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

   1 /*******************************************************************************
   2  *
   3  * Intel Ethernet Controller XL710 Family Linux Driver
   4  * Copyright(c) 2013 - 2016 Intel Corporation.
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms and conditions of the GNU General Public License,
   8  * version 2, as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope it will be useful, but WITHOUT
  11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13  * more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
  17  *
  18  * The full GNU General Public License is included in this distribution in
  19  * the file called "COPYING".
  20  *
  21  * Contact Information:
  22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
  24  *
  25  ******************************************************************************/
  26
  27 #include <linux/etherdevice.h>
  28 #include <linux/of_net.h>
  29 #include <linux/pci.h>
  30
  31 /* Local includes */
  32 #include "i40e.h"
  33 #include "i40e_diag.h"
  34 #include <net/udp_tunnel.h>
  35 /* All i40e tracepoints are defined by the include below, which
  36  * must be included exactly once across the whole kernel with
  37  * CREATE_TRACE_POINTS defined
  38  */
  39 #define CREATE_TRACE_POINTS
  40 #include "i40e_trace.h"
  41
  42 const char i40e_driver_name[] = "i40e";
  43 static const char i40e_driver_string[] =
  44                         "Intel(R) Ethernet Connection XL710 Network Driver";
  45
  46 #define DRV_KERN "-k"
  47
  48 #define DRV_VERSION_MAJOR 2
  49 #define DRV_VERSION_MINOR 1
  50 #define DRV_VERSION_BUILD 14
  51 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  52              __stringify(DRV_VERSION_MINOR) "." \
  53              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  54 const char i40e_driver_version_str[] = DRV_VERSION;
  55 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  56
  57 /* a bit of forward declarations */
  58 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  59 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
  60 static int i40e_add_vsi(struct i40e_vsi *vsi);
  61 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  62 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  63 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  64 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  65 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  66 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
  67 static int i40e_reset(struct i40e_pf *pf);
  68 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
  69 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  70 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  71
  72 /* i40e_pci_tbl - PCI Device ID Table
  73  *
  74  * Last entry must be all 0s
  75  *
  76  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  77  *   Class, Class Mask, private data (not used) }
  78  */
  79 static const struct pci_device_id i40e_pci_tbl[] = {
  80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
  90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
  91         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  92         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  93         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  94         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
  95         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  96         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  97         {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
  98         {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
  99         /* required last entry */
 100         {0, }
 101 };
 102 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
 103
 104 #define I40E_MAX_VF_COUNT 128
 105 static int debug = -1;
 106 module_param(debug, uint, 0);
 107 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
 108
 109 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 110 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 111 MODULE_LICENSE("GPL");
 112 MODULE_VERSION(DRV_VERSION);
 113
 114 static struct workqueue_struct *i40e_wq;
 115
 116 /**
 117  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 118  * @hw:   pointer to the HW structure
 119  * @mem:  ptr to mem struct to fill out
 120  * @size: size of memory requested
 121  * @alignment: what to align the allocation to
 122  **/
 123 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 124                             u64 size, u32 alignment)
 125 {
 126         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 127
 128         mem->size = ALIGN(size, alignment);
 129         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 130                                       &mem->pa, GFP_KERNEL);
 131         if (!mem->va)
 132                 return -ENOMEM;
 133
 134         return 0;
 135 }
 136
 137 /**
 138  * i40e_free_dma_mem_d - OS specific memory free for shared code
 139  * @hw:   pointer to the HW structure
 140  * @mem:  ptr to mem struct to free
 141  **/
 142 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 143 {
 144         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 145
 146         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 147         mem->va = NULL;
 148         mem->pa = 0;
 149         mem->size = 0;
 150
 151         return 0;
 152 }
 153
 154 /**
 155  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 156  * @hw:   pointer to the HW structure
 157  * @mem:  ptr to mem struct to fill out
 158  * @size: size of memory requested
 159  **/
 160 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 161                              u32 size)
 162 {
 163         mem->size = size;
 164         mem->va = kzalloc(size, GFP_KERNEL);
 165
 166         if (!mem->va)
 167                 return -ENOMEM;
 168
 169         return 0;
 170 }
 171
 172 /**
 173  * i40e_free_virt_mem_d - OS specific memory free for shared code
 174  * @hw:   pointer to the HW structure
 175  * @mem:  ptr to mem struct to free
 176  **/
 177 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 178 {
 179         /* it's ok to kfree a NULL pointer */
 180         kfree(mem->va);
 181         mem->va = NULL;
 182         mem->size = 0;
 183
 184         return 0;
 185 }
 186
 187 /**
 188  * i40e_get_lump - find a lump of free generic resource
 189  * @pf: board private structure
 190  * @pile: the pile of resource to search
 191  * @needed: the number of items needed
 192  * @id: an owner id to stick on the items assigned
 193  *
 194  * Returns the base item index of the lump, or negative for error
 195  *
 196  * The search_hint trick and lack of advanced fit-finding only work
 197  * because we're highly likely to have all the same size lump requests.
 198  * Linear search time and any fragmentation should be minimal.
 199  **/
 200 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 201                          u16 needed, u16 id)
 202 {
 203         int ret = -ENOMEM;
 204         int i, j;
 205
 206         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 207                 dev_info(&pf->pdev->dev,
 208                          "param err: pile=%p needed=%d id=0x%04x\n",
 209                          pile, needed, id);
 210                 return -EINVAL;
 211         }
 212
 213         /* start the linear search with an imperfect hint */
 214         i = pile->search_hint;
 215         while (i < pile->num_entries) {
 216                 /* skip already allocated entries */
 217                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 218                         i++;
 219                         continue;
 220                 }
 221
 222                 /* do we have enough in this lump? */
 223                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 224                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 225                                 break;
 226                 }
 227
 228                 if (j == needed) {
 229                         /* there was enough, so assign it to the requestor */
 230                         for (j = 0; j < needed; j++)
 231                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 232                         ret = i;
 233                         pile->search_hint = i + j;
 234                         break;
 235                 }
 236
 237                 /* not enough, so skip over it and continue looking */
 238                 i += j;
 239         }
 240
 241         return ret;
 242 }
 243
 244 /**
 245  * i40e_put_lump - return a lump of generic resource
 246  * @pile: the pile of resource to search
 247  * @index: the base item index
 248  * @id: the owner id of the items assigned
 249  *
 250  * Returns the count of items in the lump
 251  **/
 252 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 253 {
 254         int valid_id = (id | I40E_PILE_VALID_BIT);
 255         int count = 0;
 256         int i;
 257
 258         if (!pile || index >= pile->num_entries)
 259                 return -EINVAL;
 260
 261         for (i = index;
 262              i < pile->num_entries && pile->list[i] == valid_id;
 263              i++) {
 264                 pile->list[i] = 0;
 265                 count++;
 266         }
 267
 268         if (count && index < pile->search_hint)
 269                 pile->search_hint = index;
 270
 271         return count;
 272 }
 273
 274 /**
 275  * i40e_find_vsi_from_id - searches for the vsi with the given id
 276  * @pf - the pf structure to search for the vsi
 277  * @id - id of the vsi it is searching for
 278  **/
 279 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 280 {
 281         int i;
 282
 283         for (i = 0; i < pf->num_alloc_vsi; i++)
 284                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 285                         return pf->vsi[i];
 286
 287         return NULL;
 288 }
 289
 290 /**
 291  * i40e_service_event_schedule - Schedule the service task to wake up
 292  * @pf: board private structure
 293  *
 294  * If not already scheduled, this puts the task into the work queue
 295  **/
 296 void i40e_service_event_schedule(struct i40e_pf *pf)
 297 {
 298         if (!test_bit(__I40E_VSI_DOWN, pf->state) &&
 299             !test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
 300                 queue_work(i40e_wq, &pf->service_task);
 301 }
 302
 303 /**
 304  * i40e_tx_timeout - Respond to a Tx Hang
 305  * @netdev: network interface device structure
 306  *
 307  * If any port has noticed a Tx timeout, it is likely that the whole
 308  * device is munged, not just the one netdev port, so go for the full
 309  * reset.
 310  **/
 311 static void i40e_tx_timeout(struct net_device *netdev)
 312 {
 313         struct i40e_netdev_priv *np = netdev_priv(netdev);
 314         struct i40e_vsi *vsi = np->vsi;
 315         struct i40e_pf *pf = vsi->back;
 316         struct i40e_ring *tx_ring = NULL;
 317         unsigned int i, hung_queue = 0;
 318         u32 head, val;
 319
 320         pf->tx_timeout_count++;
 321
 322         /* find the stopped queue the same way the stack does */
 323         for (i = 0; i < netdev->num_tx_queues; i++) {
 324                 struct netdev_queue *q;
 325                 unsigned long trans_start;
 326
 327                 q = netdev_get_tx_queue(netdev, i);
 328                 trans_start = q->trans_start;
 329                 if (netif_xmit_stopped(q) &&
 330                     time_after(jiffies,
 331                                (trans_start + netdev->watchdog_timeo))) {
 332                         hung_queue = i;
 333                         break;
 334                 }
 335         }
 336
 337         if (i == netdev->num_tx_queues) {
 338                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 339         } else {
 340                 /* now that we have an index, find the tx_ring struct */
 341                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 342                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 343                                 if (hung_queue ==
 344                                     vsi->tx_rings[i]->queue_index) {
 345                                         tx_ring = vsi->tx_rings[i];
 346                                         break;
 347                                 }
 348                         }
 349                 }
 350         }
 351
 352         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 353                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 354         else if (time_before(jiffies,
 355                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 356                 return;   /* don't do any new action before the next timeout */
 357
 358         if (tx_ring) {
 359                 head = i40e_get_head(tx_ring);
 360                 /* Read interrupt register */
 361                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 362                         val = rd32(&pf->hw,
 363                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 364                                                 tx_ring->vsi->base_vector - 1));
 365                 else
 366                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 367
 368                 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",
 369                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 370                             head, tx_ring->next_to_use,
 371                             readl(tx_ring->tail), val);
 372         }
 373
 374         pf->tx_timeout_last_recovery = jiffies;
 375         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 376                     pf->tx_timeout_recovery_level, hung_queue);
 377
 378         switch (pf->tx_timeout_recovery_level) {
 379         case 1:
 380                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
 381                 break;
 382         case 2:
 383                 set_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
 384                 break;
 385         case 3:
 386                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
 387                 break;
 388         default:
 389                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 390                 break;
 391         }
 392
 393         i40e_service_event_schedule(pf);
 394         pf->tx_timeout_recovery_level++;
 395 }
 396
 397 /**
 398  * i40e_get_vsi_stats_struct - Get System Network Statistics
 399  * @vsi: the VSI we care about
 400  *
 401  * Returns the address of the device statistics structure.
 402  * The statistics are actually updated from the service task.
 403  **/
 404 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 405 {
 406         return &vsi->net_stats;
 407 }
 408
 409 /**
 410  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 411  * @netdev: network interface device structure
 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_ring *tx_ring, *rx_ring;
 421         struct i40e_vsi *vsi = np->vsi;
 422         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 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                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 437                 if (!tx_ring)
 438                         continue;
 439
 440                 do {
 441                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 442                         packets = tx_ring->stats.packets;
 443                         bytes   = tx_ring->stats.bytes;
 444                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 445
 446                 stats->tx_packets += packets;
 447                 stats->tx_bytes   += bytes;
 448                 rx_ring = &tx_ring[1];
 449
 450                 do {
 451                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 452                         packets = rx_ring->stats.packets;
 453                         bytes   = rx_ring->stats.bytes;
 454                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 455
 456                 stats->rx_packets += packets;
 457                 stats->rx_bytes   += bytes;
 458         }
 459         rcu_read_unlock();
 460
 461         /* following stats updated by i40e_watchdog_subtask() */
 462         stats->multicast        = vsi_stats->multicast;
 463         stats->tx_errors        = vsi_stats->tx_errors;
 464         stats->tx_dropped       = vsi_stats->tx_dropped;
 465         stats->rx_errors        = vsi_stats->rx_errors;
 466         stats->rx_dropped       = vsi_stats->rx_dropped;
 467         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 468         stats->rx_length_errors = vsi_stats->rx_length_errors;
 469 }
 470
 471 /**
 472  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 473  * @vsi: the VSI to have its stats reset
 474  **/
 475 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 476 {
 477         struct rtnl_link_stats64 *ns;
 478         int i;
 479
 480         if (!vsi)
 481                 return;
 482
 483         ns = i40e_get_vsi_stats_struct(vsi);
 484         memset(ns, 0, sizeof(*ns));
 485         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 486         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 487         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 488         if (vsi->rx_rings && vsi->rx_rings[0]) {
 489                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 490                         memset(&vsi->rx_rings[i]->stats, 0,
 491                                sizeof(vsi->rx_rings[i]->stats));
 492                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 493                                sizeof(vsi->rx_rings[i]->rx_stats));
 494                         memset(&vsi->tx_rings[i]->stats, 0,
 495                                sizeof(vsi->tx_rings[i]->stats));
 496                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 497                                sizeof(vsi->tx_rings[i]->tx_stats));
 498                 }
 499         }
 500         vsi->stat_offsets_loaded = false;
 501 }
 502
 503 /**
 504  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 505  * @pf: the PF to be reset
 506  **/
 507 void i40e_pf_reset_stats(struct i40e_pf *pf)
 508 {
 509         int i;
 510
 511         memset(&pf->stats, 0, sizeof(pf->stats));
 512         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 513         pf->stat_offsets_loaded = false;
 514
 515         for (i = 0; i < I40E_MAX_VEB; i++) {
 516                 if (pf->veb[i]) {
 517                         memset(&pf->veb[i]->stats, 0,
 518                                sizeof(pf->veb[i]->stats));
 519                         memset(&pf->veb[i]->stats_offsets, 0,
 520                                sizeof(pf->veb[i]->stats_offsets));
 521                         pf->veb[i]->stat_offsets_loaded = false;
 522                 }
 523         }
 524         pf->hw_csum_rx_error = 0;
 525 }
 526
 527 /**
 528  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 529  * @hw: ptr to the hardware info
 530  * @hireg: the high 32 bit reg to read
 531  * @loreg: the low 32 bit reg to read
 532  * @offset_loaded: has the initial offset been loaded yet
 533  * @offset: ptr to current offset value
 534  * @stat: ptr to the stat
 535  *
 536  * Since the device stats are not reset at PFReset, they likely will not
 537  * be zeroed when the driver starts.  We'll save the first values read
 538  * and use them as offsets to be subtracted from the raw values in order
 539  * to report stats that count from zero.  In the process, we also manage
 540  * the potential roll-over.
 541  **/
 542 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 543                                bool offset_loaded, u64 *offset, u64 *stat)
 544 {
 545         u64 new_data;
 546
 547         if (hw->device_id == I40E_DEV_ID_QEMU) {
 548                 new_data = rd32(hw, loreg);
 549                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 550         } else {
 551                 new_data = rd64(hw, loreg);
 552         }
 553         if (!offset_loaded)
 554                 *offset = new_data;
 555         if (likely(new_data >= *offset))
 556                 *stat = new_data - *offset;
 557         else
 558                 *stat = (new_data + BIT_ULL(48)) - *offset;
 559         *stat &= 0xFFFFFFFFFFFFULL;
 560 }
 561
 562 /**
 563  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 564  * @hw: ptr to the hardware info
 565  * @reg: the hw reg to read
 566  * @offset_loaded: has the initial offset been loaded yet
 567  * @offset: ptr to current offset value
 568  * @stat: ptr to the stat
 569  **/
 570 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 571                                bool offset_loaded, u64 *offset, u64 *stat)
 572 {
 573         u32 new_data;
 574
 575         new_data = rd32(hw, reg);
 576         if (!offset_loaded)
 577                 *offset = new_data;
 578         if (likely(new_data >= *offset))
 579                 *stat = (u32)(new_data - *offset);
 580         else
 581                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 582 }
 583
 584 /**
 585  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 586  * @vsi: the VSI to be updated
 587  **/
 588 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 589 {
 590         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 591         struct i40e_pf *pf = vsi->back;
 592         struct i40e_hw *hw = &pf->hw;
 593         struct i40e_eth_stats *oes;
 594         struct i40e_eth_stats *es;     /* device's eth stats */
 595
 596         es = &vsi->eth_stats;
 597         oes = &vsi->eth_stats_offsets;
 598
 599         /* Gather up the stats that the hw collects */
 600         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 601                            vsi->stat_offsets_loaded,
 602                            &oes->tx_errors, &es->tx_errors);
 603         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 604                            vsi->stat_offsets_loaded,
 605                            &oes->rx_discards, &es->rx_discards);
 606         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 607                            vsi->stat_offsets_loaded,
 608                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 609         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 610                            vsi->stat_offsets_loaded,
 611                            &oes->tx_errors, &es->tx_errors);
 612
 613         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 614                            I40E_GLV_GORCL(stat_idx),
 615                            vsi->stat_offsets_loaded,
 616                            &oes->rx_bytes, &es->rx_bytes);
 617         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 618                            I40E_GLV_UPRCL(stat_idx),
 619                            vsi->stat_offsets_loaded,
 620                            &oes->rx_unicast, &es->rx_unicast);
 621         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 622                            I40E_GLV_MPRCL(stat_idx),
 623                            vsi->stat_offsets_loaded,
 624                            &oes->rx_multicast, &es->rx_multicast);
 625         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 626                            I40E_GLV_BPRCL(stat_idx),
 627                            vsi->stat_offsets_loaded,
 628                            &oes->rx_broadcast, &es->rx_broadcast);
 629
 630         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 631                            I40E_GLV_GOTCL(stat_idx),
 632                            vsi->stat_offsets_loaded,
 633                            &oes->tx_bytes, &es->tx_bytes);
 634         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 635                            I40E_GLV_UPTCL(stat_idx),
 636                            vsi->stat_offsets_loaded,
 637                            &oes->tx_unicast, &es->tx_unicast);
 638         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 639                            I40E_GLV_MPTCL(stat_idx),
 640                            vsi->stat_offsets_loaded,
 641                            &oes->tx_multicast, &es->tx_multicast);
 642         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 643                            I40E_GLV_BPTCL(stat_idx),
 644                            vsi->stat_offsets_loaded,
 645                            &oes->tx_broadcast, &es->tx_broadcast);
 646         vsi->stat_offsets_loaded = true;
 647 }
 648
 649 /**
 650  * i40e_update_veb_stats - Update Switch component statistics
 651  * @veb: the VEB being updated
 652  **/
 653 static void i40e_update_veb_stats(struct i40e_veb *veb)
 654 {
 655         struct i40e_pf *pf = veb->pf;
 656         struct i40e_hw *hw = &pf->hw;
 657         struct i40e_eth_stats *oes;
 658         struct i40e_eth_stats *es;     /* device's eth stats */
 659         struct i40e_veb_tc_stats *veb_oes;
 660         struct i40e_veb_tc_stats *veb_es;
 661         int i, idx = 0;
 662
 663         idx = veb->stats_idx;
 664         es = &veb->stats;
 665         oes = &veb->stats_offsets;
 666         veb_es = &veb->tc_stats;
 667         veb_oes = &veb->tc_stats_offsets;
 668
 669         /* Gather up the stats that the hw collects */
 670         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 671                            veb->stat_offsets_loaded,
 672                            &oes->tx_discards, &es->tx_discards);
 673         if (hw->revision_id > 0)
 674                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 675                                    veb->stat_offsets_loaded,
 676                                    &oes->rx_unknown_protocol,
 677                                    &es->rx_unknown_protocol);
 678         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 679                            veb->stat_offsets_loaded,
 680                            &oes->rx_bytes, &es->rx_bytes);
 681         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 682                            veb->stat_offsets_loaded,
 683                            &oes->rx_unicast, &es->rx_unicast);
 684         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 685                            veb->stat_offsets_loaded,
 686                            &oes->rx_multicast, &es->rx_multicast);
 687         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 688                            veb->stat_offsets_loaded,
 689                            &oes->rx_broadcast, &es->rx_broadcast);
 690
 691         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 692                            veb->stat_offsets_loaded,
 693                            &oes->tx_bytes, &es->tx_bytes);
 694         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 695                            veb->stat_offsets_loaded,
 696                            &oes->tx_unicast, &es->tx_unicast);
 697         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 698                            veb->stat_offsets_loaded,
 699                            &oes->tx_multicast, &es->tx_multicast);
 700         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 701                            veb->stat_offsets_loaded,
 702                            &oes->tx_broadcast, &es->tx_broadcast);
 703         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 704                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 705                                    I40E_GLVEBTC_RPCL(i, idx),
 706                                    veb->stat_offsets_loaded,
 707                                    &veb_oes->tc_rx_packets[i],
 708                                    &veb_es->tc_rx_packets[i]);
 709                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 710                                    I40E_GLVEBTC_RBCL(i, idx),
 711                                    veb->stat_offsets_loaded,
 712                                    &veb_oes->tc_rx_bytes[i],
 713                                    &veb_es->tc_rx_bytes[i]);
 714                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 715                                    I40E_GLVEBTC_TPCL(i, idx),
 716                                    veb->stat_offsets_loaded,
 717                                    &veb_oes->tc_tx_packets[i],
 718                                    &veb_es->tc_tx_packets[i]);
 719                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 720                                    I40E_GLVEBTC_TBCL(i, idx),
 721                                    veb->stat_offsets_loaded,
 722                                    &veb_oes->tc_tx_bytes[i],
 723                                    &veb_es->tc_tx_bytes[i]);
 724         }
 725         veb->stat_offsets_loaded = true;
 726 }
 727
 728 /**
 729  * i40e_update_vsi_stats - Update the vsi statistics counters.
 730  * @vsi: the VSI to be updated
 731  *
 732  * There are a few instances where we store the same stat in a
 733  * couple of different structs.  This is partly because we have
 734  * the netdev stats that need to be filled out, which is slightly
 735  * different from the "eth_stats" defined by the chip and used in
 736  * VF communications.  We sort it out here.
 737  **/
 738 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 739 {
 740         struct i40e_pf *pf = vsi->back;
 741         struct rtnl_link_stats64 *ons;
 742         struct rtnl_link_stats64 *ns;   /* netdev stats */
 743         struct i40e_eth_stats *oes;
 744         struct i40e_eth_stats *es;     /* device's eth stats */
 745         u32 tx_restart, tx_busy;
 746         struct i40e_ring *p;
 747         u32 rx_page, rx_buf;
 748         u64 bytes, packets;
 749         unsigned int start;
 750         u64 tx_linearize;
 751         u64 tx_force_wb;
 752         u64 rx_p, rx_b;
 753         u64 tx_p, tx_b;
 754         u16 q;
 755
 756         if (test_bit(__I40E_VSI_DOWN, vsi->state) ||
 757             test_bit(__I40E_CONFIG_BUSY, pf->state))
 758                 return;
 759
 760         ns = i40e_get_vsi_stats_struct(vsi);
 761         ons = &vsi->net_stats_offsets;
 762         es = &vsi->eth_stats;
 763         oes = &vsi->eth_stats_offsets;
 764
 765         /* Gather up the netdev and vsi stats that the driver collects
 766          * on the fly during packet processing
 767          */
 768         rx_b = rx_p = 0;
 769         tx_b = tx_p = 0;
 770         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 771         rx_page = 0;
 772         rx_buf = 0;
 773         rcu_read_lock();
 774         for (q = 0; q < vsi->num_queue_pairs; q++) {
 775                 /* locate Tx ring */
 776                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 777
 778                 do {
 779                         start = u64_stats_fetch_begin_irq(&p->syncp);
 780                         packets = p->stats.packets;
 781                         bytes = p->stats.bytes;
 782                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 783                 tx_b += bytes;
 784                 tx_p += packets;
 785                 tx_restart += p->tx_stats.restart_queue;
 786                 tx_busy += p->tx_stats.tx_busy;
 787                 tx_linearize += p->tx_stats.tx_linearize;
 788                 tx_force_wb += p->tx_stats.tx_force_wb;
 789
 790                 /* Rx queue is part of the same block as Tx queue */
 791                 p = &p[1];
 792                 do {
 793                         start = u64_stats_fetch_begin_irq(&p->syncp);
 794                         packets = p->stats.packets;
 795                         bytes = p->stats.bytes;
 796                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 797                 rx_b += bytes;
 798                 rx_p += packets;
 799                 rx_buf += p->rx_stats.alloc_buff_failed;
 800                 rx_page += p->rx_stats.alloc_page_failed;
 801         }
 802         rcu_read_unlock();
 803         vsi->tx_restart = tx_restart;
 804         vsi->tx_busy = tx_busy;
 805         vsi->tx_linearize = tx_linearize;
 806         vsi->tx_force_wb = tx_force_wb;
 807         vsi->rx_page_failed = rx_page;
 808         vsi->rx_buf_failed = rx_buf;
 809
 810         ns->rx_packets = rx_p;
 811         ns->rx_bytes = rx_b;
 812         ns->tx_packets = tx_p;
 813         ns->tx_bytes = tx_b;
 814
 815         /* update netdev stats from eth stats */
 816         i40e_update_eth_stats(vsi);
 817         ons->tx_errors = oes->tx_errors;
 818         ns->tx_errors = es->tx_errors;
 819         ons->multicast = oes->rx_multicast;
 820         ns->multicast = es->rx_multicast;
 821         ons->rx_dropped = oes->rx_discards;
 822         ns->rx_dropped = es->rx_discards;
 823         ons->tx_dropped = oes->tx_discards;
 824         ns->tx_dropped = es->tx_discards;
 825
 826         /* pull in a couple PF stats if this is the main vsi */
 827         if (vsi == pf->vsi[pf->lan_vsi]) {
 828                 ns->rx_crc_errors = pf->stats.crc_errors;
 829                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 830                 ns->rx_length_errors = pf->stats.rx_length_errors;
 831         }
 832 }
 833
 834 /**
 835  * i40e_update_pf_stats - Update the PF statistics counters.
 836  * @pf: the PF to be updated
 837  **/
 838 static void i40e_update_pf_stats(struct i40e_pf *pf)
 839 {
 840         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 841         struct i40e_hw_port_stats *nsd = &pf->stats;
 842         struct i40e_hw *hw = &pf->hw;
 843         u32 val;
 844         int i;
 845
 846         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 847                            I40E_GLPRT_GORCL(hw->port),
 848                            pf->stat_offsets_loaded,
 849                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 850         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 851                            I40E_GLPRT_GOTCL(hw->port),
 852                            pf->stat_offsets_loaded,
 853                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 854         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 855                            pf->stat_offsets_loaded,
 856                            &osd->eth.rx_discards,
 857                            &nsd->eth.rx_discards);
 858         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 859                            I40E_GLPRT_UPRCL(hw->port),
 860                            pf->stat_offsets_loaded,
 861                            &osd->eth.rx_unicast,
 862                            &nsd->eth.rx_unicast);
 863         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 864                            I40E_GLPRT_MPRCL(hw->port),
 865                            pf->stat_offsets_loaded,
 866                            &osd->eth.rx_multicast,
 867                            &nsd->eth.rx_multicast);
 868         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 869                            I40E_GLPRT_BPRCL(hw->port),
 870                            pf->stat_offsets_loaded,
 871                            &osd->eth.rx_broadcast,
 872                            &nsd->eth.rx_broadcast);
 873         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 874                            I40E_GLPRT_UPTCL(hw->port),
 875                            pf->stat_offsets_loaded,
 876                            &osd->eth.tx_unicast,
 877                            &nsd->eth.tx_unicast);
 878         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 879                            I40E_GLPRT_MPTCL(hw->port),
 880                            pf->stat_offsets_loaded,
 881                            &osd->eth.tx_multicast,
 882                            &nsd->eth.tx_multicast);
 883         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 884                            I40E_GLPRT_BPTCL(hw->port),
 885                            pf->stat_offsets_loaded,
 886                            &osd->eth.tx_broadcast,
 887                            &nsd->eth.tx_broadcast);
 888
 889         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 890                            pf->stat_offsets_loaded,
 891                            &osd->tx_dropped_link_down,
 892                            &nsd->tx_dropped_link_down);
 893
 894         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 895                            pf->stat_offsets_loaded,
 896                            &osd->crc_errors, &nsd->crc_errors);
 897
 898         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 899                            pf->stat_offsets_loaded,
 900                            &osd->illegal_bytes, &nsd->illegal_bytes);
 901
 902         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 903                            pf->stat_offsets_loaded,
 904                            &osd->mac_local_faults,
 905                            &nsd->mac_local_faults);
 906         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 907                            pf->stat_offsets_loaded,
 908                            &osd->mac_remote_faults,
 909                            &nsd->mac_remote_faults);
 910
 911         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 912                            pf->stat_offsets_loaded,
 913                            &osd->rx_length_errors,
 914                            &nsd->rx_length_errors);
 915
 916         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 917                            pf->stat_offsets_loaded,
 918                            &osd->link_xon_rx, &nsd->link_xon_rx);
 919         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 920                            pf->stat_offsets_loaded,
 921                            &osd->link_xon_tx, &nsd->link_xon_tx);
 922         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 923                            pf->stat_offsets_loaded,
 924                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 925         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 926                            pf->stat_offsets_loaded,
 927                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 928
 929         for (i = 0; i < 8; i++) {
 930                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 931                                    pf->stat_offsets_loaded,
 932                                    &osd->priority_xoff_rx[i],
 933                                    &nsd->priority_xoff_rx[i]);
 934                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 935                                    pf->stat_offsets_loaded,
 936                                    &osd->priority_xon_rx[i],
 937                                    &nsd->priority_xon_rx[i]);
 938                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
 939                                    pf->stat_offsets_loaded,
 940                                    &osd->priority_xon_tx[i],
 941                                    &nsd->priority_xon_tx[i]);
 942                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
 943                                    pf->stat_offsets_loaded,
 944                                    &osd->priority_xoff_tx[i],
 945                                    &nsd->priority_xoff_tx[i]);
 946                 i40e_stat_update32(hw,
 947                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
 948                                    pf->stat_offsets_loaded,
 949                                    &osd->priority_xon_2_xoff[i],
 950                                    &nsd->priority_xon_2_xoff[i]);
 951         }
 952
 953         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
 954                            I40E_GLPRT_PRC64L(hw->port),
 955                            pf->stat_offsets_loaded,
 956                            &osd->rx_size_64, &nsd->rx_size_64);
 957         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
 958                            I40E_GLPRT_PRC127L(hw->port),
 959                            pf->stat_offsets_loaded,
 960                            &osd->rx_size_127, &nsd->rx_size_127);
 961         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
 962                            I40E_GLPRT_PRC255L(hw->port),
 963                            pf->stat_offsets_loaded,
 964                            &osd->rx_size_255, &nsd->rx_size_255);
 965         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
 966                            I40E_GLPRT_PRC511L(hw->port),
 967                            pf->stat_offsets_loaded,
 968                            &osd->rx_size_511, &nsd->rx_size_511);
 969         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
 970                            I40E_GLPRT_PRC1023L(hw->port),
 971                            pf->stat_offsets_loaded,
 972                            &osd->rx_size_1023, &nsd->rx_size_1023);
 973         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
 974                            I40E_GLPRT_PRC1522L(hw->port),
 975                            pf->stat_offsets_loaded,
 976                            &osd->rx_size_1522, &nsd->rx_size_1522);
 977         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
 978                            I40E_GLPRT_PRC9522L(hw->port),
 979                            pf->stat_offsets_loaded,
 980                            &osd->rx_size_big, &nsd->rx_size_big);
 981
 982         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
 983                            I40E_GLPRT_PTC64L(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->tx_size_64, &nsd->tx_size_64);
 986         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
 987                            I40E_GLPRT_PTC127L(hw->port),
 988                            pf->stat_offsets_loaded,
 989                            &osd->tx_size_127, &nsd->tx_size_127);
 990         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
 991                            I40E_GLPRT_PTC255L(hw->port),
 992                            pf->stat_offsets_loaded,
 993                            &osd->tx_size_255, &nsd->tx_size_255);
 994         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
 995                            I40E_GLPRT_PTC511L(hw->port),
 996                            pf->stat_offsets_loaded,
 997                            &osd->tx_size_511, &nsd->tx_size_511);
 998         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
 999                            I40E_GLPRT_PTC1023L(hw->port),
1000                            pf->stat_offsets_loaded,
1001                            &osd->tx_size_1023, &nsd->tx_size_1023);
1002         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1003                            I40E_GLPRT_PTC1522L(hw->port),
1004                            pf->stat_offsets_loaded,
1005                            &osd->tx_size_1522, &nsd->tx_size_1522);
1006         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1007                            I40E_GLPRT_PTC9522L(hw->port),
1008                            pf->stat_offsets_loaded,
1009                            &osd->tx_size_big, &nsd->tx_size_big);
1010
1011         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1012                            pf->stat_offsets_loaded,
1013                            &osd->rx_undersize, &nsd->rx_undersize);
1014         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1015                            pf->stat_offsets_loaded,
1016                            &osd->rx_fragments, &nsd->rx_fragments);
1017         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1018                            pf->stat_offsets_loaded,
1019                            &osd->rx_oversize, &nsd->rx_oversize);
1020         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1021                            pf->stat_offsets_loaded,
1022                            &osd->rx_jabber, &nsd->rx_jabber);
1023
1024         /* FDIR stats */
1025         i40e_stat_update32(hw,
1026                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1027                            pf->stat_offsets_loaded,
1028                            &osd->fd_atr_match, &nsd->fd_atr_match);
1029         i40e_stat_update32(hw,
1030                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1031                            pf->stat_offsets_loaded,
1032                            &osd->fd_sb_match, &nsd->fd_sb_match);
1033         i40e_stat_update32(hw,
1034                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1035                       pf->stat_offsets_loaded,
1036                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1037
1038         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1039         nsd->tx_lpi_status =
1040                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1041                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1042         nsd->rx_lpi_status =
1043                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1044                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1045         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1046                            pf->stat_offsets_loaded,
1047                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1048         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1049                            pf->stat_offsets_loaded,
1050                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1051
1052         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1053             !(pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED))
1054                 nsd->fd_sb_status = true;
1055         else
1056                 nsd->fd_sb_status = false;
1057
1058         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1059             !(pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED))
1060                 nsd->fd_atr_status = true;
1061         else
1062                 nsd->fd_atr_status = false;
1063
1064         pf->stat_offsets_loaded = true;
1065 }
1066
1067 /**
1068  * i40e_update_stats - Update the various statistics counters.
1069  * @vsi: the VSI to be updated
1070  *
1071  * Update the various stats for this VSI and its related entities.
1072  **/
1073 void i40e_update_stats(struct i40e_vsi *vsi)
1074 {
1075         struct i40e_pf *pf = vsi->back;
1076
1077         if (vsi == pf->vsi[pf->lan_vsi])
1078                 i40e_update_pf_stats(pf);
1079
1080         i40e_update_vsi_stats(vsi);
1081 }
1082
1083 /**
1084  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1085  * @vsi: the VSI to be searched
1086  * @macaddr: the MAC address
1087  * @vlan: the vlan
1088  *
1089  * Returns ptr to the filter object or NULL
1090  **/
1091 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1092                                                 const u8 *macaddr, s16 vlan)
1093 {
1094         struct i40e_mac_filter *f;
1095         u64 key;
1096
1097         if (!vsi || !macaddr)
1098                 return NULL;
1099
1100         key = i40e_addr_to_hkey(macaddr);
1101         hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1102                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1103                     (vlan == f->vlan))
1104                         return f;
1105         }
1106         return NULL;
1107 }
1108
1109 /**
1110  * i40e_find_mac - Find a mac addr in the macvlan filters list
1111  * @vsi: the VSI to be searched
1112  * @macaddr: the MAC address we are searching for
1113  *
1114  * Returns the first filter with the provided MAC address or NULL if
1115  * MAC address was not found
1116  **/
1117 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1118 {
1119         struct i40e_mac_filter *f;
1120         u64 key;
1121
1122         if (!vsi || !macaddr)
1123                 return NULL;
1124
1125         key = i40e_addr_to_hkey(macaddr);
1126         hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1127                 if ((ether_addr_equal(macaddr, f->macaddr)))
1128                         return f;
1129         }
1130         return NULL;
1131 }
1132
1133 /**
1134  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1135  * @vsi: the VSI to be searched
1136  *
1137  * Returns true if VSI is in vlan mode or false otherwise
1138  **/
1139 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1140 {
1141         /* If we have a PVID, always operate in VLAN mode */
1142         if (vsi->info.pvid)
1143                 return true;
1144
1145         /* We need to operate in VLAN mode whenever we have any filters with
1146          * a VLAN other than I40E_VLAN_ALL. We could check the table each
1147          * time, incurring search cost repeatedly. However, we can notice two
1148          * things:
1149          *
1150          * 1) the only place where we can gain a VLAN filter is in
1151          *    i40e_add_filter.
1152          *
1153          * 2) the only place where filters are actually removed is in
1154          *    i40e_sync_filters_subtask.
1155          *
1156          * Thus, we can simply use a boolean value, has_vlan_filters which we
1157          * will set to true when we add a VLAN filter in i40e_add_filter. Then
1158          * we have to perform the full search after deleting filters in
1159          * i40e_sync_filters_subtask, but we already have to search
1160          * filters here and can perform the check at the same time. This
1161          * results in avoiding embedding a loop for VLAN mode inside another
1162          * loop over all the filters, and should maintain correctness as noted
1163          * above.
1164          */
1165         return vsi->has_vlan_filter;
1166 }
1167
1168 /**
1169  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1170  * @vsi: the VSI to configure
1171  * @tmp_add_list: list of filters ready to be added
1172  * @tmp_del_list: list of filters ready to be deleted
1173  * @vlan_filters: the number of active VLAN filters
1174  *
1175  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1176  * behave as expected. If we have any active VLAN filters remaining or about
1177  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1178  * so that they only match against untagged traffic. If we no longer have any
1179  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1180  * so that they match against both tagged and untagged traffic. In this way,
1181  * we ensure that we correctly receive the desired traffic. This ensures that
1182  * when we have an active VLAN we will receive only untagged traffic and
1183  * traffic matching active VLANs. If we have no active VLANs then we will
1184  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1185  *
1186  * Finally, in a similar fashion, this function also corrects filters when
1187  * there is an active PVID assigned to this VSI.
1188  *
1189  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1190  *
1191  * This function is only expected to be called from within
1192  * i40e_sync_vsi_filters.
1193  *
1194  * NOTE: This function expects to be called while under the
1195  * mac_filter_hash_lock
1196  */
1197 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1198                                          struct hlist_head *tmp_add_list,
1199                                          struct hlist_head *tmp_del_list,
1200                                          int vlan_filters)
1201 {
1202         s16 pvid = le16_to_cpu(vsi->info.pvid);
1203         struct i40e_mac_filter *f, *add_head;
1204         struct i40e_new_mac_filter *new;
1205         struct hlist_node *h;
1206         int bkt, new_vlan;
1207
1208         /* To determine if a particular filter needs to be replaced we
1209          * have the three following conditions:
1210          *
1211          * a) if we have a PVID assigned, then all filters which are
1212          *    not marked as VLAN=PVID must be replaced with filters that
1213          *    are.
1214          * b) otherwise, if we have any active VLANS, all filters
1215          *    which are marked as VLAN=-1 must be replaced with
1216          *    filters marked as VLAN=0
1217          * c) finally, if we do not have any active VLANS, all filters
1218          *    which are marked as VLAN=0 must be replaced with filters
1219          *    marked as VLAN=-1
1220          */
1221
1222         /* Update the filters about to be added in place */
1223         hlist_for_each_entry(new, tmp_add_list, hlist) {
1224                 if (pvid && new->f->vlan != pvid)
1225                         new->f->vlan = pvid;
1226                 else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1227                         new->f->vlan = 0;
1228                 else if (!vlan_filters && new->f->vlan == 0)
1229                         new->f->vlan = I40E_VLAN_ANY;
1230         }
1231
1232         /* Update the remaining active filters */
1233         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1234                 /* Combine the checks for whether a filter needs to be changed
1235                  * and then determine the new VLAN inside the if block, in
1236                  * order to avoid duplicating code for adding the new filter
1237                  * then deleting the old filter.
1238                  */
1239                 if ((pvid && f->vlan != pvid) ||
1240                     (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1241                     (!vlan_filters && f->vlan == 0)) {
1242                         /* Determine the new vlan we will be adding */
1243                         if (pvid)
1244                                 new_vlan = pvid;
1245                         else if (vlan_filters)
1246                                 new_vlan = 0;
1247                         else
1248                                 new_vlan = I40E_VLAN_ANY;
1249
1250                         /* Create the new filter */
1251                         add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1252                         if (!add_head)
1253                                 return -ENOMEM;
1254
1255                         /* Create a temporary i40e_new_mac_filter */
1256                         new = kzalloc(sizeof(*new), GFP_ATOMIC);
1257                         if (!new)
1258                                 return -ENOMEM;
1259
1260                         new->f = add_head;
1261                         new->state = add_head->state;
1262
1263                         /* Add the new filter to the tmp list */
1264                         hlist_add_head(&new->hlist, tmp_add_list);
1265
1266                         /* Put the original filter into the delete list */
1267                         f->state = I40E_FILTER_REMOVE;
1268                         hash_del(&f->hlist);
1269                         hlist_add_head(&f->hlist, tmp_del_list);
1270                 }
1271         }
1272
1273         vsi->has_vlan_filter = !!vlan_filters;
1274
1275         return 0;
1276 }
1277
1278 /**
1279  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1280  * @vsi: the PF Main VSI - inappropriate for any other VSI
1281  * @macaddr: the MAC address
1282  *
1283  * Remove whatever filter the firmware set up so the driver can manage
1284  * its own filtering intelligently.
1285  **/
1286 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1287 {
1288         struct i40e_aqc_remove_macvlan_element_data element;
1289         struct i40e_pf *pf = vsi->back;
1290
1291         /* Only appropriate for the PF main VSI */
1292         if (vsi->type != I40E_VSI_MAIN)
1293                 return;
1294
1295         memset(&element, 0, sizeof(element));
1296         ether_addr_copy(element.mac_addr, macaddr);
1297         element.vlan_tag = 0;
1298         /* Ignore error returns, some firmware does it this way... */
1299         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1300         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1301
1302         memset(&element, 0, sizeof(element));
1303         ether_addr_copy(element.mac_addr, macaddr);
1304         element.vlan_tag = 0;
1305         /* ...and some firmware does it this way. */
1306         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1307                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1308         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1309 }
1310
1311 /**
1312  * i40e_add_filter - Add a mac/vlan filter to the VSI
1313  * @vsi: the VSI to be searched
1314  * @macaddr: the MAC address
1315  * @vlan: the vlan
1316  *
1317  * Returns ptr to the filter object or NULL when no memory available.
1318  *
1319  * NOTE: This function is expected to be called with mac_filter_hash_lock
1320  * being held.
1321  **/
1322 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1323                                         const u8 *macaddr, s16 vlan)
1324 {
1325         struct i40e_mac_filter *f;
1326         u64 key;
1327
1328         if (!vsi || !macaddr)
1329                 return NULL;
1330
1331         f = i40e_find_filter(vsi, macaddr, vlan);
1332         if (!f) {
1333                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1334                 if (!f)
1335                         return NULL;
1336
1337                 /* Update the boolean indicating if we need to function in
1338                  * VLAN mode.
1339                  */
1340                 if (vlan >= 0)
1341                         vsi->has_vlan_filter = true;
1342
1343                 ether_addr_copy(f->macaddr, macaddr);
1344                 f->vlan = vlan;
1345                 /* If we're in overflow promisc mode, set the state directly
1346                  * to failed, so we don't bother to try sending the filter
1347                  * to the hardware.
1348                  */
1349                 if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))
1350                         f->state = I40E_FILTER_FAILED;
1351                 else
1352                         f->state = I40E_FILTER_NEW;
1353                 INIT_HLIST_NODE(&f->hlist);
1354
1355                 key = i40e_addr_to_hkey(macaddr);
1356                 hash_add(vsi->mac_filter_hash, &f->hlist, key);
1357
1358                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1359                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1360         }
1361
1362         /* If we're asked to add a filter that has been marked for removal, it
1363          * is safe to simply restore it to active state. __i40e_del_filter
1364          * will have simply deleted any filters which were previously marked
1365          * NEW or FAILED, so if it is currently marked REMOVE it must have
1366          * previously been ACTIVE. Since we haven't yet run the sync filters
1367          * task, just restore this filter to the ACTIVE state so that the
1368          * sync task leaves it in place
1369          */
1370         if (f->state == I40E_FILTER_REMOVE)
1371                 f->state = I40E_FILTER_ACTIVE;
1372
1373         return f;
1374 }
1375
1376 /**
1377  * __i40e_del_filter - Remove a specific filter from the VSI
1378  * @vsi: VSI to remove from
1379  * @f: the filter to remove from the list
1380  *
1381  * This function should be called instead of i40e_del_filter only if you know
1382  * the exact filter you will remove already, such as via i40e_find_filter or
1383  * i40e_find_mac.
