743  * This function reconfigures RSS by reprogramming LUTs using 'rss_size'
5744  **/
5745 static int i40e_vsi_reconfig_rss(struct i40e_vsi *vsi, u16 rss_size)
5746 {
5747         struct i40e_pf *pf = vsi->back;
5748         u8 seed[I40E_HKEY_ARRAY_SIZE];
5749         struct i40e_hw *hw = &pf->hw;
5750         int local_rss_size;
5751         u8 *lut;
5752         int ret;
5753
5754         if (!vsi->rss_size)
5755                 return -EINVAL;
5756
5757         if (rss_size > vsi->rss_size)
5758                 return -EINVAL;
5759
5760         local_rss_size = min_t(int, vsi->rss_size, rss_size);
5761         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
5762         if (!lut)
5763                 return -ENOMEM;
5764
5765         /* Ignoring user configured lut if there is one */
5766         i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, local_rss_size);
5767
5768         /* Use user configured hash key if there is one, otherwise
5769          * use default.
5770          */
5771         if (vsi->rss_hkey_user)
5772                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
5773         else
5774                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
5775
5776         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
5777         if (ret) {
5778                 dev_info(&pf->pdev->dev,
5779                          "Cannot set RSS lut, err %s aq_err %s\n",
5780                          i40e_stat_str(hw, ret),
5781                          i40e_aq_str(hw, hw->aq.asq_last_status));
5782                 kfree(lut);
5783                 return ret;
5784         }
5785         kfree(lut);
5786
5787         /* Do the update w.r.t. storing rss_size */
5788         if (!vsi->orig_rss_size)
5789                 vsi->orig_rss_size = vsi->rss_size;
5790         vsi->current_rss_size = local_rss_size;
5791
5792         return ret;
5793 }
5794
5795 /**
5796  * i40e_channel_setup_queue_map - Setup a channel queue map
5797  * @pf: ptr to PF device
5798  * @vsi: the VSI being setup
5799  * @ctxt: VSI context structure
5800  * @ch: ptr to channel structure
5801  *
5802  * Setup queue map for a specific channel
5803  **/
5804 static void i40e_channel_setup_queue_map(struct i40e_pf *pf,
5805                                          struct i40e_vsi_context *ctxt,
5806                                          struct i40e_channel *ch)
5807 {
5808         u16 qcount, qmap, sections = 0;
5809         u8 offset = 0;
5810         int pow;
5811
5812         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
5813         sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
5814
5815         qcount = min_t(int, ch->num_queue_pairs, pf->num_lan_msix);
5816         ch->num_queue_pairs = qcount;
5817
5818         /* find the next higher power-of-2 of num queue pairs */
5819         pow = ilog2(qcount);
5820         if (!is_power_of_2(qcount))
5821                 pow++;
5822
5823         qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
5824                 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
5825
5826         /* Setup queue TC[0].qmap for given VSI context */
5827         ctxt->info.tc_mapping[0] = cpu_to_le16(qmap);
5828
5829         ctxt->info.up_enable_bits = 0x1; /* TC0 enabled */
5830         ctxt->info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
5831         ctxt->info.queue_mapping[0] = cpu_to_le16(ch->base_queue);
5832         ctxt->info.valid_sections |= cpu_to_le16(sections);
5833 }
5834
5835 /**
5836  * i40e_add_channel - add a channel by adding VSI
5837  * @pf: ptr to PF device
5838  * @uplink_seid: underlying HW switching element (VEB) ID
5839  * @ch: ptr to channel structure
5840  *
5841  * Add a channel (VSI) using add_vsi and queue_map
5842  **/
5843 static int i40e_add_channel(struct i40e_pf *pf, u16 uplink_seid,
5844                             struct i40e_channel *ch)
5845 {
5846         struct i40e_hw *hw = &pf->hw;
5847         struct i40e_vsi_context ctxt;
5848         u8 enabled_tc = 0x1; /* TC0 enabled */
5849         int ret;
5850
5851         if (ch->type != I40E_VSI_VMDQ2) {
5852                 dev_info(&pf->pdev->dev,
5853                          "add new vsi failed, ch->type %d\n", ch->type);
5854                 return -EINVAL;
5855         }
5856
5857         memset(&ctxt, 0, sizeof(ctxt));
5858         ctxt.pf_num = hw->pf_id;
5859         ctxt.vf_num = 0;
5860         ctxt.uplink_seid = uplink_seid;
5861         ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
5862         if (ch->type == I40E_VSI_VMDQ2)
5863                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5864
5865         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED) {
5866                 ctxt.info.valid_sections |=
5867                      cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5868                 ctxt.info.switch_id =
5869                    cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5870         }
5871
5872         /* Set queue map for a given VSI context */
5873         i40e_channel_setup_queue_map(pf, &ctxt, ch);
5874
5875         /* Now time to create VSI */
5876         ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5877         if (ret) {
5878                 dev_info(&pf->pdev->dev,
5879                          "add new vsi failed, err %s aq_err %s\n",
5880                          i40e_stat_str(&pf->hw, ret),
5881                          i40e_aq_str(&pf->hw,
5882                                      pf->hw.aq.asq_last_status));
5883                 return -ENOENT;
5884         }
5885
5886         /* Success, update channel, set enabled_tc only if the channel
5887          * is not a macvlan
5888          */
5889         ch->enabled_tc = !i40e_is_channel_macvlan(ch) && enabled_tc;
5890         ch->seid = ctxt.seid;
5891         ch->vsi_number = ctxt.vsi_number;
5892         ch->stat_counter_idx = cpu_to_le16(ctxt.info.stat_counter_idx);
5893
5894         /* copy just the sections touched not the entire info
5895          * since not all sections are valid as returned by
5896          * update vsi params
5897          */
5898         ch->info.mapping_flags = ctxt.info.mapping_flags;
5899         memcpy(&ch->info.queue_mapping,
5900                &ctxt.info.queue_mapping, sizeof(ctxt.info.queue_mapping));
5901         memcpy(&ch->info.tc_mapping, ctxt.info.tc_mapping,
5902                sizeof(ctxt.info.tc_mapping));
5903
5904         return 0;
5905 }
5906
5907 static int i40e_channel_config_bw(struct i40e_vsi *vsi, struct i40e_channel *ch,
5908                                   u8 *bw_share)
5909 {
5910         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
5911         i40e_status ret;
5912         int i;
5913
5914         bw_data.tc_valid_bits = ch->enabled_tc;
5915         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5916                 bw_data.tc_bw_credits[i] = bw_share[i];
5917
5918         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, ch->seid,
5919                                        &bw_data, NULL);
5920         if (ret) {
5921                 dev_info(&vsi->back->pdev->dev,
5922                          "Config VSI BW allocation per TC failed, aq_err: %d for new_vsi->seid %u\n",
5923                          vsi->back->hw.aq.asq_last_status, ch->seid);
5924                 return -EINVAL;
5925         }
5926
5927         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
5928                 ch->info.qs_handle[i] = bw_data.qs_handles[i];
5929
5930         return 0;
5931 }
5932
5933 /**
5934  * i40e_channel_config_tx_ring - config TX ring associated with new channel
5935  * @pf: ptr to PF device
5936  * @vsi: the VSI being setup
5937  * @ch: ptr to channel structure
5938  *
5939  * Configure TX rings associated with channel (VSI) since queues are being
5940  * from parent VSI.
5941  **/
5942 static int i40e_channel_config_tx_ring(struct i40e_pf *pf,
5943                                        struct i40e_vsi *vsi,
5944                                        struct i40e_channel *ch)
5945 {
5946         i40e_status ret;
5947         int i;
5948         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
5949
5950         /* Enable ETS TCs with equal BW Share for now across all VSIs */
5951         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5952                 if (ch->enabled_tc & BIT(i))
5953                         bw_share[i] = 1;
5954         }
5955
5956         /* configure BW for new VSI */
5957         ret = i40e_channel_config_bw(vsi, ch, bw_share);
5958         if (ret) {
5959                 dev_info(&vsi->back->pdev->dev,
5960                          "Failed configuring TC map %d for channel (seid %u)\n",
5961                          ch->enabled_tc, ch->seid);
5962                 return ret;
5963         }
5964
5965         for (i = 0; i < ch->num_queue_pairs; i++) {
5966                 struct i40e_ring *tx_ring, *rx_ring;
5967                 u16 pf_q;
5968
5969                 pf_q = ch->base_queue + i;
5970
5971                 /* Get to TX ring ptr of main VSI, for re-setup TX queue
5972                  * context
5973                  */
5974                 tx_ring = vsi->tx_rings[pf_q];
5975                 tx_ring->ch = ch;
5976
5977                 /* Get the RX ring ptr */
5978                 rx_ring = vsi->rx_rings[pf_q];
5979                 rx_ring->ch = ch;
5980         }
5981
5982         return 0;
5983 }
5984
5985 /**
5986  * i40e_setup_hw_channel - setup new channel
5987  * @pf: ptr to PF device
5988  * @vsi: the VSI being setup
5989  * @ch: ptr to channel structure
5990  * @uplink_seid: underlying HW switching element (VEB) ID
5991  * @type: type of channel to be created (VMDq2/VF)
5992  *
5993  * Setup new channel (VSI) based on specified type (VMDq2/VF)
5994  * and configures TX rings accordingly
5995  **/
5996 static inline int i40e_setup_hw_channel(struct i40e_pf *pf,
5997                                         struct i40e_vsi *vsi,
5998                                         struct i40e_channel *ch,
5999                                         u16 uplink_seid, u8 type)
6000 {
6001         int ret;
6002
6003         ch->initialized = false;
6004         ch->base_queue = vsi->next_base_queue;
6005         ch->type = type;
6006
6007         /* Proceed with creation of channel (VMDq2) VSI */
6008         ret = i40e_add_channel(pf, uplink_seid, ch);
6009         if (ret) {
6010                 dev_info(&pf->pdev->dev,
6011                          "failed to add_channel using uplink_seid %u\n",
6012                          uplink_seid);
6013                 return ret;
6014         }
6015
6016         /* Mark the successful creation of channel */
6017         ch->initialized = true;
6018
6019         /* Reconfigure TX queues using QTX_CTL register */
6020         ret = i40e_channel_config_tx_ring(pf, vsi, ch);
6021         if (ret) {
6022                 dev_info(&pf->pdev->dev,
6023                          "failed to configure TX rings for channel %u\n",
6024                          ch->seid);
6025                 return ret;
6026         }
6027
6028         /* update 'next_base_queue' */
6029         vsi->next_base_queue = vsi->next_base_queue + ch->num_queue_pairs;
6030         dev_dbg(&pf->pdev->dev,
6031                 "Added channel: vsi_seid %u, vsi_number %u, stat_counter_idx %u, num_queue_pairs %u, pf->next_base_queue %d\n",
6032                 ch->seid, ch->vsi_number, ch->stat_counter_idx,
6033                 ch->num_queue_pairs,
6034                 vsi->next_base_queue);
6035         return ret;
6036 }
6037
6038 /**
6039  * i40e_setup_channel - setup new channel using uplink element
6040  * @pf: ptr to PF device
6041  * @type: type of channel to be created (VMDq2/VF)
6042  * @uplink_seid: underlying HW switching element (VEB) ID
6043  * @ch: ptr to channel structure
6044  *
6045  * Setup new channel (VSI) based on specified type (VMDq2/VF)
6046  * and uplink switching element (uplink_seid)
6047  **/
6048 static bool i40e_setup_channel(struct i40e_pf *pf, struct i40e_vsi *vsi,
6049                                struct i40e_channel *ch)
6050 {
6051         u8 vsi_type;
6052         u16 seid;
6053         int ret;
6054
6055         if (vsi->type == I40E_VSI_MAIN) {
6056                 vsi_type = I40E_VSI_VMDQ2;
6057         } else {
6058                 dev_err(&pf->pdev->dev, "unsupported parent vsi type(%d)\n",
6059                         vsi->type);
6060                 return false;
6061         }
6062
6063         /* underlying switching element */
6064         seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6065
6066         /* create channel (VSI), configure TX rings */
6067         ret = i40e_setup_hw_channel(pf, vsi, ch, seid, vsi_type);
6068         if (ret) {
6069                 dev_err(&pf->pdev->dev, "failed to setup hw_channel\n");
6070                 return false;
6071         }
6072
6073         return ch->initialized ? true : false;
6074 }
6075
6076 /**
6077  * i40e_validate_and_set_switch_mode - sets up switch mode correctly
6078  * @vsi: ptr to VSI which has PF backing
6079  *
6080  * Sets up switch mode correctly if it needs to be changed and perform
6081  * what are allowed modes.
6082  **/
6083 static int i40e_validate_and_set_switch_mode(struct i40e_vsi *vsi)
6084 {
6085         u8 mode;
6086         struct i40e_pf *pf = vsi->back;
6087         struct i40e_hw *hw = &pf->hw;
6088         int ret;
6089
6090         ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_dev_capabilities);
6091         if (ret)
6092                 return -EINVAL;
6093
6094         if (hw->dev_caps.switch_mode) {
6095                 /* if switch mode is set, support mode2 (non-tunneled for
6096                  * cloud filter) for now
6097                  */
6098                 u32 switch_mode = hw->dev_caps.switch_mode &
6099                                   I40E_SWITCH_MODE_MASK;
6100                 if (switch_mode >= I40E_CLOUD_FILTER_MODE1) {
6101                         if (switch_mode == I40E_CLOUD_FILTER_MODE2)
6102                                 return 0;
6103                         dev_err(&pf->pdev->dev,
6104                                 "Invalid switch_mode (%d), only non-tunneled mode for cloud filter is supported\n",
6105                                 hw->dev_caps.switch_mode);
6106                         return -EINVAL;
6107                 }
6108         }
6109
6110         /* Set Bit 7 to be valid */
6111         mode = I40E_AQ_SET_SWITCH_BIT7_VALID;
6112
6113         /* Set L4type for TCP support */
6114         mode |= I40E_AQ_SET_SWITCH_L4_TYPE_TCP;
6115
6116         /* Set cloud filter mode */
6117         mode |= I40E_AQ_SET_SWITCH_MODE_NON_TUNNEL;
6118
6119         /* Prep mode field for set_switch_config */
6120         ret = i40e_aq_set_switch_config(hw, pf->last_sw_conf_flags,
6121                                         pf->last_sw_conf_valid_flags,
6122                                         mode, NULL);
6123         if (ret && hw->aq.asq_last_status != I40E_AQ_RC_ESRCH)
6124                 dev_err(&pf->pdev->dev,
6125                         "couldn't set switch config bits, err %s aq_err %s\n",
6126                         i40e_stat_str(hw, ret),
6127                         i40e_aq_str(hw,
6128                                     hw->aq.asq_last_status));
6129
6130         return ret;
6131 }
6132
6133 /**
6134  * i40e_create_queue_channel - function to create channel
6135  * @vsi: VSI to be configured
6136  * @ch: ptr to channel (it contains channel specific params)
6137  *
6138  * This function creates channel (VSI) using num_queues specified by user,
6139  * reconfigs RSS if needed.
6140  **/
6141 int i40e_create_queue_channel(struct i40e_vsi *vsi,
6142                               struct i40e_channel *ch)
6143 {
6144         struct i40e_pf *pf = vsi->back;
6145         bool reconfig_rss;
6146         int err;
6147
6148         if (!ch)
6149                 return -EINVAL;
6150
6151         if (!ch->num_queue_pairs) {
6152                 dev_err(&pf->pdev->dev, "Invalid num_queues requested: %d\n",
6153                         ch->num_queue_pairs);
6154                 return -EINVAL;
6155         }
6156
6157         /* validate user requested num_queues for channel */
6158         err = i40e_validate_num_queues(pf, ch->num_queue_pairs, vsi,
6159                                        &reconfig_rss);
6160         if (err) {
6161                 dev_info(&pf->pdev->dev, "Failed to validate num_queues (%d)\n",
6162                          ch->num_queue_pairs);
6163                 return -EINVAL;
6164         }
6165
6166         /* By default we are in VEPA mode, if this is the first VF/VMDq
6167          * VSI to be added switch to VEB mode.
6168          */
6169         if ((!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) ||
6170             (!i40e_is_any_channel(vsi))) {
6171                 if (!is_power_of_2(vsi->tc_config.tc_info[0].qcount)) {
6172                         dev_dbg(&pf->pdev->dev,
6173                                 "Failed to create channel. Override queues (%u) not power of 2\n",
6174                                 vsi->tc_config.tc_info[0].qcount);
6175                         return -EINVAL;
6176                 }
6177
6178                 if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
6179                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
6180
6181                         if (vsi->type == I40E_VSI_MAIN) {
6182                                 if (pf->flags & I40E_FLAG_TC_MQPRIO)
6183                                         i40e_do_reset(pf, I40E_PF_RESET_FLAG,
6184                                                       true);
6185                                 else
6186                                         i40e_do_reset_safe(pf,
6187                                                            I40E_PF_RESET_FLAG);
6188                         }
6189                 }
6190                 /* now onwards for main VSI, number of queues will be value
6191                  * of TC0's queue count
6192                  */
6193         }
6194
6195         /* By this time, vsi->cnt_q_avail shall be set to non-zero and
6196          * it should be more than num_queues
6197          */
6198         if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_queue_pairs) {
6199                 dev_dbg(&pf->pdev->dev,
6200                         "Error: cnt_q_avail (%u) less than num_queues %d\n",
6201                         vsi->cnt_q_avail, ch->num_queue_pairs);
6202                 return -EINVAL;
6203         }
6204
6205         /* reconfig_rss only if vsi type is MAIN_VSI */
6206         if (reconfig_rss && (vsi->type == I40E_VSI_MAIN)) {
6207                 err = i40e_vsi_reconfig_rss(vsi, ch->num_queue_pairs);
6208                 if (err) {
6209                         dev_info(&pf->pdev->dev,
6210                                  "Error: unable to reconfig rss for num_queues (%u)\n",
6211                                  ch->num_queue_pairs);
6212                         return -EINVAL;
6213                 }
6214         }
6215
6216         if (!i40e_setup_channel(pf, vsi, ch)) {
6217                 dev_info(&pf->pdev->dev, "Failed to setup channel\n");
6218                 return -EINVAL;
6219         }
6220
6221         dev_info(&pf->pdev->dev,
6222                  "Setup channel (id:%u) utilizing num_queues %d\n",
6223                  ch->seid, ch->num_queue_pairs);
6224
6225         /* configure VSI for BW limit */
6226         if (ch->max_tx_rate) {
6227                 u64 credits = ch->max_tx_rate;
6228
6229                 if (i40e_set_bw_limit(vsi, ch->seid, ch->max_tx_rate))
6230                         return -EINVAL;
6231
6232                 do_div(credits, I40E_BW_CREDIT_DIVISOR);
6233                 dev_dbg(&pf->pdev->dev,
6234                         "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
6235                         ch->max_tx_rate,
6236                         credits,
6237                         ch->seid);
6238         }
6239
6240         /* in case of VF, this will be main SRIOV VSI */
6241         ch->parent_vsi = vsi;
6242
6243         /* and update main_vsi's count for queue_available to use */
6244         vsi->cnt_q_avail -= ch->num_queue_pairs;
6245
6246         return 0;
6247 }
6248
6249 /**
6250  * i40e_configure_queue_channels - Add queue channel for the given TCs
6251  * @vsi: VSI to be configured
6252  *
6253  * Configures queue channel mapping to the given TCs
6254  **/
6255 static int i40e_configure_queue_channels(struct i40e_vsi *vsi)
6256 {
6257         struct i40e_channel *ch;
6258         u64 max_rate = 0;
6259         int ret = 0, i;
6260
6261         /* Create app vsi with the TCs. Main VSI with TC0 is already set up */
6262         vsi->tc_seid_map[0] = vsi->seid;
6263         for (i = 1; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6264                 if (vsi->tc_config.enabled_tc & BIT(i)) {
6265                         ch = kzalloc(sizeof(*ch), GFP_KERNEL);
6266                         if (!ch) {
6267                                 ret = -ENOMEM;
6268                                 goto err_free;
6269                         }
6270
6271                         INIT_LIST_HEAD(&ch->list);
6272                         ch->num_queue_pairs =
6273                                 vsi->tc_config.tc_info[i].qcount;
6274                         ch->base_queue =
6275                                 vsi->tc_config.tc_info[i].qoffset;
6276
6277                         /* Bandwidth limit through tc interface is in bytes/s,
6278                          * change to Mbit/s
6279                          */
6280                         max_rate = vsi->mqprio_qopt.max_rate[i];
6281                         do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6282                         ch->max_tx_rate = max_rate;
6283
6284                         list_add_tail(&ch->list, &vsi->ch_list);
6285
6286                         ret = i40e_create_queue_channel(vsi, ch);
6287                         if (ret) {
6288                                 dev_err(&vsi->back->pdev->dev,
6289                                         "Failed creating queue channel with TC%d: queues %d\n",
6290                                         i, ch->num_queue_pairs);
6291                                 goto err_free;
6292                         }
6293                         vsi->tc_seid_map[i] = ch->seid;
6294                 }
6295         }
6296         return ret;
6297
6298 err_free:
6299         i40e_remove_queue_channels(vsi);
6300         return ret;
6301 }
6302
6303 /**
6304  * i40e_veb_config_tc - Configure TCs for given VEB
6305  * @veb: given VEB
6306  * @enabled_tc: TC bitmap
6307  *
6308  * Configures given TC bitmap for VEB (switching) element
6309  **/
6310 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
6311 {
6312         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
6313         struct i40e_pf *pf = veb->pf;
6314         int ret = 0;
6315         int i;
6316
6317         /* No TCs or already enabled TCs just return */
6318         if (!enabled_tc || veb->enabled_tc == enabled_tc)
6319                 return ret;
6320
6321         bw_data.tc_valid_bits = enabled_tc;
6322         /* bw_data.absolute_credits is not set (relative) */
6323
6324         /* Enable ETS TCs with equal BW Share for now */
6325         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6326                 if (enabled_tc & BIT(i))
6327                         bw_data.tc_bw_share_credits[i] = 1;
6328         }
6329
6330         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
6331                                                    &bw_data, NULL);
6332         if (ret) {
6333                 dev_info(&pf->pdev->dev,
6334                          "VEB bw config failed, err %s aq_err %s\n",
6335                          i40e_stat_str(&pf->hw, ret),
6336                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6337                 goto out;
6338         }
6339
6340         /* Update the BW information */
6341         ret = i40e_veb_get_bw_info(veb);
6342         if (ret) {
6343                 dev_info(&pf->pdev->dev,
6344                          "Failed getting veb bw config, err %s aq_err %s\n",
6345                          i40e_stat_str(&pf->hw, ret),
6346                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6347         }
6348
6349 out:
6350         return ret;
6351 }
6352
6353 #ifdef CONFIG_I40E_DCB
6354 /**
6355  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
6356  * @pf: PF struct
6357  *
6358  * Reconfigure VEB/VSIs on a given PF; it is assumed that
6359  * the caller would've quiesce all the VSIs before calling
6360  * this function
6361  **/
6362 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
6363 {
6364         u8 tc_map = 0;
6365         int ret;
6366         u8 v;
6367
6368         /* Enable the TCs available on PF to all VEBs */
6369         tc_map = i40e_pf_get_tc_map(pf);
6370         for (v = 0; v < I40E_MAX_VEB; v++) {
6371                 if (!pf->veb[v])
6372                         continue;
6373                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
6374                 if (ret) {
6375                         dev_info(&pf->pdev->dev,
6376                                  "Failed configuring TC for VEB seid=%d\n",
6377                                  pf->veb[v]->seid);
6378                         /* Will try to configure as many components */
6379                 }
6380         }
6381
6382         /* Update each VSI */
6383         for (v = 0; v < pf->num_alloc_vsi; v++) {
6384                 if (!pf->vsi[v])
6385                         continue;
6386
6387                 /* - Enable all TCs for the LAN VSI
6388                  * - For all others keep them at TC0 for now
6389                  */
6390                 if (v == pf->lan_vsi)
6391                         tc_map = i40e_pf_get_tc_map(pf);
6392                 else
6393                         tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
6394
6395                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
6396                 if (ret) {
6397                         dev_info(&pf->pdev->dev,
6398                                  "Failed configuring TC for VSI seid=%d\n",
6399                                  pf->vsi[v]->seid);
6400                         /* Will try to configure as many components */
6401                 } else {
6402                         /* Re-configure VSI vectors based on updated TC map */
6403                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
6404                         if (pf->vsi[v]->netdev)
6405                                 i40e_dcbnl_set_all(pf->vsi[v]);
6406                 }
6407         }
6408 }
6409
6410 /**
6411  * i40e_resume_port_tx - Resume port Tx
6412  * @pf: PF struct
6413  *
6414  * Resume a port's Tx and issue a PF reset in case of failure to
6415  * resume.
6416  **/
6417 static int i40e_resume_port_tx(struct i40e_pf *pf)
6418 {
6419         struct i40e_hw *hw = &pf->hw;
6420         int ret;
6421
6422         ret = i40e_aq_resume_port_tx(hw, NULL);
6423         if (ret) {
6424                 dev_info(&pf->pdev->dev,
6425                          "Resume Port Tx failed, err %s aq_err %s\n",
6426                           i40e_stat_str(&pf->hw, ret),
6427                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6428                 /* Schedule PF reset to recover */
6429                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6430                 i40e_service_event_schedule(pf);
6431         }
6432
6433         return ret;
6434 }
6435
6436 /**
6437  * i40e_init_pf_dcb - Initialize DCB configuration
6438  * @pf: PF being configured
6439  *
6440  * Query the current DCB configuration and cache it
6441  * in the hardware structure
6442  **/
6443 static int i40e_init_pf_dcb(struct i40e_pf *pf)
6444 {
6445         struct i40e_hw *hw = &pf->hw;
6446         int err = 0;
6447
6448         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable
6449          * Also do not enable DCBx if FW LLDP agent is disabled
6450          */
6451         if ((pf->hw_features & I40E_HW_NO_DCB_SUPPORT) ||
6452             (pf->flags & I40E_FLAG_DISABLE_FW_LLDP)) {
6453                 dev_info(&pf->pdev->dev, "DCB is not supported or FW LLDP is disabled\n");
6454                 err = I40E_NOT_SUPPORTED;
6455                 goto out;
6456         }
6457
6458         err = i40e_init_dcb(hw, true);
6459         if (!err) {
6460                 /* Device/Function is not DCBX capable */
6461                 if ((!hw->func_caps.dcb) ||
6462                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
6463                         dev_info(&pf->pdev->dev,
6464                                  "DCBX offload is not supported or is disabled for this PF.\n");
6465                 } else {
6466                         /* When status is not DISABLED then DCBX in FW */
6467                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
6468                                        DCB_CAP_DCBX_VER_IEEE;
6469
6470                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
6471                         /* Enable DCB tagging only when more than one TC
6472                          * or explicitly disable if only one TC
6473                          */
6474                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
6475                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
6476                         else
6477                                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
6478                         dev_dbg(&pf->pdev->dev,
6479                                 "DCBX offload is supported for this PF.\n");
6480                 }
6481         } else if (pf->hw.aq.asq_last_status == I40E_AQ_RC_EPERM) {
6482                 dev_info(&pf->pdev->dev, "FW LLDP disabled for this PF.\n");
6483                 pf->flags |= I40E_FLAG_DISABLE_FW_LLDP;
6484         } else {
6485                 dev_info(&pf->pdev->dev,
6486                          "Query for DCB configuration failed, err %s aq_err %s\n",
6487                          i40e_stat_str(&pf->hw, err),
6488                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6489         }
6490
6491 out:
6492         return err;
6493 }
6494 #endif /* CONFIG_I40E_DCB */
6495 #define SPEED_SIZE 14
6496 #define FC_SIZE 8
6497 /**
6498  * i40e_print_link_message - print link up or down
6499  * @vsi: the VSI for which link needs a message
6500  * @isup: true of link is up, false otherwise
6501  */
6502 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
6503 {
6504         enum i40e_aq_link_speed new_speed;
6505         struct i40e_pf *pf = vsi->back;
6506         char *speed = "Unknown";
6507         char *fc = "Unknown";
6508         char *fec = "";
6509         char *req_fec = "";
6510         char *an = "";
6511
6512         if (isup)
6513                 new_speed = pf->hw.phy.link_info.link_speed;
6514         else
6515                 new_speed = I40E_LINK_SPEED_UNKNOWN;
6516
6517         if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
6518                 return;
6519         vsi->current_isup = isup;
6520         vsi->current_speed = new_speed;
6521         if (!isup) {
6522                 netdev_info(vsi->netdev, "NIC Link is Down\n");
6523                 return;
6524         }
6525
6526         /* Warn user if link speed on NPAR enabled partition is not at
6527          * least 10GB
6528          */
6529         if (pf->hw.func_caps.npar_enable &&
6530             (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
6531              pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
6532                 netdev_warn(vsi->netdev,
6533                             "The partition detected link speed that is less than 10Gbps\n");
6534
6535         switch (pf->hw.phy.link_info.link_speed) {
6536         case I40E_LINK_SPEED_40GB:
6537                 speed = "40 G";
6538                 break;
6539         case I40E_LINK_SPEED_20GB:
6540                 speed = "20 G";
6541                 break;
6542         case I40E_LINK_SPEED_25GB:
6543                 speed = "25 G";
6544                 break;
6545         case I40E_LINK_SPEED_10GB:
6546                 speed = "10 G";
6547                 break;
6548         case I40E_LINK_SPEED_5GB:
6549                 speed = "5 G";
6550                 break;
6551         case I40E_LINK_SPEED_2_5GB:
6552                 speed = "2.5 G";
6553                 break;
6554         case I40E_LINK_SPEED_1GB:
6555                 speed = "1000 M";
6556                 break;
6557         case I40E_LINK_SPEED_100MB:
6558                 speed = "100 M";
6559                 break;
6560         default:
6561                 break;
6562         }
6563
6564         switch (pf->hw.fc.current_mode) {
6565         case I40E_FC_FULL:
6566                 fc = "RX/TX";
6567                 break;
6568         case I40E_FC_TX_PAUSE:
6569                 fc = "TX";
6570                 break;
6571         case I40E_FC_RX_PAUSE:
6572                 fc = "RX";
6573                 break;
6574         default:
6575                 fc = "None";
6576                 break;
6577         }
6578
6579         if (pf->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
6580                 req_fec = "None";
6581                 fec = "None";
6582                 an = "False";
6583
6584                 if (pf->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
6585                         an = "True";
6586
6587                 if (pf->hw.phy.link_info.fec_info &
6588                     I40E_AQ_CONFIG_FEC_KR_ENA)
6589                         fec = "CL74 FC-FEC/BASE-R";
6590                 else if (pf->hw.phy.link_info.fec_info &
6591                          I40E_AQ_CONFIG_FEC_RS_ENA)
6592                         fec = "CL108 RS-FEC";
6593
6594                 /* 'CL108 RS-FEC' should be displayed when RS is requested, or
6595                  * both RS and FC are requested
6596                  */
6597                 if (vsi->back->hw.phy.link_info.req_fec_info &
6598                     (I40E_AQ_REQUEST_FEC_KR | I40E_AQ_REQUEST_FEC_RS)) {
6599                         if (vsi->back->hw.phy.link_info.req_fec_info &
6600                             I40E_AQ_REQUEST_FEC_RS)
6601                                 req_fec = "CL108 RS-FEC";
6602                         else
6603                                 req_fec = "CL74 FC-FEC/BASE-R";
6604                 }
6605                 netdev_info(vsi->netdev,
6606                             "NIC Link is Up, %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg: %s, Flow Control: %s\n",
6607                             speed, req_fec, fec, an, fc);
6608         } else {
6609                 netdev_info(vsi->netdev,
6610                             "NIC Link is Up, %sbps Full Duplex, Flow Control: %s\n",
6611                             speed, fc);
6612         }
6613
6614 }
6615
6616 /**
6617  * i40e_up_complete - Finish the last steps of bringing up a connection
6618  * @vsi: the VSI being configured
6619  **/
6620 static int i40e_up_complete(struct i40e_vsi *vsi)
6621 {
6622         struct i40e_pf *pf = vsi->back;
6623         int err;
6624
6625         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6626                 i40e_vsi_configure_msix(vsi);
6627         else
6628                 i40e_configure_msi_and_legacy(vsi);
6629
6630         /* start rings */
6631         err = i40e_vsi_start_rings(vsi);
6632         if (err)
6633                 return err;
6634
6635         clear_bit(__I40E_VSI_DOWN, vsi->state);
6636         i40e_napi_enable_all(vsi);
6637         i40e_vsi_enable_irq(vsi);
6638
6639         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
6640             (vsi->netdev)) {
6641                 i40e_print_link_message(vsi, true);
6642                 netif_tx_start_all_queues(vsi->netdev);
6643                 netif_carrier_on(vsi->netdev);
6644         }
6645
6646         /* replay FDIR SB filters */
6647         if (vsi->type == I40E_VSI_FDIR) {
6648                 /* reset fd counters */
6649                 pf->fd_add_err = 0;
6650                 pf->fd_atr_cnt = 0;
6651                 i40e_fdir_filter_restore(vsi);
6652         }
6653
6654         /* On the next run of the service_task, notify any clients of the new
6655          * opened netdev
6656          */
6657         set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
6658         i40e_service_event_schedule(pf);
6659
6660         return 0;
6661 }
6662
6663 /**
6664  * i40e_vsi_reinit_locked - Reset the VSI
6665  * @vsi: the VSI being configured
6666  *
6667  * Rebuild the ring structs after some configuration
6668  * has changed, e.g. MTU size.
6669  **/
6670 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
6671 {
6672         struct i40e_pf *pf = vsi->back;
6673
6674         WARN_ON(in_interrupt());
6675         while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
6676                 usleep_range(1000, 2000);
6677         i40e_down(vsi);
6678
6679         i40e_up(vsi);
6680         clear_bit(__I40E_CONFIG_BUSY, pf->state);
6681 }
6682
6683 /**
6684  * i40e_up - Bring the connection back up after being down
6685  * @vsi: the VSI being configured
6686  **/
6687 int i40e_up(struct i40e_vsi *vsi)
6688 {
6689         int err;
6690
6691         err = i40e_vsi_configure(vsi);
6692         if (!err)
6693                 err = i40e_up_complete(vsi);
6694
6695         return err;
6696 }
6697
6698 /**
6699  * i40e_force_link_state - Force the link status
6700  * @pf: board private structure
6701  * @is_up: whether the link state should be forced up or down
6702  **/
6703 static i40e_status i40e_force_link_state(struct i40e_pf *pf, bool is_up)
6704 {
6705         struct i40e_aq_get_phy_abilities_resp abilities;
6706         struct i40e_aq_set_phy_config config = {0};
6707         struct i40e_hw *hw = &pf->hw;
6708         i40e_status err;
6709         u64 mask;
6710         u8 speed;
6711
6712         /* Card might've been put in an unstable state by other drivers
6713          * and applications, which causes incorrect speed values being
6714          * set on startup. In order to clear speed registers, we call
6715          * get_phy_capabilities twice, once to get initial state of
6716          * available speeds, and once to get current PHY config.
6717          */
6718         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities,
6719                                            NULL);
6720         if (err) {
6721                 dev_err(&pf->pdev->dev,
6722                         "failed to get phy cap., ret =  %s last_status =  %s\n",
6723                         i40e_stat_str(hw, err),
6724                         i40e_aq_str(hw, hw->aq.asq_last_status));
6725                 return err;
6726         }
6727         speed = abilities.link_speed;
6728
6729         /* Get the current phy config */
6730         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities,
6731                                            NULL);
6732         if (err) {
6733                 dev_err(&pf->pdev->dev,
6734                         "failed to get phy cap., ret =  %s last_status =  %s\n",
6735                         i40e_stat_str(hw, err),
6736                         i40e_aq_str(hw, hw->aq.asq_last_status));
6737                 return err;
6738         }
6739
6740         /* If link needs to go up, but was not forced to go down,
6741          * and its speed values are OK, no need for a flap
6742          */
6743         if (is_up && abilities.phy_type != 0 && abilities.link_speed != 0)
6744                 return I40E_SUCCESS;
6745
6746         /* To force link we need to set bits for all supported PHY types,
6747          * but there are now more than 32, so we need to split the bitmap
6748          * across two fields.
6749          */
6750         mask = I40E_PHY_TYPES_BITMASK;
6751         config.phy_type = is_up ? cpu_to_le32((u32)(mask & 0xffffffff)) : 0;
6752         config.phy_type_ext = is_up ? (u8)((mask >> 32) & 0xff) : 0;
6753         /* Copy the old settings, except of phy_type */
6754         config.abilities = abilities.abilities;
6755         if (abilities.link_speed != 0)
6756                 config.link_speed = abilities.link_speed;
6757         else
6758                 config.link_speed = speed;
6759         config.eee_capability = abilities.eee_capability;
6760         config.eeer = abilities.eeer_val;
6761         config.low_power_ctrl = abilities.d3_lpan;
6762         config.fec_config = abilities.fec_cfg_curr_mod_ext_info &
6763                             I40E_AQ_PHY_FEC_CONFIG_MASK;
6764         err = i40e_aq_set_phy_config(hw, &config, NULL);
6765
6766         if (err) {
6767                 dev_err(&pf->pdev->dev,
6768                         "set phy config ret =  %s last_status =  %s\n",
6769                         i40e_stat_str(&pf->hw, err),
6770                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6771                 return err;
6772         }
6773
6774         /* Update the link info */
6775         err = i40e_update_link_info(hw);
6776         if (err) {
6777                 /* Wait a little bit (on 40G cards it sometimes takes a really
6778                  * long time for link to come back from the atomic reset)
6779                  * and try once more
6780                  */
6781                 msleep(1000);
6782                 i40e_update_link_info(hw);
6783         }
6784
6785         i40e_aq_set_link_restart_an(hw, true, NULL);
6786
6787         return I40E_SUCCESS;
6788 }
6789
6790 /**
6791  * i40e_down - Shutdown the connection processing
6792  * @vsi: the VSI being stopped
6793  **/
6794 void i40e_down(struct i40e_vsi *vsi)
6795 {
6796         int i;
6797
6798         /* It is assumed that the caller of this function
6799          * sets the vsi->state __I40E_VSI_DOWN bit.
6800          */
6801         if (vsi->netdev) {
6802                 netif_carrier_off(vsi->netdev);
6803                 netif_tx_disable(vsi->netdev);
6804         }
6805         i40e_vsi_disable_irq(vsi);
6806         i40e_vsi_stop_rings(vsi);
6807         if (vsi->type == I40E_VSI_MAIN &&
6808             vsi->back->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED)
6809                 i40e_force_link_state(vsi->back, false);
6810         i40e_napi_disable_all(vsi);
6811
6812         for (i = 0; i < vsi->num_queue_pairs; i++) {
6813                 i40e_clean_tx_ring(vsi->tx_rings[i]);
6814                 if (i40e_enabled_xdp_vsi(vsi)) {
6815                         /* Make sure that in-progress ndo_xdp_xmit and
6816                          * ndo_xsk_wakeup calls are completed.
6817                          */
6818                         synchronize_rcu();
6819                         i40e_clean_tx_ring(vsi->xdp_rings[i]);
6820                 }
6821                 i40e_clean_rx_ring(vsi->rx_rings[i]);
6822         }
6823
6824 }
6825
6826 /**
6827  * i40e_validate_mqprio_qopt- validate queue mapping info
6828  * @vsi: the VSI being configured
6829  * @mqprio_qopt: queue parametrs
6830  **/
6831 static int i40e_validate_mqprio_qopt(struct i40e_vsi *vsi,
6832                                      struct tc_mqprio_qopt_offload *mqprio_qopt)
6833 {
6834         u64 sum_max_rate = 0;
6835         u64 max_rate = 0;
6836         int i;
6837
6838         if (mqprio_qopt->qopt.offset[0] != 0 ||
6839             mqprio_qopt->qopt.num_tc < 1 ||
6840             mqprio_qopt->qopt.num_tc > I40E_MAX_TRAFFIC_CLASS)
6841                 return -EINVAL;
6842         for (i = 0; ; i++) {
6843                 if (!mqprio_qopt->qopt.count[i])
6844                         return -EINVAL;
6845                 if (mqprio_qopt->min_rate[i]) {
6846                         dev_err(&vsi->back->pdev->dev,
6847                                 "Invalid min tx rate (greater than 0) specified\n");
6848                         return -EINVAL;
6849                 }
6850                 max_rate = mqprio_qopt->max_rate[i];
6851                 do_div(max_rate, I40E_BW_MBPS_DIVISOR);
6852                 sum_max_rate += max_rate;
6853
6854                 if (i >= mqprio_qopt->qopt.num_tc - 1)
6855                         break;
6856                 if (mqprio_qopt->qopt.offset[i + 1] !=
6857                     (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
6858                         return -EINVAL;
6859         }
6860         if (vsi->num_queue_pairs <
6861             (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i])) {
6862                 return -EINVAL;
6863         }
6864         if (sum_max_rate > i40e_get_link_speed(vsi)) {
6865                 dev_err(&vsi->back->pdev->dev,
6866                         "Invalid max tx rate specified\n");
6867                 return -EINVAL;
6868         }
6869         return 0;
6870 }
6871
6872 /**
6873  * i40e_vsi_set_default_tc_config - set default values for tc configuration
6874  * @vsi: the VSI being configured
6875  **/
6876 static void i40e_vsi_set_default_tc_config(struct i40e_vsi *vsi)
6877 {
6878         u16 qcount;
6879         int i;
6880
6881         /* Only TC0 is enabled */
6882         vsi->tc_config.numtc = 1;
6883         vsi->tc_config.enabled_tc = 1;
6884         qcount = min_t(int, vsi->alloc_queue_pairs,
6885                        i40e_pf_get_max_q_per_tc(vsi->back));
6886         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6887                 /* For the TC that is not enabled set the offset to to default
6888                  * queue and allocate one queue for the given TC.
6889                  */
6890                 vsi->tc_config.tc_info[i].qoffset = 0;
6891                 if (i == 0)
6892                         vsi->tc_config.tc_info[i].qcount = qcount;
6893                 else
6894                         vsi->tc_config.tc_info[i].qcount = 1;
6895                 vsi->tc_config.tc_info[i].netdev_tc = 0;
6896         }
6897 }
6898
6899 /**
6900  * i40e_del_macvlan_filter
6901  * @hw: pointer to the HW structure
6902  * @seid: seid of the channel VSI
6903  * @macaddr: the mac address to apply as a filter
6904  * @aq_err: store the admin Q error
6905  *
6906  * This function deletes a mac filter on the channel VSI which serves as the
6907  * macvlan. Returns 0 on success.
6908  **/
6909 static i40e_status i40e_del_macvlan_filter(struct i40e_hw *hw, u16 seid,
6910                                            const u8 *macaddr, int *aq_err)
6911 {
6912         struct i40e_aqc_remove_macvlan_element_data element;
6913         i40e_status status;
6914
6915         memset(&element, 0, sizeof(element));
6916         ether_addr_copy(element.mac_addr, macaddr);
6917         element.vlan_tag = 0;
6918         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
6919         status = i40e_aq_remove_macvlan(hw, seid, &element, 1, NULL);
6920         *aq_err = hw->aq.asq_last_status;
6921
6922         return status;
6923 }
6924
6925 /**
6926  * i40e_add_macvlan_filter
6927  * @hw: pointer to the HW structure
6928  * @seid: seid of the channel VSI
6929  * @macaddr: the mac address to apply as a filter
6930  * @aq_err: store the admin Q error
6931  *
6932  * This function adds a mac filter on the channel VSI which serves as the
6933  * macvlan. Returns 0 on success.
