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

   1 /*******************************************************************************
   2  *
   3  * Intel Ethernet Controller XL710 Family Linux Driver
   4  * Copyright(c) 2013 - 2015 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 #ifdef CONFIG_SPARC
  32 #include <asm/idprom.h>
  33 #include <asm/prom.h>
  34 #endif
  35
  36 /* Local includes */
  37 #include "i40e.h"
  38 #include "i40e_diag.h"
  39 #if IS_ENABLED(CONFIG_VXLAN)
  40 #include <net/vxlan.h>
  41 #endif
  42 #if IS_ENABLED(CONFIG_GENEVE)
  43 #include <net/geneve.h>
  44 #endif
  45
  46 const char i40e_driver_name[] = "i40e";
  47 static const char i40e_driver_string[] =
  48                         "Intel(R) Ethernet Connection XL710 Network Driver";
  49
  50 #define DRV_KERN "-k"
  51
  52 #define DRV_VERSION_MAJOR 1
  53 #define DRV_VERSION_MINOR 4
  54 #define DRV_VERSION_BUILD 8
  55 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
  56              __stringify(DRV_VERSION_MINOR) "." \
  57              __stringify(DRV_VERSION_BUILD)    DRV_KERN
  58 const char i40e_driver_version_str[] = DRV_VERSION;
  59 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
  60
  61 /* a bit of forward declarations */
  62 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
  63 static void i40e_handle_reset_warning(struct i40e_pf *pf);
  64 static int i40e_add_vsi(struct i40e_vsi *vsi);
  65 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
  66 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
  67 static int i40e_setup_misc_vector(struct i40e_pf *pf);
  68 static void i40e_determine_queue_usage(struct i40e_pf *pf);
  69 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
  70 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
  71                               u16 rss_table_size, u16 rss_size);
  72 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
  73 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
  74
  75 /* i40e_pci_tbl - PCI Device ID Table
  76  *
  77  * Last entry must be all 0s
  78  *
  79  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
  80  *   Class, Class Mask, private data (not used) }
  81  */
  82 static const struct pci_device_id i40e_pci_tbl[] = {
  83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
  84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
  85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
  86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
  87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
  88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
  89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
  90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
  91         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
  92         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  93         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
  94         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
  95         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
  96         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
  97         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
  98         /* required last entry */
  99         {0, }
 100 };
 101 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
 102
 103 #define I40E_MAX_VF_COUNT 128
 104 static int debug = -1;
 105 module_param(debug, int, 0);
 106 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 107
 108 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
 109 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
 110 MODULE_LICENSE("GPL");
 111 MODULE_VERSION(DRV_VERSION);
 112
 113 /**
 114  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
 115  * @hw:   pointer to the HW structure
 116  * @mem:  ptr to mem struct to fill out
 117  * @size: size of memory requested
 118  * @alignment: what to align the allocation to
 119  **/
 120 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
 121                             u64 size, u32 alignment)
 122 {
 123         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 124
 125         mem->size = ALIGN(size, alignment);
 126         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
 127                                       &mem->pa, GFP_KERNEL);
 128         if (!mem->va)
 129                 return -ENOMEM;
 130
 131         return 0;
 132 }
 133
 134 /**
 135  * i40e_free_dma_mem_d - OS specific memory free for shared code
 136  * @hw:   pointer to the HW structure
 137  * @mem:  ptr to mem struct to free
 138  **/
 139 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
 140 {
 141         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
 142
 143         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
 144         mem->va = NULL;
 145         mem->pa = 0;
 146         mem->size = 0;
 147
 148         return 0;
 149 }
 150
 151 /**
 152  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
 153  * @hw:   pointer to the HW structure
 154  * @mem:  ptr to mem struct to fill out
 155  * @size: size of memory requested
 156  **/
 157 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
 158                              u32 size)
 159 {
 160         mem->size = size;
 161         mem->va = kzalloc(size, GFP_KERNEL);
 162
 163         if (!mem->va)
 164                 return -ENOMEM;
 165
 166         return 0;
 167 }
 168
 169 /**
 170  * i40e_free_virt_mem_d - OS specific memory free for shared code
 171  * @hw:   pointer to the HW structure
 172  * @mem:  ptr to mem struct to free
 173  **/
 174 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
 175 {
 176         /* it's ok to kfree a NULL pointer */
 177         kfree(mem->va);
 178         mem->va = NULL;
 179         mem->size = 0;
 180
 181         return 0;
 182 }
 183
 184 /**
 185  * i40e_get_lump - find a lump of free generic resource
 186  * @pf: board private structure
 187  * @pile: the pile of resource to search
 188  * @needed: the number of items needed
 189  * @id: an owner id to stick on the items assigned
 190  *
 191  * Returns the base item index of the lump, or negative for error
 192  *
 193  * The search_hint trick and lack of advanced fit-finding only work
 194  * because we're highly likely to have all the same size lump requests.
 195  * Linear search time and any fragmentation should be minimal.
 196  **/
 197 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
 198                          u16 needed, u16 id)
 199 {
 200         int ret = -ENOMEM;
 201         int i, j;
 202
 203         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
 204                 dev_info(&pf->pdev->dev,
 205                          "param err: pile=%p needed=%d id=0x%04x\n",
 206                          pile, needed, id);
 207                 return -EINVAL;
 208         }
 209
 210         /* start the linear search with an imperfect hint */
 211         i = pile->search_hint;
 212         while (i < pile->num_entries) {
 213                 /* skip already allocated entries */
 214                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
 215                         i++;
 216                         continue;
 217                 }
 218
 219                 /* do we have enough in this lump? */
 220                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
 221                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
 222                                 break;
 223                 }
 224
 225                 if (j == needed) {
 226                         /* there was enough, so assign it to the requestor */
 227                         for (j = 0; j < needed; j++)
 228                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
 229                         ret = i;
 230                         pile->search_hint = i + j;
 231                         break;
 232                 }
 233
 234                 /* not enough, so skip over it and continue looking */
 235                 i += j;
 236         }
 237
 238         return ret;
 239 }
 240
 241 /**
 242  * i40e_put_lump - return a lump of generic resource
 243  * @pile: the pile of resource to search
 244  * @index: the base item index
 245  * @id: the owner id of the items assigned
 246  *
 247  * Returns the count of items in the lump
 248  **/
 249 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
 250 {
 251         int valid_id = (id | I40E_PILE_VALID_BIT);
 252         int count = 0;
 253         int i;
 254
 255         if (!pile || index >= pile->num_entries)
 256                 return -EINVAL;
 257
 258         for (i = index;
 259              i < pile->num_entries && pile->list[i] == valid_id;
 260              i++) {
 261                 pile->list[i] = 0;
 262                 count++;
 263         }
 264
 265         if (count && index < pile->search_hint)
 266                 pile->search_hint = index;
 267
 268         return count;
 269 }
 270
 271 /**
 272  * i40e_find_vsi_from_id - searches for the vsi with the given id
 273  * @pf - the pf structure to search for the vsi
 274  * @id - id of the vsi it is searching for
 275  **/
 276 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
 277 {
 278         int i;
 279
 280         for (i = 0; i < pf->num_alloc_vsi; i++)
 281                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
 282                         return pf->vsi[i];
 283
 284         return NULL;
 285 }
 286
 287 /**
 288  * i40e_service_event_schedule - Schedule the service task to wake up
 289  * @pf: board private structure
 290  *
 291  * If not already scheduled, this puts the task into the work queue
 292  **/
 293 void i40e_service_event_schedule(struct i40e_pf *pf)
 294 {
 295         if (!test_bit(__I40E_DOWN, &pf->state) &&
 296             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
 297             !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
 298                 schedule_work(&pf->service_task);
 299 }
 300
 301 /**
 302  * i40e_tx_timeout - Respond to a Tx Hang
 303  * @netdev: network interface device structure
 304  *
 305  * If any port has noticed a Tx timeout, it is likely that the whole
 306  * device is munged, not just the one netdev port, so go for the full
 307  * reset.
 308  **/
 309 #ifdef I40E_FCOE
 310 void i40e_tx_timeout(struct net_device *netdev)
 311 #else
 312 static void i40e_tx_timeout(struct net_device *netdev)
 313 #endif
 314 {
 315         struct i40e_netdev_priv *np = netdev_priv(netdev);
 316         struct i40e_vsi *vsi = np->vsi;
 317         struct i40e_pf *pf = vsi->back;
 318         struct i40e_ring *tx_ring = NULL;
 319         unsigned int i, hung_queue = 0;
 320         u32 head, val;
 321
 322         pf->tx_timeout_count++;
 323
 324         /* find the stopped queue the same way the stack does */
 325         for (i = 0; i < netdev->num_tx_queues; i++) {
 326                 struct netdev_queue *q;
 327                 unsigned long trans_start;
 328
 329                 q = netdev_get_tx_queue(netdev, i);
 330                 trans_start = q->trans_start ? : netdev->trans_start;
 331                 if (netif_xmit_stopped(q) &&
 332                     time_after(jiffies,
 333                                (trans_start + netdev->watchdog_timeo))) {
 334                         hung_queue = i;
 335                         break;
 336                 }
 337         }
 338
 339         if (i == netdev->num_tx_queues) {
 340                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
 341         } else {
 342                 /* now that we have an index, find the tx_ring struct */
 343                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 344                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
 345                                 if (hung_queue ==
 346                                     vsi->tx_rings[i]->queue_index) {
 347                                         tx_ring = vsi->tx_rings[i];
 348                                         break;
 349                                 }
 350                         }
 351                 }
 352         }
 353
 354         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
 355                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
 356         else if (time_before(jiffies,
 357                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
 358                 return;   /* don't do any new action before the next timeout */
 359
 360         if (tx_ring) {
 361                 head = i40e_get_head(tx_ring);
 362                 /* Read interrupt register */
 363                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
 364                         val = rd32(&pf->hw,
 365                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
 366                                                 tx_ring->vsi->base_vector - 1));
 367                 else
 368                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
 369
 370                 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",
 371                             vsi->seid, hung_queue, tx_ring->next_to_clean,
 372                             head, tx_ring->next_to_use,
 373                             readl(tx_ring->tail), val);
 374         }
 375
 376         pf->tx_timeout_last_recovery = jiffies;
 377         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
 378                     pf->tx_timeout_recovery_level, hung_queue);
 379
 380         switch (pf->tx_timeout_recovery_level) {
 381         case 1:
 382                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
 383                 break;
 384         case 2:
 385                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
 386                 break;
 387         case 3:
 388                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
 389                 break;
 390         default:
 391                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
 392                 break;
 393         }
 394
 395         i40e_service_event_schedule(pf);
 396         pf->tx_timeout_recovery_level++;
 397 }
 398
 399 /**
 400  * i40e_release_rx_desc - Store the new tail and head values
 401  * @rx_ring: ring to bump
 402  * @val: new head index
 403  **/
 404 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
 405 {
 406         rx_ring->next_to_use = val;
 407
 408         /* Force memory writes to complete before letting h/w
 409          * know there are new descriptors to fetch.  (Only
 410          * applicable for weak-ordered memory model archs,
 411          * such as IA-64).
 412          */
 413         wmb();
 414         writel(val, rx_ring->tail);
 415 }
 416
 417 /**
 418  * i40e_get_vsi_stats_struct - Get System Network Statistics
 419  * @vsi: the VSI we care about
 420  *
 421  * Returns the address of the device statistics structure.
 422  * The statistics are actually updated from the service task.
 423  **/
 424 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
 425 {
 426         return &vsi->net_stats;
 427 }
 428
 429 /**
 430  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
 431  * @netdev: network interface device structure
 432  *
 433  * Returns the address of the device statistics structure.
 434  * The statistics are actually updated from the service task.
 435  **/
 436 #ifdef I40E_FCOE
 437 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 438                                              struct net_device *netdev,
 439                                              struct rtnl_link_stats64 *stats)
 440 #else
 441 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
 442                                              struct net_device *netdev,
 443                                              struct rtnl_link_stats64 *stats)
 444 #endif
 445 {
 446         struct i40e_netdev_priv *np = netdev_priv(netdev);
 447         struct i40e_ring *tx_ring, *rx_ring;
 448         struct i40e_vsi *vsi = np->vsi;
 449         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
 450         int i;
 451
 452         if (test_bit(__I40E_DOWN, &vsi->state))
 453                 return stats;
 454
 455         if (!vsi->tx_rings)
 456                 return stats;
 457
 458         rcu_read_lock();
 459         for (i = 0; i < vsi->num_queue_pairs; i++) {
 460                 u64 bytes, packets;
 461                 unsigned int start;
 462
 463                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
 464                 if (!tx_ring)
 465                         continue;
 466
 467                 do {
 468                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
 469                         packets = tx_ring->stats.packets;
 470                         bytes   = tx_ring->stats.bytes;
 471                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
 472
 473                 stats->tx_packets += packets;
 474                 stats->tx_bytes   += bytes;
 475                 rx_ring = &tx_ring[1];
 476
 477                 do {
 478                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
 479                         packets = rx_ring->stats.packets;
 480                         bytes   = rx_ring->stats.bytes;
 481                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
 482
 483                 stats->rx_packets += packets;
 484                 stats->rx_bytes   += bytes;
 485         }
 486         rcu_read_unlock();
 487
 488         /* following stats updated by i40e_watchdog_subtask() */
 489         stats->multicast        = vsi_stats->multicast;
 490         stats->tx_errors        = vsi_stats->tx_errors;
 491         stats->tx_dropped       = vsi_stats->tx_dropped;
 492         stats->rx_errors        = vsi_stats->rx_errors;
 493         stats->rx_dropped       = vsi_stats->rx_dropped;
 494         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
 495         stats->rx_length_errors = vsi_stats->rx_length_errors;
 496
 497         return stats;
 498 }
 499
 500 /**
 501  * i40e_vsi_reset_stats - Resets all stats of the given vsi
 502  * @vsi: the VSI to have its stats reset
 503  **/
 504 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
 505 {
 506         struct rtnl_link_stats64 *ns;
 507         int i;
 508
 509         if (!vsi)
 510                 return;
 511
 512         ns = i40e_get_vsi_stats_struct(vsi);
 513         memset(ns, 0, sizeof(*ns));
 514         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
 515         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
 516         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
 517         if (vsi->rx_rings && vsi->rx_rings[0]) {
 518                 for (i = 0; i < vsi->num_queue_pairs; i++) {
 519                         memset(&vsi->rx_rings[i]->stats, 0,
 520                                sizeof(vsi->rx_rings[i]->stats));
 521                         memset(&vsi->rx_rings[i]->rx_stats, 0,
 522                                sizeof(vsi->rx_rings[i]->rx_stats));
 523                         memset(&vsi->tx_rings[i]->stats, 0,
 524                                sizeof(vsi->tx_rings[i]->stats));
 525                         memset(&vsi->tx_rings[i]->tx_stats, 0,
 526                                sizeof(vsi->tx_rings[i]->tx_stats));
 527                 }
 528         }
 529         vsi->stat_offsets_loaded = false;
 530 }
 531
 532 /**
 533  * i40e_pf_reset_stats - Reset all of the stats for the given PF
 534  * @pf: the PF to be reset
 535  **/
 536 void i40e_pf_reset_stats(struct i40e_pf *pf)
 537 {
 538         int i;
 539
 540         memset(&pf->stats, 0, sizeof(pf->stats));
 541         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
 542         pf->stat_offsets_loaded = false;
 543
 544         for (i = 0; i < I40E_MAX_VEB; i++) {
 545                 if (pf->veb[i]) {
 546                         memset(&pf->veb[i]->stats, 0,
 547                                sizeof(pf->veb[i]->stats));
 548                         memset(&pf->veb[i]->stats_offsets, 0,
 549                                sizeof(pf->veb[i]->stats_offsets));
 550                         pf->veb[i]->stat_offsets_loaded = false;
 551                 }
 552         }
 553 }
 554
 555 /**
 556  * i40e_stat_update48 - read and update a 48 bit stat from the chip
 557  * @hw: ptr to the hardware info
 558  * @hireg: the high 32 bit reg to read
 559  * @loreg: the low 32 bit reg to read
 560  * @offset_loaded: has the initial offset been loaded yet
 561  * @offset: ptr to current offset value
 562  * @stat: ptr to the stat
 563  *
 564  * Since the device stats are not reset at PFReset, they likely will not
 565  * be zeroed when the driver starts.  We'll save the first values read
 566  * and use them as offsets to be subtracted from the raw values in order
 567  * to report stats that count from zero.  In the process, we also manage
 568  * the potential roll-over.
