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

   1 /*******************************************************************************
   2  *
   3  * Intel Ethernet Controller XL710 Family Linux Driver
   4  * Copyright(c) 2013 - 2016 Intel Corporation.
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms and conditions of the GNU General Public License,
   8  * version 2, as published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope it will be useful, but WITHOUT
  11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13  * more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
  17  *
  18  * The full GNU General Public License is included in this distribution in
  19  * the file called "COPYING".
  20  *
  21  * Contact Information:
  22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
  23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
  24  *
  25  ******************************************************************************/
  26
  27 #include <linux/etherdevice.h>
  28 #include <linux/of_net.h>
  29 #include <linux/pci.h>
  30
  31 /* Local includes */
  32 #include "i40e.h"
  33 #include "i40e_diag.h"
  34 #if IS_ENABLED(CONFIG_VXLAN)
  35 #include <net/vxlan.h>
  36 #endif
  37 #if IS_ENABLED(CONFIG_GENEVE)
  38 #include <net/geneve.h>
  39 #endif
  40
  41 const char i40e_driver_name[] = "i40e";
  42 static const char i40e_driver_string[] =
  43                         "Intel(R) Ethernet Connection XL710 Network Driver";
  44
  45 #define DRV_KERN "-k"
  46
  47 #define DRV_VERSION_MAJOR 1
  48 #define DRV_VERSION_MINOR 5
  49 #define DRV_VERSION_BUILD 10
  50 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  51              __stringify(DRV_VERSION_MINOR) "." \
  52              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  53 const char i40e_driver_version_str[] = DRV_VERSION;
  54 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  55
  56 /* a bit of forward declarations */
  57 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  58 static void i40e_handle_reset_warning(struct i40e_pf *pf);
  59 static int i40e_add_vsi(struct i40e_vsi *vsi);
  60 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  61 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  62 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  63 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  64 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  65 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
  66                               u16 rss_table_size, u16 rss_size);
  67 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  68 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  69
  70 /* i40e_pci_tbl - PCI Device ID Table
  71  *
  72  * Last entry must be all 0s
  73  *
  74  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  75  *   Class, Class Mask, private data (not used) }
  76  */
  77 static const struct pci_device_id i40e_pci_tbl[] = {
  78         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  79         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
  90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  91         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  92         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  93         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
  94         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_I_X722), 0},
  95         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  96         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  97         /* required last entry */
  98         {0, }
  99 };
 100 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
 101
 102 #define I40E_MAX_VF_COUNT 128
 103 static int debug = -1;
 104 module_param(debug, int, 0);
 105 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 106
 107 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 108 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 109 MODULE_LICENSE("GPL");
 110 MODULE_VERSION(DRV_VERSION);
 111
 112 static struct workqueue_struct *i40e_wq;
 113
 114 /**
 115  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 116  * @hw:   pointer to the HW structure
 117  * @mem:  ptr to mem struct to fill out
 118  * @size: size of memory requested
 119  * @alignment: what to align the allocation to
 120  **/
 121 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 122                             u64 size, u32 alignment)
 123 {
 124         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 125
 126         mem->size = ALIGN(size, alignment);
 127         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 128                                       &mem->pa, GFP_KERNEL);
 129         if (!mem->va)
 130                 return -ENOMEM;
 131
 132         return 0;
 133 }
 134
 135 /**
 136  * i40e_free_dma_mem_d - OS specific memory free for shared code
 137  * @hw:   pointer to the HW structure
 138  * @mem:  ptr to mem struct to free
 139  **/
 140 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 141 {
 142         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 143
 144         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 145         mem->va = NULL;
 146         mem->pa = 0;
 147         mem->size = 0;
 148
 149         return 0;
 150 }
 151
 152 /**
 153  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 154  * @hw:   pointer to the HW structure
 155  * @mem:  ptr to mem struct to fill out
 156  * @size: size of memory requested
 157  **/
 158 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 159                              u32 size)
 160 {
 161         mem->size = size;
 162         mem->va = kzalloc(size, GFP_KERNEL);
 163
 164         if (!mem->va)
 165                 return -ENOMEM;
 166
 167         return 0;
 168 }
 169
 170 /**
 171  * i40e_free_virt_mem_d - OS specific memory free for shared code
 172  * @hw:   pointer to the HW structure
 173  * @mem:  ptr to mem struct to free
 174  **/
 175 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 176 {
 177         /* it's ok to kfree a NULL pointer */
 178         kfree(mem->va);
 179         mem->va = NULL;
 180         mem->size = 0;
 181
 182         return 0;
 183 }
 184
 185 /**
 186  * i40e_get_lump - find a lump of free generic resource
 187  * @pf: board private structure
 188  * @pile: the pile of resource to search
 189  * @needed: the number of items needed
 190  * @id: an owner id to stick on the items assigned
 191  *
 192  * Returns the base item index of the lump, or negative for error
 193  *
 194  * The search_hint trick and lack of advanced fit-finding only work
 195  * because we're highly likely to have all the same size lump requests.
 196  * Linear search time and any fragmentation should be minimal.
 197  **/
 198 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 199                          u16 needed, u16 id)
 200 {
 201         int ret = -ENOMEM;
 202         int i, j;
 203
 204         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 205                 dev_info(&pf->pdev->dev,
 206                          "param err: pile=%p needed=%d id=0x%04x\n",
 207                          pile, needed, id);
 208                 return -EINVAL;
 209         }
 210
 211         /* start the linear search with an imperfect hint */
 212         i = pile->search_hint;
 213         while (i < pile->num_entries) {
 214                 /* skip already allocated entries */
 215                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 216                         i++;
 217                         continue;
 218                 }
 219
 220                 /* do we have enough in this lump? */
 221                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 222                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 223                                 break;
 224                 }
 225
 226                 if (j == needed) {
 227                         /* there was enough, so assign it to the requestor */
 228                         for (j = 0; j < needed; j++)
 229                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 230                         ret = i;
 231                         pile->search_hint = i + j;
 232                         break;
 233                 }
 234
 235                 /* not enough, so skip over it and continue looking */
 236                 i += j;
 237         }
 238
 239         return ret;
 240 }
 241
 242 /**
 243  * i40e_put_lump - return a lump of generic resource
 244  * @pile: the pile of resource to search
 245  * @index: the base item index
 246  * @id: the owner id of the items assigned
 247  *
 248  * Returns the count of items in the lump
 249  **/
 250 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 251 {
 252         int valid_id = (id | I40E_PILE_VALID_BIT);
 253         int count = 0;
 254         int i;
 255
 256         if (!pile || index >= pile->num_entries)
 257                 return -EINVAL;
 258
 259         for (i = index;
 260              i < pile->num_entries && pile->list[i] == valid_id;
 261              i++) {
 262                 pile->list[i] = 0;
 263                 count++;
 264         }
 265
 266         if (count && index < pile->search_hint)
 267                 pile->search_hint = index;
 268
 269         return count;
 270 }
 271
 272 /**
 273  * i40e_find_vsi_from_id - searches for the vsi with the given id
 274  * @pf - the pf structure to search for the vsi
 275  * @id - id of the vsi it is searching for
 276  **/
 277 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 278 {
 279         int i;
 280
 281         for (i = 0; i < pf->num_alloc_vsi; i++)
 282                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 283                         return pf->vsi[i];
 284
 285         return NULL;
 286 }
 287
 288 /**
 289  * i40e_service_event_schedule - Schedule the service task to wake up
 290  * @pf: board private structure
 291  *
 292  * If not already scheduled, this puts the task into the work queue
 293  **/
 294 void i40e_service_event_schedule(struct i40e_pf *pf)
 295 {
 296         if (!test_bit(__I40E_DOWN, &pf->state) &&
 297             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
 298             !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
 299                 queue_work(i40e_wq, &pf->service_task);
 300 }
 301
 302 /**
 303  * i40e_tx_timeout - Respond to a Tx Hang
 304  * @netdev: network interface device structure
 305  *
 306  * If any port has noticed a Tx timeout, it is likely that the whole
 307  * device is munged, not just the one netdev port, so go for the full
 308  * reset.
 309  **/
 310 #ifdef I40E_FCOE
 311 void i40e_tx_timeout(struct net_device *netdev)
 312 #else
 313 static void i40e_tx_timeout(struct net_device *netdev)
 314 #endif
 315 {
 316         struct i40e_netdev_priv *np = netdev_priv(netdev);
 317         struct i40e_vsi *vsi = np->vsi;
 318         struct i40e_pf *pf = vsi->back;
 319         struct i40e_ring *tx_ring = NULL;
 320         unsigned int i, hung_queue = 0;
 321         u32 head, val;
 322
 323         pf->tx_timeout_count++;
 324
 325         /* find the stopped queue the same way the stack does */
 326         for (i = 0; i < netdev->num_tx_queues; i++) {
 327                 struct netdev_queue *q;
 328                 unsigned long trans_start;
 329
 330                 q = netdev_get_tx_queue(netdev, i);
 331                 trans_start = q->trans_start ? : netdev->trans_start;
 332                 if (netif_xmit_stopped(q) &&
 333                     time_after(jiffies,
 334                                (trans_start + netdev->watchdog_timeo))) {
 335                         hung_queue = i;
 336                         break;
 337                 }
 338         }
 339
 340         if (i == netdev->num_tx_queues) {
 341                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 342         } else {
 343                 /* now that we have an index, find the tx_ring struct */
 344                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 345                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 346                                 if (hung_queue ==
 347                                     vsi->tx_rings[i]->queue_index) {
 348                                         tx_ring = vsi->tx_rings[i];
 349                                         break;
 350                                 }
 351                         }
 352                 }
 353         }
 354
 355         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 356                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 357         else if (time_before(jiffies,
 358                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 359                 return;   /* don't do any new action before the next timeout */
 360
 361         if (tx_ring) {
 362                 head = i40e_get_head(tx_ring);
 363                 /* Read interrupt register */
 364                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 365                         val = rd32(&pf->hw,
 366                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 367                                                 tx_ring->vsi->base_vector - 1));
 368                 else
 369                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 370
 371                 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
 372                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 373                             head, tx_ring->next_to_use,
 374                             readl(tx_ring->tail), val);
 375         }
 376
 377         pf->tx_timeout_last_recovery = jiffies;
 378         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 379                     pf->tx_timeout_recovery_level, hung_queue);
 380
 381         switch (pf->tx_timeout_recovery_level) {
 382         case 1:
 383                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
 384                 break;
 385         case 2:
 386                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
 387                 break;
 388         case 3:
 389                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
 390                 break;
 391         default:
 392                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 393                 break;
 394         }
 395
 396         i40e_service_event_schedule(pf);
 397         pf->tx_timeout_recovery_level++;
 398 }
 399
 400 /**
 401  * i40e_get_vsi_stats_struct - Get System Network Statistics
 402  * @vsi: the VSI we care about
 403  *
 404  * Returns the address of the device statistics structure.
 405  * The statistics are actually updated from the service task.
 406  **/
 407 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 408 {
 409         return &vsi->net_stats;
 410 }
 411
 412 /**
 413  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 414  * @netdev: network interface device structure
 415  *
 416  * Returns the address of the device statistics structure.
 417  * The statistics are actually updated from the service task.
 418  **/
 419 #ifdef I40E_FCOE
 420 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 421                                              struct net_device *netdev,
 422                                              struct rtnl_link_stats64 *stats)
 423 #else
 424 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 425                                              struct net_device *netdev,
 426                                              struct rtnl_link_stats64 *stats)
 427 #endif
 428 {
 429         struct i40e_netdev_priv *np = netdev_priv(netdev);
 430         struct i40e_ring *tx_ring, *rx_ring;
 431         struct i40e_vsi *vsi = np->vsi;
 432         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 433         int i;
 434
 435         if (test_bit(__I40E_DOWN, &vsi->state))
 436                 return stats;
 437
 438         if (!vsi->tx_rings)
 439                 return stats;
 440
 441         rcu_read_lock();
 442         for (i = 0; i < vsi->num_queue_pairs; i++) {
 443                 u64 bytes, packets;
 444                 unsigned int start;
 445
 446                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 447                 if (!tx_ring)
 448                         continue;
 449
 450                 do {
 451                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 452                         packets = tx_ring->stats.packets;
 453                         bytes   = tx_ring->stats.bytes;
 454                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 455
 456                 stats->tx_packets += packets;
 457                 stats->tx_bytes   += bytes;
 458                 rx_ring = &tx_ring[1];
 459
 460                 do {
 461                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 462                         packets = rx_ring->stats.packets;
 463                         bytes   = rx_ring->stats.bytes;
 464                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 465
 466                 stats->rx_packets += packets;
 467                 stats->rx_bytes   += bytes;
 468         }
 469         rcu_read_unlock();
 470
 471         /* following stats updated by i40e_watchdog_subtask() */
 472         stats->multicast        = vsi_stats->multicast;
 473         stats->tx_errors        = vsi_stats->tx_errors;
 474         stats->tx_dropped       = vsi_stats->tx_dropped;
 475         stats->rx_errors        = vsi_stats->rx_errors;
 476         stats->rx_dropped       = vsi_stats->rx_dropped;
 477         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 478         stats->rx_length_errors = vsi_stats->rx_length_errors;
 479
 480         return stats;
 481 }
 482
 483 /**
 484  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 485  * @vsi: the VSI to have its stats reset
 486  **/
 487 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 488 {
 489         struct rtnl_link_stats64 *ns;
 490         int i;
 491
 492         if (!vsi)
 493                 return;
 494
 495         ns = i40e_get_vsi_stats_struct(vsi);
 496         memset(ns, 0, sizeof(*ns));
 497         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 498         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 499         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 500         if (vsi->rx_rings && vsi->rx_rings[0]) {
 501                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 502                         memset(&vsi->rx_rings[i]->stats, 0,
 503                                sizeof(vsi->rx_rings[i]->stats));
 504                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 505                                sizeof(vsi->rx_rings[i]->rx_stats));
 506                         memset(&vsi->tx_rings[i]->stats, 0,
 507                                sizeof(vsi->tx_rings[i]->stats));
 508                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 509                                sizeof(vsi->tx_rings[i]->tx_stats));
 510                 }
 511         }
 512         vsi->stat_offsets_loaded = false;
 513 }
 514
 515 /**
 516  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 517  * @pf: the PF to be reset
 518  **/
 519 void i40e_pf_reset_stats(struct i40e_pf *pf)
 520 {
 521         int i;
 522
 523         memset(&pf->stats, 0, sizeof(pf->stats));
 524         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 525         pf->stat_offsets_loaded = false;
 526
 527         for (i = 0; i < I40E_MAX_VEB; i++) {
 528                 if (pf->veb[i]) {
 529                         memset(&pf->veb[i]->stats, 0,
 530                                sizeof(pf->veb[i]->stats));
 531                         memset(&pf->veb[i]->stats_offsets, 0,
 532                                sizeof(pf->veb[i]->stats_offsets));
 533                         pf->veb[i]->stat_offsets_loaded = false;
 534                 }
 535         }
 536 }
 537
 538 /**
 539  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 540  * @hw: ptr to the hardware info
 541  * @hireg: the high 32 bit reg to read
 542  * @loreg: the low 32 bit reg to read
 543  * @offset_loaded: has the initial offset been loaded yet
 544  * @offset: ptr to current offset value
 545  * @stat: ptr to the stat
 546  *
 547  * Since the device stats are not reset at PFReset, they likely will not
 548  * be zeroed when the driver starts.  We'll save the first values read
 549  * and use them as offsets to be subtracted from the raw values in order
 550  * to report stats that count from zero.  In the process, we also manage
 551  * the potential roll-over.
 552  **/
 553 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 554                                bool offset_loaded, u64 *offset, u64 *stat)
 555 {
 556         u64 new_data;
 557
 558         if (hw->device_id == I40E_DEV_ID_QEMU) {
 559                 new_data = rd32(hw, loreg);
 560                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 561         } else {
 562                 new_data = rd64(hw, loreg);
 563         }
 564         if (!offset_loaded)
 565                 *offset = new_data;
 566         if (likely(new_data >= *offset))
 567                 *stat = new_data - *offset;
 568         else
 569                 *stat = (new_data + BIT_ULL(48)) - *offset;
 570         *stat &= 0xFFFFFFFFFFFFULL;
 571 }
 572
 573 /**
 574  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 575  * @hw: ptr to the hardware info
 576  * @reg: the hw reg to read
 577  * @offset_loaded: has the initial offset been loaded yet
 578  * @offset: ptr to current offset value
 579  * @stat: ptr to the stat
 580  **/
 581 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 582                                bool offset_loaded, u64 *offset, u64 *stat)
 583 {
 584         u32 new_data;
 585
 586         new_data = rd32(hw, reg);
 587         if (!offset_loaded)
 588                 *offset = new_data;
 589         if (likely(new_data >= *offset))
 590                 *stat = (u32)(new_data - *offset);
 591         else
 592                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 593 }
 594
 595 /**
 596  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 597  * @vsi: the VSI to be updated
 598  **/
 599 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 600 {
 601         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 602         struct i40e_pf *pf = vsi->back;
 603         struct i40e_hw *hw = &pf->hw;
 604         struct i40e_eth_stats *oes;
 605         struct i40e_eth_stats *es;     /* device's eth stats */
 606
 607         es = &vsi->eth_stats;
 608         oes = &vsi->eth_stats_offsets;
 609
 610         /* Gather up the stats that the hw collects */
 611         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 612                            vsi->stat_offsets_loaded,
 613                            &oes->tx_errors, &es->tx_errors);
 614         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 615                            vsi->stat_offsets_loaded,
 616                            &oes->rx_discards, &es->rx_discards);
 617         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 618                            vsi->stat_offsets_loaded,
 619                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 620         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 621                            vsi->stat_offsets_loaded,
 622                            &oes->tx_errors, &es->tx_errors);
 623
 624         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 625                            I40E_GLV_GORCL(stat_idx),
 626                            vsi->stat_offsets_loaded,
 627                            &oes->rx_bytes, &es->rx_bytes);
 628         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 629                            I40E_GLV_UPRCL(stat_idx),
 630                            vsi->stat_offsets_loaded,
 631                            &oes->rx_unicast, &es->rx_unicast);
 632         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 633                            I40E_GLV_MPRCL(stat_idx),
 634                            vsi->stat_offsets_loaded,
 635                            &oes->rx_multicast, &es->rx_multicast);
 636         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 637                            I40E_GLV_BPRCL(stat_idx),
 638                            vsi->stat_offsets_loaded,
 639                            &oes->rx_broadcast, &es->rx_broadcast);
 640
 641         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 642                            I40E_GLV_GOTCL(stat_idx),
 643                            vsi->stat_offsets_loaded,
 644                            &oes->tx_bytes, &es->tx_bytes);
 645         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 646                            I40E_GLV_UPTCL(stat_idx),
 647                            vsi->stat_offsets_loaded,
 648                            &oes->tx_unicast, &es->tx_unicast);
 649         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 650                            I40E_GLV_MPTCL(stat_idx),
 651                            vsi->stat_offsets_loaded,
 652                            &oes->tx_multicast, &es->tx_multicast);
 653         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 654                            I40E_GLV_BPTCL(stat_idx),
 655                            vsi->stat_offsets_loaded,
 656                            &oes->tx_broadcast, &es->tx_broadcast);
 657         vsi->stat_offsets_loaded = true;
 658 }
 659
 660 /**
 661  * i40e_update_veb_stats - Update Switch component statistics
 662  * @veb: the VEB being updated
 663  **/
 664 static void i40e_update_veb_stats(struct i40e_veb *veb)
 665 {
 666         struct i40e_pf *pf = veb->pf;
 667         struct i40e_hw *hw = &pf->hw;
 668         struct i40e_eth_stats *oes;
 669         struct i40e_eth_stats *es;     /* device's eth stats */
 670         struct i40e_veb_tc_stats *veb_oes;
 671         struct i40e_veb_tc_stats *veb_es;
 672         int i, idx = 0;
 673
 674         idx = veb->stats_idx;
 675         es = &veb->stats;
 676         oes = &veb->stats_offsets;
 677         veb_es = &veb->tc_stats;
 678         veb_oes = &veb->tc_stats_offsets;
 679
 680         /* Gather up the stats that the hw collects */
 681         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 682                            veb->stat_offsets_loaded,
 683                            &oes->tx_discards, &es->tx_discards);
 684         if (hw->revision_id > 0)
 685                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 686                                    veb->stat_offsets_loaded,
 687                                    &oes->rx_unknown_protocol,
 688                                    &es->rx_unknown_protocol);
 689         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 690                            veb->stat_offsets_loaded,
 691                            &oes->rx_bytes, &es->rx_bytes);
 692         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 693                            veb->stat_offsets_loaded,
 694                            &oes->rx_unicast, &es->rx_unicast);
 695         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 696                            veb->stat_offsets_loaded,
 697                            &oes->rx_multicast, &es->rx_multicast);
 698         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 699                            veb->stat_offsets_loaded,
 700                            &oes->rx_broadcast, &es->rx_broadcast);
 701
 702         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 703                            veb->stat_offsets_loaded,
 704                            &oes->tx_bytes, &es->tx_bytes);
 705         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 706                            veb->stat_offsets_loaded,
 707                            &oes->tx_unicast, &es->tx_unicast);
 708         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 709                            veb->stat_offsets_loaded,
 710                            &oes->tx_multicast, &es->tx_multicast);
 711         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 712                            veb->stat_offsets_loaded,
 713                            &oes->tx_broadcast, &es->tx_broadcast);
 714         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 715                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 716                                    I40E_GLVEBTC_RPCL(i, idx),
 717                                    veb->stat_offsets_loaded,
 718                                    &veb_oes->tc_rx_packets[i],
 719                                    &veb_es->tc_rx_packets[i]);
 720                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 721                                    I40E_GLVEBTC_RBCL(i, idx),
 722                                    veb->stat_offsets_loaded,
 723                                    &veb_oes->tc_rx_bytes[i],
 724                                    &veb_es->tc_rx_bytes[i]);
 725                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 726                                    I40E_GLVEBTC_TPCL(i, idx),
 727                                    veb->stat_offsets_loaded,
 728                                    &veb_oes->tc_tx_packets[i],
 729                                    &veb_es->tc_tx_packets[i]);
 730                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 731                                    I40E_GLVEBTC_TBCL(i, idx),
 732                                    veb->stat_offsets_loaded,
 733                                    &veb_oes->tc_tx_bytes[i],
 734                                    &veb_es->tc_tx_bytes[i]);
 735         }
 736         veb->stat_offsets_loaded = true;
 737 }
 738
 739 #ifdef I40E_FCOE
 740 /**
 741  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
 742  * @vsi: the VSI that is capable of doing FCoE
 743  **/
 744 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
 745 {
 746         struct i40e_pf *pf = vsi->back;
 747         struct i40e_hw *hw = &pf->hw;
 748         struct i40e_fcoe_stats *ofs;
 749         struct i40e_fcoe_stats *fs;     /* device's eth stats */
 750         int idx;
 751
 752         if (vsi->type != I40E_VSI_FCOE)
 753                 return;
 754
 755         idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
 756         fs = &vsi->fcoe_stats;
 757         ofs = &vsi->fcoe_stats_offsets;
 758
 759         i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
 760                            vsi->fcoe_stat_offsets_loaded,
 761                            &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
 762         i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
 763                            vsi->fcoe_stat_offsets_loaded,
 764                            &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
 765         i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
 766                            vsi->fcoe_stat_offsets_loaded,
 767                            &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
 768         i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
 769                            vsi->fcoe_stat_offsets_loaded,
 770                            &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
 771         i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
 772                            vsi->fcoe_stat_offsets_loaded,
 773                            &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
 774         i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
 775                            vsi->fcoe_stat_offsets_loaded,
 776                            &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
 777         i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
 778                            vsi->fcoe_stat_offsets_loaded,
 779                            &ofs->fcoe_last_error, &fs->fcoe_last_error);
 780         i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
 781                            vsi->fcoe_stat_offsets_loaded,
 782                            &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
 783
 784         vsi->fcoe_stat_offsets_loaded = true;
 785 }
 786
 787 #endif
 788 /**
 789  * i40e_update_vsi_stats - Update the vsi statistics counters.
 790  * @vsi: the VSI to be updated
 791  *
 792  * There are a few instances where we store the same stat in a
 793  * couple of different structs.  This is partly because we have
 794  * the netdev stats that need to be filled out, which is slightly
 795  * different from the "eth_stats" defined by the chip and used in
 796  * VF communications.  We sort it out here.
 797  **/
 798 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 799 {
 800         struct i40e_pf *pf = vsi->back;
 801         struct rtnl_link_stats64 *ons;
 802         struct rtnl_link_stats64 *ns;   /* netdev stats */
 803         struct i40e_eth_stats *oes;
 804         struct i40e_eth_stats *es;     /* device's eth stats */
 805         u32 tx_restart, tx_busy;
 806         u64 tx_lost_interrupt;
 807         struct i40e_ring *p;
 808         u32 rx_page, rx_buf;
 809         u64 bytes, packets;
 810         unsigned int start;
 811         u64 tx_linearize;
 812         u64 tx_force_wb;
 813         u64 rx_p, rx_b;
 814         u64 tx_p, tx_b;
 815         u16 q;
 816
 817         if (test_bit(__I40E_DOWN, &vsi->state) ||
 818             test_bit(__I40E_CONFIG_BUSY, &pf->state))
 819                 return;
 820
 821         ns = i40e_get_vsi_stats_struct(vsi);
 822         ons = &vsi->net_stats_offsets;
 823         es = &vsi->eth_stats;
 824         oes = &vsi->eth_stats_offsets;
 825
 826         /* Gather up the netdev and vsi stats that the driver collects
 827          * on the fly during packet processing
 828          */
 829         rx_b = rx_p = 0;
 830         tx_b = tx_p = 0;
 831         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 832         tx_lost_interrupt = 0;
 833         rx_page = 0;
 834         rx_buf = 0;
 835         rcu_read_lock();
 836         for (q = 0; q < vsi->num_queue_pairs; q++) {
 837                 /* locate Tx ring */
 838                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 839
 840                 do {
 841                         start = u64_stats_fetch_begin_irq(&p->syncp);
 842                         packets = p->stats.packets;
 843                         bytes = p->stats.bytes;
 844                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 845                 tx_b += bytes;
 846                 tx_p += packets;
 847                 tx_restart += p->tx_stats.restart_queue;
 848                 tx_busy += p->tx_stats.tx_busy;
 849                 tx_linearize += p->tx_stats.tx_linearize;
 850                 tx_force_wb += p->tx_stats.tx_force_wb;
 851                 tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
 852
 853                 /* Rx queue is part of the same block as Tx queue */
 854                 p = &p[1];
 855                 do {
 856                         start = u64_stats_fetch_begin_irq(&p->syncp);
 857                         packets = p->stats.packets;
 858                         bytes = p->stats.bytes;
 859                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 860                 rx_b += bytes;
 861                 rx_p += packets;
 862                 rx_buf += p->rx_stats.alloc_buff_failed;
 863                 rx_page += p->rx_stats.alloc_page_failed;
 864         }
 865         rcu_read_unlock();
 866         vsi->tx_restart = tx_restart;
 867         vsi->tx_busy = tx_busy;
 868         vsi->tx_linearize = tx_linearize;
 869         vsi->tx_force_wb = tx_force_wb;
 870         vsi->tx_lost_interrupt = tx_lost_interrupt;
 871         vsi->rx_page_failed = rx_page;
 872         vsi->rx_buf_failed = rx_buf;
 873
 874         ns->rx_packets = rx_p;
 875         ns->rx_bytes = rx_b;
 876         ns->tx_packets = tx_p;
 877         ns->tx_bytes = tx_b;
 878
 879         /* update netdev stats from eth stats */
 880         i40e_update_eth_stats(vsi);
 881         ons->tx_errors = oes->tx_errors;
 882         ns->tx_errors = es->tx_errors;
 883         ons->multicast = oes->rx_multicast;
 884         ns->multicast = es->rx_multicast;
 885         ons->rx_dropped = oes->rx_discards;
 886         ns->rx_dropped = es->rx_discards;
 887         ons->tx_dropped = oes->tx_discards;
 888         ns->tx_dropped = es->tx_discards;
 889
 890         /* pull in a couple PF stats if this is the main vsi */
 891         if (vsi == pf->vsi[pf->lan_vsi]) {
 892                 ns->rx_crc_errors = pf->stats.crc_errors;
 893                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 894                 ns->rx_length_errors = pf->stats.rx_length_errors;
 895         }
 896 }
 897
 898 /**
 899  * i40e_update_pf_stats - Update the PF statistics counters.
 900  * @pf: the PF to be updated
 901  **/
 902 static void i40e_update_pf_stats(struct i40e_pf *pf)
 903 {
 904         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 905         struct i40e_hw_port_stats *nsd = &pf->stats;
 906         struct i40e_hw *hw = &pf->hw;
 907         u32 val;
 908         int i;
 909
 910         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 911                            I40E_GLPRT_GORCL(hw->port),
 912                            pf->stat_offsets_loaded,
 913                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 914         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 915                            I40E_GLPRT_GOTCL(hw->port),
 916                            pf->stat_offsets_loaded,
 917                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 918         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 919                            pf->stat_offsets_loaded,
 920                            &osd->eth.rx_discards,
 921                            &nsd->eth.rx_discards);
 922         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 923                            I40E_GLPRT_UPRCL(hw->port),
 924                            pf->stat_offsets_loaded,
 925                            &osd->eth.rx_unicast,
 926                            &nsd->eth.rx_unicast);
 927         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 928                            I40E_GLPRT_MPRCL(hw->port),
 929                            pf->stat_offsets_loaded,
 930                            &osd->eth.rx_multicast,
 931                            &nsd->eth.rx_multicast);
 932         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 933                            I40E_GLPRT_BPRCL(hw->port),
 934                            pf->stat_offsets_loaded,
 935                            &osd->eth.rx_broadcast,
 936                            &nsd->eth.rx_broadcast);
 937         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 938                            I40E_GLPRT_UPTCL(hw->port),
 939                            pf->stat_offsets_loaded,
 940                            &osd->eth.tx_unicast,
 941                            &nsd->eth.tx_unicast);
 942         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 943                            I40E_GLPRT_MPTCL(hw->port),
 944                            pf->stat_offsets_loaded,
 945                            &osd->eth.tx_multicast,
 946                            &nsd->eth.tx_multicast);
 947         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 948                            I40E_GLPRT_BPTCL(hw->port),
 949                            pf->stat_offsets_loaded,
 950                            &osd->eth.tx_broadcast,
 951                            &nsd->eth.tx_broadcast);
 952
 953         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 954                            pf->stat_offsets_loaded,
 955                            &osd->tx_dropped_link_down,
 956                            &nsd->tx_dropped_link_down);
 957
 958         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 959                            pf->stat_offsets_loaded,
 960                            &osd->crc_errors, &nsd->crc_errors);
 961
 962         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 963                            pf->stat_offsets_loaded,
 964                            &osd->illegal_bytes, &nsd->illegal_bytes);
 965
 966         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 967                            pf->stat_offsets_loaded,
 968                            &osd->mac_local_faults,
 969                            &nsd->mac_local_faults);
 970         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 971                            pf->stat_offsets_loaded,
 972                            &osd->mac_remote_faults,
 973                            &nsd->mac_remote_faults);
 974
 975         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 976                            pf->stat_offsets_loaded,
 977                            &osd->rx_length_errors,
 978                            &nsd->rx_length_errors);
 979
 980         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 981                            pf->stat_offsets_loaded,
 982                            &osd->link_xon_rx, &nsd->link_xon_rx);
 983         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->link_xon_tx, &nsd->link_xon_tx);
 986         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
 987                            pf->stat_offsets_loaded,
 988                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
 989         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
 990                            pf->stat_offsets_loaded,
 991                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
 992
 993         for (i = 0; i < 8; i++) {
 994                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
 995                                    pf->stat_offsets_loaded,
 996                                    &osd->priority_xoff_rx[i],
 997                                    &nsd->priority_xoff_rx[i]);
 998                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
 999                                    pf->stat_offsets_loaded,
1000                                    &osd->priority_xon_rx[i],
1001                                    &nsd->priority_xon_rx[i]);
1002                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1003                                    pf->stat_offsets_loaded,
1004                                    &osd->priority_xon_tx[i],
1005                                    &nsd->priority_xon_tx[i]);
1006                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1007                                    pf->stat_offsets_loaded,
1008                                    &osd->priority_xoff_tx[i],
1009                                    &nsd->priority_xoff_tx[i]);
1010                 i40e_stat_update32(hw,
1011                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1012                                    pf->stat_offsets_loaded,
1013                                    &osd->priority_xon_2_xoff[i],
1014                                    &nsd->priority_xon_2_xoff[i]);
1015         }
1016
1017         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1018                            I40E_GLPRT_PRC64L(hw->port),
1019                            pf->stat_offsets_loaded,
1020                            &osd->rx_size_64, &nsd->rx_size_64);
1021         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1022                            I40E_GLPRT_PRC127L(hw->port),
1023                            pf->stat_offsets_loaded,
1024                            &osd->rx_size_127, &nsd->rx_size_127);
1025         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1026                            I40E_GLPRT_PRC255L(hw->port),
1027                            pf->stat_offsets_loaded,
1028                            &osd->rx_size_255, &nsd->rx_size_255);
1029         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1030                            I40E_GLPRT_PRC511L(hw->port),
1031                            pf->stat_offsets_loaded,
1032                            &osd->rx_size_511, &nsd->rx_size_511);
1033         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1034                            I40E_GLPRT_PRC1023L(hw->port),
1035                            pf->stat_offsets_loaded,
1036                            &osd->rx_size_1023, &nsd->rx_size_1023);
1037         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1038                            I40E_GLPRT_PRC1522L(hw->port),
1039                            pf->stat_offsets_loaded,
1040                            &osd->rx_size_1522, &nsd->rx_size_1522);
1041         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1042                            I40E_GLPRT_PRC9522L(hw->port),
1043                            pf->stat_offsets_loaded,
1044                            &osd->rx_size_big, &nsd->rx_size_big);
1045
1046         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1047                            I40E_GLPRT_PTC64L(hw->port),
1048                            pf->stat_offsets_loaded,
1049                            &osd->tx_size_64, &nsd->tx_size_64);
1050         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1051                            I40E_GLPRT_PTC127L(hw->port),
1052                            pf->stat_offsets_loaded,
1053                            &osd->tx_size_127, &nsd->tx_size_127);
1054         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1055                            I40E_GLPRT_PTC255L(hw->port),
1056                            pf->stat_offsets_loaded,
1057                            &osd->tx_size_255, &nsd->tx_size_255);
1058         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1059                            I40E_GLPRT_PTC511L(hw->port),
1060                            pf->stat_offsets_loaded,
1061                            &osd->tx_size_511, &nsd->tx_size_511);
1062         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1063                            I40E_GLPRT_PTC1023L(hw->port),
1064                            pf->stat_offsets_loaded,
1065                            &osd->tx_size_1023, &nsd->tx_size_1023);
1066         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1067                            I40E_GLPRT_PTC1522L(hw->port),
1068                            pf->stat_offsets_loaded,
1069                            &osd->tx_size_1522, &nsd->tx_size_1522);
1070         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1071                            I40E_GLPRT_PTC9522L(hw->port),
1072                            pf->stat_offsets_loaded,
1073                            &osd->tx_size_big, &nsd->tx_size_big);
1074
1075         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1076                            pf->stat_offsets_loaded,
1077                            &osd->rx_undersize, &nsd->rx_undersize);
1078         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1079                            pf->stat_offsets_loaded,
1080                            &osd->rx_fragments, &nsd->rx_fragments);
1081         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1082                            pf->stat_offsets_loaded,
1083                            &osd->rx_oversize, &nsd->rx_oversize);
1084         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1085                            pf->stat_offsets_loaded,
1086                            &osd->rx_jabber, &nsd->rx_jabber);
1087
1088         /* FDIR stats */
1089         i40e_stat_update32(hw,
1090                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1091                            pf->stat_offsets_loaded,
1092                            &osd->fd_atr_match, &nsd->fd_atr_match);
1093         i40e_stat_update32(hw,
1094                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1095                            pf->stat_offsets_loaded,
1096                            &osd->fd_sb_match, &nsd->fd_sb_match);
1097         i40e_stat_update32(hw,
1098                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1099                       pf->stat_offsets_loaded,
1100                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1101
1102         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1103         nsd->tx_lpi_status =
1104                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1105                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1106         nsd->rx_lpi_status =
1107                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1108                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1109         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1110                            pf->stat_offsets_loaded,
1111                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1112         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1113                            pf->stat_offsets_loaded,
1114                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1115
1116         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1117             !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1118                 nsd->fd_sb_status = true;
1119         else
1120                 nsd->fd_sb_status = false;
1121
1122         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1123             !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1124                 nsd->fd_atr_status = true;
1125         else
1126                 nsd->fd_atr_status = false;
1127
1128         pf->stat_offsets_loaded = true;
1129 }
1130
1131 /**
1132  * i40e_update_stats - Update the various statistics counters.
1133  * @vsi: the VSI to be updated
1134  *
1135  * Update the various stats for this VSI and its related entities.
1136  **/
1137 void i40e_update_stats(struct i40e_vsi *vsi)
1138 {
1139         struct i40e_pf *pf = vsi->back;
1140
1141         if (vsi == pf->vsi[pf->lan_vsi])
1142                 i40e_update_pf_stats(pf);
1143
1144         i40e_update_vsi_stats(vsi);
1145 #ifdef I40E_FCOE
1146         i40e_update_fcoe_stats(vsi);
1147 #endif
1148 }
1149
1150 /**
1151  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1152  * @vsi: the VSI to be searched
1153  * @macaddr: the MAC address
1154  * @vlan: the vlan
1155  * @is_vf: make sure its a VF filter, else doesn't matter
1156  * @is_netdev: make sure its a netdev filter, else doesn't matter
1157  *
1158  * Returns ptr to the filter object or NULL
1159  **/
1160 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1161                                                 u8 *macaddr, s16 vlan,
1162                                                 bool is_vf, bool is_netdev)
1163 {
1164         struct i40e_mac_filter *f;
1165
1166         if (!vsi || !macaddr)
1167                 return NULL;
1168
1169         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1170                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1171                     (vlan == f->vlan)    &&
1172                     (!is_vf || f->is_vf) &&
1173                     (!is_netdev || f->is_netdev))
1174                         return f;
1175         }
1176         return NULL;
1177 }
1178
1179 /**
1180  * i40e_find_mac - Find a mac addr in the macvlan filters list
1181  * @vsi: the VSI to be searched
1182  * @macaddr: the MAC address we are searching for
1183  * @is_vf: make sure its a VF filter, else doesn't matter
1184  * @is_netdev: make sure its a netdev filter, else doesn't matter
1185  *
1186  * Returns the first filter with the provided MAC address or NULL if
1187  * MAC address was not found
1188  **/
1189 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1190                                       bool is_vf, bool is_netdev)
1191 {
1192         struct i40e_mac_filter *f;
1193
1194         if (!vsi || !macaddr)
1195                 return NULL;
1196
1197         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1198                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1199                     (!is_vf || f->is_vf) &&
1200                     (!is_netdev || f->is_netdev))
1201                         return f;
1202         }
1203         return NULL;
1204 }
1205
1206 /**
1207  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1208  * @vsi: the VSI to be searched
1209  *
1210  * Returns true if VSI is in vlan mode or false otherwise
1211  **/
1212 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1213 {
1214         struct i40e_mac_filter *f;
1215
1216         /* Only -1 for all the filters denotes not in vlan mode
1217          * so we have to go through all the list in order to make sure
1218          */
1219         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1220                 if (f->vlan >= 0 || vsi->info.pvid)
1221                         return true;
1222         }
1223
1224         return false;
1225 }
1226
1227 /**
1228  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1229  * @vsi: the VSI to be searched
1230  * @macaddr: the mac address to be filtered
1231  * @is_vf: true if it is a VF
1232  * @is_netdev: true if it is a netdev
1233  *
1234  * Goes through all the macvlan filters and adds a
1235  * macvlan filter for each unique vlan that already exists
1236  *
1237  * Returns first filter found on success, else NULL
1238  **/
1239 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1240                                              bool is_vf, bool is_netdev)
1241 {
1242         struct i40e_mac_filter *f;
1243
1244         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1245                 if (vsi->info.pvid)
1246                         f->vlan = le16_to_cpu(vsi->info.pvid);
1247                 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1248                                       is_vf, is_netdev)) {
1249                         if (!i40e_add_filter(vsi, macaddr, f->vlan,
1250                                              is_vf, is_netdev))
1251                                 return NULL;
1252                 }
1253         }
1254
1255         return list_first_entry_or_null(&vsi->mac_filter_list,
1256                                         struct i40e_mac_filter, list);
1257 }
1258
1259 /**
1260  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1261  * @vsi: the VSI to be searched
1262  * @macaddr: the mac address to be removed
1263  * @is_vf: true if it is a VF
1264  * @is_netdev: true if it is a netdev
1265  *
1266  * Removes a given MAC address from a VSI, regardless of VLAN
1267  *
1268  * Returns 0 for success, or error
1269  **/
1270 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1271                           bool is_vf, bool is_netdev)
1272 {
1273         struct i40e_mac_filter *f = NULL;
1274         int changed = 0;
1275
1276         WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1277              "Missing mac_filter_list_lock\n");
1278         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1279                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1280                     (is_vf == f->is_vf) &&
1281                     (is_netdev == f->is_netdev)) {
1282                         f->counter--;
1283                         f->changed = true;
1284                         changed = 1;
1285                 }
1286         }
1287         if (changed) {
1288                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1289                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1290                 return 0;
1291         }
1292         return -ENOENT;
1293 }
1294
1295 /**
1296  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1297  * @vsi: the PF Main VSI - inappropriate for any other VSI
1298  * @macaddr: the MAC address
1299  *
1300  * Some older firmware configurations set up a default promiscuous VLAN
1301  * filter that needs to be removed.
1302  **/
1303 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1304 {
1305         struct i40e_aqc_remove_macvlan_element_data element;
1306         struct i40e_pf *pf = vsi->back;
1307         i40e_status ret;
1308
1309         /* Only appropriate for the PF main VSI */
1310         if (vsi->type != I40E_VSI_MAIN)
1311                 return -EINVAL;
1312
1313         memset(&element, 0, sizeof(element));
1314         ether_addr_copy(element.mac_addr, macaddr);
1315         element.vlan_tag = 0;
1316         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1317                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1318         ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1319         if (ret)
1320                 return -ENOENT;
1321
1322         return 0;
1323 }
1324
1325 /**
1326  * i40e_add_filter - Add a mac/vlan filter to the VSI
1327  * @vsi: the VSI to be searched
1328  * @macaddr: the MAC address
1329  * @vlan: the vlan
1330  * @is_vf: make sure its a VF filter, else doesn't matter
1331  * @is_netdev: make sure its a netdev filter, else doesn't matter
1332  *
1333  * Returns ptr to the filter object or NULL when no memory available.
