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

   1 /*******************************************************************************
   2  *
   3  * Intel Ethernet Controller XL710 Family Linux Driver
   4  * Copyright(c) 2013 - 2016 Intel Corporation.
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms and conditions of the GNU General Public License,
   8  * version 2, as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope it will be useful, but WITHOUT
  11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13  * more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
  17  *
  18  * The full GNU General Public License is included in this distribution in
  19  * the file called "COPYING".
  20  *
  21  * Contact Information:
  22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
  24  *
  25  ******************************************************************************/
  26
  27 #include <linux/etherdevice.h>
  28 #include <linux/of_net.h>
  29 #include <linux/pci.h>
  30
  31 /* Local includes */
  32 #include "i40e.h"
  33 #include "i40e_diag.h"
  34 #include <net/udp_tunnel.h>
  35 /* All i40e tracepoints are defined by the include below, which
  36  * must be included exactly once across the whole kernel with
  37  * CREATE_TRACE_POINTS defined
  38  */
  39 #define CREATE_TRACE_POINTS
  40 #include "i40e_trace.h"
  41
  42 const char i40e_driver_name[] = "i40e";
  43 static const char i40e_driver_string[] =
  44                         "Intel(R) Ethernet Connection XL710 Network Driver";
  45
  46 #define DRV_KERN "-k"
  47
  48 #define DRV_VERSION_MAJOR 2
  49 #define DRV_VERSION_MINOR 1
  50 #define DRV_VERSION_BUILD 7
  51 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  52              __stringify(DRV_VERSION_MINOR) "." \
  53              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  54 const char i40e_driver_version_str[] = DRV_VERSION;
  55 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  56
  57 /* a bit of forward declarations */
  58 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  59 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired);
  60 static int i40e_add_vsi(struct i40e_vsi *vsi);
  61 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  62 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  63 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  64 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  65 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  66 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired);
  67 static int i40e_reset(struct i40e_pf *pf);
  68 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired);
  69 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  70 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  71
  72 /* i40e_pci_tbl - PCI Device ID Table
  73  *
  74  * Last entry must be all 0s
  75  *
  76  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  77  *   Class, Class Mask, private data (not used) }
  78  */
  79 static const struct pci_device_id i40e_pci_tbl[] = {
  80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
  90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
  91         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  92         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  93         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  94         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
  95         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  96         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  97         {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_B), 0},
  98         {PCI_VDEVICE(INTEL, I40E_DEV_ID_25G_SFP28), 0},
  99         /* required last entry */
 100         {0, }
 101 };
 102 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
 103
 104 #define I40E_MAX_VF_COUNT 128
 105 static int debug = -1;
 106 module_param(debug, uint, 0);
 107 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all), Debug mask (0x8XXXXXXX)");
 108
 109 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 110 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 111 MODULE_LICENSE("GPL");
 112 MODULE_VERSION(DRV_VERSION);
 113
 114 static struct workqueue_struct *i40e_wq;
 115
 116 /**
 117  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 118  * @hw:   pointer to the HW structure
 119  * @mem:  ptr to mem struct to fill out
 120  * @size: size of memory requested
 121  * @alignment: what to align the allocation to
 122  **/
 123 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 124                             u64 size, u32 alignment)
 125 {
 126         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 127
 128         mem->size = ALIGN(size, alignment);
 129         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 130                                       &mem->pa, GFP_KERNEL);
 131         if (!mem->va)
 132                 return -ENOMEM;
 133
 134         return 0;
 135 }
 136
 137 /**
 138  * i40e_free_dma_mem_d - OS specific memory free for shared code
 139  * @hw:   pointer to the HW structure
 140  * @mem:  ptr to mem struct to free
 141  **/
 142 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 143 {
 144         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 145
 146         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 147         mem->va = NULL;
 148         mem->pa = 0;
 149         mem->size = 0;
 150
 151         return 0;
 152 }
 153
 154 /**
 155  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 156  * @hw:   pointer to the HW structure
 157  * @mem:  ptr to mem struct to fill out
 158  * @size: size of memory requested
 159  **/
 160 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 161                              u32 size)
 162 {
 163         mem->size = size;
 164         mem->va = kzalloc(size, GFP_KERNEL);
 165
 166         if (!mem->va)
 167                 return -ENOMEM;
 168
 169         return 0;
 170 }
 171
 172 /**
 173  * i40e_free_virt_mem_d - OS specific memory free for shared code
 174  * @hw:   pointer to the HW structure
 175  * @mem:  ptr to mem struct to free
 176  **/
 177 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 178 {
 179         /* it's ok to kfree a NULL pointer */
 180         kfree(mem->va);
 181         mem->va = NULL;
 182         mem->size = 0;
 183
 184         return 0;
 185 }
 186
 187 /**
 188  * i40e_get_lump - find a lump of free generic resource
 189  * @pf: board private structure
 190  * @pile: the pile of resource to search
 191  * @needed: the number of items needed
 192  * @id: an owner id to stick on the items assigned
 193  *
 194  * Returns the base item index of the lump, or negative for error
 195  *
 196  * The search_hint trick and lack of advanced fit-finding only work
 197  * because we're highly likely to have all the same size lump requests.
 198  * Linear search time and any fragmentation should be minimal.
 199  **/
 200 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 201                          u16 needed, u16 id)
 202 {
 203         int ret = -ENOMEM;
 204         int i, j;
 205
 206         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 207                 dev_info(&pf->pdev->dev,
 208                          "param err: pile=%p needed=%d id=0x%04x\n",
 209                          pile, needed, id);
 210                 return -EINVAL;
 211         }
 212
 213         /* start the linear search with an imperfect hint */
 214         i = pile->search_hint;
 215         while (i < pile->num_entries) {
 216                 /* skip already allocated entries */
 217                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 218                         i++;
 219                         continue;
 220                 }
 221
 222                 /* do we have enough in this lump? */
 223                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 224                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 225                                 break;
 226                 }
 227
 228                 if (j == needed) {
 229                         /* there was enough, so assign it to the requestor */
 230                         for (j = 0; j < needed; j++)
 231                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 232                         ret = i;
 233                         pile->search_hint = i + j;
 234                         break;
 235                 }
 236
 237                 /* not enough, so skip over it and continue looking */
 238                 i += j;
 239         }
 240
 241         return ret;
 242 }
 243
 244 /**
 245  * i40e_put_lump - return a lump of generic resource
 246  * @pile: the pile of resource to search
 247  * @index: the base item index
 248  * @id: the owner id of the items assigned
 249  *
 250  * Returns the count of items in the lump
 251  **/
 252 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 253 {
 254         int valid_id = (id | I40E_PILE_VALID_BIT);
 255         int count = 0;
 256         int i;
 257
 258         if (!pile || index >= pile->num_entries)
 259                 return -EINVAL;
 260
 261         for (i = index;
 262              i < pile->num_entries && pile->list[i] == valid_id;
 263              i++) {
 264                 pile->list[i] = 0;
 265                 count++;
 266         }
 267
 268         if (count && index < pile->search_hint)
 269                 pile->search_hint = index;
 270
 271         return count;
 272 }
 273
 274 /**
 275  * i40e_find_vsi_from_id - searches for the vsi with the given id
 276  * @pf - the pf structure to search for the vsi
 277  * @id - id of the vsi it is searching for
 278  **/
 279 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 280 {
 281         int i;
 282
 283         for (i = 0; i < pf->num_alloc_vsi; i++)
 284                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 285                         return pf->vsi[i];
 286
 287         return NULL;
 288 }
 289
 290 /**
 291  * i40e_service_event_schedule - Schedule the service task to wake up
 292  * @pf: board private structure
 293  *
 294  * If not already scheduled, this puts the task into the work queue
 295  **/
 296 void i40e_service_event_schedule(struct i40e_pf *pf)
 297 {
 298         if (!test_bit(__I40E_DOWN, &pf->state) &&
 299             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
 300                 queue_work(i40e_wq, &pf->service_task);
 301 }
 302
 303 /**
 304  * i40e_tx_timeout - Respond to a Tx Hang
 305  * @netdev: network interface device structure
 306  *
 307  * If any port has noticed a Tx timeout, it is likely that the whole
 308  * device is munged, not just the one netdev port, so go for the full
 309  * reset.
 310  **/
 311 static void i40e_tx_timeout(struct net_device *netdev)
 312 {
 313         struct i40e_netdev_priv *np = netdev_priv(netdev);
 314         struct i40e_vsi *vsi = np->vsi;
 315         struct i40e_pf *pf = vsi->back;
 316         struct i40e_ring *tx_ring = NULL;
 317         unsigned int i, hung_queue = 0;
 318         u32 head, val;
 319
 320         pf->tx_timeout_count++;
 321
 322         /* find the stopped queue the same way the stack does */
 323         for (i = 0; i < netdev->num_tx_queues; i++) {
 324                 struct netdev_queue *q;
 325                 unsigned long trans_start;
 326
 327                 q = netdev_get_tx_queue(netdev, i);
 328                 trans_start = q->trans_start;
 329                 if (netif_xmit_stopped(q) &&
 330                     time_after(jiffies,
 331                                (trans_start + netdev->watchdog_timeo))) {
 332                         hung_queue = i;
 333                         break;
 334                 }
 335         }
 336
 337         if (i == netdev->num_tx_queues) {
 338                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 339         } else {
 340                 /* now that we have an index, find the tx_ring struct */
 341                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 342                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 343                                 if (hung_queue ==
 344                                     vsi->tx_rings[i]->queue_index) {
 345                                         tx_ring = vsi->tx_rings[i];
 346                                         break;
 347                                 }
 348                         }
 349                 }
 350         }
 351
 352         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 353                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 354         else if (time_before(jiffies,
 355                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 356                 return;   /* don't do any new action before the next timeout */
 357
 358         if (tx_ring) {
 359                 head = i40e_get_head(tx_ring);
 360                 /* Read interrupt register */
 361                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 362                         val = rd32(&pf->hw,
 363                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 364                                                 tx_ring->vsi->base_vector - 1));
 365                 else
 366                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 367
 368                 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
 369                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 370                             head, tx_ring->next_to_use,
 371                             readl(tx_ring->tail), val);
 372         }
 373
 374         pf->tx_timeout_last_recovery = jiffies;
 375         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 376                     pf->tx_timeout_recovery_level, hung_queue);
 377
 378         switch (pf->tx_timeout_recovery_level) {
 379         case 1:
 380                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
 381                 break;
 382         case 2:
 383                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
 384                 break;
 385         case 3:
 386                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
 387                 break;
 388         default:
 389                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 390                 break;
 391         }
 392
 393         i40e_service_event_schedule(pf);
 394         pf->tx_timeout_recovery_level++;
 395 }
 396
 397 /**
 398  * i40e_get_vsi_stats_struct - Get System Network Statistics
 399  * @vsi: the VSI we care about
 400  *
 401  * Returns the address of the device statistics structure.
 402  * The statistics are actually updated from the service task.
 403  **/
 404 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 405 {
 406         return &vsi->net_stats;
 407 }
 408
 409 /**
 410  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 411  * @netdev: network interface device structure
 412  *
 413  * Returns the address of the device statistics structure.
 414  * The statistics are actually updated from the service task.
 415  **/
 416 static void i40e_get_netdev_stats_struct(struct net_device *netdev,
 417                                   struct rtnl_link_stats64 *stats)
 418 {
 419         struct i40e_netdev_priv *np = netdev_priv(netdev);
 420         struct i40e_ring *tx_ring, *rx_ring;
 421         struct i40e_vsi *vsi = np->vsi;
 422         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 423         int i;
 424
 425         if (test_bit(__I40E_DOWN, &vsi->state))
 426                 return;
 427
 428         if (!vsi->tx_rings)
 429                 return;
 430
 431         rcu_read_lock();
 432         for (i = 0; i < vsi->num_queue_pairs; i++) {
 433                 u64 bytes, packets;
 434                 unsigned int start;
 435
 436                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 437                 if (!tx_ring)
 438                         continue;
 439
 440                 do {
 441                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 442                         packets = tx_ring->stats.packets;
 443                         bytes   = tx_ring->stats.bytes;
 444                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 445
 446                 stats->tx_packets += packets;
 447                 stats->tx_bytes   += bytes;
 448                 rx_ring = &tx_ring[1];
 449
 450                 do {
 451                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 452                         packets = rx_ring->stats.packets;
 453                         bytes   = rx_ring->stats.bytes;
 454                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 455
 456                 stats->rx_packets += packets;
 457                 stats->rx_bytes   += bytes;
 458         }
 459         rcu_read_unlock();
 460
 461         /* following stats updated by i40e_watchdog_subtask() */
 462         stats->multicast        = vsi_stats->multicast;
 463         stats->tx_errors        = vsi_stats->tx_errors;
 464         stats->tx_dropped       = vsi_stats->tx_dropped;
 465         stats->rx_errors        = vsi_stats->rx_errors;
 466         stats->rx_dropped       = vsi_stats->rx_dropped;
 467         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 468         stats->rx_length_errors = vsi_stats->rx_length_errors;
 469 }
 470
 471 /**
 472  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 473  * @vsi: the VSI to have its stats reset
 474  **/
 475 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 476 {
 477         struct rtnl_link_stats64 *ns;
 478         int i;
 479
 480         if (!vsi)
 481                 return;
 482
 483         ns = i40e_get_vsi_stats_struct(vsi);
 484         memset(ns, 0, sizeof(*ns));
 485         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 486         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 487         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 488         if (vsi->rx_rings && vsi->rx_rings[0]) {
 489                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 490                         memset(&vsi->rx_rings[i]->stats, 0,
 491                                sizeof(vsi->rx_rings[i]->stats));
 492                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 493                                sizeof(vsi->rx_rings[i]->rx_stats));
 494                         memset(&vsi->tx_rings[i]->stats, 0,
 495                                sizeof(vsi->tx_rings[i]->stats));
 496                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 497                                sizeof(vsi->tx_rings[i]->tx_stats));
 498                 }
 499         }
 500         vsi->stat_offsets_loaded = false;
 501 }
 502
 503 /**
 504  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 505  * @pf: the PF to be reset
 506  **/
 507 void i40e_pf_reset_stats(struct i40e_pf *pf)
 508 {
 509         int i;
 510
 511         memset(&pf->stats, 0, sizeof(pf->stats));
 512         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 513         pf->stat_offsets_loaded = false;
 514
 515         for (i = 0; i < I40E_MAX_VEB; i++) {
 516                 if (pf->veb[i]) {
 517                         memset(&pf->veb[i]->stats, 0,
 518                                sizeof(pf->veb[i]->stats));
 519                         memset(&pf->veb[i]->stats_offsets, 0,
 520                                sizeof(pf->veb[i]->stats_offsets));
 521                         pf->veb[i]->stat_offsets_loaded = false;
 522                 }
 523         }
 524         pf->hw_csum_rx_error = 0;
 525 }
 526
 527 /**
 528  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 529  * @hw: ptr to the hardware info
 530  * @hireg: the high 32 bit reg to read
 531  * @loreg: the low 32 bit reg to read
 532  * @offset_loaded: has the initial offset been loaded yet
 533  * @offset: ptr to current offset value
 534  * @stat: ptr to the stat
 535  *
 536  * Since the device stats are not reset at PFReset, they likely will not
 537  * be zeroed when the driver starts.  We'll save the first values read
 538  * and use them as offsets to be subtracted from the raw values in order
 539  * to report stats that count from zero.  In the process, we also manage
 540  * the potential roll-over.
 541  **/
 542 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 543                                bool offset_loaded, u64 *offset, u64 *stat)
 544 {
 545         u64 new_data;
 546
 547         if (hw->device_id == I40E_DEV_ID_QEMU) {
 548                 new_data = rd32(hw, loreg);
 549                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 550         } else {
 551                 new_data = rd64(hw, loreg);
 552         }
 553         if (!offset_loaded)
 554                 *offset = new_data;
 555         if (likely(new_data >= *offset))
 556                 *stat = new_data - *offset;
 557         else
 558                 *stat = (new_data + BIT_ULL(48)) - *offset;
 559         *stat &= 0xFFFFFFFFFFFFULL;
 560 }
 561
 562 /**
 563  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 564  * @hw: ptr to the hardware info
 565  * @reg: the hw reg to read
 566  * @offset_loaded: has the initial offset been loaded yet
 567  * @offset: ptr to current offset value
 568  * @stat: ptr to the stat
 569  **/
 570 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 571                                bool offset_loaded, u64 *offset, u64 *stat)
 572 {
 573         u32 new_data;
 574
 575         new_data = rd32(hw, reg);
 576         if (!offset_loaded)
 577                 *offset = new_data;
 578         if (likely(new_data >= *offset))
 579                 *stat = (u32)(new_data - *offset);
 580         else
 581                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 582 }
 583
 584 /**
 585  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 586  * @vsi: the VSI to be updated
 587  **/
 588 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 589 {
 590         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 591         struct i40e_pf *pf = vsi->back;
 592         struct i40e_hw *hw = &pf->hw;
 593         struct i40e_eth_stats *oes;
 594         struct i40e_eth_stats *es;     /* device's eth stats */
 595
 596         es = &vsi->eth_stats;
 597         oes = &vsi->eth_stats_offsets;
 598
 599         /* Gather up the stats that the hw collects */
 600         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 601                            vsi->stat_offsets_loaded,
 602                            &oes->tx_errors, &es->tx_errors);
 603         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 604                            vsi->stat_offsets_loaded,
 605                            &oes->rx_discards, &es->rx_discards);
 606         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 607                            vsi->stat_offsets_loaded,
 608                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 609         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 610                            vsi->stat_offsets_loaded,
 611                            &oes->tx_errors, &es->tx_errors);
 612
 613         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 614                            I40E_GLV_GORCL(stat_idx),
 615                            vsi->stat_offsets_loaded,
 616                            &oes->rx_bytes, &es->rx_bytes);
 617         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 618                            I40E_GLV_UPRCL(stat_idx),
 619                            vsi->stat_offsets_loaded,
 620                            &oes->rx_unicast, &es->rx_unicast);
 621         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 622                            I40E_GLV_MPRCL(stat_idx),
 623                            vsi->stat_offsets_loaded,
 624                            &oes->rx_multicast, &es->rx_multicast);
 625         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 626                            I40E_GLV_BPRCL(stat_idx),
 627                            vsi->stat_offsets_loaded,
 628                            &oes->rx_broadcast, &es->rx_broadcast);
 629
 630         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 631                            I40E_GLV_GOTCL(stat_idx),
 632                            vsi->stat_offsets_loaded,
 633                            &oes->tx_bytes, &es->tx_bytes);
 634         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 635                            I40E_GLV_UPTCL(stat_idx),
 636                            vsi->stat_offsets_loaded,
 637                            &oes->tx_unicast, &es->tx_unicast);
 638         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 639                            I40E_GLV_MPTCL(stat_idx),
 640                            vsi->stat_offsets_loaded,
 641                            &oes->tx_multicast, &es->tx_multicast);
 642         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 643                            I40E_GLV_BPTCL(stat_idx),
 644                            vsi->stat_offsets_loaded,
 645                            &oes->tx_broadcast, &es->tx_broadcast);
 646         vsi->stat_offsets_loaded = true;
 647 }
 648
 649 /**
 650  * i40e_update_veb_stats - Update Switch component statistics
 651  * @veb: the VEB being updated
 652  **/
 653 static void i40e_update_veb_stats(struct i40e_veb *veb)
 654 {
 655         struct i40e_pf *pf = veb->pf;
 656         struct i40e_hw *hw = &pf->hw;
 657         struct i40e_eth_stats *oes;
 658         struct i40e_eth_stats *es;     /* device's eth stats */
 659         struct i40e_veb_tc_stats *veb_oes;
 660         struct i40e_veb_tc_stats *veb_es;
 661         int i, idx = 0;
 662
 663         idx = veb->stats_idx;
 664         es = &veb->stats;
 665         oes = &veb->stats_offsets;
 666         veb_es = &veb->tc_stats;
 667         veb_oes = &veb->tc_stats_offsets;
 668
 669         /* Gather up the stats that the hw collects */
 670         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 671                            veb->stat_offsets_loaded,
 672                            &oes->tx_discards, &es->tx_discards);
 673         if (hw->revision_id > 0)
 674                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 675                                    veb->stat_offsets_loaded,
 676                                    &oes->rx_unknown_protocol,
 677                                    &es->rx_unknown_protocol);
 678         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 679                            veb->stat_offsets_loaded,
 680                            &oes->rx_bytes, &es->rx_bytes);
 681         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 682                            veb->stat_offsets_loaded,
 683                            &oes->rx_unicast, &es->rx_unicast);
 684         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 685                            veb->stat_offsets_loaded,
 686                            &oes->rx_multicast, &es->rx_multicast);
 687         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 688                            veb->stat_offsets_loaded,
 689                            &oes->rx_broadcast, &es->rx_broadcast);
 690
 691         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 692                            veb->stat_offsets_loaded,
 693                            &oes->tx_bytes, &es->tx_bytes);
 694         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 695                            veb->stat_offsets_loaded,
 696                            &oes->tx_unicast, &es->tx_unicast);
 697         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 698                            veb->stat_offsets_loaded,
 699                            &oes->tx_multicast, &es->tx_multicast);
 700         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 701                            veb->stat_offsets_loaded,
 702                            &oes->tx_broadcast, &es->tx_broadcast);
 703         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 704                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 705                                    I40E_GLVEBTC_RPCL(i, idx),
 706                                    veb->stat_offsets_loaded,
 707                                    &veb_oes->tc_rx_packets[i],
 708                                    &veb_es->tc_rx_packets[i]);
 709                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 710                                    I40E_GLVEBTC_RBCL(i, idx),
 711                                    veb->stat_offsets_loaded,
 712                                    &veb_oes->tc_rx_bytes[i],
 713                                    &veb_es->tc_rx_bytes[i]);
 714                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 715                                    I40E_GLVEBTC_TPCL(i, idx),
 716                                    veb->stat_offsets_loaded,
 717                                    &veb_oes->tc_tx_packets[i],
 718                                    &veb_es->tc_tx_packets[i]);
 719                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 720                                    I40E_GLVEBTC_TBCL(i, idx),
 721                                    veb->stat_offsets_loaded,
 722                                    &veb_oes->tc_tx_bytes[i],
 723                                    &veb_es->tc_tx_bytes[i]);
 724         }
 725         veb->stat_offsets_loaded = true;
 726 }
 727
 728 /**
 729  * i40e_update_vsi_stats - Update the vsi statistics counters.
 730  * @vsi: the VSI to be updated
 731  *
 732  * There are a few instances where we store the same stat in a
 733  * couple of different structs.  This is partly because we have
 734  * the netdev stats that need to be filled out, which is slightly
 735  * different from the "eth_stats" defined by the chip and used in
 736  * VF communications.  We sort it out here.
 737  **/
 738 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 739 {
 740         struct i40e_pf *pf = vsi->back;
 741         struct rtnl_link_stats64 *ons;
 742         struct rtnl_link_stats64 *ns;   /* netdev stats */
 743         struct i40e_eth_stats *oes;
 744         struct i40e_eth_stats *es;     /* device's eth stats */
 745         u32 tx_restart, tx_busy;
 746         struct i40e_ring *p;
 747         u32 rx_page, rx_buf;
 748         u64 bytes, packets;
 749         unsigned int start;
 750         u64 tx_linearize;
 751         u64 tx_force_wb;
 752         u64 rx_p, rx_b;
 753         u64 tx_p, tx_b;
 754         u16 q;
 755
 756         if (test_bit(__I40E_DOWN, &vsi->state) ||
 757             test_bit(__I40E_CONFIG_BUSY, &pf->state))
 758                 return;
 759
 760         ns = i40e_get_vsi_stats_struct(vsi);
 761         ons = &vsi->net_stats_offsets;
 762         es = &vsi->eth_stats;
 763         oes = &vsi->eth_stats_offsets;
 764
 765         /* Gather up the netdev and vsi stats that the driver collects
 766          * on the fly during packet processing
 767          */
 768         rx_b = rx_p = 0;
 769         tx_b = tx_p = 0;
 770         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 771         rx_page = 0;
 772         rx_buf = 0;
 773         rcu_read_lock();
 774         for (q = 0; q < vsi->num_queue_pairs; q++) {
 775                 /* locate Tx ring */
 776                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 777
 778                 do {
 779                         start = u64_stats_fetch_begin_irq(&p->syncp);
 780                         packets = p->stats.packets;
 781                         bytes = p->stats.bytes;
 782                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 783                 tx_b += bytes;
 784                 tx_p += packets;
 785                 tx_restart += p->tx_stats.restart_queue;
 786                 tx_busy += p->tx_stats.tx_busy;
 787                 tx_linearize += p->tx_stats.tx_linearize;
 788                 tx_force_wb += p->tx_stats.tx_force_wb;
 789
 790                 /* Rx queue is part of the same block as Tx queue */
 791                 p = &p[1];
 792                 do {
 793                         start = u64_stats_fetch_begin_irq(&p->syncp);
 794                         packets = p->stats.packets;
 795                         bytes = p->stats.bytes;
 796                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 797                 rx_b += bytes;
 798                 rx_p += packets;
 799                 rx_buf += p->rx_stats.alloc_buff_failed;
 800                 rx_page += p->rx_stats.alloc_page_failed;
 801         }
 802         rcu_read_unlock();
 803         vsi->tx_restart = tx_restart;
 804         vsi->tx_busy = tx_busy;
 805         vsi->tx_linearize = tx_linearize;
 806         vsi->tx_force_wb = tx_force_wb;
 807         vsi->rx_page_failed = rx_page;
 808         vsi->rx_buf_failed = rx_buf;
 809
 810         ns->rx_packets = rx_p;
 811         ns->rx_bytes = rx_b;
 812         ns->tx_packets = tx_p;
 813         ns->tx_bytes = tx_b;
 814
 815         /* update netdev stats from eth stats */
 816         i40e_update_eth_stats(vsi);
 817         ons->tx_errors = oes->tx_errors;
 818         ns->tx_errors = es->tx_errors;
 819         ons->multicast = oes->rx_multicast;
 820         ns->multicast = es->rx_multicast;
 821         ons->rx_dropped = oes->rx_discards;
 822         ns->rx_dropped = es->rx_discards;
 823         ons->tx_dropped = oes->tx_discards;
 824         ns->tx_dropped = es->tx_discards;
 825
 826         /* pull in a couple PF stats if this is the main vsi */
 827         if (vsi == pf->vsi[pf->lan_vsi]) {
 828                 ns->rx_crc_errors = pf->stats.crc_errors;
 829                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 830                 ns->rx_length_errors = pf->stats.rx_length_errors;
 831         }
 832 }
 833
 834 /**
 835  * i40e_update_pf_stats - Update the PF statistics counters.
 836  * @pf: the PF to be updated
 837  **/
 838 static void i40e_update_pf_stats(struct i40e_pf *pf)
 839 {
 840         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 841         struct i40e_hw_port_stats *nsd = &pf->stats;
 842         struct i40e_hw *hw = &pf->hw;
 843         u32 val;
 844         int i;
 845
 846         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 847                            I40E_GLPRT_GORCL(hw->port),
 848                            pf->stat_offsets_loaded,
 849                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 850         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 851                            I40E_GLPRT_GOTCL(hw->port),
 852                            pf->stat_offsets_loaded,
 853                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 854         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 855                            pf->stat_offsets_loaded,
 856                            &osd->eth.rx_discards,
 857                            &nsd->eth.rx_discards);
 858         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 859                            I40E_GLPRT_UPRCL(hw->port),
 860                            pf->stat_offsets_loaded,
 861                            &osd->eth.rx_unicast,
 862                            &nsd->eth.rx_unicast);
 863         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 864                            I40E_GLPRT_MPRCL(hw->port),
 865                            pf->stat_offsets_loaded,
 866                            &osd->eth.rx_multicast,
 867                            &nsd->eth.rx_multicast);
 868         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 869                            I40E_GLPRT_BPRCL(hw->port),
 870                            pf->stat_offsets_loaded,
 871                            &osd->eth.rx_broadcast,
 872                            &nsd->eth.rx_broadcast);
 873         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 874                            I40E_GLPRT_UPTCL(hw->port),
 875                            pf->stat_offsets_loaded,
 876                            &osd->eth.tx_unicast,
 877                            &nsd->eth.tx_unicast);
 878         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 879                            I40E_GLPRT_MPTCL(hw->port),
 880                            pf->stat_offsets_loaded,
 881                            &osd->eth.tx_multicast,
 882                            &nsd->eth.tx_multicast);
 883         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 884                            I40E_GLPRT_BPTCL(hw->port),
 885                            pf->stat_offsets_loaded,
 886                            &osd->eth.tx_broadcast,
 887                            &nsd->eth.tx_broadcast);
 888
 889         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 890                            pf->stat_offsets_loaded,
 891                            &osd->tx_dropped_link_down,
 892                            &nsd->tx_dropped_link_down);
 893
 894         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 895                            pf->stat_offsets_loaded,
 896                            &osd->crc_errors, &nsd->crc_errors);
 897
 898         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 899                            pf->stat_offsets_loaded,
 900                            &osd->illegal_bytes, &nsd->illegal_bytes);
 901
 902         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 903                            pf->stat_offsets_loaded,
 904                            &osd->mac_local_faults,
 905                            &nsd->mac_local_faults);
 906         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 907                            pf->stat_offsets_loaded,
 908                            &osd->mac_remote_faults,
 909                            &nsd->mac_remote_faults);
 910
 911         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 912                            pf->stat_offsets_loaded,
 913                            &osd->rx_length_errors,
 914                            &nsd->rx_length_errors);
 915
 916         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 917                            pf->stat_offsets_loaded,
 918                            &osd->link_xon_rx, &nsd->link_xon_rx);
 919         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 920                            pf->stat_offsets_loaded,
 921                            &osd->link_xon_tx, &nsd->link_xon_tx);
 922         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 923                            pf->stat_offsets_loaded,
 924                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 925         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 926                            pf->stat_offsets_loaded,
 927                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 928
 929         for (i = 0; i < 8; i++) {
 930                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 931                                    pf->stat_offsets_loaded,
 932                                    &osd->priority_xoff_rx[i],
 933                                    &nsd->priority_xoff_rx[i]);
 934                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 935                                    pf->stat_offsets_loaded,
 936                                    &osd->priority_xon_rx[i],
 937                                    &nsd->priority_xon_rx[i]);
 938                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
 939                                    pf->stat_offsets_loaded,
 940                                    &osd->priority_xon_tx[i],
 941                                    &nsd->priority_xon_tx[i]);
 942                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
 943                                    pf->stat_offsets_loaded,
 944                                    &osd->priority_xoff_tx[i],
 945                                    &nsd->priority_xoff_tx[i]);
 946                 i40e_stat_update32(hw,
 947                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
 948                                    pf->stat_offsets_loaded,
 949                                    &osd->priority_xon_2_xoff[i],
 950                                    &nsd->priority_xon_2_xoff[i]);
 951         }
 952
 953         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
 954                            I40E_GLPRT_PRC64L(hw->port),
 955                            pf->stat_offsets_loaded,
 956                            &osd->rx_size_64, &nsd->rx_size_64);
 957         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
 958                            I40E_GLPRT_PRC127L(hw->port),
 959                            pf->stat_offsets_loaded,
 960                            &osd->rx_size_127, &nsd->rx_size_127);
 961         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
 962                            I40E_GLPRT_PRC255L(hw->port),
 963                            pf->stat_offsets_loaded,
 964                            &osd->rx_size_255, &nsd->rx_size_255);
 965         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
 966                            I40E_GLPRT_PRC511L(hw->port),
 967                            pf->stat_offsets_loaded,
 968                            &osd->rx_size_511, &nsd->rx_size_511);
 969         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
 970                            I40E_GLPRT_PRC1023L(hw->port),
 971                            pf->stat_offsets_loaded,
 972                            &osd->rx_size_1023, &nsd->rx_size_1023);
 973         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
 974                            I40E_GLPRT_PRC1522L(hw->port),
 975                            pf->stat_offsets_loaded,
 976                            &osd->rx_size_1522, &nsd->rx_size_1522);
 977         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
 978                            I40E_GLPRT_PRC9522L(hw->port),
 979                            pf->stat_offsets_loaded,
 980                            &osd->rx_size_big, &nsd->rx_size_big);
 981
 982         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
 983                            I40E_GLPRT_PTC64L(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->tx_size_64, &nsd->tx_size_64);
 986         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
 987                            I40E_GLPRT_PTC127L(hw->port),
 988                            pf->stat_offsets_loaded,
 989                            &osd->tx_size_127, &nsd->tx_size_127);
 990         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
 991                            I40E_GLPRT_PTC255L(hw->port),
 992                            pf->stat_offsets_loaded,
 993                            &osd->tx_size_255, &nsd->tx_size_255);
 994         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
 995                            I40E_GLPRT_PTC511L(hw->port),
 996                            pf->stat_offsets_loaded,
 997                            &osd->tx_size_511, &nsd->tx_size_511);
 998         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
 999                            I40E_GLPRT_PTC1023L(hw->port),
1000                            pf->stat_offsets_loaded,
1001                            &osd->tx_size_1023, &nsd->tx_size_1023);
1002         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1003                            I40E_GLPRT_PTC1522L(hw->port),
1004                            pf->stat_offsets_loaded,
1005                            &osd->tx_size_1522, &nsd->tx_size_1522);
1006         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1007                            I40E_GLPRT_PTC9522L(hw->port),
1008                            pf->stat_offsets_loaded,
1009                            &osd->tx_size_big, &nsd->tx_size_big);
1010
1011         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1012                            pf->stat_offsets_loaded,
1013                            &osd->rx_undersize, &nsd->rx_undersize);
1014         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1015                            pf->stat_offsets_loaded,
1016                            &osd->rx_fragments, &nsd->rx_fragments);
1017         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1018                            pf->stat_offsets_loaded,
1019                            &osd->rx_oversize, &nsd->rx_oversize);
1020         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1021                            pf->stat_offsets_loaded,
1022                            &osd->rx_jabber, &nsd->rx_jabber);
1023
1024         /* FDIR stats */
1025         i40e_stat_update32(hw,
1026                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1027                            pf->stat_offsets_loaded,
1028                            &osd->fd_atr_match, &nsd->fd_atr_match);
1029         i40e_stat_update32(hw,
1030                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1031                            pf->stat_offsets_loaded,
1032                            &osd->fd_sb_match, &nsd->fd_sb_match);
1033         i40e_stat_update32(hw,
1034                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1035                       pf->stat_offsets_loaded,
1036                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1037
1038         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1039         nsd->tx_lpi_status =
1040                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1041                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1042         nsd->rx_lpi_status =
1043                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1044                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1045         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1046                            pf->stat_offsets_loaded,
1047                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1048         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1049                            pf->stat_offsets_loaded,
1050                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1051
1052         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1053             !(pf->hw_disabled_flags & I40E_FLAG_FD_SB_ENABLED))
1054                 nsd->fd_sb_status = true;
1055         else
1056                 nsd->fd_sb_status = false;
1057
1058         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1059             !(pf->hw_disabled_flags & I40E_FLAG_FD_ATR_ENABLED))
1060                 nsd->fd_atr_status = true;
1061         else
1062                 nsd->fd_atr_status = false;
1063
1064         pf->stat_offsets_loaded = true;
1065 }
1066
1067 /**
1068  * i40e_update_stats - Update the various statistics counters.
1069  * @vsi: the VSI to be updated
1070  *
1071  * Update the various stats for this VSI and its related entities.
1072  **/
1073 void i40e_update_stats(struct i40e_vsi *vsi)
1074 {
1075         struct i40e_pf *pf = vsi->back;
1076
1077         if (vsi == pf->vsi[pf->lan_vsi])
1078                 i40e_update_pf_stats(pf);
1079
1080         i40e_update_vsi_stats(vsi);
1081 }
1082
1083 /**
1084  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1085  * @vsi: the VSI to be searched
1086  * @macaddr: the MAC address
1087  * @vlan: the vlan
1088  *
1089  * Returns ptr to the filter object or NULL
1090  **/
1091 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1092                                                 const u8 *macaddr, s16 vlan)
1093 {
1094         struct i40e_mac_filter *f;
1095         u64 key;
1096
1097         if (!vsi || !macaddr)
1098                 return NULL;
1099
1100         key = i40e_addr_to_hkey(macaddr);
1101         hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1102                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1103                     (vlan == f->vlan))
1104                         return f;
1105         }
1106         return NULL;
1107 }
1108
1109 /**
1110  * i40e_find_mac - Find a mac addr in the macvlan filters list
1111  * @vsi: the VSI to be searched
1112  * @macaddr: the MAC address we are searching for
1113  *
1114  * Returns the first filter with the provided MAC address or NULL if
1115  * MAC address was not found
1116  **/
1117 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, const u8 *macaddr)
1118 {
1119         struct i40e_mac_filter *f;
1120         u64 key;
1121
1122         if (!vsi || !macaddr)
1123                 return NULL;
1124
1125         key = i40e_addr_to_hkey(macaddr);
1126         hash_for_each_possible(vsi->mac_filter_hash, f, hlist, key) {
1127                 if ((ether_addr_equal(macaddr, f->macaddr)))
1128                         return f;
1129         }
1130         return NULL;
1131 }
1132
1133 /**
1134  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1135  * @vsi: the VSI to be searched
1136  *
1137  * Returns true if VSI is in vlan mode or false otherwise
1138  **/
1139 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1140 {
1141         /* If we have a PVID, always operate in VLAN mode */
1142         if (vsi->info.pvid)
1143                 return true;
1144
1145         /* We need to operate in VLAN mode whenever we have any filters with
1146          * a VLAN other than I40E_VLAN_ALL. We could check the table each
1147          * time, incurring search cost repeatedly. However, we can notice two
1148          * things:
1149          *
1150          * 1) the only place where we can gain a VLAN filter is in
1151          *    i40e_add_filter.
1152          *
1153          * 2) the only place where filters are actually removed is in
1154          *    i40e_sync_filters_subtask.
1155          *
1156          * Thus, we can simply use a boolean value, has_vlan_filters which we
1157          * will set to true when we add a VLAN filter in i40e_add_filter. Then
1158          * we have to perform the full search after deleting filters in
1159          * i40e_sync_filters_subtask, but we already have to search
1160          * filters here and can perform the check at the same time. This
1161          * results in avoiding embedding a loop for VLAN mode inside another
1162          * loop over all the filters, and should maintain correctness as noted
1163          * above.
1164          */
1165         return vsi->has_vlan_filter;
1166 }
1167
1168 /**
1169  * i40e_correct_mac_vlan_filters - Correct non-VLAN filters if necessary
1170  * @vsi: the VSI to configure
1171  * @tmp_add_list: list of filters ready to be added
1172  * @tmp_del_list: list of filters ready to be deleted
1173  * @vlan_filters: the number of active VLAN filters
1174  *
1175  * Update VLAN=0 and VLAN=-1 (I40E_VLAN_ANY) filters properly so that they
1176  * behave as expected. If we have any active VLAN filters remaining or about
1177  * to be added then we need to update non-VLAN filters to be marked as VLAN=0
1178  * so that they only match against untagged traffic. If we no longer have any
1179  * active VLAN filters, we need to make all non-VLAN filters marked as VLAN=-1
1180  * so that they match against both tagged and untagged traffic. In this way,
1181  * we ensure that we correctly receive the desired traffic. This ensures that
1182  * when we have an active VLAN we will receive only untagged traffic and
1183  * traffic matching active VLANs. If we have no active VLANs then we will
1184  * operate in non-VLAN mode and receive all traffic, tagged or untagged.
1185  *
1186  * Finally, in a similar fashion, this function also corrects filters when
1187  * there is an active PVID assigned to this VSI.
1188  *
1189  * In case of memory allocation failure return -ENOMEM. Otherwise, return 0.
1190  *
1191  * This function is only expected to be called from within
1192  * i40e_sync_vsi_filters.
1193  *
1194  * NOTE: This function expects to be called while under the
1195  * mac_filter_hash_lock
1196  */
1197 static int i40e_correct_mac_vlan_filters(struct i40e_vsi *vsi,
1198                                          struct hlist_head *tmp_add_list,
1199                                          struct hlist_head *tmp_del_list,
1200                                          int vlan_filters)
1201 {
1202         s16 pvid = le16_to_cpu(vsi->info.pvid);
1203         struct i40e_mac_filter *f, *add_head;
1204         struct i40e_new_mac_filter *new;
1205         struct hlist_node *h;
1206         int bkt, new_vlan;
1207
1208         /* To determine if a particular filter needs to be replaced we
1209          * have the three following conditions:
1210          *
1211          * a) if we have a PVID assigned, then all filters which are
1212          *    not marked as VLAN=PVID must be replaced with filters that
1213          *    are.
1214          * b) otherwise, if we have any active VLANS, all filters
1215          *    which are marked as VLAN=-1 must be replaced with
1216          *    filters marked as VLAN=0
1217          * c) finally, if we do not have any active VLANS, all filters
1218          *    which are marked as VLAN=0 must be replaced with filters
1219          *    marked as VLAN=-1
1220          */
1221
1222         /* Update the filters about to be added in place */
1223         hlist_for_each_entry(new, tmp_add_list, hlist) {
1224                 if (pvid && new->f->vlan != pvid)
1225                         new->f->vlan = pvid;
1226                 else if (vlan_filters && new->f->vlan == I40E_VLAN_ANY)
1227                         new->f->vlan = 0;
1228                 else if (!vlan_filters && new->f->vlan == 0)
1229                         new->f->vlan = I40E_VLAN_ANY;
1230         }
1231
1232         /* Update the remaining active filters */
1233         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1234                 /* Combine the checks for whether a filter needs to be changed
1235                  * and then determine the new VLAN inside the if block, in
1236                  * order to avoid duplicating code for adding the new filter
1237                  * then deleting the old filter.
1238                  */
1239                 if ((pvid && f->vlan != pvid) ||
1240                     (vlan_filters && f->vlan == I40E_VLAN_ANY) ||
1241                     (!vlan_filters && f->vlan == 0)) {
1242                         /* Determine the new vlan we will be adding */
1243                         if (pvid)
1244                                 new_vlan = pvid;
1245                         else if (vlan_filters)
1246                                 new_vlan = 0;
1247                         else
1248                                 new_vlan = I40E_VLAN_ANY;
1249
1250                         /* Create the new filter */
1251                         add_head = i40e_add_filter(vsi, f->macaddr, new_vlan);
1252                         if (!add_head)
1253                                 return -ENOMEM;
1254
1255                         /* Create a temporary i40e_new_mac_filter */
1256                         new = kzalloc(sizeof(*new), GFP_ATOMIC);
1257                         if (!new)
1258                                 return -ENOMEM;
1259
1260                         new->f = add_head;
1261                         new->state = add_head->state;
1262
1263                         /* Add the new filter to the tmp list */
1264                         hlist_add_head(&new->hlist, tmp_add_list);
1265
1266                         /* Put the original filter into the delete list */
1267                         f->state = I40E_FILTER_REMOVE;
1268                         hash_del(&f->hlist);
1269                         hlist_add_head(&f->hlist, tmp_del_list);
1270                 }
1271         }
1272
1273         vsi->has_vlan_filter = !!vlan_filters;
1274
1275         return 0;
1276 }
1277
1278 /**
1279  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1280  * @vsi: the PF Main VSI - inappropriate for any other VSI
1281  * @macaddr: the MAC address
1282  *
1283  * Remove whatever filter the firmware set up so the driver can manage
1284  * its own filtering intelligently.
1285  **/
1286 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1287 {
1288         struct i40e_aqc_remove_macvlan_element_data element;
1289         struct i40e_pf *pf = vsi->back;
1290
1291         /* Only appropriate for the PF main VSI */
1292         if (vsi->type != I40E_VSI_MAIN)
1293                 return;
1294
1295         memset(&element, 0, sizeof(element));
1296         ether_addr_copy(element.mac_addr, macaddr);
1297         element.vlan_tag = 0;
1298         /* Ignore error returns, some firmware does it this way... */
1299         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1300         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1301
1302         memset(&element, 0, sizeof(element));
1303         ether_addr_copy(element.mac_addr, macaddr);
1304         element.vlan_tag = 0;
1305         /* ...and some firmware does it this way. */
1306         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1307                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1308         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1309 }
1310
1311 /**
1312  * i40e_add_filter - Add a mac/vlan filter to the VSI
1313  * @vsi: the VSI to be searched
1314  * @macaddr: the MAC address
1315  * @vlan: the vlan
1316  *
1317  * Returns ptr to the filter object or NULL when no memory available.
1318  *
1319  * NOTE: This function is expected to be called with mac_filter_hash_lock
1320  * being held.
1321  **/
1322 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1323                                         const u8 *macaddr, s16 vlan)
1324 {
1325         struct i40e_mac_filter *f;
1326         u64 key;
1327
1328         if (!vsi || !macaddr)
1329                 return NULL;
1330
1331         f = i40e_find_filter(vsi, macaddr, vlan);
1332         if (!f) {
1333                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1334                 if (!f)
1335                         return NULL;
1336
1337                 /* Update the boolean indicating if we need to function in
1338                  * VLAN mode.
1339                  */
1340                 if (vlan >= 0)
1341                         vsi->has_vlan_filter = true;
1342
1343                 ether_addr_copy(f->macaddr, macaddr);
1344                 f->vlan = vlan;
1345                 /* If we're in overflow promisc mode, set the state directly
1346                  * to failed, so we don't bother to try sending the filter
1347                  * to the hardware.