6934  **/
6935 static i40e_status i40e_add_macvlan_filter(struct i40e_hw *hw, u16 seid,
6936                                            const u8 *macaddr, int *aq_err)
6937 {
6938         struct i40e_aqc_add_macvlan_element_data element;
6939         i40e_status status;
6940         u16 cmd_flags = 0;
6941
6942         ether_addr_copy(element.mac_addr, macaddr);
6943         element.vlan_tag = 0;
6944         element.queue_number = 0;
6945         element.match_method = I40E_AQC_MM_ERR_NO_RES;
6946         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
6947         element.flags = cpu_to_le16(cmd_flags);
6948         status = i40e_aq_add_macvlan(hw, seid, &element, 1, NULL);
6949         *aq_err = hw->aq.asq_last_status;
6950
6951         return status;
6952 }
6953
6954 /**
6955  * i40e_reset_ch_rings - Reset the queue contexts in a channel
6956  * @vsi: the VSI we want to access
6957  * @ch: the channel we want to access
6958  */
6959 static void i40e_reset_ch_rings(struct i40e_vsi *vsi, struct i40e_channel *ch)
6960 {
6961         struct i40e_ring *tx_ring, *rx_ring;
6962         u16 pf_q;
6963         int i;
6964
6965         for (i = 0; i < ch->num_queue_pairs; i++) {
6966                 pf_q = ch->base_queue + i;
6967                 tx_ring = vsi->tx_rings[pf_q];
6968                 tx_ring->ch = NULL;
6969                 rx_ring = vsi->rx_rings[pf_q];
6970                 rx_ring->ch = NULL;
6971         }
6972 }
6973
6974 /**
6975  * i40e_free_macvlan_channels
6976  * @vsi: the VSI we want to access
6977  *
6978  * This function frees the Qs of the channel VSI from
6979  * the stack and also deletes the channel VSIs which
6980  * serve as macvlans.
6981  */
6982 static void i40e_free_macvlan_channels(struct i40e_vsi *vsi)
6983 {
6984         struct i40e_channel *ch, *ch_tmp;
6985         int ret;
6986
6987         if (list_empty(&vsi->macvlan_list))
6988                 return;
6989
6990         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
6991                 struct i40e_vsi *parent_vsi;
6992
6993                 if (i40e_is_channel_macvlan(ch)) {
6994                         i40e_reset_ch_rings(vsi, ch);
6995                         clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
6996                         netdev_unbind_sb_channel(vsi->netdev, ch->fwd->netdev);
6997                         netdev_set_sb_channel(ch->fwd->netdev, 0);
6998                         kfree(ch->fwd);
6999                         ch->fwd = NULL;
7000                 }
7001
7002                 list_del(&ch->list);
7003                 parent_vsi = ch->parent_vsi;
7004                 if (!parent_vsi || !ch->initialized) {
7005                         kfree(ch);
7006                         continue;
7007                 }
7008
7009                 /* remove the VSI */
7010                 ret = i40e_aq_delete_element(&vsi->back->hw, ch->seid,
7011                                              NULL);
7012                 if (ret)
7013                         dev_err(&vsi->back->pdev->dev,
7014                                 "unable to remove channel (%d) for parent VSI(%d)\n",
7015                                 ch->seid, parent_vsi->seid);
7016                 kfree(ch);
7017         }
7018         vsi->macvlan_cnt = 0;
7019 }
7020
7021 /**
7022  * i40e_fwd_ring_up - bring the macvlan device up
7023  * @vsi: the VSI we want to access
7024  * @vdev: macvlan netdevice
7025  * @fwd: the private fwd structure
7026  */
7027 static int i40e_fwd_ring_up(struct i40e_vsi *vsi, struct net_device *vdev,
7028                             struct i40e_fwd_adapter *fwd)
7029 {
7030         int ret = 0, num_tc = 1,  i, aq_err;
7031         struct i40e_channel *ch, *ch_tmp;
7032         struct i40e_pf *pf = vsi->back;
7033         struct i40e_hw *hw = &pf->hw;
7034
7035         if (list_empty(&vsi->macvlan_list))
7036                 return -EINVAL;
7037
7038         /* Go through the list and find an available channel */
7039         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7040                 if (!i40e_is_channel_macvlan(ch)) {
7041                         ch->fwd = fwd;
7042                         /* record configuration for macvlan interface in vdev */
7043                         for (i = 0; i < num_tc; i++)
7044                                 netdev_bind_sb_channel_queue(vsi->netdev, vdev,
7045                                                              i,
7046                                                              ch->num_queue_pairs,
7047                                                              ch->base_queue);
7048                         for (i = 0; i < ch->num_queue_pairs; i++) {
7049                                 struct i40e_ring *tx_ring, *rx_ring;
7050                                 u16 pf_q;
7051
7052                                 pf_q = ch->base_queue + i;
7053
7054                                 /* Get to TX ring ptr */
7055                                 tx_ring = vsi->tx_rings[pf_q];
7056                                 tx_ring->ch = ch;
7057
7058                                 /* Get the RX ring ptr */
7059                                 rx_ring = vsi->rx_rings[pf_q];
7060                                 rx_ring->ch = ch;
7061                         }
7062                         break;
7063                 }
7064         }
7065
7066         /* Guarantee all rings are updated before we update the
7067          * MAC address filter.
7068          */
7069         wmb();
7070
7071         /* Add a mac filter */
7072         ret = i40e_add_macvlan_filter(hw, ch->seid, vdev->dev_addr, &aq_err);
7073         if (ret) {
7074                 /* if we cannot add the MAC rule then disable the offload */
7075                 macvlan_release_l2fw_offload(vdev);
7076                 for (i = 0; i < ch->num_queue_pairs; i++) {
7077                         struct i40e_ring *rx_ring;
7078                         u16 pf_q;
7079
7080                         pf_q = ch->base_queue + i;
7081                         rx_ring = vsi->rx_rings[pf_q];
7082                         rx_ring->netdev = NULL;
7083                 }
7084                 dev_info(&pf->pdev->dev,
7085                          "Error adding mac filter on macvlan err %s, aq_err %s\n",
7086                           i40e_stat_str(hw, ret),
7087                           i40e_aq_str(hw, aq_err));
7088                 netdev_err(vdev, "L2fwd offload disabled to L2 filter error\n");
7089         }
7090
7091         return ret;
7092 }
7093
7094 /**
7095  * i40e_setup_macvlans - create the channels which will be macvlans
7096  * @vsi: the VSI we want to access
7097  * @macvlan_cnt: no. of macvlans to be setup
7098  * @qcnt: no. of Qs per macvlan
7099  * @vdev: macvlan netdevice
7100  */
7101 static int i40e_setup_macvlans(struct i40e_vsi *vsi, u16 macvlan_cnt, u16 qcnt,
7102                                struct net_device *vdev)
7103 {
7104         struct i40e_pf *pf = vsi->back;
7105         struct i40e_hw *hw = &pf->hw;
7106         struct i40e_vsi_context ctxt;
7107         u16 sections, qmap, num_qps;
7108         struct i40e_channel *ch;
7109         int i, pow, ret = 0;
7110         u8 offset = 0;
7111
7112         if (vsi->type != I40E_VSI_MAIN || !macvlan_cnt)
7113                 return -EINVAL;
7114
7115         num_qps = vsi->num_queue_pairs - (macvlan_cnt * qcnt);
7116
7117         /* find the next higher power-of-2 of num queue pairs */
7118         pow = fls(roundup_pow_of_two(num_qps) - 1);
7119
7120         qmap = (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
7121                 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
7122
7123         /* Setup context bits for the main VSI */
7124         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
7125         sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
7126         memset(&ctxt, 0, sizeof(ctxt));
7127         ctxt.seid = vsi->seid;
7128         ctxt.pf_num = vsi->back->hw.pf_id;
7129         ctxt.vf_num = 0;
7130         ctxt.uplink_seid = vsi->uplink_seid;
7131         ctxt.info = vsi->info;
7132         ctxt.info.tc_mapping[0] = cpu_to_le16(qmap);
7133         ctxt.info.mapping_flags |= cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
7134         ctxt.info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
7135         ctxt.info.valid_sections |= cpu_to_le16(sections);
7136
7137         /* Reconfigure RSS for main VSI with new max queue count */
7138         vsi->rss_size = max_t(u16, num_qps, qcnt);
7139         ret = i40e_vsi_config_rss(vsi);
7140         if (ret) {
7141                 dev_info(&pf->pdev->dev,
7142                          "Failed to reconfig RSS for num_queues (%u)\n",
7143                          vsi->rss_size);
7144                 return ret;
7145         }
7146         vsi->reconfig_rss = true;
7147         dev_dbg(&vsi->back->pdev->dev,
7148                 "Reconfigured RSS with num_queues (%u)\n", vsi->rss_size);
7149         vsi->next_base_queue = num_qps;
7150         vsi->cnt_q_avail = vsi->num_queue_pairs - num_qps;
7151
7152         /* Update the VSI after updating the VSI queue-mapping
7153          * information
7154          */
7155         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
7156         if (ret) {
7157                 dev_info(&pf->pdev->dev,
7158                          "Update vsi tc config failed, err %s aq_err %s\n",
7159                          i40e_stat_str(hw, ret),
7160                          i40e_aq_str(hw, hw->aq.asq_last_status));
7161                 return ret;
7162         }
7163         /* update the local VSI info with updated queue map */
7164         i40e_vsi_update_queue_map(vsi, &ctxt);
7165         vsi->info.valid_sections = 0;
7166
7167         /* Create channels for macvlans */
7168         INIT_LIST_HEAD(&vsi->macvlan_list);
7169         for (i = 0; i < macvlan_cnt; i++) {
7170                 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
7171                 if (!ch) {
7172                         ret = -ENOMEM;
7173                         goto err_free;
7174                 }
7175                 INIT_LIST_HEAD(&ch->list);
7176                 ch->num_queue_pairs = qcnt;
7177                 if (!i40e_setup_channel(pf, vsi, ch)) {
7178                         ret = -EINVAL;
7179                         kfree(ch);
7180                         goto err_free;
7181                 }
7182                 ch->parent_vsi = vsi;
7183                 vsi->cnt_q_avail -= ch->num_queue_pairs;
7184                 vsi->macvlan_cnt++;
7185                 list_add_tail(&ch->list, &vsi->macvlan_list);
7186         }
7187
7188         return ret;
7189
7190 err_free:
7191         dev_info(&pf->pdev->dev, "Failed to setup macvlans\n");
7192         i40e_free_macvlan_channels(vsi);
7193
7194         return ret;
7195 }
7196
7197 /**
7198  * i40e_fwd_add - configure macvlans
7199  * @netdev: net device to configure
7200  * @vdev: macvlan netdevice
7201  **/
7202 static void *i40e_fwd_add(struct net_device *netdev, struct net_device *vdev)
7203 {
7204         struct i40e_netdev_priv *np = netdev_priv(netdev);
7205         u16 q_per_macvlan = 0, macvlan_cnt = 0, vectors;
7206         struct i40e_vsi *vsi = np->vsi;
7207         struct i40e_pf *pf = vsi->back;
7208         struct i40e_fwd_adapter *fwd;
7209         int avail_macvlan, ret;
7210
7211         if ((pf->flags & I40E_FLAG_DCB_ENABLED)) {
7212                 netdev_info(netdev, "Macvlans are not supported when DCB is enabled\n");
7213                 return ERR_PTR(-EINVAL);
7214         }
7215         if ((pf->flags & I40E_FLAG_TC_MQPRIO)) {
7216                 netdev_info(netdev, "Macvlans are not supported when HW TC offload is on\n");
7217                 return ERR_PTR(-EINVAL);
7218         }
7219         if (pf->num_lan_msix < I40E_MIN_MACVLAN_VECTORS) {
7220                 netdev_info(netdev, "Not enough vectors available to support macvlans\n");
7221                 return ERR_PTR(-EINVAL);
7222         }
7223
7224         /* The macvlan device has to be a single Q device so that the
7225          * tc_to_txq field can be reused to pick the tx queue.
7226          */
7227         if (netif_is_multiqueue(vdev))
7228                 return ERR_PTR(-ERANGE);
7229
7230         if (!vsi->macvlan_cnt) {
7231                 /* reserve bit 0 for the pf device */
7232                 set_bit(0, vsi->fwd_bitmask);
7233
7234                 /* Try to reserve as many queues as possible for macvlans. First
7235                  * reserve 3/4th of max vectors, then half, then quarter and
7236                  * calculate Qs per macvlan as you go
7237                  */
7238                 vectors = pf->num_lan_msix;
7239                 if (vectors <= I40E_MAX_MACVLANS && vectors > 64) {
7240                         /* allocate 4 Qs per macvlan and 32 Qs to the PF*/
7241                         q_per_macvlan = 4;
7242                         macvlan_cnt = (vectors - 32) / 4;
7243                 } else if (vectors <= 64 && vectors > 32) {
7244                         /* allocate 2 Qs per macvlan and 16 Qs to the PF*/
7245                         q_per_macvlan = 2;
7246                         macvlan_cnt = (vectors - 16) / 2;
7247                 } else if (vectors <= 32 && vectors > 16) {
7248                         /* allocate 1 Q per macvlan and 16 Qs to the PF*/
7249                         q_per_macvlan = 1;
7250                         macvlan_cnt = vectors - 16;
7251                 } else if (vectors <= 16 && vectors > 8) {
7252                         /* allocate 1 Q per macvlan and 8 Qs to the PF */
7253                         q_per_macvlan = 1;
7254                         macvlan_cnt = vectors - 8;
7255                 } else {
7256                         /* allocate 1 Q per macvlan and 1 Q to the PF */
7257                         q_per_macvlan = 1;
7258                         macvlan_cnt = vectors - 1;
7259                 }
7260
7261                 if (macvlan_cnt == 0)
7262                         return ERR_PTR(-EBUSY);
7263
7264                 /* Quiesce VSI queues */
7265                 i40e_quiesce_vsi(vsi);
7266
7267                 /* sets up the macvlans but does not "enable" them */
7268                 ret = i40e_setup_macvlans(vsi, macvlan_cnt, q_per_macvlan,
7269                                           vdev);
7270                 if (ret)
7271                         return ERR_PTR(ret);
7272
7273                 /* Unquiesce VSI */
7274                 i40e_unquiesce_vsi(vsi);
7275         }
7276         avail_macvlan = find_first_zero_bit(vsi->fwd_bitmask,
7277                                             vsi->macvlan_cnt);
7278         if (avail_macvlan >= I40E_MAX_MACVLANS)
7279                 return ERR_PTR(-EBUSY);
7280
7281         /* create the fwd struct */
7282         fwd = kzalloc(sizeof(*fwd), GFP_KERNEL);
7283         if (!fwd)
7284                 return ERR_PTR(-ENOMEM);
7285
7286         set_bit(avail_macvlan, vsi->fwd_bitmask);
7287         fwd->bit_no = avail_macvlan;
7288         netdev_set_sb_channel(vdev, avail_macvlan);
7289         fwd->netdev = vdev;
7290
7291         if (!netif_running(netdev))
7292                 return fwd;
7293
7294         /* Set fwd ring up */
7295         ret = i40e_fwd_ring_up(vsi, vdev, fwd);
7296         if (ret) {
7297                 /* unbind the queues and drop the subordinate channel config */
7298                 netdev_unbind_sb_channel(netdev, vdev);
7299                 netdev_set_sb_channel(vdev, 0);
7300
7301                 kfree(fwd);
7302                 return ERR_PTR(-EINVAL);
7303         }
7304
7305         return fwd;
7306 }
7307
7308 /**
7309  * i40e_del_all_macvlans - Delete all the mac filters on the channels
7310  * @vsi: the VSI we want to access
7311  */
7312 static void i40e_del_all_macvlans(struct i40e_vsi *vsi)
7313 {
7314         struct i40e_channel *ch, *ch_tmp;
7315         struct i40e_pf *pf = vsi->back;
7316         struct i40e_hw *hw = &pf->hw;
7317         int aq_err, ret = 0;
7318
7319         if (list_empty(&vsi->macvlan_list))
7320                 return;
7321
7322         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7323                 if (i40e_is_channel_macvlan(ch)) {
7324                         ret = i40e_del_macvlan_filter(hw, ch->seid,
7325                                                       i40e_channel_mac(ch),
7326                                                       &aq_err);
7327                         if (!ret) {
7328                                 /* Reset queue contexts */
7329                                 i40e_reset_ch_rings(vsi, ch);
7330                                 clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7331                                 netdev_unbind_sb_channel(vsi->netdev,
7332                                                          ch->fwd->netdev);
7333                                 netdev_set_sb_channel(ch->fwd->netdev, 0);
7334                                 kfree(ch->fwd);
7335                                 ch->fwd = NULL;
7336                         }
7337                 }
7338         }
7339 }
7340
7341 /**
7342  * i40e_fwd_del - delete macvlan interfaces
7343  * @netdev: net device to configure
7344  * @vdev: macvlan netdevice
7345  */
7346 static void i40e_fwd_del(struct net_device *netdev, void *vdev)
7347 {
7348         struct i40e_netdev_priv *np = netdev_priv(netdev);
7349         struct i40e_fwd_adapter *fwd = vdev;
7350         struct i40e_channel *ch, *ch_tmp;
7351         struct i40e_vsi *vsi = np->vsi;
7352         struct i40e_pf *pf = vsi->back;
7353         struct i40e_hw *hw = &pf->hw;
7354         int aq_err, ret = 0;
7355
7356         /* Find the channel associated with the macvlan and del mac filter */
7357         list_for_each_entry_safe(ch, ch_tmp, &vsi->macvlan_list, list) {
7358                 if (i40e_is_channel_macvlan(ch) &&
7359                     ether_addr_equal(i40e_channel_mac(ch),
7360                                      fwd->netdev->dev_addr)) {
7361                         ret = i40e_del_macvlan_filter(hw, ch->seid,
7362                                                       i40e_channel_mac(ch),
7363                                                       &aq_err);
7364                         if (!ret) {
7365                                 /* Reset queue contexts */
7366                                 i40e_reset_ch_rings(vsi, ch);
7367                                 clear_bit(ch->fwd->bit_no, vsi->fwd_bitmask);
7368                                 netdev_unbind_sb_channel(netdev, fwd->netdev);
7369                                 netdev_set_sb_channel(fwd->netdev, 0);
7370                                 kfree(ch->fwd);
7371                                 ch->fwd = NULL;
7372                         } else {
7373                                 dev_info(&pf->pdev->dev,
7374                                          "Error deleting mac filter on macvlan err %s, aq_err %s\n",
7375                                           i40e_stat_str(hw, ret),
7376                                           i40e_aq_str(hw, aq_err));
7377                         }
7378                         break;
7379                 }
7380         }
7381 }
7382
7383 /**
7384  * i40e_setup_tc - configure multiple traffic classes
7385  * @netdev: net device to configure
7386  * @type_data: tc offload data
7387  **/
7388 static int i40e_setup_tc(struct net_device *netdev, void *type_data)
7389 {
7390         struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
7391         struct i40e_netdev_priv *np = netdev_priv(netdev);
7392         struct i40e_vsi *vsi = np->vsi;
7393         struct i40e_pf *pf = vsi->back;
7394         u8 enabled_tc = 0, num_tc, hw;
7395         bool need_reset = false;
7396         int old_queue_pairs;
7397         int ret = -EINVAL;
7398         u16 mode;
7399         int i;
7400
7401         old_queue_pairs = vsi->num_queue_pairs;
7402         num_tc = mqprio_qopt->qopt.num_tc;
7403         hw = mqprio_qopt->qopt.hw;
7404         mode = mqprio_qopt->mode;
7405         if (!hw) {
7406                 pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7407                 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
7408                 goto config_tc;
7409         }
7410
7411         /* Check if MFP enabled */
7412         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
7413                 netdev_info(netdev,
7414                             "Configuring TC not supported in MFP mode\n");
7415                 return ret;
7416         }
7417         switch (mode) {
7418         case TC_MQPRIO_MODE_DCB:
7419                 pf->flags &= ~I40E_FLAG_TC_MQPRIO;
7420
7421                 /* Check if DCB enabled to continue */
7422                 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
7423                         netdev_info(netdev,
7424                                     "DCB is not enabled for adapter\n");
7425                         return ret;
7426                 }
7427
7428                 /* Check whether tc count is within enabled limit */
7429                 if (num_tc > i40e_pf_get_num_tc(pf)) {
7430                         netdev_info(netdev,
7431                                     "TC count greater than enabled on link for adapter\n");
7432                         return ret;
7433                 }
7434                 break;
7435         case TC_MQPRIO_MODE_CHANNEL:
7436                 if (pf->flags & I40E_FLAG_DCB_ENABLED) {
7437                         netdev_info(netdev,
7438                                     "Full offload of TC Mqprio options is not supported when DCB is enabled\n");
7439                         return ret;
7440                 }
7441                 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7442                         return ret;
7443                 ret = i40e_validate_mqprio_qopt(vsi, mqprio_qopt);
7444                 if (ret)
7445                         return ret;
7446                 memcpy(&vsi->mqprio_qopt, mqprio_qopt,
7447                        sizeof(*mqprio_qopt));
7448                 pf->flags |= I40E_FLAG_TC_MQPRIO;
7449                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
7450                 break;
7451         default:
7452                 return -EINVAL;
7453         }
7454
7455 config_tc:
7456         /* Generate TC map for number of tc requested */
7457         for (i = 0; i < num_tc; i++)
7458                 enabled_tc |= BIT(i);
7459
7460         /* Requesting same TC configuration as already enabled */
7461         if (enabled_tc == vsi->tc_config.enabled_tc &&
7462             mode != TC_MQPRIO_MODE_CHANNEL)
7463                 return 0;
7464
7465         /* Quiesce VSI queues */
7466         i40e_quiesce_vsi(vsi);
7467
7468         if (!hw && !(pf->flags & I40E_FLAG_TC_MQPRIO))
7469                 i40e_remove_queue_channels(vsi);
7470
7471         /* Configure VSI for enabled TCs */
7472         ret = i40e_vsi_config_tc(vsi, enabled_tc);
7473         if (ret) {
7474                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
7475                             vsi->seid);
7476                 need_reset = true;
7477                 goto exit;
7478         } else {
7479                 dev_info(&vsi->back->pdev->dev,
7480                          "Setup channel (id:%u) utilizing num_queues %d\n",
7481                          vsi->seid, vsi->tc_config.tc_info[0].qcount);
7482         }
7483
7484         if (pf->flags & I40E_FLAG_TC_MQPRIO) {
7485                 if (vsi->mqprio_qopt.max_rate[0]) {
7486                         u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
7487
7488                         do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
7489                         ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
7490                         if (!ret) {
7491                                 u64 credits = max_tx_rate;
7492
7493                                 do_div(credits, I40E_BW_CREDIT_DIVISOR);
7494                                 dev_dbg(&vsi->back->pdev->dev,
7495                                         "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
7496                                         max_tx_rate,
7497                                         credits,
7498                                         vsi->seid);
7499                         } else {
7500                                 need_reset = true;
7501                                 goto exit;
7502                         }
7503                 }
7504                 ret = i40e_configure_queue_channels(vsi);
7505                 if (ret) {
7506                         vsi->num_queue_pairs = old_queue_pairs;
7507                         netdev_info(netdev,
7508                                     "Failed configuring queue channels\n");
7509                         need_reset = true;
7510                         goto exit;
7511                 }
7512         }
7513
7514 exit:
7515         /* Reset the configuration data to defaults, only TC0 is enabled */
7516         if (need_reset) {
7517                 i40e_vsi_set_default_tc_config(vsi);
7518                 need_reset = false;
7519         }
7520
7521         /* Unquiesce VSI */
7522         i40e_unquiesce_vsi(vsi);
7523         return ret;
7524 }
7525
7526 /**
7527  * i40e_set_cld_element - sets cloud filter element data
7528  * @filter: cloud filter rule
7529  * @cld: ptr to cloud filter element data
7530  *
7531  * This is helper function to copy data into cloud filter element
7532  **/
7533 static inline void
7534 i40e_set_cld_element(struct i40e_cloud_filter *filter,
7535                      struct i40e_aqc_cloud_filters_element_data *cld)
7536 {
7537         int i, j;
7538         u32 ipa;
7539
7540         memset(cld, 0, sizeof(*cld));
7541         ether_addr_copy(cld->outer_mac, filter->dst_mac);
7542         ether_addr_copy(cld->inner_mac, filter->src_mac);
7543
7544         if (filter->n_proto != ETH_P_IP && filter->n_proto != ETH_P_IPV6)
7545                 return;
7546
7547         if (filter->n_proto == ETH_P_IPV6) {
7548 #define IPV6_MAX_INDEX  (ARRAY_SIZE(filter->dst_ipv6) - 1)
7549                 for (i = 0, j = 0; i < ARRAY_SIZE(filter->dst_ipv6);
7550                      i++, j += 2) {
7551                         ipa = be32_to_cpu(filter->dst_ipv6[IPV6_MAX_INDEX - i]);
7552                         ipa = cpu_to_le32(ipa);
7553                         memcpy(&cld->ipaddr.raw_v6.data[j], &ipa, sizeof(ipa));
7554                 }
7555         } else {
7556                 ipa = be32_to_cpu(filter->dst_ipv4);
7557                 memcpy(&cld->ipaddr.v4.data, &ipa, sizeof(ipa));
7558         }
7559
7560         cld->inner_vlan = cpu_to_le16(ntohs(filter->vlan_id));
7561
7562         /* tenant_id is not supported by FW now, once the support is enabled
7563          * fill the cld->tenant_id with cpu_to_le32(filter->tenant_id)
7564          */
7565         if (filter->tenant_id)
7566                 return;
7567 }
7568
7569 /**
7570  * i40e_add_del_cloud_filter - Add/del cloud filter
7571  * @vsi: pointer to VSI
7572  * @filter: cloud filter rule
7573  * @add: if true, add, if false, delete
7574  *
7575  * Add or delete a cloud filter for a specific flow spec.
7576  * Returns 0 if the filter were successfully added.
7577  **/
7578 int i40e_add_del_cloud_filter(struct i40e_vsi *vsi,
7579                               struct i40e_cloud_filter *filter, bool add)
7580 {
7581         struct i40e_aqc_cloud_filters_element_data cld_filter;
7582         struct i40e_pf *pf = vsi->back;
7583         int ret;
7584         static const u16 flag_table[128] = {
7585                 [I40E_CLOUD_FILTER_FLAGS_OMAC]  =
7586                         I40E_AQC_ADD_CLOUD_FILTER_OMAC,
7587                 [I40E_CLOUD_FILTER_FLAGS_IMAC]  =
7588                         I40E_AQC_ADD_CLOUD_FILTER_IMAC,
7589                 [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN]  =
7590                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN,
7591                 [I40E_CLOUD_FILTER_FLAGS_IMAC_TEN_ID] =
7592                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_TEN_ID,
7593                 [I40E_CLOUD_FILTER_FLAGS_OMAC_TEN_ID_IMAC] =
7594                         I40E_AQC_ADD_CLOUD_FILTER_OMAC_TEN_ID_IMAC,
7595                 [I40E_CLOUD_FILTER_FLAGS_IMAC_IVLAN_TEN_ID] =
7596                         I40E_AQC_ADD_CLOUD_FILTER_IMAC_IVLAN_TEN_ID,
7597                 [I40E_CLOUD_FILTER_FLAGS_IIP] =
7598                         I40E_AQC_ADD_CLOUD_FILTER_IIP,
7599         };
7600
7601         if (filter->flags >= ARRAY_SIZE(flag_table))
7602                 return I40E_ERR_CONFIG;
7603
7604         /* copy element needed to add cloud filter from filter */
7605         i40e_set_cld_element(filter, &cld_filter);
7606
7607         if (filter->tunnel_type != I40E_CLOUD_TNL_TYPE_NONE)
7608                 cld_filter.flags = cpu_to_le16(filter->tunnel_type <<
7609                                              I40E_AQC_ADD_CLOUD_TNL_TYPE_SHIFT);
7610
7611         if (filter->n_proto == ETH_P_IPV6)
7612                 cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7613                                                 I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7614         else
7615                 cld_filter.flags |= cpu_to_le16(flag_table[filter->flags] |
7616                                                 I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7617
7618         if (add)
7619                 ret = i40e_aq_add_cloud_filters(&pf->hw, filter->seid,
7620                                                 &cld_filter, 1);
7621         else
7622                 ret = i40e_aq_rem_cloud_filters(&pf->hw, filter->seid,
7623                                                 &cld_filter, 1);
7624         if (ret)
7625                 dev_dbg(&pf->pdev->dev,
7626                         "Failed to %s cloud filter using l4 port %u, err %d aq_err %d\n",
7627                         add ? "add" : "delete", filter->dst_port, ret,
7628                         pf->hw.aq.asq_last_status);
7629         else
7630                 dev_info(&pf->pdev->dev,
7631                          "%s cloud filter for VSI: %d\n",
7632                          add ? "Added" : "Deleted", filter->seid);
7633         return ret;
7634 }
7635
7636 /**
7637  * i40e_add_del_cloud_filter_big_buf - Add/del cloud filter using big_buf
7638  * @vsi: pointer to VSI
7639  * @filter: cloud filter rule
7640  * @add: if true, add, if false, delete
7641  *
7642  * Add or delete a cloud filter for a specific flow spec using big buffer.
7643  * Returns 0 if the filter were successfully added.
7644  **/
7645 int i40e_add_del_cloud_filter_big_buf(struct i40e_vsi *vsi,
7646                                       struct i40e_cloud_filter *filter,
7647                                       bool add)
7648 {
7649         struct i40e_aqc_cloud_filters_element_bb cld_filter;
7650         struct i40e_pf *pf = vsi->back;
7651         int ret;
7652
7653         /* Both (src/dst) valid mac_addr are not supported */
7654         if ((is_valid_ether_addr(filter->dst_mac) &&
7655              is_valid_ether_addr(filter->src_mac)) ||
7656             (is_multicast_ether_addr(filter->dst_mac) &&
7657              is_multicast_ether_addr(filter->src_mac)))
7658                 return -EOPNOTSUPP;
7659
7660         /* Big buffer cloud filter needs 'L4 port' to be non-zero. Also, UDP
7661          * ports are not supported via big buffer now.
7662          */
7663         if (!filter->dst_port || filter->ip_proto == IPPROTO_UDP)
7664                 return -EOPNOTSUPP;
7665
7666         /* adding filter using src_port/src_ip is not supported at this stage */
7667         if (filter->src_port || filter->src_ipv4 ||
7668             !ipv6_addr_any(&filter->ip.v6.src_ip6))
7669                 return -EOPNOTSUPP;
7670
7671         /* copy element needed to add cloud filter from filter */
7672         i40e_set_cld_element(filter, &cld_filter.element);
7673
7674         if (is_valid_ether_addr(filter->dst_mac) ||
7675             is_valid_ether_addr(filter->src_mac) ||
7676             is_multicast_ether_addr(filter->dst_mac) ||
7677             is_multicast_ether_addr(filter->src_mac)) {
7678                 /* MAC + IP : unsupported mode */
7679                 if (filter->dst_ipv4)
7680                         return -EOPNOTSUPP;
7681
7682                 /* since we validated that L4 port must be valid before
7683                  * we get here, start with respective "flags" value
7684                  * and update if vlan is present or not
7685                  */
7686                 cld_filter.element.flags =
7687                         cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_PORT);
7688
7689                 if (filter->vlan_id) {
7690                         cld_filter.element.flags =
7691                         cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_MAC_VLAN_PORT);
7692                 }
7693
7694         } else if (filter->dst_ipv4 ||
7695                    !ipv6_addr_any(&filter->ip.v6.dst_ip6)) {
7696                 cld_filter.element.flags =
7697                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FILTER_IP_PORT);
7698                 if (filter->n_proto == ETH_P_IPV6)
7699                         cld_filter.element.flags |=
7700                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV6);
7701                 else
7702                         cld_filter.element.flags |=
7703                                 cpu_to_le16(I40E_AQC_ADD_CLOUD_FLAGS_IPV4);
7704         } else {
7705                 dev_err(&pf->pdev->dev,
7706                         "either mac or ip has to be valid for cloud filter\n");
7707                 return -EINVAL;
7708         }
7709
7710         /* Now copy L4 port in Byte 6..7 in general fields */
7711         cld_filter.general_fields[I40E_AQC_ADD_CLOUD_FV_FLU_0X16_WORD0] =
7712                                                 be16_to_cpu(filter->dst_port);
7713
7714         if (add) {
7715                 /* Validate current device switch mode, change if necessary */
7716                 ret = i40e_validate_and_set_switch_mode(vsi);
7717                 if (ret) {
7718                         dev_err(&pf->pdev->dev,
7719                                 "failed to set switch mode, ret %d\n",
7720                                 ret);
7721                         return ret;
7722                 }
7723
7724                 ret = i40e_aq_add_cloud_filters_bb(&pf->hw, filter->seid,
7725                                                    &cld_filter, 1);
7726         } else {
7727                 ret = i40e_aq_rem_cloud_filters_bb(&pf->hw, filter->seid,
7728                                                    &cld_filter, 1);
7729         }
7730
7731         if (ret)
7732                 dev_dbg(&pf->pdev->dev,
7733                         "Failed to %s cloud filter(big buffer) err %d aq_err %d\n",
7734                         add ? "add" : "delete", ret, pf->hw.aq.asq_last_status);
7735         else
7736                 dev_info(&pf->pdev->dev,
7737                          "%s cloud filter for VSI: %d, L4 port: %d\n",
7738                          add ? "add" : "delete", filter->seid,
7739                          ntohs(filter->dst_port));
7740         return ret;
7741 }
7742
7743 /**
7744  * i40e_parse_cls_flower - Parse tc flower filters provided by kernel
7745  * @vsi: Pointer to VSI
7746  * @cls_flower: Pointer to struct flow_cls_offload
7747  * @filter: Pointer to cloud filter structure
7748  *
7749  **/
7750 static int i40e_parse_cls_flower(struct i40e_vsi *vsi,
7751                                  struct flow_cls_offload *f,
7752                                  struct i40e_cloud_filter *filter)
7753 {
7754         struct flow_rule *rule = flow_cls_offload_flow_rule(f);
7755         struct flow_dissector *dissector = rule->match.dissector;
7756         u16 n_proto_mask = 0, n_proto_key = 0, addr_type = 0;
7757         struct i40e_pf *pf = vsi->back;
7758         u8 field_flags = 0;
7759
7760         if (dissector->used_keys &
7761             ~(BIT(FLOW_DISSECTOR_KEY_CONTROL) |
7762               BIT(FLOW_DISSECTOR_KEY_BASIC) |
7763               BIT(FLOW_DISSECTOR_KEY_ETH_ADDRS) |
7764               BIT(FLOW_DISSECTOR_KEY_VLAN) |
7765               BIT(FLOW_DISSECTOR_KEY_IPV4_ADDRS) |
7766               BIT(FLOW_DISSECTOR_KEY_IPV6_ADDRS) |
7767               BIT(FLOW_DISSECTOR_KEY_PORTS) |
7768               BIT(FLOW_DISSECTOR_KEY_ENC_KEYID))) {
7769                 dev_err(&pf->pdev->dev, "Unsupported key used: 0x%x\n",
7770                         dissector->used_keys);
7771                 return -EOPNOTSUPP;
7772         }
7773
7774         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
7775                 struct flow_match_enc_keyid match;
7776
7777                 flow_rule_match_enc_keyid(rule, &match);
7778                 if (match.mask->keyid != 0)
7779                         field_flags |= I40E_CLOUD_FIELD_TEN_ID;
7780
7781                 filter->tenant_id = be32_to_cpu(match.key->keyid);
7782         }
7783
7784         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
7785                 struct flow_match_basic match;
7786
7787                 flow_rule_match_basic(rule, &match);
7788                 n_proto_key = ntohs(match.key->n_proto);
7789                 n_proto_mask = ntohs(match.mask->n_proto);
7790
7791                 if (n_proto_key == ETH_P_ALL) {
7792                         n_proto_key = 0;
7793                         n_proto_mask = 0;
7794                 }
7795                 filter->n_proto = n_proto_key & n_proto_mask;
7796                 filter->ip_proto = match.key->ip_proto;
7797         }
7798
7799         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
7800                 struct flow_match_eth_addrs match;
7801
7802                 flow_rule_match_eth_addrs(rule, &match);
7803
7804                 /* use is_broadcast and is_zero to check for all 0xf or 0 */
7805                 if (!is_zero_ether_addr(match.mask->dst)) {
7806                         if (is_broadcast_ether_addr(match.mask->dst)) {
7807                                 field_flags |= I40E_CLOUD_FIELD_OMAC;
7808                         } else {
7809                                 dev_err(&pf->pdev->dev, "Bad ether dest mask %pM\n",
7810                                         match.mask->dst);
7811                                 return I40E_ERR_CONFIG;
7812                         }
7813                 }
7814
7815                 if (!is_zero_ether_addr(match.mask->src)) {
7816                         if (is_broadcast_ether_addr(match.mask->src)) {
7817                                 field_flags |= I40E_CLOUD_FIELD_IMAC;
7818                         } else {
7819                                 dev_err(&pf->pdev->dev, "Bad ether src mask %pM\n",
7820                                         match.mask->src);
7821                                 return I40E_ERR_CONFIG;
7822                         }
7823                 }
7824                 ether_addr_copy(filter->dst_mac, match.key->dst);
7825                 ether_addr_copy(filter->src_mac, match.key->src);
7826         }
7827
7828         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_VLAN)) {
7829                 struct flow_match_vlan match;
7830
7831                 flow_rule_match_vlan(rule, &match);
7832                 if (match.mask->vlan_id) {
7833                         if (match.mask->vlan_id == VLAN_VID_MASK) {
7834                                 field_flags |= I40E_CLOUD_FIELD_IVLAN;
7835
7836                         } else {
7837                                 dev_err(&pf->pdev->dev, "Bad vlan mask 0x%04x\n",
7838                                         match.mask->vlan_id);
7839                                 return I40E_ERR_CONFIG;
7840                         }
7841                 }
7842
7843                 filter->vlan_id = cpu_to_be16(match.key->vlan_id);
7844         }
7845
7846         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_CONTROL)) {
7847                 struct flow_match_control match;
7848
7849                 flow_rule_match_control(rule, &match);
7850                 addr_type = match.key->addr_type;
7851         }
7852
7853         if (addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
7854                 struct flow_match_ipv4_addrs match;
7855
7856                 flow_rule_match_ipv4_addrs(rule, &match);
7857                 if (match.mask->dst) {
7858                         if (match.mask->dst == cpu_to_be32(0xffffffff)) {
7859                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7860                         } else {
7861                                 dev_err(&pf->pdev->dev, "Bad ip dst mask %pI4b\n",
7862                                         &match.mask->dst);
7863                                 return I40E_ERR_CONFIG;
7864                         }
7865                 }
7866
7867                 if (match.mask->src) {
7868                         if (match.mask->src == cpu_to_be32(0xffffffff)) {
7869                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7870                         } else {
7871                                 dev_err(&pf->pdev->dev, "Bad ip src mask %pI4b\n",
7872                                         &match.mask->src);
7873                                 return I40E_ERR_CONFIG;
7874                         }
7875                 }
7876
7877                 if (field_flags & I40E_CLOUD_FIELD_TEN_ID) {
7878                         dev_err(&pf->pdev->dev, "Tenant id not allowed for ip filter\n");
7879                         return I40E_ERR_CONFIG;
7880                 }
7881                 filter->dst_ipv4 = match.key->dst;
7882                 filter->src_ipv4 = match.key->src;
7883         }
7884
7885         if (addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
7886                 struct flow_match_ipv6_addrs match;
7887
7888                 flow_rule_match_ipv6_addrs(rule, &match);
7889
7890                 /* src and dest IPV6 address should not be LOOPBACK
7891                  * (0:0:0:0:0:0:0:1), which can be represented as ::1
7892                  */
7893                 if (ipv6_addr_loopback(&match.key->dst) ||
7894                     ipv6_addr_loopback(&match.key->src)) {
7895                         dev_err(&pf->pdev->dev,
7896                                 "Bad ipv6, addr is LOOPBACK\n");
7897                         return I40E_ERR_CONFIG;
7898                 }
7899                 if (!ipv6_addr_any(&match.mask->dst) ||
7900                     !ipv6_addr_any(&match.mask->src))
7901                         field_flags |= I40E_CLOUD_FIELD_IIP;
7902
7903                 memcpy(&filter->src_ipv6, &match.key->src.s6_addr32,
7904                        sizeof(filter->src_ipv6));
7905                 memcpy(&filter->dst_ipv6, &match.key->dst.s6_addr32,
7906                        sizeof(filter->dst_ipv6));
7907         }
7908
7909         if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_PORTS)) {
7910                 struct flow_match_ports match;
7911
7912                 flow_rule_match_ports(rule, &match);
7913                 if (match.mask->src) {
7914                         if (match.mask->src == cpu_to_be16(0xffff)) {
7915                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7916                         } else {
7917                                 dev_err(&pf->pdev->dev, "Bad src port mask 0x%04x\n",
7918                                         be16_to_cpu(match.mask->src));
7919                                 return I40E_ERR_CONFIG;
7920                         }
7921                 }
7922
7923                 if (match.mask->dst) {
7924                         if (match.mask->dst == cpu_to_be16(0xffff)) {
7925                                 field_flags |= I40E_CLOUD_FIELD_IIP;
7926                         } else {
7927                                 dev_err(&pf->pdev->dev, "Bad dst port mask 0x%04x\n",
7928                                         be16_to_cpu(match.mask->dst));
7929                                 return I40E_ERR_CONFIG;
7930                         }
7931                 }
7932
7933                 filter->dst_port = match.key->dst;
7934                 filter->src_port = match.key->src;
7935
7936                 switch (filter->ip_proto) {
7937                 case IPPROTO_TCP:
7938                 case IPPROTO_UDP:
7939                         break;
7940                 default:
7941                         dev_err(&pf->pdev->dev,
7942                                 "Only UDP and TCP transport are supported\n");
7943                         return -EINVAL;
7944                 }
7945         }
7946         filter->flags = field_flags;
7947         return 0;
7948 }
7949
7950 /**
7951  * i40e_handle_tclass: Forward to a traffic class on the device
7952  * @vsi: Pointer to VSI
7953  * @tc: traffic class index on the device
7954  * @filter: Pointer to cloud filter structure
7955  *
7956  **/
7957 static int i40e_handle_tclass(struct i40e_vsi *vsi, u32 tc,
7958                               struct i40e_cloud_filter *filter)
7959 {
7960         struct i40e_channel *ch, *ch_tmp;
7961
7962         /* direct to a traffic class on the same device */
7963         if (tc == 0) {
7964                 filter->seid = vsi->seid;
7965                 return 0;
7966         } else if (vsi->tc_config.enabled_tc & BIT(tc)) {
7967                 if (!filter->dst_port) {
7968                         dev_err(&vsi->back->pdev->dev,
7969                                 "Specify destination port to direct to traffic class that is not default\n");
7970                         return -EINVAL;
7971                 }
7972                 if (list_empty(&vsi->ch_list))
7973                         return -EINVAL;
7974                 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list,
7975                                          list) {
7976                         if (ch->seid == vsi->tc_seid_map[tc])
7977                                 filter->seid = ch->seid;
7978                 }
7979                 return 0;
7980         }
7981         dev_err(&vsi->back->pdev->dev, "TC is not enabled\n");
7982         return -EINVAL;
7983 }
7984
7985 /**
7986  * i40e_configure_clsflower - Configure tc flower filters
7987  * @vsi: Pointer to VSI
7988  * @cls_flower: Pointer to struct flow_cls_offload
7989  *
7990  **/
7991 static int i40e_configure_clsflower(struct i40e_vsi *vsi,
7992                                     struct flow_cls_offload *cls_flower)
7993 {
7994         int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
7995         struct i40e_cloud_filter *filter = NULL;
7996         struct i40e_pf *pf = vsi->back;
7997         int err = 0;
7998
7999         if (tc < 0) {
8000                 dev_err(&vsi->back->pdev->dev, "Invalid traffic class\n");
8001                 return -EOPNOTSUPP;
8002         }
8003
8004         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
8005             test_bit(__I40E_RESET_INTR_RECEIVED, pf->state))
8006                 return -EBUSY;
8007
8008         if (pf->fdir_pf_active_filters ||
8009             (!hlist_empty(&pf->fdir_filter_list))) {
8010                 dev_err(&vsi->back->pdev->dev,
8011                         "Flow Director Sideband filters exists, turn ntuple off to configure cloud filters\n");
8012                 return -EINVAL;
8013         }
8014
8015         if (vsi->back->flags & I40E_FLAG_FD_SB_ENABLED) {
8016                 dev_err(&vsi->back->pdev->dev,
8017                         "Disable Flow Director Sideband, configuring Cloud filters via tc-flower\n");
8018                 vsi->back->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8019                 vsi->back->flags |= I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8020         }
8021
8022         filter = kzalloc(sizeof(*filter), GFP_KERNEL);
8023         if (!filter)
8024                 return -ENOMEM;
8025
8026         filter->cookie = cls_flower->cookie;
8027
8028         err = i40e_parse_cls_flower(vsi, cls_flower, filter);
8029         if (err < 0)
8030                 goto err;
8031
8032         err = i40e_handle_tclass(vsi, tc, filter);
8033         if (err < 0)
8034                 goto err;
8035
8036         /* Add cloud filter */
8037         if (filter->dst_port)
8038                 err = i40e_add_del_cloud_filter_big_buf(vsi, filter, true);
8039         else
8040                 err = i40e_add_del_cloud_filter(vsi, filter, true);
8041
8042         if (err) {
8043                 dev_err(&pf->pdev->dev,
8044                         "Failed to add cloud filter, err %s\n",
8045                         i40e_stat_str(&pf->hw, err));
8046                 goto err;
8047         }
8048
8049         /* add filter to the ordered list */
8050         INIT_HLIST_NODE(&filter->cloud_node);
8051
8052         hlist_add_head(&filter->cloud_node, &pf->cloud_filter_list);
8053
8054         pf->num_cloud_filters++;
8055
8056         return err;
8057 err:
8058         kfree(filter);
8059         return err;
8060 }
8061
8062 /**
8063  * i40e_find_cloud_filter - Find the could filter in the list
8064  * @vsi: Pointer to VSI
8065  * @cookie: filter specific cookie
8066  *
8067  **/
8068 static struct i40e_cloud_filter *i40e_find_cloud_filter(struct i40e_vsi *vsi,
8069                                                         unsigned long *cookie)
8070 {
8071         struct i40e_cloud_filter *filter = NULL;
8072         struct hlist_node *node2;
8073
8074         hlist_for_each_entry_safe(filter, node2,
8075                                   &vsi->back->cloud_filter_list, cloud_node)
8076                 if (!memcmp(cookie, &filter->cookie, sizeof(filter->cookie)))
8077                         return filter;
8078         return NULL;
8079 }
8080
8081 /**
8082  * i40e_delete_clsflower - Remove tc flower filters
8083  * @vsi: Pointer to VSI
8084  * @cls_flower: Pointer to struct flow_cls_offload
8085  *
8086  **/
8087 static int i40e_delete_clsflower(struct i40e_vsi *vsi,
8088                                  struct flow_cls_offload *cls_flower)
8089 {
8090         struct i40e_cloud_filter *filter = NULL;
8091         struct i40e_pf *pf = vsi->back;
8092         int err = 0;
8093
8094         filter = i40e_find_cloud_filter(vsi, &cls_flower->cookie);
8095
8096         if (!filter)
8097                 return -EINVAL;
8098
8099         hash_del(&filter->cloud_node);
8100
8101         if (filter->dst_port)
8102                 err = i40e_add_del_cloud_filter_big_buf(vsi, filter, false);
8103         else
8104                 err = i40e_add_del_cloud_filter(vsi, filter, false);
8105
8106         kfree(filter);
8107         if (err) {
8108                 dev_err(&pf->pdev->dev,
8109                         "Failed to delete cloud filter, err %s\n",
8110                         i40e_stat_str(&pf->hw, err));
8111                 return i40e_aq_rc_to_posix(err, pf->hw.aq.asq_last_status);
8112         }
8113
8114         pf->num_cloud_filters--;
8115         if (!pf->num_cloud_filters)
8116                 if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8117                     !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8118                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8119                         pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8120                         pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8121                 }
8122         return 0;
8123 }
8124
8125 /**
8126  * i40e_setup_tc_cls_flower - flower classifier offloads
8127  * @netdev: net device to configure
8128  * @type_data: offload data
8129  **/
8130 static int i40e_setup_tc_cls_flower(struct i40e_netdev_priv *np,
8131                                     struct flow_cls_offload *cls_flower)
8132 {
8133         struct i40e_vsi *vsi = np->vsi;
8134
8135         switch (cls_flower->command) {
8136         case FLOW_CLS_REPLACE:
8137                 return i40e_configure_clsflower(vsi, cls_flower);
8138         case FLOW_CLS_DESTROY:
8139                 return i40e_delete_clsflower(vsi, cls_flower);
8140         case FLOW_CLS_STATS:
8141                 return -EOPNOTSUPP;
8142         default:
8143                 return -EOPNOTSUPP;
8144         }
8145 }
8146
8147 static int i40e_setup_tc_block_cb(enum tc_setup_type type, void *type_data,
8148                                   void *cb_priv)
8149 {
8150         struct i40e_netdev_priv *np = cb_priv;
8151
8152         if (!tc_cls_can_offload_and_chain0(np->vsi->netdev, type_data))
8153                 return -EOPNOTSUPP;
8154
8155         switch (type) {
8156         case TC_SETUP_CLSFLOWER:
8157                 return i40e_setup_tc_cls_flower(np, type_data);
8158
8159         default:
8160                 return -EOPNOTSUPP;
8161         }
8162 }
8163
8164 static LIST_HEAD(i40e_block_cb_list);
8165
8166 static int __i40e_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8167                            void *type_data)
8168 {
8169         struct i40e_netdev_priv *np = netdev_priv(netdev);
8170
8171         switch (type) {
8172         case TC_SETUP_QDISC_MQPRIO:
8173                 return i40e_setup_tc(netdev, type_data);
8174         case TC_SETUP_BLOCK:
8175                 return flow_block_cb_setup_simple(type_data,
8176                                                   &i40e_block_cb_list,
8177                                                   i40e_setup_tc_block_cb,
8178                                                   np, np, true);
8179         default:
8180                 return -EOPNOTSUPP;
8181         }
8182 }
8183
8184 /**
8185  * i40e_open - Called when a network interface is made active
8186  * @netdev: network interface device structure
8187  *
8188  * The open entry point is called when a network interface is made
8189  * active by the system (IFF_UP).  At this point all resources needed
8190  * for transmit and receive operations are allocated, the interrupt
8191  * handler is registered with the OS, the netdev watchdog subtask is
8192  * enabled, and the stack is notified that the interface is ready.