 569  **/
 570 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
 571                                bool offset_loaded, u64 *offset, u64 *stat)
 572 {
 573         u64 new_data;
 574
 575         if (hw->device_id == I40E_DEV_ID_QEMU) {
 576                 new_data = rd32(hw, loreg);
 577                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
 578         } else {
 579                 new_data = rd64(hw, loreg);
 580         }
 581         if (!offset_loaded)
 582                 *offset = new_data;
 583         if (likely(new_data >= *offset))
 584                 *stat = new_data - *offset;
 585         else
 586                 *stat = (new_data + BIT_ULL(48)) - *offset;
 587         *stat &= 0xFFFFFFFFFFFFULL;
 588 }
 589
 590 /**
 591  * i40e_stat_update32 - read and update a 32 bit stat from the chip
 592  * @hw: ptr to the hardware info
 593  * @reg: the hw reg to read
 594  * @offset_loaded: has the initial offset been loaded yet
 595  * @offset: ptr to current offset value
 596  * @stat: ptr to the stat
 597  **/
 598 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
 599                                bool offset_loaded, u64 *offset, u64 *stat)
 600 {
 601         u32 new_data;
 602
 603         new_data = rd32(hw, reg);
 604         if (!offset_loaded)
 605                 *offset = new_data;
 606         if (likely(new_data >= *offset))
 607                 *stat = (u32)(new_data - *offset);
 608         else
 609                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
 610 }
 611
 612 /**
 613  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
 614  * @vsi: the VSI to be updated
 615  **/
 616 void i40e_update_eth_stats(struct i40e_vsi *vsi)
 617 {
 618         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
 619         struct i40e_pf *pf = vsi->back;
 620         struct i40e_hw *hw = &pf->hw;
 621         struct i40e_eth_stats *oes;
 622         struct i40e_eth_stats *es;     /* device's eth stats */
 623
 624         es = &vsi->eth_stats;
 625         oes = &vsi->eth_stats_offsets;
 626
 627         /* Gather up the stats that the hw collects */
 628         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 629                            vsi->stat_offsets_loaded,
 630                            &oes->tx_errors, &es->tx_errors);
 631         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
 632                            vsi->stat_offsets_loaded,
 633                            &oes->rx_discards, &es->rx_discards);
 634         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
 635                            vsi->stat_offsets_loaded,
 636                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
 637         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
 638                            vsi->stat_offsets_loaded,
 639                            &oes->tx_errors, &es->tx_errors);
 640
 641         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
 642                            I40E_GLV_GORCL(stat_idx),
 643                            vsi->stat_offsets_loaded,
 644                            &oes->rx_bytes, &es->rx_bytes);
 645         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
 646                            I40E_GLV_UPRCL(stat_idx),
 647                            vsi->stat_offsets_loaded,
 648                            &oes->rx_unicast, &es->rx_unicast);
 649         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
 650                            I40E_GLV_MPRCL(stat_idx),
 651                            vsi->stat_offsets_loaded,
 652                            &oes->rx_multicast, &es->rx_multicast);
 653         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
 654                            I40E_GLV_BPRCL(stat_idx),
 655                            vsi->stat_offsets_loaded,
 656                            &oes->rx_broadcast, &es->rx_broadcast);
 657
 658         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
 659                            I40E_GLV_GOTCL(stat_idx),
 660                            vsi->stat_offsets_loaded,
 661                            &oes->tx_bytes, &es->tx_bytes);
 662         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
 663                            I40E_GLV_UPTCL(stat_idx),
 664                            vsi->stat_offsets_loaded,
 665                            &oes->tx_unicast, &es->tx_unicast);
 666         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
 667                            I40E_GLV_MPTCL(stat_idx),
 668                            vsi->stat_offsets_loaded,
 669                            &oes->tx_multicast, &es->tx_multicast);
 670         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
 671                            I40E_GLV_BPTCL(stat_idx),
 672                            vsi->stat_offsets_loaded,
 673                            &oes->tx_broadcast, &es->tx_broadcast);
 674         vsi->stat_offsets_loaded = true;
 675 }
 676
 677 /**
 678  * i40e_update_veb_stats - Update Switch component statistics
 679  * @veb: the VEB being updated
 680  **/
 681 static void i40e_update_veb_stats(struct i40e_veb *veb)
 682 {
 683         struct i40e_pf *pf = veb->pf;
 684         struct i40e_hw *hw = &pf->hw;
 685         struct i40e_eth_stats *oes;
 686         struct i40e_eth_stats *es;     /* device's eth stats */
 687         struct i40e_veb_tc_stats *veb_oes;
 688         struct i40e_veb_tc_stats *veb_es;
 689         int i, idx = 0;
 690
 691         idx = veb->stats_idx;
 692         es = &veb->stats;
 693         oes = &veb->stats_offsets;
 694         veb_es = &veb->tc_stats;
 695         veb_oes = &veb->tc_stats_offsets;
 696
 697         /* Gather up the stats that the hw collects */
 698         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
 699                            veb->stat_offsets_loaded,
 700                            &oes->tx_discards, &es->tx_discards);
 701         if (hw->revision_id > 0)
 702                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
 703                                    veb->stat_offsets_loaded,
 704                                    &oes->rx_unknown_protocol,
 705                                    &es->rx_unknown_protocol);
 706         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
 707                            veb->stat_offsets_loaded,
 708                            &oes->rx_bytes, &es->rx_bytes);
 709         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
 710                            veb->stat_offsets_loaded,
 711                            &oes->rx_unicast, &es->rx_unicast);
 712         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
 713                            veb->stat_offsets_loaded,
 714                            &oes->rx_multicast, &es->rx_multicast);
 715         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
 716                            veb->stat_offsets_loaded,
 717                            &oes->rx_broadcast, &es->rx_broadcast);
 718
 719         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
 720                            veb->stat_offsets_loaded,
 721                            &oes->tx_bytes, &es->tx_bytes);
 722         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
 723                            veb->stat_offsets_loaded,
 724                            &oes->tx_unicast, &es->tx_unicast);
 725         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
 726                            veb->stat_offsets_loaded,
 727                            &oes->tx_multicast, &es->tx_multicast);
 728         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
 729                            veb->stat_offsets_loaded,
 730                            &oes->tx_broadcast, &es->tx_broadcast);
 731         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
 732                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
 733                                    I40E_GLVEBTC_RPCL(i, idx),
 734                                    veb->stat_offsets_loaded,
 735                                    &veb_oes->tc_rx_packets[i],
 736                                    &veb_es->tc_rx_packets[i]);
 737                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
 738                                    I40E_GLVEBTC_RBCL(i, idx),
 739                                    veb->stat_offsets_loaded,
 740                                    &veb_oes->tc_rx_bytes[i],
 741                                    &veb_es->tc_rx_bytes[i]);
 742                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
 743                                    I40E_GLVEBTC_TPCL(i, idx),
 744                                    veb->stat_offsets_loaded,
 745                                    &veb_oes->tc_tx_packets[i],
 746                                    &veb_es->tc_tx_packets[i]);
 747                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
 748                                    I40E_GLVEBTC_TBCL(i, idx),
 749                                    veb->stat_offsets_loaded,
 750                                    &veb_oes->tc_tx_bytes[i],
 751                                    &veb_es->tc_tx_bytes[i]);
 752         }
 753         veb->stat_offsets_loaded = true;
 754 }
 755
 756 #ifdef I40E_FCOE
 757 /**
 758  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
 759  * @vsi: the VSI that is capable of doing FCoE
 760  **/
 761 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
 762 {
 763         struct i40e_pf *pf = vsi->back;
 764         struct i40e_hw *hw = &pf->hw;
 765         struct i40e_fcoe_stats *ofs;
 766         struct i40e_fcoe_stats *fs;     /* device's eth stats */
 767         int idx;
 768
 769         if (vsi->type != I40E_VSI_FCOE)
 770                 return;
 771
 772         idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
 773         fs = &vsi->fcoe_stats;
 774         ofs = &vsi->fcoe_stats_offsets;
 775
 776         i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
 777                            vsi->fcoe_stat_offsets_loaded,
 778                            &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
 779         i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
 780                            vsi->fcoe_stat_offsets_loaded,
 781                            &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
 782         i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
 783                            vsi->fcoe_stat_offsets_loaded,
 784                            &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
 785         i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
 786                            vsi->fcoe_stat_offsets_loaded,
 787                            &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
 788         i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
 789                            vsi->fcoe_stat_offsets_loaded,
 790                            &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
 791         i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
 792                            vsi->fcoe_stat_offsets_loaded,
 793                            &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
 794         i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
 795                            vsi->fcoe_stat_offsets_loaded,
 796                            &ofs->fcoe_last_error, &fs->fcoe_last_error);
 797         i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
 798                            vsi->fcoe_stat_offsets_loaded,
 799                            &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
 800
 801         vsi->fcoe_stat_offsets_loaded = true;
 802 }
 803
 804 #endif
 805 /**
 806  * i40e_update_vsi_stats - Update the vsi statistics counters.
 807  * @vsi: the VSI to be updated
 808  *
 809  * There are a few instances where we store the same stat in a
 810  * couple of different structs.  This is partly because we have
 811  * the netdev stats that need to be filled out, which is slightly
 812  * different from the "eth_stats" defined by the chip and used in
 813  * VF communications.  We sort it out here.
 814  **/
 815 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
 816 {
 817         struct i40e_pf *pf = vsi->back;
 818         struct rtnl_link_stats64 *ons;
 819         struct rtnl_link_stats64 *ns;   /* netdev stats */
 820         struct i40e_eth_stats *oes;
 821         struct i40e_eth_stats *es;     /* device's eth stats */
 822         u32 tx_restart, tx_busy;
 823         struct i40e_ring *p;
 824         u32 rx_page, rx_buf;
 825         u64 bytes, packets;
 826         unsigned int start;
 827         u64 tx_linearize;
 828         u64 tx_force_wb;
 829         u64 rx_p, rx_b;
 830         u64 tx_p, tx_b;
 831         u16 q;
 832
 833         if (test_bit(__I40E_DOWN, &vsi->state) ||
 834             test_bit(__I40E_CONFIG_BUSY, &pf->state))
 835                 return;
 836
 837         ns = i40e_get_vsi_stats_struct(vsi);
 838         ons = &vsi->net_stats_offsets;
 839         es = &vsi->eth_stats;
 840         oes = &vsi->eth_stats_offsets;
 841
 842         /* Gather up the netdev and vsi stats that the driver collects
 843          * on the fly during packet processing
 844          */
 845         rx_b = rx_p = 0;
 846         tx_b = tx_p = 0;
 847         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
 848         rx_page = 0;
 849         rx_buf = 0;
 850         rcu_read_lock();
 851         for (q = 0; q < vsi->num_queue_pairs; q++) {
 852                 /* locate Tx ring */
 853                 p = ACCESS_ONCE(vsi->tx_rings[q]);
 854
 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                 tx_b += bytes;
 861                 tx_p += packets;
 862                 tx_restart += p->tx_stats.restart_queue;
 863                 tx_busy += p->tx_stats.tx_busy;
 864                 tx_linearize += p->tx_stats.tx_linearize;
 865                 tx_force_wb += p->tx_stats.tx_force_wb;
 866
 867                 /* Rx queue is part of the same block as Tx queue */
 868                 p = &p[1];
 869                 do {
 870                         start = u64_stats_fetch_begin_irq(&p->syncp);
 871                         packets = p->stats.packets;
 872                         bytes = p->stats.bytes;
 873                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
 874                 rx_b += bytes;
 875                 rx_p += packets;
 876                 rx_buf += p->rx_stats.alloc_buff_failed;
 877                 rx_page += p->rx_stats.alloc_page_failed;
 878         }
 879         rcu_read_unlock();
 880         vsi->tx_restart = tx_restart;
 881         vsi->tx_busy = tx_busy;
 882         vsi->tx_linearize = tx_linearize;
 883         vsi->tx_force_wb = tx_force_wb;
 884         vsi->rx_page_failed = rx_page;
 885         vsi->rx_buf_failed = rx_buf;
 886
 887         ns->rx_packets = rx_p;
 888         ns->rx_bytes = rx_b;
 889         ns->tx_packets = tx_p;
 890         ns->tx_bytes = tx_b;
 891
 892         /* update netdev stats from eth stats */
 893         i40e_update_eth_stats(vsi);
 894         ons->tx_errors = oes->tx_errors;
 895         ns->tx_errors = es->tx_errors;
 896         ons->multicast = oes->rx_multicast;
 897         ns->multicast = es->rx_multicast;
 898         ons->rx_dropped = oes->rx_discards;
 899         ns->rx_dropped = es->rx_discards;
 900         ons->tx_dropped = oes->tx_discards;
 901         ns->tx_dropped = es->tx_discards;
 902
 903         /* pull in a couple PF stats if this is the main vsi */
 904         if (vsi == pf->vsi[pf->lan_vsi]) {
 905                 ns->rx_crc_errors = pf->stats.crc_errors;
 906                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
 907                 ns->rx_length_errors = pf->stats.rx_length_errors;
 908         }
 909 }
 910
 911 /**
 912  * i40e_update_pf_stats - Update the PF statistics counters.
 913  * @pf: the PF to be updated
 914  **/
 915 static void i40e_update_pf_stats(struct i40e_pf *pf)
 916 {
 917         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
 918         struct i40e_hw_port_stats *nsd = &pf->stats;
 919         struct i40e_hw *hw = &pf->hw;
 920         u32 val;
 921         int i;
 922
 923         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
 924                            I40E_GLPRT_GORCL(hw->port),
 925                            pf->stat_offsets_loaded,
 926                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
 927         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
 928                            I40E_GLPRT_GOTCL(hw->port),
 929                            pf->stat_offsets_loaded,
 930                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
 931         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
 932                            pf->stat_offsets_loaded,
 933                            &osd->eth.rx_discards,
 934                            &nsd->eth.rx_discards);
 935         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
 936                            I40E_GLPRT_UPRCL(hw->port),
 937                            pf->stat_offsets_loaded,
 938                            &osd->eth.rx_unicast,
 939                            &nsd->eth.rx_unicast);
 940         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
 941                            I40E_GLPRT_MPRCL(hw->port),
 942                            pf->stat_offsets_loaded,
 943                            &osd->eth.rx_multicast,
 944                            &nsd->eth.rx_multicast);
 945         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
 946                            I40E_GLPRT_BPRCL(hw->port),
 947                            pf->stat_offsets_loaded,
 948                            &osd->eth.rx_broadcast,
 949                            &nsd->eth.rx_broadcast);
 950         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
 951                            I40E_GLPRT_UPTCL(hw->port),
 952                            pf->stat_offsets_loaded,
 953                            &osd->eth.tx_unicast,
 954                            &nsd->eth.tx_unicast);
 955         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
 956                            I40E_GLPRT_MPTCL(hw->port),
 957                            pf->stat_offsets_loaded,
 958                            &osd->eth.tx_multicast,
 959                            &nsd->eth.tx_multicast);
 960         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
 961                            I40E_GLPRT_BPTCL(hw->port),
 962                            pf->stat_offsets_loaded,
 963                            &osd->eth.tx_broadcast,
 964                            &nsd->eth.tx_broadcast);
 965
 966         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
 967                            pf->stat_offsets_loaded,
 968                            &osd->tx_dropped_link_down,
 969                            &nsd->tx_dropped_link_down);
 970
 971         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
 972                            pf->stat_offsets_loaded,
 973                            &osd->crc_errors, &nsd->crc_errors);
 974
 975         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
 976                            pf->stat_offsets_loaded,
 977                            &osd->illegal_bytes, &nsd->illegal_bytes);
 978
 979         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
 980                            pf->stat_offsets_loaded,
 981                            &osd->mac_local_faults,
 982                            &nsd->mac_local_faults);
 983         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
 984                            pf->stat_offsets_loaded,
 985                            &osd->mac_remote_faults,
 986                            &nsd->mac_remote_faults);
 987
 988         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
 989                            pf->stat_offsets_loaded,
 990                            &osd->rx_length_errors,
 991                            &nsd->rx_length_errors);
 992
 993         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
 994                            pf->stat_offsets_loaded,
 995                            &osd->link_xon_rx, &nsd->link_xon_rx);
 996         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
 997                            pf->stat_offsets_loaded,
 998                            &osd->link_xon_tx, &nsd->link_xon_tx);
 999         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
1000                            pf->stat_offsets_loaded,
1001                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
1002         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
1003                            pf->stat_offsets_loaded,
1004                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
1005
1006         for (i = 0; i < 8; i++) {
1007                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
1008                                    pf->stat_offsets_loaded,
1009                                    &osd->priority_xoff_rx[i],
1010                                    &nsd->priority_xoff_rx[i]);
1011                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1012                                    pf->stat_offsets_loaded,
1013                                    &osd->priority_xon_rx[i],
1014                                    &nsd->priority_xon_rx[i]);
1015                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1016                                    pf->stat_offsets_loaded,
1017                                    &osd->priority_xon_tx[i],
1018                                    &nsd->priority_xon_tx[i]);
1019                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1020                                    pf->stat_offsets_loaded,
1021                                    &osd->priority_xoff_tx[i],
1022                                    &nsd->priority_xoff_tx[i]);
1023                 i40e_stat_update32(hw,
1024                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1025                                    pf->stat_offsets_loaded,
1026                                    &osd->priority_xon_2_xoff[i],
1027                                    &nsd->priority_xon_2_xoff[i]);
1028         }
1029
1030         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1031                            I40E_GLPRT_PRC64L(hw->port),
1032                            pf->stat_offsets_loaded,
1033                            &osd->rx_size_64, &nsd->rx_size_64);
1034         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1035                            I40E_GLPRT_PRC127L(hw->port),
1036                            pf->stat_offsets_loaded,
1037                            &osd->rx_size_127, &nsd->rx_size_127);
1038         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1039                            I40E_GLPRT_PRC255L(hw->port),
1040                            pf->stat_offsets_loaded,
1041                            &osd->rx_size_255, &nsd->rx_size_255);
1042         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1043                            I40E_GLPRT_PRC511L(hw->port),
1044                            pf->stat_offsets_loaded,
1045                            &osd->rx_size_511, &nsd->rx_size_511);
1046         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1047                            I40E_GLPRT_PRC1023L(hw->port),
1048                            pf->stat_offsets_loaded,
1049                            &osd->rx_size_1023, &nsd->rx_size_1023);
1050         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1051                            I40E_GLPRT_PRC1522L(hw->port),
1052                            pf->stat_offsets_loaded,
1053                            &osd->rx_size_1522, &nsd->rx_size_1522);
1054         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1055                            I40E_GLPRT_PRC9522L(hw->port),
1056                            pf->stat_offsets_loaded,
1057                            &osd->rx_size_big, &nsd->rx_size_big);
1058
1059         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1060                            I40E_GLPRT_PTC64L(hw->port),
1061                            pf->stat_offsets_loaded,
1062                            &osd->tx_size_64, &nsd->tx_size_64);
1063         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1064                            I40E_GLPRT_PTC127L(hw->port),
1065                            pf->stat_offsets_loaded,
1066                            &osd->tx_size_127, &nsd->tx_size_127);
1067         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1068                            I40E_GLPRT_PTC255L(hw->port),
1069                            pf->stat_offsets_loaded,
1070                            &osd->tx_size_255, &nsd->tx_size_255);
1071         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1072                            I40E_GLPRT_PTC511L(hw->port),
1073                            pf->stat_offsets_loaded,
1074                            &osd->tx_size_511, &nsd->tx_size_511);
1075         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1076                            I40E_GLPRT_PTC1023L(hw->port),
1077                            pf->stat_offsets_loaded,
1078                            &osd->tx_size_1023, &nsd->tx_size_1023);
1079         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1080                            I40E_GLPRT_PTC1522L(hw->port),
1081                            pf->stat_offsets_loaded,
1082                            &osd->tx_size_1522, &nsd->tx_size_1522);
1083         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1084                            I40E_GLPRT_PTC9522L(hw->port),
1085                            pf->stat_offsets_loaded,
1086                            &osd->tx_size_big, &nsd->tx_size_big);
1087
1088         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1089                            pf->stat_offsets_loaded,
1090                            &osd->rx_undersize, &nsd->rx_undersize);
1091         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1092                            pf->stat_offsets_loaded,
1093                            &osd->rx_fragments, &nsd->rx_fragments);
1094         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1095                            pf->stat_offsets_loaded,
1096                            &osd->rx_oversize, &nsd->rx_oversize);
1097         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1098                            pf->stat_offsets_loaded,
1099                            &osd->rx_jabber, &nsd->rx_jabber);
1100
1101         /* FDIR stats */
1102         i40e_stat_update32(hw,
1103                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1104                            pf->stat_offsets_loaded,
1105                            &osd->fd_atr_match, &nsd->fd_atr_match);
1106         i40e_stat_update32(hw,
1107                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1108                            pf->stat_offsets_loaded,
1109                            &osd->fd_sb_match, &nsd->fd_sb_match);
1110         i40e_stat_update32(hw,
1111                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1112                       pf->stat_offsets_loaded,
1113                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1114
1115         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1116         nsd->tx_lpi_status =
1117                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1118                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1119         nsd->rx_lpi_status =
1120                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1121                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1122         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1123                            pf->stat_offsets_loaded,
1124                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1125         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1126                            pf->stat_offsets_loaded,
1127                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1128
1129         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1130             !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1131                 nsd->fd_sb_status = true;
1132         else
1133                 nsd->fd_sb_status = false;
1134
1135         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1136             !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1137                 nsd->fd_atr_status = true;
1138         else
1139                 nsd->fd_atr_status = false;
1140
1141         pf->stat_offsets_loaded = true;
1142 }
1143
1144 /**
1145  * i40e_update_stats - Update the various statistics counters.