1348                  */
1349                 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))
1350                         f->state = I40E_FILTER_FAILED;
1351                 else
1352                         f->state = I40E_FILTER_NEW;
1353                 INIT_HLIST_NODE(&f->hlist);
1354
1355                 key = i40e_addr_to_hkey(macaddr);
1356                 hash_add(vsi->mac_filter_hash, &f->hlist, key);
1357
1358                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1359                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1360         }
1361
1362         /* If we're asked to add a filter that has been marked for removal, it
1363          * is safe to simply restore it to active state. __i40e_del_filter
1364          * will have simply deleted any filters which were previously marked
1365          * NEW or FAILED, so if it is currently marked REMOVE it must have
1366          * previously been ACTIVE. Since we haven't yet run the sync filters
1367          * task, just restore this filter to the ACTIVE state so that the
1368          * sync task leaves it in place
1369          */
1370         if (f->state == I40E_FILTER_REMOVE)
1371                 f->state = I40E_FILTER_ACTIVE;
1372
1373         return f;
1374 }
1375
1376 /**
1377  * __i40e_del_filter - Remove a specific filter from the VSI
1378  * @vsi: VSI to remove from
1379  * @f: the filter to remove from the list
1380  *
1381  * This function should be called instead of i40e_del_filter only if you know
1382  * the exact filter you will remove already, such as via i40e_find_filter or
1383  * i40e_find_mac.
1384  *
1385  * NOTE: This function is expected to be called with mac_filter_hash_lock
1386  * being held.
1387  * ANOTHER NOTE: This function MUST be called from within the context of
1388  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1389  * instead of list_for_each_entry().
1390  **/
1391 void __i40e_del_filter(struct i40e_vsi *vsi, struct i40e_mac_filter *f)
1392 {
1393         if (!f)
1394                 return;
1395
1396         /* If the filter was never added to firmware then we can just delete it
1397          * directly and we don't want to set the status to remove or else an
1398          * admin queue command will unnecessarily fire.
1399          */
1400         if ((f->state == I40E_FILTER_FAILED) ||
1401             (f->state == I40E_FILTER_NEW)) {
1402                 hash_del(&f->hlist);
1403                 kfree(f);
1404         } else {
1405                 f->state = I40E_FILTER_REMOVE;
1406         }
1407
1408         vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1409         vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1410 }
1411
1412 /**
1413  * i40e_del_filter - Remove a MAC/VLAN filter from the VSI
1414  * @vsi: the VSI to be searched
1415  * @macaddr: the MAC address
1416  * @vlan: the VLAN
1417  *
1418  * NOTE: This function is expected to be called with mac_filter_hash_lock
1419  * being held.
1420  * ANOTHER NOTE: This function MUST be called from within the context of
1421  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1422  * instead of list_for_each_entry().
1423  **/
1424 void i40e_del_filter(struct i40e_vsi *vsi, const u8 *macaddr, s16 vlan)
1425 {
1426         struct i40e_mac_filter *f;
1427
1428         if (!vsi || !macaddr)
1429                 return;
1430
1431         f = i40e_find_filter(vsi, macaddr, vlan);
1432         __i40e_del_filter(vsi, f);
1433 }
1434
1435 /**
1436  * i40e_add_mac_filter - Add a MAC filter for all active VLANs
1437  * @vsi: the VSI to be searched
1438  * @macaddr: the mac address to be filtered
1439  *
1440  * If we're not in VLAN mode, just add the filter to I40E_VLAN_ANY. Otherwise,
1441  * go through all the macvlan filters and add a macvlan filter for each
1442  * unique vlan that already exists. If a PVID has been assigned, instead only
1443  * add the macaddr to that VLAN.
1444  *
1445  * Returns last filter added on success, else NULL
1446  **/
1447 struct i40e_mac_filter *i40e_add_mac_filter(struct i40e_vsi *vsi,
1448                                             const u8 *macaddr)
1449 {
1450         struct i40e_mac_filter *f, *add = NULL;
1451         struct hlist_node *h;
1452         int bkt;
1453
1454         if (vsi->info.pvid)
1455                 return i40e_add_filter(vsi, macaddr,
1456                                        le16_to_cpu(vsi->info.pvid));
1457
1458         if (!i40e_is_vsi_in_vlan(vsi))
1459                 return i40e_add_filter(vsi, macaddr, I40E_VLAN_ANY);
1460
1461         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1462                 if (f->state == I40E_FILTER_REMOVE)
1463                         continue;
1464                 add = i40e_add_filter(vsi, macaddr, f->vlan);
1465                 if (!add)
1466                         return NULL;
1467         }
1468
1469         return add;
1470 }
1471
1472 /**
1473  * i40e_del_mac_filter - Remove a MAC filter from all VLANs
1474  * @vsi: the VSI to be searched
1475  * @macaddr: the mac address to be removed
1476  *
1477  * Removes a given MAC address from a VSI regardless of what VLAN it has been
1478  * associated with.
1479  *
1480  * Returns 0 for success, or error
1481  **/
1482 int i40e_del_mac_filter(struct i40e_vsi *vsi, const u8 *macaddr)
1483 {
1484         struct i40e_mac_filter *f;
1485         struct hlist_node *h;
1486         bool found = false;
1487         int bkt;
1488
1489         WARN(!spin_is_locked(&vsi->mac_filter_hash_lock),
1490              "Missing mac_filter_hash_lock\n");
1491         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
1492                 if (ether_addr_equal(macaddr, f->macaddr)) {
1493                         __i40e_del_filter(vsi, f);
1494                         found = true;
1495                 }
1496         }
1497
1498         if (found)
1499                 return 0;
1500         else
1501                 return -ENOENT;
1502 }
1503
1504 /**
1505  * i40e_set_mac - NDO callback to set mac address
1506  * @netdev: network interface device structure
1507  * @p: pointer to an address structure
1508  *
1509  * Returns 0 on success, negative on failure
1510  **/
1511 static int i40e_set_mac(struct net_device *netdev, void *p)
1512 {
1513         struct i40e_netdev_priv *np = netdev_priv(netdev);
1514         struct i40e_vsi *vsi = np->vsi;
1515         struct i40e_pf *pf = vsi->back;
1516         struct i40e_hw *hw = &pf->hw;
1517         struct sockaddr *addr = p;
1518
1519         if (!is_valid_ether_addr(addr->sa_data))
1520                 return -EADDRNOTAVAIL;
1521
1522         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1523                 netdev_info(netdev, "already using mac address %pM\n",
1524                             addr->sa_data);
1525                 return 0;
1526         }
1527
1528         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1529             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1530                 return -EADDRNOTAVAIL;
1531
1532         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1533                 netdev_info(netdev, "returning to hw mac address %pM\n",
1534                             hw->mac.addr);
1535         else
1536                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1537
1538         spin_lock_bh(&vsi->mac_filter_hash_lock);
1539         i40e_del_mac_filter(vsi, netdev->dev_addr);
1540         i40e_add_mac_filter(vsi, addr->sa_data);
1541         spin_unlock_bh(&vsi->mac_filter_hash_lock);
1542         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1543         if (vsi->type == I40E_VSI_MAIN) {
1544                 i40e_status ret;
1545
1546                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1547                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1548                                                 addr->sa_data, NULL);
1549                 if (ret)
1550                         netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1551                                     i40e_stat_str(hw, ret),
1552                                     i40e_aq_str(hw, hw->aq.asq_last_status));
1553         }
1554
1555         /* schedule our worker thread which will take care of
1556          * applying the new filter changes
1557          */
1558         i40e_service_event_schedule(vsi->back);
1559         return 0;
1560 }
1561
1562 /**
1563  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1564  * @vsi: the VSI being setup
1565  * @ctxt: VSI context structure
1566  * @enabled_tc: Enabled TCs bitmap
1567  * @is_add: True if called before Add VSI
1568  *
1569  * Setup VSI queue mapping for enabled traffic classes.
1570  **/
1571 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1572                                      struct i40e_vsi_context *ctxt,
1573                                      u8 enabled_tc,
1574                                      bool is_add)
1575 {
1576         struct i40e_pf *pf = vsi->back;
1577         u16 sections = 0;
1578         u8 netdev_tc = 0;
1579         u16 numtc = 0;
1580         u16 qcount;
1581         u8 offset;
1582         u16 qmap;
1583         int i;
1584         u16 num_tc_qps = 0;
1585
1586         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1587         offset = 0;
1588
1589         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1590                 /* Find numtc from enabled TC bitmap */
1591                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1592                         if (enabled_tc & BIT(i)) /* TC is enabled */
1593                                 numtc++;
1594                 }
1595                 if (!numtc) {
1596                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1597                         numtc = 1;
1598                 }
1599         } else {
1600                 /* At least TC0 is enabled in case of non-DCB case */
1601                 numtc = 1;
1602         }
1603
1604         vsi->tc_config.numtc = numtc;
1605         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1606         /* Number of queues per enabled TC */
1607         qcount = vsi->alloc_queue_pairs;
1608
1609         num_tc_qps = qcount / numtc;
1610         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1611
1612         /* Setup queue offset/count for all TCs for given VSI */
1613         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1614                 /* See if the given TC is enabled for the given VSI */
1615                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1616                         /* TC is enabled */
1617                         int pow, num_qps;
1618
1619                         switch (vsi->type) {
1620                         case I40E_VSI_MAIN:
1621                                 qcount = min_t(int, pf->alloc_rss_size,
1622                                                num_tc_qps);
1623                                 break;
1624                         case I40E_VSI_FDIR:
1625                         case I40E_VSI_SRIOV:
1626                         case I40E_VSI_VMDQ2:
1627                         default:
1628                                 qcount = num_tc_qps;
1629                                 WARN_ON(i != 0);
1630                                 break;
1631                         }
1632                         vsi->tc_config.tc_info[i].qoffset = offset;
1633                         vsi->tc_config.tc_info[i].qcount = qcount;
1634
1635                         /* find the next higher power-of-2 of num queue pairs */
1636                         num_qps = qcount;
1637                         pow = 0;
1638                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1639                                 pow++;
1640                                 num_qps >>= 1;
1641                         }
1642
1643                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1644                         qmap =
1645                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1646                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1647
1648                         offset += qcount;
1649                 } else {
1650                         /* TC is not enabled so set the offset to
1651                          * default queue and allocate one queue
1652                          * for the given TC.
1653                          */
1654                         vsi->tc_config.tc_info[i].qoffset = 0;
1655                         vsi->tc_config.tc_info[i].qcount = 1;
1656                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1657
1658                         qmap = 0;
1659                 }
1660                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1661         }
1662
1663         /* Set actual Tx/Rx queue pairs */
1664         vsi->num_queue_pairs = offset;
1665         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1666                 if (vsi->req_queue_pairs > 0)
1667                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1668                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1669                         vsi->num_queue_pairs = pf->num_lan_msix;
1670         }
1671
1672         /* Scheduler section valid can only be set for ADD VSI */
1673         if (is_add) {
1674                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1675
1676                 ctxt->info.up_enable_bits = enabled_tc;
1677         }
1678         if (vsi->type == I40E_VSI_SRIOV) {
1679                 ctxt->info.mapping_flags |=
1680                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1681                 for (i = 0; i < vsi->num_queue_pairs; i++)
1682                         ctxt->info.queue_mapping[i] =
1683                                                cpu_to_le16(vsi->base_queue + i);
1684         } else {
1685                 ctxt->info.mapping_flags |=
1686                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1687                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1688         }
1689         ctxt->info.valid_sections |= cpu_to_le16(sections);
1690 }
1691
1692 /**
1693  * i40e_addr_sync - Callback for dev_(mc|uc)_sync to add address
1694  * @netdev: the netdevice
1695  * @addr: address to add
1696  *
1697  * Called by __dev_(mc|uc)_sync when an address needs to be added. We call
1698  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1699  */
1700 static int i40e_addr_sync(struct net_device *netdev, const u8 *addr)
1701 {
1702         struct i40e_netdev_priv *np = netdev_priv(netdev);
1703         struct i40e_vsi *vsi = np->vsi;
1704
1705         if (i40e_add_mac_filter(vsi, addr))
1706                 return 0;
1707         else
1708                 return -ENOMEM;
1709 }
1710
1711 /**
1712  * i40e_addr_unsync - Callback for dev_(mc|uc)_sync to remove address
1713  * @netdev: the netdevice
1714  * @addr: address to add
1715  *
1716  * Called by __dev_(mc|uc)_sync when an address needs to be removed. We call
1717  * __dev_(uc|mc)_sync from .set_rx_mode and guarantee to hold the hash lock.
1718  */
1719 static int i40e_addr_unsync(struct net_device *netdev, const u8 *addr)
1720 {
1721         struct i40e_netdev_priv *np = netdev_priv(netdev);
1722         struct i40e_vsi *vsi = np->vsi;
1723
1724         i40e_del_mac_filter(vsi, addr);
1725
1726         return 0;
1727 }
1728
1729 /**
1730  * i40e_set_rx_mode - NDO callback to set the netdev filters
1731  * @netdev: network interface device structure
1732  **/
1733 static void i40e_set_rx_mode(struct net_device *netdev)
1734 {
1735         struct i40e_netdev_priv *np = netdev_priv(netdev);
1736         struct i40e_vsi *vsi = np->vsi;
1737
1738         spin_lock_bh(&vsi->mac_filter_hash_lock);
1739
1740         __dev_uc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1741         __dev_mc_sync(netdev, i40e_addr_sync, i40e_addr_unsync);
1742
1743         spin_unlock_bh(&vsi->mac_filter_hash_lock);
1744
1745         /* check for other flag changes */
1746         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1747                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1748                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1749         }
1750
1751         /* schedule our worker thread which will take care of
1752          * applying the new filter changes
1753          */
1754         i40e_service_event_schedule(vsi->back);
1755 }
1756
1757 /**
1758  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1759  * @vsi: Pointer to VSI struct
1760  * @from: Pointer to list which contains MAC filter entries - changes to
1761  *        those entries needs to be undone.
1762  *
1763  * MAC filter entries from this list were slated for deletion.
1764  **/
1765 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1766                                          struct hlist_head *from)
1767 {
1768         struct i40e_mac_filter *f;
1769         struct hlist_node *h;
1770
1771         hlist_for_each_entry_safe(f, h, from, hlist) {
1772                 u64 key = i40e_addr_to_hkey(f->macaddr);
1773
1774                 /* Move the element back into MAC filter list*/
1775                 hlist_del(&f->hlist);
1776                 hash_add(vsi->mac_filter_hash, &f->hlist, key);
1777         }
1778 }
1779
1780 /**
1781  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1782  * @vsi: Pointer to vsi struct
1783  * @from: Pointer to list which contains MAC filter entries - changes to
1784  *        those entries needs to be undone.
1785  *
1786  * MAC filter entries from this list were slated for addition.
1787  **/
1788 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi,
1789                                          struct hlist_head *from)
1790 {
1791         struct i40e_new_mac_filter *new;
1792         struct hlist_node *h;
1793
1794         hlist_for_each_entry_safe(new, h, from, hlist) {
1795                 /* We can simply free the wrapper structure */
1796                 hlist_del(&new->hlist);
1797                 kfree(new);
1798         }
1799 }
1800
1801 /**
1802  * i40e_next_entry - Get the next non-broadcast filter from a list
1803  * @next: pointer to filter in list
1804  *
1805  * Returns the next non-broadcast filter in the list. Required so that we
1806  * ignore broadcast filters within the list, since these are not handled via
1807  * the normal firmware update path.
1808  */
1809 static
1810 struct i40e_new_mac_filter *i40e_next_filter(struct i40e_new_mac_filter *next)
1811 {
1812         hlist_for_each_entry_continue(next, hlist) {
1813                 if (!is_broadcast_ether_addr(next->f->macaddr))
1814                         return next;
1815         }
1816
1817         return NULL;
1818 }
1819
1820 /**
1821  * i40e_update_filter_state - Update filter state based on return data
1822  * from firmware
1823  * @count: Number of filters added
1824  * @add_list: return data from fw
1825  * @head: pointer to first filter in current batch
1826  *
1827  * MAC filter entries from list were slated to be added to device. Returns
1828  * number of successful filters. Note that 0 does NOT mean success!
1829  **/
1830 static int
1831 i40e_update_filter_state(int count,
1832                          struct i40e_aqc_add_macvlan_element_data *add_list,
1833                          struct i40e_new_mac_filter *add_head)
1834 {
1835         int retval = 0;
1836         int i;
1837
1838         for (i = 0; i < count; i++) {
1839                 /* Always check status of each filter. We don't need to check
1840                  * the firmware return status because we pre-set the filter
1841                  * status to I40E_AQC_MM_ERR_NO_RES when sending the filter
1842                  * request to the adminq. Thus, if it no longer matches then
1843                  * we know the filter is active.
1844                  */
1845                 if (add_list[i].match_method == I40E_AQC_MM_ERR_NO_RES) {
1846                         add_head->state = I40E_FILTER_FAILED;
1847                 } else {
1848                         add_head->state = I40E_FILTER_ACTIVE;
1849                         retval++;
1850                 }
1851
1852                 add_head = i40e_next_filter(add_head);
1853                 if (!add_head)
1854                         break;
1855         }
1856
1857         return retval;
1858 }
1859
1860 /**
1861  * i40e_aqc_del_filters - Request firmware to delete a set of filters
1862  * @vsi: ptr to the VSI
1863  * @vsi_name: name to display in messages
1864  * @list: the list of filters to send to firmware
1865  * @num_del: the number of filters to delete
1866  * @retval: Set to -EIO on failure to delete
1867  *
1868  * Send a request to firmware via AdminQ to delete a set of filters. Uses
1869  * *retval instead of a return value so that success does not force ret_val to
1870  * be set to 0. This ensures that a sequence of calls to this function
1871  * preserve the previous value of *retval on successful delete.
1872  */
1873 static
1874 void i40e_aqc_del_filters(struct i40e_vsi *vsi, const char *vsi_name,
1875                           struct i40e_aqc_remove_macvlan_element_data *list,
1876                           int num_del, int *retval)
1877 {
1878         struct i40e_hw *hw = &vsi->back->hw;
1879         i40e_status aq_ret;
1880         int aq_err;
1881
1882         aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, list, num_del, NULL);
1883         aq_err = hw->aq.asq_last_status;
1884
1885         /* Explicitly ignore and do not report when firmware returns ENOENT */
1886         if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1887                 *retval = -EIO;
1888                 dev_info(&vsi->back->pdev->dev,
1889                          "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1890                          vsi_name, i40e_stat_str(hw, aq_ret),
1891                          i40e_aq_str(hw, aq_err));
1892         }
1893 }
1894
1895 /**
1896  * i40e_aqc_add_filters - Request firmware to add a set of filters
1897  * @vsi: ptr to the VSI
1898  * @vsi_name: name to display in messages
1899  * @list: the list of filters to send to firmware
1900  * @add_head: Position in the add hlist
1901  * @num_add: the number of filters to add
1902  * @promisc_change: set to true on exit if promiscuous mode was forced on
1903  *
1904  * Send a request to firmware via AdminQ to add a chunk of filters. Will set
1905  * promisc_changed to true if the firmware has run out of space for more
1906  * filters.
1907  */
1908 static
1909 void i40e_aqc_add_filters(struct i40e_vsi *vsi, const char *vsi_name,
1910                           struct i40e_aqc_add_macvlan_element_data *list,
1911                           struct i40e_new_mac_filter *add_head,
1912                           int num_add, bool *promisc_changed)
1913 {
1914         struct i40e_hw *hw = &vsi->back->hw;
1915         int aq_err, fcnt;
1916
1917         i40e_aq_add_macvlan(hw, vsi->seid, list, num_add, NULL);
1918         aq_err = hw->aq.asq_last_status;
1919         fcnt = i40e_update_filter_state(num_add, list, add_head);
1920
1921         if (fcnt != num_add) {
1922                 *promisc_changed = true;
1923                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
1924                 dev_warn(&vsi->back->pdev->dev,
1925                          "Error %s adding RX filters on %s, promiscuous mode forced on\n",
1926                          i40e_aq_str(hw, aq_err),
1927                          vsi_name);
1928         }
1929 }
1930
1931 /**
1932  * i40e_aqc_broadcast_filter - Set promiscuous broadcast flags
1933  * @vsi: pointer to the VSI
1934  * @f: filter data
1935  *
1936  * This function sets or clears the promiscuous broadcast flags for VLAN
1937  * filters in order to properly receive broadcast frames. Assumes that only
1938  * broadcast filters are passed.
1939  *
1940  * Returns status indicating success or failure;
1941  **/
1942 static i40e_status
1943 i40e_aqc_broadcast_filter(struct i40e_vsi *vsi, const char *vsi_name,
1944                           struct i40e_mac_filter *f)
1945 {
1946         bool enable = f->state == I40E_FILTER_NEW;
1947         struct i40e_hw *hw = &vsi->back->hw;
1948         i40e_status aq_ret;
1949
1950         if (f->vlan == I40E_VLAN_ANY) {
1951                 aq_ret = i40e_aq_set_vsi_broadcast(hw,
1952                                                    vsi->seid,
1953                                                    enable,
1954                                                    NULL);
1955         } else {
1956                 aq_ret = i40e_aq_set_vsi_bc_promisc_on_vlan(hw,
1957                                                             vsi->seid,
1958                                                             enable,
1959                                                             f->vlan,
1960                                                             NULL);
1961         }
1962
1963         if (aq_ret)
1964                 dev_warn(&vsi->back->pdev->dev,
1965                          "Error %s setting broadcast promiscuous mode on %s\n",
1966                          i40e_aq_str(hw, hw->aq.asq_last_status),
1967                          vsi_name);
1968
1969         return aq_ret;
1970 }
1971
1972 /**
1973  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1974  * @vsi: ptr to the VSI
1975  *
1976  * Push any outstanding VSI filter changes through the AdminQ.
1977  *
1978  * Returns 0 or error value
1979  **/
1980 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1981 {
1982         struct hlist_head tmp_add_list, tmp_del_list;
1983         struct i40e_mac_filter *f;
1984         struct i40e_new_mac_filter *new, *add_head = NULL;
1985         struct i40e_hw *hw = &vsi->back->hw;
1986         unsigned int failed_filters = 0;
1987         unsigned int vlan_filters = 0;
1988         bool promisc_changed = false;
1989         char vsi_name[16] = "PF";
1990         int filter_list_len = 0;
1991         i40e_status aq_ret = 0;
1992         u32 changed_flags = 0;
1993         struct hlist_node *h;
1994         struct i40e_pf *pf;
1995         int num_add = 0;
1996         int num_del = 0;
1997         int retval = 0;
1998         u16 cmd_flags;
1999         int list_size;
2000         int bkt;
2001
2002         /* empty array typed pointers, kcalloc later */
2003         struct i40e_aqc_add_macvlan_element_data *add_list;
2004         struct i40e_aqc_remove_macvlan_element_data *del_list;
2005
2006         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
2007                 usleep_range(1000, 2000);
2008         pf = vsi->back;
2009
2010         if (vsi->netdev) {
2011                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
2012                 vsi->current_netdev_flags = vsi->netdev->flags;
2013         }
2014
2015         INIT_HLIST_HEAD(&tmp_add_list);
2016         INIT_HLIST_HEAD(&tmp_del_list);
2017
2018         if (vsi->type == I40E_VSI_SRIOV)
2019                 snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
2020         else if (vsi->type != I40E_VSI_MAIN)
2021                 snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
2022
2023         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
2024                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
2025
2026                 spin_lock_bh(&vsi->mac_filter_hash_lock);
2027                 /* Create a list of filters to delete. */
2028                 hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2029                         if (f->state == I40E_FILTER_REMOVE) {
2030                                 /* Move the element into temporary del_list */
2031                                 hash_del(&f->hlist);
2032                                 hlist_add_head(&f->hlist, &tmp_del_list);
2033
2034                                 /* Avoid counting removed filters */
2035                                 continue;
2036                         }
2037                         if (f->state == I40E_FILTER_NEW) {
2038                                 /* Create a temporary i40e_new_mac_filter */
2039                                 new = kzalloc(sizeof(*new), GFP_ATOMIC);
2040                                 if (!new)
2041                                         goto err_no_memory_locked;
2042
2043                                 /* Store pointer to the real filter */
2044                                 new->f = f;
2045                                 new->state = f->state;
2046
2047                                 /* Add it to the hash list */
2048                                 hlist_add_head(&new->hlist, &tmp_add_list);
2049                         }
2050
2051                         /* Count the number of active (current and new) VLAN
2052                          * filters we have now. Does not count filters which
2053                          * are marked for deletion.
2054                          */
2055                         if (f->vlan > 0)
2056                                 vlan_filters++;
2057                 }
2058
2059                 retval = i40e_correct_mac_vlan_filters(vsi,
2060                                                        &tmp_add_list,
2061                                                        &tmp_del_list,
2062                                                        vlan_filters);
2063                 if (retval)
2064                         goto err_no_memory_locked;
2065
2066                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
2067         }
2068
2069         /* Now process 'del_list' outside the lock */
2070         if (!hlist_empty(&tmp_del_list)) {
2071                 filter_list_len = hw->aq.asq_buf_size /
2072                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
2073                 list_size = filter_list_len *
2074                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
2075                 del_list = kzalloc(list_size, GFP_ATOMIC);
2076                 if (!del_list)
2077                         goto err_no_memory;
2078
2079                 hlist_for_each_entry_safe(f, h, &tmp_del_list, hlist) {
2080                         cmd_flags = 0;
2081
2082                         /* handle broadcast filters by updating the broadcast
2083                          * promiscuous flag and release filter list.
2084                          */
2085                         if (is_broadcast_ether_addr(f->macaddr)) {
2086                                 i40e_aqc_broadcast_filter(vsi, vsi_name, f);
2087
2088                                 hlist_del(&f->hlist);
2089                                 kfree(f);
2090                                 continue;
2091                         }
2092
2093                         /* add to delete list */
2094                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
2095                         if (f->vlan == I40E_VLAN_ANY) {
2096                                 del_list[num_del].vlan_tag = 0;
2097                                 cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
2098                         } else {
2099                                 del_list[num_del].vlan_tag =
2100                                         cpu_to_le16((u16)(f->vlan));
2101                         }
2102
2103                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
2104                         del_list[num_del].flags = cmd_flags;
2105                         num_del++;
2106
2107                         /* flush a full buffer */
2108                         if (num_del == filter_list_len) {
2109                                 i40e_aqc_del_filters(vsi, vsi_name, del_list,
2110                                                      num_del, &retval);
2111                                 memset(del_list, 0, list_size);
2112                                 num_del = 0;
2113                         }
2114                         /* Release memory for MAC filter entries which were
2115                          * synced up with HW.
2116                          */
2117                         hlist_del(&f->hlist);
2118                         kfree(f);
2119                 }
2120
2121                 if (num_del) {
2122                         i40e_aqc_del_filters(vsi, vsi_name, del_list,
2123                                              num_del, &retval);
2124                 }
2125
2126                 kfree(del_list);
2127                 del_list = NULL;
2128         }
2129
2130         if (!hlist_empty(&tmp_add_list)) {
2131                 /* Do all the adds now. */
2132                 filter_list_len = hw->aq.asq_buf_size /
2133                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2134                 list_size = filter_list_len *
2135                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2136                 add_list = kzalloc(list_size, GFP_ATOMIC);
2137                 if (!add_list)
2138                         goto err_no_memory;
2139
2140                 num_add = 0;
2141                 hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2142                         if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2143                                      &vsi->state)) {
2144                                 new->state = I40E_FILTER_FAILED;
2145                                 continue;
2146                         }
2147
2148                         /* handle broadcast filters by updating the broadcast
2149                          * promiscuous flag instead of adding a MAC filter.
2150                          */
2151                         if (is_broadcast_ether_addr(new->f->macaddr)) {
2152                                 if (i40e_aqc_broadcast_filter(vsi, vsi_name,
2153                                                               new->f))
2154                                         new->state = I40E_FILTER_FAILED;
2155                                 else
2156                                         new->state = I40E_FILTER_ACTIVE;
2157                                 continue;
2158                         }
2159
2160                         /* add to add array */
2161                         if (num_add == 0)
2162                                 add_head = new;
2163                         cmd_flags = 0;
2164                         ether_addr_copy(add_list[num_add].mac_addr,
2165                                         new->f->macaddr);
2166                         if (new->f->vlan == I40E_VLAN_ANY) {
2167                                 add_list[num_add].vlan_tag = 0;
2168                                 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2169                         } else {
2170                                 add_list[num_add].vlan_tag =
2171                                         cpu_to_le16((u16)(new->f->vlan));
2172                         }
2173                         add_list[num_add].queue_number = 0;
2174                         /* set invalid match method for later detection */
2175                         add_list[num_add].match_method = I40E_AQC_MM_ERR_NO_RES;
2176                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2177                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2178                         num_add++;
2179
2180                         /* flush a full buffer */
2181                         if (num_add == filter_list_len) {
2182                                 i40e_aqc_add_filters(vsi, vsi_name, add_list,
2183                                                      add_head, num_add,
2184                                                      &promisc_changed);
2185                                 memset(add_list, 0, list_size);
2186                                 num_add = 0;
2187                         }
2188                 }
2189                 if (num_add) {
2190                         i40e_aqc_add_filters(vsi, vsi_name, add_list, add_head,
2191                                              num_add, &promisc_changed);
2192                 }
2193                 /* Now move all of the filters from the temp add list back to
2194                  * the VSI's list.
2195                  */
2196                 spin_lock_bh(&vsi->mac_filter_hash_lock);
2197                 hlist_for_each_entry_safe(new, h, &tmp_add_list, hlist) {
2198                         /* Only update the state if we're still NEW */
2199                         if (new->f->state == I40E_FILTER_NEW)
2200                                 new->f->state = new->state;
2201                         hlist_del(&new->hlist);
2202                         kfree(new);
2203                 }
2204                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
2205                 kfree(add_list);
2206                 add_list = NULL;
2207         }
2208
2209         /* Determine the number of active and failed filters. */
2210         spin_lock_bh(&vsi->mac_filter_hash_lock);
2211         vsi->active_filters = 0;
2212         hash_for_each(vsi->mac_filter_hash, bkt, f, hlist) {
2213                 if (f->state == I40E_FILTER_ACTIVE)
2214                         vsi->active_filters++;
2215                 else if (f->state == I40E_FILTER_FAILED)
2216                         failed_filters++;
2217         }
2218         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2219
2220         /* If promiscuous mode has changed, we need to calculate a new
2221          * threshold for when we are safe to exit
2222          */
2223         if (promisc_changed)
2224                 vsi->promisc_threshold = (vsi->active_filters * 3) / 4;
2225
2226         /* Check if we are able to exit overflow promiscuous mode. We can
2227          * safely exit if we didn't just enter, we no longer have any failed
2228          * filters, and we have reduced filters below the threshold value.
2229          */
2230         if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state) &&
2231             !promisc_changed && !failed_filters &&
2232             (vsi->active_filters < vsi->promisc_threshold)) {
2233                 dev_info(&pf->pdev->dev,
2234                          "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2235                          vsi_name);
2236                 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2237                 promisc_changed = true;
2238                 vsi->promisc_threshold = 0;
2239         }
2240
2241         /* if the VF is not trusted do not do promisc */
2242         if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2243                 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2244                 goto out;
2245         }
2246
2247         /* check for changes in promiscuous modes */
2248         if (changed_flags & IFF_ALLMULTI) {
2249                 bool cur_multipromisc;
2250
2251                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2252                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2253                                                                vsi->seid,
2254                                                                cur_multipromisc,
2255                                                                NULL);
2256                 if (aq_ret) {
2257                         retval = i40e_aq_rc_to_posix(aq_ret,
2258                                                      hw->aq.asq_last_status);
2259                         dev_info(&pf->pdev->dev,
2260                                  "set multi promisc failed on %s, err %s aq_err %s\n",
2261                                  vsi_name,
2262                                  i40e_stat_str(hw, aq_ret),
2263                                  i40e_aq_str(hw, hw->aq.asq_last_status));
2264                 }
2265         }
2266         if ((changed_flags & IFF_PROMISC) ||
2267             (promisc_changed &&
2268              test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))) {
2269                 bool cur_promisc;
2270
2271                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2272                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2273                                         &vsi->state));
2274                 if ((vsi->type == I40E_VSI_MAIN) &&
2275                     (pf->lan_veb != I40E_NO_VEB) &&
2276                     !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2277                         /* set defport ON for Main VSI instead of true promisc
2278                          * this way we will get all unicast/multicast and VLAN
2279                          * promisc behavior but will not get VF or VMDq traffic
2280                          * replicated on the Main VSI.
2281                          */
2282                         if (pf->cur_promisc != cur_promisc) {
2283                                 pf->cur_promisc = cur_promisc;
2284                                 if (cur_promisc)
2285                                         aq_ret =
2286                                               i40e_aq_set_default_vsi(hw,
2287                                                                       vsi->seid,
2288                                                                       NULL);
2289                                 else
2290                                         aq_ret =
2291                                             i40e_aq_clear_default_vsi(hw,
2292                                                                       vsi->seid,
2293                                                                       NULL);
2294                                 if (aq_ret) {
2295                                         retval = i40e_aq_rc_to_posix(aq_ret,
2296                                                         hw->aq.asq_last_status);
2297                                         dev_info(&pf->pdev->dev,
2298                                                  "Set default VSI failed on %s, err %s, aq_err %s\n",
2299                                                  vsi_name,
2300                                                  i40e_stat_str(hw, aq_ret),
2301                                                  i40e_aq_str(hw,
2302                                                      hw->aq.asq_last_status));
2303                                 }
2304                         }
2305                 } else {
2306                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2307                                                           hw,
2308                                                           vsi->seid,
2309                                                           cur_promisc, NULL,
2310                                                           true);
2311                         if (aq_ret) {
2312                                 retval =
2313                                 i40e_aq_rc_to_posix(aq_ret,
2314                                                     hw->aq.asq_last_status);
2315                                 dev_info(&pf->pdev->dev,
2316                                          "set unicast promisc failed on %s, err %s, aq_err %s\n",
2317                                          vsi_name,
2318                                          i40e_stat_str(hw, aq_ret),
2319                                          i40e_aq_str(hw,
2320                                                      hw->aq.asq_last_status));
2321                         }
2322                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2323                                                           hw,
2324                                                           vsi->seid,
2325                                                           cur_promisc, NULL);
2326                         if (aq_ret) {
2327                                 retval =
2328                                 i40e_aq_rc_to_posix(aq_ret,
2329                                                     hw->aq.asq_last_status);
2330                                 dev_info(&pf->pdev->dev,
2331                                          "set multicast promisc failed on %s, err %s, aq_err %s\n",
2332                                          vsi_name,
2333                                          i40e_stat_str(hw, aq_ret),
2334                                          i40e_aq_str(hw,
2335                                                      hw->aq.asq_last_status));
2336                         }
2337                 }
2338                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2339                                                    vsi->seid,
2340                                                    cur_promisc, NULL);
2341                 if (aq_ret) {
2342                         retval = i40e_aq_rc_to_posix(aq_ret,
2343                                                      pf->hw.aq.asq_last_status);
2344                         dev_info(&pf->pdev->dev,
2345                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2346                                          i40e_stat_str(hw, aq_ret),
2347                                          i40e_aq_str(hw,
2348                                                      hw->aq.asq_last_status));
2349                 }
2350         }
2351 out:
2352         /* if something went wrong then set the changed flag so we try again */
2353         if (retval)
2354                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2355
2356         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2357         return retval;
2358
2359 err_no_memory:
2360         /* Restore elements on the temporary add and delete lists */
2361         spin_lock_bh(&vsi->mac_filter_hash_lock);
2362 err_no_memory_locked:
2363         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
2364         i40e_undo_add_filter_entries(vsi, &tmp_add_list);
2365         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2366
2367         vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2368         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2369         return -ENOMEM;
2370 }
2371
2372 /**
2373  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2374  * @pf: board private structure
2375  **/
2376 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2377 {
2378         int v;
2379
2380         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2381                 return;
2382         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2383
2384         for (v = 0; v < pf->num_alloc_vsi; v++) {
2385                 if (pf->vsi[v] &&
2386                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2387                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2388
2389                         if (ret) {
2390                                 /* come back and try again later */
2391                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2392                                 break;
2393                         }
2394                 }
2395         }
2396 }
2397
2398 /**
2399  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2400  * @netdev: network interface device structure
2401  * @new_mtu: new value for maximum frame size
2402  *
2403  * Returns 0 on success, negative on failure
2404  **/
2405 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2406 {
2407         struct i40e_netdev_priv *np = netdev_priv(netdev);
2408         struct i40e_vsi *vsi = np->vsi;
2409         struct i40e_pf *pf = vsi->back;
2410
2411         netdev_info(netdev, "changing MTU from %d to %d\n",
2412                     netdev->mtu, new_mtu);
2413         netdev->mtu = new_mtu;
2414         if (netif_running(netdev))
2415                 i40e_vsi_reinit_locked(vsi);
2416         pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
2417                       I40E_FLAG_CLIENT_L2_CHANGE);
2418         return 0;
2419 }
2420
2421 /**
2422  * i40e_ioctl - Access the hwtstamp interface
2423  * @netdev: network interface device structure
2424  * @ifr: interface request data
2425  * @cmd: ioctl command
2426  **/
2427 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2428 {
2429         struct i40e_netdev_priv *np = netdev_priv(netdev);
2430         struct i40e_pf *pf = np->vsi->back;
2431
2432         switch (cmd) {
2433         case SIOCGHWTSTAMP:
2434                 return i40e_ptp_get_ts_config(pf, ifr);
2435         case SIOCSHWTSTAMP:
2436                 return i40e_ptp_set_ts_config(pf, ifr);
2437         default:
2438                 return -EOPNOTSUPP;
2439         }
2440 }
2441
2442 /**
2443  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2444  * @vsi: the vsi being adjusted
2445  **/
2446 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2447 {
2448         struct i40e_vsi_context ctxt;
2449         i40e_status ret;
2450
2451         if ((vsi->info.valid_sections &
2452              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2453             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2454                 return;  /* already enabled */
2455
2456         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2457         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2458                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2459
2460         ctxt.seid = vsi->seid;
2461         ctxt.info = vsi->info;
2462         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2463         if (ret) {
2464                 dev_info(&vsi->back->pdev->dev,
2465                          "update vlan stripping failed, err %s aq_err %s\n",
2466                          i40e_stat_str(&vsi->back->hw, ret),
2467                          i40e_aq_str(&vsi->back->hw,
2468                                      vsi->back->hw.aq.asq_last_status));
2469         }
2470 }
2471
2472 /**
2473  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2474  * @vsi: the vsi being adjusted
2475  **/
2476 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2477 {
2478         struct i40e_vsi_context ctxt;
2479         i40e_status ret;
2480
2481         if ((vsi->info.valid_sections &
2482              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2483             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2484              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2485                 return;  /* already disabled */
2486
2487         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2488         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2489                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2490
2491         ctxt.seid = vsi->seid;
2492         ctxt.info = vsi->info;
2493         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2494         if (ret) {
2495                 dev_info(&vsi->back->pdev->dev,
2496                          "update vlan stripping failed, err %s aq_err %s\n",
2497                          i40e_stat_str(&vsi->back->hw, ret),
2498                          i40e_aq_str(&vsi->back->hw,
2499                                      vsi->back->hw.aq.asq_last_status));
2500         }
2501 }
2502
2503 /**
2504  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2505  * @netdev: network interface to be adjusted
2506  * @features: netdev features to test if VLAN offload is enabled or not
2507  **/
2508 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2509 {
2510         struct i40e_netdev_priv *np = netdev_priv(netdev);
2511         struct i40e_vsi *vsi = np->vsi;
2512
2513         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2514                 i40e_vlan_stripping_enable(vsi);
2515         else
2516                 i40e_vlan_stripping_disable(vsi);
2517 }
2518
2519 /**
2520  * i40e_add_vlan_all_mac - Add a MAC/VLAN filter for each existing MAC address
2521  * @vsi: the vsi being configured
2522  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2523  *
2524  * This is a helper function for adding a new MAC/VLAN filter with the
2525  * specified VLAN for each existing MAC address already in the hash table.
2526  * This function does *not* perform any accounting to update filters based on
2527  * VLAN mode.
2528  *
2529  * NOTE: this function expects to be called while under the
2530  * mac_filter_hash_lock
2531  **/
2532 int i40e_add_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2533 {
2534         struct i40e_mac_filter *f, *add_f;
2535         struct hlist_node *h;
2536         int bkt;
2537
2538         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2539                 if (f->state == I40E_FILTER_REMOVE)
2540                         continue;
2541                 add_f = i40e_add_filter(vsi, f->macaddr, vid);
2542                 if (!add_f) {
2543                         dev_info(&vsi->back->pdev->dev,
2544                                  "Could not add vlan filter %d for %pM\n",
2545                                  vid, f->macaddr);
2546                         return -ENOMEM;
2547                 }
2548         }
2549
2550         return 0;
2551 }
2552
2553 /**
2554  * i40e_vsi_add_vlan - Add VSI membership for given VLAN
2555  * @vsi: the VSI being configured
2556  * @vid: VLAN id to be added
2557  **/
2558 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, u16 vid)
2559 {
2560         int err;
2561
2562         if (!vid || vsi->info.pvid)
2563                 return -EINVAL;
2564
2565         /* Locked once because all functions invoked below iterates list*/
2566         spin_lock_bh(&vsi->mac_filter_hash_lock);
2567         err = i40e_add_vlan_all_mac(vsi, vid);
2568         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2569         if (err)
2570                 return err;
2571
2572         /* schedule our worker thread which will take care of
2573          * applying the new filter changes
2574          */
2575         i40e_service_event_schedule(vsi->back);
2576         return 0;
2577 }
2578
2579 /**
2580  * i40e_rm_vlan_all_mac - Remove MAC/VLAN pair for all MAC with the given VLAN
2581  * @vsi: the vsi being configured
2582  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2583  *
2584  * This function should be used to remove all VLAN filters which match the
2585  * given VID. It does not schedule the service event and does not take the
2586  * mac_filter_hash_lock so it may be combined with other operations under
2587  * a single invocation of the mac_filter_hash_lock.
2588  *
2589  * NOTE: this function expects to be called while under the
2590  * mac_filter_hash_lock
2591  */
2592 void i40e_rm_vlan_all_mac(struct i40e_vsi *vsi, s16 vid)
2593 {
2594         struct i40e_mac_filter *f;
2595         struct hlist_node *h;
2596         int bkt;
2597
2598         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
2599                 if (f->vlan == vid)
2600                         __i40e_del_filter(vsi, f);
2601         }
2602 }
2603
2604 /**
2605  * i40e_vsi_kill_vlan - Remove VSI membership for given VLAN
2606  * @vsi: the VSI being configured
2607  * @vid: VLAN id to be removed
2608  **/
2609 void i40e_vsi_kill_vlan(struct i40e_vsi *vsi, u16 vid)
2610 {
2611         if (!vid || vsi->info.pvid)
2612                 return;
2613
2614         spin_lock_bh(&vsi->mac_filter_hash_lock);
2615         i40e_rm_vlan_all_mac(vsi, vid);
2616         spin_unlock_bh(&vsi->mac_filter_hash_lock);
2617
2618         /* schedule our worker thread which will take care of
2619          * applying the new filter changes
2620          */
2621         i40e_service_event_schedule(vsi->back);
2622 }
2623
2624 /**
2625  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2626  * @netdev: network interface to be adjusted
2627  * @vid: vlan id to be added
2628  *
2629  * net_device_ops implementation for adding vlan ids
2630  **/
2631 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2632                                 __always_unused __be16 proto, u16 vid)
2633 {
2634         struct i40e_netdev_priv *np = netdev_priv(netdev);
2635         struct i40e_vsi *vsi = np->vsi;
2636         int ret = 0;
2637
2638         if (vid >= VLAN_N_VID)
2639                 return -EINVAL;
2640
2641         /* If the network stack called us with vid = 0 then
2642          * it is asking to receive priority tagged packets with
2643          * vlan id 0.  Our HW receives them by default when configured
2644          * to receive untagged packets so there is no need to add an
2645          * extra filter for vlan 0 tagged packets.
2646          */
2647         if (vid)
2648                 ret = i40e_vsi_add_vlan(vsi, vid);
2649
2650         if (!ret)
2651                 set_bit(vid, vsi->active_vlans);
2652
2653         return ret;
2654 }
2655
2656 /**
2657  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2658  * @netdev: network interface to be adjusted
2659  * @vid: vlan id to be removed
2660  *
2661  * net_device_ops implementation for removing vlan ids
2662  **/
2663 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2664                                  __always_unused __be16 proto, u16 vid)
2665 {
2666         struct i40e_netdev_priv *np = netdev_priv(netdev);
2667         struct i40e_vsi *vsi = np->vsi;
2668
2669         /* return code is ignored as there is nothing a user
2670          * can do about failure to remove and a log message was
2671          * already printed from the other function
2672          */
2673         i40e_vsi_kill_vlan(vsi, vid);
2674
2675         clear_bit(vid, vsi->active_vlans);
2676
2677         return 0;
2678 }
2679
2680 /**
2681  * i40e_macaddr_init - explicitly write the mac address filters
2682  *
2683  * @vsi: pointer to the vsi
2684  * @macaddr: the MAC address
2685  *
2686  * This is needed when the macaddr has been obtained by other
2687  * means than the default, e.g., from Open Firmware or IDPROM.