8193  *
8194  * Returns 0 on success, negative value on failure
8195  **/
8196 int i40e_open(struct net_device *netdev)
8197 {
8198         struct i40e_netdev_priv *np = netdev_priv(netdev);
8199         struct i40e_vsi *vsi = np->vsi;
8200         struct i40e_pf *pf = vsi->back;
8201         int err;
8202
8203         /* disallow open during test or if eeprom is broken */
8204         if (test_bit(__I40E_TESTING, pf->state) ||
8205             test_bit(__I40E_BAD_EEPROM, pf->state))
8206                 return -EBUSY;
8207
8208         netif_carrier_off(netdev);
8209
8210         if (i40e_force_link_state(pf, true))
8211                 return -EAGAIN;
8212
8213         err = i40e_vsi_open(vsi);
8214         if (err)
8215                 return err;
8216
8217         /* configure global TSO hardware offload settings */
8218         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
8219                                                        TCP_FLAG_FIN) >> 16);
8220         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
8221                                                        TCP_FLAG_FIN |
8222                                                        TCP_FLAG_CWR) >> 16);
8223         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
8224
8225         udp_tunnel_get_rx_info(netdev);
8226
8227         return 0;
8228 }
8229
8230 /**
8231  * i40e_vsi_open -
8232  * @vsi: the VSI to open
8233  *
8234  * Finish initialization of the VSI.
8235  *
8236  * Returns 0 on success, negative value on failure
8237  *
8238  * Note: expects to be called while under rtnl_lock()
8239  **/
8240 int i40e_vsi_open(struct i40e_vsi *vsi)
8241 {
8242         struct i40e_pf *pf = vsi->back;
8243         char int_name[I40E_INT_NAME_STR_LEN];
8244         int err;
8245
8246         /* allocate descriptors */
8247         err = i40e_vsi_setup_tx_resources(vsi);
8248         if (err)
8249                 goto err_setup_tx;
8250         err = i40e_vsi_setup_rx_resources(vsi);
8251         if (err)
8252                 goto err_setup_rx;
8253
8254         err = i40e_vsi_configure(vsi);
8255         if (err)
8256                 goto err_setup_rx;
8257
8258         if (vsi->netdev) {
8259                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
8260                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
8261                 err = i40e_vsi_request_irq(vsi, int_name);
8262                 if (err)
8263                         goto err_setup_rx;
8264
8265                 /* Notify the stack of the actual queue counts. */
8266                 err = netif_set_real_num_tx_queues(vsi->netdev,
8267                                                    vsi->num_queue_pairs);
8268                 if (err)
8269                         goto err_set_queues;
8270
8271                 err = netif_set_real_num_rx_queues(vsi->netdev,
8272                                                    vsi->num_queue_pairs);
8273                 if (err)
8274                         goto err_set_queues;
8275
8276         } else if (vsi->type == I40E_VSI_FDIR) {
8277                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
8278                          dev_driver_string(&pf->pdev->dev),
8279                          dev_name(&pf->pdev->dev));
8280                 err = i40e_vsi_request_irq(vsi, int_name);
8281
8282         } else {
8283                 err = -EINVAL;
8284                 goto err_setup_rx;
8285         }
8286
8287         err = i40e_up_complete(vsi);
8288         if (err)
8289                 goto err_up_complete;
8290
8291         return 0;
8292
8293 err_up_complete:
8294         i40e_down(vsi);
8295 err_set_queues:
8296         i40e_vsi_free_irq(vsi);
8297 err_setup_rx:
8298         i40e_vsi_free_rx_resources(vsi);
8299 err_setup_tx:
8300         i40e_vsi_free_tx_resources(vsi);
8301         if (vsi == pf->vsi[pf->lan_vsi])
8302                 i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
8303
8304         return err;
8305 }
8306
8307 /**
8308  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
8309  * @pf: Pointer to PF
8310  *
8311  * This function destroys the hlist where all the Flow Director
8312  * filters were saved.
8313  **/
8314 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
8315 {
8316         struct i40e_fdir_filter *filter;
8317         struct i40e_flex_pit *pit_entry, *tmp;
8318         struct hlist_node *node2;
8319
8320         hlist_for_each_entry_safe(filter, node2,
8321                                   &pf->fdir_filter_list, fdir_node) {
8322                 hlist_del(&filter->fdir_node);
8323                 kfree(filter);
8324         }
8325
8326         list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
8327                 list_del(&pit_entry->list);
8328                 kfree(pit_entry);
8329         }
8330         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
8331
8332         list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
8333                 list_del(&pit_entry->list);
8334                 kfree(pit_entry);
8335         }
8336         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
8337
8338         pf->fdir_pf_active_filters = 0;
8339         pf->fd_tcp4_filter_cnt = 0;
8340         pf->fd_udp4_filter_cnt = 0;
8341         pf->fd_sctp4_filter_cnt = 0;
8342         pf->fd_ip4_filter_cnt = 0;
8343
8344         /* Reprogram the default input set for TCP/IPv4 */
8345         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8346                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8347                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8348
8349         /* Reprogram the default input set for UDP/IPv4 */
8350         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
8351                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8352                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8353
8354         /* Reprogram the default input set for SCTP/IPv4 */
8355         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
8356                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8357                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8358
8359         /* Reprogram the default input set for Other/IPv4 */
8360         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
8361                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8362
8363         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_FRAG_IPV4,
8364                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
8365 }
8366
8367 /**
8368  * i40e_cloud_filter_exit - Cleans up the cloud filters
8369  * @pf: Pointer to PF
8370  *
8371  * This function destroys the hlist where all the cloud filters
8372  * were saved.
8373  **/
8374 static void i40e_cloud_filter_exit(struct i40e_pf *pf)
8375 {
8376         struct i40e_cloud_filter *cfilter;
8377         struct hlist_node *node;
8378
8379         hlist_for_each_entry_safe(cfilter, node,
8380                                   &pf->cloud_filter_list, cloud_node) {
8381                 hlist_del(&cfilter->cloud_node);
8382                 kfree(cfilter);
8383         }
8384         pf->num_cloud_filters = 0;
8385
8386         if ((pf->flags & I40E_FLAG_FD_SB_TO_CLOUD_FILTER) &&
8387             !(pf->flags & I40E_FLAG_FD_SB_INACTIVE)) {
8388                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8389                 pf->flags &= ~I40E_FLAG_FD_SB_TO_CLOUD_FILTER;
8390                 pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
8391         }
8392 }
8393
8394 /**
8395  * i40e_close - Disables a network interface
8396  * @netdev: network interface device structure
8397  *
8398  * The close entry point is called when an interface is de-activated
8399  * by the OS.  The hardware is still under the driver's control, but
8400  * this netdev interface is disabled.
8401  *
8402  * Returns 0, this is not allowed to fail
8403  **/
8404 int i40e_close(struct net_device *netdev)
8405 {
8406         struct i40e_netdev_priv *np = netdev_priv(netdev);
8407         struct i40e_vsi *vsi = np->vsi;
8408
8409         i40e_vsi_close(vsi);
8410
8411         return 0;
8412 }
8413
8414 /**
8415  * i40e_do_reset - Start a PF or Core Reset sequence
8416  * @pf: board private structure
8417  * @reset_flags: which reset is requested
8418  * @lock_acquired: indicates whether or not the lock has been acquired
8419  * before this function was called.
8420  *
8421  * The essential difference in resets is that the PF Reset
8422  * doesn't clear the packet buffers, doesn't reset the PE
8423  * firmware, and doesn't bother the other PFs on the chip.
8424  **/
8425 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
8426 {
8427         u32 val;
8428
8429         WARN_ON(in_interrupt());
8430
8431
8432         /* do the biggest reset indicated */
8433         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
8434
8435                 /* Request a Global Reset
8436                  *
8437                  * This will start the chip's countdown to the actual full
8438                  * chip reset event, and a warning interrupt to be sent
8439                  * to all PFs, including the requestor.  Our handler
8440                  * for the warning interrupt will deal with the shutdown
8441                  * and recovery of the switch setup.
8442                  */
8443                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
8444                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8445                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
8446                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8447
8448         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
8449
8450                 /* Request a Core Reset
8451                  *
8452                  * Same as Global Reset, except does *not* include the MAC/PHY
8453                  */
8454                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
8455                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8456                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
8457                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
8458                 i40e_flush(&pf->hw);
8459
8460         } else if (reset_flags & I40E_PF_RESET_FLAG) {
8461
8462                 /* Request a PF Reset
8463                  *
8464                  * Resets only the PF-specific registers
8465                  *
8466                  * This goes directly to the tear-down and rebuild of
8467                  * the switch, since we need to do all the recovery as
8468                  * for the Core Reset.
8469                  */
8470                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
8471                 i40e_handle_reset_warning(pf, lock_acquired);
8472
8473                 dev_info(&pf->pdev->dev,
8474                          pf->flags & I40E_FLAG_DISABLE_FW_LLDP ?
8475                          "FW LLDP is disabled\n" :
8476                          "FW LLDP is enabled\n");
8477
8478         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
8479                 int v;
8480
8481                 /* Find the VSI(s) that requested a re-init */
8482                 dev_info(&pf->pdev->dev,
8483                          "VSI reinit requested\n");
8484                 for (v = 0; v < pf->num_alloc_vsi; v++) {
8485                         struct i40e_vsi *vsi = pf->vsi[v];
8486
8487                         if (vsi != NULL &&
8488                             test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
8489                                                vsi->state))
8490                                 i40e_vsi_reinit_locked(pf->vsi[v]);
8491                 }
8492         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
8493                 int v;
8494
8495                 /* Find the VSI(s) that needs to be brought down */
8496                 dev_info(&pf->pdev->dev, "VSI down requested\n");
8497                 for (v = 0; v < pf->num_alloc_vsi; v++) {
8498                         struct i40e_vsi *vsi = pf->vsi[v];
8499
8500                         if (vsi != NULL &&
8501                             test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
8502                                                vsi->state)) {
8503                                 set_bit(__I40E_VSI_DOWN, vsi->state);
8504                                 i40e_down(vsi);
8505                         }
8506                 }
8507         } else {
8508                 dev_info(&pf->pdev->dev,
8509                          "bad reset request 0x%08x\n", reset_flags);
8510         }
8511 }
8512
8513 #ifdef CONFIG_I40E_DCB
8514 /**
8515  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
8516  * @pf: board private structure
8517  * @old_cfg: current DCB config
8518  * @new_cfg: new DCB config
8519  **/
8520 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
8521                             struct i40e_dcbx_config *old_cfg,
8522                             struct i40e_dcbx_config *new_cfg)
8523 {
8524         bool need_reconfig = false;
8525
8526         /* Check if ETS configuration has changed */
8527         if (memcmp(&new_cfg->etscfg,
8528                    &old_cfg->etscfg,
8529                    sizeof(new_cfg->etscfg))) {
8530                 /* If Priority Table has changed reconfig is needed */
8531                 if (memcmp(&new_cfg->etscfg.prioritytable,
8532                            &old_cfg->etscfg.prioritytable,
8533                            sizeof(new_cfg->etscfg.prioritytable))) {
8534                         need_reconfig = true;
8535                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
8536                 }
8537
8538                 if (memcmp(&new_cfg->etscfg.tcbwtable,
8539                            &old_cfg->etscfg.tcbwtable,
8540                            sizeof(new_cfg->etscfg.tcbwtable)))
8541                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
8542
8543                 if (memcmp(&new_cfg->etscfg.tsatable,
8544                            &old_cfg->etscfg.tsatable,
8545                            sizeof(new_cfg->etscfg.tsatable)))
8546                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
8547         }
8548
8549         /* Check if PFC configuration has changed */
8550         if (memcmp(&new_cfg->pfc,
8551                    &old_cfg->pfc,
8552                    sizeof(new_cfg->pfc))) {
8553                 need_reconfig = true;
8554                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
8555         }
8556
8557         /* Check if APP Table has changed */
8558         if (memcmp(&new_cfg->app,
8559                    &old_cfg->app,
8560                    sizeof(new_cfg->app))) {
8561                 need_reconfig = true;
8562                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
8563         }
8564
8565         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
8566         return need_reconfig;
8567 }
8568
8569 /**
8570  * i40e_handle_lldp_event - Handle LLDP Change MIB event
8571  * @pf: board private structure
8572  * @e: event info posted on ARQ
8573  **/
8574 static int i40e_handle_lldp_event(struct i40e_pf *pf,
8575                                   struct i40e_arq_event_info *e)
8576 {
8577         struct i40e_aqc_lldp_get_mib *mib =
8578                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
8579         struct i40e_hw *hw = &pf->hw;
8580         struct i40e_dcbx_config tmp_dcbx_cfg;
8581         bool need_reconfig = false;
8582         int ret = 0;
8583         u8 type;
8584
8585         /* Not DCB capable or capability disabled */
8586         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
8587                 return ret;
8588
8589         /* Ignore if event is not for Nearest Bridge */
8590         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
8591                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
8592         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
8593         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
8594                 return ret;
8595
8596         /* Check MIB Type and return if event for Remote MIB update */
8597         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
8598         dev_dbg(&pf->pdev->dev,
8599                 "LLDP event mib type %s\n", type ? "remote" : "local");
8600         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
8601                 /* Update the remote cached instance and return */
8602                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
8603                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
8604                                 &hw->remote_dcbx_config);
8605                 goto exit;
8606         }
8607
8608         /* Store the old configuration */
8609         tmp_dcbx_cfg = hw->local_dcbx_config;
8610
8611         /* Reset the old DCBx configuration data */
8612         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
8613         /* Get updated DCBX data from firmware */
8614         ret = i40e_get_dcb_config(&pf->hw);
8615         if (ret) {
8616                 dev_info(&pf->pdev->dev,
8617                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
8618                          i40e_stat_str(&pf->hw, ret),
8619                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8620                 goto exit;
8621         }
8622
8623         /* No change detected in DCBX configs */
8624         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
8625                     sizeof(tmp_dcbx_cfg))) {
8626                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
8627                 goto exit;
8628         }
8629
8630         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
8631                                                &hw->local_dcbx_config);
8632
8633         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
8634
8635         if (!need_reconfig)
8636                 goto exit;
8637
8638         /* Enable DCB tagging only when more than one TC */
8639         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
8640                 pf->flags |= I40E_FLAG_DCB_ENABLED;
8641         else
8642                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8643
8644         set_bit(__I40E_PORT_SUSPENDED, pf->state);
8645         /* Reconfiguration needed quiesce all VSIs */
8646         i40e_pf_quiesce_all_vsi(pf);
8647
8648         /* Changes in configuration update VEB/VSI */
8649         i40e_dcb_reconfigure(pf);
8650
8651         ret = i40e_resume_port_tx(pf);
8652
8653         clear_bit(__I40E_PORT_SUSPENDED, pf->state);
8654         /* In case of error no point in resuming VSIs */
8655         if (ret)
8656                 goto exit;
8657
8658         /* Wait for the PF's queues to be disabled */
8659         ret = i40e_pf_wait_queues_disabled(pf);
8660         if (ret) {
8661                 /* Schedule PF reset to recover */
8662                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
8663                 i40e_service_event_schedule(pf);
8664         } else {
8665                 i40e_pf_unquiesce_all_vsi(pf);
8666                 set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
8667                 set_bit(__I40E_CLIENT_L2_CHANGE, pf->state);
8668         }
8669
8670 exit:
8671         return ret;
8672 }
8673 #endif /* CONFIG_I40E_DCB */
8674
8675 /**
8676  * i40e_do_reset_safe - Protected reset path for userland calls.
8677  * @pf: board private structure
8678  * @reset_flags: which reset is requested
8679  *
8680  **/
8681 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
8682 {
8683         rtnl_lock();
8684         i40e_do_reset(pf, reset_flags, true);
8685         rtnl_unlock();
8686 }
8687
8688 /**
8689  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
8690  * @pf: board private structure
8691  * @e: event info posted on ARQ
8692  *
8693  * Handler for LAN Queue Overflow Event generated by the firmware for PF
8694  * and VF queues
8695  **/
8696 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
8697                                            struct i40e_arq_event_info *e)
8698 {
8699         struct i40e_aqc_lan_overflow *data =
8700                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
8701         u32 queue = le32_to_cpu(data->prtdcb_rupto);
8702         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
8703         struct i40e_hw *hw = &pf->hw;
8704         struct i40e_vf *vf;
8705         u16 vf_id;
8706
8707         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
8708                 queue, qtx_ctl);
8709
8710         /* Queue belongs to VF, find the VF and issue VF reset */
8711         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
8712             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
8713                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
8714                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
8715                 vf_id -= hw->func_caps.vf_base_id;
8716                 vf = &pf->vf[vf_id];
8717                 i40e_vc_notify_vf_reset(vf);
8718                 /* Allow VF to process pending reset notification */
8719                 msleep(20);
8720                 i40e_reset_vf(vf, false);
8721         }
8722 }
8723
8724 /**
8725  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
8726  * @pf: board private structure
8727  **/
8728 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
8729 {
8730         u32 val, fcnt_prog;
8731
8732         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8733         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
8734         return fcnt_prog;
8735 }
8736
8737 /**
8738  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
8739  * @pf: board private structure
8740  **/
8741 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
8742 {
8743         u32 val, fcnt_prog;
8744
8745         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
8746         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
8747                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
8748                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
8749         return fcnt_prog;
8750 }
8751
8752 /**
8753  * i40e_get_global_fd_count - Get total FD filters programmed on device
8754  * @pf: board private structure
8755  **/
8756 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
8757 {
8758         u32 val, fcnt_prog;
8759
8760         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
8761         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
8762                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
8763                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
8764         return fcnt_prog;
8765 }
8766
8767 /**
8768  * i40e_reenable_fdir_sb - Restore FDir SB capability
8769  * @pf: board private structure
8770  **/
8771 static void i40e_reenable_fdir_sb(struct i40e_pf *pf)
8772 {
8773         if (test_and_clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state))
8774                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
8775                     (I40E_DEBUG_FD & pf->hw.debug_mask))
8776                         dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
8777 }
8778
8779 /**
8780  * i40e_reenable_fdir_atr - Restore FDir ATR capability
8781  * @pf: board private structure
8782  **/
8783 static void i40e_reenable_fdir_atr(struct i40e_pf *pf)
8784 {
8785         if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state)) {
8786                 /* ATR uses the same filtering logic as SB rules. It only
8787                  * functions properly if the input set mask is at the default
8788                  * settings. It is safe to restore the default input set
8789                  * because there are no active TCPv4 filter rules.
8790                  */
8791                 i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
8792                                         I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
8793                                         I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
8794
8795                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8796                     (I40E_DEBUG_FD & pf->hw.debug_mask))
8797                         dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
8798         }
8799 }
8800
8801 /**
8802  * i40e_delete_invalid_filter - Delete an invalid FDIR filter
8803  * @pf: board private structure
8804  * @filter: FDir filter to remove
8805  */
8806 static void i40e_delete_invalid_filter(struct i40e_pf *pf,
8807                                        struct i40e_fdir_filter *filter)
8808 {
8809         /* Update counters */
8810         pf->fdir_pf_active_filters--;
8811         pf->fd_inv = 0;
8812
8813         switch (filter->flow_type) {
8814         case TCP_V4_FLOW:
8815                 pf->fd_tcp4_filter_cnt--;
8816                 break;
8817         case UDP_V4_FLOW:
8818                 pf->fd_udp4_filter_cnt--;
8819                 break;
8820         case SCTP_V4_FLOW:
8821                 pf->fd_sctp4_filter_cnt--;
8822                 break;
8823         case IP_USER_FLOW:
8824                 switch (filter->ip4_proto) {
8825                 case IPPROTO_TCP:
8826                         pf->fd_tcp4_filter_cnt--;
8827                         break;
8828                 case IPPROTO_UDP:
8829                         pf->fd_udp4_filter_cnt--;
8830                         break;
8831                 case IPPROTO_SCTP:
8832                         pf->fd_sctp4_filter_cnt--;
8833                         break;
8834                 case IPPROTO_IP:
8835                         pf->fd_ip4_filter_cnt--;
8836                         break;
8837                 }
8838                 break;
8839         }
8840
8841         /* Remove the filter from the list and free memory */
8842         hlist_del(&filter->fdir_node);
8843         kfree(filter);
8844 }
8845
8846 /**
8847  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
8848  * @pf: board private structure
8849  **/
8850 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
8851 {
8852         struct i40e_fdir_filter *filter;
8853         u32 fcnt_prog, fcnt_avail;
8854         struct hlist_node *node;
8855
8856         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8857                 return;
8858
8859         /* Check if we have enough room to re-enable FDir SB capability. */
8860         fcnt_prog = i40e_get_global_fd_count(pf);
8861         fcnt_avail = pf->fdir_pf_filter_count;
8862         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
8863             (pf->fd_add_err == 0) ||
8864             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt))
8865                 i40e_reenable_fdir_sb(pf);
8866
8867         /* We should wait for even more space before re-enabling ATR.
8868          * Additionally, we cannot enable ATR as long as we still have TCP SB
8869          * rules active.
8870          */
8871         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
8872             (pf->fd_tcp4_filter_cnt == 0))
8873                 i40e_reenable_fdir_atr(pf);
8874
8875         /* if hw had a problem adding a filter, delete it */
8876         if (pf->fd_inv > 0) {
8877                 hlist_for_each_entry_safe(filter, node,
8878                                           &pf->fdir_filter_list, fdir_node)
8879                         if (filter->fd_id == pf->fd_inv)
8880                                 i40e_delete_invalid_filter(pf, filter);
8881         }
8882 }
8883
8884 #define I40E_MIN_FD_FLUSH_INTERVAL 10
8885 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
8886 /**
8887  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
8888  * @pf: board private structure
8889  **/
8890 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
8891 {
8892         unsigned long min_flush_time;
8893         int flush_wait_retry = 50;
8894         bool disable_atr = false;
8895         int fd_room;
8896         int reg;
8897
8898         if (!time_after(jiffies, pf->fd_flush_timestamp +
8899                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
8900                 return;
8901
8902         /* If the flush is happening too quick and we have mostly SB rules we
8903          * should not re-enable ATR for some time.
8904          */
8905         min_flush_time = pf->fd_flush_timestamp +
8906                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
8907         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
8908
8909         if (!(time_after(jiffies, min_flush_time)) &&
8910             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
8911                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8912                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
8913                 disable_atr = true;
8914         }
8915
8916         pf->fd_flush_timestamp = jiffies;
8917         set_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8918         /* flush all filters */
8919         wr32(&pf->hw, I40E_PFQF_CTL_1,
8920              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
8921         i40e_flush(&pf->hw);
8922         pf->fd_flush_cnt++;
8923         pf->fd_add_err = 0;
8924         do {
8925                 /* Check FD flush status every 5-6msec */
8926                 usleep_range(5000, 6000);
8927                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
8928                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
8929                         break;
8930         } while (flush_wait_retry--);
8931         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
8932                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
8933         } else {
8934                 /* replay sideband filters */
8935                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
8936                 if (!disable_atr && !pf->fd_tcp4_filter_cnt)
8937                         clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state);
8938                 clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
8939                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8940                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
8941         }
8942 }
8943
8944 /**
8945  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
8946  * @pf: board private structure
8947  **/
8948 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
8949 {
8950         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
8951 }
8952
8953 /* We can see up to 256 filter programming desc in transit if the filters are
8954  * being applied really fast; before we see the first
8955  * filter miss error on Rx queue 0. Accumulating enough error messages before
8956  * reacting will make sure we don't cause flush too often.
8957  */
8958 #define I40E_MAX_FD_PROGRAM_ERROR 256
8959
8960 /**
8961  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
8962  * @pf: board private structure
8963  **/
8964 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
8965 {
8966
8967         /* if interface is down do nothing */
8968         if (test_bit(__I40E_DOWN, pf->state))
8969                 return;
8970
8971         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
8972                 i40e_fdir_flush_and_replay(pf);
8973
8974         i40e_fdir_check_and_reenable(pf);
8975
8976 }
8977
8978 /**
8979  * i40e_vsi_link_event - notify VSI of a link event
8980  * @vsi: vsi to be notified
8981  * @link_up: link up or down
8982  **/
8983 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
8984 {
8985         if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
8986                 return;
8987
8988         switch (vsi->type) {
8989         case I40E_VSI_MAIN:
8990                 if (!vsi->netdev || !vsi->netdev_registered)
8991                         break;
8992
8993                 if (link_up) {
8994                         netif_carrier_on(vsi->netdev);
8995                         netif_tx_wake_all_queues(vsi->netdev);
8996                 } else {
8997                         netif_carrier_off(vsi->netdev);
8998                         netif_tx_stop_all_queues(vsi->netdev);
8999                 }
9000                 break;
9001
9002         case I40E_VSI_SRIOV:
9003         case I40E_VSI_VMDQ2:
9004         case I40E_VSI_CTRL:
9005         case I40E_VSI_IWARP:
9006         case I40E_VSI_MIRROR:
9007         default:
9008                 /* there is no notification for other VSIs */
9009                 break;
9010         }
9011 }
9012
9013 /**
9014  * i40e_veb_link_event - notify elements on the veb of a link event
9015  * @veb: veb to be notified
9016  * @link_up: link up or down
9017  **/
9018 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
9019 {
9020         struct i40e_pf *pf;
9021         int i;
9022
9023         if (!veb || !veb->pf)
9024                 return;
9025         pf = veb->pf;
9026
9027         /* depth first... */
9028         for (i = 0; i < I40E_MAX_VEB; i++)
9029                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
9030                         i40e_veb_link_event(pf->veb[i], link_up);
9031
9032         /* ... now the local VSIs */
9033         for (i = 0; i < pf->num_alloc_vsi; i++)
9034                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
9035                         i40e_vsi_link_event(pf->vsi[i], link_up);
9036 }
9037
9038 /**
9039  * i40e_link_event - Update netif_carrier status
9040  * @pf: board private structure
9041  **/
9042 static void i40e_link_event(struct i40e_pf *pf)
9043 {
9044         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9045         u8 new_link_speed, old_link_speed;
9046         i40e_status status;
9047         bool new_link, old_link;
9048
9049         /* set this to force the get_link_status call to refresh state */
9050         pf->hw.phy.get_link_info = true;
9051         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
9052         status = i40e_get_link_status(&pf->hw, &new_link);
9053
9054         /* On success, disable temp link polling */
9055         if (status == I40E_SUCCESS) {
9056                 clear_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9057         } else {
9058                 /* Enable link polling temporarily until i40e_get_link_status
9059                  * returns I40E_SUCCESS
9060                  */
9061                 set_bit(__I40E_TEMP_LINK_POLLING, pf->state);
9062                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
9063                         status);
9064                 return;
9065         }
9066
9067         old_link_speed = pf->hw.phy.link_info_old.link_speed;
9068         new_link_speed = pf->hw.phy.link_info.link_speed;
9069
9070         if (new_link == old_link &&
9071             new_link_speed == old_link_speed &&
9072             (test_bit(__I40E_VSI_DOWN, vsi->state) ||
9073              new_link == netif_carrier_ok(vsi->netdev)))
9074                 return;
9075
9076         i40e_print_link_message(vsi, new_link);
9077
9078         /* Notify the base of the switch tree connected to
9079          * the link.  Floating VEBs are not notified.
9080          */
9081         if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
9082                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
9083         else
9084                 i40e_vsi_link_event(vsi, new_link);
9085
9086         if (pf->vf)
9087                 i40e_vc_notify_link_state(pf);
9088
9089         if (pf->flags & I40E_FLAG_PTP)
9090                 i40e_ptp_set_increment(pf);
9091 }
9092
9093 /**
9094  * i40e_watchdog_subtask - periodic checks not using event driven response
9095  * @pf: board private structure
9096  **/
9097 static void i40e_watchdog_subtask(struct i40e_pf *pf)
9098 {
9099         int i;
9100
9101         /* if interface is down do nothing */
9102         if (test_bit(__I40E_DOWN, pf->state) ||
9103             test_bit(__I40E_CONFIG_BUSY, pf->state))
9104                 return;
9105
9106         /* make sure we don't do these things too often */
9107         if (time_before(jiffies, (pf->service_timer_previous +
9108                                   pf->service_timer_period)))
9109                 return;
9110         pf->service_timer_previous = jiffies;
9111
9112         if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
9113             test_bit(__I40E_TEMP_LINK_POLLING, pf->state))
9114                 i40e_link_event(pf);
9115
9116         /* Update the stats for active netdevs so the network stack
9117          * can look at updated numbers whenever it cares to
9118          */
9119         for (i = 0; i < pf->num_alloc_vsi; i++)
9120                 if (pf->vsi[i] && pf->vsi[i]->netdev)
9121                         i40e_update_stats(pf->vsi[i]);
9122
9123         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
9124                 /* Update the stats for the active switching components */
9125                 for (i = 0; i < I40E_MAX_VEB; i++)
9126                         if (pf->veb[i])
9127                                 i40e_update_veb_stats(pf->veb[i]);
9128         }
9129
9130         i40e_ptp_rx_hang(pf);
9131         i40e_ptp_tx_hang(pf);
9132 }
9133
9134 /**
9135  * i40e_reset_subtask - Set up for resetting the device and driver
9136  * @pf: board private structure
9137  **/
9138 static void i40e_reset_subtask(struct i40e_pf *pf)
9139 {
9140         u32 reset_flags = 0;
9141
9142         if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
9143                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
9144                 clear_bit(__I40E_REINIT_REQUESTED, pf->state);
9145         }
9146         if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
9147                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
9148                 clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
9149         }
9150         if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
9151                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
9152                 clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
9153         }
9154         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
9155                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
9156                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
9157         }
9158         if (test_bit(__I40E_DOWN_REQUESTED, pf->state)) {
9159                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
9160                 clear_bit(__I40E_DOWN_REQUESTED, pf->state);
9161         }
9162
9163         /* If there's a recovery already waiting, it takes
9164          * precedence before starting a new reset sequence.
9165          */
9166         if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
9167                 i40e_prep_for_reset(pf, false);
9168                 i40e_reset(pf);
9169                 i40e_rebuild(pf, false, false);
9170         }
9171
9172         /* If we're already down or resetting, just bail */
9173         if (reset_flags &&
9174             !test_bit(__I40E_DOWN, pf->state) &&
9175             !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
9176                 i40e_do_reset(pf, reset_flags, false);
9177         }
9178 }
9179
9180 /**
9181  * i40e_handle_link_event - Handle link event
9182  * @pf: board private structure
9183  * @e: event info posted on ARQ
9184  **/
9185 static void i40e_handle_link_event(struct i40e_pf *pf,
9186                                    struct i40e_arq_event_info *e)
9187 {
9188         struct i40e_aqc_get_link_status *status =
9189                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
9190
9191         /* Do a new status request to re-enable LSE reporting
9192          * and load new status information into the hw struct
9193          * This completely ignores any state information
9194          * in the ARQ event info, instead choosing to always
9195          * issue the AQ update link status command.
9196          */
9197         i40e_link_event(pf);
9198
9199         /* Check if module meets thermal requirements */
9200         if (status->phy_type == I40E_PHY_TYPE_NOT_SUPPORTED_HIGH_TEMP) {
9201                 dev_err(&pf->pdev->dev,
9202                         "Rx/Tx is disabled on this device because the module does not meet thermal requirements.\n");
9203                 dev_err(&pf->pdev->dev,
9204                         "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9205         } else {
9206                 /* check for unqualified module, if link is down, suppress
9207                  * the message if link was forced to be down.
9208                  */
9209                 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
9210                     (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
9211                     (!(status->link_info & I40E_AQ_LINK_UP)) &&
9212                     (!(pf->flags & I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED))) {
9213                         dev_err(&pf->pdev->dev,
9214                                 "Rx/Tx is disabled on this device because an unsupported SFP module type was detected.\n");
9215                         dev_err(&pf->pdev->dev,
9216                                 "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
9217                 }
9218         }
9219 }
9220
9221 /**
9222  * i40e_clean_adminq_subtask - Clean the AdminQ rings
9223  * @pf: board private structure
9224  **/
9225 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
9226 {
9227         struct i40e_arq_event_info event;
9228         struct i40e_hw *hw = &pf->hw;
9229         u16 pending, i = 0;
9230         i40e_status ret;
9231         u16 opcode;
9232         u32 oldval;
9233         u32 val;
9234
9235         /* Do not run clean AQ when PF reset fails */
9236         if (test_bit(__I40E_RESET_FAILED, pf->state))
9237                 return;
9238
9239         /* check for error indications */
9240         val = rd32(&pf->hw, pf->hw.aq.arq.len);
9241         oldval = val;
9242         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
9243                 if (hw->debug_mask & I40E_DEBUG_AQ)
9244                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
9245                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
9246         }
9247         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
9248                 if (hw->debug_mask & I40E_DEBUG_AQ)
9249                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
9250                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
9251                 pf->arq_overflows++;
9252         }
9253         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
9254                 if (hw->debug_mask & I40E_DEBUG_AQ)
9255                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
9256                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
9257         }
9258         if (oldval != val)
9259                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
9260
9261         val = rd32(&pf->hw, pf->hw.aq.asq.len);
9262         oldval = val;
9263         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
9264                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9265                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
9266                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
9267         }
9268         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
9269                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9270                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
9271                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
9272         }
9273         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
9274                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
9275                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
9276                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
9277         }
9278         if (oldval != val)
9279                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
9280
9281         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
9282         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
9283         if (!event.msg_buf)
9284                 return;
9285
9286         do {
9287                 ret = i40e_clean_arq_element(hw, &event, &pending);
9288                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
9289                         break;
9290                 else if (ret) {
9291                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
9292                         break;
9293                 }
9294
9295                 opcode = le16_to_cpu(event.desc.opcode);
9296                 switch (opcode) {
9297
9298                 case i40e_aqc_opc_get_link_status:
9299                         i40e_handle_link_event(pf, &event);
9300                         break;
9301                 case i40e_aqc_opc_send_msg_to_pf:
9302                         ret = i40e_vc_process_vf_msg(pf,
9303                                         le16_to_cpu(event.desc.retval),
9304                                         le32_to_cpu(event.desc.cookie_high),
9305                                         le32_to_cpu(event.desc.cookie_low),
9306                                         event.msg_buf,
9307                                         event.msg_len);
9308                         break;
9309                 case i40e_aqc_opc_lldp_update_mib:
9310                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
9311 #ifdef CONFIG_I40E_DCB
9312                         rtnl_lock();
9313                         ret = i40e_handle_lldp_event(pf, &event);
9314                         rtnl_unlock();
9315 #endif /* CONFIG_I40E_DCB */
9316                         break;
9317                 case i40e_aqc_opc_event_lan_overflow:
9318                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
9319                         i40e_handle_lan_overflow_event(pf, &event);
9320                         break;
9321                 case i40e_aqc_opc_send_msg_to_peer:
9322                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
9323                         break;
9324                 case i40e_aqc_opc_nvm_erase:
9325                 case i40e_aqc_opc_nvm_update:
9326                 case i40e_aqc_opc_oem_post_update:
9327                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
9328                                    "ARQ NVM operation 0x%04x completed\n",
9329                                    opcode);
9330                         break;
9331                 default:
9332                         dev_info(&pf->pdev->dev,
9333                                  "ARQ: Unknown event 0x%04x ignored\n",
9334                                  opcode);
9335                         break;
9336                 }
9337         } while (i++ < pf->adminq_work_limit);
9338
9339         if (i < pf->adminq_work_limit)
9340                 clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
9341
9342         /* re-enable Admin queue interrupt cause */
9343         val = rd32(hw, I40E_PFINT_ICR0_ENA);
9344         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
9345         wr32(hw, I40E_PFINT_ICR0_ENA, val);
9346         i40e_flush(hw);
9347
9348         kfree(event.msg_buf);
9349 }
9350
9351 /**
9352  * i40e_verify_eeprom - make sure eeprom is good to use
9353  * @pf: board private structure
9354  **/
9355 static void i40e_verify_eeprom(struct i40e_pf *pf)
9356 {
9357         int err;
9358
9359         err = i40e_diag_eeprom_test(&pf->hw);
9360         if (err) {
9361                 /* retry in case of garbage read */
9362                 err = i40e_diag_eeprom_test(&pf->hw);
9363                 if (err) {
9364                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
9365                                  err);
9366                         set_bit(__I40E_BAD_EEPROM, pf->state);
9367                 }
9368         }
9369
9370         if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
9371                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
9372                 clear_bit(__I40E_BAD_EEPROM, pf->state);
9373         }
9374 }
9375
9376 /**
9377  * i40e_enable_pf_switch_lb
9378  * @pf: pointer to the PF structure
9379  *
9380  * enable switch loop back or die - no point in a return value
9381  **/
9382 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
9383 {
9384         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9385         struct i40e_vsi_context ctxt;
9386         int ret;
9387
9388         ctxt.seid = pf->main_vsi_seid;
9389         ctxt.pf_num = pf->hw.pf_id;
9390         ctxt.vf_num = 0;
9391         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9392         if (ret) {
9393                 dev_info(&pf->pdev->dev,
9394                          "couldn't get PF vsi config, err %s aq_err %s\n",
9395                          i40e_stat_str(&pf->hw, ret),
9396                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9397                 return;
9398         }
9399         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9400         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9401         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9402
9403         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9404         if (ret) {
9405                 dev_info(&pf->pdev->dev,
9406                          "update vsi switch failed, err %s aq_err %s\n",
9407                          i40e_stat_str(&pf->hw, ret),
9408                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9409         }
9410 }
9411
9412 /**
9413  * i40e_disable_pf_switch_lb
9414  * @pf: pointer to the PF structure
9415  *
9416  * disable switch loop back or die - no point in a return value
9417  **/
9418 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
9419 {
9420         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9421         struct i40e_vsi_context ctxt;
9422         int ret;
9423
9424         ctxt.seid = pf->main_vsi_seid;
9425         ctxt.pf_num = pf->hw.pf_id;
9426         ctxt.vf_num = 0;
9427         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9428         if (ret) {
9429                 dev_info(&pf->pdev->dev,
9430                          "couldn't get PF vsi config, err %s aq_err %s\n",
9431                          i40e_stat_str(&pf->hw, ret),
9432                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9433                 return;
9434         }
9435         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9436         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9437         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9438
9439         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
9440         if (ret) {
9441                 dev_info(&pf->pdev->dev,
9442                          "update vsi switch failed, err %s aq_err %s\n",
9443                          i40e_stat_str(&pf->hw, ret),
9444                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9445         }
9446 }
9447
9448 /**
9449  * i40e_config_bridge_mode - Configure the HW bridge mode
9450  * @veb: pointer to the bridge instance
9451  *
9452  * Configure the loop back mode for the LAN VSI that is downlink to the
9453  * specified HW bridge instance. It is expected this function is called
9454  * when a new HW bridge is instantiated.