1334  *
1335  * NOTE: This function is expected to be called with mac_filter_list_lock
1336  * being held.
1337  **/
1338 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1339                                         u8 *macaddr, s16 vlan,
1340                                         bool is_vf, bool is_netdev)
1341 {
1342         struct i40e_mac_filter *f;
1343
1344         if (!vsi || !macaddr)
1345                 return NULL;
1346
1347         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1348         if (!f) {
1349                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1350                 if (!f)
1351                         goto add_filter_out;
1352
1353                 ether_addr_copy(f->macaddr, macaddr);
1354                 f->vlan = vlan;
1355                 f->changed = true;
1356
1357                 INIT_LIST_HEAD(&f->list);
1358                 list_add_tail(&f->list, &vsi->mac_filter_list);
1359         }
1360
1361         /* increment counter and add a new flag if needed */
1362         if (is_vf) {
1363                 if (!f->is_vf) {
1364                         f->is_vf = true;
1365                         f->counter++;
1366                 }
1367         } else if (is_netdev) {
1368                 if (!f->is_netdev) {
1369                         f->is_netdev = true;
1370                         f->counter++;
1371                 }
1372         } else {
1373                 f->counter++;
1374         }
1375
1376         /* changed tells sync_filters_subtask to
1377          * push the filter down to the firmware
1378          */
1379         if (f->changed) {
1380                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1381                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1382         }
1383
1384 add_filter_out:
1385         return f;
1386 }
1387
1388 /**
1389  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1390  * @vsi: the VSI to be searched
1391  * @macaddr: the MAC address
1392  * @vlan: the vlan
1393  * @is_vf: make sure it's a VF filter, else doesn't matter
1394  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1395  *
1396  * NOTE: This function is expected to be called with mac_filter_list_lock
1397  * being held.
1398  **/
1399 void i40e_del_filter(struct i40e_vsi *vsi,
1400                      u8 *macaddr, s16 vlan,
1401                      bool is_vf, bool is_netdev)
1402 {
1403         struct i40e_mac_filter *f;
1404
1405         if (!vsi || !macaddr)
1406                 return;
1407
1408         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1409         if (!f || f->counter == 0)
1410                 return;
1411
1412         if (is_vf) {
1413                 if (f->is_vf) {
1414                         f->is_vf = false;
1415                         f->counter--;
1416                 }
1417         } else if (is_netdev) {
1418                 if (f->is_netdev) {
1419                         f->is_netdev = false;
1420                         f->counter--;
1421                 }
1422         } else {
1423                 /* make sure we don't remove a filter in use by VF or netdev */
1424                 int min_f = 0;
1425
1426                 min_f += (f->is_vf ? 1 : 0);
1427                 min_f += (f->is_netdev ? 1 : 0);
1428
1429                 if (f->counter > min_f)
1430                         f->counter--;
1431         }
1432
1433         /* counter == 0 tells sync_filters_subtask to
1434          * remove the filter from the firmware's list
1435          */
1436         if (f->counter == 0) {
1437                 f->changed = true;
1438                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1439                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1440         }
1441 }
1442
1443 /**
1444  * i40e_set_mac - NDO callback to set mac address
1445  * @netdev: network interface device structure
1446  * @p: pointer to an address structure
1447  *
1448  * Returns 0 on success, negative on failure
1449  **/
1450 #ifdef I40E_FCOE
1451 int i40e_set_mac(struct net_device *netdev, void *p)
1452 #else
1453 static int i40e_set_mac(struct net_device *netdev, void *p)
1454 #endif
1455 {
1456         struct i40e_netdev_priv *np = netdev_priv(netdev);
1457         struct i40e_vsi *vsi = np->vsi;
1458         struct i40e_pf *pf = vsi->back;
1459         struct i40e_hw *hw = &pf->hw;
1460         struct sockaddr *addr = p;
1461         struct i40e_mac_filter *f;
1462
1463         if (!is_valid_ether_addr(addr->sa_data))
1464                 return -EADDRNOTAVAIL;
1465
1466         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1467                 netdev_info(netdev, "already using mac address %pM\n",
1468                             addr->sa_data);
1469                 return 0;
1470         }
1471
1472         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1473             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1474                 return -EADDRNOTAVAIL;
1475
1476         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1477                 netdev_info(netdev, "returning to hw mac address %pM\n",
1478                             hw->mac.addr);
1479         else
1480                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1481
1482         if (vsi->type == I40E_VSI_MAIN) {
1483                 i40e_status ret;
1484
1485                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1486                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1487                                                 addr->sa_data, NULL);
1488                 if (ret) {
1489                         netdev_info(netdev,
1490                                     "Addr change for Main VSI failed: %d\n",
1491                                     ret);
1492                         return -EADDRNOTAVAIL;
1493                 }
1494         }
1495
1496         if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1497                 struct i40e_aqc_remove_macvlan_element_data element;
1498
1499                 memset(&element, 0, sizeof(element));
1500                 ether_addr_copy(element.mac_addr, netdev->dev_addr);
1501                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1502                 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1503         } else {
1504                 spin_lock_bh(&vsi->mac_filter_list_lock);
1505                 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1506                                 false, false);
1507                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1508         }
1509
1510         if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1511                 struct i40e_aqc_add_macvlan_element_data element;
1512
1513                 memset(&element, 0, sizeof(element));
1514                 ether_addr_copy(element.mac_addr, hw->mac.addr);
1515                 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1516                 i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1517         } else {
1518                 spin_lock_bh(&vsi->mac_filter_list_lock);
1519                 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1520                                     false, false);
1521                 if (f)
1522                         f->is_laa = true;
1523                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1524         }
1525
1526         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1527
1528         /* schedule our worker thread which will take care of
1529          * applying the new filter changes
1530          */
1531         i40e_service_event_schedule(vsi->back);
1532         return 0;
1533 }
1534
1535 /**
1536  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1537  * @vsi: the VSI being setup
1538  * @ctxt: VSI context structure
1539  * @enabled_tc: Enabled TCs bitmap
1540  * @is_add: True if called before Add VSI
1541  *
1542  * Setup VSI queue mapping for enabled traffic classes.
1543  **/
1544 #ifdef I40E_FCOE
1545 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1546                               struct i40e_vsi_context *ctxt,
1547                               u8 enabled_tc,
1548                               bool is_add)
1549 #else
1550 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1551                                      struct i40e_vsi_context *ctxt,
1552                                      u8 enabled_tc,
1553                                      bool is_add)
1554 #endif
1555 {
1556         struct i40e_pf *pf = vsi->back;
1557         u16 sections = 0;
1558         u8 netdev_tc = 0;
1559         u16 numtc = 0;
1560         u16 qcount;
1561         u8 offset;
1562         u16 qmap;
1563         int i;
1564         u16 num_tc_qps = 0;
1565
1566         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1567         offset = 0;
1568
1569         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1570                 /* Find numtc from enabled TC bitmap */
1571                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1572                         if (enabled_tc & BIT(i)) /* TC is enabled */
1573                                 numtc++;
1574                 }
1575                 if (!numtc) {
1576                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1577                         numtc = 1;
1578                 }
1579         } else {
1580                 /* At least TC0 is enabled in case of non-DCB case */
1581                 numtc = 1;
1582         }
1583
1584         vsi->tc_config.numtc = numtc;
1585         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1586         /* Number of queues per enabled TC */
1587         /* In MFP case we can have a much lower count of MSIx
1588          * vectors available and so we need to lower the used
1589          * q count.
1590          */
1591         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1592                 qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1593         else
1594                 qcount = vsi->alloc_queue_pairs;
1595         num_tc_qps = qcount / numtc;
1596         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1597
1598         /* Setup queue offset/count for all TCs for given VSI */
1599         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1600                 /* See if the given TC is enabled for the given VSI */
1601                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1602                         /* TC is enabled */
1603                         int pow, num_qps;
1604
1605                         switch (vsi->type) {
1606                         case I40E_VSI_MAIN:
1607                                 qcount = min_t(int, pf->alloc_rss_size,
1608                                                num_tc_qps);
1609                                 break;
1610 #ifdef I40E_FCOE
1611                         case I40E_VSI_FCOE:
1612                                 qcount = num_tc_qps;
1613                                 break;
1614 #endif
1615                         case I40E_VSI_FDIR:
1616                         case I40E_VSI_SRIOV:
1617                         case I40E_VSI_VMDQ2:
1618                         default:
1619                                 qcount = num_tc_qps;
1620                                 WARN_ON(i != 0);
1621                                 break;
1622                         }
1623                         vsi->tc_config.tc_info[i].qoffset = offset;
1624                         vsi->tc_config.tc_info[i].qcount = qcount;
1625
1626                         /* find the next higher power-of-2 of num queue pairs */
1627                         num_qps = qcount;
1628                         pow = 0;
1629                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1630                                 pow++;
1631                                 num_qps >>= 1;
1632                         }
1633
1634                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1635                         qmap =
1636                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1637                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1638
1639                         offset += qcount;
1640                 } else {
1641                         /* TC is not enabled so set the offset to
1642                          * default queue and allocate one queue
1643                          * for the given TC.
1644                          */
1645                         vsi->tc_config.tc_info[i].qoffset = 0;
1646                         vsi->tc_config.tc_info[i].qcount = 1;
1647                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1648
1649                         qmap = 0;
1650                 }
1651                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1652         }
1653
1654         /* Set actual Tx/Rx queue pairs */
1655         vsi->num_queue_pairs = offset;
1656         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1657                 if (vsi->req_queue_pairs > 0)
1658                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1659                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1660                         vsi->num_queue_pairs = pf->num_lan_msix;
1661         }
1662
1663         /* Scheduler section valid can only be set for ADD VSI */
1664         if (is_add) {
1665                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1666
1667                 ctxt->info.up_enable_bits = enabled_tc;
1668         }
1669         if (vsi->type == I40E_VSI_SRIOV) {
1670                 ctxt->info.mapping_flags |=
1671                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1672                 for (i = 0; i < vsi->num_queue_pairs; i++)
1673                         ctxt->info.queue_mapping[i] =
1674                                                cpu_to_le16(vsi->base_queue + i);
1675         } else {
1676                 ctxt->info.mapping_flags |=
1677                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1678                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1679         }
1680         ctxt->info.valid_sections |= cpu_to_le16(sections);
1681 }
1682
1683 /**
1684  * i40e_set_rx_mode - NDO callback to set the netdev filters
1685  * @netdev: network interface device structure
1686  **/
1687 #ifdef I40E_FCOE
1688 void i40e_set_rx_mode(struct net_device *netdev)
1689 #else
1690 static void i40e_set_rx_mode(struct net_device *netdev)
1691 #endif
1692 {
1693         struct i40e_netdev_priv *np = netdev_priv(netdev);
1694         struct i40e_mac_filter *f, *ftmp;
1695         struct i40e_vsi *vsi = np->vsi;
1696         struct netdev_hw_addr *uca;
1697         struct netdev_hw_addr *mca;
1698         struct netdev_hw_addr *ha;
1699
1700         spin_lock_bh(&vsi->mac_filter_list_lock);
1701
1702         /* add addr if not already in the filter list */
1703         netdev_for_each_uc_addr(uca, netdev) {
1704                 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1705                         if (i40e_is_vsi_in_vlan(vsi))
1706                                 i40e_put_mac_in_vlan(vsi, uca->addr,
1707                                                      false, true);
1708                         else
1709                                 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1710                                                 false, true);
1711                 }
1712         }
1713
1714         netdev_for_each_mc_addr(mca, netdev) {
1715                 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1716                         if (i40e_is_vsi_in_vlan(vsi))
1717                                 i40e_put_mac_in_vlan(vsi, mca->addr,
1718                                                      false, true);
1719                         else
1720                                 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1721                                                 false, true);
1722                 }
1723         }
1724
1725         /* remove filter if not in netdev list */
1726         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1727
1728                 if (!f->is_netdev)
1729                         continue;
1730
1731                 netdev_for_each_mc_addr(mca, netdev)
1732                         if (ether_addr_equal(mca->addr, f->macaddr))
1733                                 goto bottom_of_search_loop;
1734
1735                 netdev_for_each_uc_addr(uca, netdev)
1736                         if (ether_addr_equal(uca->addr, f->macaddr))
1737                                 goto bottom_of_search_loop;
1738
1739                 for_each_dev_addr(netdev, ha)
1740                         if (ether_addr_equal(ha->addr, f->macaddr))
1741                                 goto bottom_of_search_loop;
1742
1743                 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1744                 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1745
1746 bottom_of_search_loop:
1747                 continue;
1748         }
1749         spin_unlock_bh(&vsi->mac_filter_list_lock);
1750
1751         /* check for other flag changes */
1752         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1753                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1754                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1755         }
1756
1757         /* schedule our worker thread which will take care of
1758          * applying the new filter changes
1759          */
1760         i40e_service_event_schedule(vsi->back);
1761 }
1762
1763 /**
1764  * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1765  * @src: source MAC filter entry to be clones
1766  *
1767  * Returns the pointer to newly cloned MAC filter entry or NULL
1768  * in case of error
1769  **/
1770 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1771                                         struct i40e_mac_filter *src)
1772 {
1773         struct i40e_mac_filter *f;
1774
1775         f = kzalloc(sizeof(*f), GFP_ATOMIC);
1776         if (!f)
1777                 return NULL;
1778         *f = *src;
1779
1780         INIT_LIST_HEAD(&f->list);
1781
1782         return f;
1783 }
1784
1785 /**
1786  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1787  * @vsi: pointer to vsi struct
1788  * @from: Pointer to list which contains MAC filter entries - changes to
1789  *        those entries needs to be undone.
1790  *
1791  * MAC filter entries from list were slated to be removed from device.
1792  **/
1793 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1794                                          struct list_head *from)
1795 {
1796         struct i40e_mac_filter *f, *ftmp;
1797
1798         list_for_each_entry_safe(f, ftmp, from, list) {
1799                 f->changed = true;
1800                 /* Move the element back into MAC filter list*/
1801                 list_move_tail(&f->list, &vsi->mac_filter_list);
1802         }
1803 }
1804
1805 /**
1806  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1807  * @vsi: pointer to vsi struct
1808  *
1809  * MAC filter entries from list were slated to be added from device.
1810  **/
1811 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1812 {
1813         struct i40e_mac_filter *f, *ftmp;
1814
1815         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1816                 if (!f->changed && f->counter)
1817                         f->changed = true;
1818         }
1819 }
1820
1821 /**
1822  * i40e_cleanup_add_list - Deletes the element from add list and release
1823  *                      memory
1824  * @add_list: Pointer to list which contains MAC filter entries
1825  **/
1826 static void i40e_cleanup_add_list(struct list_head *add_list)
1827 {
1828         struct i40e_mac_filter *f, *ftmp;
1829
1830         list_for_each_entry_safe(f, ftmp, add_list, list) {
1831                 list_del(&f->list);
1832                 kfree(f);
1833         }
1834 }
1835
1836 /**
1837  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1838  * @vsi: ptr to the VSI
1839  *
1840  * Push any outstanding VSI filter changes through the AdminQ.
1841  *
1842  * Returns 0 or error value
1843  **/
1844 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1845 {
1846         struct list_head tmp_del_list, tmp_add_list;
1847         struct i40e_mac_filter *f, *ftmp, *fclone;
1848         bool promisc_forced_on = false;
1849         bool add_happened = false;
1850         int filter_list_len = 0;
1851         u32 changed_flags = 0;
1852         i40e_status aq_ret = 0;
1853         bool err_cond = false;
1854         int retval = 0;
1855         struct i40e_pf *pf;
1856         int num_add = 0;
1857         int num_del = 0;
1858         int aq_err = 0;
1859         u16 cmd_flags;
1860
1861         /* empty array typed pointers, kcalloc later */
1862         struct i40e_aqc_add_macvlan_element_data *add_list;
1863         struct i40e_aqc_remove_macvlan_element_data *del_list;
1864
1865         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1866                 usleep_range(1000, 2000);
1867         pf = vsi->back;
1868
1869         if (vsi->netdev) {
1870                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1871                 vsi->current_netdev_flags = vsi->netdev->flags;
1872         }
1873
1874         INIT_LIST_HEAD(&tmp_del_list);
1875         INIT_LIST_HEAD(&tmp_add_list);
1876
1877         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1878                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1879
1880                 spin_lock_bh(&vsi->mac_filter_list_lock);
1881                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1882                         if (!f->changed)
1883                                 continue;
1884
1885                         if (f->counter != 0)
1886                                 continue;
1887                         f->changed = false;
1888
1889                         /* Move the element into temporary del_list */
1890                         list_move_tail(&f->list, &tmp_del_list);
1891                 }
1892
1893                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1894                         if (!f->changed)
1895                                 continue;
1896
1897                         if (f->counter == 0)
1898                                 continue;
1899                         f->changed = false;
1900
1901                         /* Clone MAC filter entry and add into temporary list */
1902                         fclone = i40e_mac_filter_entry_clone(f);
1903                         if (!fclone) {
1904                                 err_cond = true;
1905                                 break;
1906                         }
1907                         list_add_tail(&fclone->list, &tmp_add_list);
1908                 }
1909
1910                 /* if failed to clone MAC filter entry - undo */
1911                 if (err_cond) {
1912                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1913                         i40e_undo_add_filter_entries(vsi);
1914                 }
1915                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1916
1917                 if (err_cond) {
1918                         i40e_cleanup_add_list(&tmp_add_list);
1919                         retval = -ENOMEM;
1920                         goto out;
1921                 }
1922         }
1923
1924         /* Now process 'del_list' outside the lock */
1925         if (!list_empty(&tmp_del_list)) {
1926                 int del_list_size;
1927
1928                 filter_list_len = pf->hw.aq.asq_buf_size /
1929                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1930                 del_list_size = filter_list_len *
1931                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1932                 del_list = kzalloc(del_list_size, GFP_ATOMIC);
1933                 if (!del_list) {
1934                         i40e_cleanup_add_list(&tmp_add_list);
1935
1936                         /* Undo VSI's MAC filter entry element updates */
1937                         spin_lock_bh(&vsi->mac_filter_list_lock);
1938                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1939                         i40e_undo_add_filter_entries(vsi);
1940                         spin_unlock_bh(&vsi->mac_filter_list_lock);
1941                         retval = -ENOMEM;
1942                         goto out;
1943                 }
1944
1945                 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1946                         cmd_flags = 0;
1947
1948                         /* add to delete list */
1949                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1950                         del_list[num_del].vlan_tag =
1951                                 cpu_to_le16((u16)(f->vlan ==
1952                                             I40E_VLAN_ANY ? 0 : f->vlan));
1953
1954                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1955                         del_list[num_del].flags = cmd_flags;
1956                         num_del++;
1957
1958                         /* flush a full buffer */
1959                         if (num_del == filter_list_len) {
1960                                 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1961                                                                 vsi->seid,
1962                                                                 del_list,
1963                                                                 num_del,
1964                                                                 NULL);
1965                                 aq_err = pf->hw.aq.asq_last_status;
1966                                 num_del = 0;
1967                                 memset(del_list, 0, del_list_size);
1968
1969                                 if (aq_ret && aq_err != I40E_AQ_RC_ENOENT) {
1970                                         retval = -EIO;
1971                                         dev_err(&pf->pdev->dev,
1972                                                 "ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1973                                                 i40e_stat_str(&pf->hw, aq_ret),
1974                                                 i40e_aq_str(&pf->hw, aq_err));
1975                                 }
1976                         }
1977                         /* Release memory for MAC filter entries which were
1978                          * synced up with HW.
1979                          */
1980                         list_del(&f->list);
1981                         kfree(f);
1982                 }
1983
1984                 if (num_del) {
1985                         aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1986                                                         del_list, num_del,
1987                                                         NULL);
1988                         aq_err = pf->hw.aq.asq_last_status;
1989                         num_del = 0;
1990
1991                         if (aq_ret && aq_err != I40E_AQ_RC_ENOENT)
1992                                 dev_info(&pf->pdev->dev,
1993                                          "ignoring delete macvlan error, err %s aq_err %s\n",
1994                                          i40e_stat_str(&pf->hw, aq_ret),
1995                                          i40e_aq_str(&pf->hw, aq_err));
1996                 }
1997
1998                 kfree(del_list);
1999                 del_list = NULL;
2000         }
2001
2002         if (!list_empty(&tmp_add_list)) {
2003                 int add_list_size;
2004
2005                 /* do all the adds now */
2006                 filter_list_len = pf->hw.aq.asq_buf_size /
2007                                sizeof(struct i40e_aqc_add_macvlan_element_data),
2008                 add_list_size = filter_list_len *
2009                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2010                 add_list = kzalloc(add_list_size, GFP_ATOMIC);
2011                 if (!add_list) {
2012                         /* Purge element from temporary lists */
2013                         i40e_cleanup_add_list(&tmp_add_list);
2014
2015                         /* Undo add filter entries from VSI MAC filter list */
2016                         spin_lock_bh(&vsi->mac_filter_list_lock);
2017                         i40e_undo_add_filter_entries(vsi);
2018                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2019                         retval = -ENOMEM;
2020                         goto out;
2021                 }
2022
2023                 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2024
2025                         add_happened = true;
2026                         cmd_flags = 0;
2027
2028                         /* add to add array */
2029                         ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2030                         add_list[num_add].vlan_tag =
2031                                 cpu_to_le16(
2032                                  (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2033                         add_list[num_add].queue_number = 0;
2034
2035                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2036                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2037                         num_add++;
2038
2039                         /* flush a full buffer */
2040                         if (num_add == filter_list_len) {
2041                                 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2042                                                              add_list, num_add,
2043                                                              NULL);
2044                                 aq_err = pf->hw.aq.asq_last_status;
2045                                 num_add = 0;
2046
2047                                 if (aq_ret)
2048                                         break;
2049                                 memset(add_list, 0, add_list_size);
2050                         }
2051                         /* Entries from tmp_add_list were cloned from MAC
2052                          * filter list, hence clean those cloned entries
2053                          */
2054                         list_del(&f->list);
2055                         kfree(f);
2056                 }
2057
2058                 if (num_add) {
2059                         aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2060                                                      add_list, num_add, NULL);
2061                         aq_err = pf->hw.aq.asq_last_status;
2062                         num_add = 0;
2063                 }
2064                 kfree(add_list);
2065                 add_list = NULL;
2066
2067                 if (add_happened && aq_ret && aq_err != I40E_AQ_RC_EINVAL) {
2068                         retval = i40e_aq_rc_to_posix(aq_ret, aq_err);
2069                         dev_info(&pf->pdev->dev,
2070                                  "add filter failed, err %s aq_err %s\n",
2071                                  i40e_stat_str(&pf->hw, aq_ret),
2072                                  i40e_aq_str(&pf->hw, aq_err));
2073                         if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2074                             !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2075                                       &vsi->state)) {
2076                                 promisc_forced_on = true;
2077                                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2078                                         &vsi->state);
2079                                 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2080                         }
2081                 }
2082         }
2083
2084         /* if the VF is not trusted do not do promisc */
2085         if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2086                 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2087                 goto out;
2088         }
2089
2090         /* check for changes in promiscuous modes */
2091         if (changed_flags & IFF_ALLMULTI) {
2092                 bool cur_multipromisc;
2093
2094                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2095                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2096                                                                vsi->seid,
2097                                                                cur_multipromisc,
2098                                                                NULL);
2099                 if (aq_ret) {
2100                         retval = i40e_aq_rc_to_posix(aq_ret,
2101                                                      pf->hw.aq.asq_last_status);
2102                         dev_info(&pf->pdev->dev,
2103                                  "set multi promisc failed, err %s aq_err %s\n",
2104                                  i40e_stat_str(&pf->hw, aq_ret),
2105                                  i40e_aq_str(&pf->hw,
2106                                              pf->hw.aq.asq_last_status));
2107                 }
2108         }
2109         if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2110                 bool cur_promisc;
2111
2112                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2113                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2114                                         &vsi->state));
2115                 if ((vsi->type == I40E_VSI_MAIN) &&
2116                     (pf->lan_veb != I40E_NO_VEB) &&
2117                     !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2118                         /* set defport ON for Main VSI instead of true promisc
2119                          * this way we will get all unicast/multicast and VLAN
2120                          * promisc behavior but will not get VF or VMDq traffic
2121                          * replicated on the Main VSI.
2122                          */
2123                         if (pf->cur_promisc != cur_promisc) {
2124                                 pf->cur_promisc = cur_promisc;
2125                                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2126                         }
2127                 } else {
2128                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2129                                                           &vsi->back->hw,
2130                                                           vsi->seid,
2131                                                           cur_promisc, NULL);
2132                         if (aq_ret) {
2133                                 retval =
2134                                 i40e_aq_rc_to_posix(aq_ret,
2135                                                     pf->hw.aq.asq_last_status);
2136                                 dev_info(&pf->pdev->dev,
2137                                          "set unicast promisc failed, err %d, aq_err %d\n",
2138                                          aq_ret, pf->hw.aq.asq_last_status);
2139                         }
2140                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2141                                                           &vsi->back->hw,
2142                                                           vsi->seid,
2143                                                           cur_promisc, NULL);
2144                         if (aq_ret) {
2145                                 retval =
2146                                 i40e_aq_rc_to_posix(aq_ret,
2147                                                     pf->hw.aq.asq_last_status);
2148                                 dev_info(&pf->pdev->dev,
2149                                          "set multicast promisc failed, err %d, aq_err %d\n",
2150                                          aq_ret, pf->hw.aq.asq_last_status);
2151                         }
2152                 }
2153                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2154                                                    vsi->seid,
2155                                                    cur_promisc, NULL);
2156                 if (aq_ret) {
2157                         retval = i40e_aq_rc_to_posix(aq_ret,
2158                                                      pf->hw.aq.asq_last_status);
2159                         dev_info(&pf->pdev->dev,
2160                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2161                                  i40e_stat_str(&pf->hw, aq_ret),
2162                                  i40e_aq_str(&pf->hw,
2163                                              pf->hw.aq.asq_last_status));
2164                 }
2165         }
2166 out:
2167         /* if something went wrong then set the changed flag so we try again */
2168         if (retval)
2169                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2170
2171         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2172         return retval;
2173 }
2174
2175 /**
2176  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2177  * @pf: board private structure
2178  **/
2179 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2180 {
2181         int v;
2182
2183         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2184                 return;
2185         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2186
2187         for (v = 0; v < pf->num_alloc_vsi; v++) {
2188                 if (pf->vsi[v] &&
2189                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2190                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2191
2192                         if (ret) {
2193                                 /* come back and try again later */
2194                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2195                                 break;
2196                         }
2197                 }
2198         }
2199 }
2200
2201 /**
2202  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2203  * @netdev: network interface device structure
2204  * @new_mtu: new value for maximum frame size
2205  *
2206  * Returns 0 on success, negative on failure
2207  **/
2208 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2209 {
2210         struct i40e_netdev_priv *np = netdev_priv(netdev);
2211         int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2212         struct i40e_vsi *vsi = np->vsi;
2213
2214         /* MTU < 68 is an error and causes problems on some kernels */
2215         if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2216                 return -EINVAL;
2217
2218         netdev_info(netdev, "changing MTU from %d to %d\n",
2219                     netdev->mtu, new_mtu);
2220         netdev->mtu = new_mtu;
2221         if (netif_running(netdev))
2222                 i40e_vsi_reinit_locked(vsi);
2223         i40e_notify_client_of_l2_param_changes(vsi);
2224         return 0;
2225 }
2226
2227 /**
2228  * i40e_ioctl - Access the hwtstamp interface
2229  * @netdev: network interface device structure
2230  * @ifr: interface request data
2231  * @cmd: ioctl command
2232  **/
2233 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2234 {
2235         struct i40e_netdev_priv *np = netdev_priv(netdev);
2236         struct i40e_pf *pf = np->vsi->back;
2237
2238         switch (cmd) {
2239         case SIOCGHWTSTAMP:
2240                 return i40e_ptp_get_ts_config(pf, ifr);
2241         case SIOCSHWTSTAMP:
2242                 return i40e_ptp_set_ts_config(pf, ifr);
2243         default:
2244                 return -EOPNOTSUPP;
2245         }
2246 }
2247
2248 /**
2249  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2250  * @vsi: the vsi being adjusted
2251  **/
2252 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2253 {
2254         struct i40e_vsi_context ctxt;
2255         i40e_status ret;
2256
2257         if ((vsi->info.valid_sections &
2258              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2259             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2260                 return;  /* already enabled */
2261
2262         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2263         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2264                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2265
2266         ctxt.seid = vsi->seid;
2267         ctxt.info = vsi->info;
2268         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2269         if (ret) {
2270                 dev_info(&vsi->back->pdev->dev,
2271                          "update vlan stripping failed, err %s aq_err %s\n",
2272                          i40e_stat_str(&vsi->back->hw, ret),
2273                          i40e_aq_str(&vsi->back->hw,
2274                                      vsi->back->hw.aq.asq_last_status));
2275         }
2276 }
2277
2278 /**
2279  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2280  * @vsi: the vsi being adjusted
2281  **/
2282 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2283 {
2284         struct i40e_vsi_context ctxt;
2285         i40e_status ret;
2286
2287         if ((vsi->info.valid_sections &
2288              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2289             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2290              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2291                 return;  /* already disabled */
2292
2293         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2294         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2295                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2296
2297         ctxt.seid = vsi->seid;
2298         ctxt.info = vsi->info;
2299         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2300         if (ret) {
2301                 dev_info(&vsi->back->pdev->dev,
2302                          "update vlan stripping failed, err %s aq_err %s\n",
2303                          i40e_stat_str(&vsi->back->hw, ret),
2304                          i40e_aq_str(&vsi->back->hw,
2305                                      vsi->back->hw.aq.asq_last_status));
2306         }
2307 }
2308
2309 /**
2310  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2311  * @netdev: network interface to be adjusted
2312  * @features: netdev features to test if VLAN offload is enabled or not
2313  **/
2314 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2315 {
2316         struct i40e_netdev_priv *np = netdev_priv(netdev);
2317         struct i40e_vsi *vsi = np->vsi;
2318
2319         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2320                 i40e_vlan_stripping_enable(vsi);
2321         else
2322                 i40e_vlan_stripping_disable(vsi);
2323 }
2324
2325 /**
2326  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2327  * @vsi: the vsi being configured
2328  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2329  **/
2330 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2331 {
2332         struct i40e_mac_filter *f, *add_f;
2333         bool is_netdev, is_vf;
2334
2335         is_vf = (vsi->type == I40E_VSI_SRIOV);
2336         is_netdev = !!(vsi->netdev);
2337
2338         /* Locked once because all functions invoked below iterates list*/
2339         spin_lock_bh(&vsi->mac_filter_list_lock);
2340
2341         if (is_netdev) {
2342                 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2343                                         is_vf, is_netdev);
2344                 if (!add_f) {
2345                         dev_info(&vsi->back->pdev->dev,
2346                                  "Could not add vlan filter %d for %pM\n",
2347                                  vid, vsi->netdev->dev_addr);
2348                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2349                         return -ENOMEM;
2350                 }
2351         }
2352
2353         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2354                 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2355                 if (!add_f) {
2356                         dev_info(&vsi->back->pdev->dev,
2357                                  "Could not add vlan filter %d for %pM\n",
2358                                  vid, f->macaddr);
2359                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2360                         return -ENOMEM;
2361                 }
2362         }
2363
2364         /* Now if we add a vlan tag, make sure to check if it is the first
2365          * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2366          * with 0, so we now accept untagged and specified tagged traffic
2367          * (and not any taged and untagged)
2368          */
2369         if (vid > 0) {
2370                 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2371                                                   I40E_VLAN_ANY,
2372                                                   is_vf, is_netdev)) {
2373                         i40e_del_filter(vsi, vsi->netdev->dev_addr,
2374                                         I40E_VLAN_ANY, is_vf, is_netdev);
2375                         add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2376                                                 is_vf, is_netdev);
2377                         if (!add_f) {
2378                                 dev_info(&vsi->back->pdev->dev,
2379                                          "Could not add filter 0 for %pM\n",
2380                                          vsi->netdev->dev_addr);
2381                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2382                                 return -ENOMEM;
2383                         }
2384                 }
2385         }
2386
2387         /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2388         if (vid > 0 && !vsi->info.pvid) {
2389                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2390                         if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2391                                               is_vf, is_netdev))
2392                                 continue;
2393                         i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2394                                         is_vf, is_netdev);
2395                         add_f = i40e_add_filter(vsi, f->macaddr,
2396                                                 0, is_vf, is_netdev);
2397                         if (!add_f) {
2398                                 dev_info(&vsi->back->pdev->dev,
2399                                          "Could not add filter 0 for %pM\n",
2400                                         f->macaddr);
2401                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2402                                 return -ENOMEM;
2403                         }
2404                 }
2405         }
2406
2407         spin_unlock_bh(&vsi->mac_filter_list_lock);
2408
2409         /* schedule our worker thread which will take care of
2410          * applying the new filter changes
2411          */
2412         i40e_service_event_schedule(vsi->back);
2413         return 0;
2414 }
2415
2416 /**
2417  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2418  * @vsi: the vsi being configured
2419  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2420  *
2421  * Return: 0 on success or negative otherwise
2422  **/
2423 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2424 {
2425         struct net_device *netdev = vsi->netdev;
2426         struct i40e_mac_filter *f, *add_f;
2427         bool is_vf, is_netdev;
2428         int filter_count = 0;
2429
2430         is_vf = (vsi->type == I40E_VSI_SRIOV);
2431         is_netdev = !!(netdev);
2432
2433         /* Locked once because all functions invoked below iterates list */
2434         spin_lock_bh(&vsi->mac_filter_list_lock);
2435
2436         if (is_netdev)
2437                 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2438
2439         list_for_each_entry(f, &vsi->mac_filter_list, list)
2440                 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2441
2442         /* go through all the filters for this VSI and if there is only
2443          * vid == 0 it means there are no other filters, so vid 0 must
2444          * be replaced with -1. This signifies that we should from now
2445          * on accept any traffic (with any tag present, or untagged)
2446          */
2447         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2448                 if (is_netdev) {
2449                         if (f->vlan &&
2450                             ether_addr_equal(netdev->dev_addr, f->macaddr))
2451                                 filter_count++;
2452                 }
2453
2454                 if (f->vlan)
2455                         filter_count++;
2456         }
2457
2458         if (!filter_count && is_netdev) {
2459                 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2460                 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2461                                     is_vf, is_netdev);
2462                 if (!f) {
2463                         dev_info(&vsi->back->pdev->dev,
2464                                  "Could not add filter %d for %pM\n",
2465                                  I40E_VLAN_ANY, netdev->dev_addr);
2466                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2467                         return -ENOMEM;
2468                 }
2469         }
2470
2471         if (!filter_count) {
2472                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2473                         i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2474                         add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2475                                                 is_vf, is_netdev);
2476                         if (!add_f) {
2477                                 dev_info(&vsi->back->pdev->dev,
2478                                          "Could not add filter %d for %pM\n",
2479                                          I40E_VLAN_ANY, f->macaddr);
2480                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2481                                 return -ENOMEM;
2482                         }
2483                 }
2484         }
2485
2486         spin_unlock_bh(&vsi->mac_filter_list_lock);
2487
2488         /* schedule our worker thread which will take care of
2489          * applying the new filter changes
2490          */
2491         i40e_service_event_schedule(vsi->back);
2492         return 0;
2493 }
2494
2495 /**
2496  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2497  * @netdev: network interface to be adjusted
2498  * @vid: vlan id to be added
2499  *
2500  * net_device_ops implementation for adding vlan ids
2501  **/
2502 #ifdef I40E_FCOE
2503 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2504                          __always_unused __be16 proto, u16 vid)
2505 #else
2506 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2507                                 __always_unused __be16 proto, u16 vid)
2508 #endif
2509 {
2510         struct i40e_netdev_priv *np = netdev_priv(netdev);
2511         struct i40e_vsi *vsi = np->vsi;
2512         int ret = 0;
2513
2514         if (vid > 4095)
2515                 return -EINVAL;
2516
2517         netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2518
2519         /* If the network stack called us with vid = 0 then
2520          * it is asking to receive priority tagged packets with
2521          * vlan id 0.  Our HW receives them by default when configured
2522          * to receive untagged packets so there is no need to add an
2523          * extra filter for vlan 0 tagged packets.
2524          */
2525         if (vid)
2526                 ret = i40e_vsi_add_vlan(vsi, vid);
2527
2528         if (!ret && (vid < VLAN_N_VID))
2529                 set_bit(vid, vsi->active_vlans);
2530
2531         return ret;
2532 }
2533
2534 /**
2535  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2536  * @netdev: network interface to be adjusted
2537  * @vid: vlan id to be removed
2538  *
2539  * net_device_ops implementation for removing vlan ids
2540  **/
2541 #ifdef I40E_FCOE
2542 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2543                           __always_unused __be16 proto, u16 vid)
2544 #else
2545 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2546                                  __always_unused __be16 proto, u16 vid)
2547 #endif
2548 {
2549         struct i40e_netdev_priv *np = netdev_priv(netdev);
2550         struct i40e_vsi *vsi = np->vsi;
2551
2552         netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2553
2554         /* return code is ignored as there is nothing a user
2555          * can do about failure to remove and a log message was
2556          * already printed from the other function
2557          */
2558         i40e_vsi_kill_vlan(vsi, vid);
2559
2560         clear_bit(vid, vsi->active_vlans);
2561
2562         return 0;
2563 }
2564
2565 /**
2566  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2567  * @vsi: the vsi being brought back up
2568  **/
2569 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2570 {
2571         u16 vid;
2572
2573         if (!vsi->netdev)
2574                 return;
2575
2576         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2577
2578         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2579                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2580                                      vid);
2581 }
2582
2583 /**
2584  * i40e_vsi_add_pvid - Add pvid for the VSI
2585  * @vsi: the vsi being adjusted
2586  * @vid: the vlan id to set as a PVID
2587  **/
2588 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2589 {
2590         struct i40e_vsi_context ctxt;
2591         i40e_status ret;
2592
2593         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2594         vsi->info.pvid = cpu_to_le16(vid);
2595         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2596                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2597                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2598
2599         ctxt.seid = vsi->seid;
2600         ctxt.info = vsi->info;
2601         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2602         if (ret) {
2603                 dev_info(&vsi->back->pdev->dev,
2604                          "add pvid failed, err %s aq_err %s\n",
2605                          i40e_stat_str(&vsi->back->hw, ret),
2606                          i40e_aq_str(&vsi->back->hw,
2607                                      vsi->back->hw.aq.asq_last_status));
2608                 return -ENOENT;
2609         }
2610
2611         return 0;
2612 }
2613
2614 /**
2615  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2616  * @vsi: the vsi being adjusted
2617  *
2618  * Just use the vlan_rx_register() service to put it back to normal
2619  **/
2620 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2621 {
2622         i40e_vlan_stripping_disable(vsi);
2623
2624         vsi->info.pvid = 0;
2625 }
2626
2627 /**
2628  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2629  * @vsi: ptr to the VSI
2630  *
2631  * If this function returns with an error, then it's possible one or
2632  * more of the rings is populated (while the rest are not).  It is the
2633  * callers duty to clean those orphaned rings.
2634  *
2635  * Return 0 on success, negative on failure
2636  **/
2637 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2638 {
2639         int i, err = 0;
2640
2641         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2642                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2643
2644         return err;
2645 }
2646
2647 /**
2648  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2649  * @vsi: ptr to the VSI
2650  *
2651  * Free VSI's transmit software resources
2652  **/
2653 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2654 {
2655         int i;
2656
2657         if (!vsi->tx_rings)
2658                 return;
2659
2660         for (i = 0; i < vsi->num_queue_pairs; i++)
2661                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2662                         i40e_free_tx_resources(vsi->tx_rings[i]);
2663 }
2664
2665 /**
2666  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2667  * @vsi: ptr to the VSI
2668  *
2669  * If this function returns with an error, then it's possible one or
2670  * more of the rings is populated (while the rest are not).  It is the
2671  * callers duty to clean those orphaned rings.
2672  *
2673  * Return 0 on success, negative on failure
2674  **/
2675 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2676 {
2677         int i, err = 0;
2678
2679         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2680                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2681 #ifdef I40E_FCOE
2682         i40e_fcoe_setup_ddp_resources(vsi);
2683 #endif
2684         return err;
2685 }
2686
2687 /**
2688  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2689  * @vsi: ptr to the VSI
2690  *
2691  * Free all receive software resources
2692  **/
2693 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2694 {
2695         int i;
2696
2697         if (!vsi->rx_rings)
2698                 return;
2699
2700         for (i = 0; i < vsi->num_queue_pairs; i++)
2701                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2702                         i40e_free_rx_resources(vsi->rx_rings[i]);
2703 #ifdef I40E_FCOE
2704         i40e_fcoe_free_ddp_resources(vsi);
2705 #endif
2706 }
2707
2708 /**
2709  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2710  * @ring: The Tx ring to configure
2711  *
2712  * This enables/disables XPS for a given Tx descriptor ring
2713  * based on the TCs enabled for the VSI that ring belongs to.
2714  **/
2715 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2716 {
2717         struct i40e_vsi *vsi = ring->vsi;
2718         cpumask_var_t mask;
2719
2720         if (!ring->q_vector || !ring->netdev)
2721                 return;
2722
2723         /* Single TC mode enable XPS */
2724         if (vsi->tc_config.numtc <= 1) {
2725                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2726                         netif_set_xps_queue(ring->netdev,
2727                                             &ring->q_vector->affinity_mask,
2728                                             ring->queue_index);
2729         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2730                 /* Disable XPS to allow selection based on TC */
2731                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2732                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2733                 free_cpumask_var(mask);
2734         }
2735
2736         /* schedule our worker thread which will take care of
2737          * applying the new filter changes
2738          */
2739         i40e_service_event_schedule(vsi->back);
2740 }
2741
2742 /**
2743  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2744  * @ring: The Tx ring to configure
2745  *
2746  * Configure the Tx descriptor ring in the HMC context.
2747  **/
2748 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2749 {
2750         struct i40e_vsi *vsi = ring->vsi;
2751         u16 pf_q = vsi->base_queue + ring->queue_index;
2752         struct i40e_hw *hw = &vsi->back->hw;
2753         struct i40e_hmc_obj_txq tx_ctx;
2754         i40e_status err = 0;
2755         u32 qtx_ctl = 0;
2756
2757         /* some ATR related tx ring init */
2758         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2759                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2760                 ring->atr_count = 0;
2761         } else {
2762                 ring->atr_sample_rate = 0;
2763         }
2764
2765         /* configure XPS */
2766         i40e_config_xps_tx_ring(ring);
2767
2768         /* clear the context structure first */
2769         memset(&tx_ctx, 0, sizeof(tx_ctx));
2770
2771         tx_ctx.new_context = 1;
2772         tx_ctx.base = (ring->dma / 128);
2773         tx_ctx.qlen = ring->count;
2774         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2775                                                I40E_FLAG_FD_ATR_ENABLED));
2776 #ifdef I40E_FCOE
2777         tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2778 #endif
2779         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2780         /* FDIR VSI tx ring can still use RS bit and writebacks */
2781         if (vsi->type != I40E_VSI_FDIR)
2782                 tx_ctx.head_wb_ena = 1;
2783         tx_ctx.head_wb_addr = ring->dma +
2784                               (ring->count * sizeof(struct i40e_tx_desc));
2785
2786         /* As part of VSI creation/update, FW allocates certain
2787          * Tx arbitration queue sets for each TC enabled for
2788          * the VSI. The FW returns the handles to these queue
2789          * sets as part of the response buffer to Add VSI,
2790          * Update VSI, etc. AQ commands. It is expected that
2791          * these queue set handles be associated with the Tx
2792          * queues by the driver as part of the TX queue context
2793          * initialization. This has to be done regardless of
2794          * DCB as by default everything is mapped to TC0.