1384  *
1385  * NOTE: This function is expected to be called with mac_filter_hash_lock
1386  * being held.
1387  * ANOTHER NOTE: This function MUST be called from within the context of
1388  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1389  * instead of list_for_each_entry().
1390  **/
1391 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1392 {
1393         if (!f)
1394                 return;
1395
1396         /* If the filter was never added to firmware then we can just delete it
1397          * directly and we don't want to set the status to remove or else an
1398          * admin queue command will unnecessarily fire.
1399          */
1400         if ((f->state == I40E_FILTER_FAILED) ||
1401             (f->state == I40E_FILTER_NEW)) {
1402                 hash_del(&f->hlist);
1403                 kfree(f);
1404         } else {
1405                 f->state = I40E_FILTER_REMOVE;
1406         }
1407
1408         vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1409         vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1410 }
1411
1412 /**
1413  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1414  * @vsi: the VSI to be searched
1415  * @macaddr: the MAC address
1416  * @vlan: the VLAN
1417  *
1418  * NOTE: This function is expected to be called with mac_filter_hash_lock
1419  * being held.
1420  * ANOTHER NOTE: This function MUST be called from within the context of
1421  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1422  * instead of list_for_each_entry().
1423  **/
1424 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1425 {
1426         struct i40e_mac_filter *f;
1427
1428         if (!vsi || !macaddr)
1429                 return;
1430
1431         f = i40e_find_filter(vsi, macaddr, vlan);
1432         __i40e_del_filter(vsi, f);
1433 }
1434
1435 /**
1436  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1437  * @vsi: the VSI to be searched
1438  * @macaddr: the mac address to be filtered
1439  *
1440  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1441  * go through all the macvlan filters and add a macvlan filter for each
1442  * unique vlan that already exists. If a PVID has been assigned, instead only
1443  * add the macaddr to that VLAN.
1444  *
1445  * Returns last filter added on success, else NULL
1446  **/
1447 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1448                                             const u8 *macaddr)
1449 {
1450         struct i40e_mac_filter *f, *add = NULL;
1451         struct hlist_node *h;
1452         int bkt;
1453
1454         if (vsi->info.pvid)
1455                 return i40e_add_filter(vsi, macaddr,
1456                                        le16_to_cpu(vsi->info.pvid));
1457
1458         if (!i40e_is_vsi_in_vlan(vsi))
1459                 return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1460
1461         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1462                 if (f->state == I40E_FILTER_REMOVE)
1463                         continue;
1464                 add = i40e_add_filter(vsi, macaddr, f->vlan);
1465                 if (!add)
1466                         return NULL;
1467         }
1468
1469         return add;
1470 }
1471
1472 /**
1473  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1474  * @vsi: the VSI to be searched
1475  * @macaddr: the mac address to be removed
1476  *
1477  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1478  * associated with.
1479  *
1480  * Returns 0 for success, or error
1481  **/
1482 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1483 {
1484         struct i40e_mac_filter *f;
1485         struct hlist_node *h;
1486         bool found = false;
1487         int bkt;
1488
1489         WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1490              "Missing mac_filter_hash_lock\n");
1491         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1492                 if (ether_addr_equal(macaddr, f->macaddr)) {
1493                         __i40e_del_filter(vsi, f);
1494                         found = true;
1495                 }
1496         }
1497
1498         if (found)
1499                 return 0;
1500         else
1501                 return -ENOENT;
1502 }
1503
1504 /**
1505  * i40e_set_mac - NDO callback to set mac address
1506  * @netdev: network interface device structure
1507  * @p: pointer to an address structure
1508  *
1509  * Returns 0 on success, negative on failure
1510  **/
1511 static int i40e_set_mac(struct net_device *netdev, void *p)
1512 {
1513         struct i40e_netdev_priv *np = netdev_priv(netdev);
1514         struct i40e_vsi *vsi = np->vsi;
1515         struct i40e_pf *pf = vsi->back;
1516         struct i40e_hw *hw = &pf->hw;
1517         struct sockaddr *addr = p;
1518
1519         if (!is_valid_ether_addr(addr->sa_data))
1520                 return -EADDRNOTAVAIL;
1521
1522         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1523                 netdev_info(netdev, "already using mac address %pM\n",
1524                             addr->sa_data);
1525                 return 0;
1526         }
1527
1528         if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
1529             test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
1530                 return -EADDRNOTAVAIL;
1531
1532         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1533                 netdev_info(netdev, "returning to hw mac address %pM\n",
1534                             hw->mac.addr);
1535         else
1536                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1537
1538         spin_lock_bh(&vsi->mac_filter_hash_lock);
1539         i40e_del_mac_filter(vsi, netdev->dev_addr);
1540         i40e_add_mac_filter(vsi, addr->sa_data);
1541         spin_unlock_bh(&vsi->mac_filter_hash_lock);
1542         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1543         if (vsi->type == I40E_VSI_MAIN) {
1544                 i40e_status ret;
1545
1546                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1547                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1548                                                 addr->sa_data, NULL);
1549                 if (ret)
1550                         netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1551                                     i40e_stat_str(hw, ret),
1552                                     i40e_aq_str(hw, hw->aq.asq_last_status));
1553         }
1554
1555         /* schedule our worker thread which will take care of
1556          * applying the new filter changes
1557          */
1558         i40e_service_event_schedule(vsi->back);
1559         return 0;
1560 }
1561
1562 /**
1563  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1564  * @vsi: the VSI being setup
1565  * @ctxt: VSI context structure
1566  * @enabled_tc: Enabled TCs bitmap
1567  * @is_add: True if called before Add VSI
1568  *
1569  * Setup VSI queue mapping for enabled traffic classes.
1570  **/
1571 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1572                                      struct i40e_vsi_context *ctxt,
1573                                      u8 enabled_tc,
1574                                      bool is_add)
1575 {
1576         struct i40e_pf *pf = vsi->back;
1577         u16 sections = 0;
1578         u8 netdev_tc = 0;
1579         u16 numtc = 0;
1580         u16 qcount;
1581         u8 offset;
1582         u16 qmap;
1583         int i;
1584         u16 num_tc_qps = 0;
1585
1586         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1587         offset = 0;
1588
1589         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1590                 /* Find numtc from enabled TC bitmap */
1591                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1592                         if (enabled_tc & BIT(i)) /* TC is enabled */
1593                                 numtc++;
1594                 }
1595                 if (!numtc) {
1596                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1597                         numtc = 1;
1598                 }
1599         } else {
1600                 /* At least TC0 is enabled in case of non-DCB case */
1601                 numtc = 1;
1602         }
1603
1604         vsi->tc_config.numtc = numtc;
1605         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1606         /* Number of queues per enabled TC */
1607         qcount = vsi->alloc_queue_pairs;
1608
1609         num_tc_qps = qcount / numtc;
1610         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1611
1612         /* Setup queue offset/count for all TCs for given VSI */
1613         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1614                 /* See if the given TC is enabled for the given VSI */
1615                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1616                         /* TC is enabled */
1617                         int pow, num_qps;
1618
1619                         switch (vsi->type) {
1620                         case I40E_VSI_MAIN:
1621                                 qcount = min_t(int, pf->alloc_rss_size,
1622                                                num_tc_qps);
1623                                 break;
1624                         case I40E_VSI_FDIR:
1625                         case I40E_VSI_SRIOV:
1626                         case I40E_VSI_VMDQ2:
1627                         default:
1628                                 qcount = num_tc_qps;
1629                                 WARN_ON(i != 0);
1630                                 break;
1631                         }
1632                         vsi->tc_config.tc_info[i].qoffset = offset;
1633                         vsi->tc_config.tc_info[i].qcount = qcount;
1634
1635                         /* find the next higher power-of-2 of num queue pairs */
1636                         num_qps = qcount;
1637                         pow = 0;
1638                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1639                                 pow++;
1640                                 num_qps >>= 1;
1641                         }
1642
1643                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1644                         qmap =
1645                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1646                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1647
1648                         offset += qcount;
1649                 } else {
1650                         /* TC is not enabled so set the offset to
1651                          * default queue and allocate one queue
1652                          * for the given TC.
1653                          */
1654                         vsi->tc_config.tc_info[i].qoffset = 0;
1655                         vsi->tc_config.tc_info[i].qcount = 1;
1656                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1657
1658                         qmap = 0;
1659                 }
1660                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1661         }
1662
1663         /* Set actual Tx/Rx queue pairs */
1664         vsi->num_queue_pairs = offset;
1665         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1666                 if (vsi->req_queue_pairs > 0)
1667                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1668                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1669                         vsi->num_queue_pairs = pf->num_lan_msix;
1670         }
1671
1672         /* Scheduler section valid can only be set for ADD VSI */
1673         if (is_add) {
1674                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1675
1676                 ctxt->info.up_enable_bits = enabled_tc;
1677         }
1678         if (vsi->type == I40E_VSI_SRIOV) {
1679                 ctxt->info.mapping_flags |=
1680                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1681                 for (i = 0; i < vsi->num_queue_pairs; i++)
1682                         ctxt->info.queue_mapping[i] =
1683                                                cpu_to_le16(vsi->base_queue + i);
1684         } else {
1685                 ctxt->info.mapping_flags |=
1686                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1687                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1688         }
1689         ctxt->info.valid_sections |= cpu_to_le16(sections);
1690 }
1691
1692 /**
1693  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1694  * @netdev: the netdevice
1695  * @addr: address to add
1696  *
1697  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1698  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1699  */
1700 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1701 {
1702         struct i40e_netdev_priv *np = netdev_priv(netdev);
1703         struct i40e_vsi *vsi = np->vsi;
1704
1705         if (i40e_add_mac_filter(vsi, addr))
1706                 return 0;
1707         else
1708                 return -ENOMEM;
1709 }
1710
1711 /**
1712  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1713  * @netdev: the netdevice
1714  * @addr: address to add
1715  *
1716  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1717  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1718  */
1719 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1720 {
1721         struct i40e_netdev_priv *np = netdev_priv(netdev);
1722         struct i40e_vsi *vsi = np->vsi;
1723
1724         i40e_del_mac_filter(vsi, addr);
1725
1726         return 0;
1727 }
1728
1729 /**
1730  * i40e_set_rx_mode - NDO callback to set the netdev filters
1731  * @netdev: network interface device structure
1732  **/
1733 static void i40e_set_rx_mode(struct net_device *netdev)
1734 {
1735         struct i40e_netdev_priv *np = netdev_priv(netdev);
1736         struct i40e_vsi *vsi = np->vsi;
1737
1738         spin_lock_bh(&vsi->mac_filter_hash_lock);
1739
1740         __dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1741         __dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1742
1743         spin_unlock_bh(&vsi->mac_filter_hash_lock);
1744
1745         /* check for other flag changes */
1746         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1747                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1748                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1749         }
1750
1751         /* schedule our worker thread which will take care of
1752          * applying the new filter changes
1753          */
1754         i40e_service_event_schedule(vsi->back);
1755 }
1756
1757 /**
1758  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1759  * @vsi: Pointer to VSI struct
1760  * @from: Pointer to list which contains MAC filter entries - changes to
1761  *        those entries needs to be undone.
1762  *
1763  * MAC filter entries from this list were slated for deletion.
1764  **/
1765 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1766                                          struct hlist_head *from)
1767 {
1768         struct i40e_mac_filter *f;
1769         struct hlist_node *h;
1770
1771         hlist_for_each_entry_safe(f, h, from, hlist) {
1772                 u64 key = i40e_addr_to_hkey(f->macaddr);
1773
1774                 /* Move the element back into MAC filter list*/
1775                 hlist_del(&f->hlist);
1776                 hash_add(vsi->mac_filter_hash, &f->hlist, key);
1777         }
1778 }
1779
1780 /**
1781  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1782  * @vsi: Pointer to vsi struct
1783  * @from: Pointer to list which contains MAC filter entries - changes to
1784  *        those entries needs to be undone.
1785  *
1786  * MAC filter entries from this list were slated for addition.
1787  **/
1788 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1789                                          struct hlist_head *from)
1790 {
1791         struct i40e_new_mac_filter *new;
1792         struct hlist_node *h;
1793
1794         hlist_for_each_entry_safe(new, h, from, hlist) {
1795                 /* We can simply free the wrapper structure */
1796                 hlist_del(&new->hlist);
1797                 kfree(new);
1798         }
1799 }
1800
1801 /**
1802  * i40e_next_entry - Get the next non-broadcast filter from a list
1803  * @next: pointer to filter in list
1804  *
1805  * Returns the next non-broadcast filter in the list. Required so that we
1806  * ignore broadcast filters within the list, since these are not handled via
1807  * the normal firmware update path.
1808  */
1809 static
1810 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
1811 {
1812         hlist_for_each_entry_continue(next, hlist) {
1813                 if (!is_broadcast_ether_addr(next->f->macaddr))
1814                         return next;
1815         }
1816
1817         return NULL;
1818 }
1819
1820 /**
1821  * i40e_update_filter_state - Update filter state based on return data
1822  * from firmware
1823  * @count: Number of filters added
1824  * @add_list: return data from fw
1825  * @head: pointer to first filter in current batch
1826  *
1827  * MAC filter entries from list were slated to be added to device. Returns
1828  * number of successful filters. Note that 0 does NOT mean success!
1829  **/
1830 static int
1831 i40e_update_filter_state(int count,
1832                          struct i40e_aqc_add_macvlan_element_data *add_list,
1833                          struct i40e_new_mac_filter *add_head)
1834 {
1835         int retval = 0;
1836         int i;
1837
1838         for (i = 0; i < count; i++) {
1839                 /* Always check status of each filter. We don't need to check
1840                  * the firmware return status because we pre-set the filter
1841                  * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
1842                  * request to the adminq. Thus, if it no longer matches then
1843                  * we know the filter is active.
1844                  */
1845                 if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
1846                         add_head->state = I40E_FILTER_FAILED;
1847                 } else {
1848                         add_head->state = I40E_FILTER_ACTIVE;
1849                         retval++;
1850                 }
1851
1852                 add_head = i40e_next_filter(add_head);
1853                 if (!add_head)
1854                         break;
1855         }
1856
1857         return retval;
1858 }
1859
1860 /**
1861  * i40e_aqc_del_filters - Request firmware to delete a set of filters
1862  * @vsi: ptr to the VSI
1863  * @vsi_name: name to display in messages
1864  * @list: the list of filters to send to firmware
1865  * @num_del: the number of filters to delete
1866  * @retval: Set to -EIO on failure to delete
1867  *
1868  * Send a request to firmware via AdminQ to delete a set of filters. Uses
1869  * *retval instead of a return value so that success does not force ret_val to
1870  * be set to 0. This ensures that a sequence of calls to this function
1871  * preserve the previous value of *retval on successful delete.
1872  */
1873 static
1874 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
1875                           struct i40e_aqc_remove_macvlan_element_data *list,
1876                           int num_del, int *retval)
1877 {
1878         struct i40e_hw *hw = &vsi->back->hw;
1879         i40e_status aq_ret;
1880         int aq_err;
1881
1882         aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
1883         aq_err = hw->aq.asq_last_status;
1884
1885         /* Explicitly ignore and do not report when firmware returns ENOENT */
1886         if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1887                 *retval = -EIO;
1888                 dev_info(&vsi->back->pdev->dev,
1889                          "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1890                          vsi_name, i40e_stat_str(hw, aq_ret),
1891                          i40e_aq_str(hw, aq_err));
1892         }
1893 }
1894
1895 /**
1896  * i40e_aqc_add_filters - Request firmware to add a set of filters
1897  * @vsi: ptr to the VSI
1898  * @vsi_name: name to display in messages
1899  * @list: the list of filters to send to firmware
1900  * @add_head: Position in the add hlist
1901  * @num_add: the number of filters to add
1902  * @promisc_change: set to true on exit if promiscuous mode was forced on
1903  *
1904  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
1905  * promisc_changed to true if the firmware has run out of space for more
1906  * filters.
1907  */
1908 static
1909 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
1910                           struct i40e_aqc_add_macvlan_element_data *list,
1911                           struct i40e_new_mac_filter *add_head,
1912                           int num_add, bool *promisc_changed)
1913 {
1914         struct i40e_hw *hw = &vsi->back->hw;
1915         int aq_err, fcnt;
1916
1917         i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
1918         aq_err = hw->aq.asq_last_status;
1919         fcnt = i40e_update_filter_state(num_add, list, add_head);
1920
1921         if (fcnt != num_add) {
1922                 *promisc_changed = true;
1923                 set_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
1924                 dev_warn(&vsi->back->pdev->dev,
1925                          "Error %s adding RX filters on %s, promiscuous mode forced on\n",
1926                          i40e_aq_str(hw, aq_err),
1927                          vsi_name);
1928         }
1929 }
1930
1931 /**
1932  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
1933  * @vsi: pointer to the VSI
1934  * @f: filter data
1935  *
1936  * This function sets or clears the promiscuous broadcast flags for VLAN
1937  * filters in order to properly receive broadcast frames. Assumes that only
1938  * broadcast filters are passed.
1939  *
1940  * Returns status indicating success or failure;
1941  **/
1942 static i40e_status
1943 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
1944                           struct i40e_mac_filter *f)
1945 {
1946         bool enable = f->state == I40E_FILTER_NEW;
1947         struct i40e_hw *hw = &vsi->back->hw;
1948         i40e_status aq_ret;
1949
1950         if (f->vlan == I40E_VLAN_ANY) {
1951                 aq_ret = i40e_aq_set_vsi_broadcast(hw,
1952                                                    vsi->seid,
1953                                                    enable,
1954                                                    NULL);
1955         } else {
1956                 aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
1957                                                             vsi->seid,
1958                                                             enable,
1959                                                             f->vlan,
1960                                                             NULL);
1961         }
1962
1963         if (aq_ret)
1964                 dev_warn(&vsi->back->pdev->dev,
1965                          "Error %s setting broadcast promiscuous mode on %s\n",
1966                          i40e_aq_str(hw, hw->aq.asq_last_status),
1967                          vsi_name);
1968
1969         return aq_ret;
1970 }
1971
1972 /**
1973  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1974  * @vsi: ptr to the VSI
1975  *
1976  * Push any outstanding VSI filter changes through the AdminQ.
1977  *
1978  * Returns 0 or error value
1979  **/
1980 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1981 {
1982         struct hlist_head tmp_add_list, tmp_del_list;
1983         struct i40e_mac_filter *f;
1984         struct i40e_new_mac_filter *new, *add_head = NULL;
1985         struct i40e_hw *hw = &vsi->back->hw;
1986         unsigned int failed_filters = 0;
1987         unsigned int vlan_filters = 0;
1988         bool promisc_changed = false;
1989         char vsi_name[16] = "PF";
1990         int filter_list_len = 0;
1991         i40e_status aq_ret = 0;
1992         u32 changed_flags = 0;
1993         struct hlist_node *h;
1994         struct i40e_pf *pf;
1995         int num_add = 0;
1996         int num_del = 0;
1997         int retval = 0;
1998         u16 cmd_flags;
1999         int list_size;
2000         int bkt;
2001
2002         /* empty array typed pointers, kcalloc later */
2003         struct i40e_aqc_add_macvlan_element_data *add_list;
2004         struct i40e_aqc_remove_macvlan_element_data *del_list;
2005
2006         while (test_and_set_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state))
2007                 usleep_range(1000, 2000);
2008         pf = vsi->back;
2009
2010         if (vsi->netdev) {
2011                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2012                 vsi->current_netdev_flags = vsi->netdev->flags;
2013         }
2014
2015         INIT_HLIST_HEAD(&tmp_add_list);
2016         INIT_HLIST_HEAD(&tmp_del_list);
2017
2018         if (vsi->type == I40E_VSI_SRIOV)
2019                 snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2020         else if (vsi->type != I40E_VSI_MAIN)
2021                 snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2022
2023         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2024                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2025
2026                 spin_lock_bh(&vsi->mac_filter_hash_lock);
2027                 /* Create a list of filters to delete. */
2028                 hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2029                         if (f->state == I40E_FILTER_REMOVE) {
2030                                 /* Move the element into temporary del_list */
2031                                 hash_del(&f->hlist);
2032                                 hlist_add_head(&f->hlist, &tmp_del_list);
2033
2034                                 /* Avoid counting removed filters */
2035                                 continue;
2036                         }
2037                         if (f->state == I40E_FILTER_NEW) {
2038                                 /* Create a temporary i40e_new_mac_filter */
2039                                 new = kzalloc(sizeof(*new), GFP_ATOMIC);
2040                                 if (!new)
2041                                         goto err_no_memory_locked;
2042
2043                                 /* Store pointer to the real filter */
2044                                 new->f = f;
2045                                 new->state = f->state;
2046
2047                                 /* Add it to the hash list */
2048                                 hlist_add_head(&new->hlist, &tmp_add_list);
2049                         }
2050
2051                         /* Count the number of active (current and new) VLAN
2052                          * filters we have now. Does not count filters which
2053                          * are marked for deletion.
2054                          */
2055                         if (f->vlan > 0)
2056                                 vlan_filters++;
2057                 }
2058
2059                 retval = i40e_correct_mac_vlan_filters(vsi,
2060                                                        &tmp_add_list,
2061                                                        &tmp_del_list,
2062                                                        vlan_filters);
2063                 if (retval)
2064                         goto err_no_memory_locked;
2065
2066                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
2067         }
2068
2069         /* Now process 'del_list' outside the lock */
2070         if (!hlist_empty(&tmp_del_list)) {
2071                 filter_list_len = hw->aq.asq_buf_size /
2072                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
2073                 list_size = filter_list_len *
2074                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
2075                 del_list = kzalloc(list_size, GFP_ATOMIC);
2076                 if (!del_list)
2077                         goto err_no_memory;
2078
2079                 hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2080                         cmd_flags = 0;
2081
2082                         /* handle broadcast filters by updating the broadcast
2083                          * promiscuous flag and release filter list.
2084                          */
2085                         if (is_broadcast_ether_addr(f->macaddr)) {
2086                                 i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2087
2088                                 hlist_del(&f->hlist);
2089                                 kfree(f);
2090                                 continue;
2091                         }
2092
2093                         /* add to delete list */
2094                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2095                         if (f->vlan == I40E_VLAN_ANY) {
2096                                 del_list[num_del].vlan_tag = 0;
2097                                 cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2098                         } else {
2099                                 del_list[num_del].vlan_tag =
2100                                         cpu_to_le16((u16)(f->vlan));
2101                         }
2102
2103                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2104                         del_list[num_del].flags = cmd_flags;
2105                         num_del++;
2106
2107                         /* flush a full buffer */
2108                         if (num_del == filter_list_len) {
2109                                 i40e_aqc_del_filters(vsi, vsi_name, del_list,
2110                                                      num_del, &retval);
2111                                 memset(del_list, 0, list_size);
2112                                 num_del = 0;
2113                         }
2114                         /* Release memory for MAC filter entries which were
2115                          * synced up with HW.
2116                          */
2117                         hlist_del(&f->hlist);
2118                         kfree(f);
2119                 }
2120
2121                 if (num_del) {
2122                         i40e_aqc_del_filters(vsi, vsi_name, del_list,
2123                                              num_del, &retval);
2124                 }
2125
2126                 kfree(del_list);
2127                 del_list = NULL;
2128         }
2129
2130         if (!hlist_empty(&tmp_add_list)) {
2131                 /* Do all the adds now. */
2132                 filter_list_len = hw->aq.asq_buf_size /
2133                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2134                 list_size = filter_list_len *
2135                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2136                 add_list = kzalloc(list_size, GFP_ATOMIC);
2137                 if (!add_list)
2138                         goto err_no_memory;
2139
2140                 num_add = 0;
2141                 hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2142                         if (test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2143                                      vsi->state)) {
2144                                 new->state = I40E_FILTER_FAILED;
2145                                 continue;
2146                         }
2147
2148                         /* handle broadcast filters by updating the broadcast
2149                          * promiscuous flag instead of adding a MAC filter.
2150                          */
2151                         if (is_broadcast_ether_addr(new->f->macaddr)) {
2152                                 if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2153                                                               new->f))
2154                                         new->state = I40E_FILTER_FAILED;
2155                                 else
2156                                         new->state = I40E_FILTER_ACTIVE;
2157                                 continue;
2158                         }
2159
2160                         /* add to add array */
2161                         if (num_add == 0)
2162                                 add_head = new;
2163                         cmd_flags = 0;
2164                         ether_addr_copy(add_list[num_add].mac_addr,
2165                                         new->f->macaddr);
2166                         if (new->f->vlan == I40E_VLAN_ANY) {
2167                                 add_list[num_add].vlan_tag = 0;
2168                                 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2169                         } else {
2170                                 add_list[num_add].vlan_tag =
2171                                         cpu_to_le16((u16)(new->f->vlan));
2172                         }
2173                         add_list[num_add].queue_number = 0;
2174                         /* set invalid match method for later detection */
2175                         add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2176                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2177                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2178                         num_add++;
2179
2180                         /* flush a full buffer */
2181                         if (num_add == filter_list_len) {
2182                                 i40e_aqc_add_filters(vsi, vsi_name, add_list,
2183                                                      add_head, num_add,
2184                                                      &promisc_changed);
2185                                 memset(add_list, 0, list_size);
2186                                 num_add = 0;
2187                         }
2188                 }
2189                 if (num_add) {
2190                         i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2191                                              num_add, &promisc_changed);
2192                 }
2193                 /* Now move all of the filters from the temp add list back to
2194                  * the VSI's list.
2195                  */
2196                 spin_lock_bh(&vsi->mac_filter_hash_lock);
2197                 hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2198                         /* Only update the state if we're still NEW */
2199                         if (new->f->state == I40E_FILTER_NEW)
2200                                 new->f->state = new->state;
2201                         hlist_del(&new->hlist);
2202                         kfree(new);
2203                 }
2204                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
2205                 kfree(add_list);
2206                 add_list = NULL;
2207         }
2208
2209         /* Determine the number of active and failed filters. */
2210         spin_lock_bh(&vsi->mac_filter_hash_lock);
2211         vsi->active_filters = 0;
2212         hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2213                 if (f->state == I40E_FILTER_ACTIVE)
2214                         vsi->active_filters++;
2215                 else if (f->state == I40E_FILTER_FAILED)
2216                         failed_filters++;
2217         }
2218         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2219
2220         /* If promiscuous mode has changed, we need to calculate a new
2221          * threshold for when we are safe to exit
2222          */
2223         if (promisc_changed)
2224                 vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2225
2226         /* Check if we are able to exit overflow promiscuous mode. We can
2227          * safely exit if we didn't just enter, we no longer have any failed
2228          * filters, and we have reduced filters below the threshold value.
2229          */
2230         if (test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state) &&
2231             !promisc_changed && !failed_filters &&
2232             (vsi->active_filters < vsi->promisc_threshold)) {
2233                 dev_info(&pf->pdev->dev,
2234                          "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2235                          vsi_name);
2236                 clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2237                 promisc_changed = true;
2238                 vsi->promisc_threshold = 0;
2239         }
2240
2241         /* if the VF is not trusted do not do promisc */
2242         if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2243                 clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
2244                 goto out;
2245         }
2246
2247         /* check for changes in promiscuous modes */
2248         if (changed_flags & IFF_ALLMULTI) {
2249                 bool cur_multipromisc;
2250
2251                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2252                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2253                                                                vsi->seid,
2254                                                                cur_multipromisc,
2255                                                                NULL);
2256                 if (aq_ret) {
2257                         retval = i40e_aq_rc_to_posix(aq_ret,
2258                                                      hw->aq.asq_last_status);
2259                         dev_info(&pf->pdev->dev,
2260                                  "set multi promisc failed on %s, err %s aq_err %s\n",
2261                                  vsi_name,
2262                                  i40e_stat_str(hw, aq_ret),
2263                                  i40e_aq_str(hw, hw->aq.asq_last_status));
2264                 }
2265         }
2266         if ((changed_flags & IFF_PROMISC) ||
2267             (promisc_changed &&
2268              test_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state))) {
2269                 bool cur_promisc;
2270
2271                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2272                                test_bit(__I40E_VSI_OVERFLOW_PROMISC,
2273                                         vsi->state));
2274                 if ((vsi->type == I40E_VSI_MAIN) &&
2275                     (pf->lan_veb != I40E_NO_VEB) &&
2276                     !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2277                         /* set defport ON for Main VSI instead of true promisc
2278                          * this way we will get all unicast/multicast and VLAN
2279                          * promisc behavior but will not get VF or VMDq traffic
2280                          * replicated on the Main VSI.
2281                          */
2282                         if (pf->cur_promisc != cur_promisc) {
2283                                 pf->cur_promisc = cur_promisc;
2284                                 if (cur_promisc)
2285                                         aq_ret =
2286                                               i40e_aq_set_default_vsi(hw,
2287                                                                       vsi->seid,
2288                                                                       NULL);
2289                                 else
2290                                         aq_ret =
2291                                             i40e_aq_clear_default_vsi(hw,
2292                                                                       vsi->seid,
2293                                                                       NULL);
2294                                 if (aq_ret) {
2295                                         retval = i40e_aq_rc_to_posix(aq_ret,
2296                                                         hw->aq.asq_last_status);
2297                                         dev_info(&pf->pdev->dev,
2298                                                  "Set default VSI failed on %s, err %s, aq_err %s\n",
2299                                                  vsi_name,
2300                                                  i40e_stat_str(hw, aq_ret),
2301                                                  i40e_aq_str(hw,
2302                                                      hw->aq.asq_last_status));
2303                                 }
2304                         }
2305                 } else {
2306                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2307                                                           hw,
2308                                                           vsi->seid,
2309                                                           cur_promisc, NULL,
2310                                                           true);
2311                         if (aq_ret) {
2312                                 retval =
2313                                 i40e_aq_rc_to_posix(aq_ret,
2314                                                     hw->aq.asq_last_status);
2315                                 dev_info(&pf->pdev->dev,
2316                                          "set unicast promisc failed on %s, err %s, aq_err %s\n",
2317                                          vsi_name,
2318                                          i40e_stat_str(hw, aq_ret),
2319                                          i40e_aq_str(hw,
2320                                                      hw->aq.asq_last_status));
2321                         }
2322                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2323                                                           hw,
2324                                                           vsi->seid,
2325                                                           cur_promisc, NULL);
2326                         if (aq_ret) {
2327                                 retval =
2328                                 i40e_aq_rc_to_posix(aq_ret,
2329                                                     hw->aq.asq_last_status);
2330                                 dev_info(&pf->pdev->dev,
2331                                          "set multicast promisc failed on %s, err %s, aq_err %s\n",
2332                                          vsi_name,
2333                                          i40e_stat_str(hw, aq_ret),
2334                                          i40e_aq_str(hw,
2335                                                      hw->aq.asq_last_status));
2336                         }
2337                 }
2338                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2339                                                    vsi->seid,
2340                                                    cur_promisc, NULL);
2341                 if (aq_ret) {
2342                         retval = i40e_aq_rc_to_posix(aq_ret,
2343                                                      pf->hw.aq.asq_last_status);
2344                         dev_info(&pf->pdev->dev,
2345                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2346                                          i40e_stat_str(hw, aq_ret),
2347                                          i40e_aq_str(hw,
2348                                                      hw->aq.asq_last_status));
2349                 }
2350         }
2351 out:
2352         /* if something went wrong then set the changed flag so we try again */
2353         if (retval)
2354                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2355
2356         clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2357         return retval;
2358
2359 err_no_memory:
2360         /* Restore elements on the temporary add and delete lists */
2361         spin_lock_bh(&vsi->mac_filter_hash_lock);
2362 err_no_memory_locked:
2363         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2364         i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2365         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2366
2367         vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2368         clear_bit(__I40E_VSI_SYNCING_FILTERS, vsi->state);
2369         return -ENOMEM;
2370 }
2371
2372 /**
2373  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2374  * @pf: board private structure
2375  **/
2376 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2377 {
2378         int v;
2379
2380         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2381                 return;
2382         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2383
2384         for (v = 0; v < pf->num_alloc_vsi; v++) {
2385                 if (pf->vsi[v] &&
2386                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2387                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2388
2389                         if (ret) {
2390                                 /* come back and try again later */
2391                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2392                                 break;
2393                         }
2394                 }
2395         }
2396 }
2397
2398 /**
2399  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2400  * @netdev: network interface device structure
2401  * @new_mtu: new value for maximum frame size
2402  *
2403  * Returns 0 on success, negative on failure
2404  **/
2405 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2406 {
2407         struct i40e_netdev_priv *np = netdev_priv(netdev);
2408         struct i40e_vsi *vsi = np->vsi;
2409         struct i40e_pf *pf = vsi->back;
2410
2411         netdev_info(netdev, "changing MTU from %d to %d\n",
2412                     netdev->mtu, new_mtu);
2413         netdev->mtu = new_mtu;
2414         if (netif_running(netdev))
2415                 i40e_vsi_reinit_locked(vsi);
2416         pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2417                       I40E_FLAG_CLIENT_L2_CHANGE);
2418         return 0;
2419 }
2420
2421 /**
2422  * i40e_ioctl - Access the hwtstamp interface
2423  * @netdev: network interface device structure
2424  * @ifr: interface request data
2425  * @cmd: ioctl command
2426  **/
2427 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2428 {
2429         struct i40e_netdev_priv *np = netdev_priv(netdev);
2430         struct i40e_pf *pf = np->vsi->back;
2431
2432         switch (cmd) {
2433         case SIOCGHWTSTAMP:
2434                 return i40e_ptp_get_ts_config(pf, ifr);
2435         case SIOCSHWTSTAMP:
2436                 return i40e_ptp_set_ts_config(pf, ifr);
2437         default:
2438                 return -EOPNOTSUPP;
2439         }
2440 }
2441
2442 /**
2443  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2444  * @vsi: the vsi being adjusted
2445  **/
2446 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2447 {
2448         struct i40e_vsi_context ctxt;
2449         i40e_status ret;
2450
2451         if ((vsi->info.valid_sections &
2452              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2453             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2454                 return;  /* already enabled */
2455
2456         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2457         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2458                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2459
2460         ctxt.seid = vsi->seid;
2461         ctxt.info = vsi->info;
2462         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2463         if (ret) {
2464                 dev_info(&vsi->back->pdev->dev,
2465                          "update vlan stripping failed, err %s aq_err %s\n",
2466                          i40e_stat_str(&vsi->back->hw, ret),
2467                          i40e_aq_str(&vsi->back->hw,
2468                                      vsi->back->hw.aq.asq_last_status));
2469         }
2470 }
2471
2472 /**
2473  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2474  * @vsi: the vsi being adjusted
2475  **/
2476 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2477 {
2478         struct i40e_vsi_context ctxt;
2479         i40e_status ret;
2480
2481         if ((vsi->info.valid_sections &
2482              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2483             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2484              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2485                 return;  /* already disabled */
2486
2487         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2488         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2489                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2490
2491         ctxt.seid = vsi->seid;
2492         ctxt.info = vsi->info;
2493         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2494         if (ret) {
2495                 dev_info(&vsi->back->pdev->dev,
2496                          "update vlan stripping failed, err %s aq_err %s\n",
2497                          i40e_stat_str(&vsi->back->hw, ret),
2498                          i40e_aq_str(&vsi->back->hw,
2499                                      vsi->back->hw.aq.asq_last_status));
2500         }
2501 }
2502
2503 /**
2504  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2505  * @netdev: network interface to be adjusted
2506  * @features: netdev features to test if VLAN offload is enabled or not
2507  **/
2508 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2509 {
2510         struct i40e_netdev_priv *np = netdev_priv(netdev);
2511         struct i40e_vsi *vsi = np->vsi;
2512
2513         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2514                 i40e_vlan_stripping_enable(vsi);
2515         else
2516                 i40e_vlan_stripping_disable(vsi);
2517 }
2518
2519 /**
2520  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2521  * @vsi: the vsi being configured
2522  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2523  *
2524  * This is a helper function for adding a new MAC/VLAN filter with the
2525  * specified VLAN for each existing MAC address already in the hash table.
2526  * This function does *not* perform any accounting to update filters based on
2527  * VLAN mode.
2528  *
2529  * NOTE: this function expects to be called while under the
2530  * mac_filter_hash_lock
2531  **/
2532 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2533 {
2534         struct i40e_mac_filter *f, *add_f;
2535         struct hlist_node *h;
2536         int bkt;
2537
2538         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2539                 if (f->state == I40E_FILTER_REMOVE)
2540                         continue;
2541                 add_f = i40e_add_filter(vsi, f->macaddr, vid);
2542                 if (!add_f) {
2543                         dev_info(&vsi->back->pdev->dev,
2544                                  "Could not add vlan filter %d for %pM\n",
2545                                  vid, f->macaddr);
2546                         return -ENOMEM;
2547                 }
2548         }
2549
2550         return 0;
2551 }
2552
2553 /**
2554  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2555  * @vsi: the VSI being configured
2556  * @vid: VLAN id to be added
2557  **/
2558 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2559 {
2560         int err;
2561
2562         if (!vid || vsi->info.pvid)
2563                 return -EINVAL;
2564
2565         /* Locked once because all functions invoked below iterates list*/
2566         spin_lock_bh(&vsi->mac_filter_hash_lock);
2567         err = i40e_add_vlan_all_mac(vsi, vid);
2568         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2569         if (err)
2570                 return err;
2571
2572         /* schedule our worker thread which will take care of
2573          * applying the new filter changes
2574          */
2575         i40e_service_event_schedule(vsi->back);
2576         return 0;
2577 }
2578
2579 /**
2580  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2581  * @vsi: the vsi being configured
2582  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2583  *
2584  * This function should be used to remove all VLAN filters which match the
2585  * given VID. It does not schedule the service event and does not take the
2586  * mac_filter_hash_lock so it may be combined with other operations under
2587  * a single invocation of the mac_filter_hash_lock.
2588  *
2589  * NOTE: this function expects to be called while under the
2590  * mac_filter_hash_lock
2591  */
2592 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2593 {
2594         struct i40e_mac_filter *f;
2595         struct hlist_node *h;
2596         int bkt;
2597
2598         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2599                 if (f->vlan == vid)
2600                         __i40e_del_filter(vsi, f);
2601         }
2602 }
2603
2604 /**
2605  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2606  * @vsi: the VSI being configured
2607  * @vid: VLAN id to be removed
2608  **/
2609 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2610 {
2611         if (!vid || vsi->info.pvid)
2612                 return;
2613
2614         spin_lock_bh(&vsi->mac_filter_hash_lock);
2615         i40e_rm_vlan_all_mac(vsi, vid);
2616         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2617
2618         /* schedule our worker thread which will take care of
2619          * applying the new filter changes
2620          */
2621         i40e_service_event_schedule(vsi->back);
2622 }
2623
2624 /**
2625  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2626  * @netdev: network interface to be adjusted
2627  * @vid: vlan id to be added
2628  *
2629  * net_device_ops implementation for adding vlan ids
2630  **/
2631 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2632                                 __always_unused __be16 proto, u16 vid)
2633 {
2634         struct i40e_netdev_priv *np = netdev_priv(netdev);
2635         struct i40e_vsi *vsi = np->vsi;
2636         int ret = 0;
2637
2638         if (vid >= VLAN_N_VID)
2639                 return -EINVAL;
2640
2641         /* If the network stack called us with vid = 0 then
2642          * it is asking to receive priority tagged packets with
2643          * vlan id 0.  Our HW receives them by default when configured
2644          * to receive untagged packets so there is no need to add an
2645          * extra filter for vlan 0 tagged packets.
2646          */
2647         if (vid)
2648                 ret = i40e_vsi_add_vlan(vsi, vid);
2649
2650         if (!ret)
2651                 set_bit(vid, vsi->active_vlans);
2652
2653         return ret;
2654 }
2655
2656 /**
2657  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2658  * @netdev: network interface to be adjusted
2659  * @vid: vlan id to be removed
2660  *
2661  * net_device_ops implementation for removing vlan ids
2662  **/
2663 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2664                                  __always_unused __be16 proto, u16 vid)
2665 {
2666         struct i40e_netdev_priv *np = netdev_priv(netdev);
2667         struct i40e_vsi *vsi = np->vsi;
2668
2669         /* return code is ignored as there is nothing a user
2670          * can do about failure to remove and a log message was
2671          * already printed from the other function
2672          */
2673         i40e_vsi_kill_vlan(vsi, vid);
2674
2675         clear_bit(vid, vsi->active_vlans);
2676
2677         return 0;
2678 }
2679
2680 /**
2681  * i40e_macaddr_init - explicitly write the mac address filters
2682  *
2683  * @vsi: pointer to the vsi
2684  * @macaddr: the MAC address
2685  *
2686  * This is needed when the macaddr has been obtained by other
2687  * means than the default, e.g., from Open Firmware or IDPROM.
2688  * Returns 0 on success, negative on failure
2689  **/
2690 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2691 {
2692         int ret;
2693         struct i40e_aqc_add_macvlan_element_data element;
2694
2695         ret = i40e_aq_mac_address_write(&vsi->back->hw,
2696                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
2697                                         macaddr, NULL);
2698         if (ret) {
2699                 dev_info(&vsi->back->pdev->dev,
2700                          "Addr change for VSI failed: %d\n", ret);
2701                 return -EADDRNOTAVAIL;
2702         }
2703
2704         memset(&element, 0, sizeof(element));
2705         ether_addr_copy(element.mac_addr, macaddr);
2706         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2707         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2708         if (ret) {
2709                 dev_info(&vsi->back->pdev->dev,
2710                          "add filter failed err %s aq_err %s\n",
2711                          i40e_stat_str(&vsi->back->hw, ret),
2712                          i40e_aq_str(&vsi->back->hw,
2713                                      vsi->back->hw.aq.asq_last_status));
2714         }
2715         return ret;
2716 }
2717
2718 /**
2719  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2720  * @vsi: the vsi being brought back up
2721  **/
2722 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2723 {
2724         u16 vid;
2725
2726         if (!vsi->netdev)
2727                 return;
2728
2729         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2730
2731         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2732                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2733                                      vid);
2734 }
2735
2736 /**
2737  * i40e_vsi_add_pvid - Add pvid for the VSI
2738  * @vsi: the vsi being adjusted
2739  * @vid: the vlan id to set as a PVID
2740  **/
2741 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2742 {
2743         struct i40e_vsi_context ctxt;
2744         i40e_status ret;
2745
2746         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2747         vsi->info.pvid = cpu_to_le16(vid);
2748         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2749                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2750                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2751
2752         ctxt.seid = vsi->seid;
2753         ctxt.info = vsi->info;
2754         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2755         if (ret) {
2756                 dev_info(&vsi->back->pdev->dev,
2757                          "add pvid failed, err %s aq_err %s\n",
2758                          i40e_stat_str(&vsi->back->hw, ret),
2759                          i40e_aq_str(&vsi->back->hw,
2760                                      vsi->back->hw.aq.asq_last_status));
2761                 return -ENOENT;
2762         }
2763
2764         return 0;
2765 }
2766
2767 /**
2768  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2769  * @vsi: the vsi being adjusted
2770  *
2771  * Just use the vlan_rx_register() service to put it back to normal
2772  **/
2773 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2774 {
2775         i40e_vlan_stripping_disable(vsi);
2776
2777         vsi->info.pvid = 0;
2778 }
2779
2780 /**
2781  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2782  * @vsi: ptr to the VSI
2783  *
2784  * If this function returns with an error, then it's possible one or
2785  * more of the rings is populated (while the rest are not).  It is the
2786  * callers duty to clean those orphaned rings.
2787  *
2788  * Return 0 on success, negative on failure
2789  **/
2790 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2791 {
2792         int i, err = 0;
2793
2794         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2795                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2796
2797         return err;
2798 }
2799
2800 /**
2801  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2802  * @vsi: ptr to the VSI
2803  *
2804  * Free VSI's transmit software resources
2805  **/
2806 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2807 {
2808         int i;
2809
2810         if (!vsi->tx_rings)
2811                 return;
2812
2813         for (i = 0; i < vsi->num_queue_pairs; i++)
2814                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2815                         i40e_free_tx_resources(vsi->tx_rings[i]);
2816 }
2817
2818 /**
2819  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2820  * @vsi: ptr to the VSI
2821  *
2822  * If this function returns with an error, then it's possible one or
2823  * more of the rings is populated (while the rest are not).  It is the
2824  * callers duty to clean those orphaned rings.