9455  **/
9456 static void i40e_config_bridge_mode(struct i40e_veb *veb)
9457 {
9458         struct i40e_pf *pf = veb->pf;
9459
9460         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
9461                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
9462                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
9463         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
9464                 i40e_disable_pf_switch_lb(pf);
9465         else
9466                 i40e_enable_pf_switch_lb(pf);
9467 }
9468
9469 /**
9470  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
9471  * @veb: pointer to the VEB instance
9472  *
9473  * This is a recursive function that first builds the attached VSIs then
9474  * recurses in to build the next layer of VEB.  We track the connections
9475  * through our own index numbers because the seid's from the HW could
9476  * change across the reset.
9477  **/
9478 static int i40e_reconstitute_veb(struct i40e_veb *veb)
9479 {
9480         struct i40e_vsi *ctl_vsi = NULL;
9481         struct i40e_pf *pf = veb->pf;
9482         int v, veb_idx;
9483         int ret;
9484
9485         /* build VSI that owns this VEB, temporarily attached to base VEB */
9486         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
9487                 if (pf->vsi[v] &&
9488                     pf->vsi[v]->veb_idx == veb->idx &&
9489                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
9490                         ctl_vsi = pf->vsi[v];
9491                         break;
9492                 }
9493         }
9494         if (!ctl_vsi) {
9495                 dev_info(&pf->pdev->dev,
9496                          "missing owner VSI for veb_idx %d\n", veb->idx);
9497                 ret = -ENOENT;
9498                 goto end_reconstitute;
9499         }
9500         if (ctl_vsi != pf->vsi[pf->lan_vsi])
9501                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9502         ret = i40e_add_vsi(ctl_vsi);
9503         if (ret) {
9504                 dev_info(&pf->pdev->dev,
9505                          "rebuild of veb_idx %d owner VSI failed: %d\n",
9506                          veb->idx, ret);
9507                 goto end_reconstitute;
9508         }
9509         i40e_vsi_reset_stats(ctl_vsi);
9510
9511         /* create the VEB in the switch and move the VSI onto the VEB */
9512         ret = i40e_add_veb(veb, ctl_vsi);
9513         if (ret)
9514                 goto end_reconstitute;
9515
9516         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
9517                 veb->bridge_mode = BRIDGE_MODE_VEB;
9518         else
9519                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9520         i40e_config_bridge_mode(veb);
9521
9522         /* create the remaining VSIs attached to this VEB */
9523         for (v = 0; v < pf->num_alloc_vsi; v++) {
9524                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
9525                         continue;
9526
9527                 if (pf->vsi[v]->veb_idx == veb->idx) {
9528                         struct i40e_vsi *vsi = pf->vsi[v];
9529
9530                         vsi->uplink_seid = veb->seid;
9531                         ret = i40e_add_vsi(vsi);
9532                         if (ret) {
9533                                 dev_info(&pf->pdev->dev,
9534                                          "rebuild of vsi_idx %d failed: %d\n",
9535                                          v, ret);
9536                                 goto end_reconstitute;
9537                         }
9538                         i40e_vsi_reset_stats(vsi);
9539                 }
9540         }
9541
9542         /* create any VEBs attached to this VEB - RECURSION */
9543         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9544                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
9545                         pf->veb[veb_idx]->uplink_seid = veb->seid;
9546                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
9547                         if (ret)
9548                                 break;
9549                 }
9550         }
9551
9552 end_reconstitute:
9553         return ret;
9554 }
9555
9556 /**
9557  * i40e_get_capabilities - get info about the HW
9558  * @pf: the PF struct
9559  **/
9560 static int i40e_get_capabilities(struct i40e_pf *pf,
9561                                  enum i40e_admin_queue_opc list_type)
9562 {
9563         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
9564         u16 data_size;
9565         int buf_len;
9566         int err;
9567
9568         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
9569         do {
9570                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
9571                 if (!cap_buf)
9572                         return -ENOMEM;
9573
9574                 /* this loads the data into the hw struct for us */
9575                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
9576                                                     &data_size, list_type,
9577                                                     NULL);
9578                 /* data loaded, buffer no longer needed */
9579                 kfree(cap_buf);
9580
9581                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
9582                         /* retry with a larger buffer */
9583                         buf_len = data_size;
9584                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
9585                         dev_info(&pf->pdev->dev,
9586                                  "capability discovery failed, err %s aq_err %s\n",
9587                                  i40e_stat_str(&pf->hw, err),
9588                                  i40e_aq_str(&pf->hw,
9589                                              pf->hw.aq.asq_last_status));
9590                         return -ENODEV;
9591                 }
9592         } while (err);
9593
9594         if (pf->hw.debug_mask & I40E_DEBUG_USER) {
9595                 if (list_type == i40e_aqc_opc_list_func_capabilities) {
9596                         dev_info(&pf->pdev->dev,
9597                                  "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",
9598                                  pf->hw.pf_id, pf->hw.func_caps.num_vfs,
9599                                  pf->hw.func_caps.num_msix_vectors,
9600                                  pf->hw.func_caps.num_msix_vectors_vf,
9601                                  pf->hw.func_caps.fd_filters_guaranteed,
9602                                  pf->hw.func_caps.fd_filters_best_effort,
9603                                  pf->hw.func_caps.num_tx_qp,
9604                                  pf->hw.func_caps.num_vsis);
9605                 } else if (list_type == i40e_aqc_opc_list_dev_capabilities) {
9606                         dev_info(&pf->pdev->dev,
9607                                  "switch_mode=0x%04x, function_valid=0x%08x\n",
9608                                  pf->hw.dev_caps.switch_mode,
9609                                  pf->hw.dev_caps.valid_functions);
9610                         dev_info(&pf->pdev->dev,
9611                                  "SR-IOV=%d, num_vfs for all function=%u\n",
9612                                  pf->hw.dev_caps.sr_iov_1_1,
9613                                  pf->hw.dev_caps.num_vfs);
9614                         dev_info(&pf->pdev->dev,
9615                                  "num_vsis=%u, num_rx:%u, num_tx=%u\n",
9616                                  pf->hw.dev_caps.num_vsis,
9617                                  pf->hw.dev_caps.num_rx_qp,
9618                                  pf->hw.dev_caps.num_tx_qp);
9619                 }
9620         }
9621         if (list_type == i40e_aqc_opc_list_func_capabilities) {
9622 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
9623                        + pf->hw.func_caps.num_vfs)
9624                 if (pf->hw.revision_id == 0 &&
9625                     pf->hw.func_caps.num_vsis < DEF_NUM_VSI) {
9626                         dev_info(&pf->pdev->dev,
9627                                  "got num_vsis %d, setting num_vsis to %d\n",
9628                                  pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
9629                         pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
9630                 }
9631         }
9632         return 0;
9633 }
9634
9635 static int i40e_vsi_clear(struct i40e_vsi *vsi);
9636
9637 /**
9638  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
9639  * @pf: board private structure
9640  **/
9641 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
9642 {
9643         struct i40e_vsi *vsi;
9644
9645         /* quick workaround for an NVM issue that leaves a critical register
9646          * uninitialized
9647          */
9648         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
9649                 static const u32 hkey[] = {
9650                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
9651                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
9652                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
9653                         0x95b3a76d};
9654                 int i;
9655
9656                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
9657                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
9658         }
9659
9660         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
9661                 return;
9662
9663         /* find existing VSI and see if it needs configuring */
9664         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9665
9666         /* create a new VSI if none exists */
9667         if (!vsi) {
9668                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
9669                                      pf->vsi[pf->lan_vsi]->seid, 0);
9670                 if (!vsi) {
9671                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
9672                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9673                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
9674                         return;
9675                 }
9676         }
9677
9678         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
9679 }
9680
9681 /**
9682  * i40e_fdir_teardown - release the Flow Director resources
9683  * @pf: board private structure
9684  **/
9685 static void i40e_fdir_teardown(struct i40e_pf *pf)
9686 {
9687         struct i40e_vsi *vsi;
9688
9689         i40e_fdir_filter_exit(pf);
9690         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
9691         if (vsi)
9692                 i40e_vsi_release(vsi);
9693 }
9694
9695 /**
9696  * i40e_rebuild_cloud_filters - Rebuilds cloud filters for VSIs
9697  * @vsi: PF main vsi
9698  * @seid: seid of main or channel VSIs
9699  *
9700  * Rebuilds cloud filters associated with main VSI and channel VSIs if they
9701  * existed before reset
9702  **/
9703 static int i40e_rebuild_cloud_filters(struct i40e_vsi *vsi, u16 seid)
9704 {
9705         struct i40e_cloud_filter *cfilter;
9706         struct i40e_pf *pf = vsi->back;
9707         struct hlist_node *node;
9708         i40e_status ret;
9709
9710         /* Add cloud filters back if they exist */
9711         hlist_for_each_entry_safe(cfilter, node, &pf->cloud_filter_list,
9712                                   cloud_node) {
9713                 if (cfilter->seid != seid)
9714                         continue;
9715
9716                 if (cfilter->dst_port)
9717                         ret = i40e_add_del_cloud_filter_big_buf(vsi, cfilter,
9718                                                                 true);
9719                 else
9720                         ret = i40e_add_del_cloud_filter(vsi, cfilter, true);
9721
9722                 if (ret) {
9723                         dev_dbg(&pf->pdev->dev,
9724                                 "Failed to rebuild cloud filter, err %s aq_err %s\n",
9725                                 i40e_stat_str(&pf->hw, ret),
9726                                 i40e_aq_str(&pf->hw,
9727                                             pf->hw.aq.asq_last_status));
9728                         return ret;
9729                 }
9730         }
9731         return 0;
9732 }
9733
9734 /**
9735  * i40e_rebuild_channels - Rebuilds channel VSIs if they existed before reset
9736  * @vsi: PF main vsi
9737  *
9738  * Rebuilds channel VSIs if they existed before reset
9739  **/
9740 static int i40e_rebuild_channels(struct i40e_vsi *vsi)
9741 {
9742         struct i40e_channel *ch, *ch_tmp;
9743         i40e_status ret;
9744
9745         if (list_empty(&vsi->ch_list))
9746                 return 0;
9747
9748         list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
9749                 if (!ch->initialized)
9750                         break;
9751                 /* Proceed with creation of channel (VMDq2) VSI */
9752                 ret = i40e_add_channel(vsi->back, vsi->uplink_seid, ch);
9753                 if (ret) {
9754                         dev_info(&vsi->back->pdev->dev,
9755                                  "failed to rebuild channels using uplink_seid %u\n",
9756                                  vsi->uplink_seid);
9757                         return ret;
9758                 }
9759                 /* Reconfigure TX queues using QTX_CTL register */
9760                 ret = i40e_channel_config_tx_ring(vsi->back, vsi, ch);
9761                 if (ret) {
9762                         dev_info(&vsi->back->pdev->dev,
9763                                  "failed to configure TX rings for channel %u\n",
9764                                  ch->seid);
9765                         return ret;
9766                 }
9767                 /* update 'next_base_queue' */
9768                 vsi->next_base_queue = vsi->next_base_queue +
9769                                                         ch->num_queue_pairs;
9770                 if (ch->max_tx_rate) {
9771                         u64 credits = ch->max_tx_rate;
9772
9773                         if (i40e_set_bw_limit(vsi, ch->seid,
9774                                               ch->max_tx_rate))
9775                                 return -EINVAL;
9776
9777                         do_div(credits, I40E_BW_CREDIT_DIVISOR);
9778                         dev_dbg(&vsi->back->pdev->dev,
9779                                 "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
9780                                 ch->max_tx_rate,
9781                                 credits,
9782                                 ch->seid);
9783                 }
9784                 ret = i40e_rebuild_cloud_filters(vsi, ch->seid);
9785                 if (ret) {
9786                         dev_dbg(&vsi->back->pdev->dev,
9787                                 "Failed to rebuild cloud filters for channel VSI %u\n",
9788                                 ch->seid);
9789                         return ret;
9790                 }
9791         }
9792         return 0;
9793 }
9794
9795 /**
9796  * i40e_prep_for_reset - prep for the core to reset
9797  * @pf: board private structure
9798  * @lock_acquired: indicates whether or not the lock has been acquired
9799  * before this function was called.
9800  *
9801  * Close up the VFs and other things in prep for PF Reset.
9802   **/
9803 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
9804 {
9805         struct i40e_hw *hw = &pf->hw;
9806         i40e_status ret = 0;
9807         u32 v;
9808
9809         clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
9810         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
9811                 return;
9812         if (i40e_check_asq_alive(&pf->hw))
9813                 i40e_vc_notify_reset(pf);
9814
9815         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
9816
9817         /* quiesce the VSIs and their queues that are not already DOWN */
9818         /* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
9819         if (!lock_acquired)
9820                 rtnl_lock();
9821         i40e_pf_quiesce_all_vsi(pf);
9822         if (!lock_acquired)
9823                 rtnl_unlock();
9824
9825         for (v = 0; v < pf->num_alloc_vsi; v++) {
9826                 if (pf->vsi[v])
9827                         pf->vsi[v]->seid = 0;
9828         }
9829
9830         i40e_shutdown_adminq(&pf->hw);
9831
9832         /* call shutdown HMC */
9833         if (hw->hmc.hmc_obj) {
9834                 ret = i40e_shutdown_lan_hmc(hw);
9835                 if (ret)
9836                         dev_warn(&pf->pdev->dev,
9837                                  "shutdown_lan_hmc failed: %d\n", ret);
9838         }
9839
9840         /* Save the current PTP time so that we can restore the time after the
9841          * reset completes.
9842          */
9843         i40e_ptp_save_hw_time(pf);
9844 }
9845
9846 /**
9847  * i40e_send_version - update firmware with driver version
9848  * @pf: PF struct
9849  */
9850 static void i40e_send_version(struct i40e_pf *pf)
9851 {
9852         struct i40e_driver_version dv;
9853
9854         dv.major_version = DRV_VERSION_MAJOR;
9855         dv.minor_version = DRV_VERSION_MINOR;
9856         dv.build_version = DRV_VERSION_BUILD;
9857         dv.subbuild_version = 0;
9858         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
9859         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
9860 }
9861
9862 /**
9863  * i40e_get_oem_version - get OEM specific version information
9864  * @hw: pointer to the hardware structure
9865  **/
9866 static void i40e_get_oem_version(struct i40e_hw *hw)
9867 {
9868         u16 block_offset = 0xffff;
9869         u16 block_length = 0;
9870         u16 capabilities = 0;
9871         u16 gen_snap = 0;
9872         u16 release = 0;
9873
9874 #define I40E_SR_NVM_OEM_VERSION_PTR             0x1B
9875 #define I40E_NVM_OEM_LENGTH_OFFSET              0x00
9876 #define I40E_NVM_OEM_CAPABILITIES_OFFSET        0x01
9877 #define I40E_NVM_OEM_GEN_OFFSET                 0x02
9878 #define I40E_NVM_OEM_RELEASE_OFFSET             0x03
9879 #define I40E_NVM_OEM_CAPABILITIES_MASK          0x000F
9880 #define I40E_NVM_OEM_LENGTH                     3
9881
9882         /* Check if pointer to OEM version block is valid. */
9883         i40e_read_nvm_word(hw, I40E_SR_NVM_OEM_VERSION_PTR, &block_offset);
9884         if (block_offset == 0xffff)
9885                 return;
9886
9887         /* Check if OEM version block has correct length. */
9888         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_LENGTH_OFFSET,
9889                            &block_length);
9890         if (block_length < I40E_NVM_OEM_LENGTH)
9891                 return;
9892
9893         /* Check if OEM version format is as expected. */
9894         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_CAPABILITIES_OFFSET,
9895                            &capabilities);
9896         if ((capabilities & I40E_NVM_OEM_CAPABILITIES_MASK) != 0)
9897                 return;
9898
9899         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_GEN_OFFSET,
9900                            &gen_snap);
9901         i40e_read_nvm_word(hw, block_offset + I40E_NVM_OEM_RELEASE_OFFSET,
9902                            &release);
9903         hw->nvm.oem_ver = (gen_snap << I40E_OEM_SNAP_SHIFT) | release;
9904         hw->nvm.eetrack = I40E_OEM_EETRACK_ID;
9905 }
9906
9907 /**
9908  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
9909  * @pf: board private structure
9910  **/
9911 static int i40e_reset(struct i40e_pf *pf)
9912 {
9913         struct i40e_hw *hw = &pf->hw;
9914         i40e_status ret;
9915
9916         ret = i40e_pf_reset(hw);
9917         if (ret) {
9918                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
9919                 set_bit(__I40E_RESET_FAILED, pf->state);
9920                 clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
9921         } else {
9922                 pf->pfr_count++;
9923         }
9924         return ret;
9925 }
9926
9927 /**
9928  * i40e_rebuild - rebuild using a saved config
9929  * @pf: board private structure
9930  * @reinit: if the Main VSI needs to re-initialized.
9931  * @lock_acquired: indicates whether or not the lock has been acquired
9932  * before this function was called.
9933  **/
9934 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
9935 {
9936         int old_recovery_mode_bit = test_bit(__I40E_RECOVERY_MODE, pf->state);
9937         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
9938         struct i40e_hw *hw = &pf->hw;
9939         u8 set_fc_aq_fail = 0;
9940         i40e_status ret;
9941         u32 val;
9942         int v;
9943
9944         if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9945             i40e_check_recovery_mode(pf)) {
9946                 i40e_set_ethtool_ops(pf->vsi[pf->lan_vsi]->netdev);
9947         }
9948
9949         if (test_bit(__I40E_DOWN, pf->state) &&
9950             !test_bit(__I40E_RECOVERY_MODE, pf->state) &&
9951             !old_recovery_mode_bit)
9952                 goto clear_recovery;
9953         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
9954
9955         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
9956         ret = i40e_init_adminq(&pf->hw);
9957         if (ret) {
9958                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
9959                          i40e_stat_str(&pf->hw, ret),
9960                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9961                 goto clear_recovery;
9962         }
9963         i40e_get_oem_version(&pf->hw);
9964
9965         if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
9966             ((hw->aq.fw_maj_ver == 4 && hw->aq.fw_min_ver <= 33) ||
9967              hw->aq.fw_maj_ver < 4) && hw->mac.type == I40E_MAC_XL710) {
9968                 /* The following delay is necessary for 4.33 firmware and older
9969                  * to recover after EMP reset. 200 ms should suffice but we
9970                  * put here 300 ms to be sure that FW is ready to operate
9971                  * after reset.
9972                  */
9973                 mdelay(300);
9974         }
9975
9976         /* re-verify the eeprom if we just had an EMP reset */
9977         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
9978                 i40e_verify_eeprom(pf);
9979
9980         /* if we are going out of or into recovery mode we have to act
9981          * accordingly with regard to resources initialization
9982          * and deinitialization
9983          */
9984         if (test_bit(__I40E_RECOVERY_MODE, pf->state) ||
9985             old_recovery_mode_bit) {
9986                 if (i40e_get_capabilities(pf,
9987                                           i40e_aqc_opc_list_func_capabilities))
9988                         goto end_unlock;
9989
9990                 if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
9991                         /* we're staying in recovery mode so we'll reinitialize
9992                          * misc vector here
9993                          */
9994                         if (i40e_setup_misc_vector_for_recovery_mode(pf))
9995                                 goto end_unlock;
9996                 } else {
9997                         if (!lock_acquired)
9998                                 rtnl_lock();
9999                         /* we're going out of recovery mode so we'll free
10000                          * the IRQ allocated specifically for recovery mode
10001                          * and restore the interrupt scheme
10002                          */
10003                         free_irq(pf->pdev->irq, pf);
10004                         i40e_clear_interrupt_scheme(pf);
10005                         if (i40e_restore_interrupt_scheme(pf))
10006                                 goto end_unlock;
10007                 }
10008
10009                 /* tell the firmware that we're starting */
10010                 i40e_send_version(pf);
10011
10012                 /* bail out in case recovery mode was detected, as there is
10013                  * no need for further configuration.
10014                  */
10015                 goto end_unlock;
10016         }
10017
10018         i40e_clear_pxe_mode(hw);
10019         ret = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
10020         if (ret)
10021                 goto end_core_reset;
10022
10023         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10024                                 hw->func_caps.num_rx_qp, 0, 0);
10025         if (ret) {
10026                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
10027                 goto end_core_reset;
10028         }
10029         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10030         if (ret) {
10031                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
10032                 goto end_core_reset;
10033         }
10034
10035         /* Enable FW to write a default DCB config on link-up */
10036         i40e_aq_set_dcb_parameters(hw, true, NULL);
10037
10038 #ifdef CONFIG_I40E_DCB
10039         ret = i40e_init_pf_dcb(pf);
10040         if (ret) {
10041                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
10042                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10043                 /* Continue without DCB enabled */
10044         }
10045 #endif /* CONFIG_I40E_DCB */
10046         /* do basic switch setup */
10047         if (!lock_acquired)
10048                 rtnl_lock();
10049         ret = i40e_setup_pf_switch(pf, reinit);
10050         if (ret)
10051                 goto end_unlock;
10052
10053         /* The driver only wants link up/down and module qualification
10054          * reports from firmware.  Note the negative logic.
10055          */
10056         ret = i40e_aq_set_phy_int_mask(&pf->hw,
10057                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
10058                                          I40E_AQ_EVENT_MEDIA_NA |
10059                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
10060         if (ret)
10061                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10062                          i40e_stat_str(&pf->hw, ret),
10063                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10064
10065         /* make sure our flow control settings are restored */
10066         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
10067         if (ret)
10068                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
10069                         i40e_stat_str(&pf->hw, ret),
10070                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10071
10072         /* Rebuild the VSIs and VEBs that existed before reset.
10073          * They are still in our local switch element arrays, so only
10074          * need to rebuild the switch model in the HW.
10075          *
10076          * If there were VEBs but the reconstitution failed, we'll try
10077          * try to recover minimal use by getting the basic PF VSI working.
10078          */
10079         if (vsi->uplink_seid != pf->mac_seid) {
10080                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
10081                 /* find the one VEB connected to the MAC, and find orphans */
10082                 for (v = 0; v < I40E_MAX_VEB; v++) {
10083                         if (!pf->veb[v])
10084                                 continue;
10085
10086                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
10087                             pf->veb[v]->uplink_seid == 0) {
10088                                 ret = i40e_reconstitute_veb(pf->veb[v]);
10089
10090                                 if (!ret)
10091                                         continue;
10092
10093                                 /* If Main VEB failed, we're in deep doodoo,
10094                                  * so give up rebuilding the switch and set up
10095                                  * for minimal rebuild of PF VSI.
10096                                  * If orphan failed, we'll report the error
10097                                  * but try to keep going.
10098                                  */
10099                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
10100                                         dev_info(&pf->pdev->dev,
10101                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
10102                                                  ret);
10103                                         vsi->uplink_seid = pf->mac_seid;
10104                                         break;
10105                                 } else if (pf->veb[v]->uplink_seid == 0) {
10106                                         dev_info(&pf->pdev->dev,
10107                                                  "rebuild of orphan VEB failed: %d\n",
10108                                                  ret);
10109                                 }
10110                         }
10111                 }
10112         }
10113
10114         if (vsi->uplink_seid == pf->mac_seid) {
10115                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
10116                 /* no VEB, so rebuild only the Main VSI */
10117                 ret = i40e_add_vsi(vsi);
10118                 if (ret) {
10119                         dev_info(&pf->pdev->dev,
10120                                  "rebuild of Main VSI failed: %d\n", ret);
10121                         goto end_unlock;
10122                 }
10123         }
10124
10125         if (vsi->mqprio_qopt.max_rate[0]) {
10126                 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
10127                 u64 credits = 0;
10128
10129                 do_div(max_tx_rate, I40E_BW_MBPS_DIVISOR);
10130                 ret = i40e_set_bw_limit(vsi, vsi->seid, max_tx_rate);
10131                 if (ret)
10132                         goto end_unlock;
10133
10134                 credits = max_tx_rate;
10135                 do_div(credits, I40E_BW_CREDIT_DIVISOR);
10136                 dev_dbg(&vsi->back->pdev->dev,
10137                         "Set tx rate of %llu Mbps (count of 50Mbps %llu) for vsi->seid %u\n",
10138                         max_tx_rate,
10139                         credits,
10140                         vsi->seid);
10141         }
10142
10143         ret = i40e_rebuild_cloud_filters(vsi, vsi->seid);
10144         if (ret)
10145                 goto end_unlock;
10146
10147         /* PF Main VSI is rebuild by now, go ahead and rebuild channel VSIs
10148          * for this main VSI if they exist
10149          */
10150         ret = i40e_rebuild_channels(vsi);
10151         if (ret)
10152                 goto end_unlock;
10153
10154         /* Reconfigure hardware for allowing smaller MSS in the case
10155          * of TSO, so that we avoid the MDD being fired and causing
10156          * a reset in the case of small MSS+TSO.
10157          */
10158 #define I40E_REG_MSS          0x000E64DC
10159 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
10160 #define I40E_64BYTE_MSS       0x400000
10161         val = rd32(hw, I40E_REG_MSS);
10162         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10163                 val &= ~I40E_REG_MSS_MIN_MASK;
10164                 val |= I40E_64BYTE_MSS;
10165                 wr32(hw, I40E_REG_MSS, val);
10166         }
10167
10168         if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
10169                 msleep(75);
10170                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10171                 if (ret)
10172                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10173                                  i40e_stat_str(&pf->hw, ret),
10174                                  i40e_aq_str(&pf->hw,
10175                                              pf->hw.aq.asq_last_status));
10176         }
10177         /* reinit the misc interrupt */
10178         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10179                 ret = i40e_setup_misc_vector(pf);
10180
10181         /* Add a filter to drop all Flow control frames from any VSI from being
10182          * transmitted. By doing so we stop a malicious VF from sending out
10183          * PAUSE or PFC frames and potentially controlling traffic for other
10184          * PF/VF VSIs.
10185          * The FW can still send Flow control frames if enabled.
10186          */
10187         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
10188                                                        pf->main_vsi_seid);
10189
10190         /* restart the VSIs that were rebuilt and running before the reset */
10191         i40e_pf_unquiesce_all_vsi(pf);
10192
10193         /* Release the RTNL lock before we start resetting VFs */
10194         if (!lock_acquired)
10195                 rtnl_unlock();
10196
10197         /* Restore promiscuous settings */
10198         ret = i40e_set_promiscuous(pf, pf->cur_promisc);
10199         if (ret)
10200                 dev_warn(&pf->pdev->dev,
10201                          "Failed to restore promiscuous setting: %s, err %s aq_err %s\n",
10202                          pf->cur_promisc ? "on" : "off",
10203                          i40e_stat_str(&pf->hw, ret),
10204                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10205
10206         i40e_reset_all_vfs(pf, true);
10207
10208         /* tell the firmware that we're starting */
10209         i40e_send_version(pf);
10210
10211         /* We've already released the lock, so don't do it again */
10212         goto end_core_reset;
10213
10214 end_unlock:
10215         if (!lock_acquired)
10216                 rtnl_unlock();
10217 end_core_reset:
10218         clear_bit(__I40E_RESET_FAILED, pf->state);
10219 clear_recovery:
10220         clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
10221         clear_bit(__I40E_TIMEOUT_RECOVERY_PENDING, pf->state);
10222 }
10223
10224 /**
10225  * i40e_reset_and_rebuild - reset and rebuild using a saved config
10226  * @pf: board private structure
10227  * @reinit: if the Main VSI needs to re-initialized.
10228  * @lock_acquired: indicates whether or not the lock has been acquired
10229  * before this function was called.
10230  **/
10231 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
10232                                    bool lock_acquired)
10233 {
10234         int ret;
10235         /* Now we wait for GRST to settle out.
10236          * We don't have to delete the VEBs or VSIs from the hw switch
10237          * because the reset will make them disappear.
10238          */
10239         ret = i40e_reset(pf);
10240         if (!ret)
10241                 i40e_rebuild(pf, reinit, lock_acquired);
10242 }
10243
10244 /**
10245  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
10246  * @pf: board private structure
10247  *
10248  * Close up the VFs and other things in prep for a Core Reset,
10249  * then get ready to rebuild the world.
10250  * @lock_acquired: indicates whether or not the lock has been acquired
10251  * before this function was called.
10252  **/
10253 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
10254 {
10255         i40e_prep_for_reset(pf, lock_acquired);
10256         i40e_reset_and_rebuild(pf, false, lock_acquired);
10257 }
10258
10259 /**
10260  * i40e_handle_mdd_event
10261  * @pf: pointer to the PF structure
10262  *
10263  * Called from the MDD irq handler to identify possibly malicious vfs
10264  **/
10265 static void i40e_handle_mdd_event(struct i40e_pf *pf)
10266 {
10267         struct i40e_hw *hw = &pf->hw;
10268         bool mdd_detected = false;
10269         struct i40e_vf *vf;
10270         u32 reg;
10271         int i;
10272
10273         if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
10274                 return;
10275
10276         /* find what triggered the MDD event */
10277         reg = rd32(hw, I40E_GL_MDET_TX);
10278         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
10279                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
10280                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
10281                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
10282                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
10283                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
10284                                 I40E_GL_MDET_TX_EVENT_SHIFT;
10285                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
10286                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
10287                                 pf->hw.func_caps.base_queue;
10288                 if (netif_msg_tx_err(pf))
10289                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
10290                                  event, queue, pf_num, vf_num);
10291                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
10292                 mdd_detected = true;
10293         }
10294         reg = rd32(hw, I40E_GL_MDET_RX);
10295         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
10296                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
10297                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
10298                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
10299                                 I40E_GL_MDET_RX_EVENT_SHIFT;
10300                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
10301                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
10302                                 pf->hw.func_caps.base_queue;
10303                 if (netif_msg_rx_err(pf))
10304                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
10305                                  event, queue, func);
10306                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
10307                 mdd_detected = true;
10308         }
10309
10310         if (mdd_detected) {
10311                 reg = rd32(hw, I40E_PF_MDET_TX);
10312                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
10313                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
10314                         dev_dbg(&pf->pdev->dev, "TX driver issue detected on PF\n");
10315                 }
10316                 reg = rd32(hw, I40E_PF_MDET_RX);
10317                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
10318                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
10319                         dev_dbg(&pf->pdev->dev, "RX driver issue detected on PF\n");
10320                 }
10321         }
10322
10323         /* see if one of the VFs needs its hand slapped */
10324         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
10325                 vf = &(pf->vf[i]);
10326                 reg = rd32(hw, I40E_VP_MDET_TX(i));
10327                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
10328                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
10329                         vf->num_mdd_events++;
10330                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
10331                                  i);
10332                         dev_info(&pf->pdev->dev,
10333                                  "Use PF Control I/F to re-enable the VF\n");
10334                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10335                 }
10336
10337                 reg = rd32(hw, I40E_VP_MDET_RX(i));
10338                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
10339                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
10340                         vf->num_mdd_events++;
10341                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
10342                                  i);
10343                         dev_info(&pf->pdev->dev,
10344                                  "Use PF Control I/F to re-enable the VF\n");
10345                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
10346                 }
10347         }
10348
10349         /* re-enable mdd interrupt cause */
10350         clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
10351         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
10352         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
10353         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
10354         i40e_flush(hw);
10355 }
10356
10357 static const char *i40e_tunnel_name(u8 type)
10358 {
10359         switch (type) {
10360         case UDP_TUNNEL_TYPE_VXLAN:
10361                 return "vxlan";
10362         case UDP_TUNNEL_TYPE_GENEVE:
10363                 return "geneve";
10364         default:
10365                 return "unknown";
10366         }
10367 }
10368
10369 /**
10370  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
10371  * @pf: board private structure
10372  **/
10373 static void i40e_sync_udp_filters(struct i40e_pf *pf)
10374 {
10375         int i;
10376
10377         /* loop through and set pending bit for all active UDP filters */
10378         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10379                 if (pf->udp_ports[i].port)
10380                         pf->pending_udp_bitmap |= BIT_ULL(i);
10381         }
10382
10383         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
10384 }
10385
10386 /**
10387  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
10388  * @pf: board private structure
10389  **/
10390 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
10391 {
10392         struct i40e_hw *hw = &pf->hw;
10393         u8 filter_index, type;
10394         u16 port;
10395         int i;
10396
10397         if (!test_and_clear_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state))
10398                 return;
10399
10400         /* acquire RTNL to maintain state of flags and port requests */
10401         rtnl_lock();
10402
10403         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
10404                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
10405                         struct i40e_udp_port_config *udp_port;
10406                         i40e_status ret = 0;
10407
10408                         udp_port = &pf->udp_ports[i];
10409                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
10410
10411                         port = READ_ONCE(udp_port->port);
10412                         type = READ_ONCE(udp_port->type);
10413                         filter_index = READ_ONCE(udp_port->filter_index);
10414
10415                         /* release RTNL while we wait on AQ command */
10416                         rtnl_unlock();
10417
10418                         if (port)
10419                                 ret = i40e_aq_add_udp_tunnel(hw, port,
10420                                                              type,
10421                                                              &filter_index,
10422                                                              NULL);
10423                         else if (filter_index != I40E_UDP_PORT_INDEX_UNUSED)
10424                                 ret = i40e_aq_del_udp_tunnel(hw, filter_index,
10425                                                              NULL);
10426
10427                         /* reacquire RTNL so we can update filter_index */
10428                         rtnl_lock();
10429
10430                         if (ret) {
10431                                 dev_info(&pf->pdev->dev,
10432                                          "%s %s port %d, index %d failed, err %s aq_err %s\n",
10433                                          i40e_tunnel_name(type),
10434                                          port ? "add" : "delete",
10435                                          port,
10436                                          filter_index,
10437                                          i40e_stat_str(&pf->hw, ret),
10438                                          i40e_aq_str(&pf->hw,
10439                                                      pf->hw.aq.asq_last_status));
10440                                 if (port) {
10441                                         /* failed to add, just reset port,
10442                                          * drop pending bit for any deletion
10443                                          */
10444                                         udp_port->port = 0;
10445                                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
10446                                 }
10447                         } else if (port) {
10448                                 /* record filter index on success */
10449                                 udp_port->filter_index = filter_index;
10450                         }
10451                 }
10452         }
10453
10454         rtnl_unlock();
10455 }
10456
10457 /**
10458  * i40e_service_task - Run the driver's async subtasks
10459  * @work: pointer to work_struct containing our data
10460  **/
10461 static void i40e_service_task(struct work_struct *work)
10462 {
10463         struct i40e_pf *pf = container_of(work,
10464                                           struct i40e_pf,
10465                                           service_task);
10466         unsigned long start_time = jiffies;
10467
10468         /* don't bother with service tasks if a reset is in progress */
10469         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state) ||
10470             test_bit(__I40E_SUSPENDED, pf->state))
10471                 return;
10472
10473         if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
10474                 return;
10475
10476         if (!test_bit(__I40E_RECOVERY_MODE, pf->state)) {
10477                 i40e_detect_recover_hung(pf->vsi[pf->lan_vsi]);
10478                 i40e_sync_filters_subtask(pf);
10479                 i40e_reset_subtask(pf);
10480                 i40e_handle_mdd_event(pf);
10481                 i40e_vc_process_vflr_event(pf);
10482                 i40e_watchdog_subtask(pf);
10483                 i40e_fdir_reinit_subtask(pf);
10484                 if (test_and_clear_bit(__I40E_CLIENT_RESET, pf->state)) {
10485                         /* Client subtask will reopen next time through. */
10486                         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi],
10487                                                            true);
10488                 } else {
10489                         i40e_client_subtask(pf);
10490                         if (test_and_clear_bit(__I40E_CLIENT_L2_CHANGE,
10491                                                pf->state))
10492                                 i40e_notify_client_of_l2_param_changes(
10493                                                                 pf->vsi[pf->lan_vsi]);
10494                 }
10495                 i40e_sync_filters_subtask(pf);
10496                 i40e_sync_udp_filters_subtask(pf);
10497         } else {
10498                 i40e_reset_subtask(pf);
10499         }
10500
10501         i40e_clean_adminq_subtask(pf);
10502
10503         /* flush memory to make sure state is correct before next watchdog */
10504         smp_mb__before_atomic();
10505         clear_bit(__I40E_SERVICE_SCHED, pf->state);
10506
10507         /* If the tasks have taken longer than one timer cycle or there
10508          * is more work to be done, reschedule the service task now
10509          * rather than wait for the timer to tick again.
10510          */
10511         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
10512             test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)             ||
10513             test_bit(__I40E_MDD_EVENT_PENDING, pf->state)                ||
10514             test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
10515                 i40e_service_event_schedule(pf);
10516 }
10517
10518 /**
10519  * i40e_service_timer - timer callback
10520  * @data: pointer to PF struct
10521  **/
10522 static void i40e_service_timer(struct timer_list *t)
10523 {
10524         struct i40e_pf *pf = from_timer(pf, t, service_timer);
10525
10526         mod_timer(&pf->service_timer,
10527                   round_jiffies(jiffies + pf->service_timer_period));
10528         i40e_service_event_schedule(pf);
10529 }
10530
10531 /**
10532  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
10533  * @vsi: the VSI being configured
10534  **/
10535 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
10536 {
10537         struct i40e_pf *pf = vsi->back;
10538
10539         switch (vsi->type) {
10540         case I40E_VSI_MAIN:
10541                 vsi->alloc_queue_pairs = pf->num_lan_qps;
10542                 if (!vsi->num_tx_desc)
10543                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10544                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10545                 if (!vsi->num_rx_desc)
10546                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10547                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10548                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
10549                         vsi->num_q_vectors = pf->num_lan_msix;
10550                 else
10551                         vsi->num_q_vectors = 1;
10552
10553                 break;
10554
10555         case I40E_VSI_FDIR:
10556                 vsi->alloc_queue_pairs = 1;
10557                 vsi->num_tx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10558                                          I40E_REQ_DESCRIPTOR_MULTIPLE);
10559                 vsi->num_rx_desc = ALIGN(I40E_FDIR_RING_COUNT,
10560                                          I40E_REQ_DESCRIPTOR_MULTIPLE);
10561                 vsi->num_q_vectors = pf->num_fdsb_msix;
10562                 break;
10563
10564         case I40E_VSI_VMDQ2:
10565                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
10566                 if (!vsi->num_tx_desc)
10567                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10568                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10569                 if (!vsi->num_rx_desc)
10570                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10571                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10572                 vsi->num_q_vectors = pf->num_vmdq_msix;
10573                 break;
10574
10575         case I40E_VSI_SRIOV:
10576                 vsi->alloc_queue_pairs = pf->num_vf_qps;
10577                 if (!vsi->num_tx_desc)
10578                         vsi->num_tx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10579                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10580                 if (!vsi->num_rx_desc)
10581                         vsi->num_rx_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
10582                                                  I40E_REQ_DESCRIPTOR_MULTIPLE);
10583                 break;
10584
10585         default:
10586                 WARN_ON(1);
10587                 return -ENODATA;
10588         }
10589
10590         return 0;
10591 }
10592
10593 /**
10594  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
10595  * @vsi: VSI pointer
10596  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
10597  *
10598  * On error: returns error code (negative)
10599  * On success: returns 0
10600  **/
10601 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
10602 {
10603         struct i40e_ring **next_rings;
10604         int size;
10605         int ret = 0;
10606
10607         /* allocate memory for both Tx, XDP Tx and Rx ring pointers */
10608         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs *
10609                (i40e_enabled_xdp_vsi(vsi) ? 3 : 2);
10610         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
10611         if (!vsi->tx_rings)
10612                 return -ENOMEM;
10613         next_rings = vsi->tx_rings + vsi->alloc_queue_pairs;
10614         if (i40e_enabled_xdp_vsi(vsi)) {
10615                 vsi->xdp_rings = next_rings;
10616                 next_rings += vsi->alloc_queue_pairs;
10617         }
10618         vsi->rx_rings = next_rings;
10619
10620         if (alloc_qvectors) {
10621                 /* allocate memory for q_vector pointers */
10622                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
10623                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
10624                 if (!vsi->q_vectors) {
10625                         ret = -ENOMEM;
10626                         goto err_vectors;
10627                 }
10628         }
10629         return ret;
10630
10631 err_vectors:
10632         kfree(vsi->tx_rings);
10633         return ret;
10634 }
10635
10636 /**
10637  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
10638  * @pf: board private structure
10639  * @type: type of VSI
10640  *
10641  * On error: returns error code (negative)
10642  * On success: returns vsi index in PF (positive)
10643  **/
10644 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
10645 {
10646         int ret = -ENODEV;
10647         struct i40e_vsi *vsi;
10648         int vsi_idx;
10649         int i;
10650
10651         /* Need to protect the allocation of the VSIs at the PF level */
10652         mutex_lock(&pf->switch_mutex);
10653
10654         /* VSI list may be fragmented if VSI creation/destruction has
10655          * been happening.  We can afford to do a quick scan to look
10656          * for any free VSIs in the list.