2795          */
2796         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2797         tx_ctx.rdylist_act = 0;
2798
2799         /* clear the context in the HMC */
2800         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2801         if (err) {
2802                 dev_info(&vsi->back->pdev->dev,
2803                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2804                          ring->queue_index, pf_q, err);
2805                 return -ENOMEM;
2806         }
2807
2808         /* set the context in the HMC */
2809         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2810         if (err) {
2811                 dev_info(&vsi->back->pdev->dev,
2812                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2813                          ring->queue_index, pf_q, err);
2814                 return -ENOMEM;
2815         }
2816
2817         /* Now associate this queue with this PCI function */
2818         if (vsi->type == I40E_VSI_VMDQ2) {
2819                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2820                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2821                            I40E_QTX_CTL_VFVM_INDX_MASK;
2822         } else {
2823                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2824         }
2825
2826         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2827                     I40E_QTX_CTL_PF_INDX_MASK);
2828         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2829         i40e_flush(hw);
2830
2831         /* cache tail off for easier writes later */
2832         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2833
2834         return 0;
2835 }
2836
2837 /**
2838  * i40e_configure_rx_ring - Configure a receive ring context
2839  * @ring: The Rx ring to configure
2840  *
2841  * Configure the Rx descriptor ring in the HMC context.
2842  **/
2843 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2844 {
2845         struct i40e_vsi *vsi = ring->vsi;
2846         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2847         u16 pf_q = vsi->base_queue + ring->queue_index;
2848         struct i40e_hw *hw = &vsi->back->hw;
2849         struct i40e_hmc_obj_rxq rx_ctx;
2850         i40e_status err = 0;
2851
2852         ring->state = 0;
2853
2854         /* clear the context structure first */
2855         memset(&rx_ctx, 0, sizeof(rx_ctx));
2856
2857         ring->rx_buf_len = vsi->rx_buf_len;
2858         ring->rx_hdr_len = vsi->rx_hdr_len;
2859
2860         rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2861         rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2862
2863         rx_ctx.base = (ring->dma / 128);
2864         rx_ctx.qlen = ring->count;
2865
2866         if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2867                 set_ring_16byte_desc_enabled(ring);
2868                 rx_ctx.dsize = 0;
2869         } else {
2870                 rx_ctx.dsize = 1;
2871         }
2872
2873         rx_ctx.dtype = vsi->dtype;
2874         if (vsi->dtype) {
2875                 set_ring_ps_enabled(ring);
2876                 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2877                                   I40E_RX_SPLIT_IP      |
2878                                   I40E_RX_SPLIT_TCP_UDP |
2879                                   I40E_RX_SPLIT_SCTP;
2880         } else {
2881                 rx_ctx.hsplit_0 = 0;
2882         }
2883
2884         rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2885                                   (chain_len * ring->rx_buf_len));
2886         if (hw->revision_id == 0)
2887                 rx_ctx.lrxqthresh = 0;
2888         else
2889                 rx_ctx.lrxqthresh = 2;
2890         rx_ctx.crcstrip = 1;
2891         rx_ctx.l2tsel = 1;
2892         /* this controls whether VLAN is stripped from inner headers */
2893         rx_ctx.showiv = 0;
2894 #ifdef I40E_FCOE
2895         rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2896 #endif
2897         /* set the prefena field to 1 because the manual says to */
2898         rx_ctx.prefena = 1;
2899
2900         /* clear the context in the HMC */
2901         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2902         if (err) {
2903                 dev_info(&vsi->back->pdev->dev,
2904                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2905                          ring->queue_index, pf_q, err);
2906                 return -ENOMEM;
2907         }
2908
2909         /* set the context in the HMC */
2910         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2911         if (err) {
2912                 dev_info(&vsi->back->pdev->dev,
2913                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2914                          ring->queue_index, pf_q, err);
2915                 return -ENOMEM;
2916         }
2917
2918         /* cache tail for quicker writes, and clear the reg before use */
2919         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2920         writel(0, ring->tail);
2921
2922         if (ring_is_ps_enabled(ring)) {
2923                 i40e_alloc_rx_headers(ring);
2924                 i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2925         } else {
2926                 i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2927         }
2928
2929         return 0;
2930 }
2931
2932 /**
2933  * i40e_vsi_configure_tx - Configure the VSI for Tx
2934  * @vsi: VSI structure describing this set of rings and resources
2935  *
2936  * Configure the Tx VSI for operation.
2937  **/
2938 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2939 {
2940         int err = 0;
2941         u16 i;
2942
2943         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2944                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2945
2946         return err;
2947 }
2948
2949 /**
2950  * i40e_vsi_configure_rx - Configure the VSI for Rx
2951  * @vsi: the VSI being configured
2952  *
2953  * Configure the Rx VSI for operation.
2954  **/
2955 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2956 {
2957         int err = 0;
2958         u16 i;
2959
2960         if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2961                 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2962                                + ETH_FCS_LEN + VLAN_HLEN;
2963         else
2964                 vsi->max_frame = I40E_RXBUFFER_2048;
2965
2966         /* figure out correct receive buffer length */
2967         switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2968                                     I40E_FLAG_RX_PS_ENABLED)) {
2969         case I40E_FLAG_RX_1BUF_ENABLED:
2970                 vsi->rx_hdr_len = 0;
2971                 vsi->rx_buf_len = vsi->max_frame;
2972                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2973                 break;
2974         case I40E_FLAG_RX_PS_ENABLED:
2975                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2976                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2977                 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2978                 break;
2979         default:
2980                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2981                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2982                 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2983                 break;
2984         }
2985
2986 #ifdef I40E_FCOE
2987         /* setup rx buffer for FCoE */
2988         if ((vsi->type == I40E_VSI_FCOE) &&
2989             (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2990                 vsi->rx_hdr_len = 0;
2991                 vsi->rx_buf_len = I40E_RXBUFFER_3072;
2992                 vsi->max_frame = I40E_RXBUFFER_3072;
2993                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2994         }
2995
2996 #endif /* I40E_FCOE */
2997         /* round up for the chip's needs */
2998         vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2999                                 BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT));
3000         vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
3001                                 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3002
3003         /* set up individual rings */
3004         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3005                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3006
3007         return err;
3008 }
3009
3010 /**
3011  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3012  * @vsi: ptr to the VSI
3013  **/
3014 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3015 {
3016         struct i40e_ring *tx_ring, *rx_ring;
3017         u16 qoffset, qcount;
3018         int i, n;
3019
3020         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3021                 /* Reset the TC information */
3022                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3023                         rx_ring = vsi->rx_rings[i];
3024                         tx_ring = vsi->tx_rings[i];
3025                         rx_ring->dcb_tc = 0;
3026                         tx_ring->dcb_tc = 0;
3027                 }
3028         }
3029
3030         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3031                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3032                         continue;
3033
3034                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3035                 qcount = vsi->tc_config.tc_info[n].qcount;
3036                 for (i = qoffset; i < (qoffset + qcount); i++) {
3037                         rx_ring = vsi->rx_rings[i];
3038                         tx_ring = vsi->tx_rings[i];
3039                         rx_ring->dcb_tc = n;
3040                         tx_ring->dcb_tc = n;
3041                 }
3042         }
3043 }
3044
3045 /**
3046  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3047  * @vsi: ptr to the VSI
3048  **/
3049 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3050 {
3051         if (vsi->netdev)
3052                 i40e_set_rx_mode(vsi->netdev);
3053 }
3054
3055 /**
3056  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3057  * @vsi: Pointer to the targeted VSI
3058  *
3059  * This function replays the hlist on the hw where all the SB Flow Director
3060  * filters were saved.
3061  **/
3062 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3063 {
3064         struct i40e_fdir_filter *filter;
3065         struct i40e_pf *pf = vsi->back;
3066         struct hlist_node *node;
3067
3068         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3069                 return;
3070
3071         hlist_for_each_entry_safe(filter, node,
3072                                   &pf->fdir_filter_list, fdir_node) {
3073                 i40e_add_del_fdir(vsi, filter, true);
3074         }
3075 }
3076
3077 /**
3078  * i40e_vsi_configure - Set up the VSI for action
3079  * @vsi: the VSI being configured
3080  **/
3081 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3082 {
3083         int err;
3084
3085         i40e_set_vsi_rx_mode(vsi);
3086         i40e_restore_vlan(vsi);
3087         i40e_vsi_config_dcb_rings(vsi);
3088         err = i40e_vsi_configure_tx(vsi);
3089         if (!err)
3090                 err = i40e_vsi_configure_rx(vsi);
3091
3092         return err;
3093 }
3094
3095 /**
3096  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3097  * @vsi: the VSI being configured
3098  **/
3099 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3100 {
3101         struct i40e_pf *pf = vsi->back;
3102         struct i40e_hw *hw = &pf->hw;
3103         u16 vector;
3104         int i, q;
3105         u32 qp;
3106
3107         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3108          * and PFINT_LNKLSTn registers, e.g.:
3109          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3110          */
3111         qp = vsi->base_queue;
3112         vector = vsi->base_vector;
3113         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3114                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3115
3116                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3117                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3118                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3119                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3120                      q_vector->rx.itr);
3121                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3122                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3123                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3124                      q_vector->tx.itr);
3125                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3126                      INTRL_USEC_TO_REG(vsi->int_rate_limit));
3127
3128                 /* Linked list for the queuepairs assigned to this vector */
3129                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3130                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3131                         u32 val;
3132
3133                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3134                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3135                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3136                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3137                               (I40E_QUEUE_TYPE_TX
3138                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3139
3140                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3141
3142                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3143                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3144                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3145                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3146                               (I40E_QUEUE_TYPE_RX
3147                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3148
3149                         /* Terminate the linked list */
3150                         if (q == (q_vector->num_ringpairs - 1))
3151                                 val |= (I40E_QUEUE_END_OF_LIST
3152                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3153
3154                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3155                         qp++;
3156                 }
3157         }
3158
3159         i40e_flush(hw);
3160 }
3161
3162 /**
3163  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3164  * @hw: ptr to the hardware info
3165  **/
3166 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3167 {
3168         struct i40e_hw *hw = &pf->hw;
3169         u32 val;
3170
3171         /* clear things first */
3172         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3173         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3174
3175         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3176               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3177               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3178               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3179               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3180               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3181               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3182               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3183
3184         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3185                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3186
3187         if (pf->flags & I40E_FLAG_PTP)
3188                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3189
3190         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3191
3192         /* SW_ITR_IDX = 0, but don't change INTENA */
3193         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3194                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3195
3196         /* OTHER_ITR_IDX = 0 */
3197         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3198 }
3199
3200 /**
3201  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3202  * @vsi: the VSI being configured
3203  **/
3204 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3205 {
3206         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3207         struct i40e_pf *pf = vsi->back;
3208         struct i40e_hw *hw = &pf->hw;
3209         u32 val;
3210
3211         /* set the ITR configuration */
3212         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3213         q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3214         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3215         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3216         q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3217         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3218         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3219
3220         i40e_enable_misc_int_causes(pf);
3221
3222         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3223         wr32(hw, I40E_PFINT_LNKLST0, 0);
3224
3225         /* Associate the queue pair to the vector and enable the queue int */
3226         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3227               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3228               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3229
3230         wr32(hw, I40E_QINT_RQCTL(0), val);
3231
3232         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3233               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3234               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3235
3236         wr32(hw, I40E_QINT_TQCTL(0), val);
3237         i40e_flush(hw);
3238 }
3239
3240 /**
3241  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3242  * @pf: board private structure
3243  **/
3244 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3245 {
3246         struct i40e_hw *hw = &pf->hw;
3247
3248         wr32(hw, I40E_PFINT_DYN_CTL0,
3249              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3250         i40e_flush(hw);
3251 }
3252
3253 /**
3254  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3255  * @pf: board private structure
3256  * @clearpba: true when all pending interrupt events should be cleared
3257  **/
3258 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3259 {
3260         struct i40e_hw *hw = &pf->hw;
3261         u32 val;
3262
3263         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3264               (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3265               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3266
3267         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3268         i40e_flush(hw);
3269 }
3270
3271 /**
3272  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3273  * @irq: interrupt number
3274  * @data: pointer to a q_vector
3275  **/
3276 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3277 {
3278         struct i40e_q_vector *q_vector = data;
3279
3280         if (!q_vector->tx.ring && !q_vector->rx.ring)
3281                 return IRQ_HANDLED;
3282
3283         napi_schedule_irqoff(&q_vector->napi);
3284
3285         return IRQ_HANDLED;
3286 }
3287
3288 /**
3289  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3290  * @vsi: the VSI being configured
3291  * @basename: name for the vector
3292  *
3293  * Allocates MSI-X vectors and requests interrupts from the kernel.
3294  **/
3295 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3296 {
3297         int q_vectors = vsi->num_q_vectors;
3298         struct i40e_pf *pf = vsi->back;
3299         int base = vsi->base_vector;
3300         int rx_int_idx = 0;
3301         int tx_int_idx = 0;
3302         int vector, err;
3303
3304         for (vector = 0; vector < q_vectors; vector++) {
3305                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3306
3307                 if (q_vector->tx.ring && q_vector->rx.ring) {
3308                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3309                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3310                         tx_int_idx++;
3311                 } else if (q_vector->rx.ring) {
3312                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3313                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3314                 } else if (q_vector->tx.ring) {
3315                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3316                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3317                 } else {
3318                         /* skip this unused q_vector */
3319                         continue;
3320                 }
3321                 err = request_irq(pf->msix_entries[base + vector].vector,
3322                                   vsi->irq_handler,
3323                                   0,
3324                                   q_vector->name,
3325                                   q_vector);
3326                 if (err) {
3327                         dev_info(&pf->pdev->dev,
3328                                  "MSIX request_irq failed, error: %d\n", err);
3329                         goto free_queue_irqs;
3330                 }
3331                 /* assign the mask for this irq */
3332                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3333                                       &q_vector->affinity_mask);
3334         }
3335
3336         vsi->irqs_ready = true;
3337         return 0;
3338
3339 free_queue_irqs:
3340         while (vector) {
3341                 vector--;
3342                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3343                                       NULL);
3344                 free_irq(pf->msix_entries[base + vector].vector,
3345                          &(vsi->q_vectors[vector]));
3346         }
3347         return err;
3348 }
3349
3350 /**
3351  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3352  * @vsi: the VSI being un-configured
3353  **/
3354 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3355 {
3356         struct i40e_pf *pf = vsi->back;
3357         struct i40e_hw *hw = &pf->hw;
3358         int base = vsi->base_vector;
3359         int i;
3360
3361         for (i = 0; i < vsi->num_queue_pairs; i++) {
3362                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3363                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3364         }
3365
3366         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3367                 for (i = vsi->base_vector;
3368                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3369                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3370
3371                 i40e_flush(hw);
3372                 for (i = 0; i < vsi->num_q_vectors; i++)
3373                         synchronize_irq(pf->msix_entries[i + base].vector);
3374         } else {
3375                 /* Legacy and MSI mode - this stops all interrupt handling */
3376                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3377                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3378                 i40e_flush(hw);
3379                 synchronize_irq(pf->pdev->irq);
3380         }
3381 }
3382
3383 /**
3384  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3385  * @vsi: the VSI being configured
3386  **/
3387 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3388 {
3389         struct i40e_pf *pf = vsi->back;
3390         int i;
3391
3392         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3393                 for (i = 0; i < vsi->num_q_vectors; i++)
3394                         i40e_irq_dynamic_enable(vsi, i);
3395         } else {
3396                 i40e_irq_dynamic_enable_icr0(pf, true);
3397         }
3398
3399         i40e_flush(&pf->hw);
3400         return 0;
3401 }
3402
3403 /**
3404  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3405  * @pf: board private structure
3406  **/
3407 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3408 {
3409         /* Disable ICR 0 */
3410         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3411         i40e_flush(&pf->hw);
3412 }
3413
3414 /**
3415  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3416  * @irq: interrupt number
3417  * @data: pointer to a q_vector
3418  *
3419  * This is the handler used for all MSI/Legacy interrupts, and deals
3420  * with both queue and non-queue interrupts.  This is also used in
3421  * MSIX mode to handle the non-queue interrupts.
3422  **/
3423 static irqreturn_t i40e_intr(int irq, void *data)
3424 {
3425         struct i40e_pf *pf = (struct i40e_pf *)data;
3426         struct i40e_hw *hw = &pf->hw;
3427         irqreturn_t ret = IRQ_NONE;
3428         u32 icr0, icr0_remaining;
3429         u32 val, ena_mask;
3430
3431         icr0 = rd32(hw, I40E_PFINT_ICR0);
3432         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3433
3434         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3435         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3436                 goto enable_intr;
3437
3438         /* if interrupt but no bits showing, must be SWINT */
3439         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3440             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3441                 pf->sw_int_count++;
3442
3443         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3444             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3445                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3446                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3447                 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3448         }
3449
3450         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3451         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3452                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3453                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3454
3455                 /* We do not have a way to disarm Queue causes while leaving
3456                  * interrupt enabled for all other causes, ideally
3457                  * interrupt should be disabled while we are in NAPI but
3458                  * this is not a performance path and napi_schedule()
3459                  * can deal with rescheduling.
3460                  */
3461                 if (!test_bit(__I40E_DOWN, &pf->state))
3462                         napi_schedule_irqoff(&q_vector->napi);
3463         }
3464
3465         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3466                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3467                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3468                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3469         }
3470
3471         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3472                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3473                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3474         }
3475
3476         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3477                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3478                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3479         }
3480
3481         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3482                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3483                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3484                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3485                 val = rd32(hw, I40E_GLGEN_RSTAT);
3486                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3487                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3488                 if (val == I40E_RESET_CORER) {
3489                         pf->corer_count++;
3490                 } else if (val == I40E_RESET_GLOBR) {
3491                         pf->globr_count++;
3492                 } else if (val == I40E_RESET_EMPR) {
3493                         pf->empr_count++;
3494                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3495                 }
3496         }
3497
3498         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3499                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3500                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3501                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3502                          rd32(hw, I40E_PFHMC_ERRORINFO),
3503                          rd32(hw, I40E_PFHMC_ERRORDATA));
3504         }
3505
3506         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3507                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3508
3509                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3510                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3511                         i40e_ptp_tx_hwtstamp(pf);
3512                 }
3513         }
3514
3515         /* If a critical error is pending we have no choice but to reset the
3516          * device.
3517          * Report and mask out any remaining unexpected interrupts.
3518          */
3519         icr0_remaining = icr0 & ena_mask;
3520         if (icr0_remaining) {
3521                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3522                          icr0_remaining);
3523                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3524                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3525                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3526                         dev_info(&pf->pdev->dev, "device will be reset\n");
3527                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3528                         i40e_service_event_schedule(pf);
3529                 }
3530                 ena_mask &= ~icr0_remaining;
3531         }
3532         ret = IRQ_HANDLED;
3533
3534 enable_intr:
3535         /* re-enable interrupt causes */
3536         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3537         if (!test_bit(__I40E_DOWN, &pf->state)) {
3538                 i40e_service_event_schedule(pf);
3539                 i40e_irq_dynamic_enable_icr0(pf, false);
3540         }
3541
3542         return ret;
3543 }
3544
3545 /**
3546  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3547  * @tx_ring:  tx ring to clean
3548  * @budget:   how many cleans we're allowed
3549  *
3550  * Returns true if there's any budget left (e.g. the clean is finished)
3551  **/
3552 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3553 {
3554         struct i40e_vsi *vsi = tx_ring->vsi;
3555         u16 i = tx_ring->next_to_clean;
3556         struct i40e_tx_buffer *tx_buf;
3557         struct i40e_tx_desc *tx_desc;
3558
3559         tx_buf = &tx_ring->tx_bi[i];
3560         tx_desc = I40E_TX_DESC(tx_ring, i);
3561         i -= tx_ring->count;
3562
3563         do {
3564                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3565
3566                 /* if next_to_watch is not set then there is no work pending */
3567                 if (!eop_desc)
3568                         break;
3569
3570                 /* prevent any other reads prior to eop_desc */
3571                 read_barrier_depends();
3572
3573                 /* if the descriptor isn't done, no work yet to do */
3574                 if (!(eop_desc->cmd_type_offset_bsz &
3575                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3576                         break;
3577
3578                 /* clear next_to_watch to prevent false hangs */
3579                 tx_buf->next_to_watch = NULL;
3580
3581                 tx_desc->buffer_addr = 0;
3582                 tx_desc->cmd_type_offset_bsz = 0;
3583                 /* move past filter desc */
3584                 tx_buf++;
3585                 tx_desc++;
3586                 i++;
3587                 if (unlikely(!i)) {
3588                         i -= tx_ring->count;
3589                         tx_buf = tx_ring->tx_bi;
3590                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3591                 }
3592                 /* unmap skb header data */
3593                 dma_unmap_single(tx_ring->dev,
3594                                  dma_unmap_addr(tx_buf, dma),
3595                                  dma_unmap_len(tx_buf, len),
3596                                  DMA_TO_DEVICE);
3597                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3598                         kfree(tx_buf->raw_buf);
3599
3600                 tx_buf->raw_buf = NULL;
3601                 tx_buf->tx_flags = 0;
3602                 tx_buf->next_to_watch = NULL;
3603                 dma_unmap_len_set(tx_buf, len, 0);
3604                 tx_desc->buffer_addr = 0;
3605                 tx_desc->cmd_type_offset_bsz = 0;
3606
3607                 /* move us past the eop_desc for start of next FD desc */
3608                 tx_buf++;
3609                 tx_desc++;
3610                 i++;
3611                 if (unlikely(!i)) {
3612                         i -= tx_ring->count;
3613                         tx_buf = tx_ring->tx_bi;
3614                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3615                 }
3616
3617                 /* update budget accounting */
3618                 budget--;
3619         } while (likely(budget));
3620
3621         i += tx_ring->count;
3622         tx_ring->next_to_clean = i;
3623
3624         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3625                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3626
3627         return budget > 0;
3628 }
3629
3630 /**
3631  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3632  * @irq: interrupt number
3633  * @data: pointer to a q_vector
3634  **/
3635 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3636 {
3637         struct i40e_q_vector *q_vector = data;
3638         struct i40e_vsi *vsi;
3639
3640         if (!q_vector->tx.ring)
3641                 return IRQ_HANDLED;
3642
3643         vsi = q_vector->tx.ring->vsi;
3644         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3645
3646         return IRQ_HANDLED;
3647 }
3648
3649 /**
3650  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3651  * @vsi: the VSI being configured
3652  * @v_idx: vector index
3653  * @qp_idx: queue pair index
3654  **/
3655 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3656 {
3657         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3658         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3659         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3660
3661         tx_ring->q_vector = q_vector;
3662         tx_ring->next = q_vector->tx.ring;
3663         q_vector->tx.ring = tx_ring;
3664         q_vector->tx.count++;
3665
3666         rx_ring->q_vector = q_vector;
3667         rx_ring->next = q_vector->rx.ring;
3668         q_vector->rx.ring = rx_ring;
3669         q_vector->rx.count++;
3670 }
3671
3672 /**
3673  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3674  * @vsi: the VSI being configured
3675  *
3676  * This function maps descriptor rings to the queue-specific vectors
3677  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3678  * one vector per queue pair, but on a constrained vector budget, we
3679  * group the queue pairs as "efficiently" as possible.
3680  **/
3681 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3682 {
3683         int qp_remaining = vsi->num_queue_pairs;
3684         int q_vectors = vsi->num_q_vectors;
3685         int num_ringpairs;
3686         int v_start = 0;
3687         int qp_idx = 0;
3688
3689         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3690          * group them so there are multiple queues per vector.
3691          * It is also important to go through all the vectors available to be
3692          * sure that if we don't use all the vectors, that the remaining vectors
3693          * are cleared. This is especially important when decreasing the
3694          * number of queues in use.
3695          */
3696         for (; v_start < q_vectors; v_start++) {
3697                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3698
3699                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3700
3701                 q_vector->num_ringpairs = num_ringpairs;
3702
3703                 q_vector->rx.count = 0;
3704                 q_vector->tx.count = 0;
3705                 q_vector->rx.ring = NULL;
3706                 q_vector->tx.ring = NULL;
3707
3708                 while (num_ringpairs--) {
3709                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3710                         qp_idx++;
3711                         qp_remaining--;
3712                 }
3713         }
3714 }
3715
3716 /**
3717  * i40e_vsi_request_irq - Request IRQ from the OS
3718  * @vsi: the VSI being configured
3719  * @basename: name for the vector
3720  **/
3721 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3722 {
3723         struct i40e_pf *pf = vsi->back;
3724         int err;
3725
3726         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3727                 err = i40e_vsi_request_irq_msix(vsi, basename);
3728         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3729                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3730                                   pf->int_name, pf);
3731         else
3732                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3733                                   pf->int_name, pf);
3734
3735         if (err)
3736                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3737
3738         return err;
3739 }
3740
3741 #ifdef CONFIG_NET_POLL_CONTROLLER
3742 /**
3743  * i40e_netpoll - A Polling 'interrupt' handler
3744  * @netdev: network interface device structure
3745  *
3746  * This is used by netconsole to send skbs without having to re-enable
3747  * interrupts.  It's not called while the normal interrupt routine is executing.
3748  **/
3749 #ifdef I40E_FCOE
3750 void i40e_netpoll(struct net_device *netdev)
3751 #else
3752 static void i40e_netpoll(struct net_device *netdev)
3753 #endif
3754 {
3755         struct i40e_netdev_priv *np = netdev_priv(netdev);
3756         struct i40e_vsi *vsi = np->vsi;
3757         struct i40e_pf *pf = vsi->back;
3758         int i;
3759
3760         /* if interface is down do nothing */
3761         if (test_bit(__I40E_DOWN, &vsi->state))
3762                 return;
3763
3764         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3765                 for (i = 0; i < vsi->num_q_vectors; i++)
3766                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3767         } else {
3768                 i40e_intr(pf->pdev->irq, netdev);
3769         }
3770 }
3771 #endif
3772
3773 /**
3774  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3775  * @pf: the PF being configured
3776  * @pf_q: the PF queue
3777  * @enable: enable or disable state of the queue
3778  *
3779  * This routine will wait for the given Tx queue of the PF to reach the
3780  * enabled or disabled state.
3781  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3782  * multiple retries; else will return 0 in case of success.
3783  **/
3784 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3785 {
3786         int i;
3787         u32 tx_reg;
3788
3789         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3790                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3791                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3792                         break;
3793
3794                 usleep_range(10, 20);
3795         }
3796         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3797                 return -ETIMEDOUT;
3798
3799         return 0;
3800 }
3801
3802 /**
3803  * i40e_vsi_control_tx - Start or stop a VSI's rings
3804  * @vsi: the VSI being configured
3805  * @enable: start or stop the rings
3806  **/
3807 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3808 {
3809         struct i40e_pf *pf = vsi->back;
3810         struct i40e_hw *hw = &pf->hw;
3811         int i, j, pf_q, ret = 0;
3812         u32 tx_reg;
3813
3814         pf_q = vsi->base_queue;
3815         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3816
3817                 /* warn the TX unit of coming changes */
3818                 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3819                 if (!enable)
3820                         usleep_range(10, 20);
3821
3822                 for (j = 0; j < 50; j++) {
3823                         tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3824                         if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3825                             ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3826                                 break;
3827                         usleep_range(1000, 2000);
3828                 }
3829                 /* Skip if the queue is already in the requested state */
3830                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3831                         continue;
3832
3833                 /* turn on/off the queue */
3834                 if (enable) {
3835                         wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3836                         tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3837                 } else {
3838                         tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3839                 }
3840
3841                 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3842                 /* No waiting for the Tx queue to disable */
3843                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3844                         continue;
3845
3846                 /* wait for the change to finish */
3847                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3848                 if (ret) {
3849                         dev_info(&pf->pdev->dev,
3850                                  "VSI seid %d Tx ring %d %sable timeout\n",
3851                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3852                         break;
3853                 }
3854         }
3855
3856         if (hw->revision_id == 0)
3857                 mdelay(50);
3858         return ret;
3859 }
3860
3861 /**
3862  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3863  * @pf: the PF being configured
3864  * @pf_q: the PF queue
3865  * @enable: enable or disable state of the queue
3866  *
3867  * This routine will wait for the given Rx queue of the PF to reach the
3868  * enabled or disabled state.
3869  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3870  * multiple retries; else will return 0 in case of success.
3871  **/
3872 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3873 {
3874         int i;
3875         u32 rx_reg;
3876
3877         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3878                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3879                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3880                         break;
3881
3882                 usleep_range(10, 20);
3883         }
3884         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3885                 return -ETIMEDOUT;
3886
3887         return 0;
3888 }
3889
3890 /**
3891  * i40e_vsi_control_rx - Start or stop a VSI's rings
3892  * @vsi: the VSI being configured
3893  * @enable: start or stop the rings
3894  **/
3895 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3896 {
3897         struct i40e_pf *pf = vsi->back;
3898         struct i40e_hw *hw = &pf->hw;
3899         int i, j, pf_q, ret = 0;
3900         u32 rx_reg;
3901
3902         pf_q = vsi->base_queue;
3903         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3904                 for (j = 0; j < 50; j++) {
3905                         rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3906                         if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3907                             ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3908                                 break;
3909                         usleep_range(1000, 2000);
3910                 }
3911
3912                 /* Skip if the queue is already in the requested state */
3913                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3914                         continue;
3915
3916                 /* turn on/off the queue */
3917                 if (enable)
3918                         rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3919                 else
3920                         rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3921                 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3922                 /* No waiting for the Tx queue to disable */
3923                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3924                         continue;
3925
3926                 /* wait for the change to finish */
3927                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3928                 if (ret) {
3929                         dev_info(&pf->pdev->dev,
3930                                  "VSI seid %d Rx ring %d %sable timeout\n",
3931                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3932                         break;
3933                 }
3934         }
3935
3936         return ret;
3937 }
3938
3939 /**
3940  * i40e_vsi_control_rings - Start or stop a VSI's rings
3941  * @vsi: the VSI being configured
3942  * @enable: start or stop the rings
3943  **/
3944 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3945 {
3946         int ret = 0;
3947
3948         /* do rx first for enable and last for disable */
3949         if (request) {
3950                 ret = i40e_vsi_control_rx(vsi, request);
3951                 if (ret)
3952                         return ret;
3953                 ret = i40e_vsi_control_tx(vsi, request);
3954         } else {
3955                 /* Ignore return value, we need to shutdown whatever we can */
3956                 i40e_vsi_control_tx(vsi, request);
3957                 i40e_vsi_control_rx(vsi, request);
3958         }
3959
3960         return ret;
3961 }
3962
3963 /**
3964  * i40e_vsi_free_irq - Free the irq association with the OS
3965  * @vsi: the VSI being configured
3966  **/
3967 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3968 {
3969         struct i40e_pf *pf = vsi->back;
3970         struct i40e_hw *hw = &pf->hw;
3971         int base = vsi->base_vector;
3972         u32 val, qp;
3973         int i;
3974
3975         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3976                 if (!vsi->q_vectors)
3977                         return;
3978
3979                 if (!vsi->irqs_ready)
3980                         return;
3981
3982                 vsi->irqs_ready = false;
3983                 for (i = 0; i < vsi->num_q_vectors; i++) {
3984                         u16 vector = i + base;
3985
3986                         /* free only the irqs that were actually requested */
3987                         if (!vsi->q_vectors[i] ||
3988                             !vsi->q_vectors[i]->num_ringpairs)
3989                                 continue;
3990
3991                         /* clear the affinity_mask in the IRQ descriptor */
3992                         irq_set_affinity_hint(pf->msix_entries[vector].vector,
3993                                               NULL);
3994                         free_irq(pf->msix_entries[vector].vector,
3995                                  vsi->q_vectors[i]);
3996
3997                         /* Tear down the interrupt queue link list
3998                          *
3999                          * We know that they come in pairs and always
4000                          * the Rx first, then the Tx.  To clear the
4001                          * link list, stick the EOL value into the
4002                          * next_q field of the registers.
4003                          */
4004                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4005                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4006                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4007                         val |= I40E_QUEUE_END_OF_LIST
4008                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4009                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4010
4011                         while (qp != I40E_QUEUE_END_OF_LIST) {
4012                                 u32 next;
4013
4014                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4015
4016                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4017                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4018                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4019                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4020
4021                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4022                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4023
4024                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4025
4026                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4027
4028                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4029                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4030
4031                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4032                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4033                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4034                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4035
4036                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4037                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4038
4039                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4040                                 qp = next;
4041                         }
4042                 }
4043         } else {
4044                 free_irq(pf->pdev->irq, pf);
4045
4046                 val = rd32(hw, I40E_PFINT_LNKLST0);
4047                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4048                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4049                 val |= I40E_QUEUE_END_OF_LIST
4050                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4051                 wr32(hw, I40E_PFINT_LNKLST0, val);
4052
4053                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4054                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4055                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4056                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4057                          I40E_QINT_RQCTL_INTEVENT_MASK);
4058
4059                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4060                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4061
4062                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4063
4064                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4065
4066                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4067                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4068                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4069                          I40E_QINT_TQCTL_INTEVENT_MASK);
4070
4071                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4072                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4073
4074                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4075         }
4076 }
4077
4078 /**
4079  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4080  * @vsi: the VSI being configured
4081  * @v_idx: Index of vector to be freed
4082  *
4083  * This function frees the memory allocated to the q_vector.  In addition if
4084  * NAPI is enabled it will delete any references to the NAPI struct prior
4085  * to freeing the q_vector.
4086  **/
4087 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4088 {
4089         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4090         struct i40e_ring *ring;
4091
4092         if (!q_vector)
4093                 return;
4094
4095         /* disassociate q_vector from rings */
4096         i40e_for_each_ring(ring, q_vector->tx)
4097                 ring->q_vector = NULL;
4098
4099         i40e_for_each_ring(ring, q_vector->rx)
4100                 ring->q_vector = NULL;
4101
4102         /* only VSI w/ an associated netdev is set up w/ NAPI */
4103         if (vsi->netdev)
4104                 netif_napi_del(&q_vector->napi);
4105
4106         vsi->q_vectors[v_idx] = NULL;
4107
4108         kfree_rcu(q_vector, rcu);
4109 }
4110
4111 /**
4112  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4113  * @vsi: the VSI being un-configured
4114  *
4115  * This frees the memory allocated to the q_vectors and
4116  * deletes references to the NAPI struct.
4117  **/
4118 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4119 {
4120         int v_idx;
4121
4122         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4123                 i40e_free_q_vector(vsi, v_idx);
4124 }
4125
4126 /**
4127  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4128  * @pf: board private structure
4129  **/
4130 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4131 {
4132         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4133         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4134                 pci_disable_msix(pf->pdev);
4135                 kfree(pf->msix_entries);
4136                 pf->msix_entries = NULL;
4137                 kfree(pf->irq_pile);
4138                 pf->irq_pile = NULL;
4139         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4140                 pci_disable_msi(pf->pdev);
4141         }
4142         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4143 }
4144
4145 /**
4146  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4147  * @pf: board private structure
4148  *
4149  * We go through and clear interrupt specific resources and reset the structure
4150  * to pre-load conditions
4151  **/
4152 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4153 {
4154         int i;
4155
4156         i40e_stop_misc_vector(pf);
4157         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4158                 synchronize_irq(pf->msix_entries[0].vector);
4159                 free_irq(pf->msix_entries[0].vector, pf);
4160         }
4161
4162         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4163                       I40E_IWARP_IRQ_PILE_ID);
4164
4165         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4166         for (i = 0; i < pf->num_alloc_vsi; i++)
4167                 if (pf->vsi[i])
4168                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4169         i40e_reset_interrupt_capability(pf);
4170 }
4171
4172 /**
4173  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4174  * @vsi: the VSI being configured
4175  **/
4176 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4177 {
4178         int q_idx;
4179
4180         if (!vsi->netdev)
4181                 return;
4182
4183         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4184                 napi_enable(&vsi->q_vectors[q_idx]->napi);
4185 }
4186
4187 /**
4188  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4189  * @vsi: the VSI being configured
4190  **/
4191 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4192 {
4193         int q_idx;
4194
4195         if (!vsi->netdev)
4196                 return;
4197
4198         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4199                 napi_disable(&vsi->q_vectors[q_idx]->napi);
4200 }
4201
4202 /**
4203  * i40e_vsi_close - Shut down a VSI
4204  * @vsi: the vsi to be quelled
4205  **/
4206 static void i40e_vsi_close(struct i40e_vsi *vsi)
4207 {
4208         bool reset = false;
4209
4210         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4211                 i40e_down(vsi);
4212         i40e_vsi_free_irq(vsi);
4213         i40e_vsi_free_tx_resources(vsi);
4214         i40e_vsi_free_rx_resources(vsi);
4215         vsi->current_netdev_flags = 0;
4216         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4217                 reset = true;
4218         i40e_notify_client_of_netdev_close(vsi, reset);
4219 }
4220
4221 /**
4222  * i40e_quiesce_vsi - Pause a given VSI
4223  * @vsi: the VSI being paused
4224  **/
4225 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4226 {
4227         if (test_bit(__I40E_DOWN, &vsi->state))
4228                 return;
4229
4230         /* No need to disable FCoE VSI when Tx suspended */
4231         if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4232             vsi->type == I40E_VSI_FCOE) {
4233                 dev_dbg(&vsi->back->pdev->dev,
4234                          "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4235                 return;
4236         }
4237
4238         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4239         if (vsi->netdev && netif_running(vsi->netdev))
4240                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4241         else
4242                 i40e_vsi_close(vsi);
4243 }
4244
4245 /**
4246  * i40e_unquiesce_vsi - Resume a given VSI
4247  * @vsi: the VSI being resumed
4248  **/
4249 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4250 {
4251         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4252                 return;
4253
4254         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4255         if (vsi->netdev && netif_running(vsi->netdev))
4256                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4257         else
4258                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4259 }
4260
4261 /**
4262  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4263  * @pf: the PF
4264  **/
4265 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4266 {
4267         int v;
4268
4269         for (v = 0; v < pf->num_alloc_vsi; v++) {
4270                 if (pf->vsi[v])
4271                         i40e_quiesce_vsi(pf->vsi[v]);
4272         }
4273 }
4274
4275 /**
4276  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4277  * @pf: the PF
4278  **/
4279 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4280 {
4281         int v;
4282
4283         for (v = 0; v < pf->num_alloc_vsi; v++) {
4284                 if (pf->vsi[v])
4285                         i40e_unquiesce_vsi(pf->vsi[v]);
4286         }
4287 }
4288
4289 #ifdef CONFIG_I40E_DCB
4290 /**
4291  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4292  * @vsi: the VSI being configured
4293  *
4294  * This function waits for the given VSI's queues to be disabled.
4295  **/
4296 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4297 {
4298         struct i40e_pf *pf = vsi->back;
4299         int i, pf_q, ret;
4300
4301         pf_q = vsi->base_queue;
4302         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4303                 /* Check and wait for the disable status of the queue */
4304                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4305                 if (ret) {
4306                         dev_info(&pf->pdev->dev,
4307                                  "VSI seid %d Tx ring %d disable timeout\n",
4308                                  vsi->seid, pf_q);
4309                         return ret;
4310                 }
4311         }
4312
4313         pf_q = vsi->base_queue;
4314         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4315                 /* Check and wait for the disable status of the queue */
4316                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4317                 if (ret) {
4318                         dev_info(&pf->pdev->dev,
4319                                  "VSI seid %d Rx ring %d disable timeout\n",
4320                                  vsi->seid, pf_q);
4321                         return ret;
4322                 }
4323         }
4324
4325         return 0;
4326 }
4327
4328 /**
4329  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4330  * @pf: the PF
4331  *
4332  * This function waits for the queues to be in disabled state for all the
4333  * VSIs that are managed by this PF.
4334  **/
4335 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4336 {
4337         int v, ret = 0;
4338
4339         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4340                 /* No need to wait for FCoE VSI queues */
4341                 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4342                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4343                         if (ret)
4344                                 break;
4345                 }
4346         }
4347
4348         return ret;
4349 }
4350
4351 #endif
4352
4353 /**
4354  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4355  * @q_idx: TX queue number
4356  * @vsi: Pointer to VSI struct
4357  *
4358  * This function checks specified queue for given VSI. Detects hung condition.
4359  * Sets hung bit since it is two step process. Before next run of service task
4360  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4361  * hung condition remain unchanged and during subsequent run, this function
4362  * issues SW interrupt to recover from hung condition.
4363  **/
4364 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4365 {
4366         struct i40e_ring *tx_ring = NULL;
4367         struct i40e_pf  *pf;
4368         u32 head, val, tx_pending_hw;
4369         int i;
4370
4371         pf = vsi->back;
4372
4373         /* now that we have an index, find the tx_ring struct */
4374         for (i = 0; i < vsi->num_queue_pairs; i++) {
4375                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4376                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4377                                 tx_ring = vsi->tx_rings[i];
4378                                 break;
4379                         }
4380                 }
4381         }
4382
4383         if (!tx_ring)
4384                 return;
4385
4386         /* Read interrupt register */
4387         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4388                 val = rd32(&pf->hw,
4389                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4390                                                tx_ring->vsi->base_vector - 1));
4391         else
4392                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4393
4394         head = i40e_get_head(tx_ring);
4395
4396         tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4397
4398         /* HW is done executing descriptors, updated HEAD write back,
4399          * but SW hasn't processed those descriptors. If interrupt is
4400          * not generated from this point ON, it could result into
4401          * dev_watchdog detecting timeout on those netdev_queue,
4402          * hence proactively trigger SW interrupt.
4403          */
4404         if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4405                 /* NAPI Poll didn't run and clear since it was set */
4406                 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4407                                        &tx_ring->q_vector->hung_detected)) {
4408                         netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4409                                     vsi->seid, q_idx, tx_pending_hw,
4410                                     tx_ring->next_to_clean, head,
4411                                     tx_ring->next_to_use,
4412                                     readl(tx_ring->tail));
4413                         netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4414                                     vsi->seid, q_idx, val);
4415                         i40e_force_wb(vsi, tx_ring->q_vector);
4416                 } else {
4417                         /* First Chance - detected possible hung */
4418                         set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4419                                 &tx_ring->q_vector->hung_detected);
4420                 }
4421         }
4422
4423         /* This is the case where we have interrupts missing,
4424          * so the tx_pending in HW will most likely be 0, but we
4425          * will have tx_pending in SW since the WB happened but the
4426          * interrupt got lost.
