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 #if IS_ENABLED(CONFIG_VXLAN)
  35 #include <net/vxlan.h>
  36 #endif
  37 #if IS_ENABLED(CONFIG_GENEVE)
  38 #include <net/geneve.h>
  39 #endif
  40
  41 const char i40e_driver_name[] = "i40e";
  42 static const char i40e_driver_string[] =
  43                         "Intel(R) Ethernet Connection XL710 Network Driver";
  44
  45 #define DRV_KERN "-k"
  46
  47 #define DRV_VERSION_MAJOR 1
  48 #define DRV_VERSION_MINOR 5
  49 #define DRV_VERSION_BUILD 16
  50 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  51              __stringify(DRV_VERSION_MINOR) "." \
  52              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  53 const char i40e_driver_version_str[] = DRV_VERSION;
  54 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  55
  56 /* a bit of forward declarations */
  57 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  58 static void i40e_handle_reset_warning(struct i40e_pf *pf);
  59 static int i40e_add_vsi(struct i40e_vsi *vsi);
  60 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  61 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  62 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  63 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  64 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  65 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
  66                               u16 rss_table_size, u16 rss_size);
  67 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  68 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  69
  70 /* i40e_pci_tbl - PCI Device ID Table
  71  *
  72  * Last entry must be all 0s
  73  *
  74  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  75  *   Class, Class Mask, private data (not used) }
  76  */
  77 static const struct pci_device_id i40e_pci_tbl[] = {
  78         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  79         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
  90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  91         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  92         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  93         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
  94         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_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         /* required last entry */
  98         {0, }
  99 };
 100 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
 101
 102 #define I40E_MAX_VF_COUNT 128
 103 static int debug = -1;
 104 module_param(debug, int, 0);
 105 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 106
 107 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 108 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 109 MODULE_LICENSE("GPL");
 110 MODULE_VERSION(DRV_VERSION);
 111
 112 static struct workqueue_struct *i40e_wq;
 113
 114 /**
 115  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 116  * @hw:   pointer to the HW structure
 117  * @mem:  ptr to mem struct to fill out
 118  * @size: size of memory requested
 119  * @alignment: what to align the allocation to
 120  **/
 121 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 122                             u64 size, u32 alignment)
 123 {
 124         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 125
 126         mem->size = ALIGN(size, alignment);
 127         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 128                                       &mem->pa, GFP_KERNEL);
 129         if (!mem->va)
 130                 return -ENOMEM;
 131
 132         return 0;
 133 }
 134
 135 /**
 136  * i40e_free_dma_mem_d - OS specific memory free for shared code
 137  * @hw:   pointer to the HW structure
 138  * @mem:  ptr to mem struct to free
 139  **/
 140 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 141 {
 142         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 143
 144         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 145         mem->va = NULL;
 146         mem->pa = 0;
 147         mem->size = 0;
 148
 149         return 0;
 150 }
 151
 152 /**
 153  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 154  * @hw:   pointer to the HW structure
 155  * @mem:  ptr to mem struct to fill out
 156  * @size: size of memory requested
 157  **/
 158 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 159                              u32 size)
 160 {
 161         mem->size = size;
 162         mem->va = kzalloc(size, GFP_KERNEL);
 163
 164         if (!mem->va)
 165                 return -ENOMEM;
 166
 167         return 0;
 168 }
 169
 170 /**
 171  * i40e_free_virt_mem_d - OS specific memory free for shared code
 172  * @hw:   pointer to the HW structure
 173  * @mem:  ptr to mem struct to free
 174  **/
 175 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 176 {
 177         /* it's ok to kfree a NULL pointer */
 178         kfree(mem->va);
 179         mem->va = NULL;
 180         mem->size = 0;
 181
 182         return 0;
 183 }
 184
 185 /**
 186  * i40e_get_lump - find a lump of free generic resource
 187  * @pf: board private structure
 188  * @pile: the pile of resource to search
 189  * @needed: the number of items needed
 190  * @id: an owner id to stick on the items assigned
 191  *
 192  * Returns the base item index of the lump, or negative for error
 193  *
 194  * The search_hint trick and lack of advanced fit-finding only work
 195  * because we're highly likely to have all the same size lump requests.
 196  * Linear search time and any fragmentation should be minimal.
 197  **/
 198 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 199                          u16 needed, u16 id)
 200 {
 201         int ret = -ENOMEM;
 202         int i, j;
 203
 204         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 205                 dev_info(&pf->pdev->dev,
 206                          "param err: pile=%p needed=%d id=0x%04x\n",
 207                          pile, needed, id);
 208                 return -EINVAL;
 209         }
 210
 211         /* start the linear search with an imperfect hint */
 212         i = pile->search_hint;
 213         while (i < pile->num_entries) {
 214                 /* skip already allocated entries */
 215                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 216                         i++;
 217                         continue;
 218                 }
 219
 220                 /* do we have enough in this lump? */
 221                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 222                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 223                                 break;
 224                 }
 225
 226                 if (j == needed) {
 227                         /* there was enough, so assign it to the requestor */
 228                         for (j = 0; j < needed; j++)
 229                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 230                         ret = i;
 231                         pile->search_hint = i + j;
 232                         break;
 233                 }
 234
 235                 /* not enough, so skip over it and continue looking */
 236                 i += j;
 237         }
 238
 239         return ret;
 240 }
 241
 242 /**
 243  * i40e_put_lump - return a lump of generic resource
 244  * @pile: the pile of resource to search
 245  * @index: the base item index
 246  * @id: the owner id of the items assigned
 247  *
 248  * Returns the count of items in the lump
 249  **/
 250 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 251 {
 252         int valid_id = (id | I40E_PILE_VALID_BIT);
 253         int count = 0;
 254         int i;
 255
 256         if (!pile || index >= pile->num_entries)
 257                 return -EINVAL;
 258
 259         for (i = index;
 260              i < pile->num_entries && pile->list[i] == valid_id;
 261              i++) {
 262                 pile->list[i] = 0;
 263                 count++;
 264         }
 265
 266         if (count && index < pile->search_hint)
 267                 pile->search_hint = index;
 268
 269         return count;
 270 }
 271
 272 /**
 273  * i40e_find_vsi_from_id - searches for the vsi with the given id
 274  * @pf - the pf structure to search for the vsi
 275  * @id - id of the vsi it is searching for
 276  **/
 277 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 278 {
 279         int i;
 280
 281         for (i = 0; i < pf->num_alloc_vsi; i++)
 282                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 283                         return pf->vsi[i];
 284
 285         return NULL;
 286 }
 287
 288 /**
 289  * i40e_service_event_schedule - Schedule the service task to wake up
 290  * @pf: board private structure
 291  *
 292  * If not already scheduled, this puts the task into the work queue
 293  **/
 294 void i40e_service_event_schedule(struct i40e_pf *pf)
 295 {
 296         if (!test_bit(__I40E_DOWN, &pf->state) &&
 297             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
 298             !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
 299                 queue_work(i40e_wq, &pf->service_task);
 300 }
 301
 302 /**
 303  * i40e_tx_timeout - Respond to a Tx Hang
 304  * @netdev: network interface device structure
 305  *
 306  * If any port has noticed a Tx timeout, it is likely that the whole
 307  * device is munged, not just the one netdev port, so go for the full
 308  * reset.
 309  **/
 310 #ifdef I40E_FCOE
 311 void i40e_tx_timeout(struct net_device *netdev)
 312 #else
 313 static void i40e_tx_timeout(struct net_device *netdev)
 314 #endif
 315 {
 316         struct i40e_netdev_priv *np = netdev_priv(netdev);
 317         struct i40e_vsi *vsi = np->vsi;
 318         struct i40e_pf *pf = vsi->back;
 319         struct i40e_ring *tx_ring = NULL;
 320         unsigned int i, hung_queue = 0;
 321         u32 head, val;
 322
 323         pf->tx_timeout_count++;
 324
 325         /* find the stopped queue the same way the stack does */
 326         for (i = 0; i < netdev->num_tx_queues; i++) {
 327                 struct netdev_queue *q;
 328                 unsigned long trans_start;
 329
 330                 q = netdev_get_tx_queue(netdev, i);
 331                 trans_start = q->trans_start;
 332                 if (netif_xmit_stopped(q) &&
 333                     time_after(jiffies,
 334                                (trans_start + netdev->watchdog_timeo))) {
 335                         hung_queue = i;
 336                         break;
 337                 }
 338         }
 339
 340         if (i == netdev->num_tx_queues) {
 341                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 342         } else {
 343                 /* now that we have an index, find the tx_ring struct */
 344                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 345                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 346                                 if (hung_queue ==
 347                                     vsi->tx_rings[i]->queue_index) {
 348                                         tx_ring = vsi->tx_rings[i];
 349                                         break;
 350                                 }
 351                         }
 352                 }
 353         }
 354
 355         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 356                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 357         else if (time_before(jiffies,
 358                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 359                 return;   /* don't do any new action before the next timeout */
 360
 361         if (tx_ring) {
 362                 head = i40e_get_head(tx_ring);
 363                 /* Read interrupt register */
 364                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 365                         val = rd32(&pf->hw,
 366                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 367                                                 tx_ring->vsi->base_vector - 1));
 368                 else
 369                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 370
 371                 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",
 372                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 373                             head, tx_ring->next_to_use,
 374                             readl(tx_ring->tail), val);
 375         }
 376
 377         pf->tx_timeout_last_recovery = jiffies;
 378         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 379                     pf->tx_timeout_recovery_level, hung_queue);
 380
 381         switch (pf->tx_timeout_recovery_level) {
 382         case 1:
 383                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
 384                 break;
 385         case 2:
 386                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
 387                 break;
 388         case 3:
 389                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
 390                 break;
 391         default:
 392                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 393                 break;
 394         }
 395
 396         i40e_service_event_schedule(pf);
 397         pf->tx_timeout_recovery_level++;
 398 }
 399
 400 /**
 401  * i40e_get_vsi_stats_struct - Get System Network Statistics
 402  * @vsi: the VSI we care about
 403  *
 404  * Returns the address of the device statistics structure.
 405  * The statistics are actually updated from the service task.
 406  **/
 407 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 408 {
 409         return &vsi->net_stats;
 410 }
 411
 412 /**
 413  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 414  * @netdev: network interface device structure
 415  *
 416  * Returns the address of the device statistics structure.
 417  * The statistics are actually updated from the service task.
 418  **/
 419 #ifdef I40E_FCOE
 420 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 421                                              struct net_device *netdev,
 422                                              struct rtnl_link_stats64 *stats)
 423 #else
 424 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 425                                              struct net_device *netdev,
 426                                              struct rtnl_link_stats64 *stats)
 427 #endif
 428 {
 429         struct i40e_netdev_priv *np = netdev_priv(netdev);
 430         struct i40e_ring *tx_ring, *rx_ring;
 431         struct i40e_vsi *vsi = np->vsi;
 432         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 433         int i;
 434
 435         if (test_bit(__I40E_DOWN, &vsi->state))
 436                 return stats;
 437
 438         if (!vsi->tx_rings)
 439                 return stats;
 440
 441         rcu_read_lock();
 442         for (i = 0; i < vsi->num_queue_pairs; i++) {
 443                 u64 bytes, packets;
 444                 unsigned int start;
 445
 446                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 447                 if (!tx_ring)
 448                         continue;
 449
 450                 do {
 451                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 452                         packets = tx_ring->stats.packets;
 453                         bytes   = tx_ring->stats.bytes;
 454                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 455
 456                 stats->tx_packets += packets;
 457                 stats->tx_bytes   += bytes;
 458                 rx_ring = &tx_ring[1];
 459
 460                 do {
 461                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 462                         packets = rx_ring->stats.packets;
 463                         bytes   = rx_ring->stats.bytes;
 464                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 465
 466                 stats->rx_packets += packets;
 467                 stats->rx_bytes   += bytes;
 468         }
 469         rcu_read_unlock();
 470
 471         /* following stats updated by i40e_watchdog_subtask() */
 472         stats->multicast        = vsi_stats->multicast;
 473         stats->tx_errors        = vsi_stats->tx_errors;
 474         stats->tx_dropped       = vsi_stats->tx_dropped;
 475         stats->rx_errors        = vsi_stats->rx_errors;
 476         stats->rx_dropped       = vsi_stats->rx_dropped;
 477         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 478         stats->rx_length_errors = vsi_stats->rx_length_errors;
 479
 480         return stats;
 481 }
 482
 483 /**
 484  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 485  * @vsi: the VSI to have its stats reset
 486  **/
 487 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 488 {
 489         struct rtnl_link_stats64 *ns;
 490         int i;
 491
 492         if (!vsi)
 493                 return;
 494
 495         ns = i40e_get_vsi_stats_struct(vsi);
 496         memset(ns, 0, sizeof(*ns));
 497         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 498         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 499         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 500         if (vsi->rx_rings && vsi->rx_rings[0]) {
 501                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 502                         memset(&vsi->rx_rings[i]->stats, 0,
 503                                sizeof(vsi->rx_rings[i]->stats));
 504                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 505                                sizeof(vsi->rx_rings[i]->rx_stats));
 506                         memset(&vsi->tx_rings[i]->stats, 0,
 507                                sizeof(vsi->tx_rings[i]->stats));
 508                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 509                                sizeof(vsi->tx_rings[i]->tx_stats));
 510                 }
 511         }
 512         vsi->stat_offsets_loaded = false;
 513 }
 514
 515 /**
 516  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 517  * @pf: the PF to be reset
 518  **/
 519 void i40e_pf_reset_stats(struct i40e_pf *pf)
 520 {
 521         int i;
 522
 523         memset(&pf->stats, 0, sizeof(pf->stats));
 524         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 525         pf->stat_offsets_loaded = false;
 526
 527         for (i = 0; i < I40E_MAX_VEB; i++) {
 528                 if (pf->veb[i]) {
 529                         memset(&pf->veb[i]->stats, 0,
 530                                sizeof(pf->veb[i]->stats));
 531                         memset(&pf->veb[i]->stats_offsets, 0,
 532                                sizeof(pf->veb[i]->stats_offsets));
 533                         pf->veb[i]->stat_offsets_loaded = false;
 534                 }
 535         }
 536 }
 537
 538 /**
 539  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 540  * @hw: ptr to the hardware info
 541  * @hireg: the high 32 bit reg to read
 542  * @loreg: the low 32 bit reg to read
 543  * @offset_loaded: has the initial offset been loaded yet
 544  * @offset: ptr to current offset value
 545  * @stat: ptr to the stat
 546  *
 547  * Since the device stats are not reset at PFReset, they likely will not
 548  * be zeroed when the driver starts.  We'll save the first values read
 549  * and use them as offsets to be subtracted from the raw values in order
 550  * to report stats that count from zero.  In the process, we also manage
 551  * the potential roll-over.
 552  **/
 553 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 554                                bool offset_loaded, u64 *offset, u64 *stat)
 555 {
 556         u64 new_data;
 557
 558         if (hw->device_id == I40E_DEV_ID_QEMU) {
 559                 new_data = rd32(hw, loreg);
 560                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 561         } else {
 562                 new_data = rd64(hw, loreg);
 563         }
 564         if (!offset_loaded)
 565                 *offset = new_data;
 566         if (likely(new_data >= *offset))
 567                 *stat = new_data - *offset;
 568         else
 569                 *stat = (new_data + BIT_ULL(48)) - *offset;
 570         *stat &= 0xFFFFFFFFFFFFULL;
 571 }
 572
 573 /**
 574  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 575  * @hw: ptr to the hardware info
 576  * @reg: the hw reg to read
 577  * @offset_loaded: has the initial offset been loaded yet
 578  * @offset: ptr to current offset value
 579  * @stat: ptr to the stat
 580  **/
 581 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 582                                bool offset_loaded, u64 *offset, u64 *stat)
 583 {
 584         u32 new_data;
 585
 586         new_data = rd32(hw, reg);
 587         if (!offset_loaded)
 588                 *offset = new_data;
 589         if (likely(new_data >= *offset))
 590                 *stat = (u32)(new_data - *offset);
 591         else
 592                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 593 }
 594
 595 /**
 596  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 597  * @vsi: the VSI to be updated
 598  **/
 599 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 600 {
 601         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 602         struct i40e_pf *pf = vsi->back;
 603         struct i40e_hw *hw = &pf->hw;
 604         struct i40e_eth_stats *oes;
 605         struct i40e_eth_stats *es;     /* device's eth stats */
 606
 607         es = &vsi->eth_stats;
 608         oes = &vsi->eth_stats_offsets;
 609
 610         /* Gather up the stats that the hw collects */
 611         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 612                            vsi->stat_offsets_loaded,
 613                            &oes->tx_errors, &es->tx_errors);
 614         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 615                            vsi->stat_offsets_loaded,
 616                            &oes->rx_discards, &es->rx_discards);
 617         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 618                            vsi->stat_offsets_loaded,
 619                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 620         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 621                            vsi->stat_offsets_loaded,
 622                            &oes->tx_errors, &es->tx_errors);
 623
 624         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 625                            I40E_GLV_GORCL(stat_idx),
 626                            vsi->stat_offsets_loaded,
 627                            &oes->rx_bytes, &es->rx_bytes);
 628         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 629                            I40E_GLV_UPRCL(stat_idx),
 630                            vsi->stat_offsets_loaded,
 631                            &oes->rx_unicast, &es->rx_unicast);
 632         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 633                            I40E_GLV_MPRCL(stat_idx),
 634                            vsi->stat_offsets_loaded,
 635                            &oes->rx_multicast, &es->rx_multicast);
 636         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 637                            I40E_GLV_BPRCL(stat_idx),
 638                            vsi->stat_offsets_loaded,
 639                            &oes->rx_broadcast, &es->rx_broadcast);
 640
 641         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 642                            I40E_GLV_GOTCL(stat_idx),
 643                            vsi->stat_offsets_loaded,
 644                            &oes->tx_bytes, &es->tx_bytes);
 645         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 646                            I40E_GLV_UPTCL(stat_idx),
 647                            vsi->stat_offsets_loaded,
 648                            &oes->tx_unicast, &es->tx_unicast);
 649         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 650                            I40E_GLV_MPTCL(stat_idx),
 651                            vsi->stat_offsets_loaded,
 652                            &oes->tx_multicast, &es->tx_multicast);
 653         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 654                            I40E_GLV_BPTCL(stat_idx),
 655                            vsi->stat_offsets_loaded,
 656                            &oes->tx_broadcast, &es->tx_broadcast);
 657         vsi->stat_offsets_loaded = true;
 658 }
 659
 660 /**
 661  * i40e_update_veb_stats - Update Switch component statistics
 662  * @veb: the VEB being updated
 663  **/
 664 static void i40e_update_veb_stats(struct i40e_veb *veb)
 665 {
 666         struct i40e_pf *pf = veb->pf;
 667         struct i40e_hw *hw = &pf->hw;
 668         struct i40e_eth_stats *oes;
 669         struct i40e_eth_stats *es;     /* device's eth stats */
 670         struct i40e_veb_tc_stats *veb_oes;
 671         struct i40e_veb_tc_stats *veb_es;
 672         int i, idx = 0;
 673
 674         idx = veb->stats_idx;
 675         es = &veb->stats;
 676         oes = &veb->stats_offsets;
 677         veb_es = &veb->tc_stats;
 678         veb_oes = &veb->tc_stats_offsets;
 679
 680         /* Gather up the stats that the hw collects */
 681         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 682                            veb->stat_offsets_loaded,
 683                            &oes->tx_discards, &es->tx_discards);
 684         if (hw->revision_id > 0)
 685                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 686                                    veb->stat_offsets_loaded,
 687                                    &oes->rx_unknown_protocol,
 688                                    &es->rx_unknown_protocol);
 689         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 690                            veb->stat_offsets_loaded,
 691                            &oes->rx_bytes, &es->rx_bytes);
 692         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 693                            veb->stat_offsets_loaded,
 694                            &oes->rx_unicast, &es->rx_unicast);
 695         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 696                            veb->stat_offsets_loaded,
 697                            &oes->rx_multicast, &es->rx_multicast);
 698         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 699                            veb->stat_offsets_loaded,
 700                            &oes->rx_broadcast, &es->rx_broadcast);
 701
 702         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 703                            veb->stat_offsets_loaded,
 704                            &oes->tx_bytes, &es->tx_bytes);
 705         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 706                            veb->stat_offsets_loaded,
 707                            &oes->tx_unicast, &es->tx_unicast);
 708         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 709                            veb->stat_offsets_loaded,
 710                            &oes->tx_multicast, &es->tx_multicast);
 711         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 712                            veb->stat_offsets_loaded,
 713                            &oes->tx_broadcast, &es->tx_broadcast);
 714         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 715                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 716                                    I40E_GLVEBTC_RPCL(i, idx),
 717                                    veb->stat_offsets_loaded,
 718                                    &veb_oes->tc_rx_packets[i],
 719                                    &veb_es->tc_rx_packets[i]);
 720                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 721                                    I40E_GLVEBTC_RBCL(i, idx),
 722                                    veb->stat_offsets_loaded,
 723                                    &veb_oes->tc_rx_bytes[i],
 724                                    &veb_es->tc_rx_bytes[i]);
 725                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 726                                    I40E_GLVEBTC_TPCL(i, idx),
 727                                    veb->stat_offsets_loaded,
 728                                    &veb_oes->tc_tx_packets[i],
 729                                    &veb_es->tc_tx_packets[i]);
 730                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 731                                    I40E_GLVEBTC_TBCL(i, idx),
 732                                    veb->stat_offsets_loaded,
 733                                    &veb_oes->tc_tx_bytes[i],
 734                                    &veb_es->tc_tx_bytes[i]);
 735         }
 736         veb->stat_offsets_loaded = true;
 737 }
 738
 739 #ifdef I40E_FCOE
 740 /**
 741  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
 742  * @vsi: the VSI that is capable of doing FCoE
 743  **/
 744 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
 745 {
 746         struct i40e_pf *pf = vsi->back;
 747         struct i40e_hw *hw = &pf->hw;
 748         struct i40e_fcoe_stats *ofs;
 749         struct i40e_fcoe_stats *fs;     /* device's eth stats */
 750         int idx;
 751
 752         if (vsi->type != I40E_VSI_FCOE)
 753                 return;
 754
 755         idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
 756         fs = &vsi->fcoe_stats;
 757         ofs = &vsi->fcoe_stats_offsets;
 758
 759         i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
 760                            vsi->fcoe_stat_offsets_loaded,
 761                            &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
 762         i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
 763                            vsi->fcoe_stat_offsets_loaded,
 764                            &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
 765         i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
 766                            vsi->fcoe_stat_offsets_loaded,
 767                            &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
 768         i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
 769                            vsi->fcoe_stat_offsets_loaded,
 770                            &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
 771         i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
 772                            vsi->fcoe_stat_offsets_loaded,
 773                            &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
 774         i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
 775                            vsi->fcoe_stat_offsets_loaded,
 776                            &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
 777         i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
 778                            vsi->fcoe_stat_offsets_loaded,
 779                            &ofs->fcoe_last_error, &fs->fcoe_last_error);
 780         i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
 781                            vsi->fcoe_stat_offsets_loaded,
 782                            &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
 783
 784         vsi->fcoe_stat_offsets_loaded = true;
 785 }
 786
 787 #endif
 788 /**
 789  * i40e_update_vsi_stats - Update the vsi statistics counters.
 790  * @vsi: the VSI to be updated
 791  *
 792  * There are a few instances where we store the same stat in a
 793  * couple of different structs.  This is partly because we have
 794  * the netdev stats that need to be filled out, which is slightly
 795  * different from the "eth_stats" defined by the chip and used in
 796  * VF communications.  We sort it out here.
 797  **/
 798 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 799 {
 800         struct i40e_pf *pf = vsi->back;
 801         struct rtnl_link_stats64 *ons;
 802         struct rtnl_link_stats64 *ns;   /* netdev stats */
 803         struct i40e_eth_stats *oes;
 804         struct i40e_eth_stats *es;     /* device's eth stats */
 805         u32 tx_restart, tx_busy;
 806         u64 tx_lost_interrupt;
 807         struct i40e_ring *p;
 808         u32 rx_page, rx_buf;
 809         u64 bytes, packets;
 810         unsigned int start;
 811         u64 tx_linearize;
 812         u64 tx_force_wb;
 813         u64 rx_p, rx_b;
 814         u64 tx_p, tx_b;
 815         u16 q;
 816
 817         if (test_bit(__I40E_DOWN, &vsi->state) ||
 818             test_bit(__I40E_CONFIG_BUSY, &pf->state))
 819                 return;
 820
 821         ns = i40e_get_vsi_stats_struct(vsi);
 822         ons = &vsi->net_stats_offsets;
 823         es = &vsi->eth_stats;
 824         oes = &vsi->eth_stats_offsets;
 825
 826         /* Gather up the netdev and vsi stats that the driver collects
 827          * on the fly during packet processing
 828          */
 829         rx_b = rx_p = 0;
 830         tx_b = tx_p = 0;
 831         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 832         tx_lost_interrupt = 0;
 833         rx_page = 0;
 834         rx_buf = 0;
 835         rcu_read_lock();
 836         for (q = 0; q < vsi->num_queue_pairs; q++) {
 837                 /* locate Tx ring */
 838                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 839
 840                 do {
 841                         start = u64_stats_fetch_begin_irq(&p->syncp);
 842                         packets = p->stats.packets;
 843                         bytes = p->stats.bytes;
 844                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 845                 tx_b += bytes;
 846                 tx_p += packets;
 847                 tx_restart += p->tx_stats.restart_queue;
 848                 tx_busy += p->tx_stats.tx_busy;
 849                 tx_linearize += p->tx_stats.tx_linearize;
 850                 tx_force_wb += p->tx_stats.tx_force_wb;
 851                 tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
 852
 853                 /* Rx queue is part of the same block as Tx queue */
 854                 p = &p[1];
 855                 do {
 856                         start = u64_stats_fetch_begin_irq(&p->syncp);
 857                         packets = p->stats.packets;
 858                         bytes = p->stats.bytes;
 859                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 860                 rx_b += bytes;
 861                 rx_p += packets;
 862                 rx_buf += p->rx_stats.alloc_buff_failed;
 863                 rx_page += p->rx_stats.alloc_page_failed;
 864         }
 865         rcu_read_unlock();
 866         vsi->tx_restart = tx_restart;
 867         vsi->tx_busy = tx_busy;
 868         vsi->tx_linearize = tx_linearize;
 869         vsi->tx_force_wb = tx_force_wb;
 870         vsi->tx_lost_interrupt = tx_lost_interrupt;
 871         vsi->rx_page_failed = rx_page;
 872         vsi->rx_buf_failed = rx_buf;
 873
 874         ns->rx_packets = rx_p;
 875         ns->rx_bytes = rx_b;
 876         ns->tx_packets = tx_p;
 877         ns->tx_bytes = tx_b;
 878
 879         /* update netdev stats from eth stats */
 880         i40e_update_eth_stats(vsi);
 881         ons->tx_errors = oes->tx_errors;
 882         ns->tx_errors = es->tx_errors;
 883         ons->multicast = oes->rx_multicast;
 884         ns->multicast = es->rx_multicast;
 885         ons->rx_dropped = oes->rx_discards;
 886         ns->rx_dropped = es->rx_discards;
 887         ons->tx_dropped = oes->tx_discards;
 888         ns->tx_dropped = es->tx_discards;
 889
 890         /* pull in a couple PF stats if this is the main vsi */
 891         if (vsi == pf->vsi[pf->lan_vsi]) {
 892                 ns->rx_crc_errors = pf->stats.crc_errors;
 893                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 894                 ns->rx_length_errors = pf->stats.rx_length_errors;
 895         }
 896 }
 897
 898 /**
 899  * i40e_update_pf_stats - Update the PF statistics counters.
 900  * @pf: the PF to be updated
 901  **/
 902 static void i40e_update_pf_stats(struct i40e_pf *pf)
 903 {
 904         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 905         struct i40e_hw_port_stats *nsd = &pf->stats;
 906         struct i40e_hw *hw = &pf->hw;
 907         u32 val;
 908         int i;
 909
 910         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 911                            I40E_GLPRT_GORCL(hw->port),
 912                            pf->stat_offsets_loaded,
 913                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 914         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 915                            I40E_GLPRT_GOTCL(hw->port),
 916                            pf->stat_offsets_loaded,
 917                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 918         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 919                            pf->stat_offsets_loaded,
 920                            &osd->eth.rx_discards,
 921                            &nsd->eth.rx_discards);
 922         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 923                            I40E_GLPRT_UPRCL(hw->port),
 924                            pf->stat_offsets_loaded,
 925                            &osd->eth.rx_unicast,
 926                            &nsd->eth.rx_unicast);
 927         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 928                            I40E_GLPRT_MPRCL(hw->port),
 929                            pf->stat_offsets_loaded,
 930                            &osd->eth.rx_multicast,
 931                            &nsd->eth.rx_multicast);
 932         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 933                            I40E_GLPRT_BPRCL(hw->port),
 934                            pf->stat_offsets_loaded,
 935                            &osd->eth.rx_broadcast,
 936                            &nsd->eth.rx_broadcast);
 937         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 938                            I40E_GLPRT_UPTCL(hw->port),
 939                            pf->stat_offsets_loaded,
 940                            &osd->eth.tx_unicast,
 941                            &nsd->eth.tx_unicast);
 942         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 943                            I40E_GLPRT_MPTCL(hw->port),
 944                            pf->stat_offsets_loaded,
 945                            &osd->eth.tx_multicast,
 946                            &nsd->eth.tx_multicast);
 947         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 948                            I40E_GLPRT_BPTCL(hw->port),
 949                            pf->stat_offsets_loaded,
 950                            &osd->eth.tx_broadcast,
 951                            &nsd->eth.tx_broadcast);
 952
 953         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 954                            pf->stat_offsets_loaded,
 955                            &osd->tx_dropped_link_down,
 956                            &nsd->tx_dropped_link_down);
 957
 958         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 959                            pf->stat_offsets_loaded,
 960                            &osd->crc_errors, &nsd->crc_errors);
 961
 962         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 963                            pf->stat_offsets_loaded,
 964                            &osd->illegal_bytes, &nsd->illegal_bytes);
 965
 966         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 967                            pf->stat_offsets_loaded,
 968                            &osd->mac_local_faults,
 969                            &nsd->mac_local_faults);
 970         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 971                            pf->stat_offsets_loaded,
 972                            &osd->mac_remote_faults,
 973                            &nsd->mac_remote_faults);
 974
 975         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 976                            pf->stat_offsets_loaded,
 977                            &osd->rx_length_errors,
 978                            &nsd->rx_length_errors);
 979
 980         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 981                            pf->stat_offsets_loaded,
 982                            &osd->link_xon_rx, &nsd->link_xon_rx);
 983         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->link_xon_tx, &nsd->link_xon_tx);
 986         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 987                            pf->stat_offsets_loaded,
 988                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 989         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 990                            pf->stat_offsets_loaded,
 991                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 992
 993         for (i = 0; i < 8; i++) {
 994                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 995                                    pf->stat_offsets_loaded,
 996                                    &osd->priority_xoff_rx[i],
 997                                    &nsd->priority_xoff_rx[i]);
 998                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 999                                    pf->stat_offsets_loaded,
1000                                    &osd->priority_xon_rx[i],
1001                                    &nsd->priority_xon_rx[i]);
1002                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1003                                    pf->stat_offsets_loaded,
1004                                    &osd->priority_xon_tx[i],
1005                                    &nsd->priority_xon_tx[i]);
1006                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1007                                    pf->stat_offsets_loaded,
1008                                    &osd->priority_xoff_tx[i],
1009                                    &nsd->priority_xoff_tx[i]);
1010                 i40e_stat_update32(hw,
1011                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1012                                    pf->stat_offsets_loaded,
1013                                    &osd->priority_xon_2_xoff[i],
1014                                    &nsd->priority_xon_2_xoff[i]);
1015         }
1016
1017         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1018                            I40E_GLPRT_PRC64L(hw->port),
1019                            pf->stat_offsets_loaded,
1020                            &osd->rx_size_64, &nsd->rx_size_64);
1021         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1022                            I40E_GLPRT_PRC127L(hw->port),
1023                            pf->stat_offsets_loaded,
1024                            &osd->rx_size_127, &nsd->rx_size_127);
1025         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1026                            I40E_GLPRT_PRC255L(hw->port),
1027                            pf->stat_offsets_loaded,
1028                            &osd->rx_size_255, &nsd->rx_size_255);
1029         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1030                            I40E_GLPRT_PRC511L(hw->port),
1031                            pf->stat_offsets_loaded,
1032                            &osd->rx_size_511, &nsd->rx_size_511);
1033         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1034                            I40E_GLPRT_PRC1023L(hw->port),
1035                            pf->stat_offsets_loaded,
1036                            &osd->rx_size_1023, &nsd->rx_size_1023);
1037         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1038                            I40E_GLPRT_PRC1522L(hw->port),
1039                            pf->stat_offsets_loaded,
1040                            &osd->rx_size_1522, &nsd->rx_size_1522);
1041         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1042                            I40E_GLPRT_PRC9522L(hw->port),
1043                            pf->stat_offsets_loaded,
1044                            &osd->rx_size_big, &nsd->rx_size_big);
1045
1046         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1047                            I40E_GLPRT_PTC64L(hw->port),
1048                            pf->stat_offsets_loaded,
1049                            &osd->tx_size_64, &nsd->tx_size_64);
1050         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1051                            I40E_GLPRT_PTC127L(hw->port),
1052                            pf->stat_offsets_loaded,
1053                            &osd->tx_size_127, &nsd->tx_size_127);
1054         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1055                            I40E_GLPRT_PTC255L(hw->port),
1056                            pf->stat_offsets_loaded,
1057                            &osd->tx_size_255, &nsd->tx_size_255);
1058         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1059                            I40E_GLPRT_PTC511L(hw->port),
1060                            pf->stat_offsets_loaded,
1061                            &osd->tx_size_511, &nsd->tx_size_511);
1062         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1063                            I40E_GLPRT_PTC1023L(hw->port),
1064                            pf->stat_offsets_loaded,
1065                            &osd->tx_size_1023, &nsd->tx_size_1023);
1066         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1067                            I40E_GLPRT_PTC1522L(hw->port),
1068                            pf->stat_offsets_loaded,
1069                            &osd->tx_size_1522, &nsd->tx_size_1522);
1070         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1071                            I40E_GLPRT_PTC9522L(hw->port),
1072                            pf->stat_offsets_loaded,
1073                            &osd->tx_size_big, &nsd->tx_size_big);
1074
1075         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1076                            pf->stat_offsets_loaded,
1077                            &osd->rx_undersize, &nsd->rx_undersize);
1078         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1079                            pf->stat_offsets_loaded,
1080                            &osd->rx_fragments, &nsd->rx_fragments);
1081         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1082                            pf->stat_offsets_loaded,
1083                            &osd->rx_oversize, &nsd->rx_oversize);
1084         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1085                            pf->stat_offsets_loaded,
1086                            &osd->rx_jabber, &nsd->rx_jabber);
1087
1088         /* FDIR stats */
1089         i40e_stat_update32(hw,
1090                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1091                            pf->stat_offsets_loaded,
1092                            &osd->fd_atr_match, &nsd->fd_atr_match);
1093         i40e_stat_update32(hw,
1094                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1095                            pf->stat_offsets_loaded,
1096                            &osd->fd_sb_match, &nsd->fd_sb_match);
1097         i40e_stat_update32(hw,
1098                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1099                       pf->stat_offsets_loaded,
1100                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1101
1102         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1103         nsd->tx_lpi_status =
1104                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1105                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1106         nsd->rx_lpi_status =
1107                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1108                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1109         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1110                            pf->stat_offsets_loaded,
1111                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1112         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1113                            pf->stat_offsets_loaded,
1114                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1115
1116         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1117             !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1118                 nsd->fd_sb_status = true;
1119         else
1120                 nsd->fd_sb_status = false;
1121
1122         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1123             !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1124                 nsd->fd_atr_status = true;
1125         else
1126                 nsd->fd_atr_status = false;
1127
1128         pf->stat_offsets_loaded = true;
1129 }
1130
1131 /**
1132  * i40e_update_stats - Update the various statistics counters.
1133  * @vsi: the VSI to be updated
1134  *
1135  * Update the various stats for this VSI and its related entities.
1136  **/
1137 void i40e_update_stats(struct i40e_vsi *vsi)
1138 {
1139         struct i40e_pf *pf = vsi->back;
1140
1141         if (vsi == pf->vsi[pf->lan_vsi])
1142                 i40e_update_pf_stats(pf);
1143
1144         i40e_update_vsi_stats(vsi);
1145 #ifdef I40E_FCOE
1146         i40e_update_fcoe_stats(vsi);
1147 #endif
1148 }
1149
1150 /**
1151  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1152  * @vsi: the VSI to be searched
1153  * @macaddr: the MAC address
1154  * @vlan: the vlan
1155  * @is_vf: make sure its a VF filter, else doesn't matter
1156  * @is_netdev: make sure its a netdev filter, else doesn't matter
1157  *
1158  * Returns ptr to the filter object or NULL
1159  **/
1160 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1161                                                 u8 *macaddr, s16 vlan,
1162                                                 bool is_vf, bool is_netdev)
1163 {
1164         struct i40e_mac_filter *f;
1165
1166         if (!vsi || !macaddr)
1167                 return NULL;
1168
1169         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1170                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1171                     (vlan == f->vlan)    &&
1172                     (!is_vf || f->is_vf) &&
1173                     (!is_netdev || f->is_netdev))
1174                         return f;
1175         }
1176         return NULL;
1177 }
1178
1179 /**
1180  * i40e_find_mac - Find a mac addr in the macvlan filters list
1181  * @vsi: the VSI to be searched
1182  * @macaddr: the MAC address we are searching for
1183  * @is_vf: make sure its a VF filter, else doesn't matter
1184  * @is_netdev: make sure its a netdev filter, else doesn't matter
1185  *
1186  * Returns the first filter with the provided MAC address or NULL if
1187  * MAC address was not found
1188  **/
1189 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1190                                       bool is_vf, bool is_netdev)
1191 {
1192         struct i40e_mac_filter *f;
1193
1194         if (!vsi || !macaddr)
1195                 return NULL;
1196
1197         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1198                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1199                     (!is_vf || f->is_vf) &&
1200                     (!is_netdev || f->is_netdev))
1201                         return f;
1202         }
1203         return NULL;
1204 }
1205
1206 /**
1207  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1208  * @vsi: the VSI to be searched
1209  *
1210  * Returns true if VSI is in vlan mode or false otherwise
1211  **/
1212 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1213 {
1214         struct i40e_mac_filter *f;
1215
1216         /* Only -1 for all the filters denotes not in vlan mode
1217          * so we have to go through all the list in order to make sure
1218          */
1219         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1220                 if (f->vlan >= 0 || vsi->info.pvid)
1221                         return true;
1222         }
1223
1224         return false;
1225 }
1226
1227 /**
1228  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1229  * @vsi: the VSI to be searched
1230  * @macaddr: the mac address to be filtered
1231  * @is_vf: true if it is a VF
1232  * @is_netdev: true if it is a netdev
1233  *
1234  * Goes through all the macvlan filters and adds a
1235  * macvlan filter for each unique vlan that already exists
1236  *
1237  * Returns first filter found on success, else NULL
1238  **/
1239 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1240                                              bool is_vf, bool is_netdev)
1241 {
1242         struct i40e_mac_filter *f;
1243
1244         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1245                 if (vsi->info.pvid)
1246                         f->vlan = le16_to_cpu(vsi->info.pvid);
1247                 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1248                                       is_vf, is_netdev)) {
1249                         if (!i40e_add_filter(vsi, macaddr, f->vlan,
1250                                              is_vf, is_netdev))
1251                                 return NULL;
1252                 }
1253         }
1254
1255         return list_first_entry_or_null(&vsi->mac_filter_list,
1256                                         struct i40e_mac_filter, list);
1257 }
1258
1259 /**
1260  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1261  * @vsi: the VSI to be searched
1262  * @macaddr: the mac address to be removed
1263  * @is_vf: true if it is a VF
1264  * @is_netdev: true if it is a netdev
1265  *
1266  * Removes a given MAC address from a VSI, regardless of VLAN
1267  *
1268  * Returns 0 for success, or error
1269  **/
1270 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1271                           bool is_vf, bool is_netdev)
1272 {
1273         struct i40e_mac_filter *f = NULL;
1274         int changed = 0;
1275
1276         WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1277              "Missing mac_filter_list_lock\n");
1278         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1279                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1280                     (is_vf == f->is_vf) &&
1281                     (is_netdev == f->is_netdev)) {
1282                         f->counter--;
1283                         f->changed = true;
1284                         changed = 1;
1285                 }
1286         }
1287         if (changed) {
1288                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1289                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1290                 return 0;
1291         }
1292         return -ENOENT;
1293 }
1294
1295 /**
1296  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1297  * @vsi: the PF Main VSI - inappropriate for any other VSI
1298  * @macaddr: the MAC address
1299  *
1300  * Some older firmware configurations set up a default promiscuous VLAN
1301  * filter that needs to be removed.
1302  **/
1303 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1304 {
1305         struct i40e_aqc_remove_macvlan_element_data element;
1306         struct i40e_pf *pf = vsi->back;
1307         i40e_status ret;
1308
1309         /* Only appropriate for the PF main VSI */
1310         if (vsi->type != I40E_VSI_MAIN)
1311                 return -EINVAL;
1312
1313         memset(&element, 0, sizeof(element));
1314         ether_addr_copy(element.mac_addr, macaddr);
1315         element.vlan_tag = 0;
1316         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1317                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1318         ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1319         if (ret)
1320                 return -ENOENT;
1321
1322         return 0;
1323 }
1324
1325 /**
1326  * i40e_add_filter - Add a mac/vlan filter to the VSI
1327  * @vsi: the VSI to be searched
1328  * @macaddr: the MAC address
1329  * @vlan: the vlan
1330  * @is_vf: make sure its a VF filter, else doesn't matter
1331  * @is_netdev: make sure its a netdev filter, else doesn't matter
1332  *
1333  * Returns ptr to the filter object or NULL when no memory available.
1334  *
1335  * NOTE: This function is expected to be called with mac_filter_list_lock
1336  * being held.