2688  * Returns 0 on success, negative on failure
2689  **/
2690 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2691 {
2692         int ret;
2693         struct i40e_aqc_add_macvlan_element_data element;
2694
2695         ret = i40e_aq_mac_address_write(&vsi->back->hw,
2696                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
2697                                         macaddr, NULL);
2698         if (ret) {
2699                 dev_info(&vsi->back->pdev->dev,
2700                          "Addr change for VSI failed: %d\n", ret);
2701                 return -EADDRNOTAVAIL;
2702         }
2703
2704         memset(&element, 0, sizeof(element));
2705         ether_addr_copy(element.mac_addr, macaddr);
2706         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2707         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2708         if (ret) {
2709                 dev_info(&vsi->back->pdev->dev,
2710                          "add filter failed err %s aq_err %s\n",
2711                          i40e_stat_str(&vsi->back->hw, ret),
2712                          i40e_aq_str(&vsi->back->hw,
2713                                      vsi->back->hw.aq.asq_last_status));
2714         }
2715         return ret;
2716 }
2717
2718 /**
2719  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2720  * @vsi: the vsi being brought back up
2721  **/
2722 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2723 {
2724         u16 vid;
2725
2726         if (!vsi->netdev)
2727                 return;
2728
2729         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2730
2731         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2732                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2733                                      vid);
2734 }
2735
2736 /**
2737  * i40e_vsi_add_pvid - Add pvid for the VSI
2738  * @vsi: the vsi being adjusted
2739  * @vid: the vlan id to set as a PVID
2740  **/
2741 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2742 {
2743         struct i40e_vsi_context ctxt;
2744         i40e_status ret;
2745
2746         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2747         vsi->info.pvid = cpu_to_le16(vid);
2748         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2749                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2750                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2751
2752         ctxt.seid = vsi->seid;
2753         ctxt.info = vsi->info;
2754         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2755         if (ret) {
2756                 dev_info(&vsi->back->pdev->dev,
2757                          "add pvid failed, err %s aq_err %s\n",
2758                          i40e_stat_str(&vsi->back->hw, ret),
2759                          i40e_aq_str(&vsi->back->hw,
2760                                      vsi->back->hw.aq.asq_last_status));
2761                 return -ENOENT;
2762         }
2763
2764         return 0;
2765 }
2766
2767 /**
2768  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2769  * @vsi: the vsi being adjusted
2770  *
2771  * Just use the vlan_rx_register() service to put it back to normal
2772  **/
2773 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2774 {
2775         i40e_vlan_stripping_disable(vsi);
2776
2777         vsi->info.pvid = 0;
2778 }
2779
2780 /**
2781  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2782  * @vsi: ptr to the VSI
2783  *
2784  * If this function returns with an error, then it's possible one or
2785  * more of the rings is populated (while the rest are not).  It is the
2786  * callers duty to clean those orphaned rings.
2787  *
2788  * Return 0 on success, negative on failure
2789  **/
2790 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2791 {
2792         int i, err = 0;
2793
2794         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2795                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2796
2797         return err;
2798 }
2799
2800 /**
2801  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2802  * @vsi: ptr to the VSI
2803  *
2804  * Free VSI's transmit software resources
2805  **/
2806 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2807 {
2808         int i;
2809
2810         if (!vsi->tx_rings)
2811                 return;
2812
2813         for (i = 0; i < vsi->num_queue_pairs; i++)
2814                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2815                         i40e_free_tx_resources(vsi->tx_rings[i]);
2816 }
2817
2818 /**
2819  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2820  * @vsi: ptr to the VSI
2821  *
2822  * If this function returns with an error, then it's possible one or
2823  * more of the rings is populated (while the rest are not).  It is the
2824  * callers duty to clean those orphaned rings.
2825  *
2826  * Return 0 on success, negative on failure
2827  **/
2828 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2829 {
2830         int i, err = 0;
2831
2832         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2833                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2834         return err;
2835 }
2836
2837 /**
2838  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2839  * @vsi: ptr to the VSI
2840  *
2841  * Free all receive software resources
2842  **/
2843 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2844 {
2845         int i;
2846
2847         if (!vsi->rx_rings)
2848                 return;
2849
2850         for (i = 0; i < vsi->num_queue_pairs; i++)
2851                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2852                         i40e_free_rx_resources(vsi->rx_rings[i]);
2853 }
2854
2855 /**
2856  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2857  * @ring: The Tx ring to configure
2858  *
2859  * This enables/disables XPS for a given Tx descriptor ring
2860  * based on the TCs enabled for the VSI that ring belongs to.
2861  **/
2862 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2863 {
2864         struct i40e_vsi *vsi = ring->vsi;
2865         cpumask_var_t mask;
2866
2867         if (!ring->q_vector || !ring->netdev)
2868                 return;
2869
2870         /* Single TC mode enable XPS */
2871         if (vsi->tc_config.numtc <= 1) {
2872                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2873                         netif_set_xps_queue(ring->netdev,
2874                                             &ring->q_vector->affinity_mask,
2875                                             ring->queue_index);
2876         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2877                 /* Disable XPS to allow selection based on TC */
2878                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2879                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2880                 free_cpumask_var(mask);
2881         }
2882
2883         /* schedule our worker thread which will take care of
2884          * applying the new filter changes
2885          */
2886         i40e_service_event_schedule(vsi->back);
2887 }
2888
2889 /**
2890  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2891  * @ring: The Tx ring to configure
2892  *
2893  * Configure the Tx descriptor ring in the HMC context.
2894  **/
2895 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2896 {
2897         struct i40e_vsi *vsi = ring->vsi;
2898         u16 pf_q = vsi->base_queue + ring->queue_index;
2899         struct i40e_hw *hw = &vsi->back->hw;
2900         struct i40e_hmc_obj_txq tx_ctx;
2901         i40e_status err = 0;
2902         u32 qtx_ctl = 0;
2903
2904         /* some ATR related tx ring init */
2905         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2906                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2907                 ring->atr_count = 0;
2908         } else {
2909                 ring->atr_sample_rate = 0;
2910         }
2911
2912         /* configure XPS */
2913         i40e_config_xps_tx_ring(ring);
2914
2915         /* clear the context structure first */
2916         memset(&tx_ctx, 0, sizeof(tx_ctx));
2917
2918         tx_ctx.new_context = 1;
2919         tx_ctx.base = (ring->dma / 128);
2920         tx_ctx.qlen = ring->count;
2921         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2922                                                I40E_FLAG_FD_ATR_ENABLED));
2923         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2924         /* FDIR VSI tx ring can still use RS bit and writebacks */
2925         if (vsi->type != I40E_VSI_FDIR)
2926                 tx_ctx.head_wb_ena = 1;
2927         tx_ctx.head_wb_addr = ring->dma +
2928                               (ring->count * sizeof(struct i40e_tx_desc));
2929
2930         /* As part of VSI creation/update, FW allocates certain
2931          * Tx arbitration queue sets for each TC enabled for
2932          * the VSI. The FW returns the handles to these queue
2933          * sets as part of the response buffer to Add VSI,
2934          * Update VSI, etc. AQ commands. It is expected that
2935          * these queue set handles be associated with the Tx
2936          * queues by the driver as part of the TX queue context
2937          * initialization. This has to be done regardless of
2938          * DCB as by default everything is mapped to TC0.
2939          */
2940         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2941         tx_ctx.rdylist_act = 0;
2942
2943         /* clear the context in the HMC */
2944         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2945         if (err) {
2946                 dev_info(&vsi->back->pdev->dev,
2947                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2948                          ring->queue_index, pf_q, err);
2949                 return -ENOMEM;
2950         }
2951
2952         /* set the context in the HMC */
2953         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2954         if (err) {
2955                 dev_info(&vsi->back->pdev->dev,
2956                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2957                          ring->queue_index, pf_q, err);
2958                 return -ENOMEM;
2959         }
2960
2961         /* Now associate this queue with this PCI function */
2962         if (vsi->type == I40E_VSI_VMDQ2) {
2963                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2964                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2965                            I40E_QTX_CTL_VFVM_INDX_MASK;
2966         } else {
2967                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2968         }
2969
2970         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2971                     I40E_QTX_CTL_PF_INDX_MASK);
2972         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2973         i40e_flush(hw);
2974
2975         /* cache tail off for easier writes later */
2976         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2977
2978         return 0;
2979 }
2980
2981 /**
2982  * i40e_configure_rx_ring - Configure a receive ring context
2983  * @ring: The Rx ring to configure
2984  *
2985  * Configure the Rx descriptor ring in the HMC context.
2986  **/
2987 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2988 {
2989         struct i40e_vsi *vsi = ring->vsi;
2990         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2991         u16 pf_q = vsi->base_queue + ring->queue_index;
2992         struct i40e_hw *hw = &vsi->back->hw;
2993         struct i40e_hmc_obj_rxq rx_ctx;
2994         i40e_status err = 0;
2995
2996         ring->state = 0;
2997
2998         /* clear the context structure first */
2999         memset(&rx_ctx, 0, sizeof(rx_ctx));
3000
3001         ring->rx_buf_len = vsi->rx_buf_len;
3002
3003         rx_ctx.dbuff = DIV_ROUND_UP(ring->rx_buf_len,
3004                                     BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3005
3006         rx_ctx.base = (ring->dma / 128);
3007         rx_ctx.qlen = ring->count;
3008
3009         /* use 32 byte descriptors */
3010         rx_ctx.dsize = 1;
3011
3012         /* descriptor type is always zero
3013          * rx_ctx.dtype = 0;
3014          */
3015         rx_ctx.hsplit_0 = 0;
3016
3017         rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
3018         if (hw->revision_id == 0)
3019                 rx_ctx.lrxqthresh = 0;
3020         else
3021                 rx_ctx.lrxqthresh = 2;
3022         rx_ctx.crcstrip = 1;
3023         rx_ctx.l2tsel = 1;
3024         /* this controls whether VLAN is stripped from inner headers */
3025         rx_ctx.showiv = 0;
3026         /* set the prefena field to 1 because the manual says to */
3027         rx_ctx.prefena = 1;
3028
3029         /* clear the context in the HMC */
3030         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
3031         if (err) {
3032                 dev_info(&vsi->back->pdev->dev,
3033                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3034                          ring->queue_index, pf_q, err);
3035                 return -ENOMEM;
3036         }
3037
3038         /* set the context in the HMC */
3039         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
3040         if (err) {
3041                 dev_info(&vsi->back->pdev->dev,
3042                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
3043                          ring->queue_index, pf_q, err);
3044                 return -ENOMEM;
3045         }
3046
3047         /* configure Rx buffer alignment */
3048         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX))
3049                 clear_ring_build_skb_enabled(ring);
3050         else
3051                 set_ring_build_skb_enabled(ring);
3052
3053         /* cache tail for quicker writes, and clear the reg before use */
3054         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
3055         writel(0, ring->tail);
3056
3057         i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
3058
3059         return 0;
3060 }
3061
3062 /**
3063  * i40e_vsi_configure_tx - Configure the VSI for Tx
3064  * @vsi: VSI structure describing this set of rings and resources
3065  *
3066  * Configure the Tx VSI for operation.
3067  **/
3068 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
3069 {
3070         int err = 0;
3071         u16 i;
3072
3073         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
3074                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
3075
3076         return err;
3077 }
3078
3079 /**
3080  * i40e_vsi_configure_rx - Configure the VSI for Rx
3081  * @vsi: the VSI being configured
3082  *
3083  * Configure the Rx VSI for operation.
3084  **/
3085 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3086 {
3087         int err = 0;
3088         u16 i;
3089
3090         if (!vsi->netdev || (vsi->back->flags & I40E_FLAG_LEGACY_RX)) {
3091                 vsi->max_frame = I40E_MAX_RXBUFFER;
3092                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
3093 #if (PAGE_SIZE < 8192)
3094         } else if (!I40E_2K_TOO_SMALL_WITH_PADDING &&
3095                    (vsi->netdev->mtu <= ETH_DATA_LEN)) {
3096                 vsi->max_frame = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3097                 vsi->rx_buf_len = I40E_RXBUFFER_1536 - NET_IP_ALIGN;
3098 #endif
3099         } else {
3100                 vsi->max_frame = I40E_MAX_RXBUFFER;
3101                 vsi->rx_buf_len = (PAGE_SIZE < 8192) ? I40E_RXBUFFER_3072 :
3102                                                        I40E_RXBUFFER_2048;
3103         }
3104
3105         /* set up individual rings */
3106         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3107                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3108
3109         return err;
3110 }
3111
3112 /**
3113  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3114  * @vsi: ptr to the VSI
3115  **/
3116 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3117 {
3118         struct i40e_ring *tx_ring, *rx_ring;
3119         u16 qoffset, qcount;
3120         int i, n;
3121
3122         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3123                 /* Reset the TC information */
3124                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3125                         rx_ring = vsi->rx_rings[i];
3126                         tx_ring = vsi->tx_rings[i];
3127                         rx_ring->dcb_tc = 0;
3128                         tx_ring->dcb_tc = 0;
3129                 }
3130         }
3131
3132         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3133                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3134                         continue;
3135
3136                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3137                 qcount = vsi->tc_config.tc_info[n].qcount;
3138                 for (i = qoffset; i < (qoffset + qcount); i++) {
3139                         rx_ring = vsi->rx_rings[i];
3140                         tx_ring = vsi->tx_rings[i];
3141                         rx_ring->dcb_tc = n;
3142                         tx_ring->dcb_tc = n;
3143                 }
3144         }
3145 }
3146
3147 /**
3148  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3149  * @vsi: ptr to the VSI
3150  **/
3151 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3152 {
3153         struct i40e_pf *pf = vsi->back;
3154         int err;
3155
3156         if (vsi->netdev)
3157                 i40e_set_rx_mode(vsi->netdev);
3158
3159         if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3160                 err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3161                 if (err) {
3162                         dev_warn(&pf->pdev->dev,
3163                                  "could not set up macaddr; err %d\n", err);
3164                 }
3165         }
3166 }
3167
3168 /**
3169  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3170  * @vsi: Pointer to the targeted VSI
3171  *
3172  * This function replays the hlist on the hw where all the SB Flow Director
3173  * filters were saved.
3174  **/
3175 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3176 {
3177         struct i40e_fdir_filter *filter;
3178         struct i40e_pf *pf = vsi->back;
3179         struct hlist_node *node;
3180
3181         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3182                 return;
3183
3184         /* Reset FDir counters as we're replaying all existing filters */
3185         pf->fd_tcp4_filter_cnt = 0;
3186         pf->fd_udp4_filter_cnt = 0;
3187         pf->fd_sctp4_filter_cnt = 0;
3188         pf->fd_ip4_filter_cnt = 0;
3189
3190         hlist_for_each_entry_safe(filter, node,
3191                                   &pf->fdir_filter_list, fdir_node) {
3192                 i40e_add_del_fdir(vsi, filter, true);
3193         }
3194 }
3195
3196 /**
3197  * i40e_vsi_configure - Set up the VSI for action
3198  * @vsi: the VSI being configured
3199  **/
3200 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3201 {
3202         int err;
3203
3204         i40e_set_vsi_rx_mode(vsi);
3205         i40e_restore_vlan(vsi);
3206         i40e_vsi_config_dcb_rings(vsi);
3207         err = i40e_vsi_configure_tx(vsi);
3208         if (!err)
3209                 err = i40e_vsi_configure_rx(vsi);
3210
3211         return err;
3212 }
3213
3214 /**
3215  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3216  * @vsi: the VSI being configured
3217  **/
3218 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3219 {
3220         struct i40e_pf *pf = vsi->back;
3221         struct i40e_hw *hw = &pf->hw;
3222         u16 vector;
3223         int i, q;
3224         u32 qp;
3225
3226         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3227          * and PFINT_LNKLSTn registers, e.g.:
3228          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3229          */
3230         qp = vsi->base_queue;
3231         vector = vsi->base_vector;
3232         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3233                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3234
3235                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3236                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3237                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3238                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3239                      q_vector->rx.itr);
3240                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3241                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3242                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3243                      q_vector->tx.itr);
3244                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3245                      i40e_intrl_usec_to_reg(vsi->int_rate_limit));
3246
3247                 /* Linked list for the queuepairs assigned to this vector */
3248                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3249                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3250                         u32 val;
3251
3252                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3253                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3254                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3255                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3256                               (I40E_QUEUE_TYPE_TX
3257                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3258
3259                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3260
3261                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3262                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3263                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3264                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3265                               (I40E_QUEUE_TYPE_RX
3266                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3267
3268                         /* Terminate the linked list */
3269                         if (q == (q_vector->num_ringpairs - 1))
3270                                 val |= (I40E_QUEUE_END_OF_LIST
3271                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3272
3273                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3274                         qp++;
3275                 }
3276         }
3277
3278         i40e_flush(hw);
3279 }
3280
3281 /**
3282  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3283  * @hw: ptr to the hardware info
3284  **/
3285 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3286 {
3287         struct i40e_hw *hw = &pf->hw;
3288         u32 val;
3289
3290         /* clear things first */
3291         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3292         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3293
3294         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3295               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3296               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3297               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3298               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3299               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3300               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3301               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3302
3303         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3304                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3305
3306         if (pf->flags & I40E_FLAG_PTP)
3307                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3308
3309         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3310
3311         /* SW_ITR_IDX = 0, but don't change INTENA */
3312         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3313                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3314
3315         /* OTHER_ITR_IDX = 0 */
3316         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3317 }
3318
3319 /**
3320  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3321  * @vsi: the VSI being configured
3322  **/
3323 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3324 {
3325         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3326         struct i40e_pf *pf = vsi->back;
3327         struct i40e_hw *hw = &pf->hw;
3328         u32 val;
3329
3330         /* set the ITR configuration */
3331         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3332         q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3333         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3334         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3335         q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3336         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3337         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3338
3339         i40e_enable_misc_int_causes(pf);
3340
3341         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3342         wr32(hw, I40E_PFINT_LNKLST0, 0);
3343
3344         /* Associate the queue pair to the vector and enable the queue int */
3345         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3346               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3347               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3348
3349         wr32(hw, I40E_QINT_RQCTL(0), val);
3350
3351         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3352               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3353               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3354
3355         wr32(hw, I40E_QINT_TQCTL(0), val);
3356         i40e_flush(hw);
3357 }
3358
3359 /**
3360  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3361  * @pf: board private structure
3362  **/
3363 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3364 {
3365         struct i40e_hw *hw = &pf->hw;
3366
3367         wr32(hw, I40E_PFINT_DYN_CTL0,
3368              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3369         i40e_flush(hw);
3370 }
3371
3372 /**
3373  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3374  * @pf: board private structure
3375  * @clearpba: true when all pending interrupt events should be cleared
3376  **/
3377 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3378 {
3379         struct i40e_hw *hw = &pf->hw;
3380         u32 val;
3381
3382         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3383               (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3384               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3385
3386         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3387         i40e_flush(hw);
3388 }
3389
3390 /**
3391  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3392  * @irq: interrupt number
3393  * @data: pointer to a q_vector
3394  **/
3395 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3396 {
3397         struct i40e_q_vector *q_vector = data;
3398
3399         if (!q_vector->tx.ring && !q_vector->rx.ring)
3400                 return IRQ_HANDLED;
3401
3402         napi_schedule_irqoff(&q_vector->napi);
3403
3404         return IRQ_HANDLED;
3405 }
3406
3407 /**
3408  * i40e_irq_affinity_notify - Callback for affinity changes
3409  * @notify: context as to what irq was changed
3410  * @mask: the new affinity mask
3411  *
3412  * This is a callback function used by the irq_set_affinity_notifier function
3413  * so that we may register to receive changes to the irq affinity masks.
3414  **/
3415 static void i40e_irq_affinity_notify(struct irq_affinity_notify *notify,
3416                                      const cpumask_t *mask)
3417 {
3418         struct i40e_q_vector *q_vector =
3419                 container_of(notify, struct i40e_q_vector, affinity_notify);
3420
3421         q_vector->affinity_mask = *mask;
3422 }
3423
3424 /**
3425  * i40e_irq_affinity_release - Callback for affinity notifier release
3426  * @ref: internal core kernel usage
3427  *
3428  * This is a callback function used by the irq_set_affinity_notifier function
3429  * to inform the current notification subscriber that they will no longer
3430  * receive notifications.
3431  **/
3432 static void i40e_irq_affinity_release(struct kref *ref) {}
3433
3434 /**
3435  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3436  * @vsi: the VSI being configured
3437  * @basename: name for the vector
3438  *
3439  * Allocates MSI-X vectors and requests interrupts from the kernel.
3440  **/
3441 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3442 {
3443         int q_vectors = vsi->num_q_vectors;
3444         struct i40e_pf *pf = vsi->back;
3445         int base = vsi->base_vector;
3446         int rx_int_idx = 0;
3447         int tx_int_idx = 0;
3448         int vector, err;
3449         int irq_num;
3450
3451         for (vector = 0; vector < q_vectors; vector++) {
3452                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3453
3454                 irq_num = pf->msix_entries[base + vector].vector;
3455
3456                 if (q_vector->tx.ring && q_vector->rx.ring) {
3457                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3458                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3459                         tx_int_idx++;
3460                 } else if (q_vector->rx.ring) {
3461                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3462                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3463                 } else if (q_vector->tx.ring) {
3464                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3465                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3466                 } else {
3467                         /* skip this unused q_vector */
3468                         continue;
3469                 }
3470                 err = request_irq(irq_num,
3471                                   vsi->irq_handler,
3472                                   0,
3473                                   q_vector->name,
3474                                   q_vector);
3475                 if (err) {
3476                         dev_info(&pf->pdev->dev,
3477                                  "MSIX request_irq failed, error: %d\n", err);
3478                         goto free_queue_irqs;
3479                 }
3480
3481                 /* register for affinity change notifications */
3482                 q_vector->affinity_notify.notify = i40e_irq_affinity_notify;
3483                 q_vector->affinity_notify.release = i40e_irq_affinity_release;
3484                 irq_set_affinity_notifier(irq_num, &q_vector->affinity_notify);
3485                 /* assign the mask for this irq */
3486                 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
3487         }
3488
3489         vsi->irqs_ready = true;
3490         return 0;
3491
3492 free_queue_irqs:
3493         while (vector) {
3494                 vector--;
3495                 irq_num = pf->msix_entries[base + vector].vector;
3496                 irq_set_affinity_notifier(irq_num, NULL);
3497                 irq_set_affinity_hint(irq_num, NULL);
3498                 free_irq(irq_num, &vsi->q_vectors[vector]);
3499         }
3500         return err;
3501 }
3502
3503 /**
3504  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3505  * @vsi: the VSI being un-configured
3506  **/
3507 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3508 {
3509         struct i40e_pf *pf = vsi->back;
3510         struct i40e_hw *hw = &pf->hw;
3511         int base = vsi->base_vector;
3512         int i;
3513
3514         for (i = 0; i < vsi->num_queue_pairs; i++) {
3515                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3516                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3517         }
3518
3519         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3520                 for (i = vsi->base_vector;
3521                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3522                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3523
3524                 i40e_flush(hw);
3525                 for (i = 0; i < vsi->num_q_vectors; i++)
3526                         synchronize_irq(pf->msix_entries[i + base].vector);
3527         } else {
3528                 /* Legacy and MSI mode - this stops all interrupt handling */
3529                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3530                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3531                 i40e_flush(hw);
3532                 synchronize_irq(pf->pdev->irq);
3533         }
3534 }
3535
3536 /**
3537  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3538  * @vsi: the VSI being configured
3539  **/
3540 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3541 {
3542         struct i40e_pf *pf = vsi->back;
3543         int i;
3544
3545         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3546                 for (i = 0; i < vsi->num_q_vectors; i++)
3547                         i40e_irq_dynamic_enable(vsi, i);
3548         } else {
3549                 i40e_irq_dynamic_enable_icr0(pf, true);
3550         }
3551
3552         i40e_flush(&pf->hw);
3553         return 0;
3554 }
3555
3556 /**
3557  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3558  * @pf: board private structure
3559  **/
3560 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3561 {
3562         /* Disable ICR 0 */
3563         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3564         i40e_flush(&pf->hw);
3565 }
3566
3567 /**
3568  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3569  * @irq: interrupt number
3570  * @data: pointer to a q_vector
3571  *
3572  * This is the handler used for all MSI/Legacy interrupts, and deals
3573  * with both queue and non-queue interrupts.  This is also used in
3574  * MSIX mode to handle the non-queue interrupts.
3575  **/
3576 static irqreturn_t i40e_intr(int irq, void *data)
3577 {
3578         struct i40e_pf *pf = (struct i40e_pf *)data;
3579         struct i40e_hw *hw = &pf->hw;
3580         irqreturn_t ret = IRQ_NONE;
3581         u32 icr0, icr0_remaining;
3582         u32 val, ena_mask;
3583
3584         icr0 = rd32(hw, I40E_PFINT_ICR0);
3585         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3586
3587         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3588         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3589                 goto enable_intr;
3590
3591         /* if interrupt but no bits showing, must be SWINT */
3592         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3593             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3594                 pf->sw_int_count++;
3595
3596         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3597             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3598                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3599                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3600                 dev_dbg(&pf->pdev->dev, "cleared PE_CRITERR\n");
3601         }
3602
3603         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3604         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3605                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3606                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3607
3608                 /* We do not have a way to disarm Queue causes while leaving
3609                  * interrupt enabled for all other causes, ideally
3610                  * interrupt should be disabled while we are in NAPI but
3611                  * this is not a performance path and napi_schedule()
3612                  * can deal with rescheduling.
3613                  */
3614                 if (!test_bit(__I40E_DOWN, &pf->state))
3615                         napi_schedule_irqoff(&q_vector->napi);
3616         }
3617
3618         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3619                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3620                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3621                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3622         }
3623
3624         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3625                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3626                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3627         }
3628
3629         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3630                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3631                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3632         }
3633
3634         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3635                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3636                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3637                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3638                 val = rd32(hw, I40E_GLGEN_RSTAT);
3639                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3640                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3641                 if (val == I40E_RESET_CORER) {
3642                         pf->corer_count++;
3643                 } else if (val == I40E_RESET_GLOBR) {
3644                         pf->globr_count++;
3645                 } else if (val == I40E_RESET_EMPR) {
3646                         pf->empr_count++;
3647                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3648                 }
3649         }
3650
3651         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3652                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3653                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3654                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3655                          rd32(hw, I40E_PFHMC_ERRORINFO),
3656                          rd32(hw, I40E_PFHMC_ERRORDATA));
3657         }
3658
3659         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3660                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3661
3662                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3663                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3664                         i40e_ptp_tx_hwtstamp(pf);
3665                 }
3666         }
3667
3668         /* If a critical error is pending we have no choice but to reset the
3669          * device.
3670          * Report and mask out any remaining unexpected interrupts.
3671          */
3672         icr0_remaining = icr0 & ena_mask;
3673         if (icr0_remaining) {
3674                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3675                          icr0_remaining);
3676                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3677                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3678                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3679                         dev_info(&pf->pdev->dev, "device will be reset\n");
3680                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3681                         i40e_service_event_schedule(pf);
3682                 }
3683                 ena_mask &= ~icr0_remaining;
3684         }
3685         ret = IRQ_HANDLED;
3686
3687 enable_intr:
3688         /* re-enable interrupt causes */
3689         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3690         if (!test_bit(__I40E_DOWN, &pf->state)) {
3691                 i40e_service_event_schedule(pf);
3692                 i40e_irq_dynamic_enable_icr0(pf, false);
3693         }
3694
3695         return ret;
3696 }
3697
3698 /**
3699  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3700  * @tx_ring:  tx ring to clean
3701  * @budget:   how many cleans we're allowed
3702  *
3703  * Returns true if there's any budget left (e.g. the clean is finished)
3704  **/
3705 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3706 {
3707         struct i40e_vsi *vsi = tx_ring->vsi;
3708         u16 i = tx_ring->next_to_clean;
3709         struct i40e_tx_buffer *tx_buf;
3710         struct i40e_tx_desc *tx_desc;
3711
3712         tx_buf = &tx_ring->tx_bi[i];
3713         tx_desc = I40E_TX_DESC(tx_ring, i);
3714         i -= tx_ring->count;
3715
3716         do {
3717                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3718
3719                 /* if next_to_watch is not set then there is no work pending */
3720                 if (!eop_desc)
3721                         break;
3722
3723                 /* prevent any other reads prior to eop_desc */
3724                 read_barrier_depends();
3725
3726                 /* if the descriptor isn't done, no work yet to do */
3727                 if (!(eop_desc->cmd_type_offset_bsz &
3728                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3729                         break;
3730
3731                 /* clear next_to_watch to prevent false hangs */
3732                 tx_buf->next_to_watch = NULL;
3733
3734                 tx_desc->buffer_addr = 0;
3735                 tx_desc->cmd_type_offset_bsz = 0;
3736                 /* move past filter desc */
3737                 tx_buf++;
3738                 tx_desc++;
3739                 i++;
3740                 if (unlikely(!i)) {
3741                         i -= tx_ring->count;
3742                         tx_buf = tx_ring->tx_bi;
3743                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3744                 }
3745                 /* unmap skb header data */
3746                 dma_unmap_single(tx_ring->dev,
3747                                  dma_unmap_addr(tx_buf, dma),
3748                                  dma_unmap_len(tx_buf, len),
3749                                  DMA_TO_DEVICE);
3750                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3751                         kfree(tx_buf->raw_buf);
3752
3753                 tx_buf->raw_buf = NULL;
3754                 tx_buf->tx_flags = 0;
3755                 tx_buf->next_to_watch = NULL;
3756                 dma_unmap_len_set(tx_buf, len, 0);
3757                 tx_desc->buffer_addr = 0;
3758                 tx_desc->cmd_type_offset_bsz = 0;
3759
3760                 /* move us past the eop_desc for start of next FD desc */
3761                 tx_buf++;
3762                 tx_desc++;
3763                 i++;
3764                 if (unlikely(!i)) {
3765                         i -= tx_ring->count;
3766                         tx_buf = tx_ring->tx_bi;
3767                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3768                 }
3769
3770                 /* update budget accounting */
3771                 budget--;
3772         } while (likely(budget));
3773
3774         i += tx_ring->count;
3775         tx_ring->next_to_clean = i;
3776
3777         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3778                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3779
3780         return budget > 0;
3781 }
3782
3783 /**
3784  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3785  * @irq: interrupt number
3786  * @data: pointer to a q_vector
3787  **/
3788 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3789 {
3790         struct i40e_q_vector *q_vector = data;
3791         struct i40e_vsi *vsi;
3792
3793         if (!q_vector->tx.ring)
3794                 return IRQ_HANDLED;
3795
3796         vsi = q_vector->tx.ring->vsi;
3797         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3798
3799         return IRQ_HANDLED;
3800 }
3801
3802 /**
3803  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3804  * @vsi: the VSI being configured
3805  * @v_idx: vector index
3806  * @qp_idx: queue pair index
3807  **/
3808 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3809 {
3810         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3811         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3812         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3813
3814         tx_ring->q_vector = q_vector;
3815         tx_ring->next = q_vector->tx.ring;
3816         q_vector->tx.ring = tx_ring;
3817         q_vector->tx.count++;
3818
3819         rx_ring->q_vector = q_vector;
3820         rx_ring->next = q_vector->rx.ring;
3821         q_vector->rx.ring = rx_ring;
3822         q_vector->rx.count++;
3823 }
3824
3825 /**
3826  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3827  * @vsi: the VSI being configured
3828  *
3829  * This function maps descriptor rings to the queue-specific vectors
3830  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3831  * one vector per queue pair, but on a constrained vector budget, we
3832  * group the queue pairs as "efficiently" as possible.
3833  **/
3834 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3835 {
3836         int qp_remaining = vsi->num_queue_pairs;
3837         int q_vectors = vsi->num_q_vectors;
3838         int num_ringpairs;
3839         int v_start = 0;
3840         int qp_idx = 0;
3841
3842         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3843          * group them so there are multiple queues per vector.
3844          * It is also important to go through all the vectors available to be
3845          * sure that if we don't use all the vectors, that the remaining vectors
3846          * are cleared. This is especially important when decreasing the
3847          * number of queues in use.
3848          */
3849         for (; v_start < q_vectors; v_start++) {
3850                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3851
3852                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3853
3854                 q_vector->num_ringpairs = num_ringpairs;
3855
3856                 q_vector->rx.count = 0;
3857                 q_vector->tx.count = 0;
3858                 q_vector->rx.ring = NULL;
3859                 q_vector->tx.ring = NULL;
3860
3861                 while (num_ringpairs--) {
3862                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3863                         qp_idx++;
3864                         qp_remaining--;
3865                 }
3866         }
3867 }
3868
3869 /**
3870  * i40e_vsi_request_irq - Request IRQ from the OS
3871  * @vsi: the VSI being configured
3872  * @basename: name for the vector
3873  **/
3874 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3875 {
3876         struct i40e_pf *pf = vsi->back;
3877         int err;
3878
3879         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3880                 err = i40e_vsi_request_irq_msix(vsi, basename);
3881         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3882                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3883                                   pf->int_name, pf);
3884         else
3885                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3886                                   pf->int_name, pf);
3887
3888         if (err)
3889                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3890
3891         return err;
3892 }
3893
3894 #ifdef CONFIG_NET_POLL_CONTROLLER
3895 /**
3896  * i40e_netpoll - A Polling 'interrupt' handler
3897  * @netdev: network interface device structure
3898  *
3899  * This is used by netconsole to send skbs without having to re-enable
3900  * interrupts.  It's not called while the normal interrupt routine is executing.
3901  **/
3902 static void i40e_netpoll(struct net_device *netdev)
3903 {
3904         struct i40e_netdev_priv *np = netdev_priv(netdev);
3905         struct i40e_vsi *vsi = np->vsi;
3906         struct i40e_pf *pf = vsi->back;
3907         int i;
3908
3909         /* if interface is down do nothing */
3910         if (test_bit(__I40E_DOWN, &vsi->state))
3911                 return;
3912
3913         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3914                 for (i = 0; i < vsi->num_q_vectors; i++)
3915                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3916         } else {
3917                 i40e_intr(pf->pdev->irq, netdev);
3918         }
3919 }
3920 #endif
3921
3922 #define I40E_QTX_ENA_WAIT_COUNT 50
3923
3924 /**
3925  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3926  * @pf: the PF being configured
3927  * @pf_q: the PF queue
3928  * @enable: enable or disable state of the queue
3929  *
3930  * This routine will wait for the given Tx queue of the PF to reach the
3931  * enabled or disabled state.
3932  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3933  * multiple retries; else will return 0 in case of success.
3934  **/
3935 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3936 {
3937         int i;
3938         u32 tx_reg;
3939
3940         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3941                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3942                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3943                         break;
3944
3945                 usleep_range(10, 20);
3946         }
3947         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3948                 return -ETIMEDOUT;
3949
3950         return 0;
3951 }
3952
3953 /**
3954  * i40e_control_tx_q - Start or stop a particular Tx queue
3955  * @pf: the PF structure
3956  * @pf_q: the PF queue to configure
3957  * @enable: start or stop the queue
3958  *
3959  * This function enables or disables a single queue. Note that any delay
3960  * required after the operation is expected to be handled by the caller of
3961  * this function.
3962  **/
3963 static void i40e_control_tx_q(struct i40e_pf *pf, int pf_q, bool enable)
3964 {
3965         struct i40e_hw *hw = &pf->hw;
3966         u32 tx_reg;
3967         int i;
3968
3969         /* warn the TX unit of coming changes */
3970         i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3971         if (!enable)
3972                 usleep_range(10, 20);
3973
3974         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
3975                 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3976                 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3977                     ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3978                         break;
3979                 usleep_range(1000, 2000);
3980         }
3981
3982         /* Skip if the queue is already in the requested state */
3983         if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3984                 return;
3985
3986         /* turn on/off the queue */
3987         if (enable) {
3988                 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3989                 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3990         } else {
3991                 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3992         }
3993
3994         wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3995 }
3996
3997 /**
3998  * i40e_vsi_control_tx - Start or stop a VSI's rings
3999  * @vsi: the VSI being configured
4000  * @enable: start or stop the rings
4001  **/
4002 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
4003 {
4004         struct i40e_pf *pf = vsi->back;
4005         int i, pf_q, ret = 0;
4006
4007         pf_q = vsi->base_queue;
4008         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4009                 i40e_control_tx_q(pf, pf_q, enable);
4010
4011                 /* wait for the change to finish */
4012                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
4013                 if (ret) {
4014                         dev_info(&pf->pdev->dev,
4015                                  "VSI seid %d Tx ring %d %sable timeout\n",
4016                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
4017                         break;
4018                 }
4019         }
4020
4021         return ret;
4022 }
4023
4024 /**
4025  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
4026  * @pf: the PF being configured
4027  * @pf_q: the PF queue
4028  * @enable: enable or disable state of the queue
4029  *
4030  * This routine will wait for the given Rx queue of the PF to reach the
4031  * enabled or disabled state.
4032  * Returns -ETIMEDOUT in case of failing to reach the requested state after
4033  * multiple retries; else will return 0 in case of success.
4034  **/
4035 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
4036 {
4037         int i;
4038         u32 rx_reg;
4039
4040         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
4041                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
4042                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4043                         break;
4044
4045                 usleep_range(10, 20);
4046         }
4047         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
4048                 return -ETIMEDOUT;
4049
4050         return 0;
4051 }
4052
4053 /**
4054  * i40e_control_rx_q - Start or stop a particular Rx queue
4055  * @pf: the PF structure
4056  * @pf_q: the PF queue to configure
4057  * @enable: start or stop the queue
4058  *
4059  * This function enables or disables a single queue. Note that any delay
4060  * required after the operation is expected to be handled by the caller of
4061  * this function.
4062  **/
4063 static void i40e_control_rx_q(struct i40e_pf *pf, int pf_q, bool enable)
4064 {
4065         struct i40e_hw *hw = &pf->hw;
4066         u32 rx_reg;
4067         int i;
4068
4069         for (i = 0; i < I40E_QTX_ENA_WAIT_COUNT; i++) {
4070                 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
4071                 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
4072                     ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
4073                         break;
4074                 usleep_range(1000, 2000);
4075         }
4076
4077         /* Skip if the queue is already in the requested state */
4078         if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
4079                 return;
4080
4081         /* turn on/off the queue */
4082         if (enable)
4083                 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
4084         else
4085                 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
4086
4087         wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
4088 }
4089
4090 /**
4091  * i40e_vsi_control_rx - Start or stop a VSI's rings
4092  * @vsi: the VSI being configured
4093  * @enable: start or stop the rings
4094  **/
4095 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
4096 {
4097         struct i40e_pf *pf = vsi->back;
4098         int i, pf_q, ret = 0;
4099
4100         pf_q = vsi->base_queue;
4101         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4102                 i40e_control_rx_q(pf, pf_q, enable);
4103
4104                 /* wait for the change to finish */
4105                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
4106                 if (ret) {
4107                         dev_info(&pf->pdev->dev,
4108                                  "VSI seid %d Rx ring %d %sable timeout\n",
4109                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
4110                         break;
4111                 }
4112         }
4113
4114         /* Due to HW errata, on Rx disable only, the register can indicate done
4115          * before it really is. Needs 50ms to be sure
4116          */
4117         if (!enable)
4118                 mdelay(50);
4119
4120         return ret;
4121 }
4122
4123 /**
4124  * i40e_vsi_start_rings - Start a VSI's rings
4125  * @vsi: the VSI being configured
4126  **/
4127 int i40e_vsi_start_rings(struct i40e_vsi *vsi)
4128 {
4129         int ret = 0;
4130
4131         /* do rx first for enable and last for disable */
4132         ret = i40e_vsi_control_rx(vsi, true);
4133         if (ret)
4134                 return ret;
4135         ret = i40e_vsi_control_tx(vsi, true);
4136
4137         return ret;
4138 }
4139
4140 /**
4141  * i40e_vsi_stop_rings - Stop a VSI's rings
4142  * @vsi: the VSI being configured
4143  **/
4144 void i40e_vsi_stop_rings(struct i40e_vsi *vsi)
4145 {
4146         /* When port TX is suspended, don't wait */
4147         if (test_bit(__I40E_PORT_SUSPENDED, &vsi->back->state))
4148                 return i40e_vsi_stop_rings_no_wait(vsi);
4149
4150         /* do rx first for enable and last for disable
4151          * Ignore return value, we need to shutdown whatever we can
4152          */
4153         i40e_vsi_control_tx(vsi, false);
4154         i40e_vsi_control_rx(vsi, false);
4155 }
4156
4157 /**
4158  * i40e_vsi_stop_rings_no_wait - Stop a VSI's rings and do not delay
4159  * @vsi: the VSI being shutdown
4160  *
4161  * This function stops all the rings for a VSI but does not delay to verify
4162  * that rings have been disabled. It is expected that the caller is shutting
4163  * down multiple VSIs at once and will delay together for all the VSIs after
4164  * initiating the shutdown. This is particularly useful for shutting down lots
4165  * of VFs together. Otherwise, a large delay can be incurred while configuring
4166  * each VSI in serial.
4167  **/
4168 void i40e_vsi_stop_rings_no_wait(struct i40e_vsi *vsi)
4169 {
4170         struct i40e_pf *pf = vsi->back;
4171         int i, pf_q;
4172
4173         pf_q = vsi->base_queue;
4174         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4175                 i40e_control_tx_q(pf, pf_q, false);
4176                 i40e_control_rx_q(pf, pf_q, false);
4177         }
4178 }
4179
4180 /**
4181  * i40e_vsi_free_irq - Free the irq association with the OS
4182  * @vsi: the VSI being configured
4183  **/
4184 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4185 {
4186         struct i40e_pf *pf = vsi->back;
4187         struct i40e_hw *hw = &pf->hw;
4188         int base = vsi->base_vector;
4189         u32 val, qp;
4190         int i;
4191
4192         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4193                 if (!vsi->q_vectors)
4194                         return;
4195
4196                 if (!vsi->irqs_ready)
4197                         return;
4198
4199                 vsi->irqs_ready = false;
4200                 for (i = 0; i < vsi->num_q_vectors; i++) {
4201                         int irq_num;
4202                         u16 vector;
4203
4204                         vector = i + base;
4205                         irq_num = pf->msix_entries[vector].vector;
4206
4207                         /* free only the irqs that were actually requested */
4208                         if (!vsi->q_vectors[i] ||
4209                             !vsi->q_vectors[i]->num_ringpairs)
4210                                 continue;
4211
4212                         /* clear the affinity notifier in the IRQ descriptor */
4213                         irq_set_affinity_notifier(irq_num, NULL);
4214                         /* clear the affinity_mask in the IRQ descriptor */
4215                         irq_set_affinity_hint(irq_num, NULL);
4216                         synchronize_irq(irq_num);
4217                         free_irq(irq_num, vsi->q_vectors[i]);
4218
4219                         /* Tear down the interrupt queue link list
4220                          *
4221                          * We know that they come in pairs and always
4222                          * the Rx first, then the Tx.  To clear the
4223                          * link list, stick the EOL value into the
4224                          * next_q field of the registers.
4225                          */
4226                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4227                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4228                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4229                         val |= I40E_QUEUE_END_OF_LIST
4230                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4231                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4232
4233                         while (qp != I40E_QUEUE_END_OF_LIST) {
4234                                 u32 next;
4235
4236                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4237
4238                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4239                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4240                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4241                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4242
4243                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4244                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4245
4246                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4247
4248                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4249
4250                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4251                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4252
4253                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4254                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4255                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4256                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4257
4258                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4259                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4260
4261                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4262                                 qp = next;
4263                         }
4264                 }
4265         } else {
4266                 free_irq(pf->pdev->irq, pf);
4267
4268                 val = rd32(hw, I40E_PFINT_LNKLST0);
4269                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4270                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4271                 val |= I40E_QUEUE_END_OF_LIST
4272                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4273                 wr32(hw, I40E_PFINT_LNKLST0, val);
4274
4275                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4276                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4277                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4278                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4279                          I40E_QINT_RQCTL_INTEVENT_MASK);
4280
4281                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4282                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4283
4284                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4285
4286                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4287
4288                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4289                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4290                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4291                          I40E_QINT_TQCTL_INTEVENT_MASK);
4292
4293                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4294                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4295
4296                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4297         }
4298 }
4299
4300 /**
4301  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4302  * @vsi: the VSI being configured
4303  * @v_idx: Index of vector to be freed
4304  *
4305  * This function frees the memory allocated to the q_vector.  In addition if
4306  * NAPI is enabled it will delete any references to the NAPI struct prior
4307  * to freeing the q_vector.
4308  **/
4309 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4310 {
4311         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4312         struct i40e_ring *ring;
4313
4314         if (!q_vector)
4315                 return;
4316
4317         /* disassociate q_vector from rings */
4318         i40e_for_each_ring(ring, q_vector->tx)
4319                 ring->q_vector = NULL;
4320
4321         i40e_for_each_ring(ring, q_vector->rx)
4322                 ring->q_vector = NULL;
4323
4324         /* only VSI w/ an associated netdev is set up w/ NAPI */
4325         if (vsi->netdev)
4326                 netif_napi_del(&q_vector->napi);
4327
4328         vsi->q_vectors[v_idx] = NULL;
4329
4330         kfree_rcu(q_vector, rcu);
4331 }
4332
4333 /**
4334  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4335  * @vsi: the VSI being un-configured
4336  *
4337  * This frees the memory allocated to the q_vectors and
4338  * deletes references to the NAPI struct.