2825  *
2826  * Return 0 on success, negative on failure
2827  **/
2828 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2829 {
2830         int i, err = 0;
2831
2832         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2833                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2834         return err;
2835 }
2836
2837 /**
2838  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2839  * @vsi: ptr to the VSI
2840  *
2841  * Free all receive software resources
2842  **/
2843 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2844 {
2845         int i;
2846
2847         if (!vsi->rx_rings)
2848                 return;
2849
2850         for (i = 0; i < vsi->num_queue_pairs; i++)
2851                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2852                         i40e_free_rx_resources(vsi->rx_rings[i]);
2853 }
2854
2855 /**
2856  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2857  * @ring: The Tx ring to configure
2858  *
2859  * This enables/disables XPS for a given Tx descriptor ring
2860  * based on the TCs enabled for the VSI that ring belongs to.
2861  **/
2862 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2863 {
2864         struct i40e_vsi *vsi = ring->vsi;
2865         cpumask_var_t mask;
2866
2867         if (!ring->q_vector || !ring->netdev)
2868                 return;
2869
2870         /* Single TC mode enable XPS */
2871         if (vsi->tc_config.numtc <= 1) {
2872                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2873                         netif_set_xps_queue(ring->netdev,
2874                                             &ring->q_vector->affinity_mask,
2875                                             ring->queue_index);
2876         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2877                 /* Disable XPS to allow selection based on TC */
2878                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2879                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2880                 free_cpumask_var(mask);
2881         }
2882
2883         /* schedule our worker thread which will take care of
2884          * applying the new filter changes
2885          */
2886         i40e_service_event_schedule(vsi->back);
2887 }
2888
2889 /**
2890  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2891  * @ring: The Tx ring to configure
2892  *
2893  * Configure the Tx descriptor ring in the HMC context.
2894  **/
2895 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2896 {
2897         struct i40e_vsi *vsi = ring->vsi;
2898         u16 pf_q = vsi->base_queue + ring->queue_index;
2899         struct i40e_hw *hw = &vsi->back->hw;
2900         struct i40e_hmc_obj_txq tx_ctx;
2901         i40e_status err = 0;
2902         u32 qtx_ctl = 0;
2903
2904         /* some ATR related tx ring init */
2905         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2906                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2907                 ring->atr_count = 0;
2908         } else {
2909                 ring->atr_sample_rate = 0;
2910         }
2911
2912         /* configure XPS */
2913         i40e_config_xps_tx_ring(ring);
2914
2915         /* clear the context structure first */
2916         memset(&tx_ctx, 0, sizeof(tx_ctx));
2917
2918         tx_ctx.new_context = 1;
2919         tx_ctx.base = (ring->dma / 128);
2920         tx_ctx.qlen = ring->count;
2921         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2922                                                I40E_FLAG_FD_ATR_ENABLED));
2923         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2924         /* FDIR VSI tx ring can still use RS bit and writebacks */
2925         if (vsi->type != I40E_VSI_FDIR)
2926                 tx_ctx.head_wb_ena = 1;
2927         tx_ctx.head_wb_addr = ring->dma +
2928                               (ring->count * sizeof(struct i40e_tx_desc));
2929
2930         /* As part of VSI creation/update, FW allocates certain
2931          * Tx arbitration queue sets for each TC enabled for
2932          * the VSI. The FW returns the handles to these queue
2933          * sets as part of the response buffer to Add VSI,
2934          * Update VSI, etc. AQ commands. It is expected that
2935          * these queue set handles be associated with the Tx
2936          * queues by the driver as part of the TX queue context
2937          * initialization. This has to be done regardless of
2938          * DCB as by default everything is mapped to TC0.
2939          */
2940         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2941         tx_ctx.rdylist_act = 0;
2942
2943         /* clear the context in the HMC */
2944         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2945         if (err) {
2946                 dev_info(&vsi->back->pdev->dev,
2947                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2948                          ring->queue_index, pf_q, err);
2949                 return -ENOMEM;
2950         }
2951
2952         /* set the context in the HMC */
2953         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2954         if (err) {
2955                 dev_info(&vsi->back->pdev->dev,
2956                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2957                          ring->queue_index, pf_q, err);
2958                 return -ENOMEM;
2959         }
2960
2961         /* Now associate this queue with this PCI function */
2962         if (vsi->type == I40E_VSI_VMDQ2) {
2963                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2964                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2965                            I40E_QTX_CTL_VFVM_INDX_MASK;
2966         } else {
2967                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2968         }
2969
2970         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2971                     I40E_QTX_CTL_PF_INDX_MASK);
2972         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2973         i40e_flush(hw);
2974
2975         /* cache tail off for easier writes later */
2976         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2977
2978         return 0;
2979 }
2980
2981 /**
2982  * i40e_configure_rx_ring - Configure a receive ring context
2983  * @ring: The Rx ring to configure
2984  *
2985  * Configure the Rx descriptor ring in the HMC context.
2986  **/
2987 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2988 {
2989         struct i40e_vsi *vsi = ring->vsi;
2990         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2991         u16 pf_q = vsi->base_queue + ring->queue_index;
2992         struct i40e_hw *hw = &vsi->back->hw;
2993         struct i40e_hmc_obj_rxq rx_ctx;
2994         i40e_status err = 0;
2995
2996         ring->state = 0;
2997
2998         /* clear the context structure first */
2999         memset(&rx_ctx, 0, sizeof(rx_ctx));
3000
3001         ring->rx_buf_len = vsi->rx_buf_len;
3002
3003         rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3004                                     BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3005
3006         rx_ctx.base = (ring->dma / 128);
3007         rx_ctx.qlen = ring->count;
3008
3009         /* use 32 byte descriptors */
3010         rx_ctx.dsize = 1;
3011
3012         /* descriptor type is always zero
3013          * rx_ctx.dtype = 0;
3014          */
3015         rx_ctx.hsplit_0 = 0;
3016
3017         rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3018         if (hw->revision_id == 0)
3019                 rx_ctx.lrxqthresh = 0;
3020         else
3021                 rx_ctx.lrxqthresh = 2;
3022         rx_ctx.crcstrip = 1;
3023         rx_ctx.l2tsel = 1;
3024         /* this controls whether VLAN is stripped from inner headers */
3025         rx_ctx.showiv = 0;
3026         /* set the prefena field to 1 because the manual says to */
3027         rx_ctx.prefena = 1;
3028
3029         /* clear the context in the HMC */
3030         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3031         if (err) {
3032                 dev_info(&vsi->back->pdev->dev,
3033                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3034                          ring->queue_index, pf_q, err);
3035                 return -ENOMEM;
3036         }
3037
3038         /* set the context in the HMC */
3039         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3040         if (err) {
3041                 dev_info(&vsi->back->pdev->dev,
3042                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3043                          ring->queue_index, pf_q, err);
3044                 return -ENOMEM;
3045         }
3046
3047         /* configure Rx buffer alignment */
3048         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3049                 clear_ring_build_skb_enabled(ring);
3050         else
3051                 set_ring_build_skb_enabled(ring);
3052
3053         /* cache tail for quicker writes, and clear the reg before use */
3054         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3055         writel(0, ring->tail);
3056
3057         i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3058
3059         return 0;
3060 }
3061
3062 /**
3063  * i40e_vsi_configure_tx - Configure the VSI for Tx
3064  * @vsi: VSI structure describing this set of rings and resources
3065  *
3066  * Configure the Tx VSI for operation.
3067  **/
3068 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3069 {
3070         int err = 0;
3071         u16 i;
3072
3073         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3074                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3075
3076         return err;
3077 }
3078
3079 /**
3080  * i40e_vsi_configure_rx - Configure the VSI for Rx
3081  * @vsi: the VSI being configured
3082  *
3083  * Configure the Rx VSI for operation.
3084  **/
3085 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3086 {
3087         int err = 0;
3088         u16 i;
3089
3090         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3091                 vsi->max_frame = I40E_MAX_RXBUFFER;
3092                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
3093 #if (PAGE_SIZE < 8192)
3094         } else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3095                    (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3096                 vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3097                 vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3098 #endif
3099         } else {
3100                 vsi->max_frame = I40E_MAX_RXBUFFER;
3101                 vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3102                                                        I40E_RXBUFFER_2048;
3103         }
3104
3105         /* set up individual rings */
3106         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3107                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3108
3109         return err;
3110 }
3111
3112 /**
3113  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3114  * @vsi: ptr to the VSI
3115  **/
3116 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3117 {
3118         struct i40e_ring *tx_ring, *rx_ring;
3119         u16 qoffset, qcount;
3120         int i, n;
3121
3122         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3123                 /* Reset the TC information */
3124                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3125                         rx_ring = vsi->rx_rings[i];
3126                         tx_ring = vsi->tx_rings[i];
3127                         rx_ring->dcb_tc = 0;
3128                         tx_ring->dcb_tc = 0;
3129                 }
3130         }
3131
3132         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3133                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3134                         continue;
3135
3136                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3137                 qcount = vsi->tc_config.tc_info[n].qcount;
3138                 for (i = qoffset; i < (qoffset + qcount); i++) {
3139                         rx_ring = vsi->rx_rings[i];
3140                         tx_ring = vsi->tx_rings[i];
3141                         rx_ring->dcb_tc = n;
3142                         tx_ring->dcb_tc = n;
3143                 }
3144         }
3145 }
3146
3147 /**
3148  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3149  * @vsi: ptr to the VSI
3150  **/
3151 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3152 {
3153         struct i40e_pf *pf = vsi->back;
3154         int err;
3155
3156         if (vsi->netdev)
3157                 i40e_set_rx_mode(vsi->netdev);
3158
3159         if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3160                 err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3161                 if (err) {
3162                         dev_warn(&pf->pdev->dev,
3163                                  "could not set up macaddr; err %d\n", err);
3164                 }
3165         }
3166 }
3167
3168 /**
3169  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3170  * @vsi: Pointer to the targeted VSI
3171  *
3172  * This function replays the hlist on the hw where all the SB Flow Director
3173  * filters were saved.
3174  **/
3175 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3176 {
3177         struct i40e_fdir_filter *filter;
3178         struct i40e_pf *pf = vsi->back;
3179         struct hlist_node *node;
3180
3181         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3182                 return;
3183
3184         /* Reset FDir counters as we're replaying all existing filters */
3185         pf->fd_tcp4_filter_cnt = 0;
3186         pf->fd_udp4_filter_cnt = 0;
3187         pf->fd_sctp4_filter_cnt = 0;
3188         pf->fd_ip4_filter_cnt = 0;
3189
3190         hlist_for_each_entry_safe(filter, node,
3191                                   &pf->fdir_filter_list, fdir_node) {
3192                 i40e_add_del_fdir(vsi, filter, true);
3193         }
3194 }
3195
3196 /**
3197  * i40e_vsi_configure - Set up the VSI for action
3198  * @vsi: the VSI being configured
3199  **/
3200 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3201 {
3202         int err;
3203
3204         i40e_set_vsi_rx_mode(vsi);
3205         i40e_restore_vlan(vsi);
3206         i40e_vsi_config_dcb_rings(vsi);
3207         err = i40e_vsi_configure_tx(vsi);
3208         if (!err)
3209                 err = i40e_vsi_configure_rx(vsi);
3210
3211         return err;
3212 }
3213
3214 /**
3215  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3216  * @vsi: the VSI being configured
3217  **/
3218 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3219 {
3220         struct i40e_pf *pf = vsi->back;
3221         struct i40e_hw *hw = &pf->hw;
3222         u16 vector;
3223         int i, q;
3224         u32 qp;
3225
3226         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3227          * and PFINT_LNKLSTn registers, e.g.:
3228          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3229          */
3230         qp = vsi->base_queue;
3231         vector = vsi->base_vector;
3232         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3233                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3234
3235                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3236                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3237                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3238                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3239                      q_vector->rx.itr);
3240                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3241                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3242                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3243                      q_vector->tx.itr);
3244                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3245                      i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3246
3247                 /* Linked list for the queuepairs assigned to this vector */
3248                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3249                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3250                         u32 val;
3251
3252                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3253                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3254                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3255                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3256                               (I40E_QUEUE_TYPE_TX
3257                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3258
3259                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3260
3261                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3262                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3263                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3264                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3265                               (I40E_QUEUE_TYPE_RX
3266                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3267
3268                         /* Terminate the linked list */
3269                         if (q == (q_vector->num_ringpairs - 1))
3270                                 val |= (I40E_QUEUE_END_OF_LIST
3271                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3272
3273                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3274                         qp++;
3275                 }
3276         }
3277
3278         i40e_flush(hw);
3279 }
3280
3281 /**
3282  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3283  * @hw: ptr to the hardware info
3284  **/
3285 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3286 {
3287         struct i40e_hw *hw = &pf->hw;
3288         u32 val;
3289
3290         /* clear things first */
3291         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3292         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3293
3294         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3295               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3296               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3297               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3298               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3299               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3300               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3301               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3302
3303         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3304                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3305
3306         if (pf->flags & I40E_FLAG_PTP)
3307                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3308
3309         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3310
3311         /* SW_ITR_IDX = 0, but don't change INTENA */
3312         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3313                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3314
3315         /* OTHER_ITR_IDX = 0 */
3316         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3317 }
3318
3319 /**
3320  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3321  * @vsi: the VSI being configured
3322  **/
3323 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3324 {
3325         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3326         struct i40e_pf *pf = vsi->back;
3327         struct i40e_hw *hw = &pf->hw;
3328         u32 val;
3329
3330         /* set the ITR configuration */
3331         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3332         q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3333         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3334         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3335         q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3336         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3337         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3338
3339         i40e_enable_misc_int_causes(pf);
3340
3341         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3342         wr32(hw, I40E_PFINT_LNKLST0, 0);
3343
3344         /* Associate the queue pair to the vector and enable the queue int */
3345         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3346               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3347               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3348
3349         wr32(hw, I40E_QINT_RQCTL(0), val);
3350
3351         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3352               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3353               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3354
3355         wr32(hw, I40E_QINT_TQCTL(0), val);
3356         i40e_flush(hw);
3357 }
3358
3359 /**
3360  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3361  * @pf: board private structure
3362  **/
3363 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3364 {
3365         struct i40e_hw *hw = &pf->hw;
3366
3367         wr32(hw, I40E_PFINT_DYN_CTL0,
3368              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3369         i40e_flush(hw);
3370 }
3371
3372 /**
3373  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3374  * @pf: board private structure
3375  * @clearpba: true when all pending interrupt events should be cleared
3376  **/
3377 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3378 {
3379         struct i40e_hw *hw = &pf->hw;
3380         u32 val;
3381
3382         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3383               (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3384               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3385
3386         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3387         i40e_flush(hw);
3388 }
3389
3390 /**
3391  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3392  * @irq: interrupt number
3393  * @data: pointer to a q_vector
3394  **/
3395 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3396 {
3397         struct i40e_q_vector *q_vector = data;
3398
3399         if (!q_vector->tx.ring && !q_vector->rx.ring)
3400                 return IRQ_HANDLED;
3401
3402         napi_schedule_irqoff(&q_vector->napi);
3403
3404         return IRQ_HANDLED;
3405 }
3406
3407 /**
3408  * i40e_irq_affinity_notify - Callback for affinity changes
3409  * @notify: context as to what irq was changed
3410  * @mask: the new affinity mask
3411  *
3412  * This is a callback function used by the irq_set_affinity_notifier function
3413  * so that we may register to receive changes to the irq affinity masks.
3414  **/
3415 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3416                                      const cpumask_t *mask)
3417 {
3418         struct i40e_q_vector *q_vector =
3419                 container_of(notify, struct i40e_q_vector, affinity_notify);
3420
3421         q_vector->affinity_mask = *mask;
3422 }
3423
3424 /**
3425  * i40e_irq_affinity_release - Callback for affinity notifier release
3426  * @ref: internal core kernel usage
3427  *
3428  * This is a callback function used by the irq_set_affinity_notifier function
3429  * to inform the current notification subscriber that they will no longer
3430  * receive notifications.
3431  **/
3432 static void i40e_irq_affinity_release(struct kref *ref) {}
3433
3434 /**
3435  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3436  * @vsi: the VSI being configured
3437  * @basename: name for the vector
3438  *
3439  * Allocates MSI-X vectors and requests interrupts from the kernel.
3440  **/
3441 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3442 {
3443         int q_vectors = vsi->num_q_vectors;
3444         struct i40e_pf *pf = vsi->back;
3445         int base = vsi->base_vector;
3446         int rx_int_idx = 0;
3447         int tx_int_idx = 0;
3448         int vector, err;
3449         int irq_num;
3450
3451         for (vector = 0; vector < q_vectors; vector++) {
3452                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3453
3454                 irq_num = pf->msix_entries[base + vector].vector;
3455
3456                 if (q_vector->tx.ring && q_vector->rx.ring) {
3457                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3458                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3459                         tx_int_idx++;
3460                 } else if (q_vector->rx.ring) {
3461                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3462                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3463                 } else if (q_vector->tx.ring) {
3464                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3465                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3466                 } else {
3467                         /* skip this unused q_vector */
3468                         continue;
3469                 }
3470                 err = request_irq(irq_num,
3471                                   vsi->irq_handler,
3472                                   0,
3473                                   q_vector->name,
3474                                   q_vector);
3475                 if (err) {
3476                         dev_info(&pf->pdev->dev,
3477                                  "MSIX request_irq failed, error: %d\n", err);
3478                         goto free_queue_irqs;
3479                 }
3480
3481                 /* register for affinity change notifications */
3482                 q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3483                 q_vector->affinity_notify.release = i40e_irq_affinity_release;
3484                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3485                 /* assign the mask for this irq */
3486                 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
3487         }
3488
3489         vsi->irqs_ready = true;
3490         return 0;
3491
3492 free_queue_irqs:
3493         while (vector) {
3494                 vector--;
3495                 irq_num = pf->msix_entries[base + vector].vector;
3496                 irq_set_affinity_notifier(irq_num, NULL);
3497                 irq_set_affinity_hint(irq_num, NULL);
3498                 free_irq(irq_num, &vsi->q_vectors[vector]);
3499         }
3500         return err;
3501 }
3502
3503 /**
3504  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3505  * @vsi: the VSI being un-configured
3506  **/
3507 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3508 {
3509         struct i40e_pf *pf = vsi->back;
3510         struct i40e_hw *hw = &pf->hw;
3511         int base = vsi->base_vector;
3512         int i;
3513
3514         for (i = 0; i < vsi->num_queue_pairs; i++) {
3515                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3516                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3517         }
3518
3519         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3520                 for (i = vsi->base_vector;
3521                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3522                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3523
3524                 i40e_flush(hw);
3525                 for (i = 0; i < vsi->num_q_vectors; i++)
3526                         synchronize_irq(pf->msix_entries[i + base].vector);
3527         } else {
3528                 /* Legacy and MSI mode - this stops all interrupt handling */
3529                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3530                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3531                 i40e_flush(hw);
3532                 synchronize_irq(pf->pdev->irq);
3533         }
3534 }
3535
3536 /**
3537  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3538  * @vsi: the VSI being configured
3539  **/
3540 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3541 {
3542         struct i40e_pf *pf = vsi->back;
3543         int i;
3544
3545         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3546                 for (i = 0; i < vsi->num_q_vectors; i++)
3547                         i40e_irq_dynamic_enable(vsi, i);
3548         } else {
3549                 i40e_irq_dynamic_enable_icr0(pf, true);
3550         }
3551
3552         i40e_flush(&pf->hw);
3553         return 0;
3554 }
3555
3556 /**
3557  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3558  * @pf: board private structure
3559  **/
3560 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3561 {
3562         /* Disable ICR 0 */
3563         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3564         i40e_flush(&pf->hw);
3565 }
3566
3567 /**
3568  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3569  * @irq: interrupt number
3570  * @data: pointer to a q_vector
3571  *
3572  * This is the handler used for all MSI/Legacy interrupts, and deals
3573  * with both queue and non-queue interrupts.  This is also used in
3574  * MSIX mode to handle the non-queue interrupts.
3575  **/
3576 static irqreturn_t i40e_intr(int irq, void *data)
3577 {
3578         struct i40e_pf *pf = (struct i40e_pf *)data;
3579         struct i40e_hw *hw = &pf->hw;
3580         irqreturn_t ret = IRQ_NONE;
3581         u32 icr0, icr0_remaining;
3582         u32 val, ena_mask;
3583
3584         icr0 = rd32(hw, I40E_PFINT_ICR0);
3585         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3586
3587         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3588         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3589                 goto enable_intr;
3590
3591         /* if interrupt but no bits showing, must be SWINT */
3592         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3593             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3594                 pf->sw_int_count++;
3595
3596         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3597             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3598                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3599                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3600                 dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3601         }
3602
3603         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3604         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3605                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3606                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3607
3608                 /* We do not have a way to disarm Queue causes while leaving
3609                  * interrupt enabled for all other causes, ideally
3610                  * interrupt should be disabled while we are in NAPI but
3611                  * this is not a performance path and napi_schedule()
3612                  * can deal with rescheduling.
3613                  */
3614                 if (!test_bit(__I40E_VSI_DOWN, pf->state))
3615                         napi_schedule_irqoff(&q_vector->napi);
3616         }
3617
3618         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3619                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3620                 set_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
3621                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3622         }
3623
3624         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3625                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3626                 set_bit(__I40E_MDD_EVENT_PENDING, pf->state);
3627         }
3628
3629         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3630                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3631                 set_bit(__I40E_VFLR_EVENT_PENDING, pf->state);
3632         }
3633
3634         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3635                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
3636                         set_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
3637                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3638                 val = rd32(hw, I40E_GLGEN_RSTAT);
3639                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3640                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3641                 if (val == I40E_RESET_CORER) {
3642                         pf->corer_count++;
3643                 } else if (val == I40E_RESET_GLOBR) {
3644                         pf->globr_count++;
3645                 } else if (val == I40E_RESET_EMPR) {
3646                         pf->empr_count++;
3647                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state);
3648                 }
3649         }
3650
3651         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3652                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3653                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3654                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3655                          rd32(hw, I40E_PFHMC_ERRORINFO),
3656                          rd32(hw, I40E_PFHMC_ERRORDATA));
3657         }
3658
3659         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3660                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3661
3662                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3663                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3664                         i40e_ptp_tx_hwtstamp(pf);
3665                 }
3666         }
3667
3668         /* If a critical error is pending we have no choice but to reset the
3669          * device.
3670          * Report and mask out any remaining unexpected interrupts.
3671          */
3672         icr0_remaining = icr0 & ena_mask;
3673         if (icr0_remaining) {
3674                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3675                          icr0_remaining);
3676                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3677                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3678                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3679                         dev_info(&pf->pdev->dev, "device will be reset\n");
3680                         set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
3681                         i40e_service_event_schedule(pf);
3682                 }
3683                 ena_mask &= ~icr0_remaining;
3684         }
3685         ret = IRQ_HANDLED;
3686
3687 enable_intr:
3688         /* re-enable interrupt causes */
3689         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3690         if (!test_bit(__I40E_VSI_DOWN, pf->state)) {
3691                 i40e_service_event_schedule(pf);
3692                 i40e_irq_dynamic_enable_icr0(pf, false);
3693         }
3694
3695         return ret;
3696 }
3697
3698 /**
3699  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3700  * @tx_ring:  tx ring to clean
3701  * @budget:   how many cleans we're allowed
3702  *
3703  * Returns true if there's any budget left (e.g. the clean is finished)
3704  **/
3705 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3706 {
3707         struct i40e_vsi *vsi = tx_ring->vsi;
3708         u16 i = tx_ring->next_to_clean;
3709         struct i40e_tx_buffer *tx_buf;
3710         struct i40e_tx_desc *tx_desc;
3711
3712         tx_buf = &tx_ring->tx_bi[i];
3713         tx_desc = I40E_TX_DESC(tx_ring, i);
3714         i -= tx_ring->count;
3715
3716         do {
3717                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3718
3719                 /* if next_to_watch is not set then there is no work pending */
3720                 if (!eop_desc)
3721                         break;
3722
3723                 /* prevent any other reads prior to eop_desc */
3724                 read_barrier_depends();
3725
3726                 /* if the descriptor isn't done, no work yet to do */
3727                 if (!(eop_desc->cmd_type_offset_bsz &
3728                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3729                         break;
3730
3731                 /* clear next_to_watch to prevent false hangs */
3732                 tx_buf->next_to_watch = NULL;
3733
3734                 tx_desc->buffer_addr = 0;
3735                 tx_desc->cmd_type_offset_bsz = 0;
3736                 /* move past filter desc */
3737                 tx_buf++;
3738                 tx_desc++;
3739                 i++;
3740                 if (unlikely(!i)) {
3741                         i -= tx_ring->count;
3742                         tx_buf = tx_ring->tx_bi;
3743                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3744                 }
3745                 /* unmap skb header data */
3746                 dma_unmap_single(tx_ring->dev,
3747                                  dma_unmap_addr(tx_buf, dma),
3748                                  dma_unmap_len(tx_buf, len),
3749                                  DMA_TO_DEVICE);
3750                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3751                         kfree(tx_buf->raw_buf);
3752
3753                 tx_buf->raw_buf = NULL;
3754                 tx_buf->tx_flags = 0;
3755                 tx_buf->next_to_watch = NULL;
3756                 dma_unmap_len_set(tx_buf, len, 0);
3757                 tx_desc->buffer_addr = 0;
3758                 tx_desc->cmd_type_offset_bsz = 0;
3759
3760                 /* move us past the eop_desc for start of next FD desc */
3761                 tx_buf++;
3762                 tx_desc++;
3763                 i++;
3764                 if (unlikely(!i)) {
3765                         i -= tx_ring->count;
3766                         tx_buf = tx_ring->tx_bi;
3767                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3768                 }
3769
3770                 /* update budget accounting */
3771                 budget--;
3772         } while (likely(budget));
3773
3774         i += tx_ring->count;
3775         tx_ring->next_to_clean = i;
3776
3777         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3778                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3779
3780         return budget > 0;
3781 }
3782
3783 /**
3784  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3785  * @irq: interrupt number
3786  * @data: pointer to a q_vector
3787  **/
3788 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3789 {
3790         struct i40e_q_vector *q_vector = data;
3791         struct i40e_vsi *vsi;
3792
3793         if (!q_vector->tx.ring)
3794                 return IRQ_HANDLED;
3795
3796         vsi = q_vector->tx.ring->vsi;
3797         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3798
3799         return IRQ_HANDLED;
3800 }
3801
3802 /**
3803  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3804  * @vsi: the VSI being configured
3805  * @v_idx: vector index
3806  * @qp_idx: queue pair index
3807  **/
3808 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3809 {
3810         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3811         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3812         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3813
3814         tx_ring->q_vector = q_vector;
3815         tx_ring->next = q_vector->tx.ring;
3816         q_vector->tx.ring = tx_ring;
3817         q_vector->tx.count++;
3818
3819         rx_ring->q_vector = q_vector;
3820         rx_ring->next = q_vector->rx.ring;
3821         q_vector->rx.ring = rx_ring;
3822         q_vector->rx.count++;
3823 }
3824
3825 /**
3826  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3827  * @vsi: the VSI being configured
3828  *
3829  * This function maps descriptor rings to the queue-specific vectors
3830  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3831  * one vector per queue pair, but on a constrained vector budget, we
3832  * group the queue pairs as "efficiently" as possible.
3833  **/
3834 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3835 {
3836         int qp_remaining = vsi->num_queue_pairs;
3837         int q_vectors = vsi->num_q_vectors;
3838         int num_ringpairs;
3839         int v_start = 0;
3840         int qp_idx = 0;
3841
3842         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3843          * group them so there are multiple queues per vector.
3844          * It is also important to go through all the vectors available to be
3845          * sure that if we don't use all the vectors, that the remaining vectors
3846          * are cleared. This is especially important when decreasing the
3847          * number of queues in use.
3848          */
3849         for (; v_start < q_vectors; v_start++) {
3850                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3851
3852                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3853
3854                 q_vector->num_ringpairs = num_ringpairs;
3855
3856                 q_vector->rx.count = 0;
3857                 q_vector->tx.count = 0;
3858                 q_vector->rx.ring = NULL;
3859                 q_vector->tx.ring = NULL;
3860
3861                 while (num_ringpairs--) {
3862                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3863                         qp_idx++;
3864                         qp_remaining--;
3865                 }
3866         }
3867 }
3868
3869 /**
3870  * i40e_vsi_request_irq - Request IRQ from the OS
3871  * @vsi: the VSI being configured
3872  * @basename: name for the vector
3873  **/
3874 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3875 {
3876         struct i40e_pf *pf = vsi->back;
3877         int err;
3878
3879         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3880                 err = i40e_vsi_request_irq_msix(vsi, basename);
3881         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3882                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3883                                   pf->int_name, pf);
3884         else
3885                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3886                                   pf->int_name, pf);
3887
3888         if (err)
3889                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3890
3891         return err;
3892 }
3893
3894 #ifdef CONFIG_NET_POLL_CONTROLLER
3895 /**
3896  * i40e_netpoll - A Polling 'interrupt' handler
3897  * @netdev: network interface device structure
3898  *
3899  * This is used by netconsole to send skbs without having to re-enable
3900  * interrupts.  It's not called while the normal interrupt routine is executing.
3901  **/
3902 static void i40e_netpoll(struct net_device *netdev)
3903 {
3904         struct i40e_netdev_priv *np = netdev_priv(netdev);
3905         struct i40e_vsi *vsi = np->vsi;
3906         struct i40e_pf *pf = vsi->back;
3907         int i;
3908
3909         /* if interface is down do nothing */
3910         if (test_bit(__I40E_VSI_DOWN, vsi->state))
3911                 return;
3912
3913         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3914                 for (i = 0; i < vsi->num_q_vectors; i++)
3915                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3916         } else {
3917                 i40e_intr(pf->pdev->irq, netdev);
3918         }
3919 }
3920 #endif
3921
3922 #define I40E_QTX_ENA_WAIT_COUNT 50
3923
3924 /**
3925  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3926  * @pf: the PF being configured
3927  * @pf_q: the PF queue
3928  * @enable: enable or disable state of the queue
3929  *
3930  * This routine will wait for the given Tx queue of the PF to reach the
3931  * enabled or disabled state.
3932  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3933  * multiple retries; else will return 0 in case of success.
3934  **/
3935 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3936 {
3937         int i;
3938         u32 tx_reg;
3939
3940         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3941                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3942                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3943                         break;
3944
3945                 usleep_range(10, 20);
3946         }
3947         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3948                 return -ETIMEDOUT;
3949
3950         return 0;
3951 }
3952
3953 /**
3954  * i40e_control_tx_q - Start or stop a particular Tx queue
3955  * @pf: the PF structure
3956  * @pf_q: the PF queue to configure
3957  * @enable: start or stop the queue
3958  *
3959  * This function enables or disables a single queue. Note that any delay
3960  * required after the operation is expected to be handled by the caller of
3961  * this function.
3962  **/
3963 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
3964 {
3965         struct i40e_hw *hw = &pf->hw;
3966         u32 tx_reg;
3967         int i;
3968
3969         /* warn the TX unit of coming changes */
3970         i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3971         if (!enable)
3972                 usleep_range(10, 20);
3973
3974         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
3975                 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3976                 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3977                     ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3978                         break;
3979                 usleep_range(1000, 2000);
3980         }
3981
3982         /* Skip if the queue is already in the requested state */
3983         if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3984                 return;
3985
3986         /* turn on/off the queue */
3987         if (enable) {
3988                 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3989                 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3990         } else {
3991                 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3992         }
3993
3994         wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3995 }
3996
3997 /**
3998  * i40e_vsi_control_tx - Start or stop a VSI's rings
3999  * @vsi: the VSI being configured
4000  * @enable: start or stop the rings
4001  **/
4002 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4003 {
4004         struct i40e_pf *pf = vsi->back;
4005         int i, pf_q, ret = 0;
4006
4007         pf_q = vsi->base_queue;
4008         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4009                 i40e_control_tx_q(pf, pf_q, enable);
4010
4011                 /* wait for the change to finish */
4012                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
4013                 if (ret) {
4014                         dev_info(&pf->pdev->dev,
4015                                  "VSI seid %d Tx ring %d %sable timeout\n",
4016                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
4017                         break;
4018                 }
4019         }
4020
4021         return ret;
4022 }
4023
4024 /**
4025  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4026  * @pf: the PF being configured
4027  * @pf_q: the PF queue
4028  * @enable: enable or disable state of the queue
4029  *
4030  * This routine will wait for the given Rx queue of the PF to reach the
4031  * enabled or disabled state.
4032  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4033  * multiple retries; else will return 0 in case of success.
4034  **/
4035 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4036 {
4037         int i;
4038         u32 rx_reg;
4039
4040         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4041                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4042                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4043                         break;
4044
4045                 usleep_range(10, 20);
4046         }
4047         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4048                 return -ETIMEDOUT;
4049
4050         return 0;
4051 }
4052
4053 /**
4054  * i40e_control_rx_q - Start or stop a particular Rx queue
4055  * @pf: the PF structure
4056  * @pf_q: the PF queue to configure
4057  * @enable: start or stop the queue
4058  *
4059  * This function enables or disables a single queue. Note that any delay
4060  * required after the operation is expected to be handled by the caller of
4061  * this function.
4062  **/
4063 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4064 {
4065         struct i40e_hw *hw = &pf->hw;
4066         u32 rx_reg;
4067         int i;
4068
4069         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4070                 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4071                 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4072                     ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4073                         break;
4074                 usleep_range(1000, 2000);
4075         }
4076
4077         /* Skip if the queue is already in the requested state */
4078         if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4079                 return;
4080
4081         /* turn on/off the queue */
4082         if (enable)
4083                 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4084         else
4085                 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4086
4087         wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4088 }
4089
4090 /**
4091  * i40e_vsi_control_rx - Start or stop a VSI's rings
4092  * @vsi: the VSI being configured
4093  * @enable: start or stop the rings
4094  **/
4095 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4096 {
4097         struct i40e_pf *pf = vsi->back;
4098         int i, pf_q, ret = 0;
4099
4100         pf_q = vsi->base_queue;
4101         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4102                 i40e_control_rx_q(pf, pf_q, enable);
4103
4104                 /* wait for the change to finish */
4105                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4106                 if (ret) {
4107                         dev_info(&pf->pdev->dev,
4108                                  "VSI seid %d Rx ring %d %sable timeout\n",
4109                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
4110                         break;
4111                 }
4112         }
4113
4114         /* Due to HW errata, on Rx disable only, the register can indicate done
4115          * before it really is. Needs 50ms to be sure
4116          */
4117         if (!enable)
4118                 mdelay(50);
4119
4120         return ret;
4121 }
4122
4123 /**
4124  * i40e_vsi_start_rings - Start a VSI's rings
4125  * @vsi: the VSI being configured
4126  **/
4127 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4128 {
4129         int ret = 0;
4130
4131         /* do rx first for enable and last for disable */
4132         ret = i40e_vsi_control_rx(vsi, true);
4133         if (ret)
4134                 return ret;
4135         ret = i40e_vsi_control_tx(vsi, true);
4136
4137         return ret;
4138 }
4139
4140 /**
4141  * i40e_vsi_stop_rings - Stop a VSI's rings
4142  * @vsi: the VSI being configured
4143  **/
4144 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4145 {
4146         /* When port TX is suspended, don't wait */
4147         if (test_bit(__I40E_PORT_SUSPENDED, vsi->back->state))
4148                 return i40e_vsi_stop_rings_no_wait(vsi);
4149
4150         /* do rx first for enable and last for disable
4151          * Ignore return value, we need to shutdown whatever we can
4152          */
4153         i40e_vsi_control_tx(vsi, false);
4154         i40e_vsi_control_rx(vsi, false);
4155 }
4156
4157 /**
4158  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4159  * @vsi: the VSI being shutdown
4160  *
4161  * This function stops all the rings for a VSI but does not delay to verify
4162  * that rings have been disabled. It is expected that the caller is shutting
4163  * down multiple VSIs at once and will delay together for all the VSIs after
4164  * initiating the shutdown. This is particularly useful for shutting down lots
4165  * of VFs together. Otherwise, a large delay can be incurred while configuring
4166  * each VSI in serial.
4167  **/
4168 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4169 {
4170         struct i40e_pf *pf = vsi->back;
4171         int i, pf_q;
4172
4173         pf_q = vsi->base_queue;
4174         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4175                 i40e_control_tx_q(pf, pf_q, false);
4176                 i40e_control_rx_q(pf, pf_q, false);
4177         }
4178 }
4179
4180 /**
4181  * i40e_vsi_free_irq - Free the irq association with the OS
4182  * @vsi: the VSI being configured
4183  **/
4184 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4185 {
4186         struct i40e_pf *pf = vsi->back;
4187         struct i40e_hw *hw = &pf->hw;
4188         int base = vsi->base_vector;
4189         u32 val, qp;
4190         int i;
4191
4192         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4193                 if (!vsi->q_vectors)
4194                         return;
4195
4196                 if (!vsi->irqs_ready)
4197                         return;
4198
4199                 vsi->irqs_ready = false;
4200                 for (i = 0; i < vsi->num_q_vectors; i++) {
4201                         int irq_num;
4202                         u16 vector;
4203
4204                         vector = i + base;
4205                         irq_num = pf->msix_entries[vector].vector;
4206
4207                         /* free only the irqs that were actually requested */
4208                         if (!vsi->q_vectors[i] ||
4209                             !vsi->q_vectors[i]->num_ringpairs)
4210                                 continue;
4211
4212                         /* clear the affinity notifier in the IRQ descriptor */
4213                         irq_set_affinity_notifier(irq_num, NULL);
4214                         /* clear the affinity_mask in the IRQ descriptor */
4215                         irq_set_affinity_hint(irq_num, NULL);
4216                         synchronize_irq(irq_num);
4217                         free_irq(irq_num, vsi->q_vectors[i]);
4218
4219                         /* Tear down the interrupt queue link list
4220                          *
4221                          * We know that they come in pairs and always
4222                          * the Rx first, then the Tx.  To clear the
4223                          * link list, stick the EOL value into the
4224                          * next_q field of the registers.
4225                          */
4226                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4227                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4228                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4229                         val |= I40E_QUEUE_END_OF_LIST
4230                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4231                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4232
4233                         while (qp != I40E_QUEUE_END_OF_LIST) {
4234                                 u32 next;
4235
4236                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4237
4238                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4239                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4240                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4241                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4242
4243                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4244                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4245
4246                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4247
4248                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4249
4250                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4251                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4252
4253                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4254                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4255                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4256                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4257
4258                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4259                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4260
4261                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4262                                 qp = next;
4263                         }
4264                 }
4265         } else {
4266                 free_irq(pf->pdev->irq, pf);
4267
4268                 val = rd32(hw, I40E_PFINT_LNKLST0);
4269                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4270                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4271                 val |= I40E_QUEUE_END_OF_LIST
4272                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4273                 wr32(hw, I40E_PFINT_LNKLST0, val);
4274
4275                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4276                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4277                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4278                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4279                          I40E_QINT_RQCTL_INTEVENT_MASK);
4280
4281                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4282                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4283
4284                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4285
4286                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4287
4288                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4289                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4290                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4291                          I40E_QINT_TQCTL_INTEVENT_MASK);
4292
4293                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4294                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4295
4296                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4297         }
4298 }
4299
4300 /**
4301  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4302  * @vsi: the VSI being configured
4303  * @v_idx: Index of vector to be freed
4304  *
4305  * This function frees the memory allocated to the q_vector.  In addition if
4306  * NAPI is enabled it will delete any references to the NAPI struct prior
4307  * to freeing the q_vector.
4308  **/
4309 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4310 {
4311         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4312         struct i40e_ring *ring;
4313
4314         if (!q_vector)
4315                 return;
4316
4317         /* disassociate q_vector from rings */
4318         i40e_for_each_ring(ring, q_vector->tx)
4319                 ring->q_vector = NULL;
4320
4321         i40e_for_each_ring(ring, q_vector->rx)
4322                 ring->q_vector = NULL;
4323
4324         /* only VSI w/ an associated netdev is set up w/ NAPI */
4325         if (vsi->netdev)
4326                 netif_napi_del(&q_vector->napi);
4327
4328         vsi->q_vectors[v_idx] = NULL;
4329
4330         kfree_rcu(q_vector, rcu);
4331 }
4332
4333 /**
4334  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4335  * @vsi: the VSI being un-configured
4336  *
4337  * This frees the memory allocated to the q_vectors and
4338  * deletes references to the NAPI struct.
4339  **/
4340 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4341 {
4342         int v_idx;
4343
4344         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4345                 i40e_free_q_vector(vsi, v_idx);
4346 }
4347
4348 /**
4349  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4350  * @pf: board private structure
4351  **/
4352 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4353 {
4354         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4355         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4356                 pci_disable_msix(pf->pdev);
4357                 kfree(pf->msix_entries);
4358                 pf->msix_entries = NULL;
4359                 kfree(pf->irq_pile);
4360                 pf->irq_pile = NULL;
4361         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4362                 pci_disable_msi(pf->pdev);
4363         }
4364         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4365 }
4366
4367 /**
4368  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4369  * @pf: board private structure
4370  *
4371  * We go through and clear interrupt specific resources and reset the structure
4372  * to pre-load conditions
4373  **/
4374 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4375 {
4376         int i;
4377
4378         i40e_stop_misc_vector(pf);
4379         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4380                 synchronize_irq(pf->msix_entries[0].vector);
4381                 free_irq(pf->msix_entries[0].vector, pf);
4382         }
4383
4384         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4385                       I40E_IWARP_IRQ_PILE_ID);
4386
4387         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4388         for (i = 0; i < pf->num_alloc_vsi; i++)
4389                 if (pf->vsi[i])
4390                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4391         i40e_reset_interrupt_capability(pf);
4392 }
4393
4394 /**
4395  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4396  * @vsi: the VSI being configured
4397  **/
4398 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4399 {
4400         int q_idx;
4401
4402         if (!vsi->netdev)
4403                 return;
4404
4405         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4406                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4407
4408                 if (q_vector->rx.ring || q_vector->tx.ring)
4409                         napi_enable(&q_vector->napi);
4410         }
4411 }
4412
4413 /**
4414  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4415  * @vsi: the VSI being configured
4416  **/
4417 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4418 {
4419         int q_idx;
4420
4421         if (!vsi->netdev)
4422                 return;
4423
4424         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4425                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4426
4427                 if (q_vector->rx.ring || q_vector->tx.ring)
4428                         napi_disable(&q_vector->napi);
4429         }
4430 }
4431
4432 /**
4433  * i40e_vsi_close - Shut down a VSI
4434  * @vsi: the vsi to be quelled
4435  **/
4436 static void i40e_vsi_close(struct i40e_vsi *vsi)
4437 {
4438         struct i40e_pf *pf = vsi->back;
4439         if (!test_and_set_bit(__I40E_VSI_DOWN, vsi->state))
4440                 i40e_down(vsi);
4441         i40e_vsi_free_irq(vsi);
4442         i40e_vsi_free_tx_resources(vsi);
4443         i40e_vsi_free_rx_resources(vsi);
4444         vsi->current_netdev_flags = 0;
4445         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4446         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
4447                 pf->flags |=  I40E_FLAG_CLIENT_RESET;
4448 }
4449
4450 /**
4451  * i40e_quiesce_vsi - Pause a given VSI
4452  * @vsi: the VSI being paused
4453  **/
4454 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4455 {
4456         if (test_bit(__I40E_VSI_DOWN, vsi->state))
4457                 return;
4458
4459         set_bit(__I40E_VSI_NEEDS_RESTART, vsi->state);
4460         if (vsi->netdev && netif_running(vsi->netdev))
4461                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4462         else
4463                 i40e_vsi_close(vsi);
4464 }
4465
4466 /**
4467  * i40e_unquiesce_vsi - Resume a given VSI
4468  * @vsi: the VSI being resumed
4469  **/
4470 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4471 {
4472         if (!test_and_clear_bit(__I40E_VSI_NEEDS_RESTART, vsi->state))
4473                 return;
4474
4475         if (vsi->netdev && netif_running(vsi->netdev))
4476                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4477         else
4478                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4479 }
4480
4481 /**
4482  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4483  * @pf: the PF
4484  **/
4485 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4486 {
4487         int v;
4488
4489         for (v = 0; v < pf->num_alloc_vsi; v++) {
4490                 if (pf->vsi[v])
4491                         i40e_quiesce_vsi(pf->vsi[v]);
4492         }
4493 }
4494
4495 /**
4496  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4497  * @pf: the PF
4498  **/
4499 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4500 {
4501         int v;
4502
4503         for (v = 0; v < pf->num_alloc_vsi; v++) {
4504                 if (pf->vsi[v])
4505                         i40e_unquiesce_vsi(pf->vsi[v]);
4506         }
4507 }
4508
4509 /**
4510  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4511  * @vsi: the VSI being configured
4512  *
4513  * Wait until all queues on a given VSI have been disabled.