1337  **/
1338 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1339                                         u8 *macaddr, s16 vlan,
1340                                         bool is_vf, bool is_netdev)
1341 {
1342         struct i40e_mac_filter *f;
1343
1344         if (!vsi || !macaddr)
1345                 return NULL;
1346
1347         /* Do not allow broadcast filter to be added since broadcast filter
1348          * is added as part of add VSI for any newly created VSI except
1349          * FDIR VSI
1350          */
1351         if (is_broadcast_ether_addr(macaddr))
1352                 return NULL;
1353
1354         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1355         if (!f) {
1356                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1357                 if (!f)
1358                         goto add_filter_out;
1359
1360                 ether_addr_copy(f->macaddr, macaddr);
1361                 f->vlan = vlan;
1362                 f->changed = true;
1363
1364                 INIT_LIST_HEAD(&f->list);
1365                 list_add_tail(&f->list, &vsi->mac_filter_list);
1366         }
1367
1368         /* increment counter and add a new flag if needed */
1369         if (is_vf) {
1370                 if (!f->is_vf) {
1371                         f->is_vf = true;
1372                         f->counter++;
1373                 }
1374         } else if (is_netdev) {
1375                 if (!f->is_netdev) {
1376                         f->is_netdev = true;
1377                         f->counter++;
1378                 }
1379         } else {
1380                 f->counter++;
1381         }
1382
1383         /* changed tells sync_filters_subtask to
1384          * push the filter down to the firmware
1385          */
1386         if (f->changed) {
1387                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1388                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1389         }
1390
1391 add_filter_out:
1392         return f;
1393 }
1394
1395 /**
1396  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1397  * @vsi: the VSI to be searched
1398  * @macaddr: the MAC address
1399  * @vlan: the vlan
1400  * @is_vf: make sure it's a VF filter, else doesn't matter
1401  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1402  *
1403  * NOTE: This function is expected to be called with mac_filter_list_lock
1404  * being held.
1405  **/
1406 void i40e_del_filter(struct i40e_vsi *vsi,
1407                      u8 *macaddr, s16 vlan,
1408                      bool is_vf, bool is_netdev)
1409 {
1410         struct i40e_mac_filter *f;
1411
1412         if (!vsi || !macaddr)
1413                 return;
1414
1415         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1416         if (!f || f->counter == 0)
1417                 return;
1418
1419         if (is_vf) {
1420                 if (f->is_vf) {
1421                         f->is_vf = false;
1422                         f->counter--;
1423                 }
1424         } else if (is_netdev) {
1425                 if (f->is_netdev) {
1426                         f->is_netdev = false;
1427                         f->counter--;
1428                 }
1429         } else {
1430                 /* make sure we don't remove a filter in use by VF or netdev */
1431                 int min_f = 0;
1432
1433                 min_f += (f->is_vf ? 1 : 0);
1434                 min_f += (f->is_netdev ? 1 : 0);
1435
1436                 if (f->counter > min_f)
1437                         f->counter--;
1438         }
1439
1440         /* counter == 0 tells sync_filters_subtask to
1441          * remove the filter from the firmware's list
1442          */
1443         if (f->counter == 0) {
1444                 f->changed = true;
1445                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1446                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1447         }
1448 }
1449
1450 /**
1451  * i40e_set_mac - NDO callback to set mac address
1452  * @netdev: network interface device structure
1453  * @p: pointer to an address structure
1454  *
1455  * Returns 0 on success, negative on failure
1456  **/
1457 #ifdef I40E_FCOE
1458 int i40e_set_mac(struct net_device *netdev, void *p)
1459 #else
1460 static int i40e_set_mac(struct net_device *netdev, void *p)
1461 #endif
1462 {
1463         struct i40e_netdev_priv *np = netdev_priv(netdev);
1464         struct i40e_vsi *vsi = np->vsi;
1465         struct i40e_pf *pf = vsi->back;
1466         struct i40e_hw *hw = &pf->hw;
1467         struct sockaddr *addr = p;
1468         struct i40e_mac_filter *f;
1469
1470         if (!is_valid_ether_addr(addr->sa_data))
1471                 return -EADDRNOTAVAIL;
1472
1473         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1474                 netdev_info(netdev, "already using mac address %pM\n",
1475                             addr->sa_data);
1476                 return 0;
1477         }
1478
1479         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1480             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1481                 return -EADDRNOTAVAIL;
1482
1483         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1484                 netdev_info(netdev, "returning to hw mac address %pM\n",
1485                             hw->mac.addr);
1486         else
1487                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1488
1489         if (vsi->type == I40E_VSI_MAIN) {
1490                 i40e_status ret;
1491
1492                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1493                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1494                                                 addr->sa_data, NULL);
1495                 if (ret) {
1496                         netdev_info(netdev,
1497                                     "Addr change for Main VSI failed: %d\n",
1498                                     ret);
1499                         return -EADDRNOTAVAIL;
1500                 }
1501         }
1502
1503         if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1504                 struct i40e_aqc_remove_macvlan_element_data element;
1505
1506                 memset(&element, 0, sizeof(element));
1507                 ether_addr_copy(element.mac_addr, netdev->dev_addr);
1508                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1509                 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1510         } else {
1511                 spin_lock_bh(&vsi->mac_filter_list_lock);
1512                 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1513                                 false, false);
1514                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1515         }
1516
1517         if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1518                 struct i40e_aqc_add_macvlan_element_data element;
1519
1520                 memset(&element, 0, sizeof(element));
1521                 ether_addr_copy(element.mac_addr, hw->mac.addr);
1522                 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1523                 i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1524         } else {
1525                 spin_lock_bh(&vsi->mac_filter_list_lock);
1526                 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1527                                     false, false);
1528                 if (f)
1529                         f->is_laa = true;
1530                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1531         }
1532
1533         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1534
1535         /* schedule our worker thread which will take care of
1536          * applying the new filter changes
1537          */
1538         i40e_service_event_schedule(vsi->back);
1539         return 0;
1540 }
1541
1542 /**
1543  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1544  * @vsi: the VSI being setup
1545  * @ctxt: VSI context structure
1546  * @enabled_tc: Enabled TCs bitmap
1547  * @is_add: True if called before Add VSI
1548  *
1549  * Setup VSI queue mapping for enabled traffic classes.
1550  **/
1551 #ifdef I40E_FCOE
1552 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1553                               struct i40e_vsi_context *ctxt,
1554                               u8 enabled_tc,
1555                               bool is_add)
1556 #else
1557 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1558                                      struct i40e_vsi_context *ctxt,
1559                                      u8 enabled_tc,
1560                                      bool is_add)
1561 #endif
1562 {
1563         struct i40e_pf *pf = vsi->back;
1564         u16 sections = 0;
1565         u8 netdev_tc = 0;
1566         u16 numtc = 0;
1567         u16 qcount;
1568         u8 offset;
1569         u16 qmap;
1570         int i;
1571         u16 num_tc_qps = 0;
1572
1573         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1574         offset = 0;
1575
1576         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1577                 /* Find numtc from enabled TC bitmap */
1578                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1579                         if (enabled_tc & BIT(i)) /* TC is enabled */
1580                                 numtc++;
1581                 }
1582                 if (!numtc) {
1583                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1584                         numtc = 1;
1585                 }
1586         } else {
1587                 /* At least TC0 is enabled in case of non-DCB case */
1588                 numtc = 1;
1589         }
1590
1591         vsi->tc_config.numtc = numtc;
1592         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1593         /* Number of queues per enabled TC */
1594         /* In MFP case we can have a much lower count of MSIx
1595          * vectors available and so we need to lower the used
1596          * q count.
1597          */
1598         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1599                 qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1600         else
1601                 qcount = vsi->alloc_queue_pairs;
1602         num_tc_qps = qcount / numtc;
1603         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1604
1605         /* Setup queue offset/count for all TCs for given VSI */
1606         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1607                 /* See if the given TC is enabled for the given VSI */
1608                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1609                         /* TC is enabled */
1610                         int pow, num_qps;
1611
1612                         switch (vsi->type) {
1613                         case I40E_VSI_MAIN:
1614                                 qcount = min_t(int, pf->alloc_rss_size,
1615                                                num_tc_qps);
1616                                 break;
1617 #ifdef I40E_FCOE
1618                         case I40E_VSI_FCOE:
1619                                 qcount = num_tc_qps;
1620                                 break;
1621 #endif
1622                         case I40E_VSI_FDIR:
1623                         case I40E_VSI_SRIOV:
1624                         case I40E_VSI_VMDQ2:
1625                         default:
1626                                 qcount = num_tc_qps;
1627                                 WARN_ON(i != 0);
1628                                 break;
1629                         }
1630                         vsi->tc_config.tc_info[i].qoffset = offset;
1631                         vsi->tc_config.tc_info[i].qcount = qcount;
1632
1633                         /* find the next higher power-of-2 of num queue pairs */
1634                         num_qps = qcount;
1635                         pow = 0;
1636                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1637                                 pow++;
1638                                 num_qps >>= 1;
1639                         }
1640
1641                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1642                         qmap =
1643                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1644                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1645
1646                         offset += qcount;
1647                 } else {
1648                         /* TC is not enabled so set the offset to
1649                          * default queue and allocate one queue
1650                          * for the given TC.
1651                          */
1652                         vsi->tc_config.tc_info[i].qoffset = 0;
1653                         vsi->tc_config.tc_info[i].qcount = 1;
1654                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1655
1656                         qmap = 0;
1657                 }
1658                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1659         }
1660
1661         /* Set actual Tx/Rx queue pairs */
1662         vsi->num_queue_pairs = offset;
1663         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1664                 if (vsi->req_queue_pairs > 0)
1665                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1666                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1667                         vsi->num_queue_pairs = pf->num_lan_msix;
1668         }
1669
1670         /* Scheduler section valid can only be set for ADD VSI */
1671         if (is_add) {
1672                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1673
1674                 ctxt->info.up_enable_bits = enabled_tc;
1675         }
1676         if (vsi->type == I40E_VSI_SRIOV) {
1677                 ctxt->info.mapping_flags |=
1678                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1679                 for (i = 0; i < vsi->num_queue_pairs; i++)
1680                         ctxt->info.queue_mapping[i] =
1681                                                cpu_to_le16(vsi->base_queue + i);
1682         } else {
1683                 ctxt->info.mapping_flags |=
1684                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1685                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1686         }
1687         ctxt->info.valid_sections |= cpu_to_le16(sections);
1688 }
1689
1690 /**
1691  * i40e_set_rx_mode - NDO callback to set the netdev filters
1692  * @netdev: network interface device structure
1693  **/
1694 #ifdef I40E_FCOE
1695 void i40e_set_rx_mode(struct net_device *netdev)
1696 #else
1697 static void i40e_set_rx_mode(struct net_device *netdev)
1698 #endif
1699 {
1700         struct i40e_netdev_priv *np = netdev_priv(netdev);
1701         struct i40e_mac_filter *f, *ftmp;
1702         struct i40e_vsi *vsi = np->vsi;
1703         struct netdev_hw_addr *uca;
1704         struct netdev_hw_addr *mca;
1705         struct netdev_hw_addr *ha;
1706
1707         spin_lock_bh(&vsi->mac_filter_list_lock);
1708
1709         /* add addr if not already in the filter list */
1710         netdev_for_each_uc_addr(uca, netdev) {
1711                 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1712                         if (i40e_is_vsi_in_vlan(vsi))
1713                                 i40e_put_mac_in_vlan(vsi, uca->addr,
1714                                                      false, true);
1715                         else
1716                                 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1717                                                 false, true);
1718                 }
1719         }
1720
1721         netdev_for_each_mc_addr(mca, netdev) {
1722                 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1723                         if (i40e_is_vsi_in_vlan(vsi))
1724                                 i40e_put_mac_in_vlan(vsi, mca->addr,
1725                                                      false, true);
1726                         else
1727                                 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1728                                                 false, true);
1729                 }
1730         }
1731
1732         /* remove filter if not in netdev list */
1733         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1734
1735                 if (!f->is_netdev)
1736                         continue;
1737
1738                 netdev_for_each_mc_addr(mca, netdev)
1739                         if (ether_addr_equal(mca->addr, f->macaddr))
1740                                 goto bottom_of_search_loop;
1741
1742                 netdev_for_each_uc_addr(uca, netdev)
1743                         if (ether_addr_equal(uca->addr, f->macaddr))
1744                                 goto bottom_of_search_loop;
1745
1746                 for_each_dev_addr(netdev, ha)
1747                         if (ether_addr_equal(ha->addr, f->macaddr))
1748                                 goto bottom_of_search_loop;
1749
1750                 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1751                 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1752
1753 bottom_of_search_loop:
1754                 continue;
1755         }
1756         spin_unlock_bh(&vsi->mac_filter_list_lock);
1757
1758         /* check for other flag changes */
1759         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1760                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1761                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1762         }
1763
1764         /* schedule our worker thread which will take care of
1765          * applying the new filter changes
1766          */
1767         i40e_service_event_schedule(vsi->back);
1768 }
1769
1770 /**
1771  * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1772  * @src: source MAC filter entry to be clones
1773  *
1774  * Returns the pointer to newly cloned MAC filter entry or NULL
1775  * in case of error
1776  **/
1777 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1778                                         struct i40e_mac_filter *src)
1779 {
1780         struct i40e_mac_filter *f;
1781
1782         f = kzalloc(sizeof(*f), GFP_ATOMIC);
1783         if (!f)
1784                 return NULL;
1785         *f = *src;
1786
1787         INIT_LIST_HEAD(&f->list);
1788
1789         return f;
1790 }
1791
1792 /**
1793  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1794  * @vsi: pointer to vsi struct
1795  * @from: Pointer to list which contains MAC filter entries - changes to
1796  *        those entries needs to be undone.
1797  *
1798  * MAC filter entries from list were slated to be removed from device.
1799  **/
1800 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1801                                          struct list_head *from)
1802 {
1803         struct i40e_mac_filter *f, *ftmp;
1804
1805         list_for_each_entry_safe(f, ftmp, from, list) {
1806                 f->changed = true;
1807                 /* Move the element back into MAC filter list*/
1808                 list_move_tail(&f->list, &vsi->mac_filter_list);
1809         }
1810 }
1811
1812 /**
1813  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1814  * @vsi: pointer to vsi struct
1815  *
1816  * MAC filter entries from list were slated to be added from device.
1817  **/
1818 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1819 {
1820         struct i40e_mac_filter *f, *ftmp;
1821
1822         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1823                 if (!f->changed && f->counter)
1824                         f->changed = true;
1825         }
1826 }
1827
1828 /**
1829  * i40e_cleanup_add_list - Deletes the element from add list and release
1830  *                      memory
1831  * @add_list: Pointer to list which contains MAC filter entries
1832  **/
1833 static void i40e_cleanup_add_list(struct list_head *add_list)
1834 {
1835         struct i40e_mac_filter *f, *ftmp;
1836
1837         list_for_each_entry_safe(f, ftmp, add_list, list) {
1838                 list_del(&f->list);
1839                 kfree(f);
1840         }
1841 }
1842
1843 /**
1844  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1845  * @vsi: ptr to the VSI
1846  *
1847  * Push any outstanding VSI filter changes through the AdminQ.
1848  *
1849  * Returns 0 or error value
1850  **/
1851 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1852 {
1853         struct list_head tmp_del_list, tmp_add_list;
1854         struct i40e_mac_filter *f, *ftmp, *fclone;
1855         bool promisc_forced_on = false;
1856         bool add_happened = false;
1857         int filter_list_len = 0;
1858         u32 changed_flags = 0;
1859         i40e_status aq_ret = 0;
1860         bool err_cond = false;
1861         int retval = 0;
1862         struct i40e_pf *pf;
1863         int num_add = 0;
1864         int num_del = 0;
1865         int aq_err = 0;
1866         u16 cmd_flags;
1867
1868         /* empty array typed pointers, kcalloc later */
1869         struct i40e_aqc_add_macvlan_element_data *add_list;
1870         struct i40e_aqc_remove_macvlan_element_data *del_list;
1871
1872         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1873                 usleep_range(1000, 2000);
1874         pf = vsi->back;
1875
1876         if (vsi->netdev) {
1877                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1878                 vsi->current_netdev_flags = vsi->netdev->flags;
1879         }
1880
1881         INIT_LIST_HEAD(&tmp_del_list);
1882         INIT_LIST_HEAD(&tmp_add_list);
1883
1884         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1885                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1886
1887                 spin_lock_bh(&vsi->mac_filter_list_lock);
1888                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1889                         if (!f->changed)
1890                                 continue;
1891
1892                         if (f->counter != 0)
1893                                 continue;
1894                         f->changed = false;
1895
1896                         /* Move the element into temporary del_list */
1897                         list_move_tail(&f->list, &tmp_del_list);
1898                 }
1899
1900                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1901                         if (!f->changed)
1902                                 continue;
1903
1904                         if (f->counter == 0)
1905                                 continue;
1906                         f->changed = false;
1907
1908                         /* Clone MAC filter entry and add into temporary list */
1909                         fclone = i40e_mac_filter_entry_clone(f);
1910                         if (!fclone) {
1911                                 err_cond = true;
1912                                 break;
1913                         }
1914                         list_add_tail(&fclone->list, &tmp_add_list);
1915                 }
1916
1917                 /* if failed to clone MAC filter entry - undo */
1918                 if (err_cond) {
1919                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1920                         i40e_undo_add_filter_entries(vsi);
1921                 }
1922                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1923
1924                 if (err_cond) {
1925                         i40e_cleanup_add_list(&tmp_add_list);
1926                         retval = -ENOMEM;
1927                         goto out;
1928                 }
1929         }
1930
1931         /* Now process 'del_list' outside the lock */
1932         if (!list_empty(&tmp_del_list)) {
1933                 int del_list_size;
1934
1935                 filter_list_len = pf->hw.aq.asq_buf_size /
1936                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1937                 del_list_size = filter_list_len *
1938                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1939                 del_list = kzalloc(del_list_size, GFP_ATOMIC);
1940                 if (!del_list) {
1941                         i40e_cleanup_add_list(&tmp_add_list);
1942
1943                         /* Undo VSI's MAC filter entry element updates */
1944                         spin_lock_bh(&vsi->mac_filter_list_lock);
1945                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1946                         i40e_undo_add_filter_entries(vsi);
1947                         spin_unlock_bh(&vsi->mac_filter_list_lock);
1948                         retval = -ENOMEM;
1949                         goto out;
1950                 }
1951
1952                 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1953                         cmd_flags = 0;
1954
1955                         /* add to delete list */
1956                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1957                         del_list[num_del].vlan_tag =
1958                                 cpu_to_le16((u16)(f->vlan ==
1959                                             I40E_VLAN_ANY ? 0 : f->vlan));
1960
1961                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1962                         del_list[num_del].flags = cmd_flags;
1963                         num_del++;
1964
1965                         /* flush a full buffer */
1966                         if (num_del == filter_list_len) {
1967                                 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1968                                                                 vsi->seid,
1969                                                                 del_list,
1970                                                                 num_del,
1971                                                                 NULL);
1972                                 aq_err = pf->hw.aq.asq_last_status;
1973                                 num_del = 0;
1974                                 memset(del_list, 0, del_list_size);
1975
1976                                 if (aq_ret && aq_err != I40E_AQ_RC_ENOENT) {
1977                                         retval = -EIO;
1978                                         dev_err(&pf->pdev->dev,
1979                                                 "ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1980                                                 i40e_stat_str(&pf->hw, aq_ret),
1981                                                 i40e_aq_str(&pf->hw, aq_err));
1982                                 }
1983                         }
1984                         /* Release memory for MAC filter entries which were
1985                          * synced up with HW.
1986                          */
1987                         list_del(&f->list);
1988                         kfree(f);
1989                 }
1990
1991                 if (num_del) {
1992                         aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1993                                                         del_list, num_del,
1994                                                         NULL);
1995                         aq_err = pf->hw.aq.asq_last_status;
1996                         num_del = 0;
1997
1998                         if (aq_ret && aq_err != I40E_AQ_RC_ENOENT)
1999                                 dev_info(&pf->pdev->dev,
2000                                          "ignoring delete macvlan error, err %s aq_err %s\n",
2001                                          i40e_stat_str(&pf->hw, aq_ret),
2002                                          i40e_aq_str(&pf->hw, aq_err));
2003                 }
2004
2005                 kfree(del_list);
2006                 del_list = NULL;
2007         }
2008
2009         if (!list_empty(&tmp_add_list)) {
2010                 int add_list_size;
2011
2012                 /* do all the adds now */
2013                 filter_list_len = pf->hw.aq.asq_buf_size /
2014                                sizeof(struct i40e_aqc_add_macvlan_element_data),
2015                 add_list_size = filter_list_len *
2016                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2017                 add_list = kzalloc(add_list_size, GFP_ATOMIC);
2018                 if (!add_list) {
2019                         /* Purge element from temporary lists */
2020                         i40e_cleanup_add_list(&tmp_add_list);
2021
2022                         /* Undo add filter entries from VSI MAC filter list */
2023                         spin_lock_bh(&vsi->mac_filter_list_lock);
2024                         i40e_undo_add_filter_entries(vsi);
2025                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2026                         retval = -ENOMEM;
2027                         goto out;
2028                 }
2029
2030                 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2031
2032                         add_happened = true;
2033                         cmd_flags = 0;
2034
2035                         /* add to add array */
2036                         ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2037                         add_list[num_add].vlan_tag =
2038                                 cpu_to_le16(
2039                                  (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2040                         add_list[num_add].queue_number = 0;
2041
2042                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2043                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2044                         num_add++;
2045
2046                         /* flush a full buffer */
2047                         if (num_add == filter_list_len) {
2048                                 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2049                                                              add_list, num_add,
2050                                                              NULL);
2051                                 aq_err = pf->hw.aq.asq_last_status;
2052                                 num_add = 0;
2053
2054                                 if (aq_ret)
2055                                         break;
2056                                 memset(add_list, 0, add_list_size);
2057                         }
2058                         /* Entries from tmp_add_list were cloned from MAC
2059                          * filter list, hence clean those cloned entries
2060                          */
2061                         list_del(&f->list);
2062                         kfree(f);
2063                 }
2064
2065                 if (num_add) {
2066                         aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2067                                                      add_list, num_add, NULL);
2068                         aq_err = pf->hw.aq.asq_last_status;
2069                         num_add = 0;
2070                 }
2071                 kfree(add_list);
2072                 add_list = NULL;
2073
2074                 if (add_happened && aq_ret && aq_err != I40E_AQ_RC_EINVAL) {
2075                         retval = i40e_aq_rc_to_posix(aq_ret, aq_err);
2076                         dev_info(&pf->pdev->dev,
2077                                  "add filter failed, err %s aq_err %s\n",
2078                                  i40e_stat_str(&pf->hw, aq_ret),
2079                                  i40e_aq_str(&pf->hw, aq_err));
2080                         if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2081                             !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2082                                       &vsi->state)) {
2083                                 promisc_forced_on = true;
2084                                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2085                                         &vsi->state);
2086                                 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2087                         }
2088                 }
2089         }
2090
2091         /* if the VF is not trusted do not do promisc */
2092         if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2093                 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2094                 goto out;
2095         }
2096
2097         /* check for changes in promiscuous modes */
2098         if (changed_flags & IFF_ALLMULTI) {
2099                 bool cur_multipromisc;
2100
2101                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2102                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2103                                                                vsi->seid,
2104                                                                cur_multipromisc,
2105                                                                NULL);
2106                 if (aq_ret) {
2107                         retval = i40e_aq_rc_to_posix(aq_ret,
2108                                                      pf->hw.aq.asq_last_status);
2109                         dev_info(&pf->pdev->dev,
2110                                  "set multi promisc failed, err %s aq_err %s\n",
2111                                  i40e_stat_str(&pf->hw, aq_ret),
2112                                  i40e_aq_str(&pf->hw,
2113                                              pf->hw.aq.asq_last_status));
2114                 }
2115         }
2116         if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2117                 bool cur_promisc;
2118
2119                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2120                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2121                                         &vsi->state));
2122                 if ((vsi->type == I40E_VSI_MAIN) &&
2123                     (pf->lan_veb != I40E_NO_VEB) &&
2124                     !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2125                         /* set defport ON for Main VSI instead of true promisc
2126                          * this way we will get all unicast/multicast and VLAN
2127                          * promisc behavior but will not get VF or VMDq traffic
2128                          * replicated on the Main VSI.
2129                          */
2130                         if (pf->cur_promisc != cur_promisc) {
2131                                 pf->cur_promisc = cur_promisc;
2132                                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2133                         }
2134                 } else {
2135                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2136                                                           &vsi->back->hw,
2137                                                           vsi->seid,
2138                                                           cur_promisc, NULL,
2139                                                           true);
2140                         if (aq_ret) {
2141                                 retval =
2142                                 i40e_aq_rc_to_posix(aq_ret,
2143                                                     pf->hw.aq.asq_last_status);
2144                                 dev_info(&pf->pdev->dev,
2145                                          "set unicast promisc failed, err %d, aq_err %d\n",
2146                                          aq_ret, pf->hw.aq.asq_last_status);
2147                         }
2148                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2149                                                           &vsi->back->hw,
2150                                                           vsi->seid,
2151                                                           cur_promisc, NULL);
2152                         if (aq_ret) {
2153                                 retval =
2154                                 i40e_aq_rc_to_posix(aq_ret,
2155                                                     pf->hw.aq.asq_last_status);
2156                                 dev_info(&pf->pdev->dev,
2157                                          "set multicast promisc failed, err %d, aq_err %d\n",
2158                                          aq_ret, pf->hw.aq.asq_last_status);
2159                         }
2160                 }
2161         }
2162 out:
2163         /* if something went wrong then set the changed flag so we try again */
2164         if (retval)
2165                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2166
2167         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2168         return retval;
2169 }
2170
2171 /**
2172  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2173  * @pf: board private structure
2174  **/
2175 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2176 {
2177         int v;
2178
2179         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2180                 return;
2181         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2182
2183         for (v = 0; v < pf->num_alloc_vsi; v++) {
2184                 if (pf->vsi[v] &&
2185                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2186                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2187
2188                         if (ret) {
2189                                 /* come back and try again later */
2190                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2191                                 break;
2192                         }
2193                 }
2194         }
2195 }
2196
2197 /**
2198  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2199  * @netdev: network interface device structure
2200  * @new_mtu: new value for maximum frame size
2201  *
2202  * Returns 0 on success, negative on failure
2203  **/
2204 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2205 {
2206         struct i40e_netdev_priv *np = netdev_priv(netdev);
2207         int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2208         struct i40e_vsi *vsi = np->vsi;
2209
2210         /* MTU < 68 is an error and causes problems on some kernels */
2211         if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2212                 return -EINVAL;
2213
2214         netdev_info(netdev, "changing MTU from %d to %d\n",
2215                     netdev->mtu, new_mtu);
2216         netdev->mtu = new_mtu;
2217         if (netif_running(netdev))
2218                 i40e_vsi_reinit_locked(vsi);
2219         i40e_notify_client_of_l2_param_changes(vsi);
2220         return 0;
2221 }
2222
2223 /**
2224  * i40e_ioctl - Access the hwtstamp interface
2225  * @netdev: network interface device structure
2226  * @ifr: interface request data
2227  * @cmd: ioctl command
2228  **/
2229 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2230 {
2231         struct i40e_netdev_priv *np = netdev_priv(netdev);
2232         struct i40e_pf *pf = np->vsi->back;
2233
2234         switch (cmd) {
2235         case SIOCGHWTSTAMP:
2236                 return i40e_ptp_get_ts_config(pf, ifr);
2237         case SIOCSHWTSTAMP:
2238                 return i40e_ptp_set_ts_config(pf, ifr);
2239         default:
2240                 return -EOPNOTSUPP;
2241         }
2242 }
2243
2244 /**
2245  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2246  * @vsi: the vsi being adjusted
2247  **/
2248 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2249 {
2250         struct i40e_vsi_context ctxt;
2251         i40e_status ret;
2252
2253         if ((vsi->info.valid_sections &
2254              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2255             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2256                 return;  /* already enabled */
2257
2258         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2259         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2260                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2261
2262         ctxt.seid = vsi->seid;
2263         ctxt.info = vsi->info;
2264         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2265         if (ret) {
2266                 dev_info(&vsi->back->pdev->dev,
2267                          "update vlan stripping failed, err %s aq_err %s\n",
2268                          i40e_stat_str(&vsi->back->hw, ret),
2269                          i40e_aq_str(&vsi->back->hw,
2270                                      vsi->back->hw.aq.asq_last_status));
2271         }
2272 }
2273
2274 /**
2275  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2276  * @vsi: the vsi being adjusted
2277  **/
2278 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2279 {
2280         struct i40e_vsi_context ctxt;
2281         i40e_status ret;
2282
2283         if ((vsi->info.valid_sections &
2284              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2285             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2286              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2287                 return;  /* already disabled */
2288
2289         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2290         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2291                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2292
2293         ctxt.seid = vsi->seid;
2294         ctxt.info = vsi->info;
2295         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2296         if (ret) {
2297                 dev_info(&vsi->back->pdev->dev,
2298                          "update vlan stripping failed, err %s aq_err %s\n",
2299                          i40e_stat_str(&vsi->back->hw, ret),
2300                          i40e_aq_str(&vsi->back->hw,
2301                                      vsi->back->hw.aq.asq_last_status));
2302         }
2303 }
2304
2305 /**
2306  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2307  * @netdev: network interface to be adjusted
2308  * @features: netdev features to test if VLAN offload is enabled or not
2309  **/
2310 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2311 {
2312         struct i40e_netdev_priv *np = netdev_priv(netdev);
2313         struct i40e_vsi *vsi = np->vsi;
2314
2315         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2316                 i40e_vlan_stripping_enable(vsi);
2317         else
2318                 i40e_vlan_stripping_disable(vsi);
2319 }
2320
2321 /**
2322  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2323  * @vsi: the vsi being configured
2324  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2325  **/
2326 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2327 {
2328         struct i40e_mac_filter *f, *add_f;
2329         bool is_netdev, is_vf;
2330
2331         is_vf = (vsi->type == I40E_VSI_SRIOV);
2332         is_netdev = !!(vsi->netdev);
2333
2334         /* Locked once because all functions invoked below iterates list*/
2335         spin_lock_bh(&vsi->mac_filter_list_lock);
2336
2337         if (is_netdev) {
2338                 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2339                                         is_vf, is_netdev);
2340                 if (!add_f) {
2341                         dev_info(&vsi->back->pdev->dev,
2342                                  "Could not add vlan filter %d for %pM\n",
2343                                  vid, vsi->netdev->dev_addr);
2344                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2345                         return -ENOMEM;
2346                 }
2347         }
2348
2349         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2350                 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2351                 if (!add_f) {
2352                         dev_info(&vsi->back->pdev->dev,
2353                                  "Could not add vlan filter %d for %pM\n",
2354                                  vid, f->macaddr);
2355                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2356                         return -ENOMEM;
2357                 }
2358         }
2359
2360         /* Now if we add a vlan tag, make sure to check if it is the first
2361          * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2362          * with 0, so we now accept untagged and specified tagged traffic
2363          * (and not any taged and untagged)
2364          */
2365         if (vid > 0) {
2366                 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2367                                                   I40E_VLAN_ANY,
2368                                                   is_vf, is_netdev)) {
2369                         i40e_del_filter(vsi, vsi->netdev->dev_addr,
2370                                         I40E_VLAN_ANY, is_vf, is_netdev);
2371                         add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2372                                                 is_vf, is_netdev);
2373                         if (!add_f) {
2374                                 dev_info(&vsi->back->pdev->dev,
2375                                          "Could not add filter 0 for %pM\n",
2376                                          vsi->netdev->dev_addr);
2377                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2378                                 return -ENOMEM;
2379                         }
2380                 }
2381         }
2382
2383         /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2384         if (vid > 0 && !vsi->info.pvid) {
2385                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2386                         if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2387                                               is_vf, is_netdev))
2388                                 continue;
2389                         i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2390                                         is_vf, is_netdev);
2391                         add_f = i40e_add_filter(vsi, f->macaddr,
2392                                                 0, is_vf, is_netdev);
2393                         if (!add_f) {
2394                                 dev_info(&vsi->back->pdev->dev,
2395                                          "Could not add filter 0 for %pM\n",
2396                                         f->macaddr);
2397                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2398                                 return -ENOMEM;
2399                         }
2400                 }
2401         }
2402
2403         spin_unlock_bh(&vsi->mac_filter_list_lock);
2404
2405         /* schedule our worker thread which will take care of
2406          * applying the new filter changes
2407          */
2408         i40e_service_event_schedule(vsi->back);
2409         return 0;
2410 }
2411
2412 /**
2413  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2414  * @vsi: the vsi being configured
2415  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2416  *
2417  * Return: 0 on success or negative otherwise
2418  **/
2419 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2420 {
2421         struct net_device *netdev = vsi->netdev;
2422         struct i40e_mac_filter *f, *add_f;
2423         bool is_vf, is_netdev;
2424         int filter_count = 0;
2425
2426         is_vf = (vsi->type == I40E_VSI_SRIOV);
2427         is_netdev = !!(netdev);
2428
2429         /* Locked once because all functions invoked below iterates list */
2430         spin_lock_bh(&vsi->mac_filter_list_lock);
2431
2432         if (is_netdev)
2433                 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2434
2435         list_for_each_entry(f, &vsi->mac_filter_list, list)
2436                 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2437
2438         /* go through all the filters for this VSI and if there is only
2439          * vid == 0 it means there are no other filters, so vid 0 must
2440          * be replaced with -1. This signifies that we should from now
2441          * on accept any traffic (with any tag present, or untagged)
2442          */
2443         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2444                 if (is_netdev) {
2445                         if (f->vlan &&
2446                             ether_addr_equal(netdev->dev_addr, f->macaddr))
2447                                 filter_count++;
2448                 }
2449
2450                 if (f->vlan)
2451                         filter_count++;
2452         }
2453
2454         if (!filter_count && is_netdev) {
2455                 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2456                 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2457                                     is_vf, is_netdev);
2458                 if (!f) {
2459                         dev_info(&vsi->back->pdev->dev,
2460                                  "Could not add filter %d for %pM\n",
2461                                  I40E_VLAN_ANY, netdev->dev_addr);
2462                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2463                         return -ENOMEM;
2464                 }
2465         }
2466
2467         if (!filter_count) {
2468                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2469                         i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2470                         add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2471                                                 is_vf, is_netdev);
2472                         if (!add_f) {
2473                                 dev_info(&vsi->back->pdev->dev,
2474                                          "Could not add filter %d for %pM\n",
2475                                          I40E_VLAN_ANY, f->macaddr);
2476                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2477                                 return -ENOMEM;
2478                         }
2479                 }
2480         }
2481
2482         spin_unlock_bh(&vsi->mac_filter_list_lock);
2483
2484         /* schedule our worker thread which will take care of
2485          * applying the new filter changes
2486          */
2487         i40e_service_event_schedule(vsi->back);
2488         return 0;
2489 }
2490
2491 /**
2492  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2493  * @netdev: network interface to be adjusted
2494  * @vid: vlan id to be added
2495  *
2496  * net_device_ops implementation for adding vlan ids
2497  **/
2498 #ifdef I40E_FCOE
2499 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2500                          __always_unused __be16 proto, u16 vid)
2501 #else
2502 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2503                                 __always_unused __be16 proto, u16 vid)
2504 #endif
2505 {
2506         struct i40e_netdev_priv *np = netdev_priv(netdev);
2507         struct i40e_vsi *vsi = np->vsi;
2508         int ret = 0;
2509
2510         if (vid > 4095)
2511                 return -EINVAL;
2512
2513         netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2514
2515         /* If the network stack called us with vid = 0 then
2516          * it is asking to receive priority tagged packets with
2517          * vlan id 0.  Our HW receives them by default when configured
2518          * to receive untagged packets so there is no need to add an
2519          * extra filter for vlan 0 tagged packets.
2520          */
2521         if (vid)
2522                 ret = i40e_vsi_add_vlan(vsi, vid);
2523
2524         if (!ret && (vid < VLAN_N_VID))
2525                 set_bit(vid, vsi->active_vlans);
2526
2527         return ret;
2528 }
2529
2530 /**
2531  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2532  * @netdev: network interface to be adjusted
2533  * @vid: vlan id to be removed
2534  *
2535  * net_device_ops implementation for removing vlan ids
2536  **/
2537 #ifdef I40E_FCOE
2538 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2539                           __always_unused __be16 proto, u16 vid)
2540 #else
2541 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2542                                  __always_unused __be16 proto, u16 vid)
2543 #endif
2544 {
2545         struct i40e_netdev_priv *np = netdev_priv(netdev);
2546         struct i40e_vsi *vsi = np->vsi;
2547
2548         netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2549
2550         /* return code is ignored as there is nothing a user
2551          * can do about failure to remove and a log message was
2552          * already printed from the other function
2553          */
2554         i40e_vsi_kill_vlan(vsi, vid);
2555
2556         clear_bit(vid, vsi->active_vlans);
2557
2558         return 0;
2559 }
2560
2561 /**
2562  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2563  * @vsi: the vsi being brought back up
2564  **/
2565 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2566 {
2567         u16 vid;
2568
2569         if (!vsi->netdev)
2570                 return;
2571
2572         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2573
2574         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2575                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2576                                      vid);
2577 }
2578
2579 /**
2580  * i40e_vsi_add_pvid - Add pvid for the VSI
2581  * @vsi: the vsi being adjusted
2582  * @vid: the vlan id to set as a PVID
2583  **/
2584 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2585 {
2586         struct i40e_vsi_context ctxt;
2587         i40e_status ret;
2588
2589         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2590         vsi->info.pvid = cpu_to_le16(vid);
2591         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2592                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2593                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2594
2595         ctxt.seid = vsi->seid;
2596         ctxt.info = vsi->info;
2597         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2598         if (ret) {
2599                 dev_info(&vsi->back->pdev->dev,
2600                          "add pvid failed, err %s aq_err %s\n",
2601                          i40e_stat_str(&vsi->back->hw, ret),
2602                          i40e_aq_str(&vsi->back->hw,
2603                                      vsi->back->hw.aq.asq_last_status));
2604                 return -ENOENT;
2605         }
2606
2607         return 0;
2608 }
2609
2610 /**
2611  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2612  * @vsi: the vsi being adjusted
2613  *
2614  * Just use the vlan_rx_register() service to put it back to normal
2615  **/
2616 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2617 {
2618         i40e_vlan_stripping_disable(vsi);
2619
2620         vsi->info.pvid = 0;
2621 }
2622
2623 /**
2624  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2625  * @vsi: ptr to the VSI
2626  *
2627  * If this function returns with an error, then it's possible one or
2628  * more of the rings is populated (while the rest are not).  It is the
2629  * callers duty to clean those orphaned rings.
2630  *
2631  * Return 0 on success, negative on failure
2632  **/
2633 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2634 {
2635         int i, err = 0;
2636
2637         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2638                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2639
2640         return err;
2641 }
2642
2643 /**
2644  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2645  * @vsi: ptr to the VSI
2646  *
2647  * Free VSI's transmit software resources
2648  **/
2649 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2650 {
2651         int i;
2652
2653         if (!vsi->tx_rings)
2654                 return;
2655
2656         for (i = 0; i < vsi->num_queue_pairs; i++)
2657                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2658                         i40e_free_tx_resources(vsi->tx_rings[i]);
2659 }
2660
2661 /**
2662  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2663  * @vsi: ptr to the VSI
2664  *
2665  * If this function returns with an error, then it's possible one or
2666  * more of the rings is populated (while the rest are not).  It is the
2667  * callers duty to clean those orphaned rings.
2668  *
2669  * Return 0 on success, negative on failure
2670  **/
2671 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2672 {
2673         int i, err = 0;
2674
2675         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2676                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2677 #ifdef I40E_FCOE
2678         i40e_fcoe_setup_ddp_resources(vsi);
2679 #endif
2680         return err;
2681 }
2682
2683 /**
2684  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2685  * @vsi: ptr to the VSI
2686  *
2687  * Free all receive software resources
2688  **/
2689 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2690 {
2691         int i;
2692
2693         if (!vsi->rx_rings)
2694                 return;
2695
2696         for (i = 0; i < vsi->num_queue_pairs; i++)
2697                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2698                         i40e_free_rx_resources(vsi->rx_rings[i]);
2699 #ifdef I40E_FCOE
2700         i40e_fcoe_free_ddp_resources(vsi);
2701 #endif
2702 }
2703
2704 /**
2705  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2706  * @ring: The Tx ring to configure
2707  *
2708  * This enables/disables XPS for a given Tx descriptor ring
2709  * based on the TCs enabled for the VSI that ring belongs to.
2710  **/
2711 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2712 {
2713         struct i40e_vsi *vsi = ring->vsi;
2714         cpumask_var_t mask;
2715
2716         if (!ring->q_vector || !ring->netdev)
2717                 return;
2718
2719         /* Single TC mode enable XPS */
2720         if (vsi->tc_config.numtc <= 1) {
2721                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2722                         netif_set_xps_queue(ring->netdev,
2723                                             &ring->q_vector->affinity_mask,
2724                                             ring->queue_index);
2725         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2726                 /* Disable XPS to allow selection based on TC */
2727                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2728                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2729                 free_cpumask_var(mask);
2730         }
2731
2732         /* schedule our worker thread which will take care of
2733          * applying the new filter changes
2734          */
2735         i40e_service_event_schedule(vsi->back);
2736 }
2737
2738 /**
2739  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2740  * @ring: The Tx ring to configure
2741  *
2742  * Configure the Tx descriptor ring in the HMC context.
2743  **/
2744 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2745 {
2746         struct i40e_vsi *vsi = ring->vsi;
2747         u16 pf_q = vsi->base_queue + ring->queue_index;
2748         struct i40e_hw *hw = &vsi->back->hw;
2749         struct i40e_hmc_obj_txq tx_ctx;
2750         i40e_status err = 0;
2751         u32 qtx_ctl = 0;
2752
2753         /* some ATR related tx ring init */
2754         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2755                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2756                 ring->atr_count = 0;
2757         } else {
2758                 ring->atr_sample_rate = 0;
2759         }
2760
2761         /* configure XPS */
2762         i40e_config_xps_tx_ring(ring);
2763
2764         /* clear the context structure first */
2765         memset(&tx_ctx, 0, sizeof(tx_ctx));
2766
2767         tx_ctx.new_context = 1;
2768         tx_ctx.base = (ring->dma / 128);
2769         tx_ctx.qlen = ring->count;
2770         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2771                                                I40E_FLAG_FD_ATR_ENABLED));
2772 #ifdef I40E_FCOE
2773         tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2774 #endif
2775         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2776         /* FDIR VSI tx ring can still use RS bit and writebacks */
2777         if (vsi->type != I40E_VSI_FDIR)
2778                 tx_ctx.head_wb_ena = 1;
2779         tx_ctx.head_wb_addr = ring->dma +
2780                               (ring->count * sizeof(struct i40e_tx_desc));
2781
2782         /* As part of VSI creation/update, FW allocates certain
2783          * Tx arbitration queue sets for each TC enabled for
2784          * the VSI. The FW returns the handles to these queue
2785          * sets as part of the response buffer to Add VSI,
2786          * Update VSI, etc. AQ commands. It is expected that
2787          * these queue set handles be associated with the Tx
2788          * queues by the driver as part of the TX queue context
2789          * initialization. This has to be done regardless of
2790          * DCB as by default everything is mapped to TC0.