1146  * @vsi: the VSI to be updated
1147  *
1148  * Update the various stats for this VSI and its related entities.
1149  **/
1150 void i40e_update_stats(struct i40e_vsi *vsi)
1151 {
1152         struct i40e_pf *pf = vsi->back;
1153
1154         if (vsi == pf->vsi[pf->lan_vsi])
1155                 i40e_update_pf_stats(pf);
1156
1157         i40e_update_vsi_stats(vsi);
1158 #ifdef I40E_FCOE
1159         i40e_update_fcoe_stats(vsi);
1160 #endif
1161 }
1162
1163 /**
1164  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1165  * @vsi: the VSI to be searched
1166  * @macaddr: the MAC address
1167  * @vlan: the vlan
1168  * @is_vf: make sure its a VF filter, else doesn't matter
1169  * @is_netdev: make sure its a netdev filter, else doesn't matter
1170  *
1171  * Returns ptr to the filter object or NULL
1172  **/
1173 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1174                                                 u8 *macaddr, s16 vlan,
1175                                                 bool is_vf, bool is_netdev)
1176 {
1177         struct i40e_mac_filter *f;
1178
1179         if (!vsi || !macaddr)
1180                 return NULL;
1181
1182         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1183                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1184                     (vlan == f->vlan)    &&
1185                     (!is_vf || f->is_vf) &&
1186                     (!is_netdev || f->is_netdev))
1187                         return f;
1188         }
1189         return NULL;
1190 }
1191
1192 /**
1193  * i40e_find_mac - Find a mac addr in the macvlan filters list
1194  * @vsi: the VSI to be searched
1195  * @macaddr: the MAC address we are searching for
1196  * @is_vf: make sure its a VF filter, else doesn't matter
1197  * @is_netdev: make sure its a netdev filter, else doesn't matter
1198  *
1199  * Returns the first filter with the provided MAC address or NULL if
1200  * MAC address was not found
1201  **/
1202 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1203                                       bool is_vf, bool is_netdev)
1204 {
1205         struct i40e_mac_filter *f;
1206
1207         if (!vsi || !macaddr)
1208                 return NULL;
1209
1210         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1211                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1212                     (!is_vf || f->is_vf) &&
1213                     (!is_netdev || f->is_netdev))
1214                         return f;
1215         }
1216         return NULL;
1217 }
1218
1219 /**
1220  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1221  * @vsi: the VSI to be searched
1222  *
1223  * Returns true if VSI is in vlan mode or false otherwise
1224  **/
1225 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1226 {
1227         struct i40e_mac_filter *f;
1228
1229         /* Only -1 for all the filters denotes not in vlan mode
1230          * so we have to go through all the list in order to make sure
1231          */
1232         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1233                 if (f->vlan >= 0 || vsi->info.pvid)
1234                         return true;
1235         }
1236
1237         return false;
1238 }
1239
1240 /**
1241  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1242  * @vsi: the VSI to be searched
1243  * @macaddr: the mac address to be filtered
1244  * @is_vf: true if it is a VF
1245  * @is_netdev: true if it is a netdev
1246  *
1247  * Goes through all the macvlan filters and adds a
1248  * macvlan filter for each unique vlan that already exists
1249  *
1250  * Returns first filter found on success, else NULL
1251  **/
1252 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1253                                              bool is_vf, bool is_netdev)
1254 {
1255         struct i40e_mac_filter *f;
1256
1257         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1258                 if (vsi->info.pvid)
1259                         f->vlan = le16_to_cpu(vsi->info.pvid);
1260                 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1261                                       is_vf, is_netdev)) {
1262                         if (!i40e_add_filter(vsi, macaddr, f->vlan,
1263                                              is_vf, is_netdev))
1264                                 return NULL;
1265                 }
1266         }
1267
1268         return list_first_entry_or_null(&vsi->mac_filter_list,
1269                                         struct i40e_mac_filter, list);
1270 }
1271
1272 /**
1273  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1274  * @vsi: the VSI to be searched
1275  * @macaddr: the mac address to be removed
1276  * @is_vf: true if it is a VF
1277  * @is_netdev: true if it is a netdev
1278  *
1279  * Removes a given MAC address from a VSI, regardless of VLAN
1280  *
1281  * Returns 0 for success, or error
1282  **/
1283 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1284                           bool is_vf, bool is_netdev)
1285 {
1286         struct i40e_mac_filter *f = NULL;
1287         int changed = 0;
1288
1289         WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1290              "Missing mac_filter_list_lock\n");
1291         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1292                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1293                     (is_vf == f->is_vf) &&
1294                     (is_netdev == f->is_netdev)) {
1295                         f->counter--;
1296                         f->changed = true;
1297                         changed = 1;
1298                 }
1299         }
1300         if (changed) {
1301                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1302                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1303                 return 0;
1304         }
1305         return -ENOENT;
1306 }
1307
1308 /**
1309  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1310  * @vsi: the PF Main VSI - inappropriate for any other VSI
1311  * @macaddr: the MAC address
1312  *
1313  * Some older firmware configurations set up a default promiscuous VLAN
1314  * filter that needs to be removed.
1315  **/
1316 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1317 {
1318         struct i40e_aqc_remove_macvlan_element_data element;
1319         struct i40e_pf *pf = vsi->back;
1320         i40e_status ret;
1321
1322         /* Only appropriate for the PF main VSI */
1323         if (vsi->type != I40E_VSI_MAIN)
1324                 return -EINVAL;
1325
1326         memset(&element, 0, sizeof(element));
1327         ether_addr_copy(element.mac_addr, macaddr);
1328         element.vlan_tag = 0;
1329         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1330                         I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1331         ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1332         if (ret)
1333                 return -ENOENT;
1334
1335         return 0;
1336 }
1337
1338 /**
1339  * i40e_add_filter - Add a mac/vlan filter to the VSI
1340  * @vsi: the VSI to be searched
1341  * @macaddr: the MAC address
1342  * @vlan: the vlan
1343  * @is_vf: make sure its a VF filter, else doesn't matter
1344  * @is_netdev: make sure its a netdev filter, else doesn't matter
1345  *
1346  * Returns ptr to the filter object or NULL when no memory available.
1347  *
1348  * NOTE: This function is expected to be called with mac_filter_list_lock
1349  * being held.
1350  **/
1351 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1352                                         u8 *macaddr, s16 vlan,
1353                                         bool is_vf, bool is_netdev)
1354 {
1355         struct i40e_mac_filter *f;
1356
1357         if (!vsi || !macaddr)
1358                 return NULL;
1359
1360         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1361         if (!f) {
1362                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1363                 if (!f)
1364                         goto add_filter_out;
1365
1366                 ether_addr_copy(f->macaddr, macaddr);
1367                 f->vlan = vlan;
1368                 f->changed = true;
1369
1370                 INIT_LIST_HEAD(&f->list);
1371                 list_add(&f->list, &vsi->mac_filter_list);
1372         }
1373
1374         /* increment counter and add a new flag if needed */
1375         if (is_vf) {
1376                 if (!f->is_vf) {
1377                         f->is_vf = true;
1378                         f->counter++;
1379                 }
1380         } else if (is_netdev) {
1381                 if (!f->is_netdev) {
1382                         f->is_netdev = true;
1383                         f->counter++;
1384                 }
1385         } else {
1386                 f->counter++;
1387         }
1388
1389         /* changed tells sync_filters_subtask to
1390          * push the filter down to the firmware
1391          */
1392         if (f->changed) {
1393                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1394                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1395         }
1396
1397 add_filter_out:
1398         return f;
1399 }
1400
1401 /**
1402  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1403  * @vsi: the VSI to be searched
1404  * @macaddr: the MAC address
1405  * @vlan: the vlan
1406  * @is_vf: make sure it's a VF filter, else doesn't matter
1407  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1408  *
1409  * NOTE: This function is expected to be called with mac_filter_list_lock
1410  * being held.
1411  **/
1412 void i40e_del_filter(struct i40e_vsi *vsi,
1413                      u8 *macaddr, s16 vlan,
1414                      bool is_vf, bool is_netdev)
1415 {
1416         struct i40e_mac_filter *f;
1417
1418         if (!vsi || !macaddr)
1419                 return;
1420
1421         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1422         if (!f || f->counter == 0)
1423                 return;
1424
1425         if (is_vf) {
1426                 if (f->is_vf) {
1427                         f->is_vf = false;
1428                         f->counter--;
1429                 }
1430         } else if (is_netdev) {
1431                 if (f->is_netdev) {
1432                         f->is_netdev = false;
1433                         f->counter--;
1434                 }
1435         } else {
1436                 /* make sure we don't remove a filter in use by VF or netdev */
1437                 int min_f = 0;
1438
1439                 min_f += (f->is_vf ? 1 : 0);
1440                 min_f += (f->is_netdev ? 1 : 0);
1441
1442                 if (f->counter > min_f)
1443                         f->counter--;
1444         }
1445
1446         /* counter == 0 tells sync_filters_subtask to
1447          * remove the filter from the firmware's list
1448          */
1449         if (f->counter == 0) {
1450                 f->changed = true;
1451                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1452                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1453         }
1454 }
1455
1456 /**
1457  * i40e_set_mac - NDO callback to set mac address
1458  * @netdev: network interface device structure
1459  * @p: pointer to an address structure
1460  *
1461  * Returns 0 on success, negative on failure
1462  **/
1463 #ifdef I40E_FCOE
1464 int i40e_set_mac(struct net_device *netdev, void *p)
1465 #else
1466 static int i40e_set_mac(struct net_device *netdev, void *p)
1467 #endif
1468 {
1469         struct i40e_netdev_priv *np = netdev_priv(netdev);
1470         struct i40e_vsi *vsi = np->vsi;
1471         struct i40e_pf *pf = vsi->back;
1472         struct i40e_hw *hw = &pf->hw;
1473         struct sockaddr *addr = p;
1474         struct i40e_mac_filter *f;
1475
1476         if (!is_valid_ether_addr(addr->sa_data))
1477                 return -EADDRNOTAVAIL;
1478
1479         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1480                 netdev_info(netdev, "already using mac address %pM\n",
1481                             addr->sa_data);
1482                 return 0;
1483         }
1484
1485         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1486             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1487                 return -EADDRNOTAVAIL;
1488
1489         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1490                 netdev_info(netdev, "returning to hw mac address %pM\n",
1491                             hw->mac.addr);
1492         else
1493                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1494
1495         if (vsi->type == I40E_VSI_MAIN) {
1496                 i40e_status ret;
1497
1498                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1499                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1500                                                 addr->sa_data, NULL);
1501                 if (ret) {
1502                         netdev_info(netdev,
1503                                     "Addr change for Main VSI failed: %d\n",
1504                                     ret);
1505                         return -EADDRNOTAVAIL;
1506                 }
1507         }
1508
1509         if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1510                 struct i40e_aqc_remove_macvlan_element_data element;
1511
1512                 memset(&element, 0, sizeof(element));
1513                 ether_addr_copy(element.mac_addr, netdev->dev_addr);
1514                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1515                 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1516         } else {
1517                 spin_lock_bh(&vsi->mac_filter_list_lock);
1518                 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1519                                 false, false);
1520                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1521         }
1522
1523         if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1524                 struct i40e_aqc_add_macvlan_element_data element;
1525
1526                 memset(&element, 0, sizeof(element));
1527                 ether_addr_copy(element.mac_addr, hw->mac.addr);
1528                 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1529                 i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1530         } else {
1531                 spin_lock_bh(&vsi->mac_filter_list_lock);
1532                 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1533                                     false, false);
1534                 if (f)
1535                         f->is_laa = true;
1536                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1537         }
1538
1539         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1540
1541         return i40e_sync_vsi_filters(vsi);
1542 }
1543
1544 /**
1545  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1546  * @vsi: the VSI being setup
1547  * @ctxt: VSI context structure
1548  * @enabled_tc: Enabled TCs bitmap
1549  * @is_add: True if called before Add VSI
1550  *
1551  * Setup VSI queue mapping for enabled traffic classes.
1552  **/
1553 #ifdef I40E_FCOE
1554 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1555                               struct i40e_vsi_context *ctxt,
1556                               u8 enabled_tc,
1557                               bool is_add)
1558 #else
1559 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1560                                      struct i40e_vsi_context *ctxt,
1561                                      u8 enabled_tc,
1562                                      bool is_add)
1563 #endif
1564 {
1565         struct i40e_pf *pf = vsi->back;
1566         u16 sections = 0;
1567         u8 netdev_tc = 0;
1568         u16 numtc = 0;
1569         u16 qcount;
1570         u8 offset;
1571         u16 qmap;
1572         int i;
1573         u16 num_tc_qps = 0;
1574
1575         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1576         offset = 0;
1577
1578         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1579                 /* Find numtc from enabled TC bitmap */
1580                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1581                         if (enabled_tc & BIT(i)) /* TC is enabled */
1582                                 numtc++;
1583                 }
1584                 if (!numtc) {
1585                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1586                         numtc = 1;
1587                 }
1588         } else {
1589                 /* At least TC0 is enabled in case of non-DCB case */
1590                 numtc = 1;
1591         }
1592
1593         vsi->tc_config.numtc = numtc;
1594         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1595         /* Number of queues per enabled TC */
1596         /* In MFP case we can have a much lower count of MSIx
1597          * vectors available and so we need to lower the used
1598          * q count.
1599          */
1600         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1601                 qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1602         else
1603                 qcount = vsi->alloc_queue_pairs;
1604         num_tc_qps = qcount / numtc;
1605         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1606
1607         /* Setup queue offset/count for all TCs for given VSI */
1608         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1609                 /* See if the given TC is enabled for the given VSI */
1610                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1611                         /* TC is enabled */
1612                         int pow, num_qps;
1613
1614                         switch (vsi->type) {
1615                         case I40E_VSI_MAIN:
1616                                 qcount = min_t(int, pf->alloc_rss_size,
1617                                                num_tc_qps);
1618                                 break;
1619 #ifdef I40E_FCOE
1620                         case I40E_VSI_FCOE:
1621                                 qcount = num_tc_qps;
1622                                 break;
1623 #endif
1624                         case I40E_VSI_FDIR:
1625                         case I40E_VSI_SRIOV:
1626                         case I40E_VSI_VMDQ2:
1627                         default:
1628                                 qcount = num_tc_qps;
1629                                 WARN_ON(i != 0);
1630                                 break;
1631                         }
1632                         vsi->tc_config.tc_info[i].qoffset = offset;
1633                         vsi->tc_config.tc_info[i].qcount = qcount;
1634
1635                         /* find the next higher power-of-2 of num queue pairs */
1636                         num_qps = qcount;
1637                         pow = 0;
1638                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1639                                 pow++;
1640                                 num_qps >>= 1;
1641                         }
1642
1643                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1644                         qmap =
1645                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1646                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1647
1648                         offset += qcount;
1649                 } else {
1650                         /* TC is not enabled so set the offset to
1651                          * default queue and allocate one queue
1652                          * for the given TC.
1653                          */
1654                         vsi->tc_config.tc_info[i].qoffset = 0;
1655                         vsi->tc_config.tc_info[i].qcount = 1;
1656                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1657
1658                         qmap = 0;
1659                 }
1660                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1661         }
1662
1663         /* Set actual Tx/Rx queue pairs */
1664         vsi->num_queue_pairs = offset;
1665         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1666                 if (vsi->req_queue_pairs > 0)
1667                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1668                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1669                         vsi->num_queue_pairs = pf->num_lan_msix;
1670         }
1671
1672         /* Scheduler section valid can only be set for ADD VSI */
1673         if (is_add) {
1674                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1675
1676                 ctxt->info.up_enable_bits = enabled_tc;
1677         }
1678         if (vsi->type == I40E_VSI_SRIOV) {
1679                 ctxt->info.mapping_flags |=
1680                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1681                 for (i = 0; i < vsi->num_queue_pairs; i++)
1682                         ctxt->info.queue_mapping[i] =
1683                                                cpu_to_le16(vsi->base_queue + i);
1684         } else {
1685                 ctxt->info.mapping_flags |=
1686                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1687                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1688         }
1689         ctxt->info.valid_sections |= cpu_to_le16(sections);
1690 }
1691
1692 /**
1693  * i40e_set_rx_mode - NDO callback to set the netdev filters
1694  * @netdev: network interface device structure
1695  **/
1696 #ifdef I40E_FCOE
1697 void i40e_set_rx_mode(struct net_device *netdev)
1698 #else
1699 static void i40e_set_rx_mode(struct net_device *netdev)
1700 #endif
1701 {
1702         struct i40e_netdev_priv *np = netdev_priv(netdev);
1703         struct i40e_mac_filter *f, *ftmp;
1704         struct i40e_vsi *vsi = np->vsi;
1705         struct netdev_hw_addr *uca;
1706         struct netdev_hw_addr *mca;
1707         struct netdev_hw_addr *ha;
1708
1709         spin_lock_bh(&vsi->mac_filter_list_lock);
1710
1711         /* add addr if not already in the filter list */
1712         netdev_for_each_uc_addr(uca, netdev) {
1713                 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1714                         if (i40e_is_vsi_in_vlan(vsi))
1715                                 i40e_put_mac_in_vlan(vsi, uca->addr,
1716                                                      false, true);
1717                         else
1718                                 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1719                                                 false, true);
1720                 }
1721         }
1722
1723         netdev_for_each_mc_addr(mca, netdev) {
1724                 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1725                         if (i40e_is_vsi_in_vlan(vsi))
1726                                 i40e_put_mac_in_vlan(vsi, mca->addr,
1727                                                      false, true);
1728                         else
1729                                 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1730                                                 false, true);
1731                 }
1732         }
1733
1734         /* remove filter if not in netdev list */
1735         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1736
1737                 if (!f->is_netdev)
1738                         continue;
1739
1740                 netdev_for_each_mc_addr(mca, netdev)
1741                         if (ether_addr_equal(mca->addr, f->macaddr))
1742                                 goto bottom_of_search_loop;
1743
1744                 netdev_for_each_uc_addr(uca, netdev)
1745                         if (ether_addr_equal(uca->addr, f->macaddr))
1746                                 goto bottom_of_search_loop;
1747
1748                 for_each_dev_addr(netdev, ha)
1749                         if (ether_addr_equal(ha->addr, f->macaddr))
1750                                 goto bottom_of_search_loop;
1751
1752                 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1753                 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1754
1755 bottom_of_search_loop:
1756                 continue;
1757         }
1758         spin_unlock_bh(&vsi->mac_filter_list_lock);
1759
1760         /* check for other flag changes */
1761         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1762                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1763                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1764         }
1765 }
1766
1767 /**
1768  * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1769  * @src: source MAC filter entry to be clones
1770  *
1771  * Returns the pointer to newly cloned MAC filter entry or NULL
1772  * in case of error
1773  **/
1774 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1775                                         struct i40e_mac_filter *src)
1776 {
1777         struct i40e_mac_filter *f;
1778
1779         f = kzalloc(sizeof(*f), GFP_ATOMIC);
1780         if (!f)
1781                 return NULL;
1782         *f = *src;
1783
1784         INIT_LIST_HEAD(&f->list);
1785
1786         return f;
1787 }
1788
1789 /**
1790  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1791  * @vsi: pointer to vsi struct
1792  * @from: Pointer to list which contains MAC filter entries - changes to
1793  *        those entries needs to be undone.