4514  **/
4515 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4516 {
4517         struct i40e_pf *pf = vsi->back;
4518         int i, pf_q, ret;
4519
4520         pf_q = vsi->base_queue;
4521         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4522                 /* Check and wait for the Tx queue */
4523                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4524                 if (ret) {
4525                         dev_info(&pf->pdev->dev,
4526                                  "VSI seid %d Tx ring %d disable timeout\n",
4527                                  vsi->seid, pf_q);
4528                         return ret;
4529                 }
4530                 /* Check and wait for the Tx queue */
4531                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4532                 if (ret) {
4533                         dev_info(&pf->pdev->dev,
4534                                  "VSI seid %d Rx ring %d disable timeout\n",
4535                                  vsi->seid, pf_q);
4536                         return ret;
4537                 }
4538         }
4539
4540         return 0;
4541 }
4542
4543 #ifdef CONFIG_I40E_DCB
4544 /**
4545  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4546  * @pf: the PF
4547  *
4548  * This function waits for the queues to be in disabled state for all the
4549  * VSIs that are managed by this PF.
4550  **/
4551 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4552 {
4553         int v, ret = 0;
4554
4555         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4556                 if (pf->vsi[v]) {
4557                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4558                         if (ret)
4559                                 break;
4560                 }
4561         }
4562
4563         return ret;
4564 }
4565
4566 #endif
4567
4568 /**
4569  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4570  * @q_idx: TX queue number
4571  * @vsi: Pointer to VSI struct
4572  *
4573  * This function checks specified queue for given VSI. Detects hung condition.
4574  * We proactively detect hung TX queues by checking if interrupts are disabled
4575  * but there are pending descriptors.  If it appears hung, attempt to recover
4576  * by triggering a SW interrupt.
4577  **/
4578 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4579 {
4580         struct i40e_ring *tx_ring = NULL;
4581         struct i40e_pf  *pf;
4582         u32 val, tx_pending;
4583         int i;
4584
4585         pf = vsi->back;
4586
4587         /* now that we have an index, find the tx_ring struct */
4588         for (i = 0; i < vsi->num_queue_pairs; i++) {
4589                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4590                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4591                                 tx_ring = vsi->tx_rings[i];
4592                                 break;
4593                         }
4594                 }
4595         }
4596
4597         if (!tx_ring)
4598                 return;
4599
4600         /* Read interrupt register */
4601         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4602                 val = rd32(&pf->hw,
4603                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4604                                                tx_ring->vsi->base_vector - 1));
4605         else
4606                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4607
4608         tx_pending = i40e_get_tx_pending(tx_ring);
4609
4610         /* Interrupts are disabled and TX pending is non-zero,
4611          * trigger the SW interrupt (don't wait). Worst case
4612          * there will be one extra interrupt which may result
4613          * into not cleaning any queues because queues are cleaned.
4614          */
4615         if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4616                 i40e_force_wb(vsi, tx_ring->q_vector);
4617 }
4618
4619 /**
4620  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4621  * @pf:  pointer to PF struct
4622  *
4623  * LAN VSI has netdev and netdev has TX queues. This function is to check
4624  * each of those TX queues if they are hung, trigger recovery by issuing
4625  * SW interrupt.
4626  **/
4627 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4628 {
4629         struct net_device *netdev;
4630         struct i40e_vsi *vsi;
4631         int i;
4632
4633         /* Only for LAN VSI */
4634         vsi = pf->vsi[pf->lan_vsi];
4635
4636         if (!vsi)
4637                 return;
4638
4639         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4640         if (test_bit(__I40E_VSI_DOWN, vsi->back->state) ||
4641             test_bit(__I40E_RESET_RECOVERY_PENDING, vsi->back->state))
4642                 return;
4643
4644         /* Make sure type is MAIN VSI */
4645         if (vsi->type != I40E_VSI_MAIN)
4646                 return;
4647
4648         netdev = vsi->netdev;
4649         if (!netdev)
4650                 return;
4651
4652         /* Bail out if netif_carrier is not OK */
4653         if (!netif_carrier_ok(netdev))
4654                 return;
4655
4656         /* Go thru' TX queues for netdev */
4657         for (i = 0; i < netdev->num_tx_queues; i++) {
4658                 struct netdev_queue *q;
4659
4660                 q = netdev_get_tx_queue(netdev, i);
4661                 if (q)
4662                         i40e_detect_recover_hung_queue(i, vsi);
4663         }
4664 }
4665
4666 /**
4667  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4668  * @pf: pointer to PF
4669  *
4670  * Get TC map for ISCSI PF type that will include iSCSI TC
4671  * and LAN TC.
4672  **/
4673 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4674 {
4675         struct i40e_dcb_app_priority_table app;
4676         struct i40e_hw *hw = &pf->hw;
4677         u8 enabled_tc = 1; /* TC0 is always enabled */
4678         u8 tc, i;
4679         /* Get the iSCSI APP TLV */
4680         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4681
4682         for (i = 0; i < dcbcfg->numapps; i++) {
4683                 app = dcbcfg->app[i];
4684                 if (app.selector == I40E_APP_SEL_TCPIP &&
4685                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4686                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4687                         enabled_tc |= BIT(tc);
4688                         break;
4689                 }
4690         }
4691
4692         return enabled_tc;
4693 }
4694
4695 /**
4696  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4697  * @dcbcfg: the corresponding DCBx configuration structure
4698  *
4699  * Return the number of TCs from given DCBx configuration
4700  **/
4701 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4702 {
4703         int i, tc_unused = 0;
4704         u8 num_tc = 0;
4705         u8 ret = 0;
4706
4707         /* Scan the ETS Config Priority Table to find
4708          * traffic class enabled for a given priority
4709          * and create a bitmask of enabled TCs
4710          */
4711         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4712                 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4713
4714         /* Now scan the bitmask to check for
4715          * contiguous TCs starting with TC0
4716          */
4717         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4718                 if (num_tc & BIT(i)) {
4719                         if (!tc_unused) {
4720                                 ret++;
4721                         } else {
4722                                 pr_err("Non-contiguous TC - Disabling DCB\n");
4723                                 return 1;
4724                         }
4725                 } else {
4726                         tc_unused = 1;
4727                 }
4728         }
4729
4730         /* There is always at least TC0 */
4731         if (!ret)
4732                 ret = 1;
4733
4734         return ret;
4735 }
4736
4737 /**
4738  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4739  * @dcbcfg: the corresponding DCBx configuration structure
4740  *
4741  * Query the current DCB configuration and return the number of
4742  * traffic classes enabled from the given DCBX config
4743  **/
4744 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4745 {
4746         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4747         u8 enabled_tc = 1;
4748         u8 i;
4749
4750         for (i = 0; i < num_tc; i++)
4751                 enabled_tc |= BIT(i);
4752
4753         return enabled_tc;
4754 }
4755
4756 /**
4757  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4758  * @pf: PF being queried
4759  *
4760  * Return number of traffic classes enabled for the given PF
4761  **/
4762 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4763 {
4764         struct i40e_hw *hw = &pf->hw;
4765         u8 i, enabled_tc = 1;
4766         u8 num_tc = 0;
4767         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4768
4769         /* If DCB is not enabled then always in single TC */
4770         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4771                 return 1;
4772
4773         /* SFP mode will be enabled for all TCs on port */
4774         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4775                 return i40e_dcb_get_num_tc(dcbcfg);
4776
4777         /* MFP mode return count of enabled TCs for this PF */
4778         if (pf->hw.func_caps.iscsi)
4779                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4780         else
4781                 return 1; /* Only TC0 */
4782
4783         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4784                 if (enabled_tc & BIT(i))
4785                         num_tc++;
4786         }
4787         return num_tc;
4788 }
4789
4790 /**
4791  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4792  * @pf: PF being queried
4793  *
4794  * Return a bitmap for enabled traffic classes for this PF.
4795  **/
4796 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4797 {
4798         /* If DCB is not enabled for this PF then just return default TC */
4799         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4800                 return I40E_DEFAULT_TRAFFIC_CLASS;
4801
4802         /* SFP mode we want PF to be enabled for all TCs */
4803         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4804                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4805
4806         /* MFP enabled and iSCSI PF type */
4807         if (pf->hw.func_caps.iscsi)
4808                 return i40e_get_iscsi_tc_map(pf);
4809         else
4810                 return I40E_DEFAULT_TRAFFIC_CLASS;
4811 }
4812
4813 /**
4814  * i40e_vsi_get_bw_info - Query VSI BW Information
4815  * @vsi: the VSI being queried
4816  *
4817  * Returns 0 on success, negative value on failure
4818  **/
4819 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4820 {
4821         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4822         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4823         struct i40e_pf *pf = vsi->back;
4824         struct i40e_hw *hw = &pf->hw;
4825         i40e_status ret;
4826         u32 tc_bw_max;
4827         int i;
4828
4829         /* Get the VSI level BW configuration */
4830         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4831         if (ret) {
4832                 dev_info(&pf->pdev->dev,
4833                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4834                          i40e_stat_str(&pf->hw, ret),
4835                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4836                 return -EINVAL;
4837         }
4838
4839         /* Get the VSI level BW configuration per TC */
4840         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4841                                                NULL);
4842         if (ret) {
4843                 dev_info(&pf->pdev->dev,
4844                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4845                          i40e_stat_str(&pf->hw, ret),
4846                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4847                 return -EINVAL;
4848         }
4849
4850         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4851                 dev_info(&pf->pdev->dev,
4852                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4853                          bw_config.tc_valid_bits,
4854                          bw_ets_config.tc_valid_bits);
4855                 /* Still continuing */
4856         }
4857
4858         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4859         vsi->bw_max_quanta = bw_config.max_bw;
4860         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4861                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4862         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4863                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4864                 vsi->bw_ets_limit_credits[i] =
4865                                         le16_to_cpu(bw_ets_config.credits[i]);
4866                 /* 3 bits out of 4 for each TC */
4867                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4868         }
4869
4870         return 0;
4871 }
4872
4873 /**
4874  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4875  * @vsi: the VSI being configured
4876  * @enabled_tc: TC bitmap
4877  * @bw_credits: BW shared credits per TC
4878  *
4879  * Returns 0 on success, negative value on failure
4880  **/
4881 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4882                                        u8 *bw_share)
4883 {
4884         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4885         i40e_status ret;
4886         int i;
4887
4888         bw_data.tc_valid_bits = enabled_tc;
4889         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4890                 bw_data.tc_bw_credits[i] = bw_share[i];
4891
4892         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4893                                        NULL);
4894         if (ret) {
4895                 dev_info(&vsi->back->pdev->dev,
4896                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4897                          vsi->back->hw.aq.asq_last_status);
4898                 return -EINVAL;
4899         }
4900
4901         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4902                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4903
4904         return 0;
4905 }
4906
4907 /**
4908  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4909  * @vsi: the VSI being configured
4910  * @enabled_tc: TC map to be enabled
4911  *
4912  **/
4913 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4914 {
4915         struct net_device *netdev = vsi->netdev;
4916         struct i40e_pf *pf = vsi->back;
4917         struct i40e_hw *hw = &pf->hw;
4918         u8 netdev_tc = 0;
4919         int i;
4920         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4921
4922         if (!netdev)
4923                 return;
4924
4925         if (!enabled_tc) {
4926                 netdev_reset_tc(netdev);
4927                 return;
4928         }
4929
4930         /* Set up actual enabled TCs on the VSI */
4931         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4932                 return;
4933
4934         /* set per TC queues for the VSI */
4935         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4936                 /* Only set TC queues for enabled tcs
4937                  *
4938                  * e.g. For a VSI that has TC0 and TC3 enabled the
4939                  * enabled_tc bitmap would be 0x00001001; the driver
4940                  * will set the numtc for netdev as 2 that will be
4941                  * referenced by the netdev layer as TC 0 and 1.
4942                  */
4943                 if (vsi->tc_config.enabled_tc & BIT(i))
4944                         netdev_set_tc_queue(netdev,
4945                                         vsi->tc_config.tc_info[i].netdev_tc,
4946                                         vsi->tc_config.tc_info[i].qcount,
4947                                         vsi->tc_config.tc_info[i].qoffset);
4948         }
4949
4950         /* Assign UP2TC map for the VSI */
4951         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4952                 /* Get the actual TC# for the UP */
4953                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4954                 /* Get the mapped netdev TC# for the UP */
4955                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4956                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4957         }
4958 }
4959
4960 /**
4961  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4962  * @vsi: the VSI being configured
4963  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4964  **/
4965 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4966                                       struct i40e_vsi_context *ctxt)
4967 {
4968         /* copy just the sections touched not the entire info
4969          * since not all sections are valid as returned by
4970          * update vsi params
4971          */
4972         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4973         memcpy(&vsi->info.queue_mapping,
4974                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4975         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4976                sizeof(vsi->info.tc_mapping));
4977 }
4978
4979 /**
4980  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4981  * @vsi: VSI to be configured
4982  * @enabled_tc: TC bitmap
4983  *
4984  * This configures a particular VSI for TCs that are mapped to the
4985  * given TC bitmap. It uses default bandwidth share for TCs across
4986  * VSIs to configure TC for a particular VSI.
4987  *
4988  * NOTE:
4989  * It is expected that the VSI queues have been quisced before calling
4990  * this function.
4991  **/
4992 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4993 {
4994         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4995         struct i40e_vsi_context ctxt;
4996         int ret = 0;
4997         int i;
4998
4999         /* Check if enabled_tc is same as existing or new TCs */
5000         if (vsi->tc_config.enabled_tc == enabled_tc)
5001                 return ret;
5002
5003         /* Enable ETS TCs with equal BW Share for now across all VSIs */
5004         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5005                 if (enabled_tc & BIT(i))
5006                         bw_share[i] = 1;
5007         }
5008
5009         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5010         if (ret) {
5011                 dev_info(&vsi->back->pdev->dev,
5012                          "Failed configuring TC map %d for VSI %d\n",
5013                          enabled_tc, vsi->seid);
5014                 goto out;
5015         }
5016
5017         /* Update Queue Pairs Mapping for currently enabled UPs */
5018         ctxt.seid = vsi->seid;
5019         ctxt.pf_num = vsi->back->hw.pf_id;
5020         ctxt.vf_num = 0;
5021         ctxt.uplink_seid = vsi->uplink_seid;
5022         ctxt.info = vsi->info;
5023         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5024
5025         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5026                 ctxt.info.valid_sections |=
5027                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5028                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5029         }
5030
5031         /* Update the VSI after updating the VSI queue-mapping information */
5032         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5033         if (ret) {
5034                 dev_info(&vsi->back->pdev->dev,
5035                          "Update vsi tc config failed, err %s aq_err %s\n",
5036                          i40e_stat_str(&vsi->back->hw, ret),
5037                          i40e_aq_str(&vsi->back->hw,
5038                                      vsi->back->hw.aq.asq_last_status));
5039                 goto out;
5040         }
5041         /* update the local VSI info with updated queue map */
5042         i40e_vsi_update_queue_map(vsi, &ctxt);
5043         vsi->info.valid_sections = 0;
5044
5045         /* Update current VSI BW information */
5046         ret = i40e_vsi_get_bw_info(vsi);
5047         if (ret) {
5048                 dev_info(&vsi->back->pdev->dev,
5049                          "Failed updating vsi bw info, err %s aq_err %s\n",
5050                          i40e_stat_str(&vsi->back->hw, ret),
5051                          i40e_aq_str(&vsi->back->hw,
5052                                      vsi->back->hw.aq.asq_last_status));
5053                 goto out;
5054         }
5055
5056         /* Update the netdev TC setup */
5057         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5058 out:
5059         return ret;
5060 }
5061
5062 /**
5063  * i40e_veb_config_tc - Configure TCs for given VEB
5064  * @veb: given VEB
5065  * @enabled_tc: TC bitmap
5066  *
5067  * Configures given TC bitmap for VEB (switching) element
5068  **/
5069 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
5070 {
5071         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
5072         struct i40e_pf *pf = veb->pf;
5073         int ret = 0;
5074         int i;
5075
5076         /* No TCs or already enabled TCs just return */
5077         if (!enabled_tc || veb->enabled_tc == enabled_tc)
5078                 return ret;
5079
5080         bw_data.tc_valid_bits = enabled_tc;
5081         /* bw_data.absolute_credits is not set (relative) */
5082
5083         /* Enable ETS TCs with equal BW Share for now */
5084         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5085                 if (enabled_tc & BIT(i))
5086                         bw_data.tc_bw_share_credits[i] = 1;
5087         }
5088
5089         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
5090                                                    &bw_data, NULL);
5091         if (ret) {
5092                 dev_info(&pf->pdev->dev,
5093                          "VEB bw config failed, err %s aq_err %s\n",
5094                          i40e_stat_str(&pf->hw, ret),
5095                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5096                 goto out;
5097         }
5098
5099         /* Update the BW information */
5100         ret = i40e_veb_get_bw_info(veb);
5101         if (ret) {
5102                 dev_info(&pf->pdev->dev,
5103                          "Failed getting veb bw config, err %s aq_err %s\n",
5104                          i40e_stat_str(&pf->hw, ret),
5105                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5106         }
5107
5108 out:
5109         return ret;
5110 }
5111
5112 #ifdef CONFIG_I40E_DCB
5113 /**
5114  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
5115  * @pf: PF struct
5116  *
5117  * Reconfigure VEB/VSIs on a given PF; it is assumed that
5118  * the caller would've quiesce all the VSIs before calling
5119  * this function
5120  **/
5121 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5122 {
5123         u8 tc_map = 0;
5124         int ret;
5125         u8 v;
5126
5127         /* Enable the TCs available on PF to all VEBs */
5128         tc_map = i40e_pf_get_tc_map(pf);
5129         for (v = 0; v < I40E_MAX_VEB; v++) {
5130                 if (!pf->veb[v])
5131                         continue;
5132                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5133                 if (ret) {
5134                         dev_info(&pf->pdev->dev,
5135                                  "Failed configuring TC for VEB seid=%d\n",
5136                                  pf->veb[v]->seid);
5137                         /* Will try to configure as many components */
5138                 }
5139         }
5140
5141         /* Update each VSI */
5142         for (v = 0; v < pf->num_alloc_vsi; v++) {
5143                 if (!pf->vsi[v])
5144                         continue;
5145
5146                 /* - Enable all TCs for the LAN VSI
5147                  * - For all others keep them at TC0 for now
5148                  */
5149                 if (v == pf->lan_vsi)
5150                         tc_map = i40e_pf_get_tc_map(pf);
5151                 else
5152                         tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5153
5154                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5155                 if (ret) {
5156                         dev_info(&pf->pdev->dev,
5157                                  "Failed configuring TC for VSI seid=%d\n",
5158                                  pf->vsi[v]->seid);
5159                         /* Will try to configure as many components */
5160                 } else {
5161                         /* Re-configure VSI vectors based on updated TC map */
5162                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5163                         if (pf->vsi[v]->netdev)
5164                                 i40e_dcbnl_set_all(pf->vsi[v]);
5165                 }
5166         }
5167 }
5168
5169 /**
5170  * i40e_resume_port_tx - Resume port Tx
5171  * @pf: PF struct
5172  *
5173  * Resume a port's Tx and issue a PF reset in case of failure to
5174  * resume.
5175  **/
5176 static int i40e_resume_port_tx(struct i40e_pf *pf)
5177 {
5178         struct i40e_hw *hw = &pf->hw;
5179         int ret;
5180
5181         ret = i40e_aq_resume_port_tx(hw, NULL);
5182         if (ret) {
5183                 dev_info(&pf->pdev->dev,
5184                          "Resume Port Tx failed, err %s aq_err %s\n",
5185                           i40e_stat_str(&pf->hw, ret),
5186                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5187                 /* Schedule PF reset to recover */
5188                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
5189                 i40e_service_event_schedule(pf);
5190         }
5191
5192         return ret;
5193 }
5194
5195 /**
5196  * i40e_init_pf_dcb - Initialize DCB configuration
5197  * @pf: PF being configured
5198  *
5199  * Query the current DCB configuration and cache it
5200  * in the hardware structure
5201  **/
5202 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5203 {
5204         struct i40e_hw *hw = &pf->hw;
5205         int err = 0;
5206
5207         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5208         if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5209                 goto out;
5210
5211         /* Get the initial DCB configuration */
5212         err = i40e_init_dcb(hw);
5213         if (!err) {
5214                 /* Device/Function is not DCBX capable */
5215                 if ((!hw->func_caps.dcb) ||
5216                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5217                         dev_info(&pf->pdev->dev,
5218                                  "DCBX offload is not supported or is disabled for this PF.\n");
5219                 } else {
5220                         /* When status is not DISABLED then DCBX in FW */
5221                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5222                                        DCB_CAP_DCBX_VER_IEEE;
5223
5224                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5225                         /* Enable DCB tagging only when more than one TC
5226                          * or explicitly disable if only one TC
5227                          */
5228                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5229                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5230                         else
5231                                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5232                         dev_dbg(&pf->pdev->dev,
5233                                 "DCBX offload is supported for this PF.\n");
5234                 }
5235         } else {
5236                 dev_info(&pf->pdev->dev,
5237                          "Query for DCB configuration failed, err %s aq_err %s\n",
5238                          i40e_stat_str(&pf->hw, err),
5239                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5240         }
5241
5242 out:
5243         return err;
5244 }
5245 #endif /* CONFIG_I40E_DCB */
5246 #define SPEED_SIZE 14
5247 #define FC_SIZE 8
5248 /**
5249  * i40e_print_link_message - print link up or down
5250  * @vsi: the VSI for which link needs a message
5251  */
5252 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5253 {
5254         enum i40e_aq_link_speed new_speed;
5255         char *speed = "Unknown";
5256         char *fc = "Unknown";
5257         char *fec = "";
5258         char *an = "";
5259
5260         new_speed = vsi->back->hw.phy.link_info.link_speed;
5261
5262         if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
5263                 return;
5264         vsi->current_isup = isup;
5265         vsi->current_speed = new_speed;
5266         if (!isup) {
5267                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5268                 return;
5269         }
5270
5271         /* Warn user if link speed on NPAR enabled partition is not at
5272          * least 10GB
5273          */
5274         if (vsi->back->hw.func_caps.npar_enable &&
5275             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5276              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5277                 netdev_warn(vsi->netdev,
5278                             "The partition detected link speed that is less than 10Gbps\n");
5279
5280         switch (vsi->back->hw.phy.link_info.link_speed) {
5281         case I40E_LINK_SPEED_40GB:
5282                 speed = "40 G";
5283                 break;
5284         case I40E_LINK_SPEED_20GB:
5285                 speed = "20 G";
5286                 break;
5287         case I40E_LINK_SPEED_25GB:
5288                 speed = "25 G";
5289                 break;
5290         case I40E_LINK_SPEED_10GB:
5291                 speed = "10 G";
5292                 break;
5293         case I40E_LINK_SPEED_1GB:
5294                 speed = "1000 M";
5295                 break;
5296         case I40E_LINK_SPEED_100MB:
5297                 speed = "100 M";
5298                 break;
5299         default:
5300                 break;
5301         }
5302
5303         switch (vsi->back->hw.fc.current_mode) {
5304         case I40E_FC_FULL:
5305                 fc = "RX/TX";
5306                 break;
5307         case I40E_FC_TX_PAUSE:
5308                 fc = "TX";
5309                 break;
5310         case I40E_FC_RX_PAUSE:
5311                 fc = "RX";
5312                 break;
5313         default:
5314                 fc = "None";
5315                 break;
5316         }
5317
5318         if (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
5319                 fec = ", FEC: None";
5320                 an = ", Autoneg: False";
5321
5322                 if (vsi->back->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
5323                         an = ", Autoneg: True";
5324
5325                 if (vsi->back->hw.phy.link_info.fec_info &
5326                     I40E_AQ_CONFIG_FEC_KR_ENA)
5327                         fec = ", FEC: CL74 FC-FEC/BASE-R";
5328                 else if (vsi->back->hw.phy.link_info.fec_info &
5329                          I40E_AQ_CONFIG_FEC_RS_ENA)
5330                         fec = ", FEC: CL108 RS-FEC";
5331         }
5332
5333         netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s, Flow Control: %s\n",
5334                     speed, fec, an, fc);
5335 }
5336
5337 /**
5338  * i40e_up_complete - Finish the last steps of bringing up a connection
5339  * @vsi: the VSI being configured
5340  **/
5341 static int i40e_up_complete(struct i40e_vsi *vsi)
5342 {
5343         struct i40e_pf *pf = vsi->back;
5344         int err;
5345
5346         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5347                 i40e_vsi_configure_msix(vsi);
5348         else
5349                 i40e_configure_msi_and_legacy(vsi);
5350
5351         /* start rings */
5352         err = i40e_vsi_start_rings(vsi);
5353         if (err)
5354                 return err;
5355
5356         clear_bit(__I40E_VSI_DOWN, vsi->state);
5357         i40e_napi_enable_all(vsi);
5358         i40e_vsi_enable_irq(vsi);
5359
5360         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5361             (vsi->netdev)) {
5362                 i40e_print_link_message(vsi, true);
5363                 netif_tx_start_all_queues(vsi->netdev);
5364                 netif_carrier_on(vsi->netdev);
5365         } else if (vsi->netdev) {
5366                 i40e_print_link_message(vsi, false);
5367                 /* need to check for qualified module here*/
5368                 if ((pf->hw.phy.link_info.link_info &
5369                         I40E_AQ_MEDIA_AVAILABLE) &&
5370                     (!(pf->hw.phy.link_info.an_info &
5371                         I40E_AQ_QUALIFIED_MODULE)))
5372                         netdev_err(vsi->netdev,
5373                                    "the driver failed to link because an unqualified module was detected.");
5374         }
5375
5376         /* replay FDIR SB filters */
5377         if (vsi->type == I40E_VSI_FDIR) {
5378                 /* reset fd counters */
5379                 pf->fd_add_err = 0;
5380                 pf->fd_atr_cnt = 0;
5381                 i40e_fdir_filter_restore(vsi);
5382         }
5383
5384         /* On the next run of the service_task, notify any clients of the new
5385          * opened netdev
5386          */
5387         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5388         i40e_service_event_schedule(pf);
5389
5390         return 0;
5391 }
5392
5393 /**
5394  * i40e_vsi_reinit_locked - Reset the VSI
5395  * @vsi: the VSI being configured
5396  *
5397  * Rebuild the ring structs after some configuration
5398  * has changed, e.g. MTU size.
5399  **/
5400 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5401 {
5402         struct i40e_pf *pf = vsi->back;
5403
5404         WARN_ON(in_interrupt());
5405         while (test_and_set_bit(__I40E_CONFIG_BUSY, pf->state))
5406                 usleep_range(1000, 2000);
5407         i40e_down(vsi);
5408
5409         i40e_up(vsi);
5410         clear_bit(__I40E_CONFIG_BUSY, pf->state);
5411 }
5412
5413 /**
5414  * i40e_up - Bring the connection back up after being down
5415  * @vsi: the VSI being configured
5416  **/
5417 int i40e_up(struct i40e_vsi *vsi)
5418 {
5419         int err;
5420
5421         err = i40e_vsi_configure(vsi);
5422         if (!err)
5423                 err = i40e_up_complete(vsi);
5424
5425         return err;
5426 }
5427
5428 /**
5429  * i40e_down - Shutdown the connection processing
5430  * @vsi: the VSI being stopped
5431  **/
5432 void i40e_down(struct i40e_vsi *vsi)
5433 {
5434         int i;
5435
5436         /* It is assumed that the caller of this function
5437          * sets the vsi->state __I40E_VSI_DOWN bit.
5438          */
5439         if (vsi->netdev) {
5440                 netif_carrier_off(vsi->netdev);
5441                 netif_tx_disable(vsi->netdev);
5442         }
5443         i40e_vsi_disable_irq(vsi);
5444         i40e_vsi_stop_rings(vsi);
5445         i40e_napi_disable_all(vsi);
5446
5447         for (i = 0; i < vsi->num_queue_pairs; i++) {
5448                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5449                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5450         }
5451
5452 }
5453
5454 /**
5455  * i40e_setup_tc - configure multiple traffic classes
5456  * @netdev: net device to configure
5457  * @tc: number of traffic classes to enable
5458  **/
5459 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5460 {
5461         struct i40e_netdev_priv *np = netdev_priv(netdev);
5462         struct i40e_vsi *vsi = np->vsi;
5463         struct i40e_pf *pf = vsi->back;
5464         u8 enabled_tc = 0;
5465         int ret = -EINVAL;
5466         int i;
5467
5468         /* Check if DCB enabled to continue */
5469         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5470                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5471                 goto exit;
5472         }
5473
5474         /* Check if MFP enabled */
5475         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5476                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5477                 goto exit;
5478         }
5479
5480         /* Check whether tc count is within enabled limit */
5481         if (tc > i40e_pf_get_num_tc(pf)) {
5482                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5483                 goto exit;
5484         }
5485
5486         /* Generate TC map for number of tc requested */
5487         for (i = 0; i < tc; i++)
5488                 enabled_tc |= BIT(i);
5489
5490         /* Requesting same TC configuration as already enabled */
5491         if (enabled_tc == vsi->tc_config.enabled_tc)
5492                 return 0;
5493
5494         /* Quiesce VSI queues */
5495         i40e_quiesce_vsi(vsi);
5496
5497         /* Configure VSI for enabled TCs */
5498         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5499         if (ret) {
5500                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5501                             vsi->seid);
5502                 goto exit;
5503         }
5504
5505         /* Unquiesce VSI */
5506         i40e_unquiesce_vsi(vsi);
5507
5508 exit:
5509         return ret;
5510 }
5511
5512 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5513                            struct tc_to_netdev *tc)
5514 {
5515         if (tc->type != TC_SETUP_MQPRIO)
5516                 return -EINVAL;
5517
5518         tc->mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
5519
5520         return i40e_setup_tc(netdev, tc->mqprio->num_tc);
5521 }
5522
5523 /**
5524  * i40e_open - Called when a network interface is made active
5525  * @netdev: network interface device structure
5526  *
5527  * The open entry point is called when a network interface is made
5528  * active by the system (IFF_UP).  At this point all resources needed
5529  * for transmit and receive operations are allocated, the interrupt
5530  * handler is registered with the OS, the netdev watchdog subtask is
5531  * enabled, and the stack is notified that the interface is ready.
5532  *
5533  * Returns 0 on success, negative value on failure
5534  **/
5535 int i40e_open(struct net_device *netdev)
5536 {
5537         struct i40e_netdev_priv *np = netdev_priv(netdev);
5538         struct i40e_vsi *vsi = np->vsi;
5539         struct i40e_pf *pf = vsi->back;
5540         int err;
5541
5542         /* disallow open during test or if eeprom is broken */
5543         if (test_bit(__I40E_TESTING, pf->state) ||
5544             test_bit(__I40E_BAD_EEPROM, pf->state))
5545                 return -EBUSY;
5546
5547         netif_carrier_off(netdev);
5548
5549         err = i40e_vsi_open(vsi);
5550         if (err)
5551                 return err;
5552
5553         /* configure global TSO hardware offload settings */
5554         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5555                                                        TCP_FLAG_FIN) >> 16);
5556         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5557                                                        TCP_FLAG_FIN |
5558                                                        TCP_FLAG_CWR) >> 16);
5559         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5560
5561         udp_tunnel_get_rx_info(netdev);
5562
5563         return 0;
5564 }
5565
5566 /**
5567  * i40e_vsi_open -
5568  * @vsi: the VSI to open
5569  *
5570  * Finish initialization of the VSI.
5571  *
5572  * Returns 0 on success, negative value on failure
5573  *
5574  * Note: expects to be called while under rtnl_lock()
5575  **/
5576 int i40e_vsi_open(struct i40e_vsi *vsi)
5577 {
5578         struct i40e_pf *pf = vsi->back;
5579         char int_name[I40E_INT_NAME_STR_LEN];
5580         int err;
5581
5582         /* allocate descriptors */
5583         err = i40e_vsi_setup_tx_resources(vsi);
5584         if (err)
5585                 goto err_setup_tx;
5586         err = i40e_vsi_setup_rx_resources(vsi);
5587         if (err)
5588                 goto err_setup_rx;
5589
5590         err = i40e_vsi_configure(vsi);
5591         if (err)
5592                 goto err_setup_rx;
5593
5594         if (vsi->netdev) {
5595                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5596                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5597                 err = i40e_vsi_request_irq(vsi, int_name);
5598                 if (err)
5599                         goto err_setup_rx;
5600
5601                 /* Notify the stack of the actual queue counts. */
5602                 err = netif_set_real_num_tx_queues(vsi->netdev,
5603                                                    vsi->num_queue_pairs);
5604                 if (err)
5605                         goto err_set_queues;
5606
5607                 err = netif_set_real_num_rx_queues(vsi->netdev,
5608                                                    vsi->num_queue_pairs);
5609                 if (err)
5610                         goto err_set_queues;
5611
5612         } else if (vsi->type == I40E_VSI_FDIR) {
5613                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5614                          dev_driver_string(&pf->pdev->dev),
5615                          dev_name(&pf->pdev->dev));
5616                 err = i40e_vsi_request_irq(vsi, int_name);
5617
5618         } else {
5619                 err = -EINVAL;
5620                 goto err_setup_rx;
5621         }
5622
5623         err = i40e_up_complete(vsi);
5624         if (err)
5625                 goto err_up_complete;
5626
5627         return 0;
5628
5629 err_up_complete:
5630         i40e_down(vsi);
5631 err_set_queues:
5632         i40e_vsi_free_irq(vsi);
5633 err_setup_rx:
5634         i40e_vsi_free_rx_resources(vsi);
5635 err_setup_tx:
5636         i40e_vsi_free_tx_resources(vsi);
5637         if (vsi == pf->vsi[pf->lan_vsi])
5638                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
5639
5640         return err;
5641 }
5642
5643 /**
5644  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5645  * @pf: Pointer to PF
5646  *
5647  * This function destroys the hlist where all the Flow Director
5648  * filters were saved.
5649  **/
5650 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5651 {
5652         struct i40e_fdir_filter *filter;
5653         struct i40e_flex_pit *pit_entry, *tmp;
5654         struct hlist_node *node2;
5655
5656         hlist_for_each_entry_safe(filter, node2,
5657                                   &pf->fdir_filter_list, fdir_node) {
5658                 hlist_del(&filter->fdir_node);
5659                 kfree(filter);
5660         }
5661
5662         list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
5663                 list_del(&pit_entry->list);
5664                 kfree(pit_entry);
5665         }
5666         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
5667
5668         list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
5669                 list_del(&pit_entry->list);
5670                 kfree(pit_entry);
5671         }
5672         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
5673
5674         pf->fdir_pf_active_filters = 0;
5675         pf->fd_tcp4_filter_cnt = 0;
5676         pf->fd_udp4_filter_cnt = 0;
5677         pf->fd_sctp4_filter_cnt = 0;
5678         pf->fd_ip4_filter_cnt = 0;
5679
5680         /* Reprogram the default input set for TCP/IPv4 */
5681         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
5682                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5683                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5684
5685         /* Reprogram the default input set for UDP/IPv4 */
5686         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
5687                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5688                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5689
5690         /* Reprogram the default input set for SCTP/IPv4 */
5691         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
5692                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5693                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5694
5695         /* Reprogram the default input set for Other/IPv4 */
5696         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
5697                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
5698 }
5699
5700 /**
5701  * i40e_close - Disables a network interface
5702  * @netdev: network interface device structure
5703  *
5704  * The close entry point is called when an interface is de-activated
5705  * by the OS.  The hardware is still under the driver's control, but
5706  * this netdev interface is disabled.
5707  *
5708  * Returns 0, this is not allowed to fail
5709  **/
5710 int i40e_close(struct net_device *netdev)
5711 {
5712         struct i40e_netdev_priv *np = netdev_priv(netdev);
5713         struct i40e_vsi *vsi = np->vsi;
5714
5715         i40e_vsi_close(vsi);
5716
5717         return 0;
5718 }
5719
5720 /**
5721  * i40e_do_reset - Start a PF or Core Reset sequence
5722  * @pf: board private structure
5723  * @reset_flags: which reset is requested
5724  * @lock_acquired: indicates whether or not the lock has been acquired
5725  * before this function was called.
5726  *
5727  * The essential difference in resets is that the PF Reset
5728  * doesn't clear the packet buffers, doesn't reset the PE
5729  * firmware, and doesn't bother the other PFs on the chip.
5730  **/
5731 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
5732 {
5733         u32 val;
5734
5735         WARN_ON(in_interrupt());
5736
5737
5738         /* do the biggest reset indicated */
5739         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5740
5741                 /* Request a Global Reset
5742                  *
5743                  * This will start the chip's countdown to the actual full
5744                  * chip reset event, and a warning interrupt to be sent
5745                  * to all PFs, including the requestor.  Our handler
5746                  * for the warning interrupt will deal with the shutdown
5747                  * and recovery of the switch setup.
5748                  */
5749                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5750                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5751                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5752                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5753
5754         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5755
5756                 /* Request a Core Reset
5757                  *
5758                  * Same as Global Reset, except does *not* include the MAC/PHY
5759                  */
5760                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5761                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5762                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5763                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5764                 i40e_flush(&pf->hw);
5765
5766         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5767
5768                 /* Request a PF Reset
5769                  *
5770                  * Resets only the PF-specific registers
5771                  *
5772                  * This goes directly to the tear-down and rebuild of
5773                  * the switch, since we need to do all the recovery as
5774                  * for the Core Reset.
5775                  */
5776                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5777                 i40e_handle_reset_warning(pf, lock_acquired);
5778
5779         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5780                 int v;
5781
5782                 /* Find the VSI(s) that requested a re-init */
5783                 dev_info(&pf->pdev->dev,
5784                          "VSI reinit requested\n");
5785                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5786                         struct i40e_vsi *vsi = pf->vsi[v];
5787
5788                         if (vsi != NULL &&
5789                             test_and_clear_bit(__I40E_VSI_REINIT_REQUESTED,
5790                                                vsi->state))
5791                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5792                 }
5793         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5794                 int v;
5795
5796                 /* Find the VSI(s) that needs to be brought down */
5797                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5798                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5799                         struct i40e_vsi *vsi = pf->vsi[v];
5800
5801                         if (vsi != NULL &&
5802                             test_and_clear_bit(__I40E_VSI_DOWN_REQUESTED,
5803                                                vsi->state)) {
5804                                 set_bit(__I40E_VSI_DOWN, vsi->state);
5805                                 i40e_down(vsi);
5806                         }
5807                 }
5808         } else {
5809                 dev_info(&pf->pdev->dev,
5810                          "bad reset request 0x%08x\n", reset_flags);
5811         }
5812 }
5813
5814 #ifdef CONFIG_I40E_DCB
5815 /**
5816  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5817  * @pf: board private structure
5818  * @old_cfg: current DCB config
5819  * @new_cfg: new DCB config
5820  **/
5821 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5822                             struct i40e_dcbx_config *old_cfg,
5823                             struct i40e_dcbx_config *new_cfg)
5824 {
5825         bool need_reconfig = false;
5826
5827         /* Check if ETS configuration has changed */
5828         if (memcmp(&new_cfg->etscfg,
5829                    &old_cfg->etscfg,
5830                    sizeof(new_cfg->etscfg))) {
5831                 /* If Priority Table has changed reconfig is needed */
5832                 if (memcmp(&new_cfg->etscfg.prioritytable,
5833                            &old_cfg->etscfg.prioritytable,
5834                            sizeof(new_cfg->etscfg.prioritytable))) {
5835                         need_reconfig = true;
5836                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5837                 }
5838
5839                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5840                            &old_cfg->etscfg.tcbwtable,
5841                            sizeof(new_cfg->etscfg.tcbwtable)))
5842                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5843
5844                 if (memcmp(&new_cfg->etscfg.tsatable,
5845                            &old_cfg->etscfg.tsatable,
5846                            sizeof(new_cfg->etscfg.tsatable)))
5847                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5848         }
5849
5850         /* Check if PFC configuration has changed */
5851         if (memcmp(&new_cfg->pfc,
5852                    &old_cfg->pfc,
5853                    sizeof(new_cfg->pfc))) {
5854                 need_reconfig = true;
5855                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5856         }
5857
5858         /* Check if APP Table has changed */
5859         if (memcmp(&new_cfg->app,
5860                    &old_cfg->app,
5861                    sizeof(new_cfg->app))) {
5862                 need_reconfig = true;
5863                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5864         }
5865
5866         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5867         return need_reconfig;
5868 }
5869
5870 /**
5871  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5872  * @pf: board private structure
5873  * @e: event info posted on ARQ
5874  **/
5875 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5876                                   struct i40e_arq_event_info *e)
5877 {
5878         struct i40e_aqc_lldp_get_mib *mib =
5879                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5880         struct i40e_hw *hw = &pf->hw;
5881         struct i40e_dcbx_config tmp_dcbx_cfg;
5882         bool need_reconfig = false;
5883         int ret = 0;
5884         u8 type;
5885
5886         /* Not DCB capable or capability disabled */
5887         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5888                 return ret;
5889
5890         /* Ignore if event is not for Nearest Bridge */
5891         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5892                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5893         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5894         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5895                 return ret;
5896
5897         /* Check MIB Type and return if event for Remote MIB update */
5898         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5899         dev_dbg(&pf->pdev->dev,
5900                 "LLDP event mib type %s\n", type ? "remote" : "local");
5901         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5902                 /* Update the remote cached instance and return */
5903                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5904                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5905                                 &hw->remote_dcbx_config);
5906                 goto exit;
5907         }
5908
5909         /* Store the old configuration */
5910         tmp_dcbx_cfg = hw->local_dcbx_config;
5911
5912         /* Reset the old DCBx configuration data */
5913         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5914         /* Get updated DCBX data from firmware */
5915         ret = i40e_get_dcb_config(&pf->hw);
5916         if (ret) {
5917                 dev_info(&pf->pdev->dev,
5918                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5919                          i40e_stat_str(&pf->hw, ret),
5920                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5921                 goto exit;
5922         }
5923
5924         /* No change detected in DCBX configs */
5925         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5926                     sizeof(tmp_dcbx_cfg))) {
5927                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5928                 goto exit;
5929         }
5930
5931         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5932                                                &hw->local_dcbx_config);
5933
5934         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5935
5936         if (!need_reconfig)
5937                 goto exit;
5938
5939         /* Enable DCB tagging only when more than one TC */
5940         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5941                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5942         else
5943                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5944
5945         set_bit(__I40E_PORT_SUSPENDED, pf->state);
5946         /* Reconfiguration needed quiesce all VSIs */
5947         i40e_pf_quiesce_all_vsi(pf);
5948
5949         /* Changes in configuration update VEB/VSI */
5950         i40e_dcb_reconfigure(pf);
5951
5952         ret = i40e_resume_port_tx(pf);
5953
5954         clear_bit(__I40E_PORT_SUSPENDED, pf->state);
5955         /* In case of error no point in resuming VSIs */
5956         if (ret)
5957                 goto exit;
5958
5959         /* Wait for the PF's queues to be disabled */
5960         ret = i40e_pf_wait_queues_disabled(pf);
5961         if (ret) {
5962                 /* Schedule PF reset to recover */
5963                 set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
5964                 i40e_service_event_schedule(pf);
5965         } else {
5966                 i40e_pf_unquiesce_all_vsi(pf);
5967         pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
5968                       I40E_FLAG_CLIENT_L2_CHANGE);
5969         }
5970
5971 exit:
5972         return ret;
5973 }
5974 #endif /* CONFIG_I40E_DCB */
5975
5976 /**
5977  * i40e_do_reset_safe - Protected reset path for userland calls.