2791          */
2792         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2793         tx_ctx.rdylist_act = 0;
2794
2795         /* clear the context in the HMC */
2796         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2797         if (err) {
2798                 dev_info(&vsi->back->pdev->dev,
2799                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2800                          ring->queue_index, pf_q, err);
2801                 return -ENOMEM;
2802         }
2803
2804         /* set the context in the HMC */
2805         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2806         if (err) {
2807                 dev_info(&vsi->back->pdev->dev,
2808                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2809                          ring->queue_index, pf_q, err);
2810                 return -ENOMEM;
2811         }
2812
2813         /* Now associate this queue with this PCI function */
2814         if (vsi->type == I40E_VSI_VMDQ2) {
2815                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2816                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2817                            I40E_QTX_CTL_VFVM_INDX_MASK;
2818         } else {
2819                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2820         }
2821
2822         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2823                     I40E_QTX_CTL_PF_INDX_MASK);
2824         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2825         i40e_flush(hw);
2826
2827         /* cache tail off for easier writes later */
2828         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2829
2830         return 0;
2831 }
2832
2833 /**
2834  * i40e_configure_rx_ring - Configure a receive ring context
2835  * @ring: The Rx ring to configure
2836  *
2837  * Configure the Rx descriptor ring in the HMC context.
2838  **/
2839 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2840 {
2841         struct i40e_vsi *vsi = ring->vsi;
2842         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2843         u16 pf_q = vsi->base_queue + ring->queue_index;
2844         struct i40e_hw *hw = &vsi->back->hw;
2845         struct i40e_hmc_obj_rxq rx_ctx;
2846         i40e_status err = 0;
2847
2848         ring->state = 0;
2849
2850         /* clear the context structure first */
2851         memset(&rx_ctx, 0, sizeof(rx_ctx));
2852
2853         ring->rx_buf_len = vsi->rx_buf_len;
2854
2855         rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2856
2857         rx_ctx.base = (ring->dma / 128);
2858         rx_ctx.qlen = ring->count;
2859
2860         /* use 32 byte descriptors */
2861         rx_ctx.dsize = 1;
2862
2863         /* descriptor type is always zero
2864          * rx_ctx.dtype = 0;
2865          */
2866         rx_ctx.hsplit_0 = 0;
2867
2868         rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
2869         if (hw->revision_id == 0)
2870                 rx_ctx.lrxqthresh = 0;
2871         else
2872                 rx_ctx.lrxqthresh = 2;
2873         rx_ctx.crcstrip = 1;
2874         rx_ctx.l2tsel = 1;
2875         /* this controls whether VLAN is stripped from inner headers */
2876         rx_ctx.showiv = 0;
2877 #ifdef I40E_FCOE
2878         rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2879 #endif
2880         /* set the prefena field to 1 because the manual says to */
2881         rx_ctx.prefena = 1;
2882
2883         /* clear the context in the HMC */
2884         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2885         if (err) {
2886                 dev_info(&vsi->back->pdev->dev,
2887                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2888                          ring->queue_index, pf_q, err);
2889                 return -ENOMEM;
2890         }
2891
2892         /* set the context in the HMC */
2893         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2894         if (err) {
2895                 dev_info(&vsi->back->pdev->dev,
2896                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2897                          ring->queue_index, pf_q, err);
2898                 return -ENOMEM;
2899         }
2900
2901         /* cache tail for quicker writes, and clear the reg before use */
2902         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2903         writel(0, ring->tail);
2904
2905         i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2906
2907         return 0;
2908 }
2909
2910 /**
2911  * i40e_vsi_configure_tx - Configure the VSI for Tx
2912  * @vsi: VSI structure describing this set of rings and resources
2913  *
2914  * Configure the Tx VSI for operation.
2915  **/
2916 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2917 {
2918         int err = 0;
2919         u16 i;
2920
2921         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2922                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2923
2924         return err;
2925 }
2926
2927 /**
2928  * i40e_vsi_configure_rx - Configure the VSI for Rx
2929  * @vsi: the VSI being configured
2930  *
2931  * Configure the Rx VSI for operation.
2932  **/
2933 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2934 {
2935         int err = 0;
2936         u16 i;
2937
2938         if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2939                 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2940                                + ETH_FCS_LEN + VLAN_HLEN;
2941         else
2942                 vsi->max_frame = I40E_RXBUFFER_2048;
2943
2944         vsi->rx_buf_len = I40E_RXBUFFER_2048;
2945
2946 #ifdef I40E_FCOE
2947         /* setup rx buffer for FCoE */
2948         if ((vsi->type == I40E_VSI_FCOE) &&
2949             (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2950                 vsi->rx_buf_len = I40E_RXBUFFER_3072;
2951                 vsi->max_frame = I40E_RXBUFFER_3072;
2952         }
2953
2954 #endif /* I40E_FCOE */
2955         /* round up for the chip's needs */
2956         vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2957                                 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
2958
2959         /* set up individual rings */
2960         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2961                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2962
2963         return err;
2964 }
2965
2966 /**
2967  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2968  * @vsi: ptr to the VSI
2969  **/
2970 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2971 {
2972         struct i40e_ring *tx_ring, *rx_ring;
2973         u16 qoffset, qcount;
2974         int i, n;
2975
2976         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
2977                 /* Reset the TC information */
2978                 for (i = 0; i < vsi->num_queue_pairs; i++) {
2979                         rx_ring = vsi->rx_rings[i];
2980                         tx_ring = vsi->tx_rings[i];
2981                         rx_ring->dcb_tc = 0;
2982                         tx_ring->dcb_tc = 0;
2983                 }
2984         }
2985
2986         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2987                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
2988                         continue;
2989
2990                 qoffset = vsi->tc_config.tc_info[n].qoffset;
2991                 qcount = vsi->tc_config.tc_info[n].qcount;
2992                 for (i = qoffset; i < (qoffset + qcount); i++) {
2993                         rx_ring = vsi->rx_rings[i];
2994                         tx_ring = vsi->tx_rings[i];
2995                         rx_ring->dcb_tc = n;
2996                         tx_ring->dcb_tc = n;
2997                 }
2998         }
2999 }
3000
3001 /**
3002  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3003  * @vsi: ptr to the VSI
3004  **/
3005 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3006 {
3007         if (vsi->netdev)
3008                 i40e_set_rx_mode(vsi->netdev);
3009 }
3010
3011 /**
3012  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3013  * @vsi: Pointer to the targeted VSI
3014  *
3015  * This function replays the hlist on the hw where all the SB Flow Director
3016  * filters were saved.
3017  **/
3018 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3019 {
3020         struct i40e_fdir_filter *filter;
3021         struct i40e_pf *pf = vsi->back;
3022         struct hlist_node *node;
3023
3024         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3025                 return;
3026
3027         hlist_for_each_entry_safe(filter, node,
3028                                   &pf->fdir_filter_list, fdir_node) {
3029                 i40e_add_del_fdir(vsi, filter, true);
3030         }
3031 }
3032
3033 /**
3034  * i40e_vsi_configure - Set up the VSI for action
3035  * @vsi: the VSI being configured
3036  **/
3037 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3038 {
3039         int err;
3040
3041         i40e_set_vsi_rx_mode(vsi);
3042         i40e_restore_vlan(vsi);
3043         i40e_vsi_config_dcb_rings(vsi);
3044         err = i40e_vsi_configure_tx(vsi);
3045         if (!err)
3046                 err = i40e_vsi_configure_rx(vsi);
3047
3048         return err;
3049 }
3050
3051 /**
3052  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3053  * @vsi: the VSI being configured
3054  **/
3055 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3056 {
3057         struct i40e_pf *pf = vsi->back;
3058         struct i40e_hw *hw = &pf->hw;
3059         u16 vector;
3060         int i, q;
3061         u32 qp;
3062
3063         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3064          * and PFINT_LNKLSTn registers, e.g.:
3065          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3066          */
3067         qp = vsi->base_queue;
3068         vector = vsi->base_vector;
3069         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3070                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3071
3072                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3073                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3074                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3075                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3076                      q_vector->rx.itr);
3077                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3078                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3079                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3080                      q_vector->tx.itr);
3081                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3082                      INTRL_USEC_TO_REG(vsi->int_rate_limit));
3083
3084                 /* Linked list for the queuepairs assigned to this vector */
3085                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3086                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3087                         u32 val;
3088
3089                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3090                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3091                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3092                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3093                               (I40E_QUEUE_TYPE_TX
3094                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3095
3096                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3097
3098                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3099                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3100                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3101                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3102                               (I40E_QUEUE_TYPE_RX
3103                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3104
3105                         /* Terminate the linked list */
3106                         if (q == (q_vector->num_ringpairs - 1))
3107                                 val |= (I40E_QUEUE_END_OF_LIST
3108                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3109
3110                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3111                         qp++;
3112                 }
3113         }
3114
3115         i40e_flush(hw);
3116 }
3117
3118 /**
3119  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3120  * @hw: ptr to the hardware info
3121  **/
3122 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3123 {
3124         struct i40e_hw *hw = &pf->hw;
3125         u32 val;
3126
3127         /* clear things first */
3128         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3129         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3130
3131         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3132               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3133               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3134               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3135               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3136               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3137               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3138               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3139
3140         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3141                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3142
3143         if (pf->flags & I40E_FLAG_PTP)
3144                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3145
3146         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3147
3148         /* SW_ITR_IDX = 0, but don't change INTENA */
3149         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3150                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3151
3152         /* OTHER_ITR_IDX = 0 */
3153         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3154 }
3155
3156 /**
3157  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3158  * @vsi: the VSI being configured
3159  **/
3160 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3161 {
3162         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3163         struct i40e_pf *pf = vsi->back;
3164         struct i40e_hw *hw = &pf->hw;
3165         u32 val;
3166
3167         /* set the ITR configuration */
3168         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3169         q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3170         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3171         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3172         q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3173         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3174         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3175
3176         i40e_enable_misc_int_causes(pf);
3177
3178         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3179         wr32(hw, I40E_PFINT_LNKLST0, 0);
3180
3181         /* Associate the queue pair to the vector and enable the queue int */
3182         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3183               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3184               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3185
3186         wr32(hw, I40E_QINT_RQCTL(0), val);
3187
3188         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3189               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3190               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3191
3192         wr32(hw, I40E_QINT_TQCTL(0), val);
3193         i40e_flush(hw);
3194 }
3195
3196 /**
3197  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3198  * @pf: board private structure
3199  **/
3200 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3201 {
3202         struct i40e_hw *hw = &pf->hw;
3203
3204         wr32(hw, I40E_PFINT_DYN_CTL0,
3205              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3206         i40e_flush(hw);
3207 }
3208
3209 /**
3210  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3211  * @pf: board private structure
3212  * @clearpba: true when all pending interrupt events should be cleared
3213  **/
3214 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3215 {
3216         struct i40e_hw *hw = &pf->hw;
3217         u32 val;
3218
3219         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3220               (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3221               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3222
3223         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3224         i40e_flush(hw);
3225 }
3226
3227 /**
3228  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3229  * @irq: interrupt number
3230  * @data: pointer to a q_vector
3231  **/
3232 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3233 {
3234         struct i40e_q_vector *q_vector = data;
3235
3236         if (!q_vector->tx.ring && !q_vector->rx.ring)
3237                 return IRQ_HANDLED;
3238
3239         napi_schedule_irqoff(&q_vector->napi);
3240
3241         return IRQ_HANDLED;
3242 }
3243
3244 /**
3245  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3246  * @vsi: the VSI being configured
3247  * @basename: name for the vector
3248  *
3249  * Allocates MSI-X vectors and requests interrupts from the kernel.
3250  **/
3251 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3252 {
3253         int q_vectors = vsi->num_q_vectors;
3254         struct i40e_pf *pf = vsi->back;
3255         int base = vsi->base_vector;
3256         int rx_int_idx = 0;
3257         int tx_int_idx = 0;
3258         int vector, err;
3259
3260         for (vector = 0; vector < q_vectors; vector++) {
3261                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3262
3263                 if (q_vector->tx.ring && q_vector->rx.ring) {
3264                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3265                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3266                         tx_int_idx++;
3267                 } else if (q_vector->rx.ring) {
3268                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3269                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3270                 } else if (q_vector->tx.ring) {
3271                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3272                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3273                 } else {
3274                         /* skip this unused q_vector */
3275                         continue;
3276                 }
3277                 err = request_irq(pf->msix_entries[base + vector].vector,
3278                                   vsi->irq_handler,
3279                                   0,
3280                                   q_vector->name,
3281                                   q_vector);
3282                 if (err) {
3283                         dev_info(&pf->pdev->dev,
3284                                  "MSIX request_irq failed, error: %d\n", err);
3285                         goto free_queue_irqs;
3286                 }
3287                 /* assign the mask for this irq */
3288                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3289                                       &q_vector->affinity_mask);
3290         }
3291
3292         vsi->irqs_ready = true;
3293         return 0;
3294
3295 free_queue_irqs:
3296         while (vector) {
3297                 vector--;
3298                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3299                                       NULL);
3300                 free_irq(pf->msix_entries[base + vector].vector,
3301                          &(vsi->q_vectors[vector]));
3302         }
3303         return err;
3304 }
3305
3306 /**
3307  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3308  * @vsi: the VSI being un-configured
3309  **/
3310 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3311 {
3312         struct i40e_pf *pf = vsi->back;
3313         struct i40e_hw *hw = &pf->hw;
3314         int base = vsi->base_vector;
3315         int i;
3316
3317         for (i = 0; i < vsi->num_queue_pairs; i++) {
3318                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3319                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3320         }
3321
3322         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3323                 for (i = vsi->base_vector;
3324                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3325                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3326
3327                 i40e_flush(hw);
3328                 for (i = 0; i < vsi->num_q_vectors; i++)
3329                         synchronize_irq(pf->msix_entries[i + base].vector);
3330         } else {
3331                 /* Legacy and MSI mode - this stops all interrupt handling */
3332                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3333                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3334                 i40e_flush(hw);
3335                 synchronize_irq(pf->pdev->irq);
3336         }
3337 }
3338
3339 /**
3340  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3341  * @vsi: the VSI being configured
3342  **/
3343 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3344 {
3345         struct i40e_pf *pf = vsi->back;
3346         int i;
3347
3348         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3349                 for (i = 0; i < vsi->num_q_vectors; i++)
3350                         i40e_irq_dynamic_enable(vsi, i);
3351         } else {
3352                 i40e_irq_dynamic_enable_icr0(pf, true);
3353         }
3354
3355         i40e_flush(&pf->hw);
3356         return 0;
3357 }
3358
3359 /**
3360  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3361  * @pf: board private structure
3362  **/
3363 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3364 {
3365         /* Disable ICR 0 */
3366         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3367         i40e_flush(&pf->hw);
3368 }
3369
3370 /**
3371  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3372  * @irq: interrupt number
3373  * @data: pointer to a q_vector
3374  *
3375  * This is the handler used for all MSI/Legacy interrupts, and deals
3376  * with both queue and non-queue interrupts.  This is also used in
3377  * MSIX mode to handle the non-queue interrupts.
3378  **/
3379 static irqreturn_t i40e_intr(int irq, void *data)
3380 {
3381         struct i40e_pf *pf = (struct i40e_pf *)data;
3382         struct i40e_hw *hw = &pf->hw;
3383         irqreturn_t ret = IRQ_NONE;
3384         u32 icr0, icr0_remaining;
3385         u32 val, ena_mask;
3386
3387         icr0 = rd32(hw, I40E_PFINT_ICR0);
3388         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3389
3390         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3391         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3392                 goto enable_intr;
3393
3394         /* if interrupt but no bits showing, must be SWINT */
3395         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3396             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3397                 pf->sw_int_count++;
3398
3399         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3400             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3401                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3402                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3403                 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3404         }
3405
3406         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3407         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3408                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3409                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3410
3411                 /* We do not have a way to disarm Queue causes while leaving
3412                  * interrupt enabled for all other causes, ideally
3413                  * interrupt should be disabled while we are in NAPI but
3414                  * this is not a performance path and napi_schedule()
3415                  * can deal with rescheduling.
3416                  */
3417                 if (!test_bit(__I40E_DOWN, &pf->state))
3418                         napi_schedule_irqoff(&q_vector->napi);
3419         }
3420
3421         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3422                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3423                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3424                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3425         }
3426
3427         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3428                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3429                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3430         }
3431
3432         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3433                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3434                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3435         }
3436
3437         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3438                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3439                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3440                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3441                 val = rd32(hw, I40E_GLGEN_RSTAT);
3442                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3443                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3444                 if (val == I40E_RESET_CORER) {
3445                         pf->corer_count++;
3446                 } else if (val == I40E_RESET_GLOBR) {
3447                         pf->globr_count++;
3448                 } else if (val == I40E_RESET_EMPR) {
3449                         pf->empr_count++;
3450                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3451                 }
3452         }
3453
3454         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3455                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3456                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3457                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3458                          rd32(hw, I40E_PFHMC_ERRORINFO),
3459                          rd32(hw, I40E_PFHMC_ERRORDATA));
3460         }
3461
3462         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3463                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3464
3465                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3466                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3467                         i40e_ptp_tx_hwtstamp(pf);
3468                 }
3469         }
3470
3471         /* If a critical error is pending we have no choice but to reset the
3472          * device.
3473          * Report and mask out any remaining unexpected interrupts.
3474          */
3475         icr0_remaining = icr0 & ena_mask;
3476         if (icr0_remaining) {
3477                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3478                          icr0_remaining);
3479                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3480                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3481                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3482                         dev_info(&pf->pdev->dev, "device will be reset\n");
3483                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3484                         i40e_service_event_schedule(pf);
3485                 }
3486                 ena_mask &= ~icr0_remaining;
3487         }
3488         ret = IRQ_HANDLED;
3489
3490 enable_intr:
3491         /* re-enable interrupt causes */
3492         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3493         if (!test_bit(__I40E_DOWN, &pf->state)) {
3494                 i40e_service_event_schedule(pf);
3495                 i40e_irq_dynamic_enable_icr0(pf, false);
3496         }
3497
3498         return ret;
3499 }
3500
3501 /**
3502  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3503  * @tx_ring:  tx ring to clean
3504  * @budget:   how many cleans we're allowed
3505  *
3506  * Returns true if there's any budget left (e.g. the clean is finished)
3507  **/
3508 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3509 {
3510         struct i40e_vsi *vsi = tx_ring->vsi;
3511         u16 i = tx_ring->next_to_clean;
3512         struct i40e_tx_buffer *tx_buf;
3513         struct i40e_tx_desc *tx_desc;
3514
3515         tx_buf = &tx_ring->tx_bi[i];
3516         tx_desc = I40E_TX_DESC(tx_ring, i);
3517         i -= tx_ring->count;
3518
3519         do {
3520                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3521
3522                 /* if next_to_watch is not set then there is no work pending */
3523                 if (!eop_desc)
3524                         break;
3525
3526                 /* prevent any other reads prior to eop_desc */
3527                 read_barrier_depends();
3528
3529                 /* if the descriptor isn't done, no work yet to do */
3530                 if (!(eop_desc->cmd_type_offset_bsz &
3531                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3532                         break;
3533
3534                 /* clear next_to_watch to prevent false hangs */
3535                 tx_buf->next_to_watch = NULL;
3536
3537                 tx_desc->buffer_addr = 0;
3538                 tx_desc->cmd_type_offset_bsz = 0;
3539                 /* move past filter desc */
3540                 tx_buf++;
3541                 tx_desc++;
3542                 i++;
3543                 if (unlikely(!i)) {
3544                         i -= tx_ring->count;
3545                         tx_buf = tx_ring->tx_bi;
3546                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3547                 }
3548                 /* unmap skb header data */
3549                 dma_unmap_single(tx_ring->dev,
3550                                  dma_unmap_addr(tx_buf, dma),
3551                                  dma_unmap_len(tx_buf, len),
3552                                  DMA_TO_DEVICE);
3553                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3554                         kfree(tx_buf->raw_buf);
3555
3556                 tx_buf->raw_buf = NULL;
3557                 tx_buf->tx_flags = 0;
3558                 tx_buf->next_to_watch = NULL;
3559                 dma_unmap_len_set(tx_buf, len, 0);
3560                 tx_desc->buffer_addr = 0;
3561                 tx_desc->cmd_type_offset_bsz = 0;
3562
3563                 /* move us past the eop_desc for start of next FD desc */
3564                 tx_buf++;
3565                 tx_desc++;
3566                 i++;
3567                 if (unlikely(!i)) {
3568                         i -= tx_ring->count;
3569                         tx_buf = tx_ring->tx_bi;
3570                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3571                 }
3572
3573                 /* update budget accounting */
3574                 budget--;
3575         } while (likely(budget));
3576
3577         i += tx_ring->count;
3578         tx_ring->next_to_clean = i;
3579
3580         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3581                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3582
3583         return budget > 0;
3584 }
3585
3586 /**
3587  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3588  * @irq: interrupt number
3589  * @data: pointer to a q_vector
3590  **/
3591 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3592 {
3593         struct i40e_q_vector *q_vector = data;
3594         struct i40e_vsi *vsi;
3595
3596         if (!q_vector->tx.ring)
3597                 return IRQ_HANDLED;
3598
3599         vsi = q_vector->tx.ring->vsi;
3600         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3601
3602         return IRQ_HANDLED;
3603 }
3604
3605 /**
3606  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3607  * @vsi: the VSI being configured
3608  * @v_idx: vector index
3609  * @qp_idx: queue pair index
3610  **/
3611 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3612 {
3613         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3614         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3615         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3616
3617         tx_ring->q_vector = q_vector;
3618         tx_ring->next = q_vector->tx.ring;
3619         q_vector->tx.ring = tx_ring;
3620         q_vector->tx.count++;
3621
3622         rx_ring->q_vector = q_vector;
3623         rx_ring->next = q_vector->rx.ring;
3624         q_vector->rx.ring = rx_ring;
3625         q_vector->rx.count++;
3626 }
3627
3628 /**
3629  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3630  * @vsi: the VSI being configured
3631  *
3632  * This function maps descriptor rings to the queue-specific vectors
3633  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3634  * one vector per queue pair, but on a constrained vector budget, we
3635  * group the queue pairs as "efficiently" as possible.
3636  **/
3637 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3638 {
3639         int qp_remaining = vsi->num_queue_pairs;
3640         int q_vectors = vsi->num_q_vectors;
3641         int num_ringpairs;
3642         int v_start = 0;
3643         int qp_idx = 0;
3644
3645         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3646          * group them so there are multiple queues per vector.
3647          * It is also important to go through all the vectors available to be
3648          * sure that if we don't use all the vectors, that the remaining vectors
3649          * are cleared. This is especially important when decreasing the
3650          * number of queues in use.
3651          */
3652         for (; v_start < q_vectors; v_start++) {
3653                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3654
3655                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3656
3657                 q_vector->num_ringpairs = num_ringpairs;
3658
3659                 q_vector->rx.count = 0;
3660                 q_vector->tx.count = 0;
3661                 q_vector->rx.ring = NULL;
3662                 q_vector->tx.ring = NULL;
3663
3664                 while (num_ringpairs--) {
3665                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3666                         qp_idx++;
3667                         qp_remaining--;
3668                 }
3669         }
3670 }
3671
3672 /**
3673  * i40e_vsi_request_irq - Request IRQ from the OS
3674  * @vsi: the VSI being configured
3675  * @basename: name for the vector
3676  **/
3677 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3678 {
3679         struct i40e_pf *pf = vsi->back;
3680         int err;
3681
3682         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3683                 err = i40e_vsi_request_irq_msix(vsi, basename);
3684         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3685                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3686                                   pf->int_name, pf);
3687         else
3688                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3689                                   pf->int_name, pf);
3690
3691         if (err)
3692                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3693
3694         return err;
3695 }
3696
3697 #ifdef CONFIG_NET_POLL_CONTROLLER
3698 /**
3699  * i40e_netpoll - A Polling 'interrupt' handler
3700  * @netdev: network interface device structure
3701  *
3702  * This is used by netconsole to send skbs without having to re-enable
3703  * interrupts.  It's not called while the normal interrupt routine is executing.
3704  **/
3705 #ifdef I40E_FCOE
3706 void i40e_netpoll(struct net_device *netdev)
3707 #else
3708 static void i40e_netpoll(struct net_device *netdev)
3709 #endif
3710 {
3711         struct i40e_netdev_priv *np = netdev_priv(netdev);
3712         struct i40e_vsi *vsi = np->vsi;
3713         struct i40e_pf *pf = vsi->back;
3714         int i;
3715
3716         /* if interface is down do nothing */
3717         if (test_bit(__I40E_DOWN, &vsi->state))
3718                 return;
3719
3720         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3721                 for (i = 0; i < vsi->num_q_vectors; i++)
3722                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3723         } else {
3724                 i40e_intr(pf->pdev->irq, netdev);
3725         }
3726 }
3727 #endif
3728
3729 /**
3730  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3731  * @pf: the PF being configured
3732  * @pf_q: the PF queue
3733  * @enable: enable or disable state of the queue
3734  *
3735  * This routine will wait for the given Tx queue of the PF to reach the
3736  * enabled or disabled state.
3737  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3738  * multiple retries; else will return 0 in case of success.
3739  **/
3740 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3741 {
3742         int i;
3743         u32 tx_reg;
3744
3745         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3746                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3747                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3748                         break;
3749
3750                 usleep_range(10, 20);
3751         }
3752         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3753                 return -ETIMEDOUT;
3754
3755         return 0;
3756 }
3757
3758 /**
3759  * i40e_vsi_control_tx - Start or stop a VSI's rings
3760  * @vsi: the VSI being configured
3761  * @enable: start or stop the rings
3762  **/
3763 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3764 {
3765         struct i40e_pf *pf = vsi->back;
3766         struct i40e_hw *hw = &pf->hw;
3767         int i, j, pf_q, ret = 0;
3768         u32 tx_reg;
3769
3770         pf_q = vsi->base_queue;
3771         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3772
3773                 /* warn the TX unit of coming changes */
3774                 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3775                 if (!enable)
3776                         usleep_range(10, 20);
3777
3778                 for (j = 0; j < 50; j++) {
3779                         tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3780                         if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3781                             ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3782                                 break;
3783                         usleep_range(1000, 2000);
3784                 }
3785                 /* Skip if the queue is already in the requested state */
3786                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3787                         continue;
3788
3789                 /* turn on/off the queue */
3790                 if (enable) {
3791                         wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3792                         tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3793                 } else {
3794                         tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3795                 }
3796
3797                 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3798                 /* No waiting for the Tx queue to disable */
3799                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3800                         continue;
3801
3802                 /* wait for the change to finish */
3803                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3804                 if (ret) {
3805                         dev_info(&pf->pdev->dev,
3806                                  "VSI seid %d Tx ring %d %sable timeout\n",
3807                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3808                         break;
3809                 }
3810         }
3811
3812         if (hw->revision_id == 0)
3813                 mdelay(50);
3814         return ret;
3815 }
3816
3817 /**
3818  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3819  * @pf: the PF being configured
3820  * @pf_q: the PF queue
3821  * @enable: enable or disable state of the queue
3822  *
3823  * This routine will wait for the given Rx queue of the PF to reach the
3824  * enabled or disabled state.
3825  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3826  * multiple retries; else will return 0 in case of success.
3827  **/
3828 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3829 {
3830         int i;
3831         u32 rx_reg;
3832
3833         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3834                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3835                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3836                         break;
3837
3838                 usleep_range(10, 20);
3839         }
3840         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3841                 return -ETIMEDOUT;
3842
3843         return 0;
3844 }
3845
3846 /**
3847  * i40e_vsi_control_rx - Start or stop a VSI's rings
3848  * @vsi: the VSI being configured
3849  * @enable: start or stop the rings
3850  **/
3851 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3852 {
3853         struct i40e_pf *pf = vsi->back;
3854         struct i40e_hw *hw = &pf->hw;
3855         int i, j, pf_q, ret = 0;
3856         u32 rx_reg;
3857
3858         pf_q = vsi->base_queue;
3859         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3860                 for (j = 0; j < 50; j++) {
3861                         rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3862                         if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3863                             ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3864                                 break;
3865                         usleep_range(1000, 2000);
3866                 }
3867
3868                 /* Skip if the queue is already in the requested state */
3869                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3870                         continue;
3871
3872                 /* turn on/off the queue */
3873                 if (enable)
3874                         rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3875                 else
3876                         rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3877                 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3878                 /* No waiting for the Tx queue to disable */
3879                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3880                         continue;
3881
3882                 /* wait for the change to finish */
3883                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3884                 if (ret) {
3885                         dev_info(&pf->pdev->dev,
3886                                  "VSI seid %d Rx ring %d %sable timeout\n",
3887                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3888                         break;
3889                 }
3890         }
3891
3892         return ret;
3893 }
3894
3895 /**
3896  * i40e_vsi_control_rings - Start or stop a VSI's rings
3897  * @vsi: the VSI being configured
3898  * @enable: start or stop the rings
3899  **/
3900 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3901 {
3902         int ret = 0;
3903
3904         /* do rx first for enable and last for disable */
3905         if (request) {
3906                 ret = i40e_vsi_control_rx(vsi, request);
3907                 if (ret)
3908                         return ret;
3909                 ret = i40e_vsi_control_tx(vsi, request);
3910         } else {
3911                 /* Ignore return value, we need to shutdown whatever we can */
3912                 i40e_vsi_control_tx(vsi, request);
3913                 i40e_vsi_control_rx(vsi, request);
3914         }
3915
3916         return ret;
3917 }
3918
3919 /**
3920  * i40e_vsi_free_irq - Free the irq association with the OS
3921  * @vsi: the VSI being configured
3922  **/
3923 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3924 {
3925         struct i40e_pf *pf = vsi->back;
3926         struct i40e_hw *hw = &pf->hw;
3927         int base = vsi->base_vector;
3928         u32 val, qp;
3929         int i;
3930
3931         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3932                 if (!vsi->q_vectors)
3933                         return;
3934
3935                 if (!vsi->irqs_ready)
3936                         return;
3937
3938                 vsi->irqs_ready = false;
3939                 for (i = 0; i < vsi->num_q_vectors; i++) {
3940                         u16 vector = i + base;
3941
3942                         /* free only the irqs that were actually requested */
3943                         if (!vsi->q_vectors[i] ||
3944                             !vsi->q_vectors[i]->num_ringpairs)
3945                                 continue;
3946
3947                         /* clear the affinity_mask in the IRQ descriptor */
3948                         irq_set_affinity_hint(pf->msix_entries[vector].vector,
3949                                               NULL);
3950                         free_irq(pf->msix_entries[vector].vector,
3951                                  vsi->q_vectors[i]);
3952
3953                         /* Tear down the interrupt queue link list
3954                          *
3955                          * We know that they come in pairs and always
3956                          * the Rx first, then the Tx.  To clear the
3957                          * link list, stick the EOL value into the
3958                          * next_q field of the registers.
3959                          */
3960                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3961                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3962                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3963                         val |= I40E_QUEUE_END_OF_LIST
3964                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3965                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3966
3967                         while (qp != I40E_QUEUE_END_OF_LIST) {
3968                                 u32 next;
3969
3970                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
3971
3972                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3973                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3974                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3975                                          I40E_QINT_RQCTL_INTEVENT_MASK);
3976
3977                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3978                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3979
3980                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
3981
3982                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
3983
3984                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3985                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3986
3987                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3988                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3989                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3990                                          I40E_QINT_TQCTL_INTEVENT_MASK);
3991
3992                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3993                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3994
3995                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
3996                                 qp = next;
3997                         }
3998                 }
3999         } else {
4000                 free_irq(pf->pdev->irq, pf);
4001
4002                 val = rd32(hw, I40E_PFINT_LNKLST0);
4003                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4004                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4005                 val |= I40E_QUEUE_END_OF_LIST
4006                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4007                 wr32(hw, I40E_PFINT_LNKLST0, val);
4008
4009                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4010                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4011                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4012                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4013                          I40E_QINT_RQCTL_INTEVENT_MASK);
4014
4015                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4016                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4017
4018                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4019
4020                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4021
4022                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4023                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4024                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4025                          I40E_QINT_TQCTL_INTEVENT_MASK);
4026
4027                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4028                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4029
4030                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4031         }
4032 }
4033
4034 /**
4035  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4036  * @vsi: the VSI being configured
4037  * @v_idx: Index of vector to be freed
4038  *
4039  * This function frees the memory allocated to the q_vector.  In addition if
4040  * NAPI is enabled it will delete any references to the NAPI struct prior
4041  * to freeing the q_vector.
4042  **/
4043 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4044 {
4045         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4046         struct i40e_ring *ring;
4047
4048         if (!q_vector)
4049                 return;
4050
4051         /* disassociate q_vector from rings */
4052         i40e_for_each_ring(ring, q_vector->tx)
4053                 ring->q_vector = NULL;
4054
4055         i40e_for_each_ring(ring, q_vector->rx)
4056                 ring->q_vector = NULL;
4057
4058         /* only VSI w/ an associated netdev is set up w/ NAPI */
4059         if (vsi->netdev)
4060                 netif_napi_del(&q_vector->napi);
4061
4062         vsi->q_vectors[v_idx] = NULL;
4063
4064         kfree_rcu(q_vector, rcu);
4065 }
4066
4067 /**
4068  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4069  * @vsi: the VSI being un-configured
4070  *
4071  * This frees the memory allocated to the q_vectors and
4072  * deletes references to the NAPI struct.
4073  **/
4074 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4075 {
4076         int v_idx;
4077
4078         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4079                 i40e_free_q_vector(vsi, v_idx);
4080 }
4081
4082 /**
4083  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4084  * @pf: board private structure
4085  **/
4086 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4087 {
4088         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4089         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4090                 pci_disable_msix(pf->pdev);
4091                 kfree(pf->msix_entries);
4092                 pf->msix_entries = NULL;
4093                 kfree(pf->irq_pile);
4094                 pf->irq_pile = NULL;
4095         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4096                 pci_disable_msi(pf->pdev);
4097         }
4098         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4099 }
4100
4101 /**
4102  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4103  * @pf: board private structure
4104  *
4105  * We go through and clear interrupt specific resources and reset the structure
4106  * to pre-load conditions
4107  **/
4108 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4109 {
4110         int i;
4111
4112         i40e_stop_misc_vector(pf);
4113         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4114                 synchronize_irq(pf->msix_entries[0].vector);
4115                 free_irq(pf->msix_entries[0].vector, pf);
4116         }
4117
4118         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4119                       I40E_IWARP_IRQ_PILE_ID);
4120
4121         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4122         for (i = 0; i < pf->num_alloc_vsi; i++)
4123                 if (pf->vsi[i])
4124                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4125         i40e_reset_interrupt_capability(pf);
4126 }
4127
4128 /**
4129  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4130  * @vsi: the VSI being configured
4131  **/
4132 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4133 {
4134         int q_idx;
4135
4136         if (!vsi->netdev)
4137                 return;
4138
4139         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4140                 napi_enable(&vsi->q_vectors[q_idx]->napi);
4141 }
4142
4143 /**
4144  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4145  * @vsi: the VSI being configured
4146  **/
4147 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4148 {
4149         int q_idx;
4150
4151         if (!vsi->netdev)
4152                 return;
4153
4154         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4155                 napi_disable(&vsi->q_vectors[q_idx]->napi);
4156 }
4157
4158 /**
4159  * i40e_vsi_close - Shut down a VSI
4160  * @vsi: the vsi to be quelled
4161  **/
4162 static void i40e_vsi_close(struct i40e_vsi *vsi)
4163 {
4164         bool reset = false;
4165
4166         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4167                 i40e_down(vsi);
4168         i40e_vsi_free_irq(vsi);
4169         i40e_vsi_free_tx_resources(vsi);
4170         i40e_vsi_free_rx_resources(vsi);
4171         vsi->current_netdev_flags = 0;
4172         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4173                 reset = true;
4174         i40e_notify_client_of_netdev_close(vsi, reset);
4175 }
4176
4177 /**
4178  * i40e_quiesce_vsi - Pause a given VSI
4179  * @vsi: the VSI being paused
4180  **/
4181 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4182 {
4183         if (test_bit(__I40E_DOWN, &vsi->state))
4184                 return;
4185
4186         /* No need to disable FCoE VSI when Tx suspended */
4187         if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4188             vsi->type == I40E_VSI_FCOE) {
4189                 dev_dbg(&vsi->back->pdev->dev,
4190                          "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4191                 return;
4192         }
4193
4194         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4195         if (vsi->netdev && netif_running(vsi->netdev))
4196                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4197         else
4198                 i40e_vsi_close(vsi);
4199 }
4200
4201 /**
4202  * i40e_unquiesce_vsi - Resume a given VSI
4203  * @vsi: the VSI being resumed
4204  **/
4205 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4206 {
4207         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4208                 return;
4209
4210         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4211         if (vsi->netdev && netif_running(vsi->netdev))
4212                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4213         else
4214                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4215 }
4216
4217 /**
4218  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4219  * @pf: the PF
4220  **/
4221 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4222 {
4223         int v;
4224
4225         for (v = 0; v < pf->num_alloc_vsi; v++) {
4226                 if (pf->vsi[v])
4227                         i40e_quiesce_vsi(pf->vsi[v]);
4228         }
4229 }
4230
4231 /**
4232  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4233  * @pf: the PF
4234  **/
4235 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4236 {
4237         int v;
4238
4239         for (v = 0; v < pf->num_alloc_vsi; v++) {
4240                 if (pf->vsi[v])
4241                         i40e_unquiesce_vsi(pf->vsi[v]);
4242         }
4243 }
4244
4245 #ifdef CONFIG_I40E_DCB
4246 /**
4247  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4248  * @vsi: the VSI being configured
4249  *
4250  * This function waits for the given VSI's queues to be disabled.
4251  **/
4252 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4253 {
4254         struct i40e_pf *pf = vsi->back;
4255         int i, pf_q, ret;
4256
4257         pf_q = vsi->base_queue;
4258         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4259                 /* Check and wait for the disable status of the queue */
4260                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4261                 if (ret) {
4262                         dev_info(&pf->pdev->dev,
4263                                  "VSI seid %d Tx ring %d disable timeout\n",
4264                                  vsi->seid, pf_q);
4265                         return ret;
4266                 }
4267         }
4268
4269         pf_q = vsi->base_queue;
4270         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4271                 /* Check and wait for the disable status of the queue */
4272                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4273                 if (ret) {
4274                         dev_info(&pf->pdev->dev,
4275                                  "VSI seid %d Rx ring %d disable timeout\n",
4276                                  vsi->seid, pf_q);
4277                         return ret;
4278                 }
4279         }
4280
4281         return 0;
4282 }
4283
4284 /**
4285  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4286  * @pf: the PF
4287  *
4288  * This function waits for the queues to be in disabled state for all the
4289  * VSIs that are managed by this PF.
4290  **/
4291 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4292 {
4293         int v, ret = 0;
4294
4295         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4296                 /* No need to wait for FCoE VSI queues */
4297                 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4298                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4299                         if (ret)
4300                                 break;
4301                 }
4302         }
4303
4304         return ret;
4305 }
4306
4307 #endif
4308
4309 /**
4310  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4311  * @q_idx: TX queue number
4312  * @vsi: Pointer to VSI struct
4313  *
4314  * This function checks specified queue for given VSI. Detects hung condition.
4315  * Sets hung bit since it is two step process. Before next run of service task
4316  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4317  * hung condition remain unchanged and during subsequent run, this function
4318  * issues SW interrupt to recover from hung condition.
4319  **/
4320 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4321 {
4322         struct i40e_ring *tx_ring = NULL;
4323         struct i40e_pf  *pf;
4324         u32 head, val, tx_pending_hw;
4325         int i;
4326
4327         pf = vsi->back;
4328
4329         /* now that we have an index, find the tx_ring struct */
4330         for (i = 0; i < vsi->num_queue_pairs; i++) {
4331                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4332                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4333                                 tx_ring = vsi->tx_rings[i];
4334                                 break;
4335                         }
4336                 }
4337         }
4338
4339         if (!tx_ring)
4340                 return;
4341
4342         /* Read interrupt register */
4343         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4344                 val = rd32(&pf->hw,
4345                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4346                                                tx_ring->vsi->base_vector - 1));
4347         else
4348                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4349
4350         head = i40e_get_head(tx_ring);
4351
4352         tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4353
4354         /* HW is done executing descriptors, updated HEAD write back,
4355          * but SW hasn't processed those descriptors. If interrupt is
4356          * not generated from this point ON, it could result into
4357          * dev_watchdog detecting timeout on those netdev_queue,
4358          * hence proactively trigger SW interrupt.
4359          */
4360         if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4361                 /* NAPI Poll didn't run and clear since it was set */
4362                 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4363                                        &tx_ring->q_vector->hung_detected)) {
4364                         netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4365                                     vsi->seid, q_idx, tx_pending_hw,
4366                                     tx_ring->next_to_clean, head,
4367                                     tx_ring->next_to_use,
4368                                     readl(tx_ring->tail));
4369                         netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4370                                     vsi->seid, q_idx, val);
4371                         i40e_force_wb(vsi, tx_ring->q_vector);
4372                 } else {
4373                         /* First Chance - detected possible hung */
4374                         set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4375                                 &tx_ring->q_vector->hung_detected);
4376                 }
4377         }
4378
4379         /* This is the case where we have interrupts missing,
4380          * so the tx_pending in HW will most likely be 0, but we
4381          * will have tx_pending in SW since the WB happened but the
4382          * interrupt got lost.
4383          */
4384         if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4385             (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4386                 if (napi_reschedule(&tx_ring->q_vector->napi))
4387                         tx_ring->tx_stats.tx_lost_interrupt++;
4388         }
4389 }
4390
4391 /**
4392  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4393  * @pf:  pointer to PF struct
4394  *
4395  * LAN VSI has netdev and netdev has TX queues. This function is to check
4396  * each of those TX queues if they are hung, trigger recovery by issuing
4397  * SW interrupt.
4398  **/
4399 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4400 {
4401         struct net_device *netdev;
4402         struct i40e_vsi *vsi;
4403         int i;
4404
4405         /* Only for LAN VSI */
4406         vsi = pf->vsi[pf->lan_vsi];
4407
4408         if (!vsi)
4409                 return;
4410
4411         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4412         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4413             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4414                 return;
4415
4416         /* Make sure type is MAIN VSI */
4417         if (vsi->type != I40E_VSI_MAIN)
4418                 return;
4419
4420         netdev = vsi->netdev;
4421         if (!netdev)
4422                 return;
4423
4424         /* Bail out if netif_carrier is not OK */
4425         if (!netif_carrier_ok(netdev))
4426                 return;
4427
4428         /* Go thru' TX queues for netdev */
4429         for (i = 0; i < netdev->num_tx_queues; i++) {
4430                 struct netdev_queue *q;
4431
4432                 q = netdev_get_tx_queue(netdev, i);
4433                 if (q)
4434                         i40e_detect_recover_hung_queue(i, vsi);
4435         }
4436 }
4437
4438 /**
4439  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4440  * @pf: pointer to PF
4441  *
4442  * Get TC map for ISCSI PF type that will include iSCSI TC
4443  * and LAN TC.