1794  *
1795  * MAC filter entries from list were slated to be removed from device.
1796  **/
1797 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1798                                          struct list_head *from)
1799 {
1800         struct i40e_mac_filter *f, *ftmp;
1801
1802         list_for_each_entry_safe(f, ftmp, from, list) {
1803                 f->changed = true;
1804                 /* Move the element back into MAC filter list*/
1805                 list_move_tail(&f->list, &vsi->mac_filter_list);
1806         }
1807 }
1808
1809 /**
1810  * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1811  * @vsi: pointer to vsi struct
1812  *
1813  * MAC filter entries from list were slated to be added from device.
1814  **/
1815 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1816 {
1817         struct i40e_mac_filter *f, *ftmp;
1818
1819         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1820                 if (!f->changed && f->counter)
1821                         f->changed = true;
1822         }
1823 }
1824
1825 /**
1826  * i40e_cleanup_add_list - Deletes the element from add list and release
1827  *                      memory
1828  * @add_list: Pointer to list which contains MAC filter entries
1829  **/
1830 static void i40e_cleanup_add_list(struct list_head *add_list)
1831 {
1832         struct i40e_mac_filter *f, *ftmp;
1833
1834         list_for_each_entry_safe(f, ftmp, add_list, list) {
1835                 list_del(&f->list);
1836                 kfree(f);
1837         }
1838 }
1839
1840 /**
1841  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1842  * @vsi: ptr to the VSI
1843  *
1844  * Push any outstanding VSI filter changes through the AdminQ.
1845  *
1846  * Returns 0 or error value
1847  **/
1848 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1849 {
1850         struct list_head tmp_del_list, tmp_add_list;
1851         struct i40e_mac_filter *f, *ftmp, *fclone;
1852         bool promisc_forced_on = false;
1853         bool add_happened = false;
1854         int filter_list_len = 0;
1855         u32 changed_flags = 0;
1856         i40e_status aq_ret = 0;
1857         bool err_cond = false;
1858         int retval = 0;
1859         struct i40e_pf *pf;
1860         int num_add = 0;
1861         int num_del = 0;
1862         int aq_err = 0;
1863         u16 cmd_flags;
1864
1865         /* empty array typed pointers, kcalloc later */
1866         struct i40e_aqc_add_macvlan_element_data *add_list;
1867         struct i40e_aqc_remove_macvlan_element_data *del_list;
1868
1869         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1870                 usleep_range(1000, 2000);
1871         pf = vsi->back;
1872
1873         if (vsi->netdev) {
1874                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1875                 vsi->current_netdev_flags = vsi->netdev->flags;
1876         }
1877
1878         INIT_LIST_HEAD(&tmp_del_list);
1879         INIT_LIST_HEAD(&tmp_add_list);
1880
1881         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1882                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1883
1884                 spin_lock_bh(&vsi->mac_filter_list_lock);
1885                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1886                         if (!f->changed)
1887                                 continue;
1888
1889                         if (f->counter != 0)
1890                                 continue;
1891                         f->changed = false;
1892
1893                         /* Move the element into temporary del_list */
1894                         list_move_tail(&f->list, &tmp_del_list);
1895                 }
1896
1897                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1898                         if (!f->changed)
1899                                 continue;
1900
1901                         if (f->counter == 0)
1902                                 continue;
1903                         f->changed = false;
1904
1905                         /* Clone MAC filter entry and add into temporary list */
1906                         fclone = i40e_mac_filter_entry_clone(f);
1907                         if (!fclone) {
1908                                 err_cond = true;
1909                                 break;
1910                         }
1911                         list_add_tail(&fclone->list, &tmp_add_list);
1912                 }
1913
1914                 /* if failed to clone MAC filter entry - undo */
1915                 if (err_cond) {
1916                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1917                         i40e_undo_add_filter_entries(vsi);
1918                 }
1919                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1920
1921                 if (err_cond) {
1922                         i40e_cleanup_add_list(&tmp_add_list);
1923                         retval = -ENOMEM;
1924                         goto out;
1925                 }
1926         }
1927
1928         /* Now process 'del_list' outside the lock */
1929         if (!list_empty(&tmp_del_list)) {
1930                 int del_list_size;
1931
1932                 filter_list_len = pf->hw.aq.asq_buf_size /
1933                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1934                 del_list_size = filter_list_len *
1935                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1936                 del_list = kzalloc(del_list_size, GFP_KERNEL);
1937                 if (!del_list) {
1938                         i40e_cleanup_add_list(&tmp_add_list);
1939
1940                         /* Undo VSI's MAC filter entry element updates */
1941                         spin_lock_bh(&vsi->mac_filter_list_lock);
1942                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1943                         i40e_undo_add_filter_entries(vsi);
1944                         spin_unlock_bh(&vsi->mac_filter_list_lock);
1945                         retval = -ENOMEM;
1946                         goto out;
1947                 }
1948
1949                 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1950                         cmd_flags = 0;
1951
1952                         /* add to delete list */
1953                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1954                         del_list[num_del].vlan_tag =
1955                                 cpu_to_le16((u16)(f->vlan ==
1956                                             I40E_VLAN_ANY ? 0 : f->vlan));
1957
1958                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1959                         del_list[num_del].flags = cmd_flags;
1960                         num_del++;
1961
1962                         /* flush a full buffer */
1963                         if (num_del == filter_list_len) {
1964                                 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1965                                                                 vsi->seid,
1966                                                                 del_list,
1967                                                                 num_del,
1968                                                                 NULL);
1969                                 aq_err = pf->hw.aq.asq_last_status;
1970                                 num_del = 0;
1971                                 memset(del_list, 0, del_list_size);
1972
1973                                 if (aq_ret && aq_err != I40E_AQ_RC_ENOENT) {
1974                                         retval = -EIO;
1975                                         dev_err(&pf->pdev->dev,
1976                                                 "ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1977                                                 i40e_stat_str(&pf->hw, aq_ret),
1978                                                 i40e_aq_str(&pf->hw, aq_err));
1979                                 }
1980                         }
1981                         /* Release memory for MAC filter entries which were
1982                          * synced up with HW.
1983                          */
1984                         list_del(&f->list);
1985                         kfree(f);
1986                 }
1987
1988                 if (num_del) {
1989                         aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1990                                                         del_list, num_del,
1991                                                         NULL);
1992                         aq_err = pf->hw.aq.asq_last_status;
1993                         num_del = 0;
1994
1995                         if (aq_ret && aq_err != I40E_AQ_RC_ENOENT)
1996                                 dev_info(&pf->pdev->dev,
1997                                          "ignoring delete macvlan error, err %s aq_err %s\n",
1998                                          i40e_stat_str(&pf->hw, aq_ret),
1999                                          i40e_aq_str(&pf->hw, aq_err));
2000                 }
2001
2002                 kfree(del_list);
2003                 del_list = NULL;
2004         }
2005
2006         if (!list_empty(&tmp_add_list)) {
2007                 int add_list_size;
2008
2009                 /* do all the adds now */
2010                 filter_list_len = pf->hw.aq.asq_buf_size /
2011                                sizeof(struct i40e_aqc_add_macvlan_element_data),
2012                 add_list_size = filter_list_len *
2013                                sizeof(struct i40e_aqc_add_macvlan_element_data);
2014                 add_list = kzalloc(add_list_size, GFP_KERNEL);
2015                 if (!add_list) {
2016                         /* Purge element from temporary lists */
2017                         i40e_cleanup_add_list(&tmp_add_list);
2018
2019                         /* Undo add filter entries from VSI MAC filter list */
2020                         spin_lock_bh(&vsi->mac_filter_list_lock);
2021                         i40e_undo_add_filter_entries(vsi);
2022                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2023                         retval = -ENOMEM;
2024                         goto out;
2025                 }
2026
2027                 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2028
2029                         add_happened = true;
2030                         cmd_flags = 0;
2031
2032                         /* add to add array */
2033                         ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2034                         add_list[num_add].vlan_tag =
2035                                 cpu_to_le16(
2036                                  (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2037                         add_list[num_add].queue_number = 0;
2038
2039                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2040                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2041                         num_add++;
2042
2043                         /* flush a full buffer */
2044                         if (num_add == filter_list_len) {
2045                                 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2046                                                              add_list, num_add,
2047                                                              NULL);
2048                                 aq_err = pf->hw.aq.asq_last_status;
2049                                 num_add = 0;
2050
2051                                 if (aq_ret)
2052                                         break;
2053                                 memset(add_list, 0, add_list_size);
2054                         }
2055                         /* Entries from tmp_add_list were cloned from MAC
2056                          * filter list, hence clean those cloned entries
2057                          */
2058                         list_del(&f->list);
2059                         kfree(f);
2060                 }
2061
2062                 if (num_add) {
2063                         aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2064                                                      add_list, num_add, NULL);
2065                         aq_err = pf->hw.aq.asq_last_status;
2066                         num_add = 0;
2067                 }
2068                 kfree(add_list);
2069                 add_list = NULL;
2070
2071                 if (add_happened && aq_ret && aq_err != I40E_AQ_RC_EINVAL) {
2072                         retval = i40e_aq_rc_to_posix(aq_ret, aq_err);
2073                         dev_info(&pf->pdev->dev,
2074                                  "add filter failed, err %s aq_err %s\n",
2075                                  i40e_stat_str(&pf->hw, aq_ret),
2076                                  i40e_aq_str(&pf->hw, aq_err));
2077                         if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2078                             !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2079                                       &vsi->state)) {
2080                                 promisc_forced_on = true;
2081                                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2082                                         &vsi->state);
2083                                 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2084                         }
2085                 }
2086         }
2087
2088         /* check for changes in promiscuous modes */
2089         if (changed_flags & IFF_ALLMULTI) {
2090                 bool cur_multipromisc;
2091
2092                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2093                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2094                                                                vsi->seid,
2095                                                                cur_multipromisc,
2096                                                                NULL);
2097                 if (aq_ret) {
2098                         retval = i40e_aq_rc_to_posix(aq_ret,
2099                                                      pf->hw.aq.asq_last_status);
2100                         dev_info(&pf->pdev->dev,
2101                                  "set multi promisc failed, err %s aq_err %s\n",
2102                                  i40e_stat_str(&pf->hw, aq_ret),
2103                                  i40e_aq_str(&pf->hw,
2104                                              pf->hw.aq.asq_last_status));
2105                 }
2106         }
2107         if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2108                 bool cur_promisc;
2109
2110                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2111                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2112                                         &vsi->state));
2113                 if (vsi->type == I40E_VSI_MAIN && pf->lan_veb != I40E_NO_VEB) {
2114                         /* set defport ON for Main VSI instead of true promisc
2115                          * this way we will get all unicast/multicast and VLAN
2116                          * promisc behavior but will not get VF or VMDq traffic
2117                          * replicated on the Main VSI.
2118                          */
2119                         if (pf->cur_promisc != cur_promisc) {
2120                                 pf->cur_promisc = cur_promisc;
2121                                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2122                         }
2123                 } else {
2124                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2125                                                           &vsi->back->hw,
2126                                                           vsi->seid,
2127                                                           cur_promisc, NULL);
2128                         if (aq_ret) {
2129                                 retval =
2130                                 i40e_aq_rc_to_posix(aq_ret,
2131                                                     pf->hw.aq.asq_last_status);
2132                                 dev_info(&pf->pdev->dev,
2133                                          "set unicast promisc failed, err %d, aq_err %d\n",
2134                                          aq_ret, pf->hw.aq.asq_last_status);
2135                         }
2136                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2137                                                           &vsi->back->hw,
2138                                                           vsi->seid,
2139                                                           cur_promisc, NULL);
2140                         if (aq_ret) {
2141                                 retval =
2142                                 i40e_aq_rc_to_posix(aq_ret,
2143                                                     pf->hw.aq.asq_last_status);
2144                                 dev_info(&pf->pdev->dev,
2145                                          "set multicast promisc failed, err %d, aq_err %d\n",
2146                                          aq_ret, pf->hw.aq.asq_last_status);
2147                         }
2148                 }
2149                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2150                                                    vsi->seid,
2151                                                    cur_promisc, NULL);
2152                 if (aq_ret) {
2153                         retval = i40e_aq_rc_to_posix(aq_ret,
2154                                                      pf->hw.aq.asq_last_status);
2155                         dev_info(&pf->pdev->dev,
2156                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2157                                  i40e_stat_str(&pf->hw, aq_ret),
2158                                  i40e_aq_str(&pf->hw,
2159                                              pf->hw.aq.asq_last_status));
2160                 }
2161         }
2162 out:
2163         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2164         return retval;
2165 }
2166
2167 /**
2168  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2169  * @pf: board private structure
2170  **/
2171 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2172 {
2173         int v;
2174
2175         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2176                 return;
2177         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2178
2179         for (v = 0; v < pf->num_alloc_vsi; v++) {
2180                 if (pf->vsi[v] &&
2181                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2182                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2183
2184                         if (ret) {
2185                                 /* come back and try again later */
2186                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2187                                 break;
2188                         }
2189                 }
2190         }
2191 }
2192
2193 /**
2194  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2195  * @netdev: network interface device structure
2196  * @new_mtu: new value for maximum frame size
2197  *
2198  * Returns 0 on success, negative on failure
2199  **/
2200 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2201 {
2202         struct i40e_netdev_priv *np = netdev_priv(netdev);
2203         int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2204         struct i40e_vsi *vsi = np->vsi;
2205
2206         /* MTU < 68 is an error and causes problems on some kernels */
2207         if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2208                 return -EINVAL;
2209
2210         netdev_info(netdev, "changing MTU from %d to %d\n",
2211                     netdev->mtu, new_mtu);
2212         netdev->mtu = new_mtu;
2213         if (netif_running(netdev))
2214                 i40e_vsi_reinit_locked(vsi);
2215         i40e_notify_client_of_l2_param_changes(vsi);
2216         return 0;
2217 }
2218
2219 /**
2220  * i40e_ioctl - Access the hwtstamp interface
2221  * @netdev: network interface device structure
2222  * @ifr: interface request data
2223  * @cmd: ioctl command
2224  **/
2225 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2226 {
2227         struct i40e_netdev_priv *np = netdev_priv(netdev);
2228         struct i40e_pf *pf = np->vsi->back;
2229
2230         switch (cmd) {
2231         case SIOCGHWTSTAMP:
2232                 return i40e_ptp_get_ts_config(pf, ifr);
2233         case SIOCSHWTSTAMP:
2234                 return i40e_ptp_set_ts_config(pf, ifr);
2235         default:
2236                 return -EOPNOTSUPP;
2237         }
2238 }
2239
2240 /**
2241  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2242  * @vsi: the vsi being adjusted
2243  **/
2244 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2245 {
2246         struct i40e_vsi_context ctxt;
2247         i40e_status ret;
2248
2249         if ((vsi->info.valid_sections &
2250              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2251             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2252                 return;  /* already enabled */
2253
2254         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2255         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2256                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2257
2258         ctxt.seid = vsi->seid;
2259         ctxt.info = vsi->info;
2260         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2261         if (ret) {
2262                 dev_info(&vsi->back->pdev->dev,
2263                          "update vlan stripping failed, err %s aq_err %s\n",
2264                          i40e_stat_str(&vsi->back->hw, ret),
2265                          i40e_aq_str(&vsi->back->hw,
2266                                      vsi->back->hw.aq.asq_last_status));
2267         }
2268 }
2269
2270 /**
2271  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2272  * @vsi: the vsi being adjusted
2273  **/
2274 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2275 {
2276         struct i40e_vsi_context ctxt;
2277         i40e_status ret;
2278
2279         if ((vsi->info.valid_sections &
2280              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2281             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2282              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2283                 return;  /* already disabled */
2284
2285         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2286         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2287                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2288
2289         ctxt.seid = vsi->seid;
2290         ctxt.info = vsi->info;
2291         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2292         if (ret) {
2293                 dev_info(&vsi->back->pdev->dev,
2294                          "update vlan stripping failed, err %s aq_err %s\n",
2295                          i40e_stat_str(&vsi->back->hw, ret),
2296                          i40e_aq_str(&vsi->back->hw,
2297                                      vsi->back->hw.aq.asq_last_status));
2298         }
2299 }
2300
2301 /**
2302  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2303  * @netdev: network interface to be adjusted
2304  * @features: netdev features to test if VLAN offload is enabled or not
2305  **/
2306 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2307 {
2308         struct i40e_netdev_priv *np = netdev_priv(netdev);
2309         struct i40e_vsi *vsi = np->vsi;
2310
2311         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2312                 i40e_vlan_stripping_enable(vsi);
2313         else
2314                 i40e_vlan_stripping_disable(vsi);
2315 }
2316
2317 /**
2318  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2319  * @vsi: the vsi being configured
2320  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2321  **/
2322 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2323 {
2324         struct i40e_mac_filter *f, *add_f;
2325         bool is_netdev, is_vf;
2326
2327         is_vf = (vsi->type == I40E_VSI_SRIOV);
2328         is_netdev = !!(vsi->netdev);
2329
2330         /* Locked once because all functions invoked below iterates list*/
2331         spin_lock_bh(&vsi->mac_filter_list_lock);
2332
2333         if (is_netdev) {
2334                 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2335                                         is_vf, is_netdev);
2336                 if (!add_f) {
2337                         dev_info(&vsi->back->pdev->dev,
2338                                  "Could not add vlan filter %d for %pM\n",
2339                                  vid, vsi->netdev->dev_addr);
2340                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2341                         return -ENOMEM;
2342                 }
2343         }
2344
2345         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2346                 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2347                 if (!add_f) {
2348                         dev_info(&vsi->back->pdev->dev,
2349                                  "Could not add vlan filter %d for %pM\n",
2350                                  vid, f->macaddr);
2351                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2352                         return -ENOMEM;
2353                 }
2354         }
2355
2356         /* Now if we add a vlan tag, make sure to check if it is the first
2357          * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2358          * with 0, so we now accept untagged and specified tagged traffic
2359          * (and not any taged and untagged)
2360          */
2361         if (vid > 0) {
2362                 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2363                                                   I40E_VLAN_ANY,
2364                                                   is_vf, is_netdev)) {
2365                         i40e_del_filter(vsi, vsi->netdev->dev_addr,
2366                                         I40E_VLAN_ANY, is_vf, is_netdev);
2367                         add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2368                                                 is_vf, is_netdev);
2369                         if (!add_f) {
2370                                 dev_info(&vsi->back->pdev->dev,
2371                                          "Could not add filter 0 for %pM\n",
2372                                          vsi->netdev->dev_addr);
2373                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2374                                 return -ENOMEM;
2375                         }
2376                 }
2377         }
2378
2379         /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2380         if (vid > 0 && !vsi->info.pvid) {
2381                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2382                         if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2383                                               is_vf, is_netdev))
2384                                 continue;
2385                         i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2386                                         is_vf, is_netdev);
2387                         add_f = i40e_add_filter(vsi, f->macaddr,
2388                                                 0, is_vf, is_netdev);
2389                         if (!add_f) {
2390                                 dev_info(&vsi->back->pdev->dev,
2391                                          "Could not add filter 0 for %pM\n",
2392                                         f->macaddr);
2393                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2394                                 return -ENOMEM;
2395                         }
2396                 }
2397         }
2398
2399         spin_unlock_bh(&vsi->mac_filter_list_lock);
2400
2401         /* schedule our worker thread which will take care of
2402          * applying the new filter changes
2403          */
2404         i40e_service_event_schedule(vsi->back);
2405         return 0;
2406 }
2407
2408 /**
2409  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2410  * @vsi: the vsi being configured
2411  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2412  *
2413  * Return: 0 on success or negative otherwise
2414  **/
2415 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2416 {
2417         struct net_device *netdev = vsi->netdev;
2418         struct i40e_mac_filter *f, *add_f;
2419         bool is_vf, is_netdev;
2420         int filter_count = 0;
2421
2422         is_vf = (vsi->type == I40E_VSI_SRIOV);
2423         is_netdev = !!(netdev);
2424
2425         /* Locked once because all functions invoked below iterates list */
2426         spin_lock_bh(&vsi->mac_filter_list_lock);
2427
2428         if (is_netdev)
2429                 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2430
2431         list_for_each_entry(f, &vsi->mac_filter_list, list)
2432                 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2433
2434         /* go through all the filters for this VSI and if there is only
2435          * vid == 0 it means there are no other filters, so vid 0 must
2436          * be replaced with -1. This signifies that we should from now
2437          * on accept any traffic (with any tag present, or untagged)
2438          */
2439         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2440                 if (is_netdev) {
2441                         if (f->vlan &&
2442                             ether_addr_equal(netdev->dev_addr, f->macaddr))
2443                                 filter_count++;
2444                 }
2445
2446                 if (f->vlan)
2447                         filter_count++;
2448         }
2449
2450         if (!filter_count && is_netdev) {
2451                 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2452                 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2453                                     is_vf, is_netdev);
2454                 if (!f) {
2455                         dev_info(&vsi->back->pdev->dev,
2456                                  "Could not add filter %d for %pM\n",
2457                                  I40E_VLAN_ANY, netdev->dev_addr);
2458                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2459                         return -ENOMEM;
2460                 }
2461         }
2462
2463         if (!filter_count) {
2464                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2465                         i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2466                         add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2467                                                 is_vf, is_netdev);
2468                         if (!add_f) {
2469                                 dev_info(&vsi->back->pdev->dev,
2470                                          "Could not add filter %d for %pM\n",
2471                                          I40E_VLAN_ANY, f->macaddr);
2472                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2473                                 return -ENOMEM;
2474                         }
2475                 }
2476         }
2477
2478         spin_unlock_bh(&vsi->mac_filter_list_lock);
2479
2480         /* schedule our worker thread which will take care of
2481          * applying the new filter changes
2482          */
2483         i40e_service_event_schedule(vsi->back);
2484         return 0;
2485 }
2486
2487 /**
2488  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2489  * @netdev: network interface to be adjusted
2490  * @vid: vlan id to be added
2491  *
2492  * net_device_ops implementation for adding vlan ids
2493  **/
2494 #ifdef I40E_FCOE
2495 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2496                          __always_unused __be16 proto, u16 vid)
2497 #else
2498 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2499                                 __always_unused __be16 proto, u16 vid)
2500 #endif
2501 {
2502         struct i40e_netdev_priv *np = netdev_priv(netdev);
2503         struct i40e_vsi *vsi = np->vsi;
2504         int ret = 0;
2505
2506         if (vid > 4095)
2507                 return -EINVAL;
2508
2509         netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2510
2511         /* If the network stack called us with vid = 0 then
2512          * it is asking to receive priority tagged packets with
2513          * vlan id 0.  Our HW receives them by default when configured
2514          * to receive untagged packets so there is no need to add an
2515          * extra filter for vlan 0 tagged packets.