10657          *
10658          * find next empty vsi slot, looping back around if necessary
10659          */
10660         i = pf->next_vsi;
10661         while (i < pf->num_alloc_vsi && pf->vsi[i])
10662                 i++;
10663         if (i >= pf->num_alloc_vsi) {
10664                 i = 0;
10665                 while (i < pf->next_vsi && pf->vsi[i])
10666                         i++;
10667         }
10668
10669         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
10670                 vsi_idx = i;             /* Found one! */
10671         } else {
10672                 ret = -ENODEV;
10673                 goto unlock_pf;  /* out of VSI slots! */
10674         }
10675         pf->next_vsi = ++i;
10676
10677         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
10678         if (!vsi) {
10679                 ret = -ENOMEM;
10680                 goto unlock_pf;
10681         }
10682         vsi->type = type;
10683         vsi->back = pf;
10684         set_bit(__I40E_VSI_DOWN, vsi->state);
10685         vsi->flags = 0;
10686         vsi->idx = vsi_idx;
10687         vsi->int_rate_limit = 0;
10688         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
10689                                 pf->rss_table_size : 64;
10690         vsi->netdev_registered = false;
10691         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
10692         hash_init(vsi->mac_filter_hash);
10693         vsi->irqs_ready = false;
10694
10695         if (type == I40E_VSI_MAIN) {
10696                 vsi->af_xdp_zc_qps = bitmap_zalloc(pf->num_lan_qps, GFP_KERNEL);
10697                 if (!vsi->af_xdp_zc_qps)
10698                         goto err_rings;
10699         }
10700
10701         ret = i40e_set_num_rings_in_vsi(vsi);
10702         if (ret)
10703                 goto err_rings;
10704
10705         ret = i40e_vsi_alloc_arrays(vsi, true);
10706         if (ret)
10707                 goto err_rings;
10708
10709         /* Setup default MSIX irq handler for VSI */
10710         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
10711
10712         /* Initialize VSI lock */
10713         spin_lock_init(&vsi->mac_filter_hash_lock);
10714         pf->vsi[vsi_idx] = vsi;
10715         ret = vsi_idx;
10716         goto unlock_pf;
10717
10718 err_rings:
10719         bitmap_free(vsi->af_xdp_zc_qps);
10720         pf->next_vsi = i - 1;
10721         kfree(vsi);
10722 unlock_pf:
10723         mutex_unlock(&pf->switch_mutex);
10724         return ret;
10725 }
10726
10727 /**
10728  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
10729  * @vsi: VSI pointer
10730  * @free_qvectors: a bool to specify if q_vectors need to be freed.
10731  *
10732  * On error: returns error code (negative)
10733  * On success: returns 0
10734  **/
10735 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
10736 {
10737         /* free the ring and vector containers */
10738         if (free_qvectors) {
10739                 kfree(vsi->q_vectors);
10740                 vsi->q_vectors = NULL;
10741         }
10742         kfree(vsi->tx_rings);
10743         vsi->tx_rings = NULL;
10744         vsi->rx_rings = NULL;
10745         vsi->xdp_rings = NULL;
10746 }
10747
10748 /**
10749  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
10750  * and lookup table
10751  * @vsi: Pointer to VSI structure
10752  */
10753 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
10754 {
10755         if (!vsi)
10756                 return;
10757
10758         kfree(vsi->rss_hkey_user);
10759         vsi->rss_hkey_user = NULL;
10760
10761         kfree(vsi->rss_lut_user);
10762         vsi->rss_lut_user = NULL;
10763 }
10764
10765 /**
10766  * i40e_vsi_clear - Deallocate the VSI provided
10767  * @vsi: the VSI being un-configured
10768  **/
10769 static int i40e_vsi_clear(struct i40e_vsi *vsi)
10770 {
10771         struct i40e_pf *pf;
10772
10773         if (!vsi)
10774                 return 0;
10775
10776         if (!vsi->back)
10777                 goto free_vsi;
10778         pf = vsi->back;
10779
10780         mutex_lock(&pf->switch_mutex);
10781         if (!pf->vsi[vsi->idx]) {
10782                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](type %d)\n",
10783                         vsi->idx, vsi->idx, vsi->type);
10784                 goto unlock_vsi;
10785         }
10786
10787         if (pf->vsi[vsi->idx] != vsi) {
10788                 dev_err(&pf->pdev->dev,
10789                         "pf->vsi[%d](type %d) != vsi[%d](type %d): no free!\n",
10790                         pf->vsi[vsi->idx]->idx,
10791                         pf->vsi[vsi->idx]->type,
10792                         vsi->idx, vsi->type);
10793                 goto unlock_vsi;
10794         }
10795
10796         /* updates the PF for this cleared vsi */
10797         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
10798         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
10799
10800         bitmap_free(vsi->af_xdp_zc_qps);
10801         i40e_vsi_free_arrays(vsi, true);
10802         i40e_clear_rss_config_user(vsi);
10803
10804         pf->vsi[vsi->idx] = NULL;
10805         if (vsi->idx < pf->next_vsi)
10806                 pf->next_vsi = vsi->idx;
10807
10808 unlock_vsi:
10809         mutex_unlock(&pf->switch_mutex);
10810 free_vsi:
10811         kfree(vsi);
10812
10813         return 0;
10814 }
10815
10816 /**
10817  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
10818  * @vsi: the VSI being cleaned
10819  **/
10820 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
10821 {
10822         int i;
10823
10824         if (vsi->tx_rings && vsi->tx_rings[0]) {
10825                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10826                         kfree_rcu(vsi->tx_rings[i], rcu);
10827                         vsi->tx_rings[i] = NULL;
10828                         vsi->rx_rings[i] = NULL;
10829                         if (vsi->xdp_rings)
10830                                 vsi->xdp_rings[i] = NULL;
10831                 }
10832         }
10833 }
10834
10835 /**
10836  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
10837  * @vsi: the VSI being configured
10838  **/
10839 static int i40e_alloc_rings(struct i40e_vsi *vsi)
10840 {
10841         int i, qpv = i40e_enabled_xdp_vsi(vsi) ? 3 : 2;
10842         struct i40e_pf *pf = vsi->back;
10843         struct i40e_ring *ring;
10844
10845         /* Set basic values in the rings to be used later during open() */
10846         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
10847                 /* allocate space for both Tx and Rx in one shot */
10848                 ring = kcalloc(qpv, sizeof(struct i40e_ring), GFP_KERNEL);
10849                 if (!ring)
10850                         goto err_out;
10851
10852                 ring->queue_index = i;
10853                 ring->reg_idx = vsi->base_queue + i;
10854                 ring->ring_active = false;
10855                 ring->vsi = vsi;
10856                 ring->netdev = vsi->netdev;
10857                 ring->dev = &pf->pdev->dev;
10858                 ring->count = vsi->num_tx_desc;
10859                 ring->size = 0;
10860                 ring->dcb_tc = 0;
10861                 if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10862                         ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10863                 ring->itr_setting = pf->tx_itr_default;
10864                 vsi->tx_rings[i] = ring++;
10865
10866                 if (!i40e_enabled_xdp_vsi(vsi))
10867                         goto setup_rx;
10868
10869                 ring->queue_index = vsi->alloc_queue_pairs + i;
10870                 ring->reg_idx = vsi->base_queue + ring->queue_index;
10871                 ring->ring_active = false;
10872                 ring->vsi = vsi;
10873                 ring->netdev = NULL;
10874                 ring->dev = &pf->pdev->dev;
10875                 ring->count = vsi->num_tx_desc;
10876                 ring->size = 0;
10877                 ring->dcb_tc = 0;
10878                 if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
10879                         ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
10880                 set_ring_xdp(ring);
10881                 ring->itr_setting = pf->tx_itr_default;
10882                 vsi->xdp_rings[i] = ring++;
10883
10884 setup_rx:
10885                 ring->queue_index = i;
10886                 ring->reg_idx = vsi->base_queue + i;
10887                 ring->ring_active = false;
10888                 ring->vsi = vsi;
10889                 ring->netdev = vsi->netdev;
10890                 ring->dev = &pf->pdev->dev;
10891                 ring->count = vsi->num_rx_desc;
10892                 ring->size = 0;
10893                 ring->dcb_tc = 0;
10894                 ring->itr_setting = pf->rx_itr_default;
10895                 vsi->rx_rings[i] = ring;
10896         }
10897
10898         return 0;
10899
10900 err_out:
10901         i40e_vsi_clear_rings(vsi);
10902         return -ENOMEM;
10903 }
10904
10905 /**
10906  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
10907  * @pf: board private structure
10908  * @vectors: the number of MSI-X vectors to request
10909  *
10910  * Returns the number of vectors reserved, or error
10911  **/
10912 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
10913 {
10914         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
10915                                         I40E_MIN_MSIX, vectors);
10916         if (vectors < 0) {
10917                 dev_info(&pf->pdev->dev,
10918                          "MSI-X vector reservation failed: %d\n", vectors);
10919                 vectors = 0;
10920         }
10921
10922         return vectors;
10923 }
10924
10925 /**
10926  * i40e_init_msix - Setup the MSIX capability
10927  * @pf: board private structure
10928  *
10929  * Work with the OS to set up the MSIX vectors needed.
10930  *
10931  * Returns the number of vectors reserved or negative on failure
10932  **/
10933 static int i40e_init_msix(struct i40e_pf *pf)
10934 {
10935         struct i40e_hw *hw = &pf->hw;
10936         int cpus, extra_vectors;
10937         int vectors_left;
10938         int v_budget, i;
10939         int v_actual;
10940         int iwarp_requested = 0;
10941
10942         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
10943                 return -ENODEV;
10944
10945         /* The number of vectors we'll request will be comprised of:
10946          *   - Add 1 for "other" cause for Admin Queue events, etc.
10947          *   - The number of LAN queue pairs
10948          *      - Queues being used for RSS.
10949          *              We don't need as many as max_rss_size vectors.
10950          *              use rss_size instead in the calculation since that
10951          *              is governed by number of cpus in the system.
10952          *      - assumes symmetric Tx/Rx pairing
10953          *   - The number of VMDq pairs
10954          *   - The CPU count within the NUMA node if iWARP is enabled
10955          * Once we count this up, try the request.
10956          *
10957          * If we can't get what we want, we'll simplify to nearly nothing
10958          * and try again.  If that still fails, we punt.
10959          */
10960         vectors_left = hw->func_caps.num_msix_vectors;
10961         v_budget = 0;
10962
10963         /* reserve one vector for miscellaneous handler */
10964         if (vectors_left) {
10965                 v_budget++;
10966                 vectors_left--;
10967         }
10968
10969         /* reserve some vectors for the main PF traffic queues. Initially we
10970          * only reserve at most 50% of the available vectors, in the case that
10971          * the number of online CPUs is large. This ensures that we can enable
10972          * extra features as well. Once we've enabled the other features, we
10973          * will use any remaining vectors to reach as close as we can to the
10974          * number of online CPUs.
10975          */
10976         cpus = num_online_cpus();
10977         pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
10978         vectors_left -= pf->num_lan_msix;
10979
10980         /* reserve one vector for sideband flow director */
10981         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10982                 if (vectors_left) {
10983                         pf->num_fdsb_msix = 1;
10984                         v_budget++;
10985                         vectors_left--;
10986                 } else {
10987                         pf->num_fdsb_msix = 0;
10988                 }
10989         }
10990
10991         /* can we reserve enough for iWARP? */
10992         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
10993                 iwarp_requested = pf->num_iwarp_msix;
10994
10995                 if (!vectors_left)
10996                         pf->num_iwarp_msix = 0;
10997                 else if (vectors_left < pf->num_iwarp_msix)
10998                         pf->num_iwarp_msix = 1;
10999                 v_budget += pf->num_iwarp_msix;
11000                 vectors_left -= pf->num_iwarp_msix;
11001         }
11002
11003         /* any vectors left over go for VMDq support */
11004         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
11005                 if (!vectors_left) {
11006                         pf->num_vmdq_msix = 0;
11007                         pf->num_vmdq_qps = 0;
11008                 } else {
11009                         int vmdq_vecs_wanted =
11010                                 pf->num_vmdq_vsis * pf->num_vmdq_qps;
11011                         int vmdq_vecs =
11012                                 min_t(int, vectors_left, vmdq_vecs_wanted);
11013
11014                         /* if we're short on vectors for what's desired, we limit
11015                          * the queues per vmdq.  If this is still more than are
11016                          * available, the user will need to change the number of
11017                          * queues/vectors used by the PF later with the ethtool
11018                          * channels command
11019                          */
11020                         if (vectors_left < vmdq_vecs_wanted) {
11021                                 pf->num_vmdq_qps = 1;
11022                                 vmdq_vecs_wanted = pf->num_vmdq_vsis;
11023                                 vmdq_vecs = min_t(int,
11024                                                   vectors_left,
11025                                                   vmdq_vecs_wanted);
11026                         }
11027                         pf->num_vmdq_msix = pf->num_vmdq_qps;
11028
11029                         v_budget += vmdq_vecs;
11030                         vectors_left -= vmdq_vecs;
11031                 }
11032         }
11033
11034         /* On systems with a large number of SMP cores, we previously limited
11035          * the number of vectors for num_lan_msix to be at most 50% of the
11036          * available vectors, to allow for other features. Now, we add back
11037          * the remaining vectors. However, we ensure that the total
11038          * num_lan_msix will not exceed num_online_cpus(). To do this, we
11039          * calculate the number of vectors we can add without going over the
11040          * cap of CPUs. For systems with a small number of CPUs this will be
11041          * zero.
11042          */
11043         extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
11044         pf->num_lan_msix += extra_vectors;
11045         vectors_left -= extra_vectors;
11046
11047         WARN(vectors_left < 0,
11048              "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
11049
11050         v_budget += pf->num_lan_msix;
11051         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
11052                                    GFP_KERNEL);
11053         if (!pf->msix_entries)
11054                 return -ENOMEM;
11055
11056         for (i = 0; i < v_budget; i++)
11057                 pf->msix_entries[i].entry = i;
11058         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
11059
11060         if (v_actual < I40E_MIN_MSIX) {
11061                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
11062                 kfree(pf->msix_entries);
11063                 pf->msix_entries = NULL;
11064                 pci_disable_msix(pf->pdev);
11065                 return -ENODEV;
11066
11067         } else if (v_actual == I40E_MIN_MSIX) {
11068                 /* Adjust for minimal MSIX use */
11069                 pf->num_vmdq_vsis = 0;
11070                 pf->num_vmdq_qps = 0;
11071                 pf->num_lan_qps = 1;
11072                 pf->num_lan_msix = 1;
11073
11074         } else if (v_actual != v_budget) {
11075                 /* If we have limited resources, we will start with no vectors
11076                  * for the special features and then allocate vectors to some
11077                  * of these features based on the policy and at the end disable
11078                  * the features that did not get any vectors.
11079                  */
11080                 int vec;
11081
11082                 dev_info(&pf->pdev->dev,
11083                          "MSI-X vector limit reached with %d, wanted %d, attempting to redistribute vectors\n",
11084                          v_actual, v_budget);
11085                 /* reserve the misc vector */
11086                 vec = v_actual - 1;
11087
11088                 /* Scale vector usage down */
11089                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
11090                 pf->num_vmdq_vsis = 1;
11091                 pf->num_vmdq_qps = 1;
11092
11093                 /* partition out the remaining vectors */
11094                 switch (vec) {
11095                 case 2:
11096                         pf->num_lan_msix = 1;
11097                         break;
11098                 case 3:
11099                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11100                                 pf->num_lan_msix = 1;
11101                                 pf->num_iwarp_msix = 1;
11102                         } else {
11103                                 pf->num_lan_msix = 2;
11104                         }
11105                         break;
11106                 default:
11107                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11108                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
11109                                                  iwarp_requested);
11110                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
11111                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
11112                         } else {
11113                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
11114                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
11115                         }
11116                         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
11117                                 pf->num_fdsb_msix = 1;
11118                                 vec--;
11119                         }
11120                         pf->num_lan_msix = min_t(int,
11121                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
11122                                                               pf->num_lan_msix);
11123                         pf->num_lan_qps = pf->num_lan_msix;
11124                         break;
11125                 }
11126         }
11127
11128         if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
11129             (pf->num_fdsb_msix == 0)) {
11130                 dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
11131                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
11132                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11133         }
11134         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
11135             (pf->num_vmdq_msix == 0)) {
11136                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
11137                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
11138         }
11139
11140         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
11141             (pf->num_iwarp_msix == 0)) {
11142                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
11143                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11144         }
11145         i40e_debug(&pf->hw, I40E_DEBUG_INIT,
11146                    "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
11147                    pf->num_lan_msix,
11148                    pf->num_vmdq_msix * pf->num_vmdq_vsis,
11149                    pf->num_fdsb_msix,
11150                    pf->num_iwarp_msix);
11151
11152         return v_actual;
11153 }
11154
11155 /**
11156  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
11157  * @vsi: the VSI being configured
11158  * @v_idx: index of the vector in the vsi struct
11159  * @cpu: cpu to be used on affinity_mask
11160  *
11161  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
11162  **/
11163 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
11164 {
11165         struct i40e_q_vector *q_vector;
11166
11167         /* allocate q_vector */
11168         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
11169         if (!q_vector)
11170                 return -ENOMEM;
11171
11172         q_vector->vsi = vsi;
11173         q_vector->v_idx = v_idx;
11174         cpumask_copy(&q_vector->affinity_mask, cpu_possible_mask);
11175
11176         if (vsi->netdev)
11177                 netif_napi_add(vsi->netdev, &q_vector->napi,
11178                                i40e_napi_poll, NAPI_POLL_WEIGHT);
11179
11180         /* tie q_vector and vsi together */
11181         vsi->q_vectors[v_idx] = q_vector;
11182
11183         return 0;
11184 }
11185
11186 /**
11187  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
11188  * @vsi: the VSI being configured
11189  *
11190  * We allocate one q_vector per queue interrupt.  If allocation fails we
11191  * return -ENOMEM.
11192  **/
11193 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
11194 {
11195         struct i40e_pf *pf = vsi->back;
11196         int err, v_idx, num_q_vectors, current_cpu;
11197
11198         /* if not MSIX, give the one vector only to the LAN VSI */
11199         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
11200                 num_q_vectors = vsi->num_q_vectors;
11201         else if (vsi == pf->vsi[pf->lan_vsi])
11202                 num_q_vectors = 1;
11203         else
11204                 return -EINVAL;
11205
11206         current_cpu = cpumask_first(cpu_online_mask);
11207
11208         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
11209                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
11210                 if (err)
11211                         goto err_out;
11212                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
11213                 if (unlikely(current_cpu >= nr_cpu_ids))
11214                         current_cpu = cpumask_first(cpu_online_mask);
11215         }
11216
11217         return 0;
11218
11219 err_out:
11220         while (v_idx--)
11221                 i40e_free_q_vector(vsi, v_idx);
11222
11223         return err;
11224 }
11225
11226 /**
11227  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
11228  * @pf: board private structure to initialize
11229  **/
11230 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
11231 {
11232         int vectors = 0;
11233         ssize_t size;
11234
11235         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11236                 vectors = i40e_init_msix(pf);
11237                 if (vectors < 0) {
11238                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
11239                                        I40E_FLAG_IWARP_ENABLED  |
11240                                        I40E_FLAG_RSS_ENABLED    |
11241                                        I40E_FLAG_DCB_CAPABLE    |
11242                                        I40E_FLAG_DCB_ENABLED    |
11243                                        I40E_FLAG_SRIOV_ENABLED  |
11244                                        I40E_FLAG_FD_SB_ENABLED  |
11245                                        I40E_FLAG_FD_ATR_ENABLED |
11246                                        I40E_FLAG_VMDQ_ENABLED);
11247                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
11248
11249                         /* rework the queue expectations without MSIX */
11250                         i40e_determine_queue_usage(pf);
11251                 }
11252         }
11253
11254         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11255             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
11256                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
11257                 vectors = pci_enable_msi(pf->pdev);
11258                 if (vectors < 0) {
11259                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
11260                                  vectors);
11261                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
11262                 }
11263                 vectors = 1;  /* one MSI or Legacy vector */
11264         }
11265
11266         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
11267                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
11268
11269         /* set up vector assignment tracking */
11270         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
11271         pf->irq_pile = kzalloc(size, GFP_KERNEL);
11272         if (!pf->irq_pile)
11273                 return -ENOMEM;
11274
11275         pf->irq_pile->num_entries = vectors;
11276         pf->irq_pile->search_hint = 0;
11277
11278         /* track first vector for misc interrupts, ignore return */
11279         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
11280
11281         return 0;
11282 }
11283
11284 /**
11285  * i40e_restore_interrupt_scheme - Restore the interrupt scheme
11286  * @pf: private board data structure
11287  *
11288  * Restore the interrupt scheme that was cleared when we suspended the
11289  * device. This should be called during resume to re-allocate the q_vectors
11290  * and reacquire IRQs.
11291  */
11292 static int i40e_restore_interrupt_scheme(struct i40e_pf *pf)
11293 {
11294         int err, i;
11295
11296         /* We cleared the MSI and MSI-X flags when disabling the old interrupt
11297          * scheme. We need to re-enabled them here in order to attempt to
11298          * re-acquire the MSI or MSI-X vectors
11299          */
11300         pf->flags |= (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
11301
11302         err = i40e_init_interrupt_scheme(pf);
11303         if (err)
11304                 return err;
11305
11306         /* Now that we've re-acquired IRQs, we need to remap the vectors and
11307          * rings together again.
11308          */
11309         for (i = 0; i < pf->num_alloc_vsi; i++) {
11310                 if (pf->vsi[i]) {
11311                         err = i40e_vsi_alloc_q_vectors(pf->vsi[i]);
11312                         if (err)
11313                                 goto err_unwind;
11314                         i40e_vsi_map_rings_to_vectors(pf->vsi[i]);
11315                 }
11316         }
11317
11318         err = i40e_setup_misc_vector(pf);
11319         if (err)
11320                 goto err_unwind;
11321
11322         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
11323                 i40e_client_update_msix_info(pf);
11324
11325         return 0;
11326
11327 err_unwind:
11328         while (i--) {
11329                 if (pf->vsi[i])
11330                         i40e_vsi_free_q_vectors(pf->vsi[i]);
11331         }
11332
11333         return err;
11334 }
11335
11336 /**
11337  * i40e_setup_misc_vector_for_recovery_mode - Setup the misc vector to handle
11338  * non queue events in recovery mode
11339  * @pf: board private structure
11340  *
11341  * This sets up the handler for MSIX 0 or MSI/legacy, which is used to manage
11342  * the non-queue interrupts, e.g. AdminQ and errors in recovery mode.
11343  * This is handled differently than in recovery mode since no Tx/Rx resources
11344  * are being allocated.
11345  **/
11346 static int i40e_setup_misc_vector_for_recovery_mode(struct i40e_pf *pf)
11347 {
11348         int err;
11349
11350         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11351                 err = i40e_setup_misc_vector(pf);
11352
11353                 if (err) {
11354                         dev_info(&pf->pdev->dev,
11355                                  "MSI-X misc vector request failed, error %d\n",
11356                                  err);
11357                         return err;
11358                 }
11359         } else {
11360                 u32 flags = pf->flags & I40E_FLAG_MSI_ENABLED ? 0 : IRQF_SHARED;
11361
11362                 err = request_irq(pf->pdev->irq, i40e_intr, flags,
11363                                   pf->int_name, pf);
11364
11365                 if (err) {
11366                         dev_info(&pf->pdev->dev,
11367                                  "MSI/legacy misc vector request failed, error %d\n",
11368                                  err);
11369                         return err;
11370                 }
11371                 i40e_enable_misc_int_causes(pf);
11372                 i40e_irq_dynamic_enable_icr0(pf);
11373         }
11374
11375         return 0;
11376 }
11377
11378 /**
11379  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
11380  * @pf: board private structure
11381  *
11382  * This sets up the handler for MSIX 0, which is used to manage the
11383  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
11384  * when in MSI or Legacy interrupt mode.
11385  **/
11386 static int i40e_setup_misc_vector(struct i40e_pf *pf)
11387 {
11388         struct i40e_hw *hw = &pf->hw;
11389         int err = 0;
11390
11391         /* Only request the IRQ once, the first time through. */
11392         if (!test_and_set_bit(__I40E_MISC_IRQ_REQUESTED, pf->state)) {
11393                 err = request_irq(pf->msix_entries[0].vector,
11394                                   i40e_intr, 0, pf->int_name, pf);
11395                 if (err) {
11396                         clear_bit(__I40E_MISC_IRQ_REQUESTED, pf->state);
11397                         dev_info(&pf->pdev->dev,
11398                                  "request_irq for %s failed: %d\n",
11399                                  pf->int_name, err);
11400                         return -EFAULT;
11401                 }
11402         }
11403
11404         i40e_enable_misc_int_causes(pf);
11405
11406         /* associate no queues to the misc vector */
11407         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
11408         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K >> 1);
11409
11410         i40e_flush(hw);
11411
11412         i40e_irq_dynamic_enable_icr0(pf);
11413
11414         return err;
11415 }
11416
11417 /**
11418  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
11419  * @vsi: Pointer to vsi structure
11420  * @seed: Buffter to store the hash keys
11421  * @lut: Buffer to store the lookup table entries
11422  * @lut_size: Size of buffer to store the lookup table entries
11423  *
11424  * Return 0 on success, negative on failure
11425  */
11426 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
11427                            u8 *lut, u16 lut_size)
11428 {
11429         struct i40e_pf *pf = vsi->back;
11430         struct i40e_hw *hw = &pf->hw;
11431         int ret = 0;
11432
11433         if (seed) {
11434                 ret = i40e_aq_get_rss_key(hw, vsi->id,
11435                         (struct i40e_aqc_get_set_rss_key_data *)seed);
11436                 if (ret) {
11437                         dev_info(&pf->pdev->dev,
11438                                  "Cannot get RSS key, err %s aq_err %s\n",
11439                                  i40e_stat_str(&pf->hw, ret),
11440                                  i40e_aq_str(&pf->hw,
11441                                              pf->hw.aq.asq_last_status));
11442                         return ret;
11443                 }
11444         }
11445
11446         if (lut) {
11447                 bool pf_lut = vsi->type == I40E_VSI_MAIN;
11448
11449                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
11450                 if (ret) {
11451                         dev_info(&pf->pdev->dev,
11452                                  "Cannot get RSS lut, err %s aq_err %s\n",
11453                                  i40e_stat_str(&pf->hw, ret),
11454                                  i40e_aq_str(&pf->hw,
11455                                              pf->hw.aq.asq_last_status));
11456                         return ret;
11457                 }
11458         }
11459
11460         return ret;
11461 }
11462
11463 /**
11464  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
11465  * @vsi: Pointer to vsi structure
11466  * @seed: RSS hash seed
11467  * @lut: Lookup table
11468  * @lut_size: Lookup table size
11469  *
11470  * Returns 0 on success, negative on failure
11471  **/
11472 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
11473                                const u8 *lut, u16 lut_size)
11474 {
11475         struct i40e_pf *pf = vsi->back;
11476         struct i40e_hw *hw = &pf->hw;
11477         u16 vf_id = vsi->vf_id;
11478         u8 i;
11479
11480         /* Fill out hash function seed */
11481         if (seed) {
11482                 u32 *seed_dw = (u32 *)seed;
11483
11484                 if (vsi->type == I40E_VSI_MAIN) {
11485                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11486                                 wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
11487                 } else if (vsi->type == I40E_VSI_SRIOV) {
11488                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
11489                                 wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
11490                 } else {
11491                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
11492                 }
11493         }
11494
11495         if (lut) {
11496                 u32 *lut_dw = (u32 *)lut;
11497
11498                 if (vsi->type == I40E_VSI_MAIN) {
11499                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
11500                                 return -EINVAL;
11501                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11502                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
11503                 } else if (vsi->type == I40E_VSI_SRIOV) {
11504                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
11505                                 return -EINVAL;
11506                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
11507                                 wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
11508                 } else {
11509                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
11510                 }
11511         }
11512         i40e_flush(hw);
11513
11514         return 0;
11515 }
11516
11517 /**
11518  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
11519  * @vsi: Pointer to VSI structure
11520  * @seed: Buffer to store the keys
11521  * @lut: Buffer to store the lookup table entries
11522  * @lut_size: Size of buffer to store the lookup table entries
11523  *
11524  * Returns 0 on success, negative on failure
11525  */
11526 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
11527                             u8 *lut, u16 lut_size)
11528 {
11529         struct i40e_pf *pf = vsi->back;
11530         struct i40e_hw *hw = &pf->hw;
11531         u16 i;
11532
11533         if (seed) {
11534                 u32 *seed_dw = (u32 *)seed;
11535
11536                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
11537                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
11538         }
11539         if (lut) {
11540                 u32 *lut_dw = (u32 *)lut;
11541
11542                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
11543                         return -EINVAL;
11544                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
11545                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
11546         }
11547
11548         return 0;
11549 }
11550
11551 /**
11552  * i40e_config_rss - Configure RSS keys and lut
11553  * @vsi: Pointer to VSI structure
11554  * @seed: RSS hash seed
11555  * @lut: Lookup table
11556  * @lut_size: Lookup table size
11557  *
11558  * Returns 0 on success, negative on failure
11559  */
11560 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11561 {
11562         struct i40e_pf *pf = vsi->back;
11563
11564         if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11565                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
11566         else
11567                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
11568 }
11569
11570 /**
11571  * i40e_get_rss - Get RSS keys and lut
11572  * @vsi: Pointer to VSI structure
11573  * @seed: Buffer to store the keys
11574  * @lut: Buffer to store the lookup table entries
11575  * @lut_size: Size of buffer to store the lookup table entries
11576  *
11577  * Returns 0 on success, negative on failure
11578  */
11579 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
11580 {
11581         struct i40e_pf *pf = vsi->back;
11582
11583         if (pf->hw_features & I40E_HW_RSS_AQ_CAPABLE)
11584                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
11585         else
11586                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
11587 }
11588
11589 /**
11590  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
11591  * @pf: Pointer to board private structure
11592  * @lut: Lookup table
11593  * @rss_table_size: Lookup table size
11594  * @rss_size: Range of queue number for hashing
11595  */
11596 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
11597                        u16 rss_table_size, u16 rss_size)
11598 {
11599         u16 i;
11600
11601         for (i = 0; i < rss_table_size; i++)
11602                 lut[i] = i % rss_size;
11603 }
11604
11605 /**
11606  * i40e_pf_config_rss - Prepare for RSS if used
11607  * @pf: board private structure
11608  **/
11609 static int i40e_pf_config_rss(struct i40e_pf *pf)
11610 {
11611         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11612         u8 seed[I40E_HKEY_ARRAY_SIZE];
11613         u8 *lut;
11614         struct i40e_hw *hw = &pf->hw;
11615         u32 reg_val;
11616         u64 hena;
11617         int ret;
11618
11619         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
11620         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
11621                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
11622         hena |= i40e_pf_get_default_rss_hena(pf);
11623
11624         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
11625         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
11626
11627         /* Determine the RSS table size based on the hardware capabilities */
11628         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
11629         reg_val = (pf->rss_table_size == 512) ?
11630                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
11631                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
11632         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
11633
11634         /* Determine the RSS size of the VSI */
11635         if (!vsi->rss_size) {
11636                 u16 qcount;
11637                 /* If the firmware does something weird during VSI init, we
11638                  * could end up with zero TCs. Check for that to avoid
11639                  * divide-by-zero. It probably won't pass traffic, but it also
11640                  * won't panic.
11641                  */
11642                 qcount = vsi->num_queue_pairs /
11643                          (vsi->tc_config.numtc ? vsi->tc_config.numtc : 1);
11644                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11645         }
11646         if (!vsi->rss_size)
11647                 return -EINVAL;
11648
11649         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
11650         if (!lut)
11651                 return -ENOMEM;
11652
11653         /* Use user configured lut if there is one, otherwise use default */
11654         if (vsi->rss_lut_user)
11655                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
11656         else
11657                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
11658
11659         /* Use user configured hash key if there is one, otherwise
11660          * use default.
11661          */
11662         if (vsi->rss_hkey_user)
11663                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
11664         else
11665                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
11666         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
11667         kfree(lut);
11668
11669         return ret;
11670 }
11671
11672 /**
11673  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
11674  * @pf: board private structure
11675  * @queue_count: the requested queue count for rss.
11676  *
11677  * returns 0 if rss is not enabled, if enabled returns the final rss queue
11678  * count which may be different from the requested queue count.
11679  * Note: expects to be called while under rtnl_lock()
11680  **/
11681 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
11682 {
11683         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
11684         int new_rss_size;
11685
11686         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
11687                 return 0;
11688
11689         queue_count = min_t(int, queue_count, num_online_cpus());
11690         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
11691
11692         if (queue_count != vsi->num_queue_pairs) {
11693                 u16 qcount;
11694
11695                 vsi->req_queue_pairs = queue_count;
11696                 i40e_prep_for_reset(pf, true);
11697
11698                 pf->alloc_rss_size = new_rss_size;
11699
11700                 i40e_reset_and_rebuild(pf, true, true);
11701
11702                 /* Discard the user configured hash keys and lut, if less
11703                  * queues are enabled.
11704                  */
11705                 if (queue_count < vsi->rss_size) {
11706                         i40e_clear_rss_config_user(vsi);
11707                         dev_dbg(&pf->pdev->dev,
11708                                 "discard user configured hash keys and lut\n");
11709                 }
11710
11711                 /* Reset vsi->rss_size, as number of enabled queues changed */
11712                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
11713                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
11714
11715                 i40e_pf_config_rss(pf);
11716         }
11717         dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
11718                  vsi->req_queue_pairs, pf->rss_size_max);
11719         return pf->alloc_rss_size;
11720 }
11721
11722 /**
11723  * i40e_get_partition_bw_setting - Retrieve BW settings for this PF partition
11724  * @pf: board private structure
11725  **/
11726 i40e_status i40e_get_partition_bw_setting(struct i40e_pf *pf)
11727 {
11728         i40e_status status;
11729         bool min_valid, max_valid;
11730         u32 max_bw, min_bw;
11731
11732         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
11733                                            &min_valid, &max_valid);
11734
11735         if (!status) {
11736                 if (min_valid)
11737                         pf->min_bw = min_bw;
11738                 if (max_valid)
11739                         pf->max_bw = max_bw;
11740         }
11741
11742         return status;
11743 }
11744
11745 /**
11746  * i40e_set_partition_bw_setting - Set BW settings for this PF partition
11747  * @pf: board private structure
11748  **/
11749 i40e_status i40e_set_partition_bw_setting(struct i40e_pf *pf)
11750 {
11751         struct i40e_aqc_configure_partition_bw_data bw_data;
11752         i40e_status status;
11753
11754         /* Set the valid bit for this PF */
11755         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
11756         bw_data.max_bw[pf->hw.pf_id] = pf->max_bw & I40E_ALT_BW_VALUE_MASK;
11757         bw_data.min_bw[pf->hw.pf_id] = pf->min_bw & I40E_ALT_BW_VALUE_MASK;
11758
11759         /* Set the new bandwidths */
11760         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
11761
11762         return status;
11763 }
11764
11765 /**
11766  * i40e_commit_partition_bw_setting - Commit BW settings for this PF partition
11767  * @pf: board private structure
11768  **/
11769 i40e_status i40e_commit_partition_bw_setting(struct i40e_pf *pf)
11770 {
11771         /* Commit temporary BW setting to permanent NVM image */
11772         enum i40e_admin_queue_err last_aq_status;
11773         i40e_status ret;
11774         u16 nvm_word;
11775
11776         if (pf->hw.partition_id != 1) {
11777                 dev_info(&pf->pdev->dev,
11778                          "Commit BW only works on partition 1! This is partition %d",
11779                          pf->hw.partition_id);
11780                 ret = I40E_NOT_SUPPORTED;
11781                 goto bw_commit_out;
11782         }
11783
11784         /* Acquire NVM for read access */
11785         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
11786         last_aq_status = pf->hw.aq.asq_last_status;
11787         if (ret) {
11788                 dev_info(&pf->pdev->dev,
11789                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
11790                          i40e_stat_str(&pf->hw, ret),
11791                          i40e_aq_str(&pf->hw, last_aq_status));
11792                 goto bw_commit_out;
11793         }
11794
11795         /* Read word 0x10 of NVM - SW compatibility word 1 */
11796         ret = i40e_aq_read_nvm(&pf->hw,
11797                                I40E_SR_NVM_CONTROL_WORD,
11798                                0x10, sizeof(nvm_word), &nvm_word,
11799                                false, NULL);
11800         /* Save off last admin queue command status before releasing
11801          * the NVM
11802          */
11803         last_aq_status = pf->hw.aq.asq_last_status;
11804         i40e_release_nvm(&pf->hw);
11805         if (ret) {
11806                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
11807                          i40e_stat_str(&pf->hw, ret),
11808                          i40e_aq_str(&pf->hw, last_aq_status));
11809                 goto bw_commit_out;
11810         }
11811
11812         /* Wait a bit for NVM release to complete */
11813         msleep(50);
11814
11815         /* Acquire NVM for write access */
11816         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
11817         last_aq_status = pf->hw.aq.asq_last_status;
11818         if (ret) {
11819                 dev_info(&pf->pdev->dev,
11820                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
11821                          i40e_stat_str(&pf->hw, ret),
11822                          i40e_aq_str(&pf->hw, last_aq_status));
11823                 goto bw_commit_out;
11824         }
11825         /* Write it back out unchanged to initiate update NVM,
11826          * which will force a write of the shadow (alt) RAM to
11827          * the NVM - thus storing the bandwidth values permanently.
11828          */
11829         ret = i40e_aq_update_nvm(&pf->hw,
11830                                  I40E_SR_NVM_CONTROL_WORD,
11831                                  0x10, sizeof(nvm_word),
11832                                  &nvm_word, true, 0, NULL);
11833         /* Save off last admin queue command status before releasing
11834          * the NVM
11835          */
11836         last_aq_status = pf->hw.aq.asq_last_status;
11837         i40e_release_nvm(&pf->hw);
11838         if (ret)
11839                 dev_info(&pf->pdev->dev,
11840                          "BW settings NOT SAVED, err %s aq_err %s\n",
11841                          i40e_stat_str(&pf->hw, ret),
11842                          i40e_aq_str(&pf->hw, last_aq_status));
11843 bw_commit_out:
11844
11845         return ret;
11846 }
11847
11848 /**
11849  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
11850  * @pf: board private structure to initialize
11851  *
11852  * i40e_sw_init initializes the Adapter private data structure.
11853  * Fields are initialized based on PCI device information and
11854  * OS network device settings (MTU size).
11855  **/
11856 static int i40e_sw_init(struct i40e_pf *pf)
11857 {
11858         int err = 0;
11859         int size;
11860
11861         /* Set default capability flags */
11862         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
11863                     I40E_FLAG_MSI_ENABLED     |
11864                     I40E_FLAG_MSIX_ENABLED;
11865
11866         /* Set default ITR */
11867         pf->rx_itr_default = I40E_ITR_RX_DEF;
11868         pf->tx_itr_default = I40E_ITR_TX_DEF;
11869
11870         /* Depending on PF configurations, it is possible that the RSS
11871          * maximum might end up larger than the available queues
11872          */
11873         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
11874         pf->alloc_rss_size = 1;
11875         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
11876         pf->rss_size_max = min_t(int, pf->rss_size_max,
11877                                  pf->hw.func_caps.num_tx_qp);
11878         if (pf->hw.func_caps.rss) {
11879                 pf->flags |= I40E_FLAG_RSS_ENABLED;
11880                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
11881                                            num_online_cpus());
11882         }
11883
11884         /* MFP mode enabled */
11885         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
11886                 pf->flags |= I40E_FLAG_MFP_ENABLED;
11887                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
11888                 if (i40e_get_partition_bw_setting(pf)) {
11889                         dev_warn(&pf->pdev->dev,
11890                                  "Could not get partition bw settings\n");
11891                 } else {
11892                         dev_info(&pf->pdev->dev,
11893                                  "Partition BW Min = %8.8x, Max = %8.8x\n",
11894                                  pf->min_bw, pf->max_bw);
11895
11896                         /* nudge the Tx scheduler */
11897                         i40e_set_partition_bw_setting(pf);
11898                 }
11899         }
11900
11901         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
11902             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
11903                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
11904                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
11905                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
11906                     pf->hw.num_partitions > 1)
11907                         dev_info(&pf->pdev->dev,
11908                                  "Flow Director Sideband mode Disabled in MFP mode\n");
11909                 else
11910                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
11911                 pf->fdir_pf_filter_count =
11912                                  pf->hw.func_caps.fd_filters_guaranteed;
11913                 pf->hw.fdir_shared_filter_count =
11914                                  pf->hw.func_caps.fd_filters_best_effort;
11915         }
11916
11917         if (pf->hw.mac.type == I40E_MAC_X722) {
11918                 pf->hw_features |= (I40E_HW_RSS_AQ_CAPABLE |
11919                                     I40E_HW_128_QP_RSS_CAPABLE |
11920                                     I40E_HW_ATR_EVICT_CAPABLE |
11921                                     I40E_HW_WB_ON_ITR_CAPABLE |
11922                                     I40E_HW_MULTIPLE_TCP_UDP_RSS_PCTYPE |
11923                                     I40E_HW_NO_PCI_LINK_CHECK |
11924                                     I40E_HW_USE_SET_LLDP_MIB |
11925                                     I40E_HW_GENEVE_OFFLOAD_CAPABLE |
11926                                     I40E_HW_PTP_L4_CAPABLE |
11927                                     I40E_HW_WOL_MC_MAGIC_PKT_WAKE |
11928                                     I40E_HW_OUTER_UDP_CSUM_CAPABLE);
11929
11930 #define I40E_FDEVICT_PCTYPE_DEFAULT 0xc03
11931                 if (rd32(&pf->hw, I40E_GLQF_FDEVICTENA(1)) !=
11932                     I40E_FDEVICT_PCTYPE_DEFAULT) {
11933                         dev_warn(&pf->pdev->dev,
11934                                  "FD EVICT PCTYPES are not right, disable FD HW EVICT\n");
11935                         pf->hw_features &= ~I40E_HW_ATR_EVICT_CAPABLE;
11936                 }
11937         } else if ((pf->hw.aq.api_maj_ver > 1) ||
11938                    ((pf->hw.aq.api_maj_ver == 1) &&
11939                     (pf->hw.aq.api_min_ver > 4))) {
11940                 /* Supported in FW API version higher than 1.4 */
11941                 pf->hw_features |= I40E_HW_GENEVE_OFFLOAD_CAPABLE;
11942         }
11943
11944         /* Enable HW ATR eviction if possible */
11945         if (pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE)
11946                 pf->flags |= I40E_FLAG_HW_ATR_EVICT_ENABLED;
11947
11948         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11949             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
11950             (pf->hw.aq.fw_maj_ver < 4))) {
11951                 pf->hw_features |= I40E_HW_RESTART_AUTONEG;
11952                 /* No DCB support  for FW < v4.33 */
11953                 pf->hw_features |= I40E_HW_NO_DCB_SUPPORT;
11954         }
11955
11956         /* Disable FW LLDP if FW < v4.3 */
11957         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11958             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
11959             (pf->hw.aq.fw_maj_ver < 4)))
11960                 pf->hw_features |= I40E_HW_STOP_FW_LLDP;
11961
11962         /* Use the FW Set LLDP MIB API if FW > v4.40 */
11963         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
11964             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
11965             (pf->hw.aq.fw_maj_ver >= 5)))
11966                 pf->hw_features |= I40E_HW_USE_SET_LLDP_MIB;
11967
11968         /* Enable PTP L4 if FW > v6.0 */
11969         if (pf->hw.mac.type == I40E_MAC_XL710 &&
11970             pf->hw.aq.fw_maj_ver >= 6)
11971                 pf->hw_features |= I40E_HW_PTP_L4_CAPABLE;
11972
11973         if (pf->hw.func_caps.vmdq && num_online_cpus() != 1) {
11974                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
11975                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
11976                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
11977         }
11978
11979         if (pf->hw.func_caps.iwarp && num_online_cpus() != 1) {
11980                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
11981                 /* IWARP needs one extra vector for CQP just like MISC.*/
11982                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
11983         }
11984         /* Stopping FW LLDP engine is supported on XL710 and X722
11985          * starting from FW versions determined in i40e_init_adminq.
11986          * Stopping the FW LLDP engine is not supported on XL710
11987          * if NPAR is functioning so unset this hw flag in this case.