5978  * @pf: board private structure
5979  * @reset_flags: which reset is requested
5980  *
5981  **/
5982 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5983 {
5984         rtnl_lock();
5985         i40e_do_reset(pf, reset_flags, true);
5986         rtnl_unlock();
5987 }
5988
5989 /**
5990  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5991  * @pf: board private structure
5992  * @e: event info posted on ARQ
5993  *
5994  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5995  * and VF queues
5996  **/
5997 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5998                                            struct i40e_arq_event_info *e)
5999 {
6000         struct i40e_aqc_lan_overflow *data =
6001                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
6002         u32 queue = le32_to_cpu(data->prtdcb_rupto);
6003         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
6004         struct i40e_hw *hw = &pf->hw;
6005         struct i40e_vf *vf;
6006         u16 vf_id;
6007
6008         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
6009                 queue, qtx_ctl);
6010
6011         /* Queue belongs to VF, find the VF and issue VF reset */
6012         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
6013             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
6014                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
6015                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
6016                 vf_id -= hw->func_caps.vf_base_id;
6017                 vf = &pf->vf[vf_id];
6018                 i40e_vc_notify_vf_reset(vf);
6019                 /* Allow VF to process pending reset notification */
6020                 msleep(20);
6021                 i40e_reset_vf(vf, false);
6022         }
6023 }
6024
6025 /**
6026  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
6027  * @pf: board private structure
6028  **/
6029 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
6030 {
6031         u32 val, fcnt_prog;
6032
6033         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6034         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
6035         return fcnt_prog;
6036 }
6037
6038 /**
6039  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
6040  * @pf: board private structure
6041  **/
6042 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
6043 {
6044         u32 val, fcnt_prog;
6045
6046         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6047         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
6048                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
6049                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
6050         return fcnt_prog;
6051 }
6052
6053 /**
6054  * i40e_get_global_fd_count - Get total FD filters programmed on device
6055  * @pf: board private structure
6056  **/
6057 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
6058 {
6059         u32 val, fcnt_prog;
6060
6061         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
6062         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
6063                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
6064                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
6065         return fcnt_prog;
6066 }
6067
6068 /**
6069  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
6070  * @pf: board private structure
6071  **/
6072 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
6073 {
6074         struct i40e_fdir_filter *filter;
6075         u32 fcnt_prog, fcnt_avail;
6076         struct hlist_node *node;
6077
6078         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6079                 return;
6080
6081         /* Check if we have enough room to re-enable FDir SB capability. */
6082         fcnt_prog = i40e_get_global_fd_count(pf);
6083         fcnt_avail = pf->fdir_pf_filter_count;
6084         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
6085             (pf->fd_add_err == 0) ||
6086             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
6087                 if (pf->flags & I40E_FLAG_FD_SB_AUTO_DISABLED) {
6088                         pf->flags &= ~I40E_FLAG_FD_SB_AUTO_DISABLED;
6089                         if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
6090                             (I40E_DEBUG_FD & pf->hw.debug_mask))
6091                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
6092                 }
6093         }
6094
6095         /* We should wait for even more space before re-enabling ATR.
6096          * Additionally, we cannot enable ATR as long as we still have TCP SB
6097          * rules active.
6098          */
6099         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) &&
6100             (pf->fd_tcp4_filter_cnt == 0)) {
6101                 if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
6102                         pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6103                         if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
6104                             (I40E_DEBUG_FD & pf->hw.debug_mask))
6105                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
6106                 }
6107         }
6108
6109         /* if hw had a problem adding a filter, delete it */
6110         if (pf->fd_inv > 0) {
6111                 hlist_for_each_entry_safe(filter, node,
6112                                           &pf->fdir_filter_list, fdir_node) {
6113                         if (filter->fd_id == pf->fd_inv) {
6114                                 hlist_del(&filter->fdir_node);
6115                                 kfree(filter);
6116                                 pf->fdir_pf_active_filters--;
6117                         }
6118                 }
6119         }
6120 }
6121
6122 #define I40E_MIN_FD_FLUSH_INTERVAL 10
6123 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
6124 /**
6125  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
6126  * @pf: board private structure
6127  **/
6128 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
6129 {
6130         unsigned long min_flush_time;
6131         int flush_wait_retry = 50;
6132         bool disable_atr = false;
6133         int fd_room;
6134         int reg;
6135
6136         if (!time_after(jiffies, pf->fd_flush_timestamp +
6137                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
6138                 return;
6139
6140         /* If the flush is happening too quick and we have mostly SB rules we
6141          * should not re-enable ATR for some time.
6142          */
6143         min_flush_time = pf->fd_flush_timestamp +
6144                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
6145         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
6146
6147         if (!(time_after(jiffies, min_flush_time)) &&
6148             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
6149                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6150                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6151                 disable_atr = true;
6152         }
6153
6154         pf->fd_flush_timestamp = jiffies;
6155         pf->flags |= I40E_FLAG_FD_ATR_AUTO_DISABLED;
6156         /* flush all filters */
6157         wr32(&pf->hw, I40E_PFQF_CTL_1,
6158              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6159         i40e_flush(&pf->hw);
6160         pf->fd_flush_cnt++;
6161         pf->fd_add_err = 0;
6162         do {
6163                 /* Check FD flush status every 5-6msec */
6164                 usleep_range(5000, 6000);
6165                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6166                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6167                         break;
6168         } while (flush_wait_retry--);
6169         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6170                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6171         } else {
6172                 /* replay sideband filters */
6173                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6174                 if (!disable_atr && !pf->fd_tcp4_filter_cnt)
6175                         pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
6176                 clear_bit(__I40E_FD_FLUSH_REQUESTED, pf->state);
6177                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6178                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6179         }
6180 }
6181
6182 /**
6183  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6184  * @pf: board private structure
6185  **/
6186 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6187 {
6188         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6189 }
6190
6191 /* We can see up to 256 filter programming desc in transit if the filters are
6192  * being applied really fast; before we see the first
6193  * filter miss error on Rx queue 0. Accumulating enough error messages before
6194  * reacting will make sure we don't cause flush too often.
6195  */
6196 #define I40E_MAX_FD_PROGRAM_ERROR 256
6197
6198 /**
6199  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6200  * @pf: board private structure
6201  **/
6202 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6203 {
6204
6205         /* if interface is down do nothing */
6206         if (test_bit(__I40E_VSI_DOWN, pf->state))
6207                 return;
6208
6209         if (test_bit(__I40E_FD_FLUSH_REQUESTED, pf->state))
6210                 i40e_fdir_flush_and_replay(pf);
6211
6212         i40e_fdir_check_and_reenable(pf);
6213
6214 }
6215
6216 /**
6217  * i40e_vsi_link_event - notify VSI of a link event
6218  * @vsi: vsi to be notified
6219  * @link_up: link up or down
6220  **/
6221 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6222 {
6223         if (!vsi || test_bit(__I40E_VSI_DOWN, vsi->state))
6224                 return;
6225
6226         switch (vsi->type) {
6227         case I40E_VSI_MAIN:
6228                 if (!vsi->netdev || !vsi->netdev_registered)
6229                         break;
6230
6231                 if (link_up) {
6232                         netif_carrier_on(vsi->netdev);
6233                         netif_tx_wake_all_queues(vsi->netdev);
6234                 } else {
6235                         netif_carrier_off(vsi->netdev);
6236                         netif_tx_stop_all_queues(vsi->netdev);
6237                 }
6238                 break;
6239
6240         case I40E_VSI_SRIOV:
6241         case I40E_VSI_VMDQ2:
6242         case I40E_VSI_CTRL:
6243         case I40E_VSI_IWARP:
6244         case I40E_VSI_MIRROR:
6245         default:
6246                 /* there is no notification for other VSIs */
6247                 break;
6248         }
6249 }
6250
6251 /**
6252  * i40e_veb_link_event - notify elements on the veb of a link event
6253  * @veb: veb to be notified
6254  * @link_up: link up or down
6255  **/
6256 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6257 {
6258         struct i40e_pf *pf;
6259         int i;
6260
6261         if (!veb || !veb->pf)
6262                 return;
6263         pf = veb->pf;
6264
6265         /* depth first... */
6266         for (i = 0; i < I40E_MAX_VEB; i++)
6267                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6268                         i40e_veb_link_event(pf->veb[i], link_up);
6269
6270         /* ... now the local VSIs */
6271         for (i = 0; i < pf->num_alloc_vsi; i++)
6272                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6273                         i40e_vsi_link_event(pf->vsi[i], link_up);
6274 }
6275
6276 /**
6277  * i40e_link_event - Update netif_carrier status
6278  * @pf: board private structure
6279  **/
6280 static void i40e_link_event(struct i40e_pf *pf)
6281 {
6282         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6283         u8 new_link_speed, old_link_speed;
6284         i40e_status status;
6285         bool new_link, old_link;
6286
6287         /* save off old link status information */
6288         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6289
6290         /* set this to force the get_link_status call to refresh state */
6291         pf->hw.phy.get_link_info = true;
6292
6293         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6294
6295         status = i40e_get_link_status(&pf->hw, &new_link);
6296
6297         /* On success, disable temp link polling */
6298         if (status == I40E_SUCCESS) {
6299                 if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
6300                         pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
6301         } else {
6302                 /* Enable link polling temporarily until i40e_get_link_status
6303                  * returns I40E_SUCCESS
6304                  */
6305                 pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
6306                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6307                         status);
6308                 return;
6309         }
6310
6311         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6312         new_link_speed = pf->hw.phy.link_info.link_speed;
6313
6314         if (new_link == old_link &&
6315             new_link_speed == old_link_speed &&
6316             (test_bit(__I40E_VSI_DOWN, vsi->state) ||
6317              new_link == netif_carrier_ok(vsi->netdev)))
6318                 return;
6319
6320         if (!test_bit(__I40E_VSI_DOWN, vsi->state))
6321                 i40e_print_link_message(vsi, new_link);
6322
6323         /* Notify the base of the switch tree connected to
6324          * the link.  Floating VEBs are not notified.
6325          */
6326         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6327                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6328         else
6329                 i40e_vsi_link_event(vsi, new_link);
6330
6331         if (pf->vf)
6332                 i40e_vc_notify_link_state(pf);
6333
6334         if (pf->flags & I40E_FLAG_PTP)
6335                 i40e_ptp_set_increment(pf);
6336 }
6337
6338 /**
6339  * i40e_watchdog_subtask - periodic checks not using event driven response
6340  * @pf: board private structure
6341  **/
6342 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6343 {
6344         int i;
6345
6346         /* if interface is down do nothing */
6347         if (test_bit(__I40E_VSI_DOWN, pf->state) ||
6348             test_bit(__I40E_CONFIG_BUSY, pf->state))
6349                 return;
6350
6351         /* make sure we don't do these things too often */
6352         if (time_before(jiffies, (pf->service_timer_previous +
6353                                   pf->service_timer_period)))
6354                 return;
6355         pf->service_timer_previous = jiffies;
6356
6357         if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
6358             (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
6359                 i40e_link_event(pf);
6360
6361         /* Update the stats for active netdevs so the network stack
6362          * can look at updated numbers whenever it cares to
6363          */
6364         for (i = 0; i < pf->num_alloc_vsi; i++)
6365                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6366                         i40e_update_stats(pf->vsi[i]);
6367
6368         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6369                 /* Update the stats for the active switching components */
6370                 for (i = 0; i < I40E_MAX_VEB; i++)
6371                         if (pf->veb[i])
6372                                 i40e_update_veb_stats(pf->veb[i]);
6373         }
6374
6375         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6376 }
6377
6378 /**
6379  * i40e_reset_subtask - Set up for resetting the device and driver
6380  * @pf: board private structure
6381  **/
6382 static void i40e_reset_subtask(struct i40e_pf *pf)
6383 {
6384         u32 reset_flags = 0;
6385
6386         if (test_bit(__I40E_REINIT_REQUESTED, pf->state)) {
6387                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6388                 clear_bit(__I40E_REINIT_REQUESTED, pf->state);
6389         }
6390         if (test_bit(__I40E_PF_RESET_REQUESTED, pf->state)) {
6391                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6392                 clear_bit(__I40E_PF_RESET_REQUESTED, pf->state);
6393         }
6394         if (test_bit(__I40E_CORE_RESET_REQUESTED, pf->state)) {
6395                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6396                 clear_bit(__I40E_CORE_RESET_REQUESTED, pf->state);
6397         }
6398         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state)) {
6399                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6400                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, pf->state);
6401         }
6402         if (test_bit(__I40E_VSI_DOWN_REQUESTED, pf->state)) {
6403                 reset_flags |= BIT(__I40E_VSI_DOWN_REQUESTED);
6404                 clear_bit(__I40E_VSI_DOWN_REQUESTED, pf->state);
6405         }
6406
6407         /* If there's a recovery already waiting, it takes
6408          * precedence before starting a new reset sequence.
6409          */
6410         if (test_bit(__I40E_RESET_INTR_RECEIVED, pf->state)) {
6411                 i40e_prep_for_reset(pf, false);
6412                 i40e_reset(pf);
6413                 i40e_rebuild(pf, false, false);
6414         }
6415
6416         /* If we're already down or resetting, just bail */
6417         if (reset_flags &&
6418             !test_bit(__I40E_VSI_DOWN, pf->state) &&
6419             !test_bit(__I40E_CONFIG_BUSY, pf->state)) {
6420                 rtnl_lock();
6421                 i40e_do_reset(pf, reset_flags, true);
6422                 rtnl_unlock();
6423         }
6424 }
6425
6426 /**
6427  * i40e_handle_link_event - Handle link event
6428  * @pf: board private structure
6429  * @e: event info posted on ARQ
6430  **/
6431 static void i40e_handle_link_event(struct i40e_pf *pf,
6432                                    struct i40e_arq_event_info *e)
6433 {
6434         struct i40e_aqc_get_link_status *status =
6435                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6436
6437         /* Do a new status request to re-enable LSE reporting
6438          * and load new status information into the hw struct
6439          * This completely ignores any state information
6440          * in the ARQ event info, instead choosing to always
6441          * issue the AQ update link status command.
6442          */
6443         i40e_link_event(pf);
6444
6445         /* check for unqualified module, if link is down */
6446         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6447             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6448             (!(status->link_info & I40E_AQ_LINK_UP)))
6449                 dev_err(&pf->pdev->dev,
6450                         "The driver failed to link because an unqualified module was detected.\n");
6451 }
6452
6453 /**
6454  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6455  * @pf: board private structure
6456  **/
6457 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6458 {
6459         struct i40e_arq_event_info event;
6460         struct i40e_hw *hw = &pf->hw;
6461         u16 pending, i = 0;
6462         i40e_status ret;
6463         u16 opcode;
6464         u32 oldval;
6465         u32 val;
6466
6467         /* Do not run clean AQ when PF reset fails */
6468         if (test_bit(__I40E_RESET_FAILED, pf->state))
6469                 return;
6470
6471         /* check for error indications */
6472         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6473         oldval = val;
6474         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6475                 if (hw->debug_mask & I40E_DEBUG_AQ)
6476                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6477                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6478         }
6479         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6480                 if (hw->debug_mask & I40E_DEBUG_AQ)
6481                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6482                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6483                 pf->arq_overflows++;
6484         }
6485         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6486                 if (hw->debug_mask & I40E_DEBUG_AQ)
6487                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6488                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6489         }
6490         if (oldval != val)
6491                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6492
6493         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6494         oldval = val;
6495         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6496                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6497                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6498                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6499         }
6500         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6501                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6502                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6503                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6504         }
6505         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6506                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6507                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6508                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6509         }
6510         if (oldval != val)
6511                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6512
6513         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6514         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6515         if (!event.msg_buf)
6516                 return;
6517
6518         do {
6519                 ret = i40e_clean_arq_element(hw, &event, &pending);
6520                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6521                         break;
6522                 else if (ret) {
6523                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6524                         break;
6525                 }
6526
6527                 opcode = le16_to_cpu(event.desc.opcode);
6528                 switch (opcode) {
6529
6530                 case i40e_aqc_opc_get_link_status:
6531                         i40e_handle_link_event(pf, &event);
6532                         break;
6533                 case i40e_aqc_opc_send_msg_to_pf:
6534                         ret = i40e_vc_process_vf_msg(pf,
6535                                         le16_to_cpu(event.desc.retval),
6536                                         le32_to_cpu(event.desc.cookie_high),
6537                                         le32_to_cpu(event.desc.cookie_low),
6538                                         event.msg_buf,
6539                                         event.msg_len);
6540                         break;
6541                 case i40e_aqc_opc_lldp_update_mib:
6542                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6543 #ifdef CONFIG_I40E_DCB
6544                         rtnl_lock();
6545                         ret = i40e_handle_lldp_event(pf, &event);
6546                         rtnl_unlock();
6547 #endif /* CONFIG_I40E_DCB */
6548                         break;
6549                 case i40e_aqc_opc_event_lan_overflow:
6550                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6551                         i40e_handle_lan_overflow_event(pf, &event);
6552                         break;
6553                 case i40e_aqc_opc_send_msg_to_peer:
6554                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6555                         break;
6556                 case i40e_aqc_opc_nvm_erase:
6557                 case i40e_aqc_opc_nvm_update:
6558                 case i40e_aqc_opc_oem_post_update:
6559                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6560                                    "ARQ NVM operation 0x%04x completed\n",
6561                                    opcode);
6562                         break;
6563                 default:
6564                         dev_info(&pf->pdev->dev,
6565                                  "ARQ: Unknown event 0x%04x ignored\n",
6566                                  opcode);
6567                         break;
6568                 }
6569         } while (i++ < pf->adminq_work_limit);
6570
6571         if (i < pf->adminq_work_limit)
6572                 clear_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state);
6573
6574         /* re-enable Admin queue interrupt cause */
6575         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6576         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6577         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6578         i40e_flush(hw);
6579
6580         kfree(event.msg_buf);
6581 }
6582
6583 /**
6584  * i40e_verify_eeprom - make sure eeprom is good to use
6585  * @pf: board private structure
6586  **/
6587 static void i40e_verify_eeprom(struct i40e_pf *pf)
6588 {
6589         int err;
6590
6591         err = i40e_diag_eeprom_test(&pf->hw);
6592         if (err) {
6593                 /* retry in case of garbage read */
6594                 err = i40e_diag_eeprom_test(&pf->hw);
6595                 if (err) {
6596                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6597                                  err);
6598                         set_bit(__I40E_BAD_EEPROM, pf->state);
6599                 }
6600         }
6601
6602         if (!err && test_bit(__I40E_BAD_EEPROM, pf->state)) {
6603                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6604                 clear_bit(__I40E_BAD_EEPROM, pf->state);
6605         }
6606 }
6607
6608 /**
6609  * i40e_enable_pf_switch_lb
6610  * @pf: pointer to the PF structure
6611  *
6612  * enable switch loop back or die - no point in a return value
6613  **/
6614 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6615 {
6616         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6617         struct i40e_vsi_context ctxt;
6618         int ret;
6619
6620         ctxt.seid = pf->main_vsi_seid;
6621         ctxt.pf_num = pf->hw.pf_id;
6622         ctxt.vf_num = 0;
6623         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6624         if (ret) {
6625                 dev_info(&pf->pdev->dev,
6626                          "couldn't get PF vsi config, err %s aq_err %s\n",
6627                          i40e_stat_str(&pf->hw, ret),
6628                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6629                 return;
6630         }
6631         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6632         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6633         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6634
6635         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6636         if (ret) {
6637                 dev_info(&pf->pdev->dev,
6638                          "update vsi switch failed, err %s aq_err %s\n",
6639                          i40e_stat_str(&pf->hw, ret),
6640                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6641         }
6642 }
6643
6644 /**
6645  * i40e_disable_pf_switch_lb
6646  * @pf: pointer to the PF structure
6647  *
6648  * disable switch loop back or die - no point in a return value
6649  **/
6650 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6651 {
6652         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6653         struct i40e_vsi_context ctxt;
6654         int ret;
6655
6656         ctxt.seid = pf->main_vsi_seid;
6657         ctxt.pf_num = pf->hw.pf_id;
6658         ctxt.vf_num = 0;
6659         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6660         if (ret) {
6661                 dev_info(&pf->pdev->dev,
6662                          "couldn't get PF vsi config, err %s aq_err %s\n",
6663                          i40e_stat_str(&pf->hw, ret),
6664                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6665                 return;
6666         }
6667         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6668         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6669         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6670
6671         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6672         if (ret) {
6673                 dev_info(&pf->pdev->dev,
6674                          "update vsi switch failed, err %s aq_err %s\n",
6675                          i40e_stat_str(&pf->hw, ret),
6676                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6677         }
6678 }
6679
6680 /**
6681  * i40e_config_bridge_mode - Configure the HW bridge mode
6682  * @veb: pointer to the bridge instance
6683  *
6684  * Configure the loop back mode for the LAN VSI that is downlink to the
6685  * specified HW bridge instance. It is expected this function is called
6686  * when a new HW bridge is instantiated.
6687  **/
6688 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6689 {
6690         struct i40e_pf *pf = veb->pf;
6691
6692         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6693                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6694                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6695         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6696                 i40e_disable_pf_switch_lb(pf);
6697         else
6698                 i40e_enable_pf_switch_lb(pf);
6699 }
6700
6701 /**
6702  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6703  * @veb: pointer to the VEB instance
6704  *
6705  * This is a recursive function that first builds the attached VSIs then
6706  * recurses in to build the next layer of VEB.  We track the connections
6707  * through our own index numbers because the seid's from the HW could
6708  * change across the reset.
6709  **/
6710 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6711 {
6712         struct i40e_vsi *ctl_vsi = NULL;
6713         struct i40e_pf *pf = veb->pf;
6714         int v, veb_idx;
6715         int ret;
6716
6717         /* build VSI that owns this VEB, temporarily attached to base VEB */
6718         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6719                 if (pf->vsi[v] &&
6720                     pf->vsi[v]->veb_idx == veb->idx &&
6721                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6722                         ctl_vsi = pf->vsi[v];
6723                         break;
6724                 }
6725         }
6726         if (!ctl_vsi) {
6727                 dev_info(&pf->pdev->dev,
6728                          "missing owner VSI for veb_idx %d\n", veb->idx);
6729                 ret = -ENOENT;
6730                 goto end_reconstitute;
6731         }
6732         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6733                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6734         ret = i40e_add_vsi(ctl_vsi);
6735         if (ret) {
6736                 dev_info(&pf->pdev->dev,
6737                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6738                          veb->idx, ret);
6739                 goto end_reconstitute;
6740         }
6741         i40e_vsi_reset_stats(ctl_vsi);
6742
6743         /* create the VEB in the switch and move the VSI onto the VEB */
6744         ret = i40e_add_veb(veb, ctl_vsi);
6745         if (ret)
6746                 goto end_reconstitute;
6747
6748         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6749                 veb->bridge_mode = BRIDGE_MODE_VEB;
6750         else
6751                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6752         i40e_config_bridge_mode(veb);
6753
6754         /* create the remaining VSIs attached to this VEB */
6755         for (v = 0; v < pf->num_alloc_vsi; v++) {
6756                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6757                         continue;
6758
6759                 if (pf->vsi[v]->veb_idx == veb->idx) {
6760                         struct i40e_vsi *vsi = pf->vsi[v];
6761
6762                         vsi->uplink_seid = veb->seid;
6763                         ret = i40e_add_vsi(vsi);
6764                         if (ret) {
6765                                 dev_info(&pf->pdev->dev,
6766                                          "rebuild of vsi_idx %d failed: %d\n",
6767                                          v, ret);
6768                                 goto end_reconstitute;
6769                         }
6770                         i40e_vsi_reset_stats(vsi);
6771                 }
6772         }
6773
6774         /* create any VEBs attached to this VEB - RECURSION */
6775         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6776                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6777                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6778                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6779                         if (ret)
6780                                 break;
6781                 }
6782         }
6783
6784 end_reconstitute:
6785         return ret;
6786 }
6787
6788 /**
6789  * i40e_get_capabilities - get info about the HW
6790  * @pf: the PF struct
6791  **/
6792 static int i40e_get_capabilities(struct i40e_pf *pf)
6793 {
6794         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6795         u16 data_size;
6796         int buf_len;
6797         int err;
6798
6799         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6800         do {
6801                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6802                 if (!cap_buf)
6803                         return -ENOMEM;
6804
6805                 /* this loads the data into the hw struct for us */
6806                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6807                                             &data_size,
6808                                             i40e_aqc_opc_list_func_capabilities,
6809                                             NULL);
6810                 /* data loaded, buffer no longer needed */
6811                 kfree(cap_buf);
6812
6813                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6814                         /* retry with a larger buffer */
6815                         buf_len = data_size;
6816                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6817                         dev_info(&pf->pdev->dev,
6818                                  "capability discovery failed, err %s aq_err %s\n",
6819                                  i40e_stat_str(&pf->hw, err),
6820                                  i40e_aq_str(&pf->hw,
6821                                              pf->hw.aq.asq_last_status));
6822                         return -ENODEV;
6823                 }
6824         } while (err);
6825
6826         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6827                 dev_info(&pf->pdev->dev,
6828                          "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",
6829                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6830                          pf->hw.func_caps.num_msix_vectors,
6831                          pf->hw.func_caps.num_msix_vectors_vf,
6832                          pf->hw.func_caps.fd_filters_guaranteed,
6833                          pf->hw.func_caps.fd_filters_best_effort,
6834                          pf->hw.func_caps.num_tx_qp,
6835                          pf->hw.func_caps.num_vsis);
6836
6837 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6838                        + pf->hw.func_caps.num_vfs)
6839         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6840                 dev_info(&pf->pdev->dev,
6841                          "got num_vsis %d, setting num_vsis to %d\n",
6842                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6843                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6844         }
6845
6846         return 0;
6847 }
6848
6849 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6850
6851 /**
6852  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6853  * @pf: board private structure
6854  **/
6855 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6856 {
6857         struct i40e_vsi *vsi;
6858
6859         /* quick workaround for an NVM issue that leaves a critical register
6860          * uninitialized
6861          */
6862         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6863                 static const u32 hkey[] = {
6864                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6865                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6866                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6867                         0x95b3a76d};
6868                 int i;
6869
6870                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6871                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6872         }
6873
6874         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6875                 return;
6876
6877         /* find existing VSI and see if it needs configuring */
6878         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6879
6880         /* create a new VSI if none exists */
6881         if (!vsi) {
6882                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6883                                      pf->vsi[pf->lan_vsi]->seid, 0);
6884                 if (!vsi) {
6885                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6886                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6887                         return;
6888                 }
6889         }
6890
6891         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6892 }
6893
6894 /**
6895  * i40e_fdir_teardown - release the Flow Director resources
6896  * @pf: board private structure
6897  **/
6898 static void i40e_fdir_teardown(struct i40e_pf *pf)
6899 {
6900         struct i40e_vsi *vsi;
6901
6902         i40e_fdir_filter_exit(pf);
6903         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6904         if (vsi)
6905                 i40e_vsi_release(vsi);
6906 }
6907
6908 /**
6909  * i40e_prep_for_reset - prep for the core to reset
6910  * @pf: board private structure
6911  * @lock_acquired: indicates whether or not the lock has been acquired
6912  * before this function was called.
6913  *
6914  * Close up the VFs and other things in prep for PF Reset.
6915   **/
6916 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
6917 {
6918         struct i40e_hw *hw = &pf->hw;
6919         i40e_status ret = 0;
6920         u32 v;
6921
6922         clear_bit(__I40E_RESET_INTR_RECEIVED, pf->state);
6923         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
6924                 return;
6925         if (i40e_check_asq_alive(&pf->hw))
6926                 i40e_vc_notify_reset(pf);
6927
6928         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6929
6930         /* quiesce the VSIs and their queues that are not already DOWN */
6931         /* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
6932         if (!lock_acquired)
6933                 rtnl_lock();
6934         i40e_pf_quiesce_all_vsi(pf);
6935         if (!lock_acquired)
6936                 rtnl_unlock();
6937
6938         for (v = 0; v < pf->num_alloc_vsi; v++) {
6939                 if (pf->vsi[v])
6940                         pf->vsi[v]->seid = 0;
6941         }
6942
6943         i40e_shutdown_adminq(&pf->hw);
6944
6945         /* call shutdown HMC */
6946         if (hw->hmc.hmc_obj) {
6947                 ret = i40e_shutdown_lan_hmc(hw);
6948                 if (ret)
6949                         dev_warn(&pf->pdev->dev,
6950                                  "shutdown_lan_hmc failed: %d\n", ret);
6951         }
6952 }
6953
6954 /**
6955  * i40e_send_version - update firmware with driver version
6956  * @pf: PF struct
6957  */
6958 static void i40e_send_version(struct i40e_pf *pf)
6959 {
6960         struct i40e_driver_version dv;
6961
6962         dv.major_version = DRV_VERSION_MAJOR;
6963         dv.minor_version = DRV_VERSION_MINOR;
6964         dv.build_version = DRV_VERSION_BUILD;
6965         dv.subbuild_version = 0;
6966         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6967         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6968 }
6969
6970 /**
6971  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
6972  * @pf: board private structure
6973  **/
6974 static int i40e_reset(struct i40e_pf *pf)
6975 {
6976         struct i40e_hw *hw = &pf->hw;
6977         i40e_status ret;
6978
6979         ret = i40e_pf_reset(hw);
6980         if (ret) {
6981                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6982                 set_bit(__I40E_RESET_FAILED, pf->state);
6983                 clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
6984         } else {
6985                 pf->pfr_count++;
6986         }
6987         return ret;
6988 }
6989
6990 /**
6991  * i40e_rebuild - rebuild using a saved config
6992  * @pf: board private structure
6993  * @reinit: if the Main VSI needs to re-initialized.
6994  * @lock_acquired: indicates whether or not the lock has been acquired
6995  * before this function was called.
6996  **/
6997 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
6998 {
6999         struct i40e_hw *hw = &pf->hw;
7000         u8 set_fc_aq_fail = 0;
7001         i40e_status ret;
7002         u32 val;
7003         int v;
7004
7005         if (test_bit(__I40E_VSI_DOWN, pf->state))
7006                 goto clear_recovery;
7007         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
7008
7009         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
7010         ret = i40e_init_adminq(&pf->hw);
7011         if (ret) {
7012                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
7013                          i40e_stat_str(&pf->hw, ret),
7014                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7015                 goto clear_recovery;
7016         }
7017
7018         /* re-verify the eeprom if we just had an EMP reset */
7019         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
7020                 i40e_verify_eeprom(pf);
7021
7022         i40e_clear_pxe_mode(hw);
7023         ret = i40e_get_capabilities(pf);
7024         if (ret)
7025                 goto end_core_reset;
7026
7027         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7028                                 hw->func_caps.num_rx_qp, 0, 0);
7029         if (ret) {
7030                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
7031                 goto end_core_reset;
7032         }
7033         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7034         if (ret) {
7035                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
7036                 goto end_core_reset;
7037         }
7038
7039 #ifdef CONFIG_I40E_DCB
7040         ret = i40e_init_pf_dcb(pf);
7041         if (ret) {
7042                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
7043                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
7044                 /* Continue without DCB enabled */
7045         }
7046 #endif /* CONFIG_I40E_DCB */
7047         /* do basic switch setup */
7048         if (!lock_acquired)
7049                 rtnl_lock();
7050         ret = i40e_setup_pf_switch(pf, reinit);
7051         if (ret)
7052                 goto end_unlock;
7053
7054         /* The driver only wants link up/down and module qualification
7055          * reports from firmware.  Note the negative logic.
7056          */
7057         ret = i40e_aq_set_phy_int_mask(&pf->hw,
7058                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
7059                                          I40E_AQ_EVENT_MEDIA_NA |
7060                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
7061         if (ret)
7062                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
7063                          i40e_stat_str(&pf->hw, ret),
7064                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7065
7066         /* make sure our flow control settings are restored */
7067         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
7068         if (ret)
7069                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
7070                         i40e_stat_str(&pf->hw, ret),
7071                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7072
7073         /* Rebuild the VSIs and VEBs that existed before reset.
7074          * They are still in our local switch element arrays, so only
7075          * need to rebuild the switch model in the HW.
7076          *
7077          * If there were VEBs but the reconstitution failed, we'll try
7078          * try to recover minimal use by getting the basic PF VSI working.
7079          */
7080         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
7081                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
7082                 /* find the one VEB connected to the MAC, and find orphans */
7083                 for (v = 0; v < I40E_MAX_VEB; v++) {
7084                         if (!pf->veb[v])
7085                                 continue;
7086
7087                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
7088                             pf->veb[v]->uplink_seid == 0) {
7089                                 ret = i40e_reconstitute_veb(pf->veb[v]);
7090
7091                                 if (!ret)
7092                                         continue;
7093
7094                                 /* If Main VEB failed, we're in deep doodoo,
7095                                  * so give up rebuilding the switch and set up
7096                                  * for minimal rebuild of PF VSI.
7097                                  * If orphan failed, we'll report the error
7098                                  * but try to keep going.
7099                                  */
7100                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
7101                                         dev_info(&pf->pdev->dev,
7102                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
7103                                                  ret);
7104                                         pf->vsi[pf->lan_vsi]->uplink_seid
7105                                                                 = pf->mac_seid;
7106                                         break;
7107                                 } else if (pf->veb[v]->uplink_seid == 0) {
7108                                         dev_info(&pf->pdev->dev,
7109                                                  "rebuild of orphan VEB failed: %d\n",
7110                                                  ret);
7111                                 }
7112                         }
7113                 }
7114         }
7115
7116         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
7117                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
7118                 /* no VEB, so rebuild only the Main VSI */
7119                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
7120                 if (ret) {
7121                         dev_info(&pf->pdev->dev,
7122                                  "rebuild of Main VSI failed: %d\n", ret);
7123                         goto end_unlock;
7124                 }
7125         }
7126
7127         /* Reconfigure hardware for allowing smaller MSS in the case
7128          * of TSO, so that we avoid the MDD being fired and causing
7129          * a reset in the case of small MSS+TSO.
7130          */
7131 #define I40E_REG_MSS          0x000E64DC
7132 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
7133 #define I40E_64BYTE_MSS       0x400000
7134         val = rd32(hw, I40E_REG_MSS);
7135         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
7136                 val &= ~I40E_REG_MSS_MIN_MASK;
7137                 val |= I40E_64BYTE_MSS;
7138                 wr32(hw, I40E_REG_MSS, val);
7139         }
7140
7141         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
7142                 msleep(75);
7143                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
7144                 if (ret)
7145                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
7146                                  i40e_stat_str(&pf->hw, ret),
7147                                  i40e_aq_str(&pf->hw,
7148                                              pf->hw.aq.asq_last_status));
7149         }
7150         /* reinit the misc interrupt */
7151         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7152                 ret = i40e_setup_misc_vector(pf);
7153
7154         /* Add a filter to drop all Flow control frames from any VSI from being
7155          * transmitted. By doing so we stop a malicious VF from sending out
7156          * PAUSE or PFC frames and potentially controlling traffic for other
7157          * PF/VF VSIs.
7158          * The FW can still send Flow control frames if enabled.
7159          */
7160         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7161                                                        pf->main_vsi_seid);
7162
7163         /* restart the VSIs that were rebuilt and running before the reset */
7164         i40e_pf_unquiesce_all_vsi(pf);
7165
7166         /* Release the RTNL lock before we start resetting VFs */
7167         if (!lock_acquired)
7168                 rtnl_unlock();
7169
7170         i40e_reset_all_vfs(pf, true);
7171
7172         /* tell the firmware that we're starting */
7173         i40e_send_version(pf);
7174
7175         /* We've already released the lock, so don't do it again */
7176         goto end_core_reset;
7177
7178 end_unlock:
7179         if (!lock_acquired)
7180                 rtnl_unlock();
7181 end_core_reset:
7182         clear_bit(__I40E_RESET_FAILED, pf->state);
7183 clear_recovery:
7184         clear_bit(__I40E_RESET_RECOVERY_PENDING, pf->state);
7185 }
7186
7187 /**
7188  * i40e_reset_and_rebuild - reset and rebuild using a saved config
7189  * @pf: board private structure
7190  * @reinit: if the Main VSI needs to re-initialized.
7191  * @lock_acquired: indicates whether or not the lock has been acquired
7192  * before this function was called.
7193  **/
7194 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
7195                                    bool lock_acquired)
7196 {
7197         int ret;
7198         /* Now we wait for GRST to settle out.
7199          * We don't have to delete the VEBs or VSIs from the hw switch
7200          * because the reset will make them disappear.
7201          */
7202         ret = i40e_reset(pf);
7203         if (!ret)
7204                 i40e_rebuild(pf, reinit, lock_acquired);
7205 }
7206
7207 /**
7208  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7209  * @pf: board private structure
7210  *
7211  * Close up the VFs and other things in prep for a Core Reset,
7212  * then get ready to rebuild the world.
7213  * @lock_acquired: indicates whether or not the lock has been acquired
7214  * before this function was called.
7215  **/
7216 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
7217 {
7218         i40e_prep_for_reset(pf, lock_acquired);
7219         i40e_reset_and_rebuild(pf, false, lock_acquired);
7220 }
7221
7222 /**
7223  * i40e_handle_mdd_event
7224  * @pf: pointer to the PF structure
7225  *
7226  * Called from the MDD irq handler to identify possibly malicious vfs
7227  **/
7228 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7229 {
7230         struct i40e_hw *hw = &pf->hw;
7231         bool mdd_detected = false;
7232         bool pf_mdd_detected = false;
7233         struct i40e_vf *vf;
7234         u32 reg;
7235         int i;
7236
7237         if (!test_bit(__I40E_MDD_EVENT_PENDING, pf->state))
7238                 return;
7239
7240         /* find what triggered the MDD event */
7241         reg = rd32(hw, I40E_GL_MDET_TX);
7242         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7243                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7244                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7245                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7246                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7247                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7248                                 I40E_GL_MDET_TX_EVENT_SHIFT;
7249                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7250                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7251                                 pf->hw.func_caps.base_queue;
7252                 if (netif_msg_tx_err(pf))
7253                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7254                                  event, queue, pf_num, vf_num);
7255                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7256                 mdd_detected = true;
7257         }
7258         reg = rd32(hw, I40E_GL_MDET_RX);
7259         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7260                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7261                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7262                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7263                                 I40E_GL_MDET_RX_EVENT_SHIFT;
7264                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7265                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7266                                 pf->hw.func_caps.base_queue;
7267                 if (netif_msg_rx_err(pf))
7268                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7269                                  event, queue, func);
7270                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7271                 mdd_detected = true;
7272         }
7273
7274         if (mdd_detected) {
7275                 reg = rd32(hw, I40E_PF_MDET_TX);
7276                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7277                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7278                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7279                         pf_mdd_detected = true;
7280                 }
7281                 reg = rd32(hw, I40E_PF_MDET_RX);
7282                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7283                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7284                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7285                         pf_mdd_detected = true;
7286                 }
7287                 /* Queue belongs to the PF, initiate a reset */
7288                 if (pf_mdd_detected) {
7289                         set_bit(__I40E_PF_RESET_REQUESTED, pf->state);
7290                         i40e_service_event_schedule(pf);
7291                 }
7292         }
7293
7294         /* see if one of the VFs needs its hand slapped */
7295         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7296                 vf = &(pf->vf[i]);
7297                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7298                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7299                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7300                         vf->num_mdd_events++;
7301                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7302                                  i);
7303                 }
7304
7305                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7306                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7307                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7308                         vf->num_mdd_events++;
7309                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7310                                  i);
7311                 }
7312
7313                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7314                         dev_info(&pf->pdev->dev,
7315                                  "Too many MDD events on VF %d, disabled\n", i);
7316                         dev_info(&pf->pdev->dev,
7317                                  "Use PF Control I/F to re-enable the VF\n");
7318                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
7319                 }
7320         }
7321
7322         /* re-enable mdd interrupt cause */
7323         clear_bit(__I40E_MDD_EVENT_PENDING, pf->state);
7324         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7325         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7326         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7327         i40e_flush(hw);
7328 }
7329
7330 /**
7331  * i40e_sync_udp_filters - Trigger a sync event for existing UDP filters
7332  * @pf: board private structure
7333  **/
7334 static void i40e_sync_udp_filters(struct i40e_pf *pf)
7335 {
7336         int i;
7337
7338         /* loop through and set pending bit for all active UDP filters */
7339         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7340                 if (pf->udp_ports[i].port)
7341                         pf->pending_udp_bitmap |= BIT_ULL(i);
7342         }
7343
7344         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
7345 }
7346
7347 /**
7348  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7349  * @pf: board private structure
7350  **/
7351 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7352 {
7353         struct i40e_hw *hw = &pf->hw;
7354         i40e_status ret;
7355         u16 port;
7356         int i;
7357
7358         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7359                 return;
7360
7361         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7362
7363         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7364                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7365                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7366                         port = pf->udp_ports[i].port;
7367                         if (port)
7368                                 ret = i40e_aq_add_udp_tunnel(hw, port,
7369                                                         pf->udp_ports[i].type,
7370                                                         NULL, NULL);
7371                         else
7372                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7373
7374                         if (ret) {
7375                                 dev_dbg(&pf->pdev->dev,
7376                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
7377                                         pf->udp_ports[i].type ? "vxlan" : "geneve",
7378                                         port ? "add" : "delete",
7379                                         port, i,
7380                                         i40e_stat_str(&pf->hw, ret),
7381                                         i40e_aq_str(&pf->hw,
7382                                                     pf->hw.aq.asq_last_status));
7383                                 pf->udp_ports[i].port = 0;
7384                         }
7385                 }
7386         }
7387 }
7388
7389 /**
7390  * i40e_service_task - Run the driver's async subtasks
7391  * @work: pointer to work_struct containing our data
7392  **/
7393 static void i40e_service_task(struct work_struct *work)
7394 {
7395         struct i40e_pf *pf = container_of(work,
7396                                           struct i40e_pf,
7397                                           service_task);
7398         unsigned long start_time = jiffies;
7399
7400         /* don't bother with service tasks if a reset is in progress */
7401         if (test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state))
7402                 return;
7403
7404         if (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
7405                 return;
7406
7407         i40e_detect_recover_hung(pf);
7408         i40e_sync_filters_subtask(pf);
7409         i40e_reset_subtask(pf);
7410         i40e_handle_mdd_event(pf);
7411         i40e_vc_process_vflr_event(pf);
7412         i40e_watchdog_subtask(pf);
7413         i40e_fdir_reinit_subtask(pf);
7414         if (pf->flags & I40E_FLAG_CLIENT_RESET) {
7415                 /* Client subtask will reopen next time through. */
7416                 i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
7417                 pf->flags &= ~I40E_FLAG_CLIENT_RESET;
7418         } else {
7419                 i40e_client_subtask(pf);
7420                 if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
7421                         i40e_notify_client_of_l2_param_changes(
7422                                                         pf->vsi[pf->lan_vsi]);
7423                         pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
7424                 }
7425         }
7426         i40e_sync_filters_subtask(pf);
7427         i40e_sync_udp_filters_subtask(pf);
7428         i40e_clean_adminq_subtask(pf);
7429
7430         /* flush memory to make sure state is correct before next watchdog */
7431         smp_mb__before_atomic();
7432         clear_bit(__I40E_SERVICE_SCHED, pf->state);
7433
7434         /* If the tasks have taken longer than one timer cycle or there
7435          * is more work to be done, reschedule the service task now
7436          * rather than wait for the timer to tick again.
7437          */
7438         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7439             test_bit(__I40E_ADMINQ_EVENT_PENDING, pf->state)             ||
7440             test_bit(__I40E_MDD_EVENT_PENDING, pf->state)                ||
7441             test_bit(__I40E_VFLR_EVENT_PENDING, pf->state))
7442                 i40e_service_event_schedule(pf);
7443 }
7444
7445 /**
7446  * i40e_service_timer - timer callback
7447  * @data: pointer to PF struct
7448  **/
7449 static void i40e_service_timer(unsigned long data)
7450 {
7451         struct i40e_pf *pf = (struct i40e_pf *)data;
7452
7453         mod_timer(&pf->service_timer,
7454                   round_jiffies(jiffies + pf->service_timer_period));
7455         i40e_service_event_schedule(pf);
7456 }
7457
7458 /**
7459  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7460  * @vsi: the VSI being configured
7461  **/
7462 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7463 {
7464         struct i40e_pf *pf = vsi->back;
7465
7466         switch (vsi->type) {
7467         case I40E_VSI_MAIN:
7468                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7469                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7470                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7471                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7472                         vsi->num_q_vectors = pf->num_lan_msix;
7473                 else
7474                         vsi->num_q_vectors = 1;
7475
7476                 break;
7477
7478         case I40E_VSI_FDIR:
7479                 vsi->alloc_queue_pairs = 1;
7480                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7481                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7482                 vsi->num_q_vectors = pf->num_fdsb_msix;
7483                 break;
7484
7485         case I40E_VSI_VMDQ2:
7486                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7487                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7488                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7489                 vsi->num_q_vectors = pf->num_vmdq_msix;
7490                 break;
7491
7492         case I40E_VSI_SRIOV:
7493                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7494                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7495                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7496                 break;
7497
7498         default:
7499                 WARN_ON(1);
7500                 return -ENODATA;
7501         }
7502
7503         return 0;
7504 }
7505
7506 /**
7507  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7508  * @type: VSI pointer
7509  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7510  *
7511  * On error: returns error code (negative)
7512  * On success: returns 0
7513  **/
7514 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7515 {
7516         int size;
7517         int ret = 0;
7518
7519         /* allocate memory for both Tx and Rx ring pointers */
7520         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7521         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7522         if (!vsi->tx_rings)
7523                 return -ENOMEM;
7524         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7525
7526         if (alloc_qvectors) {
7527                 /* allocate memory for q_vector pointers */
7528                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7529                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7530                 if (!vsi->q_vectors) {
7531                         ret = -ENOMEM;
7532                         goto err_vectors;
7533                 }
7534         }
7535         return ret;
7536
7537 err_vectors:
7538         kfree(vsi->tx_rings);
7539         return ret;
7540 }
7541
7542 /**
7543  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7544  * @pf: board private structure
7545  * @type: type of VSI
7546  *
7547  * On error: returns error code (negative)
7548  * On success: returns vsi index in PF (positive)
7549  **/
7550 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7551 {
7552         int ret = -ENODEV;
7553         struct i40e_vsi *vsi;
7554         int vsi_idx;
7555         int i;
7556
7557         /* Need to protect the allocation of the VSIs at the PF level */
7558         mutex_lock(&pf->switch_mutex);
7559
7560         /* VSI list may be fragmented if VSI creation/destruction has
7561          * been happening.  We can afford to do a quick scan to look
7562          * for any free VSIs in the list.