4444  **/
4445 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4446 {
4447         struct i40e_dcb_app_priority_table app;
4448         struct i40e_hw *hw = &pf->hw;
4449         u8 enabled_tc = 1; /* TC0 is always enabled */
4450         u8 tc, i;
4451         /* Get the iSCSI APP TLV */
4452         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4453
4454         for (i = 0; i < dcbcfg->numapps; i++) {
4455                 app = dcbcfg->app[i];
4456                 if (app.selector == I40E_APP_SEL_TCPIP &&
4457                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4458                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4459                         enabled_tc |= BIT(tc);
4460                         break;
4461                 }
4462         }
4463
4464         return enabled_tc;
4465 }
4466
4467 /**
4468  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4469  * @dcbcfg: the corresponding DCBx configuration structure
4470  *
4471  * Return the number of TCs from given DCBx configuration
4472  **/
4473 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4474 {
4475         u8 num_tc = 0;
4476         int i;
4477
4478         /* Scan the ETS Config Priority Table to find
4479          * traffic class enabled for a given priority
4480          * and use the traffic class index to get the
4481          * number of traffic classes enabled
4482          */
4483         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4484                 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4485                         num_tc = dcbcfg->etscfg.prioritytable[i];
4486         }
4487
4488         /* Traffic class index starts from zero so
4489          * increment to return the actual count
4490          */
4491         return num_tc + 1;
4492 }
4493
4494 /**
4495  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4496  * @dcbcfg: the corresponding DCBx configuration structure
4497  *
4498  * Query the current DCB configuration and return the number of
4499  * traffic classes enabled from the given DCBX config
4500  **/
4501 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4502 {
4503         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4504         u8 enabled_tc = 1;
4505         u8 i;
4506
4507         for (i = 0; i < num_tc; i++)
4508                 enabled_tc |= BIT(i);
4509
4510         return enabled_tc;
4511 }
4512
4513 /**
4514  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4515  * @pf: PF being queried
4516  *
4517  * Return number of traffic classes enabled for the given PF
4518  **/
4519 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4520 {
4521         struct i40e_hw *hw = &pf->hw;
4522         u8 i, enabled_tc;
4523         u8 num_tc = 0;
4524         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4525
4526         /* If DCB is not enabled then always in single TC */
4527         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4528                 return 1;
4529
4530         /* SFP mode will be enabled for all TCs on port */
4531         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4532                 return i40e_dcb_get_num_tc(dcbcfg);
4533
4534         /* MFP mode return count of enabled TCs for this PF */
4535         if (pf->hw.func_caps.iscsi)
4536                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4537         else
4538                 return 1; /* Only TC0 */
4539
4540         /* At least have TC0 */
4541         enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4542         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4543                 if (enabled_tc & BIT(i))
4544                         num_tc++;
4545         }
4546         return num_tc;
4547 }
4548
4549 /**
4550  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4551  * @pf: PF being queried
4552  *
4553  * Return a bitmap for first enabled traffic class for this PF.
4554  **/
4555 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4556 {
4557         u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4558         u8 i = 0;
4559
4560         if (!enabled_tc)
4561                 return 0x1; /* TC0 */
4562
4563         /* Find the first enabled TC */
4564         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4565                 if (enabled_tc & BIT(i))
4566                         break;
4567         }
4568
4569         return BIT(i);
4570 }
4571
4572 /**
4573  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4574  * @pf: PF being queried
4575  *
4576  * Return a bitmap for enabled traffic classes for this PF.
4577  **/
4578 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4579 {
4580         /* If DCB is not enabled for this PF then just return default TC */
4581         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4582                 return i40e_pf_get_default_tc(pf);
4583
4584         /* SFP mode we want PF to be enabled for all TCs */
4585         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4586                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4587
4588         /* MFP enabled and iSCSI PF type */
4589         if (pf->hw.func_caps.iscsi)
4590                 return i40e_get_iscsi_tc_map(pf);
4591         else
4592                 return i40e_pf_get_default_tc(pf);
4593 }
4594
4595 /**
4596  * i40e_vsi_get_bw_info - Query VSI BW Information
4597  * @vsi: the VSI being queried
4598  *
4599  * Returns 0 on success, negative value on failure
4600  **/
4601 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4602 {
4603         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4604         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4605         struct i40e_pf *pf = vsi->back;
4606         struct i40e_hw *hw = &pf->hw;
4607         i40e_status ret;
4608         u32 tc_bw_max;
4609         int i;
4610
4611         /* Get the VSI level BW configuration */
4612         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4613         if (ret) {
4614                 dev_info(&pf->pdev->dev,
4615                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4616                          i40e_stat_str(&pf->hw, ret),
4617                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4618                 return -EINVAL;
4619         }
4620
4621         /* Get the VSI level BW configuration per TC */
4622         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4623                                                NULL);
4624         if (ret) {
4625                 dev_info(&pf->pdev->dev,
4626                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4627                          i40e_stat_str(&pf->hw, ret),
4628                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4629                 return -EINVAL;
4630         }
4631
4632         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4633                 dev_info(&pf->pdev->dev,
4634                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4635                          bw_config.tc_valid_bits,
4636                          bw_ets_config.tc_valid_bits);
4637                 /* Still continuing */
4638         }
4639
4640         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4641         vsi->bw_max_quanta = bw_config.max_bw;
4642         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4643                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4644         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4645                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4646                 vsi->bw_ets_limit_credits[i] =
4647                                         le16_to_cpu(bw_ets_config.credits[i]);
4648                 /* 3 bits out of 4 for each TC */
4649                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4650         }
4651
4652         return 0;
4653 }
4654
4655 /**
4656  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4657  * @vsi: the VSI being configured
4658  * @enabled_tc: TC bitmap
4659  * @bw_credits: BW shared credits per TC
4660  *
4661  * Returns 0 on success, negative value on failure
4662  **/
4663 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4664                                        u8 *bw_share)
4665 {
4666         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4667         i40e_status ret;
4668         int i;
4669
4670         bw_data.tc_valid_bits = enabled_tc;
4671         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4672                 bw_data.tc_bw_credits[i] = bw_share[i];
4673
4674         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4675                                        NULL);
4676         if (ret) {
4677                 dev_info(&vsi->back->pdev->dev,
4678                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4679                          vsi->back->hw.aq.asq_last_status);
4680                 return -EINVAL;
4681         }
4682
4683         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4684                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4685
4686         return 0;
4687 }
4688
4689 /**
4690  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4691  * @vsi: the VSI being configured
4692  * @enabled_tc: TC map to be enabled
4693  *
4694  **/
4695 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4696 {
4697         struct net_device *netdev = vsi->netdev;
4698         struct i40e_pf *pf = vsi->back;
4699         struct i40e_hw *hw = &pf->hw;
4700         u8 netdev_tc = 0;
4701         int i;
4702         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4703
4704         if (!netdev)
4705                 return;
4706
4707         if (!enabled_tc) {
4708                 netdev_reset_tc(netdev);
4709                 return;
4710         }
4711
4712         /* Set up actual enabled TCs on the VSI */
4713         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4714                 return;
4715
4716         /* set per TC queues for the VSI */
4717         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4718                 /* Only set TC queues for enabled tcs
4719                  *
4720                  * e.g. For a VSI that has TC0 and TC3 enabled the
4721                  * enabled_tc bitmap would be 0x00001001; the driver
4722                  * will set the numtc for netdev as 2 that will be
4723                  * referenced by the netdev layer as TC 0 and 1.
4724                  */
4725                 if (vsi->tc_config.enabled_tc & BIT(i))
4726                         netdev_set_tc_queue(netdev,
4727                                         vsi->tc_config.tc_info[i].netdev_tc,
4728                                         vsi->tc_config.tc_info[i].qcount,
4729                                         vsi->tc_config.tc_info[i].qoffset);
4730         }
4731
4732         /* Assign UP2TC map for the VSI */
4733         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4734                 /* Get the actual TC# for the UP */
4735                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4736                 /* Get the mapped netdev TC# for the UP */
4737                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4738                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4739         }
4740 }
4741
4742 /**
4743  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4744  * @vsi: the VSI being configured
4745  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4746  **/
4747 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4748                                       struct i40e_vsi_context *ctxt)
4749 {
4750         /* copy just the sections touched not the entire info
4751          * since not all sections are valid as returned by
4752          * update vsi params
4753          */
4754         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4755         memcpy(&vsi->info.queue_mapping,
4756                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4757         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4758                sizeof(vsi->info.tc_mapping));
4759 }
4760
4761 /**
4762  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4763  * @vsi: VSI to be configured
4764  * @enabled_tc: TC bitmap
4765  *
4766  * This configures a particular VSI for TCs that are mapped to the
4767  * given TC bitmap. It uses default bandwidth share for TCs across
4768  * VSIs to configure TC for a particular VSI.
4769  *
4770  * NOTE:
4771  * It is expected that the VSI queues have been quisced before calling
4772  * this function.
4773  **/
4774 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4775 {
4776         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4777         struct i40e_vsi_context ctxt;
4778         int ret = 0;
4779         int i;
4780
4781         /* Check if enabled_tc is same as existing or new TCs */
4782         if (vsi->tc_config.enabled_tc == enabled_tc)
4783                 return ret;
4784
4785         /* Enable ETS TCs with equal BW Share for now across all VSIs */
4786         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4787                 if (enabled_tc & BIT(i))
4788                         bw_share[i] = 1;
4789         }
4790
4791         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4792         if (ret) {
4793                 dev_info(&vsi->back->pdev->dev,
4794                          "Failed configuring TC map %d for VSI %d\n",
4795                          enabled_tc, vsi->seid);
4796                 goto out;
4797         }
4798
4799         /* Update Queue Pairs Mapping for currently enabled UPs */
4800         ctxt.seid = vsi->seid;
4801         ctxt.pf_num = vsi->back->hw.pf_id;
4802         ctxt.vf_num = 0;
4803         ctxt.uplink_seid = vsi->uplink_seid;
4804         ctxt.info = vsi->info;
4805         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4806
4807         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4808                 ctxt.info.valid_sections |=
4809                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4810                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4811         }
4812
4813         /* Update the VSI after updating the VSI queue-mapping information */
4814         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4815         if (ret) {
4816                 dev_info(&vsi->back->pdev->dev,
4817                          "Update vsi tc config failed, err %s aq_err %s\n",
4818                          i40e_stat_str(&vsi->back->hw, ret),
4819                          i40e_aq_str(&vsi->back->hw,
4820                                      vsi->back->hw.aq.asq_last_status));
4821                 goto out;
4822         }
4823         /* update the local VSI info with updated queue map */
4824         i40e_vsi_update_queue_map(vsi, &ctxt);
4825         vsi->info.valid_sections = 0;
4826
4827         /* Update current VSI BW information */
4828         ret = i40e_vsi_get_bw_info(vsi);
4829         if (ret) {
4830                 dev_info(&vsi->back->pdev->dev,
4831                          "Failed updating vsi bw info, err %s aq_err %s\n",
4832                          i40e_stat_str(&vsi->back->hw, ret),
4833                          i40e_aq_str(&vsi->back->hw,
4834                                      vsi->back->hw.aq.asq_last_status));
4835                 goto out;
4836         }
4837
4838         /* Update the netdev TC setup */
4839         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4840 out:
4841         return ret;
4842 }
4843
4844 /**
4845  * i40e_veb_config_tc - Configure TCs for given VEB
4846  * @veb: given VEB
4847  * @enabled_tc: TC bitmap
4848  *
4849  * Configures given TC bitmap for VEB (switching) element
4850  **/
4851 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4852 {
4853         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4854         struct i40e_pf *pf = veb->pf;
4855         int ret = 0;
4856         int i;
4857
4858         /* No TCs or already enabled TCs just return */
4859         if (!enabled_tc || veb->enabled_tc == enabled_tc)
4860                 return ret;
4861
4862         bw_data.tc_valid_bits = enabled_tc;
4863         /* bw_data.absolute_credits is not set (relative) */
4864
4865         /* Enable ETS TCs with equal BW Share for now */
4866         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4867                 if (enabled_tc & BIT(i))
4868                         bw_data.tc_bw_share_credits[i] = 1;
4869         }
4870
4871         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4872                                                    &bw_data, NULL);
4873         if (ret) {
4874                 dev_info(&pf->pdev->dev,
4875                          "VEB bw config failed, err %s aq_err %s\n",
4876                          i40e_stat_str(&pf->hw, ret),
4877                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4878                 goto out;
4879         }
4880
4881         /* Update the BW information */
4882         ret = i40e_veb_get_bw_info(veb);
4883         if (ret) {
4884                 dev_info(&pf->pdev->dev,
4885                          "Failed getting veb bw config, err %s aq_err %s\n",
4886                          i40e_stat_str(&pf->hw, ret),
4887                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4888         }
4889
4890 out:
4891         return ret;
4892 }
4893
4894 #ifdef CONFIG_I40E_DCB
4895 /**
4896  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4897  * @pf: PF struct
4898  *
4899  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4900  * the caller would've quiesce all the VSIs before calling
4901  * this function
4902  **/
4903 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4904 {
4905         u8 tc_map = 0;
4906         int ret;
4907         u8 v;
4908
4909         /* Enable the TCs available on PF to all VEBs */
4910         tc_map = i40e_pf_get_tc_map(pf);
4911         for (v = 0; v < I40E_MAX_VEB; v++) {
4912                 if (!pf->veb[v])
4913                         continue;
4914                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4915                 if (ret) {
4916                         dev_info(&pf->pdev->dev,
4917                                  "Failed configuring TC for VEB seid=%d\n",
4918                                  pf->veb[v]->seid);
4919                         /* Will try to configure as many components */
4920                 }
4921         }
4922
4923         /* Update each VSI */
4924         for (v = 0; v < pf->num_alloc_vsi; v++) {
4925                 if (!pf->vsi[v])
4926                         continue;
4927
4928                 /* - Enable all TCs for the LAN VSI
4929 #ifdef I40E_FCOE
4930                  * - For FCoE VSI only enable the TC configured
4931                  *   as per the APP TLV
4932 #endif
4933                  * - For all others keep them at TC0 for now
4934                  */
4935                 if (v == pf->lan_vsi)
4936                         tc_map = i40e_pf_get_tc_map(pf);
4937                 else
4938                         tc_map = i40e_pf_get_default_tc(pf);
4939 #ifdef I40E_FCOE
4940                 if (pf->vsi[v]->type == I40E_VSI_FCOE)
4941                         tc_map = i40e_get_fcoe_tc_map(pf);
4942 #endif /* #ifdef I40E_FCOE */
4943
4944                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4945                 if (ret) {
4946                         dev_info(&pf->pdev->dev,
4947                                  "Failed configuring TC for VSI seid=%d\n",
4948                                  pf->vsi[v]->seid);
4949                         /* Will try to configure as many components */
4950                 } else {
4951                         /* Re-configure VSI vectors based on updated TC map */
4952                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4953                         if (pf->vsi[v]->netdev)
4954                                 i40e_dcbnl_set_all(pf->vsi[v]);
4955                 }
4956                 i40e_notify_client_of_l2_param_changes(pf->vsi[v]);
4957         }
4958 }
4959
4960 /**
4961  * i40e_resume_port_tx - Resume port Tx
4962  * @pf: PF struct
4963  *
4964  * Resume a port's Tx and issue a PF reset in case of failure to
4965  * resume.
4966  **/
4967 static int i40e_resume_port_tx(struct i40e_pf *pf)
4968 {
4969         struct i40e_hw *hw = &pf->hw;
4970         int ret;
4971
4972         ret = i40e_aq_resume_port_tx(hw, NULL);
4973         if (ret) {
4974                 dev_info(&pf->pdev->dev,
4975                          "Resume Port Tx failed, err %s aq_err %s\n",
4976                           i40e_stat_str(&pf->hw, ret),
4977                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4978                 /* Schedule PF reset to recover */
4979                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4980                 i40e_service_event_schedule(pf);
4981         }
4982
4983         return ret;
4984 }
4985
4986 /**
4987  * i40e_init_pf_dcb - Initialize DCB configuration
4988  * @pf: PF being configured
4989  *
4990  * Query the current DCB configuration and cache it
4991  * in the hardware structure
4992  **/
4993 static int i40e_init_pf_dcb(struct i40e_pf *pf)
4994 {
4995         struct i40e_hw *hw = &pf->hw;
4996         int err = 0;
4997
4998         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
4999         if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5000                 goto out;
5001
5002         /* Get the initial DCB configuration */
5003         err = i40e_init_dcb(hw);
5004         if (!err) {
5005                 /* Device/Function is not DCBX capable */
5006                 if ((!hw->func_caps.dcb) ||
5007                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5008                         dev_info(&pf->pdev->dev,
5009                                  "DCBX offload is not supported or is disabled for this PF.\n");
5010
5011                         if (pf->flags & I40E_FLAG_MFP_ENABLED)
5012                                 goto out;
5013
5014                 } else {
5015                         /* When status is not DISABLED then DCBX in FW */
5016                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5017                                        DCB_CAP_DCBX_VER_IEEE;
5018
5019                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5020                         /* Enable DCB tagging only when more than one TC */
5021                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5022                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5023                         dev_dbg(&pf->pdev->dev,
5024                                 "DCBX offload is supported for this PF.\n");
5025                 }
5026         } else {
5027                 dev_info(&pf->pdev->dev,
5028                          "Query for DCB configuration failed, err %s aq_err %s\n",
5029                          i40e_stat_str(&pf->hw, err),
5030                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5031         }
5032
5033 out:
5034         return err;
5035 }
5036 #endif /* CONFIG_I40E_DCB */
5037 #define SPEED_SIZE 14
5038 #define FC_SIZE 8
5039 /**
5040  * i40e_print_link_message - print link up or down
5041  * @vsi: the VSI for which link needs a message
5042  */
5043 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5044 {
5045         char *speed = "Unknown";
5046         char *fc = "Unknown";
5047
5048         if (vsi->current_isup == isup)
5049                 return;
5050         vsi->current_isup = isup;
5051         if (!isup) {
5052                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5053                 return;
5054         }
5055
5056         /* Warn user if link speed on NPAR enabled partition is not at
5057          * least 10GB
5058          */
5059         if (vsi->back->hw.func_caps.npar_enable &&
5060             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5061              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5062                 netdev_warn(vsi->netdev,
5063                             "The partition detected link speed that is less than 10Gbps\n");
5064
5065         switch (vsi->back->hw.phy.link_info.link_speed) {
5066         case I40E_LINK_SPEED_40GB:
5067                 speed = "40 G";
5068                 break;
5069         case I40E_LINK_SPEED_20GB:
5070                 speed = "20 G";
5071                 break;
5072         case I40E_LINK_SPEED_10GB:
5073                 speed = "10 G";
5074                 break;
5075         case I40E_LINK_SPEED_1GB:
5076                 speed = "1000 M";
5077                 break;
5078         case I40E_LINK_SPEED_100MB:
5079                 speed = "100 M";
5080                 break;
5081         default:
5082                 break;
5083         }
5084
5085         switch (vsi->back->hw.fc.current_mode) {
5086         case I40E_FC_FULL:
5087                 fc = "RX/TX";
5088                 break;
5089         case I40E_FC_TX_PAUSE:
5090                 fc = "TX";
5091                 break;
5092         case I40E_FC_RX_PAUSE:
5093                 fc = "RX";
5094                 break;
5095         default:
5096                 fc = "None";
5097                 break;
5098         }
5099
5100         netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5101                     speed, fc);
5102 }
5103
5104 /**
5105  * i40e_up_complete - Finish the last steps of bringing up a connection
5106  * @vsi: the VSI being configured
5107  **/
5108 static int i40e_up_complete(struct i40e_vsi *vsi)
5109 {
5110         struct i40e_pf *pf = vsi->back;
5111         int err;
5112
5113         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5114                 i40e_vsi_configure_msix(vsi);
5115         else
5116                 i40e_configure_msi_and_legacy(vsi);
5117
5118         /* start rings */
5119         err = i40e_vsi_control_rings(vsi, true);
5120         if (err)
5121                 return err;
5122
5123         clear_bit(__I40E_DOWN, &vsi->state);
5124         i40e_napi_enable_all(vsi);
5125         i40e_vsi_enable_irq(vsi);
5126
5127         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5128             (vsi->netdev)) {
5129                 i40e_print_link_message(vsi, true);
5130                 netif_tx_start_all_queues(vsi->netdev);
5131                 netif_carrier_on(vsi->netdev);
5132         } else if (vsi->netdev) {
5133                 i40e_print_link_message(vsi, false);
5134                 /* need to check for qualified module here*/
5135                 if ((pf->hw.phy.link_info.link_info &
5136                         I40E_AQ_MEDIA_AVAILABLE) &&
5137                     (!(pf->hw.phy.link_info.an_info &
5138                         I40E_AQ_QUALIFIED_MODULE)))
5139                         netdev_err(vsi->netdev,
5140                                    "the driver failed to link because an unqualified module was detected.");
5141         }
5142
5143         /* replay FDIR SB filters */
5144         if (vsi->type == I40E_VSI_FDIR) {
5145                 /* reset fd counters */
5146                 pf->fd_add_err = pf->fd_atr_cnt = 0;
5147                 if (pf->fd_tcp_rule > 0) {
5148                         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5149                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5150                                 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5151                         pf->fd_tcp_rule = 0;
5152                 }
5153                 i40e_fdir_filter_restore(vsi);
5154         }
5155
5156         /* On the next run of the service_task, notify any clients of the new
5157          * opened netdev
5158          */
5159         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5160         i40e_service_event_schedule(pf);
5161
5162         return 0;
5163 }
5164
5165 /**
5166  * i40e_vsi_reinit_locked - Reset the VSI
5167  * @vsi: the VSI being configured
5168  *
5169  * Rebuild the ring structs after some configuration
5170  * has changed, e.g. MTU size.
5171  **/
5172 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5173 {
5174         struct i40e_pf *pf = vsi->back;
5175
5176         WARN_ON(in_interrupt());
5177         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5178                 usleep_range(1000, 2000);
5179         i40e_down(vsi);
5180
5181         /* Give a VF some time to respond to the reset.  The
5182          * two second wait is based upon the watchdog cycle in
5183          * the VF driver.
5184          */
5185         if (vsi->type == I40E_VSI_SRIOV)
5186                 msleep(2000);
5187         i40e_up(vsi);
5188         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5189 }
5190
5191 /**
5192  * i40e_up - Bring the connection back up after being down
5193  * @vsi: the VSI being configured
5194  **/
5195 int i40e_up(struct i40e_vsi *vsi)
5196 {
5197         int err;
5198
5199         err = i40e_vsi_configure(vsi);
5200         if (!err)
5201                 err = i40e_up_complete(vsi);
5202
5203         return err;
5204 }
5205
5206 /**
5207  * i40e_down - Shutdown the connection processing
5208  * @vsi: the VSI being stopped
5209  **/
5210 void i40e_down(struct i40e_vsi *vsi)
5211 {
5212         int i;
5213
5214         /* It is assumed that the caller of this function
5215          * sets the vsi->state __I40E_DOWN bit.
5216          */
5217         if (vsi->netdev) {
5218                 netif_carrier_off(vsi->netdev);
5219                 netif_tx_disable(vsi->netdev);
5220         }
5221         i40e_vsi_disable_irq(vsi);
5222         i40e_vsi_control_rings(vsi, false);
5223         i40e_napi_disable_all(vsi);
5224
5225         for (i = 0; i < vsi->num_queue_pairs; i++) {
5226                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5227                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5228         }
5229 }
5230
5231 /**
5232  * i40e_setup_tc - configure multiple traffic classes
5233  * @netdev: net device to configure
5234  * @tc: number of traffic classes to enable
5235  **/
5236 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5237 {
5238         struct i40e_netdev_priv *np = netdev_priv(netdev);
5239         struct i40e_vsi *vsi = np->vsi;
5240         struct i40e_pf *pf = vsi->back;
5241         u8 enabled_tc = 0;
5242         int ret = -EINVAL;
5243         int i;
5244
5245         /* Check if DCB enabled to continue */
5246         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5247                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5248                 goto exit;
5249         }
5250
5251         /* Check if MFP enabled */
5252         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5253                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5254                 goto exit;
5255         }
5256
5257         /* Check whether tc count is within enabled limit */
5258         if (tc > i40e_pf_get_num_tc(pf)) {
5259                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5260                 goto exit;
5261         }
5262
5263         /* Generate TC map for number of tc requested */
5264         for (i = 0; i < tc; i++)
5265                 enabled_tc |= BIT(i);
5266
5267         /* Requesting same TC configuration as already enabled */
5268         if (enabled_tc == vsi->tc_config.enabled_tc)
5269                 return 0;
5270
5271         /* Quiesce VSI queues */
5272         i40e_quiesce_vsi(vsi);
5273
5274         /* Configure VSI for enabled TCs */
5275         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5276         if (ret) {
5277                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5278                             vsi->seid);
5279                 goto exit;
5280         }
5281
5282         /* Unquiesce VSI */
5283         i40e_unquiesce_vsi(vsi);
5284
5285 exit:
5286         return ret;
5287 }
5288
5289 #ifdef I40E_FCOE
5290 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5291                     struct tc_to_netdev *tc)
5292 #else
5293 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5294                            struct tc_to_netdev *tc)
5295 #endif
5296 {
5297         if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5298                 return -EINVAL;
5299         return i40e_setup_tc(netdev, tc->tc);
5300 }
5301
5302 /**
5303  * i40e_open - Called when a network interface is made active
5304  * @netdev: network interface device structure
5305  *
5306  * The open entry point is called when a network interface is made
5307  * active by the system (IFF_UP).  At this point all resources needed
5308  * for transmit and receive operations are allocated, the interrupt
5309  * handler is registered with the OS, the netdev watchdog subtask is
5310  * enabled, and the stack is notified that the interface is ready.
5311  *
5312  * Returns 0 on success, negative value on failure
5313  **/
5314 int i40e_open(struct net_device *netdev)
5315 {
5316         struct i40e_netdev_priv *np = netdev_priv(netdev);
5317         struct i40e_vsi *vsi = np->vsi;
5318         struct i40e_pf *pf = vsi->back;
5319         int err;
5320
5321         /* disallow open during test or if eeprom is broken */
5322         if (test_bit(__I40E_TESTING, &pf->state) ||
5323             test_bit(__I40E_BAD_EEPROM, &pf->state))
5324                 return -EBUSY;
5325
5326         netif_carrier_off(netdev);
5327
5328         err = i40e_vsi_open(vsi);
5329         if (err)
5330                 return err;
5331
5332         /* configure global TSO hardware offload settings */
5333         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5334                                                        TCP_FLAG_FIN) >> 16);
5335         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5336                                                        TCP_FLAG_FIN |
5337                                                        TCP_FLAG_CWR) >> 16);
5338         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5339
5340 #ifdef CONFIG_I40E_VXLAN
5341         vxlan_get_rx_port(netdev);
5342 #endif
5343 #ifdef CONFIG_I40E_GENEVE
5344         if (pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE)
5345                 geneve_get_rx_port(netdev);
5346 #endif
5347
5348         i40e_notify_client_of_netdev_open(vsi);
5349
5350         return 0;
5351 }
5352
5353 /**
5354  * i40e_vsi_open -
5355  * @vsi: the VSI to open
5356  *
5357  * Finish initialization of the VSI.
5358  *
5359  * Returns 0 on success, negative value on failure
5360  **/
5361 int i40e_vsi_open(struct i40e_vsi *vsi)
5362 {
5363         struct i40e_pf *pf = vsi->back;
5364         char int_name[I40E_INT_NAME_STR_LEN];
5365         int err;
5366
5367         /* allocate descriptors */
5368         err = i40e_vsi_setup_tx_resources(vsi);
5369         if (err)
5370                 goto err_setup_tx;
5371         err = i40e_vsi_setup_rx_resources(vsi);
5372         if (err)
5373                 goto err_setup_rx;
5374
5375         err = i40e_vsi_configure(vsi);
5376         if (err)
5377                 goto err_setup_rx;
5378
5379         if (vsi->netdev) {
5380                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5381                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5382                 err = i40e_vsi_request_irq(vsi, int_name);
5383                 if (err)
5384                         goto err_setup_rx;
5385
5386                 /* Notify the stack of the actual queue counts. */
5387                 err = netif_set_real_num_tx_queues(vsi->netdev,
5388                                                    vsi->num_queue_pairs);
5389                 if (err)
5390                         goto err_set_queues;
5391
5392                 err = netif_set_real_num_rx_queues(vsi->netdev,
5393                                                    vsi->num_queue_pairs);
5394                 if (err)
5395                         goto err_set_queues;
5396
5397         } else if (vsi->type == I40E_VSI_FDIR) {
5398                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5399                          dev_driver_string(&pf->pdev->dev),
5400                          dev_name(&pf->pdev->dev));
5401                 err = i40e_vsi_request_irq(vsi, int_name);
5402
5403         } else {
5404                 err = -EINVAL;
5405                 goto err_setup_rx;
5406         }
5407
5408         err = i40e_up_complete(vsi);
5409         if (err)
5410                 goto err_up_complete;
5411
5412         return 0;
5413
5414 err_up_complete:
5415         i40e_down(vsi);
5416 err_set_queues:
5417         i40e_vsi_free_irq(vsi);
5418 err_setup_rx:
5419         i40e_vsi_free_rx_resources(vsi);
5420 err_setup_tx:
5421         i40e_vsi_free_tx_resources(vsi);
5422         if (vsi == pf->vsi[pf->lan_vsi])
5423                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5424
5425         return err;
5426 }
5427
5428 /**
5429  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5430  * @pf: Pointer to PF
5431  *
5432  * This function destroys the hlist where all the Flow Director
5433  * filters were saved.
5434  **/
5435 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5436 {
5437         struct i40e_fdir_filter *filter;
5438         struct hlist_node *node2;
5439
5440         hlist_for_each_entry_safe(filter, node2,
5441                                   &pf->fdir_filter_list, fdir_node) {
5442                 hlist_del(&filter->fdir_node);
5443                 kfree(filter);
5444         }
5445         pf->fdir_pf_active_filters = 0;
5446 }
5447
5448 /**
5449  * i40e_close - Disables a network interface
5450  * @netdev: network interface device structure
5451  *
5452  * The close entry point is called when an interface is de-activated
5453  * by the OS.  The hardware is still under the driver's control, but
5454  * this netdev interface is disabled.
5455  *
5456  * Returns 0, this is not allowed to fail
5457  **/
5458 int i40e_close(struct net_device *netdev)
5459 {
5460         struct i40e_netdev_priv *np = netdev_priv(netdev);
5461         struct i40e_vsi *vsi = np->vsi;
5462
5463         i40e_vsi_close(vsi);
5464
5465         return 0;
5466 }
5467
5468 /**
5469  * i40e_do_reset - Start a PF or Core Reset sequence
5470  * @pf: board private structure
5471  * @reset_flags: which reset is requested
5472  *
5473  * The essential difference in resets is that the PF Reset
5474  * doesn't clear the packet buffers, doesn't reset the PE
5475  * firmware, and doesn't bother the other PFs on the chip.
5476  **/
5477 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5478 {
5479         u32 val;
5480
5481         WARN_ON(in_interrupt());
5482
5483
5484         /* do the biggest reset indicated */
5485         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5486
5487                 /* Request a Global Reset
5488                  *
5489                  * This will start the chip's countdown to the actual full
5490                  * chip reset event, and a warning interrupt to be sent
5491                  * to all PFs, including the requestor.  Our handler
5492                  * for the warning interrupt will deal with the shutdown
5493                  * and recovery of the switch setup.
5494                  */
5495                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5496                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5497                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5498                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5499
5500         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5501
5502                 /* Request a Core Reset
5503                  *
5504                  * Same as Global Reset, except does *not* include the MAC/PHY
5505                  */
5506                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5507                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5508                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5509                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5510                 i40e_flush(&pf->hw);
5511
5512         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5513
5514                 /* Request a PF Reset
5515                  *
5516                  * Resets only the PF-specific registers
5517                  *
5518                  * This goes directly to the tear-down and rebuild of
5519                  * the switch, since we need to do all the recovery as
5520                  * for the Core Reset.
5521                  */
5522                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5523                 i40e_handle_reset_warning(pf);
5524
5525         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5526                 int v;
5527
5528                 /* Find the VSI(s) that requested a re-init */
5529                 dev_info(&pf->pdev->dev,
5530                          "VSI reinit requested\n");
5531                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5532                         struct i40e_vsi *vsi = pf->vsi[v];
5533
5534                         if (vsi != NULL &&
5535                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5536                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5537                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5538                         }
5539                 }
5540         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5541                 int v;
5542
5543                 /* Find the VSI(s) that needs to be brought down */
5544                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5545                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5546                         struct i40e_vsi *vsi = pf->vsi[v];
5547
5548                         if (vsi != NULL &&
5549                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5550                                 set_bit(__I40E_DOWN, &vsi->state);
5551                                 i40e_down(vsi);
5552                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5553                         }
5554                 }
5555         } else {
5556                 dev_info(&pf->pdev->dev,
5557                          "bad reset request 0x%08x\n", reset_flags);
5558         }
5559 }
5560
5561 #ifdef CONFIG_I40E_DCB
5562 /**
5563  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5564  * @pf: board private structure
5565  * @old_cfg: current DCB config
5566  * @new_cfg: new DCB config
5567  **/
5568 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5569                             struct i40e_dcbx_config *old_cfg,
5570                             struct i40e_dcbx_config *new_cfg)
5571 {
5572         bool need_reconfig = false;
5573
5574         /* Check if ETS configuration has changed */
5575         if (memcmp(&new_cfg->etscfg,
5576                    &old_cfg->etscfg,
5577                    sizeof(new_cfg->etscfg))) {
5578                 /* If Priority Table has changed reconfig is needed */
5579                 if (memcmp(&new_cfg->etscfg.prioritytable,
5580                            &old_cfg->etscfg.prioritytable,
5581                            sizeof(new_cfg->etscfg.prioritytable))) {
5582                         need_reconfig = true;
5583                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5584                 }
5585
5586                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5587                            &old_cfg->etscfg.tcbwtable,
5588                            sizeof(new_cfg->etscfg.tcbwtable)))
5589                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5590
5591                 if (memcmp(&new_cfg->etscfg.tsatable,
5592                            &old_cfg->etscfg.tsatable,
5593                            sizeof(new_cfg->etscfg.tsatable)))
5594                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5595         }
5596
5597         /* Check if PFC configuration has changed */
5598         if (memcmp(&new_cfg->pfc,
5599                    &old_cfg->pfc,
5600                    sizeof(new_cfg->pfc))) {
5601                 need_reconfig = true;
5602                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5603         }
5604
5605         /* Check if APP Table has changed */
5606         if (memcmp(&new_cfg->app,
5607                    &old_cfg->app,
5608                    sizeof(new_cfg->app))) {
5609                 need_reconfig = true;
5610                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5611         }
5612
5613         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5614         return need_reconfig;
5615 }
5616
5617 /**
5618  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5619  * @pf: board private structure
5620  * @e: event info posted on ARQ
5621  **/
5622 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5623                                   struct i40e_arq_event_info *e)
5624 {
5625         struct i40e_aqc_lldp_get_mib *mib =
5626                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5627         struct i40e_hw *hw = &pf->hw;
5628         struct i40e_dcbx_config tmp_dcbx_cfg;
5629         bool need_reconfig = false;
5630         int ret = 0;
5631         u8 type;
5632
5633         /* Not DCB capable or capability disabled */
5634         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5635                 return ret;
5636
5637         /* Ignore if event is not for Nearest Bridge */
5638         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5639                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5640         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5641         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5642                 return ret;
5643
5644         /* Check MIB Type and return if event for Remote MIB update */
5645         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5646         dev_dbg(&pf->pdev->dev,
5647                 "LLDP event mib type %s\n", type ? "remote" : "local");
5648         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5649                 /* Update the remote cached instance and return */
5650                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5651                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5652                                 &hw->remote_dcbx_config);
5653                 goto exit;
5654         }
5655
5656         /* Store the old configuration */
5657         tmp_dcbx_cfg = hw->local_dcbx_config;
5658
5659         /* Reset the old DCBx configuration data */
5660         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5661         /* Get updated DCBX data from firmware */
5662         ret = i40e_get_dcb_config(&pf->hw);
5663         if (ret) {
5664                 dev_info(&pf->pdev->dev,
5665                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5666                          i40e_stat_str(&pf->hw, ret),
5667                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5668                 goto exit;
5669         }
5670
5671         /* No change detected in DCBX configs */
5672         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5673                     sizeof(tmp_dcbx_cfg))) {
5674                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5675                 goto exit;
5676         }
5677
5678         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5679                                                &hw->local_dcbx_config);
5680
5681         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5682
5683         if (!need_reconfig)
5684                 goto exit;
5685
5686         /* Enable DCB tagging only when more than one TC */
5687         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5688                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5689         else
5690                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5691
5692         set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5693         /* Reconfiguration needed quiesce all VSIs */
5694         i40e_pf_quiesce_all_vsi(pf);
5695
5696         /* Changes in configuration update VEB/VSI */
5697         i40e_dcb_reconfigure(pf);
5698
5699         ret = i40e_resume_port_tx(pf);
5700
5701         clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5702         /* In case of error no point in resuming VSIs */
5703         if (ret)
5704                 goto exit;
5705
5706         /* Wait for the PF's queues to be disabled */
5707         ret = i40e_pf_wait_queues_disabled(pf);
5708         if (ret) {
5709                 /* Schedule PF reset to recover */
5710                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5711                 i40e_service_event_schedule(pf);
5712         } else {
5713                 i40e_pf_unquiesce_all_vsi(pf);
5714         }
5715
5716 exit:
5717         return ret;
5718 }
5719 #endif /* CONFIG_I40E_DCB */
5720
5721 /**
5722  * i40e_do_reset_safe - Protected reset path for userland calls.
5723  * @pf: board private structure
5724  * @reset_flags: which reset is requested
5725  *
5726  **/
5727 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5728 {
5729         rtnl_lock();
5730         i40e_do_reset(pf, reset_flags);
5731         rtnl_unlock();
5732 }
5733
5734 /**
5735  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5736  * @pf: board private structure
5737  * @e: event info posted on ARQ
5738  *
5739  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5740  * and VF queues
5741  **/
5742 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5743                                            struct i40e_arq_event_info *e)
5744 {
5745         struct i40e_aqc_lan_overflow *data =
5746                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5747         u32 queue = le32_to_cpu(data->prtdcb_rupto);
5748         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5749         struct i40e_hw *hw = &pf->hw;
5750         struct i40e_vf *vf;
5751         u16 vf_id;
5752
5753         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5754                 queue, qtx_ctl);
5755
5756         /* Queue belongs to VF, find the VF and issue VF reset */
5757         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5758             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5759                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5760                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5761                 vf_id -= hw->func_caps.vf_base_id;
5762                 vf = &pf->vf[vf_id];
5763                 i40e_vc_notify_vf_reset(vf);
5764                 /* Allow VF to process pending reset notification */
5765                 msleep(20);
5766                 i40e_reset_vf(vf, false);
5767         }
5768 }
5769
5770 /**
5771  * i40e_service_event_complete - Finish up the service event
5772  * @pf: board private structure
5773  **/
5774 static void i40e_service_event_complete(struct i40e_pf *pf)
5775 {
5776         WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5777
5778         /* flush memory to make sure state is correct before next watchog */
5779         smp_mb__before_atomic();
5780         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5781 }
5782
5783 /**
5784  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5785  * @pf: board private structure
5786  **/
5787 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5788 {
5789         u32 val, fcnt_prog;
5790
5791         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5792         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5793         return fcnt_prog;
5794 }
5795
5796 /**
5797  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5798  * @pf: board private structure
5799  **/
5800 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5801 {
5802         u32 val, fcnt_prog;
5803
5804         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5805         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5806                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5807                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5808         return fcnt_prog;
5809 }
5810
5811 /**
5812  * i40e_get_global_fd_count - Get total FD filters programmed on device
5813  * @pf: board private structure
5814  **/
5815 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5816 {
5817         u32 val, fcnt_prog;
5818
5819         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5820         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5821                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5822                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5823         return fcnt_prog;
5824 }
5825
5826 /**
5827  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5828  * @pf: board private structure
5829  **/
5830 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5831 {
5832         struct i40e_fdir_filter *filter;
5833         u32 fcnt_prog, fcnt_avail;
5834         struct hlist_node *node;
5835
5836         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5837                 return;
5838
5839         /* Check if, FD SB or ATR was auto disabled and if there is enough room
5840          * to re-enable
5841          */
5842         fcnt_prog = i40e_get_global_fd_count(pf);
5843         fcnt_avail = pf->fdir_pf_filter_count;
5844         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5845             (pf->fd_add_err == 0) ||
5846             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5847                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5848                     (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5849                         pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5850                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5851                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5852                 }
5853         }
5854         /* Wait for some more space to be available to turn on ATR */
5855         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5856                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5857                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5858                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5859                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5860                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5861                 }
5862         }
5863
5864         /* if hw had a problem adding a filter, delete it */
5865         if (pf->fd_inv > 0) {
5866                 hlist_for_each_entry_safe(filter, node,
5867                                           &pf->fdir_filter_list, fdir_node) {
5868                         if (filter->fd_id == pf->fd_inv) {
5869                                 hlist_del(&filter->fdir_node);
5870                                 kfree(filter);
5871                                 pf->fdir_pf_active_filters--;
5872                         }
5873                 }
5874         }
5875 }
5876
5877 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5878 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5879 /**
5880  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5881  * @pf: board private structure
5882  **/
5883 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5884 {
5885         unsigned long min_flush_time;
5886         int flush_wait_retry = 50;
5887         bool disable_atr = false;
5888         int fd_room;
5889         int reg;
5890
5891         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5892                 return;
5893
5894         if (!time_after(jiffies, pf->fd_flush_timestamp +
5895                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5896                 return;
5897
5898         /* If the flush is happening too quick and we have mostly SB rules we
5899          * should not re-enable ATR for some time.
5900          */
5901         min_flush_time = pf->fd_flush_timestamp +
5902                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5903         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5904
5905         if (!(time_after(jiffies, min_flush_time)) &&
5906             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5907                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5908                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5909                 disable_atr = true;
5910         }
5911
5912         pf->fd_flush_timestamp = jiffies;
5913         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5914         /* flush all filters */
5915         wr32(&pf->hw, I40E_PFQF_CTL_1,
5916              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5917         i40e_flush(&pf->hw);
5918         pf->fd_flush_cnt++;
5919         pf->fd_add_err = 0;
5920         do {
5921                 /* Check FD flush status every 5-6msec */
5922                 usleep_range(5000, 6000);
5923                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5924                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5925                         break;
5926         } while (flush_wait_retry--);
5927         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5928                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5929         } else {
5930                 /* replay sideband filters */
5931                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5932                 if (!disable_atr)
5933                         pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5934                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5935                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5936                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5937         }
5938
5939 }
5940
5941 /**
5942  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5943  * @pf: board private structure
5944  **/
5945 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5946 {
5947         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5948 }
5949
5950 /* We can see up to 256 filter programming desc in transit if the filters are
5951  * being applied really fast; before we see the first
5952  * filter miss error on Rx queue 0. Accumulating enough error messages before
5953  * reacting will make sure we don't cause flush too often.