2516          */
2517         if (vid)
2518                 ret = i40e_vsi_add_vlan(vsi, vid);
2519
2520         if (!ret && (vid < VLAN_N_VID))
2521                 set_bit(vid, vsi->active_vlans);
2522
2523         return ret;
2524 }
2525
2526 /**
2527  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2528  * @netdev: network interface to be adjusted
2529  * @vid: vlan id to be removed
2530  *
2531  * net_device_ops implementation for removing vlan ids
2532  **/
2533 #ifdef I40E_FCOE
2534 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2535                           __always_unused __be16 proto, u16 vid)
2536 #else
2537 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2538                                  __always_unused __be16 proto, u16 vid)
2539 #endif
2540 {
2541         struct i40e_netdev_priv *np = netdev_priv(netdev);
2542         struct i40e_vsi *vsi = np->vsi;
2543
2544         netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2545
2546         /* return code is ignored as there is nothing a user
2547          * can do about failure to remove and a log message was
2548          * already printed from the other function
2549          */
2550         i40e_vsi_kill_vlan(vsi, vid);
2551
2552         clear_bit(vid, vsi->active_vlans);
2553
2554         return 0;
2555 }
2556
2557 /**
2558  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2559  * @vsi: the vsi being brought back up
2560  **/
2561 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2562 {
2563         u16 vid;
2564
2565         if (!vsi->netdev)
2566                 return;
2567
2568         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2569
2570         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2571                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2572                                      vid);
2573 }
2574
2575 /**
2576  * i40e_vsi_add_pvid - Add pvid for the VSI
2577  * @vsi: the vsi being adjusted
2578  * @vid: the vlan id to set as a PVID
2579  **/
2580 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2581 {
2582         struct i40e_vsi_context ctxt;
2583         i40e_status ret;
2584
2585         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2586         vsi->info.pvid = cpu_to_le16(vid);
2587         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2588                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2589                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2590
2591         ctxt.seid = vsi->seid;
2592         ctxt.info = vsi->info;
2593         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2594         if (ret) {
2595                 dev_info(&vsi->back->pdev->dev,
2596                          "add pvid failed, err %s aq_err %s\n",
2597                          i40e_stat_str(&vsi->back->hw, ret),
2598                          i40e_aq_str(&vsi->back->hw,
2599                                      vsi->back->hw.aq.asq_last_status));
2600                 return -ENOENT;
2601         }
2602
2603         return 0;
2604 }
2605
2606 /**
2607  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2608  * @vsi: the vsi being adjusted
2609  *
2610  * Just use the vlan_rx_register() service to put it back to normal
2611  **/
2612 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2613 {
2614         i40e_vlan_stripping_disable(vsi);
2615
2616         vsi->info.pvid = 0;
2617 }
2618
2619 /**
2620  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2621  * @vsi: ptr to the VSI
2622  *
2623  * If this function returns with an error, then it's possible one or
2624  * more of the rings is populated (while the rest are not).  It is the
2625  * callers duty to clean those orphaned rings.
2626  *
2627  * Return 0 on success, negative on failure
2628  **/
2629 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2630 {
2631         int i, err = 0;
2632
2633         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2634                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2635
2636         return err;
2637 }
2638
2639 /**
2640  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2641  * @vsi: ptr to the VSI
2642  *
2643  * Free VSI's transmit software resources
2644  **/
2645 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2646 {
2647         int i;
2648
2649         if (!vsi->tx_rings)
2650                 return;
2651
2652         for (i = 0; i < vsi->num_queue_pairs; i++)
2653                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2654                         i40e_free_tx_resources(vsi->tx_rings[i]);
2655 }
2656
2657 /**
2658  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2659  * @vsi: ptr to the VSI
2660  *
2661  * If this function returns with an error, then it's possible one or
2662  * more of the rings is populated (while the rest are not).  It is the
2663  * callers duty to clean those orphaned rings.
2664  *
2665  * Return 0 on success, negative on failure
2666  **/
2667 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2668 {
2669         int i, err = 0;
2670
2671         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2672                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2673 #ifdef I40E_FCOE
2674         i40e_fcoe_setup_ddp_resources(vsi);
2675 #endif
2676         return err;
2677 }
2678
2679 /**
2680  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2681  * @vsi: ptr to the VSI
2682  *
2683  * Free all receive software resources
2684  **/
2685 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2686 {
2687         int i;
2688
2689         if (!vsi->rx_rings)
2690                 return;
2691
2692         for (i = 0; i < vsi->num_queue_pairs; i++)
2693                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2694                         i40e_free_rx_resources(vsi->rx_rings[i]);
2695 #ifdef I40E_FCOE
2696         i40e_fcoe_free_ddp_resources(vsi);
2697 #endif
2698 }
2699
2700 /**
2701  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2702  * @ring: The Tx ring to configure
2703  *
2704  * This enables/disables XPS for a given Tx descriptor ring
2705  * based on the TCs enabled for the VSI that ring belongs to.
2706  **/
2707 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2708 {
2709         struct i40e_vsi *vsi = ring->vsi;
2710         cpumask_var_t mask;
2711
2712         if (!ring->q_vector || !ring->netdev)
2713                 return;
2714
2715         /* Single TC mode enable XPS */
2716         if (vsi->tc_config.numtc <= 1) {
2717                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2718                         netif_set_xps_queue(ring->netdev,
2719                                             &ring->q_vector->affinity_mask,
2720                                             ring->queue_index);
2721         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2722                 /* Disable XPS to allow selection based on TC */
2723                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2724                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2725                 free_cpumask_var(mask);
2726         }
2727
2728         /* schedule our worker thread which will take care of
2729          * applying the new filter changes
2730          */
2731         i40e_service_event_schedule(vsi->back);
2732 }
2733
2734 /**
2735  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2736  * @ring: The Tx ring to configure
2737  *
2738  * Configure the Tx descriptor ring in the HMC context.
2739  **/
2740 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2741 {
2742         struct i40e_vsi *vsi = ring->vsi;
2743         u16 pf_q = vsi->base_queue + ring->queue_index;
2744         struct i40e_hw *hw = &vsi->back->hw;
2745         struct i40e_hmc_obj_txq tx_ctx;
2746         i40e_status err = 0;
2747         u32 qtx_ctl = 0;
2748
2749         /* some ATR related tx ring init */
2750         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2751                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2752                 ring->atr_count = 0;
2753         } else {
2754                 ring->atr_sample_rate = 0;
2755         }
2756
2757         /* configure XPS */
2758         i40e_config_xps_tx_ring(ring);
2759
2760         /* clear the context structure first */
2761         memset(&tx_ctx, 0, sizeof(tx_ctx));
2762
2763         tx_ctx.new_context = 1;
2764         tx_ctx.base = (ring->dma / 128);
2765         tx_ctx.qlen = ring->count;
2766         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2767                                                I40E_FLAG_FD_ATR_ENABLED));
2768 #ifdef I40E_FCOE
2769         tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2770 #endif
2771         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2772         /* FDIR VSI tx ring can still use RS bit and writebacks */
2773         if (vsi->type != I40E_VSI_FDIR)
2774                 tx_ctx.head_wb_ena = 1;
2775         tx_ctx.head_wb_addr = ring->dma +
2776                               (ring->count * sizeof(struct i40e_tx_desc));
2777
2778         /* As part of VSI creation/update, FW allocates certain
2779          * Tx arbitration queue sets for each TC enabled for
2780          * the VSI. The FW returns the handles to these queue
2781          * sets as part of the response buffer to Add VSI,
2782          * Update VSI, etc. AQ commands. It is expected that
2783          * these queue set handles be associated with the Tx
2784          * queues by the driver as part of the TX queue context
2785          * initialization. This has to be done regardless of
2786          * DCB as by default everything is mapped to TC0.
2787          */
2788         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2789         tx_ctx.rdylist_act = 0;
2790
2791         /* clear the context in the HMC */
2792         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2793         if (err) {
2794                 dev_info(&vsi->back->pdev->dev,
2795                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2796                          ring->queue_index, pf_q, err);
2797                 return -ENOMEM;
2798         }
2799
2800         /* set the context in the HMC */
2801         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2802         if (err) {
2803                 dev_info(&vsi->back->pdev->dev,
2804                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2805                          ring->queue_index, pf_q, err);
2806                 return -ENOMEM;
2807         }
2808
2809         /* Now associate this queue with this PCI function */
2810         if (vsi->type == I40E_VSI_VMDQ2) {
2811                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2812                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2813                            I40E_QTX_CTL_VFVM_INDX_MASK;
2814         } else {
2815                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2816         }
2817
2818         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2819                     I40E_QTX_CTL_PF_INDX_MASK);
2820         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2821         i40e_flush(hw);
2822
2823         /* cache tail off for easier writes later */
2824         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2825
2826         return 0;
2827 }
2828
2829 /**
2830  * i40e_configure_rx_ring - Configure a receive ring context
2831  * @ring: The Rx ring to configure
2832  *
2833  * Configure the Rx descriptor ring in the HMC context.
2834  **/
2835 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2836 {
2837         struct i40e_vsi *vsi = ring->vsi;
2838         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2839         u16 pf_q = vsi->base_queue + ring->queue_index;
2840         struct i40e_hw *hw = &vsi->back->hw;
2841         struct i40e_hmc_obj_rxq rx_ctx;
2842         i40e_status err = 0;
2843
2844         ring->state = 0;
2845
2846         /* clear the context structure first */
2847         memset(&rx_ctx, 0, sizeof(rx_ctx));
2848
2849         ring->rx_buf_len = vsi->rx_buf_len;
2850         ring->rx_hdr_len = vsi->rx_hdr_len;
2851
2852         rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2853         rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2854
2855         rx_ctx.base = (ring->dma / 128);
2856         rx_ctx.qlen = ring->count;
2857
2858         if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2859                 set_ring_16byte_desc_enabled(ring);
2860                 rx_ctx.dsize = 0;
2861         } else {
2862                 rx_ctx.dsize = 1;
2863         }
2864
2865         rx_ctx.dtype = vsi->dtype;
2866         if (vsi->dtype) {
2867                 set_ring_ps_enabled(ring);
2868                 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2869                                   I40E_RX_SPLIT_IP      |
2870                                   I40E_RX_SPLIT_TCP_UDP |
2871                                   I40E_RX_SPLIT_SCTP;
2872         } else {
2873                 rx_ctx.hsplit_0 = 0;
2874         }
2875
2876         rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2877                                   (chain_len * ring->rx_buf_len));
2878         if (hw->revision_id == 0)
2879                 rx_ctx.lrxqthresh = 0;
2880         else
2881                 rx_ctx.lrxqthresh = 2;
2882         rx_ctx.crcstrip = 1;
2883         rx_ctx.l2tsel = 1;
2884         /* this controls whether VLAN is stripped from inner headers */
2885         rx_ctx.showiv = 0;
2886 #ifdef I40E_FCOE
2887         rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2888 #endif
2889         /* set the prefena field to 1 because the manual says to */
2890         rx_ctx.prefena = 1;
2891
2892         /* clear the context in the HMC */
2893         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2894         if (err) {
2895                 dev_info(&vsi->back->pdev->dev,
2896                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2897                          ring->queue_index, pf_q, err);
2898                 return -ENOMEM;
2899         }
2900
2901         /* set the context in the HMC */
2902         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2903         if (err) {
2904                 dev_info(&vsi->back->pdev->dev,
2905                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2906                          ring->queue_index, pf_q, err);
2907                 return -ENOMEM;
2908         }
2909
2910         /* cache tail for quicker writes, and clear the reg before use */
2911         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2912         writel(0, ring->tail);
2913
2914         if (ring_is_ps_enabled(ring)) {
2915                 i40e_alloc_rx_headers(ring);
2916                 i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2917         } else {
2918                 i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2919         }
2920
2921         return 0;
2922 }
2923
2924 /**
2925  * i40e_vsi_configure_tx - Configure the VSI for Tx
2926  * @vsi: VSI structure describing this set of rings and resources
2927  *
2928  * Configure the Tx VSI for operation.
2929  **/
2930 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2931 {
2932         int err = 0;
2933         u16 i;
2934
2935         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2936                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2937
2938         return err;
2939 }
2940
2941 /**
2942  * i40e_vsi_configure_rx - Configure the VSI for Rx
2943  * @vsi: the VSI being configured
2944  *
2945  * Configure the Rx VSI for operation.
2946  **/
2947 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2948 {
2949         int err = 0;
2950         u16 i;
2951
2952         if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2953                 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2954                                + ETH_FCS_LEN + VLAN_HLEN;
2955         else
2956                 vsi->max_frame = I40E_RXBUFFER_2048;
2957
2958         /* figure out correct receive buffer length */
2959         switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2960                                     I40E_FLAG_RX_PS_ENABLED)) {
2961         case I40E_FLAG_RX_1BUF_ENABLED:
2962                 vsi->rx_hdr_len = 0;
2963                 vsi->rx_buf_len = vsi->max_frame;
2964                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2965                 break;
2966         case I40E_FLAG_RX_PS_ENABLED:
2967                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2968                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2969                 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2970                 break;
2971         default:
2972                 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2973                 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2974                 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2975                 break;
2976         }
2977
2978 #ifdef I40E_FCOE
2979         /* setup rx buffer for FCoE */
2980         if ((vsi->type == I40E_VSI_FCOE) &&
2981             (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2982                 vsi->rx_hdr_len = 0;
2983                 vsi->rx_buf_len = I40E_RXBUFFER_3072;
2984                 vsi->max_frame = I40E_RXBUFFER_3072;
2985                 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2986         }
2987
2988 #endif /* I40E_FCOE */
2989         /* round up for the chip's needs */
2990         vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2991                                 BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT));
2992         vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2993                                 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
2994
2995         /* set up individual rings */
2996         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2997                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2998
2999         return err;
3000 }
3001
3002 /**
3003  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3004  * @vsi: ptr to the VSI
3005  **/
3006 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3007 {
3008         struct i40e_ring *tx_ring, *rx_ring;
3009         u16 qoffset, qcount;
3010         int i, n;
3011
3012         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3013                 /* Reset the TC information */
3014                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3015                         rx_ring = vsi->rx_rings[i];
3016                         tx_ring = vsi->tx_rings[i];
3017                         rx_ring->dcb_tc = 0;
3018                         tx_ring->dcb_tc = 0;
3019                 }
3020         }
3021
3022         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3023                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3024                         continue;
3025
3026                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3027                 qcount = vsi->tc_config.tc_info[n].qcount;
3028                 for (i = qoffset; i < (qoffset + qcount); i++) {
3029                         rx_ring = vsi->rx_rings[i];
3030                         tx_ring = vsi->tx_rings[i];
3031                         rx_ring->dcb_tc = n;
3032                         tx_ring->dcb_tc = n;
3033                 }
3034         }
3035 }
3036
3037 /**
3038  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3039  * @vsi: ptr to the VSI
3040  **/
3041 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3042 {
3043         if (vsi->netdev)
3044                 i40e_set_rx_mode(vsi->netdev);
3045 }
3046
3047 /**
3048  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3049  * @vsi: Pointer to the targeted VSI
3050  *
3051  * This function replays the hlist on the hw where all the SB Flow Director
3052  * filters were saved.