7563          *
7564          * find next empty vsi slot, looping back around if necessary
7565          */
7566         i = pf->next_vsi;
7567         while (i < pf->num_alloc_vsi && pf->vsi[i])
7568                 i++;
7569         if (i >= pf->num_alloc_vsi) {
7570                 i = 0;
7571                 while (i < pf->next_vsi && pf->vsi[i])
7572                         i++;
7573         }
7574
7575         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7576                 vsi_idx = i;             /* Found one! */
7577         } else {
7578                 ret = -ENODEV;
7579                 goto unlock_pf;  /* out of VSI slots! */
7580         }
7581         pf->next_vsi = ++i;
7582
7583         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7584         if (!vsi) {
7585                 ret = -ENOMEM;
7586                 goto unlock_pf;
7587         }
7588         vsi->type = type;
7589         vsi->back = pf;
7590         set_bit(__I40E_VSI_DOWN, vsi->state);
7591         vsi->flags = 0;
7592         vsi->idx = vsi_idx;
7593         vsi->int_rate_limit = 0;
7594         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7595                                 pf->rss_table_size : 64;
7596         vsi->netdev_registered = false;
7597         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7598         hash_init(vsi->mac_filter_hash);
7599         vsi->irqs_ready = false;
7600
7601         ret = i40e_set_num_rings_in_vsi(vsi);
7602         if (ret)
7603                 goto err_rings;
7604
7605         ret = i40e_vsi_alloc_arrays(vsi, true);
7606         if (ret)
7607                 goto err_rings;
7608
7609         /* Setup default MSIX irq handler for VSI */
7610         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7611
7612         /* Initialize VSI lock */
7613         spin_lock_init(&vsi->mac_filter_hash_lock);
7614         pf->vsi[vsi_idx] = vsi;
7615         ret = vsi_idx;
7616         goto unlock_pf;
7617
7618 err_rings:
7619         pf->next_vsi = i - 1;
7620         kfree(vsi);
7621 unlock_pf:
7622         mutex_unlock(&pf->switch_mutex);
7623         return ret;
7624 }
7625
7626 /**
7627  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7628  * @type: VSI pointer
7629  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7630  *
7631  * On error: returns error code (negative)
7632  * On success: returns 0
7633  **/
7634 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7635 {
7636         /* free the ring and vector containers */
7637         if (free_qvectors) {
7638                 kfree(vsi->q_vectors);
7639                 vsi->q_vectors = NULL;
7640         }
7641         kfree(vsi->tx_rings);
7642         vsi->tx_rings = NULL;
7643         vsi->rx_rings = NULL;
7644 }
7645
7646 /**
7647  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7648  * and lookup table
7649  * @vsi: Pointer to VSI structure
7650  */
7651 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7652 {
7653         if (!vsi)
7654                 return;
7655
7656         kfree(vsi->rss_hkey_user);
7657         vsi->rss_hkey_user = NULL;
7658
7659         kfree(vsi->rss_lut_user);
7660         vsi->rss_lut_user = NULL;
7661 }
7662
7663 /**
7664  * i40e_vsi_clear - Deallocate the VSI provided
7665  * @vsi: the VSI being un-configured
7666  **/
7667 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7668 {
7669         struct i40e_pf *pf;
7670
7671         if (!vsi)
7672                 return 0;
7673
7674         if (!vsi->back)
7675                 goto free_vsi;
7676         pf = vsi->back;
7677
7678         mutex_lock(&pf->switch_mutex);
7679         if (!pf->vsi[vsi->idx]) {
7680                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7681                         vsi->idx, vsi->idx, vsi, vsi->type);
7682                 goto unlock_vsi;
7683         }
7684
7685         if (pf->vsi[vsi->idx] != vsi) {
7686                 dev_err(&pf->pdev->dev,
7687                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7688                         pf->vsi[vsi->idx]->idx,
7689                         pf->vsi[vsi->idx],
7690                         pf->vsi[vsi->idx]->type,
7691                         vsi->idx, vsi, vsi->type);
7692                 goto unlock_vsi;
7693         }
7694
7695         /* updates the PF for this cleared vsi */
7696         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7697         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7698
7699         i40e_vsi_free_arrays(vsi, true);
7700         i40e_clear_rss_config_user(vsi);
7701
7702         pf->vsi[vsi->idx] = NULL;
7703         if (vsi->idx < pf->next_vsi)
7704                 pf->next_vsi = vsi->idx;
7705
7706 unlock_vsi:
7707         mutex_unlock(&pf->switch_mutex);
7708 free_vsi:
7709         kfree(vsi);
7710
7711         return 0;
7712 }
7713
7714 /**
7715  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7716  * @vsi: the VSI being cleaned
7717  **/
7718 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7719 {
7720         int i;
7721
7722         if (vsi->tx_rings && vsi->tx_rings[0]) {
7723                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7724                         kfree_rcu(vsi->tx_rings[i], rcu);
7725                         vsi->tx_rings[i] = NULL;
7726                         vsi->rx_rings[i] = NULL;
7727                 }
7728         }
7729 }
7730
7731 /**
7732  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7733  * @vsi: the VSI being configured
7734  **/
7735 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7736 {
7737         struct i40e_ring *tx_ring, *rx_ring;
7738         struct i40e_pf *pf = vsi->back;
7739         int i;
7740
7741         /* Set basic values in the rings to be used later during open() */
7742         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7743                 /* allocate space for both Tx and Rx in one shot */
7744                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7745                 if (!tx_ring)
7746                         goto err_out;
7747
7748                 tx_ring->queue_index = i;
7749                 tx_ring->reg_idx = vsi->base_queue + i;
7750                 tx_ring->ring_active = false;
7751                 tx_ring->vsi = vsi;
7752                 tx_ring->netdev = vsi->netdev;
7753                 tx_ring->dev = &pf->pdev->dev;
7754                 tx_ring->count = vsi->num_desc;
7755                 tx_ring->size = 0;
7756                 tx_ring->dcb_tc = 0;
7757                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7758                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7759                 tx_ring->tx_itr_setting = pf->tx_itr_default;
7760                 vsi->tx_rings[i] = tx_ring;
7761
7762                 rx_ring = &tx_ring[1];
7763                 rx_ring->queue_index = i;
7764                 rx_ring->reg_idx = vsi->base_queue + i;
7765                 rx_ring->ring_active = false;
7766                 rx_ring->vsi = vsi;
7767                 rx_ring->netdev = vsi->netdev;
7768                 rx_ring->dev = &pf->pdev->dev;
7769                 rx_ring->count = vsi->num_desc;
7770                 rx_ring->size = 0;
7771                 rx_ring->dcb_tc = 0;
7772                 rx_ring->rx_itr_setting = pf->rx_itr_default;
7773                 vsi->rx_rings[i] = rx_ring;
7774         }
7775
7776         return 0;
7777
7778 err_out:
7779         i40e_vsi_clear_rings(vsi);
7780         return -ENOMEM;
7781 }
7782
7783 /**
7784  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7785  * @pf: board private structure
7786  * @vectors: the number of MSI-X vectors to request
7787  *
7788  * Returns the number of vectors reserved, or error
7789  **/
7790 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7791 {
7792         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7793                                         I40E_MIN_MSIX, vectors);
7794         if (vectors < 0) {
7795                 dev_info(&pf->pdev->dev,
7796                          "MSI-X vector reservation failed: %d\n", vectors);
7797                 vectors = 0;
7798         }
7799
7800         return vectors;
7801 }
7802
7803 /**
7804  * i40e_init_msix - Setup the MSIX capability
7805  * @pf: board private structure
7806  *
7807  * Work with the OS to set up the MSIX vectors needed.
7808  *
7809  * Returns the number of vectors reserved or negative on failure
7810  **/
7811 static int i40e_init_msix(struct i40e_pf *pf)
7812 {
7813         struct i40e_hw *hw = &pf->hw;
7814         int cpus, extra_vectors;
7815         int vectors_left;
7816         int v_budget, i;
7817         int v_actual;
7818         int iwarp_requested = 0;
7819
7820         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7821                 return -ENODEV;
7822
7823         /* The number of vectors we'll request will be comprised of:
7824          *   - Add 1 for "other" cause for Admin Queue events, etc.
7825          *   - The number of LAN queue pairs
7826          *      - Queues being used for RSS.
7827          *              We don't need as many as max_rss_size vectors.
7828          *              use rss_size instead in the calculation since that
7829          *              is governed by number of cpus in the system.
7830          *      - assumes symmetric Tx/Rx pairing
7831          *   - The number of VMDq pairs
7832          *   - The CPU count within the NUMA node if iWARP is enabled
7833          * Once we count this up, try the request.
7834          *
7835          * If we can't get what we want, we'll simplify to nearly nothing
7836          * and try again.  If that still fails, we punt.
7837          */
7838         vectors_left = hw->func_caps.num_msix_vectors;
7839         v_budget = 0;
7840
7841         /* reserve one vector for miscellaneous handler */
7842         if (vectors_left) {
7843                 v_budget++;
7844                 vectors_left--;
7845         }
7846
7847         /* reserve some vectors for the main PF traffic queues. Initially we
7848          * only reserve at most 50% of the available vectors, in the case that
7849          * the number of online CPUs is large. This ensures that we can enable
7850          * extra features as well. Once we've enabled the other features, we
7851          * will use any remaining vectors to reach as close as we can to the
7852          * number of online CPUs.
7853          */
7854         cpus = num_online_cpus();
7855         pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
7856         vectors_left -= pf->num_lan_msix;
7857
7858         /* reserve one vector for sideband flow director */
7859         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7860                 if (vectors_left) {
7861                         pf->num_fdsb_msix = 1;
7862                         v_budget++;
7863                         vectors_left--;
7864                 } else {
7865                         pf->num_fdsb_msix = 0;
7866                 }
7867         }
7868
7869         /* can we reserve enough for iWARP? */
7870         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7871                 iwarp_requested = pf->num_iwarp_msix;
7872
7873                 if (!vectors_left)
7874                         pf->num_iwarp_msix = 0;
7875                 else if (vectors_left < pf->num_iwarp_msix)
7876                         pf->num_iwarp_msix = 1;
7877                 v_budget += pf->num_iwarp_msix;
7878                 vectors_left -= pf->num_iwarp_msix;
7879         }
7880
7881         /* any vectors left over go for VMDq support */
7882         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7883                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7884                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7885
7886                 if (!vectors_left) {
7887                         pf->num_vmdq_msix = 0;
7888                         pf->num_vmdq_qps = 0;
7889                 } else {
7890                         /* if we're short on vectors for what's desired, we limit
7891                          * the queues per vmdq.  If this is still more than are
7892                          * available, the user will need to change the number of
7893                          * queues/vectors used by the PF later with the ethtool
7894                          * channels command
7895                          */
7896                         if (vmdq_vecs < vmdq_vecs_wanted)
7897                                 pf->num_vmdq_qps = 1;
7898                         pf->num_vmdq_msix = pf->num_vmdq_qps;
7899
7900                         v_budget += vmdq_vecs;
7901                         vectors_left -= vmdq_vecs;
7902                 }
7903         }
7904
7905         /* On systems with a large number of SMP cores, we previously limited
7906          * the number of vectors for num_lan_msix to be at most 50% of the
7907          * available vectors, to allow for other features. Now, we add back
7908          * the remaining vectors. However, we ensure that the total
7909          * num_lan_msix will not exceed num_online_cpus(). To do this, we
7910          * calculate the number of vectors we can add without going over the
7911          * cap of CPUs. For systems with a small number of CPUs this will be
7912          * zero.
7913          */
7914         extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
7915         pf->num_lan_msix += extra_vectors;
7916         vectors_left -= extra_vectors;
7917
7918         WARN(vectors_left < 0,
7919              "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
7920
7921         v_budget += pf->num_lan_msix;
7922         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7923                                    GFP_KERNEL);
7924         if (!pf->msix_entries)
7925                 return -ENOMEM;
7926
7927         for (i = 0; i < v_budget; i++)
7928                 pf->msix_entries[i].entry = i;
7929         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7930
7931         if (v_actual < I40E_MIN_MSIX) {
7932                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7933                 kfree(pf->msix_entries);
7934                 pf->msix_entries = NULL;
7935                 pci_disable_msix(pf->pdev);
7936                 return -ENODEV;
7937
7938         } else if (v_actual == I40E_MIN_MSIX) {
7939                 /* Adjust for minimal MSIX use */
7940                 pf->num_vmdq_vsis = 0;
7941                 pf->num_vmdq_qps = 0;
7942                 pf->num_lan_qps = 1;
7943                 pf->num_lan_msix = 1;
7944
7945         } else if (!vectors_left) {
7946                 /* If we have limited resources, we will start with no vectors
7947                  * for the special features and then allocate vectors to some
7948                  * of these features based on the policy and at the end disable
7949                  * the features that did not get any vectors.
7950                  */
7951                 int vec;
7952
7953                 dev_info(&pf->pdev->dev,
7954                          "MSI-X vector limit reached, attempting to redistribute vectors\n");
7955                 /* reserve the misc vector */
7956                 vec = v_actual - 1;
7957
7958                 /* Scale vector usage down */
7959                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7960                 pf->num_vmdq_vsis = 1;
7961                 pf->num_vmdq_qps = 1;
7962
7963                 /* partition out the remaining vectors */
7964                 switch (vec) {
7965                 case 2:
7966                         pf->num_lan_msix = 1;
7967                         break;
7968                 case 3:
7969                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7970                                 pf->num_lan_msix = 1;
7971                                 pf->num_iwarp_msix = 1;
7972                         } else {
7973                                 pf->num_lan_msix = 2;
7974                         }
7975                         break;
7976                 default:
7977                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7978                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7979                                                  iwarp_requested);
7980                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7981                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7982                         } else {
7983                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7984                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7985                         }
7986                         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7987                                 pf->num_fdsb_msix = 1;
7988                                 vec--;
7989                         }
7990                         pf->num_lan_msix = min_t(int,
7991                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7992                                                               pf->num_lan_msix);
7993                         pf->num_lan_qps = pf->num_lan_msix;
7994                         break;
7995                 }
7996         }
7997
7998         if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
7999             (pf->num_fdsb_msix == 0)) {
8000                 dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
8001                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8002         }
8003         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
8004             (pf->num_vmdq_msix == 0)) {
8005                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
8006                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
8007         }
8008
8009         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
8010             (pf->num_iwarp_msix == 0)) {
8011                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
8012                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
8013         }
8014         i40e_debug(&pf->hw, I40E_DEBUG_INIT,
8015                    "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
8016                    pf->num_lan_msix,
8017                    pf->num_vmdq_msix * pf->num_vmdq_vsis,
8018                    pf->num_fdsb_msix,
8019                    pf->num_iwarp_msix);
8020
8021         return v_actual;
8022 }
8023
8024 /**
8025  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
8026  * @vsi: the VSI being configured
8027  * @v_idx: index of the vector in the vsi struct
8028  * @cpu: cpu to be used on affinity_mask
8029  *
8030  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
8031  **/
8032 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
8033 {
8034         struct i40e_q_vector *q_vector;
8035
8036         /* allocate q_vector */
8037         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
8038         if (!q_vector)
8039                 return -ENOMEM;
8040
8041         q_vector->vsi = vsi;
8042         q_vector->v_idx = v_idx;
8043         cpumask_set_cpu(cpu, &q_vector->affinity_mask);
8044
8045         if (vsi->netdev)
8046                 netif_napi_add(vsi->netdev, &q_vector->napi,
8047                                i40e_napi_poll, NAPI_POLL_WEIGHT);
8048
8049         q_vector->rx.latency_range = I40E_LOW_LATENCY;
8050         q_vector->tx.latency_range = I40E_LOW_LATENCY;
8051
8052         /* tie q_vector and vsi together */
8053         vsi->q_vectors[v_idx] = q_vector;
8054
8055         return 0;
8056 }
8057
8058 /**
8059  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
8060  * @vsi: the VSI being configured
8061  *
8062  * We allocate one q_vector per queue interrupt.  If allocation fails we
8063  * return -ENOMEM.
8064  **/
8065 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
8066 {
8067         struct i40e_pf *pf = vsi->back;
8068         int err, v_idx, num_q_vectors, current_cpu;
8069
8070         /* if not MSIX, give the one vector only to the LAN VSI */
8071         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8072                 num_q_vectors = vsi->num_q_vectors;
8073         else if (vsi == pf->vsi[pf->lan_vsi])
8074                 num_q_vectors = 1;
8075         else
8076                 return -EINVAL;
8077
8078         current_cpu = cpumask_first(cpu_online_mask);
8079
8080         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
8081                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
8082                 if (err)
8083                         goto err_out;
8084                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
8085                 if (unlikely(current_cpu >= nr_cpu_ids))
8086                         current_cpu = cpumask_first(cpu_online_mask);
8087         }
8088
8089         return 0;
8090
8091 err_out:
8092         while (v_idx--)
8093                 i40e_free_q_vector(vsi, v_idx);
8094
8095         return err;
8096 }
8097
8098 /**
8099  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
8100  * @pf: board private structure to initialize
8101  **/
8102 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
8103 {
8104         int vectors = 0;
8105         ssize_t size;
8106
8107         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8108                 vectors = i40e_init_msix(pf);
8109                 if (vectors < 0) {
8110                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
8111                                        I40E_FLAG_IWARP_ENABLED  |
8112                                        I40E_FLAG_RSS_ENABLED    |
8113                                        I40E_FLAG_DCB_CAPABLE    |
8114                                        I40E_FLAG_DCB_ENABLED    |
8115                                        I40E_FLAG_SRIOV_ENABLED  |
8116                                        I40E_FLAG_FD_SB_ENABLED  |
8117                                        I40E_FLAG_FD_ATR_ENABLED |
8118                                        I40E_FLAG_VMDQ_ENABLED);
8119
8120                         /* rework the queue expectations without MSIX */
8121                         i40e_determine_queue_usage(pf);
8122                 }
8123         }
8124
8125         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
8126             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
8127                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
8128                 vectors = pci_enable_msi(pf->pdev);
8129                 if (vectors < 0) {
8130                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
8131                                  vectors);
8132                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
8133                 }
8134                 vectors = 1;  /* one MSI or Legacy vector */
8135         }
8136
8137         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
8138                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
8139
8140         /* set up vector assignment tracking */
8141         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
8142         pf->irq_pile = kzalloc(size, GFP_KERNEL);
8143         if (!pf->irq_pile) {
8144                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
8145                 return -ENOMEM;
8146         }
8147         pf->irq_pile->num_entries = vectors;
8148         pf->irq_pile->search_hint = 0;
8149
8150         /* track first vector for misc interrupts, ignore return */
8151         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
8152
8153         return 0;
8154 }
8155
8156 /**
8157  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
8158  * @pf: board private structure
8159  *
8160  * This sets up the handler for MSIX 0, which is used to manage the
8161  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
8162  * when in MSI or Legacy interrupt mode.
8163  **/
8164 static int i40e_setup_misc_vector(struct i40e_pf *pf)
8165 {
8166         struct i40e_hw *hw = &pf->hw;
8167         int err = 0;
8168
8169         /* Only request the irq if this is the first time through, and
8170          * not when we're rebuilding after a Reset
8171          */
8172         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, pf->state)) {
8173                 err = request_irq(pf->msix_entries[0].vector,
8174                                   i40e_intr, 0, pf->int_name, pf);
8175                 if (err) {
8176                         dev_info(&pf->pdev->dev,
8177                                  "request_irq for %s failed: %d\n",
8178                                  pf->int_name, err);
8179                         return -EFAULT;
8180                 }
8181         }
8182
8183         i40e_enable_misc_int_causes(pf);
8184
8185         /* associate no queues to the misc vector */
8186         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
8187         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
8188
8189         i40e_flush(hw);
8190
8191         i40e_irq_dynamic_enable_icr0(pf, true);
8192
8193         return err;
8194 }
8195
8196 /**
8197  * i40e_config_rss_aq - Prepare for RSS using AQ commands
8198  * @vsi: vsi structure
8199  * @seed: RSS hash seed
8200  **/
8201 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8202                               u8 *lut, u16 lut_size)
8203 {
8204         struct i40e_pf *pf = vsi->back;
8205         struct i40e_hw *hw = &pf->hw;
8206         int ret = 0;
8207
8208         if (seed) {
8209                 struct i40e_aqc_get_set_rss_key_data *seed_dw =
8210                         (struct i40e_aqc_get_set_rss_key_data *)seed;
8211                 ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8212                 if (ret) {
8213                         dev_info(&pf->pdev->dev,
8214                                  "Cannot set RSS key, err %s aq_err %s\n",
8215                                  i40e_stat_str(hw, ret),
8216                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8217                         return ret;
8218                 }
8219         }
8220         if (lut) {
8221                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8222
8223                 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8224                 if (ret) {
8225                         dev_info(&pf->pdev->dev,
8226                                  "Cannot set RSS lut, err %s aq_err %s\n",
8227                                  i40e_stat_str(hw, ret),
8228                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8229                         return ret;
8230                 }
8231         }
8232         return ret;
8233 }
8234
8235 /**
8236  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8237  * @vsi: Pointer to vsi structure
8238  * @seed: Buffter to store the hash keys
8239  * @lut: Buffer to store the lookup table entries
8240  * @lut_size: Size of buffer to store the lookup table entries
8241  *
8242  * Return 0 on success, negative on failure
8243  */
8244 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8245                            u8 *lut, u16 lut_size)
8246 {
8247         struct i40e_pf *pf = vsi->back;
8248         struct i40e_hw *hw = &pf->hw;
8249         int ret = 0;
8250
8251         if (seed) {
8252                 ret = i40e_aq_get_rss_key(hw, vsi->id,
8253                         (struct i40e_aqc_get_set_rss_key_data *)seed);
8254                 if (ret) {
8255                         dev_info(&pf->pdev->dev,
8256                                  "Cannot get RSS key, err %s aq_err %s\n",
8257                                  i40e_stat_str(&pf->hw, ret),
8258                                  i40e_aq_str(&pf->hw,
8259                                              pf->hw.aq.asq_last_status));
8260                         return ret;
8261                 }
8262         }
8263
8264         if (lut) {
8265                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8266
8267                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8268                 if (ret) {
8269                         dev_info(&pf->pdev->dev,
8270                                  "Cannot get RSS lut, err %s aq_err %s\n",
8271                                  i40e_stat_str(&pf->hw, ret),
8272                                  i40e_aq_str(&pf->hw,
8273                                              pf->hw.aq.asq_last_status));
8274                         return ret;
8275                 }
8276         }
8277
8278         return ret;
8279 }
8280
8281 /**
8282  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8283  * @vsi: VSI structure
8284  **/
8285 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8286 {
8287         u8 seed[I40E_HKEY_ARRAY_SIZE];
8288         struct i40e_pf *pf = vsi->back;
8289         u8 *lut;
8290         int ret;
8291
8292         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8293                 return 0;
8294
8295         if (!vsi->rss_size)
8296                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8297                                       vsi->num_queue_pairs);
8298         if (!vsi->rss_size)
8299                 return -EINVAL;
8300
8301         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8302         if (!lut)
8303                 return -ENOMEM;
8304         /* Use the user configured hash keys and lookup table if there is one,
8305          * otherwise use default
8306          */
8307         if (vsi->rss_lut_user)
8308                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8309         else
8310                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8311         if (vsi->rss_hkey_user)
8312                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8313         else
8314                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8315         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8316         kfree(lut);
8317
8318         return ret;
8319 }
8320
8321 /**
8322  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8323  * @vsi: Pointer to vsi structure
8324  * @seed: RSS hash seed
8325  * @lut: Lookup table
8326  * @lut_size: Lookup table size
8327  *
8328  * Returns 0 on success, negative on failure
8329  **/
8330 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8331                                const u8 *lut, u16 lut_size)
8332 {
8333         struct i40e_pf *pf = vsi->back;
8334         struct i40e_hw *hw = &pf->hw;
8335         u16 vf_id = vsi->vf_id;
8336         u8 i;
8337
8338         /* Fill out hash function seed */
8339         if (seed) {
8340                 u32 *seed_dw = (u32 *)seed;
8341
8342                 if (vsi->type == I40E_VSI_MAIN) {
8343                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8344                                 wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
8345                 } else if (vsi->type == I40E_VSI_SRIOV) {
8346                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8347                                 wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
8348                 } else {
8349                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8350                 }
8351         }
8352
8353         if (lut) {
8354                 u32 *lut_dw = (u32 *)lut;
8355
8356                 if (vsi->type == I40E_VSI_MAIN) {
8357                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
8358                                 return -EINVAL;
8359                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8360                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8361                 } else if (vsi->type == I40E_VSI_SRIOV) {
8362                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8363                                 return -EINVAL;
8364                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8365                                 wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
8366                 } else {
8367                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8368                 }
8369         }
8370         i40e_flush(hw);
8371
8372         return 0;
8373 }
8374
8375 /**
8376  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8377  * @vsi: Pointer to VSI structure
8378  * @seed: Buffer to store the keys
8379  * @lut: Buffer to store the lookup table entries
8380  * @lut_size: Size of buffer to store the lookup table entries
8381  *
8382  * Returns 0 on success, negative on failure
8383  */
8384 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8385                             u8 *lut, u16 lut_size)
8386 {
8387         struct i40e_pf *pf = vsi->back;
8388         struct i40e_hw *hw = &pf->hw;
8389         u16 i;
8390
8391         if (seed) {
8392                 u32 *seed_dw = (u32 *)seed;
8393
8394                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8395                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8396         }
8397         if (lut) {
8398                 u32 *lut_dw = (u32 *)lut;
8399
8400                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8401                         return -EINVAL;
8402                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8403                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8404         }
8405
8406         return 0;
8407 }
8408
8409 /**
8410  * i40e_config_rss - Configure RSS keys and lut
8411  * @vsi: Pointer to VSI structure
8412  * @seed: RSS hash seed
8413  * @lut: Lookup table
8414  * @lut_size: Lookup table size
8415  *
8416  * Returns 0 on success, negative on failure
8417  */
8418 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8419 {
8420         struct i40e_pf *pf = vsi->back;
8421
8422         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8423                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8424         else
8425                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8426 }
8427
8428 /**
8429  * i40e_get_rss - Get RSS keys and lut
8430  * @vsi: Pointer to VSI structure
8431  * @seed: Buffer to store the keys
8432  * @lut: Buffer to store the lookup table entries
8433  * lut_size: Size of buffer to store the lookup table entries
8434  *
8435  * Returns 0 on success, negative on failure
8436  */
8437 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8438 {
8439         struct i40e_pf *pf = vsi->back;
8440
8441         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8442                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8443         else
8444                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8445 }
8446
8447 /**
8448  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8449  * @pf: Pointer to board private structure
8450  * @lut: Lookup table
8451  * @rss_table_size: Lookup table size
8452  * @rss_size: Range of queue number for hashing
8453  */
8454 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8455                        u16 rss_table_size, u16 rss_size)
8456 {
8457         u16 i;
8458
8459         for (i = 0; i < rss_table_size; i++)
8460                 lut[i] = i % rss_size;
8461 }
8462
8463 /**
8464  * i40e_pf_config_rss - Prepare for RSS if used
8465  * @pf: board private structure
8466  **/
8467 static int i40e_pf_config_rss(struct i40e_pf *pf)
8468 {
8469         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8470         u8 seed[I40E_HKEY_ARRAY_SIZE];
8471         u8 *lut;
8472         struct i40e_hw *hw = &pf->hw;
8473         u32 reg_val;
8474         u64 hena;
8475         int ret;
8476
8477         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8478         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8479                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8480         hena |= i40e_pf_get_default_rss_hena(pf);
8481
8482         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8483         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8484
8485         /* Determine the RSS table size based on the hardware capabilities */
8486         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8487         reg_val = (pf->rss_table_size == 512) ?
8488                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8489                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8490         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8491
8492         /* Determine the RSS size of the VSI */
8493         if (!vsi->rss_size) {
8494                 u16 qcount;
8495
8496                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8497                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8498         }
8499         if (!vsi->rss_size)
8500                 return -EINVAL;
8501
8502         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8503         if (!lut)
8504                 return -ENOMEM;
8505
8506         /* Use user configured lut if there is one, otherwise use default */
8507         if (vsi->rss_lut_user)
8508                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8509         else
8510                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8511
8512         /* Use user configured hash key if there is one, otherwise
8513          * use default.
8514          */
8515         if (vsi->rss_hkey_user)
8516                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8517         else
8518                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8519         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8520         kfree(lut);
8521
8522         return ret;
8523 }
8524
8525 /**
8526  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8527  * @pf: board private structure
8528  * @queue_count: the requested queue count for rss.
8529  *
8530  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8531  * count which may be different from the requested queue count.
8532  * Note: expects to be called while under rtnl_lock()
8533  **/
8534 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8535 {
8536         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8537         int new_rss_size;
8538
8539         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8540                 return 0;
8541
8542         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8543
8544         if (queue_count != vsi->num_queue_pairs) {
8545                 u16 qcount;
8546
8547                 vsi->req_queue_pairs = queue_count;
8548                 i40e_prep_for_reset(pf, true);
8549
8550                 pf->alloc_rss_size = new_rss_size;
8551
8552                 i40e_reset_and_rebuild(pf, true, true);
8553
8554                 /* Discard the user configured hash keys and lut, if less
8555                  * queues are enabled.
8556                  */
8557                 if (queue_count < vsi->rss_size) {
8558                         i40e_clear_rss_config_user(vsi);
8559                         dev_dbg(&pf->pdev->dev,
8560                                 "discard user configured hash keys and lut\n");
8561                 }
8562
8563                 /* Reset vsi->rss_size, as number of enabled queues changed */
8564                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8565                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8566
8567                 i40e_pf_config_rss(pf);
8568         }
8569         dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
8570                  vsi->req_queue_pairs, pf->rss_size_max);
8571         return pf->alloc_rss_size;
8572 }
8573
8574 /**
8575  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8576  * @pf: board private structure
8577  **/
8578 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8579 {
8580         i40e_status status;
8581         bool min_valid, max_valid;
8582         u32 max_bw, min_bw;
8583
8584         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8585                                            &min_valid, &max_valid);
8586
8587         if (!status) {
8588                 if (min_valid)
8589                         pf->npar_min_bw = min_bw;
8590                 if (max_valid)
8591                         pf->npar_max_bw = max_bw;
8592         }
8593
8594         return status;
8595 }
8596
8597 /**
8598  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8599  * @pf: board private structure
8600  **/
8601 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8602 {
8603         struct i40e_aqc_configure_partition_bw_data bw_data;
8604         i40e_status status;
8605
8606         /* Set the valid bit for this PF */
8607         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8608         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8609         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8610
8611         /* Set the new bandwidths */
8612         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8613
8614         return status;
8615 }
8616
8617 /**
8618  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8619  * @pf: board private structure
8620  **/
8621 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8622 {
8623         /* Commit temporary BW setting to permanent NVM image */
8624         enum i40e_admin_queue_err last_aq_status;
8625         i40e_status ret;
8626         u16 nvm_word;
8627
8628         if (pf->hw.partition_id != 1) {
8629                 dev_info(&pf->pdev->dev,
8630                          "Commit BW only works on partition 1! This is partition %d",
8631                          pf->hw.partition_id);
8632                 ret = I40E_NOT_SUPPORTED;
8633                 goto bw_commit_out;
8634         }
8635
8636         /* Acquire NVM for read access */
8637         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8638         last_aq_status = pf->hw.aq.asq_last_status;
8639         if (ret) {
8640                 dev_info(&pf->pdev->dev,
8641                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8642                          i40e_stat_str(&pf->hw, ret),
8643                          i40e_aq_str(&pf->hw, last_aq_status));
8644                 goto bw_commit_out;
8645         }
8646
8647         /* Read word 0x10 of NVM - SW compatibility word 1 */
8648         ret = i40e_aq_read_nvm(&pf->hw,
8649                                I40E_SR_NVM_CONTROL_WORD,
8650                                0x10, sizeof(nvm_word), &nvm_word,
8651                                false, NULL);
8652         /* Save off last admin queue command status before releasing
8653          * the NVM
8654          */
8655         last_aq_status = pf->hw.aq.asq_last_status;
8656         i40e_release_nvm(&pf->hw);
8657         if (ret) {
8658                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8659                          i40e_stat_str(&pf->hw, ret),
8660                          i40e_aq_str(&pf->hw, last_aq_status));
8661                 goto bw_commit_out;
8662         }
8663
8664         /* Wait a bit for NVM release to complete */
8665         msleep(50);
8666
8667         /* Acquire NVM for write access */
8668         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8669         last_aq_status = pf->hw.aq.asq_last_status;
8670         if (ret) {
8671                 dev_info(&pf->pdev->dev,
8672                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8673                          i40e_stat_str(&pf->hw, ret),
8674                          i40e_aq_str(&pf->hw, last_aq_status));
8675                 goto bw_commit_out;
8676         }
8677         /* Write it back out unchanged to initiate update NVM,
8678          * which will force a write of the shadow (alt) RAM to
8679          * the NVM - thus storing the bandwidth values permanently.
8680          */
8681         ret = i40e_aq_update_nvm(&pf->hw,
8682                                  I40E_SR_NVM_CONTROL_WORD,
8683                                  0x10, sizeof(nvm_word),
8684                                  &nvm_word, true, NULL);
8685         /* Save off last admin queue command status before releasing
8686          * the NVM
8687          */
8688         last_aq_status = pf->hw.aq.asq_last_status;
8689         i40e_release_nvm(&pf->hw);
8690         if (ret)
8691                 dev_info(&pf->pdev->dev,
8692                          "BW settings NOT SAVED, err %s aq_err %s\n",
8693                          i40e_stat_str(&pf->hw, ret),
8694                          i40e_aq_str(&pf->hw, last_aq_status));
8695 bw_commit_out:
8696
8697         return ret;
8698 }
8699
8700 /**
8701  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8702  * @pf: board private structure to initialize
8703  *
8704  * i40e_sw_init initializes the Adapter private data structure.
8705  * Fields are initialized based on PCI device information and
8706  * OS network device settings (MTU size).
8707  **/
8708 static int i40e_sw_init(struct i40e_pf *pf)
8709 {
8710         int err = 0;
8711         int size;
8712
8713         /* Set default capability flags */
8714         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8715                     I40E_FLAG_MSI_ENABLED     |
8716                     I40E_FLAG_MSIX_ENABLED;
8717
8718         /* Set default ITR */
8719         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8720         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8721
8722         /* Depending on PF configurations, it is possible that the RSS
8723          * maximum might end up larger than the available queues
8724          */
8725         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8726         pf->alloc_rss_size = 1;
8727         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8728         pf->rss_size_max = min_t(int, pf->rss_size_max,
8729                                  pf->hw.func_caps.num_tx_qp);
8730         if (pf->hw.func_caps.rss) {
8731                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8732                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8733                                            num_online_cpus());
8734         }
8735
8736         /* MFP mode enabled */
8737         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8738                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8739                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8740                 if (i40e_get_npar_bw_setting(pf))
8741                         dev_warn(&pf->pdev->dev,
8742                                  "Could not get NPAR bw settings\n");
8743                 else
8744                         dev_info(&pf->pdev->dev,
8745                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8746                                  pf->npar_min_bw, pf->npar_max_bw);
8747         }
8748
8749         /* FW/NVM is not yet fixed in this regard */
8750         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8751             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8752                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8753                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8754                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8755                     pf->hw.num_partitions > 1)
8756                         dev_info(&pf->pdev->dev,
8757                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8758                 else
8759                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8760                 pf->fdir_pf_filter_count =
8761                                  pf->hw.func_caps.fd_filters_guaranteed;
8762                 pf->hw.fdir_shared_filter_count =
8763                                  pf->hw.func_caps.fd_filters_best_effort;
8764         }
8765
8766         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8767             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8768             (pf->hw.aq.fw_maj_ver < 4))) {
8769                 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8770                 /* No DCB support  for FW < v4.33 */
8771                 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8772         }
8773
8774         /* Disable FW LLDP if FW < v4.3 */
8775         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8776             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8777             (pf->hw.aq.fw_maj_ver < 4)))
8778                 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8779
8780         /* Use the FW Set LLDP MIB API if FW > v4.40 */
8781         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8782             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8783             (pf->hw.aq.fw_maj_ver >= 5)))
8784                 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8785
8786         if (pf->hw.func_caps.vmdq) {
8787                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8788                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8789                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8790         }
8791
8792         if (pf->hw.func_caps.iwarp) {
8793                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8794                 /* IWARP needs one extra vector for CQP just like MISC.*/
8795                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8796         }
8797
8798 #ifdef CONFIG_PCI_IOV
8799         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8800                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8801                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8802                 pf->num_req_vfs = min_t(int,
8803                                         pf->hw.func_caps.num_vfs,
8804                                         I40E_MAX_VF_COUNT);
8805         }
8806 #endif /* CONFIG_PCI_IOV */
8807         if (pf->hw.mac.type == I40E_MAC_X722) {
8808                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE
8809                              | I40E_FLAG_128_QP_RSS_CAPABLE
8810                              | I40E_FLAG_HW_ATR_EVICT_CAPABLE
8811                              | I40E_FLAG_OUTER_UDP_CSUM_CAPABLE
8812                              | I40E_FLAG_WB_ON_ITR_CAPABLE
8813                              | I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE
8814                              | I40E_FLAG_NO_PCI_LINK_CHECK
8815                              | I40E_FLAG_USE_SET_LLDP_MIB
8816                              | I40E_FLAG_GENEVE_OFFLOAD_CAPABLE
8817                              | I40E_FLAG_PTP_L4_CAPABLE
8818                              | I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE;
8819         } else if ((pf->hw.aq.api_maj_ver > 1) ||
8820                    ((pf->hw.aq.api_maj_ver == 1) &&
8821                     (pf->hw.aq.api_min_ver > 4))) {
8822                 /* Supported in FW API version higher than 1.4 */
8823                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8824                 pf->flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8825         } else {
8826                 pf->flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8827         }
8828
8829         pf->eeprom_version = 0xDEAD;
8830         pf->lan_veb = I40E_NO_VEB;
8831         pf->lan_vsi = I40E_NO_VSI;
8832
8833         /* By default FW has this off for performance reasons */
8834         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8835
8836         /* set up queue assignment tracking */
8837         size = sizeof(struct i40e_lump_tracking)
8838                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8839         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8840         if (!pf->qp_pile) {
8841                 err = -ENOMEM;
8842                 goto sw_init_done;
8843         }
8844         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8845         pf->qp_pile->search_hint = 0;
8846
8847         pf->tx_timeout_recovery_level = 1;
8848
8849         mutex_init(&pf->switch_mutex);
8850
8851         /* If NPAR is enabled nudge the Tx scheduler */
8852         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8853                 i40e_set_npar_bw_setting(pf);
8854
8855 sw_init_done:
8856         return err;
8857 }
8858
8859 /**
8860  * i40e_set_ntuple - set the ntuple feature flag and take action
8861  * @pf: board private structure to initialize
8862  * @features: the feature set that the stack is suggesting
8863  *
8864  * returns a bool to indicate if reset needs to happen
8865  **/
8866 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8867 {
8868         bool need_reset = false;
8869
8870         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8871          * the state changed, we need to reset.
8872          */
8873         if (features & NETIF_F_NTUPLE) {
8874                 /* Enable filters and mark for reset */
8875                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8876                         need_reset = true;
8877                 /* enable FD_SB only if there is MSI-X vector */
8878                 if (pf->num_fdsb_msix > 0)
8879                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8880         } else {
8881                 /* turn off filters, mark for reset and clear SW filter list */
8882                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8883                         need_reset = true;
8884                         i40e_fdir_filter_exit(pf);
8885                 }
8886                 pf->flags &= ~(I40E_FLAG_FD_SB_ENABLED |
8887                                I40E_FLAG_FD_SB_AUTO_DISABLED);
8888                 /* reset fd counters */
8889                 pf->fd_add_err = 0;
8890                 pf->fd_atr_cnt = 0;
8891                 /* if ATR was auto disabled it can be re-enabled. */
8892                 if (pf->flags & I40E_FLAG_FD_ATR_AUTO_DISABLED) {
8893                         pf->flags &= ~I40E_FLAG_FD_ATR_AUTO_DISABLED;
8894                         if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8895                             (I40E_DEBUG_FD & pf->hw.debug_mask))
8896                                 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8897                 }
8898         }
8899         return need_reset;
8900 }
8901
8902 /**
8903  * i40e_clear_rss_lut - clear the rx hash lookup table
8904  * @vsi: the VSI being configured
8905  **/
8906 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
8907 {
8908         struct i40e_pf *pf = vsi->back;
8909         struct i40e_hw *hw = &pf->hw;
8910         u16 vf_id = vsi->vf_id;
8911         u8 i;
8912
8913         if (vsi->type == I40E_VSI_MAIN) {
8914                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8915                         wr32(hw, I40E_PFQF_HLUT(i), 0);
8916         } else if (vsi->type == I40E_VSI_SRIOV) {
8917                 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8918                         i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
8919         } else {
8920                 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8921         }
8922 }
8923
8924 /**
8925  * i40e_set_features - set the netdev feature flags
8926  * @netdev: ptr to the netdev being adjusted
8927  * @features: the feature set that the stack is suggesting
8928  * Note: expects to be called while under rtnl_lock()
8929  **/
8930 static int i40e_set_features(struct net_device *netdev,
8931                              netdev_features_t features)
8932 {
8933         struct i40e_netdev_priv *np = netdev_priv(netdev);
8934         struct i40e_vsi *vsi = np->vsi;
8935         struct i40e_pf *pf = vsi->back;
8936         bool need_reset;
8937
8938         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
8939                 i40e_pf_config_rss(pf);
8940         else if (!(features & NETIF_F_RXHASH) &&
8941                  netdev->features & NETIF_F_RXHASH)
8942                 i40e_clear_rss_lut(vsi);
8943
8944         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8945                 i40e_vlan_stripping_enable(vsi);
8946         else
8947                 i40e_vlan_stripping_disable(vsi);
8948
8949         need_reset = i40e_set_ntuple(pf, features);
8950
8951         if (need_reset)
8952                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
8953
8954         return 0;
8955 }
8956
8957 /**
8958  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8959  * @pf: board private structure
8960  * @port: The UDP port to look up
8961  *
8962  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8963  **/
8964 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
8965 {
8966         u8 i;
8967
8968         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8969                 if (pf->udp_ports[i].port == port)
8970                         return i;
8971         }
8972
8973         return i;
8974 }
8975
8976 /**
8977  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8978  * @netdev: This physical port's netdev
8979  * @ti: Tunnel endpoint information
8980  **/
8981 static void i40e_udp_tunnel_add(struct net_device *netdev,
8982                                 struct udp_tunnel_info *ti)
8983 {
8984         struct i40e_netdev_priv *np = netdev_priv(netdev);
8985         struct i40e_vsi *vsi = np->vsi;
8986         struct i40e_pf *pf = vsi->back;
8987         u16 port = ntohs(ti->port);
8988         u8 next_idx;
8989         u8 idx;
8990
8991         idx = i40e_get_udp_port_idx(pf, port);
8992
8993         /* Check if port already exists */
8994         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8995                 netdev_info(netdev, "port %d already offloaded\n", port);
8996                 return;
8997         }
8998
8999         /* Now check if there is space to add the new port */
9000         next_idx = i40e_get_udp_port_idx(pf, 0);
9001
9002         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
9003                 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
9004                             port);
9005                 return;
9006         }
9007
9008         switch (ti->type) {
9009         case UDP_TUNNEL_TYPE_VXLAN:
9010                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
9011                 break;
9012         case UDP_TUNNEL_TYPE_GENEVE:
9013                 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
9014                         return;
9015                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
9016                 break;
9017         default:
9018                 return;
9019         }
9020
9021         /* New port: add it and mark its index in the bitmap */
9022         pf->udp_ports[next_idx].port = port;
9023         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
9024         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9025 }
9026
9027 /**
9028  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
9029  * @netdev: This physical port's netdev
9030  * @ti: Tunnel endpoint information
9031  **/
9032 static void i40e_udp_tunnel_del(struct net_device *netdev,
9033                                 struct udp_tunnel_info *ti)
9034 {
9035         struct i40e_netdev_priv *np = netdev_priv(netdev);
9036         struct i40e_vsi *vsi = np->vsi;
9037         struct i40e_pf *pf = vsi->back;
9038         u16 port = ntohs(ti->port);
9039         u8 idx;
9040
9041         idx = i40e_get_udp_port_idx(pf, port);
9042
9043         /* Check if port already exists */
9044         if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
9045                 goto not_found;
9046
9047         switch (ti->type) {
9048         case UDP_TUNNEL_TYPE_VXLAN:
9049                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
9050                         goto not_found;
9051                 break;
9052         case UDP_TUNNEL_TYPE_GENEVE:
9053                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
9054                         goto not_found;
9055                 break;
9056         default:
9057                 goto not_found;
9058         }
9059
9060         /* if port exists, set it to 0 (mark for deletion)
9061          * and make it pending
9062          */
9063         pf->udp_ports[idx].port = 0;
9064         pf->pending_udp_bitmap |= BIT_ULL(idx);
9065         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9066
9067         return;
9068 not_found:
9069         netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
9070                     port);
9071 }
9072
9073 static int i40e_get_phys_port_id(struct net_device *netdev,
9074                                  struct netdev_phys_item_id *ppid)
9075 {
9076         struct i40e_netdev_priv *np = netdev_priv(netdev);
9077         struct i40e_pf *pf = np->vsi->back;
9078         struct i40e_hw *hw = &pf->hw;
9079
9080         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
9081                 return -EOPNOTSUPP;
9082
9083         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
9084         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
9085
9086         return 0;
9087 }
9088
9089 /**
9090  * i40e_ndo_fdb_add - add an entry to the hardware database
9091  * @ndm: the input from the stack
9092  * @tb: pointer to array of nladdr (unused)
9093  * @dev: the net device pointer
9094  * @addr: the MAC address entry being added
9095  * @flags: instructions from stack about fdb operation
9096  */
9097 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9098                             struct net_device *dev,
9099                             const unsigned char *addr, u16 vid,
9100                             u16 flags)
9101 {
9102         struct i40e_netdev_priv *np = netdev_priv(dev);
9103         struct i40e_pf *pf = np->vsi->back;
9104         int err = 0;
9105
9106         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
9107                 return -EOPNOTSUPP;
9108
9109         if (vid) {
9110                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
9111                 return -EINVAL;
9112         }
9113
9114         /* Hardware does not support aging addresses so if a
9115          * ndm_state is given only allow permanent addresses
9116          */
9117         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
9118                 netdev_info(dev, "FDB only supports static addresses\n");
9119                 return -EINVAL;
9120         }
9121
9122         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
9123                 err = dev_uc_add_excl(dev, addr);
9124         else if (is_multicast_ether_addr(addr))
9125                 err = dev_mc_add_excl(dev, addr);
9126         else
9127                 err = -EINVAL;
9128
9129         /* Only return duplicate errors if NLM_F_EXCL is set */
9130         if (err == -EEXIST && !(flags & NLM_F_EXCL))
9131                 err = 0;
9132
9133         return err;
9134 }
9135
9136 /**
9137  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
9138  * @dev: the netdev being configured
9139  * @nlh: RTNL message
9140  *
9141  * Inserts a new hardware bridge if not already created and
9142  * enables the bridging mode requested (VEB or VEPA). If the
9143  * hardware bridge has already been inserted and the request
9144  * is to change the mode then that requires a PF reset to
9145  * allow rebuild of the components with required hardware
9146  * bridge mode enabled.