5954  */
5955 #define I40E_MAX_FD_PROGRAM_ERROR 256
5956
5957 /**
5958  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5959  * @pf: board private structure
5960  **/
5961 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5962 {
5963
5964         /* if interface is down do nothing */
5965         if (test_bit(__I40E_DOWN, &pf->state))
5966                 return;
5967
5968         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5969                 return;
5970
5971         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5972                 i40e_fdir_flush_and_replay(pf);
5973
5974         i40e_fdir_check_and_reenable(pf);
5975
5976 }
5977
5978 /**
5979  * i40e_vsi_link_event - notify VSI of a link event
5980  * @vsi: vsi to be notified
5981  * @link_up: link up or down
5982  **/
5983 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5984 {
5985         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5986                 return;
5987
5988         switch (vsi->type) {
5989         case I40E_VSI_MAIN:
5990 #ifdef I40E_FCOE
5991         case I40E_VSI_FCOE:
5992 #endif
5993                 if (!vsi->netdev || !vsi->netdev_registered)
5994                         break;
5995
5996                 if (link_up) {
5997                         netif_carrier_on(vsi->netdev);
5998                         netif_tx_wake_all_queues(vsi->netdev);
5999                 } else {
6000                         netif_carrier_off(vsi->netdev);
6001                         netif_tx_stop_all_queues(vsi->netdev);
6002                 }
6003                 break;
6004
6005         case I40E_VSI_SRIOV:
6006         case I40E_VSI_VMDQ2:
6007         case I40E_VSI_CTRL:
6008         case I40E_VSI_IWARP:
6009         case I40E_VSI_MIRROR:
6010         default:
6011                 /* there is no notification for other VSIs */
6012                 break;
6013         }
6014 }
6015
6016 /**
6017  * i40e_veb_link_event - notify elements on the veb of a link event
6018  * @veb: veb to be notified
6019  * @link_up: link up or down
6020  **/
6021 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6022 {
6023         struct i40e_pf *pf;
6024         int i;
6025
6026         if (!veb || !veb->pf)
6027                 return;
6028         pf = veb->pf;
6029
6030         /* depth first... */
6031         for (i = 0; i < I40E_MAX_VEB; i++)
6032                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6033                         i40e_veb_link_event(pf->veb[i], link_up);
6034
6035         /* ... now the local VSIs */
6036         for (i = 0; i < pf->num_alloc_vsi; i++)
6037                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6038                         i40e_vsi_link_event(pf->vsi[i], link_up);
6039 }
6040
6041 /**
6042  * i40e_link_event - Update netif_carrier status
6043  * @pf: board private structure
6044  **/
6045 static void i40e_link_event(struct i40e_pf *pf)
6046 {
6047         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6048         u8 new_link_speed, old_link_speed;
6049         i40e_status status;
6050         bool new_link, old_link;
6051
6052         /* save off old link status information */
6053         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6054
6055         /* set this to force the get_link_status call to refresh state */
6056         pf->hw.phy.get_link_info = true;
6057
6058         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6059
6060         status = i40e_get_link_status(&pf->hw, &new_link);
6061         if (status) {
6062                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6063                         status);
6064                 return;
6065         }
6066
6067         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6068         new_link_speed = pf->hw.phy.link_info.link_speed;
6069
6070         if (new_link == old_link &&
6071             new_link_speed == old_link_speed &&
6072             (test_bit(__I40E_DOWN, &vsi->state) ||
6073              new_link == netif_carrier_ok(vsi->netdev)))
6074                 return;
6075
6076         if (!test_bit(__I40E_DOWN, &vsi->state))
6077                 i40e_print_link_message(vsi, new_link);
6078
6079         /* Notify the base of the switch tree connected to
6080          * the link.  Floating VEBs are not notified.
6081          */
6082         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6083                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6084         else
6085                 i40e_vsi_link_event(vsi, new_link);
6086
6087         if (pf->vf)
6088                 i40e_vc_notify_link_state(pf);
6089
6090         if (pf->flags & I40E_FLAG_PTP)
6091                 i40e_ptp_set_increment(pf);
6092 }
6093
6094 /**
6095  * i40e_watchdog_subtask - periodic checks not using event driven response
6096  * @pf: board private structure
6097  **/
6098 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6099 {
6100         int i;
6101
6102         /* if interface is down do nothing */
6103         if (test_bit(__I40E_DOWN, &pf->state) ||
6104             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6105                 return;
6106
6107         /* make sure we don't do these things too often */
6108         if (time_before(jiffies, (pf->service_timer_previous +
6109                                   pf->service_timer_period)))
6110                 return;
6111         pf->service_timer_previous = jiffies;
6112
6113         if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6114                 i40e_link_event(pf);
6115
6116         /* Update the stats for active netdevs so the network stack
6117          * can look at updated numbers whenever it cares to
6118          */
6119         for (i = 0; i < pf->num_alloc_vsi; i++)
6120                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6121                         i40e_update_stats(pf->vsi[i]);
6122
6123         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6124                 /* Update the stats for the active switching components */
6125                 for (i = 0; i < I40E_MAX_VEB; i++)
6126                         if (pf->veb[i])
6127                                 i40e_update_veb_stats(pf->veb[i]);
6128         }
6129
6130         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6131 }
6132
6133 /**
6134  * i40e_reset_subtask - Set up for resetting the device and driver
6135  * @pf: board private structure
6136  **/
6137 static void i40e_reset_subtask(struct i40e_pf *pf)
6138 {
6139         u32 reset_flags = 0;
6140
6141         rtnl_lock();
6142         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6143                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6144                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6145         }
6146         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6147                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6148                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6149         }
6150         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6151                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6152                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6153         }
6154         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6155                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6156                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6157         }
6158         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6159                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6160                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6161         }
6162
6163         /* If there's a recovery already waiting, it takes
6164          * precedence before starting a new reset sequence.
6165          */
6166         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6167                 i40e_handle_reset_warning(pf);
6168                 goto unlock;
6169         }
6170
6171         /* If we're already down or resetting, just bail */
6172         if (reset_flags &&
6173             !test_bit(__I40E_DOWN, &pf->state) &&
6174             !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6175                 i40e_do_reset(pf, reset_flags);
6176
6177 unlock:
6178         rtnl_unlock();
6179 }
6180
6181 /**
6182  * i40e_handle_link_event - Handle link event
6183  * @pf: board private structure
6184  * @e: event info posted on ARQ
6185  **/
6186 static void i40e_handle_link_event(struct i40e_pf *pf,
6187                                    struct i40e_arq_event_info *e)
6188 {
6189         struct i40e_aqc_get_link_status *status =
6190                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6191
6192         /* Do a new status request to re-enable LSE reporting
6193          * and load new status information into the hw struct
6194          * This completely ignores any state information
6195          * in the ARQ event info, instead choosing to always
6196          * issue the AQ update link status command.
6197          */
6198         i40e_link_event(pf);
6199
6200         /* check for unqualified module, if link is down */
6201         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6202             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6203             (!(status->link_info & I40E_AQ_LINK_UP)))
6204                 dev_err(&pf->pdev->dev,
6205                         "The driver failed to link because an unqualified module was detected.\n");
6206 }
6207
6208 /**
6209  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6210  * @pf: board private structure
6211  **/
6212 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6213 {
6214         struct i40e_arq_event_info event;
6215         struct i40e_hw *hw = &pf->hw;
6216         u16 pending, i = 0;
6217         i40e_status ret;
6218         u16 opcode;
6219         u32 oldval;
6220         u32 val;
6221
6222         /* Do not run clean AQ when PF reset fails */
6223         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6224                 return;
6225
6226         /* check for error indications */
6227         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6228         oldval = val;
6229         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6230                 if (hw->debug_mask & I40E_DEBUG_AQ)
6231                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6232                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6233         }
6234         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6235                 if (hw->debug_mask & I40E_DEBUG_AQ)
6236                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6237                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6238                 pf->arq_overflows++;
6239         }
6240         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6241                 if (hw->debug_mask & I40E_DEBUG_AQ)
6242                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6243                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6244         }
6245         if (oldval != val)
6246                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6247
6248         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6249         oldval = val;
6250         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6251                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6252                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6253                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6254         }
6255         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6256                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6257                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6258                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6259         }
6260         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6261                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6262                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6263                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6264         }
6265         if (oldval != val)
6266                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6267
6268         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6269         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6270         if (!event.msg_buf)
6271                 return;
6272
6273         do {
6274                 ret = i40e_clean_arq_element(hw, &event, &pending);
6275                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6276                         break;
6277                 else if (ret) {
6278                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6279                         break;
6280                 }
6281
6282                 opcode = le16_to_cpu(event.desc.opcode);
6283                 switch (opcode) {
6284
6285                 case i40e_aqc_opc_get_link_status:
6286                         i40e_handle_link_event(pf, &event);
6287                         break;
6288                 case i40e_aqc_opc_send_msg_to_pf:
6289                         ret = i40e_vc_process_vf_msg(pf,
6290                                         le16_to_cpu(event.desc.retval),
6291                                         le32_to_cpu(event.desc.cookie_high),
6292                                         le32_to_cpu(event.desc.cookie_low),
6293                                         event.msg_buf,
6294                                         event.msg_len);
6295                         break;
6296                 case i40e_aqc_opc_lldp_update_mib:
6297                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6298 #ifdef CONFIG_I40E_DCB
6299                         rtnl_lock();
6300                         ret = i40e_handle_lldp_event(pf, &event);
6301                         rtnl_unlock();
6302 #endif /* CONFIG_I40E_DCB */
6303                         break;
6304                 case i40e_aqc_opc_event_lan_overflow:
6305                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6306                         i40e_handle_lan_overflow_event(pf, &event);
6307                         break;
6308                 case i40e_aqc_opc_send_msg_to_peer:
6309                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6310                         break;
6311                 case i40e_aqc_opc_nvm_erase:
6312                 case i40e_aqc_opc_nvm_update:
6313                 case i40e_aqc_opc_oem_post_update:
6314                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6315                                    "ARQ NVM operation 0x%04x completed\n",
6316                                    opcode);
6317                         break;
6318                 default:
6319                         dev_info(&pf->pdev->dev,
6320                                  "ARQ: Unknown event 0x%04x ignored\n",
6321                                  opcode);
6322                         break;
6323                 }
6324         } while (pending && (i++ < pf->adminq_work_limit));
6325
6326         clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6327         /* re-enable Admin queue interrupt cause */
6328         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6329         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6330         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6331         i40e_flush(hw);
6332
6333         kfree(event.msg_buf);
6334 }
6335
6336 /**
6337  * i40e_verify_eeprom - make sure eeprom is good to use
6338  * @pf: board private structure
6339  **/
6340 static void i40e_verify_eeprom(struct i40e_pf *pf)
6341 {
6342         int err;
6343
6344         err = i40e_diag_eeprom_test(&pf->hw);
6345         if (err) {
6346                 /* retry in case of garbage read */
6347                 err = i40e_diag_eeprom_test(&pf->hw);
6348                 if (err) {
6349                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6350                                  err);
6351                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6352                 }
6353         }
6354
6355         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6356                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6357                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6358         }
6359 }
6360
6361 /**
6362  * i40e_enable_pf_switch_lb
6363  * @pf: pointer to the PF structure
6364  *
6365  * enable switch loop back or die - no point in a return value
6366  **/
6367 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6368 {
6369         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6370         struct i40e_vsi_context ctxt;
6371         int ret;
6372
6373         ctxt.seid = pf->main_vsi_seid;
6374         ctxt.pf_num = pf->hw.pf_id;
6375         ctxt.vf_num = 0;
6376         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6377         if (ret) {
6378                 dev_info(&pf->pdev->dev,
6379                          "couldn't get PF vsi config, err %s aq_err %s\n",
6380                          i40e_stat_str(&pf->hw, ret),
6381                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6382                 return;
6383         }
6384         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6385         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6386         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6387
6388         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6389         if (ret) {
6390                 dev_info(&pf->pdev->dev,
6391                          "update vsi switch failed, err %s aq_err %s\n",
6392                          i40e_stat_str(&pf->hw, ret),
6393                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6394         }
6395 }
6396
6397 /**
6398  * i40e_disable_pf_switch_lb
6399  * @pf: pointer to the PF structure
6400  *
6401  * disable switch loop back or die - no point in a return value
6402  **/
6403 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6404 {
6405         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6406         struct i40e_vsi_context ctxt;
6407         int ret;
6408
6409         ctxt.seid = pf->main_vsi_seid;
6410         ctxt.pf_num = pf->hw.pf_id;
6411         ctxt.vf_num = 0;
6412         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6413         if (ret) {
6414                 dev_info(&pf->pdev->dev,
6415                          "couldn't get PF vsi config, err %s aq_err %s\n",
6416                          i40e_stat_str(&pf->hw, ret),
6417                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6418                 return;
6419         }
6420         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6421         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6422         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6423
6424         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6425         if (ret) {
6426                 dev_info(&pf->pdev->dev,
6427                          "update vsi switch failed, err %s aq_err %s\n",
6428                          i40e_stat_str(&pf->hw, ret),
6429                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6430         }
6431 }
6432
6433 /**
6434  * i40e_config_bridge_mode - Configure the HW bridge mode
6435  * @veb: pointer to the bridge instance
6436  *
6437  * Configure the loop back mode for the LAN VSI that is downlink to the
6438  * specified HW bridge instance. It is expected this function is called
6439  * when a new HW bridge is instantiated.
6440  **/
6441 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6442 {
6443         struct i40e_pf *pf = veb->pf;
6444
6445         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6446                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6447                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6448         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6449                 i40e_disable_pf_switch_lb(pf);
6450         else
6451                 i40e_enable_pf_switch_lb(pf);
6452 }
6453
6454 /**
6455  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6456  * @veb: pointer to the VEB instance
6457  *
6458  * This is a recursive function that first builds the attached VSIs then
6459  * recurses in to build the next layer of VEB.  We track the connections
6460  * through our own index numbers because the seid's from the HW could
6461  * change across the reset.
6462  **/
6463 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6464 {
6465         struct i40e_vsi *ctl_vsi = NULL;
6466         struct i40e_pf *pf = veb->pf;
6467         int v, veb_idx;
6468         int ret;
6469
6470         /* build VSI that owns this VEB, temporarily attached to base VEB */
6471         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6472                 if (pf->vsi[v] &&
6473                     pf->vsi[v]->veb_idx == veb->idx &&
6474                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6475                         ctl_vsi = pf->vsi[v];
6476                         break;
6477                 }
6478         }
6479         if (!ctl_vsi) {
6480                 dev_info(&pf->pdev->dev,
6481                          "missing owner VSI for veb_idx %d\n", veb->idx);
6482                 ret = -ENOENT;
6483                 goto end_reconstitute;
6484         }
6485         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6486                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6487         ret = i40e_add_vsi(ctl_vsi);
6488         if (ret) {
6489                 dev_info(&pf->pdev->dev,
6490                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6491                          veb->idx, ret);
6492                 goto end_reconstitute;
6493         }
6494         i40e_vsi_reset_stats(ctl_vsi);
6495
6496         /* create the VEB in the switch and move the VSI onto the VEB */
6497         ret = i40e_add_veb(veb, ctl_vsi);
6498         if (ret)
6499                 goto end_reconstitute;
6500
6501         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6502                 veb->bridge_mode = BRIDGE_MODE_VEB;
6503         else
6504                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6505         i40e_config_bridge_mode(veb);
6506
6507         /* create the remaining VSIs attached to this VEB */
6508         for (v = 0; v < pf->num_alloc_vsi; v++) {
6509                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6510                         continue;
6511
6512                 if (pf->vsi[v]->veb_idx == veb->idx) {
6513                         struct i40e_vsi *vsi = pf->vsi[v];
6514
6515                         vsi->uplink_seid = veb->seid;
6516                         ret = i40e_add_vsi(vsi);
6517                         if (ret) {
6518                                 dev_info(&pf->pdev->dev,
6519                                          "rebuild of vsi_idx %d failed: %d\n",
6520                                          v, ret);
6521                                 goto end_reconstitute;
6522                         }
6523                         i40e_vsi_reset_stats(vsi);
6524                 }
6525         }
6526
6527         /* create any VEBs attached to this VEB - RECURSION */
6528         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6529                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6530                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6531                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6532                         if (ret)
6533                                 break;
6534                 }
6535         }
6536
6537 end_reconstitute:
6538         return ret;
6539 }
6540
6541 /**
6542  * i40e_get_capabilities - get info about the HW
6543  * @pf: the PF struct
6544  **/
6545 static int i40e_get_capabilities(struct i40e_pf *pf)
6546 {
6547         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6548         u16 data_size;
6549         int buf_len;
6550         int err;
6551
6552         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6553         do {
6554                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6555                 if (!cap_buf)
6556                         return -ENOMEM;
6557
6558                 /* this loads the data into the hw struct for us */
6559                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6560                                             &data_size,
6561                                             i40e_aqc_opc_list_func_capabilities,
6562                                             NULL);
6563                 /* data loaded, buffer no longer needed */
6564                 kfree(cap_buf);
6565
6566                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6567                         /* retry with a larger buffer */
6568                         buf_len = data_size;
6569                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6570                         dev_info(&pf->pdev->dev,
6571                                  "capability discovery failed, err %s aq_err %s\n",
6572                                  i40e_stat_str(&pf->hw, err),
6573                                  i40e_aq_str(&pf->hw,
6574                                              pf->hw.aq.asq_last_status));
6575                         return -ENODEV;
6576                 }
6577         } while (err);
6578
6579         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6580                 dev_info(&pf->pdev->dev,
6581                          "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",
6582                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6583                          pf->hw.func_caps.num_msix_vectors,
6584                          pf->hw.func_caps.num_msix_vectors_vf,
6585                          pf->hw.func_caps.fd_filters_guaranteed,
6586                          pf->hw.func_caps.fd_filters_best_effort,
6587                          pf->hw.func_caps.num_tx_qp,
6588                          pf->hw.func_caps.num_vsis);
6589
6590 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6591                        + pf->hw.func_caps.num_vfs)
6592         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6593                 dev_info(&pf->pdev->dev,
6594                          "got num_vsis %d, setting num_vsis to %d\n",
6595                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6596                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6597         }
6598
6599         return 0;
6600 }
6601
6602 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6603
6604 /**
6605  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6606  * @pf: board private structure
6607  **/
6608 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6609 {
6610         struct i40e_vsi *vsi;
6611         int i;
6612
6613         /* quick workaround for an NVM issue that leaves a critical register
6614          * uninitialized
6615          */
6616         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6617                 static const u32 hkey[] = {
6618                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6619                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6620                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6621                         0x95b3a76d};
6622
6623                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6624                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6625         }
6626
6627         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6628                 return;
6629
6630         /* find existing VSI and see if it needs configuring */
6631         vsi = NULL;
6632         for (i = 0; i < pf->num_alloc_vsi; i++) {
6633                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6634                         vsi = pf->vsi[i];
6635                         break;
6636                 }
6637         }
6638
6639         /* create a new VSI if none exists */
6640         if (!vsi) {
6641                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6642                                      pf->vsi[pf->lan_vsi]->seid, 0);
6643                 if (!vsi) {
6644                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6645                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6646                         return;
6647                 }
6648         }
6649
6650         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6651 }
6652
6653 /**
6654  * i40e_fdir_teardown - release the Flow Director resources
6655  * @pf: board private structure
6656  **/
6657 static void i40e_fdir_teardown(struct i40e_pf *pf)
6658 {
6659         int i;
6660
6661         i40e_fdir_filter_exit(pf);
6662         for (i = 0; i < pf->num_alloc_vsi; i++) {
6663                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6664                         i40e_vsi_release(pf->vsi[i]);
6665                         break;
6666                 }
6667         }
6668 }
6669
6670 /**
6671  * i40e_prep_for_reset - prep for the core to reset
6672  * @pf: board private structure
6673  *
6674  * Close up the VFs and other things in prep for PF Reset.
6675   **/
6676 static void i40e_prep_for_reset(struct i40e_pf *pf)
6677 {
6678         struct i40e_hw *hw = &pf->hw;
6679         i40e_status ret = 0;
6680         u32 v;
6681
6682         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6683         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6684                 return;
6685         if (i40e_check_asq_alive(&pf->hw))
6686                 i40e_vc_notify_reset(pf);
6687
6688         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6689
6690         /* quiesce the VSIs and their queues that are not already DOWN */
6691         i40e_pf_quiesce_all_vsi(pf);
6692
6693         for (v = 0; v < pf->num_alloc_vsi; v++) {
6694                 if (pf->vsi[v])
6695                         pf->vsi[v]->seid = 0;
6696         }
6697
6698         i40e_shutdown_adminq(&pf->hw);
6699
6700         /* call shutdown HMC */
6701         if (hw->hmc.hmc_obj) {
6702                 ret = i40e_shutdown_lan_hmc(hw);
6703                 if (ret)
6704                         dev_warn(&pf->pdev->dev,
6705                                  "shutdown_lan_hmc failed: %d\n", ret);
6706         }
6707 }
6708
6709 /**
6710  * i40e_send_version - update firmware with driver version
6711  * @pf: PF struct
6712  */
6713 static void i40e_send_version(struct i40e_pf *pf)
6714 {
6715         struct i40e_driver_version dv;
6716
6717         dv.major_version = DRV_VERSION_MAJOR;
6718         dv.minor_version = DRV_VERSION_MINOR;
6719         dv.build_version = DRV_VERSION_BUILD;
6720         dv.subbuild_version = 0;
6721         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6722         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6723 }
6724
6725 /**
6726  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6727  * @pf: board private structure
6728  * @reinit: if the Main VSI needs to re-initialized.
6729  **/
6730 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6731 {
6732         struct i40e_hw *hw = &pf->hw;
6733         u8 set_fc_aq_fail = 0;
6734         i40e_status ret;
6735         u32 val;
6736         u32 v;
6737
6738         /* Now we wait for GRST to settle out.
6739          * We don't have to delete the VEBs or VSIs from the hw switch
6740          * because the reset will make them disappear.
6741          */
6742         ret = i40e_pf_reset(hw);
6743         if (ret) {
6744                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6745                 set_bit(__I40E_RESET_FAILED, &pf->state);
6746                 goto clear_recovery;
6747         }
6748         pf->pfr_count++;
6749
6750         if (test_bit(__I40E_DOWN, &pf->state))
6751                 goto clear_recovery;
6752         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6753
6754         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6755         ret = i40e_init_adminq(&pf->hw);
6756         if (ret) {
6757                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6758                          i40e_stat_str(&pf->hw, ret),
6759                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6760                 goto clear_recovery;
6761         }
6762
6763         /* re-verify the eeprom if we just had an EMP reset */
6764         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6765                 i40e_verify_eeprom(pf);
6766
6767         i40e_clear_pxe_mode(hw);
6768         ret = i40e_get_capabilities(pf);
6769         if (ret)
6770                 goto end_core_reset;
6771
6772         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6773                                 hw->func_caps.num_rx_qp,
6774                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6775         if (ret) {
6776                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6777                 goto end_core_reset;
6778         }
6779         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6780         if (ret) {
6781                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6782                 goto end_core_reset;
6783         }
6784
6785 #ifdef CONFIG_I40E_DCB
6786         ret = i40e_init_pf_dcb(pf);
6787         if (ret) {
6788                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6789                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6790                 /* Continue without DCB enabled */
6791         }
6792 #endif /* CONFIG_I40E_DCB */
6793 #ifdef I40E_FCOE
6794         i40e_init_pf_fcoe(pf);
6795
6796 #endif
6797         /* do basic switch setup */
6798         ret = i40e_setup_pf_switch(pf, reinit);
6799         if (ret)
6800                 goto end_core_reset;
6801
6802         /* The driver only wants link up/down and module qualification
6803          * reports from firmware.  Note the negative logic.
6804          */
6805         ret = i40e_aq_set_phy_int_mask(&pf->hw,
6806                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
6807                                          I40E_AQ_EVENT_MEDIA_NA |
6808                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6809         if (ret)
6810                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6811                          i40e_stat_str(&pf->hw, ret),
6812                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6813
6814         /* make sure our flow control settings are restored */
6815         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6816         if (ret)
6817                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6818                         i40e_stat_str(&pf->hw, ret),
6819                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6820
6821         /* Rebuild the VSIs and VEBs that existed before reset.
6822          * They are still in our local switch element arrays, so only
6823          * need to rebuild the switch model in the HW.
6824          *
6825          * If there were VEBs but the reconstitution failed, we'll try
6826          * try to recover minimal use by getting the basic PF VSI working.
6827          */
6828         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6829                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6830                 /* find the one VEB connected to the MAC, and find orphans */
6831                 for (v = 0; v < I40E_MAX_VEB; v++) {
6832                         if (!pf->veb[v])
6833                                 continue;
6834
6835                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6836                             pf->veb[v]->uplink_seid == 0) {
6837                                 ret = i40e_reconstitute_veb(pf->veb[v]);
6838
6839                                 if (!ret)
6840                                         continue;
6841
6842                                 /* If Main VEB failed, we're in deep doodoo,
6843                                  * so give up rebuilding the switch and set up
6844                                  * for minimal rebuild of PF VSI.
6845                                  * If orphan failed, we'll report the error
6846                                  * but try to keep going.
6847                                  */
6848                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6849                                         dev_info(&pf->pdev->dev,
6850                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6851                                                  ret);
6852                                         pf->vsi[pf->lan_vsi]->uplink_seid
6853                                                                 = pf->mac_seid;
6854                                         break;
6855                                 } else if (pf->veb[v]->uplink_seid == 0) {
6856                                         dev_info(&pf->pdev->dev,
6857                                                  "rebuild of orphan VEB failed: %d\n",
6858                                                  ret);
6859                                 }
6860                         }
6861                 }
6862         }
6863
6864         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6865                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6866                 /* no VEB, so rebuild only the Main VSI */
6867                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6868                 if (ret) {
6869                         dev_info(&pf->pdev->dev,
6870                                  "rebuild of Main VSI failed: %d\n", ret);
6871                         goto end_core_reset;
6872                 }
6873         }
6874
6875         /* Reconfigure hardware for allowing smaller MSS in the case
6876          * of TSO, so that we avoid the MDD being fired and causing
6877          * a reset in the case of small MSS+TSO.
6878          */
6879 #define I40E_REG_MSS          0x000E64DC
6880 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6881 #define I40E_64BYTE_MSS       0x400000
6882         val = rd32(hw, I40E_REG_MSS);
6883         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6884                 val &= ~I40E_REG_MSS_MIN_MASK;
6885                 val |= I40E_64BYTE_MSS;
6886                 wr32(hw, I40E_REG_MSS, val);
6887         }
6888
6889         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6890                 msleep(75);
6891                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6892                 if (ret)
6893                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6894                                  i40e_stat_str(&pf->hw, ret),
6895                                  i40e_aq_str(&pf->hw,
6896                                              pf->hw.aq.asq_last_status));
6897         }
6898         /* reinit the misc interrupt */
6899         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6900                 ret = i40e_setup_misc_vector(pf);
6901
6902         /* Add a filter to drop all Flow control frames from any VSI from being
6903          * transmitted. By doing so we stop a malicious VF from sending out
6904          * PAUSE or PFC frames and potentially controlling traffic for other
6905          * PF/VF VSIs.
6906          * The FW can still send Flow control frames if enabled.
6907          */
6908         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6909                                                        pf->main_vsi_seid);
6910
6911         /* restart the VSIs that were rebuilt and running before the reset */
6912         i40e_pf_unquiesce_all_vsi(pf);
6913
6914         if (pf->num_alloc_vfs) {
6915                 for (v = 0; v < pf->num_alloc_vfs; v++)
6916                         i40e_reset_vf(&pf->vf[v], true);
6917         }
6918
6919         /* tell the firmware that we're starting */
6920         i40e_send_version(pf);
6921
6922 end_core_reset:
6923         clear_bit(__I40E_RESET_FAILED, &pf->state);
6924 clear_recovery:
6925         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6926 }
6927
6928 /**
6929  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6930  * @pf: board private structure
6931  *
6932  * Close up the VFs and other things in prep for a Core Reset,
6933  * then get ready to rebuild the world.
6934  **/
6935 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6936 {
6937         i40e_prep_for_reset(pf);
6938         i40e_reset_and_rebuild(pf, false);
6939 }
6940
6941 /**
6942  * i40e_handle_mdd_event
6943  * @pf: pointer to the PF structure
6944  *
6945  * Called from the MDD irq handler to identify possibly malicious vfs
6946  **/
6947 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6948 {
6949         struct i40e_hw *hw = &pf->hw;
6950         bool mdd_detected = false;
6951         bool pf_mdd_detected = false;
6952         struct i40e_vf *vf;
6953         u32 reg;
6954         int i;
6955
6956         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6957                 return;
6958
6959         /* find what triggered the MDD event */
6960         reg = rd32(hw, I40E_GL_MDET_TX);
6961         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6962                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6963                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
6964                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6965                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
6966                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6967                                 I40E_GL_MDET_TX_EVENT_SHIFT;
6968                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6969                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
6970                                 pf->hw.func_caps.base_queue;
6971                 if (netif_msg_tx_err(pf))
6972                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6973                                  event, queue, pf_num, vf_num);
6974                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6975                 mdd_detected = true;
6976         }
6977         reg = rd32(hw, I40E_GL_MDET_RX);
6978         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6979                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6980                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
6981                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6982                                 I40E_GL_MDET_RX_EVENT_SHIFT;
6983                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6984                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
6985                                 pf->hw.func_caps.base_queue;
6986                 if (netif_msg_rx_err(pf))
6987                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6988                                  event, queue, func);
6989                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6990                 mdd_detected = true;
6991         }
6992
6993         if (mdd_detected) {
6994                 reg = rd32(hw, I40E_PF_MDET_TX);
6995                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
6996                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
6997                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
6998                         pf_mdd_detected = true;
6999                 }
7000                 reg = rd32(hw, I40E_PF_MDET_RX);
7001                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7002                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7003                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7004                         pf_mdd_detected = true;
7005                 }
7006                 /* Queue belongs to the PF, initiate a reset */
7007                 if (pf_mdd_detected) {
7008                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7009                         i40e_service_event_schedule(pf);
7010                 }
7011         }
7012
7013         /* see if one of the VFs needs its hand slapped */
7014         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7015                 vf = &(pf->vf[i]);
7016                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7017                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7018                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7019                         vf->num_mdd_events++;
7020                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7021                                  i);
7022                 }
7023
7024                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7025                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7026                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7027                         vf->num_mdd_events++;
7028                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7029                                  i);
7030                 }
7031
7032                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7033                         dev_info(&pf->pdev->dev,
7034                                  "Too many MDD events on VF %d, disabled\n", i);
7035                         dev_info(&pf->pdev->dev,
7036                                  "Use PF Control I/F to re-enable the VF\n");
7037                         set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7038                 }
7039         }
7040
7041         /* re-enable mdd interrupt cause */
7042         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7043         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7044         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7045         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7046         i40e_flush(hw);
7047 }
7048
7049 /**
7050  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7051  * @pf: board private structure
7052  **/
7053 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7054 {
7055 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7056         struct i40e_hw *hw = &pf->hw;
7057         i40e_status ret;
7058         __be16 port;
7059         int i;
7060
7061         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7062                 return;
7063
7064         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7065
7066         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7067                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7068                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7069                         port = pf->udp_ports[i].index;
7070                         if (port)
7071                                 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7072                                                      pf->udp_ports[i].type,
7073                                                      NULL, NULL);
7074                         else
7075                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7076
7077                         if (ret) {
7078                                 dev_dbg(&pf->pdev->dev,
7079                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
7080                                         pf->udp_ports[i].type ? "vxlan" : "geneve",
7081                                         port ? "add" : "delete",
7082                                         ntohs(port), i,
7083                                         i40e_stat_str(&pf->hw, ret),
7084                                         i40e_aq_str(&pf->hw,
7085                                                     pf->hw.aq.asq_last_status));
7086                                 pf->udp_ports[i].index = 0;
7087                         }
7088                 }
7089         }
7090 #endif
7091 }
7092
7093 /**
7094  * i40e_service_task - Run the driver's async subtasks
7095  * @work: pointer to work_struct containing our data
7096  **/
7097 static void i40e_service_task(struct work_struct *work)
7098 {
7099         struct i40e_pf *pf = container_of(work,
7100                                           struct i40e_pf,
7101                                           service_task);
7102         unsigned long start_time = jiffies;
7103
7104         /* don't bother with service tasks if a reset is in progress */
7105         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7106                 i40e_service_event_complete(pf);
7107                 return;
7108         }
7109
7110         i40e_detect_recover_hung(pf);
7111         i40e_sync_filters_subtask(pf);
7112         i40e_reset_subtask(pf);
7113         i40e_handle_mdd_event(pf);
7114         i40e_vc_process_vflr_event(pf);
7115         i40e_watchdog_subtask(pf);
7116         i40e_fdir_reinit_subtask(pf);
7117         i40e_client_subtask(pf);
7118         i40e_sync_filters_subtask(pf);
7119         i40e_sync_udp_filters_subtask(pf);
7120         i40e_clean_adminq_subtask(pf);
7121
7122         i40e_service_event_complete(pf);
7123
7124         /* If the tasks have taken longer than one timer cycle or there
7125          * is more work to be done, reschedule the service task now
7126          * rather than wait for the timer to tick again.
7127          */
7128         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7129             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7130             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7131             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7132                 i40e_service_event_schedule(pf);
7133 }
7134
7135 /**
7136  * i40e_service_timer - timer callback
7137  * @data: pointer to PF struct
7138  **/
7139 static void i40e_service_timer(unsigned long data)
7140 {
7141         struct i40e_pf *pf = (struct i40e_pf *)data;
7142
7143         mod_timer(&pf->service_timer,
7144                   round_jiffies(jiffies + pf->service_timer_period));
7145         i40e_service_event_schedule(pf);
7146 }
7147
7148 /**
7149  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7150  * @vsi: the VSI being configured
7151  **/
7152 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7153 {
7154         struct i40e_pf *pf = vsi->back;
7155
7156         switch (vsi->type) {
7157         case I40E_VSI_MAIN:
7158                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7159                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7160                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7161                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7162                         vsi->num_q_vectors = pf->num_lan_msix;
7163                 else
7164                         vsi->num_q_vectors = 1;
7165
7166                 break;
7167
7168         case I40E_VSI_FDIR:
7169                 vsi->alloc_queue_pairs = 1;
7170                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7171                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7172                 vsi->num_q_vectors = 1;
7173                 break;
7174
7175         case I40E_VSI_VMDQ2:
7176                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7177                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7178                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7179                 vsi->num_q_vectors = pf->num_vmdq_msix;
7180                 break;
7181
7182         case I40E_VSI_SRIOV:
7183                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7184                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7185                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7186                 break;
7187
7188 #ifdef I40E_FCOE
7189         case I40E_VSI_FCOE:
7190                 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7191                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7192                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7193                 vsi->num_q_vectors = pf->num_fcoe_msix;
7194                 break;
7195
7196 #endif /* I40E_FCOE */
7197         default:
7198                 WARN_ON(1);
7199                 return -ENODATA;
7200         }
7201
7202         return 0;
7203 }
7204
7205 /**
7206  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7207  * @type: VSI pointer
7208  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7209  *
7210  * On error: returns error code (negative)
7211  * On success: returns 0
7212  **/
7213 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7214 {
7215         int size;
7216         int ret = 0;
7217
7218         /* allocate memory for both Tx and Rx ring pointers */
7219         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7220         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7221         if (!vsi->tx_rings)
7222                 return -ENOMEM;
7223         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7224
7225         if (alloc_qvectors) {
7226                 /* allocate memory for q_vector pointers */
7227                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7228                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7229                 if (!vsi->q_vectors) {
7230                         ret = -ENOMEM;
7231                         goto err_vectors;
7232                 }
7233         }
7234         return ret;
7235
7236 err_vectors:
7237         kfree(vsi->tx_rings);
7238         return ret;
7239 }
7240
7241 /**
7242  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7243  * @pf: board private structure
7244  * @type: type of VSI
7245  *
7246  * On error: returns error code (negative)
7247  * On success: returns vsi index in PF (positive)
7248  **/
7249 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7250 {
7251         int ret = -ENODEV;
7252         struct i40e_vsi *vsi;
7253         int vsi_idx;
7254         int i;
7255
7256         /* Need to protect the allocation of the VSIs at the PF level */
7257         mutex_lock(&pf->switch_mutex);
7258
7259         /* VSI list may be fragmented if VSI creation/destruction has
7260          * been happening.  We can afford to do a quick scan to look
7261          * for any free VSIs in the list.
7262          *
7263          * find next empty vsi slot, looping back around if necessary
7264          */
7265         i = pf->next_vsi;
7266         while (i < pf->num_alloc_vsi && pf->vsi[i])
7267                 i++;
7268         if (i >= pf->num_alloc_vsi) {
7269                 i = 0;
7270                 while (i < pf->next_vsi && pf->vsi[i])
7271                         i++;
7272         }
7273
7274         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7275                 vsi_idx = i;             /* Found one! */
7276         } else {
7277                 ret = -ENODEV;
7278                 goto unlock_pf;  /* out of VSI slots! */
7279         }
7280         pf->next_vsi = ++i;
7281
7282         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7283         if (!vsi) {
7284                 ret = -ENOMEM;
7285                 goto unlock_pf;
7286         }
7287         vsi->type = type;
7288         vsi->back = pf;
7289         set_bit(__I40E_DOWN, &vsi->state);
7290         vsi->flags = 0;
7291         vsi->idx = vsi_idx;
7292         vsi->int_rate_limit = 0;
7293         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7294                                 pf->rss_table_size : 64;
7295         vsi->netdev_registered = false;
7296         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7297         INIT_LIST_HEAD(&vsi->mac_filter_list);
7298         vsi->irqs_ready = false;
7299
7300         ret = i40e_set_num_rings_in_vsi(vsi);
7301         if (ret)
7302                 goto err_rings;
7303
7304         ret = i40e_vsi_alloc_arrays(vsi, true);
7305         if (ret)
7306                 goto err_rings;
7307
7308         /* Setup default MSIX irq handler for VSI */
7309         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7310
7311         /* Initialize VSI lock */
7312         spin_lock_init(&vsi->mac_filter_list_lock);
7313         pf->vsi[vsi_idx] = vsi;
7314         ret = vsi_idx;
7315         goto unlock_pf;
7316
7317 err_rings:
7318         pf->next_vsi = i - 1;
7319         kfree(vsi);
7320 unlock_pf:
7321         mutex_unlock(&pf->switch_mutex);
7322         return ret;
7323 }
7324
7325 /**
7326  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7327  * @type: VSI pointer
7328  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7329  *
7330  * On error: returns error code (negative)
7331  * On success: returns 0
7332  **/
7333 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7334 {
7335         /* free the ring and vector containers */
7336         if (free_qvectors) {
7337                 kfree(vsi->q_vectors);
7338                 vsi->q_vectors = NULL;
7339         }
7340         kfree(vsi->tx_rings);
7341         vsi->tx_rings = NULL;
7342         vsi->rx_rings = NULL;
7343 }
7344
7345 /**
7346  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7347  * and lookup table
7348  * @vsi: Pointer to VSI structure
7349  */
7350 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7351 {
7352         if (!vsi)
7353                 return;
7354
7355         kfree(vsi->rss_hkey_user);
7356         vsi->rss_hkey_user = NULL;
7357
7358         kfree(vsi->rss_lut_user);
7359         vsi->rss_lut_user = NULL;
7360 }
7361
7362 /**
7363  * i40e_vsi_clear - Deallocate the VSI provided
7364  * @vsi: the VSI being un-configured
7365  **/
7366 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7367 {
7368         struct i40e_pf *pf;
7369
7370         if (!vsi)
7371                 return 0;
7372
7373         if (!vsi->back)
7374                 goto free_vsi;
7375         pf = vsi->back;
7376
7377         mutex_lock(&pf->switch_mutex);
7378         if (!pf->vsi[vsi->idx]) {
7379                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7380                         vsi->idx, vsi->idx, vsi, vsi->type);
7381                 goto unlock_vsi;
7382         }
7383
7384         if (pf->vsi[vsi->idx] != vsi) {
7385                 dev_err(&pf->pdev->dev,
7386                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7387                         pf->vsi[vsi->idx]->idx,
7388                         pf->vsi[vsi->idx],
7389                         pf->vsi[vsi->idx]->type,
7390                         vsi->idx, vsi, vsi->type);
7391                 goto unlock_vsi;
7392         }
7393
7394         /* updates the PF for this cleared vsi */
7395         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7396         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7397
7398         i40e_vsi_free_arrays(vsi, true);
7399         i40e_clear_rss_config_user(vsi);
7400
7401         pf->vsi[vsi->idx] = NULL;
7402         if (vsi->idx < pf->next_vsi)
7403                 pf->next_vsi = vsi->idx;
7404
7405 unlock_vsi:
7406         mutex_unlock(&pf->switch_mutex);
7407 free_vsi:
7408         kfree(vsi);
7409
7410         return 0;
7411 }
7412
7413 /**
7414  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7415  * @vsi: the VSI being cleaned
7416  **/
7417 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7418 {
7419         int i;
7420
7421         if (vsi->tx_rings && vsi->tx_rings[0]) {
7422                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7423                         kfree_rcu(vsi->tx_rings[i], rcu);
7424                         vsi->tx_rings[i] = NULL;
7425                         vsi->rx_rings[i] = NULL;
7426                 }
7427         }
7428 }
7429
7430 /**
7431  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7432  * @vsi: the VSI being configured
7433  **/
7434 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7435 {
7436         struct i40e_ring *tx_ring, *rx_ring;
7437         struct i40e_pf *pf = vsi->back;
7438         int i;
7439
7440         /* Set basic values in the rings to be used later during open() */
7441         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7442                 /* allocate space for both Tx and Rx in one shot */
7443                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7444                 if (!tx_ring)
7445                         goto err_out;
7446
7447                 tx_ring->queue_index = i;
7448                 tx_ring->reg_idx = vsi->base_queue + i;
7449                 tx_ring->ring_active = false;
7450                 tx_ring->vsi = vsi;
7451                 tx_ring->netdev = vsi->netdev;
7452                 tx_ring->dev = &pf->pdev->dev;
7453                 tx_ring->count = vsi->num_desc;
7454                 tx_ring->size = 0;
7455                 tx_ring->dcb_tc = 0;
7456                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7457                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7458                 tx_ring->tx_itr_setting = pf->tx_itr_default;
7459                 vsi->tx_rings[i] = tx_ring;
7460
7461                 rx_ring = &tx_ring[1];
7462                 rx_ring->queue_index = i;
7463                 rx_ring->reg_idx = vsi->base_queue + i;
7464                 rx_ring->ring_active = false;
7465                 rx_ring->vsi = vsi;
7466                 rx_ring->netdev = vsi->netdev;
7467                 rx_ring->dev = &pf->pdev->dev;
7468                 rx_ring->count = vsi->num_desc;
7469                 rx_ring->size = 0;
7470                 rx_ring->dcb_tc = 0;
7471                 rx_ring->rx_itr_setting = pf->rx_itr_default;
7472                 vsi->rx_rings[i] = rx_ring;
7473         }
7474
7475         return 0;
7476
7477 err_out:
7478         i40e_vsi_clear_rings(vsi);
7479         return -ENOMEM;
7480 }
7481
7482 /**
7483  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7484  * @pf: board private structure
7485  * @vectors: the number of MSI-X vectors to request
7486  *
7487  * Returns the number of vectors reserved, or error
7488  **/
7489 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7490 {
7491         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7492                                         I40E_MIN_MSIX, vectors);
7493         if (vectors < 0) {
7494                 dev_info(&pf->pdev->dev,
7495                          "MSI-X vector reservation failed: %d\n", vectors);
7496                 vectors = 0;
7497         }
7498
7499         return vectors;
7500 }
7501
7502 /**
7503  * i40e_init_msix - Setup the MSIX capability
7504  * @pf: board private structure
7505  *
7506  * Work with the OS to set up the MSIX vectors needed.