11988          */
11989         if (pf->hw.mac.type == I40E_MAC_XL710 &&
11990             pf->hw.func_caps.npar_enable &&
11991             (pf->hw.flags & I40E_HW_FLAG_FW_LLDP_STOPPABLE))
11992                 pf->hw.flags &= ~I40E_HW_FLAG_FW_LLDP_STOPPABLE;
11993
11994 #ifdef CONFIG_PCI_IOV
11995         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
11996                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
11997                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
11998                 pf->num_req_vfs = min_t(int,
11999                                         pf->hw.func_caps.num_vfs,
12000                                         I40E_MAX_VF_COUNT);
12001         }
12002 #endif /* CONFIG_PCI_IOV */
12003         pf->eeprom_version = 0xDEAD;
12004         pf->lan_veb = I40E_NO_VEB;
12005         pf->lan_vsi = I40E_NO_VSI;
12006
12007         /* By default FW has this off for performance reasons */
12008         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
12009
12010         /* set up queue assignment tracking */
12011         size = sizeof(struct i40e_lump_tracking)
12012                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
12013         pf->qp_pile = kzalloc(size, GFP_KERNEL);
12014         if (!pf->qp_pile) {
12015                 err = -ENOMEM;
12016                 goto sw_init_done;
12017         }
12018         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
12019         pf->qp_pile->search_hint = 0;
12020
12021         pf->tx_timeout_recovery_level = 1;
12022
12023         mutex_init(&pf->switch_mutex);
12024
12025 sw_init_done:
12026         return err;
12027 }
12028
12029 /**
12030  * i40e_set_ntuple - set the ntuple feature flag and take action
12031  * @pf: board private structure to initialize
12032  * @features: the feature set that the stack is suggesting
12033  *
12034  * returns a bool to indicate if reset needs to happen
12035  **/
12036 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
12037 {
12038         bool need_reset = false;
12039
12040         /* Check if Flow Director n-tuple support was enabled or disabled.  If
12041          * the state changed, we need to reset.
12042          */
12043         if (features & NETIF_F_NTUPLE) {
12044                 /* Enable filters and mark for reset */
12045                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
12046                         need_reset = true;
12047                 /* enable FD_SB only if there is MSI-X vector and no cloud
12048                  * filters exist
12049                  */
12050                 if (pf->num_fdsb_msix > 0 && !pf->num_cloud_filters) {
12051                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
12052                         pf->flags &= ~I40E_FLAG_FD_SB_INACTIVE;
12053                 }
12054         } else {
12055                 /* turn off filters, mark for reset and clear SW filter list */
12056                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
12057                         need_reset = true;
12058                         i40e_fdir_filter_exit(pf);
12059                 }
12060                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
12061                 clear_bit(__I40E_FD_SB_AUTO_DISABLED, pf->state);
12062                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
12063
12064                 /* reset fd counters */
12065                 pf->fd_add_err = 0;
12066                 pf->fd_atr_cnt = 0;
12067                 /* if ATR was auto disabled it can be re-enabled. */
12068                 if (test_and_clear_bit(__I40E_FD_ATR_AUTO_DISABLED, pf->state))
12069                         if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
12070                             (I40E_DEBUG_FD & pf->hw.debug_mask))
12071                                 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
12072         }
12073         return need_reset;
12074 }
12075
12076 /**
12077  * i40e_clear_rss_lut - clear the rx hash lookup table
12078  * @vsi: the VSI being configured
12079  **/
12080 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
12081 {
12082         struct i40e_pf *pf = vsi->back;
12083         struct i40e_hw *hw = &pf->hw;
12084         u16 vf_id = vsi->vf_id;
12085         u8 i;
12086
12087         if (vsi->type == I40E_VSI_MAIN) {
12088                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
12089                         wr32(hw, I40E_PFQF_HLUT(i), 0);
12090         } else if (vsi->type == I40E_VSI_SRIOV) {
12091                 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
12092                         i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
12093         } else {
12094                 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
12095         }
12096 }
12097
12098 /**
12099  * i40e_set_features - set the netdev feature flags
12100  * @netdev: ptr to the netdev being adjusted
12101  * @features: the feature set that the stack is suggesting
12102  * Note: expects to be called while under rtnl_lock()
12103  **/
12104 static int i40e_set_features(struct net_device *netdev,
12105                              netdev_features_t features)
12106 {
12107         struct i40e_netdev_priv *np = netdev_priv(netdev);
12108         struct i40e_vsi *vsi = np->vsi;
12109         struct i40e_pf *pf = vsi->back;
12110         bool need_reset;
12111
12112         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
12113                 i40e_pf_config_rss(pf);
12114         else if (!(features & NETIF_F_RXHASH) &&
12115                  netdev->features & NETIF_F_RXHASH)
12116                 i40e_clear_rss_lut(vsi);
12117
12118         if (features & NETIF_F_HW_VLAN_CTAG_RX)
12119                 i40e_vlan_stripping_enable(vsi);
12120         else
12121                 i40e_vlan_stripping_disable(vsi);
12122
12123         if (!(features & NETIF_F_HW_TC) && pf->num_cloud_filters) {
12124                 dev_err(&pf->pdev->dev,
12125                         "Offloaded tc filters active, can't turn hw_tc_offload off");
12126                 return -EINVAL;
12127         }
12128
12129         if (!(features & NETIF_F_HW_L2FW_DOFFLOAD) && vsi->macvlan_cnt)
12130                 i40e_del_all_macvlans(vsi);
12131
12132         need_reset = i40e_set_ntuple(pf, features);
12133
12134         if (need_reset)
12135                 i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12136
12137         return 0;
12138 }
12139
12140 /**
12141  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
12142  * @pf: board private structure
12143  * @port: The UDP port to look up
12144  *
12145  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
12146  **/
12147 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
12148 {
12149         u8 i;
12150
12151         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
12152                 /* Do not report ports with pending deletions as
12153                  * being available.
12154                  */
12155                 if (!port && (pf->pending_udp_bitmap & BIT_ULL(i)))
12156                         continue;
12157                 if (pf->udp_ports[i].port == port)
12158                         return i;
12159         }
12160
12161         return i;
12162 }
12163
12164 /**
12165  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
12166  * @netdev: This physical port's netdev
12167  * @ti: Tunnel endpoint information
12168  **/
12169 static void i40e_udp_tunnel_add(struct net_device *netdev,
12170                                 struct udp_tunnel_info *ti)
12171 {
12172         struct i40e_netdev_priv *np = netdev_priv(netdev);
12173         struct i40e_vsi *vsi = np->vsi;
12174         struct i40e_pf *pf = vsi->back;
12175         u16 port = ntohs(ti->port);
12176         u8 next_idx;
12177         u8 idx;
12178
12179         idx = i40e_get_udp_port_idx(pf, port);
12180
12181         /* Check if port already exists */
12182         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12183                 netdev_info(netdev, "port %d already offloaded\n", port);
12184                 return;
12185         }
12186
12187         /* Now check if there is space to add the new port */
12188         next_idx = i40e_get_udp_port_idx(pf, 0);
12189
12190         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
12191                 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
12192                             port);
12193                 return;
12194         }
12195
12196         switch (ti->type) {
12197         case UDP_TUNNEL_TYPE_VXLAN:
12198                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
12199                 break;
12200         case UDP_TUNNEL_TYPE_GENEVE:
12201                 if (!(pf->hw_features & I40E_HW_GENEVE_OFFLOAD_CAPABLE))
12202                         return;
12203                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
12204                 break;
12205         default:
12206                 return;
12207         }
12208
12209         /* New port: add it and mark its index in the bitmap */
12210         pf->udp_ports[next_idx].port = port;
12211         pf->udp_ports[next_idx].filter_index = I40E_UDP_PORT_INDEX_UNUSED;
12212         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
12213         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12214 }
12215
12216 /**
12217  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
12218  * @netdev: This physical port's netdev
12219  * @ti: Tunnel endpoint information
12220  **/
12221 static void i40e_udp_tunnel_del(struct net_device *netdev,
12222                                 struct udp_tunnel_info *ti)
12223 {
12224         struct i40e_netdev_priv *np = netdev_priv(netdev);
12225         struct i40e_vsi *vsi = np->vsi;
12226         struct i40e_pf *pf = vsi->back;
12227         u16 port = ntohs(ti->port);
12228         u8 idx;
12229
12230         idx = i40e_get_udp_port_idx(pf, port);
12231
12232         /* Check if port already exists */
12233         if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
12234                 goto not_found;
12235
12236         switch (ti->type) {
12237         case UDP_TUNNEL_TYPE_VXLAN:
12238                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
12239                         goto not_found;
12240                 break;
12241         case UDP_TUNNEL_TYPE_GENEVE:
12242                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
12243                         goto not_found;
12244                 break;
12245         default:
12246                 goto not_found;
12247         }
12248
12249         /* if port exists, set it to 0 (mark for deletion)
12250          * and make it pending
12251          */
12252         pf->udp_ports[idx].port = 0;
12253
12254         /* Toggle pending bit instead of setting it. This way if we are
12255          * deleting a port that has yet to be added we just clear the pending
12256          * bit and don't have to worry about it.
12257          */
12258         pf->pending_udp_bitmap ^= BIT_ULL(idx);
12259         set_bit(__I40E_UDP_FILTER_SYNC_PENDING, pf->state);
12260
12261         return;
12262 not_found:
12263         netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
12264                     port);
12265 }
12266
12267 static int i40e_get_phys_port_id(struct net_device *netdev,
12268                                  struct netdev_phys_item_id *ppid)
12269 {
12270         struct i40e_netdev_priv *np = netdev_priv(netdev);
12271         struct i40e_pf *pf = np->vsi->back;
12272         struct i40e_hw *hw = &pf->hw;
12273
12274         if (!(pf->hw_features & I40E_HW_PORT_ID_VALID))
12275                 return -EOPNOTSUPP;
12276
12277         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
12278         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
12279
12280         return 0;
12281 }
12282
12283 /**
12284  * i40e_ndo_fdb_add - add an entry to the hardware database
12285  * @ndm: the input from the stack
12286  * @tb: pointer to array of nladdr (unused)
12287  * @dev: the net device pointer
12288  * @addr: the MAC address entry being added
12289  * @vid: VLAN ID
12290  * @flags: instructions from stack about fdb operation
12291  */
12292 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
12293                             struct net_device *dev,
12294                             const unsigned char *addr, u16 vid,
12295                             u16 flags,
12296                             struct netlink_ext_ack *extack)
12297 {
12298         struct i40e_netdev_priv *np = netdev_priv(dev);
12299         struct i40e_pf *pf = np->vsi->back;
12300         int err = 0;
12301
12302         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
12303                 return -EOPNOTSUPP;
12304
12305         if (vid) {
12306                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
12307                 return -EINVAL;
12308         }
12309
12310         /* Hardware does not support aging addresses so if a
12311          * ndm_state is given only allow permanent addresses
12312          */
12313         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
12314                 netdev_info(dev, "FDB only supports static addresses\n");
12315                 return -EINVAL;
12316         }
12317
12318         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
12319                 err = dev_uc_add_excl(dev, addr);
12320         else if (is_multicast_ether_addr(addr))
12321                 err = dev_mc_add_excl(dev, addr);
12322         else
12323                 err = -EINVAL;
12324
12325         /* Only return duplicate errors if NLM_F_EXCL is set */
12326         if (err == -EEXIST && !(flags & NLM_F_EXCL))
12327                 err = 0;
12328
12329         return err;
12330 }
12331
12332 /**
12333  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
12334  * @dev: the netdev being configured
12335  * @nlh: RTNL message
12336  * @flags: bridge flags
12337  * @extack: netlink extended ack
12338  *
12339  * Inserts a new hardware bridge if not already created and
12340  * enables the bridging mode requested (VEB or VEPA). If the
12341  * hardware bridge has already been inserted and the request
12342  * is to change the mode then that requires a PF reset to
12343  * allow rebuild of the components with required hardware
12344  * bridge mode enabled.
12345  *
12346  * Note: expects to be called while under rtnl_lock()
12347  **/
12348 static int i40e_ndo_bridge_setlink(struct net_device *dev,
12349                                    struct nlmsghdr *nlh,
12350                                    u16 flags,
12351                                    struct netlink_ext_ack *extack)
12352 {
12353         struct i40e_netdev_priv *np = netdev_priv(dev);
12354         struct i40e_vsi *vsi = np->vsi;
12355         struct i40e_pf *pf = vsi->back;
12356         struct i40e_veb *veb = NULL;
12357         struct nlattr *attr, *br_spec;
12358         int i, rem;
12359
12360         /* Only for PF VSI for now */
12361         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12362                 return -EOPNOTSUPP;
12363
12364         /* Find the HW bridge for PF VSI */
12365         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12366                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12367                         veb = pf->veb[i];
12368         }
12369
12370         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
12371
12372         nla_for_each_nested(attr, br_spec, rem) {
12373                 __u16 mode;
12374
12375                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
12376                         continue;
12377
12378                 mode = nla_get_u16(attr);
12379                 if ((mode != BRIDGE_MODE_VEPA) &&
12380                     (mode != BRIDGE_MODE_VEB))
12381                         return -EINVAL;
12382
12383                 /* Insert a new HW bridge */
12384                 if (!veb) {
12385                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
12386                                              vsi->tc_config.enabled_tc);
12387                         if (veb) {
12388                                 veb->bridge_mode = mode;
12389                                 i40e_config_bridge_mode(veb);
12390                         } else {
12391                                 /* No Bridge HW offload available */
12392                                 return -ENOENT;
12393                         }
12394                         break;
12395                 } else if (mode != veb->bridge_mode) {
12396                         /* Existing HW bridge but different mode needs reset */
12397                         veb->bridge_mode = mode;
12398                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
12399                         if (mode == BRIDGE_MODE_VEB)
12400                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
12401                         else
12402                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
12403                         i40e_do_reset(pf, I40E_PF_RESET_FLAG, true);
12404                         break;
12405                 }
12406         }
12407
12408         return 0;
12409 }
12410
12411 /**
12412  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
12413  * @skb: skb buff
12414  * @pid: process id
12415  * @seq: RTNL message seq #
12416  * @dev: the netdev being configured
12417  * @filter_mask: unused
12418  * @nlflags: netlink flags passed in
12419  *
12420  * Return the mode in which the hardware bridge is operating in
12421  * i.e VEB or VEPA.
12422  **/
12423 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
12424                                    struct net_device *dev,
12425                                    u32 __always_unused filter_mask,
12426                                    int nlflags)
12427 {
12428         struct i40e_netdev_priv *np = netdev_priv(dev);
12429         struct i40e_vsi *vsi = np->vsi;
12430         struct i40e_pf *pf = vsi->back;
12431         struct i40e_veb *veb = NULL;
12432         int i;
12433
12434         /* Only for PF VSI for now */
12435         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
12436                 return -EOPNOTSUPP;
12437
12438         /* Find the HW bridge for the PF VSI */
12439         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
12440                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
12441                         veb = pf->veb[i];
12442         }
12443
12444         if (!veb)
12445                 return 0;
12446
12447         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
12448                                        0, 0, nlflags, filter_mask, NULL);
12449 }
12450
12451 /**
12452  * i40e_features_check - Validate encapsulated packet conforms to limits
12453  * @skb: skb buff
12454  * @dev: This physical port's netdev
12455  * @features: Offload features that the stack believes apply
12456  **/
12457 static netdev_features_t i40e_features_check(struct sk_buff *skb,
12458                                              struct net_device *dev,
12459                                              netdev_features_t features)
12460 {
12461         size_t len;
12462
12463         /* No point in doing any of this if neither checksum nor GSO are
12464          * being requested for this frame.  We can rule out both by just
12465          * checking for CHECKSUM_PARTIAL
12466          */
12467         if (skb->ip_summed != CHECKSUM_PARTIAL)
12468                 return features;
12469
12470         /* We cannot support GSO if the MSS is going to be less than
12471          * 64 bytes.  If it is then we need to drop support for GSO.
12472          */
12473         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
12474                 features &= ~NETIF_F_GSO_MASK;
12475
12476         /* MACLEN can support at most 63 words */
12477         len = skb_network_header(skb) - skb->data;
12478         if (len & ~(63 * 2))
12479                 goto out_err;
12480
12481         /* IPLEN and EIPLEN can support at most 127 dwords */
12482         len = skb_transport_header(skb) - skb_network_header(skb);
12483         if (len & ~(127 * 4))
12484                 goto out_err;
12485
12486         if (skb->encapsulation) {
12487                 /* L4TUNLEN can support 127 words */
12488                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
12489                 if (len & ~(127 * 2))
12490                         goto out_err;
12491
12492                 /* IPLEN can support at most 127 dwords */
12493                 len = skb_inner_transport_header(skb) -
12494                       skb_inner_network_header(skb);
12495                 if (len & ~(127 * 4))
12496                         goto out_err;
12497         }
12498
12499         /* No need to validate L4LEN as TCP is the only protocol with a
12500          * a flexible value and we support all possible values supported
12501          * by TCP, which is at most 15 dwords
12502          */
12503
12504         return features;
12505 out_err:
12506         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
12507 }
12508
12509 /**
12510  * i40e_xdp_setup - add/remove an XDP program
12511  * @vsi: VSI to changed
12512  * @prog: XDP program
12513  **/
12514 static int i40e_xdp_setup(struct i40e_vsi *vsi,
12515                           struct bpf_prog *prog)
12516 {
12517         int frame_size = vsi->netdev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
12518         struct i40e_pf *pf = vsi->back;
12519         struct bpf_prog *old_prog;
12520         bool need_reset;
12521         int i;
12522
12523         /* Don't allow frames that span over multiple buffers */
12524         if (frame_size > vsi->rx_buf_len)
12525                 return -EINVAL;
12526
12527         if (!i40e_enabled_xdp_vsi(vsi) && !prog)
12528                 return 0;
12529
12530         /* When turning XDP on->off/off->on we reset and rebuild the rings. */
12531         need_reset = (i40e_enabled_xdp_vsi(vsi) != !!prog);
12532
12533         if (need_reset)
12534                 i40e_prep_for_reset(pf, true);
12535
12536         old_prog = xchg(&vsi->xdp_prog, prog);
12537
12538         if (need_reset) {
12539                 if (!prog)
12540                         /* Wait until ndo_xsk_wakeup completes. */
12541                         synchronize_rcu();
12542                 i40e_reset_and_rebuild(pf, true, true);
12543         }
12544
12545         for (i = 0; i < vsi->num_queue_pairs; i++)
12546                 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
12547
12548         if (old_prog)
12549                 bpf_prog_put(old_prog);
12550
12551         /* Kick start the NAPI context if there is an AF_XDP socket open
12552          * on that queue id. This so that receiving will start.
12553          */
12554         if (need_reset && prog)
12555                 for (i = 0; i < vsi->num_queue_pairs; i++)
12556                         if (vsi->xdp_rings[i]->xsk_umem)
12557                                 (void)i40e_xsk_wakeup(vsi->netdev, i,
12558                                                       XDP_WAKEUP_RX);
12559
12560         return 0;
12561 }
12562
12563 /**
12564  * i40e_enter_busy_conf - Enters busy config state
12565  * @vsi: vsi
12566  *
12567  * Returns 0 on success, <0 for failure.
12568  **/
12569 static int i40e_enter_busy_conf(struct i40e_vsi *vsi)
12570 {
12571         struct i40e_pf *pf = vsi->back;
12572         int timeout = 50;
12573
12574         while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state)) {
12575                 timeout--;
12576                 if (!timeout)
12577                         return -EBUSY;
12578                 usleep_range(1000, 2000);
12579         }
12580
12581         return 0;
12582 }
12583
12584 /**
12585  * i40e_exit_busy_conf - Exits busy config state
12586  * @vsi: vsi
12587  **/
12588 static void i40e_exit_busy_conf(struct i40e_vsi *vsi)
12589 {
12590         struct i40e_pf *pf = vsi->back;
12591
12592         clear_bit(__I40E_CONFIG_BUSY, pf->state);
12593 }
12594
12595 /**
12596  * i40e_queue_pair_reset_stats - Resets all statistics for a queue pair
12597  * @vsi: vsi
12598  * @queue_pair: queue pair
12599  **/
12600 static void i40e_queue_pair_reset_stats(struct i40e_vsi *vsi, int queue_pair)
12601 {
12602         memset(&vsi->rx_rings[queue_pair]->rx_stats, 0,
12603                sizeof(vsi->rx_rings[queue_pair]->rx_stats));
12604         memset(&vsi->tx_rings[queue_pair]->stats, 0,
12605                sizeof(vsi->tx_rings[queue_pair]->stats));
12606         if (i40e_enabled_xdp_vsi(vsi)) {
12607                 memset(&vsi->xdp_rings[queue_pair]->stats, 0,
12608                        sizeof(vsi->xdp_rings[queue_pair]->stats));
12609         }
12610 }
12611
12612 /**
12613  * i40e_queue_pair_clean_rings - Cleans all the rings of a queue pair
12614  * @vsi: vsi
12615  * @queue_pair: queue pair
12616  **/
12617 static void i40e_queue_pair_clean_rings(struct i40e_vsi *vsi, int queue_pair)
12618 {
12619         i40e_clean_tx_ring(vsi->tx_rings[queue_pair]);
12620         if (i40e_enabled_xdp_vsi(vsi)) {
12621                 /* Make sure that in-progress ndo_xdp_xmit calls are
12622                  * completed.
12623                  */
12624                 synchronize_rcu();
12625                 i40e_clean_tx_ring(vsi->xdp_rings[queue_pair]);
12626         }
12627         i40e_clean_rx_ring(vsi->rx_rings[queue_pair]);
12628 }
12629
12630 /**
12631  * i40e_queue_pair_toggle_napi - Enables/disables NAPI for a queue pair
12632  * @vsi: vsi
12633  * @queue_pair: queue pair
12634  * @enable: true for enable, false for disable
12635  **/
12636 static void i40e_queue_pair_toggle_napi(struct i40e_vsi *vsi, int queue_pair,
12637                                         bool enable)
12638 {
12639         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12640         struct i40e_q_vector *q_vector = rxr->q_vector;
12641
12642         if (!vsi->netdev)
12643                 return;
12644
12645         /* All rings in a qp belong to the same qvector. */
12646         if (q_vector->rx.ring || q_vector->tx.ring) {
12647                 if (enable)
12648                         napi_enable(&q_vector->napi);
12649                 else
12650                         napi_disable(&q_vector->napi);
12651         }
12652 }
12653
12654 /**
12655  * i40e_queue_pair_toggle_rings - Enables/disables all rings for a queue pair
12656  * @vsi: vsi
12657  * @queue_pair: queue pair
12658  * @enable: true for enable, false for disable
12659  *
12660  * Returns 0 on success, <0 on failure.
12661  **/
12662 static int i40e_queue_pair_toggle_rings(struct i40e_vsi *vsi, int queue_pair,
12663                                         bool enable)
12664 {
12665         struct i40e_pf *pf = vsi->back;
12666         int pf_q, ret = 0;
12667
12668         pf_q = vsi->base_queue + queue_pair;
12669         ret = i40e_control_wait_tx_q(vsi->seid, pf, pf_q,
12670                                      false /*is xdp*/, enable);
12671         if (ret) {
12672                 dev_info(&pf->pdev->dev,
12673                          "VSI seid %d Tx ring %d %sable timeout\n",
12674                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12675                 return ret;
12676         }
12677
12678         i40e_control_rx_q(pf, pf_q, enable);
12679         ret = i40e_pf_rxq_wait(pf, pf_q, enable);
12680         if (ret) {
12681                 dev_info(&pf->pdev->dev,
12682                          "VSI seid %d Rx ring %d %sable timeout\n",
12683                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12684                 return ret;
12685         }
12686
12687         /* Due to HW errata, on Rx disable only, the register can
12688          * indicate done before it really is. Needs 50ms to be sure
12689          */
12690         if (!enable)
12691                 mdelay(50);
12692
12693         if (!i40e_enabled_xdp_vsi(vsi))
12694                 return ret;
12695
12696         ret = i40e_control_wait_tx_q(vsi->seid, pf,
12697                                      pf_q + vsi->alloc_queue_pairs,
12698                                      true /*is xdp*/, enable);
12699         if (ret) {
12700                 dev_info(&pf->pdev->dev,
12701                          "VSI seid %d XDP Tx ring %d %sable timeout\n",
12702                          vsi->seid, pf_q, (enable ? "en" : "dis"));
12703         }
12704
12705         return ret;
12706 }
12707
12708 /**
12709  * i40e_queue_pair_enable_irq - Enables interrupts for a queue pair
12710  * @vsi: vsi
12711  * @queue_pair: queue_pair
12712  **/
12713 static void i40e_queue_pair_enable_irq(struct i40e_vsi *vsi, int queue_pair)
12714 {
12715         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12716         struct i40e_pf *pf = vsi->back;
12717         struct i40e_hw *hw = &pf->hw;
12718
12719         /* All rings in a qp belong to the same qvector. */
12720         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
12721                 i40e_irq_dynamic_enable(vsi, rxr->q_vector->v_idx);
12722         else
12723                 i40e_irq_dynamic_enable_icr0(pf);
12724
12725         i40e_flush(hw);
12726 }
12727
12728 /**
12729  * i40e_queue_pair_disable_irq - Disables interrupts for a queue pair
12730  * @vsi: vsi
12731  * @queue_pair: queue_pair
12732  **/
12733 static void i40e_queue_pair_disable_irq(struct i40e_vsi *vsi, int queue_pair)
12734 {
12735         struct i40e_ring *rxr = vsi->rx_rings[queue_pair];
12736         struct i40e_pf *pf = vsi->back;
12737         struct i40e_hw *hw = &pf->hw;
12738
12739         /* For simplicity, instead of removing the qp interrupt causes
12740          * from the interrupt linked list, we simply disable the interrupt, and
12741          * leave the list intact.
12742          *
12743          * All rings in a qp belong to the same qvector.
12744          */
12745         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
12746                 u32 intpf = vsi->base_vector + rxr->q_vector->v_idx;
12747
12748                 wr32(hw, I40E_PFINT_DYN_CTLN(intpf - 1), 0);
12749                 i40e_flush(hw);
12750                 synchronize_irq(pf->msix_entries[intpf].vector);
12751         } else {
12752                 /* Legacy and MSI mode - this stops all interrupt handling */
12753                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
12754                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
12755                 i40e_flush(hw);
12756                 synchronize_irq(pf->pdev->irq);
12757         }
12758 }
12759
12760 /**
12761  * i40e_queue_pair_disable - Disables a queue pair
12762  * @vsi: vsi
12763  * @queue_pair: queue pair
12764  *
12765  * Returns 0 on success, <0 on failure.
12766  **/
12767 int i40e_queue_pair_disable(struct i40e_vsi *vsi, int queue_pair)
12768 {
12769         int err;
12770
12771         err = i40e_enter_busy_conf(vsi);
12772         if (err)
12773                 return err;
12774
12775         i40e_queue_pair_disable_irq(vsi, queue_pair);
12776         err = i40e_queue_pair_toggle_rings(vsi, queue_pair, false /* off */);
12777         i40e_queue_pair_toggle_napi(vsi, queue_pair, false /* off */);
12778         i40e_queue_pair_clean_rings(vsi, queue_pair);
12779         i40e_queue_pair_reset_stats(vsi, queue_pair);
12780
12781         return err;
12782 }
12783
12784 /**
12785  * i40e_queue_pair_enable - Enables a queue pair
12786  * @vsi: vsi
12787  * @queue_pair: queue pair
12788  *
12789  * Returns 0 on success, <0 on failure.
12790  **/
12791 int i40e_queue_pair_enable(struct i40e_vsi *vsi, int queue_pair)
12792 {
12793         int err;
12794
12795         err = i40e_configure_tx_ring(vsi->tx_rings[queue_pair]);
12796         if (err)
12797                 return err;
12798
12799         if (i40e_enabled_xdp_vsi(vsi)) {
12800                 err = i40e_configure_tx_ring(vsi->xdp_rings[queue_pair]);
12801                 if (err)
12802                         return err;
12803         }
12804
12805         err = i40e_configure_rx_ring(vsi->rx_rings[queue_pair]);
12806         if (err)
12807                 return err;
12808
12809         err = i40e_queue_pair_toggle_rings(vsi, queue_pair, true /* on */);
12810         i40e_queue_pair_toggle_napi(vsi, queue_pair, true /* on */);
12811         i40e_queue_pair_enable_irq(vsi, queue_pair);
12812
12813         i40e_exit_busy_conf(vsi);
12814
12815         return err;
12816 }
12817
12818 /**
12819  * i40e_xdp - implements ndo_bpf for i40e
12820  * @dev: netdevice
12821  * @xdp: XDP command
12822  **/
12823 static int i40e_xdp(struct net_device *dev,
12824                     struct netdev_bpf *xdp)
12825 {
12826         struct i40e_netdev_priv *np = netdev_priv(dev);
12827         struct i40e_vsi *vsi = np->vsi;
12828
12829         if (vsi->type != I40E_VSI_MAIN)
12830                 return -EINVAL;
12831
12832         switch (xdp->command) {
12833         case XDP_SETUP_PROG:
12834                 return i40e_xdp_setup(vsi, xdp->prog);
12835         case XDP_QUERY_PROG:
12836                 xdp->prog_id = vsi->xdp_prog ? vsi->xdp_prog->aux->id : 0;
12837                 return 0;
12838         case XDP_SETUP_XSK_UMEM:
12839                 return i40e_xsk_umem_setup(vsi, xdp->xsk.umem,
12840                                            xdp->xsk.queue_id);
12841         default:
12842                 return -EINVAL;
12843         }
12844 }
12845
12846 static const struct net_device_ops i40e_netdev_ops = {
12847         .ndo_open               = i40e_open,
12848         .ndo_stop               = i40e_close,
12849         .ndo_start_xmit         = i40e_lan_xmit_frame,
12850         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
12851         .ndo_set_rx_mode        = i40e_set_rx_mode,
12852         .ndo_validate_addr      = eth_validate_addr,
12853         .ndo_set_mac_address    = i40e_set_mac,
12854         .ndo_change_mtu         = i40e_change_mtu,
12855         .ndo_do_ioctl           = i40e_ioctl,
12856         .ndo_tx_timeout         = i40e_tx_timeout,
12857         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
12858         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
12859 #ifdef CONFIG_NET_POLL_CONTROLLER
12860         .ndo_poll_controller    = i40e_netpoll,
12861 #endif
12862         .ndo_setup_tc           = __i40e_setup_tc,
12863         .ndo_set_features       = i40e_set_features,
12864         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
12865         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
12866         .ndo_get_vf_stats       = i40e_get_vf_stats,
12867         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
12868         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
12869         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
12870         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
12871         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
12872         .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
12873         .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
12874         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
12875         .ndo_fdb_add            = i40e_ndo_fdb_add,
12876         .ndo_features_check     = i40e_features_check,
12877         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
12878         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
12879         .ndo_bpf                = i40e_xdp,
12880         .ndo_xdp_xmit           = i40e_xdp_xmit,
12881         .ndo_xsk_wakeup         = i40e_xsk_wakeup,
12882         .ndo_dfwd_add_station   = i40e_fwd_add,
12883         .ndo_dfwd_del_station   = i40e_fwd_del,
12884 };
12885
12886 /**
12887  * i40e_config_netdev - Setup the netdev flags
12888  * @vsi: the VSI being configured
12889  *
12890  * Returns 0 on success, negative value on failure
12891  **/
12892 static int i40e_config_netdev(struct i40e_vsi *vsi)
12893 {
12894         struct i40e_pf *pf = vsi->back;
12895         struct i40e_hw *hw = &pf->hw;
12896         struct i40e_netdev_priv *np;
12897         struct net_device *netdev;
12898         u8 broadcast[ETH_ALEN];
12899         u8 mac_addr[ETH_ALEN];
12900         int etherdev_size;
12901         netdev_features_t hw_enc_features;
12902         netdev_features_t hw_features;
12903
12904         etherdev_size = sizeof(struct i40e_netdev_priv);
12905         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
12906         if (!netdev)
12907                 return -ENOMEM;
12908
12909         vsi->netdev = netdev;
12910         np = netdev_priv(netdev);
12911         np->vsi = vsi;
12912
12913         hw_enc_features = NETIF_F_SG                    |
12914                           NETIF_F_IP_CSUM               |
12915                           NETIF_F_IPV6_CSUM             |
12916                           NETIF_F_HIGHDMA               |
12917                           NETIF_F_SOFT_FEATURES         |
12918                           NETIF_F_TSO                   |
12919                           NETIF_F_TSO_ECN               |
12920                           NETIF_F_TSO6                  |
12921                           NETIF_F_GSO_GRE               |
12922                           NETIF_F_GSO_GRE_CSUM          |
12923                           NETIF_F_GSO_PARTIAL           |
12924                           NETIF_F_GSO_IPXIP4            |
12925                           NETIF_F_GSO_IPXIP6            |
12926                           NETIF_F_GSO_UDP_TUNNEL        |
12927                           NETIF_F_GSO_UDP_TUNNEL_CSUM   |
12928                           NETIF_F_GSO_UDP_L4            |
12929                           NETIF_F_SCTP_CRC              |
12930                           NETIF_F_RXHASH                |
12931                           NETIF_F_RXCSUM                |
12932                           0;
12933
12934         if (!(pf->hw_features & I40E_HW_OUTER_UDP_CSUM_CAPABLE))
12935                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
12936
12937         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
12938
12939         netdev->hw_enc_features |= hw_enc_features;
12940
12941         /* record features VLANs can make use of */
12942         netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
12943
12944         /* enable macvlan offloads */
12945         netdev->hw_features |= NETIF_F_HW_L2FW_DOFFLOAD;
12946
12947         hw_features = hw_enc_features           |
12948                       NETIF_F_HW_VLAN_CTAG_TX   |
12949                       NETIF_F_HW_VLAN_CTAG_RX;
12950
12951         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
12952                 hw_features |= NETIF_F_NTUPLE | NETIF_F_HW_TC;
12953
12954         netdev->hw_features |= hw_features;
12955
12956         netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
12957         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
12958
12959         if (vsi->type == I40E_VSI_MAIN) {
12960                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
12961                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
12962                 /* The following steps are necessary for two reasons. First,
12963                  * some older NVM configurations load a default MAC-VLAN
12964                  * filter that will accept any tagged packet, and we want to
12965                  * replace this with a normal filter. Additionally, it is
12966                  * possible our MAC address was provided by the platform using
12967                  * Open Firmware or similar.
12968                  *
12969                  * Thus, we need to remove the default filter and install one
12970                  * specific to the MAC address.
12971                  */
12972                 i40e_rm_default_mac_filter(vsi, mac_addr);
12973                 spin_lock_bh(&vsi->mac_filter_hash_lock);
12974                 i40e_add_mac_filter(vsi, mac_addr);
12975                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
12976         } else {
12977                 /* Relate the VSI_VMDQ name to the VSI_MAIN name. Note that we
12978                  * are still limited by IFNAMSIZ, but we're adding 'v%d\0' to
12979                  * the end, which is 4 bytes long, so force truncation of the
12980                  * original name by IFNAMSIZ - 4
12981                  */
12982                 snprintf(netdev->name, IFNAMSIZ, "%.*sv%%d",
12983                          IFNAMSIZ - 4,
12984                          pf->vsi[pf->lan_vsi]->netdev->name);
12985                 eth_random_addr(mac_addr);
12986
12987                 spin_lock_bh(&vsi->mac_filter_hash_lock);
12988                 i40e_add_mac_filter(vsi, mac_addr);
12989                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
12990         }
12991
12992         /* Add the broadcast filter so that we initially will receive
12993          * broadcast packets. Note that when a new VLAN is first added the
12994          * driver will convert all filters marked I40E_VLAN_ANY into VLAN
12995          * specific filters as part of transitioning into "vlan" operation.
12996          * When more VLANs are added, the driver will copy each existing MAC
12997          * filter and add it for the new VLAN.
12998          *
12999          * Broadcast filters are handled specially by
13000          * i40e_sync_filters_subtask, as the driver must to set the broadcast
13001          * promiscuous bit instead of adding this directly as a MAC/VLAN
13002          * filter. The subtask will update the correct broadcast promiscuous
13003          * bits as VLANs become active or inactive.
13004          */
13005         eth_broadcast_addr(broadcast);
13006         spin_lock_bh(&vsi->mac_filter_hash_lock);
13007         i40e_add_mac_filter(vsi, broadcast);
13008         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13009
13010         ether_addr_copy(netdev->dev_addr, mac_addr);
13011         ether_addr_copy(netdev->perm_addr, mac_addr);
13012
13013         /* i40iw_net_event() reads 16 bytes from neigh->primary_key */
13014         netdev->neigh_priv_len = sizeof(u32) * 4;
13015
13016         netdev->priv_flags |= IFF_UNICAST_FLT;
13017         netdev->priv_flags |= IFF_SUPP_NOFCS;
13018         /* Setup netdev TC information */
13019         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
13020
13021         netdev->netdev_ops = &i40e_netdev_ops;
13022         netdev->watchdog_timeo = 5 * HZ;
13023         i40e_set_ethtool_ops(netdev);
13024
13025         /* MTU range: 68 - 9706 */
13026         netdev->min_mtu = ETH_MIN_MTU;
13027         netdev->max_mtu = I40E_MAX_RXBUFFER - I40E_PACKET_HDR_PAD;
13028
13029         return 0;
13030 }
13031
13032 /**
13033  * i40e_vsi_delete - Delete a VSI from the switch
13034  * @vsi: the VSI being removed
13035  *
13036  * Returns 0 on success, negative value on failure
13037  **/
13038 static void i40e_vsi_delete(struct i40e_vsi *vsi)
13039 {
13040         /* remove default VSI is not allowed */
13041         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
13042                 return;
13043
13044         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
13045 }
13046
13047 /**
13048  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
13049  * @vsi: the VSI being queried
13050  *
13051  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
13052  **/
13053 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
13054 {
13055         struct i40e_veb *veb;
13056         struct i40e_pf *pf = vsi->back;
13057
13058         /* Uplink is not a bridge so default to VEB */
13059         if (vsi->veb_idx >= I40E_MAX_VEB)
13060                 return 1;
13061
13062         veb = pf->veb[vsi->veb_idx];
13063         if (!veb) {
13064                 dev_info(&pf->pdev->dev,
13065                          "There is no veb associated with the bridge\n");
13066                 return -ENOENT;
13067         }
13068
13069         /* Uplink is a bridge in VEPA mode */
13070         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
13071                 return 0;
13072         } else {
13073                 /* Uplink is a bridge in VEB mode */
13074                 return 1;
13075         }
13076
13077         /* VEPA is now default bridge, so return 0 */
13078         return 0;
13079 }
13080
13081 /**
13082  * i40e_add_vsi - Add a VSI to the switch
13083  * @vsi: the VSI being configured
13084  *
13085  * This initializes a VSI context depending on the VSI type to be added and
13086  * passes it down to the add_vsi aq command.
13087  **/
13088 static int i40e_add_vsi(struct i40e_vsi *vsi)
13089 {
13090         int ret = -ENODEV;
13091         struct i40e_pf *pf = vsi->back;
13092         struct i40e_hw *hw = &pf->hw;
13093         struct i40e_vsi_context ctxt;
13094         struct i40e_mac_filter *f;
13095         struct hlist_node *h;
13096         int bkt;
13097
13098         u8 enabled_tc = 0x1; /* TC0 enabled */
13099         int f_count = 0;
13100
13101         memset(&ctxt, 0, sizeof(ctxt));
13102         switch (vsi->type) {
13103         case I40E_VSI_MAIN:
13104                 /* The PF's main VSI is already setup as part of the
13105                  * device initialization, so we'll not bother with
13106                  * the add_vsi call, but we will retrieve the current
13107                  * VSI context.
13108                  */
13109                 ctxt.seid = pf->main_vsi_seid;
13110                 ctxt.pf_num = pf->hw.pf_id;
13111                 ctxt.vf_num = 0;
13112                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
13113                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13114                 if (ret) {
13115                         dev_info(&pf->pdev->dev,
13116                                  "couldn't get PF vsi config, err %s aq_err %s\n",
13117                                  i40e_stat_str(&pf->hw, ret),
13118                                  i40e_aq_str(&pf->hw,
13119                                              pf->hw.aq.asq_last_status));
13120                         return -ENOENT;
13121                 }
13122                 vsi->info = ctxt.info;
13123                 vsi->info.valid_sections = 0;
13124
13125                 vsi->seid = ctxt.seid;
13126                 vsi->id = ctxt.vsi_number;
13127
13128                 enabled_tc = i40e_pf_get_tc_map(pf);
13129
13130                 /* Source pruning is enabled by default, so the flag is
13131                  * negative logic - if it's set, we need to fiddle with
13132                  * the VSI to disable source pruning.
13133                  */
13134                 if (pf->flags & I40E_FLAG_SOURCE_PRUNING_DISABLED) {
13135                         memset(&ctxt, 0, sizeof(ctxt));
13136                         ctxt.seid = pf->main_vsi_seid;
13137                         ctxt.pf_num = pf->hw.pf_id;
13138                         ctxt.vf_num = 0;
13139                         ctxt.info.valid_sections |=
13140                                      cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13141                         ctxt.info.switch_id =
13142                                    cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
13143                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13144                         if (ret) {
13145                                 dev_info(&pf->pdev->dev,
13146                                          "update vsi failed, err %s aq_err %s\n",
13147                                          i40e_stat_str(&pf->hw, ret),
13148                                          i40e_aq_str(&pf->hw,
13149                                                      pf->hw.aq.asq_last_status));
13150                                 ret = -ENOENT;
13151                                 goto err;
13152                         }
13153                 }
13154
13155                 /* MFP mode setup queue map and update VSI */
13156                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
13157                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
13158                         memset(&ctxt, 0, sizeof(ctxt));
13159                         ctxt.seid = pf->main_vsi_seid;
13160                         ctxt.pf_num = pf->hw.pf_id;
13161                         ctxt.vf_num = 0;
13162                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
13163                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
13164                         if (ret) {
13165                                 dev_info(&pf->pdev->dev,
13166                                          "update vsi failed, err %s aq_err %s\n",
13167                                          i40e_stat_str(&pf->hw, ret),
13168                                          i40e_aq_str(&pf->hw,
13169                                                     pf->hw.aq.asq_last_status));
13170                                 ret = -ENOENT;
13171                                 goto err;
13172                         }
13173                         /* update the local VSI info queue map */
13174                         i40e_vsi_update_queue_map(vsi, &ctxt);
13175                         vsi->info.valid_sections = 0;
13176                 } else {
13177                         /* Default/Main VSI is only enabled for TC0
13178                          * reconfigure it to enable all TCs that are
13179                          * available on the port in SFP mode.
13180                          * For MFP case the iSCSI PF would use this
13181                          * flow to enable LAN+iSCSI TC.
13182                          */
13183                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
13184                         if (ret) {
13185                                 /* Single TC condition is not fatal,
13186                                  * message and continue
13187                                  */
13188                                 dev_info(&pf->pdev->dev,
13189                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
13190                                          enabled_tc,
13191                                          i40e_stat_str(&pf->hw, ret),
13192                                          i40e_aq_str(&pf->hw,
13193                                                     pf->hw.aq.asq_last_status));
13194                         }
13195                 }
13196                 break;
13197
13198         case I40E_VSI_FDIR:
13199                 ctxt.pf_num = hw->pf_id;
13200                 ctxt.vf_num = 0;
13201                 ctxt.uplink_seid = vsi->uplink_seid;
13202                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13203                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
13204                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
13205                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
13206                         ctxt.info.valid_sections |=
13207                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13208                         ctxt.info.switch_id =
13209                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13210                 }
13211                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13212                 break;
13213
13214         case I40E_VSI_VMDQ2:
13215                 ctxt.pf_num = hw->pf_id;
13216                 ctxt.vf_num = 0;
13217                 ctxt.uplink_seid = vsi->uplink_seid;
13218                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13219                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
13220
13221                 /* This VSI is connected to VEB so the switch_id
13222                  * should be set to zero by default.