9147  *
9148  * Note: expects to be called while under rtnl_lock()
9149  **/
9150 static int i40e_ndo_bridge_setlink(struct net_device *dev,
9151                                    struct nlmsghdr *nlh,
9152                                    u16 flags)
9153 {
9154         struct i40e_netdev_priv *np = netdev_priv(dev);
9155         struct i40e_vsi *vsi = np->vsi;
9156         struct i40e_pf *pf = vsi->back;
9157         struct i40e_veb *veb = NULL;
9158         struct nlattr *attr, *br_spec;
9159         int i, rem;
9160
9161         /* Only for PF VSI for now */
9162         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9163                 return -EOPNOTSUPP;
9164
9165         /* Find the HW bridge for PF VSI */
9166         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9167                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9168                         veb = pf->veb[i];
9169         }
9170
9171         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9172
9173         nla_for_each_nested(attr, br_spec, rem) {
9174                 __u16 mode;
9175
9176                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
9177                         continue;
9178
9179                 mode = nla_get_u16(attr);
9180                 if ((mode != BRIDGE_MODE_VEPA) &&
9181                     (mode != BRIDGE_MODE_VEB))
9182                         return -EINVAL;
9183
9184                 /* Insert a new HW bridge */
9185                 if (!veb) {
9186                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9187                                              vsi->tc_config.enabled_tc);
9188                         if (veb) {
9189                                 veb->bridge_mode = mode;
9190                                 i40e_config_bridge_mode(veb);
9191                         } else {
9192                                 /* No Bridge HW offload available */
9193                                 return -ENOENT;
9194                         }
9195                         break;
9196                 } else if (mode != veb->bridge_mode) {
9197                         /* Existing HW bridge but different mode needs reset */
9198                         veb->bridge_mode = mode;
9199                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
9200                         if (mode == BRIDGE_MODE_VEB)
9201                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9202                         else
9203                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9204                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED),
9205                                       true);
9206                         break;
9207                 }
9208         }
9209
9210         return 0;
9211 }
9212
9213 /**
9214  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9215  * @skb: skb buff
9216  * @pid: process id
9217  * @seq: RTNL message seq #
9218  * @dev: the netdev being configured
9219  * @filter_mask: unused
9220  * @nlflags: netlink flags passed in
9221  *
9222  * Return the mode in which the hardware bridge is operating in
9223  * i.e VEB or VEPA.
9224  **/
9225 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9226                                    struct net_device *dev,
9227                                    u32 __always_unused filter_mask,
9228                                    int nlflags)
9229 {
9230         struct i40e_netdev_priv *np = netdev_priv(dev);
9231         struct i40e_vsi *vsi = np->vsi;
9232         struct i40e_pf *pf = vsi->back;
9233         struct i40e_veb *veb = NULL;
9234         int i;
9235
9236         /* Only for PF VSI for now */
9237         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9238                 return -EOPNOTSUPP;
9239
9240         /* Find the HW bridge for the PF VSI */
9241         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9242                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9243                         veb = pf->veb[i];
9244         }
9245
9246         if (!veb)
9247                 return 0;
9248
9249         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9250                                        0, 0, nlflags, filter_mask, NULL);
9251 }
9252
9253 /**
9254  * i40e_features_check - Validate encapsulated packet conforms to limits
9255  * @skb: skb buff
9256  * @dev: This physical port's netdev
9257  * @features: Offload features that the stack believes apply
9258  **/
9259 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9260                                              struct net_device *dev,
9261                                              netdev_features_t features)
9262 {
9263         size_t len;
9264
9265         /* No point in doing any of this if neither checksum nor GSO are
9266          * being requested for this frame.  We can rule out both by just
9267          * checking for CHECKSUM_PARTIAL
9268          */
9269         if (skb->ip_summed != CHECKSUM_PARTIAL)
9270                 return features;
9271
9272         /* We cannot support GSO if the MSS is going to be less than
9273          * 64 bytes.  If it is then we need to drop support for GSO.
9274          */
9275         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
9276                 features &= ~NETIF_F_GSO_MASK;
9277
9278         /* MACLEN can support at most 63 words */
9279         len = skb_network_header(skb) - skb->data;
9280         if (len & ~(63 * 2))
9281                 goto out_err;
9282
9283         /* IPLEN and EIPLEN can support at most 127 dwords */
9284         len = skb_transport_header(skb) - skb_network_header(skb);
9285         if (len & ~(127 * 4))
9286                 goto out_err;
9287
9288         if (skb->encapsulation) {
9289                 /* L4TUNLEN can support 127 words */
9290                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
9291                 if (len & ~(127 * 2))
9292                         goto out_err;
9293
9294                 /* IPLEN can support at most 127 dwords */
9295                 len = skb_inner_transport_header(skb) -
9296                       skb_inner_network_header(skb);
9297                 if (len & ~(127 * 4))
9298                         goto out_err;
9299         }
9300
9301         /* No need to validate L4LEN as TCP is the only protocol with a
9302          * a flexible value and we support all possible values supported
9303          * by TCP, which is at most 15 dwords
9304          */
9305
9306         return features;
9307 out_err:
9308         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9309 }
9310
9311 static const struct net_device_ops i40e_netdev_ops = {
9312         .ndo_open               = i40e_open,
9313         .ndo_stop               = i40e_close,
9314         .ndo_start_xmit         = i40e_lan_xmit_frame,
9315         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
9316         .ndo_set_rx_mode        = i40e_set_rx_mode,
9317         .ndo_validate_addr      = eth_validate_addr,
9318         .ndo_set_mac_address    = i40e_set_mac,
9319         .ndo_change_mtu         = i40e_change_mtu,
9320         .ndo_do_ioctl           = i40e_ioctl,
9321         .ndo_tx_timeout         = i40e_tx_timeout,
9322         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
9323         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
9324 #ifdef CONFIG_NET_POLL_CONTROLLER
9325         .ndo_poll_controller    = i40e_netpoll,
9326 #endif
9327         .ndo_setup_tc           = __i40e_setup_tc,
9328         .ndo_set_features       = i40e_set_features,
9329         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
9330         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
9331         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
9332         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
9333         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
9334         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
9335         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
9336         .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
9337         .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
9338         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
9339         .ndo_fdb_add            = i40e_ndo_fdb_add,
9340         .ndo_features_check     = i40e_features_check,
9341         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
9342         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
9343 };
9344
9345 /**
9346  * i40e_config_netdev - Setup the netdev flags
9347  * @vsi: the VSI being configured
9348  *
9349  * Returns 0 on success, negative value on failure
9350  **/
9351 static int i40e_config_netdev(struct i40e_vsi *vsi)
9352 {
9353         struct i40e_pf *pf = vsi->back;
9354         struct i40e_hw *hw = &pf->hw;
9355         struct i40e_netdev_priv *np;
9356         struct net_device *netdev;
9357         u8 broadcast[ETH_ALEN];
9358         u8 mac_addr[ETH_ALEN];
9359         int etherdev_size;
9360         netdev_features_t hw_enc_features;
9361         netdev_features_t hw_features;
9362
9363         etherdev_size = sizeof(struct i40e_netdev_priv);
9364         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9365         if (!netdev)
9366                 return -ENOMEM;
9367
9368         vsi->netdev = netdev;
9369         np = netdev_priv(netdev);
9370         np->vsi = vsi;
9371
9372         hw_enc_features = NETIF_F_SG                    |
9373                           NETIF_F_IP_CSUM               |
9374                           NETIF_F_IPV6_CSUM             |
9375                           NETIF_F_HIGHDMA               |
9376                           NETIF_F_SOFT_FEATURES         |
9377                           NETIF_F_TSO                   |
9378                           NETIF_F_TSO_ECN               |
9379                           NETIF_F_TSO6                  |
9380                           NETIF_F_GSO_GRE               |
9381                           NETIF_F_GSO_GRE_CSUM          |
9382                           NETIF_F_GSO_PARTIAL           |
9383                           NETIF_F_GSO_UDP_TUNNEL        |
9384                           NETIF_F_GSO_UDP_TUNNEL_CSUM   |
9385                           NETIF_F_SCTP_CRC              |
9386                           NETIF_F_RXHASH                |
9387                           NETIF_F_RXCSUM                |
9388                           0;
9389
9390         if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9391                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9392
9393         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9394
9395         netdev->hw_enc_features |= hw_enc_features;
9396
9397         /* record features VLANs can make use of */
9398         netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
9399
9400         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9401                 netdev->hw_features |= NETIF_F_NTUPLE;
9402         hw_features = hw_enc_features           |
9403                       NETIF_F_HW_VLAN_CTAG_TX   |
9404                       NETIF_F_HW_VLAN_CTAG_RX;
9405
9406         netdev->hw_features |= hw_features;
9407
9408         netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9409         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9410
9411         if (vsi->type == I40E_VSI_MAIN) {
9412                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9413                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9414                 /* The following steps are necessary for two reasons. First,
9415                  * some older NVM configurations load a default MAC-VLAN
9416                  * filter that will accept any tagged packet, and we want to
9417                  * replace this with a normal filter. Additionally, it is
9418                  * possible our MAC address was provided by the platform using
9419                  * Open Firmware or similar.
9420                  *
9421                  * Thus, we need to remove the default filter and install one
9422                  * specific to the MAC address.
9423                  */
9424                 i40e_rm_default_mac_filter(vsi, mac_addr);
9425                 spin_lock_bh(&vsi->mac_filter_hash_lock);
9426                 i40e_add_mac_filter(vsi, mac_addr);
9427                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
9428         } else {
9429                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9430                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9431                          pf->vsi[pf->lan_vsi]->netdev->name);
9432                 random_ether_addr(mac_addr);
9433
9434                 spin_lock_bh(&vsi->mac_filter_hash_lock);
9435                 i40e_add_mac_filter(vsi, mac_addr);
9436                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
9437         }
9438
9439         /* Add the broadcast filter so that we initially will receive
9440          * broadcast packets. Note that when a new VLAN is first added the
9441          * driver will convert all filters marked I40E_VLAN_ANY into VLAN
9442          * specific filters as part of transitioning into "vlan" operation.
9443          * When more VLANs are added, the driver will copy each existing MAC
9444          * filter and add it for the new VLAN.
9445          *
9446          * Broadcast filters are handled specially by
9447          * i40e_sync_filters_subtask, as the driver must to set the broadcast
9448          * promiscuous bit instead of adding this directly as a MAC/VLAN
9449          * filter. The subtask will update the correct broadcast promiscuous
9450          * bits as VLANs become active or inactive.
9451          */
9452         eth_broadcast_addr(broadcast);
9453         spin_lock_bh(&vsi->mac_filter_hash_lock);
9454         i40e_add_mac_filter(vsi, broadcast);
9455         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9456
9457         ether_addr_copy(netdev->dev_addr, mac_addr);
9458         ether_addr_copy(netdev->perm_addr, mac_addr);
9459
9460         netdev->priv_flags |= IFF_UNICAST_FLT;
9461         netdev->priv_flags |= IFF_SUPP_NOFCS;
9462         /* Setup netdev TC information */
9463         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9464
9465         netdev->netdev_ops = &i40e_netdev_ops;
9466         netdev->watchdog_timeo = 5 * HZ;
9467         i40e_set_ethtool_ops(netdev);
9468
9469         /* MTU range: 68 - 9706 */
9470         netdev->min_mtu = ETH_MIN_MTU;
9471         netdev->max_mtu = I40E_MAX_RXBUFFER -
9472                           (ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
9473
9474         return 0;
9475 }
9476
9477 /**
9478  * i40e_vsi_delete - Delete a VSI from the switch
9479  * @vsi: the VSI being removed
9480  *
9481  * Returns 0 on success, negative value on failure
9482  **/
9483 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9484 {
9485         /* remove default VSI is not allowed */
9486         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9487                 return;
9488
9489         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9490 }
9491
9492 /**
9493  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9494  * @vsi: the VSI being queried
9495  *
9496  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9497  **/
9498 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9499 {
9500         struct i40e_veb *veb;
9501         struct i40e_pf *pf = vsi->back;
9502
9503         /* Uplink is not a bridge so default to VEB */
9504         if (vsi->veb_idx == I40E_NO_VEB)
9505                 return 1;
9506
9507         veb = pf->veb[vsi->veb_idx];
9508         if (!veb) {
9509                 dev_info(&pf->pdev->dev,
9510                          "There is no veb associated with the bridge\n");
9511                 return -ENOENT;
9512         }
9513
9514         /* Uplink is a bridge in VEPA mode */
9515         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9516                 return 0;
9517         } else {
9518                 /* Uplink is a bridge in VEB mode */
9519                 return 1;
9520         }
9521
9522         /* VEPA is now default bridge, so return 0 */
9523         return 0;
9524 }
9525
9526 /**
9527  * i40e_add_vsi - Add a VSI to the switch
9528  * @vsi: the VSI being configured
9529  *
9530  * This initializes a VSI context depending on the VSI type to be added and
9531  * passes it down to the add_vsi aq command.
9532  **/
9533 static int i40e_add_vsi(struct i40e_vsi *vsi)
9534 {
9535         int ret = -ENODEV;
9536         struct i40e_pf *pf = vsi->back;
9537         struct i40e_hw *hw = &pf->hw;
9538         struct i40e_vsi_context ctxt;
9539         struct i40e_mac_filter *f;
9540         struct hlist_node *h;
9541         int bkt;
9542
9543         u8 enabled_tc = 0x1; /* TC0 enabled */
9544         int f_count = 0;
9545
9546         memset(&ctxt, 0, sizeof(ctxt));
9547         switch (vsi->type) {
9548         case I40E_VSI_MAIN:
9549                 /* The PF's main VSI is already setup as part of the
9550                  * device initialization, so we'll not bother with
9551                  * the add_vsi call, but we will retrieve the current
9552                  * VSI context.
9553                  */
9554                 ctxt.seid = pf->main_vsi_seid;
9555                 ctxt.pf_num = pf->hw.pf_id;
9556                 ctxt.vf_num = 0;
9557                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9558                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9559                 if (ret) {
9560                         dev_info(&pf->pdev->dev,
9561                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9562                                  i40e_stat_str(&pf->hw, ret),
9563                                  i40e_aq_str(&pf->hw,
9564                                              pf->hw.aq.asq_last_status));
9565                         return -ENOENT;
9566                 }
9567                 vsi->info = ctxt.info;
9568                 vsi->info.valid_sections = 0;
9569
9570                 vsi->seid = ctxt.seid;
9571                 vsi->id = ctxt.vsi_number;
9572
9573                 enabled_tc = i40e_pf_get_tc_map(pf);
9574
9575                 /* MFP mode setup queue map and update VSI */
9576                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9577                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9578                         memset(&ctxt, 0, sizeof(ctxt));
9579                         ctxt.seid = pf->main_vsi_seid;
9580                         ctxt.pf_num = pf->hw.pf_id;
9581                         ctxt.vf_num = 0;
9582                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9583                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9584                         if (ret) {
9585                                 dev_info(&pf->pdev->dev,
9586                                          "update vsi failed, err %s aq_err %s\n",
9587                                          i40e_stat_str(&pf->hw, ret),
9588                                          i40e_aq_str(&pf->hw,
9589                                                     pf->hw.aq.asq_last_status));
9590                                 ret = -ENOENT;
9591                                 goto err;
9592                         }
9593                         /* update the local VSI info queue map */
9594                         i40e_vsi_update_queue_map(vsi, &ctxt);
9595                         vsi->info.valid_sections = 0;
9596                 } else {
9597                         /* Default/Main VSI is only enabled for TC0
9598                          * reconfigure it to enable all TCs that are
9599                          * available on the port in SFP mode.
9600                          * For MFP case the iSCSI PF would use this
9601                          * flow to enable LAN+iSCSI TC.
9602                          */
9603                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9604                         if (ret) {
9605                                 dev_info(&pf->pdev->dev,
9606                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9607                                          enabled_tc,
9608                                          i40e_stat_str(&pf->hw, ret),
9609                                          i40e_aq_str(&pf->hw,
9610                                                     pf->hw.aq.asq_last_status));
9611                                 ret = -ENOENT;
9612                         }
9613                 }
9614                 break;
9615
9616         case I40E_VSI_FDIR:
9617                 ctxt.pf_num = hw->pf_id;
9618                 ctxt.vf_num = 0;
9619                 ctxt.uplink_seid = vsi->uplink_seid;
9620                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9621                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9622                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9623                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9624                         ctxt.info.valid_sections |=
9625                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9626                         ctxt.info.switch_id =
9627                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9628                 }
9629                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9630                 break;
9631
9632         case I40E_VSI_VMDQ2:
9633                 ctxt.pf_num = hw->pf_id;
9634                 ctxt.vf_num = 0;
9635                 ctxt.uplink_seid = vsi->uplink_seid;
9636                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9637                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9638
9639                 /* This VSI is connected to VEB so the switch_id
9640                  * should be set to zero by default.
9641                  */
9642                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9643                         ctxt.info.valid_sections |=
9644                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9645                         ctxt.info.switch_id =
9646                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9647                 }
9648
9649                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9650                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9651                 break;
9652
9653         case I40E_VSI_SRIOV:
9654                 ctxt.pf_num = hw->pf_id;
9655                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9656                 ctxt.uplink_seid = vsi->uplink_seid;
9657                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9658                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9659
9660                 /* This VSI is connected to VEB so the switch_id
9661                  * should be set to zero by default.
9662                  */
9663                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9664                         ctxt.info.valid_sections |=
9665                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9666                         ctxt.info.switch_id =
9667                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9668                 }
9669
9670                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9671                         ctxt.info.valid_sections |=
9672                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9673                         ctxt.info.queueing_opt_flags |=
9674                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9675                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9676                 }
9677
9678                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9679                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9680                 if (pf->vf[vsi->vf_id].spoofchk) {
9681                         ctxt.info.valid_sections |=
9682                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9683                         ctxt.info.sec_flags |=
9684                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9685                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9686                 }
9687                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9688                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9689                 break;
9690
9691         case I40E_VSI_IWARP:
9692                 /* send down message to iWARP */
9693                 break;
9694
9695         default:
9696                 return -ENODEV;
9697         }
9698
9699         if (vsi->type != I40E_VSI_MAIN) {
9700                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9701                 if (ret) {
9702                         dev_info(&vsi->back->pdev->dev,
9703                                  "add vsi failed, err %s aq_err %s\n",
9704                                  i40e_stat_str(&pf->hw, ret),
9705                                  i40e_aq_str(&pf->hw,
9706                                              pf->hw.aq.asq_last_status));
9707                         ret = -ENOENT;
9708                         goto err;
9709                 }
9710                 vsi->info = ctxt.info;
9711                 vsi->info.valid_sections = 0;
9712                 vsi->seid = ctxt.seid;
9713                 vsi->id = ctxt.vsi_number;
9714         }
9715
9716         vsi->active_filters = 0;
9717         clear_bit(__I40E_VSI_OVERFLOW_PROMISC, vsi->state);
9718         spin_lock_bh(&vsi->mac_filter_hash_lock);
9719         /* If macvlan filters already exist, force them to get loaded */
9720         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
9721                 f->state = I40E_FILTER_NEW;
9722                 f_count++;
9723         }
9724         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9725
9726         if (f_count) {
9727                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9728                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9729         }
9730
9731         /* Update VSI BW information */
9732         ret = i40e_vsi_get_bw_info(vsi);
9733         if (ret) {
9734                 dev_info(&pf->pdev->dev,
9735                          "couldn't get vsi bw info, err %s aq_err %s\n",
9736                          i40e_stat_str(&pf->hw, ret),
9737                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9738                 /* VSI is already added so not tearing that up */
9739                 ret = 0;
9740         }
9741
9742 err:
9743         return ret;
9744 }
9745
9746 /**
9747  * i40e_vsi_release - Delete a VSI and free its resources
9748  * @vsi: the VSI being removed
9749  *
9750  * Returns 0 on success or < 0 on error
9751  **/
9752 int i40e_vsi_release(struct i40e_vsi *vsi)
9753 {
9754         struct i40e_mac_filter *f;
9755         struct hlist_node *h;
9756         struct i40e_veb *veb = NULL;
9757         struct i40e_pf *pf;
9758         u16 uplink_seid;
9759         int i, n, bkt;
9760
9761         pf = vsi->back;
9762
9763         /* release of a VEB-owner or last VSI is not allowed */
9764         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9765                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9766                          vsi->seid, vsi->uplink_seid);
9767                 return -ENODEV;
9768         }
9769         if (vsi == pf->vsi[pf->lan_vsi] &&
9770             !test_bit(__I40E_VSI_DOWN, pf->state)) {
9771                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9772                 return -ENODEV;
9773         }
9774
9775         uplink_seid = vsi->uplink_seid;
9776         if (vsi->type != I40E_VSI_SRIOV) {
9777                 if (vsi->netdev_registered) {
9778                         vsi->netdev_registered = false;
9779                         if (vsi->netdev) {
9780                                 /* results in a call to i40e_close() */
9781                                 unregister_netdev(vsi->netdev);
9782                         }
9783                 } else {
9784                         i40e_vsi_close(vsi);
9785                 }
9786                 i40e_vsi_disable_irq(vsi);
9787         }
9788
9789         spin_lock_bh(&vsi->mac_filter_hash_lock);
9790
9791         /* clear the sync flag on all filters */
9792         if (vsi->netdev) {
9793                 __dev_uc_unsync(vsi->netdev, NULL);
9794                 __dev_mc_unsync(vsi->netdev, NULL);
9795         }
9796
9797         /* make sure any remaining filters are marked for deletion */
9798         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
9799                 __i40e_del_filter(vsi, f);
9800
9801         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9802
9803         i40e_sync_vsi_filters(vsi);
9804
9805         i40e_vsi_delete(vsi);
9806         i40e_vsi_free_q_vectors(vsi);
9807         if (vsi->netdev) {
9808                 free_netdev(vsi->netdev);
9809                 vsi->netdev = NULL;
9810         }
9811         i40e_vsi_clear_rings(vsi);
9812         i40e_vsi_clear(vsi);
9813
9814         /* If this was the last thing on the VEB, except for the
9815          * controlling VSI, remove the VEB, which puts the controlling
9816          * VSI onto the next level down in the switch.
9817          *
9818          * Well, okay, there's one more exception here: don't remove
9819          * the orphan VEBs yet.  We'll wait for an explicit remove request
9820          * from up the network stack.
9821          */
9822         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9823                 if (pf->vsi[i] &&
9824                     pf->vsi[i]->uplink_seid == uplink_seid &&
9825                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9826                         n++;      /* count the VSIs */
9827                 }
9828         }
9829         for (i = 0; i < I40E_MAX_VEB; i++) {
9830                 if (!pf->veb[i])
9831                         continue;
9832                 if (pf->veb[i]->uplink_seid == uplink_seid)
9833                         n++;     /* count the VEBs */
9834                 if (pf->veb[i]->seid == uplink_seid)
9835                         veb = pf->veb[i];
9836         }
9837         if (n == 0 && veb && veb->uplink_seid != 0)
9838                 i40e_veb_release(veb);
9839
9840         return 0;
9841 }
9842
9843 /**
9844  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9845  * @vsi: ptr to the VSI
9846  *
9847  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9848  * corresponding SW VSI structure and initializes num_queue_pairs for the
9849  * newly allocated VSI.
9850  *
9851  * Returns 0 on success or negative on failure
9852  **/
9853 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9854 {
9855         int ret = -ENOENT;
9856         struct i40e_pf *pf = vsi->back;
9857
9858         if (vsi->q_vectors[0]) {
9859                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9860                          vsi->seid);
9861                 return -EEXIST;
9862         }
9863
9864         if (vsi->base_vector) {
9865                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9866                          vsi->seid, vsi->base_vector);
9867                 return -EEXIST;
9868         }
9869
9870         ret = i40e_vsi_alloc_q_vectors(vsi);
9871         if (ret) {
9872                 dev_info(&pf->pdev->dev,
9873                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9874                          vsi->num_q_vectors, vsi->seid, ret);
9875                 vsi->num_q_vectors = 0;
9876                 goto vector_setup_out;
9877         }
9878
9879         /* In Legacy mode, we do not have to get any other vector since we
9880          * piggyback on the misc/ICR0 for queue interrupts.
9881         */
9882         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9883                 return ret;
9884         if (vsi->num_q_vectors)
9885                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9886                                                  vsi->num_q_vectors, vsi->idx);
9887         if (vsi->base_vector < 0) {
9888                 dev_info(&pf->pdev->dev,
9889                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9890                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9891                 i40e_vsi_free_q_vectors(vsi);
9892                 ret = -ENOENT;
9893                 goto vector_setup_out;
9894         }
9895
9896 vector_setup_out:
9897         return ret;
9898 }
9899
9900 /**
9901  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9902  * @vsi: pointer to the vsi.
9903  *
9904  * This re-allocates a vsi's queue resources.
9905  *
9906  * Returns pointer to the successfully allocated and configured VSI sw struct
9907  * on success, otherwise returns NULL on failure.
9908  **/
9909 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9910 {
9911         struct i40e_pf *pf;
9912         u8 enabled_tc;
9913         int ret;
9914
9915         if (!vsi)
9916                 return NULL;
9917
9918         pf = vsi->back;
9919
9920         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9921         i40e_vsi_clear_rings(vsi);
9922
9923         i40e_vsi_free_arrays(vsi, false);
9924         i40e_set_num_rings_in_vsi(vsi);
9925         ret = i40e_vsi_alloc_arrays(vsi, false);
9926         if (ret)
9927                 goto err_vsi;
9928
9929         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9930         if (ret < 0) {
9931                 dev_info(&pf->pdev->dev,
9932                          "failed to get tracking for %d queues for VSI %d err %d\n",
9933                          vsi->alloc_queue_pairs, vsi->seid, ret);
9934                 goto err_vsi;
9935         }
9936         vsi->base_queue = ret;
9937
9938         /* Update the FW view of the VSI. Force a reset of TC and queue
9939          * layout configurations.
9940          */
9941         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9942         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9943         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9944         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9945         if (vsi->type == I40E_VSI_MAIN)
9946                 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9947
9948         /* assign it some queues */
9949         ret = i40e_alloc_rings(vsi);
9950         if (ret)
9951                 goto err_rings;
9952
9953         /* map all of the rings to the q_vectors */
9954         i40e_vsi_map_rings_to_vectors(vsi);
9955         return vsi;
9956
9957 err_rings:
9958         i40e_vsi_free_q_vectors(vsi);
9959         if (vsi->netdev_registered) {
9960                 vsi->netdev_registered = false;
9961                 unregister_netdev(vsi->netdev);
9962                 free_netdev(vsi->netdev);
9963                 vsi->netdev = NULL;
9964         }
9965         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9966 err_vsi:
9967         i40e_vsi_clear(vsi);
9968         return NULL;
9969 }
9970
9971 /**
9972  * i40e_vsi_setup - Set up a VSI by a given type
9973  * @pf: board private structure
9974  * @type: VSI type
9975  * @uplink_seid: the switch element to link to
9976  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9977  *
9978  * This allocates the sw VSI structure and its queue resources, then add a VSI
9979  * to the identified VEB.
9980  *
9981  * Returns pointer to the successfully allocated and configure VSI sw struct on
9982  * success, otherwise returns NULL on failure.
9983  **/
9984 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9985                                 u16 uplink_seid, u32 param1)
9986 {
9987         struct i40e_vsi *vsi = NULL;
9988         struct i40e_veb *veb = NULL;
9989         int ret, i;
9990         int v_idx;
9991
9992         /* The requested uplink_seid must be either
9993          *     - the PF's port seid
9994          *              no VEB is needed because this is the PF
9995          *              or this is a Flow Director special case VSI
9996          *     - seid of an existing VEB
9997          *     - seid of a VSI that owns an existing VEB
9998          *     - seid of a VSI that doesn't own a VEB
9999          *              a new VEB is created and the VSI becomes the owner
10000          *     - seid of the PF VSI, which is what creates the first VEB
10001          *              this is a special case of the previous
10002          *
10003          * Find which uplink_seid we were given and create a new VEB if needed
10004          */
10005         for (i = 0; i < I40E_MAX_VEB; i++) {
10006                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
10007                         veb = pf->veb[i];
10008                         break;
10009                 }
10010         }
10011
10012         if (!veb && uplink_seid != pf->mac_seid) {
10013
10014                 for (i = 0; i < pf->num_alloc_vsi; i++) {
10015                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
10016                                 vsi = pf->vsi[i];
10017                                 break;
10018                         }
10019                 }
10020                 if (!vsi) {
10021                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
10022                                  uplink_seid);
10023                         return NULL;
10024                 }
10025
10026                 if (vsi->uplink_seid == pf->mac_seid)
10027                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
10028                                              vsi->tc_config.enabled_tc);
10029                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
10030                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
10031                                              vsi->tc_config.enabled_tc);
10032                 if (veb) {
10033                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
10034                                 dev_info(&vsi->back->pdev->dev,
10035                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
10036                                 return NULL;
10037                         }
10038                         /* We come up by default in VEPA mode if SRIOV is not
10039                          * already enabled, in which case we can't force VEPA
10040                          * mode.
10041                          */
10042                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
10043                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
10044                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
10045                         }
10046                         i40e_config_bridge_mode(veb);
10047                 }
10048                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10049                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10050                                 veb = pf->veb[i];
10051                 }
10052                 if (!veb) {
10053                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
10054                         return NULL;
10055                 }
10056
10057                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10058                 uplink_seid = veb->seid;
10059         }
10060
10061         /* get vsi sw struct */
10062         v_idx = i40e_vsi_mem_alloc(pf, type);
10063         if (v_idx < 0)
10064                 goto err_alloc;
10065         vsi = pf->vsi[v_idx];
10066         if (!vsi)
10067                 goto err_alloc;
10068         vsi->type = type;
10069         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
10070
10071         if (type == I40E_VSI_MAIN)
10072                 pf->lan_vsi = v_idx;
10073         else if (type == I40E_VSI_SRIOV)
10074                 vsi->vf_id = param1;
10075         /* assign it some queues */
10076         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
10077                                 vsi->idx);
10078         if (ret < 0) {
10079                 dev_info(&pf->pdev->dev,
10080                          "failed to get tracking for %d queues for VSI %d err=%d\n",
10081                          vsi->alloc_queue_pairs, vsi->seid, ret);
10082                 goto err_vsi;
10083         }
10084         vsi->base_queue = ret;
10085
10086         /* get a VSI from the hardware */
10087         vsi->uplink_seid = uplink_seid;
10088         ret = i40e_add_vsi(vsi);
10089         if (ret)
10090                 goto err_vsi;
10091
10092         switch (vsi->type) {
10093         /* setup the netdev if needed */
10094         case I40E_VSI_MAIN:
10095                 /* Apply relevant filters if a platform-specific mac
10096                  * address was selected.
10097                  */
10098                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
10099                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
10100                         if (ret) {
10101                                 dev_warn(&pf->pdev->dev,
10102                                          "could not set up macaddr; err %d\n",
10103                                          ret);
10104                         }
10105                 }
10106         case I40E_VSI_VMDQ2:
10107                 ret = i40e_config_netdev(vsi);
10108                 if (ret)
10109                         goto err_netdev;
10110                 ret = register_netdev(vsi->netdev);
10111                 if (ret)
10112                         goto err_netdev;
10113                 vsi->netdev_registered = true;
10114                 netif_carrier_off(vsi->netdev);
10115 #ifdef CONFIG_I40E_DCB
10116                 /* Setup DCB netlink interface */
10117                 i40e_dcbnl_setup(vsi);
10118 #endif /* CONFIG_I40E_DCB */
10119                 /* fall through */
10120
10121         case I40E_VSI_FDIR:
10122                 /* set up vectors and rings if needed */
10123                 ret = i40e_vsi_setup_vectors(vsi);
10124                 if (ret)
10125                         goto err_msix;
10126
10127                 ret = i40e_alloc_rings(vsi);
10128                 if (ret)
10129                         goto err_rings;
10130
10131                 /* map all of the rings to the q_vectors */
10132                 i40e_vsi_map_rings_to_vectors(vsi);
10133
10134                 i40e_vsi_reset_stats(vsi);
10135                 break;
10136
10137         default:
10138                 /* no netdev or rings for the other VSI types */
10139                 break;
10140         }
10141
10142         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
10143             (vsi->type == I40E_VSI_VMDQ2)) {
10144                 ret = i40e_vsi_config_rss(vsi);
10145         }
10146         return vsi;
10147
10148 err_rings:
10149         i40e_vsi_free_q_vectors(vsi);
10150 err_msix:
10151         if (vsi->netdev_registered) {
10152                 vsi->netdev_registered = false;
10153                 unregister_netdev(vsi->netdev);
10154                 free_netdev(vsi->netdev);
10155                 vsi->netdev = NULL;
10156         }
10157 err_netdev:
10158         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10159 err_vsi:
10160         i40e_vsi_clear(vsi);
10161 err_alloc:
10162         return NULL;
10163 }
10164
10165 /**
10166  * i40e_veb_get_bw_info - Query VEB BW information
10167  * @veb: the veb to query
10168  *
10169  * Query the Tx scheduler BW configuration data for given VEB
10170  **/
10171 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
10172 {
10173         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
10174         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
10175         struct i40e_pf *pf = veb->pf;
10176         struct i40e_hw *hw = &pf->hw;
10177         u32 tc_bw_max;
10178         int ret = 0;
10179         int i;
10180
10181         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
10182                                                   &bw_data, NULL);
10183         if (ret) {
10184                 dev_info(&pf->pdev->dev,
10185                          "query veb bw config failed, err %s aq_err %s\n",
10186                          i40e_stat_str(&pf->hw, ret),
10187                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10188                 goto out;
10189         }
10190
10191         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
10192                                                    &ets_data, NULL);
10193         if (ret) {
10194                 dev_info(&pf->pdev->dev,
10195                          "query veb bw ets config failed, err %s aq_err %s\n",
10196                          i40e_stat_str(&pf->hw, ret),
10197                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10198                 goto out;
10199         }
10200
10201         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
10202         veb->bw_max_quanta = ets_data.tc_bw_max;
10203         veb->is_abs_credits = bw_data.absolute_credits_enable;
10204         veb->enabled_tc = ets_data.tc_valid_bits;
10205         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
10206                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
10207         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
10208                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
10209                 veb->bw_tc_limit_credits[i] =
10210                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
10211                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
10212         }
10213
10214 out:
10215         return ret;
10216 }
10217
10218 /**
10219  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10220  * @pf: board private structure
10221  *
10222  * On error: returns error code (negative)
10223  * On success: returns vsi index in PF (positive)
10224  **/
10225 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10226 {
10227         int ret = -ENOENT;
10228         struct i40e_veb *veb;
10229         int i;
10230
10231         /* Need to protect the allocation of switch elements at the PF level */
10232         mutex_lock(&pf->switch_mutex);
10233
10234         /* VEB list may be fragmented if VEB creation/destruction has
10235          * been happening.  We can afford to do a quick scan to look
10236          * for any free slots in the list.
10237          *
10238          * find next empty veb slot, looping back around if necessary
10239          */
10240         i = 0;
10241         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10242                 i++;
10243         if (i >= I40E_MAX_VEB) {
10244                 ret = -ENOMEM;
10245                 goto err_alloc_veb;  /* out of VEB slots! */
10246         }
10247
10248         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10249         if (!veb) {
10250                 ret = -ENOMEM;
10251                 goto err_alloc_veb;
10252         }
10253         veb->pf = pf;
10254         veb->idx = i;
10255         veb->enabled_tc = 1;
10256
10257         pf->veb[i] = veb;
10258         ret = i;
10259 err_alloc_veb:
10260         mutex_unlock(&pf->switch_mutex);
10261         return ret;
10262 }
10263
10264 /**
10265  * i40e_switch_branch_release - Delete a branch of the switch tree
10266  * @branch: where to start deleting
10267  *
10268  * This uses recursion to find the tips of the branch to be
10269  * removed, deleting until we get back to and can delete this VEB.
10270  **/
10271 static void i40e_switch_branch_release(struct i40e_veb *branch)
10272 {
10273         struct i40e_pf *pf = branch->pf;
10274         u16 branch_seid = branch->seid;
10275         u16 veb_idx = branch->idx;
10276         int i;
10277
10278         /* release any VEBs on this VEB - RECURSION */
10279         for (i = 0; i < I40E_MAX_VEB; i++) {
10280                 if (!pf->veb[i])
10281                         continue;
10282                 if (pf->veb[i]->uplink_seid == branch->seid)
10283                         i40e_switch_branch_release(pf->veb[i]);
10284         }
10285
10286         /* Release the VSIs on this VEB, but not the owner VSI.
10287          *
10288          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10289          *       the VEB itself, so don't use (*branch) after this loop.
10290          */
10291         for (i = 0; i < pf->num_alloc_vsi; i++) {
10292                 if (!pf->vsi[i])
10293                         continue;
10294                 if (pf->vsi[i]->uplink_seid == branch_seid &&
10295                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10296                         i40e_vsi_release(pf->vsi[i]);
10297                 }
10298         }
10299
10300         /* There's one corner case where the VEB might not have been
10301          * removed, so double check it here and remove it if needed.
10302          * This case happens if the veb was created from the debugfs
10303          * commands and no VSIs were added to it.
10304          */
10305         if (pf->veb[veb_idx])
10306                 i40e_veb_release(pf->veb[veb_idx]);
10307 }
10308
10309 /**
10310  * i40e_veb_clear - remove veb struct
10311  * @veb: the veb to remove
10312  **/
10313 static void i40e_veb_clear(struct i40e_veb *veb)
10314 {
10315         if (!veb)
10316                 return;
10317
10318         if (veb->pf) {
10319                 struct i40e_pf *pf = veb->pf;
10320
10321                 mutex_lock(&pf->switch_mutex);
10322                 if (pf->veb[veb->idx] == veb)
10323                         pf->veb[veb->idx] = NULL;
10324                 mutex_unlock(&pf->switch_mutex);
10325         }
10326
10327         kfree(veb);
10328 }
10329
10330 /**
10331  * i40e_veb_release - Delete a VEB and free its resources
10332  * @veb: the VEB being removed
10333  **/
10334 void i40e_veb_release(struct i40e_veb *veb)
10335 {
10336         struct i40e_vsi *vsi = NULL;
10337         struct i40e_pf *pf;
10338         int i, n = 0;
10339
10340         pf = veb->pf;
10341
10342         /* find the remaining VSI and check for extras */
10343         for (i = 0; i < pf->num_alloc_vsi; i++) {
10344                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10345                         n++;
10346                         vsi = pf->vsi[i];
10347                 }
10348         }
10349         if (n != 1) {
10350                 dev_info(&pf->pdev->dev,
10351                          "can't remove VEB %d with %d VSIs left\n",
10352                          veb->seid, n);
10353                 return;
10354         }
10355
10356         /* move the remaining VSI to uplink veb */
10357         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10358         if (veb->uplink_seid) {
10359                 vsi->uplink_seid = veb->uplink_seid;
10360                 if (veb->uplink_seid == pf->mac_seid)
10361                         vsi->veb_idx = I40E_NO_VEB;
10362                 else
10363                         vsi->veb_idx = veb->veb_idx;
10364         } else {
10365                 /* floating VEB */
10366                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10367                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10368         }
10369
10370         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10371         i40e_veb_clear(veb);
10372 }
10373
10374 /**
10375  * i40e_add_veb - create the VEB in the switch
10376  * @veb: the VEB to be instantiated
10377  * @vsi: the controlling VSI
10378  **/
10379 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10380 {
10381         struct i40e_pf *pf = veb->pf;
10382         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10383         int ret;
10384
10385         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10386                               veb->enabled_tc, false,
10387                               &veb->seid, enable_stats, NULL);
10388
10389         /* get a VEB from the hardware */
10390         if (ret) {
10391                 dev_info(&pf->pdev->dev,
10392                          "couldn't add VEB, err %s aq_err %s\n",
10393                          i40e_stat_str(&pf->hw, ret),
10394                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10395                 return -EPERM;
10396         }
10397
10398         /* get statistics counter */
10399         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10400                                          &veb->stats_idx, NULL, NULL, NULL);
10401         if (ret) {
10402                 dev_info(&pf->pdev->dev,
10403                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10404                          i40e_stat_str(&pf->hw, ret),
10405                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10406                 return -EPERM;
10407         }
10408         ret = i40e_veb_get_bw_info(veb);
10409         if (ret) {
10410                 dev_info(&pf->pdev->dev,
10411                          "couldn't get VEB bw info, err %s aq_err %s\n",
10412                          i40e_stat_str(&pf->hw, ret),
10413                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10414                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10415                 return -ENOENT;
10416         }
10417
10418         vsi->uplink_seid = veb->seid;
10419         vsi->veb_idx = veb->idx;
10420         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10421
10422         return 0;
10423 }
10424
10425 /**
10426  * i40e_veb_setup - Set up a VEB
10427  * @pf: board private structure
10428  * @flags: VEB setup flags
10429  * @uplink_seid: the switch element to link to
10430  * @vsi_seid: the initial VSI seid
10431  * @enabled_tc: Enabled TC bit-map
10432  *
10433  * This allocates the sw VEB structure and links it into the switch
10434  * It is possible and legal for this to be a duplicate of an already
10435  * existing VEB.  It is also possible for both uplink and vsi seids
10436  * to be zero, in order to create a floating VEB.
10437  *
10438  * Returns pointer to the successfully allocated VEB sw struct on
10439  * success, otherwise returns NULL on failure.