4427          */
4428         if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4429             (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4430                 if (napi_reschedule(&tx_ring->q_vector->napi))
4431                         tx_ring->tx_stats.tx_lost_interrupt++;
4432         }
4433 }
4434
4435 /**
4436  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4437  * @pf:  pointer to PF struct
4438  *
4439  * LAN VSI has netdev and netdev has TX queues. This function is to check
4440  * each of those TX queues if they are hung, trigger recovery by issuing
4441  * SW interrupt.
4442  **/
4443 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4444 {
4445         struct net_device *netdev;
4446         struct i40e_vsi *vsi;
4447         int i;
4448
4449         /* Only for LAN VSI */
4450         vsi = pf->vsi[pf->lan_vsi];
4451
4452         if (!vsi)
4453                 return;
4454
4455         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4456         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4457             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4458                 return;
4459
4460         /* Make sure type is MAIN VSI */
4461         if (vsi->type != I40E_VSI_MAIN)
4462                 return;
4463
4464         netdev = vsi->netdev;
4465         if (!netdev)
4466                 return;
4467
4468         /* Bail out if netif_carrier is not OK */
4469         if (!netif_carrier_ok(netdev))
4470                 return;
4471
4472         /* Go thru' TX queues for netdev */
4473         for (i = 0; i < netdev->num_tx_queues; i++) {
4474                 struct netdev_queue *q;
4475
4476                 q = netdev_get_tx_queue(netdev, i);
4477                 if (q)
4478                         i40e_detect_recover_hung_queue(i, vsi);
4479         }
4480 }
4481
4482 /**
4483  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4484  * @pf: pointer to PF
4485  *
4486  * Get TC map for ISCSI PF type that will include iSCSI TC
4487  * and LAN TC.
4488  **/
4489 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4490 {
4491         struct i40e_dcb_app_priority_table app;
4492         struct i40e_hw *hw = &pf->hw;
4493         u8 enabled_tc = 1; /* TC0 is always enabled */
4494         u8 tc, i;
4495         /* Get the iSCSI APP TLV */
4496         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4497
4498         for (i = 0; i < dcbcfg->numapps; i++) {
4499                 app = dcbcfg->app[i];
4500                 if (app.selector == I40E_APP_SEL_TCPIP &&
4501                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4502                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4503                         enabled_tc |= BIT(tc);
4504                         break;
4505                 }
4506         }
4507
4508         return enabled_tc;
4509 }
4510
4511 /**
4512  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4513  * @dcbcfg: the corresponding DCBx configuration structure
4514  *
4515  * Return the number of TCs from given DCBx configuration
4516  **/
4517 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4518 {
4519         u8 num_tc = 0;
4520         int i;
4521
4522         /* Scan the ETS Config Priority Table to find
4523          * traffic class enabled for a given priority
4524          * and use the traffic class index to get the
4525          * number of traffic classes enabled
4526          */
4527         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4528                 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4529                         num_tc = dcbcfg->etscfg.prioritytable[i];
4530         }
4531
4532         /* Traffic class index starts from zero so
4533          * increment to return the actual count
4534          */
4535         return num_tc + 1;
4536 }
4537
4538 /**
4539  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4540  * @dcbcfg: the corresponding DCBx configuration structure
4541  *
4542  * Query the current DCB configuration and return the number of
4543  * traffic classes enabled from the given DCBX config
4544  **/
4545 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4546 {
4547         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4548         u8 enabled_tc = 1;
4549         u8 i;
4550
4551         for (i = 0; i < num_tc; i++)
4552                 enabled_tc |= BIT(i);
4553
4554         return enabled_tc;
4555 }
4556
4557 /**
4558  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4559  * @pf: PF being queried
4560  *
4561  * Return number of traffic classes enabled for the given PF
4562  **/
4563 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4564 {
4565         struct i40e_hw *hw = &pf->hw;
4566         u8 i, enabled_tc;
4567         u8 num_tc = 0;
4568         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4569
4570         /* If DCB is not enabled then always in single TC */
4571         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4572                 return 1;
4573
4574         /* SFP mode will be enabled for all TCs on port */
4575         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4576                 return i40e_dcb_get_num_tc(dcbcfg);
4577
4578         /* MFP mode return count of enabled TCs for this PF */
4579         if (pf->hw.func_caps.iscsi)
4580                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4581         else
4582                 return 1; /* Only TC0 */
4583
4584         /* At least have TC0 */
4585         enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4586         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4587                 if (enabled_tc & BIT(i))
4588                         num_tc++;
4589         }
4590         return num_tc;
4591 }
4592
4593 /**
4594  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4595  * @pf: PF being queried
4596  *
4597  * Return a bitmap for first enabled traffic class for this PF.
4598  **/
4599 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4600 {
4601         u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4602         u8 i = 0;
4603
4604         if (!enabled_tc)
4605                 return 0x1; /* TC0 */
4606
4607         /* Find the first enabled TC */
4608         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4609                 if (enabled_tc & BIT(i))
4610                         break;
4611         }
4612
4613         return BIT(i);
4614 }
4615
4616 /**
4617  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4618  * @pf: PF being queried
4619  *
4620  * Return a bitmap for enabled traffic classes for this PF.
4621  **/
4622 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4623 {
4624         /* If DCB is not enabled for this PF then just return default TC */
4625         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4626                 return i40e_pf_get_default_tc(pf);
4627
4628         /* SFP mode we want PF to be enabled for all TCs */
4629         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4630                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4631
4632         /* MFP enabled and iSCSI PF type */
4633         if (pf->hw.func_caps.iscsi)
4634                 return i40e_get_iscsi_tc_map(pf);
4635         else
4636                 return i40e_pf_get_default_tc(pf);
4637 }
4638
4639 /**
4640  * i40e_vsi_get_bw_info - Query VSI BW Information
4641  * @vsi: the VSI being queried
4642  *
4643  * Returns 0 on success, negative value on failure
4644  **/
4645 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4646 {
4647         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4648         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4649         struct i40e_pf *pf = vsi->back;
4650         struct i40e_hw *hw = &pf->hw;
4651         i40e_status ret;
4652         u32 tc_bw_max;
4653         int i;
4654
4655         /* Get the VSI level BW configuration */
4656         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4657         if (ret) {
4658                 dev_info(&pf->pdev->dev,
4659                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4660                          i40e_stat_str(&pf->hw, ret),
4661                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4662                 return -EINVAL;
4663         }
4664
4665         /* Get the VSI level BW configuration per TC */
4666         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4667                                                NULL);
4668         if (ret) {
4669                 dev_info(&pf->pdev->dev,
4670                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4671                          i40e_stat_str(&pf->hw, ret),
4672                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4673                 return -EINVAL;
4674         }
4675
4676         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4677                 dev_info(&pf->pdev->dev,
4678                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4679                          bw_config.tc_valid_bits,
4680                          bw_ets_config.tc_valid_bits);
4681                 /* Still continuing */
4682         }
4683
4684         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4685         vsi->bw_max_quanta = bw_config.max_bw;
4686         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4687                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4688         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4689                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4690                 vsi->bw_ets_limit_credits[i] =
4691                                         le16_to_cpu(bw_ets_config.credits[i]);
4692                 /* 3 bits out of 4 for each TC */
4693                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4694         }
4695
4696         return 0;
4697 }
4698
4699 /**
4700  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4701  * @vsi: the VSI being configured
4702  * @enabled_tc: TC bitmap
4703  * @bw_credits: BW shared credits per TC
4704  *
4705  * Returns 0 on success, negative value on failure
4706  **/
4707 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4708                                        u8 *bw_share)
4709 {
4710         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4711         i40e_status ret;
4712         int i;
4713
4714         bw_data.tc_valid_bits = enabled_tc;
4715         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4716                 bw_data.tc_bw_credits[i] = bw_share[i];
4717
4718         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4719                                        NULL);
4720         if (ret) {
4721                 dev_info(&vsi->back->pdev->dev,
4722                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4723                          vsi->back->hw.aq.asq_last_status);
4724                 return -EINVAL;
4725         }
4726
4727         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4728                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4729
4730         return 0;
4731 }
4732
4733 /**
4734  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4735  * @vsi: the VSI being configured
4736  * @enabled_tc: TC map to be enabled
4737  *
4738  **/
4739 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4740 {
4741         struct net_device *netdev = vsi->netdev;
4742         struct i40e_pf *pf = vsi->back;
4743         struct i40e_hw *hw = &pf->hw;
4744         u8 netdev_tc = 0;
4745         int i;
4746         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4747
4748         if (!netdev)
4749                 return;
4750
4751         if (!enabled_tc) {
4752                 netdev_reset_tc(netdev);
4753                 return;
4754         }
4755
4756         /* Set up actual enabled TCs on the VSI */
4757         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4758                 return;
4759
4760         /* set per TC queues for the VSI */
4761         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4762                 /* Only set TC queues for enabled tcs
4763                  *
4764                  * e.g. For a VSI that has TC0 and TC3 enabled the
4765                  * enabled_tc bitmap would be 0x00001001; the driver
4766                  * will set the numtc for netdev as 2 that will be
4767                  * referenced by the netdev layer as TC 0 and 1.
4768                  */
4769                 if (vsi->tc_config.enabled_tc & BIT(i))
4770                         netdev_set_tc_queue(netdev,
4771                                         vsi->tc_config.tc_info[i].netdev_tc,
4772                                         vsi->tc_config.tc_info[i].qcount,
4773                                         vsi->tc_config.tc_info[i].qoffset);
4774         }
4775
4776         /* Assign UP2TC map for the VSI */
4777         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4778                 /* Get the actual TC# for the UP */
4779                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4780                 /* Get the mapped netdev TC# for the UP */
4781                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4782                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4783         }
4784 }
4785
4786 /**
4787  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4788  * @vsi: the VSI being configured
4789  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4790  **/
4791 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4792                                       struct i40e_vsi_context *ctxt)
4793 {
4794         /* copy just the sections touched not the entire info
4795          * since not all sections are valid as returned by
4796          * update vsi params
4797          */
4798         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4799         memcpy(&vsi->info.queue_mapping,
4800                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4801         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4802                sizeof(vsi->info.tc_mapping));
4803 }
4804
4805 /**
4806  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4807  * @vsi: VSI to be configured
4808  * @enabled_tc: TC bitmap
4809  *
4810  * This configures a particular VSI for TCs that are mapped to the
4811  * given TC bitmap. It uses default bandwidth share for TCs across
4812  * VSIs to configure TC for a particular VSI.
4813  *
4814  * NOTE:
4815  * It is expected that the VSI queues have been quisced before calling
4816  * this function.
4817  **/
4818 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4819 {
4820         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4821         struct i40e_vsi_context ctxt;
4822         int ret = 0;
4823         int i;
4824
4825         /* Check if enabled_tc is same as existing or new TCs */
4826         if (vsi->tc_config.enabled_tc == enabled_tc)
4827                 return ret;
4828
4829         /* Enable ETS TCs with equal BW Share for now across all VSIs */
4830         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4831                 if (enabled_tc & BIT(i))
4832                         bw_share[i] = 1;
4833         }
4834
4835         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4836         if (ret) {
4837                 dev_info(&vsi->back->pdev->dev,
4838                          "Failed configuring TC map %d for VSI %d\n",
4839                          enabled_tc, vsi->seid);
4840                 goto out;
4841         }
4842
4843         /* Update Queue Pairs Mapping for currently enabled UPs */
4844         ctxt.seid = vsi->seid;
4845         ctxt.pf_num = vsi->back->hw.pf_id;
4846         ctxt.vf_num = 0;
4847         ctxt.uplink_seid = vsi->uplink_seid;
4848         ctxt.info = vsi->info;
4849         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4850
4851         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4852                 ctxt.info.valid_sections |=
4853                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4854                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4855         }
4856
4857         /* Update the VSI after updating the VSI queue-mapping information */
4858         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4859         if (ret) {
4860                 dev_info(&vsi->back->pdev->dev,
4861                          "Update vsi tc config failed, err %s aq_err %s\n",
4862                          i40e_stat_str(&vsi->back->hw, ret),
4863                          i40e_aq_str(&vsi->back->hw,
4864                                      vsi->back->hw.aq.asq_last_status));
4865                 goto out;
4866         }
4867         /* update the local VSI info with updated queue map */
4868         i40e_vsi_update_queue_map(vsi, &ctxt);
4869         vsi->info.valid_sections = 0;
4870
4871         /* Update current VSI BW information */
4872         ret = i40e_vsi_get_bw_info(vsi);
4873         if (ret) {
4874                 dev_info(&vsi->back->pdev->dev,
4875                          "Failed updating vsi bw info, err %s aq_err %s\n",
4876                          i40e_stat_str(&vsi->back->hw, ret),
4877                          i40e_aq_str(&vsi->back->hw,
4878                                      vsi->back->hw.aq.asq_last_status));
4879                 goto out;
4880         }
4881
4882         /* Update the netdev TC setup */
4883         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4884 out:
4885         return ret;
4886 }
4887
4888 /**
4889  * i40e_veb_config_tc - Configure TCs for given VEB
4890  * @veb: given VEB
4891  * @enabled_tc: TC bitmap
4892  *
4893  * Configures given TC bitmap for VEB (switching) element
4894  **/
4895 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4896 {
4897         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4898         struct i40e_pf *pf = veb->pf;
4899         int ret = 0;
4900         int i;
4901
4902         /* No TCs or already enabled TCs just return */
4903         if (!enabled_tc || veb->enabled_tc == enabled_tc)
4904                 return ret;
4905
4906         bw_data.tc_valid_bits = enabled_tc;
4907         /* bw_data.absolute_credits is not set (relative) */
4908
4909         /* Enable ETS TCs with equal BW Share for now */
4910         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4911                 if (enabled_tc & BIT(i))
4912                         bw_data.tc_bw_share_credits[i] = 1;
4913         }
4914
4915         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4916                                                    &bw_data, NULL);
4917         if (ret) {
4918                 dev_info(&pf->pdev->dev,
4919                          "VEB bw config failed, err %s aq_err %s\n",
4920                          i40e_stat_str(&pf->hw, ret),
4921                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4922                 goto out;
4923         }
4924
4925         /* Update the BW information */
4926         ret = i40e_veb_get_bw_info(veb);
4927         if (ret) {
4928                 dev_info(&pf->pdev->dev,
4929                          "Failed getting veb bw config, err %s aq_err %s\n",
4930                          i40e_stat_str(&pf->hw, ret),
4931                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4932         }
4933
4934 out:
4935         return ret;
4936 }
4937
4938 #ifdef CONFIG_I40E_DCB
4939 /**
4940  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4941  * @pf: PF struct
4942  *
4943  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4944  * the caller would've quiesce all the VSIs before calling
4945  * this function
4946  **/
4947 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4948 {
4949         u8 tc_map = 0;
4950         int ret;
4951         u8 v;
4952
4953         /* Enable the TCs available on PF to all VEBs */
4954         tc_map = i40e_pf_get_tc_map(pf);
4955         for (v = 0; v < I40E_MAX_VEB; v++) {
4956                 if (!pf->veb[v])
4957                         continue;
4958                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4959                 if (ret) {
4960                         dev_info(&pf->pdev->dev,
4961                                  "Failed configuring TC for VEB seid=%d\n",
4962                                  pf->veb[v]->seid);
4963                         /* Will try to configure as many components */
4964                 }
4965         }
4966
4967         /* Update each VSI */
4968         for (v = 0; v < pf->num_alloc_vsi; v++) {
4969                 if (!pf->vsi[v])
4970                         continue;
4971
4972                 /* - Enable all TCs for the LAN VSI
4973 #ifdef I40E_FCOE
4974                  * - For FCoE VSI only enable the TC configured
4975                  *   as per the APP TLV
4976 #endif
4977                  * - For all others keep them at TC0 for now
4978                  */
4979                 if (v == pf->lan_vsi)
4980                         tc_map = i40e_pf_get_tc_map(pf);
4981                 else
4982                         tc_map = i40e_pf_get_default_tc(pf);
4983 #ifdef I40E_FCOE
4984                 if (pf->vsi[v]->type == I40E_VSI_FCOE)
4985                         tc_map = i40e_get_fcoe_tc_map(pf);
4986 #endif /* #ifdef I40E_FCOE */
4987
4988                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4989                 if (ret) {
4990                         dev_info(&pf->pdev->dev,
4991                                  "Failed configuring TC for VSI seid=%d\n",
4992                                  pf->vsi[v]->seid);
4993                         /* Will try to configure as many components */
4994                 } else {
4995                         /* Re-configure VSI vectors based on updated TC map */
4996                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4997                         if (pf->vsi[v]->netdev)
4998                                 i40e_dcbnl_set_all(pf->vsi[v]);
4999                 }
5000                 i40e_notify_client_of_l2_param_changes(pf->vsi[v]);
5001         }
5002 }
5003
5004 /**
5005  * i40e_resume_port_tx - Resume port Tx
5006  * @pf: PF struct
5007  *
5008  * Resume a port's Tx and issue a PF reset in case of failure to
5009  * resume.
5010  **/
5011 static int i40e_resume_port_tx(struct i40e_pf *pf)
5012 {
5013         struct i40e_hw *hw = &pf->hw;
5014         int ret;
5015
5016         ret = i40e_aq_resume_port_tx(hw, NULL);
5017         if (ret) {
5018                 dev_info(&pf->pdev->dev,
5019                          "Resume Port Tx failed, err %s aq_err %s\n",
5020                           i40e_stat_str(&pf->hw, ret),
5021                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5022                 /* Schedule PF reset to recover */
5023                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5024                 i40e_service_event_schedule(pf);
5025         }
5026
5027         return ret;
5028 }
5029
5030 /**
5031  * i40e_init_pf_dcb - Initialize DCB configuration
5032  * @pf: PF being configured
5033  *
5034  * Query the current DCB configuration and cache it
5035  * in the hardware structure
5036  **/
5037 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5038 {
5039         struct i40e_hw *hw = &pf->hw;
5040         int err = 0;
5041
5042         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5043         if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5044                 goto out;
5045
5046         /* Get the initial DCB configuration */
5047         err = i40e_init_dcb(hw);
5048         if (!err) {
5049                 /* Device/Function is not DCBX capable */
5050                 if ((!hw->func_caps.dcb) ||
5051                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5052                         dev_info(&pf->pdev->dev,
5053                                  "DCBX offload is not supported or is disabled for this PF.\n");
5054
5055                         if (pf->flags & I40E_FLAG_MFP_ENABLED)
5056                                 goto out;
5057
5058                 } else {
5059                         /* When status is not DISABLED then DCBX in FW */
5060                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5061                                        DCB_CAP_DCBX_VER_IEEE;
5062
5063                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5064                         /* Enable DCB tagging only when more than one TC */
5065                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5066                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5067                         dev_dbg(&pf->pdev->dev,
5068                                 "DCBX offload is supported for this PF.\n");
5069                 }
5070         } else {
5071                 dev_info(&pf->pdev->dev,
5072                          "Query for DCB configuration failed, err %s aq_err %s\n",
5073                          i40e_stat_str(&pf->hw, err),
5074                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5075         }
5076
5077 out:
5078         return err;
5079 }
5080 #endif /* CONFIG_I40E_DCB */
5081 #define SPEED_SIZE 14
5082 #define FC_SIZE 8
5083 /**
5084  * i40e_print_link_message - print link up or down
5085  * @vsi: the VSI for which link needs a message
5086  */
5087 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5088 {
5089         char *speed = "Unknown";
5090         char *fc = "Unknown";
5091
5092         if (vsi->current_isup == isup)
5093                 return;
5094         vsi->current_isup = isup;
5095         if (!isup) {
5096                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5097                 return;
5098         }
5099
5100         /* Warn user if link speed on NPAR enabled partition is not at
5101          * least 10GB
5102          */
5103         if (vsi->back->hw.func_caps.npar_enable &&
5104             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5105              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5106                 netdev_warn(vsi->netdev,
5107                             "The partition detected link speed that is less than 10Gbps\n");
5108
5109         switch (vsi->back->hw.phy.link_info.link_speed) {
5110         case I40E_LINK_SPEED_40GB:
5111                 speed = "40 G";
5112                 break;
5113         case I40E_LINK_SPEED_20GB:
5114                 speed = "20 G";
5115                 break;
5116         case I40E_LINK_SPEED_10GB:
5117                 speed = "10 G";
5118                 break;
5119         case I40E_LINK_SPEED_1GB:
5120                 speed = "1000 M";
5121                 break;
5122         case I40E_LINK_SPEED_100MB:
5123                 speed = "100 M";
5124                 break;
5125         default:
5126                 break;
5127         }
5128
5129         switch (vsi->back->hw.fc.current_mode) {
5130         case I40E_FC_FULL:
5131                 fc = "RX/TX";
5132                 break;
5133         case I40E_FC_TX_PAUSE:
5134                 fc = "TX";
5135                 break;
5136         case I40E_FC_RX_PAUSE:
5137                 fc = "RX";
5138                 break;
5139         default:
5140                 fc = "None";
5141                 break;
5142         }
5143
5144         netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5145                     speed, fc);
5146 }
5147
5148 /**
5149  * i40e_up_complete - Finish the last steps of bringing up a connection
5150  * @vsi: the VSI being configured
5151  **/
5152 static int i40e_up_complete(struct i40e_vsi *vsi)
5153 {
5154         struct i40e_pf *pf = vsi->back;
5155         int err;
5156
5157         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5158                 i40e_vsi_configure_msix(vsi);
5159         else
5160                 i40e_configure_msi_and_legacy(vsi);
5161
5162         /* start rings */
5163         err = i40e_vsi_control_rings(vsi, true);
5164         if (err)
5165                 return err;
5166
5167         clear_bit(__I40E_DOWN, &vsi->state);
5168         i40e_napi_enable_all(vsi);
5169         i40e_vsi_enable_irq(vsi);
5170
5171         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5172             (vsi->netdev)) {
5173                 i40e_print_link_message(vsi, true);
5174                 netif_tx_start_all_queues(vsi->netdev);
5175                 netif_carrier_on(vsi->netdev);
5176         } else if (vsi->netdev) {
5177                 i40e_print_link_message(vsi, false);
5178                 /* need to check for qualified module here*/
5179                 if ((pf->hw.phy.link_info.link_info &
5180                         I40E_AQ_MEDIA_AVAILABLE) &&
5181                     (!(pf->hw.phy.link_info.an_info &
5182                         I40E_AQ_QUALIFIED_MODULE)))
5183                         netdev_err(vsi->netdev,
5184                                    "the driver failed to link because an unqualified module was detected.");
5185         }
5186
5187         /* replay FDIR SB filters */
5188         if (vsi->type == I40E_VSI_FDIR) {
5189                 /* reset fd counters */
5190                 pf->fd_add_err = pf->fd_atr_cnt = 0;
5191                 if (pf->fd_tcp_rule > 0) {
5192                         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5193                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5194                                 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5195                         pf->fd_tcp_rule = 0;
5196                 }
5197                 i40e_fdir_filter_restore(vsi);
5198         }
5199
5200         /* On the next run of the service_task, notify any clients of the new
5201          * opened netdev
5202          */
5203         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5204         i40e_service_event_schedule(pf);
5205
5206         return 0;
5207 }
5208
5209 /**
5210  * i40e_vsi_reinit_locked - Reset the VSI
5211  * @vsi: the VSI being configured
5212  *
5213  * Rebuild the ring structs after some configuration
5214  * has changed, e.g. MTU size.
5215  **/
5216 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5217 {
5218         struct i40e_pf *pf = vsi->back;
5219
5220         WARN_ON(in_interrupt());
5221         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5222                 usleep_range(1000, 2000);
5223         i40e_down(vsi);
5224
5225         /* Give a VF some time to respond to the reset.  The
5226          * two second wait is based upon the watchdog cycle in
5227          * the VF driver.
5228          */
5229         if (vsi->type == I40E_VSI_SRIOV)
5230                 msleep(2000);
5231         i40e_up(vsi);
5232         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5233 }
5234
5235 /**
5236  * i40e_up - Bring the connection back up after being down
5237  * @vsi: the VSI being configured
5238  **/
5239 int i40e_up(struct i40e_vsi *vsi)
5240 {
5241         int err;
5242
5243         err = i40e_vsi_configure(vsi);
5244         if (!err)
5245                 err = i40e_up_complete(vsi);
5246
5247         return err;
5248 }
5249
5250 /**
5251  * i40e_down - Shutdown the connection processing
5252  * @vsi: the VSI being stopped
5253  **/
5254 void i40e_down(struct i40e_vsi *vsi)
5255 {
5256         int i;
5257
5258         /* It is assumed that the caller of this function
5259          * sets the vsi->state __I40E_DOWN bit.
5260          */
5261         if (vsi->netdev) {
5262                 netif_carrier_off(vsi->netdev);
5263                 netif_tx_disable(vsi->netdev);
5264         }
5265         i40e_vsi_disable_irq(vsi);
5266         i40e_vsi_control_rings(vsi, false);
5267         i40e_napi_disable_all(vsi);
5268
5269         for (i = 0; i < vsi->num_queue_pairs; i++) {
5270                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5271                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5272         }
5273 }
5274
5275 /**
5276  * i40e_setup_tc - configure multiple traffic classes
5277  * @netdev: net device to configure
5278  * @tc: number of traffic classes to enable
5279  **/
5280 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5281 {
5282         struct i40e_netdev_priv *np = netdev_priv(netdev);
5283         struct i40e_vsi *vsi = np->vsi;
5284         struct i40e_pf *pf = vsi->back;
5285         u8 enabled_tc = 0;
5286         int ret = -EINVAL;
5287         int i;
5288
5289         /* Check if DCB enabled to continue */
5290         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5291                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5292                 goto exit;
5293         }
5294
5295         /* Check if MFP enabled */
5296         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5297                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5298                 goto exit;
5299         }
5300
5301         /* Check whether tc count is within enabled limit */
5302         if (tc > i40e_pf_get_num_tc(pf)) {
5303                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5304                 goto exit;
5305         }
5306
5307         /* Generate TC map for number of tc requested */
5308         for (i = 0; i < tc; i++)
5309                 enabled_tc |= BIT(i);
5310
5311         /* Requesting same TC configuration as already enabled */
5312         if (enabled_tc == vsi->tc_config.enabled_tc)
5313                 return 0;
5314
5315         /* Quiesce VSI queues */
5316         i40e_quiesce_vsi(vsi);
5317
5318         /* Configure VSI for enabled TCs */
5319         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5320         if (ret) {
5321                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5322                             vsi->seid);
5323                 goto exit;
5324         }
5325
5326         /* Unquiesce VSI */
5327         i40e_unquiesce_vsi(vsi);
5328
5329 exit:
5330         return ret;
5331 }
5332
5333 #ifdef I40E_FCOE
5334 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5335                     struct tc_to_netdev *tc)
5336 #else
5337 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5338                            struct tc_to_netdev *tc)
5339 #endif
5340 {
5341         if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5342                 return -EINVAL;
5343         return i40e_setup_tc(netdev, tc->tc);
5344 }
5345
5346 /**
5347  * i40e_open - Called when a network interface is made active
5348  * @netdev: network interface device structure
5349  *
5350  * The open entry point is called when a network interface is made
5351  * active by the system (IFF_UP).  At this point all resources needed
5352  * for transmit and receive operations are allocated, the interrupt
5353  * handler is registered with the OS, the netdev watchdog subtask is
5354  * enabled, and the stack is notified that the interface is ready.
5355  *
5356  * Returns 0 on success, negative value on failure
5357  **/
5358 int i40e_open(struct net_device *netdev)
5359 {
5360         struct i40e_netdev_priv *np = netdev_priv(netdev);
5361         struct i40e_vsi *vsi = np->vsi;
5362         struct i40e_pf *pf = vsi->back;
5363         int err;
5364
5365         /* disallow open during test or if eeprom is broken */
5366         if (test_bit(__I40E_TESTING, &pf->state) ||
5367             test_bit(__I40E_BAD_EEPROM, &pf->state))
5368                 return -EBUSY;
5369
5370         netif_carrier_off(netdev);
5371
5372         err = i40e_vsi_open(vsi);
5373         if (err)
5374                 return err;
5375
5376         /* configure global TSO hardware offload settings */
5377         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5378                                                        TCP_FLAG_FIN) >> 16);
5379         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5380                                                        TCP_FLAG_FIN |
5381                                                        TCP_FLAG_CWR) >> 16);
5382         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5383
5384 #ifdef CONFIG_I40E_VXLAN
5385         vxlan_get_rx_port(netdev);
5386 #endif
5387 #ifdef CONFIG_I40E_GENEVE
5388         if (pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE)
5389                 geneve_get_rx_port(netdev);
5390 #endif
5391
5392         i40e_notify_client_of_netdev_open(vsi);
5393
5394         return 0;
5395 }
5396
5397 /**
5398  * i40e_vsi_open -
5399  * @vsi: the VSI to open
5400  *
5401  * Finish initialization of the VSI.
5402  *
5403  * Returns 0 on success, negative value on failure
5404  **/
5405 int i40e_vsi_open(struct i40e_vsi *vsi)
5406 {
5407         struct i40e_pf *pf = vsi->back;
5408         char int_name[I40E_INT_NAME_STR_LEN];
5409         int err;
5410
5411         /* allocate descriptors */
5412         err = i40e_vsi_setup_tx_resources(vsi);
5413         if (err)
5414                 goto err_setup_tx;
5415         err = i40e_vsi_setup_rx_resources(vsi);
5416         if (err)
5417                 goto err_setup_rx;
5418
5419         err = i40e_vsi_configure(vsi);
5420         if (err)
5421                 goto err_setup_rx;
5422
5423         if (vsi->netdev) {
5424                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5425                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5426                 err = i40e_vsi_request_irq(vsi, int_name);
5427                 if (err)
5428                         goto err_setup_rx;
5429
5430                 /* Notify the stack of the actual queue counts. */
5431                 err = netif_set_real_num_tx_queues(vsi->netdev,
5432                                                    vsi->num_queue_pairs);
5433                 if (err)
5434                         goto err_set_queues;
5435
5436                 err = netif_set_real_num_rx_queues(vsi->netdev,
5437                                                    vsi->num_queue_pairs);
5438                 if (err)
5439                         goto err_set_queues;
5440
5441         } else if (vsi->type == I40E_VSI_FDIR) {
5442                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5443                          dev_driver_string(&pf->pdev->dev),
5444                          dev_name(&pf->pdev->dev));
5445                 err = i40e_vsi_request_irq(vsi, int_name);
5446
5447         } else {
5448                 err = -EINVAL;
5449                 goto err_setup_rx;
5450         }
5451
5452         err = i40e_up_complete(vsi);
5453         if (err)
5454                 goto err_up_complete;
5455
5456         return 0;
5457
5458 err_up_complete:
5459         i40e_down(vsi);
5460 err_set_queues:
5461         i40e_vsi_free_irq(vsi);
5462 err_setup_rx:
5463         i40e_vsi_free_rx_resources(vsi);
5464 err_setup_tx:
5465         i40e_vsi_free_tx_resources(vsi);
5466         if (vsi == pf->vsi[pf->lan_vsi])
5467                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5468
5469         return err;
5470 }
5471
5472 /**
5473  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5474  * @pf: Pointer to PF
5475  *
5476  * This function destroys the hlist where all the Flow Director
5477  * filters were saved.
5478  **/
5479 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5480 {
5481         struct i40e_fdir_filter *filter;
5482         struct hlist_node *node2;
5483
5484         hlist_for_each_entry_safe(filter, node2,
5485                                   &pf->fdir_filter_list, fdir_node) {
5486                 hlist_del(&filter->fdir_node);
5487                 kfree(filter);
5488         }
5489         pf->fdir_pf_active_filters = 0;
5490 }
5491
5492 /**
5493  * i40e_close - Disables a network interface
5494  * @netdev: network interface device structure
5495  *
5496  * The close entry point is called when an interface is de-activated
5497  * by the OS.  The hardware is still under the driver's control, but
5498  * this netdev interface is disabled.
5499  *
5500  * Returns 0, this is not allowed to fail
5501  **/
5502 int i40e_close(struct net_device *netdev)
5503 {
5504         struct i40e_netdev_priv *np = netdev_priv(netdev);
5505         struct i40e_vsi *vsi = np->vsi;
5506
5507         i40e_vsi_close(vsi);
5508
5509         return 0;
5510 }
5511
5512 /**
5513  * i40e_do_reset - Start a PF or Core Reset sequence
5514  * @pf: board private structure
5515  * @reset_flags: which reset is requested
5516  *
5517  * The essential difference in resets is that the PF Reset
5518  * doesn't clear the packet buffers, doesn't reset the PE
5519  * firmware, and doesn't bother the other PFs on the chip.
5520  **/
5521 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5522 {
5523         u32 val;
5524
5525         WARN_ON(in_interrupt());
5526
5527
5528         /* do the biggest reset indicated */
5529         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5530
5531                 /* Request a Global Reset
5532                  *
5533                  * This will start the chip's countdown to the actual full
5534                  * chip reset event, and a warning interrupt to be sent
5535                  * to all PFs, including the requestor.  Our handler
5536                  * for the warning interrupt will deal with the shutdown
5537                  * and recovery of the switch setup.
5538                  */
5539                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5540                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5541                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5542                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5543
5544         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5545
5546                 /* Request a Core Reset
5547                  *
5548                  * Same as Global Reset, except does *not* include the MAC/PHY
5549                  */
5550                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5551                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5552                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5553                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5554                 i40e_flush(&pf->hw);
5555
5556         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5557
5558                 /* Request a PF Reset
5559                  *
5560                  * Resets only the PF-specific registers
5561                  *
5562                  * This goes directly to the tear-down and rebuild of
5563                  * the switch, since we need to do all the recovery as
5564                  * for the Core Reset.
5565                  */
5566                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5567                 i40e_handle_reset_warning(pf);
5568
5569         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5570                 int v;
5571
5572                 /* Find the VSI(s) that requested a re-init */
5573                 dev_info(&pf->pdev->dev,
5574                          "VSI reinit requested\n");
5575                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5576                         struct i40e_vsi *vsi = pf->vsi[v];
5577
5578                         if (vsi != NULL &&
5579                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5580                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5581                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5582                         }
5583                 }
5584         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5585                 int v;
5586
5587                 /* Find the VSI(s) that needs to be brought down */
5588                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5589                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5590                         struct i40e_vsi *vsi = pf->vsi[v];
5591
5592                         if (vsi != NULL &&
5593                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5594                                 set_bit(__I40E_DOWN, &vsi->state);
5595                                 i40e_down(vsi);
5596                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5597                         }
5598                 }
5599         } else {
5600                 dev_info(&pf->pdev->dev,
5601                          "bad reset request 0x%08x\n", reset_flags);
5602         }
5603 }
5604
5605 #ifdef CONFIG_I40E_DCB
5606 /**
5607  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5608  * @pf: board private structure
5609  * @old_cfg: current DCB config
5610  * @new_cfg: new DCB config
5611  **/
5612 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5613                             struct i40e_dcbx_config *old_cfg,
5614                             struct i40e_dcbx_config *new_cfg)
5615 {
5616         bool need_reconfig = false;
5617
5618         /* Check if ETS configuration has changed */
5619         if (memcmp(&new_cfg->etscfg,
5620                    &old_cfg->etscfg,
5621                    sizeof(new_cfg->etscfg))) {
5622                 /* If Priority Table has changed reconfig is needed */
5623                 if (memcmp(&new_cfg->etscfg.prioritytable,
5624                            &old_cfg->etscfg.prioritytable,
5625                            sizeof(new_cfg->etscfg.prioritytable))) {
5626                         need_reconfig = true;
5627                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5628                 }
5629
5630                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5631                            &old_cfg->etscfg.tcbwtable,
5632                            sizeof(new_cfg->etscfg.tcbwtable)))
5633                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5634
5635                 if (memcmp(&new_cfg->etscfg.tsatable,
5636                            &old_cfg->etscfg.tsatable,
5637                            sizeof(new_cfg->etscfg.tsatable)))
5638                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5639         }
5640
5641         /* Check if PFC configuration has changed */
5642         if (memcmp(&new_cfg->pfc,
5643                    &old_cfg->pfc,
5644                    sizeof(new_cfg->pfc))) {
5645                 need_reconfig = true;
5646                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5647         }
5648
5649         /* Check if APP Table has changed */
5650         if (memcmp(&new_cfg->app,
5651                    &old_cfg->app,
5652                    sizeof(new_cfg->app))) {
5653                 need_reconfig = true;
5654                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5655         }
5656
5657         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5658         return need_reconfig;
5659 }
5660
5661 /**
5662  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5663  * @pf: board private structure
5664  * @e: event info posted on ARQ
5665  **/
5666 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5667                                   struct i40e_arq_event_info *e)
5668 {
5669         struct i40e_aqc_lldp_get_mib *mib =
5670                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5671         struct i40e_hw *hw = &pf->hw;
5672         struct i40e_dcbx_config tmp_dcbx_cfg;
5673         bool need_reconfig = false;
5674         int ret = 0;
5675         u8 type;
5676
5677         /* Not DCB capable or capability disabled */
5678         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5679                 return ret;
5680
5681         /* Ignore if event is not for Nearest Bridge */
5682         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5683                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5684         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5685         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5686                 return ret;
5687
5688         /* Check MIB Type and return if event for Remote MIB update */
5689         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5690         dev_dbg(&pf->pdev->dev,
5691                 "LLDP event mib type %s\n", type ? "remote" : "local");
5692         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5693                 /* Update the remote cached instance and return */
5694                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5695                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5696                                 &hw->remote_dcbx_config);
5697                 goto exit;
5698         }
5699
5700         /* Store the old configuration */
5701         tmp_dcbx_cfg = hw->local_dcbx_config;
5702
5703         /* Reset the old DCBx configuration data */
5704         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5705         /* Get updated DCBX data from firmware */
5706         ret = i40e_get_dcb_config(&pf->hw);
5707         if (ret) {
5708                 dev_info(&pf->pdev->dev,
5709                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5710                          i40e_stat_str(&pf->hw, ret),
5711                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5712                 goto exit;
5713         }
5714
5715         /* No change detected in DCBX configs */
5716         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5717                     sizeof(tmp_dcbx_cfg))) {
5718                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5719                 goto exit;
5720         }
5721
5722         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5723                                                &hw->local_dcbx_config);
5724
5725         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5726
5727         if (!need_reconfig)
5728                 goto exit;
5729
5730         /* Enable DCB tagging only when more than one TC */
5731         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5732                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5733         else
5734                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5735
5736         set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5737         /* Reconfiguration needed quiesce all VSIs */
5738         i40e_pf_quiesce_all_vsi(pf);
5739
5740         /* Changes in configuration update VEB/VSI */
5741         i40e_dcb_reconfigure(pf);
5742
5743         ret = i40e_resume_port_tx(pf);
5744
5745         clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5746         /* In case of error no point in resuming VSIs */
5747         if (ret)
5748                 goto exit;
5749
5750         /* Wait for the PF's queues to be disabled */
5751         ret = i40e_pf_wait_queues_disabled(pf);
5752         if (ret) {
5753                 /* Schedule PF reset to recover */
5754                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5755                 i40e_service_event_schedule(pf);
5756         } else {
5757                 i40e_pf_unquiesce_all_vsi(pf);
5758         }
5759
5760 exit:
5761         return ret;
5762 }
5763 #endif /* CONFIG_I40E_DCB */
5764
5765 /**
5766  * i40e_do_reset_safe - Protected reset path for userland calls.
5767  * @pf: board private structure
5768  * @reset_flags: which reset is requested
5769  *
5770  **/
5771 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5772 {
5773         rtnl_lock();
5774         i40e_do_reset(pf, reset_flags);
5775         rtnl_unlock();
5776 }
5777
5778 /**
5779  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5780  * @pf: board private structure
5781  * @e: event info posted on ARQ
5782  *
5783  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5784  * and VF queues
5785  **/
5786 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5787                                            struct i40e_arq_event_info *e)
5788 {
5789         struct i40e_aqc_lan_overflow *data =
5790                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5791         u32 queue = le32_to_cpu(data->prtdcb_rupto);
5792         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5793         struct i40e_hw *hw = &pf->hw;
5794         struct i40e_vf *vf;
5795         u16 vf_id;
5796
5797         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5798                 queue, qtx_ctl);
5799
5800         /* Queue belongs to VF, find the VF and issue VF reset */
5801         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5802             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5803                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5804                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5805                 vf_id -= hw->func_caps.vf_base_id;
5806                 vf = &pf->vf[vf_id];
5807                 i40e_vc_notify_vf_reset(vf);
5808                 /* Allow VF to process pending reset notification */
5809                 msleep(20);
5810                 i40e_reset_vf(vf, false);
5811         }
5812 }
5813
5814 /**
5815  * i40e_service_event_complete - Finish up the service event
5816  * @pf: board private structure
5817  **/
5818 static void i40e_service_event_complete(struct i40e_pf *pf)
5819 {
5820         WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5821
5822         /* flush memory to make sure state is correct before next watchog */
5823         smp_mb__before_atomic();
5824         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5825 }
5826
5827 /**
5828  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5829  * @pf: board private structure
5830  **/
5831 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5832 {
5833         u32 val, fcnt_prog;
5834
5835         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5836         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5837         return fcnt_prog;
5838 }
5839
5840 /**
5841  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5842  * @pf: board private structure
5843  **/
5844 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5845 {
5846         u32 val, fcnt_prog;
5847
5848         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5849         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5850                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5851                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5852         return fcnt_prog;
5853 }
5854
5855 /**
5856  * i40e_get_global_fd_count - Get total FD filters programmed on device
5857  * @pf: board private structure
5858  **/
5859 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5860 {
5861         u32 val, fcnt_prog;
5862
5863         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5864         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5865                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5866                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5867         return fcnt_prog;
5868 }
5869
5870 /**
5871  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5872  * @pf: board private structure
5873  **/
5874 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5875 {
5876         struct i40e_fdir_filter *filter;
5877         u32 fcnt_prog, fcnt_avail;
5878         struct hlist_node *node;
5879
5880         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5881                 return;
5882
5883         /* Check if, FD SB or ATR was auto disabled and if there is enough room
5884          * to re-enable
5885          */
5886         fcnt_prog = i40e_get_global_fd_count(pf);
5887         fcnt_avail = pf->fdir_pf_filter_count;
5888         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5889             (pf->fd_add_err == 0) ||
5890             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5891                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5892                     (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5893                         pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5894                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5895                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5896                 }
5897         }
5898         /* Wait for some more space to be available to turn on ATR */
5899         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5900                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5901                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5902                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5903                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5904                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5905                 }
5906         }
5907
5908         /* if hw had a problem adding a filter, delete it */
5909         if (pf->fd_inv > 0) {
5910                 hlist_for_each_entry_safe(filter, node,
5911                                           &pf->fdir_filter_list, fdir_node) {
5912                         if (filter->fd_id == pf->fd_inv) {
5913                                 hlist_del(&filter->fdir_node);
5914                                 kfree(filter);
5915                                 pf->fdir_pf_active_filters--;
5916                         }
5917                 }
5918         }
5919 }
5920
5921 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5922 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5923 /**
5924  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5925  * @pf: board private structure
5926  **/
5927 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5928 {
5929         unsigned long min_flush_time;
5930         int flush_wait_retry = 50;
5931         bool disable_atr = false;
5932         int fd_room;
5933         int reg;
5934
5935         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5936                 return;
5937
5938         if (!time_after(jiffies, pf->fd_flush_timestamp +
5939                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5940                 return;
5941
5942         /* If the flush is happening too quick and we have mostly SB rules we
5943          * should not re-enable ATR for some time.