3053  **/
3054 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3055 {
3056         struct i40e_fdir_filter *filter;
3057         struct i40e_pf *pf = vsi->back;
3058         struct hlist_node *node;
3059
3060         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3061                 return;
3062
3063         hlist_for_each_entry_safe(filter, node,
3064                                   &pf->fdir_filter_list, fdir_node) {
3065                 i40e_add_del_fdir(vsi, filter, true);
3066         }
3067 }
3068
3069 /**
3070  * i40e_vsi_configure - Set up the VSI for action
3071  * @vsi: the VSI being configured
3072  **/
3073 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3074 {
3075         int err;
3076
3077         i40e_set_vsi_rx_mode(vsi);
3078         i40e_restore_vlan(vsi);
3079         i40e_vsi_config_dcb_rings(vsi);
3080         err = i40e_vsi_configure_tx(vsi);
3081         if (!err)
3082                 err = i40e_vsi_configure_rx(vsi);
3083
3084         return err;
3085 }
3086
3087 /**
3088  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3089  * @vsi: the VSI being configured
3090  **/
3091 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3092 {
3093         struct i40e_pf *pf = vsi->back;
3094         struct i40e_hw *hw = &pf->hw;
3095         u16 vector;
3096         int i, q;
3097         u32 qp;
3098
3099         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3100          * and PFINT_LNKLSTn registers, e.g.:
3101          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3102          */
3103         qp = vsi->base_queue;
3104         vector = vsi->base_vector;
3105         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3106                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3107
3108                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3109                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3110                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3111                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3112                      q_vector->rx.itr);
3113                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3114                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3115                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3116                      q_vector->tx.itr);
3117                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3118                      INTRL_USEC_TO_REG(vsi->int_rate_limit));
3119
3120                 /* Linked list for the queuepairs assigned to this vector */
3121                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3122                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3123                         u32 val;
3124
3125                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3126                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3127                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3128                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3129                               (I40E_QUEUE_TYPE_TX
3130                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3131
3132                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3133
3134                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3135                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3136                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3137                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3138                               (I40E_QUEUE_TYPE_RX
3139                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3140
3141                         /* Terminate the linked list */
3142                         if (q == (q_vector->num_ringpairs - 1))
3143                                 val |= (I40E_QUEUE_END_OF_LIST
3144                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3145
3146                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3147                         qp++;
3148                 }
3149         }
3150
3151         i40e_flush(hw);
3152 }
3153
3154 /**
3155  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3156  * @hw: ptr to the hardware info
3157  **/
3158 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3159 {
3160         struct i40e_hw *hw = &pf->hw;
3161         u32 val;
3162
3163         /* clear things first */
3164         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3165         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3166
3167         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3168               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3169               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3170               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3171               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3172               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3173               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3174               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3175
3176         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3177                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3178
3179         if (pf->flags & I40E_FLAG_PTP)
3180                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3181
3182         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3183
3184         /* SW_ITR_IDX = 0, but don't change INTENA */
3185         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3186                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3187
3188         /* OTHER_ITR_IDX = 0 */
3189         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3190 }
3191
3192 /**
3193  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3194  * @vsi: the VSI being configured
3195  **/
3196 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3197 {
3198         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3199         struct i40e_pf *pf = vsi->back;
3200         struct i40e_hw *hw = &pf->hw;
3201         u32 val;
3202
3203         /* set the ITR configuration */
3204         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3205         q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3206         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3207         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3208         q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3209         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3210         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3211
3212         i40e_enable_misc_int_causes(pf);
3213
3214         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3215         wr32(hw, I40E_PFINT_LNKLST0, 0);
3216
3217         /* Associate the queue pair to the vector and enable the queue int */
3218         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3219               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3220               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3221
3222         wr32(hw, I40E_QINT_RQCTL(0), val);
3223
3224         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3225               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3226               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3227
3228         wr32(hw, I40E_QINT_TQCTL(0), val);
3229         i40e_flush(hw);
3230 }
3231
3232 /**
3233  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3234  * @pf: board private structure
3235  **/
3236 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3237 {
3238         struct i40e_hw *hw = &pf->hw;
3239
3240         wr32(hw, I40E_PFINT_DYN_CTL0,
3241              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3242         i40e_flush(hw);
3243 }
3244
3245 /**
3246  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3247  * @pf: board private structure
3248  **/
3249 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3250 {
3251         struct i40e_hw *hw = &pf->hw;
3252         u32 val;
3253
3254         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3255               I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3256               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3257
3258         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3259         i40e_flush(hw);
3260 }
3261
3262 /**
3263  * i40e_irq_dynamic_disable - Disable default interrupt generation settings
3264  * @vsi: pointer to a vsi
3265  * @vector: disable a particular Hw Interrupt vector
3266  **/
3267 void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
3268 {
3269         struct i40e_pf *pf = vsi->back;
3270         struct i40e_hw *hw = &pf->hw;
3271         u32 val;
3272
3273         val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
3274         wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
3275         i40e_flush(hw);
3276 }
3277
3278 /**
3279  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3280  * @irq: interrupt number
3281  * @data: pointer to a q_vector
3282  **/
3283 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3284 {
3285         struct i40e_q_vector *q_vector = data;
3286
3287         if (!q_vector->tx.ring && !q_vector->rx.ring)
3288                 return IRQ_HANDLED;
3289
3290         napi_schedule_irqoff(&q_vector->napi);
3291
3292         return IRQ_HANDLED;
3293 }
3294
3295 /**
3296  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3297  * @vsi: the VSI being configured
3298  * @basename: name for the vector
3299  *
3300  * Allocates MSI-X vectors and requests interrupts from the kernel.
3301  **/
3302 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3303 {
3304         int q_vectors = vsi->num_q_vectors;
3305         struct i40e_pf *pf = vsi->back;
3306         int base = vsi->base_vector;
3307         int rx_int_idx = 0;
3308         int tx_int_idx = 0;
3309         int vector, err;
3310
3311         for (vector = 0; vector < q_vectors; vector++) {
3312                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3313
3314                 if (q_vector->tx.ring && q_vector->rx.ring) {
3315                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3316                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3317                         tx_int_idx++;
3318                 } else if (q_vector->rx.ring) {
3319                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3320                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3321                 } else if (q_vector->tx.ring) {
3322                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3323                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3324                 } else {
3325                         /* skip this unused q_vector */
3326                         continue;
3327                 }
3328                 err = request_irq(pf->msix_entries[base + vector].vector,
3329                                   vsi->irq_handler,
3330                                   0,
3331                                   q_vector->name,
3332                                   q_vector);
3333                 if (err) {
3334                         dev_info(&pf->pdev->dev,
3335                                  "MSIX request_irq failed, error: %d\n", err);
3336                         goto free_queue_irqs;
3337                 }
3338                 /* assign the mask for this irq */
3339                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3340                                       &q_vector->affinity_mask);
3341         }
3342
3343         vsi->irqs_ready = true;
3344         return 0;
3345
3346 free_queue_irqs:
3347         while (vector) {
3348                 vector--;
3349                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3350                                       NULL);
3351                 free_irq(pf->msix_entries[base + vector].vector,
3352                          &(vsi->q_vectors[vector]));
3353         }
3354         return err;
3355 }
3356
3357 /**
3358  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3359  * @vsi: the VSI being un-configured
3360  **/
3361 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3362 {
3363         struct i40e_pf *pf = vsi->back;
3364         struct i40e_hw *hw = &pf->hw;
3365         int base = vsi->base_vector;
3366         int i;
3367
3368         for (i = 0; i < vsi->num_queue_pairs; i++) {
3369                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3370                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3371         }
3372
3373         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3374                 for (i = vsi->base_vector;
3375                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3376                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3377
3378                 i40e_flush(hw);
3379                 for (i = 0; i < vsi->num_q_vectors; i++)
3380                         synchronize_irq(pf->msix_entries[i + base].vector);
3381         } else {
3382                 /* Legacy and MSI mode - this stops all interrupt handling */
3383                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3384                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3385                 i40e_flush(hw);
3386                 synchronize_irq(pf->pdev->irq);
3387         }
3388 }
3389
3390 /**
3391  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3392  * @vsi: the VSI being configured
3393  **/
3394 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3395 {
3396         struct i40e_pf *pf = vsi->back;
3397         int i;
3398
3399         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3400                 for (i = 0; i < vsi->num_q_vectors; i++)
3401                         i40e_irq_dynamic_enable(vsi, i);
3402         } else {
3403                 i40e_irq_dynamic_enable_icr0(pf);
3404         }
3405
3406         i40e_flush(&pf->hw);
3407         return 0;
3408 }
3409
3410 /**
3411  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3412  * @pf: board private structure
3413  **/
3414 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3415 {
3416         /* Disable ICR 0 */
3417         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3418         i40e_flush(&pf->hw);
3419 }
3420
3421 /**
3422  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3423  * @irq: interrupt number
3424  * @data: pointer to a q_vector
3425  *
3426  * This is the handler used for all MSI/Legacy interrupts, and deals
3427  * with both queue and non-queue interrupts.  This is also used in
3428  * MSIX mode to handle the non-queue interrupts.
3429  **/
3430 static irqreturn_t i40e_intr(int irq, void *data)
3431 {
3432         struct i40e_pf *pf = (struct i40e_pf *)data;
3433         struct i40e_hw *hw = &pf->hw;
3434         irqreturn_t ret = IRQ_NONE;
3435         u32 icr0, icr0_remaining;
3436         u32 val, ena_mask;
3437
3438         icr0 = rd32(hw, I40E_PFINT_ICR0);
3439         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3440
3441         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3442         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3443                 goto enable_intr;
3444
3445         /* if interrupt but no bits showing, must be SWINT */
3446         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3447             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3448                 pf->sw_int_count++;
3449
3450         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3451             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3452                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3453                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3454                 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3455         }
3456
3457         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3458         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3459                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3460                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3461
3462                 /* temporarily disable queue cause for NAPI processing */
3463                 u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3464
3465                 qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3466                 wr32(hw, I40E_QINT_RQCTL(0), qval);
3467
3468                 qval = rd32(hw, I40E_QINT_TQCTL(0));
3469                 qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3470                 wr32(hw, I40E_QINT_TQCTL(0), qval);
3471
3472                 if (!test_bit(__I40E_DOWN, &pf->state))
3473                         napi_schedule_irqoff(&q_vector->napi);
3474         }
3475
3476         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3477                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3478                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3479         }
3480
3481         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3482                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3483                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3484         }
3485
3486         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3487                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3488                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3489         }
3490
3491         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3492                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3493                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3494                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3495                 val = rd32(hw, I40E_GLGEN_RSTAT);
3496                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3497                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3498                 if (val == I40E_RESET_CORER) {
3499                         pf->corer_count++;
3500                 } else if (val == I40E_RESET_GLOBR) {
3501                         pf->globr_count++;
3502                 } else if (val == I40E_RESET_EMPR) {
3503                         pf->empr_count++;
3504                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3505                 }
3506         }
3507
3508         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3509                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3510                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3511                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3512                          rd32(hw, I40E_PFHMC_ERRORINFO),
3513                          rd32(hw, I40E_PFHMC_ERRORDATA));
3514         }
3515
3516         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3517                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3518
3519                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3520                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3521                         i40e_ptp_tx_hwtstamp(pf);
3522                 }
3523         }
3524
3525         /* If a critical error is pending we have no choice but to reset the
3526          * device.
3527          * Report and mask out any remaining unexpected interrupts.
3528          */
3529         icr0_remaining = icr0 & ena_mask;
3530         if (icr0_remaining) {
3531                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3532                          icr0_remaining);
3533                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3534                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3535                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3536                         dev_info(&pf->pdev->dev, "device will be reset\n");
3537                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3538                         i40e_service_event_schedule(pf);
3539                 }
3540                 ena_mask &= ~icr0_remaining;
3541         }
3542         ret = IRQ_HANDLED;
3543
3544 enable_intr:
3545         /* re-enable interrupt causes */
3546         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3547         if (!test_bit(__I40E_DOWN, &pf->state)) {
3548                 i40e_service_event_schedule(pf);
3549                 i40e_irq_dynamic_enable_icr0(pf);
3550         }
3551
3552         return ret;
3553 }
3554
3555 /**
3556  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3557  * @tx_ring:  tx ring to clean
3558  * @budget:   how many cleans we're allowed
3559  *
3560  * Returns true if there's any budget left (e.g. the clean is finished)
3561  **/
3562 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3563 {
3564         struct i40e_vsi *vsi = tx_ring->vsi;
3565         u16 i = tx_ring->next_to_clean;
3566         struct i40e_tx_buffer *tx_buf;
3567         struct i40e_tx_desc *tx_desc;
3568
3569         tx_buf = &tx_ring->tx_bi[i];
3570         tx_desc = I40E_TX_DESC(tx_ring, i);
3571         i -= tx_ring->count;
3572
3573         do {
3574                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3575
3576                 /* if next_to_watch is not set then there is no work pending */
3577                 if (!eop_desc)
3578                         break;
3579
3580                 /* prevent any other reads prior to eop_desc */
3581                 read_barrier_depends();
3582
3583                 /* if the descriptor isn't done, no work yet to do */
3584                 if (!(eop_desc->cmd_type_offset_bsz &
3585                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3586                         break;
3587
3588                 /* clear next_to_watch to prevent false hangs */
3589                 tx_buf->next_to_watch = NULL;
3590
3591                 tx_desc->buffer_addr = 0;
3592                 tx_desc->cmd_type_offset_bsz = 0;
3593                 /* move past filter desc */
3594                 tx_buf++;
3595                 tx_desc++;
3596                 i++;
3597                 if (unlikely(!i)) {
3598                         i -= tx_ring->count;
3599                         tx_buf = tx_ring->tx_bi;
3600                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3601                 }
3602                 /* unmap skb header data */
3603                 dma_unmap_single(tx_ring->dev,
3604                                  dma_unmap_addr(tx_buf, dma),
3605                                  dma_unmap_len(tx_buf, len),
3606                                  DMA_TO_DEVICE);
3607                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3608                         kfree(tx_buf->raw_buf);
3609
3610                 tx_buf->raw_buf = NULL;
3611                 tx_buf->tx_flags = 0;
3612                 tx_buf->next_to_watch = NULL;
3613                 dma_unmap_len_set(tx_buf, len, 0);
3614                 tx_desc->buffer_addr = 0;
3615                 tx_desc->cmd_type_offset_bsz = 0;
3616
3617                 /* move us past the eop_desc for start of next FD desc */
3618                 tx_buf++;
3619                 tx_desc++;
3620                 i++;
3621                 if (unlikely(!i)) {
3622                         i -= tx_ring->count;
3623                         tx_buf = tx_ring->tx_bi;
3624                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3625                 }
3626
3627                 /* update budget accounting */
3628                 budget--;
3629         } while (likely(budget));
3630
3631         i += tx_ring->count;
3632         tx_ring->next_to_clean = i;
3633
3634         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3635                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3636
3637         return budget > 0;
3638 }
3639
3640 /**
3641  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3642  * @irq: interrupt number
3643  * @data: pointer to a q_vector
3644  **/
3645 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3646 {
3647         struct i40e_q_vector *q_vector = data;
3648         struct i40e_vsi *vsi;
3649
3650         if (!q_vector->tx.ring)
3651                 return IRQ_HANDLED;
3652
3653         vsi = q_vector->tx.ring->vsi;
3654         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3655
3656         return IRQ_HANDLED;
3657 }
3658
3659 /**
3660  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3661  * @vsi: the VSI being configured
3662  * @v_idx: vector index
3663  * @qp_idx: queue pair index
3664  **/
3665 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3666 {
3667         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3668         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3669         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3670
3671         tx_ring->q_vector = q_vector;
3672         tx_ring->next = q_vector->tx.ring;
3673         q_vector->tx.ring = tx_ring;
3674         q_vector->tx.count++;
3675
3676         rx_ring->q_vector = q_vector;
3677         rx_ring->next = q_vector->rx.ring;
3678         q_vector->rx.ring = rx_ring;
3679         q_vector->rx.count++;
3680 }
3681
3682 /**
3683  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3684  * @vsi: the VSI being configured
3685  *
3686  * This function maps descriptor rings to the queue-specific vectors
3687  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3688  * one vector per queue pair, but on a constrained vector budget, we
3689  * group the queue pairs as "efficiently" as possible.
3690  **/
3691 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3692 {
3693         int qp_remaining = vsi->num_queue_pairs;
3694         int q_vectors = vsi->num_q_vectors;
3695         int num_ringpairs;
3696         int v_start = 0;
3697         int qp_idx = 0;
3698
3699         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3700          * group them so there are multiple queues per vector.
3701          * It is also important to go through all the vectors available to be
3702          * sure that if we don't use all the vectors, that the remaining vectors
3703          * are cleared. This is especially important when decreasing the
3704          * number of queues in use.
3705          */
3706         for (; v_start < q_vectors; v_start++) {
3707                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3708
3709                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3710
3711                 q_vector->num_ringpairs = num_ringpairs;
3712
3713                 q_vector->rx.count = 0;
3714                 q_vector->tx.count = 0;
3715                 q_vector->rx.ring = NULL;
3716                 q_vector->tx.ring = NULL;
3717
3718                 while (num_ringpairs--) {
3719                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3720                         qp_idx++;
3721                         qp_remaining--;
3722                 }
3723         }
3724 }
3725
3726 /**
3727  * i40e_vsi_request_irq - Request IRQ from the OS
3728  * @vsi: the VSI being configured
3729  * @basename: name for the vector
3730  **/
3731 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3732 {
3733         struct i40e_pf *pf = vsi->back;
3734         int err;
3735
3736         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3737                 err = i40e_vsi_request_irq_msix(vsi, basename);
3738         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3739                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3740                                   pf->int_name, pf);
3741         else
3742                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3743                                   pf->int_name, pf);
3744
3745         if (err)
3746                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3747
3748         return err;
3749 }
3750
3751 #ifdef CONFIG_NET_POLL_CONTROLLER
3752 /**
3753  * i40e_netpoll - A Polling 'interrupt'handler
3754  * @netdev: network interface device structure
3755  *
3756  * This is used by netconsole to send skbs without having to re-enable
3757  * interrupts.  It's not called while the normal interrupt routine is executing.
3758  **/
3759 #ifdef I40E_FCOE
3760 void i40e_netpoll(struct net_device *netdev)
3761 #else
3762 static void i40e_netpoll(struct net_device *netdev)
3763 #endif
3764 {
3765         struct i40e_netdev_priv *np = netdev_priv(netdev);
3766         struct i40e_vsi *vsi = np->vsi;
3767         struct i40e_pf *pf = vsi->back;
3768         int i;
3769
3770         /* if interface is down do nothing */
3771         if (test_bit(__I40E_DOWN, &vsi->state))
3772                 return;
3773
3774         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3775                 for (i = 0; i < vsi->num_q_vectors; i++)
3776                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3777         } else {
3778                 i40e_intr(pf->pdev->irq, netdev);
3779         }
3780 }
3781 #endif
3782
3783 /**
3784  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3785  * @pf: the PF being configured
3786  * @pf_q: the PF queue
3787  * @enable: enable or disable state of the queue
3788  *
3789  * This routine will wait for the given Tx queue of the PF to reach the
3790  * enabled or disabled state.
3791  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3792  * multiple retries; else will return 0 in case of success.