7507  *
7508  * Returns the number of vectors reserved or negative on failure
7509  **/
7510 static int i40e_init_msix(struct i40e_pf *pf)
7511 {
7512         struct i40e_hw *hw = &pf->hw;
7513         int vectors_left;
7514         int v_budget, i;
7515         int v_actual;
7516         int iwarp_requested = 0;
7517
7518         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7519                 return -ENODEV;
7520
7521         /* The number of vectors we'll request will be comprised of:
7522          *   - Add 1 for "other" cause for Admin Queue events, etc.
7523          *   - The number of LAN queue pairs
7524          *      - Queues being used for RSS.
7525          *              We don't need as many as max_rss_size vectors.
7526          *              use rss_size instead in the calculation since that
7527          *              is governed by number of cpus in the system.
7528          *      - assumes symmetric Tx/Rx pairing
7529          *   - The number of VMDq pairs
7530          *   - The CPU count within the NUMA node if iWARP is enabled
7531 #ifdef I40E_FCOE
7532          *   - The number of FCOE qps.
7533 #endif
7534          * Once we count this up, try the request.
7535          *
7536          * If we can't get what we want, we'll simplify to nearly nothing
7537          * and try again.  If that still fails, we punt.
7538          */
7539         vectors_left = hw->func_caps.num_msix_vectors;
7540         v_budget = 0;
7541
7542         /* reserve one vector for miscellaneous handler */
7543         if (vectors_left) {
7544                 v_budget++;
7545                 vectors_left--;
7546         }
7547
7548         /* reserve vectors for the main PF traffic queues */
7549         pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7550         vectors_left -= pf->num_lan_msix;
7551         v_budget += pf->num_lan_msix;
7552
7553         /* reserve one vector for sideband flow director */
7554         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7555                 if (vectors_left) {
7556                         v_budget++;
7557                         vectors_left--;
7558                 } else {
7559                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7560                 }
7561         }
7562
7563 #ifdef I40E_FCOE
7564         /* can we reserve enough for FCoE? */
7565         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7566                 if (!vectors_left)
7567                         pf->num_fcoe_msix = 0;
7568                 else if (vectors_left >= pf->num_fcoe_qps)
7569                         pf->num_fcoe_msix = pf->num_fcoe_qps;
7570                 else
7571                         pf->num_fcoe_msix = 1;
7572                 v_budget += pf->num_fcoe_msix;
7573                 vectors_left -= pf->num_fcoe_msix;
7574         }
7575
7576 #endif
7577         /* can we reserve enough for iWARP? */
7578         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7579                 if (!vectors_left)
7580                         pf->num_iwarp_msix = 0;
7581                 else if (vectors_left < pf->num_iwarp_msix)
7582                         pf->num_iwarp_msix = 1;
7583                 v_budget += pf->num_iwarp_msix;
7584                 vectors_left -= pf->num_iwarp_msix;
7585         }
7586
7587         /* any vectors left over go for VMDq support */
7588         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7589                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7590                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7591
7592                 /* if we're short on vectors for what's desired, we limit
7593                  * the queues per vmdq.  If this is still more than are
7594                  * available, the user will need to change the number of
7595                  * queues/vectors used by the PF later with the ethtool
7596                  * channels command
7597                  */
7598                 if (vmdq_vecs < vmdq_vecs_wanted)
7599                         pf->num_vmdq_qps = 1;
7600                 pf->num_vmdq_msix = pf->num_vmdq_qps;
7601
7602                 v_budget += vmdq_vecs;
7603                 vectors_left -= vmdq_vecs;
7604         }
7605
7606         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7607                                    GFP_KERNEL);
7608         if (!pf->msix_entries)
7609                 return -ENOMEM;
7610
7611         for (i = 0; i < v_budget; i++)
7612                 pf->msix_entries[i].entry = i;
7613         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7614
7615         if (v_actual != v_budget) {
7616                 /* If we have limited resources, we will start with no vectors
7617                  * for the special features and then allocate vectors to some
7618                  * of these features based on the policy and at the end disable
7619                  * the features that did not get any vectors.
7620                  */
7621                 iwarp_requested = pf->num_iwarp_msix;
7622                 pf->num_iwarp_msix = 0;
7623 #ifdef I40E_FCOE
7624                 pf->num_fcoe_qps = 0;
7625                 pf->num_fcoe_msix = 0;
7626 #endif
7627                 pf->num_vmdq_msix = 0;
7628         }
7629
7630         if (v_actual < I40E_MIN_MSIX) {
7631                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7632                 kfree(pf->msix_entries);
7633                 pf->msix_entries = NULL;
7634                 return -ENODEV;
7635
7636         } else if (v_actual == I40E_MIN_MSIX) {
7637                 /* Adjust for minimal MSIX use */
7638                 pf->num_vmdq_vsis = 0;
7639                 pf->num_vmdq_qps = 0;
7640                 pf->num_lan_qps = 1;
7641                 pf->num_lan_msix = 1;
7642
7643         } else if (v_actual != v_budget) {
7644                 int vec;
7645
7646                 /* reserve the misc vector */
7647                 vec = v_actual - 1;
7648
7649                 /* Scale vector usage down */
7650                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7651                 pf->num_vmdq_vsis = 1;
7652                 pf->num_vmdq_qps = 1;
7653                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7654
7655                 /* partition out the remaining vectors */
7656                 switch (vec) {
7657                 case 2:
7658                         pf->num_lan_msix = 1;
7659                         break;
7660                 case 3:
7661                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7662                                 pf->num_lan_msix = 1;
7663                                 pf->num_iwarp_msix = 1;
7664                         } else {
7665                                 pf->num_lan_msix = 2;
7666                         }
7667 #ifdef I40E_FCOE
7668                         /* give one vector to FCoE */
7669                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7670                                 pf->num_lan_msix = 1;
7671                                 pf->num_fcoe_msix = 1;
7672                         }
7673 #endif
7674                         break;
7675                 default:
7676                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7677                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7678                                                  iwarp_requested);
7679                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7680                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7681                         } else {
7682                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7683                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7684                         }
7685                         pf->num_lan_msix = min_t(int,
7686                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7687                                                               pf->num_lan_msix);
7688 #ifdef I40E_FCOE
7689                         /* give one vector to FCoE */
7690                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7691                                 pf->num_fcoe_msix = 1;
7692                                 vec--;
7693                         }
7694 #endif
7695                         break;
7696                 }
7697         }
7698
7699         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7700             (pf->num_vmdq_msix == 0)) {
7701                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7702                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7703         }
7704
7705         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7706             (pf->num_iwarp_msix == 0)) {
7707                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7708                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7709         }
7710 #ifdef I40E_FCOE
7711
7712         if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7713                 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7714                 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7715         }
7716 #endif
7717         return v_actual;
7718 }
7719
7720 /**
7721  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7722  * @vsi: the VSI being configured
7723  * @v_idx: index of the vector in the vsi struct
7724  * @cpu: cpu to be used on affinity_mask
7725  *
7726  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7727  **/
7728 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7729 {
7730         struct i40e_q_vector *q_vector;
7731
7732         /* allocate q_vector */
7733         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7734         if (!q_vector)
7735                 return -ENOMEM;
7736
7737         q_vector->vsi = vsi;
7738         q_vector->v_idx = v_idx;
7739         cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7740
7741         if (vsi->netdev)
7742                 netif_napi_add(vsi->netdev, &q_vector->napi,
7743                                i40e_napi_poll, NAPI_POLL_WEIGHT);
7744
7745         q_vector->rx.latency_range = I40E_LOW_LATENCY;
7746         q_vector->tx.latency_range = I40E_LOW_LATENCY;
7747
7748         /* tie q_vector and vsi together */
7749         vsi->q_vectors[v_idx] = q_vector;
7750
7751         return 0;
7752 }
7753
7754 /**
7755  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7756  * @vsi: the VSI being configured
7757  *
7758  * We allocate one q_vector per queue interrupt.  If allocation fails we
7759  * return -ENOMEM.
7760  **/
7761 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7762 {
7763         struct i40e_pf *pf = vsi->back;
7764         int err, v_idx, num_q_vectors, current_cpu;
7765
7766         /* if not MSIX, give the one vector only to the LAN VSI */
7767         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7768                 num_q_vectors = vsi->num_q_vectors;
7769         else if (vsi == pf->vsi[pf->lan_vsi])
7770                 num_q_vectors = 1;
7771         else
7772                 return -EINVAL;
7773
7774         current_cpu = cpumask_first(cpu_online_mask);
7775
7776         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7777                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7778                 if (err)
7779                         goto err_out;
7780                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7781                 if (unlikely(current_cpu >= nr_cpu_ids))
7782                         current_cpu = cpumask_first(cpu_online_mask);
7783         }
7784
7785         return 0;
7786
7787 err_out:
7788         while (v_idx--)
7789                 i40e_free_q_vector(vsi, v_idx);
7790
7791         return err;
7792 }
7793
7794 /**
7795  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7796  * @pf: board private structure to initialize
7797  **/
7798 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7799 {
7800         int vectors = 0;
7801         ssize_t size;
7802
7803         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7804                 vectors = i40e_init_msix(pf);
7805                 if (vectors < 0) {
7806                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
7807                                        I40E_FLAG_IWARP_ENABLED  |
7808 #ifdef I40E_FCOE
7809                                        I40E_FLAG_FCOE_ENABLED   |
7810 #endif
7811                                        I40E_FLAG_RSS_ENABLED    |
7812                                        I40E_FLAG_DCB_CAPABLE    |
7813                                        I40E_FLAG_SRIOV_ENABLED  |
7814                                        I40E_FLAG_FD_SB_ENABLED  |
7815                                        I40E_FLAG_FD_ATR_ENABLED |
7816                                        I40E_FLAG_VMDQ_ENABLED);
7817
7818                         /* rework the queue expectations without MSIX */
7819                         i40e_determine_queue_usage(pf);
7820                 }
7821         }
7822
7823         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7824             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7825                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7826                 vectors = pci_enable_msi(pf->pdev);
7827                 if (vectors < 0) {
7828                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7829                                  vectors);
7830                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7831                 }
7832                 vectors = 1;  /* one MSI or Legacy vector */
7833         }
7834
7835         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7836                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7837
7838         /* set up vector assignment tracking */
7839         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7840         pf->irq_pile = kzalloc(size, GFP_KERNEL);
7841         if (!pf->irq_pile) {
7842                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7843                 return -ENOMEM;
7844         }
7845         pf->irq_pile->num_entries = vectors;
7846         pf->irq_pile->search_hint = 0;
7847
7848         /* track first vector for misc interrupts, ignore return */
7849         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7850
7851         return 0;
7852 }
7853
7854 /**
7855  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7856  * @pf: board private structure
7857  *
7858  * This sets up the handler for MSIX 0, which is used to manage the
7859  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7860  * when in MSI or Legacy interrupt mode.
7861  **/
7862 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7863 {
7864         struct i40e_hw *hw = &pf->hw;
7865         int err = 0;
7866
7867         /* Only request the irq if this is the first time through, and
7868          * not when we're rebuilding after a Reset
7869          */
7870         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7871                 err = request_irq(pf->msix_entries[0].vector,
7872                                   i40e_intr, 0, pf->int_name, pf);
7873                 if (err) {
7874                         dev_info(&pf->pdev->dev,
7875                                  "request_irq for %s failed: %d\n",
7876                                  pf->int_name, err);
7877                         return -EFAULT;
7878                 }
7879         }
7880
7881         i40e_enable_misc_int_causes(pf);
7882
7883         /* associate no queues to the misc vector */
7884         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7885         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7886
7887         i40e_flush(hw);
7888
7889         i40e_irq_dynamic_enable_icr0(pf, true);
7890
7891         return err;
7892 }
7893
7894 /**
7895  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7896  * @vsi: vsi structure
7897  * @seed: RSS hash seed
7898  **/
7899 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7900                               u8 *lut, u16 lut_size)
7901 {
7902         struct i40e_aqc_get_set_rss_key_data rss_key;
7903         struct i40e_pf *pf = vsi->back;
7904         struct i40e_hw *hw = &pf->hw;
7905         bool pf_lut = false;
7906         u8 *rss_lut;
7907         int ret, i;
7908
7909         memset(&rss_key, 0, sizeof(rss_key));
7910         memcpy(&rss_key, seed, sizeof(rss_key));
7911
7912         rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7913         if (!rss_lut)
7914                 return -ENOMEM;
7915
7916         /* Populate the LUT with max no. of queues in round robin fashion */
7917         for (i = 0; i < vsi->rss_table_size; i++)
7918                 rss_lut[i] = i % vsi->rss_size;
7919
7920         ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7921         if (ret) {
7922                 dev_info(&pf->pdev->dev,
7923                          "Cannot set RSS key, err %s aq_err %s\n",
7924                          i40e_stat_str(&pf->hw, ret),
7925                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7926                 goto config_rss_aq_out;
7927         }
7928
7929         if (vsi->type == I40E_VSI_MAIN)
7930                 pf_lut = true;
7931
7932         ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7933                                   vsi->rss_table_size);
7934         if (ret)
7935                 dev_info(&pf->pdev->dev,
7936                          "Cannot set RSS lut, err %s aq_err %s\n",
7937                          i40e_stat_str(&pf->hw, ret),
7938                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7939
7940 config_rss_aq_out:
7941         kfree(rss_lut);
7942         return ret;
7943 }
7944
7945 /**
7946  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7947  * @vsi: VSI structure
7948  **/
7949 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7950 {
7951         u8 seed[I40E_HKEY_ARRAY_SIZE];
7952         struct i40e_pf *pf = vsi->back;
7953         u8 *lut;
7954         int ret;
7955
7956         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
7957                 return 0;
7958
7959         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
7960         if (!lut)
7961                 return -ENOMEM;
7962
7963         i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
7964         netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7965         vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
7966         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
7967         kfree(lut);
7968
7969         return ret;
7970 }
7971
7972 /**
7973  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
7974  * @vsi: Pointer to vsi structure
7975  * @seed: Buffter to store the hash keys
7976  * @lut: Buffer to store the lookup table entries
7977  * @lut_size: Size of buffer to store the lookup table entries
7978  *
7979  * Return 0 on success, negative on failure
7980  */
7981 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7982                            u8 *lut, u16 lut_size)
7983 {
7984         struct i40e_pf *pf = vsi->back;
7985         struct i40e_hw *hw = &pf->hw;
7986         int ret = 0;
7987
7988         if (seed) {
7989                 ret = i40e_aq_get_rss_key(hw, vsi->id,
7990                         (struct i40e_aqc_get_set_rss_key_data *)seed);
7991                 if (ret) {
7992                         dev_info(&pf->pdev->dev,
7993                                  "Cannot get RSS key, err %s aq_err %s\n",
7994                                  i40e_stat_str(&pf->hw, ret),
7995                                  i40e_aq_str(&pf->hw,
7996                                              pf->hw.aq.asq_last_status));
7997                         return ret;
7998                 }
7999         }
8000
8001         if (lut) {
8002                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8003
8004                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8005                 if (ret) {
8006                         dev_info(&pf->pdev->dev,
8007                                  "Cannot get RSS lut, err %s aq_err %s\n",
8008                                  i40e_stat_str(&pf->hw, ret),
8009                                  i40e_aq_str(&pf->hw,
8010                                              pf->hw.aq.asq_last_status));
8011                         return ret;
8012                 }
8013         }
8014
8015         return ret;
8016 }
8017
8018 /**
8019  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8020  * @vsi: Pointer to vsi structure
8021  * @seed: RSS hash seed
8022  * @lut: Lookup table
8023  * @lut_size: Lookup table size
8024  *
8025  * Returns 0 on success, negative on failure
8026  **/
8027 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8028                                const u8 *lut, u16 lut_size)
8029 {
8030         struct i40e_pf *pf = vsi->back;
8031         struct i40e_hw *hw = &pf->hw;
8032         u16 vf_id = vsi->vf_id;
8033         u8 i;
8034
8035         /* Fill out hash function seed */
8036         if (seed) {
8037                 u32 *seed_dw = (u32 *)seed;
8038
8039                 if (vsi->type == I40E_VSI_MAIN) {
8040                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8041                                 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8042                                                   seed_dw[i]);
8043                 } else if (vsi->type == I40E_VSI_SRIOV) {
8044                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8045                                 i40e_write_rx_ctl(hw,
8046                                                   I40E_VFQF_HKEY1(i, vf_id),
8047                                                   seed_dw[i]);
8048                 } else {
8049                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8050                 }
8051         }
8052
8053         if (lut) {
8054                 u32 *lut_dw = (u32 *)lut;
8055
8056                 if (vsi->type == I40E_VSI_MAIN) {
8057                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
8058                                 return -EINVAL;
8059                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8060                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8061                 } else if (vsi->type == I40E_VSI_SRIOV) {
8062                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8063                                 return -EINVAL;
8064                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8065                                 i40e_write_rx_ctl(hw,
8066                                                   I40E_VFQF_HLUT1(i, vf_id),
8067                                                   lut_dw[i]);
8068                 } else {
8069                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8070                 }
8071         }
8072         i40e_flush(hw);
8073
8074         return 0;
8075 }
8076
8077 /**
8078  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8079  * @vsi: Pointer to VSI structure
8080  * @seed: Buffer to store the keys
8081  * @lut: Buffer to store the lookup table entries
8082  * @lut_size: Size of buffer to store the lookup table entries
8083  *
8084  * Returns 0 on success, negative on failure
8085  */
8086 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8087                             u8 *lut, u16 lut_size)
8088 {
8089         struct i40e_pf *pf = vsi->back;
8090         struct i40e_hw *hw = &pf->hw;
8091         u16 i;
8092
8093         if (seed) {
8094                 u32 *seed_dw = (u32 *)seed;
8095
8096                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8097                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8098         }
8099         if (lut) {
8100                 u32 *lut_dw = (u32 *)lut;
8101
8102                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8103                         return -EINVAL;
8104                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8105                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8106         }
8107
8108         return 0;
8109 }
8110
8111 /**
8112  * i40e_config_rss - Configure RSS keys and lut
8113  * @vsi: Pointer to VSI structure
8114  * @seed: RSS hash seed
8115  * @lut: Lookup table
8116  * @lut_size: Lookup table size
8117  *
8118  * Returns 0 on success, negative on failure
8119  */
8120 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8121 {
8122         struct i40e_pf *pf = vsi->back;
8123
8124         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8125                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8126         else
8127                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8128 }
8129
8130 /**
8131  * i40e_get_rss - Get RSS keys and lut
8132  * @vsi: Pointer to VSI structure
8133  * @seed: Buffer to store the keys
8134  * @lut: Buffer to store the lookup table entries
8135  * lut_size: Size of buffer to store the lookup table entries
8136  *
8137  * Returns 0 on success, negative on failure
8138  */
8139 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8140 {
8141         struct i40e_pf *pf = vsi->back;
8142
8143         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8144                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8145         else
8146                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8147 }
8148
8149 /**
8150  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8151  * @pf: Pointer to board private structure
8152  * @lut: Lookup table
8153  * @rss_table_size: Lookup table size
8154  * @rss_size: Range of queue number for hashing
8155  */
8156 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8157                               u16 rss_table_size, u16 rss_size)
8158 {
8159         u16 i;
8160
8161         for (i = 0; i < rss_table_size; i++)
8162                 lut[i] = i % rss_size;
8163 }
8164
8165 /**
8166  * i40e_pf_config_rss - Prepare for RSS if used
8167  * @pf: board private structure
8168  **/
8169 static int i40e_pf_config_rss(struct i40e_pf *pf)
8170 {
8171         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8172         u8 seed[I40E_HKEY_ARRAY_SIZE];
8173         u8 *lut;
8174         struct i40e_hw *hw = &pf->hw;
8175         u32 reg_val;
8176         u64 hena;
8177         int ret;
8178
8179         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8180         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8181                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8182         hena |= i40e_pf_get_default_rss_hena(pf);
8183
8184         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8185         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8186
8187         /* Determine the RSS table size based on the hardware capabilities */
8188         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8189         reg_val = (pf->rss_table_size == 512) ?
8190                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8191                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8192         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8193
8194         /* Determine the RSS size of the VSI */
8195         if (!vsi->rss_size)
8196                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8197                                       vsi->num_queue_pairs);
8198
8199         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8200         if (!lut)
8201                 return -ENOMEM;
8202
8203         /* Use user configured lut if there is one, otherwise use default */
8204         if (vsi->rss_lut_user)
8205                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8206         else
8207                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8208
8209         /* Use user configured hash key if there is one, otherwise
8210          * use default.
8211          */
8212         if (vsi->rss_hkey_user)
8213                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8214         else
8215                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8216         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8217         kfree(lut);
8218
8219         return ret;
8220 }
8221
8222 /**
8223  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8224  * @pf: board private structure
8225  * @queue_count: the requested queue count for rss.
8226  *
8227  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8228  * count which may be different from the requested queue count.
8229  **/
8230 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8231 {
8232         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8233         int new_rss_size;
8234
8235         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8236                 return 0;
8237
8238         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8239
8240         if (queue_count != vsi->num_queue_pairs) {
8241                 vsi->req_queue_pairs = queue_count;
8242                 i40e_prep_for_reset(pf);
8243
8244                 pf->alloc_rss_size = new_rss_size;
8245
8246                 i40e_reset_and_rebuild(pf, true);
8247
8248                 /* Discard the user configured hash keys and lut, if less
8249                  * queues are enabled.
8250                  */
8251                 if (queue_count < vsi->rss_size) {
8252                         i40e_clear_rss_config_user(vsi);
8253                         dev_dbg(&pf->pdev->dev,
8254                                 "discard user configured hash keys and lut\n");
8255                 }
8256
8257                 /* Reset vsi->rss_size, as number of enabled queues changed */
8258                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8259                                       vsi->num_queue_pairs);
8260
8261                 i40e_pf_config_rss(pf);
8262         }
8263         dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8264                  pf->alloc_rss_size, pf->rss_size_max);
8265         return pf->alloc_rss_size;
8266 }
8267
8268 /**
8269  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8270  * @pf: board private structure
8271  **/
8272 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8273 {
8274         i40e_status status;
8275         bool min_valid, max_valid;
8276         u32 max_bw, min_bw;
8277
8278         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8279                                            &min_valid, &max_valid);
8280
8281         if (!status) {
8282                 if (min_valid)
8283                         pf->npar_min_bw = min_bw;
8284                 if (max_valid)
8285                         pf->npar_max_bw = max_bw;
8286         }
8287
8288         return status;
8289 }
8290
8291 /**
8292  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8293  * @pf: board private structure
8294  **/
8295 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8296 {
8297         struct i40e_aqc_configure_partition_bw_data bw_data;
8298         i40e_status status;
8299
8300         /* Set the valid bit for this PF */
8301         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8302         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8303         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8304
8305         /* Set the new bandwidths */
8306         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8307
8308         return status;
8309 }
8310
8311 /**
8312  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8313  * @pf: board private structure
8314  **/
8315 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8316 {
8317         /* Commit temporary BW setting to permanent NVM image */
8318         enum i40e_admin_queue_err last_aq_status;
8319         i40e_status ret;
8320         u16 nvm_word;
8321
8322         if (pf->hw.partition_id != 1) {
8323                 dev_info(&pf->pdev->dev,
8324                          "Commit BW only works on partition 1! This is partition %d",
8325                          pf->hw.partition_id);
8326                 ret = I40E_NOT_SUPPORTED;
8327                 goto bw_commit_out;
8328         }
8329
8330         /* Acquire NVM for read access */
8331         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8332         last_aq_status = pf->hw.aq.asq_last_status;
8333         if (ret) {
8334                 dev_info(&pf->pdev->dev,
8335                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8336                          i40e_stat_str(&pf->hw, ret),
8337                          i40e_aq_str(&pf->hw, last_aq_status));
8338                 goto bw_commit_out;
8339         }
8340
8341         /* Read word 0x10 of NVM - SW compatibility word 1 */
8342         ret = i40e_aq_read_nvm(&pf->hw,
8343                                I40E_SR_NVM_CONTROL_WORD,
8344                                0x10, sizeof(nvm_word), &nvm_word,
8345                                false, NULL);
8346         /* Save off last admin queue command status before releasing
8347          * the NVM
8348          */
8349         last_aq_status = pf->hw.aq.asq_last_status;
8350         i40e_release_nvm(&pf->hw);
8351         if (ret) {
8352                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8353                          i40e_stat_str(&pf->hw, ret),
8354                          i40e_aq_str(&pf->hw, last_aq_status));
8355                 goto bw_commit_out;
8356         }
8357
8358         /* Wait a bit for NVM release to complete */
8359         msleep(50);
8360
8361         /* Acquire NVM for write access */
8362         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8363         last_aq_status = pf->hw.aq.asq_last_status;
8364         if (ret) {
8365                 dev_info(&pf->pdev->dev,
8366                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8367                          i40e_stat_str(&pf->hw, ret),
8368                          i40e_aq_str(&pf->hw, last_aq_status));
8369                 goto bw_commit_out;
8370         }
8371         /* Write it back out unchanged to initiate update NVM,
8372          * which will force a write of the shadow (alt) RAM to
8373          * the NVM - thus storing the bandwidth values permanently.
8374          */
8375         ret = i40e_aq_update_nvm(&pf->hw,
8376                                  I40E_SR_NVM_CONTROL_WORD,
8377                                  0x10, sizeof(nvm_word),
8378                                  &nvm_word, true, NULL);
8379         /* Save off last admin queue command status before releasing
8380          * the NVM
8381          */
8382         last_aq_status = pf->hw.aq.asq_last_status;
8383         i40e_release_nvm(&pf->hw);
8384         if (ret)
8385                 dev_info(&pf->pdev->dev,
8386                          "BW settings NOT SAVED, err %s aq_err %s\n",
8387                          i40e_stat_str(&pf->hw, ret),
8388                          i40e_aq_str(&pf->hw, last_aq_status));
8389 bw_commit_out:
8390
8391         return ret;
8392 }
8393
8394 /**
8395  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8396  * @pf: board private structure to initialize
8397  *
8398  * i40e_sw_init initializes the Adapter private data structure.
8399  * Fields are initialized based on PCI device information and
8400  * OS network device settings (MTU size).
8401  **/
8402 static int i40e_sw_init(struct i40e_pf *pf)
8403 {
8404         int err = 0;
8405         int size;
8406
8407         pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8408                                 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8409         if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8410                 if (I40E_DEBUG_USER & debug)
8411                         pf->hw.debug_mask = debug;
8412                 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8413                                                 I40E_DEFAULT_MSG_ENABLE);
8414         }
8415
8416         /* Set default capability flags */
8417         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8418                     I40E_FLAG_MSI_ENABLED     |
8419                     I40E_FLAG_MSIX_ENABLED;
8420
8421         /* Set default ITR */
8422         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8423         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8424
8425         /* Depending on PF configurations, it is possible that the RSS
8426          * maximum might end up larger than the available queues
8427          */
8428         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8429         pf->alloc_rss_size = 1;
8430         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8431         pf->rss_size_max = min_t(int, pf->rss_size_max,
8432                                  pf->hw.func_caps.num_tx_qp);
8433         if (pf->hw.func_caps.rss) {
8434                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8435                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8436                                            num_online_cpus());
8437         }
8438
8439         /* MFP mode enabled */
8440         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8441                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8442                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8443                 if (i40e_get_npar_bw_setting(pf))
8444                         dev_warn(&pf->pdev->dev,
8445                                  "Could not get NPAR bw settings\n");
8446                 else
8447                         dev_info(&pf->pdev->dev,
8448                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8449                                  pf->npar_min_bw, pf->npar_max_bw);
8450         }
8451
8452         /* FW/NVM is not yet fixed in this regard */
8453         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8454             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8455                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8456                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8457                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8458                     pf->hw.num_partitions > 1)
8459                         dev_info(&pf->pdev->dev,
8460                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8461                 else
8462                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8463                 pf->fdir_pf_filter_count =
8464                                  pf->hw.func_caps.fd_filters_guaranteed;
8465                 pf->hw.fdir_shared_filter_count =
8466                                  pf->hw.func_caps.fd_filters_best_effort;
8467         }
8468
8469         if (i40e_is_mac_710(&pf->hw) &&
8470             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8471             (pf->hw.aq.fw_maj_ver < 4))) {
8472                 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8473                 /* No DCB support  for FW < v4.33 */
8474                 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8475         }
8476
8477         /* Disable FW LLDP if FW < v4.3 */
8478         if (i40e_is_mac_710(&pf->hw) &&
8479             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8480             (pf->hw.aq.fw_maj_ver < 4)))
8481                 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8482
8483         /* Use the FW Set LLDP MIB API if FW > v4.40 */
8484         if (i40e_is_mac_710(&pf->hw) &&
8485             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8486             (pf->hw.aq.fw_maj_ver >= 5)))
8487                 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8488
8489         if (pf->hw.func_caps.vmdq) {
8490                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8491                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8492                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8493         }
8494
8495         if (pf->hw.func_caps.iwarp) {
8496                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8497                 /* IWARP needs one extra vector for CQP just like MISC.*/
8498                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8499         }
8500
8501 #ifdef I40E_FCOE
8502         i40e_init_pf_fcoe(pf);
8503
8504 #endif /* I40E_FCOE */
8505 #ifdef CONFIG_PCI_IOV
8506         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8507                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8508                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8509                 pf->num_req_vfs = min_t(int,
8510                                         pf->hw.func_caps.num_vfs,
8511                                         I40E_MAX_VF_COUNT);
8512         }
8513 #endif /* CONFIG_PCI_IOV */
8514         if (pf->hw.mac.type == I40E_MAC_X722) {
8515                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8516                              I40E_FLAG_128_QP_RSS_CAPABLE |
8517                              I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8518                              I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8519                              I40E_FLAG_WB_ON_ITR_CAPABLE |
8520                              I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8521                              I40E_FLAG_NO_PCI_LINK_CHECK |
8522                              I40E_FLAG_100M_SGMII_CAPABLE |
8523                              I40E_FLAG_USE_SET_LLDP_MIB |
8524                              I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8525         } else if ((pf->hw.aq.api_maj_ver > 1) ||
8526                    ((pf->hw.aq.api_maj_ver == 1) &&
8527                     (pf->hw.aq.api_min_ver > 4))) {
8528                 /* Supported in FW API version higher than 1.4 */
8529                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8530                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8531         } else {
8532                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8533         }
8534
8535         pf->eeprom_version = 0xDEAD;
8536         pf->lan_veb = I40E_NO_VEB;
8537         pf->lan_vsi = I40E_NO_VSI;
8538
8539         /* By default FW has this off for performance reasons */
8540         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8541
8542         /* set up queue assignment tracking */
8543         size = sizeof(struct i40e_lump_tracking)
8544                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8545         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8546         if (!pf->qp_pile) {
8547                 err = -ENOMEM;
8548                 goto sw_init_done;
8549         }
8550         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8551         pf->qp_pile->search_hint = 0;
8552
8553         pf->tx_timeout_recovery_level = 1;
8554
8555         mutex_init(&pf->switch_mutex);
8556
8557         /* If NPAR is enabled nudge the Tx scheduler */
8558         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8559                 i40e_set_npar_bw_setting(pf);
8560
8561 sw_init_done:
8562         return err;
8563 }
8564
8565 /**
8566  * i40e_set_ntuple - set the ntuple feature flag and take action
8567  * @pf: board private structure to initialize
8568  * @features: the feature set that the stack is suggesting
8569  *
8570  * returns a bool to indicate if reset needs to happen
8571  **/
8572 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8573 {
8574         bool need_reset = false;
8575
8576         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8577          * the state changed, we need to reset.
8578          */
8579         if (features & NETIF_F_NTUPLE) {
8580                 /* Enable filters and mark for reset */
8581                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8582                         need_reset = true;
8583                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8584         } else {
8585                 /* turn off filters, mark for reset and clear SW filter list */
8586                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8587                         need_reset = true;
8588                         i40e_fdir_filter_exit(pf);
8589                 }
8590                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8591                 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8592                 /* reset fd counters */
8593                 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8594                 pf->fdir_pf_active_filters = 0;
8595                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8596                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8597                         dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8598                 /* if ATR was auto disabled it can be re-enabled. */
8599                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8600                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8601                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8602         }
8603         return need_reset;
8604 }
8605
8606 /**
8607  * i40e_set_features - set the netdev feature flags
8608  * @netdev: ptr to the netdev being adjusted
8609  * @features: the feature set that the stack is suggesting
8610  **/
8611 static int i40e_set_features(struct net_device *netdev,
8612                              netdev_features_t features)
8613 {
8614         struct i40e_netdev_priv *np = netdev_priv(netdev);
8615         struct i40e_vsi *vsi = np->vsi;
8616         struct i40e_pf *pf = vsi->back;
8617         bool need_reset;
8618
8619         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8620                 i40e_vlan_stripping_enable(vsi);
8621         else
8622                 i40e_vlan_stripping_disable(vsi);
8623
8624         need_reset = i40e_set_ntuple(pf, features);
8625
8626         if (need_reset)
8627                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8628
8629         return 0;
8630 }
8631
8632 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
8633 /**
8634  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8635  * @pf: board private structure
8636  * @port: The UDP port to look up
8637  *
8638  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8639  **/
8640 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8641 {
8642         u8 i;
8643
8644         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8645                 if (pf->udp_ports[i].index == port)
8646                         return i;
8647         }
8648
8649         return i;
8650 }
8651
8652 #endif
8653
8654 #if IS_ENABLED(CONFIG_VXLAN)
8655 /**
8656  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8657  * @netdev: This physical port's netdev
8658  * @sa_family: Socket Family that VXLAN is notifying us about
8659  * @port: New UDP port number that VXLAN started listening to
8660  **/
8661 static void i40e_add_vxlan_port(struct net_device *netdev,
8662                                 sa_family_t sa_family, __be16 port)
8663 {
8664         struct i40e_netdev_priv *np = netdev_priv(netdev);
8665         struct i40e_vsi *vsi = np->vsi;
8666         struct i40e_pf *pf = vsi->back;
8667         u8 next_idx;
8668         u8 idx;
8669
8670         idx = i40e_get_udp_port_idx(pf, port);
8671
8672         /* Check if port already exists */
8673         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8674                 netdev_info(netdev, "vxlan port %d already offloaded\n",
8675                             ntohs(port));
8676                 return;
8677         }
8678
8679         /* Now check if there is space to add the new port */
8680         next_idx = i40e_get_udp_port_idx(pf, 0);
8681
8682         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8683                 netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8684                             ntohs(port));
8685                 return;
8686         }
8687
8688         /* New port: add it and mark its index in the bitmap */
8689         pf->udp_ports[next_idx].index = port;
8690         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8691         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8692         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8693 }
8694
8695 /**
8696  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8697  * @netdev: This physical port's netdev
8698  * @sa_family: Socket Family that VXLAN is notifying us about
8699  * @port: UDP port number that VXLAN stopped listening to
8700  **/
8701 static void i40e_del_vxlan_port(struct net_device *netdev,
8702                                 sa_family_t sa_family, __be16 port)
8703 {
8704         struct i40e_netdev_priv *np = netdev_priv(netdev);
8705         struct i40e_vsi *vsi = np->vsi;
8706         struct i40e_pf *pf = vsi->back;
8707         u8 idx;
8708
8709         idx = i40e_get_udp_port_idx(pf, port);
8710
8711         /* Check if port already exists */
8712         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8713                 /* if port exists, set it to 0 (mark for deletion)
8714                  * and make it pending
8715                  */
8716                 pf->udp_ports[idx].index = 0;
8717                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8718                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8719         } else {
8720                 netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8721                             ntohs(port));
8722         }
8723 }
8724 #endif
8725
8726 #if IS_ENABLED(CONFIG_GENEVE)
8727 /**
8728  * i40e_add_geneve_port - Get notifications about GENEVE ports that come up
8729  * @netdev: This physical port's netdev
8730  * @sa_family: Socket Family that GENEVE is notifying us about
8731  * @port: New UDP port number that GENEVE started listening to
8732  **/
8733 static void i40e_add_geneve_port(struct net_device *netdev,
8734                                  sa_family_t sa_family, __be16 port)
8735 {
8736         struct i40e_netdev_priv *np = netdev_priv(netdev);
8737         struct i40e_vsi *vsi = np->vsi;
8738         struct i40e_pf *pf = vsi->back;
8739         u8 next_idx;
8740         u8 idx;
8741
8742         if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8743                 return;
8744
8745         idx = i40e_get_udp_port_idx(pf, port);
8746
8747         /* Check if port already exists */
8748         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8749                 netdev_info(netdev, "udp port %d already offloaded\n",
8750                             ntohs(port));
8751                 return;
8752         }
8753
8754         /* Now check if there is space to add the new port */
8755         next_idx = i40e_get_udp_port_idx(pf, 0);
8756
8757         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8758                 netdev_info(netdev, "maximum number of UDP ports reached, not adding port %d\n",
8759                             ntohs(port));
8760                 return;
8761         }
8762
8763         /* New port: add it and mark its index in the bitmap */
8764         pf->udp_ports[next_idx].index = port;
8765         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8766         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8767         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8768
8769         dev_info(&pf->pdev->dev, "adding geneve port %d\n", ntohs(port));
8770 }
8771
8772 /**
8773  * i40e_del_geneve_port - Get notifications about GENEVE ports that go away
8774  * @netdev: This physical port's netdev
8775  * @sa_family: Socket Family that GENEVE is notifying us about
8776  * @port: UDP port number that GENEVE stopped listening to
8777  **/
8778 static void i40e_del_geneve_port(struct net_device *netdev,
8779                                  sa_family_t sa_family, __be16 port)
8780 {
8781         struct i40e_netdev_priv *np = netdev_priv(netdev);
8782         struct i40e_vsi *vsi = np->vsi;
8783         struct i40e_pf *pf = vsi->back;
8784         u8 idx;
8785
8786         if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8787                 return;
8788
8789         idx = i40e_get_udp_port_idx(pf, port);
8790
8791         /* Check if port already exists */
8792         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8793                 /* if port exists, set it to 0 (mark for deletion)
8794                  * and make it pending
8795                  */
8796                 pf->udp_ports[idx].index = 0;
8797                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8798                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8799
8800                 dev_info(&pf->pdev->dev, "deleting geneve port %d\n",
8801                          ntohs(port));
8802         } else {
8803                 netdev_warn(netdev, "geneve port %d was not found, not deleting\n",
8804                             ntohs(port));
8805         }
8806 }
8807 #endif
8808
8809 static int i40e_get_phys_port_id(struct net_device *netdev,
8810                                  struct netdev_phys_item_id *ppid)
8811 {
8812         struct i40e_netdev_priv *np = netdev_priv(netdev);
8813         struct i40e_pf *pf = np->vsi->back;
8814         struct i40e_hw *hw = &pf->hw;
8815
8816         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8817                 return -EOPNOTSUPP;
8818
8819         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8820         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8821
8822         return 0;
8823 }
8824
8825 /**
8826  * i40e_ndo_fdb_add - add an entry to the hardware database
8827  * @ndm: the input from the stack
8828  * @tb: pointer to array of nladdr (unused)
8829  * @dev: the net device pointer
8830  * @addr: the MAC address entry being added
8831  * @flags: instructions from stack about fdb operation
8832  */
8833 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8834                             struct net_device *dev,
8835                             const unsigned char *addr, u16 vid,
8836                             u16 flags)
8837 {
8838         struct i40e_netdev_priv *np = netdev_priv(dev);
8839         struct i40e_pf *pf = np->vsi->back;
8840         int err = 0;
8841
8842         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8843                 return -EOPNOTSUPP;
8844
8845         if (vid) {
8846                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8847                 return -EINVAL;
8848         }
8849
8850         /* Hardware does not support aging addresses so if a
8851          * ndm_state is given only allow permanent addresses
8852          */
8853         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8854                 netdev_info(dev, "FDB only supports static addresses\n");
8855                 return -EINVAL;
8856         }
8857
8858         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8859                 err = dev_uc_add_excl(dev, addr);
8860         else if (is_multicast_ether_addr(addr))
8861                 err = dev_mc_add_excl(dev, addr);
8862         else
8863                 err = -EINVAL;
8864
8865         /* Only return duplicate errors if NLM_F_EXCL is set */
8866         if (err == -EEXIST && !(flags & NLM_F_EXCL))
8867                 err = 0;
8868
8869         return err;
8870 }
8871
8872 /**
8873  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8874  * @dev: the netdev being configured
8875  * @nlh: RTNL message
8876  *
8877  * Inserts a new hardware bridge if not already created and
8878  * enables the bridging mode requested (VEB or VEPA). If the
8879  * hardware bridge has already been inserted and the request
8880  * is to change the mode then that requires a PF reset to
8881  * allow rebuild of the components with required hardware
8882  * bridge mode enabled.