10440  **/
10441 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10442                                 u16 uplink_seid, u16 vsi_seid,
10443                                 u8 enabled_tc)
10444 {
10445         struct i40e_veb *veb, *uplink_veb = NULL;
10446         int vsi_idx, veb_idx;
10447         int ret;
10448
10449         /* if one seid is 0, the other must be 0 to create a floating relay */
10450         if ((uplink_seid == 0 || vsi_seid == 0) &&
10451             (uplink_seid + vsi_seid != 0)) {
10452                 dev_info(&pf->pdev->dev,
10453                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10454                          uplink_seid, vsi_seid);
10455                 return NULL;
10456         }
10457
10458         /* make sure there is such a vsi and uplink */
10459         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10460                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10461                         break;
10462         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10463                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10464                          vsi_seid);
10465                 return NULL;
10466         }
10467
10468         if (uplink_seid && uplink_seid != pf->mac_seid) {
10469                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10470                         if (pf->veb[veb_idx] &&
10471                             pf->veb[veb_idx]->seid == uplink_seid) {
10472                                 uplink_veb = pf->veb[veb_idx];
10473                                 break;
10474                         }
10475                 }
10476                 if (!uplink_veb) {
10477                         dev_info(&pf->pdev->dev,
10478                                  "uplink seid %d not found\n", uplink_seid);
10479                         return NULL;
10480                 }
10481         }
10482
10483         /* get veb sw struct */
10484         veb_idx = i40e_veb_mem_alloc(pf);
10485         if (veb_idx < 0)
10486                 goto err_alloc;
10487         veb = pf->veb[veb_idx];
10488         veb->flags = flags;
10489         veb->uplink_seid = uplink_seid;
10490         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10491         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10492
10493         /* create the VEB in the switch */
10494         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10495         if (ret)
10496                 goto err_veb;
10497         if (vsi_idx == pf->lan_vsi)
10498                 pf->lan_veb = veb->idx;
10499
10500         return veb;
10501
10502 err_veb:
10503         i40e_veb_clear(veb);
10504 err_alloc:
10505         return NULL;
10506 }
10507
10508 /**
10509  * i40e_setup_pf_switch_element - set PF vars based on switch type
10510  * @pf: board private structure
10511  * @ele: element we are building info from
10512  * @num_reported: total number of elements
10513  * @printconfig: should we print the contents
10514  *
10515  * helper function to assist in extracting a few useful SEID values.
10516  **/
10517 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10518                                 struct i40e_aqc_switch_config_element_resp *ele,
10519                                 u16 num_reported, bool printconfig)
10520 {
10521         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10522         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10523         u8 element_type = ele->element_type;
10524         u16 seid = le16_to_cpu(ele->seid);
10525
10526         if (printconfig)
10527                 dev_info(&pf->pdev->dev,
10528                          "type=%d seid=%d uplink=%d downlink=%d\n",
10529                          element_type, seid, uplink_seid, downlink_seid);
10530
10531         switch (element_type) {
10532         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10533                 pf->mac_seid = seid;
10534                 break;
10535         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10536                 /* Main VEB? */
10537                 if (uplink_seid != pf->mac_seid)
10538                         break;
10539                 if (pf->lan_veb == I40E_NO_VEB) {
10540                         int v;
10541
10542                         /* find existing or else empty VEB */
10543                         for (v = 0; v < I40E_MAX_VEB; v++) {
10544                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10545                                         pf->lan_veb = v;
10546                                         break;
10547                                 }
10548                         }
10549                         if (pf->lan_veb == I40E_NO_VEB) {
10550                                 v = i40e_veb_mem_alloc(pf);
10551                                 if (v < 0)
10552                                         break;
10553                                 pf->lan_veb = v;
10554                         }
10555                 }
10556
10557                 pf->veb[pf->lan_veb]->seid = seid;
10558                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10559                 pf->veb[pf->lan_veb]->pf = pf;
10560                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10561                 break;
10562         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10563                 if (num_reported != 1)
10564                         break;
10565                 /* This is immediately after a reset so we can assume this is
10566                  * the PF's VSI
10567                  */
10568                 pf->mac_seid = uplink_seid;
10569                 pf->pf_seid = downlink_seid;
10570                 pf->main_vsi_seid = seid;
10571                 if (printconfig)
10572                         dev_info(&pf->pdev->dev,
10573                                  "pf_seid=%d main_vsi_seid=%d\n",
10574                                  pf->pf_seid, pf->main_vsi_seid);
10575                 break;
10576         case I40E_SWITCH_ELEMENT_TYPE_PF:
10577         case I40E_SWITCH_ELEMENT_TYPE_VF:
10578         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10579         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10580         case I40E_SWITCH_ELEMENT_TYPE_PE:
10581         case I40E_SWITCH_ELEMENT_TYPE_PA:
10582                 /* ignore these for now */
10583                 break;
10584         default:
10585                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10586                          element_type, seid);
10587                 break;
10588         }
10589 }
10590
10591 /**
10592  * i40e_fetch_switch_configuration - Get switch config from firmware
10593  * @pf: board private structure
10594  * @printconfig: should we print the contents
10595  *
10596  * Get the current switch configuration from the device and
10597  * extract a few useful SEID values.
10598  **/
10599 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10600 {
10601         struct i40e_aqc_get_switch_config_resp *sw_config;
10602         u16 next_seid = 0;
10603         int ret = 0;
10604         u8 *aq_buf;
10605         int i;
10606
10607         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10608         if (!aq_buf)
10609                 return -ENOMEM;
10610
10611         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10612         do {
10613                 u16 num_reported, num_total;
10614
10615                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10616                                                 I40E_AQ_LARGE_BUF,
10617                                                 &next_seid, NULL);
10618                 if (ret) {
10619                         dev_info(&pf->pdev->dev,
10620                                  "get switch config failed err %s aq_err %s\n",
10621                                  i40e_stat_str(&pf->hw, ret),
10622                                  i40e_aq_str(&pf->hw,
10623                                              pf->hw.aq.asq_last_status));
10624                         kfree(aq_buf);
10625                         return -ENOENT;
10626                 }
10627
10628                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10629                 num_total = le16_to_cpu(sw_config->header.num_total);
10630
10631                 if (printconfig)
10632                         dev_info(&pf->pdev->dev,
10633                                  "header: %d reported %d total\n",
10634                                  num_reported, num_total);
10635
10636                 for (i = 0; i < num_reported; i++) {
10637                         struct i40e_aqc_switch_config_element_resp *ele =
10638                                 &sw_config->element[i];
10639
10640                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10641                                                      printconfig);
10642                 }
10643         } while (next_seid != 0);
10644
10645         kfree(aq_buf);
10646         return ret;
10647 }
10648
10649 /**
10650  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10651  * @pf: board private structure
10652  * @reinit: if the Main VSI needs to re-initialized.
10653  *
10654  * Returns 0 on success, negative value on failure
10655  **/
10656 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10657 {
10658         u16 flags = 0;
10659         int ret;
10660
10661         /* find out what's out there already */
10662         ret = i40e_fetch_switch_configuration(pf, false);
10663         if (ret) {
10664                 dev_info(&pf->pdev->dev,
10665                          "couldn't fetch switch config, err %s aq_err %s\n",
10666                          i40e_stat_str(&pf->hw, ret),
10667                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10668                 return ret;
10669         }
10670         i40e_pf_reset_stats(pf);
10671
10672         /* set the switch config bit for the whole device to
10673          * support limited promisc or true promisc
10674          * when user requests promisc. The default is limited
10675          * promisc.
10676         */
10677
10678         if ((pf->hw.pf_id == 0) &&
10679             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10680                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10681
10682         if (pf->hw.pf_id == 0) {
10683                 u16 valid_flags;
10684
10685                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10686                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10687                                                 NULL);
10688                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10689                         dev_info(&pf->pdev->dev,
10690                                  "couldn't set switch config bits, err %s aq_err %s\n",
10691                                  i40e_stat_str(&pf->hw, ret),
10692                                  i40e_aq_str(&pf->hw,
10693                                              pf->hw.aq.asq_last_status));
10694                         /* not a fatal problem, just keep going */
10695                 }
10696         }
10697
10698         /* first time setup */
10699         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10700                 struct i40e_vsi *vsi = NULL;
10701                 u16 uplink_seid;
10702
10703                 /* Set up the PF VSI associated with the PF's main VSI
10704                  * that is already in the HW switch
10705                  */
10706                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10707                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10708                 else
10709                         uplink_seid = pf->mac_seid;
10710                 if (pf->lan_vsi == I40E_NO_VSI)
10711                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10712                 else if (reinit)
10713                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10714                 if (!vsi) {
10715                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10716                         i40e_fdir_teardown(pf);
10717                         return -EAGAIN;
10718                 }
10719         } else {
10720                 /* force a reset of TC and queue layout configurations */
10721                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10722
10723                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10724                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10725                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10726         }
10727         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10728
10729         i40e_fdir_sb_setup(pf);
10730
10731         /* Setup static PF queue filter control settings */
10732         ret = i40e_setup_pf_filter_control(pf);
10733         if (ret) {
10734                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10735                          ret);
10736                 /* Failure here should not stop continuing other steps */
10737         }
10738
10739         /* enable RSS in the HW, even for only one queue, as the stack can use
10740          * the hash
10741          */
10742         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10743                 i40e_pf_config_rss(pf);
10744
10745         /* fill in link information and enable LSE reporting */
10746         i40e_link_event(pf);
10747
10748         /* Initialize user-specific link properties */
10749         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10750                                   I40E_AQ_AN_COMPLETED) ? true : false);
10751
10752         i40e_ptp_init(pf);
10753
10754         /* repopulate tunnel port filters */
10755         i40e_sync_udp_filters(pf);
10756
10757         return ret;
10758 }
10759
10760 /**
10761  * i40e_determine_queue_usage - Work out queue distribution
10762  * @pf: board private structure
10763  **/
10764 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10765 {
10766         int queues_left;
10767
10768         pf->num_lan_qps = 0;
10769
10770         /* Find the max queues to be put into basic use.  We'll always be
10771          * using TC0, whether or not DCB is running, and TC0 will get the
10772          * big RSS set.
10773          */
10774         queues_left = pf->hw.func_caps.num_tx_qp;
10775
10776         if ((queues_left == 1) ||
10777             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10778                 /* one qp for PF, no queues for anything else */
10779                 queues_left = 0;
10780                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10781
10782                 /* make sure all the fancies are disabled */
10783                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10784                                I40E_FLAG_IWARP_ENABLED  |
10785                                I40E_FLAG_FD_SB_ENABLED  |
10786                                I40E_FLAG_FD_ATR_ENABLED |
10787                                I40E_FLAG_DCB_CAPABLE    |
10788                                I40E_FLAG_DCB_ENABLED    |
10789                                I40E_FLAG_SRIOV_ENABLED  |
10790                                I40E_FLAG_VMDQ_ENABLED);
10791         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10792                                   I40E_FLAG_FD_SB_ENABLED |
10793                                   I40E_FLAG_FD_ATR_ENABLED |
10794                                   I40E_FLAG_DCB_CAPABLE))) {
10795                 /* one qp for PF */
10796                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10797                 queues_left -= pf->num_lan_qps;
10798
10799                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10800                                I40E_FLAG_IWARP_ENABLED  |
10801                                I40E_FLAG_FD_SB_ENABLED  |
10802                                I40E_FLAG_FD_ATR_ENABLED |
10803                                I40E_FLAG_DCB_ENABLED    |
10804                                I40E_FLAG_VMDQ_ENABLED);
10805         } else {
10806                 /* Not enough queues for all TCs */
10807                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10808                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10809                         pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10810                                         I40E_FLAG_DCB_ENABLED);
10811                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10812                 }
10813                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10814                                         num_online_cpus());
10815                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10816                                         pf->hw.func_caps.num_tx_qp);
10817
10818                 queues_left -= pf->num_lan_qps;
10819         }
10820
10821         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10822                 if (queues_left > 1) {
10823                         queues_left -= 1; /* save 1 queue for FD */
10824                 } else {
10825                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10826                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10827                 }
10828         }
10829
10830         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10831             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10832                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10833                                         (queues_left / pf->num_vf_qps));
10834                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10835         }
10836
10837         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10838             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10839                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10840                                           (queues_left / pf->num_vmdq_qps));
10841                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10842         }
10843
10844         pf->queues_left = queues_left;
10845         dev_dbg(&pf->pdev->dev,
10846                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10847                 pf->hw.func_caps.num_tx_qp,
10848                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10849                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10850                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10851                 queues_left);
10852 }
10853
10854 /**
10855  * i40e_setup_pf_filter_control - Setup PF static filter control
10856  * @pf: PF to be setup
10857  *
10858  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10859  * settings. If PE/FCoE are enabled then it will also set the per PF
10860  * based filter sizes required for them. It also enables Flow director,
10861  * ethertype and macvlan type filter settings for the pf.
10862  *
10863  * Returns 0 on success, negative on failure
10864  **/
10865 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10866 {
10867         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10868
10869         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10870
10871         /* Flow Director is enabled */
10872         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10873                 settings->enable_fdir = true;
10874
10875         /* Ethtype and MACVLAN filters enabled for PF */
10876         settings->enable_ethtype = true;
10877         settings->enable_macvlan = true;
10878
10879         if (i40e_set_filter_control(&pf->hw, settings))
10880                 return -ENOENT;
10881
10882         return 0;
10883 }
10884
10885 #define INFO_STRING_LEN 255
10886 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10887 static void i40e_print_features(struct i40e_pf *pf)
10888 {
10889         struct i40e_hw *hw = &pf->hw;
10890         char *buf;
10891         int i;
10892
10893         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10894         if (!buf)
10895                 return;
10896
10897         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10898 #ifdef CONFIG_PCI_IOV
10899         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10900 #endif
10901         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10902                       pf->hw.func_caps.num_vsis,
10903                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
10904         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10905                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10906         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10907                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10908         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10909                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10910                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10911         }
10912         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10913                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10914         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10915         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10916         if (pf->flags & I40E_FLAG_PTP)
10917                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10918         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10919                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10920         else
10921                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10922
10923         dev_info(&pf->pdev->dev, "%s\n", buf);
10924         kfree(buf);
10925         WARN_ON(i > INFO_STRING_LEN);
10926 }
10927
10928 /**
10929  * i40e_get_platform_mac_addr - get platform-specific MAC address
10930  * @pdev: PCI device information struct
10931  * @pf: board private structure
10932  *
10933  * Look up the MAC address for the device. First we'll try
10934  * eth_platform_get_mac_address, which will check Open Firmware, or arch
10935  * specific fallback. Otherwise, we'll default to the stored value in
10936  * firmware.
10937  **/
10938 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10939 {
10940         if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10941                 i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
10942 }
10943
10944 /**
10945  * i40e_probe - Device initialization routine
10946  * @pdev: PCI device information struct
10947  * @ent: entry in i40e_pci_tbl
10948  *
10949  * i40e_probe initializes a PF identified by a pci_dev structure.
10950  * The OS initialization, configuring of the PF private structure,
10951  * and a hardware reset occur.
10952  *
10953  * Returns 0 on success, negative on failure
10954  **/
10955 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10956 {
10957         struct i40e_aq_get_phy_abilities_resp abilities;
10958         struct i40e_pf *pf;
10959         struct i40e_hw *hw;
10960         static u16 pfs_found;
10961         u16 wol_nvm_bits;
10962         u16 link_status;
10963         int err;
10964         u32 val;
10965         u32 i;
10966         u8 set_fc_aq_fail;
10967
10968         err = pci_enable_device_mem(pdev);
10969         if (err)
10970                 return err;
10971
10972         /* set up for high or low dma */
10973         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10974         if (err) {
10975                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10976                 if (err) {
10977                         dev_err(&pdev->dev,
10978                                 "DMA configuration failed: 0x%x\n", err);
10979                         goto err_dma;
10980                 }
10981         }
10982
10983         /* set up pci connections */
10984         err = pci_request_mem_regions(pdev, i40e_driver_name);
10985         if (err) {
10986                 dev_info(&pdev->dev,
10987                          "pci_request_selected_regions failed %d\n", err);
10988                 goto err_pci_reg;
10989         }
10990
10991         pci_enable_pcie_error_reporting(pdev);
10992         pci_set_master(pdev);
10993
10994         /* Now that we have a PCI connection, we need to do the
10995          * low level device setup.  This is primarily setting up
10996          * the Admin Queue structures and then querying for the
10997          * device's current profile information.
10998          */
10999         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
11000         if (!pf) {
11001                 err = -ENOMEM;
11002                 goto err_pf_alloc;
11003         }
11004         pf->next_vsi = 0;
11005         pf->pdev = pdev;
11006         set_bit(__I40E_VSI_DOWN, pf->state);
11007
11008         hw = &pf->hw;
11009         hw->back = pf;
11010
11011         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
11012                                 I40E_MAX_CSR_SPACE);
11013
11014         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
11015         if (!hw->hw_addr) {
11016                 err = -EIO;
11017                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
11018                          (unsigned int)pci_resource_start(pdev, 0),
11019                          pf->ioremap_len, err);
11020                 goto err_ioremap;
11021         }
11022         hw->vendor_id = pdev->vendor;
11023         hw->device_id = pdev->device;
11024         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
11025         hw->subsystem_vendor_id = pdev->subsystem_vendor;
11026         hw->subsystem_device_id = pdev->subsystem_device;
11027         hw->bus.device = PCI_SLOT(pdev->devfn);
11028         hw->bus.func = PCI_FUNC(pdev->devfn);
11029         hw->bus.bus_id = pdev->bus->number;
11030         pf->instance = pfs_found;
11031
11032         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
11033         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
11034
11035         /* set up the locks for the AQ, do this only once in probe
11036          * and destroy them only once in remove
11037          */
11038         mutex_init(&hw->aq.asq_mutex);
11039         mutex_init(&hw->aq.arq_mutex);
11040
11041         pf->msg_enable = netif_msg_init(debug,
11042                                         NETIF_MSG_DRV |
11043                                         NETIF_MSG_PROBE |
11044                                         NETIF_MSG_LINK);
11045         if (debug < -1)
11046                 pf->hw.debug_mask = debug;
11047
11048         /* do a special CORER for clearing PXE mode once at init */
11049         if (hw->revision_id == 0 &&
11050             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
11051                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
11052                 i40e_flush(hw);
11053                 msleep(200);
11054                 pf->corer_count++;
11055
11056                 i40e_clear_pxe_mode(hw);
11057         }
11058
11059         /* Reset here to make sure all is clean and to define PF 'n' */
11060         i40e_clear_hw(hw);
11061         err = i40e_pf_reset(hw);
11062         if (err) {
11063                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
11064                 goto err_pf_reset;
11065         }
11066         pf->pfr_count++;
11067
11068         hw->aq.num_arq_entries = I40E_AQ_LEN;
11069         hw->aq.num_asq_entries = I40E_AQ_LEN;
11070         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11071         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11072         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
11073
11074         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
11075                  "%s-%s:misc",
11076                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
11077
11078         err = i40e_init_shared_code(hw);
11079         if (err) {
11080                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
11081                          err);
11082                 goto err_pf_reset;
11083         }
11084
11085         /* set up a default setting for link flow control */
11086         pf->hw.fc.requested_mode = I40E_FC_NONE;
11087
11088         err = i40e_init_adminq(hw);
11089         if (err) {
11090                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
11091                         dev_info(&pdev->dev,
11092                                  "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
11093                 else
11094                         dev_info(&pdev->dev,
11095                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
11096
11097                 goto err_pf_reset;
11098         }
11099
11100         /* provide nvm, fw, api versions */
11101         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
11102                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
11103                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
11104                  i40e_nvm_version_str(hw));
11105
11106         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
11107             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
11108                 dev_info(&pdev->dev,
11109                          "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
11110         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
11111                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
11112                 dev_info(&pdev->dev,
11113                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
11114
11115         i40e_verify_eeprom(pf);
11116
11117         /* Rev 0 hardware was never productized */
11118         if (hw->revision_id < 1)
11119                 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");
11120
11121         i40e_clear_pxe_mode(hw);
11122         err = i40e_get_capabilities(pf);
11123         if (err)
11124                 goto err_adminq_setup;
11125
11126         err = i40e_sw_init(pf);
11127         if (err) {
11128                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
11129                 goto err_sw_init;
11130         }
11131
11132         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
11133                                 hw->func_caps.num_rx_qp, 0, 0);
11134         if (err) {
11135                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
11136                 goto err_init_lan_hmc;
11137         }
11138
11139         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
11140         if (err) {
11141                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
11142                 err = -ENOENT;
11143                 goto err_configure_lan_hmc;
11144         }
11145
11146         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
11147          * Ignore error return codes because if it was already disabled via
11148          * hardware settings this will fail
11149          */
11150         if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
11151                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
11152                 i40e_aq_stop_lldp(hw, true, NULL);
11153         }
11154
11155         /* allow a platform config to override the HW addr */
11156         i40e_get_platform_mac_addr(pdev, pf);
11157
11158         if (!is_valid_ether_addr(hw->mac.addr)) {
11159                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
11160                 err = -EIO;
11161                 goto err_mac_addr;
11162         }
11163         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
11164         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
11165         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
11166         if (is_valid_ether_addr(hw->mac.port_addr))
11167                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
11168
11169         pci_set_drvdata(pdev, pf);
11170         pci_save_state(pdev);
11171 #ifdef CONFIG_I40E_DCB
11172         err = i40e_init_pf_dcb(pf);
11173         if (err) {
11174                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
11175                 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
11176                 /* Continue without DCB enabled */
11177         }
11178 #endif /* CONFIG_I40E_DCB */
11179
11180         /* set up periodic task facility */
11181         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11182         pf->service_timer_period = HZ;
11183
11184         INIT_WORK(&pf->service_task, i40e_service_task);
11185         clear_bit(__I40E_SERVICE_SCHED, pf->state);
11186
11187         /* NVM bit on means WoL disabled for the port */
11188         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11189         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11190                 pf->wol_en = false;
11191         else
11192                 pf->wol_en = true;
11193         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11194
11195         /* set up the main switch operations */
11196         i40e_determine_queue_usage(pf);
11197         err = i40e_init_interrupt_scheme(pf);
11198         if (err)
11199                 goto err_switch_setup;
11200
11201         /* The number of VSIs reported by the FW is the minimum guaranteed
11202          * to us; HW supports far more and we share the remaining pool with
11203          * the other PFs. We allocate space for more than the guarantee with
11204          * the understanding that we might not get them all later.
11205          */
11206         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11207                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11208         else
11209                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11210
11211         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11212         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11213                           GFP_KERNEL);
11214         if (!pf->vsi) {
11215                 err = -ENOMEM;
11216                 goto err_switch_setup;
11217         }
11218
11219 #ifdef CONFIG_PCI_IOV
11220         /* prep for VF support */
11221         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11222             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11223             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11224                 if (pci_num_vf(pdev))
11225                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11226         }
11227 #endif
11228         err = i40e_setup_pf_switch(pf, false);
11229         if (err) {
11230                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11231                 goto err_vsis;
11232         }
11233
11234         /* Make sure flow control is set according to current settings */
11235         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11236         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11237                 dev_dbg(&pf->pdev->dev,
11238                         "Set fc with err %s aq_err %s on get_phy_cap\n",
11239                         i40e_stat_str(hw, err),
11240                         i40e_aq_str(hw, hw->aq.asq_last_status));
11241         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11242                 dev_dbg(&pf->pdev->dev,
11243                         "Set fc with err %s aq_err %s on set_phy_config\n",
11244                         i40e_stat_str(hw, err),
11245                         i40e_aq_str(hw, hw->aq.asq_last_status));
11246         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11247                 dev_dbg(&pf->pdev->dev,
11248                         "Set fc with err %s aq_err %s on get_link_info\n",
11249                         i40e_stat_str(hw, err),
11250                         i40e_aq_str(hw, hw->aq.asq_last_status));
11251
11252         /* if FDIR VSI was set up, start it now */
11253         for (i = 0; i < pf->num_alloc_vsi; i++) {
11254                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11255                         i40e_vsi_open(pf->vsi[i]);
11256                         break;
11257                 }
11258         }
11259
11260         /* The driver only wants link up/down and module qualification
11261          * reports from firmware.  Note the negative logic.
11262          */
11263         err = i40e_aq_set_phy_int_mask(&pf->hw,
11264                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
11265                                          I40E_AQ_EVENT_MEDIA_NA |
11266                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11267         if (err)
11268                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11269                          i40e_stat_str(&pf->hw, err),
11270                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11271
11272         /* Reconfigure hardware for allowing smaller MSS in the case
11273          * of TSO, so that we avoid the MDD being fired and causing
11274          * a reset in the case of small MSS+TSO.
11275          */
11276         val = rd32(hw, I40E_REG_MSS);
11277         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11278                 val &= ~I40E_REG_MSS_MIN_MASK;
11279                 val |= I40E_64BYTE_MSS;
11280                 wr32(hw, I40E_REG_MSS, val);
11281         }
11282
11283         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11284                 msleep(75);
11285                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11286                 if (err)
11287                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11288                                  i40e_stat_str(&pf->hw, err),
11289                                  i40e_aq_str(&pf->hw,
11290                                              pf->hw.aq.asq_last_status));
11291         }
11292         /* The main driver is (mostly) up and happy. We need to set this state
11293          * before setting up the misc vector or we get a race and the vector
11294          * ends up disabled forever.
11295          */
11296         clear_bit(__I40E_VSI_DOWN, pf->state);
11297
11298         /* In case of MSIX we are going to setup the misc vector right here
11299          * to handle admin queue events etc. In case of legacy and MSI
11300          * the misc functionality and queue processing is combined in
11301          * the same vector and that gets setup at open.
11302          */
11303         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11304                 err = i40e_setup_misc_vector(pf);
11305                 if (err) {
11306                         dev_info(&pdev->dev,
11307                                  "setup of misc vector failed: %d\n", err);
11308                         goto err_vsis;
11309                 }
11310         }
11311
11312 #ifdef CONFIG_PCI_IOV
11313         /* prep for VF support */
11314         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11315             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11316             !test_bit(__I40E_BAD_EEPROM, pf->state)) {
11317                 /* disable link interrupts for VFs */
11318                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11319                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11320                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11321                 i40e_flush(hw);
11322
11323                 if (pci_num_vf(pdev)) {
11324                         dev_info(&pdev->dev,
11325                                  "Active VFs found, allocating resources.\n");
11326                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11327                         if (err)
11328                                 dev_info(&pdev->dev,
11329                                          "Error %d allocating resources for existing VFs\n",
11330                                          err);
11331                 }
11332         }
11333 #endif /* CONFIG_PCI_IOV */
11334
11335         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11336                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11337                                                       pf->num_iwarp_msix,
11338                                                       I40E_IWARP_IRQ_PILE_ID);
11339                 if (pf->iwarp_base_vector < 0) {
11340                         dev_info(&pdev->dev,
11341                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11342                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11343                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11344                 }
11345         }
11346
11347         i40e_dbg_pf_init(pf);
11348
11349         /* tell the firmware that we're starting */
11350         i40e_send_version(pf);
11351
11352         /* since everything's happy, start the service_task timer */
11353         mod_timer(&pf->service_timer,
11354                   round_jiffies(jiffies + pf->service_timer_period));
11355
11356         /* add this PF to client device list and launch a client service task */
11357         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11358                 err = i40e_lan_add_device(pf);
11359                 if (err)
11360                         dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11361                                  err);
11362         }
11363
11364 #define PCI_SPEED_SIZE 8
11365 #define PCI_WIDTH_SIZE 8
11366         /* Devices on the IOSF bus do not have this information
11367          * and will report PCI Gen 1 x 1 by default so don't bother
11368          * checking them.
11369          */
11370         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11371                 char speed[PCI_SPEED_SIZE] = "Unknown";
11372                 char width[PCI_WIDTH_SIZE] = "Unknown";
11373
11374                 /* Get the negotiated link width and speed from PCI config
11375                  * space
11376                  */
11377                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11378                                           &link_status);
11379
11380                 i40e_set_pci_config_data(hw, link_status);
11381
11382                 switch (hw->bus.speed) {
11383                 case i40e_bus_speed_8000:
11384                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11385                 case i40e_bus_speed_5000:
11386                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11387                 case i40e_bus_speed_2500:
11388                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11389                 default:
11390                         break;
11391                 }
11392                 switch (hw->bus.width) {
11393                 case i40e_bus_width_pcie_x8:
11394                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11395                 case i40e_bus_width_pcie_x4:
11396                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11397                 case i40e_bus_width_pcie_x2:
11398                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11399                 case i40e_bus_width_pcie_x1:
11400                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11401                 default:
11402                         break;
11403                 }
11404
11405                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11406                          speed, width);
11407
11408                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11409                     hw->bus.speed < i40e_bus_speed_8000) {
11410                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11411                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11412                 }
11413         }
11414
11415         /* get the requested speeds from the fw */
11416         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11417         if (err)
11418                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11419                         i40e_stat_str(&pf->hw, err),
11420                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11421         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11422
11423         /* get the supported phy types from the fw */
11424         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11425         if (err)
11426                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11427                         i40e_stat_str(&pf->hw, err),
11428                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11429
11430         /* Add a filter to drop all Flow control frames from any VSI from being
11431          * transmitted. By doing so we stop a malicious VF from sending out
11432          * PAUSE or PFC frames and potentially controlling traffic for other
11433          * PF/VF VSIs.
11434          * The FW can still send Flow control frames if enabled.
11435          */
11436         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11437                                                        pf->main_vsi_seid);
11438
11439         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11440                 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11441                 pf->flags |= I40E_FLAG_PHY_CONTROLS_LEDS;
11442         if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
11443                 pf->flags |= I40E_FLAG_HAVE_CRT_RETIMER;
11444         /* print a string summarizing features */
11445         i40e_print_features(pf);
11446
11447         return 0;
11448
11449         /* Unwind what we've done if something failed in the setup */
11450 err_vsis:
11451         set_bit(__I40E_VSI_DOWN, pf->state);
11452         i40e_clear_interrupt_scheme(pf);
11453         kfree(pf->vsi);
11454 err_switch_setup:
11455         i40e_reset_interrupt_capability(pf);
11456         del_timer_sync(&pf->service_timer);
11457 err_mac_addr:
11458 err_configure_lan_hmc:
11459         (void)i40e_shutdown_lan_hmc(hw);
11460 err_init_lan_hmc:
11461         kfree(pf->qp_pile);
11462 err_sw_init:
11463 err_adminq_setup:
11464 err_pf_reset:
11465         iounmap(hw->hw_addr);
11466 err_ioremap:
11467         kfree(pf);
11468 err_pf_alloc:
11469         pci_disable_pcie_error_reporting(pdev);
11470         pci_release_mem_regions(pdev);
11471 err_pci_reg:
11472 err_dma:
11473         pci_disable_device(pdev);
11474         return err;
11475 }
11476
11477 /**
11478  * i40e_remove - Device removal routine
11479  * @pdev: PCI device information struct
11480  *
11481  * i40e_remove is called by the PCI subsystem to alert the driver
11482  * that is should release a PCI device.  This could be caused by a
11483  * Hot-Plug event, or because the driver is going to be removed from
11484  * memory.
11485  **/
11486 static void i40e_remove(struct pci_dev *pdev)
11487 {
11488         struct i40e_pf *pf = pci_get_drvdata(pdev);
11489         struct i40e_hw *hw = &pf->hw;
11490         i40e_status ret_code;
11491         int i;
11492
11493         i40e_dbg_pf_exit(pf);
11494
11495         i40e_ptp_stop(pf);
11496
11497         /* Disable RSS in hw */
11498         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11499         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11500
11501         /* no more scheduling of any task */
11502         set_bit(__I40E_SUSPENDED, pf->state);
11503         set_bit(__I40E_VSI_DOWN, pf->state);
11504         if (pf->service_timer.data)
11505                 del_timer_sync(&pf->service_timer);
11506         if (pf->service_task.func)
11507                 cancel_work_sync(&pf->service_task);
11508
11509         /* Client close must be called explicitly here because the timer
11510          * has been stopped.
11511          */
11512         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11513
11514         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11515                 i40e_free_vfs(pf);
11516                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11517         }
11518
11519         i40e_fdir_teardown(pf);
11520
11521         /* If there is a switch structure or any orphans, remove them.
11522          * This will leave only the PF's VSI remaining.
11523          */
11524         for (i = 0; i < I40E_MAX_VEB; i++) {
11525                 if (!pf->veb[i])
11526                         continue;
11527
11528                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11529                     pf->veb[i]->uplink_seid == 0)
11530                         i40e_switch_branch_release(pf->veb[i]);
11531         }
11532
11533         /* Now we can shutdown the PF's VSI, just before we kill
11534          * adminq and hmc.
11535          */
11536         if (pf->vsi[pf->lan_vsi])
11537                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11538
11539         /* remove attached clients */
11540         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11541                 ret_code = i40e_lan_del_device(pf);
11542                 if (ret_code)
11543                         dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11544                                  ret_code);
11545         }
11546
11547         /* shutdown and destroy the HMC */
11548         if (hw->hmc.hmc_obj) {
11549                 ret_code = i40e_shutdown_lan_hmc(hw);
11550                 if (ret_code)
11551                         dev_warn(&pdev->dev,
11552                                  "Failed to destroy the HMC resources: %d\n",
11553                                  ret_code);
11554         }
11555
11556         /* shutdown the adminq */
11557         i40e_shutdown_adminq(hw);
11558
11559         /* destroy the locks only once, here */
11560         mutex_destroy(&hw->aq.arq_mutex);
11561         mutex_destroy(&hw->aq.asq_mutex);
11562
11563         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11564         i40e_clear_interrupt_scheme(pf);
11565         for (i = 0; i < pf->num_alloc_vsi; i++) {
11566                 if (pf->vsi[i]) {
11567                         i40e_vsi_clear_rings(pf->vsi[i]);
11568                         i40e_vsi_clear(pf->vsi[i]);
11569                         pf->vsi[i] = NULL;
11570                 }
11571         }
11572
11573         for (i = 0; i < I40E_MAX_VEB; i++) {
11574                 kfree(pf->veb[i]);
11575                 pf->veb[i] = NULL;
11576         }
11577
11578         kfree(pf->qp_pile);
11579         kfree(pf->vsi);
11580
11581         iounmap(hw->hw_addr);
11582         kfree(pf);
11583         pci_release_mem_regions(pdev);
11584
11585         pci_disable_pcie_error_reporting(pdev);
11586         pci_disable_device(pdev);
11587 }
11588
11589 /**
11590  * i40e_pci_error_detected - warning that something funky happened in PCI land
11591  * @pdev: PCI device information struct
11592  *
11593  * Called to warn that something happened and the error handling steps
11594  * are in progress.  Allows the driver to quiesce things, be ready for
11595  * remediation.
11596  **/
11597 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11598                                                 enum pci_channel_state error)
11599 {
11600         struct i40e_pf *pf = pci_get_drvdata(pdev);
11601
11602         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11603
11604         if (!pf) {
11605                 dev_info(&pdev->dev,
11606                          "Cannot recover - error happened during device probe\n");
11607                 return PCI_ERS_RESULT_DISCONNECT;
11608         }
11609
11610         /* shutdown all operations */
11611         if (!test_bit(__I40E_SUSPENDED, pf->state)) {
11612                 rtnl_lock();
11613                 i40e_prep_for_reset(pf, true);
11614                 rtnl_unlock();
11615         }
11616
11617         /* Request a slot reset */
11618         return PCI_ERS_RESULT_NEED_RESET;
11619 }
11620
11621 /**
11622  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11623  * @pdev: PCI device information struct
11624  *
11625  * Called to find if the driver can work with the device now that
11626  * the pci slot has been reset.  If a basic connection seems good
11627  * (registers are readable and have sane content) then return a
11628  * happy little PCI_ERS_RESULT_xxx.
11629  **/
11630 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11631 {
11632         struct i40e_pf *pf = pci_get_drvdata(pdev);
11633         pci_ers_result_t result;
11634         int err;
11635         u32 reg;
11636
11637         dev_dbg(&pdev->dev, "%s\n", __func__);
11638         if (pci_enable_device_mem(pdev)) {
11639                 dev_info(&pdev->dev,
11640                          "Cannot re-enable PCI device after reset.\n");
11641                 result = PCI_ERS_RESULT_DISCONNECT;
11642         } else {
11643                 pci_set_master(pdev);
11644                 pci_restore_state(pdev);
11645                 pci_save_state(pdev);
11646                 pci_wake_from_d3(pdev, false);
11647
11648                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11649                 if (reg == 0)
11650                         result = PCI_ERS_RESULT_RECOVERED;
11651                 else
11652                         result = PCI_ERS_RESULT_DISCONNECT;
11653         }
11654
11655         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11656         if (err) {
11657                 dev_info(&pdev->dev,
11658                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11659                          err);
11660                 /* non-fatal, continue */
11661         }
11662
11663         return result;
11664 }
11665
11666 /**
11667  * i40e_pci_error_resume - restart operations after PCI error recovery
11668  * @pdev: PCI device information struct
11669  *
11670  * Called to allow the driver to bring things back up after PCI error
11671  * and/or reset recovery has finished.
11672  **/
11673 static void i40e_pci_error_resume(struct pci_dev *pdev)
11674 {
11675         struct i40e_pf *pf = pci_get_drvdata(pdev);
11676
11677         dev_dbg(&pdev->dev, "%s\n", __func__);
11678         if (test_bit(__I40E_SUSPENDED, pf->state))
11679                 return;
11680
11681         rtnl_lock();
11682         i40e_handle_reset_warning(pf, true);
11683         rtnl_unlock();
11684 }
11685
11686 /**
11687  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
11688  * using the mac_address_write admin q function
11689  * @pf: pointer to i40e_pf struct
11690  **/
11691 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
11692 {
11693         struct i40e_hw *hw = &pf->hw;
11694         i40e_status ret;
11695         u8 mac_addr[6];
11696         u16 flags = 0;
11697
11698         /* Get current MAC address in case it's an LAA */
11699         if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
11700                 ether_addr_copy(mac_addr,
11701                                 pf->vsi[pf->lan_vsi]->netdev->dev_addr);
11702         } else {
11703                 dev_err(&pf->pdev->dev,
11704                         "Failed to retrieve MAC address; using default\n");
11705                 ether_addr_copy(mac_addr, hw->mac.addr);
11706         }
11707
11708         /* The FW expects the mac address write cmd to first be called with
11709          * one of these flags before calling it again with the multicast
11710          * enable flags.
11711          */
11712         flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
11713
11714         if (hw->func_caps.flex10_enable && hw->partition_id != 1)
11715                 flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
11716
11717         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
11718         if (ret) {
11719                 dev_err(&pf->pdev->dev,
11720                         "Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
11721                 return;
11722         }
11723
11724         flags = I40E_AQC_MC_MAG_EN
11725                         | I40E_AQC_WOL_PRESERVE_ON_PFR
11726                         | I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
11727         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
11728         if (ret)
11729                 dev_err(&pf->pdev->dev,
11730                         "Failed to enable Multicast Magic Packet wake up\n");
11731 }
11732
11733 /**
11734  * i40e_shutdown - PCI callback for shutting down
11735  * @pdev: PCI device information struct
11736  **/
11737 static void i40e_shutdown(struct pci_dev *pdev)
11738 {
11739         struct i40e_pf *pf = pci_get_drvdata(pdev);
11740         struct i40e_hw *hw = &pf->hw;
11741
11742         set_bit(__I40E_SUSPENDED, pf->state);
11743         set_bit(__I40E_VSI_DOWN, pf->state);
11744         rtnl_lock();
11745         i40e_prep_for_reset(pf, true);
11746         rtnl_unlock();
11747
11748         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11749         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11750
11751         del_timer_sync(&pf->service_timer);
11752         cancel_work_sync(&pf->service_task);
11753         i40e_fdir_teardown(pf);
11754
11755         /* Client close must be called explicitly here because the timer
11756          * has been stopped.
11757          */
11758         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11759
11760         if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
11761                 i40e_enable_mc_magic_wake(pf);
11762
11763         rtnl_lock();
11764         i40e_prep_for_reset(pf, true);
11765         rtnl_unlock();
11766
11767         wr32(hw, I40E_PFPM_APM,
11768              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11769         wr32(hw, I40E_PFPM_WUFC,
11770              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11771
11772         i40e_clear_interrupt_scheme(pf);
11773
11774         if (system_state == SYSTEM_POWER_OFF) {
11775                 pci_wake_from_d3(pdev, pf->wol_en);
11776                 pci_set_power_state(pdev, PCI_D3hot);
11777         }
11778 }
11779
11780 #ifdef CONFIG_PM
11781 /**
11782  * i40e_suspend - PCI callback for moving to D3
11783  * @pdev: PCI device information struct
11784  **/
11785 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11786 {
11787         struct i40e_pf *pf = pci_get_drvdata(pdev);
11788         struct i40e_hw *hw = &pf->hw;
11789         int retval = 0;
11790
11791         set_bit(__I40E_SUSPENDED, pf->state);
11792         set_bit(__I40E_VSI_DOWN, pf->state);
11793
11794         if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
11795                 i40e_enable_mc_magic_wake(pf);
11796
11797         rtnl_lock();
11798         i40e_prep_for_reset(pf, true);
11799         rtnl_unlock();
11800
11801         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11802         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11803
11804         i40e_stop_misc_vector(pf);
11805
11806         retval = pci_save_state(pdev);
11807         if (retval)
11808                 return retval;
11809
11810         pci_wake_from_d3(pdev, pf->wol_en);
11811         pci_set_power_state(pdev, PCI_D3hot);
11812
11813         return retval;
11814 }
11815
11816 /**
11817  * i40e_resume - PCI callback for waking up from D3
11818  * @pdev: PCI device information struct
11819  **/
11820 static int i40e_resume(struct pci_dev *pdev)
11821 {
11822         struct i40e_pf *pf = pci_get_drvdata(pdev);
11823         u32 err;
11824
11825         pci_set_power_state(pdev, PCI_D0);
11826         pci_restore_state(pdev);
11827         /* pci_restore_state() clears dev->state_saves, so
11828          * call pci_save_state() again to restore it.
11829          */
11830         pci_save_state(pdev);
11831
11832         err = pci_enable_device_mem(pdev);
11833         if (err) {
11834                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11835                 return err;
11836         }
11837         pci_set_master(pdev);
11838
11839         /* no wakeup events while running */
11840         pci_wake_from_d3(pdev, false);
11841
11842         /* handling the reset will rebuild the device state */
11843         if (test_and_clear_bit(__I40E_SUSPENDED, pf->state)) {
11844                 clear_bit(__I40E_VSI_DOWN, pf->state);
11845                 rtnl_lock();
11846                 i40e_reset_and_rebuild(pf, false, true);
11847                 rtnl_unlock();
11848         }
11849
11850         return 0;
11851 }
11852
11853 #endif
11854 static const struct pci_error_handlers i40e_err_handler = {
11855         .error_detected = i40e_pci_error_detected,
11856         .slot_reset = i40e_pci_error_slot_reset,
11857         .resume = i40e_pci_error_resume,
11858 };
11859
11860 static struct pci_driver i40e_driver = {
11861         .name     = i40e_driver_name,
11862         .id_table = i40e_pci_tbl,
11863         .probe    = i40e_probe,
11864         .remove   = i40e_remove,
11865 #ifdef CONFIG_PM
11866         .suspend  = i40e_suspend,
11867         .resume   = i40e_resume,
11868 #endif
11869         .shutdown = i40e_shutdown,
11870         .err_handler = &i40e_err_handler,
11871         .sriov_configure = i40e_pci_sriov_configure,
11872 };
11873
11874 /**
11875  * i40e_init_module - Driver registration routine
11876  *
11877  * i40e_init_module is the first routine called when the driver is
11878  * loaded. All it does is register with the PCI subsystem.
11879  **/
11880 static int __init i40e_init_module(void)
11881 {
11882         pr_info("%s: %s - version %s\n", i40e_driver_name,
11883                 i40e_driver_string, i40e_driver_version_str);
11884         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11885
11886         /* we will see if single thread per module is enough for now,
11887          * it can't be any worse than using the system workqueue which
11888          * was already single threaded
11889          */
11890         i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
11891                                   i40e_driver_name);
11892         if (!i40e_wq) {
11893                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11894                 return -ENOMEM;
11895         }
11896
11897         i40e_dbg_init();
11898         return pci_register_driver(&i40e_driver);
11899 }
11900 module_init(i40e_init_module);
11901
11902 /**
11903  * i40e_exit_module - Driver exit cleanup routine
11904  *
11905  * i40e_exit_module is called just before the driver is removed
11906  * from memory.
11907  **/
11908 static void __exit i40e_exit_module(void)
11909 {
11910         pci_unregister_driver(&i40e_driver);
11911         destroy_workqueue(i40e_wq);
11912         i40e_dbg_exit();
11913 }
11914 module_exit(i40e_exit_module);