13223                  */
13224                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13225                         ctxt.info.valid_sections |=
13226                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13227                         ctxt.info.switch_id =
13228                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13229                 }
13230
13231                 /* Setup the VSI tx/rx queue map for TC0 only for now */
13232                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13233                 break;
13234
13235         case I40E_VSI_SRIOV:
13236                 ctxt.pf_num = hw->pf_id;
13237                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
13238                 ctxt.uplink_seid = vsi->uplink_seid;
13239                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
13240                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
13241
13242                 /* This VSI is connected to VEB so the switch_id
13243                  * should be set to zero by default.
13244                  */
13245                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
13246                         ctxt.info.valid_sections |=
13247                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
13248                         ctxt.info.switch_id =
13249                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
13250                 }
13251
13252                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
13253                         ctxt.info.valid_sections |=
13254                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
13255                         ctxt.info.queueing_opt_flags |=
13256                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
13257                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
13258                 }
13259
13260                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
13261                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
13262                 if (pf->vf[vsi->vf_id].spoofchk) {
13263                         ctxt.info.valid_sections |=
13264                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
13265                         ctxt.info.sec_flags |=
13266                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
13267                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
13268                 }
13269                 /* Setup the VSI tx/rx queue map for TC0 only for now */
13270                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
13271                 break;
13272
13273         case I40E_VSI_IWARP:
13274                 /* send down message to iWARP */
13275                 break;
13276
13277         default:
13278                 return -ENODEV;
13279         }
13280
13281         if (vsi->type != I40E_VSI_MAIN) {
13282                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
13283                 if (ret) {
13284                         dev_info(&vsi->back->pdev->dev,
13285                                  "add vsi failed, err %s aq_err %s\n",
13286                                  i40e_stat_str(&pf->hw, ret),
13287                                  i40e_aq_str(&pf->hw,
13288                                              pf->hw.aq.asq_last_status));
13289                         ret = -ENOENT;
13290                         goto err;
13291                 }
13292                 vsi->info = ctxt.info;
13293                 vsi->info.valid_sections = 0;
13294                 vsi->seid = ctxt.seid;
13295                 vsi->id = ctxt.vsi_number;
13296         }
13297
13298         vsi->active_filters = 0;
13299         clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
13300         spin_lock_bh(&vsi->mac_filter_hash_lock);
13301         /* If macvlan filters already exist, force them to get loaded */
13302         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
13303                 f->state = I40E_FILTER_NEW;
13304                 f_count++;
13305         }
13306         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13307
13308         if (f_count) {
13309                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
13310                 set_bit(__I40E_MACVLAN_SYNC_PENDING, pf->state);
13311         }
13312
13313         /* Update VSI BW information */
13314         ret = i40e_vsi_get_bw_info(vsi);
13315         if (ret) {
13316                 dev_info(&pf->pdev->dev,
13317                          "couldn't get vsi bw info, err %s aq_err %s\n",
13318                          i40e_stat_str(&pf->hw, ret),
13319                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13320                 /* VSI is already added so not tearing that up */
13321                 ret = 0;
13322         }
13323
13324 err:
13325         return ret;
13326 }
13327
13328 /**
13329  * i40e_vsi_release - Delete a VSI and free its resources
13330  * @vsi: the VSI being removed
13331  *
13332  * Returns 0 on success or < 0 on error
13333  **/
13334 int i40e_vsi_release(struct i40e_vsi *vsi)
13335 {
13336         struct i40e_mac_filter *f;
13337         struct hlist_node *h;
13338         struct i40e_veb *veb = NULL;
13339         struct i40e_pf *pf;
13340         u16 uplink_seid;
13341         int i, n, bkt;
13342
13343         pf = vsi->back;
13344
13345         /* release of a VEB-owner or last VSI is not allowed */
13346         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
13347                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
13348                          vsi->seid, vsi->uplink_seid);
13349                 return -ENODEV;
13350         }
13351         if (vsi == pf->vsi[pf->lan_vsi] &&
13352             !test_bit(__I40E_DOWN, pf->state)) {
13353                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
13354                 return -ENODEV;
13355         }
13356
13357         uplink_seid = vsi->uplink_seid;
13358         if (vsi->type != I40E_VSI_SRIOV) {
13359                 if (vsi->netdev_registered) {
13360                         vsi->netdev_registered = false;
13361                         if (vsi->netdev) {
13362                                 /* results in a call to i40e_close() */
13363                                 unregister_netdev(vsi->netdev);
13364                         }
13365                 } else {
13366                         i40e_vsi_close(vsi);
13367                 }
13368                 i40e_vsi_disable_irq(vsi);
13369         }
13370
13371         spin_lock_bh(&vsi->mac_filter_hash_lock);
13372
13373         /* clear the sync flag on all filters */
13374         if (vsi->netdev) {
13375                 __dev_uc_unsync(vsi->netdev, NULL);
13376                 __dev_mc_unsync(vsi->netdev, NULL);
13377         }
13378
13379         /* make sure any remaining filters are marked for deletion */
13380         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
13381                 __i40e_del_filter(vsi, f);
13382
13383         spin_unlock_bh(&vsi->mac_filter_hash_lock);
13384
13385         i40e_sync_vsi_filters(vsi);
13386
13387         i40e_vsi_delete(vsi);
13388         i40e_vsi_free_q_vectors(vsi);
13389         if (vsi->netdev) {
13390                 free_netdev(vsi->netdev);
13391                 vsi->netdev = NULL;
13392         }
13393         i40e_vsi_clear_rings(vsi);
13394         i40e_vsi_clear(vsi);
13395
13396         /* If this was the last thing on the VEB, except for the
13397          * controlling VSI, remove the VEB, which puts the controlling
13398          * VSI onto the next level down in the switch.
13399          *
13400          * Well, okay, there's one more exception here: don't remove
13401          * the orphan VEBs yet.  We'll wait for an explicit remove request
13402          * from up the network stack.
13403          */
13404         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
13405                 if (pf->vsi[i] &&
13406                     pf->vsi[i]->uplink_seid == uplink_seid &&
13407                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13408                         n++;      /* count the VSIs */
13409                 }
13410         }
13411         for (i = 0; i < I40E_MAX_VEB; i++) {
13412                 if (!pf->veb[i])
13413                         continue;
13414                 if (pf->veb[i]->uplink_seid == uplink_seid)
13415                         n++;     /* count the VEBs */
13416                 if (pf->veb[i]->seid == uplink_seid)
13417                         veb = pf->veb[i];
13418         }
13419         if (n == 0 && veb && veb->uplink_seid != 0)
13420                 i40e_veb_release(veb);
13421
13422         return 0;
13423 }
13424
13425 /**
13426  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
13427  * @vsi: ptr to the VSI
13428  *
13429  * This should only be called after i40e_vsi_mem_alloc() which allocates the
13430  * corresponding SW VSI structure and initializes num_queue_pairs for the
13431  * newly allocated VSI.
13432  *
13433  * Returns 0 on success or negative on failure
13434  **/
13435 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
13436 {
13437         int ret = -ENOENT;
13438         struct i40e_pf *pf = vsi->back;
13439
13440         if (vsi->q_vectors[0]) {
13441                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
13442                          vsi->seid);
13443                 return -EEXIST;
13444         }
13445
13446         if (vsi->base_vector) {
13447                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
13448                          vsi->seid, vsi->base_vector);
13449                 return -EEXIST;
13450         }
13451
13452         ret = i40e_vsi_alloc_q_vectors(vsi);
13453         if (ret) {
13454                 dev_info(&pf->pdev->dev,
13455                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
13456                          vsi->num_q_vectors, vsi->seid, ret);
13457                 vsi->num_q_vectors = 0;
13458                 goto vector_setup_out;
13459         }
13460
13461         /* In Legacy mode, we do not have to get any other vector since we
13462          * piggyback on the misc/ICR0 for queue interrupts.
13463         */
13464         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
13465                 return ret;
13466         if (vsi->num_q_vectors)
13467                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
13468                                                  vsi->num_q_vectors, vsi->idx);
13469         if (vsi->base_vector < 0) {
13470                 dev_info(&pf->pdev->dev,
13471                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
13472                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
13473                 i40e_vsi_free_q_vectors(vsi);
13474                 ret = -ENOENT;
13475                 goto vector_setup_out;
13476         }
13477
13478 vector_setup_out:
13479         return ret;
13480 }
13481
13482 /**
13483  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
13484  * @vsi: pointer to the vsi.
13485  *
13486  * This re-allocates a vsi's queue resources.
13487  *
13488  * Returns pointer to the successfully allocated and configured VSI sw struct
13489  * on success, otherwise returns NULL on failure.
13490  **/
13491 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
13492 {
13493         u16 alloc_queue_pairs;
13494         struct i40e_pf *pf;
13495         u8 enabled_tc;
13496         int ret;
13497
13498         if (!vsi)
13499                 return NULL;
13500
13501         pf = vsi->back;
13502
13503         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
13504         i40e_vsi_clear_rings(vsi);
13505
13506         i40e_vsi_free_arrays(vsi, false);
13507         i40e_set_num_rings_in_vsi(vsi);
13508         ret = i40e_vsi_alloc_arrays(vsi, false);
13509         if (ret)
13510                 goto err_vsi;
13511
13512         alloc_queue_pairs = vsi->alloc_queue_pairs *
13513                             (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13514
13515         ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13516         if (ret < 0) {
13517                 dev_info(&pf->pdev->dev,
13518                          "failed to get tracking for %d queues for VSI %d err %d\n",
13519                          alloc_queue_pairs, vsi->seid, ret);
13520                 goto err_vsi;
13521         }
13522         vsi->base_queue = ret;
13523
13524         /* Update the FW view of the VSI. Force a reset of TC and queue
13525          * layout configurations.
13526          */
13527         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
13528         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
13529         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
13530         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
13531         if (vsi->type == I40E_VSI_MAIN)
13532                 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
13533
13534         /* assign it some queues */
13535         ret = i40e_alloc_rings(vsi);
13536         if (ret)
13537                 goto err_rings;
13538
13539         /* map all of the rings to the q_vectors */
13540         i40e_vsi_map_rings_to_vectors(vsi);
13541         return vsi;
13542
13543 err_rings:
13544         i40e_vsi_free_q_vectors(vsi);
13545         if (vsi->netdev_registered) {
13546                 vsi->netdev_registered = false;
13547                 unregister_netdev(vsi->netdev);
13548                 free_netdev(vsi->netdev);
13549                 vsi->netdev = NULL;
13550         }
13551         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13552 err_vsi:
13553         i40e_vsi_clear(vsi);
13554         return NULL;
13555 }
13556
13557 /**
13558  * i40e_vsi_setup - Set up a VSI by a given type
13559  * @pf: board private structure
13560  * @type: VSI type
13561  * @uplink_seid: the switch element to link to
13562  * @param1: usage depends upon VSI type. For VF types, indicates VF id
13563  *
13564  * This allocates the sw VSI structure and its queue resources, then add a VSI
13565  * to the identified VEB.
13566  *
13567  * Returns pointer to the successfully allocated and configure VSI sw struct on
13568  * success, otherwise returns NULL on failure.
13569  **/
13570 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
13571                                 u16 uplink_seid, u32 param1)
13572 {
13573         struct i40e_vsi *vsi = NULL;
13574         struct i40e_veb *veb = NULL;
13575         u16 alloc_queue_pairs;
13576         int ret, i;
13577         int v_idx;
13578
13579         /* The requested uplink_seid must be either
13580          *     - the PF's port seid
13581          *              no VEB is needed because this is the PF
13582          *              or this is a Flow Director special case VSI
13583          *     - seid of an existing VEB
13584          *     - seid of a VSI that owns an existing VEB
13585          *     - seid of a VSI that doesn't own a VEB
13586          *              a new VEB is created and the VSI becomes the owner
13587          *     - seid of the PF VSI, which is what creates the first VEB
13588          *              this is a special case of the previous
13589          *
13590          * Find which uplink_seid we were given and create a new VEB if needed
13591          */
13592         for (i = 0; i < I40E_MAX_VEB; i++) {
13593                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
13594                         veb = pf->veb[i];
13595                         break;
13596                 }
13597         }
13598
13599         if (!veb && uplink_seid != pf->mac_seid) {
13600
13601                 for (i = 0; i < pf->num_alloc_vsi; i++) {
13602                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
13603                                 vsi = pf->vsi[i];
13604                                 break;
13605                         }
13606                 }
13607                 if (!vsi) {
13608                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
13609                                  uplink_seid);
13610                         return NULL;
13611                 }
13612
13613                 if (vsi->uplink_seid == pf->mac_seid)
13614                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
13615                                              vsi->tc_config.enabled_tc);
13616                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
13617                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
13618                                              vsi->tc_config.enabled_tc);
13619                 if (veb) {
13620                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
13621                                 dev_info(&vsi->back->pdev->dev,
13622                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
13623                                 return NULL;
13624                         }
13625                         /* We come up by default in VEPA mode if SRIOV is not
13626                          * already enabled, in which case we can't force VEPA
13627                          * mode.
13628                          */
13629                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
13630                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
13631                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
13632                         }
13633                         i40e_config_bridge_mode(veb);
13634                 }
13635                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
13636                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
13637                                 veb = pf->veb[i];
13638                 }
13639                 if (!veb) {
13640                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
13641                         return NULL;
13642                 }
13643
13644                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13645                 uplink_seid = veb->seid;
13646         }
13647
13648         /* get vsi sw struct */
13649         v_idx = i40e_vsi_mem_alloc(pf, type);
13650         if (v_idx < 0)
13651                 goto err_alloc;
13652         vsi = pf->vsi[v_idx];
13653         if (!vsi)
13654                 goto err_alloc;
13655         vsi->type = type;
13656         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
13657
13658         if (type == I40E_VSI_MAIN)
13659                 pf->lan_vsi = v_idx;
13660         else if (type == I40E_VSI_SRIOV)
13661                 vsi->vf_id = param1;
13662         /* assign it some queues */
13663         alloc_queue_pairs = vsi->alloc_queue_pairs *
13664                             (i40e_enabled_xdp_vsi(vsi) ? 2 : 1);
13665
13666         ret = i40e_get_lump(pf, pf->qp_pile, alloc_queue_pairs, vsi->idx);
13667         if (ret < 0) {
13668                 dev_info(&pf->pdev->dev,
13669                          "failed to get tracking for %d queues for VSI %d err=%d\n",
13670                          alloc_queue_pairs, vsi->seid, ret);
13671                 goto err_vsi;
13672         }
13673         vsi->base_queue = ret;
13674
13675         /* get a VSI from the hardware */
13676         vsi->uplink_seid = uplink_seid;
13677         ret = i40e_add_vsi(vsi);
13678         if (ret)
13679                 goto err_vsi;
13680
13681         switch (vsi->type) {
13682         /* setup the netdev if needed */
13683         case I40E_VSI_MAIN:
13684         case I40E_VSI_VMDQ2:
13685                 ret = i40e_config_netdev(vsi);
13686                 if (ret)
13687                         goto err_netdev;
13688                 ret = register_netdev(vsi->netdev);
13689                 if (ret)
13690                         goto err_netdev;
13691                 vsi->netdev_registered = true;
13692                 netif_carrier_off(vsi->netdev);
13693 #ifdef CONFIG_I40E_DCB
13694                 /* Setup DCB netlink interface */
13695                 i40e_dcbnl_setup(vsi);
13696 #endif /* CONFIG_I40E_DCB */
13697                 /* fall through */
13698
13699         case I40E_VSI_FDIR:
13700                 /* set up vectors and rings if needed */
13701                 ret = i40e_vsi_setup_vectors(vsi);
13702                 if (ret)
13703                         goto err_msix;
13704
13705                 ret = i40e_alloc_rings(vsi);
13706                 if (ret)
13707                         goto err_rings;
13708
13709                 /* map all of the rings to the q_vectors */
13710                 i40e_vsi_map_rings_to_vectors(vsi);
13711
13712                 i40e_vsi_reset_stats(vsi);
13713                 break;
13714
13715         default:
13716                 /* no netdev or rings for the other VSI types */
13717                 break;
13718         }
13719
13720         if ((pf->hw_features & I40E_HW_RSS_AQ_CAPABLE) &&
13721             (vsi->type == I40E_VSI_VMDQ2)) {
13722                 ret = i40e_vsi_config_rss(vsi);
13723         }
13724         return vsi;
13725
13726 err_rings:
13727         i40e_vsi_free_q_vectors(vsi);
13728 err_msix:
13729         if (vsi->netdev_registered) {
13730                 vsi->netdev_registered = false;
13731                 unregister_netdev(vsi->netdev);
13732                 free_netdev(vsi->netdev);
13733                 vsi->netdev = NULL;
13734         }
13735 err_netdev:
13736         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
13737 err_vsi:
13738         i40e_vsi_clear(vsi);
13739 err_alloc:
13740         return NULL;
13741 }
13742
13743 /**
13744  * i40e_veb_get_bw_info - Query VEB BW information
13745  * @veb: the veb to query
13746  *
13747  * Query the Tx scheduler BW configuration data for given VEB
13748  **/
13749 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
13750 {
13751         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
13752         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
13753         struct i40e_pf *pf = veb->pf;
13754         struct i40e_hw *hw = &pf->hw;
13755         u32 tc_bw_max;
13756         int ret = 0;
13757         int i;
13758
13759         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
13760                                                   &bw_data, NULL);
13761         if (ret) {
13762                 dev_info(&pf->pdev->dev,
13763                          "query veb bw config failed, err %s aq_err %s\n",
13764                          i40e_stat_str(&pf->hw, ret),
13765                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13766                 goto out;
13767         }
13768
13769         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
13770                                                    &ets_data, NULL);
13771         if (ret) {
13772                 dev_info(&pf->pdev->dev,
13773                          "query veb bw ets config failed, err %s aq_err %s\n",
13774                          i40e_stat_str(&pf->hw, ret),
13775                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
13776                 goto out;
13777         }
13778
13779         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
13780         veb->bw_max_quanta = ets_data.tc_bw_max;
13781         veb->is_abs_credits = bw_data.absolute_credits_enable;
13782         veb->enabled_tc = ets_data.tc_valid_bits;
13783         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
13784                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
13785         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
13786                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
13787                 veb->bw_tc_limit_credits[i] =
13788                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
13789                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
13790         }
13791
13792 out:
13793         return ret;
13794 }
13795
13796 /**
13797  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
13798  * @pf: board private structure
13799  *
13800  * On error: returns error code (negative)
13801  * On success: returns vsi index in PF (positive)
13802  **/
13803 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
13804 {
13805         int ret = -ENOENT;
13806         struct i40e_veb *veb;
13807         int i;
13808
13809         /* Need to protect the allocation of switch elements at the PF level */
13810         mutex_lock(&pf->switch_mutex);
13811
13812         /* VEB list may be fragmented if VEB creation/destruction has
13813          * been happening.  We can afford to do a quick scan to look
13814          * for any free slots in the list.
13815          *
13816          * find next empty veb slot, looping back around if necessary
13817          */
13818         i = 0;
13819         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
13820                 i++;
13821         if (i >= I40E_MAX_VEB) {
13822                 ret = -ENOMEM;
13823                 goto err_alloc_veb;  /* out of VEB slots! */
13824         }
13825
13826         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
13827         if (!veb) {
13828                 ret = -ENOMEM;
13829                 goto err_alloc_veb;
13830         }
13831         veb->pf = pf;
13832         veb->idx = i;
13833         veb->enabled_tc = 1;
13834
13835         pf->veb[i] = veb;
13836         ret = i;
13837 err_alloc_veb:
13838         mutex_unlock(&pf->switch_mutex);
13839         return ret;
13840 }
13841
13842 /**
13843  * i40e_switch_branch_release - Delete a branch of the switch tree
13844  * @branch: where to start deleting
13845  *
13846  * This uses recursion to find the tips of the branch to be
13847  * removed, deleting until we get back to and can delete this VEB.
13848  **/
13849 static void i40e_switch_branch_release(struct i40e_veb *branch)
13850 {
13851         struct i40e_pf *pf = branch->pf;
13852         u16 branch_seid = branch->seid;
13853         u16 veb_idx = branch->idx;
13854         int i;
13855
13856         /* release any VEBs on this VEB - RECURSION */
13857         for (i = 0; i < I40E_MAX_VEB; i++) {
13858                 if (!pf->veb[i])
13859                         continue;
13860                 if (pf->veb[i]->uplink_seid == branch->seid)
13861                         i40e_switch_branch_release(pf->veb[i]);
13862         }
13863
13864         /* Release the VSIs on this VEB, but not the owner VSI.
13865          *
13866          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
13867          *       the VEB itself, so don't use (*branch) after this loop.
13868          */
13869         for (i = 0; i < pf->num_alloc_vsi; i++) {
13870                 if (!pf->vsi[i])
13871                         continue;
13872                 if (pf->vsi[i]->uplink_seid == branch_seid &&
13873                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
13874                         i40e_vsi_release(pf->vsi[i]);
13875                 }
13876         }
13877
13878         /* There's one corner case where the VEB might not have been
13879          * removed, so double check it here and remove it if needed.
13880          * This case happens if the veb was created from the debugfs
13881          * commands and no VSIs were added to it.
13882          */
13883         if (pf->veb[veb_idx])
13884                 i40e_veb_release(pf->veb[veb_idx]);
13885 }
13886
13887 /**
13888  * i40e_veb_clear - remove veb struct
13889  * @veb: the veb to remove
13890  **/
13891 static void i40e_veb_clear(struct i40e_veb *veb)
13892 {
13893         if (!veb)
13894                 return;
13895
13896         if (veb->pf) {
13897                 struct i40e_pf *pf = veb->pf;
13898
13899                 mutex_lock(&pf->switch_mutex);
13900                 if (pf->veb[veb->idx] == veb)
13901                         pf->veb[veb->idx] = NULL;
13902                 mutex_unlock(&pf->switch_mutex);
13903         }
13904
13905         kfree(veb);
13906 }
13907
13908 /**
13909  * i40e_veb_release - Delete a VEB and free its resources
13910  * @veb: the VEB being removed
13911  **/
13912 void i40e_veb_release(struct i40e_veb *veb)
13913 {
13914         struct i40e_vsi *vsi = NULL;
13915         struct i40e_pf *pf;
13916         int i, n = 0;
13917
13918         pf = veb->pf;
13919
13920         /* find the remaining VSI and check for extras */
13921         for (i = 0; i < pf->num_alloc_vsi; i++) {
13922                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
13923                         n++;
13924                         vsi = pf->vsi[i];
13925                 }
13926         }
13927         if (n != 1) {
13928                 dev_info(&pf->pdev->dev,
13929                          "can't remove VEB %d with %d VSIs left\n",
13930                          veb->seid, n);
13931                 return;
13932         }
13933
13934         /* move the remaining VSI to uplink veb */
13935         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
13936         if (veb->uplink_seid) {
13937                 vsi->uplink_seid = veb->uplink_seid;
13938                 if (veb->uplink_seid == pf->mac_seid)
13939                         vsi->veb_idx = I40E_NO_VEB;
13940                 else
13941                         vsi->veb_idx = veb->veb_idx;
13942         } else {
13943                 /* floating VEB */
13944                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
13945                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
13946         }
13947
13948         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13949         i40e_veb_clear(veb);
13950 }
13951
13952 /**
13953  * i40e_add_veb - create the VEB in the switch
13954  * @veb: the VEB to be instantiated
13955  * @vsi: the controlling VSI
13956  **/
13957 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
13958 {
13959         struct i40e_pf *pf = veb->pf;
13960         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
13961         int ret;
13962
13963         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
13964                               veb->enabled_tc, false,
13965                               &veb->seid, enable_stats, NULL);
13966
13967         /* get a VEB from the hardware */
13968         if (ret) {
13969                 dev_info(&pf->pdev->dev,
13970                          "couldn't add VEB, err %s aq_err %s\n",
13971                          i40e_stat_str(&pf->hw, ret),
13972                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13973                 return -EPERM;
13974         }
13975
13976         /* get statistics counter */
13977         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
13978                                          &veb->stats_idx, NULL, NULL, NULL);
13979         if (ret) {
13980                 dev_info(&pf->pdev->dev,
13981                          "couldn't get VEB statistics idx, 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                 return -EPERM;
13985         }
13986         ret = i40e_veb_get_bw_info(veb);
13987         if (ret) {
13988                 dev_info(&pf->pdev->dev,
13989                          "couldn't get VEB bw info, err %s aq_err %s\n",
13990                          i40e_stat_str(&pf->hw, ret),
13991                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
13992                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
13993                 return -ENOENT;
13994         }
13995
13996         vsi->uplink_seid = veb->seid;
13997         vsi->veb_idx = veb->idx;
13998         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
13999
14000         return 0;
14001 }
14002
14003 /**
14004  * i40e_veb_setup - Set up a VEB
14005  * @pf: board private structure
14006  * @flags: VEB setup flags
14007  * @uplink_seid: the switch element to link to
14008  * @vsi_seid: the initial VSI seid
14009  * @enabled_tc: Enabled TC bit-map
14010  *
14011  * This allocates the sw VEB structure and links it into the switch
14012  * It is possible and legal for this to be a duplicate of an already
14013  * existing VEB.  It is also possible for both uplink and vsi seids
14014  * to be zero, in order to create a floating VEB.
14015  *
14016  * Returns pointer to the successfully allocated VEB sw struct on
14017  * success, otherwise returns NULL on failure.
14018  **/
14019 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
14020                                 u16 uplink_seid, u16 vsi_seid,
14021                                 u8 enabled_tc)
14022 {
14023         struct i40e_veb *veb, *uplink_veb = NULL;
14024         int vsi_idx, veb_idx;
14025         int ret;
14026
14027         /* if one seid is 0, the other must be 0 to create a floating relay */
14028         if ((uplink_seid == 0 || vsi_seid == 0) &&
14029             (uplink_seid + vsi_seid != 0)) {
14030                 dev_info(&pf->pdev->dev,
14031                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
14032                          uplink_seid, vsi_seid);
14033                 return NULL;
14034         }
14035
14036         /* make sure there is such a vsi and uplink */
14037         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
14038                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
14039                         break;
14040         if (vsi_idx == pf->num_alloc_vsi && vsi_seid != 0) {
14041                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
14042                          vsi_seid);
14043                 return NULL;
14044         }
14045
14046         if (uplink_seid && uplink_seid != pf->mac_seid) {
14047                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
14048                         if (pf->veb[veb_idx] &&
14049                             pf->veb[veb_idx]->seid == uplink_seid) {
14050                                 uplink_veb = pf->veb[veb_idx];
14051                                 break;
14052                         }
14053                 }
14054                 if (!uplink_veb) {
14055                         dev_info(&pf->pdev->dev,
14056                                  "uplink seid %d not found\n", uplink_seid);
14057                         return NULL;
14058                 }
14059         }
14060
14061         /* get veb sw struct */
14062         veb_idx = i40e_veb_mem_alloc(pf);
14063         if (veb_idx < 0)
14064                 goto err_alloc;
14065         veb = pf->veb[veb_idx];
14066         veb->flags = flags;
14067         veb->uplink_seid = uplink_seid;
14068         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
14069         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
14070
14071         /* create the VEB in the switch */
14072         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
14073         if (ret)
14074                 goto err_veb;
14075         if (vsi_idx == pf->lan_vsi)
14076                 pf->lan_veb = veb->idx;
14077
14078         return veb;
14079
14080 err_veb:
14081         i40e_veb_clear(veb);
14082 err_alloc:
14083         return NULL;
14084 }
14085
14086 /**
14087  * i40e_setup_pf_switch_element - set PF vars based on switch type
14088  * @pf: board private structure
14089  * @ele: element we are building info from
14090  * @num_reported: total number of elements
14091  * @printconfig: should we print the contents
14092  *
14093  * helper function to assist in extracting a few useful SEID values.
14094  **/
14095 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
14096                                 struct i40e_aqc_switch_config_element_resp *ele,
14097                                 u16 num_reported, bool printconfig)
14098 {
14099         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
14100         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
14101         u8 element_type = ele->element_type;
14102         u16 seid = le16_to_cpu(ele->seid);
14103
14104         if (printconfig)
14105                 dev_info(&pf->pdev->dev,
14106                          "type=%d seid=%d uplink=%d downlink=%d\n",
14107                          element_type, seid, uplink_seid, downlink_seid);
14108
14109         switch (element_type) {
14110         case I40E_SWITCH_ELEMENT_TYPE_MAC:
14111                 pf->mac_seid = seid;
14112                 break;
14113         case I40E_SWITCH_ELEMENT_TYPE_VEB:
14114                 /* Main VEB? */
14115                 if (uplink_seid != pf->mac_seid)
14116                         break;
14117                 if (pf->lan_veb >= I40E_MAX_VEB) {
14118                         int v;
14119
14120                         /* find existing or else empty VEB */
14121                         for (v = 0; v < I40E_MAX_VEB; v++) {
14122                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
14123                                         pf->lan_veb = v;
14124                                         break;
14125                                 }
14126                         }
14127                         if (pf->lan_veb >= I40E_MAX_VEB) {
14128                                 v = i40e_veb_mem_alloc(pf);
14129                                 if (v < 0)
14130                                         break;
14131                                 pf->lan_veb = v;
14132                         }
14133                 }
14134                 if (pf->lan_veb >= I40E_MAX_VEB)
14135                         break;
14136
14137                 pf->veb[pf->lan_veb]->seid = seid;
14138                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
14139                 pf->veb[pf->lan_veb]->pf = pf;
14140                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
14141                 break;
14142         case I40E_SWITCH_ELEMENT_TYPE_VSI:
14143                 if (num_reported != 1)
14144                         break;
14145                 /* This is immediately after a reset so we can assume this is
14146                  * the PF's VSI
14147                  */
14148                 pf->mac_seid = uplink_seid;
14149                 pf->pf_seid = downlink_seid;
14150                 pf->main_vsi_seid = seid;
14151                 if (printconfig)
14152                         dev_info(&pf->pdev->dev,
14153                                  "pf_seid=%d main_vsi_seid=%d\n",
14154                                  pf->pf_seid, pf->main_vsi_seid);
14155                 break;
14156         case I40E_SWITCH_ELEMENT_TYPE_PF:
14157         case I40E_SWITCH_ELEMENT_TYPE_VF:
14158         case I40E_SWITCH_ELEMENT_TYPE_EMP:
14159         case I40E_SWITCH_ELEMENT_TYPE_BMC:
14160         case I40E_SWITCH_ELEMENT_TYPE_PE:
14161         case I40E_SWITCH_ELEMENT_TYPE_PA:
14162                 /* ignore these for now */
14163                 break;
14164         default:
14165                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
14166                          element_type, seid);
14167                 break;
14168         }
14169 }
14170
14171 /**
14172  * i40e_fetch_switch_configuration - Get switch config from firmware
14173  * @pf: board private structure
14174  * @printconfig: should we print the contents
14175  *
14176  * Get the current switch configuration from the device and
14177  * extract a few useful SEID values.
14178  **/
14179 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
14180 {
14181         struct i40e_aqc_get_switch_config_resp *sw_config;
14182         u16 next_seid = 0;
14183         int ret = 0;
14184         u8 *aq_buf;
14185         int i;
14186
14187         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
14188         if (!aq_buf)
14189                 return -ENOMEM;
14190
14191         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
14192         do {
14193                 u16 num_reported, num_total;
14194
14195                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
14196                                                 I40E_AQ_LARGE_BUF,
14197                                                 &next_seid, NULL);
14198                 if (ret) {
14199                         dev_info(&pf->pdev->dev,
14200                                  "get switch config failed err %s aq_err %s\n",
14201                                  i40e_stat_str(&pf->hw, ret),
14202                                  i40e_aq_str(&pf->hw,
14203                                              pf->hw.aq.asq_last_status));
14204                         kfree(aq_buf);
14205                         return -ENOENT;
14206                 }
14207
14208                 num_reported = le16_to_cpu(sw_config->header.num_reported);
14209                 num_total = le16_to_cpu(sw_config->header.num_total);
14210
14211                 if (printconfig)
14212                         dev_info(&pf->pdev->dev,
14213                                  "header: %d reported %d total\n",
14214                                  num_reported, num_total);
14215
14216                 for (i = 0; i < num_reported; i++) {
14217                         struct i40e_aqc_switch_config_element_resp *ele =
14218                                 &sw_config->element[i];
14219
14220                         i40e_setup_pf_switch_element(pf, ele, num_reported,
14221                                                      printconfig);
14222                 }
14223         } while (next_seid != 0);
14224
14225         kfree(aq_buf);
14226         return ret;
14227 }
14228
14229 /**
14230  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
14231  * @pf: board private structure
14232  * @reinit: if the Main VSI needs to re-initialized.
14233  *
14234  * Returns 0 on success, negative value on failure
14235  **/
14236 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
14237 {
14238         u16 flags = 0;
14239         int ret;
14240
14241         /* find out what's out there already */
14242         ret = i40e_fetch_switch_configuration(pf, false);
14243         if (ret) {
14244                 dev_info(&pf->pdev->dev,
14245                          "couldn't fetch switch config, err %s aq_err %s\n",
14246                          i40e_stat_str(&pf->hw, ret),
14247                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
14248                 return ret;
14249         }
14250         i40e_pf_reset_stats(pf);
14251
14252         /* set the switch config bit for the whole device to
14253          * support limited promisc or true promisc
14254          * when user requests promisc. The default is limited
14255          * promisc.
14256         */
14257
14258         if ((pf->hw.pf_id == 0) &&
14259             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT)) {
14260                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14261                 pf->last_sw_conf_flags = flags;
14262         }
14263
14264         if (pf->hw.pf_id == 0) {
14265                 u16 valid_flags;
14266
14267                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
14268                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags, 0,
14269                                                 NULL);
14270                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
14271                         dev_info(&pf->pdev->dev,
14272                                  "couldn't set switch config bits, err %s aq_err %s\n",
14273                                  i40e_stat_str(&pf->hw, ret),
14274                                  i40e_aq_str(&pf->hw,
14275                                              pf->hw.aq.asq_last_status));
14276                         /* not a fatal problem, just keep going */
14277                 }
14278                 pf->last_sw_conf_valid_flags = valid_flags;
14279         }
14280
14281         /* first time setup */
14282         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
14283                 struct i40e_vsi *vsi = NULL;
14284                 u16 uplink_seid;
14285
14286                 /* Set up the PF VSI associated with the PF's main VSI
14287                  * that is already in the HW switch
14288                  */
14289                 if (pf->lan_veb < I40E_MAX_VEB && pf->veb[pf->lan_veb])
14290                         uplink_seid = pf->veb[pf->lan_veb]->seid;
14291                 else
14292                         uplink_seid = pf->mac_seid;
14293                 if (pf->lan_vsi == I40E_NO_VSI)
14294                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
14295                 else if (reinit)
14296                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
14297                 if (!vsi) {
14298                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
14299                         i40e_cloud_filter_exit(pf);
14300                         i40e_fdir_teardown(pf);
14301                         return -EAGAIN;
14302                 }
14303         } else {
14304                 /* force a reset of TC and queue layout configurations */
14305                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
14306
14307                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
14308                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
14309                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
14310         }
14311         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
14312
14313         i40e_fdir_sb_setup(pf);
14314
14315         /* Setup static PF queue filter control settings */
14316         ret = i40e_setup_pf_filter_control(pf);
14317         if (ret) {
14318                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
14319                          ret);
14320                 /* Failure here should not stop continuing other steps */
14321         }
14322
14323         /* enable RSS in the HW, even for only one queue, as the stack can use
14324          * the hash
14325          */
14326         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
14327                 i40e_pf_config_rss(pf);
14328
14329         /* fill in link information and enable LSE reporting */
14330         i40e_link_event(pf);
14331
14332         /* Initialize user-specific link properties */
14333         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
14334                                   I40E_AQ_AN_COMPLETED) ? true : false);
14335
14336         i40e_ptp_init(pf);
14337
14338         /* repopulate tunnel port filters */
14339         i40e_sync_udp_filters(pf);
14340
14341         return ret;
14342 }
14343
14344 /**
14345  * i40e_determine_queue_usage - Work out queue distribution
14346  * @pf: board private structure
14347  **/
14348 static void i40e_determine_queue_usage(struct i40e_pf *pf)
14349 {
14350         int queues_left;
14351         int q_max;
14352
14353         pf->num_lan_qps = 0;
14354
14355         /* Find the max queues to be put into basic use.  We'll always be
14356          * using TC0, whether or not DCB is running, and TC0 will get the
14357          * big RSS set.
14358          */
14359         queues_left = pf->hw.func_caps.num_tx_qp;
14360
14361         if ((queues_left == 1) ||
14362             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
14363                 /* one qp for PF, no queues for anything else */
14364                 queues_left = 0;
14365                 pf->alloc_rss_size = pf->num_lan_qps = 1;
14366
14367                 /* make sure all the fancies are disabled */
14368                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14369                                I40E_FLAG_IWARP_ENABLED  |
14370                                I40E_FLAG_FD_SB_ENABLED  |
14371                                I40E_FLAG_FD_ATR_ENABLED |
14372                                I40E_FLAG_DCB_CAPABLE    |
14373                                I40E_FLAG_DCB_ENABLED    |
14374                                I40E_FLAG_SRIOV_ENABLED  |
14375                                I40E_FLAG_VMDQ_ENABLED);
14376                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14377         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
14378                                   I40E_FLAG_FD_SB_ENABLED |
14379                                   I40E_FLAG_FD_ATR_ENABLED |
14380                                   I40E_FLAG_DCB_CAPABLE))) {
14381                 /* one qp for PF */
14382                 pf->alloc_rss_size = pf->num_lan_qps = 1;
14383                 queues_left -= pf->num_lan_qps;
14384
14385                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
14386                                I40E_FLAG_IWARP_ENABLED  |
14387                                I40E_FLAG_FD_SB_ENABLED  |
14388                                I40E_FLAG_FD_ATR_ENABLED |
14389                                I40E_FLAG_DCB_ENABLED    |
14390                                I40E_FLAG_VMDQ_ENABLED);
14391                 pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14392         } else {
14393                 /* Not enough queues for all TCs */
14394                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
14395                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
14396                         pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
14397                                         I40E_FLAG_DCB_ENABLED);
14398                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
14399                 }
14400
14401                 /* limit lan qps to the smaller of qps, cpus or msix */
14402                 q_max = max_t(int, pf->rss_size_max, num_online_cpus());
14403                 q_max = min_t(int, q_max, pf->hw.func_caps.num_tx_qp);
14404                 q_max = min_t(int, q_max, pf->hw.func_caps.num_msix_vectors);
14405                 pf->num_lan_qps = q_max;
14406
14407                 queues_left -= pf->num_lan_qps;
14408         }
14409
14410         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14411                 if (queues_left > 1) {
14412                         queues_left -= 1; /* save 1 queue for FD */
14413                 } else {
14414                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
14415                         pf->flags |= I40E_FLAG_FD_SB_INACTIVE;
14416                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
14417                 }
14418         }
14419
14420         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
14421             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
14422                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
14423                                         (queues_left / pf->num_vf_qps));
14424                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
14425         }
14426
14427         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
14428             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
14429                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
14430                                           (queues_left / pf->num_vmdq_qps));
14431                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
14432         }
14433
14434         pf->queues_left = queues_left;
14435         dev_dbg(&pf->pdev->dev,
14436                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
14437                 pf->hw.func_caps.num_tx_qp,
14438                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
14439                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
14440                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
14441                 queues_left);
14442 }
14443
14444 /**
14445  * i40e_setup_pf_filter_control - Setup PF static filter control
14446  * @pf: PF to be setup
14447  *
14448  * i40e_setup_pf_filter_control sets up a PF's initial filter control
14449  * settings. If PE/FCoE are enabled then it will also set the per PF
14450  * based filter sizes required for them. It also enables Flow director,
14451  * ethertype and macvlan type filter settings for the pf.
14452  *
14453  * Returns 0 on success, negative on failure
14454  **/
14455 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
14456 {
14457         struct i40e_filter_control_settings *settings = &pf->filter_settings;
14458
14459         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
14460
14461         /* Flow Director is enabled */
14462         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
14463                 settings->enable_fdir = true;
14464
14465         /* Ethtype and MACVLAN filters enabled for PF */
14466         settings->enable_ethtype = true;
14467         settings->enable_macvlan = true;
14468
14469         if (i40e_set_filter_control(&pf->hw, settings))
14470                 return -ENOENT;
14471
14472         return 0;
14473 }
14474
14475 #define INFO_STRING_LEN 255
14476 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
14477 static void i40e_print_features(struct i40e_pf *pf)
14478 {
14479         struct i40e_hw *hw = &pf->hw;
14480         char *buf;
14481         int i;
14482
14483         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
14484         if (!buf)
14485                 return;
14486
14487         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
14488 #ifdef CONFIG_PCI_IOV
14489         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
14490 #endif
14491         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
14492                       pf->hw.func_caps.num_vsis,
14493                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
14494         if (pf->flags & I40E_FLAG_RSS_ENABLED)
14495                 i += snprintf(&buf[i], REMAIN(i), " RSS");
14496         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
14497                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
14498         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
14499                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
14500                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
14501         }
14502         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
14503                 i += snprintf(&buf[i], REMAIN(i), " DCB");
14504         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
14505         i += snprintf(&buf[i], REMAIN(i), " Geneve");
14506         if (pf->flags & I40E_FLAG_PTP)
14507                 i += snprintf(&buf[i], REMAIN(i), " PTP");
14508         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
14509                 i += snprintf(&buf[i], REMAIN(i), " VEB");
14510         else
14511                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
14512
14513         dev_info(&pf->pdev->dev, "%s\n", buf);
14514         kfree(buf);
14515         WARN_ON(i > INFO_STRING_LEN);
14516 }
14517
14518 /**
14519  * i40e_get_platform_mac_addr - get platform-specific MAC address
14520  * @pdev: PCI device information struct
14521  * @pf: board private structure
14522  *
14523  * Look up the MAC address for the device. First we'll try
14524  * eth_platform_get_mac_address, which will check Open Firmware, or arch
14525  * specific fallback. Otherwise, we'll default to the stored value in
14526  * firmware.
14527  **/
14528 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
14529 {
14530         if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
14531                 i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
14532 }
14533
14534 /**
14535  * i40e_set_fec_in_flags - helper function for setting FEC options in flags
14536  * @fec_cfg: FEC option to set in flags
14537  * @flags: ptr to flags in which we set FEC option
14538  **/
14539 void i40e_set_fec_in_flags(u8 fec_cfg, u32 *flags)
14540 {
14541         if (fec_cfg & I40E_AQ_SET_FEC_AUTO)
14542                 *flags |= I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC;
14543         if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_RS) ||
14544             (fec_cfg & I40E_AQ_SET_FEC_ABILITY_RS)) {
14545                 *flags |= I40E_FLAG_RS_FEC;
14546                 *flags &= ~I40E_FLAG_BASE_R_FEC;
14547         }
14548         if ((fec_cfg & I40E_AQ_SET_FEC_REQUEST_KR) ||
14549             (fec_cfg & I40E_AQ_SET_FEC_ABILITY_KR)) {
14550                 *flags |= I40E_FLAG_BASE_R_FEC;
14551                 *flags &= ~I40E_FLAG_RS_FEC;
14552         }
14553         if (fec_cfg == 0)
14554                 *flags &= ~(I40E_FLAG_RS_FEC | I40E_FLAG_BASE_R_FEC);
14555 }
14556
14557 /**
14558  * i40e_check_recovery_mode - check if we are running transition firmware
14559  * @pf: board private structure
14560  *
14561  * Check registers indicating the firmware runs in recovery mode. Sets the
14562  * appropriate driver state.