8883  **/
8884 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8885                                    struct nlmsghdr *nlh,
8886                                    u16 flags)
8887 {
8888         struct i40e_netdev_priv *np = netdev_priv(dev);
8889         struct i40e_vsi *vsi = np->vsi;
8890         struct i40e_pf *pf = vsi->back;
8891         struct i40e_veb *veb = NULL;
8892         struct nlattr *attr, *br_spec;
8893         int i, rem;
8894
8895         /* Only for PF VSI for now */
8896         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8897                 return -EOPNOTSUPP;
8898
8899         /* Find the HW bridge for PF VSI */
8900         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8901                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8902                         veb = pf->veb[i];
8903         }
8904
8905         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8906
8907         nla_for_each_nested(attr, br_spec, rem) {
8908                 __u16 mode;
8909
8910                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8911                         continue;
8912
8913                 mode = nla_get_u16(attr);
8914                 if ((mode != BRIDGE_MODE_VEPA) &&
8915                     (mode != BRIDGE_MODE_VEB))
8916                         return -EINVAL;
8917
8918                 /* Insert a new HW bridge */
8919                 if (!veb) {
8920                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8921                                              vsi->tc_config.enabled_tc);
8922                         if (veb) {
8923                                 veb->bridge_mode = mode;
8924                                 i40e_config_bridge_mode(veb);
8925                         } else {
8926                                 /* No Bridge HW offload available */
8927                                 return -ENOENT;
8928                         }
8929                         break;
8930                 } else if (mode != veb->bridge_mode) {
8931                         /* Existing HW bridge but different mode needs reset */
8932                         veb->bridge_mode = mode;
8933                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8934                         if (mode == BRIDGE_MODE_VEB)
8935                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8936                         else
8937                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8938                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8939                         break;
8940                 }
8941         }
8942
8943         return 0;
8944 }
8945
8946 /**
8947  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8948  * @skb: skb buff
8949  * @pid: process id
8950  * @seq: RTNL message seq #
8951  * @dev: the netdev being configured
8952  * @filter_mask: unused
8953  * @nlflags: netlink flags passed in
8954  *
8955  * Return the mode in which the hardware bridge is operating in
8956  * i.e VEB or VEPA.
8957  **/
8958 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8959                                    struct net_device *dev,
8960                                    u32 __always_unused filter_mask,
8961                                    int nlflags)
8962 {
8963         struct i40e_netdev_priv *np = netdev_priv(dev);
8964         struct i40e_vsi *vsi = np->vsi;
8965         struct i40e_pf *pf = vsi->back;
8966         struct i40e_veb *veb = NULL;
8967         int i;
8968
8969         /* Only for PF VSI for now */
8970         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8971                 return -EOPNOTSUPP;
8972
8973         /* Find the HW bridge for the PF VSI */
8974         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8975                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8976                         veb = pf->veb[i];
8977         }
8978
8979         if (!veb)
8980                 return 0;
8981
8982         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8983                                        nlflags, 0, 0, filter_mask, NULL);
8984 }
8985
8986 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
8987  * inner mac plus all inner ethertypes.
8988  */
8989 #define I40E_MAX_TUNNEL_HDR_LEN 128
8990 /**
8991  * i40e_features_check - Validate encapsulated packet conforms to limits
8992  * @skb: skb buff
8993  * @dev: This physical port's netdev
8994  * @features: Offload features that the stack believes apply
8995  **/
8996 static netdev_features_t i40e_features_check(struct sk_buff *skb,
8997                                              struct net_device *dev,
8998                                              netdev_features_t features)
8999 {
9000         if (skb->encapsulation &&
9001             ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9002              I40E_MAX_TUNNEL_HDR_LEN))
9003                 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9004
9005         return features;
9006 }
9007
9008 static const struct net_device_ops i40e_netdev_ops = {
9009         .ndo_open               = i40e_open,
9010         .ndo_stop               = i40e_close,
9011         .ndo_start_xmit         = i40e_lan_xmit_frame,
9012         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
9013         .ndo_set_rx_mode        = i40e_set_rx_mode,
9014         .ndo_validate_addr      = eth_validate_addr,
9015         .ndo_set_mac_address    = i40e_set_mac,
9016         .ndo_change_mtu         = i40e_change_mtu,
9017         .ndo_do_ioctl           = i40e_ioctl,
9018         .ndo_tx_timeout         = i40e_tx_timeout,
9019         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
9020         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
9021 #ifdef CONFIG_NET_POLL_CONTROLLER
9022         .ndo_poll_controller    = i40e_netpoll,
9023 #endif
9024         .ndo_setup_tc           = __i40e_setup_tc,
9025 #ifdef I40E_FCOE
9026         .ndo_fcoe_enable        = i40e_fcoe_enable,
9027         .ndo_fcoe_disable       = i40e_fcoe_disable,
9028 #endif
9029         .ndo_set_features       = i40e_set_features,
9030         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
9031         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
9032         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
9033         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
9034         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
9035         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
9036         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
9037 #if IS_ENABLED(CONFIG_VXLAN)
9038         .ndo_add_vxlan_port     = i40e_add_vxlan_port,
9039         .ndo_del_vxlan_port     = i40e_del_vxlan_port,
9040 #endif
9041 #if IS_ENABLED(CONFIG_GENEVE)
9042         .ndo_add_geneve_port    = i40e_add_geneve_port,
9043         .ndo_del_geneve_port    = i40e_del_geneve_port,
9044 #endif
9045         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
9046         .ndo_fdb_add            = i40e_ndo_fdb_add,
9047         .ndo_features_check     = i40e_features_check,
9048         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
9049         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
9050 };
9051
9052 /**
9053  * i40e_config_netdev - Setup the netdev flags
9054  * @vsi: the VSI being configured
9055  *
9056  * Returns 0 on success, negative value on failure
9057  **/
9058 static int i40e_config_netdev(struct i40e_vsi *vsi)
9059 {
9060         u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
9061         struct i40e_pf *pf = vsi->back;
9062         struct i40e_hw *hw = &pf->hw;
9063         struct i40e_netdev_priv *np;
9064         struct net_device *netdev;
9065         u8 mac_addr[ETH_ALEN];
9066         int etherdev_size;
9067
9068         etherdev_size = sizeof(struct i40e_netdev_priv);
9069         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9070         if (!netdev)
9071                 return -ENOMEM;
9072
9073         vsi->netdev = netdev;
9074         np = netdev_priv(netdev);
9075         np->vsi = vsi;
9076
9077         netdev->hw_enc_features |= NETIF_F_SG                   |
9078                                    NETIF_F_IP_CSUM              |
9079                                    NETIF_F_IPV6_CSUM            |
9080                                    NETIF_F_HIGHDMA              |
9081                                    NETIF_F_SOFT_FEATURES        |
9082                                    NETIF_F_TSO                  |
9083                                    NETIF_F_TSO_ECN              |
9084                                    NETIF_F_TSO6                 |
9085                                    NETIF_F_GSO_GRE              |
9086                                    NETIF_F_GSO_GRE_CSUM         |
9087                                    NETIF_F_GSO_IPXIP4           |
9088                                    NETIF_F_GSO_IPXIP6           |
9089                                    NETIF_F_GSO_UDP_TUNNEL       |
9090                                    NETIF_F_GSO_UDP_TUNNEL_CSUM  |
9091                                    NETIF_F_GSO_PARTIAL          |
9092                                    NETIF_F_SCTP_CRC             |
9093                                    NETIF_F_RXHASH               |
9094                                    NETIF_F_RXCSUM               |
9095                                    0;
9096
9097         if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9098                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9099
9100         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9101
9102         /* record features VLANs can make use of */
9103         netdev->vlan_features |= netdev->hw_enc_features |
9104                                  NETIF_F_TSO_MANGLEID;
9105
9106         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9107                 netdev->hw_features |= NETIF_F_NTUPLE;
9108
9109         netdev->hw_features |= netdev->hw_enc_features  |
9110                                NETIF_F_HW_VLAN_CTAG_TX  |
9111                                NETIF_F_HW_VLAN_CTAG_RX;
9112
9113         netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9114         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9115
9116         if (vsi->type == I40E_VSI_MAIN) {
9117                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9118                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9119                 /* The following steps are necessary to prevent reception
9120                  * of tagged packets - some older NVM configurations load a
9121                  * default a MAC-VLAN filter that accepts any tagged packet
9122                  * which must be replaced by a normal filter.
9123                  */
9124                 if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
9125                         spin_lock_bh(&vsi->mac_filter_list_lock);
9126                         i40e_add_filter(vsi, mac_addr,
9127                                         I40E_VLAN_ANY, false, true);
9128                         spin_unlock_bh(&vsi->mac_filter_list_lock);
9129                 }
9130         } else if ((pf->hw.aq.api_maj_ver > 1) ||
9131                    ((pf->hw.aq.api_maj_ver == 1) &&
9132                     (pf->hw.aq.api_min_ver > 4))) {
9133                 /* Supported in FW API version higher than 1.4 */
9134                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
9135                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
9136         } else {
9137                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9138                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9139                          pf->vsi[pf->lan_vsi]->netdev->name);
9140                 random_ether_addr(mac_addr);
9141
9142                 spin_lock_bh(&vsi->mac_filter_list_lock);
9143                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9144                 spin_unlock_bh(&vsi->mac_filter_list_lock);
9145         }
9146
9147         spin_lock_bh(&vsi->mac_filter_list_lock);
9148         i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
9149         spin_unlock_bh(&vsi->mac_filter_list_lock);
9150
9151         ether_addr_copy(netdev->dev_addr, mac_addr);
9152         ether_addr_copy(netdev->perm_addr, mac_addr);
9153
9154         netdev->priv_flags |= IFF_UNICAST_FLT;
9155         netdev->priv_flags |= IFF_SUPP_NOFCS;
9156         /* Setup netdev TC information */
9157         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9158
9159         netdev->netdev_ops = &i40e_netdev_ops;
9160         netdev->watchdog_timeo = 5 * HZ;
9161         i40e_set_ethtool_ops(netdev);
9162 #ifdef I40E_FCOE
9163         i40e_fcoe_config_netdev(netdev, vsi);
9164 #endif
9165
9166         return 0;
9167 }
9168
9169 /**
9170  * i40e_vsi_delete - Delete a VSI from the switch
9171  * @vsi: the VSI being removed
9172  *
9173  * Returns 0 on success, negative value on failure
9174  **/
9175 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9176 {
9177         /* remove default VSI is not allowed */
9178         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9179                 return;
9180
9181         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9182 }
9183
9184 /**
9185  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9186  * @vsi: the VSI being queried
9187  *
9188  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9189  **/
9190 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9191 {
9192         struct i40e_veb *veb;
9193         struct i40e_pf *pf = vsi->back;
9194
9195         /* Uplink is not a bridge so default to VEB */
9196         if (vsi->veb_idx == I40E_NO_VEB)
9197                 return 1;
9198
9199         veb = pf->veb[vsi->veb_idx];
9200         if (!veb) {
9201                 dev_info(&pf->pdev->dev,
9202                          "There is no veb associated with the bridge\n");
9203                 return -ENOENT;
9204         }
9205
9206         /* Uplink is a bridge in VEPA mode */
9207         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9208                 return 0;
9209         } else {
9210                 /* Uplink is a bridge in VEB mode */
9211                 return 1;
9212         }
9213
9214         /* VEPA is now default bridge, so return 0 */
9215         return 0;
9216 }
9217
9218 /**
9219  * i40e_add_vsi - Add a VSI to the switch
9220  * @vsi: the VSI being configured
9221  *
9222  * This initializes a VSI context depending on the VSI type to be added and
9223  * passes it down to the add_vsi aq command.
9224  **/
9225 static int i40e_add_vsi(struct i40e_vsi *vsi)
9226 {
9227         int ret = -ENODEV;
9228         i40e_status aq_ret = 0;
9229         u8 laa_macaddr[ETH_ALEN];
9230         bool found_laa_mac_filter = false;
9231         struct i40e_pf *pf = vsi->back;
9232         struct i40e_hw *hw = &pf->hw;
9233         struct i40e_vsi_context ctxt;
9234         struct i40e_mac_filter *f, *ftmp;
9235
9236         u8 enabled_tc = 0x1; /* TC0 enabled */
9237         int f_count = 0;
9238
9239         memset(&ctxt, 0, sizeof(ctxt));
9240         switch (vsi->type) {
9241         case I40E_VSI_MAIN:
9242                 /* The PF's main VSI is already setup as part of the
9243                  * device initialization, so we'll not bother with
9244                  * the add_vsi call, but we will retrieve the current
9245                  * VSI context.
9246                  */
9247                 ctxt.seid = pf->main_vsi_seid;
9248                 ctxt.pf_num = pf->hw.pf_id;
9249                 ctxt.vf_num = 0;
9250                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9251                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9252                 if (ret) {
9253                         dev_info(&pf->pdev->dev,
9254                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9255                                  i40e_stat_str(&pf->hw, ret),
9256                                  i40e_aq_str(&pf->hw,
9257                                              pf->hw.aq.asq_last_status));
9258                         return -ENOENT;
9259                 }
9260                 vsi->info = ctxt.info;
9261                 vsi->info.valid_sections = 0;
9262
9263                 vsi->seid = ctxt.seid;
9264                 vsi->id = ctxt.vsi_number;
9265
9266                 enabled_tc = i40e_pf_get_tc_map(pf);
9267
9268                 /* MFP mode setup queue map and update VSI */
9269                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9270                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9271                         memset(&ctxt, 0, sizeof(ctxt));
9272                         ctxt.seid = pf->main_vsi_seid;
9273                         ctxt.pf_num = pf->hw.pf_id;
9274                         ctxt.vf_num = 0;
9275                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9276                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9277                         if (ret) {
9278                                 dev_info(&pf->pdev->dev,
9279                                          "update vsi failed, err %s aq_err %s\n",
9280                                          i40e_stat_str(&pf->hw, ret),
9281                                          i40e_aq_str(&pf->hw,
9282                                                     pf->hw.aq.asq_last_status));
9283                                 ret = -ENOENT;
9284                                 goto err;
9285                         }
9286                         /* update the local VSI info queue map */
9287                         i40e_vsi_update_queue_map(vsi, &ctxt);
9288                         vsi->info.valid_sections = 0;
9289                 } else {
9290                         /* Default/Main VSI is only enabled for TC0
9291                          * reconfigure it to enable all TCs that are
9292                          * available on the port in SFP mode.
9293                          * For MFP case the iSCSI PF would use this
9294                          * flow to enable LAN+iSCSI TC.
9295                          */
9296                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9297                         if (ret) {
9298                                 dev_info(&pf->pdev->dev,
9299                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9300                                          enabled_tc,
9301                                          i40e_stat_str(&pf->hw, ret),
9302                                          i40e_aq_str(&pf->hw,
9303                                                     pf->hw.aq.asq_last_status));
9304                                 ret = -ENOENT;
9305                         }
9306                 }
9307                 break;
9308
9309         case I40E_VSI_FDIR:
9310                 ctxt.pf_num = hw->pf_id;
9311                 ctxt.vf_num = 0;
9312                 ctxt.uplink_seid = vsi->uplink_seid;
9313                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9314                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9315                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9316                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9317                         ctxt.info.valid_sections |=
9318                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9319                         ctxt.info.switch_id =
9320                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9321                 }
9322                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9323                 break;
9324
9325         case I40E_VSI_VMDQ2:
9326                 ctxt.pf_num = hw->pf_id;
9327                 ctxt.vf_num = 0;
9328                 ctxt.uplink_seid = vsi->uplink_seid;
9329                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9330                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9331
9332                 /* This VSI is connected to VEB so the switch_id
9333                  * should be set to zero by default.
9334                  */
9335                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9336                         ctxt.info.valid_sections |=
9337                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9338                         ctxt.info.switch_id =
9339                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9340                 }
9341
9342                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9343                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9344                 break;
9345
9346         case I40E_VSI_SRIOV:
9347                 ctxt.pf_num = hw->pf_id;
9348                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9349                 ctxt.uplink_seid = vsi->uplink_seid;
9350                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9351                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9352
9353                 /* This VSI is connected to VEB so the switch_id
9354                  * should be set to zero by default.
9355                  */
9356                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9357                         ctxt.info.valid_sections |=
9358                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9359                         ctxt.info.switch_id =
9360                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9361                 }
9362
9363                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9364                         ctxt.info.valid_sections |=
9365                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9366                         ctxt.info.queueing_opt_flags |=
9367                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9368                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9369                 }
9370
9371                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9372                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9373                 if (pf->vf[vsi->vf_id].spoofchk) {
9374                         ctxt.info.valid_sections |=
9375                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9376                         ctxt.info.sec_flags |=
9377                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9378                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9379                 }
9380                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9381                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9382                 break;
9383
9384 #ifdef I40E_FCOE
9385         case I40E_VSI_FCOE:
9386                 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9387                 if (ret) {
9388                         dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9389                         return ret;
9390                 }
9391                 break;
9392
9393 #endif /* I40E_FCOE */
9394         case I40E_VSI_IWARP:
9395                 /* send down message to iWARP */
9396                 break;
9397
9398         default:
9399                 return -ENODEV;
9400         }
9401
9402         if (vsi->type != I40E_VSI_MAIN) {
9403                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9404                 if (ret) {
9405                         dev_info(&vsi->back->pdev->dev,
9406                                  "add vsi failed, err %s aq_err %s\n",
9407                                  i40e_stat_str(&pf->hw, ret),
9408                                  i40e_aq_str(&pf->hw,
9409                                              pf->hw.aq.asq_last_status));
9410                         ret = -ENOENT;
9411                         goto err;
9412                 }
9413                 vsi->info = ctxt.info;
9414                 vsi->info.valid_sections = 0;
9415                 vsi->seid = ctxt.seid;
9416                 vsi->id = ctxt.vsi_number;
9417         }
9418         /* Except FDIR VSI, for all othet VSI set the broadcast filter */
9419         if (vsi->type != I40E_VSI_FDIR) {
9420                 aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9421                 if (aq_ret) {
9422                         ret = i40e_aq_rc_to_posix(aq_ret,
9423                                                   hw->aq.asq_last_status);
9424                         dev_info(&pf->pdev->dev,
9425                                  "set brdcast promisc failed, err %s, aq_err %s\n",
9426                                  i40e_stat_str(hw, aq_ret),
9427                                  i40e_aq_str(hw, hw->aq.asq_last_status));
9428                 }
9429         }
9430
9431         spin_lock_bh(&vsi->mac_filter_list_lock);
9432         /* If macvlan filters already exist, force them to get loaded */
9433         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9434                 f->changed = true;
9435                 f_count++;
9436
9437                 /* Expected to have only one MAC filter entry for LAA in list */
9438                 if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
9439                         ether_addr_copy(laa_macaddr, f->macaddr);
9440                         found_laa_mac_filter = true;
9441                 }
9442         }
9443         spin_unlock_bh(&vsi->mac_filter_list_lock);
9444
9445         if (found_laa_mac_filter) {
9446                 struct i40e_aqc_remove_macvlan_element_data element;
9447
9448                 memset(&element, 0, sizeof(element));
9449                 ether_addr_copy(element.mac_addr, laa_macaddr);
9450                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
9451                 ret = i40e_aq_remove_macvlan(hw, vsi->seid,
9452                                              &element, 1, NULL);
9453                 if (ret) {
9454                         /* some older FW has a different default */
9455                         element.flags |=
9456                                        I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
9457                         i40e_aq_remove_macvlan(hw, vsi->seid,
9458                                                &element, 1, NULL);
9459                 }
9460
9461                 i40e_aq_mac_address_write(hw,
9462                                           I40E_AQC_WRITE_TYPE_LAA_WOL,
9463                                           laa_macaddr, NULL);
9464         }
9465
9466         if (f_count) {
9467                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9468                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9469         }
9470
9471         /* Update VSI BW information */
9472         ret = i40e_vsi_get_bw_info(vsi);
9473         if (ret) {
9474                 dev_info(&pf->pdev->dev,
9475                          "couldn't get vsi bw info, err %s aq_err %s\n",
9476                          i40e_stat_str(&pf->hw, ret),
9477                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9478                 /* VSI is already added so not tearing that up */
9479                 ret = 0;
9480         }
9481
9482 err:
9483         return ret;
9484 }
9485
9486 /**
9487  * i40e_vsi_release - Delete a VSI and free its resources
9488  * @vsi: the VSI being removed
9489  *
9490  * Returns 0 on success or < 0 on error
9491  **/
9492 int i40e_vsi_release(struct i40e_vsi *vsi)
9493 {
9494         struct i40e_mac_filter *f, *ftmp;
9495         struct i40e_veb *veb = NULL;
9496         struct i40e_pf *pf;
9497         u16 uplink_seid;
9498         int i, n;
9499
9500         pf = vsi->back;
9501
9502         /* release of a VEB-owner or last VSI is not allowed */
9503         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9504                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9505                          vsi->seid, vsi->uplink_seid);
9506                 return -ENODEV;
9507         }
9508         if (vsi == pf->vsi[pf->lan_vsi] &&
9509             !test_bit(__I40E_DOWN, &pf->state)) {
9510                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9511                 return -ENODEV;
9512         }
9513
9514         uplink_seid = vsi->uplink_seid;
9515         if (vsi->type != I40E_VSI_SRIOV) {
9516                 if (vsi->netdev_registered) {
9517                         vsi->netdev_registered = false;
9518                         if (vsi->netdev) {
9519                                 /* results in a call to i40e_close() */
9520                                 unregister_netdev(vsi->netdev);
9521                         }
9522                 } else {
9523                         i40e_vsi_close(vsi);
9524                 }
9525                 i40e_vsi_disable_irq(vsi);
9526         }
9527
9528         spin_lock_bh(&vsi->mac_filter_list_lock);
9529         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9530                 i40e_del_filter(vsi, f->macaddr, f->vlan,
9531                                 f->is_vf, f->is_netdev);
9532         spin_unlock_bh(&vsi->mac_filter_list_lock);
9533
9534         i40e_sync_vsi_filters(vsi);
9535
9536         i40e_vsi_delete(vsi);
9537         i40e_vsi_free_q_vectors(vsi);
9538         if (vsi->netdev) {
9539                 free_netdev(vsi->netdev);
9540                 vsi->netdev = NULL;
9541         }
9542         i40e_vsi_clear_rings(vsi);
9543         i40e_vsi_clear(vsi);
9544
9545         /* If this was the last thing on the VEB, except for the
9546          * controlling VSI, remove the VEB, which puts the controlling
9547          * VSI onto the next level down in the switch.
9548          *
9549          * Well, okay, there's one more exception here: don't remove
9550          * the orphan VEBs yet.  We'll wait for an explicit remove request
9551          * from up the network stack.
9552          */
9553         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9554                 if (pf->vsi[i] &&
9555                     pf->vsi[i]->uplink_seid == uplink_seid &&
9556                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9557                         n++;      /* count the VSIs */
9558                 }
9559         }
9560         for (i = 0; i < I40E_MAX_VEB; i++) {
9561                 if (!pf->veb[i])
9562                         continue;
9563                 if (pf->veb[i]->uplink_seid == uplink_seid)
9564                         n++;     /* count the VEBs */
9565                 if (pf->veb[i]->seid == uplink_seid)
9566                         veb = pf->veb[i];
9567         }
9568         if (n == 0 && veb && veb->uplink_seid != 0)
9569                 i40e_veb_release(veb);
9570
9571         return 0;
9572 }
9573
9574 /**
9575  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9576  * @vsi: ptr to the VSI
9577  *
9578  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9579  * corresponding SW VSI structure and initializes num_queue_pairs for the
9580  * newly allocated VSI.
9581  *
9582  * Returns 0 on success or negative on failure
9583  **/
9584 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9585 {
9586         int ret = -ENOENT;
9587         struct i40e_pf *pf = vsi->back;
9588
9589         if (vsi->q_vectors[0]) {
9590                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9591                          vsi->seid);
9592                 return -EEXIST;
9593         }
9594
9595         if (vsi->base_vector) {
9596                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9597                          vsi->seid, vsi->base_vector);
9598                 return -EEXIST;
9599         }
9600
9601         ret = i40e_vsi_alloc_q_vectors(vsi);
9602         if (ret) {
9603                 dev_info(&pf->pdev->dev,
9604                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9605                          vsi->num_q_vectors, vsi->seid, ret);
9606                 vsi->num_q_vectors = 0;
9607                 goto vector_setup_out;
9608         }
9609
9610         /* In Legacy mode, we do not have to get any other vector since we
9611          * piggyback on the misc/ICR0 for queue interrupts.
9612         */
9613         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9614                 return ret;
9615         if (vsi->num_q_vectors)
9616                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9617                                                  vsi->num_q_vectors, vsi->idx);
9618         if (vsi->base_vector < 0) {
9619                 dev_info(&pf->pdev->dev,
9620                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9621                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9622                 i40e_vsi_free_q_vectors(vsi);
9623                 ret = -ENOENT;
9624                 goto vector_setup_out;
9625         }
9626
9627 vector_setup_out:
9628         return ret;
9629 }
9630
9631 /**
9632  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9633  * @vsi: pointer to the vsi.
9634  *
9635  * This re-allocates a vsi's queue resources.
9636  *
9637  * Returns pointer to the successfully allocated and configured VSI sw struct
9638  * on success, otherwise returns NULL on failure.
9639  **/
9640 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9641 {
9642         struct i40e_pf *pf;
9643         u8 enabled_tc;
9644         int ret;
9645
9646         if (!vsi)
9647                 return NULL;
9648
9649         pf = vsi->back;
9650
9651         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9652         i40e_vsi_clear_rings(vsi);
9653
9654         i40e_vsi_free_arrays(vsi, false);
9655         i40e_set_num_rings_in_vsi(vsi);
9656         ret = i40e_vsi_alloc_arrays(vsi, false);
9657         if (ret)
9658                 goto err_vsi;
9659
9660         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9661         if (ret < 0) {
9662                 dev_info(&pf->pdev->dev,
9663                          "failed to get tracking for %d queues for VSI %d err %d\n",
9664                          vsi->alloc_queue_pairs, vsi->seid, ret);
9665                 goto err_vsi;
9666         }
9667         vsi->base_queue = ret;
9668
9669         /* Update the FW view of the VSI. Force a reset of TC and queue
9670          * layout configurations.
9671          */
9672         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9673         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9674         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9675         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9676
9677         /* assign it some queues */
9678         ret = i40e_alloc_rings(vsi);
9679         if (ret)
9680                 goto err_rings;
9681
9682         /* map all of the rings to the q_vectors */
9683         i40e_vsi_map_rings_to_vectors(vsi);
9684         return vsi;
9685
9686 err_rings:
9687         i40e_vsi_free_q_vectors(vsi);
9688         if (vsi->netdev_registered) {
9689                 vsi->netdev_registered = false;
9690                 unregister_netdev(vsi->netdev);
9691                 free_netdev(vsi->netdev);
9692                 vsi->netdev = NULL;
9693         }
9694         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9695 err_vsi:
9696         i40e_vsi_clear(vsi);
9697         return NULL;
9698 }
9699
9700 /**
9701  * i40e_macaddr_init - explicitly write the mac address filters.
9702  *
9703  * @vsi: pointer to the vsi.
9704  * @macaddr: the MAC address
9705  *
9706  * This is needed when the macaddr has been obtained by other
9707  * means than the default, e.g., from Open Firmware or IDPROM.
9708  * Returns 0 on success, negative on failure
9709  **/
9710 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
9711 {
9712         int ret;
9713         struct i40e_aqc_add_macvlan_element_data element;
9714
9715         ret = i40e_aq_mac_address_write(&vsi->back->hw,
9716                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
9717                                         macaddr, NULL);
9718         if (ret) {
9719                 dev_info(&vsi->back->pdev->dev,
9720                          "Addr change for VSI failed: %d\n", ret);
9721                 return -EADDRNOTAVAIL;
9722         }
9723
9724         memset(&element, 0, sizeof(element));
9725         ether_addr_copy(element.mac_addr, macaddr);
9726         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
9727         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
9728         if (ret) {
9729                 dev_info(&vsi->back->pdev->dev,
9730                          "add filter failed err %s aq_err %s\n",
9731                          i40e_stat_str(&vsi->back->hw, ret),
9732                          i40e_aq_str(&vsi->back->hw,
9733                                      vsi->back->hw.aq.asq_last_status));
9734         }
9735         return ret;
9736 }
9737
9738 /**
9739  * i40e_vsi_setup - Set up a VSI by a given type
9740  * @pf: board private structure
9741  * @type: VSI type
9742  * @uplink_seid: the switch element to link to
9743  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9744  *
9745  * This allocates the sw VSI structure and its queue resources, then add a VSI
9746  * to the identified VEB.
9747  *
9748  * Returns pointer to the successfully allocated and configure VSI sw struct on
9749  * success, otherwise returns NULL on failure.
9750  **/
9751 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9752                                 u16 uplink_seid, u32 param1)
9753 {
9754         struct i40e_vsi *vsi = NULL;
9755         struct i40e_veb *veb = NULL;
9756         int ret, i;
9757         int v_idx;
9758
9759         /* The requested uplink_seid must be either
9760          *     - the PF's port seid
9761          *              no VEB is needed because this is the PF
9762          *              or this is a Flow Director special case VSI
9763          *     - seid of an existing VEB
9764          *     - seid of a VSI that owns an existing VEB
9765          *     - seid of a VSI that doesn't own a VEB
9766          *              a new VEB is created and the VSI becomes the owner
9767          *     - seid of the PF VSI, which is what creates the first VEB
9768          *              this is a special case of the previous
9769          *
9770          * Find which uplink_seid we were given and create a new VEB if needed
9771          */
9772         for (i = 0; i < I40E_MAX_VEB; i++) {
9773                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9774                         veb = pf->veb[i];
9775                         break;
9776                 }
9777         }
9778
9779         if (!veb && uplink_seid != pf->mac_seid) {
9780
9781                 for (i = 0; i < pf->num_alloc_vsi; i++) {
9782                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9783                                 vsi = pf->vsi[i];
9784                                 break;
9785                         }
9786                 }
9787                 if (!vsi) {
9788                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9789                                  uplink_seid);
9790                         return NULL;
9791                 }
9792
9793                 if (vsi->uplink_seid == pf->mac_seid)
9794                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9795                                              vsi->tc_config.enabled_tc);
9796                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9797                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9798                                              vsi->tc_config.enabled_tc);
9799                 if (veb) {
9800                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9801                                 dev_info(&vsi->back->pdev->dev,
9802                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
9803                                 return NULL;
9804                         }
9805                         /* We come up by default in VEPA mode if SRIOV is not
9806                          * already enabled, in which case we can't force VEPA
9807                          * mode.
9808                          */
9809                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9810                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9811                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9812                         }
9813                         i40e_config_bridge_mode(veb);
9814                 }
9815                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9816                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9817                                 veb = pf->veb[i];
9818                 }
9819                 if (!veb) {
9820                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9821                         return NULL;
9822                 }
9823
9824                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9825                 uplink_seid = veb->seid;
9826         }
9827
9828         /* get vsi sw struct */
9829         v_idx = i40e_vsi_mem_alloc(pf, type);
9830         if (v_idx < 0)
9831                 goto err_alloc;
9832         vsi = pf->vsi[v_idx];
9833         if (!vsi)
9834                 goto err_alloc;
9835         vsi->type = type;
9836         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9837
9838         if (type == I40E_VSI_MAIN)
9839                 pf->lan_vsi = v_idx;
9840         else if (type == I40E_VSI_SRIOV)
9841                 vsi->vf_id = param1;
9842         /* assign it some queues */
9843         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9844                                 vsi->idx);
9845         if (ret < 0) {
9846                 dev_info(&pf->pdev->dev,
9847                          "failed to get tracking for %d queues for VSI %d err=%d\n",
9848                          vsi->alloc_queue_pairs, vsi->seid, ret);
9849                 goto err_vsi;
9850         }
9851         vsi->base_queue = ret;
9852
9853         /* get a VSI from the hardware */
9854         vsi->uplink_seid = uplink_seid;
9855         ret = i40e_add_vsi(vsi);
9856         if (ret)
9857                 goto err_vsi;
9858
9859         switch (vsi->type) {
9860         /* setup the netdev if needed */
9861         case I40E_VSI_MAIN:
9862                 /* Apply relevant filters if a platform-specific mac
9863                  * address was selected.
9864                  */
9865                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9866                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9867                         if (ret) {
9868                                 dev_warn(&pf->pdev->dev,
9869                                          "could not set up macaddr; err %d\n",
9870                                          ret);
9871                         }
9872                 }
9873         case I40E_VSI_VMDQ2:
9874         case I40E_VSI_FCOE:
9875                 ret = i40e_config_netdev(vsi);
9876                 if (ret)
9877                         goto err_netdev;
9878                 ret = register_netdev(vsi->netdev);
9879                 if (ret)
9880                         goto err_netdev;
9881                 vsi->netdev_registered = true;
9882                 netif_carrier_off(vsi->netdev);
9883 #ifdef CONFIG_I40E_DCB
9884                 /* Setup DCB netlink interface */
9885                 i40e_dcbnl_setup(vsi);
9886 #endif /* CONFIG_I40E_DCB */
9887                 /* fall through */
9888
9889         case I40E_VSI_FDIR:
9890                 /* set up vectors and rings if needed */
9891                 ret = i40e_vsi_setup_vectors(vsi);
9892                 if (ret)
9893                         goto err_msix;
9894
9895                 ret = i40e_alloc_rings(vsi);
9896                 if (ret)
9897                         goto err_rings;
9898
9899                 /* map all of the rings to the q_vectors */
9900                 i40e_vsi_map_rings_to_vectors(vsi);
9901
9902                 i40e_vsi_reset_stats(vsi);
9903                 break;
9904
9905         default:
9906                 /* no netdev or rings for the other VSI types */
9907                 break;
9908         }
9909
9910         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9911             (vsi->type == I40E_VSI_VMDQ2)) {
9912                 ret = i40e_vsi_config_rss(vsi);
9913         }
9914         return vsi;
9915
9916 err_rings:
9917         i40e_vsi_free_q_vectors(vsi);
9918 err_msix:
9919         if (vsi->netdev_registered) {
9920                 vsi->netdev_registered = false;
9921                 unregister_netdev(vsi->netdev);
9922                 free_netdev(vsi->netdev);
9923                 vsi->netdev = NULL;
9924         }
9925 err_netdev:
9926         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9927 err_vsi:
9928         i40e_vsi_clear(vsi);
9929 err_alloc:
9930         return NULL;
9931 }
9932
9933 /**
9934  * i40e_veb_get_bw_info - Query VEB BW information
9935  * @veb: the veb to query
9936  *
9937  * Query the Tx scheduler BW configuration data for given VEB
9938  **/
9939 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9940 {
9941         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9942         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9943         struct i40e_pf *pf = veb->pf;
9944         struct i40e_hw *hw = &pf->hw;
9945         u32 tc_bw_max;
9946         int ret = 0;
9947         int i;
9948
9949         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9950                                                   &bw_data, NULL);
9951         if (ret) {
9952                 dev_info(&pf->pdev->dev,
9953                          "query veb bw config failed, err %s aq_err %s\n",
9954                          i40e_stat_str(&pf->hw, ret),
9955                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9956                 goto out;
9957         }
9958
9959         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9960                                                    &ets_data, NULL);
9961         if (ret) {
9962                 dev_info(&pf->pdev->dev,
9963                          "query veb bw ets config failed, err %s aq_err %s\n",
9964                          i40e_stat_str(&pf->hw, ret),
9965                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9966                 goto out;
9967         }
9968
9969         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9970         veb->bw_max_quanta = ets_data.tc_bw_max;
9971         veb->is_abs_credits = bw_data.absolute_credits_enable;
9972         veb->enabled_tc = ets_data.tc_valid_bits;
9973         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9974                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9975         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9976                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9977                 veb->bw_tc_limit_credits[i] =
9978                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
9979                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9980         }
9981
9982 out:
9983         return ret;
9984 }
9985
9986 /**
9987  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9988  * @pf: board private structure
9989  *
9990  * On error: returns error code (negative)
9991  * On success: returns vsi index in PF (positive)
9992  **/
9993 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9994 {
9995         int ret = -ENOENT;
9996         struct i40e_veb *veb;
9997         int i;
9998
9999         /* Need to protect the allocation of switch elements at the PF level */
10000         mutex_lock(&pf->switch_mutex);
10001
10002         /* VEB list may be fragmented if VEB creation/destruction has
10003          * been happening.  We can afford to do a quick scan to look
10004          * for any free slots in the list.
10005          *
10006          * find next empty veb slot, looping back around if necessary
10007          */
10008         i = 0;
10009         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10010                 i++;
10011         if (i >= I40E_MAX_VEB) {
10012                 ret = -ENOMEM;
10013                 goto err_alloc_veb;  /* out of VEB slots! */
10014         }
10015
10016         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10017         if (!veb) {
10018                 ret = -ENOMEM;
10019                 goto err_alloc_veb;
10020         }
10021         veb->pf = pf;
10022         veb->idx = i;
10023         veb->enabled_tc = 1;
10024
10025         pf->veb[i] = veb;
10026         ret = i;
10027 err_alloc_veb:
10028         mutex_unlock(&pf->switch_mutex);
10029         return ret;
10030 }
10031
10032 /**
10033  * i40e_switch_branch_release - Delete a branch of the switch tree
10034  * @branch: where to start deleting
10035  *
10036  * This uses recursion to find the tips of the branch to be
10037  * removed, deleting until we get back to and can delete this VEB.
10038  **/
10039 static void i40e_switch_branch_release(struct i40e_veb *branch)
10040 {
10041         struct i40e_pf *pf = branch->pf;
10042         u16 branch_seid = branch->seid;
10043         u16 veb_idx = branch->idx;
10044         int i;
10045
10046         /* release any VEBs on this VEB - RECURSION */
10047         for (i = 0; i < I40E_MAX_VEB; i++) {
10048                 if (!pf->veb[i])
10049                         continue;
10050                 if (pf->veb[i]->uplink_seid == branch->seid)
10051                         i40e_switch_branch_release(pf->veb[i]);
10052         }
10053
10054         /* Release the VSIs on this VEB, but not the owner VSI.
10055          *
10056          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10057          *       the VEB itself, so don't use (*branch) after this loop.
10058          */
10059         for (i = 0; i < pf->num_alloc_vsi; i++) {
10060                 if (!pf->vsi[i])
10061                         continue;
10062                 if (pf->vsi[i]->uplink_seid == branch_seid &&
10063                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10064                         i40e_vsi_release(pf->vsi[i]);
10065                 }
10066         }
10067
10068         /* There's one corner case where the VEB might not have been
10069          * removed, so double check it here and remove it if needed.
10070          * This case happens if the veb was created from the debugfs
10071          * commands and no VSIs were added to it.
10072          */
10073         if (pf->veb[veb_idx])
10074                 i40e_veb_release(pf->veb[veb_idx]);
10075 }
10076
10077 /**
10078  * i40e_veb_clear - remove veb struct
10079  * @veb: the veb to remove
10080  **/
10081 static void i40e_veb_clear(struct i40e_veb *veb)
10082 {
10083         if (!veb)
10084                 return;
10085
10086         if (veb->pf) {
10087                 struct i40e_pf *pf = veb->pf;
10088
10089                 mutex_lock(&pf->switch_mutex);
10090                 if (pf->veb[veb->idx] == veb)
10091                         pf->veb[veb->idx] = NULL;
10092                 mutex_unlock(&pf->switch_mutex);
10093         }
10094
10095         kfree(veb);
10096 }
10097
10098 /**
10099  * i40e_veb_release - Delete a VEB and free its resources
10100  * @veb: the VEB being removed
10101  **/
10102 void i40e_veb_release(struct i40e_veb *veb)
10103 {
10104         struct i40e_vsi *vsi = NULL;
10105         struct i40e_pf *pf;
10106         int i, n = 0;
10107
10108         pf = veb->pf;
10109
10110         /* find the remaining VSI and check for extras */
10111         for (i = 0; i < pf->num_alloc_vsi; i++) {
10112                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10113                         n++;
10114                         vsi = pf->vsi[i];
10115                 }
10116         }
10117         if (n != 1) {
10118                 dev_info(&pf->pdev->dev,
10119                          "can't remove VEB %d with %d VSIs left\n",
10120                          veb->seid, n);
10121                 return;
10122         }
10123
10124         /* move the remaining VSI to uplink veb */
10125         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10126         if (veb->uplink_seid) {
10127                 vsi->uplink_seid = veb->uplink_seid;
10128                 if (veb->uplink_seid == pf->mac_seid)
10129                         vsi->veb_idx = I40E_NO_VEB;
10130                 else
10131                         vsi->veb_idx = veb->veb_idx;
10132         } else {
10133                 /* floating VEB */
10134                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10135                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10136         }
10137
10138         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10139         i40e_veb_clear(veb);
10140 }
10141
10142 /**
10143  * i40e_add_veb - create the VEB in the switch
10144  * @veb: the VEB to be instantiated
10145  * @vsi: the controlling VSI
10146  **/
10147 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10148 {
10149         struct i40e_pf *pf = veb->pf;
10150         bool is_default = veb->pf->cur_promisc;
10151         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10152         int ret;
10153
10154         /* get a VEB from the hardware */
10155         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10156                               veb->enabled_tc, is_default,
10157                               &veb->seid, enable_stats, NULL);
10158         if (ret) {
10159                 dev_info(&pf->pdev->dev,
10160                          "couldn't add VEB, err %s aq_err %s\n",
10161                          i40e_stat_str(&pf->hw, ret),
10162                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10163                 return -EPERM;
10164         }
10165
10166         /* get statistics counter */
10167         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10168                                          &veb->stats_idx, NULL, NULL, NULL);
10169         if (ret) {
10170                 dev_info(&pf->pdev->dev,
10171                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10172                          i40e_stat_str(&pf->hw, ret),
10173                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10174                 return -EPERM;
10175         }
10176         ret = i40e_veb_get_bw_info(veb);
10177         if (ret) {
10178                 dev_info(&pf->pdev->dev,
10179                          "couldn't get VEB bw info, err %s aq_err %s\n",
10180                          i40e_stat_str(&pf->hw, ret),
10181                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10182                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10183                 return -ENOENT;
10184         }
10185
10186         vsi->uplink_seid = veb->seid;
10187         vsi->veb_idx = veb->idx;
10188         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10189
10190         return 0;
10191 }
10192
10193 /**
10194  * i40e_veb_setup - Set up a VEB
10195  * @pf: board private structure
10196  * @flags: VEB setup flags
10197  * @uplink_seid: the switch element to link to
10198  * @vsi_seid: the initial VSI seid
10199  * @enabled_tc: Enabled TC bit-map
10200  *
10201  * This allocates the sw VEB structure and links it into the switch
10202  * It is possible and legal for this to be a duplicate of an already
10203  * existing VEB.  It is also possible for both uplink and vsi seids
10204  * to be zero, in order to create a floating VEB.
10205  *
10206  * Returns pointer to the successfully allocated VEB sw struct on
10207  * success, otherwise returns NULL on failure.