3793  **/
3794 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3795 {
3796         int i;
3797         u32 tx_reg;
3798
3799         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3800                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3801                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3802                         break;
3803
3804                 usleep_range(10, 20);
3805         }
3806         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3807                 return -ETIMEDOUT;
3808
3809         return 0;
3810 }
3811
3812 /**
3813  * i40e_vsi_control_tx - Start or stop a VSI's rings
3814  * @vsi: the VSI being configured
3815  * @enable: start or stop the rings
3816  **/
3817 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3818 {
3819         struct i40e_pf *pf = vsi->back;
3820         struct i40e_hw *hw = &pf->hw;
3821         int i, j, pf_q, ret = 0;
3822         u32 tx_reg;
3823
3824         pf_q = vsi->base_queue;
3825         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3826
3827                 /* warn the TX unit of coming changes */
3828                 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3829                 if (!enable)
3830                         usleep_range(10, 20);
3831
3832                 for (j = 0; j < 50; j++) {
3833                         tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3834                         if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3835                             ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3836                                 break;
3837                         usleep_range(1000, 2000);
3838                 }
3839                 /* Skip if the queue is already in the requested state */
3840                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3841                         continue;
3842
3843                 /* turn on/off the queue */
3844                 if (enable) {
3845                         wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3846                         tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3847                 } else {
3848                         tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3849                 }
3850
3851                 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3852                 /* No waiting for the Tx queue to disable */
3853                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3854                         continue;
3855
3856                 /* wait for the change to finish */
3857                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3858                 if (ret) {
3859                         dev_info(&pf->pdev->dev,
3860                                  "VSI seid %d Tx ring %d %sable timeout\n",
3861                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3862                         break;
3863                 }
3864         }
3865
3866         if (hw->revision_id == 0)
3867                 mdelay(50);
3868         return ret;
3869 }
3870
3871 /**
3872  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3873  * @pf: the PF being configured
3874  * @pf_q: the PF queue
3875  * @enable: enable or disable state of the queue
3876  *
3877  * This routine will wait for the given Rx queue of the PF to reach the
3878  * enabled or disabled state.
3879  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3880  * multiple retries; else will return 0 in case of success.
3881  **/
3882 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3883 {
3884         int i;
3885         u32 rx_reg;
3886
3887         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3888                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3889                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3890                         break;
3891
3892                 usleep_range(10, 20);
3893         }
3894         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3895                 return -ETIMEDOUT;
3896
3897         return 0;
3898 }
3899
3900 /**
3901  * i40e_vsi_control_rx - Start or stop a VSI's rings
3902  * @vsi: the VSI being configured
3903  * @enable: start or stop the rings
3904  **/
3905 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3906 {
3907         struct i40e_pf *pf = vsi->back;
3908         struct i40e_hw *hw = &pf->hw;
3909         int i, j, pf_q, ret = 0;
3910         u32 rx_reg;
3911
3912         pf_q = vsi->base_queue;
3913         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3914                 for (j = 0; j < 50; j++) {
3915                         rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3916                         if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3917                             ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3918                                 break;
3919                         usleep_range(1000, 2000);
3920                 }
3921
3922                 /* Skip if the queue is already in the requested state */
3923                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3924                         continue;
3925
3926                 /* turn on/off the queue */
3927                 if (enable)
3928                         rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3929                 else
3930                         rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3931                 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3932
3933                 /* wait for the change to finish */
3934                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3935                 if (ret) {
3936                         dev_info(&pf->pdev->dev,
3937                                  "VSI seid %d Rx ring %d %sable timeout\n",
3938                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3939                         break;
3940                 }
3941         }
3942
3943         return ret;
3944 }
3945
3946 /**
3947  * i40e_vsi_control_rings - Start or stop a VSI's rings
3948  * @vsi: the VSI being configured
3949  * @enable: start or stop the rings
3950  **/
3951 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3952 {
3953         int ret = 0;
3954
3955         /* do rx first for enable and last for disable */
3956         if (request) {
3957                 ret = i40e_vsi_control_rx(vsi, request);
3958                 if (ret)
3959                         return ret;
3960                 ret = i40e_vsi_control_tx(vsi, request);
3961         } else {
3962                 /* Ignore return value, we need to shutdown whatever we can */
3963                 i40e_vsi_control_tx(vsi, request);
3964                 i40e_vsi_control_rx(vsi, request);
3965         }
3966
3967         return ret;
3968 }
3969
3970 /**
3971  * i40e_vsi_free_irq - Free the irq association with the OS
3972  * @vsi: the VSI being configured
3973  **/
3974 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3975 {
3976         struct i40e_pf *pf = vsi->back;
3977         struct i40e_hw *hw = &pf->hw;
3978         int base = vsi->base_vector;
3979         u32 val, qp;
3980         int i;
3981
3982         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3983                 if (!vsi->q_vectors)
3984                         return;
3985
3986                 if (!vsi->irqs_ready)
3987                         return;
3988
3989                 vsi->irqs_ready = false;
3990                 for (i = 0; i < vsi->num_q_vectors; i++) {
3991                         u16 vector = i + base;
3992
3993                         /* free only the irqs that were actually requested */
3994                         if (!vsi->q_vectors[i] ||
3995                             !vsi->q_vectors[i]->num_ringpairs)
3996                                 continue;
3997
3998                         /* clear the affinity_mask in the IRQ descriptor */
3999                         irq_set_affinity_hint(pf->msix_entries[vector].vector,
4000                                               NULL);
4001                         free_irq(pf->msix_entries[vector].vector,
4002                                  vsi->q_vectors[i]);
4003
4004                         /* Tear down the interrupt queue link list
4005                          *
4006                          * We know that they come in pairs and always
4007                          * the Rx first, then the Tx.  To clear the
4008                          * link list, stick the EOL value into the
4009                          * next_q field of the registers.
4010                          */
4011                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4012                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4013                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4014                         val |= I40E_QUEUE_END_OF_LIST
4015                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4016                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4017
4018                         while (qp != I40E_QUEUE_END_OF_LIST) {
4019                                 u32 next;
4020
4021                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4022
4023                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4024                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4025                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4026                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4027
4028                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4029                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4030
4031                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4032
4033                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4034
4035                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4036                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4037
4038                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4039                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4040                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4041                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4042
4043                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4044                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4045
4046                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4047                                 qp = next;
4048                         }
4049                 }
4050         } else {
4051                 free_irq(pf->pdev->irq, pf);
4052
4053                 val = rd32(hw, I40E_PFINT_LNKLST0);
4054                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4055                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4056                 val |= I40E_QUEUE_END_OF_LIST
4057                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4058                 wr32(hw, I40E_PFINT_LNKLST0, val);
4059
4060                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4061                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4062                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4063                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4064                          I40E_QINT_RQCTL_INTEVENT_MASK);
4065
4066                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4067                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4068
4069                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4070
4071                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4072
4073                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4074                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4075                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4076                          I40E_QINT_TQCTL_INTEVENT_MASK);
4077
4078                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4079                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4080
4081                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4082         }
4083 }
4084
4085 /**
4086  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4087  * @vsi: the VSI being configured
4088  * @v_idx: Index of vector to be freed
4089  *
4090  * This function frees the memory allocated to the q_vector.  In addition if
4091  * NAPI is enabled it will delete any references to the NAPI struct prior
4092  * to freeing the q_vector.
4093  **/
4094 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4095 {
4096         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4097         struct i40e_ring *ring;
4098
4099         if (!q_vector)
4100                 return;
4101
4102         /* disassociate q_vector from rings */
4103         i40e_for_each_ring(ring, q_vector->tx)
4104                 ring->q_vector = NULL;
4105
4106         i40e_for_each_ring(ring, q_vector->rx)
4107                 ring->q_vector = NULL;
4108
4109         /* only VSI w/ an associated netdev is set up w/ NAPI */
4110         if (vsi->netdev)
4111                 netif_napi_del(&q_vector->napi);
4112
4113         vsi->q_vectors[v_idx] = NULL;
4114
4115         kfree_rcu(q_vector, rcu);
4116 }
4117
4118 /**
4119  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4120  * @vsi: the VSI being un-configured
4121  *
4122  * This frees the memory allocated to the q_vectors and
4123  * deletes references to the NAPI struct.
4124  **/
4125 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4126 {
4127         int v_idx;
4128
4129         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4130                 i40e_free_q_vector(vsi, v_idx);
4131 }
4132
4133 /**
4134  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4135  * @pf: board private structure
4136  **/
4137 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4138 {
4139         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4140         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4141                 pci_disable_msix(pf->pdev);
4142                 kfree(pf->msix_entries);
4143                 pf->msix_entries = NULL;
4144                 kfree(pf->irq_pile);
4145                 pf->irq_pile = NULL;
4146         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4147                 pci_disable_msi(pf->pdev);
4148         }
4149         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4150 }
4151
4152 /**
4153  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4154  * @pf: board private structure
4155  *
4156  * We go through and clear interrupt specific resources and reset the structure
4157  * to pre-load conditions
4158  **/
4159 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4160 {
4161         int i;
4162
4163         i40e_stop_misc_vector(pf);
4164         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4165                 synchronize_irq(pf->msix_entries[0].vector);
4166                 free_irq(pf->msix_entries[0].vector, pf);
4167         }
4168
4169         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4170                       I40E_IWARP_IRQ_PILE_ID);
4171
4172         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4173         for (i = 0; i < pf->num_alloc_vsi; i++)
4174                 if (pf->vsi[i])
4175                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4176         i40e_reset_interrupt_capability(pf);
4177 }
4178
4179 /**
4180  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4181  * @vsi: the VSI being configured
4182  **/
4183 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4184 {
4185         int q_idx;
4186
4187         if (!vsi->netdev)
4188                 return;
4189
4190         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4191                 napi_enable(&vsi->q_vectors[q_idx]->napi);
4192 }
4193
4194 /**
4195  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4196  * @vsi: the VSI being configured
4197  **/
4198 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4199 {
4200         int q_idx;
4201
4202         if (!vsi->netdev)
4203                 return;
4204
4205         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4206                 napi_disable(&vsi->q_vectors[q_idx]->napi);
4207 }
4208
4209 /**
4210  * i40e_vsi_close - Shut down a VSI
4211  * @vsi: the vsi to be quelled
4212  **/
4213 static void i40e_vsi_close(struct i40e_vsi *vsi)
4214 {
4215         bool reset = false;
4216
4217         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4218                 i40e_down(vsi);
4219         i40e_vsi_free_irq(vsi);
4220         i40e_vsi_free_tx_resources(vsi);
4221         i40e_vsi_free_rx_resources(vsi);
4222         vsi->current_netdev_flags = 0;
4223         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4224                 reset = true;
4225         i40e_notify_client_of_netdev_close(vsi, reset);
4226 }
4227
4228 /**
4229  * i40e_quiesce_vsi - Pause a given VSI
4230  * @vsi: the VSI being paused
4231  **/
4232 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4233 {
4234         if (test_bit(__I40E_DOWN, &vsi->state))
4235                 return;
4236
4237         /* No need to disable FCoE VSI when Tx suspended */
4238         if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4239             vsi->type == I40E_VSI_FCOE) {
4240                 dev_dbg(&vsi->back->pdev->dev,
4241                          "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4242                 return;
4243         }
4244
4245         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4246         if (vsi->netdev && netif_running(vsi->netdev))
4247                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4248         else
4249                 i40e_vsi_close(vsi);
4250 }
4251
4252 /**
4253  * i40e_unquiesce_vsi - Resume a given VSI
4254  * @vsi: the VSI being resumed
4255  **/
4256 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4257 {
4258         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4259                 return;
4260
4261         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4262         if (vsi->netdev && netif_running(vsi->netdev))
4263                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4264         else
4265                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4266 }
4267
4268 /**
4269  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4270  * @pf: the PF
4271  **/
4272 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4273 {
4274         int v;
4275
4276         for (v = 0; v < pf->num_alloc_vsi; v++) {
4277                 if (pf->vsi[v])
4278                         i40e_quiesce_vsi(pf->vsi[v]);
4279         }
4280 }
4281
4282 /**
4283  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4284  * @pf: the PF
4285  **/
4286 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4287 {
4288         int v;
4289
4290         for (v = 0; v < pf->num_alloc_vsi; v++) {
4291                 if (pf->vsi[v])
4292                         i40e_unquiesce_vsi(pf->vsi[v]);
4293         }
4294 }
4295
4296 #ifdef CONFIG_I40E_DCB
4297 /**
4298  * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4299  * @vsi: the VSI being configured
4300  *
4301  * This function waits for the given VSI's Tx queues to be disabled.
4302  **/
4303 static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4304 {
4305         struct i40e_pf *pf = vsi->back;
4306         int i, pf_q, ret;
4307
4308         pf_q = vsi->base_queue;
4309         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4310                 /* Check and wait for the disable status of the queue */
4311                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4312                 if (ret) {
4313                         dev_info(&pf->pdev->dev,
4314                                  "VSI seid %d Tx ring %d disable timeout\n",
4315                                  vsi->seid, pf_q);
4316                         return ret;
4317                 }
4318         }
4319
4320         return 0;
4321 }
4322
4323 /**
4324  * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4325  * @pf: the PF
4326  *
4327  * This function waits for the Tx queues to be in disabled state for all the
4328  * VSIs that are managed by this PF.
4329  **/
4330 static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4331 {
4332         int v, ret = 0;
4333
4334         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4335                 /* No need to wait for FCoE VSI queues */
4336                 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4337                         ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4338                         if (ret)
4339                                 break;
4340                 }
4341         }
4342
4343         return ret;
4344 }
4345
4346 #endif
4347
4348 /**
4349  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4350  * @q_idx: TX queue number
4351  * @vsi: Pointer to VSI struct
4352  *
4353  * This function checks specified queue for given VSI. Detects hung condition.
4354  * Sets hung bit since it is two step process. Before next run of service task
4355  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4356  * hung condition remain unchanged and during subsequent run, this function
4357  * issues SW interrupt to recover from hung condition.
4358  **/
4359 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4360 {
4361         struct i40e_ring *tx_ring = NULL;
4362         struct i40e_pf  *pf;
4363         u32 head, val, tx_pending;
4364         int i;
4365
4366         pf = vsi->back;
4367
4368         /* now that we have an index, find the tx_ring struct */
4369         for (i = 0; i < vsi->num_queue_pairs; i++) {
4370                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4371                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4372                                 tx_ring = vsi->tx_rings[i];
4373                                 break;
4374                         }
4375                 }
4376         }
4377
4378         if (!tx_ring)
4379                 return;
4380
4381         /* Read interrupt register */
4382         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4383                 val = rd32(&pf->hw,
4384                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4385                                                tx_ring->vsi->base_vector - 1));
4386         else
4387                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4388
4389         /* Bail out if interrupts are disabled because napi_poll
4390          * execution in-progress or will get scheduled soon.
4391          * napi_poll cleans TX and RX queues and updates 'next_to_clean'.
4392          */
4393         if (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))
4394                 return;
4395
4396         head = i40e_get_head(tx_ring);
4397
4398         tx_pending = i40e_get_tx_pending(tx_ring);
4399
4400         /* HW is done executing descriptors, updated HEAD write back,
4401          * but SW hasn't processed those descriptors. If interrupt is
4402          * not generated from this point ON, it could result into
4403          * dev_watchdog detecting timeout on those netdev_queue,
4404          * hence proactively trigger SW interrupt.
4405          */
4406         if (tx_pending) {
4407                 /* NAPI Poll didn't run and clear since it was set */
4408                 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4409                                        &tx_ring->q_vector->hung_detected)) {
4410                         netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4411                                     vsi->seid, q_idx, tx_pending,
4412                                     tx_ring->next_to_clean, head,
4413                                     tx_ring->next_to_use,
4414                                     readl(tx_ring->tail));
4415                         netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4416                                     vsi->seid, q_idx, val);
4417                         i40e_force_wb(vsi, tx_ring->q_vector);
4418                 } else {
4419                         /* First Chance - detected possible hung */
4420                         set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4421                                 &tx_ring->q_vector->hung_detected);
4422                 }
4423         }
4424 }
4425
4426 /**
4427  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4428  * @pf:  pointer to PF struct
4429  *
4430  * LAN VSI has netdev and netdev has TX queues. This function is to check
4431  * each of those TX queues if they are hung, trigger recovery by issuing
4432  * SW interrupt.
4433  **/
4434 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4435 {
4436         struct net_device *netdev;
4437         struct i40e_vsi *vsi;
4438         int i;
4439
4440         /* Only for LAN VSI */
4441         vsi = pf->vsi[pf->lan_vsi];
4442
4443         if (!vsi)
4444                 return;
4445
4446         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4447         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4448             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4449                 return;
4450
4451         /* Make sure type is MAIN VSI */
4452         if (vsi->type != I40E_VSI_MAIN)
4453                 return;
4454
4455         netdev = vsi->netdev;
4456         if (!netdev)
4457                 return;
4458
4459         /* Bail out if netif_carrier is not OK */
4460         if (!netif_carrier_ok(netdev))
4461                 return;
4462
4463         /* Go thru' TX queues for netdev */
4464         for (i = 0; i < netdev->num_tx_queues; i++) {
4465                 struct netdev_queue *q;
4466
4467                 q = netdev_get_tx_queue(netdev, i);
4468                 if (q)
4469                         i40e_detect_recover_hung_queue(i, vsi);
4470         }
4471 }
4472
4473 /**
4474  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4475  * @pf: pointer to PF
4476  *
4477  * Get TC map for ISCSI PF type that will include iSCSI TC
4478  * and LAN TC.
4479  **/
4480 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4481 {
4482         struct i40e_dcb_app_priority_table app;
4483         struct i40e_hw *hw = &pf->hw;
4484         u8 enabled_tc = 1; /* TC0 is always enabled */
4485         u8 tc, i;
4486         /* Get the iSCSI APP TLV */
4487         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4488
4489         for (i = 0; i < dcbcfg->numapps; i++) {
4490                 app = dcbcfg->app[i];
4491                 if (app.selector == I40E_APP_SEL_TCPIP &&
4492                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4493                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4494                         enabled_tc |= BIT(tc);
4495                         break;
4496                 }
4497         }
4498
4499         return enabled_tc;
4500 }
4501
4502 /**
4503  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4504  * @dcbcfg: the corresponding DCBx configuration structure
4505  *
4506  * Return the number of TCs from given DCBx configuration
4507  **/
4508 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4509 {
4510         u8 num_tc = 0;
4511         int i;
4512
4513         /* Scan the ETS Config Priority Table to find
4514          * traffic class enabled for a given priority
4515          * and use the traffic class index to get the
4516          * number of traffic classes enabled
4517          */
4518         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4519                 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4520                         num_tc = dcbcfg->etscfg.prioritytable[i];
4521         }
4522
4523         /* Traffic class index starts from zero so
4524          * increment to return the actual count
4525          */
4526         return num_tc + 1;
4527 }
4528
4529 /**
4530  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4531  * @dcbcfg: the corresponding DCBx configuration structure
4532  *
4533  * Query the current DCB configuration and return the number of
4534  * traffic classes enabled from the given DCBX config
4535  **/
4536 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4537 {
4538         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4539         u8 enabled_tc = 1;
4540         u8 i;
4541
4542         for (i = 0; i < num_tc; i++)
4543                 enabled_tc |= BIT(i);
4544
4545         return enabled_tc;
4546 }
4547
4548 /**
4549  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4550  * @pf: PF being queried
4551  *
4552  * Return number of traffic classes enabled for the given PF
4553  **/
4554 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4555 {
4556         struct i40e_hw *hw = &pf->hw;
4557         u8 i, enabled_tc;
4558         u8 num_tc = 0;
4559         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4560
4561         /* If DCB is not enabled then always in single TC */
4562         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4563                 return 1;
4564
4565         /* SFP mode will be enabled for all TCs on port */
4566         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4567                 return i40e_dcb_get_num_tc(dcbcfg);
4568
4569         /* MFP mode return count of enabled TCs for this PF */
4570         if (pf->hw.func_caps.iscsi)
4571                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4572         else
4573                 return 1; /* Only TC0 */
4574
4575         /* At least have TC0 */
4576         enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4577         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4578                 if (enabled_tc & BIT(i))
4579                         num_tc++;
4580         }
4581         return num_tc;
4582 }
4583
4584 /**
4585  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4586  * @pf: PF being queried
4587  *
4588  * Return a bitmap for first enabled traffic class for this PF.