4339  **/
4340 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4341 {
4342         int v_idx;
4343
4344         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4345                 i40e_free_q_vector(vsi, v_idx);
4346 }
4347
4348 /**
4349  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4350  * @pf: board private structure
4351  **/
4352 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4353 {
4354         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4355         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4356                 pci_disable_msix(pf->pdev);
4357                 kfree(pf->msix_entries);
4358                 pf->msix_entries = NULL;
4359                 kfree(pf->irq_pile);
4360                 pf->irq_pile = NULL;
4361         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4362                 pci_disable_msi(pf->pdev);
4363         }
4364         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4365 }
4366
4367 /**
4368  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4369  * @pf: board private structure
4370  *
4371  * We go through and clear interrupt specific resources and reset the structure
4372  * to pre-load conditions
4373  **/
4374 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4375 {
4376         int i;
4377
4378         i40e_stop_misc_vector(pf);
4379         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4380                 synchronize_irq(pf->msix_entries[0].vector);
4381                 free_irq(pf->msix_entries[0].vector, pf);
4382         }
4383
4384         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4385                       I40E_IWARP_IRQ_PILE_ID);
4386
4387         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4388         for (i = 0; i < pf->num_alloc_vsi; i++)
4389                 if (pf->vsi[i])
4390                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4391         i40e_reset_interrupt_capability(pf);
4392 }
4393
4394 /**
4395  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4396  * @vsi: the VSI being configured
4397  **/
4398 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4399 {
4400         int q_idx;
4401
4402         if (!vsi->netdev)
4403                 return;
4404
4405         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4406                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4407
4408                 if (q_vector->rx.ring || q_vector->tx.ring)
4409                         napi_enable(&q_vector->napi);
4410         }
4411 }
4412
4413 /**
4414  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4415  * @vsi: the VSI being configured
4416  **/
4417 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4418 {
4419         int q_idx;
4420
4421         if (!vsi->netdev)
4422                 return;
4423
4424         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4425                 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4426
4427                 if (q_vector->rx.ring || q_vector->tx.ring)
4428                         napi_disable(&q_vector->napi);
4429         }
4430 }
4431
4432 /**
4433  * i40e_vsi_close - Shut down a VSI
4434  * @vsi: the vsi to be quelled
4435  **/
4436 static void i40e_vsi_close(struct i40e_vsi *vsi)
4437 {
4438         struct i40e_pf *pf = vsi->back;
4439         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4440                 i40e_down(vsi);
4441         i40e_vsi_free_irq(vsi);
4442         i40e_vsi_free_tx_resources(vsi);
4443         i40e_vsi_free_rx_resources(vsi);
4444         vsi->current_netdev_flags = 0;
4445         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
4446         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
4447                 pf->flags |=  I40E_FLAG_CLIENT_RESET;
4448 }
4449
4450 /**
4451  * i40e_quiesce_vsi - Pause a given VSI
4452  * @vsi: the VSI being paused
4453  **/
4454 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4455 {
4456         if (test_bit(__I40E_DOWN, &vsi->state))
4457                 return;
4458
4459         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4460         if (vsi->netdev && netif_running(vsi->netdev))
4461                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4462         else
4463                 i40e_vsi_close(vsi);
4464 }
4465
4466 /**
4467  * i40e_unquiesce_vsi - Resume a given VSI
4468  * @vsi: the VSI being resumed
4469  **/
4470 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4471 {
4472         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4473                 return;
4474
4475         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4476         if (vsi->netdev && netif_running(vsi->netdev))
4477                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4478         else
4479                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4480 }
4481
4482 /**
4483  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4484  * @pf: the PF
4485  **/
4486 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4487 {
4488         int v;
4489
4490         for (v = 0; v < pf->num_alloc_vsi; v++) {
4491                 if (pf->vsi[v])
4492                         i40e_quiesce_vsi(pf->vsi[v]);
4493         }
4494 }
4495
4496 /**
4497  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4498  * @pf: the PF
4499  **/
4500 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4501 {
4502         int v;
4503
4504         for (v = 0; v < pf->num_alloc_vsi; v++) {
4505                 if (pf->vsi[v])
4506                         i40e_unquiesce_vsi(pf->vsi[v]);
4507         }
4508 }
4509
4510 /**
4511  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4512  * @vsi: the VSI being configured
4513  *
4514  * Wait until all queues on a given VSI have been disabled.
4515  **/
4516 int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4517 {
4518         struct i40e_pf *pf = vsi->back;
4519         int i, pf_q, ret;
4520
4521         pf_q = vsi->base_queue;
4522         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4523                 /* Check and wait for the Tx queue */
4524                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4525                 if (ret) {
4526                         dev_info(&pf->pdev->dev,
4527                                  "VSI seid %d Tx ring %d disable timeout\n",
4528                                  vsi->seid, pf_q);
4529                         return ret;
4530                 }
4531                 /* Check and wait for the Tx queue */
4532                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4533                 if (ret) {
4534                         dev_info(&pf->pdev->dev,
4535                                  "VSI seid %d Rx ring %d disable timeout\n",
4536                                  vsi->seid, pf_q);
4537                         return ret;
4538                 }
4539         }
4540
4541         return 0;
4542 }
4543
4544 #ifdef CONFIG_I40E_DCB
4545 /**
4546  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4547  * @pf: the PF
4548  *
4549  * This function waits for the queues to be in disabled state for all the
4550  * VSIs that are managed by this PF.
4551  **/
4552 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4553 {
4554         int v, ret = 0;
4555
4556         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4557                 if (pf->vsi[v]) {
4558                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4559                         if (ret)
4560                                 break;
4561                 }
4562         }
4563
4564         return ret;
4565 }
4566
4567 #endif
4568
4569 /**
4570  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4571  * @q_idx: TX queue number
4572  * @vsi: Pointer to VSI struct
4573  *
4574  * This function checks specified queue for given VSI. Detects hung condition.
4575  * We proactively detect hung TX queues by checking if interrupts are disabled
4576  * but there are pending descriptors.  If it appears hung, attempt to recover
4577  * by triggering a SW interrupt.
4578  **/
4579 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4580 {
4581         struct i40e_ring *tx_ring = NULL;
4582         struct i40e_pf  *pf;
4583         u32 val, tx_pending;
4584         int i;
4585
4586         pf = vsi->back;
4587
4588         /* now that we have an index, find the tx_ring struct */
4589         for (i = 0; i < vsi->num_queue_pairs; i++) {
4590                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4591                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4592                                 tx_ring = vsi->tx_rings[i];
4593                                 break;
4594                         }
4595                 }
4596         }
4597
4598         if (!tx_ring)
4599                 return;
4600
4601         /* Read interrupt register */
4602         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4603                 val = rd32(&pf->hw,
4604                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4605                                                tx_ring->vsi->base_vector - 1));
4606         else
4607                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4608
4609         tx_pending = i40e_get_tx_pending(tx_ring);
4610
4611         /* Interrupts are disabled and TX pending is non-zero,
4612          * trigger the SW interrupt (don't wait). Worst case
4613          * there will be one extra interrupt which may result
4614          * into not cleaning any queues because queues are cleaned.
4615          */
4616         if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4617                 i40e_force_wb(vsi, tx_ring->q_vector);
4618 }
4619
4620 /**
4621  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4622  * @pf:  pointer to PF struct
4623  *
4624  * LAN VSI has netdev and netdev has TX queues. This function is to check
4625  * each of those TX queues if they are hung, trigger recovery by issuing
4626  * SW interrupt.
4627  **/
4628 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4629 {
4630         struct net_device *netdev;
4631         struct i40e_vsi *vsi;
4632         int i;
4633
4634         /* Only for LAN VSI */
4635         vsi = pf->vsi[pf->lan_vsi];
4636
4637         if (!vsi)
4638                 return;
4639
4640         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4641         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4642             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4643                 return;
4644
4645         /* Make sure type is MAIN VSI */
4646         if (vsi->type != I40E_VSI_MAIN)
4647                 return;
4648
4649         netdev = vsi->netdev;
4650         if (!netdev)
4651                 return;
4652
4653         /* Bail out if netif_carrier is not OK */
4654         if (!netif_carrier_ok(netdev))
4655                 return;
4656
4657         /* Go thru' TX queues for netdev */
4658         for (i = 0; i < netdev->num_tx_queues; i++) {
4659                 struct netdev_queue *q;
4660
4661                 q = netdev_get_tx_queue(netdev, i);
4662                 if (q)
4663                         i40e_detect_recover_hung_queue(i, vsi);
4664         }
4665 }
4666
4667 /**
4668  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4669  * @pf: pointer to PF
4670  *
4671  * Get TC map for ISCSI PF type that will include iSCSI TC
4672  * and LAN TC.
4673  **/
4674 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4675 {
4676         struct i40e_dcb_app_priority_table app;
4677         struct i40e_hw *hw = &pf->hw;
4678         u8 enabled_tc = 1; /* TC0 is always enabled */
4679         u8 tc, i;
4680         /* Get the iSCSI APP TLV */
4681         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4682
4683         for (i = 0; i < dcbcfg->numapps; i++) {
4684                 app = dcbcfg->app[i];
4685                 if (app.selector == I40E_APP_SEL_TCPIP &&
4686                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4687                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4688                         enabled_tc |= BIT(tc);
4689                         break;
4690                 }
4691         }
4692
4693         return enabled_tc;
4694 }
4695
4696 /**
4697  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4698  * @dcbcfg: the corresponding DCBx configuration structure
4699  *
4700  * Return the number of TCs from given DCBx configuration
4701  **/
4702 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4703 {
4704         int i, tc_unused = 0;
4705         u8 num_tc = 0;
4706         u8 ret = 0;
4707
4708         /* Scan the ETS Config Priority Table to find
4709          * traffic class enabled for a given priority
4710          * and create a bitmask of enabled TCs
4711          */
4712         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4713                 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4714
4715         /* Now scan the bitmask to check for
4716          * contiguous TCs starting with TC0
4717          */
4718         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4719                 if (num_tc & BIT(i)) {
4720                         if (!tc_unused) {
4721                                 ret++;
4722                         } else {
4723                                 pr_err("Non-contiguous TC - Disabling DCB\n");
4724                                 return 1;
4725                         }
4726                 } else {
4727                         tc_unused = 1;
4728                 }
4729         }
4730
4731         /* There is always at least TC0 */
4732         if (!ret)
4733                 ret = 1;
4734
4735         return ret;
4736 }
4737
4738 /**
4739  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4740  * @dcbcfg: the corresponding DCBx configuration structure
4741  *
4742  * Query the current DCB configuration and return the number of
4743  * traffic classes enabled from the given DCBX config
4744  **/
4745 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4746 {
4747         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4748         u8 enabled_tc = 1;
4749         u8 i;
4750
4751         for (i = 0; i < num_tc; i++)
4752                 enabled_tc |= BIT(i);
4753
4754         return enabled_tc;
4755 }
4756
4757 /**
4758  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4759  * @pf: PF being queried
4760  *
4761  * Return number of traffic classes enabled for the given PF
4762  **/
4763 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4764 {
4765         struct i40e_hw *hw = &pf->hw;
4766         u8 i, enabled_tc = 1;
4767         u8 num_tc = 0;
4768         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4769
4770         /* If DCB is not enabled then always in single TC */
4771         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4772                 return 1;
4773
4774         /* SFP mode will be enabled for all TCs on port */
4775         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4776                 return i40e_dcb_get_num_tc(dcbcfg);
4777
4778         /* MFP mode return count of enabled TCs for this PF */
4779         if (pf->hw.func_caps.iscsi)
4780                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4781         else
4782                 return 1; /* Only TC0 */
4783
4784         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4785                 if (enabled_tc & BIT(i))
4786                         num_tc++;
4787         }
4788         return num_tc;
4789 }
4790
4791 /**
4792  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4793  * @pf: PF being queried
4794  *
4795  * Return a bitmap for enabled traffic classes for this PF.
4796  **/
4797 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4798 {
4799         /* If DCB is not enabled for this PF then just return default TC */
4800         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4801                 return I40E_DEFAULT_TRAFFIC_CLASS;
4802
4803         /* SFP mode we want PF to be enabled for all TCs */
4804         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4805                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4806
4807         /* MFP enabled and iSCSI PF type */
4808         if (pf->hw.func_caps.iscsi)
4809                 return i40e_get_iscsi_tc_map(pf);
4810         else
4811                 return I40E_DEFAULT_TRAFFIC_CLASS;
4812 }
4813
4814 /**
4815  * i40e_vsi_get_bw_info - Query VSI BW Information
4816  * @vsi: the VSI being queried
4817  *
4818  * Returns 0 on success, negative value on failure
4819  **/
4820 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4821 {
4822         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4823         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4824         struct i40e_pf *pf = vsi->back;
4825         struct i40e_hw *hw = &pf->hw;
4826         i40e_status ret;
4827         u32 tc_bw_max;
4828         int i;
4829
4830         /* Get the VSI level BW configuration */
4831         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4832         if (ret) {
4833                 dev_info(&pf->pdev->dev,
4834                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4835                          i40e_stat_str(&pf->hw, ret),
4836                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4837                 return -EINVAL;
4838         }
4839
4840         /* Get the VSI level BW configuration per TC */
4841         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4842                                                NULL);
4843         if (ret) {
4844                 dev_info(&pf->pdev->dev,
4845                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4846                          i40e_stat_str(&pf->hw, ret),
4847                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4848                 return -EINVAL;
4849         }
4850
4851         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4852                 dev_info(&pf->pdev->dev,
4853                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4854                          bw_config.tc_valid_bits,
4855                          bw_ets_config.tc_valid_bits);
4856                 /* Still continuing */
4857         }
4858
4859         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4860         vsi->bw_max_quanta = bw_config.max_bw;
4861         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4862                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4863         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4864                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4865                 vsi->bw_ets_limit_credits[i] =
4866                                         le16_to_cpu(bw_ets_config.credits[i]);
4867                 /* 3 bits out of 4 for each TC */
4868                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4869         }
4870
4871         return 0;
4872 }
4873
4874 /**
4875  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4876  * @vsi: the VSI being configured
4877  * @enabled_tc: TC bitmap
4878  * @bw_credits: BW shared credits per TC
4879  *
4880  * Returns 0 on success, negative value on failure
4881  **/
4882 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4883                                        u8 *bw_share)
4884 {
4885         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4886         i40e_status ret;
4887         int i;
4888
4889         bw_data.tc_valid_bits = enabled_tc;
4890         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4891                 bw_data.tc_bw_credits[i] = bw_share[i];
4892
4893         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4894                                        NULL);
4895         if (ret) {
4896                 dev_info(&vsi->back->pdev->dev,
4897                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4898                          vsi->back->hw.aq.asq_last_status);
4899                 return -EINVAL;
4900         }
4901
4902         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4903                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4904
4905         return 0;
4906 }
4907
4908 /**
4909  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4910  * @vsi: the VSI being configured
4911  * @enabled_tc: TC map to be enabled
4912  *
4913  **/
4914 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4915 {
4916         struct net_device *netdev = vsi->netdev;
4917         struct i40e_pf *pf = vsi->back;
4918         struct i40e_hw *hw = &pf->hw;
4919         u8 netdev_tc = 0;
4920         int i;
4921         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4922
4923         if (!netdev)
4924                 return;
4925
4926         if (!enabled_tc) {
4927                 netdev_reset_tc(netdev);
4928                 return;
4929         }
4930
4931         /* Set up actual enabled TCs on the VSI */
4932         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4933                 return;
4934
4935         /* set per TC queues for the VSI */
4936         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4937                 /* Only set TC queues for enabled tcs
4938                  *
4939                  * e.g. For a VSI that has TC0 and TC3 enabled the
4940                  * enabled_tc bitmap would be 0x00001001; the driver
4941                  * will set the numtc for netdev as 2 that will be
4942                  * referenced by the netdev layer as TC 0 and 1.
4943                  */
4944                 if (vsi->tc_config.enabled_tc & BIT(i))
4945                         netdev_set_tc_queue(netdev,
4946                                         vsi->tc_config.tc_info[i].netdev_tc,
4947                                         vsi->tc_config.tc_info[i].qcount,
4948                                         vsi->tc_config.tc_info[i].qoffset);
4949         }
4950
4951         /* Assign UP2TC map for the VSI */
4952         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4953                 /* Get the actual TC# for the UP */
4954                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4955                 /* Get the mapped netdev TC# for the UP */
4956                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4957                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4958         }
4959 }
4960
4961 /**
4962  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4963  * @vsi: the VSI being configured
4964  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4965  **/
4966 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4967                                       struct i40e_vsi_context *ctxt)
4968 {
4969         /* copy just the sections touched not the entire info
4970          * since not all sections are valid as returned by
4971          * update vsi params
4972          */
4973         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4974         memcpy(&vsi->info.queue_mapping,
4975                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4976         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4977                sizeof(vsi->info.tc_mapping));
4978 }
4979
4980 /**
4981  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4982  * @vsi: VSI to be configured
4983  * @enabled_tc: TC bitmap
4984  *
4985  * This configures a particular VSI for TCs that are mapped to the
4986  * given TC bitmap. It uses default bandwidth share for TCs across
4987  * VSIs to configure TC for a particular VSI.
4988  *
4989  * NOTE:
4990  * It is expected that the VSI queues have been quisced before calling
4991  * this function.
4992  **/
4993 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4994 {
4995         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4996         struct i40e_vsi_context ctxt;
4997         int ret = 0;
4998         int i;
4999
5000         /* Check if enabled_tc is same as existing or new TCs */
5001         if (vsi->tc_config.enabled_tc == enabled_tc)
5002                 return ret;
5003
5004         /* Enable ETS TCs with equal BW Share for now across all VSIs */
5005         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5006                 if (enabled_tc & BIT(i))
5007                         bw_share[i] = 1;
5008         }
5009
5010         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
5011         if (ret) {
5012                 dev_info(&vsi->back->pdev->dev,
5013                          "Failed configuring TC map %d for VSI %d\n",
5014                          enabled_tc, vsi->seid);
5015                 goto out;
5016         }
5017
5018         /* Update Queue Pairs Mapping for currently enabled UPs */
5019         ctxt.seid = vsi->seid;
5020         ctxt.pf_num = vsi->back->hw.pf_id;
5021         ctxt.vf_num = 0;
5022         ctxt.uplink_seid = vsi->uplink_seid;
5023         ctxt.info = vsi->info;
5024         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5025
5026         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
5027                 ctxt.info.valid_sections |=
5028                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
5029                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
5030         }
5031
5032         /* Update the VSI after updating the VSI queue-mapping information */
5033         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5034         if (ret) {
5035                 dev_info(&vsi->back->pdev->dev,
5036                          "Update vsi tc config failed, err %s aq_err %s\n",
5037                          i40e_stat_str(&vsi->back->hw, ret),
5038                          i40e_aq_str(&vsi->back->hw,
5039                                      vsi->back->hw.aq.asq_last_status));
5040                 goto out;
5041         }
5042         /* update the local VSI info with updated queue map */
5043         i40e_vsi_update_queue_map(vsi, &ctxt);
5044         vsi->info.valid_sections = 0;
5045
5046         /* Update current VSI BW information */
5047         ret = i40e_vsi_get_bw_info(vsi);
5048         if (ret) {
5049                 dev_info(&vsi->back->pdev->dev,
5050                          "Failed updating vsi bw info, err %s aq_err %s\n",
5051                          i40e_stat_str(&vsi->back->hw, ret),
5052                          i40e_aq_str(&vsi->back->hw,
5053                                      vsi->back->hw.aq.asq_last_status));
5054                 goto out;
5055         }
5056
5057         /* Update the netdev TC setup */
5058         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
5059 out:
5060         return ret;
5061 }
5062
5063 /**
5064  * i40e_veb_config_tc - Configure TCs for given VEB
5065  * @veb: given VEB
5066  * @enabled_tc: TC bitmap
5067  *
5068  * Configures given TC bitmap for VEB (switching) element
5069  **/
5070 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
5071 {
5072         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
5073         struct i40e_pf *pf = veb->pf;
5074         int ret = 0;
5075         int i;
5076
5077         /* No TCs or already enabled TCs just return */
5078         if (!enabled_tc || veb->enabled_tc == enabled_tc)
5079                 return ret;
5080
5081         bw_data.tc_valid_bits = enabled_tc;
5082         /* bw_data.absolute_credits is not set (relative) */
5083
5084         /* Enable ETS TCs with equal BW Share for now */
5085         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
5086                 if (enabled_tc & BIT(i))
5087                         bw_data.tc_bw_share_credits[i] = 1;
5088         }
5089
5090         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
5091                                                    &bw_data, NULL);
5092         if (ret) {
5093                 dev_info(&pf->pdev->dev,
5094                          "VEB bw config failed, err %s aq_err %s\n",
5095                          i40e_stat_str(&pf->hw, ret),
5096                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5097                 goto out;
5098         }
5099
5100         /* Update the BW information */
5101         ret = i40e_veb_get_bw_info(veb);
5102         if (ret) {
5103                 dev_info(&pf->pdev->dev,
5104                          "Failed getting veb bw config, err %s aq_err %s\n",
5105                          i40e_stat_str(&pf->hw, ret),
5106                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5107         }
5108
5109 out:
5110         return ret;
5111 }
5112
5113 #ifdef CONFIG_I40E_DCB
5114 /**
5115  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
5116  * @pf: PF struct
5117  *
5118  * Reconfigure VEB/VSIs on a given PF; it is assumed that
5119  * the caller would've quiesce all the VSIs before calling
5120  * this function
5121  **/
5122 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5123 {
5124         u8 tc_map = 0;
5125         int ret;
5126         u8 v;
5127
5128         /* Enable the TCs available on PF to all VEBs */
5129         tc_map = i40e_pf_get_tc_map(pf);
5130         for (v = 0; v < I40E_MAX_VEB; v++) {
5131                 if (!pf->veb[v])
5132                         continue;
5133                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5134                 if (ret) {
5135                         dev_info(&pf->pdev->dev,
5136                                  "Failed configuring TC for VEB seid=%d\n",
5137                                  pf->veb[v]->seid);
5138                         /* Will try to configure as many components */
5139                 }
5140         }
5141
5142         /* Update each VSI */
5143         for (v = 0; v < pf->num_alloc_vsi; v++) {
5144                 if (!pf->vsi[v])
5145                         continue;
5146
5147                 /* - Enable all TCs for the LAN VSI
5148                  * - For all others keep them at TC0 for now
5149                  */
5150                 if (v == pf->lan_vsi)
5151                         tc_map = i40e_pf_get_tc_map(pf);
5152                 else
5153                         tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5154
5155                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5156                 if (ret) {
5157                         dev_info(&pf->pdev->dev,
5158                                  "Failed configuring TC for VSI seid=%d\n",
5159                                  pf->vsi[v]->seid);
5160                         /* Will try to configure as many components */
5161                 } else {
5162                         /* Re-configure VSI vectors based on updated TC map */
5163                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5164                         if (pf->vsi[v]->netdev)
5165                                 i40e_dcbnl_set_all(pf->vsi[v]);
5166                 }
5167         }
5168 }
5169
5170 /**
5171  * i40e_resume_port_tx - Resume port Tx
5172  * @pf: PF struct
5173  *
5174  * Resume a port's Tx and issue a PF reset in case of failure to
5175  * resume.
5176  **/
5177 static int i40e_resume_port_tx(struct i40e_pf *pf)
5178 {
5179         struct i40e_hw *hw = &pf->hw;
5180         int ret;
5181
5182         ret = i40e_aq_resume_port_tx(hw, NULL);
5183         if (ret) {
5184                 dev_info(&pf->pdev->dev,
5185                          "Resume Port Tx failed, err %s aq_err %s\n",
5186                           i40e_stat_str(&pf->hw, ret),
5187                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5188                 /* Schedule PF reset to recover */
5189                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5190                 i40e_service_event_schedule(pf);
5191         }
5192
5193         return ret;
5194 }
5195
5196 /**
5197  * i40e_init_pf_dcb - Initialize DCB configuration
5198  * @pf: PF being configured
5199  *
5200  * Query the current DCB configuration and cache it
5201  * in the hardware structure
5202  **/
5203 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5204 {
5205         struct i40e_hw *hw = &pf->hw;
5206         int err = 0;
5207
5208         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5209         if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5210                 goto out;
5211
5212         /* Get the initial DCB configuration */
5213         err = i40e_init_dcb(hw);
5214         if (!err) {
5215                 /* Device/Function is not DCBX capable */
5216                 if ((!hw->func_caps.dcb) ||
5217                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5218                         dev_info(&pf->pdev->dev,
5219                                  "DCBX offload is not supported or is disabled for this PF.\n");
5220                 } else {
5221                         /* When status is not DISABLED then DCBX in FW */
5222                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5223                                        DCB_CAP_DCBX_VER_IEEE;
5224
5225                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5226                         /* Enable DCB tagging only when more than one TC
5227                          * or explicitly disable if only one TC
5228                          */
5229                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5230                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5231                         else
5232                                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5233                         dev_dbg(&pf->pdev->dev,
5234                                 "DCBX offload is supported for this PF.\n");
5235                 }
5236         } else {
5237                 dev_info(&pf->pdev->dev,
5238                          "Query for DCB configuration failed, err %s aq_err %s\n",
5239                          i40e_stat_str(&pf->hw, err),
5240                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5241         }
5242
5243 out:
5244         return err;
5245 }
5246 #endif /* CONFIG_I40E_DCB */
5247 #define SPEED_SIZE 14
5248 #define FC_SIZE 8
5249 /**
5250  * i40e_print_link_message - print link up or down
5251  * @vsi: the VSI for which link needs a message
5252  */
5253 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5254 {
5255         enum i40e_aq_link_speed new_speed;
5256         char *speed = "Unknown";
5257         char *fc = "Unknown";
5258         char *fec = "";
5259         char *an = "";
5260
5261         new_speed = vsi->back->hw.phy.link_info.link_speed;
5262
5263         if ((vsi->current_isup == isup) && (vsi->current_speed == new_speed))
5264                 return;
5265         vsi->current_isup = isup;
5266         vsi->current_speed = new_speed;
5267         if (!isup) {
5268                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5269                 return;
5270         }
5271
5272         /* Warn user if link speed on NPAR enabled partition is not at
5273          * least 10GB
5274          */
5275         if (vsi->back->hw.func_caps.npar_enable &&
5276             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5277              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5278                 netdev_warn(vsi->netdev,
5279                             "The partition detected link speed that is less than 10Gbps\n");
5280
5281         switch (vsi->back->hw.phy.link_info.link_speed) {
5282         case I40E_LINK_SPEED_40GB:
5283                 speed = "40 G";
5284                 break;
5285         case I40E_LINK_SPEED_20GB:
5286                 speed = "20 G";
5287                 break;
5288         case I40E_LINK_SPEED_25GB:
5289                 speed = "25 G";
5290                 break;
5291         case I40E_LINK_SPEED_10GB:
5292                 speed = "10 G";
5293                 break;
5294         case I40E_LINK_SPEED_1GB:
5295                 speed = "1000 M";
5296                 break;
5297         case I40E_LINK_SPEED_100MB:
5298                 speed = "100 M";
5299                 break;
5300         default:
5301                 break;
5302         }
5303
5304         switch (vsi->back->hw.fc.current_mode) {
5305         case I40E_FC_FULL:
5306                 fc = "RX/TX";
5307                 break;
5308         case I40E_FC_TX_PAUSE:
5309                 fc = "TX";
5310                 break;
5311         case I40E_FC_RX_PAUSE:
5312                 fc = "RX";
5313                 break;
5314         default:
5315                 fc = "None";
5316                 break;
5317         }
5318
5319         if (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_25GB) {
5320                 fec = ", FEC: None";
5321                 an = ", Autoneg: False";
5322
5323                 if (vsi->back->hw.phy.link_info.an_info & I40E_AQ_AN_COMPLETED)
5324                         an = ", Autoneg: True";
5325
5326                 if (vsi->back->hw.phy.link_info.fec_info &
5327                     I40E_AQ_CONFIG_FEC_KR_ENA)
5328                         fec = ", FEC: CL74 FC-FEC/BASE-R";
5329                 else if (vsi->back->hw.phy.link_info.fec_info &
5330                          I40E_AQ_CONFIG_FEC_RS_ENA)
5331                         fec = ", FEC: CL108 RS-FEC";
5332         }
5333
5334         netdev_info(vsi->netdev, "NIC Link is Up, %sbps Full Duplex%s%s, Flow Control: %s\n",
5335                     speed, fec, an, fc);
5336 }
5337
5338 /**
5339  * i40e_up_complete - Finish the last steps of bringing up a connection
5340  * @vsi: the VSI being configured
5341  **/
5342 static int i40e_up_complete(struct i40e_vsi *vsi)
5343 {
5344         struct i40e_pf *pf = vsi->back;
5345         int err;
5346
5347         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5348                 i40e_vsi_configure_msix(vsi);
5349         else
5350                 i40e_configure_msi_and_legacy(vsi);
5351
5352         /* start rings */
5353         err = i40e_vsi_start_rings(vsi);
5354         if (err)
5355                 return err;
5356
5357         clear_bit(__I40E_DOWN, &vsi->state);
5358         i40e_napi_enable_all(vsi);
5359         i40e_vsi_enable_irq(vsi);
5360
5361         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5362             (vsi->netdev)) {
5363                 i40e_print_link_message(vsi, true);
5364                 netif_tx_start_all_queues(vsi->netdev);
5365                 netif_carrier_on(vsi->netdev);
5366         } else if (vsi->netdev) {
5367                 i40e_print_link_message(vsi, false);
5368                 /* need to check for qualified module here*/
5369                 if ((pf->hw.phy.link_info.link_info &
5370                         I40E_AQ_MEDIA_AVAILABLE) &&
5371                     (!(pf->hw.phy.link_info.an_info &
5372                         I40E_AQ_QUALIFIED_MODULE)))
5373                         netdev_err(vsi->netdev,
5374                                    "the driver failed to link because an unqualified module was detected.");
5375         }
5376
5377         /* replay FDIR SB filters */
5378         if (vsi->type == I40E_VSI_FDIR) {
5379                 /* reset fd counters */
5380                 pf->fd_add_err = 0;
5381                 pf->fd_atr_cnt = 0;
5382                 i40e_fdir_filter_restore(vsi);
5383         }
5384
5385         /* On the next run of the service_task, notify any clients of the new
5386          * opened netdev
5387          */
5388         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5389         i40e_service_event_schedule(pf);
5390
5391         return 0;
5392 }
5393
5394 /**
5395  * i40e_vsi_reinit_locked - Reset the VSI
5396  * @vsi: the VSI being configured
5397  *
5398  * Rebuild the ring structs after some configuration
5399  * has changed, e.g. MTU size.
5400  **/
5401 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5402 {
5403         struct i40e_pf *pf = vsi->back;
5404
5405         WARN_ON(in_interrupt());
5406         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5407                 usleep_range(1000, 2000);
5408         i40e_down(vsi);
5409
5410         i40e_up(vsi);
5411         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5412 }
5413
5414 /**
5415  * i40e_up - Bring the connection back up after being down
5416  * @vsi: the VSI being configured
5417  **/
5418 int i40e_up(struct i40e_vsi *vsi)
5419 {
5420         int err;
5421
5422         err = i40e_vsi_configure(vsi);
5423         if (!err)
5424                 err = i40e_up_complete(vsi);
5425
5426         return err;
5427 }
5428
5429 /**
5430  * i40e_down - Shutdown the connection processing
5431  * @vsi: the VSI being stopped
5432  **/
5433 void i40e_down(struct i40e_vsi *vsi)
5434 {
5435         int i;
5436
5437         /* It is assumed that the caller of this function
5438          * sets the vsi->state __I40E_DOWN bit.
5439          */
5440         if (vsi->netdev) {
5441                 netif_carrier_off(vsi->netdev);
5442                 netif_tx_disable(vsi->netdev);
5443         }
5444         i40e_vsi_disable_irq(vsi);
5445         i40e_vsi_stop_rings(vsi);
5446         i40e_napi_disable_all(vsi);
5447
5448         for (i = 0; i < vsi->num_queue_pairs; i++) {
5449                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5450                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5451         }
5452
5453 }
5454
5455 /**
5456  * i40e_setup_tc - configure multiple traffic classes
5457  * @netdev: net device to configure
5458  * @tc: number of traffic classes to enable
5459  **/
5460 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5461 {
5462         struct i40e_netdev_priv *np = netdev_priv(netdev);
5463         struct i40e_vsi *vsi = np->vsi;
5464         struct i40e_pf *pf = vsi->back;
5465         u8 enabled_tc = 0;
5466         int ret = -EINVAL;
5467         int i;
5468
5469         /* Check if DCB enabled to continue */
5470         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5471                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5472                 goto exit;
5473         }
5474
5475         /* Check if MFP enabled */
5476         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5477                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5478                 goto exit;
5479         }
5480
5481         /* Check whether tc count is within enabled limit */
5482         if (tc > i40e_pf_get_num_tc(pf)) {
5483                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5484                 goto exit;
5485         }
5486
5487         /* Generate TC map for number of tc requested */
5488         for (i = 0; i < tc; i++)
5489                 enabled_tc |= BIT(i);
5490
5491         /* Requesting same TC configuration as already enabled */
5492         if (enabled_tc == vsi->tc_config.enabled_tc)
5493                 return 0;
5494
5495         /* Quiesce VSI queues */
5496         i40e_quiesce_vsi(vsi);
5497
5498         /* Configure VSI for enabled TCs */
5499         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5500         if (ret) {
5501                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5502                             vsi->seid);
5503                 goto exit;
5504         }
5505
5506         /* Unquiesce VSI */
5507         i40e_unquiesce_vsi(vsi);
5508
5509 exit:
5510         return ret;
5511 }
5512
5513 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5514                            struct tc_to_netdev *tc)
5515 {
5516         if (tc->type != TC_SETUP_MQPRIO)
5517                 return -EINVAL;
5518
5519         tc->mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
5520
5521         return i40e_setup_tc(netdev, tc->mqprio->num_tc);
5522 }
5523
5524 /**
5525  * i40e_open - Called when a network interface is made active
5526  * @netdev: network interface device structure
5527  *
5528  * The open entry point is called when a network interface is made
5529  * active by the system (IFF_UP).  At this point all resources needed
5530  * for transmit and receive operations are allocated, the interrupt
5531  * handler is registered with the OS, the netdev watchdog subtask is
5532  * enabled, and the stack is notified that the interface is ready.
5533  *
5534  * Returns 0 on success, negative value on failure
5535  **/
5536 int i40e_open(struct net_device *netdev)
5537 {
5538         struct i40e_netdev_priv *np = netdev_priv(netdev);
5539         struct i40e_vsi *vsi = np->vsi;
5540         struct i40e_pf *pf = vsi->back;
5541         int err;
5542
5543         /* disallow open during test or if eeprom is broken */
5544         if (test_bit(__I40E_TESTING, &pf->state) ||
5545             test_bit(__I40E_BAD_EEPROM, &pf->state))
5546                 return -EBUSY;
5547
5548         netif_carrier_off(netdev);
5549
5550         err = i40e_vsi_open(vsi);
5551         if (err)
5552                 return err;
5553
5554         /* configure global TSO hardware offload settings */
5555         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5556                                                        TCP_FLAG_FIN) >> 16);
5557         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5558                                                        TCP_FLAG_FIN |
5559                                                        TCP_FLAG_CWR) >> 16);
5560         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5561
5562         udp_tunnel_get_rx_info(netdev);
5563
5564         return 0;
5565 }
5566
5567 /**
5568  * i40e_vsi_open -
5569  * @vsi: the VSI to open
5570  *
5571  * Finish initialization of the VSI.
5572  *
5573  * Returns 0 on success, negative value on failure
5574  *
5575  * Note: expects to be called while under rtnl_lock()
5576  **/
5577 int i40e_vsi_open(struct i40e_vsi *vsi)
5578 {
5579         struct i40e_pf *pf = vsi->back;
5580         char int_name[I40E_INT_NAME_STR_LEN];
5581         int err;
5582
5583         /* allocate descriptors */
5584         err = i40e_vsi_setup_tx_resources(vsi);
5585         if (err)
5586                 goto err_setup_tx;
5587         err = i40e_vsi_setup_rx_resources(vsi);
5588         if (err)
5589                 goto err_setup_rx;
5590
5591         err = i40e_vsi_configure(vsi);
5592         if (err)
5593                 goto err_setup_rx;
5594
5595         if (vsi->netdev) {
5596                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5597                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5598                 err = i40e_vsi_request_irq(vsi, int_name);
5599                 if (err)
5600                         goto err_setup_rx;
5601
5602                 /* Notify the stack of the actual queue counts. */
5603                 err = netif_set_real_num_tx_queues(vsi->netdev,
5604                                                    vsi->num_queue_pairs);
5605                 if (err)
5606                         goto err_set_queues;
5607
5608                 err = netif_set_real_num_rx_queues(vsi->netdev,
5609                                                    vsi->num_queue_pairs);
5610                 if (err)
5611                         goto err_set_queues;
5612
5613         } else if (vsi->type == I40E_VSI_FDIR) {
5614                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5615                          dev_driver_string(&pf->pdev->dev),
5616                          dev_name(&pf->pdev->dev));
5617                 err = i40e_vsi_request_irq(vsi, int_name);
5618
5619         } else {
5620                 err = -EINVAL;
5621                 goto err_setup_rx;
5622         }
5623
5624         err = i40e_up_complete(vsi);
5625         if (err)
5626                 goto err_up_complete;
5627
5628         return 0;
5629
5630 err_up_complete:
5631         i40e_down(vsi);
5632 err_set_queues:
5633         i40e_vsi_free_irq(vsi);
5634 err_setup_rx:
5635         i40e_vsi_free_rx_resources(vsi);
5636 err_setup_tx:
5637         i40e_vsi_free_tx_resources(vsi);
5638         if (vsi == pf->vsi[pf->lan_vsi])
5639                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
5640
5641         return err;
5642 }
5643
5644 /**
5645  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5646  * @pf: Pointer to PF
5647  *
5648  * This function destroys the hlist where all the Flow Director
5649  * filters were saved.
5650  **/
5651 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5652 {
5653         struct i40e_fdir_filter *filter;
5654         struct i40e_flex_pit *pit_entry, *tmp;
5655         struct hlist_node *node2;
5656
5657         hlist_for_each_entry_safe(filter, node2,
5658                                   &pf->fdir_filter_list, fdir_node) {
5659                 hlist_del(&filter->fdir_node);
5660                 kfree(filter);
5661         }
5662
5663         list_for_each_entry_safe(pit_entry, tmp, &pf->l3_flex_pit_list, list) {
5664                 list_del(&pit_entry->list);
5665                 kfree(pit_entry);
5666         }
5667         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
5668
5669         list_for_each_entry_safe(pit_entry, tmp, &pf->l4_flex_pit_list, list) {
5670                 list_del(&pit_entry->list);
5671                 kfree(pit_entry);
5672         }
5673         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
5674
5675         pf->fdir_pf_active_filters = 0;
5676         pf->fd_tcp4_filter_cnt = 0;
5677         pf->fd_udp4_filter_cnt = 0;
5678         pf->fd_sctp4_filter_cnt = 0;
5679         pf->fd_ip4_filter_cnt = 0;
5680
5681         /* Reprogram the default input set for TCP/IPv4 */
5682         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_TCP,
5683                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5684                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5685
5686         /* Reprogram the default input set for UDP/IPv4 */
5687         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_UDP,
5688                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5689                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5690
5691         /* Reprogram the default input set for SCTP/IPv4 */
5692         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_SCTP,
5693                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK |
5694                                 I40E_L4_SRC_MASK | I40E_L4_DST_MASK);
5695
5696         /* Reprogram the default input set for Other/IPv4 */
5697         i40e_write_fd_input_set(pf, I40E_FILTER_PCTYPE_NONF_IPV4_OTHER,
5698                                 I40E_L3_SRC_MASK | I40E_L3_DST_MASK);
5699 }
5700
5701 /**
5702  * i40e_close - Disables a network interface
5703  * @netdev: network interface device structure
5704  *
5705  * The close entry point is called when an interface is de-activated
5706  * by the OS.  The hardware is still under the driver's control, but
5707  * this netdev interface is disabled.
5708  *
5709  * Returns 0, this is not allowed to fail
5710  **/
5711 int i40e_close(struct net_device *netdev)
5712 {
5713         struct i40e_netdev_priv *np = netdev_priv(netdev);
5714         struct i40e_vsi *vsi = np->vsi;
5715
5716         i40e_vsi_close(vsi);
5717
5718         return 0;
5719 }
5720
5721 /**
5722  * i40e_do_reset - Start a PF or Core Reset sequence
5723  * @pf: board private structure
5724  * @reset_flags: which reset is requested
5725  * @lock_acquired: indicates whether or not the lock has been acquired
5726  * before this function was called.
5727  *
5728  * The essential difference in resets is that the PF Reset
5729  * doesn't clear the packet buffers, doesn't reset the PE
5730  * firmware, and doesn't bother the other PFs on the chip.
5731  **/
5732 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags, bool lock_acquired)
5733 {
5734         u32 val;
5735
5736         WARN_ON(in_interrupt());
5737
5738
5739         /* do the biggest reset indicated */
5740         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5741
5742                 /* Request a Global Reset
5743                  *
5744                  * This will start the chip's countdown to the actual full
5745                  * chip reset event, and a warning interrupt to be sent
5746                  * to all PFs, including the requestor.  Our handler
5747                  * for the warning interrupt will deal with the shutdown
5748                  * and recovery of the switch setup.
5749                  */
5750                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5751                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5752                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5753                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5754
5755         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5756
5757                 /* Request a Core Reset
5758                  *
5759                  * Same as Global Reset, except does *not* include the MAC/PHY
5760                  */
5761                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5762                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5763                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5764                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5765                 i40e_flush(&pf->hw);
5766
5767         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5768
5769                 /* Request a PF Reset
5770                  *
5771                  * Resets only the PF-specific registers
5772                  *
5773                  * This goes directly to the tear-down and rebuild of
5774                  * the switch, since we need to do all the recovery as
5775                  * for the Core Reset.
5776                  */
5777                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5778                 i40e_handle_reset_warning(pf, lock_acquired);
5779
5780         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5781                 int v;
5782
5783                 /* Find the VSI(s) that requested a re-init */
5784                 dev_info(&pf->pdev->dev,
5785                          "VSI reinit requested\n");
5786                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5787                         struct i40e_vsi *vsi = pf->vsi[v];
5788
5789                         if (vsi != NULL &&
5790                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5791                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5792                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5793                         }
5794                 }
5795         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5796                 int v;
5797
5798                 /* Find the VSI(s) that needs to be brought down */
5799                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5800                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5801                         struct i40e_vsi *vsi = pf->vsi[v];
5802
5803                         if (vsi != NULL &&
5804                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5805                                 set_bit(__I40E_DOWN, &vsi->state);
5806                                 i40e_down(vsi);
5807                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5808                         }
5809                 }
5810         } else {
5811                 dev_info(&pf->pdev->dev,
5812                          "bad reset request 0x%08x\n", reset_flags);
5813         }
5814 }
5815
5816 #ifdef CONFIG_I40E_DCB
5817 /**
5818  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5819  * @pf: board private structure
5820  * @old_cfg: current DCB config
5821  * @new_cfg: new DCB config
5822  **/
5823 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5824                             struct i40e_dcbx_config *old_cfg,
5825                             struct i40e_dcbx_config *new_cfg)
5826 {
5827         bool need_reconfig = false;
5828
5829         /* Check if ETS configuration has changed */
5830         if (memcmp(&new_cfg->etscfg,
5831                    &old_cfg->etscfg,
5832                    sizeof(new_cfg->etscfg))) {
5833                 /* If Priority Table has changed reconfig is needed */
5834                 if (memcmp(&new_cfg->etscfg.prioritytable,
5835                            &old_cfg->etscfg.prioritytable,
5836                            sizeof(new_cfg->etscfg.prioritytable))) {
5837                         need_reconfig = true;
5838                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5839                 }
5840
5841                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5842                            &old_cfg->etscfg.tcbwtable,
5843                            sizeof(new_cfg->etscfg.tcbwtable)))
5844                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5845
5846                 if (memcmp(&new_cfg->etscfg.tsatable,
5847                            &old_cfg->etscfg.tsatable,
5848                            sizeof(new_cfg->etscfg.tsatable)))
5849                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5850         }
5851
5852         /* Check if PFC configuration has changed */
5853         if (memcmp(&new_cfg->pfc,
5854                    &old_cfg->pfc,
5855                    sizeof(new_cfg->pfc))) {
5856                 need_reconfig = true;
5857                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5858         }
5859
5860         /* Check if APP Table has changed */
5861         if (memcmp(&new_cfg->app,
5862                    &old_cfg->app,
5863                    sizeof(new_cfg->app))) {
5864                 need_reconfig = true;
5865                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5866         }
5867
5868         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5869         return need_reconfig;
5870 }
5871
5872 /**
5873  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5874  * @pf: board private structure
5875  * @e: event info posted on ARQ
5876  **/
5877 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5878                                   struct i40e_arq_event_info *e)
5879 {
5880         struct i40e_aqc_lldp_get_mib *mib =
5881                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5882         struct i40e_hw *hw = &pf->hw;
5883         struct i40e_dcbx_config tmp_dcbx_cfg;
5884         bool need_reconfig = false;
5885         int ret = 0;
5886         u8 type;
5887
5888         /* Not DCB capable or capability disabled */
5889         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5890                 return ret;
5891
5892         /* Ignore if event is not for Nearest Bridge */
5893         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5894                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5895         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5896         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5897                 return ret;
5898
5899         /* Check MIB Type and return if event for Remote MIB update */
5900         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5901         dev_dbg(&pf->pdev->dev,
5902                 "LLDP event mib type %s\n", type ? "remote" : "local");
5903         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5904                 /* Update the remote cached instance and return */
5905                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5906                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5907                                 &hw->remote_dcbx_config);
5908                 goto exit;
5909         }
5910
5911         /* Store the old configuration */
5912         tmp_dcbx_cfg = hw->local_dcbx_config;
5913
5914         /* Reset the old DCBx configuration data */
5915         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5916         /* Get updated DCBX data from firmware */
5917         ret = i40e_get_dcb_config(&pf->hw);
5918         if (ret) {
5919                 dev_info(&pf->pdev->dev,
5920                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5921                          i40e_stat_str(&pf->hw, ret),
5922                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5923                 goto exit;
5924         }
5925
5926         /* No change detected in DCBX configs */
5927         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5928                     sizeof(tmp_dcbx_cfg))) {
5929                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5930                 goto exit;
5931         }
5932
5933         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5934                                                &hw->local_dcbx_config);
5935
5936         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5937
5938         if (!need_reconfig)
5939                 goto exit;
5940
5941         /* Enable DCB tagging only when more than one TC */
5942         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5943                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5944         else
5945                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5946
5947         set_bit(__I40E_PORT_SUSPENDED, &pf->state);
5948         /* Reconfiguration needed quiesce all VSIs */
5949         i40e_pf_quiesce_all_vsi(pf);
5950
5951         /* Changes in configuration update VEB/VSI */
5952         i40e_dcb_reconfigure(pf);
5953
5954         ret = i40e_resume_port_tx(pf);
5955
5956         clear_bit(__I40E_PORT_SUSPENDED, &pf->state);
5957         /* In case of error no point in resuming VSIs */
5958         if (ret)
5959                 goto exit;
5960
5961         /* Wait for the PF's queues to be disabled */
5962         ret = i40e_pf_wait_queues_disabled(pf);
5963         if (ret) {
5964                 /* Schedule PF reset to recover */
5965                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5966                 i40e_service_event_schedule(pf);
5967         } else {
5968                 i40e_pf_unquiesce_all_vsi(pf);
5969         pf->flags |= (I40E_FLAG_SERVICE_CLIENT_REQUESTED |
5970                       I40E_FLAG_CLIENT_L2_CHANGE);
5971         }
5972
5973 exit:
5974         return ret;
5975 }
5976 #endif /* CONFIG_I40E_DCB */
5977
5978 /**
5979  * i40e_do_reset_safe - Protected reset path for userland calls.