5944          */
5945         min_flush_time = pf->fd_flush_timestamp +
5946                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5947         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5948
5949         if (!(time_after(jiffies, min_flush_time)) &&
5950             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5951                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5952                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5953                 disable_atr = true;
5954         }
5955
5956         pf->fd_flush_timestamp = jiffies;
5957         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5958         /* flush all filters */
5959         wr32(&pf->hw, I40E_PFQF_CTL_1,
5960              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5961         i40e_flush(&pf->hw);
5962         pf->fd_flush_cnt++;
5963         pf->fd_add_err = 0;
5964         do {
5965                 /* Check FD flush status every 5-6msec */
5966                 usleep_range(5000, 6000);
5967                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5968                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5969                         break;
5970         } while (flush_wait_retry--);
5971         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5972                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5973         } else {
5974                 /* replay sideband filters */
5975                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5976                 if (!disable_atr)
5977                         pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5978                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5979                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5980                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5981         }
5982
5983 }
5984
5985 /**
5986  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5987  * @pf: board private structure
5988  **/
5989 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5990 {
5991         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5992 }
5993
5994 /* We can see up to 256 filter programming desc in transit if the filters are
5995  * being applied really fast; before we see the first
5996  * filter miss error on Rx queue 0. Accumulating enough error messages before
5997  * reacting will make sure we don't cause flush too often.
5998  */
5999 #define I40E_MAX_FD_PROGRAM_ERROR 256
6000
6001 /**
6002  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6003  * @pf: board private structure
6004  **/
6005 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6006 {
6007
6008         /* if interface is down do nothing */
6009         if (test_bit(__I40E_DOWN, &pf->state))
6010                 return;
6011
6012         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
6013                 return;
6014
6015         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6016                 i40e_fdir_flush_and_replay(pf);
6017
6018         i40e_fdir_check_and_reenable(pf);
6019
6020 }
6021
6022 /**
6023  * i40e_vsi_link_event - notify VSI of a link event
6024  * @vsi: vsi to be notified
6025  * @link_up: link up or down
6026  **/
6027 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6028 {
6029         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6030                 return;
6031
6032         switch (vsi->type) {
6033         case I40E_VSI_MAIN:
6034 #ifdef I40E_FCOE
6035         case I40E_VSI_FCOE:
6036 #endif
6037                 if (!vsi->netdev || !vsi->netdev_registered)
6038                         break;
6039
6040                 if (link_up) {
6041                         netif_carrier_on(vsi->netdev);
6042                         netif_tx_wake_all_queues(vsi->netdev);
6043                 } else {
6044                         netif_carrier_off(vsi->netdev);
6045                         netif_tx_stop_all_queues(vsi->netdev);
6046                 }
6047                 break;
6048
6049         case I40E_VSI_SRIOV:
6050         case I40E_VSI_VMDQ2:
6051         case I40E_VSI_CTRL:
6052         case I40E_VSI_IWARP:
6053         case I40E_VSI_MIRROR:
6054         default:
6055                 /* there is no notification for other VSIs */
6056                 break;
6057         }
6058 }
6059
6060 /**
6061  * i40e_veb_link_event - notify elements on the veb of a link event
6062  * @veb: veb to be notified
6063  * @link_up: link up or down
6064  **/
6065 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6066 {
6067         struct i40e_pf *pf;
6068         int i;
6069
6070         if (!veb || !veb->pf)
6071                 return;
6072         pf = veb->pf;
6073
6074         /* depth first... */
6075         for (i = 0; i < I40E_MAX_VEB; i++)
6076                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6077                         i40e_veb_link_event(pf->veb[i], link_up);
6078
6079         /* ... now the local VSIs */
6080         for (i = 0; i < pf->num_alloc_vsi; i++)
6081                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6082                         i40e_vsi_link_event(pf->vsi[i], link_up);
6083 }
6084
6085 /**
6086  * i40e_link_event - Update netif_carrier status
6087  * @pf: board private structure
6088  **/
6089 static void i40e_link_event(struct i40e_pf *pf)
6090 {
6091         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6092         u8 new_link_speed, old_link_speed;
6093         i40e_status status;
6094         bool new_link, old_link;
6095
6096         /* save off old link status information */
6097         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6098
6099         /* set this to force the get_link_status call to refresh state */
6100         pf->hw.phy.get_link_info = true;
6101
6102         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6103
6104         status = i40e_get_link_status(&pf->hw, &new_link);
6105         if (status) {
6106                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6107                         status);
6108                 return;
6109         }
6110
6111         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6112         new_link_speed = pf->hw.phy.link_info.link_speed;
6113
6114         if (new_link == old_link &&
6115             new_link_speed == old_link_speed &&
6116             (test_bit(__I40E_DOWN, &vsi->state) ||
6117              new_link == netif_carrier_ok(vsi->netdev)))
6118                 return;
6119
6120         if (!test_bit(__I40E_DOWN, &vsi->state))
6121                 i40e_print_link_message(vsi, new_link);
6122
6123         /* Notify the base of the switch tree connected to
6124          * the link.  Floating VEBs are not notified.
6125          */
6126         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6127                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6128         else
6129                 i40e_vsi_link_event(vsi, new_link);
6130
6131         if (pf->vf)
6132                 i40e_vc_notify_link_state(pf);
6133
6134         if (pf->flags & I40E_FLAG_PTP)
6135                 i40e_ptp_set_increment(pf);
6136 }
6137
6138 /**
6139  * i40e_watchdog_subtask - periodic checks not using event driven response
6140  * @pf: board private structure
6141  **/
6142 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6143 {
6144         int i;
6145
6146         /* if interface is down do nothing */
6147         if (test_bit(__I40E_DOWN, &pf->state) ||
6148             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6149                 return;
6150
6151         /* make sure we don't do these things too often */
6152         if (time_before(jiffies, (pf->service_timer_previous +
6153                                   pf->service_timer_period)))
6154                 return;
6155         pf->service_timer_previous = jiffies;
6156
6157         if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6158                 i40e_link_event(pf);
6159
6160         /* Update the stats for active netdevs so the network stack
6161          * can look at updated numbers whenever it cares to
6162          */
6163         for (i = 0; i < pf->num_alloc_vsi; i++)
6164                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6165                         i40e_update_stats(pf->vsi[i]);
6166
6167         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6168                 /* Update the stats for the active switching components */
6169                 for (i = 0; i < I40E_MAX_VEB; i++)
6170                         if (pf->veb[i])
6171                                 i40e_update_veb_stats(pf->veb[i]);
6172         }
6173
6174         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6175 }
6176
6177 /**
6178  * i40e_reset_subtask - Set up for resetting the device and driver
6179  * @pf: board private structure
6180  **/
6181 static void i40e_reset_subtask(struct i40e_pf *pf)
6182 {
6183         u32 reset_flags = 0;
6184
6185         rtnl_lock();
6186         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6187                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6188                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6189         }
6190         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6191                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6192                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6193         }
6194         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6195                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6196                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6197         }
6198         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6199                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6200                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6201         }
6202         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6203                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6204                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6205         }
6206
6207         /* If there's a recovery already waiting, it takes
6208          * precedence before starting a new reset sequence.
6209          */
6210         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6211                 i40e_handle_reset_warning(pf);
6212                 goto unlock;
6213         }
6214
6215         /* If we're already down or resetting, just bail */
6216         if (reset_flags &&
6217             !test_bit(__I40E_DOWN, &pf->state) &&
6218             !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6219                 i40e_do_reset(pf, reset_flags);
6220
6221 unlock:
6222         rtnl_unlock();
6223 }
6224
6225 /**
6226  * i40e_handle_link_event - Handle link event
6227  * @pf: board private structure
6228  * @e: event info posted on ARQ
6229  **/
6230 static void i40e_handle_link_event(struct i40e_pf *pf,
6231                                    struct i40e_arq_event_info *e)
6232 {
6233         struct i40e_aqc_get_link_status *status =
6234                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6235
6236         /* Do a new status request to re-enable LSE reporting
6237          * and load new status information into the hw struct
6238          * This completely ignores any state information
6239          * in the ARQ event info, instead choosing to always
6240          * issue the AQ update link status command.
6241          */
6242         i40e_link_event(pf);
6243
6244         /* check for unqualified module, if link is down */
6245         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6246             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6247             (!(status->link_info & I40E_AQ_LINK_UP)))
6248                 dev_err(&pf->pdev->dev,
6249                         "The driver failed to link because an unqualified module was detected.\n");
6250 }
6251
6252 /**
6253  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6254  * @pf: board private structure
6255  **/
6256 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6257 {
6258         struct i40e_arq_event_info event;
6259         struct i40e_hw *hw = &pf->hw;
6260         u16 pending, i = 0;
6261         i40e_status ret;
6262         u16 opcode;
6263         u32 oldval;
6264         u32 val;
6265
6266         /* Do not run clean AQ when PF reset fails */
6267         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6268                 return;
6269
6270         /* check for error indications */
6271         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6272         oldval = val;
6273         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6274                 if (hw->debug_mask & I40E_DEBUG_AQ)
6275                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6276                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6277         }
6278         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6279                 if (hw->debug_mask & I40E_DEBUG_AQ)
6280                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6281                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6282                 pf->arq_overflows++;
6283         }
6284         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6285                 if (hw->debug_mask & I40E_DEBUG_AQ)
6286                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6287                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6288         }
6289         if (oldval != val)
6290                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6291
6292         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6293         oldval = val;
6294         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6295                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6296                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6297                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6298         }
6299         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6300                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6301                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6302                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6303         }
6304         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6305                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6306                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6307                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6308         }
6309         if (oldval != val)
6310                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6311
6312         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6313         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6314         if (!event.msg_buf)
6315                 return;
6316
6317         do {
6318                 ret = i40e_clean_arq_element(hw, &event, &pending);
6319                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6320                         break;
6321                 else if (ret) {
6322                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6323                         break;
6324                 }
6325
6326                 opcode = le16_to_cpu(event.desc.opcode);
6327                 switch (opcode) {
6328
6329                 case i40e_aqc_opc_get_link_status:
6330                         i40e_handle_link_event(pf, &event);
6331                         break;
6332                 case i40e_aqc_opc_send_msg_to_pf:
6333                         ret = i40e_vc_process_vf_msg(pf,
6334                                         le16_to_cpu(event.desc.retval),
6335                                         le32_to_cpu(event.desc.cookie_high),
6336                                         le32_to_cpu(event.desc.cookie_low),
6337                                         event.msg_buf,
6338                                         event.msg_len);
6339                         break;
6340                 case i40e_aqc_opc_lldp_update_mib:
6341                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6342 #ifdef CONFIG_I40E_DCB
6343                         rtnl_lock();
6344                         ret = i40e_handle_lldp_event(pf, &event);
6345                         rtnl_unlock();
6346 #endif /* CONFIG_I40E_DCB */
6347                         break;
6348                 case i40e_aqc_opc_event_lan_overflow:
6349                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6350                         i40e_handle_lan_overflow_event(pf, &event);
6351                         break;
6352                 case i40e_aqc_opc_send_msg_to_peer:
6353                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6354                         break;
6355                 case i40e_aqc_opc_nvm_erase:
6356                 case i40e_aqc_opc_nvm_update:
6357                 case i40e_aqc_opc_oem_post_update:
6358                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6359                                    "ARQ NVM operation 0x%04x completed\n",
6360                                    opcode);
6361                         break;
6362                 default:
6363                         dev_info(&pf->pdev->dev,
6364                                  "ARQ: Unknown event 0x%04x ignored\n",
6365                                  opcode);
6366                         break;
6367                 }
6368         } while (pending && (i++ < pf->adminq_work_limit));
6369
6370         clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6371         /* re-enable Admin queue interrupt cause */
6372         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6373         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6374         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6375         i40e_flush(hw);
6376
6377         kfree(event.msg_buf);
6378 }
6379
6380 /**
6381  * i40e_verify_eeprom - make sure eeprom is good to use
6382  * @pf: board private structure
6383  **/
6384 static void i40e_verify_eeprom(struct i40e_pf *pf)
6385 {
6386         int err;
6387
6388         err = i40e_diag_eeprom_test(&pf->hw);
6389         if (err) {
6390                 /* retry in case of garbage read */
6391                 err = i40e_diag_eeprom_test(&pf->hw);
6392                 if (err) {
6393                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6394                                  err);
6395                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6396                 }
6397         }
6398
6399         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6400                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6401                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6402         }
6403 }
6404
6405 /**
6406  * i40e_enable_pf_switch_lb
6407  * @pf: pointer to the PF structure
6408  *
6409  * enable switch loop back or die - no point in a return value
6410  **/
6411 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6412 {
6413         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6414         struct i40e_vsi_context ctxt;
6415         int ret;
6416
6417         ctxt.seid = pf->main_vsi_seid;
6418         ctxt.pf_num = pf->hw.pf_id;
6419         ctxt.vf_num = 0;
6420         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6421         if (ret) {
6422                 dev_info(&pf->pdev->dev,
6423                          "couldn't get PF vsi config, err %s aq_err %s\n",
6424                          i40e_stat_str(&pf->hw, ret),
6425                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6426                 return;
6427         }
6428         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6429         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6430         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6431
6432         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6433         if (ret) {
6434                 dev_info(&pf->pdev->dev,
6435                          "update vsi switch failed, err %s aq_err %s\n",
6436                          i40e_stat_str(&pf->hw, ret),
6437                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6438         }
6439 }
6440
6441 /**
6442  * i40e_disable_pf_switch_lb
6443  * @pf: pointer to the PF structure
6444  *
6445  * disable switch loop back or die - no point in a return value
6446  **/
6447 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6448 {
6449         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6450         struct i40e_vsi_context ctxt;
6451         int ret;
6452
6453         ctxt.seid = pf->main_vsi_seid;
6454         ctxt.pf_num = pf->hw.pf_id;
6455         ctxt.vf_num = 0;
6456         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6457         if (ret) {
6458                 dev_info(&pf->pdev->dev,
6459                          "couldn't get PF vsi config, err %s aq_err %s\n",
6460                          i40e_stat_str(&pf->hw, ret),
6461                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6462                 return;
6463         }
6464         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6465         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6466         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6467
6468         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6469         if (ret) {
6470                 dev_info(&pf->pdev->dev,
6471                          "update vsi switch failed, err %s aq_err %s\n",
6472                          i40e_stat_str(&pf->hw, ret),
6473                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6474         }
6475 }
6476
6477 /**
6478  * i40e_config_bridge_mode - Configure the HW bridge mode
6479  * @veb: pointer to the bridge instance
6480  *
6481  * Configure the loop back mode for the LAN VSI that is downlink to the
6482  * specified HW bridge instance. It is expected this function is called
6483  * when a new HW bridge is instantiated.
6484  **/
6485 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6486 {
6487         struct i40e_pf *pf = veb->pf;
6488
6489         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6490                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6491                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6492         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6493                 i40e_disable_pf_switch_lb(pf);
6494         else
6495                 i40e_enable_pf_switch_lb(pf);
6496 }
6497
6498 /**
6499  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6500  * @veb: pointer to the VEB instance
6501  *
6502  * This is a recursive function that first builds the attached VSIs then
6503  * recurses in to build the next layer of VEB.  We track the connections
6504  * through our own index numbers because the seid's from the HW could
6505  * change across the reset.
6506  **/
6507 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6508 {
6509         struct i40e_vsi *ctl_vsi = NULL;
6510         struct i40e_pf *pf = veb->pf;
6511         int v, veb_idx;
6512         int ret;
6513
6514         /* build VSI that owns this VEB, temporarily attached to base VEB */
6515         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6516                 if (pf->vsi[v] &&
6517                     pf->vsi[v]->veb_idx == veb->idx &&
6518                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6519                         ctl_vsi = pf->vsi[v];
6520                         break;
6521                 }
6522         }
6523         if (!ctl_vsi) {
6524                 dev_info(&pf->pdev->dev,
6525                          "missing owner VSI for veb_idx %d\n", veb->idx);
6526                 ret = -ENOENT;
6527                 goto end_reconstitute;
6528         }
6529         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6530                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6531         ret = i40e_add_vsi(ctl_vsi);
6532         if (ret) {
6533                 dev_info(&pf->pdev->dev,
6534                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6535                          veb->idx, ret);
6536                 goto end_reconstitute;
6537         }
6538         i40e_vsi_reset_stats(ctl_vsi);
6539
6540         /* create the VEB in the switch and move the VSI onto the VEB */
6541         ret = i40e_add_veb(veb, ctl_vsi);
6542         if (ret)
6543                 goto end_reconstitute;
6544
6545         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6546                 veb->bridge_mode = BRIDGE_MODE_VEB;
6547         else
6548                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6549         i40e_config_bridge_mode(veb);
6550
6551         /* create the remaining VSIs attached to this VEB */
6552         for (v = 0; v < pf->num_alloc_vsi; v++) {
6553                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6554                         continue;
6555
6556                 if (pf->vsi[v]->veb_idx == veb->idx) {
6557                         struct i40e_vsi *vsi = pf->vsi[v];
6558
6559                         vsi->uplink_seid = veb->seid;
6560                         ret = i40e_add_vsi(vsi);
6561                         if (ret) {
6562                                 dev_info(&pf->pdev->dev,
6563                                          "rebuild of vsi_idx %d failed: %d\n",
6564                                          v, ret);
6565                                 goto end_reconstitute;
6566                         }
6567                         i40e_vsi_reset_stats(vsi);
6568                 }
6569         }
6570
6571         /* create any VEBs attached to this VEB - RECURSION */
6572         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6573                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6574                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6575                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6576                         if (ret)
6577                                 break;
6578                 }
6579         }
6580
6581 end_reconstitute:
6582         return ret;
6583 }
6584
6585 /**
6586  * i40e_get_capabilities - get info about the HW
6587  * @pf: the PF struct
6588  **/
6589 static int i40e_get_capabilities(struct i40e_pf *pf)
6590 {
6591         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6592         u16 data_size;
6593         int buf_len;
6594         int err;
6595
6596         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6597         do {
6598                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6599                 if (!cap_buf)
6600                         return -ENOMEM;
6601
6602                 /* this loads the data into the hw struct for us */
6603                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6604                                             &data_size,
6605                                             i40e_aqc_opc_list_func_capabilities,
6606                                             NULL);
6607                 /* data loaded, buffer no longer needed */
6608                 kfree(cap_buf);
6609
6610                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6611                         /* retry with a larger buffer */
6612                         buf_len = data_size;
6613                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6614                         dev_info(&pf->pdev->dev,
6615                                  "capability discovery failed, err %s aq_err %s\n",
6616                                  i40e_stat_str(&pf->hw, err),
6617                                  i40e_aq_str(&pf->hw,
6618                                              pf->hw.aq.asq_last_status));
6619                         return -ENODEV;
6620                 }
6621         } while (err);
6622
6623         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6624                 dev_info(&pf->pdev->dev,
6625                          "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",
6626                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6627                          pf->hw.func_caps.num_msix_vectors,
6628                          pf->hw.func_caps.num_msix_vectors_vf,
6629                          pf->hw.func_caps.fd_filters_guaranteed,
6630                          pf->hw.func_caps.fd_filters_best_effort,
6631                          pf->hw.func_caps.num_tx_qp,
6632                          pf->hw.func_caps.num_vsis);
6633
6634 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6635                        + pf->hw.func_caps.num_vfs)
6636         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6637                 dev_info(&pf->pdev->dev,
6638                          "got num_vsis %d, setting num_vsis to %d\n",
6639                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6640                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6641         }
6642
6643         return 0;
6644 }
6645
6646 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6647
6648 /**
6649  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6650  * @pf: board private structure
6651  **/
6652 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6653 {
6654         struct i40e_vsi *vsi;
6655         int i;
6656
6657         /* quick workaround for an NVM issue that leaves a critical register
6658          * uninitialized
6659          */
6660         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6661                 static const u32 hkey[] = {
6662                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6663                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6664                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6665                         0x95b3a76d};
6666
6667                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6668                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6669         }
6670
6671         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6672                 return;
6673
6674         /* find existing VSI and see if it needs configuring */
6675         vsi = NULL;
6676         for (i = 0; i < pf->num_alloc_vsi; i++) {
6677                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6678                         vsi = pf->vsi[i];
6679                         break;
6680                 }
6681         }
6682
6683         /* create a new VSI if none exists */
6684         if (!vsi) {
6685                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6686                                      pf->vsi[pf->lan_vsi]->seid, 0);
6687                 if (!vsi) {
6688                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6689                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6690                         return;
6691                 }
6692         }
6693
6694         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6695 }
6696
6697 /**
6698  * i40e_fdir_teardown - release the Flow Director resources
6699  * @pf: board private structure
6700  **/
6701 static void i40e_fdir_teardown(struct i40e_pf *pf)
6702 {
6703         int i;
6704
6705         i40e_fdir_filter_exit(pf);
6706         for (i = 0; i < pf->num_alloc_vsi; i++) {
6707                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6708                         i40e_vsi_release(pf->vsi[i]);
6709                         break;
6710                 }
6711         }
6712 }
6713
6714 /**
6715  * i40e_prep_for_reset - prep for the core to reset
6716  * @pf: board private structure
6717  *
6718  * Close up the VFs and other things in prep for PF Reset.
6719   **/
6720 static void i40e_prep_for_reset(struct i40e_pf *pf)
6721 {
6722         struct i40e_hw *hw = &pf->hw;
6723         i40e_status ret = 0;
6724         u32 v;
6725
6726         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6727         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6728                 return;
6729         if (i40e_check_asq_alive(&pf->hw))
6730                 i40e_vc_notify_reset(pf);
6731
6732         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6733
6734         /* quiesce the VSIs and their queues that are not already DOWN */
6735         i40e_pf_quiesce_all_vsi(pf);
6736
6737         for (v = 0; v < pf->num_alloc_vsi; v++) {
6738                 if (pf->vsi[v])
6739                         pf->vsi[v]->seid = 0;
6740         }
6741
6742         i40e_shutdown_adminq(&pf->hw);
6743
6744         /* call shutdown HMC */
6745         if (hw->hmc.hmc_obj) {
6746                 ret = i40e_shutdown_lan_hmc(hw);
6747                 if (ret)
6748                         dev_warn(&pf->pdev->dev,
6749                                  "shutdown_lan_hmc failed: %d\n", ret);
6750         }
6751 }
6752
6753 /**
6754  * i40e_send_version - update firmware with driver version
6755  * @pf: PF struct
6756  */
6757 static void i40e_send_version(struct i40e_pf *pf)
6758 {
6759         struct i40e_driver_version dv;
6760
6761         dv.major_version = DRV_VERSION_MAJOR;
6762         dv.minor_version = DRV_VERSION_MINOR;
6763         dv.build_version = DRV_VERSION_BUILD;
6764         dv.subbuild_version = 0;
6765         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6766         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6767 }
6768
6769 /**
6770  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6771  * @pf: board private structure
6772  * @reinit: if the Main VSI needs to re-initialized.
6773  **/
6774 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6775 {
6776         struct i40e_hw *hw = &pf->hw;
6777         u8 set_fc_aq_fail = 0;
6778         i40e_status ret;
6779         u32 val;
6780         u32 v;
6781
6782         /* Now we wait for GRST to settle out.
6783          * We don't have to delete the VEBs or VSIs from the hw switch
6784          * because the reset will make them disappear.
6785          */
6786         ret = i40e_pf_reset(hw);
6787         if (ret) {
6788                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6789                 set_bit(__I40E_RESET_FAILED, &pf->state);
6790                 goto clear_recovery;
6791         }
6792         pf->pfr_count++;
6793
6794         if (test_bit(__I40E_DOWN, &pf->state))
6795                 goto clear_recovery;
6796         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6797
6798         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6799         ret = i40e_init_adminq(&pf->hw);
6800         if (ret) {
6801                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6802                          i40e_stat_str(&pf->hw, ret),
6803                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6804                 goto clear_recovery;
6805         }
6806
6807         /* re-verify the eeprom if we just had an EMP reset */
6808         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6809                 i40e_verify_eeprom(pf);
6810
6811         i40e_clear_pxe_mode(hw);
6812         ret = i40e_get_capabilities(pf);
6813         if (ret)
6814                 goto end_core_reset;
6815
6816         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6817                                 hw->func_caps.num_rx_qp,
6818                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6819         if (ret) {
6820                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6821                 goto end_core_reset;
6822         }
6823         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6824         if (ret) {
6825                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6826                 goto end_core_reset;
6827         }
6828
6829 #ifdef CONFIG_I40E_DCB
6830         ret = i40e_init_pf_dcb(pf);
6831         if (ret) {
6832                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6833                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6834                 /* Continue without DCB enabled */
6835         }
6836 #endif /* CONFIG_I40E_DCB */
6837 #ifdef I40E_FCOE
6838         i40e_init_pf_fcoe(pf);
6839
6840 #endif
6841         /* do basic switch setup */
6842         ret = i40e_setup_pf_switch(pf, reinit);
6843         if (ret)
6844                 goto end_core_reset;
6845
6846         /* The driver only wants link up/down and module qualification
6847          * reports from firmware.  Note the negative logic.
6848          */
6849         ret = i40e_aq_set_phy_int_mask(&pf->hw,
6850                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
6851                                          I40E_AQ_EVENT_MEDIA_NA |
6852                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6853         if (ret)
6854                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6855                          i40e_stat_str(&pf->hw, ret),
6856                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6857
6858         /* make sure our flow control settings are restored */
6859         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6860         if (ret)
6861                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6862                         i40e_stat_str(&pf->hw, ret),
6863                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6864
6865         /* Rebuild the VSIs and VEBs that existed before reset.
6866          * They are still in our local switch element arrays, so only
6867          * need to rebuild the switch model in the HW.
6868          *
6869          * If there were VEBs but the reconstitution failed, we'll try
6870          * try to recover minimal use by getting the basic PF VSI working.
6871          */
6872         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6873                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6874                 /* find the one VEB connected to the MAC, and find orphans */
6875                 for (v = 0; v < I40E_MAX_VEB; v++) {
6876                         if (!pf->veb[v])
6877                                 continue;
6878
6879                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6880                             pf->veb[v]->uplink_seid == 0) {
6881                                 ret = i40e_reconstitute_veb(pf->veb[v]);
6882
6883                                 if (!ret)
6884                                         continue;
6885
6886                                 /* If Main VEB failed, we're in deep doodoo,
6887                                  * so give up rebuilding the switch and set up
6888                                  * for minimal rebuild of PF VSI.
6889                                  * If orphan failed, we'll report the error
6890                                  * but try to keep going.
6891                                  */
6892                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6893                                         dev_info(&pf->pdev->dev,
6894                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6895                                                  ret);
6896                                         pf->vsi[pf->lan_vsi]->uplink_seid
6897                                                                 = pf->mac_seid;
6898                                         break;
6899                                 } else if (pf->veb[v]->uplink_seid == 0) {
6900                                         dev_info(&pf->pdev->dev,
6901                                                  "rebuild of orphan VEB failed: %d\n",
6902                                                  ret);
6903                                 }
6904                         }
6905                 }
6906         }
6907
6908         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6909                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6910                 /* no VEB, so rebuild only the Main VSI */
6911                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6912                 if (ret) {
6913                         dev_info(&pf->pdev->dev,
6914                                  "rebuild of Main VSI failed: %d\n", ret);
6915                         goto end_core_reset;
6916                 }
6917         }
6918
6919         /* Reconfigure hardware for allowing smaller MSS in the case
6920          * of TSO, so that we avoid the MDD being fired and causing
6921          * a reset in the case of small MSS+TSO.
6922          */
6923 #define I40E_REG_MSS          0x000E64DC
6924 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6925 #define I40E_64BYTE_MSS       0x400000
6926         val = rd32(hw, I40E_REG_MSS);
6927         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6928                 val &= ~I40E_REG_MSS_MIN_MASK;
6929                 val |= I40E_64BYTE_MSS;
6930                 wr32(hw, I40E_REG_MSS, val);
6931         }
6932
6933         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6934                 msleep(75);
6935                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6936                 if (ret)
6937                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6938                                  i40e_stat_str(&pf->hw, ret),
6939                                  i40e_aq_str(&pf->hw,
6940                                              pf->hw.aq.asq_last_status));
6941         }
6942         /* reinit the misc interrupt */
6943         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6944                 ret = i40e_setup_misc_vector(pf);
6945
6946         /* Add a filter to drop all Flow control frames from any VSI from being
6947          * transmitted. By doing so we stop a malicious VF from sending out
6948          * PAUSE or PFC frames and potentially controlling traffic for other
6949          * PF/VF VSIs.
6950          * The FW can still send Flow control frames if enabled.
6951          */
6952         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6953                                                        pf->main_vsi_seid);
6954
6955         /* restart the VSIs that were rebuilt and running before the reset */
6956         i40e_pf_unquiesce_all_vsi(pf);
6957
6958         if (pf->num_alloc_vfs) {
6959                 for (v = 0; v < pf->num_alloc_vfs; v++)
6960                         i40e_reset_vf(&pf->vf[v], true);
6961         }
6962
6963         /* tell the firmware that we're starting */
6964         i40e_send_version(pf);
6965
6966 end_core_reset:
6967         clear_bit(__I40E_RESET_FAILED, &pf->state);
6968 clear_recovery:
6969         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6970 }
6971
6972 /**
6973  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6974  * @pf: board private structure
6975  *
6976  * Close up the VFs and other things in prep for a Core Reset,
6977  * then get ready to rebuild the world.
6978  **/
6979 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6980 {
6981         i40e_prep_for_reset(pf);
6982         i40e_reset_and_rebuild(pf, false);
6983 }
6984
6985 /**
6986  * i40e_handle_mdd_event
6987  * @pf: pointer to the PF structure
6988  *
6989  * Called from the MDD irq handler to identify possibly malicious vfs
6990  **/
6991 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6992 {
6993         struct i40e_hw *hw = &pf->hw;
6994         bool mdd_detected = false;
6995         bool pf_mdd_detected = false;
6996         struct i40e_vf *vf;
6997         u32 reg;
6998         int i;
6999
7000         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7001                 return;
7002
7003         /* find what triggered the MDD event */
7004         reg = rd32(hw, I40E_GL_MDET_TX);
7005         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7006                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7007                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7008                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7009                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7010                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7011                                 I40E_GL_MDET_TX_EVENT_SHIFT;
7012                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7013                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7014                                 pf->hw.func_caps.base_queue;
7015                 if (netif_msg_tx_err(pf))
7016                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7017                                  event, queue, pf_num, vf_num);
7018                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7019                 mdd_detected = true;
7020         }
7021         reg = rd32(hw, I40E_GL_MDET_RX);
7022         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7023                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7024                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7025                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7026                                 I40E_GL_MDET_RX_EVENT_SHIFT;
7027                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7028                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7029                                 pf->hw.func_caps.base_queue;
7030                 if (netif_msg_rx_err(pf))
7031                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7032                                  event, queue, func);
7033                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7034                 mdd_detected = true;
7035         }
7036
7037         if (mdd_detected) {
7038                 reg = rd32(hw, I40E_PF_MDET_TX);
7039                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7040                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7041                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7042                         pf_mdd_detected = true;
7043                 }
7044                 reg = rd32(hw, I40E_PF_MDET_RX);
7045                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7046                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7047                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7048                         pf_mdd_detected = true;
7049                 }
7050                 /* Queue belongs to the PF, initiate a reset */
7051                 if (pf_mdd_detected) {
7052                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7053                         i40e_service_event_schedule(pf);
7054                 }
7055         }
7056
7057         /* see if one of the VFs needs its hand slapped */
7058         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7059                 vf = &(pf->vf[i]);
7060                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7061                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7062                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7063                         vf->num_mdd_events++;
7064                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7065                                  i);
7066                 }
7067
7068                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7069                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7070                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7071                         vf->num_mdd_events++;
7072                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7073                                  i);
7074                 }
7075
7076                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7077                         dev_info(&pf->pdev->dev,
7078                                  "Too many MDD events on VF %d, disabled\n", i);
7079                         dev_info(&pf->pdev->dev,
7080                                  "Use PF Control I/F to re-enable the VF\n");
7081                         set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7082                 }
7083         }
7084
7085         /* re-enable mdd interrupt cause */
7086         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7087         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7088         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7089         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7090         i40e_flush(hw);
7091 }
7092
7093 /**
7094  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7095  * @pf: board private structure
7096  **/
7097 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7098 {
7099 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7100         struct i40e_hw *hw = &pf->hw;
7101         i40e_status ret;
7102         __be16 port;
7103         int i;
7104
7105         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7106                 return;
7107
7108         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7109
7110         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7111                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7112                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7113                         port = pf->udp_ports[i].index;
7114                         if (port)
7115                                 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7116                                                      pf->udp_ports[i].type,
7117                                                      NULL, NULL);
7118                         else
7119                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7120
7121                         if (ret) {
7122                                 dev_dbg(&pf->pdev->dev,
7123                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
7124                                         pf->udp_ports[i].type ? "vxlan" : "geneve",
7125                                         port ? "add" : "delete",
7126                                         ntohs(port), i,
7127                                         i40e_stat_str(&pf->hw, ret),
7128                                         i40e_aq_str(&pf->hw,
7129                                                     pf->hw.aq.asq_last_status));
7130                                 pf->udp_ports[i].index = 0;
7131                         }
7132                 }
7133         }
7134 #endif
7135 }
7136
7137 /**
7138  * i40e_service_task - Run the driver's async subtasks
7139  * @work: pointer to work_struct containing our data
7140  **/
7141 static void i40e_service_task(struct work_struct *work)
7142 {
7143         struct i40e_pf *pf = container_of(work,
7144                                           struct i40e_pf,
7145                                           service_task);
7146         unsigned long start_time = jiffies;
7147
7148         /* don't bother with service tasks if a reset is in progress */
7149         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7150                 i40e_service_event_complete(pf);
7151                 return;
7152         }
7153
7154         i40e_detect_recover_hung(pf);
7155         i40e_sync_filters_subtask(pf);
7156         i40e_reset_subtask(pf);
7157         i40e_handle_mdd_event(pf);
7158         i40e_vc_process_vflr_event(pf);
7159         i40e_watchdog_subtask(pf);
7160         i40e_fdir_reinit_subtask(pf);
7161         i40e_client_subtask(pf);
7162         i40e_sync_filters_subtask(pf);
7163         i40e_sync_udp_filters_subtask(pf);
7164         i40e_clean_adminq_subtask(pf);
7165
7166         i40e_service_event_complete(pf);
7167
7168         /* If the tasks have taken longer than one timer cycle or there
7169          * is more work to be done, reschedule the service task now
7170          * rather than wait for the timer to tick again.
7171          */
7172         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7173             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7174             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7175             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7176                 i40e_service_event_schedule(pf);
7177 }
7178
7179 /**
7180  * i40e_service_timer - timer callback
7181  * @data: pointer to PF struct
7182  **/
7183 static void i40e_service_timer(unsigned long data)
7184 {
7185         struct i40e_pf *pf = (struct i40e_pf *)data;
7186
7187         mod_timer(&pf->service_timer,
7188                   round_jiffies(jiffies + pf->service_timer_period));
7189         i40e_service_event_schedule(pf);
7190 }
7191
7192 /**
7193  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7194  * @vsi: the VSI being configured
7195  **/
7196 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7197 {
7198         struct i40e_pf *pf = vsi->back;
7199
7200         switch (vsi->type) {
7201         case I40E_VSI_MAIN:
7202                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7203                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7204                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7205                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7206                         vsi->num_q_vectors = pf->num_lan_msix;
7207                 else
7208                         vsi->num_q_vectors = 1;
7209
7210                 break;
7211
7212         case I40E_VSI_FDIR:
7213                 vsi->alloc_queue_pairs = 1;
7214                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7215                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7216                 vsi->num_q_vectors = 1;
7217                 break;
7218
7219         case I40E_VSI_VMDQ2:
7220                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7221                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7222                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7223                 vsi->num_q_vectors = pf->num_vmdq_msix;
7224                 break;
7225
7226         case I40E_VSI_SRIOV:
7227                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7228                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7229                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7230                 break;
7231
7232 #ifdef I40E_FCOE
7233         case I40E_VSI_FCOE:
7234                 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7235                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7236                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7237                 vsi->num_q_vectors = pf->num_fcoe_msix;
7238                 break;
7239
7240 #endif /* I40E_FCOE */
7241         default:
7242                 WARN_ON(1);
7243                 return -ENODATA;
7244         }
7245
7246         return 0;
7247 }
7248
7249 /**
7250  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7251  * @type: VSI pointer
7252  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7253  *
7254  * On error: returns error code (negative)
7255  * On success: returns 0
7256  **/
7257 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7258 {
7259         int size;
7260         int ret = 0;
7261
7262         /* allocate memory for both Tx and Rx ring pointers */
7263         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7264         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7265         if (!vsi->tx_rings)
7266                 return -ENOMEM;
7267         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7268
7269         if (alloc_qvectors) {
7270                 /* allocate memory for q_vector pointers */
7271                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7272                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7273                 if (!vsi->q_vectors) {
7274                         ret = -ENOMEM;
7275                         goto err_vectors;
7276                 }
7277         }
7278         return ret;
7279
7280 err_vectors:
7281         kfree(vsi->tx_rings);
7282         return ret;
7283 }
7284
7285 /**
7286  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7287  * @pf: board private structure
7288  * @type: type of VSI
7289  *
7290  * On error: returns error code (negative)
7291  * On success: returns vsi index in PF (positive)
7292  **/
7293 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7294 {
7295         int ret = -ENODEV;
7296         struct i40e_vsi *vsi;
7297         int vsi_idx;
7298         int i;
7299
7300         /* Need to protect the allocation of the VSIs at the PF level */
7301         mutex_lock(&pf->switch_mutex);
7302
7303         /* VSI list may be fragmented if VSI creation/destruction has
7304          * been happening.  We can afford to do a quick scan to look
7305          * for any free VSIs in the list.
7306          *
7307          * find next empty vsi slot, looping back around if necessary
7308          */
7309         i = pf->next_vsi;
7310         while (i < pf->num_alloc_vsi && pf->vsi[i])
7311                 i++;
7312         if (i >= pf->num_alloc_vsi) {
7313                 i = 0;
7314                 while (i < pf->next_vsi && pf->vsi[i])
7315                         i++;
7316         }
7317
7318         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7319                 vsi_idx = i;             /* Found one! */
7320         } else {
7321                 ret = -ENODEV;
7322                 goto unlock_pf;  /* out of VSI slots! */
7323         }
7324         pf->next_vsi = ++i;
7325
7326         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7327         if (!vsi) {
7328                 ret = -ENOMEM;
7329                 goto unlock_pf;
7330         }
7331         vsi->type = type;
7332         vsi->back = pf;
7333         set_bit(__I40E_DOWN, &vsi->state);
7334         vsi->flags = 0;
7335         vsi->idx = vsi_idx;
7336         vsi->int_rate_limit = 0;
7337         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7338                                 pf->rss_table_size : 64;
7339         vsi->netdev_registered = false;
7340         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7341         INIT_LIST_HEAD(&vsi->mac_filter_list);
7342         vsi->irqs_ready = false;
7343
7344         ret = i40e_set_num_rings_in_vsi(vsi);
7345         if (ret)
7346                 goto err_rings;
7347
7348         ret = i40e_vsi_alloc_arrays(vsi, true);
7349         if (ret)
7350                 goto err_rings;
7351
7352         /* Setup default MSIX irq handler for VSI */
7353         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7354
7355         /* Initialize VSI lock */
7356         spin_lock_init(&vsi->mac_filter_list_lock);
7357         pf->vsi[vsi_idx] = vsi;
7358         ret = vsi_idx;
7359         goto unlock_pf;
7360
7361 err_rings:
7362         pf->next_vsi = i - 1;
7363         kfree(vsi);
7364 unlock_pf:
7365         mutex_unlock(&pf->switch_mutex);
7366         return ret;
7367 }
7368
7369 /**
7370  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7371  * @type: VSI pointer
7372  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7373  *
7374  * On error: returns error code (negative)
7375  * On success: returns 0
7376  **/
7377 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7378 {
7379         /* free the ring and vector containers */
7380         if (free_qvectors) {
7381                 kfree(vsi->q_vectors);
7382                 vsi->q_vectors = NULL;
7383         }
7384         kfree(vsi->tx_rings);
7385         vsi->tx_rings = NULL;
7386         vsi->rx_rings = NULL;
7387 }
7388
7389 /**
7390  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7391  * and lookup table
7392  * @vsi: Pointer to VSI structure
7393  */
7394 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7395 {
7396         if (!vsi)
7397                 return;
7398
7399         kfree(vsi->rss_hkey_user);
7400         vsi->rss_hkey_user = NULL;
7401
7402         kfree(vsi->rss_lut_user);
7403         vsi->rss_lut_user = NULL;
7404 }
7405
7406 /**
7407  * i40e_vsi_clear - Deallocate the VSI provided
7408  * @vsi: the VSI being un-configured
7409  **/
7410 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7411 {
7412         struct i40e_pf *pf;
7413
7414         if (!vsi)
7415                 return 0;
7416
7417         if (!vsi->back)
7418                 goto free_vsi;
7419         pf = vsi->back;
7420
7421         mutex_lock(&pf->switch_mutex);
7422         if (!pf->vsi[vsi->idx]) {
7423                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7424                         vsi->idx, vsi->idx, vsi, vsi->type);
7425                 goto unlock_vsi;
7426         }
7427
7428         if (pf->vsi[vsi->idx] != vsi) {
7429                 dev_err(&pf->pdev->dev,
7430                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7431                         pf->vsi[vsi->idx]->idx,
7432                         pf->vsi[vsi->idx],
7433                         pf->vsi[vsi->idx]->type,
7434                         vsi->idx, vsi, vsi->type);
7435                 goto unlock_vsi;
7436         }
7437
7438         /* updates the PF for this cleared vsi */
7439         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7440         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7441
7442         i40e_vsi_free_arrays(vsi, true);
7443         i40e_clear_rss_config_user(vsi);
7444
7445         pf->vsi[vsi->idx] = NULL;
7446         if (vsi->idx < pf->next_vsi)
7447                 pf->next_vsi = vsi->idx;
7448
7449 unlock_vsi:
7450         mutex_unlock(&pf->switch_mutex);
7451 free_vsi:
7452         kfree(vsi);
7453
7454         return 0;
7455 }
7456
7457 /**
7458  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7459  * @vsi: the VSI being cleaned
7460  **/
7461 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7462 {
7463         int i;
7464
7465         if (vsi->tx_rings && vsi->tx_rings[0]) {
7466                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7467                         kfree_rcu(vsi->tx_rings[i], rcu);
7468                         vsi->tx_rings[i] = NULL;
7469                         vsi->rx_rings[i] = NULL;
7470                 }
7471         }
7472 }
7473
7474 /**
7475  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7476  * @vsi: the VSI being configured
7477  **/
7478 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7479 {
7480         struct i40e_ring *tx_ring, *rx_ring;
7481         struct i40e_pf *pf = vsi->back;
7482         int i;
7483
7484         /* Set basic values in the rings to be used later during open() */
7485         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7486                 /* allocate space for both Tx and Rx in one shot */
7487                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7488                 if (!tx_ring)
7489                         goto err_out;
7490
7491                 tx_ring->queue_index = i;
7492                 tx_ring->reg_idx = vsi->base_queue + i;
7493                 tx_ring->ring_active = false;
7494                 tx_ring->vsi = vsi;
7495                 tx_ring->netdev = vsi->netdev;
7496                 tx_ring->dev = &pf->pdev->dev;
7497                 tx_ring->count = vsi->num_desc;
7498                 tx_ring->size = 0;
7499                 tx_ring->dcb_tc = 0;
7500                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7501                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7502                 tx_ring->tx_itr_setting = pf->tx_itr_default;
7503                 vsi->tx_rings[i] = tx_ring;
7504
7505                 rx_ring = &tx_ring[1];
7506                 rx_ring->queue_index = i;
7507                 rx_ring->reg_idx = vsi->base_queue + i;
7508                 rx_ring->ring_active = false;
7509                 rx_ring->vsi = vsi;
7510                 rx_ring->netdev = vsi->netdev;
7511                 rx_ring->dev = &pf->pdev->dev;
7512                 rx_ring->count = vsi->num_desc;
7513                 rx_ring->size = 0;
7514                 rx_ring->dcb_tc = 0;
7515                 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7516                         set_ring_16byte_desc_enabled(rx_ring);
7517                 else
7518                         clear_ring_16byte_desc_enabled(rx_ring);
7519                 rx_ring->rx_itr_setting = pf->rx_itr_default;
7520                 vsi->rx_rings[i] = rx_ring;
7521         }
7522
7523         return 0;
7524
7525 err_out:
7526         i40e_vsi_clear_rings(vsi);
7527         return -ENOMEM;
7528 }
7529
7530 /**
7531  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7532  * @pf: board private structure
7533  * @vectors: the number of MSI-X vectors to request
7534  *
7535  * Returns the number of vectors reserved, or error
7536  **/
7537 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7538 {
7539         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7540                                         I40E_MIN_MSIX, vectors);
7541         if (vectors < 0) {
7542                 dev_info(&pf->pdev->dev,
7543                          "MSI-X vector reservation failed: %d\n", vectors);
7544                 vectors = 0;
7545         }
7546
7547         return vectors;
7548 }
7549
7550 /**
7551  * i40e_init_msix - Setup the MSIX capability
7552  * @pf: board private structure
7553  *
7554  * Work with the OS to set up the MSIX vectors needed.