10208  **/
10209 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10210                                 u16 uplink_seid, u16 vsi_seid,
10211                                 u8 enabled_tc)
10212 {
10213         struct i40e_veb *veb, *uplink_veb = NULL;
10214         int vsi_idx, veb_idx;
10215         int ret;
10216
10217         /* if one seid is 0, the other must be 0 to create a floating relay */
10218         if ((uplink_seid == 0 || vsi_seid == 0) &&
10219             (uplink_seid + vsi_seid != 0)) {
10220                 dev_info(&pf->pdev->dev,
10221                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10222                          uplink_seid, vsi_seid);
10223                 return NULL;
10224         }
10225
10226         /* make sure there is such a vsi and uplink */
10227         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10228                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10229                         break;
10230         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10231                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10232                          vsi_seid);
10233                 return NULL;
10234         }
10235
10236         if (uplink_seid && uplink_seid != pf->mac_seid) {
10237                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10238                         if (pf->veb[veb_idx] &&
10239                             pf->veb[veb_idx]->seid == uplink_seid) {
10240                                 uplink_veb = pf->veb[veb_idx];
10241                                 break;
10242                         }
10243                 }
10244                 if (!uplink_veb) {
10245                         dev_info(&pf->pdev->dev,
10246                                  "uplink seid %d not found\n", uplink_seid);
10247                         return NULL;
10248                 }
10249         }
10250
10251         /* get veb sw struct */
10252         veb_idx = i40e_veb_mem_alloc(pf);
10253         if (veb_idx < 0)
10254                 goto err_alloc;
10255         veb = pf->veb[veb_idx];
10256         veb->flags = flags;
10257         veb->uplink_seid = uplink_seid;
10258         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10259         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10260
10261         /* create the VEB in the switch */
10262         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10263         if (ret)
10264                 goto err_veb;
10265         if (vsi_idx == pf->lan_vsi)
10266                 pf->lan_veb = veb->idx;
10267
10268         return veb;
10269
10270 err_veb:
10271         i40e_veb_clear(veb);
10272 err_alloc:
10273         return NULL;
10274 }
10275
10276 /**
10277  * i40e_setup_pf_switch_element - set PF vars based on switch type
10278  * @pf: board private structure
10279  * @ele: element we are building info from
10280  * @num_reported: total number of elements
10281  * @printconfig: should we print the contents
10282  *
10283  * helper function to assist in extracting a few useful SEID values.
10284  **/
10285 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10286                                 struct i40e_aqc_switch_config_element_resp *ele,
10287                                 u16 num_reported, bool printconfig)
10288 {
10289         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10290         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10291         u8 element_type = ele->element_type;
10292         u16 seid = le16_to_cpu(ele->seid);
10293
10294         if (printconfig)
10295                 dev_info(&pf->pdev->dev,
10296                          "type=%d seid=%d uplink=%d downlink=%d\n",
10297                          element_type, seid, uplink_seid, downlink_seid);
10298
10299         switch (element_type) {
10300         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10301                 pf->mac_seid = seid;
10302                 break;
10303         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10304                 /* Main VEB? */
10305                 if (uplink_seid != pf->mac_seid)
10306                         break;
10307                 if (pf->lan_veb == I40E_NO_VEB) {
10308                         int v;
10309
10310                         /* find existing or else empty VEB */
10311                         for (v = 0; v < I40E_MAX_VEB; v++) {
10312                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10313                                         pf->lan_veb = v;
10314                                         break;
10315                                 }
10316                         }
10317                         if (pf->lan_veb == I40E_NO_VEB) {
10318                                 v = i40e_veb_mem_alloc(pf);
10319                                 if (v < 0)
10320                                         break;
10321                                 pf->lan_veb = v;
10322                         }
10323                 }
10324
10325                 pf->veb[pf->lan_veb]->seid = seid;
10326                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10327                 pf->veb[pf->lan_veb]->pf = pf;
10328                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10329                 break;
10330         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10331                 if (num_reported != 1)
10332                         break;
10333                 /* This is immediately after a reset so we can assume this is
10334                  * the PF's VSI
10335                  */
10336                 pf->mac_seid = uplink_seid;
10337                 pf->pf_seid = downlink_seid;
10338                 pf->main_vsi_seid = seid;
10339                 if (printconfig)
10340                         dev_info(&pf->pdev->dev,
10341                                  "pf_seid=%d main_vsi_seid=%d\n",
10342                                  pf->pf_seid, pf->main_vsi_seid);
10343                 break;
10344         case I40E_SWITCH_ELEMENT_TYPE_PF:
10345         case I40E_SWITCH_ELEMENT_TYPE_VF:
10346         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10347         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10348         case I40E_SWITCH_ELEMENT_TYPE_PE:
10349         case I40E_SWITCH_ELEMENT_TYPE_PA:
10350                 /* ignore these for now */
10351                 break;
10352         default:
10353                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10354                          element_type, seid);
10355                 break;
10356         }
10357 }
10358
10359 /**
10360  * i40e_fetch_switch_configuration - Get switch config from firmware
10361  * @pf: board private structure
10362  * @printconfig: should we print the contents
10363  *
10364  * Get the current switch configuration from the device and
10365  * extract a few useful SEID values.
10366  **/
10367 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10368 {
10369         struct i40e_aqc_get_switch_config_resp *sw_config;
10370         u16 next_seid = 0;
10371         int ret = 0;
10372         u8 *aq_buf;
10373         int i;
10374
10375         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10376         if (!aq_buf)
10377                 return -ENOMEM;
10378
10379         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10380         do {
10381                 u16 num_reported, num_total;
10382
10383                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10384                                                 I40E_AQ_LARGE_BUF,
10385                                                 &next_seid, NULL);
10386                 if (ret) {
10387                         dev_info(&pf->pdev->dev,
10388                                  "get switch config failed err %s aq_err %s\n",
10389                                  i40e_stat_str(&pf->hw, ret),
10390                                  i40e_aq_str(&pf->hw,
10391                                              pf->hw.aq.asq_last_status));
10392                         kfree(aq_buf);
10393                         return -ENOENT;
10394                 }
10395
10396                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10397                 num_total = le16_to_cpu(sw_config->header.num_total);
10398
10399                 if (printconfig)
10400                         dev_info(&pf->pdev->dev,
10401                                  "header: %d reported %d total\n",
10402                                  num_reported, num_total);
10403
10404                 for (i = 0; i < num_reported; i++) {
10405                         struct i40e_aqc_switch_config_element_resp *ele =
10406                                 &sw_config->element[i];
10407
10408                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10409                                                      printconfig);
10410                 }
10411         } while (next_seid != 0);
10412
10413         kfree(aq_buf);
10414         return ret;
10415 }
10416
10417 /**
10418  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10419  * @pf: board private structure
10420  * @reinit: if the Main VSI needs to re-initialized.
10421  *
10422  * Returns 0 on success, negative value on failure
10423  **/
10424 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10425 {
10426         u16 flags = 0;
10427         int ret;
10428
10429         /* find out what's out there already */
10430         ret = i40e_fetch_switch_configuration(pf, false);
10431         if (ret) {
10432                 dev_info(&pf->pdev->dev,
10433                          "couldn't fetch switch config, err %s aq_err %s\n",
10434                          i40e_stat_str(&pf->hw, ret),
10435                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10436                 return ret;
10437         }
10438         i40e_pf_reset_stats(pf);
10439
10440         /* set the switch config bit for the whole device to
10441          * support limited promisc or true promisc
10442          * when user requests promisc. The default is limited
10443          * promisc.
10444         */
10445
10446         if ((pf->hw.pf_id == 0) &&
10447             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10448                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10449
10450         if (pf->hw.pf_id == 0) {
10451                 u16 valid_flags;
10452
10453                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10454                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10455                                                 NULL);
10456                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10457                         dev_info(&pf->pdev->dev,
10458                                  "couldn't set switch config bits, err %s aq_err %s\n",
10459                                  i40e_stat_str(&pf->hw, ret),
10460                                  i40e_aq_str(&pf->hw,
10461                                              pf->hw.aq.asq_last_status));
10462                         /* not a fatal problem, just keep going */
10463                 }
10464         }
10465
10466         /* first time setup */
10467         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10468                 struct i40e_vsi *vsi = NULL;
10469                 u16 uplink_seid;
10470
10471                 /* Set up the PF VSI associated with the PF's main VSI
10472                  * that is already in the HW switch
10473                  */
10474                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10475                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10476                 else
10477                         uplink_seid = pf->mac_seid;
10478                 if (pf->lan_vsi == I40E_NO_VSI)
10479                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10480                 else if (reinit)
10481                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10482                 if (!vsi) {
10483                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10484                         i40e_fdir_teardown(pf);
10485                         return -EAGAIN;
10486                 }
10487         } else {
10488                 /* force a reset of TC and queue layout configurations */
10489                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10490
10491                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10492                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10493                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10494         }
10495         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10496
10497         i40e_fdir_sb_setup(pf);
10498
10499         /* Setup static PF queue filter control settings */
10500         ret = i40e_setup_pf_filter_control(pf);
10501         if (ret) {
10502                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10503                          ret);
10504                 /* Failure here should not stop continuing other steps */
10505         }
10506
10507         /* enable RSS in the HW, even for only one queue, as the stack can use
10508          * the hash
10509          */
10510         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10511                 i40e_pf_config_rss(pf);
10512
10513         /* fill in link information and enable LSE reporting */
10514         i40e_update_link_info(&pf->hw);
10515         i40e_link_event(pf);
10516
10517         /* Initialize user-specific link properties */
10518         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10519                                   I40E_AQ_AN_COMPLETED) ? true : false);
10520
10521         i40e_ptp_init(pf);
10522
10523         return ret;
10524 }
10525
10526 /**
10527  * i40e_determine_queue_usage - Work out queue distribution
10528  * @pf: board private structure
10529  **/
10530 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10531 {
10532         int queues_left;
10533
10534         pf->num_lan_qps = 0;
10535 #ifdef I40E_FCOE
10536         pf->num_fcoe_qps = 0;
10537 #endif
10538
10539         /* Find the max queues to be put into basic use.  We'll always be
10540          * using TC0, whether or not DCB is running, and TC0 will get the
10541          * big RSS set.
10542          */
10543         queues_left = pf->hw.func_caps.num_tx_qp;
10544
10545         if ((queues_left == 1) ||
10546             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10547                 /* one qp for PF, no queues for anything else */
10548                 queues_left = 0;
10549                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10550
10551                 /* make sure all the fancies are disabled */
10552                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10553                                I40E_FLAG_IWARP_ENABLED  |
10554 #ifdef I40E_FCOE
10555                                I40E_FLAG_FCOE_ENABLED   |
10556 #endif
10557                                I40E_FLAG_FD_SB_ENABLED  |
10558                                I40E_FLAG_FD_ATR_ENABLED |
10559                                I40E_FLAG_DCB_CAPABLE    |
10560                                I40E_FLAG_SRIOV_ENABLED  |
10561                                I40E_FLAG_VMDQ_ENABLED);
10562         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10563                                   I40E_FLAG_FD_SB_ENABLED |
10564                                   I40E_FLAG_FD_ATR_ENABLED |
10565                                   I40E_FLAG_DCB_CAPABLE))) {
10566                 /* one qp for PF */
10567                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10568                 queues_left -= pf->num_lan_qps;
10569
10570                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10571                                I40E_FLAG_IWARP_ENABLED  |
10572 #ifdef I40E_FCOE
10573                                I40E_FLAG_FCOE_ENABLED   |
10574 #endif
10575                                I40E_FLAG_FD_SB_ENABLED  |
10576                                I40E_FLAG_FD_ATR_ENABLED |
10577                                I40E_FLAG_DCB_ENABLED    |
10578                                I40E_FLAG_VMDQ_ENABLED);
10579         } else {
10580                 /* Not enough queues for all TCs */
10581                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10582                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10583                         pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10584                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10585                 }
10586                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10587                                         num_online_cpus());
10588                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10589                                         pf->hw.func_caps.num_tx_qp);
10590
10591                 queues_left -= pf->num_lan_qps;
10592         }
10593
10594 #ifdef I40E_FCOE
10595         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10596                 if (I40E_DEFAULT_FCOE <= queues_left) {
10597                         pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10598                 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10599                         pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10600                 } else {
10601                         pf->num_fcoe_qps = 0;
10602                         pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10603                         dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10604                 }
10605
10606                 queues_left -= pf->num_fcoe_qps;
10607         }
10608
10609 #endif
10610         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10611                 if (queues_left > 1) {
10612                         queues_left -= 1; /* save 1 queue for FD */
10613                 } else {
10614                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10615                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10616                 }
10617         }
10618
10619         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10620             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10621                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10622                                         (queues_left / pf->num_vf_qps));
10623                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10624         }
10625
10626         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10627             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10628                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10629                                           (queues_left / pf->num_vmdq_qps));
10630                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10631         }
10632
10633         pf->queues_left = queues_left;
10634         dev_dbg(&pf->pdev->dev,
10635                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10636                 pf->hw.func_caps.num_tx_qp,
10637                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10638                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10639                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10640                 queues_left);
10641 #ifdef I40E_FCOE
10642         dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10643 #endif
10644 }
10645
10646 /**
10647  * i40e_setup_pf_filter_control - Setup PF static filter control
10648  * @pf: PF to be setup
10649  *
10650  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10651  * settings. If PE/FCoE are enabled then it will also set the per PF
10652  * based filter sizes required for them. It also enables Flow director,
10653  * ethertype and macvlan type filter settings for the pf.
10654  *
10655  * Returns 0 on success, negative on failure
10656  **/
10657 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10658 {
10659         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10660
10661         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10662
10663         /* Flow Director is enabled */
10664         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10665                 settings->enable_fdir = true;
10666
10667         /* Ethtype and MACVLAN filters enabled for PF */
10668         settings->enable_ethtype = true;
10669         settings->enable_macvlan = true;
10670
10671         if (i40e_set_filter_control(&pf->hw, settings))
10672                 return -ENOENT;
10673
10674         return 0;
10675 }
10676
10677 #define INFO_STRING_LEN 255
10678 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10679 static void i40e_print_features(struct i40e_pf *pf)
10680 {
10681         struct i40e_hw *hw = &pf->hw;
10682         char *buf;
10683         int i;
10684
10685         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10686         if (!buf)
10687                 return;
10688
10689         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10690 #ifdef CONFIG_PCI_IOV
10691         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10692 #endif
10693         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10694                       pf->hw.func_caps.num_vsis,
10695                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
10696         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10697                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10698         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10699                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10700         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10701                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10702                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10703         }
10704         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10705                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10706 #if IS_ENABLED(CONFIG_VXLAN)
10707         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10708 #endif
10709 #if IS_ENABLED(CONFIG_GENEVE)
10710         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10711 #endif
10712         if (pf->flags & I40E_FLAG_PTP)
10713                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10714 #ifdef I40E_FCOE
10715         if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10716                 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10717 #endif
10718         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10719                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10720         else
10721                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10722
10723         dev_info(&pf->pdev->dev, "%s\n", buf);
10724         kfree(buf);
10725         WARN_ON(i > INFO_STRING_LEN);
10726 }
10727
10728 /**
10729  * i40e_get_platform_mac_addr - get platform-specific MAC address
10730  *
10731  * @pdev: PCI device information struct
10732  * @pf: board private structure
10733  *
10734  * Look up the MAC address in Open Firmware  on systems that support it,
10735  * and use IDPROM on SPARC if no OF address is found. On return, the
10736  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10737  * has been selected.
10738  **/
10739 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10740 {
10741         pf->flags &= ~I40E_FLAG_PF_MAC;
10742         if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10743                 pf->flags |= I40E_FLAG_PF_MAC;
10744 }
10745
10746 /**
10747  * i40e_probe - Device initialization routine
10748  * @pdev: PCI device information struct
10749  * @ent: entry in i40e_pci_tbl
10750  *
10751  * i40e_probe initializes a PF identified by a pci_dev structure.
10752  * The OS initialization, configuring of the PF private structure,
10753  * and a hardware reset occur.
10754  *
10755  * Returns 0 on success, negative on failure
10756  **/
10757 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10758 {
10759         struct i40e_aq_get_phy_abilities_resp abilities;
10760         struct i40e_pf *pf;
10761         struct i40e_hw *hw;
10762         static u16 pfs_found;
10763         u16 wol_nvm_bits;
10764         u16 link_status;
10765         int err;
10766         u32 val;
10767         u32 i;
10768         u8 set_fc_aq_fail;
10769
10770         err = pci_enable_device_mem(pdev);
10771         if (err)
10772                 return err;
10773
10774         /* set up for high or low dma */
10775         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10776         if (err) {
10777                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10778                 if (err) {
10779                         dev_err(&pdev->dev,
10780                                 "DMA configuration failed: 0x%x\n", err);
10781                         goto err_dma;
10782                 }
10783         }
10784
10785         /* set up pci connections */
10786         err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10787                                            IORESOURCE_MEM), i40e_driver_name);
10788         if (err) {
10789                 dev_info(&pdev->dev,
10790                          "pci_request_selected_regions failed %d\n", err);
10791                 goto err_pci_reg;
10792         }
10793
10794         pci_enable_pcie_error_reporting(pdev);
10795         pci_set_master(pdev);
10796
10797         /* Now that we have a PCI connection, we need to do the
10798          * low level device setup.  This is primarily setting up
10799          * the Admin Queue structures and then querying for the
10800          * device's current profile information.
10801          */
10802         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10803         if (!pf) {
10804                 err = -ENOMEM;
10805                 goto err_pf_alloc;
10806         }
10807         pf->next_vsi = 0;
10808         pf->pdev = pdev;
10809         set_bit(__I40E_DOWN, &pf->state);
10810
10811         hw = &pf->hw;
10812         hw->back = pf;
10813
10814         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10815                                 I40E_MAX_CSR_SPACE);
10816
10817         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10818         if (!hw->hw_addr) {
10819                 err = -EIO;
10820                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10821                          (unsigned int)pci_resource_start(pdev, 0),
10822                          pf->ioremap_len, err);
10823                 goto err_ioremap;
10824         }
10825         hw->vendor_id = pdev->vendor;
10826         hw->device_id = pdev->device;
10827         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10828         hw->subsystem_vendor_id = pdev->subsystem_vendor;
10829         hw->subsystem_device_id = pdev->subsystem_device;
10830         hw->bus.device = PCI_SLOT(pdev->devfn);
10831         hw->bus.func = PCI_FUNC(pdev->devfn);
10832         pf->instance = pfs_found;
10833
10834         /* set up the locks for the AQ, do this only once in probe
10835          * and destroy them only once in remove
10836          */
10837         mutex_init(&hw->aq.asq_mutex);
10838         mutex_init(&hw->aq.arq_mutex);
10839
10840         if (debug != -1) {
10841                 pf->msg_enable = pf->hw.debug_mask;
10842                 pf->msg_enable = debug;
10843         }
10844
10845         /* do a special CORER for clearing PXE mode once at init */
10846         if (hw->revision_id == 0 &&
10847             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10848                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10849                 i40e_flush(hw);
10850                 msleep(200);
10851                 pf->corer_count++;
10852
10853                 i40e_clear_pxe_mode(hw);
10854         }
10855
10856         /* Reset here to make sure all is clean and to define PF 'n' */
10857         i40e_clear_hw(hw);
10858         err = i40e_pf_reset(hw);
10859         if (err) {
10860                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10861                 goto err_pf_reset;
10862         }
10863         pf->pfr_count++;
10864
10865         hw->aq.num_arq_entries = I40E_AQ_LEN;
10866         hw->aq.num_asq_entries = I40E_AQ_LEN;
10867         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10868         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10869         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10870
10871         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10872                  "%s-%s:misc",
10873                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10874
10875         err = i40e_init_shared_code(hw);
10876         if (err) {
10877                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10878                          err);
10879                 goto err_pf_reset;
10880         }
10881
10882         /* set up a default setting for link flow control */
10883         pf->hw.fc.requested_mode = I40E_FC_NONE;
10884
10885         err = i40e_init_adminq(hw);
10886         if (err) {
10887                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10888                         dev_info(&pdev->dev,
10889                                  "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");
10890                 else
10891                         dev_info(&pdev->dev,
10892                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10893
10894                 goto err_pf_reset;
10895         }
10896
10897         /* provide nvm, fw, api versions */
10898         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10899                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10900                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
10901                  i40e_nvm_version_str(hw));
10902
10903         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10904             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10905                 dev_info(&pdev->dev,
10906                          "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");
10907         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10908                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10909                 dev_info(&pdev->dev,
10910                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10911
10912         i40e_verify_eeprom(pf);
10913
10914         /* Rev 0 hardware was never productized */
10915         if (hw->revision_id < 1)
10916                 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");
10917
10918         i40e_clear_pxe_mode(hw);
10919         err = i40e_get_capabilities(pf);
10920         if (err)
10921                 goto err_adminq_setup;
10922
10923         err = i40e_sw_init(pf);
10924         if (err) {
10925                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10926                 goto err_sw_init;
10927         }
10928
10929         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10930                                 hw->func_caps.num_rx_qp,
10931                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10932         if (err) {
10933                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10934                 goto err_init_lan_hmc;
10935         }
10936
10937         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10938         if (err) {
10939                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10940                 err = -ENOENT;
10941                 goto err_configure_lan_hmc;
10942         }
10943
10944         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10945          * Ignore error return codes because if it was already disabled via
10946          * hardware settings this will fail
10947          */
10948         if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10949                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10950                 i40e_aq_stop_lldp(hw, true, NULL);
10951         }
10952
10953         i40e_get_mac_addr(hw, hw->mac.addr);
10954         /* allow a platform config to override the HW addr */
10955         i40e_get_platform_mac_addr(pdev, pf);
10956         if (!is_valid_ether_addr(hw->mac.addr)) {
10957                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10958                 err = -EIO;
10959                 goto err_mac_addr;
10960         }
10961         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10962         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10963         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10964         if (is_valid_ether_addr(hw->mac.port_addr))
10965                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10966 #ifdef I40E_FCOE
10967         err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10968         if (err)
10969                 dev_info(&pdev->dev,
10970                          "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10971         if (!is_valid_ether_addr(hw->mac.san_addr)) {
10972                 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10973                          hw->mac.san_addr);
10974                 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10975         }
10976         dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10977 #endif /* I40E_FCOE */
10978
10979         pci_set_drvdata(pdev, pf);
10980         pci_save_state(pdev);
10981 #ifdef CONFIG_I40E_DCB
10982         err = i40e_init_pf_dcb(pf);
10983         if (err) {
10984                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10985                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10986                 /* Continue without DCB enabled */
10987         }
10988 #endif /* CONFIG_I40E_DCB */
10989
10990         /* set up periodic task facility */
10991         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10992         pf->service_timer_period = HZ;
10993
10994         INIT_WORK(&pf->service_task, i40e_service_task);
10995         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10996         pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10997
10998         /* NVM bit on means WoL disabled for the port */
10999         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11000         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11001                 pf->wol_en = false;
11002         else
11003                 pf->wol_en = true;
11004         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11005
11006         /* set up the main switch operations */
11007         i40e_determine_queue_usage(pf);
11008         err = i40e_init_interrupt_scheme(pf);
11009         if (err)
11010                 goto err_switch_setup;
11011
11012         /* The number of VSIs reported by the FW is the minimum guaranteed
11013          * to us; HW supports far more and we share the remaining pool with
11014          * the other PFs. We allocate space for more than the guarantee with
11015          * the understanding that we might not get them all later.
11016          */
11017         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11018                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11019         else
11020                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11021
11022         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11023         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11024                           GFP_KERNEL);
11025         if (!pf->vsi) {
11026                 err = -ENOMEM;
11027                 goto err_switch_setup;
11028         }
11029
11030 #ifdef CONFIG_PCI_IOV
11031         /* prep for VF support */
11032         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11033             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11034             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11035                 if (pci_num_vf(pdev))
11036                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11037         }
11038 #endif
11039         err = i40e_setup_pf_switch(pf, false);
11040         if (err) {
11041                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11042                 goto err_vsis;
11043         }
11044
11045         /* Make sure flow control is set according to current settings */
11046         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11047         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11048                 dev_dbg(&pf->pdev->dev,
11049                         "Set fc with err %s aq_err %s on get_phy_cap\n",
11050                         i40e_stat_str(hw, err),
11051                         i40e_aq_str(hw, hw->aq.asq_last_status));
11052         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11053                 dev_dbg(&pf->pdev->dev,
11054                         "Set fc with err %s aq_err %s on set_phy_config\n",
11055                         i40e_stat_str(hw, err),
11056                         i40e_aq_str(hw, hw->aq.asq_last_status));
11057         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11058                 dev_dbg(&pf->pdev->dev,
11059                         "Set fc with err %s aq_err %s on get_link_info\n",
11060                         i40e_stat_str(hw, err),
11061                         i40e_aq_str(hw, hw->aq.asq_last_status));
11062
11063         /* if FDIR VSI was set up, start it now */
11064         for (i = 0; i < pf->num_alloc_vsi; i++) {
11065                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11066                         i40e_vsi_open(pf->vsi[i]);
11067                         break;
11068                 }
11069         }
11070
11071         /* The driver only wants link up/down and module qualification
11072          * reports from firmware.  Note the negative logic.
11073          */
11074         err = i40e_aq_set_phy_int_mask(&pf->hw,
11075                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
11076                                          I40E_AQ_EVENT_MEDIA_NA |
11077                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11078         if (err)
11079                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11080                          i40e_stat_str(&pf->hw, err),
11081                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11082
11083         /* Reconfigure hardware for allowing smaller MSS in the case
11084          * of TSO, so that we avoid the MDD being fired and causing
11085          * a reset in the case of small MSS+TSO.
11086          */
11087         val = rd32(hw, I40E_REG_MSS);
11088         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11089                 val &= ~I40E_REG_MSS_MIN_MASK;
11090                 val |= I40E_64BYTE_MSS;
11091                 wr32(hw, I40E_REG_MSS, val);
11092         }
11093
11094         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11095                 msleep(75);
11096                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11097                 if (err)
11098                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11099                                  i40e_stat_str(&pf->hw, err),
11100                                  i40e_aq_str(&pf->hw,
11101                                              pf->hw.aq.asq_last_status));
11102         }
11103         /* The main driver is (mostly) up and happy. We need to set this state
11104          * before setting up the misc vector or we get a race and the vector
11105          * ends up disabled forever.
11106          */
11107         clear_bit(__I40E_DOWN, &pf->state);
11108
11109         /* In case of MSIX we are going to setup the misc vector right here
11110          * to handle admin queue events etc. In case of legacy and MSI
11111          * the misc functionality and queue processing is combined in
11112          * the same vector and that gets setup at open.
11113          */
11114         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11115                 err = i40e_setup_misc_vector(pf);
11116                 if (err) {
11117                         dev_info(&pdev->dev,
11118                                  "setup of misc vector failed: %d\n", err);
11119                         goto err_vsis;
11120                 }
11121         }
11122
11123 #ifdef CONFIG_PCI_IOV
11124         /* prep for VF support */
11125         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11126             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11127             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11128                 /* disable link interrupts for VFs */
11129                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11130                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11131                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11132                 i40e_flush(hw);
11133
11134                 if (pci_num_vf(pdev)) {
11135                         dev_info(&pdev->dev,
11136                                  "Active VFs found, allocating resources.\n");
11137                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11138                         if (err)
11139                                 dev_info(&pdev->dev,
11140                                          "Error %d allocating resources for existing VFs\n",
11141                                          err);
11142                 }
11143         }
11144 #endif /* CONFIG_PCI_IOV */
11145
11146         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11147                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11148                                                       pf->num_iwarp_msix,
11149                                                       I40E_IWARP_IRQ_PILE_ID);
11150                 if (pf->iwarp_base_vector < 0) {
11151                         dev_info(&pdev->dev,
11152                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11153                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11154                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11155                 }
11156         }
11157
11158         i40e_dbg_pf_init(pf);
11159
11160         /* tell the firmware that we're starting */
11161         i40e_send_version(pf);
11162
11163         /* since everything's happy, start the service_task timer */
11164         mod_timer(&pf->service_timer,
11165                   round_jiffies(jiffies + pf->service_timer_period));
11166
11167         /* add this PF to client device list and launch a client service task */
11168         err = i40e_lan_add_device(pf);
11169         if (err)
11170                 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11171                          err);
11172
11173 #ifdef I40E_FCOE
11174         /* create FCoE interface */
11175         i40e_fcoe_vsi_setup(pf);
11176
11177 #endif
11178 #define PCI_SPEED_SIZE 8
11179 #define PCI_WIDTH_SIZE 8
11180         /* Devices on the IOSF bus do not have this information
11181          * and will report PCI Gen 1 x 1 by default so don't bother
11182          * checking them.
11183          */
11184         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11185                 char speed[PCI_SPEED_SIZE] = "Unknown";
11186                 char width[PCI_WIDTH_SIZE] = "Unknown";
11187
11188                 /* Get the negotiated link width and speed from PCI config
11189                  * space
11190                  */
11191                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11192                                           &link_status);
11193
11194                 i40e_set_pci_config_data(hw, link_status);
11195
11196                 switch (hw->bus.speed) {
11197                 case i40e_bus_speed_8000:
11198                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11199                 case i40e_bus_speed_5000:
11200                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11201                 case i40e_bus_speed_2500:
11202                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11203                 default:
11204                         break;
11205                 }
11206                 switch (hw->bus.width) {
11207                 case i40e_bus_width_pcie_x8:
11208                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11209                 case i40e_bus_width_pcie_x4:
11210                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11211                 case i40e_bus_width_pcie_x2:
11212                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11213                 case i40e_bus_width_pcie_x1:
11214                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11215                 default:
11216                         break;
11217                 }
11218
11219                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11220                          speed, width);
11221
11222                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11223                     hw->bus.speed < i40e_bus_speed_8000) {
11224                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11225                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11226                 }
11227         }
11228
11229         /* get the requested speeds from the fw */
11230         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11231         if (err)
11232                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11233                         i40e_stat_str(&pf->hw, err),
11234                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11235         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11236
11237         /* get the supported phy types from the fw */
11238         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11239         if (err)
11240                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11241                         i40e_stat_str(&pf->hw, err),
11242                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11243         pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11244
11245         /* Add a filter to drop all Flow control frames from any VSI from being
11246          * transmitted. By doing so we stop a malicious VF from sending out
11247          * PAUSE or PFC frames and potentially controlling traffic for other
11248          * PF/VF VSIs.
11249          * The FW can still send Flow control frames if enabled.
11250          */
11251         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11252                                                        pf->main_vsi_seid);
11253
11254         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11255             (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11256                 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11257
11258         /* print a string summarizing features */
11259         i40e_print_features(pf);
11260
11261         return 0;
11262
11263         /* Unwind what we've done if something failed in the setup */
11264 err_vsis:
11265         set_bit(__I40E_DOWN, &pf->state);
11266         i40e_clear_interrupt_scheme(pf);
11267         kfree(pf->vsi);
11268 err_switch_setup:
11269         i40e_reset_interrupt_capability(pf);
11270         del_timer_sync(&pf->service_timer);
11271 err_mac_addr:
11272 err_configure_lan_hmc:
11273         (void)i40e_shutdown_lan_hmc(hw);
11274 err_init_lan_hmc:
11275         kfree(pf->qp_pile);
11276 err_sw_init:
11277 err_adminq_setup:
11278 err_pf_reset:
11279         iounmap(hw->hw_addr);
11280 err_ioremap:
11281         kfree(pf);
11282 err_pf_alloc:
11283         pci_disable_pcie_error_reporting(pdev);
11284         pci_release_selected_regions(pdev,
11285                                      pci_select_bars(pdev, IORESOURCE_MEM));
11286 err_pci_reg:
11287 err_dma:
11288         pci_disable_device(pdev);
11289         return err;
11290 }
11291
11292 /**
11293  * i40e_remove - Device removal routine
11294  * @pdev: PCI device information struct
11295  *
11296  * i40e_remove is called by the PCI subsystem to alert the driver
11297  * that is should release a PCI device.  This could be caused by a
11298  * Hot-Plug event, or because the driver is going to be removed from
11299  * memory.
11300  **/
11301 static void i40e_remove(struct pci_dev *pdev)
11302 {
11303         struct i40e_pf *pf = pci_get_drvdata(pdev);
11304         struct i40e_hw *hw = &pf->hw;
11305         i40e_status ret_code;
11306         int i;
11307
11308         i40e_dbg_pf_exit(pf);
11309
11310         i40e_ptp_stop(pf);
11311
11312         /* Disable RSS in hw */
11313         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11314         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11315
11316         /* no more scheduling of any task */
11317         set_bit(__I40E_SUSPENDED, &pf->state);
11318         set_bit(__I40E_DOWN, &pf->state);
11319         if (pf->service_timer.data)
11320                 del_timer_sync(&pf->service_timer);
11321         if (pf->service_task.func)
11322                 cancel_work_sync(&pf->service_task);
11323
11324         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11325                 i40e_free_vfs(pf);
11326                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11327         }
11328
11329         i40e_fdir_teardown(pf);
11330
11331         /* If there is a switch structure or any orphans, remove them.
11332          * This will leave only the PF's VSI remaining.
11333          */
11334         for (i = 0; i < I40E_MAX_VEB; i++) {
11335                 if (!pf->veb[i])
11336                         continue;
11337
11338                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11339                     pf->veb[i]->uplink_seid == 0)
11340                         i40e_switch_branch_release(pf->veb[i]);
11341         }
11342
11343         /* Now we can shutdown the PF's VSI, just before we kill
11344          * adminq and hmc.
11345          */
11346         if (pf->vsi[pf->lan_vsi])
11347                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11348
11349         /* remove attached clients */
11350         ret_code = i40e_lan_del_device(pf);
11351         if (ret_code) {
11352                 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11353                          ret_code);
11354         }
11355
11356         /* shutdown and destroy the HMC */
11357         if (hw->hmc.hmc_obj) {
11358                 ret_code = i40e_shutdown_lan_hmc(hw);
11359                 if (ret_code)
11360                         dev_warn(&pdev->dev,
11361                                  "Failed to destroy the HMC resources: %d\n",
11362                                  ret_code);
11363         }
11364
11365         /* shutdown the adminq */
11366         ret_code = i40e_shutdown_adminq(hw);
11367         if (ret_code)
11368                 dev_warn(&pdev->dev,
11369                          "Failed to destroy the Admin Queue resources: %d\n",
11370                          ret_code);
11371
11372         /* destroy the locks only once, here */
11373         mutex_destroy(&hw->aq.arq_mutex);
11374         mutex_destroy(&hw->aq.asq_mutex);
11375
11376         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11377         i40e_clear_interrupt_scheme(pf);
11378         for (i = 0; i < pf->num_alloc_vsi; i++) {
11379                 if (pf->vsi[i]) {
11380                         i40e_vsi_clear_rings(pf->vsi[i]);
11381                         i40e_vsi_clear(pf->vsi[i]);
11382                         pf->vsi[i] = NULL;
11383                 }
11384         }
11385
11386         for (i = 0; i < I40E_MAX_VEB; i++) {
11387                 kfree(pf->veb[i]);
11388                 pf->veb[i] = NULL;
11389         }
11390
11391         kfree(pf->qp_pile);
11392         kfree(pf->vsi);
11393
11394         iounmap(hw->hw_addr);
11395         kfree(pf);
11396         pci_release_selected_regions(pdev,
11397                                      pci_select_bars(pdev, IORESOURCE_MEM));
11398
11399         pci_disable_pcie_error_reporting(pdev);
11400         pci_disable_device(pdev);
11401 }
11402
11403 /**
11404  * i40e_pci_error_detected - warning that something funky happened in PCI land
11405  * @pdev: PCI device information struct
11406  *
11407  * Called to warn that something happened and the error handling steps
11408  * are in progress.  Allows the driver to quiesce things, be ready for
11409  * remediation.
11410  **/
11411 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11412                                                 enum pci_channel_state error)
11413 {
11414         struct i40e_pf *pf = pci_get_drvdata(pdev);
11415
11416         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11417
11418         /* shutdown all operations */
11419         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11420                 rtnl_lock();
11421                 i40e_prep_for_reset(pf);
11422                 rtnl_unlock();
11423         }
11424
11425         /* Request a slot reset */
11426         return PCI_ERS_RESULT_NEED_RESET;
11427 }
11428
11429 /**
11430  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11431  * @pdev: PCI device information struct
11432  *
11433  * Called to find if the driver can work with the device now that
11434  * the pci slot has been reset.  If a basic connection seems good
11435  * (registers are readable and have sane content) then return a
11436  * happy little PCI_ERS_RESULT_xxx.
11437  **/
11438 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11439 {
11440         struct i40e_pf *pf = pci_get_drvdata(pdev);
11441         pci_ers_result_t result;
11442         int err;
11443         u32 reg;
11444
11445         dev_dbg(&pdev->dev, "%s\n", __func__);
11446         if (pci_enable_device_mem(pdev)) {
11447                 dev_info(&pdev->dev,
11448                          "Cannot re-enable PCI device after reset.\n");
11449                 result = PCI_ERS_RESULT_DISCONNECT;
11450         } else {
11451                 pci_set_master(pdev);
11452                 pci_restore_state(pdev);
11453                 pci_save_state(pdev);
11454                 pci_wake_from_d3(pdev, false);
11455
11456                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11457                 if (reg == 0)
11458                         result = PCI_ERS_RESULT_RECOVERED;
11459                 else
11460                         result = PCI_ERS_RESULT_DISCONNECT;
11461         }
11462
11463         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11464         if (err) {
11465                 dev_info(&pdev->dev,
11466                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11467                          err);
11468                 /* non-fatal, continue */
11469         }
11470
11471         return result;
11472 }
11473
11474 /**
11475  * i40e_pci_error_resume - restart operations after PCI error recovery
11476  * @pdev: PCI device information struct
11477  *
11478  * Called to allow the driver to bring things back up after PCI error
11479  * and/or reset recovery has finished.
11480  **/
11481 static void i40e_pci_error_resume(struct pci_dev *pdev)
11482 {
11483         struct i40e_pf *pf = pci_get_drvdata(pdev);
11484
11485         dev_dbg(&pdev->dev, "%s\n", __func__);
11486         if (test_bit(__I40E_SUSPENDED, &pf->state))
11487                 return;
11488
11489         rtnl_lock();
11490         i40e_handle_reset_warning(pf);
11491         rtnl_unlock();
11492 }
11493
11494 /**
11495  * i40e_shutdown - PCI callback for shutting down
11496  * @pdev: PCI device information struct
11497  **/
11498 static void i40e_shutdown(struct pci_dev *pdev)
11499 {
11500         struct i40e_pf *pf = pci_get_drvdata(pdev);
11501         struct i40e_hw *hw = &pf->hw;
11502
11503         set_bit(__I40E_SUSPENDED, &pf->state);
11504         set_bit(__I40E_DOWN, &pf->state);
11505         rtnl_lock();
11506         i40e_prep_for_reset(pf);
11507         rtnl_unlock();
11508
11509         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11510         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11511
11512         del_timer_sync(&pf->service_timer);
11513         cancel_work_sync(&pf->service_task);
11514         i40e_fdir_teardown(pf);
11515
11516         rtnl_lock();
11517         i40e_prep_for_reset(pf);
11518         rtnl_unlock();
11519
11520         wr32(hw, I40E_PFPM_APM,
11521              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11522         wr32(hw, I40E_PFPM_WUFC,
11523              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11524
11525         i40e_clear_interrupt_scheme(pf);
11526
11527         if (system_state == SYSTEM_POWER_OFF) {
11528                 pci_wake_from_d3(pdev, pf->wol_en);
11529                 pci_set_power_state(pdev, PCI_D3hot);
11530         }
11531 }
11532
11533 #ifdef CONFIG_PM
11534 /**
11535  * i40e_suspend - PCI callback for moving to D3
11536  * @pdev: PCI device information struct
11537  **/
11538 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11539 {
11540         struct i40e_pf *pf = pci_get_drvdata(pdev);
11541         struct i40e_hw *hw = &pf->hw;
11542
11543         set_bit(__I40E_SUSPENDED, &pf->state);
11544         set_bit(__I40E_DOWN, &pf->state);
11545
11546         rtnl_lock();
11547         i40e_prep_for_reset(pf);
11548         rtnl_unlock();
11549
11550         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11551         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11552
11553         pci_wake_from_d3(pdev, pf->wol_en);
11554         pci_set_power_state(pdev, PCI_D3hot);
11555
11556         return 0;
11557 }
11558
11559 /**
11560  * i40e_resume - PCI callback for waking up from D3
11561  * @pdev: PCI device information struct
11562  **/
11563 static int i40e_resume(struct pci_dev *pdev)
11564 {
11565         struct i40e_pf *pf = pci_get_drvdata(pdev);
11566         u32 err;
11567
11568         pci_set_power_state(pdev, PCI_D0);
11569         pci_restore_state(pdev);
11570         /* pci_restore_state() clears dev->state_saves, so
11571          * call pci_save_state() again to restore it.
11572          */
11573         pci_save_state(pdev);
11574
11575         err = pci_enable_device_mem(pdev);
11576         if (err) {
11577                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11578                 return err;
11579         }
11580         pci_set_master(pdev);
11581
11582         /* no wakeup events while running */
11583         pci_wake_from_d3(pdev, false);
11584
11585         /* handling the reset will rebuild the device state */
11586         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11587                 clear_bit(__I40E_DOWN, &pf->state);
11588                 rtnl_lock();
11589                 i40e_reset_and_rebuild(pf, false);
11590                 rtnl_unlock();
11591         }
11592
11593         return 0;
11594 }
11595
11596 #endif
11597 static const struct pci_error_handlers i40e_err_handler = {
11598         .error_detected = i40e_pci_error_detected,
11599         .slot_reset = i40e_pci_error_slot_reset,
11600         .resume = i40e_pci_error_resume,
11601 };
11602
11603 static struct pci_driver i40e_driver = {
11604         .name     = i40e_driver_name,
11605         .id_table = i40e_pci_tbl,
11606         .probe    = i40e_probe,
11607         .remove   = i40e_remove,
11608 #ifdef CONFIG_PM
11609         .suspend  = i40e_suspend,
11610         .resume   = i40e_resume,
11611 #endif
11612         .shutdown = i40e_shutdown,
11613         .err_handler = &i40e_err_handler,
11614         .sriov_configure = i40e_pci_sriov_configure,
11615 };
11616
11617 /**
11618  * i40e_init_module - Driver registration routine
11619  *
11620  * i40e_init_module is the first routine called when the driver is
11621  * loaded. All it does is register with the PCI subsystem.
11622  **/
11623 static int __init i40e_init_module(void)
11624 {
11625         pr_info("%s: %s - version %s\n", i40e_driver_name,
11626                 i40e_driver_string, i40e_driver_version_str);
11627         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11628
11629         /* we will see if single thread per module is enough for now,
11630          * it can't be any worse than using the system workqueue which
11631          * was already single threaded
11632          */
11633         i40e_wq = create_singlethread_workqueue(i40e_driver_name);
11634         if (!i40e_wq) {
11635                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11636                 return -ENOMEM;
11637         }
11638
11639         i40e_dbg_init();
11640         return pci_register_driver(&i40e_driver);
11641 }
11642 module_init(i40e_init_module);
11643
11644 /**
11645  * i40e_exit_module - Driver exit cleanup routine
11646  *
11647  * i40e_exit_module is called just before the driver is removed
11648  * from memory.
11649  **/
11650 static void __exit i40e_exit_module(void)
11651 {
11652         pci_unregister_driver(&i40e_driver);
11653         destroy_workqueue(i40e_wq);
11654         i40e_dbg_exit();
11655 }
11656 module_exit(i40e_exit_module);