5980  * @pf: board private structure
5981  * @reset_flags: which reset is requested
5982  *
5983  **/
5984 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5985 {
5986         rtnl_lock();
5987         i40e_do_reset(pf, reset_flags, true);
5988         rtnl_unlock();
5989 }
5990
5991 /**
5992  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5993  * @pf: board private structure
5994  * @e: event info posted on ARQ
5995  *
5996  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5997  * and VF queues
5998  **/
5999 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
6000                                            struct i40e_arq_event_info *e)
6001 {
6002         struct i40e_aqc_lan_overflow *data =
6003                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
6004         u32 queue = le32_to_cpu(data->prtdcb_rupto);
6005         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
6006         struct i40e_hw *hw = &pf->hw;
6007         struct i40e_vf *vf;
6008         u16 vf_id;
6009
6010         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
6011                 queue, qtx_ctl);
6012
6013         /* Queue belongs to VF, find the VF and issue VF reset */
6014         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
6015             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
6016                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
6017                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
6018                 vf_id -= hw->func_caps.vf_base_id;
6019                 vf = &pf->vf[vf_id];
6020                 i40e_vc_notify_vf_reset(vf);
6021                 /* Allow VF to process pending reset notification */
6022                 msleep(20);
6023                 i40e_reset_vf(vf, false);
6024         }
6025 }
6026
6027 /**
6028  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
6029  * @pf: board private structure
6030  **/
6031 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
6032 {
6033         u32 val, fcnt_prog;
6034
6035         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6036         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
6037         return fcnt_prog;
6038 }
6039
6040 /**
6041  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
6042  * @pf: board private structure
6043  **/
6044 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
6045 {
6046         u32 val, fcnt_prog;
6047
6048         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
6049         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
6050                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
6051                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
6052         return fcnt_prog;
6053 }
6054
6055 /**
6056  * i40e_get_global_fd_count - Get total FD filters programmed on device
6057  * @pf: board private structure
6058  **/
6059 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
6060 {
6061         u32 val, fcnt_prog;
6062
6063         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
6064         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
6065                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
6066                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
6067         return fcnt_prog;
6068 }
6069
6070 /**
6071  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
6072  * @pf: board private structure
6073  **/
6074 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
6075 {
6076         struct i40e_fdir_filter *filter;
6077         u32 fcnt_prog, fcnt_avail;
6078         struct hlist_node *node;
6079
6080         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6081                 return;
6082
6083         /* Check if, FD SB or ATR was auto disabled and if there is enough room
6084          * to re-enable
6085          */
6086         fcnt_prog = i40e_get_global_fd_count(pf);
6087         fcnt_avail = pf->fdir_pf_filter_count;
6088         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
6089             (pf->fd_add_err == 0) ||
6090             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
6091                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
6092                     (pf->hw_disabled_flags & I40E_FLAG_FD_SB_ENABLED)) {
6093                         pf->hw_disabled_flags &= ~I40E_FLAG_FD_SB_ENABLED;
6094                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
6095                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
6096                 }
6097         }
6098
6099         /* Wait for some more space to be available to turn on ATR. We also
6100          * must check that no existing ntuple rules for TCP are in effect
6101          */
6102         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
6103                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
6104                     (pf->hw_disabled_flags & I40E_FLAG_FD_ATR_ENABLED) &&
6105                     (pf->fd_tcp4_filter_cnt == 0)) {
6106                         pf->hw_disabled_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6107                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
6108                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
6109                 }
6110         }
6111
6112         /* if hw had a problem adding a filter, delete it */
6113         if (pf->fd_inv > 0) {
6114                 hlist_for_each_entry_safe(filter, node,
6115                                           &pf->fdir_filter_list, fdir_node) {
6116                         if (filter->fd_id == pf->fd_inv) {
6117                                 hlist_del(&filter->fdir_node);
6118                                 kfree(filter);
6119                                 pf->fdir_pf_active_filters--;
6120                         }
6121                 }
6122         }
6123 }
6124
6125 #define I40E_MIN_FD_FLUSH_INTERVAL 10
6126 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
6127 /**
6128  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
6129  * @pf: board private structure
6130  **/
6131 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
6132 {
6133         unsigned long min_flush_time;
6134         int flush_wait_retry = 50;
6135         bool disable_atr = false;
6136         int fd_room;
6137         int reg;
6138
6139         if (!time_after(jiffies, pf->fd_flush_timestamp +
6140                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
6141                 return;
6142
6143         /* If the flush is happening too quick and we have mostly SB rules we
6144          * should not re-enable ATR for some time.
6145          */
6146         min_flush_time = pf->fd_flush_timestamp +
6147                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
6148         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
6149
6150         if (!(time_after(jiffies, min_flush_time)) &&
6151             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
6152                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6153                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6154                 disable_atr = true;
6155         }
6156
6157         pf->fd_flush_timestamp = jiffies;
6158         pf->hw_disabled_flags |= I40E_FLAG_FD_ATR_ENABLED;
6159         /* flush all filters */
6160         wr32(&pf->hw, I40E_PFQF_CTL_1,
6161              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6162         i40e_flush(&pf->hw);
6163         pf->fd_flush_cnt++;
6164         pf->fd_add_err = 0;
6165         do {
6166                 /* Check FD flush status every 5-6msec */
6167                 usleep_range(5000, 6000);
6168                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6169                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6170                         break;
6171         } while (flush_wait_retry--);
6172         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6173                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6174         } else {
6175                 /* replay sideband filters */
6176                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6177                 if (!disable_atr && !pf->fd_tcp4_filter_cnt)
6178                         pf->hw_disabled_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6179                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6180                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6181                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6182         }
6183 }
6184
6185 /**
6186  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6187  * @pf: board private structure
6188  **/
6189 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6190 {
6191         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6192 }
6193
6194 /* We can see up to 256 filter programming desc in transit if the filters are
6195  * being applied really fast; before we see the first
6196  * filter miss error on Rx queue 0. Accumulating enough error messages before
6197  * reacting will make sure we don't cause flush too often.
6198  */
6199 #define I40E_MAX_FD_PROGRAM_ERROR 256
6200
6201 /**
6202  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6203  * @pf: board private structure
6204  **/
6205 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6206 {
6207
6208         /* if interface is down do nothing */
6209         if (test_bit(__I40E_DOWN, &pf->state))
6210                 return;
6211
6212         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6213                 i40e_fdir_flush_and_replay(pf);
6214
6215         i40e_fdir_check_and_reenable(pf);
6216
6217 }
6218
6219 /**
6220  * i40e_vsi_link_event - notify VSI of a link event
6221  * @vsi: vsi to be notified
6222  * @link_up: link up or down
6223  **/
6224 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6225 {
6226         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6227                 return;
6228
6229         switch (vsi->type) {
6230         case I40E_VSI_MAIN:
6231                 if (!vsi->netdev || !vsi->netdev_registered)
6232                         break;
6233
6234                 if (link_up) {
6235                         netif_carrier_on(vsi->netdev);
6236                         netif_tx_wake_all_queues(vsi->netdev);
6237                 } else {
6238                         netif_carrier_off(vsi->netdev);
6239                         netif_tx_stop_all_queues(vsi->netdev);
6240                 }
6241                 break;
6242
6243         case I40E_VSI_SRIOV:
6244         case I40E_VSI_VMDQ2:
6245         case I40E_VSI_CTRL:
6246         case I40E_VSI_IWARP:
6247         case I40E_VSI_MIRROR:
6248         default:
6249                 /* there is no notification for other VSIs */
6250                 break;
6251         }
6252 }
6253
6254 /**
6255  * i40e_veb_link_event - notify elements on the veb of a link event
6256  * @veb: veb to be notified
6257  * @link_up: link up or down
6258  **/
6259 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6260 {
6261         struct i40e_pf *pf;
6262         int i;
6263
6264         if (!veb || !veb->pf)
6265                 return;
6266         pf = veb->pf;
6267
6268         /* depth first... */
6269         for (i = 0; i < I40E_MAX_VEB; i++)
6270                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6271                         i40e_veb_link_event(pf->veb[i], link_up);
6272
6273         /* ... now the local VSIs */
6274         for (i = 0; i < pf->num_alloc_vsi; i++)
6275                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6276                         i40e_vsi_link_event(pf->vsi[i], link_up);
6277 }
6278
6279 /**
6280  * i40e_link_event - Update netif_carrier status
6281  * @pf: board private structure
6282  **/
6283 static void i40e_link_event(struct i40e_pf *pf)
6284 {
6285         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6286         u8 new_link_speed, old_link_speed;
6287         i40e_status status;
6288         bool new_link, old_link;
6289
6290         /* save off old link status information */
6291         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6292
6293         /* set this to force the get_link_status call to refresh state */
6294         pf->hw.phy.get_link_info = true;
6295
6296         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6297
6298         status = i40e_get_link_status(&pf->hw, &new_link);
6299
6300         /* On success, disable temp link polling */
6301         if (status == I40E_SUCCESS) {
6302                 if (pf->flags & I40E_FLAG_TEMP_LINK_POLLING)
6303                         pf->flags &= ~I40E_FLAG_TEMP_LINK_POLLING;
6304         } else {
6305                 /* Enable link polling temporarily until i40e_get_link_status
6306                  * returns I40E_SUCCESS
6307                  */
6308                 pf->flags |= I40E_FLAG_TEMP_LINK_POLLING;
6309                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6310                         status);
6311                 return;
6312         }
6313
6314         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6315         new_link_speed = pf->hw.phy.link_info.link_speed;
6316
6317         if (new_link == old_link &&
6318             new_link_speed == old_link_speed &&
6319             (test_bit(__I40E_DOWN, &vsi->state) ||
6320              new_link == netif_carrier_ok(vsi->netdev)))
6321                 return;
6322
6323         if (!test_bit(__I40E_DOWN, &vsi->state))
6324                 i40e_print_link_message(vsi, new_link);
6325
6326         /* Notify the base of the switch tree connected to
6327          * the link.  Floating VEBs are not notified.
6328          */
6329         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6330                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6331         else
6332                 i40e_vsi_link_event(vsi, new_link);
6333
6334         if (pf->vf)
6335                 i40e_vc_notify_link_state(pf);
6336
6337         if (pf->flags & I40E_FLAG_PTP)
6338                 i40e_ptp_set_increment(pf);
6339 }
6340
6341 /**
6342  * i40e_watchdog_subtask - periodic checks not using event driven response
6343  * @pf: board private structure
6344  **/
6345 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6346 {
6347         int i;
6348
6349         /* if interface is down do nothing */
6350         if (test_bit(__I40E_DOWN, &pf->state) ||
6351             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6352                 return;
6353
6354         /* make sure we don't do these things too often */
6355         if (time_before(jiffies, (pf->service_timer_previous +
6356                                   pf->service_timer_period)))
6357                 return;
6358         pf->service_timer_previous = jiffies;
6359
6360         if ((pf->flags & I40E_FLAG_LINK_POLLING_ENABLED) ||
6361             (pf->flags & I40E_FLAG_TEMP_LINK_POLLING))
6362                 i40e_link_event(pf);
6363
6364         /* Update the stats for active netdevs so the network stack
6365          * can look at updated numbers whenever it cares to
6366          */
6367         for (i = 0; i < pf->num_alloc_vsi; i++)
6368                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6369                         i40e_update_stats(pf->vsi[i]);
6370
6371         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6372                 /* Update the stats for the active switching components */
6373                 for (i = 0; i < I40E_MAX_VEB; i++)
6374                         if (pf->veb[i])
6375                                 i40e_update_veb_stats(pf->veb[i]);
6376         }
6377
6378         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6379 }
6380
6381 /**
6382  * i40e_reset_subtask - Set up for resetting the device and driver
6383  * @pf: board private structure
6384  **/
6385 static void i40e_reset_subtask(struct i40e_pf *pf)
6386 {
6387         u32 reset_flags = 0;
6388
6389         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6390                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6391                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6392         }
6393         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6394                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6395                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6396         }
6397         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6398                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6399                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6400         }
6401         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6402                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6403                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6404         }
6405         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6406                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6407                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6408         }
6409
6410         /* If there's a recovery already waiting, it takes
6411          * precedence before starting a new reset sequence.
6412          */
6413         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6414                 i40e_prep_for_reset(pf, false);
6415                 i40e_reset(pf);
6416                 i40e_rebuild(pf, false, false);
6417         }
6418
6419         /* If we're already down or resetting, just bail */
6420         if (reset_flags &&
6421             !test_bit(__I40E_DOWN, &pf->state) &&
6422             !test_bit(__I40E_CONFIG_BUSY, &pf->state)) {
6423                 rtnl_lock();
6424                 i40e_do_reset(pf, reset_flags, true);
6425                 rtnl_unlock();
6426         }
6427 }
6428
6429 /**
6430  * i40e_handle_link_event - Handle link event
6431  * @pf: board private structure
6432  * @e: event info posted on ARQ
6433  **/
6434 static void i40e_handle_link_event(struct i40e_pf *pf,
6435                                    struct i40e_arq_event_info *e)
6436 {
6437         struct i40e_aqc_get_link_status *status =
6438                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6439
6440         /* Do a new status request to re-enable LSE reporting
6441          * and load new status information into the hw struct
6442          * This completely ignores any state information
6443          * in the ARQ event info, instead choosing to always
6444          * issue the AQ update link status command.
6445          */
6446         i40e_link_event(pf);
6447
6448         /* check for unqualified module, if link is down */
6449         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6450             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6451             (!(status->link_info & I40E_AQ_LINK_UP)))
6452                 dev_err(&pf->pdev->dev,
6453                         "The driver failed to link because an unqualified module was detected.\n");
6454 }
6455
6456 /**
6457  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6458  * @pf: board private structure
6459  **/
6460 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6461 {
6462         struct i40e_arq_event_info event;
6463         struct i40e_hw *hw = &pf->hw;
6464         u16 pending, i = 0;
6465         i40e_status ret;
6466         u16 opcode;
6467         u32 oldval;
6468         u32 val;
6469
6470         /* Do not run clean AQ when PF reset fails */
6471         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6472                 return;
6473
6474         /* check for error indications */
6475         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6476         oldval = val;
6477         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6478                 if (hw->debug_mask & I40E_DEBUG_AQ)
6479                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6480                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6481         }
6482         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6483                 if (hw->debug_mask & I40E_DEBUG_AQ)
6484                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6485                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6486                 pf->arq_overflows++;
6487         }
6488         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6489                 if (hw->debug_mask & I40E_DEBUG_AQ)
6490                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6491                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6492         }
6493         if (oldval != val)
6494                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6495
6496         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6497         oldval = val;
6498         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6499                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6500                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6501                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6502         }
6503         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6504                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6505                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6506                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6507         }
6508         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6509                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6510                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6511                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6512         }
6513         if (oldval != val)
6514                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6515
6516         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6517         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6518         if (!event.msg_buf)
6519                 return;
6520
6521         do {
6522                 ret = i40e_clean_arq_element(hw, &event, &pending);
6523                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6524                         break;
6525                 else if (ret) {
6526                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6527                         break;
6528                 }
6529
6530                 opcode = le16_to_cpu(event.desc.opcode);
6531                 switch (opcode) {
6532
6533                 case i40e_aqc_opc_get_link_status:
6534                         i40e_handle_link_event(pf, &event);
6535                         break;
6536                 case i40e_aqc_opc_send_msg_to_pf:
6537                         ret = i40e_vc_process_vf_msg(pf,
6538                                         le16_to_cpu(event.desc.retval),
6539                                         le32_to_cpu(event.desc.cookie_high),
6540                                         le32_to_cpu(event.desc.cookie_low),
6541                                         event.msg_buf,
6542                                         event.msg_len);
6543                         break;
6544                 case i40e_aqc_opc_lldp_update_mib:
6545                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6546 #ifdef CONFIG_I40E_DCB
6547                         rtnl_lock();
6548                         ret = i40e_handle_lldp_event(pf, &event);
6549                         rtnl_unlock();
6550 #endif /* CONFIG_I40E_DCB */
6551                         break;
6552                 case i40e_aqc_opc_event_lan_overflow:
6553                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6554                         i40e_handle_lan_overflow_event(pf, &event);
6555                         break;
6556                 case i40e_aqc_opc_send_msg_to_peer:
6557                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6558                         break;
6559                 case i40e_aqc_opc_nvm_erase:
6560                 case i40e_aqc_opc_nvm_update:
6561                 case i40e_aqc_opc_oem_post_update:
6562                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6563                                    "ARQ NVM operation 0x%04x completed\n",
6564                                    opcode);
6565                         break;
6566                 default:
6567                         dev_info(&pf->pdev->dev,
6568                                  "ARQ: Unknown event 0x%04x ignored\n",
6569                                  opcode);
6570                         break;
6571                 }
6572         } while (i++ < pf->adminq_work_limit);
6573
6574         if (i < pf->adminq_work_limit)
6575                 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6576
6577         /* re-enable Admin queue interrupt cause */
6578         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6579         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6580         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6581         i40e_flush(hw);
6582
6583         kfree(event.msg_buf);
6584 }
6585
6586 /**
6587  * i40e_verify_eeprom - make sure eeprom is good to use
6588  * @pf: board private structure
6589  **/
6590 static void i40e_verify_eeprom(struct i40e_pf *pf)
6591 {
6592         int err;
6593
6594         err = i40e_diag_eeprom_test(&pf->hw);
6595         if (err) {
6596                 /* retry in case of garbage read */
6597                 err = i40e_diag_eeprom_test(&pf->hw);
6598                 if (err) {
6599                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6600                                  err);
6601                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6602                 }
6603         }
6604
6605         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6606                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6607                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6608         }
6609 }
6610
6611 /**
6612  * i40e_enable_pf_switch_lb
6613  * @pf: pointer to the PF structure
6614  *
6615  * enable switch loop back or die - no point in a return value
6616  **/
6617 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6618 {
6619         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6620         struct i40e_vsi_context ctxt;
6621         int ret;
6622
6623         ctxt.seid = pf->main_vsi_seid;
6624         ctxt.pf_num = pf->hw.pf_id;
6625         ctxt.vf_num = 0;
6626         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6627         if (ret) {
6628                 dev_info(&pf->pdev->dev,
6629                          "couldn't get PF vsi config, err %s aq_err %s\n",
6630                          i40e_stat_str(&pf->hw, ret),
6631                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6632                 return;
6633         }
6634         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6635         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6636         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6637
6638         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6639         if (ret) {
6640                 dev_info(&pf->pdev->dev,
6641                          "update vsi switch failed, err %s aq_err %s\n",
6642                          i40e_stat_str(&pf->hw, ret),
6643                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6644         }
6645 }
6646
6647 /**
6648  * i40e_disable_pf_switch_lb
6649  * @pf: pointer to the PF structure
6650  *
6651  * disable switch loop back or die - no point in a return value
6652  **/
6653 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6654 {
6655         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6656         struct i40e_vsi_context ctxt;
6657         int ret;
6658
6659         ctxt.seid = pf->main_vsi_seid;
6660         ctxt.pf_num = pf->hw.pf_id;
6661         ctxt.vf_num = 0;
6662         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6663         if (ret) {
6664                 dev_info(&pf->pdev->dev,
6665                          "couldn't get PF vsi config, err %s aq_err %s\n",
6666                          i40e_stat_str(&pf->hw, ret),
6667                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6668                 return;
6669         }
6670         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6671         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6672         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6673
6674         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6675         if (ret) {
6676                 dev_info(&pf->pdev->dev,
6677                          "update vsi switch failed, err %s aq_err %s\n",
6678                          i40e_stat_str(&pf->hw, ret),
6679                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6680         }
6681 }
6682
6683 /**
6684  * i40e_config_bridge_mode - Configure the HW bridge mode
6685  * @veb: pointer to the bridge instance
6686  *
6687  * Configure the loop back mode for the LAN VSI that is downlink to the
6688  * specified HW bridge instance. It is expected this function is called
6689  * when a new HW bridge is instantiated.
6690  **/
6691 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6692 {
6693         struct i40e_pf *pf = veb->pf;
6694
6695         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6696                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6697                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6698         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6699                 i40e_disable_pf_switch_lb(pf);
6700         else
6701                 i40e_enable_pf_switch_lb(pf);
6702 }
6703
6704 /**
6705  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6706  * @veb: pointer to the VEB instance
6707  *
6708  * This is a recursive function that first builds the attached VSIs then
6709  * recurses in to build the next layer of VEB.  We track the connections
6710  * through our own index numbers because the seid's from the HW could
6711  * change across the reset.
6712  **/
6713 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6714 {
6715         struct i40e_vsi *ctl_vsi = NULL;
6716         struct i40e_pf *pf = veb->pf;
6717         int v, veb_idx;
6718         int ret;
6719
6720         /* build VSI that owns this VEB, temporarily attached to base VEB */
6721         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6722                 if (pf->vsi[v] &&
6723                     pf->vsi[v]->veb_idx == veb->idx &&
6724                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6725                         ctl_vsi = pf->vsi[v];
6726                         break;
6727                 }
6728         }
6729         if (!ctl_vsi) {
6730                 dev_info(&pf->pdev->dev,
6731                          "missing owner VSI for veb_idx %d\n", veb->idx);
6732                 ret = -ENOENT;
6733                 goto end_reconstitute;
6734         }
6735         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6736                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6737         ret = i40e_add_vsi(ctl_vsi);
6738         if (ret) {
6739                 dev_info(&pf->pdev->dev,
6740                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6741                          veb->idx, ret);
6742                 goto end_reconstitute;
6743         }
6744         i40e_vsi_reset_stats(ctl_vsi);
6745
6746         /* create the VEB in the switch and move the VSI onto the VEB */
6747         ret = i40e_add_veb(veb, ctl_vsi);
6748         if (ret)
6749                 goto end_reconstitute;
6750
6751         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6752                 veb->bridge_mode = BRIDGE_MODE_VEB;
6753         else
6754                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6755         i40e_config_bridge_mode(veb);
6756
6757         /* create the remaining VSIs attached to this VEB */
6758         for (v = 0; v < pf->num_alloc_vsi; v++) {
6759                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6760                         continue;
6761
6762                 if (pf->vsi[v]->veb_idx == veb->idx) {
6763                         struct i40e_vsi *vsi = pf->vsi[v];
6764
6765                         vsi->uplink_seid = veb->seid;
6766                         ret = i40e_add_vsi(vsi);
6767                         if (ret) {
6768                                 dev_info(&pf->pdev->dev,
6769                                          "rebuild of vsi_idx %d failed: %d\n",
6770                                          v, ret);
6771                                 goto end_reconstitute;
6772                         }
6773                         i40e_vsi_reset_stats(vsi);
6774                 }
6775         }
6776
6777         /* create any VEBs attached to this VEB - RECURSION */
6778         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6779                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6780                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6781                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6782                         if (ret)
6783                                 break;
6784                 }
6785         }
6786
6787 end_reconstitute:
6788         return ret;
6789 }
6790
6791 /**
6792  * i40e_get_capabilities - get info about the HW
6793  * @pf: the PF struct
6794  **/
6795 static int i40e_get_capabilities(struct i40e_pf *pf)
6796 {
6797         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6798         u16 data_size;
6799         int buf_len;
6800         int err;
6801
6802         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6803         do {
6804                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6805                 if (!cap_buf)
6806                         return -ENOMEM;
6807
6808                 /* this loads the data into the hw struct for us */
6809                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6810                                             &data_size,
6811                                             i40e_aqc_opc_list_func_capabilities,
6812                                             NULL);
6813                 /* data loaded, buffer no longer needed */
6814                 kfree(cap_buf);
6815
6816                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6817                         /* retry with a larger buffer */
6818                         buf_len = data_size;
6819                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6820                         dev_info(&pf->pdev->dev,
6821                                  "capability discovery failed, err %s aq_err %s\n",
6822                                  i40e_stat_str(&pf->hw, err),
6823                                  i40e_aq_str(&pf->hw,
6824                                              pf->hw.aq.asq_last_status));
6825                         return -ENODEV;
6826                 }
6827         } while (err);
6828
6829         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6830                 dev_info(&pf->pdev->dev,
6831                          "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",
6832                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6833                          pf->hw.func_caps.num_msix_vectors,
6834                          pf->hw.func_caps.num_msix_vectors_vf,
6835                          pf->hw.func_caps.fd_filters_guaranteed,
6836                          pf->hw.func_caps.fd_filters_best_effort,
6837                          pf->hw.func_caps.num_tx_qp,
6838                          pf->hw.func_caps.num_vsis);
6839
6840 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6841                        + pf->hw.func_caps.num_vfs)
6842         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6843                 dev_info(&pf->pdev->dev,
6844                          "got num_vsis %d, setting num_vsis to %d\n",
6845                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6846                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6847         }
6848
6849         return 0;
6850 }
6851
6852 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6853
6854 /**
6855  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6856  * @pf: board private structure
6857  **/
6858 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6859 {
6860         struct i40e_vsi *vsi;
6861
6862         /* quick workaround for an NVM issue that leaves a critical register
6863          * uninitialized
6864          */
6865         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6866                 static const u32 hkey[] = {
6867                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6868                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6869                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6870                         0x95b3a76d};
6871                 int i;
6872
6873                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6874                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6875         }
6876
6877         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6878                 return;
6879
6880         /* find existing VSI and see if it needs configuring */
6881         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6882
6883         /* create a new VSI if none exists */
6884         if (!vsi) {
6885                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6886                                      pf->vsi[pf->lan_vsi]->seid, 0);
6887                 if (!vsi) {
6888                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6889                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6890                         return;
6891                 }
6892         }
6893
6894         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6895 }
6896
6897 /**
6898  * i40e_fdir_teardown - release the Flow Director resources
6899  * @pf: board private structure
6900  **/
6901 static void i40e_fdir_teardown(struct i40e_pf *pf)
6902 {
6903         struct i40e_vsi *vsi;
6904
6905         i40e_fdir_filter_exit(pf);
6906         vsi = i40e_find_vsi_by_type(pf, I40E_VSI_FDIR);
6907         if (vsi)
6908                 i40e_vsi_release(vsi);
6909 }
6910
6911 /**
6912  * i40e_prep_for_reset - prep for the core to reset
6913  * @pf: board private structure
6914  * @lock_acquired: indicates whether or not the lock has been acquired
6915  * before this function was called.
6916  *
6917  * Close up the VFs and other things in prep for PF Reset.
6918   **/
6919 static void i40e_prep_for_reset(struct i40e_pf *pf, bool lock_acquired)
6920 {
6921         struct i40e_hw *hw = &pf->hw;
6922         i40e_status ret = 0;
6923         u32 v;
6924
6925         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6926         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6927                 return;
6928         if (i40e_check_asq_alive(&pf->hw))
6929                 i40e_vc_notify_reset(pf);
6930
6931         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6932
6933         /* quiesce the VSIs and their queues that are not already DOWN */
6934         /* pf_quiesce_all_vsi modifies netdev structures -rtnl_lock needed */
6935         if (!lock_acquired)
6936                 rtnl_lock();
6937         i40e_pf_quiesce_all_vsi(pf);
6938         if (!lock_acquired)
6939                 rtnl_unlock();
6940
6941         for (v = 0; v < pf->num_alloc_vsi; v++) {
6942                 if (pf->vsi[v])
6943                         pf->vsi[v]->seid = 0;
6944         }
6945
6946         i40e_shutdown_adminq(&pf->hw);
6947
6948         /* call shutdown HMC */
6949         if (hw->hmc.hmc_obj) {
6950                 ret = i40e_shutdown_lan_hmc(hw);
6951                 if (ret)
6952                         dev_warn(&pf->pdev->dev,
6953                                  "shutdown_lan_hmc failed: %d\n", ret);
6954         }
6955 }
6956
6957 /**
6958  * i40e_send_version - update firmware with driver version
6959  * @pf: PF struct
6960  */
6961 static void i40e_send_version(struct i40e_pf *pf)
6962 {
6963         struct i40e_driver_version dv;
6964
6965         dv.major_version = DRV_VERSION_MAJOR;
6966         dv.minor_version = DRV_VERSION_MINOR;
6967         dv.build_version = DRV_VERSION_BUILD;
6968         dv.subbuild_version = 0;
6969         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6970         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6971 }
6972
6973 /**
6974  * i40e_reset - wait for core reset to finish reset, reset pf if corer not seen
6975  * @pf: board private structure
6976  **/
6977 static int i40e_reset(struct i40e_pf *pf)
6978 {
6979         struct i40e_hw *hw = &pf->hw;
6980         i40e_status ret;
6981
6982         ret = i40e_pf_reset(hw);
6983         if (ret) {
6984                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6985                 set_bit(__I40E_RESET_FAILED, &pf->state);
6986                 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6987         } else {
6988                 pf->pfr_count++;
6989         }
6990         return ret;
6991 }
6992
6993 /**
6994  * i40e_rebuild - rebuild using a saved config
6995  * @pf: board private structure
6996  * @reinit: if the Main VSI needs to re-initialized.
6997  * @lock_acquired: indicates whether or not the lock has been acquired
6998  * before this function was called.
6999  **/
7000 static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
7001 {
7002         struct i40e_hw *hw = &pf->hw;
7003         u8 set_fc_aq_fail = 0;
7004         i40e_status ret;
7005         u32 val;
7006         int v;
7007
7008         if (test_bit(__I40E_DOWN, &pf->state))
7009                 goto clear_recovery;
7010         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
7011
7012         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
7013         ret = i40e_init_adminq(&pf->hw);
7014         if (ret) {
7015                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
7016                          i40e_stat_str(&pf->hw, ret),
7017                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7018                 goto clear_recovery;
7019         }
7020
7021         /* re-verify the eeprom if we just had an EMP reset */
7022         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
7023                 i40e_verify_eeprom(pf);
7024
7025         i40e_clear_pxe_mode(hw);
7026         ret = i40e_get_capabilities(pf);
7027         if (ret)
7028                 goto end_core_reset;
7029
7030         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7031                                 hw->func_caps.num_rx_qp, 0, 0);
7032         if (ret) {
7033                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
7034                 goto end_core_reset;
7035         }
7036         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7037         if (ret) {
7038                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
7039                 goto end_core_reset;
7040         }
7041
7042 #ifdef CONFIG_I40E_DCB
7043         ret = i40e_init_pf_dcb(pf);
7044         if (ret) {
7045                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
7046                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
7047                 /* Continue without DCB enabled */
7048         }
7049 #endif /* CONFIG_I40E_DCB */
7050         /* do basic switch setup */
7051         if (!lock_acquired)
7052                 rtnl_lock();
7053         ret = i40e_setup_pf_switch(pf, reinit);
7054         if (ret)
7055                 goto end_unlock;
7056
7057         /* The driver only wants link up/down and module qualification
7058          * reports from firmware.  Note the negative logic.
7059          */
7060         ret = i40e_aq_set_phy_int_mask(&pf->hw,
7061                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
7062                                          I40E_AQ_EVENT_MEDIA_NA |
7063                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
7064         if (ret)
7065                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
7066                          i40e_stat_str(&pf->hw, ret),
7067                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7068
7069         /* make sure our flow control settings are restored */
7070         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
7071         if (ret)
7072                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
7073                         i40e_stat_str(&pf->hw, ret),
7074                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7075
7076         /* Rebuild the VSIs and VEBs that existed before reset.
7077          * They are still in our local switch element arrays, so only
7078          * need to rebuild the switch model in the HW.
7079          *
7080          * If there were VEBs but the reconstitution failed, we'll try
7081          * try to recover minimal use by getting the basic PF VSI working.
7082          */
7083         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
7084                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
7085                 /* find the one VEB connected to the MAC, and find orphans */
7086                 for (v = 0; v < I40E_MAX_VEB; v++) {
7087                         if (!pf->veb[v])
7088                                 continue;
7089
7090                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
7091                             pf->veb[v]->uplink_seid == 0) {
7092                                 ret = i40e_reconstitute_veb(pf->veb[v]);
7093
7094                                 if (!ret)
7095                                         continue;
7096
7097                                 /* If Main VEB failed, we're in deep doodoo,
7098                                  * so give up rebuilding the switch and set up
7099                                  * for minimal rebuild of PF VSI.
7100                                  * If orphan failed, we'll report the error
7101                                  * but try to keep going.
7102                                  */
7103                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
7104                                         dev_info(&pf->pdev->dev,
7105                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
7106                                                  ret);
7107                                         pf->vsi[pf->lan_vsi]->uplink_seid
7108                                                                 = pf->mac_seid;
7109                                         break;
7110                                 } else if (pf->veb[v]->uplink_seid == 0) {
7111                                         dev_info(&pf->pdev->dev,
7112                                                  "rebuild of orphan VEB failed: %d\n",
7113                                                  ret);
7114                                 }
7115                         }
7116                 }
7117         }
7118
7119         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
7120                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
7121                 /* no VEB, so rebuild only the Main VSI */
7122                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
7123                 if (ret) {
7124                         dev_info(&pf->pdev->dev,
7125                                  "rebuild of Main VSI failed: %d\n", ret);
7126                         goto end_unlock;
7127                 }
7128         }
7129
7130         /* Reconfigure hardware for allowing smaller MSS in the case
7131          * of TSO, so that we avoid the MDD being fired and causing
7132          * a reset in the case of small MSS+TSO.
7133          */
7134 #define I40E_REG_MSS          0x000E64DC
7135 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
7136 #define I40E_64BYTE_MSS       0x400000
7137         val = rd32(hw, I40E_REG_MSS);
7138         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
7139                 val &= ~I40E_REG_MSS_MIN_MASK;
7140                 val |= I40E_64BYTE_MSS;
7141                 wr32(hw, I40E_REG_MSS, val);
7142         }
7143
7144         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
7145                 msleep(75);
7146                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
7147                 if (ret)
7148                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
7149                                  i40e_stat_str(&pf->hw, ret),
7150                                  i40e_aq_str(&pf->hw,
7151                                              pf->hw.aq.asq_last_status));
7152         }
7153         /* reinit the misc interrupt */
7154         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7155                 ret = i40e_setup_misc_vector(pf);
7156
7157         /* Add a filter to drop all Flow control frames from any VSI from being
7158          * transmitted. By doing so we stop a malicious VF from sending out
7159          * PAUSE or PFC frames and potentially controlling traffic for other
7160          * PF/VF VSIs.
7161          * The FW can still send Flow control frames if enabled.
7162          */
7163         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7164                                                        pf->main_vsi_seid);
7165
7166         /* restart the VSIs that were rebuilt and running before the reset */
7167         i40e_pf_unquiesce_all_vsi(pf);
7168
7169         /* Release the RTNL lock before we start resetting VFs */
7170         if (!lock_acquired)
7171                 rtnl_unlock();
7172
7173         i40e_reset_all_vfs(pf, true);
7174
7175         /* tell the firmware that we're starting */
7176         i40e_send_version(pf);
7177
7178         /* We've already released the lock, so don't do it again */
7179         goto end_core_reset;
7180
7181 end_unlock:
7182         if (!lock_acquired)
7183                 rtnl_unlock();
7184 end_core_reset:
7185         clear_bit(__I40E_RESET_FAILED, &pf->state);
7186 clear_recovery:
7187         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
7188 }
7189
7190 /**
7191  * i40e_reset_and_rebuild - reset and rebuild using a saved config
7192  * @pf: board private structure
7193  * @reinit: if the Main VSI needs to re-initialized.
7194  * @lock_acquired: indicates whether or not the lock has been acquired
7195  * before this function was called.
7196  **/
7197 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit,
7198                                    bool lock_acquired)
7199 {
7200         int ret;
7201         /* Now we wait for GRST to settle out.
7202          * We don't have to delete the VEBs or VSIs from the hw switch
7203          * because the reset will make them disappear.
7204          */
7205         ret = i40e_reset(pf);
7206         if (!ret)
7207                 i40e_rebuild(pf, reinit, lock_acquired);
7208 }
7209
7210 /**
7211  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7212  * @pf: board private structure
7213  *
7214  * Close up the VFs and other things in prep for a Core Reset,
7215  * then get ready to rebuild the world.
7216  * @lock_acquired: indicates whether or not the lock has been acquired
7217  * before this function was called.
7218  **/
7219 static void i40e_handle_reset_warning(struct i40e_pf *pf, bool lock_acquired)
7220 {
7221         i40e_prep_for_reset(pf, lock_acquired);
7222         i40e_reset_and_rebuild(pf, false, lock_acquired);
7223 }
7224
7225 /**
7226  * i40e_handle_mdd_event
7227  * @pf: pointer to the PF structure
7228  *
7229  * Called from the MDD irq handler to identify possibly malicious vfs
7230  **/
7231 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7232 {
7233         struct i40e_hw *hw = &pf->hw;
7234         bool mdd_detected = false;
7235         bool pf_mdd_detected = false;
7236         struct i40e_vf *vf;
7237         u32 reg;
7238         int i;
7239
7240         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7241                 return;
7242
7243         /* find what triggered the MDD event */
7244         reg = rd32(hw, I40E_GL_MDET_TX);
7245         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7246                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7247                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7248                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7249                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7250                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7251                                 I40E_GL_MDET_TX_EVENT_SHIFT;
7252                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7253                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7254                                 pf->hw.func_caps.base_queue;
7255                 if (netif_msg_tx_err(pf))
7256                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7257                                  event, queue, pf_num, vf_num);
7258                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7259                 mdd_detected = true;
7260         }
7261         reg = rd32(hw, I40E_GL_MDET_RX);
7262         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7263                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7264                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7265                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7266                                 I40E_GL_MDET_RX_EVENT_SHIFT;
7267                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7268                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7269                                 pf->hw.func_caps.base_queue;
7270                 if (netif_msg_rx_err(pf))
7271                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7272                                  event, queue, func);
7273                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7274                 mdd_detected = true;
7275         }
7276
7277         if (mdd_detected) {
7278                 reg = rd32(hw, I40E_PF_MDET_TX);
7279                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7280                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7281                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7282                         pf_mdd_detected = true;
7283                 }
7284                 reg = rd32(hw, I40E_PF_MDET_RX);
7285                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7286                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7287                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7288                         pf_mdd_detected = true;
7289                 }
7290                 /* Queue belongs to the PF, initiate a reset */
7291                 if (pf_mdd_detected) {
7292                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7293                         i40e_service_event_schedule(pf);
7294                 }
7295         }
7296
7297         /* see if one of the VFs needs its hand slapped */
7298         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7299                 vf = &(pf->vf[i]);
7300                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7301                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7302                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7303                         vf->num_mdd_events++;
7304                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7305                                  i);
7306                 }
7307
7308                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7309                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7310                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7311                         vf->num_mdd_events++;
7312                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7313                                  i);
7314                 }
7315
7316                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7317                         dev_info(&pf->pdev->dev,
7318                                  "Too many MDD events on VF %d, disabled\n", i);
7319                         dev_info(&pf->pdev->dev,
7320                                  "Use PF Control I/F to re-enable the VF\n");
7321                         set_bit(I40E_VF_STATE_DISABLED, &vf->vf_states);
7322                 }
7323         }
7324
7325         /* re-enable mdd interrupt cause */
7326         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7327         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7328         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7329         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7330         i40e_flush(hw);
7331 }
7332
7333 /**
7334  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7335  * @pf: board private structure
7336  **/
7337 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7338 {
7339         struct i40e_hw *hw = &pf->hw;
7340         i40e_status ret;
7341         u16 port;
7342         int i;
7343
7344         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7345                 return;
7346
7347         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7348
7349         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7350                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7351                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7352                         port = pf->udp_ports[i].index;
7353                         if (port)
7354                                 ret = i40e_aq_add_udp_tunnel(hw, port,
7355                                                         pf->udp_ports[i].type,
7356                                                         NULL, NULL);
7357                         else
7358                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7359
7360                         if (ret) {
7361                                 dev_dbg(&pf->pdev->dev,
7362                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
7363                                         pf->udp_ports[i].type ? "vxlan" : "geneve",
7364                                         port ? "add" : "delete",
7365                                         port, i,
7366                                         i40e_stat_str(&pf->hw, ret),
7367                                         i40e_aq_str(&pf->hw,
7368                                                     pf->hw.aq.asq_last_status));
7369                                 pf->udp_ports[i].index = 0;
7370                         }
7371                 }
7372         }
7373 }
7374
7375 /**
7376  * i40e_service_task - Run the driver's async subtasks
7377  * @work: pointer to work_struct containing our data
7378  **/
7379 static void i40e_service_task(struct work_struct *work)
7380 {
7381         struct i40e_pf *pf = container_of(work,
7382                                           struct i40e_pf,
7383                                           service_task);
7384         unsigned long start_time = jiffies;
7385
7386         /* don't bother with service tasks if a reset is in progress */
7387         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7388                 return;
7389         }
7390
7391         if (test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
7392                 return;
7393
7394         i40e_detect_recover_hung(pf);
7395         i40e_sync_filters_subtask(pf);
7396         i40e_reset_subtask(pf);
7397         i40e_handle_mdd_event(pf);
7398         i40e_vc_process_vflr_event(pf);
7399         i40e_watchdog_subtask(pf);
7400         i40e_fdir_reinit_subtask(pf);
7401         if (pf->flags & I40E_FLAG_CLIENT_RESET) {
7402                 /* Client subtask will reopen next time through. */
7403                 i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], true);
7404                 pf->flags &= ~I40E_FLAG_CLIENT_RESET;
7405         } else {
7406                 i40e_client_subtask(pf);
7407                 if (pf->flags & I40E_FLAG_CLIENT_L2_CHANGE) {
7408                         i40e_notify_client_of_l2_param_changes(
7409                                                         pf->vsi[pf->lan_vsi]);
7410                         pf->flags &= ~I40E_FLAG_CLIENT_L2_CHANGE;
7411                 }
7412         }
7413         i40e_sync_filters_subtask(pf);
7414         i40e_sync_udp_filters_subtask(pf);
7415         i40e_clean_adminq_subtask(pf);
7416
7417         /* flush memory to make sure state is correct before next watchdog */
7418         smp_mb__before_atomic();
7419         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
7420
7421         /* If the tasks have taken longer than one timer cycle or there
7422          * is more work to be done, reschedule the service task now
7423          * rather than wait for the timer to tick again.
7424          */
7425         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7426             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7427             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7428             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7429                 i40e_service_event_schedule(pf);
7430 }
7431
7432 /**
7433  * i40e_service_timer - timer callback
7434  * @data: pointer to PF struct
7435  **/
7436 static void i40e_service_timer(unsigned long data)
7437 {
7438         struct i40e_pf *pf = (struct i40e_pf *)data;
7439
7440         mod_timer(&pf->service_timer,
7441                   round_jiffies(jiffies + pf->service_timer_period));
7442         i40e_service_event_schedule(pf);
7443 }
7444
7445 /**
7446  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7447  * @vsi: the VSI being configured
7448  **/
7449 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7450 {
7451         struct i40e_pf *pf = vsi->back;
7452
7453         switch (vsi->type) {
7454         case I40E_VSI_MAIN:
7455                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7456                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7457                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7458                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7459                         vsi->num_q_vectors = pf->num_lan_msix;
7460                 else
7461                         vsi->num_q_vectors = 1;
7462
7463                 break;
7464
7465         case I40E_VSI_FDIR:
7466                 vsi->alloc_queue_pairs = 1;
7467                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7468                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7469                 vsi->num_q_vectors = pf->num_fdsb_msix;
7470                 break;
7471
7472         case I40E_VSI_VMDQ2:
7473                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7474                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7475                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7476                 vsi->num_q_vectors = pf->num_vmdq_msix;
7477                 break;
7478
7479         case I40E_VSI_SRIOV:
7480                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7481                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7482                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7483                 break;
7484
7485         default:
7486                 WARN_ON(1);
7487                 return -ENODATA;
7488         }
7489
7490         return 0;
7491 }
7492
7493 /**
7494  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7495  * @type: VSI pointer
7496  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7497  *
7498  * On error: returns error code (negative)
7499  * On success: returns 0
7500  **/
7501 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7502 {
7503         int size;
7504         int ret = 0;
7505
7506         /* allocate memory for both Tx and Rx ring pointers */
7507         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7508         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7509         if (!vsi->tx_rings)
7510                 return -ENOMEM;
7511         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7512
7513         if (alloc_qvectors) {
7514                 /* allocate memory for q_vector pointers */
7515                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7516                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7517                 if (!vsi->q_vectors) {
7518                         ret = -ENOMEM;
7519                         goto err_vectors;
7520                 }
7521         }
7522         return ret;
7523
7524 err_vectors:
7525         kfree(vsi->tx_rings);
7526         return ret;
7527 }
7528
7529 /**
7530  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7531  * @pf: board private structure
7532  * @type: type of VSI
7533  *
7534  * On error: returns error code (negative)
7535  * On success: returns vsi index in PF (positive)
7536  **/
7537 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7538 {
7539         int ret = -ENODEV;
7540         struct i40e_vsi *vsi;
7541         int vsi_idx;
7542         int i;
7543
7544         /* Need to protect the allocation of the VSIs at the PF level */
7545         mutex_lock(&pf->switch_mutex);
7546
7547         /* VSI list may be fragmented if VSI creation/destruction has
7548          * been happening.  We can afford to do a quick scan to look
7549          * for any free VSIs in the list.