7555  *
7556  * Returns the number of vectors reserved or negative on failure
7557  **/
7558 static int i40e_init_msix(struct i40e_pf *pf)
7559 {
7560         struct i40e_hw *hw = &pf->hw;
7561         int vectors_left;
7562         int v_budget, i;
7563         int v_actual;
7564         int iwarp_requested = 0;
7565
7566         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7567                 return -ENODEV;
7568
7569         /* The number of vectors we'll request will be comprised of:
7570          *   - Add 1 for "other" cause for Admin Queue events, etc.
7571          *   - The number of LAN queue pairs
7572          *      - Queues being used for RSS.
7573          *              We don't need as many as max_rss_size vectors.
7574          *              use rss_size instead in the calculation since that
7575          *              is governed by number of cpus in the system.
7576          *      - assumes symmetric Tx/Rx pairing
7577          *   - The number of VMDq pairs
7578          *   - The CPU count within the NUMA node if iWARP is enabled
7579 #ifdef I40E_FCOE
7580          *   - The number of FCOE qps.
7581 #endif
7582          * Once we count this up, try the request.
7583          *
7584          * If we can't get what we want, we'll simplify to nearly nothing
7585          * and try again.  If that still fails, we punt.
7586          */
7587         vectors_left = hw->func_caps.num_msix_vectors;
7588         v_budget = 0;
7589
7590         /* reserve one vector for miscellaneous handler */
7591         if (vectors_left) {
7592                 v_budget++;
7593                 vectors_left--;
7594         }
7595
7596         /* reserve vectors for the main PF traffic queues */
7597         pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7598         vectors_left -= pf->num_lan_msix;
7599         v_budget += pf->num_lan_msix;
7600
7601         /* reserve one vector for sideband flow director */
7602         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7603                 if (vectors_left) {
7604                         v_budget++;
7605                         vectors_left--;
7606                 } else {
7607                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7608                 }
7609         }
7610
7611 #ifdef I40E_FCOE
7612         /* can we reserve enough for FCoE? */
7613         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7614                 if (!vectors_left)
7615                         pf->num_fcoe_msix = 0;
7616                 else if (vectors_left >= pf->num_fcoe_qps)
7617                         pf->num_fcoe_msix = pf->num_fcoe_qps;
7618                 else
7619                         pf->num_fcoe_msix = 1;
7620                 v_budget += pf->num_fcoe_msix;
7621                 vectors_left -= pf->num_fcoe_msix;
7622         }
7623
7624 #endif
7625         /* can we reserve enough for iWARP? */
7626         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7627                 if (!vectors_left)
7628                         pf->num_iwarp_msix = 0;
7629                 else if (vectors_left < pf->num_iwarp_msix)
7630                         pf->num_iwarp_msix = 1;
7631                 v_budget += pf->num_iwarp_msix;
7632                 vectors_left -= pf->num_iwarp_msix;
7633         }
7634
7635         /* any vectors left over go for VMDq support */
7636         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7637                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7638                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7639
7640                 /* if we're short on vectors for what's desired, we limit
7641                  * the queues per vmdq.  If this is still more than are
7642                  * available, the user will need to change the number of
7643                  * queues/vectors used by the PF later with the ethtool
7644                  * channels command
7645                  */
7646                 if (vmdq_vecs < vmdq_vecs_wanted)
7647                         pf->num_vmdq_qps = 1;
7648                 pf->num_vmdq_msix = pf->num_vmdq_qps;
7649
7650                 v_budget += vmdq_vecs;
7651                 vectors_left -= vmdq_vecs;
7652         }
7653
7654         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7655                                    GFP_KERNEL);
7656         if (!pf->msix_entries)
7657                 return -ENOMEM;
7658
7659         for (i = 0; i < v_budget; i++)
7660                 pf->msix_entries[i].entry = i;
7661         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7662
7663         if (v_actual != v_budget) {
7664                 /* If we have limited resources, we will start with no vectors
7665                  * for the special features and then allocate vectors to some
7666                  * of these features based on the policy and at the end disable
7667                  * the features that did not get any vectors.
7668                  */
7669                 iwarp_requested = pf->num_iwarp_msix;
7670                 pf->num_iwarp_msix = 0;
7671 #ifdef I40E_FCOE
7672                 pf->num_fcoe_qps = 0;
7673                 pf->num_fcoe_msix = 0;
7674 #endif
7675                 pf->num_vmdq_msix = 0;
7676         }
7677
7678         if (v_actual < I40E_MIN_MSIX) {
7679                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7680                 kfree(pf->msix_entries);
7681                 pf->msix_entries = NULL;
7682                 return -ENODEV;
7683
7684         } else if (v_actual == I40E_MIN_MSIX) {
7685                 /* Adjust for minimal MSIX use */
7686                 pf->num_vmdq_vsis = 0;
7687                 pf->num_vmdq_qps = 0;
7688                 pf->num_lan_qps = 1;
7689                 pf->num_lan_msix = 1;
7690
7691         } else if (v_actual != v_budget) {
7692                 int vec;
7693
7694                 /* reserve the misc vector */
7695                 vec = v_actual - 1;
7696
7697                 /* Scale vector usage down */
7698                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7699                 pf->num_vmdq_vsis = 1;
7700                 pf->num_vmdq_qps = 1;
7701                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7702
7703                 /* partition out the remaining vectors */
7704                 switch (vec) {
7705                 case 2:
7706                         pf->num_lan_msix = 1;
7707                         break;
7708                 case 3:
7709                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7710                                 pf->num_lan_msix = 1;
7711                                 pf->num_iwarp_msix = 1;
7712                         } else {
7713                                 pf->num_lan_msix = 2;
7714                         }
7715 #ifdef I40E_FCOE
7716                         /* give one vector to FCoE */
7717                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7718                                 pf->num_lan_msix = 1;
7719                                 pf->num_fcoe_msix = 1;
7720                         }
7721 #endif
7722                         break;
7723                 default:
7724                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7725                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7726                                                  iwarp_requested);
7727                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7728                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7729                         } else {
7730                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7731                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7732                         }
7733                         pf->num_lan_msix = min_t(int,
7734                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7735                                                               pf->num_lan_msix);
7736 #ifdef I40E_FCOE
7737                         /* give one vector to FCoE */
7738                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7739                                 pf->num_fcoe_msix = 1;
7740                                 vec--;
7741                         }
7742 #endif
7743                         break;
7744                 }
7745         }
7746
7747         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7748             (pf->num_vmdq_msix == 0)) {
7749                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7750                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7751         }
7752
7753         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7754             (pf->num_iwarp_msix == 0)) {
7755                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7756                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7757         }
7758 #ifdef I40E_FCOE
7759
7760         if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7761                 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7762                 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7763         }
7764 #endif
7765         return v_actual;
7766 }
7767
7768 /**
7769  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7770  * @vsi: the VSI being configured
7771  * @v_idx: index of the vector in the vsi struct
7772  *
7773  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7774  **/
7775 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7776 {
7777         struct i40e_q_vector *q_vector;
7778
7779         /* allocate q_vector */
7780         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7781         if (!q_vector)
7782                 return -ENOMEM;
7783
7784         q_vector->vsi = vsi;
7785         q_vector->v_idx = v_idx;
7786         cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7787         if (vsi->netdev)
7788                 netif_napi_add(vsi->netdev, &q_vector->napi,
7789                                i40e_napi_poll, NAPI_POLL_WEIGHT);
7790
7791         q_vector->rx.latency_range = I40E_LOW_LATENCY;
7792         q_vector->tx.latency_range = I40E_LOW_LATENCY;
7793
7794         /* tie q_vector and vsi together */
7795         vsi->q_vectors[v_idx] = q_vector;
7796
7797         return 0;
7798 }
7799
7800 /**
7801  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7802  * @vsi: the VSI being configured
7803  *
7804  * We allocate one q_vector per queue interrupt.  If allocation fails we
7805  * return -ENOMEM.
7806  **/
7807 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7808 {
7809         struct i40e_pf *pf = vsi->back;
7810         int v_idx, num_q_vectors;
7811         int err;
7812
7813         /* if not MSIX, give the one vector only to the LAN VSI */
7814         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7815                 num_q_vectors = vsi->num_q_vectors;
7816         else if (vsi == pf->vsi[pf->lan_vsi])
7817                 num_q_vectors = 1;
7818         else
7819                 return -EINVAL;
7820
7821         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7822                 err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7823                 if (err)
7824                         goto err_out;
7825         }
7826
7827         return 0;
7828
7829 err_out:
7830         while (v_idx--)
7831                 i40e_free_q_vector(vsi, v_idx);
7832
7833         return err;
7834 }
7835
7836 /**
7837  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7838  * @pf: board private structure to initialize
7839  **/
7840 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7841 {
7842         int vectors = 0;
7843         ssize_t size;
7844
7845         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7846                 vectors = i40e_init_msix(pf);
7847                 if (vectors < 0) {
7848                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
7849                                        I40E_FLAG_IWARP_ENABLED  |
7850 #ifdef I40E_FCOE
7851                                        I40E_FLAG_FCOE_ENABLED   |
7852 #endif
7853                                        I40E_FLAG_RSS_ENABLED    |
7854                                        I40E_FLAG_DCB_CAPABLE    |
7855                                        I40E_FLAG_SRIOV_ENABLED  |
7856                                        I40E_FLAG_FD_SB_ENABLED  |
7857                                        I40E_FLAG_FD_ATR_ENABLED |
7858                                        I40E_FLAG_VMDQ_ENABLED);
7859
7860                         /* rework the queue expectations without MSIX */
7861                         i40e_determine_queue_usage(pf);
7862                 }
7863         }
7864
7865         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7866             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7867                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7868                 vectors = pci_enable_msi(pf->pdev);
7869                 if (vectors < 0) {
7870                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7871                                  vectors);
7872                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7873                 }
7874                 vectors = 1;  /* one MSI or Legacy vector */
7875         }
7876
7877         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7878                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7879
7880         /* set up vector assignment tracking */
7881         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7882         pf->irq_pile = kzalloc(size, GFP_KERNEL);
7883         if (!pf->irq_pile) {
7884                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7885                 return -ENOMEM;
7886         }
7887         pf->irq_pile->num_entries = vectors;
7888         pf->irq_pile->search_hint = 0;
7889
7890         /* track first vector for misc interrupts, ignore return */
7891         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7892
7893         return 0;
7894 }
7895
7896 /**
7897  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7898  * @pf: board private structure
7899  *
7900  * This sets up the handler for MSIX 0, which is used to manage the
7901  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7902  * when in MSI or Legacy interrupt mode.
7903  **/
7904 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7905 {
7906         struct i40e_hw *hw = &pf->hw;
7907         int err = 0;
7908
7909         /* Only request the irq if this is the first time through, and
7910          * not when we're rebuilding after a Reset
7911          */
7912         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7913                 err = request_irq(pf->msix_entries[0].vector,
7914                                   i40e_intr, 0, pf->int_name, pf);
7915                 if (err) {
7916                         dev_info(&pf->pdev->dev,
7917                                  "request_irq for %s failed: %d\n",
7918                                  pf->int_name, err);
7919                         return -EFAULT;
7920                 }
7921         }
7922
7923         i40e_enable_misc_int_causes(pf);
7924
7925         /* associate no queues to the misc vector */
7926         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7927         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7928
7929         i40e_flush(hw);
7930
7931         i40e_irq_dynamic_enable_icr0(pf, true);
7932
7933         return err;
7934 }
7935
7936 /**
7937  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7938  * @vsi: vsi structure
7939  * @seed: RSS hash seed
7940  **/
7941 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7942                               u8 *lut, u16 lut_size)
7943 {
7944         struct i40e_aqc_get_set_rss_key_data rss_key;
7945         struct i40e_pf *pf = vsi->back;
7946         struct i40e_hw *hw = &pf->hw;
7947         bool pf_lut = false;
7948         u8 *rss_lut;
7949         int ret, i;
7950
7951         memset(&rss_key, 0, sizeof(rss_key));
7952         memcpy(&rss_key, seed, sizeof(rss_key));
7953
7954         rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7955         if (!rss_lut)
7956                 return -ENOMEM;
7957
7958         /* Populate the LUT with max no. of queues in round robin fashion */
7959         for (i = 0; i < vsi->rss_table_size; i++)
7960                 rss_lut[i] = i % vsi->rss_size;
7961
7962         ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7963         if (ret) {
7964                 dev_info(&pf->pdev->dev,
7965                          "Cannot set RSS key, err %s aq_err %s\n",
7966                          i40e_stat_str(&pf->hw, ret),
7967                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7968                 goto config_rss_aq_out;
7969         }
7970
7971         if (vsi->type == I40E_VSI_MAIN)
7972                 pf_lut = true;
7973
7974         ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7975                                   vsi->rss_table_size);
7976         if (ret)
7977                 dev_info(&pf->pdev->dev,
7978                          "Cannot set RSS lut, err %s aq_err %s\n",
7979                          i40e_stat_str(&pf->hw, ret),
7980                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7981
7982 config_rss_aq_out:
7983         kfree(rss_lut);
7984         return ret;
7985 }
7986
7987 /**
7988  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7989  * @vsi: VSI structure
7990  **/
7991 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7992 {
7993         u8 seed[I40E_HKEY_ARRAY_SIZE];
7994         struct i40e_pf *pf = vsi->back;
7995         u8 *lut;
7996         int ret;
7997
7998         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
7999                 return 0;
8000
8001         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8002         if (!lut)
8003                 return -ENOMEM;
8004
8005         i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8006         netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8007         vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
8008         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8009         kfree(lut);
8010
8011         return ret;
8012 }
8013
8014 /**
8015  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8016  * @vsi: Pointer to vsi structure
8017  * @seed: Buffter to store the hash keys
8018  * @lut: Buffer to store the lookup table entries
8019  * @lut_size: Size of buffer to store the lookup table entries
8020  *
8021  * Return 0 on success, negative on failure
8022  */
8023 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8024                            u8 *lut, u16 lut_size)
8025 {
8026         struct i40e_pf *pf = vsi->back;
8027         struct i40e_hw *hw = &pf->hw;
8028         int ret = 0;
8029
8030         if (seed) {
8031                 ret = i40e_aq_get_rss_key(hw, vsi->id,
8032                         (struct i40e_aqc_get_set_rss_key_data *)seed);
8033                 if (ret) {
8034                         dev_info(&pf->pdev->dev,
8035                                  "Cannot get RSS key, err %s aq_err %s\n",
8036                                  i40e_stat_str(&pf->hw, ret),
8037                                  i40e_aq_str(&pf->hw,
8038                                              pf->hw.aq.asq_last_status));
8039                         return ret;
8040                 }
8041         }
8042
8043         if (lut) {
8044                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8045
8046                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8047                 if (ret) {
8048                         dev_info(&pf->pdev->dev,
8049                                  "Cannot get RSS lut, err %s aq_err %s\n",
8050                                  i40e_stat_str(&pf->hw, ret),
8051                                  i40e_aq_str(&pf->hw,
8052                                              pf->hw.aq.asq_last_status));
8053                         return ret;
8054                 }
8055         }
8056
8057         return ret;
8058 }
8059
8060 /**
8061  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8062  * @vsi: Pointer to vsi structure
8063  * @seed: RSS hash seed
8064  * @lut: Lookup table
8065  * @lut_size: Lookup table size
8066  *
8067  * Returns 0 on success, negative on failure
8068  **/
8069 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8070                                const u8 *lut, u16 lut_size)
8071 {
8072         struct i40e_pf *pf = vsi->back;
8073         struct i40e_hw *hw = &pf->hw;
8074         u16 vf_id = vsi->vf_id;
8075         u8 i;
8076
8077         /* Fill out hash function seed */
8078         if (seed) {
8079                 u32 *seed_dw = (u32 *)seed;
8080
8081                 if (vsi->type == I40E_VSI_MAIN) {
8082                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8083                                 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8084                                                   seed_dw[i]);
8085                 } else if (vsi->type == I40E_VSI_SRIOV) {
8086                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8087                                 i40e_write_rx_ctl(hw,
8088                                                   I40E_VFQF_HKEY1(i, vf_id),
8089                                                   seed_dw[i]);
8090                 } else {
8091                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8092                 }
8093         }
8094
8095         if (lut) {
8096                 u32 *lut_dw = (u32 *)lut;
8097
8098                 if (vsi->type == I40E_VSI_MAIN) {
8099                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
8100                                 return -EINVAL;
8101                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8102                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8103                 } else if (vsi->type == I40E_VSI_SRIOV) {
8104                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8105                                 return -EINVAL;
8106                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8107                                 i40e_write_rx_ctl(hw,
8108                                                   I40E_VFQF_HLUT1(i, vf_id),
8109                                                   lut_dw[i]);
8110                 } else {
8111                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8112                 }
8113         }
8114         i40e_flush(hw);
8115
8116         return 0;
8117 }
8118
8119 /**
8120  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8121  * @vsi: Pointer to VSI structure
8122  * @seed: Buffer to store the keys
8123  * @lut: Buffer to store the lookup table entries
8124  * @lut_size: Size of buffer to store the lookup table entries
8125  *
8126  * Returns 0 on success, negative on failure
8127  */
8128 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8129                             u8 *lut, u16 lut_size)
8130 {
8131         struct i40e_pf *pf = vsi->back;
8132         struct i40e_hw *hw = &pf->hw;
8133         u16 i;
8134
8135         if (seed) {
8136                 u32 *seed_dw = (u32 *)seed;
8137
8138                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8139                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8140         }
8141         if (lut) {
8142                 u32 *lut_dw = (u32 *)lut;
8143
8144                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8145                         return -EINVAL;
8146                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8147                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8148         }
8149
8150         return 0;
8151 }
8152
8153 /**
8154  * i40e_config_rss - Configure RSS keys and lut
8155  * @vsi: Pointer to VSI structure
8156  * @seed: RSS hash seed
8157  * @lut: Lookup table
8158  * @lut_size: Lookup table size
8159  *
8160  * Returns 0 on success, negative on failure
8161  */
8162 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8163 {
8164         struct i40e_pf *pf = vsi->back;
8165
8166         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8167                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8168         else
8169                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8170 }
8171
8172 /**
8173  * i40e_get_rss - Get RSS keys and lut
8174  * @vsi: Pointer to VSI structure
8175  * @seed: Buffer to store the keys
8176  * @lut: Buffer to store the lookup table entries
8177  * lut_size: Size of buffer to store the lookup table entries
8178  *
8179  * Returns 0 on success, negative on failure
8180  */
8181 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8182 {
8183         struct i40e_pf *pf = vsi->back;
8184
8185         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8186                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8187         else
8188                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8189 }
8190
8191 /**
8192  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8193  * @pf: Pointer to board private structure
8194  * @lut: Lookup table
8195  * @rss_table_size: Lookup table size
8196  * @rss_size: Range of queue number for hashing
8197  */
8198 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8199                               u16 rss_table_size, u16 rss_size)
8200 {
8201         u16 i;
8202
8203         for (i = 0; i < rss_table_size; i++)
8204                 lut[i] = i % rss_size;
8205 }
8206
8207 /**
8208  * i40e_pf_config_rss - Prepare for RSS if used
8209  * @pf: board private structure
8210  **/
8211 static int i40e_pf_config_rss(struct i40e_pf *pf)
8212 {
8213         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8214         u8 seed[I40E_HKEY_ARRAY_SIZE];
8215         u8 *lut;
8216         struct i40e_hw *hw = &pf->hw;
8217         u32 reg_val;
8218         u64 hena;
8219         int ret;
8220
8221         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8222         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8223                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8224         hena |= i40e_pf_get_default_rss_hena(pf);
8225
8226         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8227         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8228
8229         /* Determine the RSS table size based on the hardware capabilities */
8230         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8231         reg_val = (pf->rss_table_size == 512) ?
8232                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8233                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8234         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8235
8236         /* Determine the RSS size of the VSI */
8237         if (!vsi->rss_size)
8238                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8239                                       vsi->num_queue_pairs);
8240
8241         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8242         if (!lut)
8243                 return -ENOMEM;
8244
8245         /* Use user configured lut if there is one, otherwise use default */
8246         if (vsi->rss_lut_user)
8247                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8248         else
8249                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8250
8251         /* Use user configured hash key if there is one, otherwise
8252          * use default.
8253          */
8254         if (vsi->rss_hkey_user)
8255                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8256         else
8257                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8258         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8259         kfree(lut);
8260
8261         return ret;
8262 }
8263
8264 /**
8265  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8266  * @pf: board private structure
8267  * @queue_count: the requested queue count for rss.
8268  *
8269  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8270  * count which may be different from the requested queue count.
8271  **/
8272 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8273 {
8274         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8275         int new_rss_size;
8276
8277         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8278                 return 0;
8279
8280         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8281
8282         if (queue_count != vsi->num_queue_pairs) {
8283                 vsi->req_queue_pairs = queue_count;
8284                 i40e_prep_for_reset(pf);
8285
8286                 pf->alloc_rss_size = new_rss_size;
8287
8288                 i40e_reset_and_rebuild(pf, true);
8289
8290                 /* Discard the user configured hash keys and lut, if less
8291                  * queues are enabled.
8292                  */
8293                 if (queue_count < vsi->rss_size) {
8294                         i40e_clear_rss_config_user(vsi);
8295                         dev_dbg(&pf->pdev->dev,
8296                                 "discard user configured hash keys and lut\n");
8297                 }
8298
8299                 /* Reset vsi->rss_size, as number of enabled queues changed */
8300                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8301                                       vsi->num_queue_pairs);
8302
8303                 i40e_pf_config_rss(pf);
8304         }
8305         dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8306                  pf->alloc_rss_size, pf->rss_size_max);
8307         return pf->alloc_rss_size;
8308 }
8309
8310 /**
8311  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8312  * @pf: board private structure
8313  **/
8314 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8315 {
8316         i40e_status status;
8317         bool min_valid, max_valid;
8318         u32 max_bw, min_bw;
8319
8320         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8321                                            &min_valid, &max_valid);
8322
8323         if (!status) {
8324                 if (min_valid)
8325                         pf->npar_min_bw = min_bw;
8326                 if (max_valid)
8327                         pf->npar_max_bw = max_bw;
8328         }
8329
8330         return status;
8331 }
8332
8333 /**
8334  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8335  * @pf: board private structure
8336  **/
8337 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8338 {
8339         struct i40e_aqc_configure_partition_bw_data bw_data;
8340         i40e_status status;
8341
8342         /* Set the valid bit for this PF */
8343         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8344         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8345         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8346
8347         /* Set the new bandwidths */
8348         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8349
8350         return status;
8351 }
8352
8353 /**
8354  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8355  * @pf: board private structure
8356  **/
8357 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8358 {
8359         /* Commit temporary BW setting to permanent NVM image */
8360         enum i40e_admin_queue_err last_aq_status;
8361         i40e_status ret;
8362         u16 nvm_word;
8363
8364         if (pf->hw.partition_id != 1) {
8365                 dev_info(&pf->pdev->dev,
8366                          "Commit BW only works on partition 1! This is partition %d",
8367                          pf->hw.partition_id);
8368                 ret = I40E_NOT_SUPPORTED;
8369                 goto bw_commit_out;
8370         }
8371
8372         /* Acquire NVM for read access */
8373         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8374         last_aq_status = pf->hw.aq.asq_last_status;
8375         if (ret) {
8376                 dev_info(&pf->pdev->dev,
8377                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8378                          i40e_stat_str(&pf->hw, ret),
8379                          i40e_aq_str(&pf->hw, last_aq_status));
8380                 goto bw_commit_out;
8381         }
8382
8383         /* Read word 0x10 of NVM - SW compatibility word 1 */
8384         ret = i40e_aq_read_nvm(&pf->hw,
8385                                I40E_SR_NVM_CONTROL_WORD,
8386                                0x10, sizeof(nvm_word), &nvm_word,
8387                                false, NULL);
8388         /* Save off last admin queue command status before releasing
8389          * the NVM
8390          */
8391         last_aq_status = pf->hw.aq.asq_last_status;
8392         i40e_release_nvm(&pf->hw);
8393         if (ret) {
8394                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8395                          i40e_stat_str(&pf->hw, ret),
8396                          i40e_aq_str(&pf->hw, last_aq_status));
8397                 goto bw_commit_out;
8398         }
8399
8400         /* Wait a bit for NVM release to complete */
8401         msleep(50);
8402
8403         /* Acquire NVM for write access */
8404         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8405         last_aq_status = pf->hw.aq.asq_last_status;
8406         if (ret) {
8407                 dev_info(&pf->pdev->dev,
8408                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8409                          i40e_stat_str(&pf->hw, ret),
8410                          i40e_aq_str(&pf->hw, last_aq_status));
8411                 goto bw_commit_out;
8412         }
8413         /* Write it back out unchanged to initiate update NVM,
8414          * which will force a write of the shadow (alt) RAM to
8415          * the NVM - thus storing the bandwidth values permanently.
8416          */
8417         ret = i40e_aq_update_nvm(&pf->hw,
8418                                  I40E_SR_NVM_CONTROL_WORD,
8419                                  0x10, sizeof(nvm_word),
8420                                  &nvm_word, true, NULL);
8421         /* Save off last admin queue command status before releasing
8422          * the NVM
8423          */
8424         last_aq_status = pf->hw.aq.asq_last_status;
8425         i40e_release_nvm(&pf->hw);
8426         if (ret)
8427                 dev_info(&pf->pdev->dev,
8428                          "BW settings NOT SAVED, err %s aq_err %s\n",
8429                          i40e_stat_str(&pf->hw, ret),
8430                          i40e_aq_str(&pf->hw, last_aq_status));
8431 bw_commit_out:
8432
8433         return ret;
8434 }
8435
8436 /**
8437  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8438  * @pf: board private structure to initialize
8439  *
8440  * i40e_sw_init initializes the Adapter private data structure.
8441  * Fields are initialized based on PCI device information and
8442  * OS network device settings (MTU size).
8443  **/
8444 static int i40e_sw_init(struct i40e_pf *pf)
8445 {
8446         int err = 0;
8447         int size;
8448
8449         pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8450                                 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8451         if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8452                 if (I40E_DEBUG_USER & debug)
8453                         pf->hw.debug_mask = debug;
8454                 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8455                                                 I40E_DEFAULT_MSG_ENABLE);
8456         }
8457
8458         /* Set default capability flags */
8459         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8460                     I40E_FLAG_MSI_ENABLED     |
8461                     I40E_FLAG_MSIX_ENABLED;
8462
8463         if (iommu_present(&pci_bus_type))
8464                 pf->flags |= I40E_FLAG_RX_PS_ENABLED;
8465         else
8466                 pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
8467
8468         /* Set default ITR */
8469         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8470         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8471
8472         /* Depending on PF configurations, it is possible that the RSS
8473          * maximum might end up larger than the available queues
8474          */
8475         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8476         pf->alloc_rss_size = 1;
8477         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8478         pf->rss_size_max = min_t(int, pf->rss_size_max,
8479                                  pf->hw.func_caps.num_tx_qp);
8480         if (pf->hw.func_caps.rss) {
8481                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8482                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8483                                            num_online_cpus());
8484         }
8485
8486         /* MFP mode enabled */
8487         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8488                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8489                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8490                 if (i40e_get_npar_bw_setting(pf))
8491                         dev_warn(&pf->pdev->dev,
8492                                  "Could not get NPAR bw settings\n");
8493                 else
8494                         dev_info(&pf->pdev->dev,
8495                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8496                                  pf->npar_min_bw, pf->npar_max_bw);
8497         }
8498
8499         /* FW/NVM is not yet fixed in this regard */
8500         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8501             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8502                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8503                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8504                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8505                     pf->hw.num_partitions > 1)
8506                         dev_info(&pf->pdev->dev,
8507                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8508                 else
8509                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8510                 pf->fdir_pf_filter_count =
8511                                  pf->hw.func_caps.fd_filters_guaranteed;
8512                 pf->hw.fdir_shared_filter_count =
8513                                  pf->hw.func_caps.fd_filters_best_effort;
8514         }
8515
8516         if (i40e_is_mac_710(&pf->hw) &&
8517             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8518             (pf->hw.aq.fw_maj_ver < 4))) {
8519                 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8520                 /* No DCB support  for FW < v4.33 */
8521                 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8522         }
8523
8524         /* Disable FW LLDP if FW < v4.3 */
8525         if (i40e_is_mac_710(&pf->hw) &&
8526             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8527             (pf->hw.aq.fw_maj_ver < 4)))
8528                 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8529
8530         /* Use the FW Set LLDP MIB API if FW > v4.40 */
8531         if (i40e_is_mac_710(&pf->hw) &&
8532             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8533             (pf->hw.aq.fw_maj_ver >= 5)))
8534                 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8535
8536         if (pf->hw.func_caps.vmdq) {
8537                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8538                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8539                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8540         }
8541
8542         if (pf->hw.func_caps.iwarp) {
8543                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8544                 /* IWARP needs one extra vector for CQP just like MISC.*/
8545                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8546         }
8547
8548 #ifdef I40E_FCOE
8549         i40e_init_pf_fcoe(pf);
8550
8551 #endif /* I40E_FCOE */
8552 #ifdef CONFIG_PCI_IOV
8553         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8554                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8555                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8556                 pf->num_req_vfs = min_t(int,
8557                                         pf->hw.func_caps.num_vfs,
8558                                         I40E_MAX_VF_COUNT);
8559         }
8560 #endif /* CONFIG_PCI_IOV */
8561         if (pf->hw.mac.type == I40E_MAC_X722) {
8562                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8563                              I40E_FLAG_128_QP_RSS_CAPABLE |
8564                              I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8565                              I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8566                              I40E_FLAG_WB_ON_ITR_CAPABLE |
8567                              I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8568                              I40E_FLAG_NO_PCI_LINK_CHECK |
8569                              I40E_FLAG_100M_SGMII_CAPABLE |
8570                              I40E_FLAG_USE_SET_LLDP_MIB |
8571                              I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8572         } else if ((pf->hw.aq.api_maj_ver > 1) ||
8573                    ((pf->hw.aq.api_maj_ver == 1) &&
8574                     (pf->hw.aq.api_min_ver > 4))) {
8575                 /* Supported in FW API version higher than 1.4 */
8576                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8577                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8578         } else {
8579                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8580         }
8581
8582         pf->eeprom_version = 0xDEAD;
8583         pf->lan_veb = I40E_NO_VEB;
8584         pf->lan_vsi = I40E_NO_VSI;
8585
8586         /* By default FW has this off for performance reasons */
8587         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8588
8589         /* set up queue assignment tracking */
8590         size = sizeof(struct i40e_lump_tracking)
8591                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8592         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8593         if (!pf->qp_pile) {
8594                 err = -ENOMEM;
8595                 goto sw_init_done;
8596         }
8597         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8598         pf->qp_pile->search_hint = 0;
8599
8600         pf->tx_timeout_recovery_level = 1;
8601
8602         mutex_init(&pf->switch_mutex);
8603
8604         /* If NPAR is enabled nudge the Tx scheduler */
8605         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8606                 i40e_set_npar_bw_setting(pf);
8607
8608 sw_init_done:
8609         return err;
8610 }
8611
8612 /**
8613  * i40e_set_ntuple - set the ntuple feature flag and take action
8614  * @pf: board private structure to initialize
8615  * @features: the feature set that the stack is suggesting
8616  *
8617  * returns a bool to indicate if reset needs to happen
8618  **/
8619 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8620 {
8621         bool need_reset = false;
8622
8623         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8624          * the state changed, we need to reset.
8625          */
8626         if (features & NETIF_F_NTUPLE) {
8627                 /* Enable filters and mark for reset */
8628                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8629                         need_reset = true;
8630                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8631         } else {
8632                 /* turn off filters, mark for reset and clear SW filter list */
8633                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8634                         need_reset = true;
8635                         i40e_fdir_filter_exit(pf);
8636                 }
8637                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8638                 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8639                 /* reset fd counters */
8640                 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8641                 pf->fdir_pf_active_filters = 0;
8642                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8643                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8644                         dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8645                 /* if ATR was auto disabled it can be re-enabled. */
8646                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8647                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8648                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8649         }
8650         return need_reset;
8651 }
8652
8653 /**
8654  * i40e_set_features - set the netdev feature flags
8655  * @netdev: ptr to the netdev being adjusted
8656  * @features: the feature set that the stack is suggesting
8657  **/
8658 static int i40e_set_features(struct net_device *netdev,
8659                              netdev_features_t features)
8660 {
8661         struct i40e_netdev_priv *np = netdev_priv(netdev);
8662         struct i40e_vsi *vsi = np->vsi;
8663         struct i40e_pf *pf = vsi->back;
8664         bool need_reset;
8665
8666         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8667                 i40e_vlan_stripping_enable(vsi);
8668         else
8669                 i40e_vlan_stripping_disable(vsi);
8670
8671         need_reset = i40e_set_ntuple(pf, features);
8672
8673         if (need_reset)
8674                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8675
8676         return 0;
8677 }
8678
8679 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
8680 /**
8681  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8682  * @pf: board private structure
8683  * @port: The UDP port to look up
8684  *
8685  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8686  **/
8687 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8688 {
8689         u8 i;
8690
8691         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8692                 if (pf->udp_ports[i].index == port)
8693                         return i;
8694         }
8695
8696         return i;
8697 }
8698
8699 #endif
8700
8701 #if IS_ENABLED(CONFIG_VXLAN)
8702 /**
8703  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8704  * @netdev: This physical port's netdev
8705  * @sa_family: Socket Family that VXLAN is notifying us about
8706  * @port: New UDP port number that VXLAN started listening to
8707  **/
8708 static void i40e_add_vxlan_port(struct net_device *netdev,
8709                                 sa_family_t sa_family, __be16 port)
8710 {
8711         struct i40e_netdev_priv *np = netdev_priv(netdev);
8712         struct i40e_vsi *vsi = np->vsi;
8713         struct i40e_pf *pf = vsi->back;
8714         u8 next_idx;
8715         u8 idx;
8716
8717         idx = i40e_get_udp_port_idx(pf, port);
8718
8719         /* Check if port already exists */
8720         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8721                 netdev_info(netdev, "vxlan port %d already offloaded\n",
8722                             ntohs(port));
8723                 return;
8724         }
8725
8726         /* Now check if there is space to add the new port */
8727         next_idx = i40e_get_udp_port_idx(pf, 0);
8728
8729         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8730                 netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8731                             ntohs(port));
8732                 return;
8733         }
8734
8735         /* New port: add it and mark its index in the bitmap */
8736         pf->udp_ports[next_idx].index = port;
8737         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8738         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8739         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8740 }
8741
8742 /**
8743  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8744  * @netdev: This physical port's netdev
8745  * @sa_family: Socket Family that VXLAN is notifying us about
8746  * @port: UDP port number that VXLAN stopped listening to
8747  **/
8748 static void i40e_del_vxlan_port(struct net_device *netdev,
8749                                 sa_family_t sa_family, __be16 port)
8750 {
8751         struct i40e_netdev_priv *np = netdev_priv(netdev);
8752         struct i40e_vsi *vsi = np->vsi;
8753         struct i40e_pf *pf = vsi->back;
8754         u8 idx;
8755
8756         idx = i40e_get_udp_port_idx(pf, port);
8757
8758         /* Check if port already exists */
8759         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8760                 /* if port exists, set it to 0 (mark for deletion)
8761                  * and make it pending
8762                  */
8763                 pf->udp_ports[idx].index = 0;
8764                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8765                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8766         } else {
8767                 netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8768                             ntohs(port));
8769         }
8770 }
8771 #endif
8772
8773 #if IS_ENABLED(CONFIG_GENEVE)
8774 /**
8775  * i40e_add_geneve_port - Get notifications about GENEVE ports that come up
8776  * @netdev: This physical port's netdev
8777  * @sa_family: Socket Family that GENEVE is notifying us about
8778  * @port: New UDP port number that GENEVE started listening to
8779  **/
8780 static void i40e_add_geneve_port(struct net_device *netdev,
8781                                  sa_family_t sa_family, __be16 port)
8782 {
8783         struct i40e_netdev_priv *np = netdev_priv(netdev);
8784         struct i40e_vsi *vsi = np->vsi;
8785         struct i40e_pf *pf = vsi->back;
8786         u8 next_idx;
8787         u8 idx;
8788
8789         if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8790                 return;
8791
8792         idx = i40e_get_udp_port_idx(pf, port);
8793
8794         /* Check if port already exists */
8795         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8796                 netdev_info(netdev, "udp port %d already offloaded\n",
8797                             ntohs(port));
8798                 return;
8799         }
8800
8801         /* Now check if there is space to add the new port */
8802         next_idx = i40e_get_udp_port_idx(pf, 0);
8803
8804         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8805                 netdev_info(netdev, "maximum number of UDP ports reached, not adding port %d\n",
8806                             ntohs(port));
8807                 return;
8808         }
8809
8810         /* New port: add it and mark its index in the bitmap */
8811         pf->udp_ports[next_idx].index = port;
8812         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8813         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8814         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8815
8816         dev_info(&pf->pdev->dev, "adding geneve port %d\n", ntohs(port));
8817 }
8818
8819 /**
8820  * i40e_del_geneve_port - Get notifications about GENEVE ports that go away
8821  * @netdev: This physical port's netdev
8822  * @sa_family: Socket Family that GENEVE is notifying us about
8823  * @port: UDP port number that GENEVE stopped listening to
8824  **/
8825 static void i40e_del_geneve_port(struct net_device *netdev,
8826                                  sa_family_t sa_family, __be16 port)
8827 {
8828         struct i40e_netdev_priv *np = netdev_priv(netdev);
8829         struct i40e_vsi *vsi = np->vsi;
8830         struct i40e_pf *pf = vsi->back;
8831         u8 idx;
8832
8833         if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8834                 return;
8835
8836         idx = i40e_get_udp_port_idx(pf, port);
8837
8838         /* Check if port already exists */
8839         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8840                 /* if port exists, set it to 0 (mark for deletion)
8841                  * and make it pending
8842                  */
8843                 pf->udp_ports[idx].index = 0;
8844                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8845                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8846
8847                 dev_info(&pf->pdev->dev, "deleting geneve port %d\n",
8848                          ntohs(port));
8849         } else {
8850                 netdev_warn(netdev, "geneve port %d was not found, not deleting\n",
8851                             ntohs(port));
8852         }
8853 }
8854 #endif
8855
8856 static int i40e_get_phys_port_id(struct net_device *netdev,
8857                                  struct netdev_phys_item_id *ppid)
8858 {
8859         struct i40e_netdev_priv *np = netdev_priv(netdev);
8860         struct i40e_pf *pf = np->vsi->back;
8861         struct i40e_hw *hw = &pf->hw;
8862
8863         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8864                 return -EOPNOTSUPP;
8865
8866         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8867         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8868
8869         return 0;
8870 }
8871
8872 /**
8873  * i40e_ndo_fdb_add - add an entry to the hardware database
8874  * @ndm: the input from the stack
8875  * @tb: pointer to array of nladdr (unused)
8876  * @dev: the net device pointer
8877  * @addr: the MAC address entry being added
8878  * @flags: instructions from stack about fdb operation
8879  */
8880 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8881                             struct net_device *dev,
8882                             const unsigned char *addr, u16 vid,
8883                             u16 flags)
8884 {
8885         struct i40e_netdev_priv *np = netdev_priv(dev);
8886         struct i40e_pf *pf = np->vsi->back;
8887         int err = 0;
8888
8889         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8890                 return -EOPNOTSUPP;
8891
8892         if (vid) {
8893                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8894                 return -EINVAL;
8895         }
8896
8897         /* Hardware does not support aging addresses so if a
8898          * ndm_state is given only allow permanent addresses
8899          */
8900         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8901                 netdev_info(dev, "FDB only supports static addresses\n");
8902                 return -EINVAL;
8903         }
8904
8905         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8906                 err = dev_uc_add_excl(dev, addr);
8907         else if (is_multicast_ether_addr(addr))
8908                 err = dev_mc_add_excl(dev, addr);
8909         else
8910                 err = -EINVAL;
8911
8912         /* Only return duplicate errors if NLM_F_EXCL is set */
8913         if (err == -EEXIST && !(flags & NLM_F_EXCL))
8914                 err = 0;
8915
8916         return err;
8917 }
8918
8919 /**
8920  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8921  * @dev: the netdev being configured
8922  * @nlh: RTNL message
8923  *
8924  * Inserts a new hardware bridge if not already created and
8925  * enables the bridging mode requested (VEB or VEPA). If the
8926  * hardware bridge has already been inserted and the request
8927  * is to change the mode then that requires a PF reset to
8928  * allow rebuild of the components with required hardware
8929  * bridge mode enabled.