14563  *
14564  * Returns true if the recovery mode was detected, false otherwise
14565  **/
14566 static bool i40e_check_recovery_mode(struct i40e_pf *pf)
14567 {
14568         u32 val = rd32(&pf->hw, I40E_GL_FWSTS) & I40E_GL_FWSTS_FWS1B_MASK;
14569         bool is_recovery_mode = false;
14570
14571         if (pf->hw.mac.type == I40E_MAC_XL710)
14572                 is_recovery_mode =
14573                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14574                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK ||
14575                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_TRANSITION_MASK ||
14576                 val == I40E_XL710_GL_FWSTS_FWS1B_REC_MOD_NVM_MASK;
14577         if (pf->hw.mac.type == I40E_MAC_X722)
14578                 is_recovery_mode =
14579                 val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_CORER_MASK ||
14580                 val == I40E_X722_GL_FWSTS_FWS1B_REC_MOD_GLOBR_MASK;
14581         if (is_recovery_mode) {
14582                 dev_notice(&pf->pdev->dev, "Firmware recovery mode detected. Limiting functionality.\n");
14583                 dev_notice(&pf->pdev->dev, "Refer to the Intel(R) Ethernet Adapters and Devices User Guide for details on firmware recovery mode.\n");
14584                 set_bit(__I40E_RECOVERY_MODE, pf->state);
14585
14586                 return true;
14587         }
14588         if (test_and_clear_bit(__I40E_RECOVERY_MODE, pf->state))
14589                 dev_info(&pf->pdev->dev, "Reinitializing in normal mode with full functionality.\n");
14590
14591         return false;
14592 }
14593
14594 /**
14595  * i40e_pf_loop_reset - perform reset in a loop.
14596  * @pf: board private structure
14597  *
14598  * This function is useful when a NIC is about to enter recovery mode.
14599  * When a NIC's internal data structures are corrupted the NIC's
14600  * firmware is going to enter recovery mode.
14601  * Right after a POR it takes about 7 minutes for firmware to enter
14602  * recovery mode. Until that time a NIC is in some kind of intermediate
14603  * state. After that time period the NIC almost surely enters
14604  * recovery mode. The only way for a driver to detect intermediate
14605  * state is to issue a series of pf-resets and check a return value.
14606  * If a PF reset returns success then the firmware could be in recovery
14607  * mode so the caller of this code needs to check for recovery mode
14608  * if this function returns success. There is a little chance that
14609  * firmware will hang in intermediate state forever.
14610  * Since waiting 7 minutes is quite a lot of time this function waits
14611  * 10 seconds and then gives up by returning an error.
14612  *
14613  * Return 0 on success, negative on failure.
14614  **/
14615 static i40e_status i40e_pf_loop_reset(struct i40e_pf *pf)
14616 {
14617         const unsigned short MAX_CNT = 1000;
14618         const unsigned short MSECS = 10;
14619         struct i40e_hw *hw = &pf->hw;
14620         i40e_status ret;
14621         int cnt;
14622
14623         for (cnt = 0; cnt < MAX_CNT; ++cnt) {
14624                 ret = i40e_pf_reset(hw);
14625                 if (!ret)
14626                         break;
14627                 msleep(MSECS);
14628         }
14629
14630         if (cnt == MAX_CNT) {
14631                 dev_info(&pf->pdev->dev, "PF reset failed: %d\n", ret);
14632                 return ret;
14633         }
14634
14635         pf->pfr_count++;
14636         return ret;
14637 }
14638
14639 /**
14640  * i40e_init_recovery_mode - initialize subsystems needed in recovery mode
14641  * @pf: board private structure
14642  * @hw: ptr to the hardware info
14643  *
14644  * This function does a minimal setup of all subsystems needed for running
14645  * recovery mode.
14646  *
14647  * Returns 0 on success, negative on failure
14648  **/
14649 static int i40e_init_recovery_mode(struct i40e_pf *pf, struct i40e_hw *hw)
14650 {
14651         struct i40e_vsi *vsi;
14652         int err;
14653         int v_idx;
14654
14655         pci_save_state(pf->pdev);
14656
14657         /* set up periodic task facility */
14658         timer_setup(&pf->service_timer, i40e_service_timer, 0);
14659         pf->service_timer_period = HZ;
14660
14661         INIT_WORK(&pf->service_task, i40e_service_task);
14662         clear_bit(__I40E_SERVICE_SCHED, pf->state);
14663
14664         err = i40e_init_interrupt_scheme(pf);
14665         if (err)
14666                 goto err_switch_setup;
14667
14668         /* The number of VSIs reported by the FW is the minimum guaranteed
14669          * to us; HW supports far more and we share the remaining pool with
14670          * the other PFs. We allocate space for more than the guarantee with
14671          * the understanding that we might not get them all later.
14672          */
14673         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
14674                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
14675         else
14676                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
14677
14678         /* Set up the vsi struct and our local tracking of the MAIN PF vsi. */
14679         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
14680                           GFP_KERNEL);
14681         if (!pf->vsi) {
14682                 err = -ENOMEM;
14683                 goto err_switch_setup;
14684         }
14685
14686         /* We allocate one VSI which is needed as absolute minimum
14687          * in order to register the netdev
14688          */
14689         v_idx = i40e_vsi_mem_alloc(pf, I40E_VSI_MAIN);
14690         if (v_idx < 0)
14691                 goto err_switch_setup;
14692         pf->lan_vsi = v_idx;
14693         vsi = pf->vsi[v_idx];
14694         if (!vsi)
14695                 goto err_switch_setup;
14696         vsi->alloc_queue_pairs = 1;
14697         err = i40e_config_netdev(vsi);
14698         if (err)
14699                 goto err_switch_setup;
14700         err = register_netdev(vsi->netdev);
14701         if (err)
14702                 goto err_switch_setup;
14703         vsi->netdev_registered = true;
14704         i40e_dbg_pf_init(pf);
14705
14706         err = i40e_setup_misc_vector_for_recovery_mode(pf);
14707         if (err)
14708                 goto err_switch_setup;
14709
14710         /* tell the firmware that we're starting */
14711         i40e_send_version(pf);
14712
14713         /* since everything's happy, start the service_task timer */
14714         mod_timer(&pf->service_timer,
14715                   round_jiffies(jiffies + pf->service_timer_period));
14716
14717         return 0;
14718
14719 err_switch_setup:
14720         i40e_reset_interrupt_capability(pf);
14721         del_timer_sync(&pf->service_timer);
14722         i40e_shutdown_adminq(hw);
14723         iounmap(hw->hw_addr);
14724         pci_disable_pcie_error_reporting(pf->pdev);
14725         pci_release_mem_regions(pf->pdev);
14726         pci_disable_device(pf->pdev);
14727         kfree(pf);
14728
14729         return err;
14730 }
14731
14732 /**
14733  * i40e_probe - Device initialization routine
14734  * @pdev: PCI device information struct
14735  * @ent: entry in i40e_pci_tbl
14736  *
14737  * i40e_probe initializes a PF identified by a pci_dev structure.
14738  * The OS initialization, configuring of the PF private structure,
14739  * and a hardware reset occur.
14740  *
14741  * Returns 0 on success, negative on failure
14742  **/
14743 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
14744 {
14745         struct i40e_aq_get_phy_abilities_resp abilities;
14746         struct i40e_pf *pf;
14747         struct i40e_hw *hw;
14748         static u16 pfs_found;
14749         u16 wol_nvm_bits;
14750         u16 link_status;
14751         int err;
14752         u32 val;
14753         u32 i;
14754         u8 set_fc_aq_fail;
14755
14756         err = pci_enable_device_mem(pdev);
14757         if (err)
14758                 return err;
14759
14760         /* set up for high or low dma */
14761         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
14762         if (err) {
14763                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
14764                 if (err) {
14765                         dev_err(&pdev->dev,
14766                                 "DMA configuration failed: 0x%x\n", err);
14767                         goto err_dma;
14768                 }
14769         }
14770
14771         /* set up pci connections */
14772         err = pci_request_mem_regions(pdev, i40e_driver_name);
14773         if (err) {
14774                 dev_info(&pdev->dev,
14775                          "pci_request_selected_regions failed %d\n", err);
14776                 goto err_pci_reg;
14777         }
14778
14779         pci_enable_pcie_error_reporting(pdev);
14780         pci_set_master(pdev);
14781
14782         /* Now that we have a PCI connection, we need to do the
14783          * low level device setup.  This is primarily setting up
14784          * the Admin Queue structures and then querying for the
14785          * device's current profile information.
14786          */
14787         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
14788         if (!pf) {
14789                 err = -ENOMEM;
14790                 goto err_pf_alloc;
14791         }
14792         pf->next_vsi = 0;
14793         pf->pdev = pdev;
14794         set_bit(__I40E_DOWN, pf->state);
14795
14796         hw = &pf->hw;
14797         hw->back = pf;
14798
14799         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
14800                                 I40E_MAX_CSR_SPACE);
14801         /* We believe that the highest register to read is
14802          * I40E_GLGEN_STAT_CLEAR, so we check if the BAR size
14803          * is not less than that before mapping to prevent a
14804          * kernel panic.
14805          */
14806         if (pf->ioremap_len < I40E_GLGEN_STAT_CLEAR) {
14807                 dev_err(&pdev->dev, "Cannot map registers, bar size 0x%X too small, aborting\n",
14808                         pf->ioremap_len);
14809                 err = -ENOMEM;
14810                 goto err_ioremap;
14811         }
14812         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
14813         if (!hw->hw_addr) {
14814                 err = -EIO;
14815                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
14816                          (unsigned int)pci_resource_start(pdev, 0),
14817                          pf->ioremap_len, err);
14818                 goto err_ioremap;
14819         }
14820         hw->vendor_id = pdev->vendor;
14821         hw->device_id = pdev->device;
14822         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
14823         hw->subsystem_vendor_id = pdev->subsystem_vendor;
14824         hw->subsystem_device_id = pdev->subsystem_device;
14825         hw->bus.device = PCI_SLOT(pdev->devfn);
14826         hw->bus.func = PCI_FUNC(pdev->devfn);
14827         hw->bus.bus_id = pdev->bus->number;
14828         pf->instance = pfs_found;
14829
14830         /* Select something other than the 802.1ad ethertype for the
14831          * switch to use internally and drop on ingress.
14832          */
14833         hw->switch_tag = 0xffff;
14834         hw->first_tag = ETH_P_8021AD;
14835         hw->second_tag = ETH_P_8021Q;
14836
14837         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
14838         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
14839         INIT_LIST_HEAD(&pf->ddp_old_prof);
14840
14841         /* set up the locks for the AQ, do this only once in probe
14842          * and destroy them only once in remove
14843          */
14844         mutex_init(&hw->aq.asq_mutex);
14845         mutex_init(&hw->aq.arq_mutex);
14846
14847         pf->msg_enable = netif_msg_init(debug,
14848                                         NETIF_MSG_DRV |
14849                                         NETIF_MSG_PROBE |
14850                                         NETIF_MSG_LINK);
14851         if (debug < -1)
14852                 pf->hw.debug_mask = debug;
14853
14854         /* do a special CORER for clearing PXE mode once at init */
14855         if (hw->revision_id == 0 &&
14856             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
14857                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
14858                 i40e_flush(hw);
14859                 msleep(200);
14860                 pf->corer_count++;
14861
14862                 i40e_clear_pxe_mode(hw);
14863         }
14864
14865         /* Reset here to make sure all is clean and to define PF 'n' */
14866         i40e_clear_hw(hw);
14867
14868         err = i40e_set_mac_type(hw);
14869         if (err) {
14870                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14871                          err);
14872                 goto err_pf_reset;
14873         }
14874
14875         err = i40e_pf_loop_reset(pf);
14876         if (err) {
14877                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
14878                 goto err_pf_reset;
14879         }
14880
14881         i40e_check_recovery_mode(pf);
14882
14883         hw->aq.num_arq_entries = I40E_AQ_LEN;
14884         hw->aq.num_asq_entries = I40E_AQ_LEN;
14885         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14886         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
14887         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
14888
14889         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
14890                  "%s-%s:misc",
14891                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
14892
14893         err = i40e_init_shared_code(hw);
14894         if (err) {
14895                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
14896                          err);
14897                 goto err_pf_reset;
14898         }
14899
14900         /* set up a default setting for link flow control */
14901         pf->hw.fc.requested_mode = I40E_FC_NONE;
14902
14903         err = i40e_init_adminq(hw);
14904         if (err) {
14905                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
14906                         dev_info(&pdev->dev,
14907                                  "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",
14908                                  hw->aq.api_maj_ver,
14909                                  hw->aq.api_min_ver,
14910                                  I40E_FW_API_VERSION_MAJOR,
14911                                  I40E_FW_MINOR_VERSION(hw));
14912                 else
14913                         dev_info(&pdev->dev,
14914                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
14915
14916                 goto err_pf_reset;
14917         }
14918         i40e_get_oem_version(hw);
14919
14920         /* provide nvm, fw, api versions, vendor:device id, subsys vendor:device id */
14921         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s [%04x:%04x] [%04x:%04x]\n",
14922                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
14923                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
14924                  i40e_nvm_version_str(hw), hw->vendor_id, hw->device_id,
14925                  hw->subsystem_vendor_id, hw->subsystem_device_id);
14926
14927         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
14928             hw->aq.api_min_ver > I40E_FW_MINOR_VERSION(hw))
14929                 dev_info(&pdev->dev,
14930                          "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",
14931                          hw->aq.api_maj_ver,
14932                          hw->aq.api_min_ver,
14933                          I40E_FW_API_VERSION_MAJOR,
14934                          I40E_FW_MINOR_VERSION(hw));
14935         else if (hw->aq.api_maj_ver == 1 && hw->aq.api_min_ver < 4)
14936                 dev_info(&pdev->dev,
14937                          "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",
14938                          hw->aq.api_maj_ver,
14939                          hw->aq.api_min_ver,
14940                          I40E_FW_API_VERSION_MAJOR,
14941                          I40E_FW_MINOR_VERSION(hw));
14942
14943         i40e_verify_eeprom(pf);
14944
14945         /* Rev 0 hardware was never productized */
14946         if (hw->revision_id < 1)
14947                 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");
14948
14949         i40e_clear_pxe_mode(hw);
14950
14951         err = i40e_get_capabilities(pf, i40e_aqc_opc_list_func_capabilities);
14952         if (err)
14953                 goto err_adminq_setup;
14954
14955         err = i40e_sw_init(pf);
14956         if (err) {
14957                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
14958                 goto err_sw_init;
14959         }
14960
14961         if (test_bit(__I40E_RECOVERY_MODE, pf->state))
14962                 return i40e_init_recovery_mode(pf, hw);
14963
14964         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
14965                                 hw->func_caps.num_rx_qp, 0, 0);
14966         if (err) {
14967                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
14968                 goto err_init_lan_hmc;
14969         }
14970
14971         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
14972         if (err) {
14973                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
14974                 err = -ENOENT;
14975                 goto err_configure_lan_hmc;
14976         }
14977
14978         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
14979          * Ignore error return codes because if it was already disabled via
14980          * hardware settings this will fail
14981          */
14982         if (pf->hw_features & I40E_HW_STOP_FW_LLDP) {
14983                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
14984                 i40e_aq_stop_lldp(hw, true, false, NULL);
14985         }
14986
14987         /* allow a platform config to override the HW addr */
14988         i40e_get_platform_mac_addr(pdev, pf);
14989
14990         if (!is_valid_ether_addr(hw->mac.addr)) {
14991                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
14992                 err = -EIO;
14993                 goto err_mac_addr;
14994         }
14995         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
14996         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
14997         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
14998         if (is_valid_ether_addr(hw->mac.port_addr))
14999                 pf->hw_features |= I40E_HW_PORT_ID_VALID;
15000
15001         pci_set_drvdata(pdev, pf);
15002         pci_save_state(pdev);
15003
15004         dev_info(&pdev->dev,
15005                  (pf->flags & I40E_FLAG_DISABLE_FW_LLDP) ?
15006                         "FW LLDP is disabled\n" :
15007                         "FW LLDP is enabled\n");
15008
15009         /* Enable FW to write default DCB config on link-up */
15010         i40e_aq_set_dcb_parameters(hw, true, NULL);
15011
15012 #ifdef CONFIG_I40E_DCB
15013         err = i40e_init_pf_dcb(pf);
15014         if (err) {
15015                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
15016                 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
15017                 /* Continue without DCB enabled */
15018         }
15019 #endif /* CONFIG_I40E_DCB */
15020
15021         /* set up periodic task facility */
15022         timer_setup(&pf->service_timer, i40e_service_timer, 0);
15023         pf->service_timer_period = HZ;
15024
15025         INIT_WORK(&pf->service_task, i40e_service_task);
15026         clear_bit(__I40E_SERVICE_SCHED, pf->state);
15027
15028         /* NVM bit on means WoL disabled for the port */
15029         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
15030         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
15031                 pf->wol_en = false;
15032         else
15033                 pf->wol_en = true;
15034         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
15035
15036         /* set up the main switch operations */
15037         i40e_determine_queue_usage(pf);
15038         err = i40e_init_interrupt_scheme(pf);
15039         if (err)
15040                 goto err_switch_setup;
15041
15042         /* The number of VSIs reported by the FW is the minimum guaranteed
15043          * to us; HW supports far more and we share the remaining pool with
15044          * the other PFs. We allocate space for more than the guarantee with
15045          * the understanding that we might not get them all later.
15046          */
15047         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
15048                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
15049         else
15050                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
15051
15052         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
15053         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
15054                           GFP_KERNEL);
15055         if (!pf->vsi) {
15056                 err = -ENOMEM;
15057                 goto err_switch_setup;
15058         }
15059
15060 #ifdef CONFIG_PCI_IOV
15061         /* prep for VF support */
15062         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15063             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15064             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15065                 if (pci_num_vf(pdev))
15066                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
15067         }
15068 #endif
15069         err = i40e_setup_pf_switch(pf, false);
15070         if (err) {
15071                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
15072                 goto err_vsis;
15073         }
15074         INIT_LIST_HEAD(&pf->vsi[pf->lan_vsi]->ch_list);
15075
15076         /* Make sure flow control is set according to current settings */
15077         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
15078         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
15079                 dev_dbg(&pf->pdev->dev,
15080                         "Set fc with err %s aq_err %s on get_phy_cap\n",
15081                         i40e_stat_str(hw, err),
15082                         i40e_aq_str(hw, hw->aq.asq_last_status));
15083         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
15084                 dev_dbg(&pf->pdev->dev,
15085                         "Set fc with err %s aq_err %s on set_phy_config\n",
15086                         i40e_stat_str(hw, err),
15087                         i40e_aq_str(hw, hw->aq.asq_last_status));
15088         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
15089                 dev_dbg(&pf->pdev->dev,
15090                         "Set fc with err %s aq_err %s on get_link_info\n",
15091                         i40e_stat_str(hw, err),
15092                         i40e_aq_str(hw, hw->aq.asq_last_status));
15093
15094         /* if FDIR VSI was set up, start it now */
15095         for (i = 0; i < pf->num_alloc_vsi; i++) {
15096                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
15097                         i40e_vsi_open(pf->vsi[i]);
15098                         break;
15099                 }
15100         }
15101
15102         /* The driver only wants link up/down and module qualification
15103          * reports from firmware.  Note the negative logic.
15104          */
15105         err = i40e_aq_set_phy_int_mask(&pf->hw,
15106                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
15107                                          I40E_AQ_EVENT_MEDIA_NA |
15108                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
15109         if (err)
15110                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
15111                          i40e_stat_str(&pf->hw, err),
15112                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15113
15114         /* Reconfigure hardware for allowing smaller MSS in the case
15115          * of TSO, so that we avoid the MDD being fired and causing
15116          * a reset in the case of small MSS+TSO.
15117          */
15118         val = rd32(hw, I40E_REG_MSS);
15119         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
15120                 val &= ~I40E_REG_MSS_MIN_MASK;
15121                 val |= I40E_64BYTE_MSS;
15122                 wr32(hw, I40E_REG_MSS, val);
15123         }
15124
15125         if (pf->hw_features & I40E_HW_RESTART_AUTONEG) {
15126                 msleep(75);
15127                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
15128                 if (err)
15129                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
15130                                  i40e_stat_str(&pf->hw, err),
15131                                  i40e_aq_str(&pf->hw,
15132                                              pf->hw.aq.asq_last_status));
15133         }
15134         /* The main driver is (mostly) up and happy. We need to set this state
15135          * before setting up the misc vector or we get a race and the vector
15136          * ends up disabled forever.
15137          */
15138         clear_bit(__I40E_DOWN, pf->state);
15139
15140         /* In case of MSIX we are going to setup the misc vector right here
15141          * to handle admin queue events etc. In case of legacy and MSI
15142          * the misc functionality and queue processing is combined in
15143          * the same vector and that gets setup at open.
15144          */
15145         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
15146                 err = i40e_setup_misc_vector(pf);
15147                 if (err) {
15148                         dev_info(&pdev->dev,
15149                                  "setup of misc vector failed: %d\n", err);
15150                         goto err_vsis;
15151                 }
15152         }
15153
15154 #ifdef CONFIG_PCI_IOV
15155         /* prep for VF support */
15156         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
15157             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
15158             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
15159                 /* disable link interrupts for VFs */
15160                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
15161                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
15162                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
15163                 i40e_flush(hw);
15164
15165                 if (pci_num_vf(pdev)) {
15166                         dev_info(&pdev->dev,
15167                                  "Active VFs found, allocating resources.\n");
15168                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
15169                         if (err)
15170                                 dev_info(&pdev->dev,
15171                                          "Error %d allocating resources for existing VFs\n",
15172                                          err);
15173                 }
15174         }
15175 #endif /* CONFIG_PCI_IOV */
15176
15177         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15178                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
15179                                                       pf->num_iwarp_msix,
15180                                                       I40E_IWARP_IRQ_PILE_ID);
15181                 if (pf->iwarp_base_vector < 0) {
15182                         dev_info(&pdev->dev,
15183                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
15184                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
15185                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
15186                 }
15187         }
15188
15189         i40e_dbg_pf_init(pf);
15190
15191         /* tell the firmware that we're starting */
15192         i40e_send_version(pf);
15193
15194         /* since everything's happy, start the service_task timer */
15195         mod_timer(&pf->service_timer,
15196                   round_jiffies(jiffies + pf->service_timer_period));
15197
15198         /* add this PF to client device list and launch a client service task */
15199         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15200                 err = i40e_lan_add_device(pf);
15201                 if (err)
15202                         dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
15203                                  err);
15204         }
15205
15206 #define PCI_SPEED_SIZE 8
15207 #define PCI_WIDTH_SIZE 8
15208         /* Devices on the IOSF bus do not have this information
15209          * and will report PCI Gen 1 x 1 by default so don't bother
15210          * checking them.
15211          */
15212         if (!(pf->hw_features & I40E_HW_NO_PCI_LINK_CHECK)) {
15213                 char speed[PCI_SPEED_SIZE] = "Unknown";
15214                 char width[PCI_WIDTH_SIZE] = "Unknown";
15215
15216                 /* Get the negotiated link width and speed from PCI config
15217                  * space
15218                  */
15219                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
15220                                           &link_status);
15221
15222                 i40e_set_pci_config_data(hw, link_status);
15223
15224                 switch (hw->bus.speed) {
15225                 case i40e_bus_speed_8000:
15226                         strlcpy(speed, "8.0", PCI_SPEED_SIZE); break;
15227                 case i40e_bus_speed_5000:
15228                         strlcpy(speed, "5.0", PCI_SPEED_SIZE); break;
15229                 case i40e_bus_speed_2500:
15230                         strlcpy(speed, "2.5", PCI_SPEED_SIZE); break;
15231                 default:
15232                         break;
15233                 }
15234                 switch (hw->bus.width) {
15235                 case i40e_bus_width_pcie_x8:
15236                         strlcpy(width, "8", PCI_WIDTH_SIZE); break;
15237                 case i40e_bus_width_pcie_x4:
15238                         strlcpy(width, "4", PCI_WIDTH_SIZE); break;
15239                 case i40e_bus_width_pcie_x2:
15240                         strlcpy(width, "2", PCI_WIDTH_SIZE); break;
15241                 case i40e_bus_width_pcie_x1:
15242                         strlcpy(width, "1", PCI_WIDTH_SIZE); break;
15243                 default:
15244                         break;
15245                 }
15246
15247                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
15248                          speed, width);
15249
15250                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
15251                     hw->bus.speed < i40e_bus_speed_8000) {
15252                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
15253                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
15254                 }
15255         }
15256
15257         /* get the requested speeds from the fw */
15258         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
15259         if (err)
15260                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
15261                         i40e_stat_str(&pf->hw, err),
15262                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15263         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
15264
15265         /* set the FEC config due to the board capabilities */
15266         i40e_set_fec_in_flags(abilities.fec_cfg_curr_mod_ext_info, &pf->flags);
15267
15268         /* get the supported phy types from the fw */
15269         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
15270         if (err)
15271                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
15272                         i40e_stat_str(&pf->hw, err),
15273                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
15274
15275         /* Add a filter to drop all Flow control frames from any VSI from being
15276          * transmitted. By doing so we stop a malicious VF from sending out
15277          * PAUSE or PFC frames and potentially controlling traffic for other
15278          * PF/VF VSIs.
15279          * The FW can still send Flow control frames if enabled.
15280          */
15281         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
15282                                                        pf->main_vsi_seid);
15283
15284         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
15285                 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
15286                 pf->hw_features |= I40E_HW_PHY_CONTROLS_LEDS;
15287         if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
15288                 pf->hw_features |= I40E_HW_HAVE_CRT_RETIMER;
15289         /* print a string summarizing features */
15290         i40e_print_features(pf);
15291
15292         return 0;
15293
15294         /* Unwind what we've done if something failed in the setup */
15295 err_vsis:
15296         set_bit(__I40E_DOWN, pf->state);
15297         i40e_clear_interrupt_scheme(pf);
15298         kfree(pf->vsi);
15299 err_switch_setup:
15300         i40e_reset_interrupt_capability(pf);
15301         del_timer_sync(&pf->service_timer);
15302 err_mac_addr:
15303 err_configure_lan_hmc:
15304         (void)i40e_shutdown_lan_hmc(hw);
15305 err_init_lan_hmc:
15306         kfree(pf->qp_pile);
15307 err_sw_init:
15308 err_adminq_setup:
15309 err_pf_reset:
15310         iounmap(hw->hw_addr);
15311 err_ioremap:
15312         kfree(pf);
15313 err_pf_alloc:
15314         pci_disable_pcie_error_reporting(pdev);
15315         pci_release_mem_regions(pdev);
15316 err_pci_reg:
15317 err_dma:
15318         pci_disable_device(pdev);
15319         return err;
15320 }
15321
15322 /**
15323  * i40e_remove - Device removal routine
15324  * @pdev: PCI device information struct
15325  *
15326  * i40e_remove is called by the PCI subsystem to alert the driver
15327  * that is should release a PCI device.  This could be caused by a
15328  * Hot-Plug event, or because the driver is going to be removed from
15329  * memory.
15330  **/
15331 static void i40e_remove(struct pci_dev *pdev)
15332 {
15333         struct i40e_pf *pf = pci_get_drvdata(pdev);
15334         struct i40e_hw *hw = &pf->hw;
15335         i40e_status ret_code;
15336         int i;
15337
15338         i40e_dbg_pf_exit(pf);
15339
15340         i40e_ptp_stop(pf);
15341
15342         /* Disable RSS in hw */
15343         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
15344         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
15345
15346         /* no more scheduling of any task */
15347         set_bit(__I40E_SUSPENDED, pf->state);
15348         set_bit(__I40E_DOWN, pf->state);
15349         if (pf->service_timer.function)
15350                 del_timer_sync(&pf->service_timer);
15351         if (pf->service_task.func)
15352                 cancel_work_sync(&pf->service_task);
15353
15354         if (test_bit(__I40E_RECOVERY_MODE, pf->state)) {
15355                 struct i40e_vsi *vsi = pf->vsi[0];
15356
15357                 /* We know that we have allocated only one vsi for this PF,
15358                  * it was just for registering netdevice, so the interface
15359                  * could be visible in the 'ifconfig' output
15360                  */
15361                 unregister_netdev(vsi->netdev);
15362                 free_netdev(vsi->netdev);
15363
15364                 goto unmap;
15365         }
15366
15367         /* Client close must be called explicitly here because the timer
15368          * has been stopped.
15369          */
15370         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15371
15372         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
15373                 i40e_free_vfs(pf);
15374                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
15375         }
15376
15377         i40e_fdir_teardown(pf);
15378
15379         /* If there is a switch structure or any orphans, remove them.
15380          * This will leave only the PF's VSI remaining.
15381          */
15382         for (i = 0; i < I40E_MAX_VEB; i++) {
15383                 if (!pf->veb[i])
15384                         continue;
15385
15386                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
15387                     pf->veb[i]->uplink_seid == 0)
15388                         i40e_switch_branch_release(pf->veb[i]);
15389         }
15390
15391         /* Now we can shutdown the PF's VSI, just before we kill
15392          * adminq and hmc.
15393          */
15394         if (pf->vsi[pf->lan_vsi])
15395                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
15396
15397         i40e_cloud_filter_exit(pf);
15398
15399         /* remove attached clients */
15400         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
15401                 ret_code = i40e_lan_del_device(pf);
15402                 if (ret_code)
15403                         dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
15404                                  ret_code);
15405         }
15406
15407         /* shutdown and destroy the HMC */
15408         if (hw->hmc.hmc_obj) {
15409                 ret_code = i40e_shutdown_lan_hmc(hw);
15410                 if (ret_code)
15411                         dev_warn(&pdev->dev,
15412                                  "Failed to destroy the HMC resources: %d\n",
15413                                  ret_code);
15414         }
15415
15416 unmap:
15417         /* Free MSI/legacy interrupt 0 when in recovery mode. */
15418         if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15419             !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15420                 free_irq(pf->pdev->irq, pf);
15421
15422         /* shutdown the adminq */
15423         i40e_shutdown_adminq(hw);
15424
15425         /* destroy the locks only once, here */
15426         mutex_destroy(&hw->aq.arq_mutex);
15427         mutex_destroy(&hw->aq.asq_mutex);
15428
15429         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
15430         rtnl_lock();
15431         i40e_clear_interrupt_scheme(pf);
15432         for (i = 0; i < pf->num_alloc_vsi; i++) {
15433                 if (pf->vsi[i]) {
15434                         if (!test_bit(__I40E_RECOVERY_MODE, pf->state))
15435                                 i40e_vsi_clear_rings(pf->vsi[i]);
15436                         i40e_vsi_clear(pf->vsi[i]);
15437                         pf->vsi[i] = NULL;
15438                 }
15439         }
15440         rtnl_unlock();
15441
15442         for (i = 0; i < I40E_MAX_VEB; i++) {
15443                 kfree(pf->veb[i]);
15444                 pf->veb[i] = NULL;
15445         }
15446
15447         kfree(pf->qp_pile);
15448         kfree(pf->vsi);
15449
15450         iounmap(hw->hw_addr);
15451         kfree(pf);
15452         pci_release_mem_regions(pdev);
15453
15454         pci_disable_pcie_error_reporting(pdev);
15455         pci_disable_device(pdev);
15456 }
15457
15458 /**
15459  * i40e_pci_error_detected - warning that something funky happened in PCI land
15460  * @pdev: PCI device information struct
15461  * @error: the type of PCI error
15462  *
15463  * Called to warn that something happened and the error handling steps
15464  * are in progress.  Allows the driver to quiesce things, be ready for
15465  * remediation.
15466  **/
15467 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
15468                                                 enum pci_channel_state error)
15469 {
15470         struct i40e_pf *pf = pci_get_drvdata(pdev);
15471
15472         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
15473
15474         if (!pf) {
15475                 dev_info(&pdev->dev,
15476                          "Cannot recover - error happened during device probe\n");
15477                 return PCI_ERS_RESULT_DISCONNECT;
15478         }
15479
15480         /* shutdown all operations */
15481         if (!test_bit(__I40E_SUSPENDED, pf->state))
15482                 i40e_prep_for_reset(pf, false);
15483
15484         /* Request a slot reset */
15485         return PCI_ERS_RESULT_NEED_RESET;
15486 }
15487
15488 /**
15489  * i40e_pci_error_slot_reset - a PCI slot reset just happened
15490  * @pdev: PCI device information struct
15491  *
15492  * Called to find if the driver can work with the device now that
15493  * the pci slot has been reset.  If a basic connection seems good
15494  * (registers are readable and have sane content) then return a
15495  * happy little PCI_ERS_RESULT_xxx.
15496  **/
15497 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
15498 {
15499         struct i40e_pf *pf = pci_get_drvdata(pdev);
15500         pci_ers_result_t result;
15501         u32 reg;
15502
15503         dev_dbg(&pdev->dev, "%s\n", __func__);
15504         if (pci_enable_device_mem(pdev)) {
15505                 dev_info(&pdev->dev,
15506                          "Cannot re-enable PCI device after reset.\n");
15507                 result = PCI_ERS_RESULT_DISCONNECT;
15508         } else {
15509                 pci_set_master(pdev);
15510                 pci_restore_state(pdev);
15511                 pci_save_state(pdev);
15512                 pci_wake_from_d3(pdev, false);
15513
15514                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
15515                 if (reg == 0)
15516                         result = PCI_ERS_RESULT_RECOVERED;
15517                 else
15518                         result = PCI_ERS_RESULT_DISCONNECT;
15519         }
15520
15521         return result;
15522 }
15523
15524 /**
15525  * i40e_pci_error_reset_prepare - prepare device driver for pci reset
15526  * @pdev: PCI device information struct
15527  */
15528 static void i40e_pci_error_reset_prepare(struct pci_dev *pdev)
15529 {
15530         struct i40e_pf *pf = pci_get_drvdata(pdev);
15531
15532         i40e_prep_for_reset(pf, false);
15533 }
15534
15535 /**
15536  * i40e_pci_error_reset_done - pci reset done, device driver reset can begin
15537  * @pdev: PCI device information struct
15538  */
15539 static void i40e_pci_error_reset_done(struct pci_dev *pdev)
15540 {
15541         struct i40e_pf *pf = pci_get_drvdata(pdev);
15542
15543         i40e_reset_and_rebuild(pf, false, false);
15544 }
15545
15546 /**
15547  * i40e_pci_error_resume - restart operations after PCI error recovery
15548  * @pdev: PCI device information struct
15549  *
15550  * Called to allow the driver to bring things back up after PCI error
15551  * and/or reset recovery has finished.
15552  **/
15553 static void i40e_pci_error_resume(struct pci_dev *pdev)
15554 {
15555         struct i40e_pf *pf = pci_get_drvdata(pdev);
15556
15557         dev_dbg(&pdev->dev, "%s\n", __func__);
15558         if (test_bit(__I40E_SUSPENDED, pf->state))
15559                 return;
15560
15561         i40e_handle_reset_warning(pf, false);
15562 }
15563
15564 /**
15565  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
15566  * using the mac_address_write admin q function
15567  * @pf: pointer to i40e_pf struct
15568  **/
15569 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
15570 {
15571         struct i40e_hw *hw = &pf->hw;
15572         i40e_status ret;
15573         u8 mac_addr[6];
15574         u16 flags = 0;
15575
15576         /* Get current MAC address in case it's an LAA */
15577         if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
15578                 ether_addr_copy(mac_addr,
15579                                 pf->vsi[pf->lan_vsi]->netdev->dev_addr);
15580         } else {
15581                 dev_err(&pf->pdev->dev,
15582                         "Failed to retrieve MAC address; using default\n");
15583                 ether_addr_copy(mac_addr, hw->mac.addr);
15584         }
15585
15586         /* The FW expects the mac address write cmd to first be called with
15587          * one of these flags before calling it again with the multicast
15588          * enable flags.
15589          */
15590         flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
15591
15592         if (hw->func_caps.flex10_enable && hw->partition_id != 1)
15593                 flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
15594
15595         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15596         if (ret) {
15597                 dev_err(&pf->pdev->dev,
15598                         "Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
15599                 return;
15600         }
15601
15602         flags = I40E_AQC_MC_MAG_EN
15603                         | I40E_AQC_WOL_PRESERVE_ON_PFR
15604                         | I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
15605         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
15606         if (ret)
15607                 dev_err(&pf->pdev->dev,
15608                         "Failed to enable Multicast Magic Packet wake up\n");
15609 }
15610
15611 /**
15612  * i40e_shutdown - PCI callback for shutting down
15613  * @pdev: PCI device information struct
15614  **/
15615 static void i40e_shutdown(struct pci_dev *pdev)
15616 {
15617         struct i40e_pf *pf = pci_get_drvdata(pdev);
15618         struct i40e_hw *hw = &pf->hw;
15619
15620         set_bit(__I40E_SUSPENDED, pf->state);
15621         set_bit(__I40E_DOWN, pf->state);
15622
15623         del_timer_sync(&pf->service_timer);
15624         cancel_work_sync(&pf->service_task);
15625         i40e_cloud_filter_exit(pf);
15626         i40e_fdir_teardown(pf);
15627
15628         /* Client close must be called explicitly here because the timer
15629          * has been stopped.
15630          */
15631         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15632
15633         if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15634                 i40e_enable_mc_magic_wake(pf);
15635
15636         i40e_prep_for_reset(pf, false);
15637
15638         wr32(hw, I40E_PFPM_APM,
15639              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15640         wr32(hw, I40E_PFPM_WUFC,
15641              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15642
15643         /* Free MSI/legacy interrupt 0 when in recovery mode. */
15644         if (test_bit(__I40E_RECOVERY_MODE, pf->state) &&
15645             !(pf->flags & I40E_FLAG_MSIX_ENABLED))
15646                 free_irq(pf->pdev->irq, pf);
15647
15648         /* Since we're going to destroy queues during the
15649          * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15650          * whole section
15651          */
15652         rtnl_lock();
15653         i40e_clear_interrupt_scheme(pf);
15654         rtnl_unlock();
15655
15656         if (system_state == SYSTEM_POWER_OFF) {
15657                 pci_wake_from_d3(pdev, pf->wol_en);
15658                 pci_set_power_state(pdev, PCI_D3hot);
15659         }
15660 }
15661
15662 /**
15663  * i40e_suspend - PM callback for moving to D3
15664  * @dev: generic device information structure
15665  **/
15666 static int __maybe_unused i40e_suspend(struct device *dev)
15667 {
15668         struct i40e_pf *pf = dev_get_drvdata(dev);
15669         struct i40e_hw *hw = &pf->hw;
15670
15671         /* If we're already suspended, then there is nothing to do */
15672         if (test_and_set_bit(__I40E_SUSPENDED, pf->state))
15673                 return 0;
15674
15675         set_bit(__I40E_DOWN, pf->state);
15676
15677         /* Ensure service task will not be running */
15678         del_timer_sync(&pf->service_timer);
15679         cancel_work_sync(&pf->service_task);
15680
15681         /* Client close must be called explicitly here because the timer
15682          * has been stopped.
15683          */
15684         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
15685
15686         if (pf->wol_en && (pf->hw_features & I40E_HW_WOL_MC_MAGIC_PKT_WAKE))
15687                 i40e_enable_mc_magic_wake(pf);
15688
15689         /* Since we're going to destroy queues during the
15690          * i40e_clear_interrupt_scheme() we should hold the RTNL lock for this
15691          * whole section
15692          */
15693         rtnl_lock();
15694
15695         i40e_prep_for_reset(pf, true);
15696
15697         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
15698         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
15699
15700         /* Clear the interrupt scheme and release our IRQs so that the system
15701          * can safely hibernate even when there are a large number of CPUs.
15702          * Otherwise hibernation might fail when mapping all the vectors back
15703          * to CPU0.
15704          */
15705         i40e_clear_interrupt_scheme(pf);
15706
15707         rtnl_unlock();
15708
15709         return 0;
15710 }
15711
15712 /**
15713  * i40e_resume - PM callback for waking up from D3
15714  * @dev: generic device information structure
15715  **/
15716 static int __maybe_unused i40e_resume(struct device *dev)
15717 {
15718         struct i40e_pf *pf = dev_get_drvdata(dev);
15719         int err;
15720
15721         /* If we're not suspended, then there is nothing to do */
15722         if (!test_bit(__I40E_SUSPENDED, pf->state))
15723                 return 0;
15724
15725         /* We need to hold the RTNL lock prior to restoring interrupt schemes,
15726          * since we're going to be restoring queues
15727          */
15728         rtnl_lock();
15729
15730         /* We cleared the interrupt scheme when we suspended, so we need to
15731          * restore it now to resume device functionality.
15732          */
15733         err = i40e_restore_interrupt_scheme(pf);
15734         if (err) {
15735                 dev_err(dev, "Cannot restore interrupt scheme: %d\n",
15736                         err);
15737         }
15738
15739         clear_bit(__I40E_DOWN, pf->state);
15740         i40e_reset_and_rebuild(pf, false, true);
15741
15742         rtnl_unlock();
15743
15744         /* Clear suspended state last after everything is recovered */
15745         clear_bit(__I40E_SUSPENDED, pf->state);
15746
15747         /* Restart the service task */
15748         mod_timer(&pf->service_timer,
15749                   round_jiffies(jiffies + pf->service_timer_period));
15750
15751         return 0;
15752 }
15753
15754 static const struct pci_error_handlers i40e_err_handler = {
15755         .error_detected = i40e_pci_error_detected,
15756         .slot_reset = i40e_pci_error_slot_reset,
15757         .reset_prepare = i40e_pci_error_reset_prepare,
15758         .reset_done = i40e_pci_error_reset_done,
15759         .resume = i40e_pci_error_resume,
15760 };
15761
15762 static SIMPLE_DEV_PM_OPS(i40e_pm_ops, i40e_suspend, i40e_resume);
15763
15764 static struct pci_driver i40e_driver = {
15765         .name     = i40e_driver_name,
15766         .id_table = i40e_pci_tbl,
15767         .probe    = i40e_probe,
15768         .remove   = i40e_remove,
15769         .driver   = {
15770                 .pm = &i40e_pm_ops,
15771         },
15772         .shutdown = i40e_shutdown,
15773         .err_handler = &i40e_err_handler,
15774         .sriov_configure = i40e_pci_sriov_configure,
15775 };
15776
15777 /**
15778  * i40e_init_module - Driver registration routine
15779  *
15780  * i40e_init_module is the first routine called when the driver is
15781  * loaded. All it does is register with the PCI subsystem.
15782  **/
15783 static int __init i40e_init_module(void)
15784 {
15785         pr_info("%s: %s - version %s\n", i40e_driver_name,
15786                 i40e_driver_string, i40e_driver_version_str);
15787         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
15788
15789         /* There is no need to throttle the number of active tasks because
15790          * each device limits its own task using a state bit for scheduling
15791          * the service task, and the device tasks do not interfere with each
15792          * other, so we don't set a max task limit. We must set WQ_MEM_RECLAIM
15793          * since we need to be able to guarantee forward progress even under
15794          * memory pressure.
15795          */
15796         i40e_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, i40e_driver_name);
15797         if (!i40e_wq) {
15798                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
15799                 return -ENOMEM;
15800         }
15801
15802         i40e_dbg_init();
15803         return pci_register_driver(&i40e_driver);
15804 }
15805 module_init(i40e_init_module);
15806
15807 /**
15808  * i40e_exit_module - Driver exit cleanup routine
15809  *
15810  * i40e_exit_module is called just before the driver is removed
15811  * from memory.
15812  **/
15813 static void __exit i40e_exit_module(void)
15814 {
15815         pci_unregister_driver(&i40e_driver);
15816         destroy_workqueue(i40e_wq);
15817         i40e_dbg_exit();
15818 }
15819 module_exit(i40e_exit_module);