7550          *
7551          * find next empty vsi slot, looping back around if necessary
7552          */
7553         i = pf->next_vsi;
7554         while (i < pf->num_alloc_vsi && pf->vsi[i])
7555                 i++;
7556         if (i >= pf->num_alloc_vsi) {
7557                 i = 0;
7558                 while (i < pf->next_vsi && pf->vsi[i])
7559                         i++;
7560         }
7561
7562         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7563                 vsi_idx = i;             /* Found one! */
7564         } else {
7565                 ret = -ENODEV;
7566                 goto unlock_pf;  /* out of VSI slots! */
7567         }
7568         pf->next_vsi = ++i;
7569
7570         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7571         if (!vsi) {
7572                 ret = -ENOMEM;
7573                 goto unlock_pf;
7574         }
7575         vsi->type = type;
7576         vsi->back = pf;
7577         set_bit(__I40E_DOWN, &vsi->state);
7578         vsi->flags = 0;
7579         vsi->idx = vsi_idx;
7580         vsi->int_rate_limit = 0;
7581         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7582                                 pf->rss_table_size : 64;
7583         vsi->netdev_registered = false;
7584         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7585         hash_init(vsi->mac_filter_hash);
7586         vsi->irqs_ready = false;
7587
7588         ret = i40e_set_num_rings_in_vsi(vsi);
7589         if (ret)
7590                 goto err_rings;
7591
7592         ret = i40e_vsi_alloc_arrays(vsi, true);
7593         if (ret)
7594                 goto err_rings;
7595
7596         /* Setup default MSIX irq handler for VSI */
7597         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7598
7599         /* Initialize VSI lock */
7600         spin_lock_init(&vsi->mac_filter_hash_lock);
7601         pf->vsi[vsi_idx] = vsi;
7602         ret = vsi_idx;
7603         goto unlock_pf;
7604
7605 err_rings:
7606         pf->next_vsi = i - 1;
7607         kfree(vsi);
7608 unlock_pf:
7609         mutex_unlock(&pf->switch_mutex);
7610         return ret;
7611 }
7612
7613 /**
7614  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7615  * @type: VSI pointer
7616  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7617  *
7618  * On error: returns error code (negative)
7619  * On success: returns 0
7620  **/
7621 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7622 {
7623         /* free the ring and vector containers */
7624         if (free_qvectors) {
7625                 kfree(vsi->q_vectors);
7626                 vsi->q_vectors = NULL;
7627         }
7628         kfree(vsi->tx_rings);
7629         vsi->tx_rings = NULL;
7630         vsi->rx_rings = NULL;
7631 }
7632
7633 /**
7634  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7635  * and lookup table
7636  * @vsi: Pointer to VSI structure
7637  */
7638 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7639 {
7640         if (!vsi)
7641                 return;
7642
7643         kfree(vsi->rss_hkey_user);
7644         vsi->rss_hkey_user = NULL;
7645
7646         kfree(vsi->rss_lut_user);
7647         vsi->rss_lut_user = NULL;
7648 }
7649
7650 /**
7651  * i40e_vsi_clear - Deallocate the VSI provided
7652  * @vsi: the VSI being un-configured
7653  **/
7654 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7655 {
7656         struct i40e_pf *pf;
7657
7658         if (!vsi)
7659                 return 0;
7660
7661         if (!vsi->back)
7662                 goto free_vsi;
7663         pf = vsi->back;
7664
7665         mutex_lock(&pf->switch_mutex);
7666         if (!pf->vsi[vsi->idx]) {
7667                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7668                         vsi->idx, vsi->idx, vsi, vsi->type);
7669                 goto unlock_vsi;
7670         }
7671
7672         if (pf->vsi[vsi->idx] != vsi) {
7673                 dev_err(&pf->pdev->dev,
7674                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7675                         pf->vsi[vsi->idx]->idx,
7676                         pf->vsi[vsi->idx],
7677                         pf->vsi[vsi->idx]->type,
7678                         vsi->idx, vsi, vsi->type);
7679                 goto unlock_vsi;
7680         }
7681
7682         /* updates the PF for this cleared vsi */
7683         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7684         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7685
7686         i40e_vsi_free_arrays(vsi, true);
7687         i40e_clear_rss_config_user(vsi);
7688
7689         pf->vsi[vsi->idx] = NULL;
7690         if (vsi->idx < pf->next_vsi)
7691                 pf->next_vsi = vsi->idx;
7692
7693 unlock_vsi:
7694         mutex_unlock(&pf->switch_mutex);
7695 free_vsi:
7696         kfree(vsi);
7697
7698         return 0;
7699 }
7700
7701 /**
7702  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7703  * @vsi: the VSI being cleaned
7704  **/
7705 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7706 {
7707         int i;
7708
7709         if (vsi->tx_rings && vsi->tx_rings[0]) {
7710                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7711                         kfree_rcu(vsi->tx_rings[i], rcu);
7712                         vsi->tx_rings[i] = NULL;
7713                         vsi->rx_rings[i] = NULL;
7714                 }
7715         }
7716 }
7717
7718 /**
7719  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7720  * @vsi: the VSI being configured
7721  **/
7722 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7723 {
7724         struct i40e_ring *tx_ring, *rx_ring;
7725         struct i40e_pf *pf = vsi->back;
7726         int i;
7727
7728         /* Set basic values in the rings to be used later during open() */
7729         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7730                 /* allocate space for both Tx and Rx in one shot */
7731                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7732                 if (!tx_ring)
7733                         goto err_out;
7734
7735                 tx_ring->queue_index = i;
7736                 tx_ring->reg_idx = vsi->base_queue + i;
7737                 tx_ring->ring_active = false;
7738                 tx_ring->vsi = vsi;
7739                 tx_ring->netdev = vsi->netdev;
7740                 tx_ring->dev = &pf->pdev->dev;
7741                 tx_ring->count = vsi->num_desc;
7742                 tx_ring->size = 0;
7743                 tx_ring->dcb_tc = 0;
7744                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7745                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7746                 tx_ring->tx_itr_setting = pf->tx_itr_default;
7747                 vsi->tx_rings[i] = tx_ring;
7748
7749                 rx_ring = &tx_ring[1];
7750                 rx_ring->queue_index = i;
7751                 rx_ring->reg_idx = vsi->base_queue + i;
7752                 rx_ring->ring_active = false;
7753                 rx_ring->vsi = vsi;
7754                 rx_ring->netdev = vsi->netdev;
7755                 rx_ring->dev = &pf->pdev->dev;
7756                 rx_ring->count = vsi->num_desc;
7757                 rx_ring->size = 0;
7758                 rx_ring->dcb_tc = 0;
7759                 rx_ring->rx_itr_setting = pf->rx_itr_default;
7760                 vsi->rx_rings[i] = rx_ring;
7761         }
7762
7763         return 0;
7764
7765 err_out:
7766         i40e_vsi_clear_rings(vsi);
7767         return -ENOMEM;
7768 }
7769
7770 /**
7771  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7772  * @pf: board private structure
7773  * @vectors: the number of MSI-X vectors to request
7774  *
7775  * Returns the number of vectors reserved, or error
7776  **/
7777 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7778 {
7779         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7780                                         I40E_MIN_MSIX, vectors);
7781         if (vectors < 0) {
7782                 dev_info(&pf->pdev->dev,
7783                          "MSI-X vector reservation failed: %d\n", vectors);
7784                 vectors = 0;
7785         }
7786
7787         return vectors;
7788 }
7789
7790 /**
7791  * i40e_init_msix - Setup the MSIX capability
7792  * @pf: board private structure
7793  *
7794  * Work with the OS to set up the MSIX vectors needed.
7795  *
7796  * Returns the number of vectors reserved or negative on failure
7797  **/
7798 static int i40e_init_msix(struct i40e_pf *pf)
7799 {
7800         struct i40e_hw *hw = &pf->hw;
7801         int cpus, extra_vectors;
7802         int vectors_left;
7803         int v_budget, i;
7804         int v_actual;
7805         int iwarp_requested = 0;
7806
7807         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7808                 return -ENODEV;
7809
7810         /* The number of vectors we'll request will be comprised of:
7811          *   - Add 1 for "other" cause for Admin Queue events, etc.
7812          *   - The number of LAN queue pairs
7813          *      - Queues being used for RSS.
7814          *              We don't need as many as max_rss_size vectors.
7815          *              use rss_size instead in the calculation since that
7816          *              is governed by number of cpus in the system.
7817          *      - assumes symmetric Tx/Rx pairing
7818          *   - The number of VMDq pairs
7819          *   - The CPU count within the NUMA node if iWARP is enabled
7820          * Once we count this up, try the request.
7821          *
7822          * If we can't get what we want, we'll simplify to nearly nothing
7823          * and try again.  If that still fails, we punt.
7824          */
7825         vectors_left = hw->func_caps.num_msix_vectors;
7826         v_budget = 0;
7827
7828         /* reserve one vector for miscellaneous handler */
7829         if (vectors_left) {
7830                 v_budget++;
7831                 vectors_left--;
7832         }
7833
7834         /* reserve some vectors for the main PF traffic queues. Initially we
7835          * only reserve at most 50% of the available vectors, in the case that
7836          * the number of online CPUs is large. This ensures that we can enable
7837          * extra features as well. Once we've enabled the other features, we
7838          * will use any remaining vectors to reach as close as we can to the
7839          * number of online CPUs.
7840          */
7841         cpus = num_online_cpus();
7842         pf->num_lan_msix = min_t(int, cpus, vectors_left / 2);
7843         vectors_left -= pf->num_lan_msix;
7844
7845         /* reserve one vector for sideband flow director */
7846         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7847                 if (vectors_left) {
7848                         pf->num_fdsb_msix = 1;
7849                         v_budget++;
7850                         vectors_left--;
7851                 } else {
7852                         pf->num_fdsb_msix = 0;
7853                 }
7854         }
7855
7856         /* can we reserve enough for iWARP? */
7857         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7858                 iwarp_requested = pf->num_iwarp_msix;
7859
7860                 if (!vectors_left)
7861                         pf->num_iwarp_msix = 0;
7862                 else if (vectors_left < pf->num_iwarp_msix)
7863                         pf->num_iwarp_msix = 1;
7864                 v_budget += pf->num_iwarp_msix;
7865                 vectors_left -= pf->num_iwarp_msix;
7866         }
7867
7868         /* any vectors left over go for VMDq support */
7869         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7870                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7871                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7872
7873                 if (!vectors_left) {
7874                         pf->num_vmdq_msix = 0;
7875                         pf->num_vmdq_qps = 0;
7876                 } else {
7877                         /* if we're short on vectors for what's desired, we limit
7878                          * the queues per vmdq.  If this is still more than are
7879                          * available, the user will need to change the number of
7880                          * queues/vectors used by the PF later with the ethtool
7881                          * channels command
7882                          */
7883                         if (vmdq_vecs < vmdq_vecs_wanted)
7884                                 pf->num_vmdq_qps = 1;
7885                         pf->num_vmdq_msix = pf->num_vmdq_qps;
7886
7887                         v_budget += vmdq_vecs;
7888                         vectors_left -= vmdq_vecs;
7889                 }
7890         }
7891
7892         /* On systems with a large number of SMP cores, we previously limited
7893          * the number of vectors for num_lan_msix to be at most 50% of the
7894          * available vectors, to allow for other features. Now, we add back
7895          * the remaining vectors. However, we ensure that the total
7896          * num_lan_msix will not exceed num_online_cpus(). To do this, we
7897          * calculate the number of vectors we can add without going over the
7898          * cap of CPUs. For systems with a small number of CPUs this will be
7899          * zero.
7900          */
7901         extra_vectors = min_t(int, cpus - pf->num_lan_msix, vectors_left);
7902         pf->num_lan_msix += extra_vectors;
7903         vectors_left -= extra_vectors;
7904
7905         WARN(vectors_left < 0,
7906              "Calculation of remaining vectors underflowed. This is an accounting bug when determining total MSI-X vectors.\n");
7907
7908         v_budget += pf->num_lan_msix;
7909         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7910                                    GFP_KERNEL);
7911         if (!pf->msix_entries)
7912                 return -ENOMEM;
7913
7914         for (i = 0; i < v_budget; i++)
7915                 pf->msix_entries[i].entry = i;
7916         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7917
7918         if (v_actual < I40E_MIN_MSIX) {
7919                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7920                 kfree(pf->msix_entries);
7921                 pf->msix_entries = NULL;
7922                 pci_disable_msix(pf->pdev);
7923                 return -ENODEV;
7924
7925         } else if (v_actual == I40E_MIN_MSIX) {
7926                 /* Adjust for minimal MSIX use */
7927                 pf->num_vmdq_vsis = 0;
7928                 pf->num_vmdq_qps = 0;
7929                 pf->num_lan_qps = 1;
7930                 pf->num_lan_msix = 1;
7931
7932         } else if (!vectors_left) {
7933                 /* If we have limited resources, we will start with no vectors
7934                  * for the special features and then allocate vectors to some
7935                  * of these features based on the policy and at the end disable
7936                  * the features that did not get any vectors.
7937                  */
7938                 int vec;
7939
7940                 dev_info(&pf->pdev->dev,
7941                          "MSI-X vector limit reached, attempting to redistribute vectors\n");
7942                 /* reserve the misc vector */
7943                 vec = v_actual - 1;
7944
7945                 /* Scale vector usage down */
7946                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7947                 pf->num_vmdq_vsis = 1;
7948                 pf->num_vmdq_qps = 1;
7949
7950                 /* partition out the remaining vectors */
7951                 switch (vec) {
7952                 case 2:
7953                         pf->num_lan_msix = 1;
7954                         break;
7955                 case 3:
7956                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7957                                 pf->num_lan_msix = 1;
7958                                 pf->num_iwarp_msix = 1;
7959                         } else {
7960                                 pf->num_lan_msix = 2;
7961                         }
7962                         break;
7963                 default:
7964                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7965                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7966                                                  iwarp_requested);
7967                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7968                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7969                         } else {
7970                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7971                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7972                         }
7973                         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7974                                 pf->num_fdsb_msix = 1;
7975                                 vec--;
7976                         }
7977                         pf->num_lan_msix = min_t(int,
7978                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7979                                                               pf->num_lan_msix);
7980                         pf->num_lan_qps = pf->num_lan_msix;
7981                         break;
7982                 }
7983         }
7984
7985         if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
7986             (pf->num_fdsb_msix == 0)) {
7987                 dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
7988                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7989         }
7990         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7991             (pf->num_vmdq_msix == 0)) {
7992                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7993                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7994         }
7995
7996         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7997             (pf->num_iwarp_msix == 0)) {
7998                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7999                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
8000         }
8001         i40e_debug(&pf->hw, I40E_DEBUG_INIT,
8002                    "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
8003                    pf->num_lan_msix,
8004                    pf->num_vmdq_msix * pf->num_vmdq_vsis,
8005                    pf->num_fdsb_msix,
8006                    pf->num_iwarp_msix);
8007
8008         return v_actual;
8009 }
8010
8011 /**
8012  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
8013  * @vsi: the VSI being configured
8014  * @v_idx: index of the vector in the vsi struct
8015  * @cpu: cpu to be used on affinity_mask
8016  *
8017  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
8018  **/
8019 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
8020 {
8021         struct i40e_q_vector *q_vector;
8022
8023         /* allocate q_vector */
8024         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
8025         if (!q_vector)
8026                 return -ENOMEM;
8027
8028         q_vector->vsi = vsi;
8029         q_vector->v_idx = v_idx;
8030         cpumask_set_cpu(cpu, &q_vector->affinity_mask);
8031
8032         if (vsi->netdev)
8033                 netif_napi_add(vsi->netdev, &q_vector->napi,
8034                                i40e_napi_poll, NAPI_POLL_WEIGHT);
8035
8036         q_vector->rx.latency_range = I40E_LOW_LATENCY;
8037         q_vector->tx.latency_range = I40E_LOW_LATENCY;
8038
8039         /* tie q_vector and vsi together */
8040         vsi->q_vectors[v_idx] = q_vector;
8041
8042         return 0;
8043 }
8044
8045 /**
8046  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
8047  * @vsi: the VSI being configured
8048  *
8049  * We allocate one q_vector per queue interrupt.  If allocation fails we
8050  * return -ENOMEM.
8051  **/
8052 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
8053 {
8054         struct i40e_pf *pf = vsi->back;
8055         int err, v_idx, num_q_vectors, current_cpu;
8056
8057         /* if not MSIX, give the one vector only to the LAN VSI */
8058         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
8059                 num_q_vectors = vsi->num_q_vectors;
8060         else if (vsi == pf->vsi[pf->lan_vsi])
8061                 num_q_vectors = 1;
8062         else
8063                 return -EINVAL;
8064
8065         current_cpu = cpumask_first(cpu_online_mask);
8066
8067         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
8068                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
8069                 if (err)
8070                         goto err_out;
8071                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
8072                 if (unlikely(current_cpu >= nr_cpu_ids))
8073                         current_cpu = cpumask_first(cpu_online_mask);
8074         }
8075
8076         return 0;
8077
8078 err_out:
8079         while (v_idx--)
8080                 i40e_free_q_vector(vsi, v_idx);
8081
8082         return err;
8083 }
8084
8085 /**
8086  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
8087  * @pf: board private structure to initialize
8088  **/
8089 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
8090 {
8091         int vectors = 0;
8092         ssize_t size;
8093
8094         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8095                 vectors = i40e_init_msix(pf);
8096                 if (vectors < 0) {
8097                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
8098                                        I40E_FLAG_IWARP_ENABLED  |
8099                                        I40E_FLAG_RSS_ENABLED    |
8100                                        I40E_FLAG_DCB_CAPABLE    |
8101                                        I40E_FLAG_DCB_ENABLED    |
8102                                        I40E_FLAG_SRIOV_ENABLED  |
8103                                        I40E_FLAG_FD_SB_ENABLED  |
8104                                        I40E_FLAG_FD_ATR_ENABLED |
8105                                        I40E_FLAG_VMDQ_ENABLED);
8106
8107                         /* rework the queue expectations without MSIX */
8108                         i40e_determine_queue_usage(pf);
8109                 }
8110         }
8111
8112         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
8113             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
8114                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
8115                 vectors = pci_enable_msi(pf->pdev);
8116                 if (vectors < 0) {
8117                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
8118                                  vectors);
8119                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
8120                 }
8121                 vectors = 1;  /* one MSI or Legacy vector */
8122         }
8123
8124         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
8125                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
8126
8127         /* set up vector assignment tracking */
8128         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
8129         pf->irq_pile = kzalloc(size, GFP_KERNEL);
8130         if (!pf->irq_pile) {
8131                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
8132                 return -ENOMEM;
8133         }
8134         pf->irq_pile->num_entries = vectors;
8135         pf->irq_pile->search_hint = 0;
8136
8137         /* track first vector for misc interrupts, ignore return */
8138         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
8139
8140         return 0;
8141 }
8142
8143 /**
8144  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
8145  * @pf: board private structure
8146  *
8147  * This sets up the handler for MSIX 0, which is used to manage the
8148  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
8149  * when in MSI or Legacy interrupt mode.
8150  **/
8151 static int i40e_setup_misc_vector(struct i40e_pf *pf)
8152 {
8153         struct i40e_hw *hw = &pf->hw;
8154         int err = 0;
8155
8156         /* Only request the irq if this is the first time through, and
8157          * not when we're rebuilding after a Reset
8158          */
8159         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
8160                 err = request_irq(pf->msix_entries[0].vector,
8161                                   i40e_intr, 0, pf->int_name, pf);
8162                 if (err) {
8163                         dev_info(&pf->pdev->dev,
8164                                  "request_irq for %s failed: %d\n",
8165                                  pf->int_name, err);
8166                         return -EFAULT;
8167                 }
8168         }
8169
8170         i40e_enable_misc_int_causes(pf);
8171
8172         /* associate no queues to the misc vector */
8173         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
8174         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
8175
8176         i40e_flush(hw);
8177
8178         i40e_irq_dynamic_enable_icr0(pf, true);
8179
8180         return err;
8181 }
8182
8183 /**
8184  * i40e_config_rss_aq - Prepare for RSS using AQ commands
8185  * @vsi: vsi structure
8186  * @seed: RSS hash seed
8187  **/
8188 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8189                               u8 *lut, u16 lut_size)
8190 {
8191         struct i40e_pf *pf = vsi->back;
8192         struct i40e_hw *hw = &pf->hw;
8193         int ret = 0;
8194
8195         if (seed) {
8196                 struct i40e_aqc_get_set_rss_key_data *seed_dw =
8197                         (struct i40e_aqc_get_set_rss_key_data *)seed;
8198                 ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8199                 if (ret) {
8200                         dev_info(&pf->pdev->dev,
8201                                  "Cannot set RSS key, err %s aq_err %s\n",
8202                                  i40e_stat_str(hw, ret),
8203                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8204                         return ret;
8205                 }
8206         }
8207         if (lut) {
8208                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8209
8210                 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8211                 if (ret) {
8212                         dev_info(&pf->pdev->dev,
8213                                  "Cannot set RSS lut, err %s aq_err %s\n",
8214                                  i40e_stat_str(hw, ret),
8215                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8216                         return ret;
8217                 }
8218         }
8219         return ret;
8220 }
8221
8222 /**
8223  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8224  * @vsi: Pointer to vsi structure
8225  * @seed: Buffter to store the hash keys
8226  * @lut: Buffer to store the lookup table entries
8227  * @lut_size: Size of buffer to store the lookup table entries
8228  *
8229  * Return 0 on success, negative on failure
8230  */
8231 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8232                            u8 *lut, u16 lut_size)
8233 {
8234         struct i40e_pf *pf = vsi->back;
8235         struct i40e_hw *hw = &pf->hw;
8236         int ret = 0;
8237
8238         if (seed) {
8239                 ret = i40e_aq_get_rss_key(hw, vsi->id,
8240                         (struct i40e_aqc_get_set_rss_key_data *)seed);
8241                 if (ret) {
8242                         dev_info(&pf->pdev->dev,
8243                                  "Cannot get RSS key, err %s aq_err %s\n",
8244                                  i40e_stat_str(&pf->hw, ret),
8245                                  i40e_aq_str(&pf->hw,
8246                                              pf->hw.aq.asq_last_status));
8247                         return ret;
8248                 }
8249         }
8250
8251         if (lut) {
8252                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8253
8254                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8255                 if (ret) {
8256                         dev_info(&pf->pdev->dev,
8257                                  "Cannot get RSS lut, err %s aq_err %s\n",
8258                                  i40e_stat_str(&pf->hw, ret),
8259                                  i40e_aq_str(&pf->hw,
8260                                              pf->hw.aq.asq_last_status));
8261                         return ret;
8262                 }
8263         }
8264
8265         return ret;
8266 }
8267
8268 /**
8269  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8270  * @vsi: VSI structure
8271  **/
8272 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8273 {
8274         u8 seed[I40E_HKEY_ARRAY_SIZE];
8275         struct i40e_pf *pf = vsi->back;
8276         u8 *lut;
8277         int ret;
8278
8279         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8280                 return 0;
8281
8282         if (!vsi->rss_size)
8283                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8284                                       vsi->num_queue_pairs);
8285         if (!vsi->rss_size)
8286                 return -EINVAL;
8287
8288         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8289         if (!lut)
8290                 return -ENOMEM;
8291         /* Use the user configured hash keys and lookup table if there is one,
8292          * otherwise use default
8293          */
8294         if (vsi->rss_lut_user)
8295                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8296         else
8297                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8298         if (vsi->rss_hkey_user)
8299                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8300         else
8301                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8302         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8303         kfree(lut);
8304
8305         return ret;
8306 }
8307
8308 /**
8309  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8310  * @vsi: Pointer to vsi structure
8311  * @seed: RSS hash seed
8312  * @lut: Lookup table
8313  * @lut_size: Lookup table size
8314  *
8315  * Returns 0 on success, negative on failure
8316  **/
8317 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8318                                const u8 *lut, u16 lut_size)
8319 {
8320         struct i40e_pf *pf = vsi->back;
8321         struct i40e_hw *hw = &pf->hw;
8322         u16 vf_id = vsi->vf_id;
8323         u8 i;
8324
8325         /* Fill out hash function seed */
8326         if (seed) {
8327                 u32 *seed_dw = (u32 *)seed;
8328
8329                 if (vsi->type == I40E_VSI_MAIN) {
8330                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8331                                 wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
8332                 } else if (vsi->type == I40E_VSI_SRIOV) {
8333                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8334                                 wr32(hw, I40E_VFQF_HKEY1(i, vf_id), seed_dw[i]);
8335                 } else {
8336                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8337                 }
8338         }
8339
8340         if (lut) {
8341                 u32 *lut_dw = (u32 *)lut;
8342
8343                 if (vsi->type == I40E_VSI_MAIN) {
8344                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
8345                                 return -EINVAL;
8346                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8347                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8348                 } else if (vsi->type == I40E_VSI_SRIOV) {
8349                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8350                                 return -EINVAL;
8351                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8352                                 wr32(hw, I40E_VFQF_HLUT1(i, vf_id), lut_dw[i]);
8353                 } else {
8354                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8355                 }
8356         }
8357         i40e_flush(hw);
8358
8359         return 0;
8360 }
8361
8362 /**
8363  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8364  * @vsi: Pointer to VSI structure
8365  * @seed: Buffer to store the keys
8366  * @lut: Buffer to store the lookup table entries
8367  * @lut_size: Size of buffer to store the lookup table entries
8368  *
8369  * Returns 0 on success, negative on failure
8370  */
8371 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8372                             u8 *lut, u16 lut_size)
8373 {
8374         struct i40e_pf *pf = vsi->back;
8375         struct i40e_hw *hw = &pf->hw;
8376         u16 i;
8377
8378         if (seed) {
8379                 u32 *seed_dw = (u32 *)seed;
8380
8381                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8382                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8383         }
8384         if (lut) {
8385                 u32 *lut_dw = (u32 *)lut;
8386
8387                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8388                         return -EINVAL;
8389                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8390                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8391         }
8392
8393         return 0;
8394 }
8395
8396 /**
8397  * i40e_config_rss - Configure RSS keys and lut
8398  * @vsi: Pointer to VSI structure
8399  * @seed: RSS hash seed
8400  * @lut: Lookup table
8401  * @lut_size: Lookup table size
8402  *
8403  * Returns 0 on success, negative on failure
8404  */
8405 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8406 {
8407         struct i40e_pf *pf = vsi->back;
8408
8409         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8410                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8411         else
8412                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8413 }
8414
8415 /**
8416  * i40e_get_rss - Get RSS keys and lut
8417  * @vsi: Pointer to VSI structure
8418  * @seed: Buffer to store the keys
8419  * @lut: Buffer to store the lookup table entries
8420  * lut_size: Size of buffer to store the lookup table entries
8421  *
8422  * Returns 0 on success, negative on failure
8423  */
8424 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8425 {
8426         struct i40e_pf *pf = vsi->back;
8427
8428         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8429                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8430         else
8431                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8432 }
8433
8434 /**
8435  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8436  * @pf: Pointer to board private structure
8437  * @lut: Lookup table
8438  * @rss_table_size: Lookup table size
8439  * @rss_size: Range of queue number for hashing
8440  */
8441 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8442                        u16 rss_table_size, u16 rss_size)
8443 {
8444         u16 i;
8445
8446         for (i = 0; i < rss_table_size; i++)
8447                 lut[i] = i % rss_size;
8448 }
8449
8450 /**
8451  * i40e_pf_config_rss - Prepare for RSS if used
8452  * @pf: board private structure
8453  **/
8454 static int i40e_pf_config_rss(struct i40e_pf *pf)
8455 {
8456         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8457         u8 seed[I40E_HKEY_ARRAY_SIZE];
8458         u8 *lut;
8459         struct i40e_hw *hw = &pf->hw;
8460         u32 reg_val;
8461         u64 hena;
8462         int ret;
8463
8464         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8465         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8466                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8467         hena |= i40e_pf_get_default_rss_hena(pf);
8468
8469         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8470         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8471
8472         /* Determine the RSS table size based on the hardware capabilities */
8473         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8474         reg_val = (pf->rss_table_size == 512) ?
8475                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8476                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8477         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8478
8479         /* Determine the RSS size of the VSI */
8480         if (!vsi->rss_size) {
8481                 u16 qcount;
8482
8483                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8484                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8485         }
8486         if (!vsi->rss_size)
8487                 return -EINVAL;
8488
8489         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8490         if (!lut)
8491                 return -ENOMEM;
8492
8493         /* Use user configured lut if there is one, otherwise use default */
8494         if (vsi->rss_lut_user)
8495                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8496         else
8497                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8498
8499         /* Use user configured hash key if there is one, otherwise
8500          * use default.
8501          */
8502         if (vsi->rss_hkey_user)
8503                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8504         else
8505                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8506         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8507         kfree(lut);
8508
8509         return ret;
8510 }
8511
8512 /**
8513  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8514  * @pf: board private structure
8515  * @queue_count: the requested queue count for rss.
8516  *
8517  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8518  * count which may be different from the requested queue count.
8519  * Note: expects to be called while under rtnl_lock()
8520  **/
8521 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8522 {
8523         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8524         int new_rss_size;
8525
8526         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8527                 return 0;
8528
8529         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8530
8531         if (queue_count != vsi->num_queue_pairs) {
8532                 u16 qcount;
8533
8534                 vsi->req_queue_pairs = queue_count;
8535                 i40e_prep_for_reset(pf, true);
8536
8537                 pf->alloc_rss_size = new_rss_size;
8538
8539                 i40e_reset_and_rebuild(pf, true, true);
8540
8541                 /* Discard the user configured hash keys and lut, if less
8542                  * queues are enabled.
8543                  */
8544                 if (queue_count < vsi->rss_size) {
8545                         i40e_clear_rss_config_user(vsi);
8546                         dev_dbg(&pf->pdev->dev,
8547                                 "discard user configured hash keys and lut\n");
8548                 }
8549
8550                 /* Reset vsi->rss_size, as number of enabled queues changed */
8551                 qcount = vsi->num_queue_pairs / vsi->tc_config.numtc;
8552                 vsi->rss_size = min_t(int, pf->alloc_rss_size, qcount);
8553
8554                 i40e_pf_config_rss(pf);
8555         }
8556         dev_info(&pf->pdev->dev, "User requested queue count/HW max RSS count:  %d/%d\n",
8557                  vsi->req_queue_pairs, pf->rss_size_max);
8558         return pf->alloc_rss_size;
8559 }
8560
8561 /**
8562  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8563  * @pf: board private structure
8564  **/
8565 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8566 {
8567         i40e_status status;
8568         bool min_valid, max_valid;
8569         u32 max_bw, min_bw;
8570
8571         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8572                                            &min_valid, &max_valid);
8573
8574         if (!status) {
8575                 if (min_valid)
8576                         pf->npar_min_bw = min_bw;
8577                 if (max_valid)
8578                         pf->npar_max_bw = max_bw;
8579         }
8580
8581         return status;
8582 }
8583
8584 /**
8585  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8586  * @pf: board private structure
8587  **/
8588 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8589 {
8590         struct i40e_aqc_configure_partition_bw_data bw_data;
8591         i40e_status status;
8592
8593         /* Set the valid bit for this PF */
8594         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8595         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8596         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8597
8598         /* Set the new bandwidths */
8599         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8600
8601         return status;
8602 }
8603
8604 /**
8605  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8606  * @pf: board private structure
8607  **/
8608 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8609 {
8610         /* Commit temporary BW setting to permanent NVM image */
8611         enum i40e_admin_queue_err last_aq_status;
8612         i40e_status ret;
8613         u16 nvm_word;
8614
8615         if (pf->hw.partition_id != 1) {
8616                 dev_info(&pf->pdev->dev,
8617                          "Commit BW only works on partition 1! This is partition %d",
8618                          pf->hw.partition_id);
8619                 ret = I40E_NOT_SUPPORTED;
8620                 goto bw_commit_out;
8621         }
8622
8623         /* Acquire NVM for read access */
8624         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8625         last_aq_status = pf->hw.aq.asq_last_status;
8626         if (ret) {
8627                 dev_info(&pf->pdev->dev,
8628                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8629                          i40e_stat_str(&pf->hw, ret),
8630                          i40e_aq_str(&pf->hw, last_aq_status));
8631                 goto bw_commit_out;
8632         }
8633
8634         /* Read word 0x10 of NVM - SW compatibility word 1 */
8635         ret = i40e_aq_read_nvm(&pf->hw,
8636                                I40E_SR_NVM_CONTROL_WORD,
8637                                0x10, sizeof(nvm_word), &nvm_word,
8638                                false, NULL);
8639         /* Save off last admin queue command status before releasing
8640          * the NVM
8641          */
8642         last_aq_status = pf->hw.aq.asq_last_status;
8643         i40e_release_nvm(&pf->hw);
8644         if (ret) {
8645                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8646                          i40e_stat_str(&pf->hw, ret),
8647                          i40e_aq_str(&pf->hw, last_aq_status));
8648                 goto bw_commit_out;
8649         }
8650
8651         /* Wait a bit for NVM release to complete */
8652         msleep(50);
8653
8654         /* Acquire NVM for write access */
8655         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8656         last_aq_status = pf->hw.aq.asq_last_status;
8657         if (ret) {
8658                 dev_info(&pf->pdev->dev,
8659                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8660                          i40e_stat_str(&pf->hw, ret),
8661                          i40e_aq_str(&pf->hw, last_aq_status));
8662                 goto bw_commit_out;
8663         }
8664         /* Write it back out unchanged to initiate update NVM,
8665          * which will force a write of the shadow (alt) RAM to
8666          * the NVM - thus storing the bandwidth values permanently.
8667          */
8668         ret = i40e_aq_update_nvm(&pf->hw,
8669                                  I40E_SR_NVM_CONTROL_WORD,
8670                                  0x10, sizeof(nvm_word),
8671                                  &nvm_word, true, NULL);
8672         /* Save off last admin queue command status before releasing
8673          * the NVM
8674          */
8675         last_aq_status = pf->hw.aq.asq_last_status;
8676         i40e_release_nvm(&pf->hw);
8677         if (ret)
8678                 dev_info(&pf->pdev->dev,
8679                          "BW settings NOT SAVED, err %s aq_err %s\n",
8680                          i40e_stat_str(&pf->hw, ret),
8681                          i40e_aq_str(&pf->hw, last_aq_status));
8682 bw_commit_out:
8683
8684         return ret;
8685 }
8686
8687 /**
8688  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8689  * @pf: board private structure to initialize
8690  *
8691  * i40e_sw_init initializes the Adapter private data structure.
8692  * Fields are initialized based on PCI device information and
8693  * OS network device settings (MTU size).
8694  **/
8695 static int i40e_sw_init(struct i40e_pf *pf)
8696 {
8697         int err = 0;
8698         int size;
8699
8700         /* Set default capability flags */
8701         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8702                     I40E_FLAG_MSI_ENABLED     |
8703                     I40E_FLAG_MSIX_ENABLED;
8704
8705         /* Set default ITR */
8706         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8707         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8708
8709         /* Depending on PF configurations, it is possible that the RSS
8710          * maximum might end up larger than the available queues
8711          */
8712         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8713         pf->alloc_rss_size = 1;
8714         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8715         pf->rss_size_max = min_t(int, pf->rss_size_max,
8716                                  pf->hw.func_caps.num_tx_qp);
8717         if (pf->hw.func_caps.rss) {
8718                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8719                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8720                                            num_online_cpus());
8721         }
8722
8723         /* MFP mode enabled */
8724         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8725                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8726                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8727                 if (i40e_get_npar_bw_setting(pf))
8728                         dev_warn(&pf->pdev->dev,
8729                                  "Could not get NPAR bw settings\n");
8730                 else
8731                         dev_info(&pf->pdev->dev,
8732                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8733                                  pf->npar_min_bw, pf->npar_max_bw);
8734         }
8735
8736         /* FW/NVM is not yet fixed in this regard */
8737         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8738             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8739                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8740                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8741                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8742                     pf->hw.num_partitions > 1)
8743                         dev_info(&pf->pdev->dev,
8744                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8745                 else
8746                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8747                 pf->fdir_pf_filter_count =
8748                                  pf->hw.func_caps.fd_filters_guaranteed;
8749                 pf->hw.fdir_shared_filter_count =
8750                                  pf->hw.func_caps.fd_filters_best_effort;
8751         }
8752
8753         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8754             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8755             (pf->hw.aq.fw_maj_ver < 4))) {
8756                 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8757                 /* No DCB support  for FW < v4.33 */
8758                 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8759         }
8760
8761         /* Disable FW LLDP if FW < v4.3 */
8762         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8763             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8764             (pf->hw.aq.fw_maj_ver < 4)))
8765                 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8766
8767         /* Use the FW Set LLDP MIB API if FW > v4.40 */
8768         if ((pf->hw.mac.type == I40E_MAC_XL710) &&
8769             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8770             (pf->hw.aq.fw_maj_ver >= 5)))
8771                 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8772
8773         if (pf->hw.func_caps.vmdq) {
8774                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8775                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8776                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8777         }
8778
8779         if (pf->hw.func_caps.iwarp) {
8780                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8781                 /* IWARP needs one extra vector for CQP just like MISC.*/
8782                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8783         }
8784
8785 #ifdef CONFIG_PCI_IOV
8786         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8787                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8788                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8789                 pf->num_req_vfs = min_t(int,
8790                                         pf->hw.func_caps.num_vfs,
8791                                         I40E_MAX_VF_COUNT);
8792         }
8793 #endif /* CONFIG_PCI_IOV */
8794         if (pf->hw.mac.type == I40E_MAC_X722) {
8795                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE
8796                              | I40E_FLAG_128_QP_RSS_CAPABLE
8797                              | I40E_FLAG_HW_ATR_EVICT_CAPABLE
8798                              | I40E_FLAG_OUTER_UDP_CSUM_CAPABLE
8799                              | I40E_FLAG_WB_ON_ITR_CAPABLE
8800                              | I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE
8801                              | I40E_FLAG_NO_PCI_LINK_CHECK
8802                              | I40E_FLAG_USE_SET_LLDP_MIB
8803                              | I40E_FLAG_GENEVE_OFFLOAD_CAPABLE
8804                              | I40E_FLAG_PTP_L4_CAPABLE
8805                              | I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE;
8806         } else if ((pf->hw.aq.api_maj_ver > 1) ||
8807                    ((pf->hw.aq.api_maj_ver == 1) &&
8808                     (pf->hw.aq.api_min_ver > 4))) {
8809                 /* Supported in FW API version higher than 1.4 */
8810                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8811                 pf->hw_disabled_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8812         } else {
8813                 pf->hw_disabled_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8814         }
8815
8816         pf->eeprom_version = 0xDEAD;
8817         pf->lan_veb = I40E_NO_VEB;
8818         pf->lan_vsi = I40E_NO_VSI;
8819
8820         /* By default FW has this off for performance reasons */
8821         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8822
8823         /* set up queue assignment tracking */
8824         size = sizeof(struct i40e_lump_tracking)
8825                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8826         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8827         if (!pf->qp_pile) {
8828                 err = -ENOMEM;
8829                 goto sw_init_done;
8830         }
8831         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8832         pf->qp_pile->search_hint = 0;
8833
8834         pf->tx_timeout_recovery_level = 1;
8835
8836         mutex_init(&pf->switch_mutex);
8837
8838         /* If NPAR is enabled nudge the Tx scheduler */
8839         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8840                 i40e_set_npar_bw_setting(pf);
8841
8842 sw_init_done:
8843         return err;
8844 }
8845
8846 /**
8847  * i40e_set_ntuple - set the ntuple feature flag and take action
8848  * @pf: board private structure to initialize
8849  * @features: the feature set that the stack is suggesting
8850  *
8851  * returns a bool to indicate if reset needs to happen
8852  **/
8853 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8854 {
8855         bool need_reset = false;
8856
8857         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8858          * the state changed, we need to reset.
8859          */
8860         if (features & NETIF_F_NTUPLE) {
8861                 /* Enable filters and mark for reset */
8862                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8863                         need_reset = true;
8864                 /* enable FD_SB only if there is MSI-X vector */
8865                 if (pf->num_fdsb_msix > 0)
8866                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8867         } else {
8868                 /* turn off filters, mark for reset and clear SW filter list */
8869                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8870                         need_reset = true;
8871                         i40e_fdir_filter_exit(pf);
8872                 }
8873                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8874                 pf->hw_disabled_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8875                 /* reset fd counters */
8876                 pf->fd_add_err = 0;
8877                 pf->fd_atr_cnt = 0;
8878                 /* if ATR was auto disabled it can be re-enabled. */
8879                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8880                     (pf->hw_disabled_flags & I40E_FLAG_FD_ATR_ENABLED)) {
8881                         pf->hw_disabled_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8882                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
8883                                 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8884                 }
8885         }
8886         return need_reset;
8887 }
8888
8889 /**
8890  * i40e_clear_rss_lut - clear the rx hash lookup table
8891  * @vsi: the VSI being configured
8892  **/
8893 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
8894 {
8895         struct i40e_pf *pf = vsi->back;
8896         struct i40e_hw *hw = &pf->hw;
8897         u16 vf_id = vsi->vf_id;
8898         u8 i;
8899
8900         if (vsi->type == I40E_VSI_MAIN) {
8901                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8902                         wr32(hw, I40E_PFQF_HLUT(i), 0);
8903         } else if (vsi->type == I40E_VSI_SRIOV) {
8904                 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8905                         i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
8906         } else {
8907                 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8908         }
8909 }
8910
8911 /**
8912  * i40e_set_features - set the netdev feature flags
8913  * @netdev: ptr to the netdev being adjusted
8914  * @features: the feature set that the stack is suggesting
8915  * Note: expects to be called while under rtnl_lock()
8916  **/
8917 static int i40e_set_features(struct net_device *netdev,
8918                              netdev_features_t features)
8919 {
8920         struct i40e_netdev_priv *np = netdev_priv(netdev);
8921         struct i40e_vsi *vsi = np->vsi;
8922         struct i40e_pf *pf = vsi->back;
8923         bool need_reset;
8924
8925         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
8926                 i40e_pf_config_rss(pf);
8927         else if (!(features & NETIF_F_RXHASH) &&
8928                  netdev->features & NETIF_F_RXHASH)
8929                 i40e_clear_rss_lut(vsi);
8930
8931         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8932                 i40e_vlan_stripping_enable(vsi);
8933         else
8934                 i40e_vlan_stripping_disable(vsi);
8935
8936         need_reset = i40e_set_ntuple(pf, features);
8937
8938         if (need_reset)
8939                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED), true);
8940
8941         return 0;
8942 }
8943
8944 /**
8945  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8946  * @pf: board private structure
8947  * @port: The UDP port to look up
8948  *
8949  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8950  **/
8951 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, u16 port)
8952 {
8953         u8 i;
8954
8955         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8956                 if (pf->udp_ports[i].index == port)
8957                         return i;
8958         }
8959
8960         return i;
8961 }
8962
8963 /**
8964  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8965  * @netdev: This physical port's netdev
8966  * @ti: Tunnel endpoint information
8967  **/
8968 static void i40e_udp_tunnel_add(struct net_device *netdev,
8969                                 struct udp_tunnel_info *ti)
8970 {
8971         struct i40e_netdev_priv *np = netdev_priv(netdev);
8972         struct i40e_vsi *vsi = np->vsi;
8973         struct i40e_pf *pf = vsi->back;
8974         u16 port = ntohs(ti->port);
8975         u8 next_idx;
8976         u8 idx;
8977
8978         idx = i40e_get_udp_port_idx(pf, port);
8979
8980         /* Check if port already exists */
8981         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8982                 netdev_info(netdev, "port %d already offloaded\n", port);
8983                 return;
8984         }
8985
8986         /* Now check if there is space to add the new port */
8987         next_idx = i40e_get_udp_port_idx(pf, 0);
8988
8989         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8990                 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8991                             port);
8992                 return;
8993         }
8994
8995         switch (ti->type) {
8996         case UDP_TUNNEL_TYPE_VXLAN:
8997                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8998                 break;
8999         case UDP_TUNNEL_TYPE_GENEVE:
9000                 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
9001                         return;
9002                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
9003                 break;
9004         default:
9005                 return;
9006         }
9007
9008         /* New port: add it and mark its index in the bitmap */
9009         pf->udp_ports[next_idx].index = port;
9010         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
9011         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9012 }
9013
9014 /**
9015  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
9016  * @netdev: This physical port's netdev
9017  * @ti: Tunnel endpoint information
9018  **/
9019 static void i40e_udp_tunnel_del(struct net_device *netdev,
9020                                 struct udp_tunnel_info *ti)
9021 {
9022         struct i40e_netdev_priv *np = netdev_priv(netdev);
9023         struct i40e_vsi *vsi = np->vsi;
9024         struct i40e_pf *pf = vsi->back;
9025         u16 port = ntohs(ti->port);
9026         u8 idx;
9027
9028         idx = i40e_get_udp_port_idx(pf, port);
9029
9030         /* Check if port already exists */
9031         if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
9032                 goto not_found;
9033
9034         switch (ti->type) {
9035         case UDP_TUNNEL_TYPE_VXLAN:
9036                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
9037                         goto not_found;
9038                 break;
9039         case UDP_TUNNEL_TYPE_GENEVE:
9040                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
9041                         goto not_found;
9042                 break;
9043         default:
9044                 goto not_found;
9045         }
9046
9047         /* if port exists, set it to 0 (mark for deletion)
9048          * and make it pending
9049          */
9050         pf->udp_ports[idx].index = 0;
9051         pf->pending_udp_bitmap |= BIT_ULL(idx);
9052         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
9053
9054         return;
9055 not_found:
9056         netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
9057                     port);
9058 }
9059
9060 static int i40e_get_phys_port_id(struct net_device *netdev,
9061                                  struct netdev_phys_item_id *ppid)
9062 {
9063         struct i40e_netdev_priv *np = netdev_priv(netdev);
9064         struct i40e_pf *pf = np->vsi->back;
9065         struct i40e_hw *hw = &pf->hw;
9066
9067         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
9068                 return -EOPNOTSUPP;
9069
9070         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
9071         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
9072
9073         return 0;
9074 }
9075
9076 /**
9077  * i40e_ndo_fdb_add - add an entry to the hardware database
9078  * @ndm: the input from the stack
9079  * @tb: pointer to array of nladdr (unused)
9080  * @dev: the net device pointer
9081  * @addr: the MAC address entry being added
9082  * @flags: instructions from stack about fdb operation
9083  */
9084 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
9085                             struct net_device *dev,
9086                             const unsigned char *addr, u16 vid,
9087                             u16 flags)
9088 {
9089         struct i40e_netdev_priv *np = netdev_priv(dev);
9090         struct i40e_pf *pf = np->vsi->back;
9091         int err = 0;
9092
9093         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
9094                 return -EOPNOTSUPP;
9095
9096         if (vid) {
9097                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
9098                 return -EINVAL;
9099         }
9100
9101         /* Hardware does not support aging addresses so if a
9102          * ndm_state is given only allow permanent addresses
9103          */
9104         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
9105                 netdev_info(dev, "FDB only supports static addresses\n");
9106                 return -EINVAL;
9107         }
9108
9109         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
9110                 err = dev_uc_add_excl(dev, addr);
9111         else if (is_multicast_ether_addr(addr))
9112                 err = dev_mc_add_excl(dev, addr);
9113         else
9114                 err = -EINVAL;
9115
9116         /* Only return duplicate errors if NLM_F_EXCL is set */
9117         if (err == -EEXIST && !(flags & NLM_F_EXCL))
9118                 err = 0;
9119
9120         return err;
9121 }
9122
9123 /**
9124  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
9125  * @dev: the netdev being configured
9126  * @nlh: RTNL message
9127  *
9128  * Inserts a new hardware bridge if not already created and
9129  * enables the bridging mode requested (VEB or VEPA). If the
9130  * hardware bridge has already been inserted and the request
9131  * is to change the mode then that requires a PF reset to
9132  * allow rebuild of the components with required hardware
9133  * bridge mode enabled.