8930  **/
8931 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8932                                    struct nlmsghdr *nlh,
8933                                    u16 flags)
8934 {
8935         struct i40e_netdev_priv *np = netdev_priv(dev);
8936         struct i40e_vsi *vsi = np->vsi;
8937         struct i40e_pf *pf = vsi->back;
8938         struct i40e_veb *veb = NULL;
8939         struct nlattr *attr, *br_spec;
8940         int i, rem;
8941
8942         /* Only for PF VSI for now */
8943         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8944                 return -EOPNOTSUPP;
8945
8946         /* Find the HW bridge for PF VSI */
8947         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8948                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8949                         veb = pf->veb[i];
8950         }
8951
8952         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8953
8954         nla_for_each_nested(attr, br_spec, rem) {
8955                 __u16 mode;
8956
8957                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8958                         continue;
8959
8960                 mode = nla_get_u16(attr);
8961                 if ((mode != BRIDGE_MODE_VEPA) &&
8962                     (mode != BRIDGE_MODE_VEB))
8963                         return -EINVAL;
8964
8965                 /* Insert a new HW bridge */
8966                 if (!veb) {
8967                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8968                                              vsi->tc_config.enabled_tc);
8969                         if (veb) {
8970                                 veb->bridge_mode = mode;
8971                                 i40e_config_bridge_mode(veb);
8972                         } else {
8973                                 /* No Bridge HW offload available */
8974                                 return -ENOENT;
8975                         }
8976                         break;
8977                 } else if (mode != veb->bridge_mode) {
8978                         /* Existing HW bridge but different mode needs reset */
8979                         veb->bridge_mode = mode;
8980                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8981                         if (mode == BRIDGE_MODE_VEB)
8982                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8983                         else
8984                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8985                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8986                         break;
8987                 }
8988         }
8989
8990         return 0;
8991 }
8992
8993 /**
8994  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8995  * @skb: skb buff
8996  * @pid: process id
8997  * @seq: RTNL message seq #
8998  * @dev: the netdev being configured
8999  * @filter_mask: unused
9000  * @nlflags: netlink flags passed in
9001  *
9002  * Return the mode in which the hardware bridge is operating in
9003  * i.e VEB or VEPA.
9004  **/
9005 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9006                                    struct net_device *dev,
9007                                    u32 __always_unused filter_mask,
9008                                    int nlflags)
9009 {
9010         struct i40e_netdev_priv *np = netdev_priv(dev);
9011         struct i40e_vsi *vsi = np->vsi;
9012         struct i40e_pf *pf = vsi->back;
9013         struct i40e_veb *veb = NULL;
9014         int i;
9015
9016         /* Only for PF VSI for now */
9017         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9018                 return -EOPNOTSUPP;
9019
9020         /* Find the HW bridge for the PF VSI */
9021         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9022                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9023                         veb = pf->veb[i];
9024         }
9025
9026         if (!veb)
9027                 return 0;
9028
9029         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9030                                        nlflags, 0, 0, filter_mask, NULL);
9031 }
9032
9033 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9034  * inner mac plus all inner ethertypes.
9035  */
9036 #define I40E_MAX_TUNNEL_HDR_LEN 128
9037 /**
9038  * i40e_features_check - Validate encapsulated packet conforms to limits
9039  * @skb: skb buff
9040  * @dev: This physical port's netdev
9041  * @features: Offload features that the stack believes apply
9042  **/
9043 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9044                                              struct net_device *dev,
9045                                              netdev_features_t features)
9046 {
9047         if (skb->encapsulation &&
9048             ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9049              I40E_MAX_TUNNEL_HDR_LEN))
9050                 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9051
9052         return features;
9053 }
9054
9055 static const struct net_device_ops i40e_netdev_ops = {
9056         .ndo_open               = i40e_open,
9057         .ndo_stop               = i40e_close,
9058         .ndo_start_xmit         = i40e_lan_xmit_frame,
9059         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
9060         .ndo_set_rx_mode        = i40e_set_rx_mode,
9061         .ndo_validate_addr      = eth_validate_addr,
9062         .ndo_set_mac_address    = i40e_set_mac,
9063         .ndo_change_mtu         = i40e_change_mtu,
9064         .ndo_do_ioctl           = i40e_ioctl,
9065         .ndo_tx_timeout         = i40e_tx_timeout,
9066         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
9067         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
9068 #ifdef CONFIG_NET_POLL_CONTROLLER
9069         .ndo_poll_controller    = i40e_netpoll,
9070 #endif
9071         .ndo_setup_tc           = __i40e_setup_tc,
9072 #ifdef I40E_FCOE
9073         .ndo_fcoe_enable        = i40e_fcoe_enable,
9074         .ndo_fcoe_disable       = i40e_fcoe_disable,
9075 #endif
9076         .ndo_set_features       = i40e_set_features,
9077         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
9078         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
9079         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
9080         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
9081         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
9082         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
9083         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
9084 #if IS_ENABLED(CONFIG_VXLAN)
9085         .ndo_add_vxlan_port     = i40e_add_vxlan_port,
9086         .ndo_del_vxlan_port     = i40e_del_vxlan_port,
9087 #endif
9088 #if IS_ENABLED(CONFIG_GENEVE)
9089         .ndo_add_geneve_port    = i40e_add_geneve_port,
9090         .ndo_del_geneve_port    = i40e_del_geneve_port,
9091 #endif
9092         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
9093         .ndo_fdb_add            = i40e_ndo_fdb_add,
9094         .ndo_features_check     = i40e_features_check,
9095         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
9096         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
9097 };
9098
9099 /**
9100  * i40e_config_netdev - Setup the netdev flags
9101  * @vsi: the VSI being configured
9102  *
9103  * Returns 0 on success, negative value on failure
9104  **/
9105 static int i40e_config_netdev(struct i40e_vsi *vsi)
9106 {
9107         u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
9108         struct i40e_pf *pf = vsi->back;
9109         struct i40e_hw *hw = &pf->hw;
9110         struct i40e_netdev_priv *np;
9111         struct net_device *netdev;
9112         u8 mac_addr[ETH_ALEN];
9113         int etherdev_size;
9114
9115         etherdev_size = sizeof(struct i40e_netdev_priv);
9116         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9117         if (!netdev)
9118                 return -ENOMEM;
9119
9120         vsi->netdev = netdev;
9121         np = netdev_priv(netdev);
9122         np->vsi = vsi;
9123
9124         netdev->hw_enc_features |= NETIF_F_SG                   |
9125                                    NETIF_F_IP_CSUM              |
9126                                    NETIF_F_IPV6_CSUM            |
9127                                    NETIF_F_HIGHDMA              |
9128                                    NETIF_F_SOFT_FEATURES        |
9129                                    NETIF_F_TSO                  |
9130                                    NETIF_F_TSO_ECN              |
9131                                    NETIF_F_TSO6                 |
9132                                    NETIF_F_GSO_GRE              |
9133                                    NETIF_F_GSO_GRE_CSUM         |
9134                                    NETIF_F_GSO_IPIP             |
9135                                    NETIF_F_GSO_SIT              |
9136                                    NETIF_F_GSO_UDP_TUNNEL       |
9137                                    NETIF_F_GSO_UDP_TUNNEL_CSUM  |
9138                                    NETIF_F_GSO_PARTIAL          |
9139                                    NETIF_F_SCTP_CRC             |
9140                                    NETIF_F_RXHASH               |
9141                                    NETIF_F_RXCSUM               |
9142                                    0;
9143
9144         if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9145                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9146
9147         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9148
9149         /* record features VLANs can make use of */
9150         netdev->vlan_features |= netdev->hw_enc_features |
9151                                  NETIF_F_TSO_MANGLEID;
9152
9153         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9154                 netdev->hw_features |= NETIF_F_NTUPLE;
9155
9156         netdev->hw_features |= netdev->hw_enc_features  |
9157                                NETIF_F_HW_VLAN_CTAG_TX  |
9158                                NETIF_F_HW_VLAN_CTAG_RX;
9159
9160         netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9161         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9162
9163         if (vsi->type == I40E_VSI_MAIN) {
9164                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9165                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9166                 /* The following steps are necessary to prevent reception
9167                  * of tagged packets - some older NVM configurations load a
9168                  * default a MAC-VLAN filter that accepts any tagged packet
9169                  * which must be replaced by a normal filter.
9170                  */
9171                 if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
9172                         spin_lock_bh(&vsi->mac_filter_list_lock);
9173                         i40e_add_filter(vsi, mac_addr,
9174                                         I40E_VLAN_ANY, false, true);
9175                         spin_unlock_bh(&vsi->mac_filter_list_lock);
9176                 }
9177         } else if ((pf->hw.aq.api_maj_ver > 1) ||
9178                    ((pf->hw.aq.api_maj_ver == 1) &&
9179                     (pf->hw.aq.api_min_ver > 4))) {
9180                 /* Supported in FW API version higher than 1.4 */
9181                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
9182                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
9183         } else {
9184                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9185                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9186                          pf->vsi[pf->lan_vsi]->netdev->name);
9187                 random_ether_addr(mac_addr);
9188
9189                 spin_lock_bh(&vsi->mac_filter_list_lock);
9190                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9191                 spin_unlock_bh(&vsi->mac_filter_list_lock);
9192         }
9193
9194         spin_lock_bh(&vsi->mac_filter_list_lock);
9195         i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
9196         spin_unlock_bh(&vsi->mac_filter_list_lock);
9197
9198         ether_addr_copy(netdev->dev_addr, mac_addr);
9199         ether_addr_copy(netdev->perm_addr, mac_addr);
9200
9201         netdev->priv_flags |= IFF_UNICAST_FLT;
9202         netdev->priv_flags |= IFF_SUPP_NOFCS;
9203         /* Setup netdev TC information */
9204         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9205
9206         netdev->netdev_ops = &i40e_netdev_ops;
9207         netdev->watchdog_timeo = 5 * HZ;
9208         i40e_set_ethtool_ops(netdev);
9209 #ifdef I40E_FCOE
9210         i40e_fcoe_config_netdev(netdev, vsi);
9211 #endif
9212
9213         return 0;
9214 }
9215
9216 /**
9217  * i40e_vsi_delete - Delete a VSI from the switch
9218  * @vsi: the VSI being removed
9219  *
9220  * Returns 0 on success, negative value on failure
9221  **/
9222 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9223 {
9224         /* remove default VSI is not allowed */
9225         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9226                 return;
9227
9228         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9229 }
9230
9231 /**
9232  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9233  * @vsi: the VSI being queried
9234  *
9235  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9236  **/
9237 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9238 {
9239         struct i40e_veb *veb;
9240         struct i40e_pf *pf = vsi->back;
9241
9242         /* Uplink is not a bridge so default to VEB */
9243         if (vsi->veb_idx == I40E_NO_VEB)
9244                 return 1;
9245
9246         veb = pf->veb[vsi->veb_idx];
9247         if (!veb) {
9248                 dev_info(&pf->pdev->dev,
9249                          "There is no veb associated with the bridge\n");
9250                 return -ENOENT;
9251         }
9252
9253         /* Uplink is a bridge in VEPA mode */
9254         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9255                 return 0;
9256         } else {
9257                 /* Uplink is a bridge in VEB mode */
9258                 return 1;
9259         }
9260
9261         /* VEPA is now default bridge, so return 0 */
9262         return 0;
9263 }
9264
9265 /**
9266  * i40e_add_vsi - Add a VSI to the switch
9267  * @vsi: the VSI being configured
9268  *
9269  * This initializes a VSI context depending on the VSI type to be added and
9270  * passes it down to the add_vsi aq command.
9271  **/
9272 static int i40e_add_vsi(struct i40e_vsi *vsi)
9273 {
9274         int ret = -ENODEV;
9275         u8 laa_macaddr[ETH_ALEN];
9276         bool found_laa_mac_filter = false;
9277         struct i40e_pf *pf = vsi->back;
9278         struct i40e_hw *hw = &pf->hw;
9279         struct i40e_vsi_context ctxt;
9280         struct i40e_mac_filter *f, *ftmp;
9281
9282         u8 enabled_tc = 0x1; /* TC0 enabled */
9283         int f_count = 0;
9284
9285         memset(&ctxt, 0, sizeof(ctxt));
9286         switch (vsi->type) {
9287         case I40E_VSI_MAIN:
9288                 /* The PF's main VSI is already setup as part of the
9289                  * device initialization, so we'll not bother with
9290                  * the add_vsi call, but we will retrieve the current
9291                  * VSI context.
9292                  */
9293                 ctxt.seid = pf->main_vsi_seid;
9294                 ctxt.pf_num = pf->hw.pf_id;
9295                 ctxt.vf_num = 0;
9296                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9297                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9298                 if (ret) {
9299                         dev_info(&pf->pdev->dev,
9300                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9301                                  i40e_stat_str(&pf->hw, ret),
9302                                  i40e_aq_str(&pf->hw,
9303                                              pf->hw.aq.asq_last_status));
9304                         return -ENOENT;
9305                 }
9306                 vsi->info = ctxt.info;
9307                 vsi->info.valid_sections = 0;
9308
9309                 vsi->seid = ctxt.seid;
9310                 vsi->id = ctxt.vsi_number;
9311
9312                 enabled_tc = i40e_pf_get_tc_map(pf);
9313
9314                 /* MFP mode setup queue map and update VSI */
9315                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9316                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9317                         memset(&ctxt, 0, sizeof(ctxt));
9318                         ctxt.seid = pf->main_vsi_seid;
9319                         ctxt.pf_num = pf->hw.pf_id;
9320                         ctxt.vf_num = 0;
9321                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9322                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9323                         if (ret) {
9324                                 dev_info(&pf->pdev->dev,
9325                                          "update vsi failed, err %s aq_err %s\n",
9326                                          i40e_stat_str(&pf->hw, ret),
9327                                          i40e_aq_str(&pf->hw,
9328                                                     pf->hw.aq.asq_last_status));
9329                                 ret = -ENOENT;
9330                                 goto err;
9331                         }
9332                         /* update the local VSI info queue map */
9333                         i40e_vsi_update_queue_map(vsi, &ctxt);
9334                         vsi->info.valid_sections = 0;
9335                 } else {
9336                         /* Default/Main VSI is only enabled for TC0
9337                          * reconfigure it to enable all TCs that are
9338                          * available on the port in SFP mode.
9339                          * For MFP case the iSCSI PF would use this
9340                          * flow to enable LAN+iSCSI TC.
9341                          */
9342                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9343                         if (ret) {
9344                                 dev_info(&pf->pdev->dev,
9345                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9346                                          enabled_tc,
9347                                          i40e_stat_str(&pf->hw, ret),
9348                                          i40e_aq_str(&pf->hw,
9349                                                     pf->hw.aq.asq_last_status));
9350                                 ret = -ENOENT;
9351                         }
9352                 }
9353                 break;
9354
9355         case I40E_VSI_FDIR:
9356                 ctxt.pf_num = hw->pf_id;
9357                 ctxt.vf_num = 0;
9358                 ctxt.uplink_seid = vsi->uplink_seid;
9359                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9360                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9361                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9362                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9363                         ctxt.info.valid_sections |=
9364                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9365                         ctxt.info.switch_id =
9366                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9367                 }
9368                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9369                 break;
9370
9371         case I40E_VSI_VMDQ2:
9372                 ctxt.pf_num = hw->pf_id;
9373                 ctxt.vf_num = 0;
9374                 ctxt.uplink_seid = vsi->uplink_seid;
9375                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9376                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9377
9378                 /* This VSI is connected to VEB so the switch_id
9379                  * should be set to zero by default.
9380                  */
9381                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9382                         ctxt.info.valid_sections |=
9383                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9384                         ctxt.info.switch_id =
9385                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9386                 }
9387
9388                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9389                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9390                 break;
9391
9392         case I40E_VSI_SRIOV:
9393                 ctxt.pf_num = hw->pf_id;
9394                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9395                 ctxt.uplink_seid = vsi->uplink_seid;
9396                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9397                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9398
9399                 /* This VSI is connected to VEB so the switch_id
9400                  * should be set to zero by default.
9401                  */
9402                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9403                         ctxt.info.valid_sections |=
9404                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9405                         ctxt.info.switch_id =
9406                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9407                 }
9408
9409                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9410                         ctxt.info.valid_sections |=
9411                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9412                         ctxt.info.queueing_opt_flags |=
9413                                                 I40E_AQ_VSI_QUE_OPT_TCP_ENA;
9414                 }
9415
9416                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9417                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9418                 if (pf->vf[vsi->vf_id].spoofchk) {
9419                         ctxt.info.valid_sections |=
9420                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9421                         ctxt.info.sec_flags |=
9422                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9423                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9424                 }
9425                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9426                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9427                 break;
9428
9429 #ifdef I40E_FCOE
9430         case I40E_VSI_FCOE:
9431                 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9432                 if (ret) {
9433                         dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9434                         return ret;
9435                 }
9436                 break;
9437
9438 #endif /* I40E_FCOE */
9439         case I40E_VSI_IWARP:
9440                 /* send down message to iWARP */
9441                 break;
9442
9443         default:
9444                 return -ENODEV;
9445         }
9446
9447         if (vsi->type != I40E_VSI_MAIN) {
9448                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9449                 if (ret) {
9450                         dev_info(&vsi->back->pdev->dev,
9451                                  "add vsi failed, err %s aq_err %s\n",
9452                                  i40e_stat_str(&pf->hw, ret),
9453                                  i40e_aq_str(&pf->hw,
9454                                              pf->hw.aq.asq_last_status));
9455                         ret = -ENOENT;
9456                         goto err;
9457                 }
9458                 vsi->info = ctxt.info;
9459                 vsi->info.valid_sections = 0;
9460                 vsi->seid = ctxt.seid;
9461                 vsi->id = ctxt.vsi_number;
9462         }
9463
9464         spin_lock_bh(&vsi->mac_filter_list_lock);
9465         /* If macvlan filters already exist, force them to get loaded */
9466         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9467                 f->changed = true;
9468                 f_count++;
9469
9470                 /* Expected to have only one MAC filter entry for LAA in list */
9471                 if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
9472                         ether_addr_copy(laa_macaddr, f->macaddr);
9473                         found_laa_mac_filter = true;
9474                 }
9475         }
9476         spin_unlock_bh(&vsi->mac_filter_list_lock);
9477
9478         if (found_laa_mac_filter) {
9479                 struct i40e_aqc_remove_macvlan_element_data element;
9480
9481                 memset(&element, 0, sizeof(element));
9482                 ether_addr_copy(element.mac_addr, laa_macaddr);
9483                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
9484                 ret = i40e_aq_remove_macvlan(hw, vsi->seid,
9485                                              &element, 1, NULL);
9486                 if (ret) {
9487                         /* some older FW has a different default */
9488                         element.flags |=
9489                                        I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
9490                         i40e_aq_remove_macvlan(hw, vsi->seid,
9491                                                &element, 1, NULL);
9492                 }
9493
9494                 i40e_aq_mac_address_write(hw,
9495                                           I40E_AQC_WRITE_TYPE_LAA_WOL,
9496                                           laa_macaddr, NULL);
9497         }
9498
9499         if (f_count) {
9500                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9501                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9502         }
9503
9504         /* Update VSI BW information */
9505         ret = i40e_vsi_get_bw_info(vsi);
9506         if (ret) {
9507                 dev_info(&pf->pdev->dev,
9508                          "couldn't get vsi bw info, err %s aq_err %s\n",
9509                          i40e_stat_str(&pf->hw, ret),
9510                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9511                 /* VSI is already added so not tearing that up */
9512                 ret = 0;
9513         }
9514
9515 err:
9516         return ret;
9517 }
9518
9519 /**
9520  * i40e_vsi_release - Delete a VSI and free its resources
9521  * @vsi: the VSI being removed
9522  *
9523  * Returns 0 on success or < 0 on error
9524  **/
9525 int i40e_vsi_release(struct i40e_vsi *vsi)
9526 {
9527         struct i40e_mac_filter *f, *ftmp;
9528         struct i40e_veb *veb = NULL;
9529         struct i40e_pf *pf;
9530         u16 uplink_seid;
9531         int i, n;
9532
9533         pf = vsi->back;
9534
9535         /* release of a VEB-owner or last VSI is not allowed */
9536         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9537                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9538                          vsi->seid, vsi->uplink_seid);
9539                 return -ENODEV;
9540         }
9541         if (vsi == pf->vsi[pf->lan_vsi] &&
9542             !test_bit(__I40E_DOWN, &pf->state)) {
9543                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9544                 return -ENODEV;
9545         }
9546
9547         uplink_seid = vsi->uplink_seid;
9548         if (vsi->type != I40E_VSI_SRIOV) {
9549                 if (vsi->netdev_registered) {
9550                         vsi->netdev_registered = false;
9551                         if (vsi->netdev) {
9552                                 /* results in a call to i40e_close() */
9553                                 unregister_netdev(vsi->netdev);
9554                         }
9555                 } else {
9556                         i40e_vsi_close(vsi);
9557                 }
9558                 i40e_vsi_disable_irq(vsi);
9559         }
9560
9561         spin_lock_bh(&vsi->mac_filter_list_lock);
9562         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9563                 i40e_del_filter(vsi, f->macaddr, f->vlan,
9564                                 f->is_vf, f->is_netdev);
9565         spin_unlock_bh(&vsi->mac_filter_list_lock);
9566
9567         i40e_sync_vsi_filters(vsi);
9568
9569         i40e_vsi_delete(vsi);
9570         i40e_vsi_free_q_vectors(vsi);
9571         if (vsi->netdev) {
9572                 free_netdev(vsi->netdev);
9573                 vsi->netdev = NULL;
9574         }
9575         i40e_vsi_clear_rings(vsi);
9576         i40e_vsi_clear(vsi);
9577
9578         /* If this was the last thing on the VEB, except for the
9579          * controlling VSI, remove the VEB, which puts the controlling
9580          * VSI onto the next level down in the switch.
9581          *
9582          * Well, okay, there's one more exception here: don't remove
9583          * the orphan VEBs yet.  We'll wait for an explicit remove request
9584          * from up the network stack.
9585          */
9586         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9587                 if (pf->vsi[i] &&
9588                     pf->vsi[i]->uplink_seid == uplink_seid &&
9589                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9590                         n++;      /* count the VSIs */
9591                 }
9592         }
9593         for (i = 0; i < I40E_MAX_VEB; i++) {
9594                 if (!pf->veb[i])
9595                         continue;
9596                 if (pf->veb[i]->uplink_seid == uplink_seid)
9597                         n++;     /* count the VEBs */
9598                 if (pf->veb[i]->seid == uplink_seid)
9599                         veb = pf->veb[i];
9600         }
9601         if (n == 0 && veb && veb->uplink_seid != 0)
9602                 i40e_veb_release(veb);
9603
9604         return 0;
9605 }
9606
9607 /**
9608  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9609  * @vsi: ptr to the VSI
9610  *
9611  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9612  * corresponding SW VSI structure and initializes num_queue_pairs for the
9613  * newly allocated VSI.
9614  *
9615  * Returns 0 on success or negative on failure
9616  **/
9617 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9618 {
9619         int ret = -ENOENT;
9620         struct i40e_pf *pf = vsi->back;
9621
9622         if (vsi->q_vectors[0]) {
9623                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9624                          vsi->seid);
9625                 return -EEXIST;
9626         }
9627
9628         if (vsi->base_vector) {
9629                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9630                          vsi->seid, vsi->base_vector);
9631                 return -EEXIST;
9632         }
9633
9634         ret = i40e_vsi_alloc_q_vectors(vsi);
9635         if (ret) {
9636                 dev_info(&pf->pdev->dev,
9637                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9638                          vsi->num_q_vectors, vsi->seid, ret);
9639                 vsi->num_q_vectors = 0;
9640                 goto vector_setup_out;
9641         }
9642
9643         /* In Legacy mode, we do not have to get any other vector since we
9644          * piggyback on the misc/ICR0 for queue interrupts.
9645         */
9646         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9647                 return ret;
9648         if (vsi->num_q_vectors)
9649                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9650                                                  vsi->num_q_vectors, vsi->idx);
9651         if (vsi->base_vector < 0) {
9652                 dev_info(&pf->pdev->dev,
9653                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9654                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9655                 i40e_vsi_free_q_vectors(vsi);
9656                 ret = -ENOENT;
9657                 goto vector_setup_out;
9658         }
9659
9660 vector_setup_out:
9661         return ret;
9662 }
9663
9664 /**
9665  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9666  * @vsi: pointer to the vsi.
9667  *
9668  * This re-allocates a vsi's queue resources.
9669  *
9670  * Returns pointer to the successfully allocated and configured VSI sw struct
9671  * on success, otherwise returns NULL on failure.
9672  **/
9673 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9674 {
9675         struct i40e_pf *pf;
9676         u8 enabled_tc;
9677         int ret;
9678
9679         if (!vsi)
9680                 return NULL;
9681
9682         pf = vsi->back;
9683
9684         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9685         i40e_vsi_clear_rings(vsi);
9686
9687         i40e_vsi_free_arrays(vsi, false);
9688         i40e_set_num_rings_in_vsi(vsi);
9689         ret = i40e_vsi_alloc_arrays(vsi, false);
9690         if (ret)
9691                 goto err_vsi;
9692
9693         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9694         if (ret < 0) {
9695                 dev_info(&pf->pdev->dev,
9696                          "failed to get tracking for %d queues for VSI %d err %d\n",
9697                          vsi->alloc_queue_pairs, vsi->seid, ret);
9698                 goto err_vsi;
9699         }
9700         vsi->base_queue = ret;
9701
9702         /* Update the FW view of the VSI. Force a reset of TC and queue
9703          * layout configurations.
9704          */
9705         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9706         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9707         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9708         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9709
9710         /* assign it some queues */
9711         ret = i40e_alloc_rings(vsi);
9712         if (ret)
9713                 goto err_rings;
9714
9715         /* map all of the rings to the q_vectors */
9716         i40e_vsi_map_rings_to_vectors(vsi);
9717         return vsi;
9718
9719 err_rings:
9720         i40e_vsi_free_q_vectors(vsi);
9721         if (vsi->netdev_registered) {
9722                 vsi->netdev_registered = false;
9723                 unregister_netdev(vsi->netdev);
9724                 free_netdev(vsi->netdev);
9725                 vsi->netdev = NULL;
9726         }
9727         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9728 err_vsi:
9729         i40e_vsi_clear(vsi);
9730         return NULL;
9731 }
9732
9733 /**
9734  * i40e_macaddr_init - explicitly write the mac address filters.
9735  *
9736  * @vsi: pointer to the vsi.
9737  * @macaddr: the MAC address
9738  *
9739  * This is needed when the macaddr has been obtained by other
9740  * means than the default, e.g., from Open Firmware or IDPROM.
9741  * Returns 0 on success, negative on failure
9742  **/
9743 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
9744 {
9745         int ret;
9746         struct i40e_aqc_add_macvlan_element_data element;
9747
9748         ret = i40e_aq_mac_address_write(&vsi->back->hw,
9749                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
9750                                         macaddr, NULL);
9751         if (ret) {
9752                 dev_info(&vsi->back->pdev->dev,
9753                          "Addr change for VSI failed: %d\n", ret);
9754                 return -EADDRNOTAVAIL;
9755         }
9756
9757         memset(&element, 0, sizeof(element));
9758         ether_addr_copy(element.mac_addr, macaddr);
9759         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
9760         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
9761         if (ret) {
9762                 dev_info(&vsi->back->pdev->dev,
9763                          "add filter failed err %s aq_err %s\n",
9764                          i40e_stat_str(&vsi->back->hw, ret),
9765                          i40e_aq_str(&vsi->back->hw,
9766                                      vsi->back->hw.aq.asq_last_status));
9767         }
9768         return ret;
9769 }
9770
9771 /**
9772  * i40e_vsi_setup - Set up a VSI by a given type
9773  * @pf: board private structure
9774  * @type: VSI type
9775  * @uplink_seid: the switch element to link to
9776  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9777  *
9778  * This allocates the sw VSI structure and its queue resources, then add a VSI
9779  * to the identified VEB.
9780  *
9781  * Returns pointer to the successfully allocated and configure VSI sw struct on
9782  * success, otherwise returns NULL on failure.
9783  **/
9784 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9785                                 u16 uplink_seid, u32 param1)
9786 {
9787         struct i40e_vsi *vsi = NULL;
9788         struct i40e_veb *veb = NULL;
9789         int ret, i;
9790         int v_idx;
9791
9792         /* The requested uplink_seid must be either
9793          *     - the PF's port seid
9794          *              no VEB is needed because this is the PF
9795          *              or this is a Flow Director special case VSI
9796          *     - seid of an existing VEB
9797          *     - seid of a VSI that owns an existing VEB
9798          *     - seid of a VSI that doesn't own a VEB
9799          *              a new VEB is created and the VSI becomes the owner
9800          *     - seid of the PF VSI, which is what creates the first VEB
9801          *              this is a special case of the previous
9802          *
9803          * Find which uplink_seid we were given and create a new VEB if needed
9804          */
9805         for (i = 0; i < I40E_MAX_VEB; i++) {
9806                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9807                         veb = pf->veb[i];
9808                         break;
9809                 }
9810         }
9811
9812         if (!veb && uplink_seid != pf->mac_seid) {
9813
9814                 for (i = 0; i < pf->num_alloc_vsi; i++) {
9815                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9816                                 vsi = pf->vsi[i];
9817                                 break;
9818                         }
9819                 }
9820                 if (!vsi) {
9821                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9822                                  uplink_seid);
9823                         return NULL;
9824                 }
9825
9826                 if (vsi->uplink_seid == pf->mac_seid)
9827                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9828                                              vsi->tc_config.enabled_tc);
9829                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9830                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9831                                              vsi->tc_config.enabled_tc);
9832                 if (veb) {
9833                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9834                                 dev_info(&vsi->back->pdev->dev,
9835                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
9836                                 return NULL;
9837                         }
9838                         /* We come up by default in VEPA mode if SRIOV is not
9839                          * already enabled, in which case we can't force VEPA
9840                          * mode.
9841                          */
9842                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9843                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9844                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9845                         }
9846                         i40e_config_bridge_mode(veb);
9847                 }
9848                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9849                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9850                                 veb = pf->veb[i];
9851                 }
9852                 if (!veb) {
9853                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9854                         return NULL;
9855                 }
9856
9857                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9858                 uplink_seid = veb->seid;
9859         }
9860
9861         /* get vsi sw struct */
9862         v_idx = i40e_vsi_mem_alloc(pf, type);
9863         if (v_idx < 0)
9864                 goto err_alloc;
9865         vsi = pf->vsi[v_idx];
9866         if (!vsi)
9867                 goto err_alloc;
9868         vsi->type = type;
9869         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9870
9871         if (type == I40E_VSI_MAIN)
9872                 pf->lan_vsi = v_idx;
9873         else if (type == I40E_VSI_SRIOV)
9874                 vsi->vf_id = param1;
9875         /* assign it some queues */
9876         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9877                                 vsi->idx);
9878         if (ret < 0) {
9879                 dev_info(&pf->pdev->dev,
9880                          "failed to get tracking for %d queues for VSI %d err=%d\n",
9881                          vsi->alloc_queue_pairs, vsi->seid, ret);
9882                 goto err_vsi;
9883         }
9884         vsi->base_queue = ret;
9885
9886         /* get a VSI from the hardware */
9887         vsi->uplink_seid = uplink_seid;
9888         ret = i40e_add_vsi(vsi);
9889         if (ret)
9890                 goto err_vsi;
9891
9892         switch (vsi->type) {
9893         /* setup the netdev if needed */
9894         case I40E_VSI_MAIN:
9895                 /* Apply relevant filters if a platform-specific mac
9896                  * address was selected.
9897                  */
9898                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9899                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9900                         if (ret) {
9901                                 dev_warn(&pf->pdev->dev,
9902                                          "could not set up macaddr; err %d\n",
9903                                          ret);
9904                         }
9905                 }
9906         case I40E_VSI_VMDQ2:
9907         case I40E_VSI_FCOE:
9908                 ret = i40e_config_netdev(vsi);
9909                 if (ret)
9910                         goto err_netdev;
9911                 ret = register_netdev(vsi->netdev);
9912                 if (ret)
9913                         goto err_netdev;
9914                 vsi->netdev_registered = true;
9915                 netif_carrier_off(vsi->netdev);
9916 #ifdef CONFIG_I40E_DCB
9917                 /* Setup DCB netlink interface */
9918                 i40e_dcbnl_setup(vsi);
9919 #endif /* CONFIG_I40E_DCB */
9920                 /* fall through */
9921
9922         case I40E_VSI_FDIR:
9923                 /* set up vectors and rings if needed */
9924                 ret = i40e_vsi_setup_vectors(vsi);
9925                 if (ret)
9926                         goto err_msix;
9927
9928                 ret = i40e_alloc_rings(vsi);
9929                 if (ret)
9930                         goto err_rings;
9931
9932                 /* map all of the rings to the q_vectors */
9933                 i40e_vsi_map_rings_to_vectors(vsi);
9934
9935                 i40e_vsi_reset_stats(vsi);
9936                 break;
9937
9938         default:
9939                 /* no netdev or rings for the other VSI types */
9940                 break;
9941         }
9942
9943         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9944             (vsi->type == I40E_VSI_VMDQ2)) {
9945                 ret = i40e_vsi_config_rss(vsi);
9946         }
9947         return vsi;
9948
9949 err_rings:
9950         i40e_vsi_free_q_vectors(vsi);
9951 err_msix:
9952         if (vsi->netdev_registered) {
9953                 vsi->netdev_registered = false;
9954                 unregister_netdev(vsi->netdev);
9955                 free_netdev(vsi->netdev);
9956                 vsi->netdev = NULL;
9957         }
9958 err_netdev:
9959         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9960 err_vsi:
9961         i40e_vsi_clear(vsi);
9962 err_alloc:
9963         return NULL;
9964 }
9965
9966 /**
9967  * i40e_veb_get_bw_info - Query VEB BW information
9968  * @veb: the veb to query
9969  *
9970  * Query the Tx scheduler BW configuration data for given VEB
9971  **/
9972 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9973 {
9974         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9975         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9976         struct i40e_pf *pf = veb->pf;
9977         struct i40e_hw *hw = &pf->hw;
9978         u32 tc_bw_max;
9979         int ret = 0;
9980         int i;
9981
9982         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9983                                                   &bw_data, NULL);
9984         if (ret) {
9985                 dev_info(&pf->pdev->dev,
9986                          "query veb bw config failed, err %s aq_err %s\n",
9987                          i40e_stat_str(&pf->hw, ret),
9988                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9989                 goto out;
9990         }
9991
9992         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9993                                                    &ets_data, NULL);
9994         if (ret) {
9995                 dev_info(&pf->pdev->dev,
9996                          "query veb bw ets config failed, err %s aq_err %s\n",
9997                          i40e_stat_str(&pf->hw, ret),
9998                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9999                 goto out;
10000         }
10001
10002         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
10003         veb->bw_max_quanta = ets_data.tc_bw_max;
10004         veb->is_abs_credits = bw_data.absolute_credits_enable;
10005         veb->enabled_tc = ets_data.tc_valid_bits;
10006         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
10007                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
10008         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
10009                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
10010                 veb->bw_tc_limit_credits[i] =
10011                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
10012                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
10013         }
10014
10015 out:
10016         return ret;
10017 }
10018
10019 /**
10020  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
10021  * @pf: board private structure
10022  *
10023  * On error: returns error code (negative)
10024  * On success: returns vsi index in PF (positive)
10025  **/
10026 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
10027 {
10028         int ret = -ENOENT;
10029         struct i40e_veb *veb;
10030         int i;
10031
10032         /* Need to protect the allocation of switch elements at the PF level */
10033         mutex_lock(&pf->switch_mutex);
10034
10035         /* VEB list may be fragmented if VEB creation/destruction has
10036          * been happening.  We can afford to do a quick scan to look
10037          * for any free slots in the list.
10038          *
10039          * find next empty veb slot, looping back around if necessary
10040          */
10041         i = 0;
10042         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
10043                 i++;
10044         if (i >= I40E_MAX_VEB) {
10045                 ret = -ENOMEM;
10046                 goto err_alloc_veb;  /* out of VEB slots! */
10047         }
10048
10049         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
10050         if (!veb) {
10051                 ret = -ENOMEM;
10052                 goto err_alloc_veb;
10053         }
10054         veb->pf = pf;
10055         veb->idx = i;
10056         veb->enabled_tc = 1;
10057
10058         pf->veb[i] = veb;
10059         ret = i;
10060 err_alloc_veb:
10061         mutex_unlock(&pf->switch_mutex);
10062         return ret;
10063 }
10064
10065 /**
10066  * i40e_switch_branch_release - Delete a branch of the switch tree
10067  * @branch: where to start deleting
10068  *
10069  * This uses recursion to find the tips of the branch to be
10070  * removed, deleting until we get back to and can delete this VEB.
10071  **/
10072 static void i40e_switch_branch_release(struct i40e_veb *branch)
10073 {
10074         struct i40e_pf *pf = branch->pf;
10075         u16 branch_seid = branch->seid;
10076         u16 veb_idx = branch->idx;
10077         int i;
10078
10079         /* release any VEBs on this VEB - RECURSION */
10080         for (i = 0; i < I40E_MAX_VEB; i++) {
10081                 if (!pf->veb[i])
10082                         continue;
10083                 if (pf->veb[i]->uplink_seid == branch->seid)
10084                         i40e_switch_branch_release(pf->veb[i]);
10085         }
10086
10087         /* Release the VSIs on this VEB, but not the owner VSI.
10088          *
10089          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10090          *       the VEB itself, so don't use (*branch) after this loop.
10091          */
10092         for (i = 0; i < pf->num_alloc_vsi; i++) {
10093                 if (!pf->vsi[i])
10094                         continue;
10095                 if (pf->vsi[i]->uplink_seid == branch_seid &&
10096                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10097                         i40e_vsi_release(pf->vsi[i]);
10098                 }
10099         }
10100
10101         /* There's one corner case where the VEB might not have been
10102          * removed, so double check it here and remove it if needed.
10103          * This case happens if the veb was created from the debugfs
10104          * commands and no VSIs were added to it.
10105          */
10106         if (pf->veb[veb_idx])
10107                 i40e_veb_release(pf->veb[veb_idx]);
10108 }
10109
10110 /**
10111  * i40e_veb_clear - remove veb struct
10112  * @veb: the veb to remove
10113  **/
10114 static void i40e_veb_clear(struct i40e_veb *veb)
10115 {
10116         if (!veb)
10117                 return;
10118
10119         if (veb->pf) {
10120                 struct i40e_pf *pf = veb->pf;
10121
10122                 mutex_lock(&pf->switch_mutex);
10123                 if (pf->veb[veb->idx] == veb)
10124                         pf->veb[veb->idx] = NULL;
10125                 mutex_unlock(&pf->switch_mutex);
10126         }
10127
10128         kfree(veb);
10129 }
10130
10131 /**
10132  * i40e_veb_release - Delete a VEB and free its resources
10133  * @veb: the VEB being removed
10134  **/
10135 void i40e_veb_release(struct i40e_veb *veb)
10136 {
10137         struct i40e_vsi *vsi = NULL;
10138         struct i40e_pf *pf;
10139         int i, n = 0;
10140
10141         pf = veb->pf;
10142
10143         /* find the remaining VSI and check for extras */
10144         for (i = 0; i < pf->num_alloc_vsi; i++) {
10145                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10146                         n++;
10147                         vsi = pf->vsi[i];
10148                 }
10149         }
10150         if (n != 1) {
10151                 dev_info(&pf->pdev->dev,
10152                          "can't remove VEB %d with %d VSIs left\n",
10153                          veb->seid, n);
10154                 return;
10155         }
10156
10157         /* move the remaining VSI to uplink veb */
10158         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10159         if (veb->uplink_seid) {
10160                 vsi->uplink_seid = veb->uplink_seid;
10161                 if (veb->uplink_seid == pf->mac_seid)
10162                         vsi->veb_idx = I40E_NO_VEB;
10163                 else
10164                         vsi->veb_idx = veb->veb_idx;
10165         } else {
10166                 /* floating VEB */
10167                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10168                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10169         }
10170
10171         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10172         i40e_veb_clear(veb);
10173 }
10174
10175 /**
10176  * i40e_add_veb - create the VEB in the switch
10177  * @veb: the VEB to be instantiated
10178  * @vsi: the controlling VSI
10179  **/
10180 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10181 {
10182         struct i40e_pf *pf = veb->pf;
10183         bool is_default = veb->pf->cur_promisc;
10184         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10185         int ret;
10186
10187         /* get a VEB from the hardware */
10188         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10189                               veb->enabled_tc, is_default,
10190                               &veb->seid, enable_stats, NULL);
10191         if (ret) {
10192                 dev_info(&pf->pdev->dev,
10193                          "couldn't add VEB, err %s aq_err %s\n",
10194                          i40e_stat_str(&pf->hw, ret),
10195                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10196                 return -EPERM;
10197         }
10198
10199         /* get statistics counter */
10200         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10201                                          &veb->stats_idx, NULL, NULL, NULL);
10202         if (ret) {
10203                 dev_info(&pf->pdev->dev,
10204                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10205                          i40e_stat_str(&pf->hw, ret),
10206                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10207                 return -EPERM;
10208         }
10209         ret = i40e_veb_get_bw_info(veb);
10210         if (ret) {
10211                 dev_info(&pf->pdev->dev,
10212                          "couldn't get VEB bw info, err %s aq_err %s\n",
10213                          i40e_stat_str(&pf->hw, ret),
10214                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10215                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10216                 return -ENOENT;
10217         }
10218
10219         vsi->uplink_seid = veb->seid;
10220         vsi->veb_idx = veb->idx;
10221         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10222
10223         return 0;
10224 }
10225
10226 /**
10227  * i40e_veb_setup - Set up a VEB
10228  * @pf: board private structure
10229  * @flags: VEB setup flags
10230  * @uplink_seid: the switch element to link to
10231  * @vsi_seid: the initial VSI seid
10232  * @enabled_tc: Enabled TC bit-map
10233  *
10234  * This allocates the sw VEB structure and links it into the switch
10235  * It is possible and legal for this to be a duplicate of an already
10236  * existing VEB.  It is also possible for both uplink and vsi seids
10237  * to be zero, in order to create a floating VEB.
10238  *
10239  * Returns pointer to the successfully allocated VEB sw struct on
10240  * success, otherwise returns NULL on failure.
10241  **/
10242 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10243                                 u16 uplink_seid, u16 vsi_seid,
10244                                 u8 enabled_tc)
10245 {
10246         struct i40e_veb *veb, *uplink_veb = NULL;
10247         int vsi_idx, veb_idx;
10248         int ret;
10249
10250         /* if one seid is 0, the other must be 0 to create a floating relay */
10251         if ((uplink_seid == 0 || vsi_seid == 0) &&
10252             (uplink_seid + vsi_seid != 0)) {
10253                 dev_info(&pf->pdev->dev,
10254                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10255                          uplink_seid, vsi_seid);
10256                 return NULL;
10257         }
10258
10259         /* make sure there is such a vsi and uplink */
10260         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10261                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10262                         break;
10263         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10264                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10265                          vsi_seid);
10266                 return NULL;
10267         }
10268
10269         if (uplink_seid && uplink_seid != pf->mac_seid) {
10270                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10271                         if (pf->veb[veb_idx] &&
10272                             pf->veb[veb_idx]->seid == uplink_seid) {
10273                                 uplink_veb = pf->veb[veb_idx];
10274                                 break;
10275                         }
10276                 }
10277                 if (!uplink_veb) {
10278                         dev_info(&pf->pdev->dev,
10279                                  "uplink seid %d not found\n", uplink_seid);
10280                         return NULL;
10281                 }
10282         }
10283
10284         /* get veb sw struct */
10285         veb_idx = i40e_veb_mem_alloc(pf);
10286         if (veb_idx < 0)
10287                 goto err_alloc;
10288         veb = pf->veb[veb_idx];
10289         veb->flags = flags;
10290         veb->uplink_seid = uplink_seid;
10291         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10292         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10293
10294         /* create the VEB in the switch */
10295         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10296         if (ret)
10297                 goto err_veb;
10298         if (vsi_idx == pf->lan_vsi)
10299                 pf->lan_veb = veb->idx;
10300
10301         return veb;
10302
10303 err_veb:
10304         i40e_veb_clear(veb);
10305 err_alloc:
10306         return NULL;
10307 }
10308
10309 /**
10310  * i40e_setup_pf_switch_element - set PF vars based on switch type
10311  * @pf: board private structure
10312  * @ele: element we are building info from
10313  * @num_reported: total number of elements
10314  * @printconfig: should we print the contents
10315  *
10316  * helper function to assist in extracting a few useful SEID values.
10317  **/
10318 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10319                                 struct i40e_aqc_switch_config_element_resp *ele,
10320                                 u16 num_reported, bool printconfig)
10321 {
10322         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10323         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10324         u8 element_type = ele->element_type;
10325         u16 seid = le16_to_cpu(ele->seid);
10326
10327         if (printconfig)
10328                 dev_info(&pf->pdev->dev,
10329                          "type=%d seid=%d uplink=%d downlink=%d\n",
10330                          element_type, seid, uplink_seid, downlink_seid);
10331
10332         switch (element_type) {
10333         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10334                 pf->mac_seid = seid;
10335                 break;
10336         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10337                 /* Main VEB? */
10338                 if (uplink_seid != pf->mac_seid)
10339                         break;
10340                 if (pf->lan_veb == I40E_NO_VEB) {
10341                         int v;
10342
10343                         /* find existing or else empty VEB */
10344                         for (v = 0; v < I40E_MAX_VEB; v++) {
10345                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10346                                         pf->lan_veb = v;
10347                                         break;
10348                                 }
10349                         }
10350                         if (pf->lan_veb == I40E_NO_VEB) {
10351                                 v = i40e_veb_mem_alloc(pf);
10352                                 if (v < 0)
10353                                         break;
10354                                 pf->lan_veb = v;
10355                         }
10356                 }
10357
10358                 pf->veb[pf->lan_veb]->seid = seid;
10359                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10360                 pf->veb[pf->lan_veb]->pf = pf;
10361                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10362                 break;
10363         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10364                 if (num_reported != 1)
10365                         break;
10366                 /* This is immediately after a reset so we can assume this is
10367                  * the PF's VSI
10368                  */
10369                 pf->mac_seid = uplink_seid;
10370                 pf->pf_seid = downlink_seid;
10371                 pf->main_vsi_seid = seid;
10372                 if (printconfig)
10373                         dev_info(&pf->pdev->dev,
10374                                  "pf_seid=%d main_vsi_seid=%d\n",
10375                                  pf->pf_seid, pf->main_vsi_seid);
10376                 break;
10377         case I40E_SWITCH_ELEMENT_TYPE_PF:
10378         case I40E_SWITCH_ELEMENT_TYPE_VF:
10379         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10380         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10381         case I40E_SWITCH_ELEMENT_TYPE_PE:
10382         case I40E_SWITCH_ELEMENT_TYPE_PA:
10383                 /* ignore these for now */
10384                 break;
10385         default:
10386                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10387                          element_type, seid);
10388                 break;
10389         }
10390 }
10391
10392 /**
10393  * i40e_fetch_switch_configuration - Get switch config from firmware
10394  * @pf: board private structure
10395  * @printconfig: should we print the contents
10396  *
10397  * Get the current switch configuration from the device and
10398  * extract a few useful SEID values.
10399  **/
10400 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10401 {
10402         struct i40e_aqc_get_switch_config_resp *sw_config;
10403         u16 next_seid = 0;
10404         int ret = 0;
10405         u8 *aq_buf;
10406         int i;
10407
10408         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10409         if (!aq_buf)
10410                 return -ENOMEM;
10411
10412         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10413         do {
10414                 u16 num_reported, num_total;
10415
10416                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10417                                                 I40E_AQ_LARGE_BUF,
10418                                                 &next_seid, NULL);
10419                 if (ret) {
10420                         dev_info(&pf->pdev->dev,
10421                                  "get switch config failed err %s aq_err %s\n",
10422                                  i40e_stat_str(&pf->hw, ret),
10423                                  i40e_aq_str(&pf->hw,
10424                                              pf->hw.aq.asq_last_status));
10425                         kfree(aq_buf);
10426                         return -ENOENT;
10427                 }
10428
10429                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10430                 num_total = le16_to_cpu(sw_config->header.num_total);
10431
10432                 if (printconfig)
10433                         dev_info(&pf->pdev->dev,
10434                                  "header: %d reported %d total\n",
10435                                  num_reported, num_total);
10436
10437                 for (i = 0; i < num_reported; i++) {
10438                         struct i40e_aqc_switch_config_element_resp *ele =
10439                                 &sw_config->element[i];
10440
10441                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10442                                                      printconfig);
10443                 }
10444         } while (next_seid != 0);
10445
10446         kfree(aq_buf);
10447         return ret;
10448 }
10449
10450 /**
10451  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10452  * @pf: board private structure
10453  * @reinit: if the Main VSI needs to re-initialized.
10454  *
10455  * Returns 0 on success, negative value on failure
10456  **/
10457 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10458 {
10459         int ret;
10460
10461         /* find out what's out there already */
10462         ret = i40e_fetch_switch_configuration(pf, false);
10463         if (ret) {
10464                 dev_info(&pf->pdev->dev,
10465                          "couldn't fetch switch config, err %s aq_err %s\n",
10466                          i40e_stat_str(&pf->hw, ret),
10467                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10468                 return ret;
10469         }
10470         i40e_pf_reset_stats(pf);
10471
10472         /* first time setup */
10473         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10474                 struct i40e_vsi *vsi = NULL;
10475                 u16 uplink_seid;
10476
10477                 /* Set up the PF VSI associated with the PF's main VSI
10478                  * that is already in the HW switch
10479                  */
10480                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10481                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10482                 else
10483                         uplink_seid = pf->mac_seid;
10484                 if (pf->lan_vsi == I40E_NO_VSI)
10485                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10486                 else if (reinit)
10487                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10488                 if (!vsi) {
10489                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10490                         i40e_fdir_teardown(pf);
10491                         return -EAGAIN;
10492                 }
10493         } else {
10494                 /* force a reset of TC and queue layout configurations */
10495                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10496
10497                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10498                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10499                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10500         }
10501         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10502
10503         i40e_fdir_sb_setup(pf);
10504
10505         /* Setup static PF queue filter control settings */
10506         ret = i40e_setup_pf_filter_control(pf);
10507         if (ret) {
10508                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10509                          ret);
10510                 /* Failure here should not stop continuing other steps */
10511         }
10512
10513         /* enable RSS in the HW, even for only one queue, as the stack can use
10514          * the hash
10515          */
10516         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10517                 i40e_pf_config_rss(pf);
10518
10519         /* fill in link information and enable LSE reporting */
10520         i40e_update_link_info(&pf->hw);
10521         i40e_link_event(pf);
10522
10523         /* Initialize user-specific link properties */
10524         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10525                                   I40E_AQ_AN_COMPLETED) ? true : false);
10526
10527         i40e_ptp_init(pf);
10528
10529         return ret;
10530 }
10531
10532 /**
10533  * i40e_determine_queue_usage - Work out queue distribution
10534  * @pf: board private structure
10535  **/
10536 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10537 {
10538         int queues_left;
10539
10540         pf->num_lan_qps = 0;
10541 #ifdef I40E_FCOE
10542         pf->num_fcoe_qps = 0;
10543 #endif
10544
10545         /* Find the max queues to be put into basic use.  We'll always be
10546          * using TC0, whether or not DCB is running, and TC0 will get the
10547          * big RSS set.
10548          */
10549         queues_left = pf->hw.func_caps.num_tx_qp;
10550
10551         if ((queues_left == 1) ||
10552             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10553                 /* one qp for PF, no queues for anything else */
10554                 queues_left = 0;
10555                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10556
10557                 /* make sure all the fancies are disabled */
10558                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10559                                I40E_FLAG_IWARP_ENABLED  |
10560 #ifdef I40E_FCOE
10561                                I40E_FLAG_FCOE_ENABLED   |
10562 #endif
10563                                I40E_FLAG_FD_SB_ENABLED  |
10564                                I40E_FLAG_FD_ATR_ENABLED |
10565                                I40E_FLAG_DCB_CAPABLE    |
10566                                I40E_FLAG_SRIOV_ENABLED  |
10567                                I40E_FLAG_VMDQ_ENABLED);
10568         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10569                                   I40E_FLAG_FD_SB_ENABLED |
10570                                   I40E_FLAG_FD_ATR_ENABLED |
10571                                   I40E_FLAG_DCB_CAPABLE))) {
10572                 /* one qp for PF */
10573                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10574                 queues_left -= pf->num_lan_qps;
10575
10576                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10577                                I40E_FLAG_IWARP_ENABLED  |
10578 #ifdef I40E_FCOE
10579                                I40E_FLAG_FCOE_ENABLED   |
10580 #endif
10581                                I40E_FLAG_FD_SB_ENABLED  |
10582                                I40E_FLAG_FD_ATR_ENABLED |
10583                                I40E_FLAG_DCB_ENABLED    |
10584                                I40E_FLAG_VMDQ_ENABLED);
10585         } else {
10586                 /* Not enough queues for all TCs */
10587                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10588                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10589                         pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10590                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10591                 }
10592                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10593                                         num_online_cpus());
10594                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10595                                         pf->hw.func_caps.num_tx_qp);
10596
10597                 queues_left -= pf->num_lan_qps;
10598         }
10599
10600 #ifdef I40E_FCOE
10601         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10602                 if (I40E_DEFAULT_FCOE <= queues_left) {
10603                         pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10604                 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10605                         pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10606                 } else {
10607                         pf->num_fcoe_qps = 0;
10608                         pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10609                         dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10610                 }
10611
10612                 queues_left -= pf->num_fcoe_qps;
10613         }
10614
10615 #endif
10616         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10617                 if (queues_left > 1) {
10618                         queues_left -= 1; /* save 1 queue for FD */
10619                 } else {
10620                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10621                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10622                 }
10623         }
10624
10625         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10626             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10627                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10628                                         (queues_left / pf->num_vf_qps));
10629                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10630         }
10631
10632         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10633             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10634                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10635                                           (queues_left / pf->num_vmdq_qps));
10636                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10637         }
10638
10639         pf->queues_left = queues_left;
10640         dev_dbg(&pf->pdev->dev,
10641                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10642                 pf->hw.func_caps.num_tx_qp,
10643                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10644                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10645                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10646                 queues_left);
10647 #ifdef I40E_FCOE
10648         dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10649 #endif
10650 }
10651
10652 /**
10653  * i40e_setup_pf_filter_control - Setup PF static filter control
10654  * @pf: PF to be setup
10655  *
10656  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10657  * settings. If PE/FCoE are enabled then it will also set the per PF
10658  * based filter sizes required for them. It also enables Flow director,
10659  * ethertype and macvlan type filter settings for the pf.
10660  *
10661  * Returns 0 on success, negative on failure
10662  **/
10663 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10664 {
10665         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10666
10667         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10668
10669         /* Flow Director is enabled */
10670         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10671                 settings->enable_fdir = true;
10672
10673         /* Ethtype and MACVLAN filters enabled for PF */
10674         settings->enable_ethtype = true;
10675         settings->enable_macvlan = true;
10676
10677         if (i40e_set_filter_control(&pf->hw, settings))
10678                 return -ENOENT;
10679
10680         return 0;
10681 }
10682
10683 #define INFO_STRING_LEN 255
10684 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10685 static void i40e_print_features(struct i40e_pf *pf)
10686 {
10687         struct i40e_hw *hw = &pf->hw;
10688         char *buf;
10689         int i;
10690
10691         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10692         if (!buf)
10693                 return;
10694
10695         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10696 #ifdef CONFIG_PCI_IOV
10697         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10698 #endif
10699         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d RX: %s",
10700                       pf->hw.func_caps.num_vsis,
10701                       pf->vsi[pf->lan_vsi]->num_queue_pairs,
10702                       pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
10703
10704         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10705                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10706         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10707                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10708         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10709                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10710                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10711         }
10712         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10713                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10714 #if IS_ENABLED(CONFIG_VXLAN)
10715         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10716 #endif
10717 #if IS_ENABLED(CONFIG_GENEVE)
10718         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10719 #endif
10720         if (pf->flags & I40E_FLAG_PTP)
10721                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10722 #ifdef I40E_FCOE
10723         if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10724                 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10725 #endif
10726         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10727                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10728         else
10729                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10730
10731         dev_info(&pf->pdev->dev, "%s\n", buf);
10732         kfree(buf);
10733         WARN_ON(i > INFO_STRING_LEN);
10734 }
10735
10736 /**
10737  * i40e_get_platform_mac_addr - get platform-specific MAC address
10738  *
10739  * @pdev: PCI device information struct
10740  * @pf: board private structure
10741  *
10742  * Look up the MAC address in Open Firmware  on systems that support it,
10743  * and use IDPROM on SPARC if no OF address is found. On return, the
10744  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10745  * has been selected.
10746  **/
10747 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10748 {
10749         pf->flags &= ~I40E_FLAG_PF_MAC;
10750         if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10751                 pf->flags |= I40E_FLAG_PF_MAC;
10752 }
10753
10754 /**
10755  * i40e_probe - Device initialization routine
10756  * @pdev: PCI device information struct
10757  * @ent: entry in i40e_pci_tbl
10758  *
10759  * i40e_probe initializes a PF identified by a pci_dev structure.
10760  * The OS initialization, configuring of the PF private structure,
10761  * and a hardware reset occur.
10762  *
10763  * Returns 0 on success, negative on failure
10764  **/
10765 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10766 {
10767         struct i40e_aq_get_phy_abilities_resp abilities;
10768         struct i40e_pf *pf;
10769         struct i40e_hw *hw;
10770         static u16 pfs_found;
10771         u16 wol_nvm_bits;
10772         u16 link_status;
10773         int err;
10774         u32 val;
10775         u32 i;
10776         u8 set_fc_aq_fail;
10777
10778         err = pci_enable_device_mem(pdev);
10779         if (err)
10780                 return err;
10781
10782         /* set up for high or low dma */
10783         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10784         if (err) {
10785                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10786                 if (err) {
10787                         dev_err(&pdev->dev,
10788                                 "DMA configuration failed: 0x%x\n", err);
10789                         goto err_dma;
10790                 }
10791         }
10792
10793         /* set up pci connections */
10794         err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10795                                            IORESOURCE_MEM), i40e_driver_name);
10796         if (err) {
10797                 dev_info(&pdev->dev,
10798                          "pci_request_selected_regions failed %d\n", err);
10799                 goto err_pci_reg;
10800         }
10801
10802         pci_enable_pcie_error_reporting(pdev);
10803         pci_set_master(pdev);
10804
10805         /* Now that we have a PCI connection, we need to do the
10806          * low level device setup.  This is primarily setting up
10807          * the Admin Queue structures and then querying for the
10808          * device's current profile information.
10809          */
10810         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10811         if (!pf) {
10812                 err = -ENOMEM;
10813                 goto err_pf_alloc;
10814         }
10815         pf->next_vsi = 0;
10816         pf->pdev = pdev;
10817         set_bit(__I40E_DOWN, &pf->state);
10818
10819         hw = &pf->hw;
10820         hw->back = pf;
10821
10822         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10823                                 I40E_MAX_CSR_SPACE);
10824
10825         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10826         if (!hw->hw_addr) {
10827                 err = -EIO;
10828                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10829                          (unsigned int)pci_resource_start(pdev, 0),
10830                          pf->ioremap_len, err);
10831                 goto err_ioremap;
10832         }
10833         hw->vendor_id = pdev->vendor;
10834         hw->device_id = pdev->device;
10835         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10836         hw->subsystem_vendor_id = pdev->subsystem_vendor;
10837         hw->subsystem_device_id = pdev->subsystem_device;
10838         hw->bus.device = PCI_SLOT(pdev->devfn);
10839         hw->bus.func = PCI_FUNC(pdev->devfn);
10840         pf->instance = pfs_found;
10841
10842         /* set up the locks for the AQ, do this only once in probe
10843          * and destroy them only once in remove
10844          */
10845         mutex_init(&hw->aq.asq_mutex);
10846         mutex_init(&hw->aq.arq_mutex);
10847
10848         if (debug != -1) {
10849                 pf->msg_enable = pf->hw.debug_mask;
10850                 pf->msg_enable = debug;
10851         }
10852
10853         /* do a special CORER for clearing PXE mode once at init */
10854         if (hw->revision_id == 0 &&
10855             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10856                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10857                 i40e_flush(hw);
10858                 msleep(200);
10859                 pf->corer_count++;
10860
10861                 i40e_clear_pxe_mode(hw);
10862         }
10863
10864         /* Reset here to make sure all is clean and to define PF 'n' */
10865         i40e_clear_hw(hw);
10866         err = i40e_pf_reset(hw);
10867         if (err) {
10868                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10869                 goto err_pf_reset;
10870         }
10871         pf->pfr_count++;
10872
10873         hw->aq.num_arq_entries = I40E_AQ_LEN;
10874         hw->aq.num_asq_entries = I40E_AQ_LEN;
10875         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10876         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10877         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10878
10879         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10880                  "%s-%s:misc",
10881                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10882
10883         err = i40e_init_shared_code(hw);
10884         if (err) {
10885                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10886                          err);
10887                 goto err_pf_reset;
10888         }
10889
10890         /* set up a default setting for link flow control */
10891         pf->hw.fc.requested_mode = I40E_FC_NONE;
10892
10893         err = i40e_init_adminq(hw);
10894         if (err) {
10895                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10896                         dev_info(&pdev->dev,
10897                                  "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");
10898                 else
10899                         dev_info(&pdev->dev,
10900                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10901
10902                 goto err_pf_reset;
10903         }
10904
10905         /* provide nvm, fw, api versions */
10906         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10907                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10908                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
10909                  i40e_nvm_version_str(hw));
10910
10911         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10912             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10913                 dev_info(&pdev->dev,
10914                          "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");
10915         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10916                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10917                 dev_info(&pdev->dev,
10918                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10919
10920         i40e_verify_eeprom(pf);
10921
10922         /* Rev 0 hardware was never productized */
10923         if (hw->revision_id < 1)
10924                 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");
10925
10926         i40e_clear_pxe_mode(hw);
10927         err = i40e_get_capabilities(pf);
10928         if (err)
10929                 goto err_adminq_setup;
10930
10931         err = i40e_sw_init(pf);
10932         if (err) {
10933                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10934                 goto err_sw_init;
10935         }
10936
10937         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10938                                 hw->func_caps.num_rx_qp,
10939                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10940         if (err) {
10941                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10942                 goto err_init_lan_hmc;
10943         }
10944
10945         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10946         if (err) {
10947                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10948                 err = -ENOENT;
10949                 goto err_configure_lan_hmc;
10950         }
10951
10952         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10953          * Ignore error return codes because if it was already disabled via
10954          * hardware settings this will fail
10955          */
10956         if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10957                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10958                 i40e_aq_stop_lldp(hw, true, NULL);
10959         }
10960
10961         i40e_get_mac_addr(hw, hw->mac.addr);
10962         /* allow a platform config to override the HW addr */
10963         i40e_get_platform_mac_addr(pdev, pf);
10964         if (!is_valid_ether_addr(hw->mac.addr)) {
10965                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10966                 err = -EIO;
10967                 goto err_mac_addr;
10968         }
10969         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10970         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10971         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10972         if (is_valid_ether_addr(hw->mac.port_addr))
10973                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10974 #ifdef I40E_FCOE
10975         err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10976         if (err)
10977                 dev_info(&pdev->dev,
10978                          "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10979         if (!is_valid_ether_addr(hw->mac.san_addr)) {
10980                 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10981                          hw->mac.san_addr);
10982                 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10983         }
10984         dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10985 #endif /* I40E_FCOE */
10986
10987         pci_set_drvdata(pdev, pf);
10988         pci_save_state(pdev);
10989 #ifdef CONFIG_I40E_DCB
10990         err = i40e_init_pf_dcb(pf);
10991         if (err) {
10992                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10993                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10994                 /* Continue without DCB enabled */
10995         }
10996 #endif /* CONFIG_I40E_DCB */
10997
10998         /* set up periodic task facility */
10999         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
11000         pf->service_timer_period = HZ;
11001
11002         INIT_WORK(&pf->service_task, i40e_service_task);
11003         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
11004         pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
11005
11006         /* NVM bit on means WoL disabled for the port */
11007         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
11008         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
11009                 pf->wol_en = false;
11010         else
11011                 pf->wol_en = true;
11012         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
11013
11014         /* set up the main switch operations */
11015         i40e_determine_queue_usage(pf);
11016         err = i40e_init_interrupt_scheme(pf);
11017         if (err)
11018                 goto err_switch_setup;
11019
11020         /* The number of VSIs reported by the FW is the minimum guaranteed
11021          * to us; HW supports far more and we share the remaining pool with
11022          * the other PFs. We allocate space for more than the guarantee with
11023          * the understanding that we might not get them all later.
11024          */
11025         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
11026                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
11027         else
11028                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
11029
11030         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11031         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11032                           GFP_KERNEL);
11033         if (!pf->vsi) {
11034                 err = -ENOMEM;
11035                 goto err_switch_setup;
11036         }
11037
11038 #ifdef CONFIG_PCI_IOV
11039         /* prep for VF support */
11040         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11041             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11042             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11043                 if (pci_num_vf(pdev))
11044                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11045         }
11046 #endif
11047         err = i40e_setup_pf_switch(pf, false);
11048         if (err) {
11049                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11050                 goto err_vsis;
11051         }
11052
11053         /* Make sure flow control is set according to current settings */
11054         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11055         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11056                 dev_dbg(&pf->pdev->dev,
11057                         "Set fc with err %s aq_err %s on get_phy_cap\n",
11058                         i40e_stat_str(hw, err),
11059                         i40e_aq_str(hw, hw->aq.asq_last_status));
11060         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11061                 dev_dbg(&pf->pdev->dev,
11062                         "Set fc with err %s aq_err %s on set_phy_config\n",
11063                         i40e_stat_str(hw, err),
11064                         i40e_aq_str(hw, hw->aq.asq_last_status));
11065         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11066                 dev_dbg(&pf->pdev->dev,
11067                         "Set fc with err %s aq_err %s on get_link_info\n",
11068                         i40e_stat_str(hw, err),
11069                         i40e_aq_str(hw, hw->aq.asq_last_status));
11070
11071         /* if FDIR VSI was set up, start it now */
11072         for (i = 0; i < pf->num_alloc_vsi; i++) {
11073                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11074                         i40e_vsi_open(pf->vsi[i]);
11075                         break;
11076                 }
11077         }
11078
11079         /* The driver only wants link up/down and module qualification
11080          * reports from firmware.  Note the negative logic.
11081          */
11082         err = i40e_aq_set_phy_int_mask(&pf->hw,
11083                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
11084                                          I40E_AQ_EVENT_MEDIA_NA |
11085                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11086         if (err)
11087                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11088                          i40e_stat_str(&pf->hw, err),
11089                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11090
11091         /* Reconfigure hardware for allowing smaller MSS in the case
11092          * of TSO, so that we avoid the MDD being fired and causing
11093          * a reset in the case of small MSS+TSO.
11094          */
11095         val = rd32(hw, I40E_REG_MSS);
11096         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11097                 val &= ~I40E_REG_MSS_MIN_MASK;
11098                 val |= I40E_64BYTE_MSS;
11099                 wr32(hw, I40E_REG_MSS, val);
11100         }
11101
11102         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11103                 msleep(75);
11104                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11105                 if (err)
11106                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11107                                  i40e_stat_str(&pf->hw, err),
11108                                  i40e_aq_str(&pf->hw,
11109                                              pf->hw.aq.asq_last_status));
11110         }
11111         /* The main driver is (mostly) up and happy. We need to set this state
11112          * before setting up the misc vector or we get a race and the vector
11113          * ends up disabled forever.
11114          */
11115         clear_bit(__I40E_DOWN, &pf->state);
11116
11117         /* In case of MSIX we are going to setup the misc vector right here
11118          * to handle admin queue events etc. In case of legacy and MSI
11119          * the misc functionality and queue processing is combined in
11120          * the same vector and that gets setup at open.
11121          */
11122         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11123                 err = i40e_setup_misc_vector(pf);
11124                 if (err) {
11125                         dev_info(&pdev->dev,
11126                                  "setup of misc vector failed: %d\n", err);
11127                         goto err_vsis;
11128                 }
11129         }
11130
11131 #ifdef CONFIG_PCI_IOV
11132         /* prep for VF support */
11133         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11134             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11135             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11136                 /* disable link interrupts for VFs */
11137                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11138                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11139                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11140                 i40e_flush(hw);
11141
11142                 if (pci_num_vf(pdev)) {
11143                         dev_info(&pdev->dev,
11144                                  "Active VFs found, allocating resources.\n");
11145                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11146                         if (err)
11147                                 dev_info(&pdev->dev,
11148                                          "Error %d allocating resources for existing VFs\n",
11149                                          err);
11150                 }
11151         }
11152 #endif /* CONFIG_PCI_IOV */
11153
11154         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11155                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11156                                                       pf->num_iwarp_msix,
11157                                                       I40E_IWARP_IRQ_PILE_ID);
11158                 if (pf->iwarp_base_vector < 0) {
11159                         dev_info(&pdev->dev,
11160                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11161                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11162                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11163                 }
11164         }
11165
11166         i40e_dbg_pf_init(pf);
11167
11168         /* tell the firmware that we're starting */
11169         i40e_send_version(pf);
11170
11171         /* since everything's happy, start the service_task timer */
11172         mod_timer(&pf->service_timer,
11173                   round_jiffies(jiffies + pf->service_timer_period));
11174
11175         /* add this PF to client device list and launch a client service task */
11176         err = i40e_lan_add_device(pf);
11177         if (err)
11178                 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11179                          err);
11180
11181 #ifdef I40E_FCOE
11182         /* create FCoE interface */
11183         i40e_fcoe_vsi_setup(pf);
11184
11185 #endif
11186 #define PCI_SPEED_SIZE 8
11187 #define PCI_WIDTH_SIZE 8
11188         /* Devices on the IOSF bus do not have this information
11189          * and will report PCI Gen 1 x 1 by default so don't bother
11190          * checking them.
11191          */
11192         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11193                 char speed[PCI_SPEED_SIZE] = "Unknown";
11194                 char width[PCI_WIDTH_SIZE] = "Unknown";
11195
11196                 /* Get the negotiated link width and speed from PCI config
11197                  * space
11198                  */
11199                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11200                                           &link_status);
11201
11202                 i40e_set_pci_config_data(hw, link_status);
11203
11204                 switch (hw->bus.speed) {
11205                 case i40e_bus_speed_8000:
11206                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11207                 case i40e_bus_speed_5000:
11208                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11209                 case i40e_bus_speed_2500:
11210                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11211                 default:
11212                         break;
11213                 }
11214                 switch (hw->bus.width) {
11215                 case i40e_bus_width_pcie_x8:
11216                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11217                 case i40e_bus_width_pcie_x4:
11218                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11219                 case i40e_bus_width_pcie_x2:
11220                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11221                 case i40e_bus_width_pcie_x1:
11222                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11223                 default:
11224                         break;
11225                 }
11226
11227                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11228                          speed, width);
11229
11230                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11231                     hw->bus.speed < i40e_bus_speed_8000) {
11232                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11233                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11234                 }
11235         }
11236
11237         /* get the requested speeds from the fw */
11238         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11239         if (err)
11240                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11241                         i40e_stat_str(&pf->hw, err),
11242                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11243         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11244
11245         /* get the supported phy types from the fw */
11246         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11247         if (err)
11248                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11249                         i40e_stat_str(&pf->hw, err),
11250                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11251         pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11252
11253         /* Add a filter to drop all Flow control frames from any VSI from being
11254          * transmitted. By doing so we stop a malicious VF from sending out
11255          * PAUSE or PFC frames and potentially controlling traffic for other
11256          * PF/VF VSIs.
11257          * The FW can still send Flow control frames if enabled.
11258          */
11259         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11260                                                        pf->main_vsi_seid);
11261
11262         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11263             (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11264                 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11265
11266         /* print a string summarizing features */
11267         i40e_print_features(pf);
11268
11269         return 0;
11270
11271         /* Unwind what we've done if something failed in the setup */
11272 err_vsis:
11273         set_bit(__I40E_DOWN, &pf->state);
11274         i40e_clear_interrupt_scheme(pf);
11275         kfree(pf->vsi);
11276 err_switch_setup:
11277         i40e_reset_interrupt_capability(pf);
11278         del_timer_sync(&pf->service_timer);
11279 err_mac_addr:
11280 err_configure_lan_hmc:
11281         (void)i40e_shutdown_lan_hmc(hw);
11282 err_init_lan_hmc:
11283         kfree(pf->qp_pile);
11284 err_sw_init:
11285 err_adminq_setup:
11286 err_pf_reset:
11287         iounmap(hw->hw_addr);
11288 err_ioremap:
11289         kfree(pf);
11290 err_pf_alloc:
11291         pci_disable_pcie_error_reporting(pdev);
11292         pci_release_selected_regions(pdev,
11293                                      pci_select_bars(pdev, IORESOURCE_MEM));
11294 err_pci_reg:
11295 err_dma:
11296         pci_disable_device(pdev);
11297         return err;
11298 }
11299
11300 /**
11301  * i40e_remove - Device removal routine
11302  * @pdev: PCI device information struct
11303  *
11304  * i40e_remove is called by the PCI subsystem to alert the driver
11305  * that is should release a PCI device.  This could be caused by a
11306  * Hot-Plug event, or because the driver is going to be removed from
11307  * memory.
11308  **/
11309 static void i40e_remove(struct pci_dev *pdev)
11310 {
11311         struct i40e_pf *pf = pci_get_drvdata(pdev);
11312         struct i40e_hw *hw = &pf->hw;
11313         i40e_status ret_code;
11314         int i;
11315
11316         i40e_dbg_pf_exit(pf);
11317
11318         i40e_ptp_stop(pf);
11319
11320         /* Disable RSS in hw */
11321         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11322         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11323
11324         /* no more scheduling of any task */
11325         set_bit(__I40E_SUSPENDED, &pf->state);
11326         set_bit(__I40E_DOWN, &pf->state);
11327         if (pf->service_timer.data)
11328                 del_timer_sync(&pf->service_timer);
11329         if (pf->service_task.func)
11330                 cancel_work_sync(&pf->service_task);
11331
11332         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11333                 i40e_free_vfs(pf);
11334                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11335         }
11336
11337         i40e_fdir_teardown(pf);
11338
11339         /* If there is a switch structure or any orphans, remove them.
11340          * This will leave only the PF's VSI remaining.
11341          */
11342         for (i = 0; i < I40E_MAX_VEB; i++) {
11343                 if (!pf->veb[i])
11344                         continue;
11345
11346                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11347                     pf->veb[i]->uplink_seid == 0)
11348                         i40e_switch_branch_release(pf->veb[i]);
11349         }
11350
11351         /* Now we can shutdown the PF's VSI, just before we kill
11352          * adminq and hmc.
11353          */
11354         if (pf->vsi[pf->lan_vsi])
11355                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11356
11357         /* remove attached clients */
11358         ret_code = i40e_lan_del_device(pf);
11359         if (ret_code) {
11360                 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11361                          ret_code);
11362         }
11363
11364         /* shutdown and destroy the HMC */
11365         if (hw->hmc.hmc_obj) {
11366                 ret_code = i40e_shutdown_lan_hmc(hw);
11367                 if (ret_code)
11368                         dev_warn(&pdev->dev,
11369                                  "Failed to destroy the HMC resources: %d\n",
11370                                  ret_code);
11371         }
11372
11373         /* shutdown the adminq */
11374         ret_code = i40e_shutdown_adminq(hw);
11375         if (ret_code)
11376                 dev_warn(&pdev->dev,
11377                          "Failed to destroy the Admin Queue resources: %d\n",
11378                          ret_code);
11379
11380         /* destroy the locks only once, here */
11381         mutex_destroy(&hw->aq.arq_mutex);
11382         mutex_destroy(&hw->aq.asq_mutex);
11383
11384         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11385         i40e_clear_interrupt_scheme(pf);
11386         for (i = 0; i < pf->num_alloc_vsi; i++) {
11387                 if (pf->vsi[i]) {
11388                         i40e_vsi_clear_rings(pf->vsi[i]);
11389                         i40e_vsi_clear(pf->vsi[i]);
11390                         pf->vsi[i] = NULL;
11391                 }
11392         }
11393
11394         for (i = 0; i < I40E_MAX_VEB; i++) {
11395                 kfree(pf->veb[i]);
11396                 pf->veb[i] = NULL;
11397         }
11398
11399         kfree(pf->qp_pile);
11400         kfree(pf->vsi);
11401
11402         iounmap(hw->hw_addr);
11403         kfree(pf);
11404         pci_release_selected_regions(pdev,
11405                                      pci_select_bars(pdev, IORESOURCE_MEM));
11406
11407         pci_disable_pcie_error_reporting(pdev);
11408         pci_disable_device(pdev);
11409 }
11410
11411 /**
11412  * i40e_pci_error_detected - warning that something funky happened in PCI land
11413  * @pdev: PCI device information struct
11414  *
11415  * Called to warn that something happened and the error handling steps
11416  * are in progress.  Allows the driver to quiesce things, be ready for
11417  * remediation.
11418  **/
11419 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11420                                                 enum pci_channel_state error)
11421 {
11422         struct i40e_pf *pf = pci_get_drvdata(pdev);
11423
11424         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11425
11426         /* shutdown all operations */
11427         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11428                 rtnl_lock();
11429                 i40e_prep_for_reset(pf);
11430                 rtnl_unlock();
11431         }
11432
11433         /* Request a slot reset */
11434         return PCI_ERS_RESULT_NEED_RESET;
11435 }
11436
11437 /**
11438  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11439  * @pdev: PCI device information struct
11440  *
11441  * Called to find if the driver can work with the device now that
11442  * the pci slot has been reset.  If a basic connection seems good
11443  * (registers are readable and have sane content) then return a
11444  * happy little PCI_ERS_RESULT_xxx.
11445  **/
11446 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11447 {
11448         struct i40e_pf *pf = pci_get_drvdata(pdev);
11449         pci_ers_result_t result;
11450         int err;
11451         u32 reg;
11452
11453         dev_dbg(&pdev->dev, "%s\n", __func__);
11454         if (pci_enable_device_mem(pdev)) {
11455                 dev_info(&pdev->dev,
11456                          "Cannot re-enable PCI device after reset.\n");
11457                 result = PCI_ERS_RESULT_DISCONNECT;
11458         } else {
11459                 pci_set_master(pdev);
11460                 pci_restore_state(pdev);
11461                 pci_save_state(pdev);
11462                 pci_wake_from_d3(pdev, false);
11463
11464                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11465                 if (reg == 0)
11466                         result = PCI_ERS_RESULT_RECOVERED;
11467                 else
11468                         result = PCI_ERS_RESULT_DISCONNECT;
11469         }
11470
11471         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11472         if (err) {
11473                 dev_info(&pdev->dev,
11474                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11475                          err);
11476                 /* non-fatal, continue */
11477         }
11478
11479         return result;
11480 }
11481
11482 /**
11483  * i40e_pci_error_resume - restart operations after PCI error recovery
11484  * @pdev: PCI device information struct
11485  *
11486  * Called to allow the driver to bring things back up after PCI error
11487  * and/or reset recovery has finished.
11488  **/
11489 static void i40e_pci_error_resume(struct pci_dev *pdev)
11490 {
11491         struct i40e_pf *pf = pci_get_drvdata(pdev);
11492
11493         dev_dbg(&pdev->dev, "%s\n", __func__);
11494         if (test_bit(__I40E_SUSPENDED, &pf->state))
11495                 return;
11496
11497         rtnl_lock();
11498         i40e_handle_reset_warning(pf);
11499         rtnl_unlock();
11500 }
11501
11502 /**
11503  * i40e_shutdown - PCI callback for shutting down
11504  * @pdev: PCI device information struct
11505  **/
11506 static void i40e_shutdown(struct pci_dev *pdev)
11507 {
11508         struct i40e_pf *pf = pci_get_drvdata(pdev);
11509         struct i40e_hw *hw = &pf->hw;
11510
11511         set_bit(__I40E_SUSPENDED, &pf->state);
11512         set_bit(__I40E_DOWN, &pf->state);
11513         rtnl_lock();
11514         i40e_prep_for_reset(pf);
11515         rtnl_unlock();
11516
11517         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11518         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11519
11520         del_timer_sync(&pf->service_timer);
11521         cancel_work_sync(&pf->service_task);
11522         i40e_fdir_teardown(pf);
11523
11524         rtnl_lock();
11525         i40e_prep_for_reset(pf);
11526         rtnl_unlock();
11527
11528         wr32(hw, I40E_PFPM_APM,
11529              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11530         wr32(hw, I40E_PFPM_WUFC,
11531              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11532
11533         i40e_clear_interrupt_scheme(pf);
11534
11535         if (system_state == SYSTEM_POWER_OFF) {
11536                 pci_wake_from_d3(pdev, pf->wol_en);
11537                 pci_set_power_state(pdev, PCI_D3hot);
11538         }
11539 }
11540
11541 #ifdef CONFIG_PM
11542 /**
11543  * i40e_suspend - PCI callback for moving to D3
11544  * @pdev: PCI device information struct
11545  **/
11546 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11547 {
11548         struct i40e_pf *pf = pci_get_drvdata(pdev);
11549         struct i40e_hw *hw = &pf->hw;
11550
11551         set_bit(__I40E_SUSPENDED, &pf->state);
11552         set_bit(__I40E_DOWN, &pf->state);
11553
11554         rtnl_lock();
11555         i40e_prep_for_reset(pf);
11556         rtnl_unlock();
11557
11558         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11559         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11560
11561         pci_wake_from_d3(pdev, pf->wol_en);
11562         pci_set_power_state(pdev, PCI_D3hot);
11563
11564         return 0;
11565 }
11566
11567 /**
11568  * i40e_resume - PCI callback for waking up from D3
11569  * @pdev: PCI device information struct
11570  **/
11571 static int i40e_resume(struct pci_dev *pdev)
11572 {
11573         struct i40e_pf *pf = pci_get_drvdata(pdev);
11574         u32 err;
11575
11576         pci_set_power_state(pdev, PCI_D0);
11577         pci_restore_state(pdev);
11578         /* pci_restore_state() clears dev->state_saves, so
11579          * call pci_save_state() again to restore it.
11580          */
11581         pci_save_state(pdev);
11582
11583         err = pci_enable_device_mem(pdev);
11584         if (err) {
11585                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11586                 return err;
11587         }
11588         pci_set_master(pdev);
11589
11590         /* no wakeup events while running */
11591         pci_wake_from_d3(pdev, false);
11592
11593         /* handling the reset will rebuild the device state */
11594         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11595                 clear_bit(__I40E_DOWN, &pf->state);
11596                 rtnl_lock();
11597                 i40e_reset_and_rebuild(pf, false);
11598                 rtnl_unlock();
11599         }
11600
11601         return 0;
11602 }
11603
11604 #endif
11605 static const struct pci_error_handlers i40e_err_handler = {
11606         .error_detected = i40e_pci_error_detected,
11607         .slot_reset = i40e_pci_error_slot_reset,
11608         .resume = i40e_pci_error_resume,
11609 };
11610
11611 static struct pci_driver i40e_driver = {
11612         .name     = i40e_driver_name,
11613         .id_table = i40e_pci_tbl,
11614         .probe    = i40e_probe,
11615         .remove   = i40e_remove,
11616 #ifdef CONFIG_PM
11617         .suspend  = i40e_suspend,
11618         .resume   = i40e_resume,
11619 #endif
11620         .shutdown = i40e_shutdown,
11621         .err_handler = &i40e_err_handler,
11622         .sriov_configure = i40e_pci_sriov_configure,
11623 };
11624
11625 /**
11626  * i40e_init_module - Driver registration routine
11627  *
11628  * i40e_init_module is the first routine called when the driver is
11629  * loaded. All it does is register with the PCI subsystem.
11630  **/
11631 static int __init i40e_init_module(void)
11632 {
11633         pr_info("%s: %s - version %s\n", i40e_driver_name,
11634                 i40e_driver_string, i40e_driver_version_str);
11635         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11636
11637         /* we will see if single thread per module is enough for now,
11638          * it can't be any worse than using the system workqueue which
11639          * was already single threaded
11640          */
11641         i40e_wq = create_singlethread_workqueue(i40e_driver_name);
11642         if (!i40e_wq) {
11643                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11644                 return -ENOMEM;
11645         }
11646
11647         i40e_dbg_init();
11648         return pci_register_driver(&i40e_driver);
11649 }
11650 module_init(i40e_init_module);
11651
11652 /**
11653  * i40e_exit_module - Driver exit cleanup routine
11654  *
11655  * i40e_exit_module is called just before the driver is removed
11656  * from memory.
11657  **/
11658 static void __exit i40e_exit_module(void)
11659 {
11660         pci_unregister_driver(&i40e_driver);
11661         destroy_workqueue(i40e_wq);
11662         i40e_dbg_exit();
11663 }
11664 module_exit(i40e_exit_module);