9134  *
9135  * Note: expects to be called while under rtnl_lock()
9136  **/
9137 static int i40e_ndo_bridge_setlink(struct net_device *dev,
9138                                    struct nlmsghdr *nlh,
9139                                    u16 flags)
9140 {
9141         struct i40e_netdev_priv *np = netdev_priv(dev);
9142         struct i40e_vsi *vsi = np->vsi;
9143         struct i40e_pf *pf = vsi->back;
9144         struct i40e_veb *veb = NULL;
9145         struct nlattr *attr, *br_spec;
9146         int i, rem;
9147
9148         /* Only for PF VSI for now */
9149         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9150                 return -EOPNOTSUPP;
9151
9152         /* Find the HW bridge for PF VSI */
9153         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9154                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9155                         veb = pf->veb[i];
9156         }
9157
9158         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
9159
9160         nla_for_each_nested(attr, br_spec, rem) {
9161                 __u16 mode;
9162
9163                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
9164                         continue;
9165
9166                 mode = nla_get_u16(attr);
9167                 if ((mode != BRIDGE_MODE_VEPA) &&
9168                     (mode != BRIDGE_MODE_VEB))
9169                         return -EINVAL;
9170
9171                 /* Insert a new HW bridge */
9172                 if (!veb) {
9173                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9174                                              vsi->tc_config.enabled_tc);
9175                         if (veb) {
9176                                 veb->bridge_mode = mode;
9177                                 i40e_config_bridge_mode(veb);
9178                         } else {
9179                                 /* No Bridge HW offload available */
9180                                 return -ENOENT;
9181                         }
9182                         break;
9183                 } else if (mode != veb->bridge_mode) {
9184                         /* Existing HW bridge but different mode needs reset */
9185                         veb->bridge_mode = mode;
9186                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
9187                         if (mode == BRIDGE_MODE_VEB)
9188                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9189                         else
9190                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9191                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED),
9192                                       true);
9193                         break;
9194                 }
9195         }
9196
9197         return 0;
9198 }
9199
9200 /**
9201  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9202  * @skb: skb buff
9203  * @pid: process id
9204  * @seq: RTNL message seq #
9205  * @dev: the netdev being configured
9206  * @filter_mask: unused
9207  * @nlflags: netlink flags passed in
9208  *
9209  * Return the mode in which the hardware bridge is operating in
9210  * i.e VEB or VEPA.
9211  **/
9212 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9213                                    struct net_device *dev,
9214                                    u32 __always_unused filter_mask,
9215                                    int nlflags)
9216 {
9217         struct i40e_netdev_priv *np = netdev_priv(dev);
9218         struct i40e_vsi *vsi = np->vsi;
9219         struct i40e_pf *pf = vsi->back;
9220         struct i40e_veb *veb = NULL;
9221         int i;
9222
9223         /* Only for PF VSI for now */
9224         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9225                 return -EOPNOTSUPP;
9226
9227         /* Find the HW bridge for the PF VSI */
9228         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9229                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9230                         veb = pf->veb[i];
9231         }
9232
9233         if (!veb)
9234                 return 0;
9235
9236         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9237                                        0, 0, nlflags, filter_mask, NULL);
9238 }
9239
9240 /**
9241  * i40e_features_check - Validate encapsulated packet conforms to limits
9242  * @skb: skb buff
9243  * @dev: This physical port's netdev
9244  * @features: Offload features that the stack believes apply
9245  **/
9246 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9247                                              struct net_device *dev,
9248                                              netdev_features_t features)
9249 {
9250         size_t len;
9251
9252         /* No point in doing any of this if neither checksum nor GSO are
9253          * being requested for this frame.  We can rule out both by just
9254          * checking for CHECKSUM_PARTIAL
9255          */
9256         if (skb->ip_summed != CHECKSUM_PARTIAL)
9257                 return features;
9258
9259         /* We cannot support GSO if the MSS is going to be less than
9260          * 64 bytes.  If it is then we need to drop support for GSO.
9261          */
9262         if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
9263                 features &= ~NETIF_F_GSO_MASK;
9264
9265         /* MACLEN can support at most 63 words */
9266         len = skb_network_header(skb) - skb->data;
9267         if (len & ~(63 * 2))
9268                 goto out_err;
9269
9270         /* IPLEN and EIPLEN can support at most 127 dwords */
9271         len = skb_transport_header(skb) - skb_network_header(skb);
9272         if (len & ~(127 * 4))
9273                 goto out_err;
9274
9275         if (skb->encapsulation) {
9276                 /* L4TUNLEN can support 127 words */
9277                 len = skb_inner_network_header(skb) - skb_transport_header(skb);
9278                 if (len & ~(127 * 2))
9279                         goto out_err;
9280
9281                 /* IPLEN can support at most 127 dwords */
9282                 len = skb_inner_transport_header(skb) -
9283                       skb_inner_network_header(skb);
9284                 if (len & ~(127 * 4))
9285                         goto out_err;
9286         }
9287
9288         /* No need to validate L4LEN as TCP is the only protocol with a
9289          * a flexible value and we support all possible values supported
9290          * by TCP, which is at most 15 dwords
9291          */
9292
9293         return features;
9294 out_err:
9295         return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9296 }
9297
9298 static const struct net_device_ops i40e_netdev_ops = {
9299         .ndo_open               = i40e_open,
9300         .ndo_stop               = i40e_close,
9301         .ndo_start_xmit         = i40e_lan_xmit_frame,
9302         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
9303         .ndo_set_rx_mode        = i40e_set_rx_mode,
9304         .ndo_validate_addr      = eth_validate_addr,
9305         .ndo_set_mac_address    = i40e_set_mac,
9306         .ndo_change_mtu         = i40e_change_mtu,
9307         .ndo_do_ioctl           = i40e_ioctl,
9308         .ndo_tx_timeout         = i40e_tx_timeout,
9309         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
9310         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
9311 #ifdef CONFIG_NET_POLL_CONTROLLER
9312         .ndo_poll_controller    = i40e_netpoll,
9313 #endif
9314         .ndo_setup_tc           = __i40e_setup_tc,
9315         .ndo_set_features       = i40e_set_features,
9316         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
9317         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
9318         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
9319         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
9320         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
9321         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
9322         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
9323         .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
9324         .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
9325         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
9326         .ndo_fdb_add            = i40e_ndo_fdb_add,
9327         .ndo_features_check     = i40e_features_check,
9328         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
9329         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
9330 };
9331
9332 /**
9333  * i40e_config_netdev - Setup the netdev flags
9334  * @vsi: the VSI being configured
9335  *
9336  * Returns 0 on success, negative value on failure
9337  **/
9338 static int i40e_config_netdev(struct i40e_vsi *vsi)
9339 {
9340         struct i40e_pf *pf = vsi->back;
9341         struct i40e_hw *hw = &pf->hw;
9342         struct i40e_netdev_priv *np;
9343         struct net_device *netdev;
9344         u8 broadcast[ETH_ALEN];
9345         u8 mac_addr[ETH_ALEN];
9346         int etherdev_size;
9347         netdev_features_t hw_enc_features;
9348         netdev_features_t hw_features;
9349
9350         etherdev_size = sizeof(struct i40e_netdev_priv);
9351         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9352         if (!netdev)
9353                 return -ENOMEM;
9354
9355         vsi->netdev = netdev;
9356         np = netdev_priv(netdev);
9357         np->vsi = vsi;
9358
9359         hw_enc_features = NETIF_F_SG                    |
9360                           NETIF_F_IP_CSUM               |
9361                           NETIF_F_IPV6_CSUM             |
9362                           NETIF_F_HIGHDMA               |
9363                           NETIF_F_SOFT_FEATURES         |
9364                           NETIF_F_TSO                   |
9365                           NETIF_F_TSO_ECN               |
9366                           NETIF_F_TSO6                  |
9367                           NETIF_F_GSO_GRE               |
9368                           NETIF_F_GSO_GRE_CSUM          |
9369                           NETIF_F_GSO_PARTIAL           |
9370                           NETIF_F_GSO_UDP_TUNNEL        |
9371                           NETIF_F_GSO_UDP_TUNNEL_CSUM   |
9372                           NETIF_F_SCTP_CRC              |
9373                           NETIF_F_RXHASH                |
9374                           NETIF_F_RXCSUM                |
9375                           0;
9376
9377         if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9378                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9379
9380         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9381
9382         netdev->hw_enc_features |= hw_enc_features;
9383
9384         /* record features VLANs can make use of */
9385         netdev->vlan_features |= hw_enc_features | NETIF_F_TSO_MANGLEID;
9386
9387         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9388                 netdev->hw_features |= NETIF_F_NTUPLE;
9389         hw_features = hw_enc_features           |
9390                       NETIF_F_HW_VLAN_CTAG_TX   |
9391                       NETIF_F_HW_VLAN_CTAG_RX;
9392
9393         netdev->hw_features |= hw_features;
9394
9395         netdev->features |= hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9396         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9397
9398         if (vsi->type == I40E_VSI_MAIN) {
9399                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9400                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9401                 /* The following steps are necessary for two reasons. First,
9402                  * some older NVM configurations load a default MAC-VLAN
9403                  * filter that will accept any tagged packet, and we want to
9404                  * replace this with a normal filter. Additionally, it is
9405                  * possible our MAC address was provided by the platform using
9406                  * Open Firmware or similar.
9407                  *
9408                  * Thus, we need to remove the default filter and install one
9409                  * specific to the MAC address.
9410                  */
9411                 i40e_rm_default_mac_filter(vsi, mac_addr);
9412                 spin_lock_bh(&vsi->mac_filter_hash_lock);
9413                 i40e_add_mac_filter(vsi, mac_addr);
9414                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
9415         } else {
9416                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9417                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9418                          pf->vsi[pf->lan_vsi]->netdev->name);
9419                 random_ether_addr(mac_addr);
9420
9421                 spin_lock_bh(&vsi->mac_filter_hash_lock);
9422                 i40e_add_mac_filter(vsi, mac_addr);
9423                 spin_unlock_bh(&vsi->mac_filter_hash_lock);
9424         }
9425
9426         /* Add the broadcast filter so that we initially will receive
9427          * broadcast packets. Note that when a new VLAN is first added the
9428          * driver will convert all filters marked I40E_VLAN_ANY into VLAN
9429          * specific filters as part of transitioning into "vlan" operation.
9430          * When more VLANs are added, the driver will copy each existing MAC
9431          * filter and add it for the new VLAN.
9432          *
9433          * Broadcast filters are handled specially by
9434          * i40e_sync_filters_subtask, as the driver must to set the broadcast
9435          * promiscuous bit instead of adding this directly as a MAC/VLAN
9436          * filter. The subtask will update the correct broadcast promiscuous
9437          * bits as VLANs become active or inactive.
9438          */
9439         eth_broadcast_addr(broadcast);
9440         spin_lock_bh(&vsi->mac_filter_hash_lock);
9441         i40e_add_mac_filter(vsi, broadcast);
9442         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9443
9444         ether_addr_copy(netdev->dev_addr, mac_addr);
9445         ether_addr_copy(netdev->perm_addr, mac_addr);
9446
9447         netdev->priv_flags |= IFF_UNICAST_FLT;
9448         netdev->priv_flags |= IFF_SUPP_NOFCS;
9449         /* Setup netdev TC information */
9450         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9451
9452         netdev->netdev_ops = &i40e_netdev_ops;
9453         netdev->watchdog_timeo = 5 * HZ;
9454         i40e_set_ethtool_ops(netdev);
9455
9456         /* MTU range: 68 - 9706 */
9457         netdev->min_mtu = ETH_MIN_MTU;
9458         netdev->max_mtu = I40E_MAX_RXBUFFER -
9459                           (ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
9460
9461         return 0;
9462 }
9463
9464 /**
9465  * i40e_vsi_delete - Delete a VSI from the switch
9466  * @vsi: the VSI being removed
9467  *
9468  * Returns 0 on success, negative value on failure
9469  **/
9470 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9471 {
9472         /* remove default VSI is not allowed */
9473         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9474                 return;
9475
9476         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9477 }
9478
9479 /**
9480  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9481  * @vsi: the VSI being queried
9482  *
9483  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9484  **/
9485 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9486 {
9487         struct i40e_veb *veb;
9488         struct i40e_pf *pf = vsi->back;
9489
9490         /* Uplink is not a bridge so default to VEB */
9491         if (vsi->veb_idx == I40E_NO_VEB)
9492                 return 1;
9493
9494         veb = pf->veb[vsi->veb_idx];
9495         if (!veb) {
9496                 dev_info(&pf->pdev->dev,
9497                          "There is no veb associated with the bridge\n");
9498                 return -ENOENT;
9499         }
9500
9501         /* Uplink is a bridge in VEPA mode */
9502         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9503                 return 0;
9504         } else {
9505                 /* Uplink is a bridge in VEB mode */
9506                 return 1;
9507         }
9508
9509         /* VEPA is now default bridge, so return 0 */
9510         return 0;
9511 }
9512
9513 /**
9514  * i40e_add_vsi - Add a VSI to the switch
9515  * @vsi: the VSI being configured
9516  *
9517  * This initializes a VSI context depending on the VSI type to be added and
9518  * passes it down to the add_vsi aq command.
9519  **/
9520 static int i40e_add_vsi(struct i40e_vsi *vsi)
9521 {
9522         int ret = -ENODEV;
9523         struct i40e_pf *pf = vsi->back;
9524         struct i40e_hw *hw = &pf->hw;
9525         struct i40e_vsi_context ctxt;
9526         struct i40e_mac_filter *f;
9527         struct hlist_node *h;
9528         int bkt;
9529
9530         u8 enabled_tc = 0x1; /* TC0 enabled */
9531         int f_count = 0;
9532
9533         memset(&ctxt, 0, sizeof(ctxt));
9534         switch (vsi->type) {
9535         case I40E_VSI_MAIN:
9536                 /* The PF's main VSI is already setup as part of the
9537                  * device initialization, so we'll not bother with
9538                  * the add_vsi call, but we will retrieve the current
9539                  * VSI context.
9540                  */
9541                 ctxt.seid = pf->main_vsi_seid;
9542                 ctxt.pf_num = pf->hw.pf_id;
9543                 ctxt.vf_num = 0;
9544                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9545                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9546                 if (ret) {
9547                         dev_info(&pf->pdev->dev,
9548                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9549                                  i40e_stat_str(&pf->hw, ret),
9550                                  i40e_aq_str(&pf->hw,
9551                                              pf->hw.aq.asq_last_status));
9552                         return -ENOENT;
9553                 }
9554                 vsi->info = ctxt.info;
9555                 vsi->info.valid_sections = 0;
9556
9557                 vsi->seid = ctxt.seid;
9558                 vsi->id = ctxt.vsi_number;
9559
9560                 enabled_tc = i40e_pf_get_tc_map(pf);
9561
9562                 /* MFP mode setup queue map and update VSI */
9563                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9564                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9565                         memset(&ctxt, 0, sizeof(ctxt));
9566                         ctxt.seid = pf->main_vsi_seid;
9567                         ctxt.pf_num = pf->hw.pf_id;
9568                         ctxt.vf_num = 0;
9569                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9570                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9571                         if (ret) {
9572                                 dev_info(&pf->pdev->dev,
9573                                          "update vsi failed, err %s aq_err %s\n",
9574                                          i40e_stat_str(&pf->hw, ret),
9575                                          i40e_aq_str(&pf->hw,
9576                                                     pf->hw.aq.asq_last_status));
9577                                 ret = -ENOENT;
9578                                 goto err;
9579                         }
9580                         /* update the local VSI info queue map */
9581                         i40e_vsi_update_queue_map(vsi, &ctxt);
9582                         vsi->info.valid_sections = 0;
9583                 } else {
9584                         /* Default/Main VSI is only enabled for TC0
9585                          * reconfigure it to enable all TCs that are
9586                          * available on the port in SFP mode.
9587                          * For MFP case the iSCSI PF would use this
9588                          * flow to enable LAN+iSCSI TC.
9589                          */
9590                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9591                         if (ret) {
9592                                 dev_info(&pf->pdev->dev,
9593                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9594                                          enabled_tc,
9595                                          i40e_stat_str(&pf->hw, ret),
9596                                          i40e_aq_str(&pf->hw,
9597                                                     pf->hw.aq.asq_last_status));
9598                                 ret = -ENOENT;
9599                         }
9600                 }
9601                 break;
9602
9603         case I40E_VSI_FDIR:
9604                 ctxt.pf_num = hw->pf_id;
9605                 ctxt.vf_num = 0;
9606                 ctxt.uplink_seid = vsi->uplink_seid;
9607                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9608                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9609                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9610                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9611                         ctxt.info.valid_sections |=
9612                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9613                         ctxt.info.switch_id =
9614                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9615                 }
9616                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9617                 break;
9618
9619         case I40E_VSI_VMDQ2:
9620                 ctxt.pf_num = hw->pf_id;
9621                 ctxt.vf_num = 0;
9622                 ctxt.uplink_seid = vsi->uplink_seid;
9623                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9624                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9625
9626                 /* This VSI is connected to VEB so the switch_id
9627                  * should be set to zero by default.
9628                  */
9629                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9630                         ctxt.info.valid_sections |=
9631                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9632                         ctxt.info.switch_id =
9633                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9634                 }
9635
9636                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9637                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9638                 break;
9639
9640         case I40E_VSI_SRIOV:
9641                 ctxt.pf_num = hw->pf_id;
9642                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9643                 ctxt.uplink_seid = vsi->uplink_seid;
9644                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9645                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9646
9647                 /* This VSI is connected to VEB so the switch_id
9648                  * should be set to zero by default.
9649                  */
9650                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9651                         ctxt.info.valid_sections |=
9652                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9653                         ctxt.info.switch_id =
9654                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9655                 }
9656
9657                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9658                         ctxt.info.valid_sections |=
9659                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9660                         ctxt.info.queueing_opt_flags |=
9661                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9662                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9663                 }
9664
9665                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9666                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9667                 if (pf->vf[vsi->vf_id].spoofchk) {
9668                         ctxt.info.valid_sections |=
9669                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9670                         ctxt.info.sec_flags |=
9671                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9672                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9673                 }
9674                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9675                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9676                 break;
9677
9678         case I40E_VSI_IWARP:
9679                 /* send down message to iWARP */
9680                 break;
9681
9682         default:
9683                 return -ENODEV;
9684         }
9685
9686         if (vsi->type != I40E_VSI_MAIN) {
9687                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9688                 if (ret) {
9689                         dev_info(&vsi->back->pdev->dev,
9690                                  "add vsi failed, err %s aq_err %s\n",
9691                                  i40e_stat_str(&pf->hw, ret),
9692                                  i40e_aq_str(&pf->hw,
9693                                              pf->hw.aq.asq_last_status));
9694                         ret = -ENOENT;
9695                         goto err;
9696                 }
9697                 vsi->info = ctxt.info;
9698                 vsi->info.valid_sections = 0;
9699                 vsi->seid = ctxt.seid;
9700                 vsi->id = ctxt.vsi_number;
9701         }
9702
9703         vsi->active_filters = 0;
9704         clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9705         spin_lock_bh(&vsi->mac_filter_hash_lock);
9706         /* If macvlan filters already exist, force them to get loaded */
9707         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist) {
9708                 f->state = I40E_FILTER_NEW;
9709                 f_count++;
9710         }
9711         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9712
9713         if (f_count) {
9714                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9715                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9716         }
9717
9718         /* Update VSI BW information */
9719         ret = i40e_vsi_get_bw_info(vsi);
9720         if (ret) {
9721                 dev_info(&pf->pdev->dev,
9722                          "couldn't get vsi bw info, err %s aq_err %s\n",
9723                          i40e_stat_str(&pf->hw, ret),
9724                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9725                 /* VSI is already added so not tearing that up */
9726                 ret = 0;
9727         }
9728
9729 err:
9730         return ret;
9731 }
9732
9733 /**
9734  * i40e_vsi_release - Delete a VSI and free its resources
9735  * @vsi: the VSI being removed
9736  *
9737  * Returns 0 on success or < 0 on error
9738  **/
9739 int i40e_vsi_release(struct i40e_vsi *vsi)
9740 {
9741         struct i40e_mac_filter *f;
9742         struct hlist_node *h;
9743         struct i40e_veb *veb = NULL;
9744         struct i40e_pf *pf;
9745         u16 uplink_seid;
9746         int i, n, bkt;
9747
9748         pf = vsi->back;
9749
9750         /* release of a VEB-owner or last VSI is not allowed */
9751         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9752                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9753                          vsi->seid, vsi->uplink_seid);
9754                 return -ENODEV;
9755         }
9756         if (vsi == pf->vsi[pf->lan_vsi] &&
9757             !test_bit(__I40E_DOWN, &pf->state)) {
9758                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9759                 return -ENODEV;
9760         }
9761
9762         uplink_seid = vsi->uplink_seid;
9763         if (vsi->type != I40E_VSI_SRIOV) {
9764                 if (vsi->netdev_registered) {
9765                         vsi->netdev_registered = false;
9766                         if (vsi->netdev) {
9767                                 /* results in a call to i40e_close() */
9768                                 unregister_netdev(vsi->netdev);
9769                         }
9770                 } else {
9771                         i40e_vsi_close(vsi);
9772                 }
9773                 i40e_vsi_disable_irq(vsi);
9774         }
9775
9776         spin_lock_bh(&vsi->mac_filter_hash_lock);
9777
9778         /* clear the sync flag on all filters */
9779         if (vsi->netdev) {
9780                 __dev_uc_unsync(vsi->netdev, NULL);
9781                 __dev_mc_unsync(vsi->netdev, NULL);
9782         }
9783
9784         /* make sure any remaining filters are marked for deletion */
9785         hash_for_each_safe(vsi->mac_filter_hash, bkt, h, f, hlist)
9786                 __i40e_del_filter(vsi, f);
9787
9788         spin_unlock_bh(&vsi->mac_filter_hash_lock);
9789
9790         i40e_sync_vsi_filters(vsi);
9791
9792         i40e_vsi_delete(vsi);
9793         i40e_vsi_free_q_vectors(vsi);
9794         if (vsi->netdev) {
9795                 free_netdev(vsi->netdev);
9796                 vsi->netdev = NULL;
9797         }
9798         i40e_vsi_clear_rings(vsi);
9799         i40e_vsi_clear(vsi);
9800
9801         /* If this was the last thing on the VEB, except for the
9802          * controlling VSI, remove the VEB, which puts the controlling
9803          * VSI onto the next level down in the switch.
9804          *
9805          * Well, okay, there's one more exception here: don't remove
9806          * the orphan VEBs yet.  We'll wait for an explicit remove request
9807          * from up the network stack.
9808          */
9809         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9810                 if (pf->vsi[i] &&
9811                     pf->vsi[i]->uplink_seid == uplink_seid &&
9812                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9813                         n++;      /* count the VSIs */
9814                 }
9815         }
9816         for (i = 0; i < I40E_MAX_VEB; i++) {
9817                 if (!pf->veb[i])
9818                         continue;
9819                 if (pf->veb[i]->uplink_seid == uplink_seid)
9820                         n++;     /* count the VEBs */
9821                 if (pf->veb[i]->seid == uplink_seid)
9822                         veb = pf->veb[i];
9823         }
9824         if (n == 0 && veb && veb->uplink_seid != 0)
9825                 i40e_veb_release(veb);
9826
9827         return 0;
9828 }
9829
9830 /**
9831  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9832  * @vsi: ptr to the VSI
9833  *
9834  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9835  * corresponding SW VSI structure and initializes num_queue_pairs for the
9836  * newly allocated VSI.
9837  *
9838  * Returns 0 on success or negative on failure
9839  **/
9840 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9841 {
9842         int ret = -ENOENT;
9843         struct i40e_pf *pf = vsi->back;
9844
9845         if (vsi->q_vectors[0]) {
9846                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9847                          vsi->seid);
9848                 return -EEXIST;
9849         }
9850
9851         if (vsi->base_vector) {
9852                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9853                          vsi->seid, vsi->base_vector);
9854                 return -EEXIST;
9855         }
9856
9857         ret = i40e_vsi_alloc_q_vectors(vsi);
9858         if (ret) {
9859                 dev_info(&pf->pdev->dev,
9860                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9861                          vsi->num_q_vectors, vsi->seid, ret);
9862                 vsi->num_q_vectors = 0;
9863                 goto vector_setup_out;
9864         }
9865
9866         /* In Legacy mode, we do not have to get any other vector since we
9867          * piggyback on the misc/ICR0 for queue interrupts.
9868         */
9869         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9870                 return ret;
9871         if (vsi->num_q_vectors)
9872                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9873                                                  vsi->num_q_vectors, vsi->idx);
9874         if (vsi->base_vector < 0) {
9875                 dev_info(&pf->pdev->dev,
9876                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9877                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9878                 i40e_vsi_free_q_vectors(vsi);
9879                 ret = -ENOENT;
9880                 goto vector_setup_out;
9881         }
9882
9883 vector_setup_out:
9884         return ret;
9885 }
9886
9887 /**
9888  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9889  * @vsi: pointer to the vsi.
9890  *
9891  * This re-allocates a vsi's queue resources.
9892  *
9893  * Returns pointer to the successfully allocated and configured VSI sw struct
9894  * on success, otherwise returns NULL on failure.
9895  **/
9896 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9897 {
9898         struct i40e_pf *pf;
9899         u8 enabled_tc;
9900         int ret;
9901
9902         if (!vsi)
9903                 return NULL;
9904
9905         pf = vsi->back;
9906
9907         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9908         i40e_vsi_clear_rings(vsi);
9909
9910         i40e_vsi_free_arrays(vsi, false);
9911         i40e_set_num_rings_in_vsi(vsi);
9912         ret = i40e_vsi_alloc_arrays(vsi, false);
9913         if (ret)
9914                 goto err_vsi;
9915
9916         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9917         if (ret < 0) {
9918                 dev_info(&pf->pdev->dev,
9919                          "failed to get tracking for %d queues for VSI %d err %d\n",
9920                          vsi->alloc_queue_pairs, vsi->seid, ret);
9921                 goto err_vsi;
9922         }
9923         vsi->base_queue = ret;
9924
9925         /* Update the FW view of the VSI. Force a reset of TC and queue
9926          * layout configurations.
9927          */
9928         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9929         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9930         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9931         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9932         if (vsi->type == I40E_VSI_MAIN)
9933                 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9934
9935         /* assign it some queues */
9936         ret = i40e_alloc_rings(vsi);
9937         if (ret)
9938                 goto err_rings;
9939
9940         /* map all of the rings to the q_vectors */
9941         i40e_vsi_map_rings_to_vectors(vsi);
9942         return vsi;
9943
9944 err_rings:
9945         i40e_vsi_free_q_vectors(vsi);
9946         if (vsi->netdev_registered) {
9947                 vsi->netdev_registered = false;
9948                 unregister_netdev(vsi->netdev);
9949                 free_netdev(vsi->netdev);
9950                 vsi->netdev = NULL;
9951         }
9952         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9953 err_vsi:
9954         i40e_vsi_clear(vsi);
9955         return NULL;
9956 }
9957
9958 /**
9959  * i40e_vsi_setup - Set up a VSI by a given type
9960  * @pf: board private structure
9961  * @type: VSI type
9962  * @uplink_seid: the switch element to link to
9963  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9964  *
9965  * This allocates the sw VSI structure and its queue resources, then add a VSI
9966  * to the identified VEB.
9967  *
9968  * Returns pointer to the successfully allocated and configure VSI sw struct on
9969  * success, otherwise returns NULL on failure.
9970  **/
9971 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9972                                 u16 uplink_seid, u32 param1)
9973 {
9974         struct i40e_vsi *vsi = NULL;
9975         struct i40e_veb *veb = NULL;
9976         int ret, i;
9977         int v_idx;
9978
9979         /* The requested uplink_seid must be either
9980          *     - the PF's port seid
9981          *              no VEB is needed because this is the PF
9982          *              or this is a Flow Director special case VSI
9983          *     - seid of an existing VEB
9984          *     - seid of a VSI that owns an existing VEB
9985          *     - seid of a VSI that doesn't own a VEB
9986          *              a new VEB is created and the VSI becomes the owner
9987          *     - seid of the PF VSI, which is what creates the first VEB
9988          *              this is a special case of the previous
9989          *
9990          * Find which uplink_seid we were given and create a new VEB if needed
9991          */
9992         for (i = 0; i < I40E_MAX_VEB; i++) {
9993                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9994                         veb = pf->veb[i];
9995                         break;
9996                 }
9997         }
9998
9999         if (!veb && uplink_seid != pf->mac_seid) {
10000
10001                 for (i = 0; i < pf->num_alloc_vsi; i++) {
10002                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
10003                                 vsi = pf->vsi[i];
10004                                 break;
10005                         }
10006                 }
10007                 if (!vsi) {
10008                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
10009                                  uplink_seid);
10010                         return NULL;
10011                 }
10012
10013                 if (vsi->uplink_seid == pf->mac_seid)
10014                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
10015                                              vsi->tc_config.enabled_tc);
10016                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
10017                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
10018                                              vsi->tc_config.enabled_tc);
10019                 if (veb) {
10020                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
10021                                 dev_info(&vsi->back->pdev->dev,
10022                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
10023                                 return NULL;
10024                         }
10025                         /* We come up by default in VEPA mode if SRIOV is not
10026                          * already enabled, in which case we can't force VEPA
10027                          * mode.
10028                          */
10029                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
10030                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
10031                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
10032                         }
10033                         i40e_config_bridge_mode(veb);
10034                 }
10035                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
10036                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
10037                                 veb = pf->veb[i];
10038                 }
10039                 if (!veb) {
10040                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
10041                         return NULL;
10042                 }
10043
10044                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10045                 uplink_seid = veb->seid;
10046         }
10047
10048         /* get vsi sw struct */
10049         v_idx = i40e_vsi_mem_alloc(pf, type);
10050         if (v_idx < 0)
10051                 goto err_alloc;
10052         vsi = pf->vsi[v_idx];
10053         if (!vsi)
10054                 goto err_alloc;
10055         vsi->type = type;
10056         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
10057
10058         if (type == I40E_VSI_MAIN)
10059                 pf->lan_vsi = v_idx;
10060         else if (type == I40E_VSI_SRIOV)
10061                 vsi->vf_id = param1;
10062         /* assign it some queues */
10063         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
10064                                 vsi->idx);
10065         if (ret < 0) {
10066                 dev_info(&pf->pdev->dev,
10067                          "failed to get tracking for %d queues for VSI %d err=%d\n",
10068                          vsi->alloc_queue_pairs, vsi->seid, ret);
10069                 goto err_vsi;
10070         }
10071         vsi->base_queue = ret;
10072
10073         /* get a VSI from the hardware */
10074         vsi->uplink_seid = uplink_seid;
10075         ret = i40e_add_vsi(vsi);
10076         if (ret)
10077                 goto err_vsi;
10078
10079         switch (vsi->type) {
10080         /* setup the netdev if needed */
10081         case I40E_VSI_MAIN:
10082                 /* Apply relevant filters if a platform-specific mac
10083                  * address was selected.
10084                  */
10085                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
10086                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
10087                         if (ret) {
10088                                 dev_warn(&pf->pdev->dev,
10089                                          "could not set up macaddr; err %d\n",
10090                                          ret);
10091                         }
10092                 }
10093         case I40E_VSI_VMDQ2:
10094                 ret = i40e_config_netdev(vsi);
10095                 if (ret)
10096                         goto err_netdev;
10097                 ret = register_netdev(vsi->netdev);
10098                 if (ret)
10099                         goto err_netdev;
10100                 vsi->netdev_registered = true;
10101                 netif_carrier_off(vsi->netdev);
10102 #ifdef CONFIG_I40E_DCB
10103                 /* Setup DCB netlink interface */
10104                 i40e_dcbnl_setup(vsi);
10105 #endif /* CONFIG_I40E_DCB */
10106                 /* fall through */
10107
10108         case I40E_VSI_FDIR:
10109                 /* set up vectors and rings if needed */
10110                 ret = i40e_vsi_setup_vectors(vsi);
10111                 if (ret)
10112                         goto err_msix;
10113
10114                 ret = i40e_alloc_rings(vsi);
10115                 if (ret)
10116                         goto err_rings;
10117
10118                 /* map all of the rings to the q_vectors */
10119                 i40e_vsi_map_rings_to_vectors(vsi);
10120
10121                 i40e_vsi_reset_stats(vsi);
10122                 break;
10123
10124         default:
10125                 /* no netdev or rings for the other VSI types */
10126                 break;
10127         }
10128
10129         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
10130             (vsi->type == I40E_VSI_VMDQ2)) {
10131                 ret = i40e_vsi_config_rss(vsi);
10132         }
10133         return vsi;
10134
10135 err_rings:
10136         i40e_vsi_free_q_vectors(vsi);
10137 err_msix:
10138         if (vsi->netdev_registered) {
10139                 vsi->netdev_registered = false;
10140                 unregister_netdev(vsi->netdev);
10141                 free_netdev(vsi->netdev);
10142                 vsi->netdev = NULL;
10143         }
10144 err_netdev:
10145         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
10146 err_vsi:
10147         i40e_vsi_clear(vsi);
10148 err_alloc:
10149         return NULL;
10150 }
10151
10152 /**
10153  * i40e_veb_get_bw_info - Query VEB BW information
10154  * @veb: the veb to query
10155  *
10156  * Query the Tx scheduler BW configuration data for given VEB
10157  **/
10158 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
10159 {
10160         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
10161         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
10162         struct i40e_pf *pf = veb->pf;
10163         struct i40e_hw *hw = &pf->hw;
10164         u32 tc_bw_max;
10165         int ret = 0;
10166         int i;
10167
10168         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
10169                                                   &bw_data, NULL);
10170         if (ret) {
10171                 dev_info(&pf->pdev->dev,
10172                          "query veb bw config failed, err %s aq_err %s\n",
10173                          i40e_stat_str(&pf->hw, ret),
10174                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10175                 goto out;
10176         }
10177
10178         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
10179                                                    &ets_data, NULL);
10180         if (ret) {
10181                 dev_info(&pf->pdev->dev,
10182                          "query veb bw ets config failed, err %s aq_err %s\n",
10183                          i40e_stat_str(&pf->hw, ret),
10184                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
10185                 goto out;
10186         }
10187
10188         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
10189         veb->bw_max_quanta = ets_data.tc_bw_max;
10190         veb->is_abs_credits = bw_data.absolute_credits_enable;
10191         veb->enabled_tc = ets_data.tc_valid_bits;
10192         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
10193                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
10194         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
10195                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
10196                 veb->bw_tc_limit_credits[i] =
10197                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
10198                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
10199         }
10200
10201 out:
10202         return ret;
10203 }
10204
10205 /**
10206  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10207  * @pf: board private structure
10208  *
10209  * On error: returns error code (negative)
10210  * On success: returns vsi index in PF (positive)
10211  **/
10212 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10213 {
10214         int ret = -ENOENT;
10215         struct i40e_veb *veb;
10216         int i;
10217
10218         /* Need to protect the allocation of switch elements at the PF level */
10219         mutex_lock(&pf->switch_mutex);
10220
10221         /* VEB list may be fragmented if VEB creation/destruction has
10222          * been happening.  We can afford to do a quick scan to look
10223          * for any free slots in the list.
10224          *
10225          * find next empty veb slot, looping back around if necessary
10226          */
10227         i = 0;
10228         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10229                 i++;
10230         if (i >= I40E_MAX_VEB) {
10231                 ret = -ENOMEM;
10232                 goto err_alloc_veb;  /* out of VEB slots! */
10233         }
10234
10235         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10236         if (!veb) {
10237                 ret = -ENOMEM;
10238                 goto err_alloc_veb;
10239         }
10240         veb->pf = pf;
10241         veb->idx = i;
10242         veb->enabled_tc = 1;
10243
10244         pf->veb[i] = veb;
10245         ret = i;
10246 err_alloc_veb:
10247         mutex_unlock(&pf->switch_mutex);
10248         return ret;
10249 }
10250
10251 /**
10252  * i40e_switch_branch_release - Delete a branch of the switch tree
10253  * @branch: where to start deleting
10254  *
10255  * This uses recursion to find the tips of the branch to be
10256  * removed, deleting until we get back to and can delete this VEB.
10257  **/
10258 static void i40e_switch_branch_release(struct i40e_veb *branch)
10259 {
10260         struct i40e_pf *pf = branch->pf;
10261         u16 branch_seid = branch->seid;
10262         u16 veb_idx = branch->idx;
10263         int i;
10264
10265         /* release any VEBs on this VEB - RECURSION */
10266         for (i = 0; i < I40E_MAX_VEB; i++) {
10267                 if (!pf->veb[i])
10268                         continue;
10269                 if (pf->veb[i]->uplink_seid == branch->seid)
10270                         i40e_switch_branch_release(pf->veb[i]);
10271         }
10272
10273         /* Release the VSIs on this VEB, but not the owner VSI.
10274          *
10275          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10276          *       the VEB itself, so don't use (*branch) after this loop.
10277          */
10278         for (i = 0; i < pf->num_alloc_vsi; i++) {
10279                 if (!pf->vsi[i])
10280                         continue;
10281                 if (pf->vsi[i]->uplink_seid == branch_seid &&
10282                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10283                         i40e_vsi_release(pf->vsi[i]);
10284                 }
10285         }
10286
10287         /* There's one corner case where the VEB might not have been
10288          * removed, so double check it here and remove it if needed.
10289          * This case happens if the veb was created from the debugfs
10290          * commands and no VSIs were added to it.
10291          */
10292         if (pf->veb[veb_idx])
10293                 i40e_veb_release(pf->veb[veb_idx]);
10294 }
10295
10296 /**
10297  * i40e_veb_clear - remove veb struct
10298  * @veb: the veb to remove
10299  **/
10300 static void i40e_veb_clear(struct i40e_veb *veb)
10301 {
10302         if (!veb)
10303                 return;
10304
10305         if (veb->pf) {
10306                 struct i40e_pf *pf = veb->pf;
10307
10308                 mutex_lock(&pf->switch_mutex);
10309                 if (pf->veb[veb->idx] == veb)
10310                         pf->veb[veb->idx] = NULL;
10311                 mutex_unlock(&pf->switch_mutex);
10312         }
10313
10314         kfree(veb);
10315 }
10316
10317 /**
10318  * i40e_veb_release - Delete a VEB and free its resources
10319  * @veb: the VEB being removed
10320  **/
10321 void i40e_veb_release(struct i40e_veb *veb)
10322 {
10323         struct i40e_vsi *vsi = NULL;
10324         struct i40e_pf *pf;
10325         int i, n = 0;
10326
10327         pf = veb->pf;
10328
10329         /* find the remaining VSI and check for extras */
10330         for (i = 0; i < pf->num_alloc_vsi; i++) {
10331                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10332                         n++;
10333                         vsi = pf->vsi[i];
10334                 }
10335         }
10336         if (n != 1) {
10337                 dev_info(&pf->pdev->dev,
10338                          "can't remove VEB %d with %d VSIs left\n",
10339                          veb->seid, n);
10340                 return;
10341         }
10342
10343         /* move the remaining VSI to uplink veb */
10344         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10345         if (veb->uplink_seid) {
10346                 vsi->uplink_seid = veb->uplink_seid;
10347                 if (veb->uplink_seid == pf->mac_seid)
10348                         vsi->veb_idx = I40E_NO_VEB;
10349                 else
10350                         vsi->veb_idx = veb->veb_idx;
10351         } else {
10352                 /* floating VEB */
10353                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10354                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10355         }
10356
10357         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10358         i40e_veb_clear(veb);
10359 }
10360
10361 /**
10362  * i40e_add_veb - create the VEB in the switch
10363  * @veb: the VEB to be instantiated
10364  * @vsi: the controlling VSI
10365  **/
10366 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10367 {
10368         struct i40e_pf *pf = veb->pf;
10369         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10370         int ret;
10371
10372         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10373                               veb->enabled_tc, false,
10374                               &veb->seid, enable_stats, NULL);
10375
10376         /* get a VEB from the hardware */
10377         if (ret) {
10378                 dev_info(&pf->pdev->dev,
10379                          "couldn't add VEB, err %s aq_err %s\n",
10380                          i40e_stat_str(&pf->hw, ret),
10381                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10382                 return -EPERM;
10383         }
10384
10385         /* get statistics counter */
10386         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10387                                          &veb->stats_idx, NULL, NULL, NULL);
10388         if (ret) {
10389                 dev_info(&pf->pdev->dev,
10390                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10391                          i40e_stat_str(&pf->hw, ret),
10392                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10393                 return -EPERM;
10394         }
10395         ret = i40e_veb_get_bw_info(veb);
10396         if (ret) {
10397                 dev_info(&pf->pdev->dev,
10398                          "couldn't get VEB bw info, err %s aq_err %s\n",
10399                          i40e_stat_str(&pf->hw, ret),
10400                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10401                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10402                 return -ENOENT;
10403         }
10404
10405         vsi->uplink_seid = veb->seid;
10406         vsi->veb_idx = veb->idx;
10407         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10408
10409         return 0;
10410 }
10411
10412 /**
10413  * i40e_veb_setup - Set up a VEB
10414  * @pf: board private structure
10415  * @flags: VEB setup flags
10416  * @uplink_seid: the switch element to link to
10417  * @vsi_seid: the initial VSI seid
10418  * @enabled_tc: Enabled TC bit-map
10419  *
10420  * This allocates the sw VEB structure and links it into the switch
10421  * It is possible and legal for this to be a duplicate of an already
10422  * existing VEB.  It is also possible for both uplink and vsi seids
10423  * to be zero, in order to create a floating VEB.
10424  *
10425  * Returns pointer to the successfully allocated VEB sw struct on
10426  * success, otherwise returns NULL on failure.
10427  **/
10428 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10429                                 u16 uplink_seid, u16 vsi_seid,
10430                                 u8 enabled_tc)
10431 {
10432         struct i40e_veb *veb, *uplink_veb = NULL;
10433         int vsi_idx, veb_idx;
10434         int ret;
10435
10436         /* if one seid is 0, the other must be 0 to create a floating relay */
10437         if ((uplink_seid == 0 || vsi_seid == 0) &&
10438             (uplink_seid + vsi_seid != 0)) {
10439                 dev_info(&pf->pdev->dev,
10440                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10441                          uplink_seid, vsi_seid);
10442                 return NULL;
10443         }
10444
10445         /* make sure there is such a vsi and uplink */
10446         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10447                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10448                         break;
10449         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10450                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10451                          vsi_seid);
10452                 return NULL;
10453         }
10454
10455         if (uplink_seid && uplink_seid != pf->mac_seid) {
10456                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10457                         if (pf->veb[veb_idx] &&
10458                             pf->veb[veb_idx]->seid == uplink_seid) {
10459                                 uplink_veb = pf->veb[veb_idx];
10460                                 break;
10461                         }
10462                 }
10463                 if (!uplink_veb) {
10464                         dev_info(&pf->pdev->dev,
10465                                  "uplink seid %d not found\n", uplink_seid);
10466                         return NULL;
10467                 }
10468         }
10469
10470         /* get veb sw struct */
10471         veb_idx = i40e_veb_mem_alloc(pf);
10472         if (veb_idx < 0)
10473                 goto err_alloc;
10474         veb = pf->veb[veb_idx];
10475         veb->flags = flags;
10476         veb->uplink_seid = uplink_seid;
10477         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10478         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10479
10480         /* create the VEB in the switch */
10481         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10482         if (ret)
10483                 goto err_veb;
10484         if (vsi_idx == pf->lan_vsi)
10485                 pf->lan_veb = veb->idx;
10486
10487         return veb;
10488
10489 err_veb:
10490         i40e_veb_clear(veb);
10491 err_alloc:
10492         return NULL;
10493 }
10494
10495 /**
10496  * i40e_setup_pf_switch_element - set PF vars based on switch type
10497  * @pf: board private structure
10498  * @ele: element we are building info from
10499  * @num_reported: total number of elements
10500  * @printconfig: should we print the contents
10501  *
10502  * helper function to assist in extracting a few useful SEID values.
10503  **/
10504 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10505                                 struct i40e_aqc_switch_config_element_resp *ele,
10506                                 u16 num_reported, bool printconfig)
10507 {
10508         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10509         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10510         u8 element_type = ele->element_type;
10511         u16 seid = le16_to_cpu(ele->seid);
10512
10513         if (printconfig)
10514                 dev_info(&pf->pdev->dev,
10515                          "type=%d seid=%d uplink=%d downlink=%d\n",
10516                          element_type, seid, uplink_seid, downlink_seid);
10517
10518         switch (element_type) {
10519         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10520                 pf->mac_seid = seid;
10521                 break;
10522         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10523                 /* Main VEB? */
10524                 if (uplink_seid != pf->mac_seid)
10525                         break;
10526                 if (pf->lan_veb == I40E_NO_VEB) {
10527                         int v;
10528
10529                         /* find existing or else empty VEB */
10530                         for (v = 0; v < I40E_MAX_VEB; v++) {
10531                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10532                                         pf->lan_veb = v;
10533                                         break;
10534                                 }
10535                         }
10536                         if (pf->lan_veb == I40E_NO_VEB) {
10537                                 v = i40e_veb_mem_alloc(pf);
10538                                 if (v < 0)
10539                                         break;
10540                                 pf->lan_veb = v;
10541                         }
10542                 }
10543
10544                 pf->veb[pf->lan_veb]->seid = seid;
10545                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10546                 pf->veb[pf->lan_veb]->pf = pf;
10547                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10548                 break;
10549         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10550                 if (num_reported != 1)
10551                         break;
10552                 /* This is immediately after a reset so we can assume this is
10553                  * the PF's VSI
10554                  */
10555                 pf->mac_seid = uplink_seid;
10556                 pf->pf_seid = downlink_seid;
10557                 pf->main_vsi_seid = seid;
10558                 if (printconfig)
10559                         dev_info(&pf->pdev->dev,
10560                                  "pf_seid=%d main_vsi_seid=%d\n",
10561                                  pf->pf_seid, pf->main_vsi_seid);
10562                 break;
10563         case I40E_SWITCH_ELEMENT_TYPE_PF:
10564         case I40E_SWITCH_ELEMENT_TYPE_VF:
10565         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10566         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10567         case I40E_SWITCH_ELEMENT_TYPE_PE:
10568         case I40E_SWITCH_ELEMENT_TYPE_PA:
10569                 /* ignore these for now */
10570                 break;
10571         default:
10572                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10573                          element_type, seid);
10574                 break;
10575         }
10576 }
10577
10578 /**
10579  * i40e_fetch_switch_configuration - Get switch config from firmware
10580  * @pf: board private structure
10581  * @printconfig: should we print the contents
10582  *
10583  * Get the current switch configuration from the device and
10584  * extract a few useful SEID values.
10585  **/
10586 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10587 {
10588         struct i40e_aqc_get_switch_config_resp *sw_config;
10589         u16 next_seid = 0;
10590         int ret = 0;
10591         u8 *aq_buf;
10592         int i;
10593
10594         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10595         if (!aq_buf)
10596                 return -ENOMEM;
10597
10598         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10599         do {
10600                 u16 num_reported, num_total;
10601
10602                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10603                                                 I40E_AQ_LARGE_BUF,
10604                                                 &next_seid, NULL);
10605                 if (ret) {
10606                         dev_info(&pf->pdev->dev,
10607                                  "get switch config failed err %s aq_err %s\n",
10608                                  i40e_stat_str(&pf->hw, ret),
10609                                  i40e_aq_str(&pf->hw,
10610                                              pf->hw.aq.asq_last_status));
10611                         kfree(aq_buf);
10612                         return -ENOENT;
10613                 }
10614
10615                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10616                 num_total = le16_to_cpu(sw_config->header.num_total);
10617
10618                 if (printconfig)
10619                         dev_info(&pf->pdev->dev,
10620                                  "header: %d reported %d total\n",
10621                                  num_reported, num_total);
10622
10623                 for (i = 0; i < num_reported; i++) {
10624                         struct i40e_aqc_switch_config_element_resp *ele =
10625                                 &sw_config->element[i];
10626
10627                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10628                                                      printconfig);
10629                 }
10630         } while (next_seid != 0);
10631
10632         kfree(aq_buf);
10633         return ret;
10634 }
10635
10636 /**
10637  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10638  * @pf: board private structure
10639  * @reinit: if the Main VSI needs to re-initialized.
10640  *
10641  * Returns 0 on success, negative value on failure
10642  **/
10643 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10644 {
10645         u16 flags = 0;
10646         int ret;
10647
10648         /* find out what's out there already */
10649         ret = i40e_fetch_switch_configuration(pf, false);
10650         if (ret) {
10651                 dev_info(&pf->pdev->dev,
10652                          "couldn't fetch switch config, err %s aq_err %s\n",
10653                          i40e_stat_str(&pf->hw, ret),
10654                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10655                 return ret;
10656         }
10657         i40e_pf_reset_stats(pf);
10658
10659         /* set the switch config bit for the whole device to
10660          * support limited promisc or true promisc
10661          * when user requests promisc. The default is limited
10662          * promisc.
10663         */
10664
10665         if ((pf->hw.pf_id == 0) &&
10666             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10667                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10668
10669         if (pf->hw.pf_id == 0) {
10670                 u16 valid_flags;
10671
10672                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10673                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10674                                                 NULL);
10675                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10676                         dev_info(&pf->pdev->dev,
10677                                  "couldn't set switch config bits, err %s aq_err %s\n",
10678                                  i40e_stat_str(&pf->hw, ret),
10679                                  i40e_aq_str(&pf->hw,
10680                                              pf->hw.aq.asq_last_status));
10681                         /* not a fatal problem, just keep going */
10682                 }
10683         }
10684
10685         /* first time setup */
10686         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10687                 struct i40e_vsi *vsi = NULL;
10688                 u16 uplink_seid;
10689
10690                 /* Set up the PF VSI associated with the PF's main VSI
10691                  * that is already in the HW switch
10692                  */
10693                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10694                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10695                 else
10696                         uplink_seid = pf->mac_seid;
10697                 if (pf->lan_vsi == I40E_NO_VSI)
10698                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10699                 else if (reinit)
10700                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10701                 if (!vsi) {
10702                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10703                         i40e_fdir_teardown(pf);
10704                         return -EAGAIN;
10705                 }
10706         } else {
10707                 /* force a reset of TC and queue layout configurations */
10708                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10709
10710                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10711                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10712                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10713         }
10714         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10715
10716         i40e_fdir_sb_setup(pf);
10717
10718         /* Setup static PF queue filter control settings */
10719         ret = i40e_setup_pf_filter_control(pf);
10720         if (ret) {
10721                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10722                          ret);
10723                 /* Failure here should not stop continuing other steps */
10724         }
10725
10726         /* enable RSS in the HW, even for only one queue, as the stack can use
10727          * the hash
10728          */
10729         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10730                 i40e_pf_config_rss(pf);
10731
10732         /* fill in link information and enable LSE reporting */
10733         i40e_link_event(pf);
10734
10735         /* Initialize user-specific link properties */
10736         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10737                                   I40E_AQ_AN_COMPLETED) ? true : false);
10738
10739         i40e_ptp_init(pf);
10740
10741         return ret;
10742 }
10743
10744 /**
10745  * i40e_determine_queue_usage - Work out queue distribution
10746  * @pf: board private structure
10747  **/
10748 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10749 {
10750         int queues_left;
10751
10752         pf->num_lan_qps = 0;
10753
10754         /* Find the max queues to be put into basic use.  We'll always be
10755          * using TC0, whether or not DCB is running, and TC0 will get the
10756          * big RSS set.
10757          */
10758         queues_left = pf->hw.func_caps.num_tx_qp;
10759
10760         if ((queues_left == 1) ||
10761             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10762                 /* one qp for PF, no queues for anything else */
10763                 queues_left = 0;
10764                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10765
10766                 /* make sure all the fancies are disabled */
10767                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10768                                I40E_FLAG_IWARP_ENABLED  |
10769                                I40E_FLAG_FD_SB_ENABLED  |
10770                                I40E_FLAG_FD_ATR_ENABLED |
10771                                I40E_FLAG_DCB_CAPABLE    |
10772                                I40E_FLAG_DCB_ENABLED    |
10773                                I40E_FLAG_SRIOV_ENABLED  |
10774                                I40E_FLAG_VMDQ_ENABLED);
10775         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10776                                   I40E_FLAG_FD_SB_ENABLED |
10777                                   I40E_FLAG_FD_ATR_ENABLED |
10778                                   I40E_FLAG_DCB_CAPABLE))) {
10779                 /* one qp for PF */
10780                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10781                 queues_left -= pf->num_lan_qps;
10782
10783                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10784                                I40E_FLAG_IWARP_ENABLED  |
10785                                I40E_FLAG_FD_SB_ENABLED  |
10786                                I40E_FLAG_FD_ATR_ENABLED |
10787                                I40E_FLAG_DCB_ENABLED    |
10788                                I40E_FLAG_VMDQ_ENABLED);
10789         } else {
10790                 /* Not enough queues for all TCs */
10791                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10792                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10793                         pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10794                                         I40E_FLAG_DCB_ENABLED);
10795                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10796                 }
10797                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10798                                         num_online_cpus());
10799                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10800                                         pf->hw.func_caps.num_tx_qp);
10801
10802                 queues_left -= pf->num_lan_qps;
10803         }
10804
10805         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10806                 if (queues_left > 1) {
10807                         queues_left -= 1; /* save 1 queue for FD */
10808                 } else {
10809                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10810                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10811                 }
10812         }
10813
10814         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10815             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10816                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10817                                         (queues_left / pf->num_vf_qps));
10818                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10819         }
10820
10821         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10822             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10823                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10824                                           (queues_left / pf->num_vmdq_qps));
10825                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10826         }
10827
10828         pf->queues_left = queues_left;
10829         dev_dbg(&pf->pdev->dev,
10830                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10831                 pf->hw.func_caps.num_tx_qp,
10832                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10833                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10834                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10835                 queues_left);
10836 }
10837
10838 /**
10839  * i40e_setup_pf_filter_control - Setup PF static filter control
10840  * @pf: PF to be setup
10841  *
10842  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10843  * settings. If PE/FCoE are enabled then it will also set the per PF
10844  * based filter sizes required for them. It also enables Flow director,
10845  * ethertype and macvlan type filter settings for the pf.
10846  *
10847  * Returns 0 on success, negative on failure
10848  **/
10849 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10850 {
10851         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10852
10853         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10854
10855         /* Flow Director is enabled */
10856         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10857                 settings->enable_fdir = true;
10858
10859         /* Ethtype and MACVLAN filters enabled for PF */
10860         settings->enable_ethtype = true;
10861         settings->enable_macvlan = true;
10862
10863         if (i40e_set_filter_control(&pf->hw, settings))
10864                 return -ENOENT;
10865
10866         return 0;
10867 }
10868
10869 #define INFO_STRING_LEN 255
10870 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10871 static void i40e_print_features(struct i40e_pf *pf)
10872 {
10873         struct i40e_hw *hw = &pf->hw;
10874         char *buf;
10875         int i;
10876
10877         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10878         if (!buf)
10879                 return;
10880
10881         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10882 #ifdef CONFIG_PCI_IOV
10883         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10884 #endif
10885         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10886                       pf->hw.func_caps.num_vsis,
10887                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
10888         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10889                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10890         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10891                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10892         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10893                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10894                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10895         }
10896         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10897                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10898         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10899         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10900         if (pf->flags & I40E_FLAG_PTP)
10901                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10902         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10903                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10904         else
10905                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10906
10907         dev_info(&pf->pdev->dev, "%s\n", buf);
10908         kfree(buf);
10909         WARN_ON(i > INFO_STRING_LEN);
10910 }
10911
10912 /**
10913  * i40e_get_platform_mac_addr - get platform-specific MAC address
10914  * @pdev: PCI device information struct
10915  * @pf: board private structure
10916  *
10917  * Look up the MAC address for the device. First we'll try
10918  * eth_platform_get_mac_address, which will check Open Firmware, or arch
10919  * specific fallback. Otherwise, we'll default to the stored value in
10920  * firmware.
10921  **/
10922 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10923 {
10924         if (eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10925                 i40e_get_mac_addr(&pf->hw, pf->hw.mac.addr);
10926 }
10927
10928 /**
10929  * i40e_probe - Device initialization routine
10930  * @pdev: PCI device information struct
10931  * @ent: entry in i40e_pci_tbl
10932  *
10933  * i40e_probe initializes a PF identified by a pci_dev structure.
10934  * The OS initialization, configuring of the PF private structure,
10935  * and a hardware reset occur.
10936  *
10937  * Returns 0 on success, negative on failure
10938  **/
10939 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10940 {
10941         struct i40e_aq_get_phy_abilities_resp abilities;
10942         struct i40e_pf *pf;
10943         struct i40e_hw *hw;
10944         static u16 pfs_found;
10945         u16 wol_nvm_bits;
10946         u16 link_status;
10947         int err;
10948         u32 val;
10949         u32 i;
10950         u8 set_fc_aq_fail;
10951
10952         err = pci_enable_device_mem(pdev);
10953         if (err)
10954                 return err;
10955
10956         /* set up for high or low dma */
10957         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10958         if (err) {
10959                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10960                 if (err) {
10961                         dev_err(&pdev->dev,
10962                                 "DMA configuration failed: 0x%x\n", err);
10963                         goto err_dma;
10964                 }
10965         }
10966
10967         /* set up pci connections */
10968         err = pci_request_mem_regions(pdev, i40e_driver_name);
10969         if (err) {
10970                 dev_info(&pdev->dev,
10971                          "pci_request_selected_regions failed %d\n", err);
10972                 goto err_pci_reg;
10973         }
10974
10975         pci_enable_pcie_error_reporting(pdev);
10976         pci_set_master(pdev);
10977
10978         /* Now that we have a PCI connection, we need to do the
10979          * low level device setup.  This is primarily setting up
10980          * the Admin Queue structures and then querying for the
10981          * device's current profile information.
10982          */
10983         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10984         if (!pf) {
10985                 err = -ENOMEM;
10986                 goto err_pf_alloc;
10987         }
10988         pf->next_vsi = 0;
10989         pf->pdev = pdev;
10990         set_bit(__I40E_DOWN, &pf->state);
10991
10992         hw = &pf->hw;
10993         hw->back = pf;
10994
10995         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10996                                 I40E_MAX_CSR_SPACE);
10997
10998         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10999         if (!hw->hw_addr) {
11000                 err = -EIO;
11001                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
11002                          (unsigned int)pci_resource_start(pdev, 0),
11003                          pf->ioremap_len, err);
11004                 goto err_ioremap;
11005         }
11006         hw->vendor_id = pdev->vendor;
11007         hw->device_id = pdev->device;
11008         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
11009         hw->subsystem_vendor_id = pdev->subsystem_vendor;
11010         hw->subsystem_device_id = pdev->subsystem_device;
11011         hw->bus.device = PCI_SLOT(pdev->devfn);
11012         hw->bus.func = PCI_FUNC(pdev->devfn);
11013         hw->bus.bus_id = pdev->bus->number;
11014         pf->instance = pfs_found;
11015
11016         INIT_LIST_HEAD(&pf->l3_flex_pit_list);
11017         INIT_LIST_HEAD(&pf->l4_flex_pit_list);
11018
11019         /* set up the locks for the AQ, do this only once in probe
11020          * and destroy them only once in remove
11021          */
11022         mutex_init(&hw->aq.asq_mutex);
11023         mutex_init(&hw->aq.arq_mutex);
11024
11025         pf->msg_enable = netif_msg_init(debug,
11026                                         NETIF_MSG_DRV |
11027                                         NETIF_MSG_PROBE |
11028                                         NETIF_MSG_LINK);
11029         if (debug < -1)
11030                 pf->hw.debug_mask = debug;
11031
11032         /* do a special CORER for clearing PXE mode once at init */
11033         if (hw->revision_id == 0 &&
11034             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
11035                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
11036                 i40e_flush(hw);
11037                 msleep(200);
11038                 pf->corer_count++;
11039
11040                 i40e_clear_pxe_mode(hw);
11041         }
11042
11043         /* Reset here to make sure all is clean and to define PF 'n' */
11044         i40e_clear_hw(hw);
11045         err = i40e_pf_reset(hw);
11046         if (err) {
11047                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
11048                 goto err_pf_reset;
11049         }
11050         pf->pfr_count++;
11051
11052         hw->aq.num_arq_entries = I40E_AQ_LEN;
11053         hw->aq.num_asq_entries = I40E_AQ_LEN;
11054         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11055         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
11056         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
11057
11058         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
11059                  "%s-%s:misc",
11060                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
11061
11062         err = i40e_init_shared_code(hw);
11063         if (err) {
11064                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
11065                          err);
11066                 goto err_pf_reset;
11067         }
11068
11069         /* set up a default setting for link flow control */
11070         pf->hw.fc.requested_mode = I40E_FC_NONE;
11071
11072         err = i40e_init_adminq(hw);
11073         if (err) {
11074                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
11075                         dev_info(&pdev->dev,
11076                                  "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");
11077                 else
11078                         dev_info(&pdev->dev,
11079                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
11080
11081                 goto err_pf_reset;
11082         }
11083
11084         /* provide nvm, fw, api versions */
11085         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
11086                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
11087                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
11088                  i40e_nvm_version_str(hw));
11089
11090         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
11091             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
11092                 dev_info(&pdev->dev,
11093                          "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");
11094         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
11095                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
11096                 dev_info(&pdev->dev,
11097                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
11098
11099         i40e_verify_eeprom(pf);
11100
11101         /* Rev 0 hardware was never productized */
11102         if (hw->revision_id < 1)
11103                 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");
11104
11105         i40e_clear_pxe_mode(hw);
11106         err = i40e_get_capabilities(pf);
11107         if (err)
11108                 goto err_adminq_setup;
11109
11110         err = i40e_sw_init(pf);
11111         if (err) {
11112                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
11113                 goto err_sw_init;
11114         }
11115
11116         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
11117                                 hw->func_caps.num_rx_qp, 0, 0);
11118         if (err) {
11119                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
11120                 goto err_init_lan_hmc;
11121         }
11122
11123         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
11124         if (err) {
11125                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
11126                 err = -ENOENT;
11127                 goto err_configure_lan_hmc;
11128         }
11129
11130         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
11131          * Ignore error return codes because if it was already disabled via
11132          * hardware settings this will fail
11133          */
11134         if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
11135                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
11136                 i40e_aq_stop_lldp(hw, true, NULL);
11137         }
11138
11139         /* allow a platform config to override the HW addr */
11140         i40e_get_platform_mac_addr(pdev, pf);
11141
11142         if (!is_valid_ether_addr(hw->mac.addr)) {
11143                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
11144                 err = -EIO;
11145                 goto err_mac_addr;
11146         }
11147         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
11148         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
11149         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
11150         if (is_valid_ether_addr(hw->mac.port_addr))
11151                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
11152
11153         pci_set_drvdata(pdev, pf);
11154         pci_save_state(pdev);
11155 #ifdef CONFIG_I40E_DCB
11156         err = i40e_init_pf_dcb(pf);
11157         if (err) {
11158                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
11159                 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE | I40E_FLAG_DCB_ENABLED);
11160                 /* Continue without DCB enabled */
11161         }
11162 #endif /* CONFIG_I40E_DCB */
11163
11164         /* set up periodic task facility */
11165         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11166         pf->service_timer_period = HZ;
11167
11168         INIT_WORK(&pf->service_task, i40e_service_task);
11169         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
11170
11171         /* NVM bit on means WoL disabled for the port */
11172         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11173         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11174                 pf->wol_en = false;
11175         else
11176                 pf->wol_en = true;
11177         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11178
11179         /* set up the main switch operations */
11180         i40e_determine_queue_usage(pf);
11181         err = i40e_init_interrupt_scheme(pf);
11182         if (err)
11183                 goto err_switch_setup;
11184
11185         /* The number of VSIs reported by the FW is the minimum guaranteed
11186          * to us; HW supports far more and we share the remaining pool with
11187          * the other PFs. We allocate space for more than the guarantee with
11188          * the understanding that we might not get them all later.
11189          */
11190         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11191                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11192         else
11193                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11194
11195         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11196         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11197                           GFP_KERNEL);
11198         if (!pf->vsi) {
11199                 err = -ENOMEM;
11200                 goto err_switch_setup;
11201         }
11202
11203 #ifdef CONFIG_PCI_IOV
11204         /* prep for VF support */
11205         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11206             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11207             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11208                 if (pci_num_vf(pdev))
11209                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11210         }
11211 #endif
11212         err = i40e_setup_pf_switch(pf, false);
11213         if (err) {
11214                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11215                 goto err_vsis;
11216         }
11217
11218         /* Make sure flow control is set according to current settings */
11219         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11220         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11221                 dev_dbg(&pf->pdev->dev,
11222                         "Set fc with err %s aq_err %s on get_phy_cap\n",
11223                         i40e_stat_str(hw, err),
11224                         i40e_aq_str(hw, hw->aq.asq_last_status));
11225         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11226                 dev_dbg(&pf->pdev->dev,
11227                         "Set fc with err %s aq_err %s on set_phy_config\n",
11228                         i40e_stat_str(hw, err),
11229                         i40e_aq_str(hw, hw->aq.asq_last_status));
11230         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11231                 dev_dbg(&pf->pdev->dev,
11232                         "Set fc with err %s aq_err %s on get_link_info\n",
11233                         i40e_stat_str(hw, err),
11234                         i40e_aq_str(hw, hw->aq.asq_last_status));
11235
11236         /* if FDIR VSI was set up, start it now */
11237         for (i = 0; i < pf->num_alloc_vsi; i++) {
11238                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11239                         i40e_vsi_open(pf->vsi[i]);
11240                         break;
11241                 }
11242         }
11243
11244         /* The driver only wants link up/down and module qualification
11245          * reports from firmware.  Note the negative logic.
11246          */
11247         err = i40e_aq_set_phy_int_mask(&pf->hw,
11248                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
11249                                          I40E_AQ_EVENT_MEDIA_NA |
11250                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11251         if (err)
11252                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11253                          i40e_stat_str(&pf->hw, err),
11254                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11255
11256         /* Reconfigure hardware for allowing smaller MSS in the case
11257          * of TSO, so that we avoid the MDD being fired and causing
11258          * a reset in the case of small MSS+TSO.
11259          */
11260         val = rd32(hw, I40E_REG_MSS);
11261         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11262                 val &= ~I40E_REG_MSS_MIN_MASK;
11263                 val |= I40E_64BYTE_MSS;
11264                 wr32(hw, I40E_REG_MSS, val);
11265         }
11266
11267         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11268                 msleep(75);
11269                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11270                 if (err)
11271                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11272                                  i40e_stat_str(&pf->hw, err),
11273                                  i40e_aq_str(&pf->hw,
11274                                              pf->hw.aq.asq_last_status));
11275         }
11276         /* The main driver is (mostly) up and happy. We need to set this state
11277          * before setting up the misc vector or we get a race and the vector
11278          * ends up disabled forever.
11279          */
11280         clear_bit(__I40E_DOWN, &pf->state);
11281
11282         /* In case of MSIX we are going to setup the misc vector right here
11283          * to handle admin queue events etc. In case of legacy and MSI
11284          * the misc functionality and queue processing is combined in
11285          * the same vector and that gets setup at open.
11286          */
11287         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11288                 err = i40e_setup_misc_vector(pf);
11289                 if (err) {
11290                         dev_info(&pdev->dev,
11291                                  "setup of misc vector failed: %d\n", err);
11292                         goto err_vsis;
11293                 }
11294         }
11295
11296 #ifdef CONFIG_PCI_IOV
11297         /* prep for VF support */
11298         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11299             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11300             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11301                 /* disable link interrupts for VFs */
11302                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11303                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11304                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11305                 i40e_flush(hw);
11306
11307                 if (pci_num_vf(pdev)) {
11308                         dev_info(&pdev->dev,
11309                                  "Active VFs found, allocating resources.\n");
11310                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11311                         if (err)
11312                                 dev_info(&pdev->dev,
11313                                          "Error %d allocating resources for existing VFs\n",
11314                                          err);
11315                 }
11316         }
11317 #endif /* CONFIG_PCI_IOV */
11318
11319         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11320                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11321                                                       pf->num_iwarp_msix,
11322                                                       I40E_IWARP_IRQ_PILE_ID);
11323                 if (pf->iwarp_base_vector < 0) {
11324                         dev_info(&pdev->dev,
11325                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11326                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11327                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11328                 }
11329         }
11330
11331         i40e_dbg_pf_init(pf);
11332
11333         /* tell the firmware that we're starting */
11334         i40e_send_version(pf);
11335
11336         /* since everything's happy, start the service_task timer */
11337         mod_timer(&pf->service_timer,
11338                   round_jiffies(jiffies + pf->service_timer_period));
11339
11340         /* add this PF to client device list and launch a client service task */
11341         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11342                 err = i40e_lan_add_device(pf);
11343                 if (err)
11344                         dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11345                                  err);
11346         }
11347
11348 #define PCI_SPEED_SIZE 8
11349 #define PCI_WIDTH_SIZE 8
11350         /* Devices on the IOSF bus do not have this information
11351          * and will report PCI Gen 1 x 1 by default so don't bother
11352          * checking them.
11353          */
11354         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11355                 char speed[PCI_SPEED_SIZE] = "Unknown";
11356                 char width[PCI_WIDTH_SIZE] = "Unknown";
11357
11358                 /* Get the negotiated link width and speed from PCI config
11359                  * space
11360                  */
11361                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11362                                           &link_status);
11363
11364                 i40e_set_pci_config_data(hw, link_status);
11365
11366                 switch (hw->bus.speed) {
11367                 case i40e_bus_speed_8000:
11368                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11369                 case i40e_bus_speed_5000:
11370                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11371                 case i40e_bus_speed_2500:
11372                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11373                 default:
11374                         break;
11375                 }
11376                 switch (hw->bus.width) {
11377                 case i40e_bus_width_pcie_x8:
11378                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11379                 case i40e_bus_width_pcie_x4:
11380                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11381                 case i40e_bus_width_pcie_x2:
11382                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11383                 case i40e_bus_width_pcie_x1:
11384                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11385                 default:
11386                         break;
11387                 }
11388
11389                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11390                          speed, width);
11391
11392                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11393                     hw->bus.speed < i40e_bus_speed_8000) {
11394                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11395                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11396                 }
11397         }
11398
11399         /* get the requested speeds from the fw */
11400         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11401         if (err)
11402                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11403                         i40e_stat_str(&pf->hw, err),
11404                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11405         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11406
11407         /* get the supported phy types from the fw */
11408         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11409         if (err)
11410                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11411                         i40e_stat_str(&pf->hw, err),
11412                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11413
11414         /* Add a filter to drop all Flow control frames from any VSI from being
11415          * transmitted. By doing so we stop a malicious VF from sending out
11416          * PAUSE or PFC frames and potentially controlling traffic for other
11417          * PF/VF VSIs.
11418          * The FW can still send Flow control frames if enabled.
11419          */
11420         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11421                                                        pf->main_vsi_seid);
11422
11423         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11424                 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11425                 pf->flags |= I40E_FLAG_PHY_CONTROLS_LEDS;
11426         if (pf->hw.device_id == I40E_DEV_ID_SFP_I_X722)
11427                 pf->flags |= I40E_FLAG_HAVE_CRT_RETIMER;
11428         /* print a string summarizing features */
11429         i40e_print_features(pf);
11430
11431         return 0;
11432
11433         /* Unwind what we've done if something failed in the setup */
11434 err_vsis:
11435         set_bit(__I40E_DOWN, &pf->state);
11436         i40e_clear_interrupt_scheme(pf);
11437         kfree(pf->vsi);
11438 err_switch_setup:
11439         i40e_reset_interrupt_capability(pf);
11440         del_timer_sync(&pf->service_timer);
11441 err_mac_addr:
11442 err_configure_lan_hmc:
11443         (void)i40e_shutdown_lan_hmc(hw);
11444 err_init_lan_hmc:
11445         kfree(pf->qp_pile);
11446 err_sw_init:
11447 err_adminq_setup:
11448 err_pf_reset:
11449         iounmap(hw->hw_addr);
11450 err_ioremap:
11451         kfree(pf);
11452 err_pf_alloc:
11453         pci_disable_pcie_error_reporting(pdev);
11454         pci_release_mem_regions(pdev);
11455 err_pci_reg:
11456 err_dma:
11457         pci_disable_device(pdev);
11458         return err;
11459 }
11460
11461 /**
11462  * i40e_remove - Device removal routine
11463  * @pdev: PCI device information struct
11464  *
11465  * i40e_remove is called by the PCI subsystem to alert the driver
11466  * that is should release a PCI device.  This could be caused by a
11467  * Hot-Plug event, or because the driver is going to be removed from
11468  * memory.
11469  **/
11470 static void i40e_remove(struct pci_dev *pdev)
11471 {
11472         struct i40e_pf *pf = pci_get_drvdata(pdev);
11473         struct i40e_hw *hw = &pf->hw;
11474         i40e_status ret_code;
11475         int i;
11476
11477         i40e_dbg_pf_exit(pf);
11478
11479         i40e_ptp_stop(pf);
11480
11481         /* Disable RSS in hw */
11482         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11483         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11484
11485         /* no more scheduling of any task */
11486         set_bit(__I40E_SUSPENDED, &pf->state);
11487         set_bit(__I40E_DOWN, &pf->state);
11488         if (pf->service_timer.data)
11489                 del_timer_sync(&pf->service_timer);
11490         if (pf->service_task.func)
11491                 cancel_work_sync(&pf->service_task);
11492
11493         /* Client close must be called explicitly here because the timer
11494          * has been stopped.
11495          */
11496         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11497
11498         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11499                 i40e_free_vfs(pf);
11500                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11501         }
11502
11503         i40e_fdir_teardown(pf);
11504
11505         /* If there is a switch structure or any orphans, remove them.
11506          * This will leave only the PF's VSI remaining.
11507          */
11508         for (i = 0; i < I40E_MAX_VEB; i++) {
11509                 if (!pf->veb[i])
11510                         continue;
11511
11512                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11513                     pf->veb[i]->uplink_seid == 0)
11514                         i40e_switch_branch_release(pf->veb[i]);
11515         }
11516
11517         /* Now we can shutdown the PF's VSI, just before we kill
11518          * adminq and hmc.
11519          */
11520         if (pf->vsi[pf->lan_vsi])
11521                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11522
11523         /* remove attached clients */
11524         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11525                 ret_code = i40e_lan_del_device(pf);
11526                 if (ret_code)
11527                         dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11528                                  ret_code);
11529         }
11530
11531         /* shutdown and destroy the HMC */
11532         if (hw->hmc.hmc_obj) {
11533                 ret_code = i40e_shutdown_lan_hmc(hw);
11534                 if (ret_code)
11535                         dev_warn(&pdev->dev,
11536                                  "Failed to destroy the HMC resources: %d\n",
11537                                  ret_code);
11538         }
11539
11540         /* shutdown the adminq */
11541         i40e_shutdown_adminq(hw);
11542
11543         /* destroy the locks only once, here */
11544         mutex_destroy(&hw->aq.arq_mutex);
11545         mutex_destroy(&hw->aq.asq_mutex);
11546
11547         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11548         i40e_clear_interrupt_scheme(pf);
11549         for (i = 0; i < pf->num_alloc_vsi; i++) {
11550                 if (pf->vsi[i]) {
11551                         i40e_vsi_clear_rings(pf->vsi[i]);
11552                         i40e_vsi_clear(pf->vsi[i]);
11553                         pf->vsi[i] = NULL;
11554                 }
11555         }
11556
11557         for (i = 0; i < I40E_MAX_VEB; i++) {
11558                 kfree(pf->veb[i]);
11559                 pf->veb[i] = NULL;
11560         }
11561
11562         kfree(pf->qp_pile);
11563         kfree(pf->vsi);
11564
11565         iounmap(hw->hw_addr);
11566         kfree(pf);
11567         pci_release_mem_regions(pdev);
11568
11569         pci_disable_pcie_error_reporting(pdev);
11570         pci_disable_device(pdev);
11571 }
11572
11573 /**
11574  * i40e_pci_error_detected - warning that something funky happened in PCI land
11575  * @pdev: PCI device information struct
11576  *
11577  * Called to warn that something happened and the error handling steps
11578  * are in progress.  Allows the driver to quiesce things, be ready for
11579  * remediation.
11580  **/
11581 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11582                                                 enum pci_channel_state error)
11583 {
11584         struct i40e_pf *pf = pci_get_drvdata(pdev);
11585
11586         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11587
11588         if (!pf) {
11589                 dev_info(&pdev->dev,
11590                          "Cannot recover - error happened during device probe\n");
11591                 return PCI_ERS_RESULT_DISCONNECT;
11592         }
11593
11594         /* shutdown all operations */
11595         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11596                 rtnl_lock();
11597                 i40e_prep_for_reset(pf, true);
11598                 rtnl_unlock();
11599         }
11600
11601         /* Request a slot reset */
11602         return PCI_ERS_RESULT_NEED_RESET;
11603 }
11604
11605 /**
11606  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11607  * @pdev: PCI device information struct
11608  *
11609  * Called to find if the driver can work with the device now that
11610  * the pci slot has been reset.  If a basic connection seems good
11611  * (registers are readable and have sane content) then return a
11612  * happy little PCI_ERS_RESULT_xxx.
11613  **/
11614 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11615 {
11616         struct i40e_pf *pf = pci_get_drvdata(pdev);
11617         pci_ers_result_t result;
11618         int err;
11619         u32 reg;
11620
11621         dev_dbg(&pdev->dev, "%s\n", __func__);
11622         if (pci_enable_device_mem(pdev)) {
11623                 dev_info(&pdev->dev,
11624                          "Cannot re-enable PCI device after reset.\n");
11625                 result = PCI_ERS_RESULT_DISCONNECT;
11626         } else {
11627                 pci_set_master(pdev);
11628                 pci_restore_state(pdev);
11629                 pci_save_state(pdev);
11630                 pci_wake_from_d3(pdev, false);
11631
11632                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11633                 if (reg == 0)
11634                         result = PCI_ERS_RESULT_RECOVERED;
11635                 else
11636                         result = PCI_ERS_RESULT_DISCONNECT;
11637         }
11638
11639         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11640         if (err) {
11641                 dev_info(&pdev->dev,
11642                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11643                          err);
11644                 /* non-fatal, continue */
11645         }
11646
11647         return result;
11648 }
11649
11650 /**
11651  * i40e_pci_error_resume - restart operations after PCI error recovery
11652  * @pdev: PCI device information struct
11653  *
11654  * Called to allow the driver to bring things back up after PCI error
11655  * and/or reset recovery has finished.
11656  **/
11657 static void i40e_pci_error_resume(struct pci_dev *pdev)
11658 {
11659         struct i40e_pf *pf = pci_get_drvdata(pdev);
11660
11661         dev_dbg(&pdev->dev, "%s\n", __func__);
11662         if (test_bit(__I40E_SUSPENDED, &pf->state))
11663                 return;
11664
11665         rtnl_lock();
11666         i40e_handle_reset_warning(pf, true);
11667         rtnl_unlock();
11668 }
11669
11670 /**
11671  * i40e_enable_mc_magic_wake - enable multicast magic packet wake up
11672  * using the mac_address_write admin q function
11673  * @pf: pointer to i40e_pf struct
11674  **/
11675 static void i40e_enable_mc_magic_wake(struct i40e_pf *pf)
11676 {
11677         struct i40e_hw *hw = &pf->hw;
11678         i40e_status ret;
11679         u8 mac_addr[6];
11680         u16 flags = 0;
11681
11682         /* Get current MAC address in case it's an LAA */
11683         if (pf->vsi[pf->lan_vsi] && pf->vsi[pf->lan_vsi]->netdev) {
11684                 ether_addr_copy(mac_addr,
11685                                 pf->vsi[pf->lan_vsi]->netdev->dev_addr);
11686         } else {
11687                 dev_err(&pf->pdev->dev,
11688                         "Failed to retrieve MAC address; using default\n");
11689                 ether_addr_copy(mac_addr, hw->mac.addr);
11690         }
11691
11692         /* The FW expects the mac address write cmd to first be called with
11693          * one of these flags before calling it again with the multicast
11694          * enable flags.
11695          */
11696         flags = I40E_AQC_WRITE_TYPE_LAA_WOL;
11697
11698         if (hw->func_caps.flex10_enable && hw->partition_id != 1)
11699                 flags = I40E_AQC_WRITE_TYPE_LAA_ONLY;
11700
11701         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
11702         if (ret) {
11703                 dev_err(&pf->pdev->dev,
11704                         "Failed to update MAC address registers; cannot enable Multicast Magic packet wake up");
11705                 return;
11706         }
11707
11708         flags = I40E_AQC_MC_MAG_EN
11709                         | I40E_AQC_WOL_PRESERVE_ON_PFR
11710                         | I40E_AQC_WRITE_TYPE_UPDATE_MC_MAG;
11711         ret = i40e_aq_mac_address_write(hw, flags, mac_addr, NULL);
11712         if (ret)
11713                 dev_err(&pf->pdev->dev,
11714                         "Failed to enable Multicast Magic Packet wake up\n");
11715 }
11716
11717 /**
11718  * i40e_shutdown - PCI callback for shutting down
11719  * @pdev: PCI device information struct
11720  **/
11721 static void i40e_shutdown(struct pci_dev *pdev)
11722 {
11723         struct i40e_pf *pf = pci_get_drvdata(pdev);
11724         struct i40e_hw *hw = &pf->hw;
11725
11726         set_bit(__I40E_SUSPENDED, &pf->state);
11727         set_bit(__I40E_DOWN, &pf->state);
11728         rtnl_lock();
11729         i40e_prep_for_reset(pf, true);
11730         rtnl_unlock();
11731
11732         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11733         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11734
11735         del_timer_sync(&pf->service_timer);
11736         cancel_work_sync(&pf->service_task);
11737         i40e_fdir_teardown(pf);
11738
11739         /* Client close must be called explicitly here because the timer
11740          * has been stopped.
11741          */
11742         i40e_notify_client_of_netdev_close(pf->vsi[pf->lan_vsi], false);
11743
11744         if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
11745                 i40e_enable_mc_magic_wake(pf);
11746
11747         rtnl_lock();
11748         i40e_prep_for_reset(pf, true);
11749         rtnl_unlock();
11750
11751         wr32(hw, I40E_PFPM_APM,
11752              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11753         wr32(hw, I40E_PFPM_WUFC,
11754              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11755
11756         i40e_clear_interrupt_scheme(pf);
11757
11758         if (system_state == SYSTEM_POWER_OFF) {
11759                 pci_wake_from_d3(pdev, pf->wol_en);
11760                 pci_set_power_state(pdev, PCI_D3hot);
11761         }
11762 }
11763
11764 #ifdef CONFIG_PM
11765 /**
11766  * i40e_suspend - PCI callback for moving to D3
11767  * @pdev: PCI device information struct
11768  **/
11769 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11770 {
11771         struct i40e_pf *pf = pci_get_drvdata(pdev);
11772         struct i40e_hw *hw = &pf->hw;
11773         int retval = 0;
11774
11775         set_bit(__I40E_SUSPENDED, &pf->state);
11776         set_bit(__I40E_DOWN, &pf->state);
11777
11778         if (pf->wol_en && (pf->flags & I40E_FLAG_WOL_MC_MAGIC_PKT_WAKE))
11779                 i40e_enable_mc_magic_wake(pf);
11780
11781         rtnl_lock();
11782         i40e_prep_for_reset(pf, true);
11783         rtnl_unlock();
11784
11785         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11786         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11787
11788         i40e_stop_misc_vector(pf);
11789
11790         retval = pci_save_state(pdev);
11791         if (retval)
11792                 return retval;
11793
11794         pci_wake_from_d3(pdev, pf->wol_en);
11795         pci_set_power_state(pdev, PCI_D3hot);
11796
11797         return retval;
11798 }
11799
11800 /**
11801  * i40e_resume - PCI callback for waking up from D3
11802  * @pdev: PCI device information struct
11803  **/
11804 static int i40e_resume(struct pci_dev *pdev)
11805 {
11806         struct i40e_pf *pf = pci_get_drvdata(pdev);
11807         u32 err;
11808
11809         pci_set_power_state(pdev, PCI_D0);
11810         pci_restore_state(pdev);
11811         /* pci_restore_state() clears dev->state_saves, so
11812          * call pci_save_state() again to restore it.
11813          */
11814         pci_save_state(pdev);
11815
11816         err = pci_enable_device_mem(pdev);
11817         if (err) {
11818                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11819                 return err;
11820         }
11821         pci_set_master(pdev);
11822
11823         /* no wakeup events while running */
11824         pci_wake_from_d3(pdev, false);
11825
11826         /* handling the reset will rebuild the device state */
11827         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11828                 clear_bit(__I40E_DOWN, &pf->state);
11829                 rtnl_lock();
11830                 i40e_reset_and_rebuild(pf, false, true);
11831                 rtnl_unlock();
11832         }
11833
11834         return 0;
11835 }
11836
11837 #endif
11838 static const struct pci_error_handlers i40e_err_handler = {
11839         .error_detected = i40e_pci_error_detected,
11840         .slot_reset = i40e_pci_error_slot_reset,
11841         .resume = i40e_pci_error_resume,
11842 };
11843
11844 static struct pci_driver i40e_driver = {
11845         .name     = i40e_driver_name,
11846         .id_table = i40e_pci_tbl,
11847         .probe    = i40e_probe,
11848         .remove   = i40e_remove,
11849 #ifdef CONFIG_PM
11850         .suspend  = i40e_suspend,
11851         .resume   = i40e_resume,
11852 #endif
11853         .shutdown = i40e_shutdown,
11854         .err_handler = &i40e_err_handler,
11855         .sriov_configure = i40e_pci_sriov_configure,
11856 };
11857
11858 /**
11859  * i40e_init_module - Driver registration routine
11860  *
11861  * i40e_init_module is the first routine called when the driver is
11862  * loaded. All it does is register with the PCI subsystem.
11863  **/
11864 static int __init i40e_init_module(void)
11865 {
11866         pr_info("%s: %s - version %s\n", i40e_driver_name,
11867                 i40e_driver_string, i40e_driver_version_str);
11868         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11869
11870         /* we will see if single thread per module is enough for now,
11871          * it can't be any worse than using the system workqueue which
11872          * was already single threaded
11873          */
11874         i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
11875                                   i40e_driver_name);
11876         if (!i40e_wq) {
11877                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11878                 return -ENOMEM;
11879         }
11880
11881         i40e_dbg_init();
11882         return pci_register_driver(&i40e_driver);
11883 }
11884 module_init(i40e_init_module);
11885
11886 /**
11887  * i40e_exit_module - Driver exit cleanup routine
11888  *
11889  * i40e_exit_module is called just before the driver is removed
11890  * from memory.
11891  **/
11892 static void __exit i40e_exit_module(void)
11893 {
11894         pci_unregister_driver(&i40e_driver);
11895         destroy_workqueue(i40e_wq);
11896         i40e_dbg_exit();
11897 }
11898 module_exit(i40e_exit_module);