4589  **/
4590 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4591 {
4592         u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4593         u8 i = 0;
4594
4595         if (!enabled_tc)
4596                 return 0x1; /* TC0 */
4597
4598         /* Find the first enabled TC */
4599         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4600                 if (enabled_tc & BIT(i))
4601                         break;
4602         }
4603
4604         return BIT(i);
4605 }
4606
4607 /**
4608  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4609  * @pf: PF being queried
4610  *
4611  * Return a bitmap for enabled traffic classes for this PF.
4612  **/
4613 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4614 {
4615         /* If DCB is not enabled for this PF then just return default TC */
4616         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4617                 return i40e_pf_get_default_tc(pf);
4618
4619         /* SFP mode we want PF to be enabled for all TCs */
4620         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4621                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4622
4623         /* MFP enabled and iSCSI PF type */
4624         if (pf->hw.func_caps.iscsi)
4625                 return i40e_get_iscsi_tc_map(pf);
4626         else
4627                 return i40e_pf_get_default_tc(pf);
4628 }
4629
4630 /**
4631  * i40e_vsi_get_bw_info - Query VSI BW Information
4632  * @vsi: the VSI being queried
4633  *
4634  * Returns 0 on success, negative value on failure
4635  **/
4636 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4637 {
4638         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4639         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4640         struct i40e_pf *pf = vsi->back;
4641         struct i40e_hw *hw = &pf->hw;
4642         i40e_status ret;
4643         u32 tc_bw_max;
4644         int i;
4645
4646         /* Get the VSI level BW configuration */
4647         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4648         if (ret) {
4649                 dev_info(&pf->pdev->dev,
4650                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4651                          i40e_stat_str(&pf->hw, ret),
4652                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4653                 return -EINVAL;
4654         }
4655
4656         /* Get the VSI level BW configuration per TC */
4657         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4658                                                NULL);
4659         if (ret) {
4660                 dev_info(&pf->pdev->dev,
4661                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4662                          i40e_stat_str(&pf->hw, ret),
4663                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4664                 return -EINVAL;
4665         }
4666
4667         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4668                 dev_info(&pf->pdev->dev,
4669                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4670                          bw_config.tc_valid_bits,
4671                          bw_ets_config.tc_valid_bits);
4672                 /* Still continuing */
4673         }
4674
4675         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4676         vsi->bw_max_quanta = bw_config.max_bw;
4677         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4678                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4679         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4680                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4681                 vsi->bw_ets_limit_credits[i] =
4682                                         le16_to_cpu(bw_ets_config.credits[i]);
4683                 /* 3 bits out of 4 for each TC */
4684                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4685         }
4686
4687         return 0;
4688 }
4689
4690 /**
4691  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4692  * @vsi: the VSI being configured
4693  * @enabled_tc: TC bitmap
4694  * @bw_credits: BW shared credits per TC
4695  *
4696  * Returns 0 on success, negative value on failure
4697  **/
4698 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4699                                        u8 *bw_share)
4700 {
4701         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4702         i40e_status ret;
4703         int i;
4704
4705         bw_data.tc_valid_bits = enabled_tc;
4706         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4707                 bw_data.tc_bw_credits[i] = bw_share[i];
4708
4709         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4710                                        NULL);
4711         if (ret) {
4712                 dev_info(&vsi->back->pdev->dev,
4713                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4714                          vsi->back->hw.aq.asq_last_status);
4715                 return -EINVAL;
4716         }
4717
4718         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4719                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4720
4721         return 0;
4722 }
4723
4724 /**
4725  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4726  * @vsi: the VSI being configured
4727  * @enabled_tc: TC map to be enabled
4728  *
4729  **/
4730 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4731 {
4732         struct net_device *netdev = vsi->netdev;
4733         struct i40e_pf *pf = vsi->back;
4734         struct i40e_hw *hw = &pf->hw;
4735         u8 netdev_tc = 0;
4736         int i;
4737         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4738
4739         if (!netdev)
4740                 return;
4741
4742         if (!enabled_tc) {
4743                 netdev_reset_tc(netdev);
4744                 return;
4745         }
4746
4747         /* Set up actual enabled TCs on the VSI */
4748         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4749                 return;
4750
4751         /* set per TC queues for the VSI */
4752         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4753                 /* Only set TC queues for enabled tcs
4754                  *
4755                  * e.g. For a VSI that has TC0 and TC3 enabled the
4756                  * enabled_tc bitmap would be 0x00001001; the driver
4757                  * will set the numtc for netdev as 2 that will be
4758                  * referenced by the netdev layer as TC 0 and 1.
4759                  */
4760                 if (vsi->tc_config.enabled_tc & BIT(i))
4761                         netdev_set_tc_queue(netdev,
4762                                         vsi->tc_config.tc_info[i].netdev_tc,
4763                                         vsi->tc_config.tc_info[i].qcount,
4764                                         vsi->tc_config.tc_info[i].qoffset);
4765         }
4766
4767         /* Assign UP2TC map for the VSI */
4768         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4769                 /* Get the actual TC# for the UP */
4770                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4771                 /* Get the mapped netdev TC# for the UP */
4772                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4773                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4774         }
4775 }
4776
4777 /**
4778  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4779  * @vsi: the VSI being configured
4780  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4781  **/
4782 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4783                                       struct i40e_vsi_context *ctxt)
4784 {
4785         /* copy just the sections touched not the entire info
4786          * since not all sections are valid as returned by
4787          * update vsi params
4788          */
4789         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4790         memcpy(&vsi->info.queue_mapping,
4791                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4792         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4793                sizeof(vsi->info.tc_mapping));
4794 }
4795
4796 /**
4797  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4798  * @vsi: VSI to be configured
4799  * @enabled_tc: TC bitmap
4800  *
4801  * This configures a particular VSI for TCs that are mapped to the
4802  * given TC bitmap. It uses default bandwidth share for TCs across
4803  * VSIs to configure TC for a particular VSI.
4804  *
4805  * NOTE:
4806  * It is expected that the VSI queues have been quisced before calling
4807  * this function.
4808  **/
4809 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4810 {
4811         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4812         struct i40e_vsi_context ctxt;
4813         int ret = 0;
4814         int i;
4815
4816         /* Check if enabled_tc is same as existing or new TCs */
4817         if (vsi->tc_config.enabled_tc == enabled_tc)
4818                 return ret;
4819
4820         /* Enable ETS TCs with equal BW Share for now across all VSIs */
4821         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4822                 if (enabled_tc & BIT(i))
4823                         bw_share[i] = 1;
4824         }
4825
4826         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4827         if (ret) {
4828                 dev_info(&vsi->back->pdev->dev,
4829                          "Failed configuring TC map %d for VSI %d\n",
4830                          enabled_tc, vsi->seid);
4831                 goto out;
4832         }
4833
4834         /* Update Queue Pairs Mapping for currently enabled UPs */
4835         ctxt.seid = vsi->seid;
4836         ctxt.pf_num = vsi->back->hw.pf_id;
4837         ctxt.vf_num = 0;
4838         ctxt.uplink_seid = vsi->uplink_seid;
4839         ctxt.info = vsi->info;
4840         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4841
4842         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4843                 ctxt.info.valid_sections |=
4844                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4845                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4846         }
4847
4848         /* Update the VSI after updating the VSI queue-mapping information */
4849         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4850         if (ret) {
4851                 dev_info(&vsi->back->pdev->dev,
4852                          "Update vsi tc config failed, err %s aq_err %s\n",
4853                          i40e_stat_str(&vsi->back->hw, ret),
4854                          i40e_aq_str(&vsi->back->hw,
4855                                      vsi->back->hw.aq.asq_last_status));
4856                 goto out;
4857         }
4858         /* update the local VSI info with updated queue map */
4859         i40e_vsi_update_queue_map(vsi, &ctxt);
4860         vsi->info.valid_sections = 0;
4861
4862         /* Update current VSI BW information */
4863         ret = i40e_vsi_get_bw_info(vsi);
4864         if (ret) {
4865                 dev_info(&vsi->back->pdev->dev,
4866                          "Failed updating vsi bw info, err %s aq_err %s\n",
4867                          i40e_stat_str(&vsi->back->hw, ret),
4868                          i40e_aq_str(&vsi->back->hw,
4869                                      vsi->back->hw.aq.asq_last_status));
4870                 goto out;
4871         }
4872
4873         /* Update the netdev TC setup */
4874         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4875 out:
4876         return ret;
4877 }
4878
4879 /**
4880  * i40e_veb_config_tc - Configure TCs for given VEB
4881  * @veb: given VEB
4882  * @enabled_tc: TC bitmap
4883  *
4884  * Configures given TC bitmap for VEB (switching) element
4885  **/
4886 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4887 {
4888         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4889         struct i40e_pf *pf = veb->pf;
4890         int ret = 0;
4891         int i;
4892
4893         /* No TCs or already enabled TCs just return */
4894         if (!enabled_tc || veb->enabled_tc == enabled_tc)
4895                 return ret;
4896
4897         bw_data.tc_valid_bits = enabled_tc;
4898         /* bw_data.absolute_credits is not set (relative) */
4899
4900         /* Enable ETS TCs with equal BW Share for now */
4901         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4902                 if (enabled_tc & BIT(i))
4903                         bw_data.tc_bw_share_credits[i] = 1;
4904         }
4905
4906         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4907                                                    &bw_data, NULL);
4908         if (ret) {
4909                 dev_info(&pf->pdev->dev,
4910                          "VEB bw config failed, err %s aq_err %s\n",
4911                          i40e_stat_str(&pf->hw, ret),
4912                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4913                 goto out;
4914         }
4915
4916         /* Update the BW information */
4917         ret = i40e_veb_get_bw_info(veb);
4918         if (ret) {
4919                 dev_info(&pf->pdev->dev,
4920                          "Failed getting veb bw config, err %s aq_err %s\n",
4921                          i40e_stat_str(&pf->hw, ret),
4922                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4923         }
4924
4925 out:
4926         return ret;
4927 }
4928
4929 #ifdef CONFIG_I40E_DCB
4930 /**
4931  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4932  * @pf: PF struct
4933  *
4934  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4935  * the caller would've quiesce all the VSIs before calling
4936  * this function
4937  **/
4938 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4939 {
4940         u8 tc_map = 0;
4941         int ret;
4942         u8 v;
4943
4944         /* Enable the TCs available on PF to all VEBs */
4945         tc_map = i40e_pf_get_tc_map(pf);
4946         for (v = 0; v < I40E_MAX_VEB; v++) {
4947                 if (!pf->veb[v])
4948                         continue;
4949                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4950                 if (ret) {
4951                         dev_info(&pf->pdev->dev,
4952                                  "Failed configuring TC for VEB seid=%d\n",
4953                                  pf->veb[v]->seid);
4954                         /* Will try to configure as many components */
4955                 }
4956         }
4957
4958         /* Update each VSI */
4959         for (v = 0; v < pf->num_alloc_vsi; v++) {
4960                 if (!pf->vsi[v])
4961                         continue;
4962
4963                 /* - Enable all TCs for the LAN VSI
4964 #ifdef I40E_FCOE
4965                  * - For FCoE VSI only enable the TC configured
4966                  *   as per the APP TLV
4967 #endif
4968                  * - For all others keep them at TC0 for now
4969                  */
4970                 if (v == pf->lan_vsi)
4971                         tc_map = i40e_pf_get_tc_map(pf);
4972                 else
4973                         tc_map = i40e_pf_get_default_tc(pf);
4974 #ifdef I40E_FCOE
4975                 if (pf->vsi[v]->type == I40E_VSI_FCOE)
4976                         tc_map = i40e_get_fcoe_tc_map(pf);
4977 #endif /* #ifdef I40E_FCOE */
4978
4979                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4980                 if (ret) {
4981                         dev_info(&pf->pdev->dev,
4982                                  "Failed configuring TC for VSI seid=%d\n",
4983                                  pf->vsi[v]->seid);
4984                         /* Will try to configure as many components */
4985                 } else {
4986                         /* Re-configure VSI vectors based on updated TC map */
4987                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4988                         if (pf->vsi[v]->netdev)
4989                                 i40e_dcbnl_set_all(pf->vsi[v]);
4990                 }
4991                 i40e_notify_client_of_l2_param_changes(pf->vsi[v]);
4992         }
4993 }
4994
4995 /**
4996  * i40e_resume_port_tx - Resume port Tx
4997  * @pf: PF struct
4998  *
4999  * Resume a port's Tx and issue a PF reset in case of failure to
5000  * resume.
5001  **/
5002 static int i40e_resume_port_tx(struct i40e_pf *pf)
5003 {
5004         struct i40e_hw *hw = &pf->hw;
5005         int ret;
5006
5007         ret = i40e_aq_resume_port_tx(hw, NULL);
5008         if (ret) {
5009                 dev_info(&pf->pdev->dev,
5010                          "Resume Port Tx failed, err %s aq_err %s\n",
5011                           i40e_stat_str(&pf->hw, ret),
5012                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5013                 /* Schedule PF reset to recover */
5014                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5015                 i40e_service_event_schedule(pf);
5016         }
5017
5018         return ret;
5019 }
5020
5021 /**
5022  * i40e_init_pf_dcb - Initialize DCB configuration
5023  * @pf: PF being configured
5024  *
5025  * Query the current DCB configuration and cache it
5026  * in the hardware structure
5027  **/
5028 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5029 {
5030         struct i40e_hw *hw = &pf->hw;
5031         int err = 0;
5032
5033         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5034         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
5035             (pf->hw.aq.fw_maj_ver < 4))
5036                 goto out;
5037
5038         /* Get the initial DCB configuration */
5039         err = i40e_init_dcb(hw);
5040         if (!err) {
5041                 /* Device/Function is not DCBX capable */
5042                 if ((!hw->func_caps.dcb) ||
5043                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5044                         dev_info(&pf->pdev->dev,
5045                                  "DCBX offload is not supported or is disabled for this PF.\n");
5046
5047                         if (pf->flags & I40E_FLAG_MFP_ENABLED)
5048                                 goto out;
5049
5050                 } else {
5051                         /* When status is not DISABLED then DCBX in FW */
5052                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5053                                        DCB_CAP_DCBX_VER_IEEE;
5054
5055                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5056                         /* Enable DCB tagging only when more than one TC */
5057                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5058                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5059                         dev_dbg(&pf->pdev->dev,
5060                                 "DCBX offload is supported for this PF.\n");
5061                 }
5062         } else {
5063                 dev_info(&pf->pdev->dev,
5064                          "Query for DCB configuration failed, err %s aq_err %s\n",
5065                          i40e_stat_str(&pf->hw, err),
5066                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5067         }
5068
5069 out:
5070         return err;
5071 }
5072 #endif /* CONFIG_I40E_DCB */
5073 #define SPEED_SIZE 14
5074 #define FC_SIZE 8
5075 /**
5076  * i40e_print_link_message - print link up or down
5077  * @vsi: the VSI for which link needs a message
5078  */
5079 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5080 {
5081         char *speed = "Unknown";
5082         char *fc = "Unknown";
5083
5084         if (vsi->current_isup == isup)
5085                 return;
5086         vsi->current_isup = isup;
5087         if (!isup) {
5088                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5089                 return;
5090         }
5091
5092         /* Warn user if link speed on NPAR enabled partition is not at
5093          * least 10GB
5094          */
5095         if (vsi->back->hw.func_caps.npar_enable &&
5096             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5097              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5098                 netdev_warn(vsi->netdev,
5099                             "The partition detected link speed that is less than 10Gbps\n");
5100
5101         switch (vsi->back->hw.phy.link_info.link_speed) {
5102         case I40E_LINK_SPEED_40GB:
5103                 speed = "40 G";
5104                 break;
5105         case I40E_LINK_SPEED_20GB:
5106                 speed = "20 G";
5107                 break;
5108         case I40E_LINK_SPEED_10GB:
5109                 speed = "10 G";
5110                 break;
5111         case I40E_LINK_SPEED_1GB:
5112                 speed = "1000 M";
5113                 break;
5114         case I40E_LINK_SPEED_100MB:
5115                 speed = "100 M";
5116                 break;
5117         default:
5118                 break;
5119         }
5120
5121         switch (vsi->back->hw.fc.current_mode) {
5122         case I40E_FC_FULL:
5123                 fc = "RX/TX";
5124                 break;
5125         case I40E_FC_TX_PAUSE:
5126                 fc = "TX";
5127                 break;
5128         case I40E_FC_RX_PAUSE:
5129                 fc = "RX";
5130                 break;
5131         default:
5132                 fc = "None";
5133                 break;
5134         }
5135
5136         netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5137                     speed, fc);
5138 }
5139
5140 /**
5141  * i40e_up_complete - Finish the last steps of bringing up a connection
5142  * @vsi: the VSI being configured
5143  **/
5144 static int i40e_up_complete(struct i40e_vsi *vsi)
5145 {
5146         struct i40e_pf *pf = vsi->back;
5147         int err;
5148
5149         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5150                 i40e_vsi_configure_msix(vsi);
5151         else
5152                 i40e_configure_msi_and_legacy(vsi);
5153
5154         /* start rings */
5155         err = i40e_vsi_control_rings(vsi, true);
5156         if (err)
5157                 return err;
5158
5159         clear_bit(__I40E_DOWN, &vsi->state);
5160         i40e_napi_enable_all(vsi);
5161         i40e_vsi_enable_irq(vsi);
5162
5163         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5164             (vsi->netdev)) {
5165                 i40e_print_link_message(vsi, true);
5166                 netif_tx_start_all_queues(vsi->netdev);
5167                 netif_carrier_on(vsi->netdev);
5168         } else if (vsi->netdev) {
5169                 i40e_print_link_message(vsi, false);
5170                 /* need to check for qualified module here*/
5171                 if ((pf->hw.phy.link_info.link_info &
5172                         I40E_AQ_MEDIA_AVAILABLE) &&
5173                     (!(pf->hw.phy.link_info.an_info &
5174                         I40E_AQ_QUALIFIED_MODULE)))
5175                         netdev_err(vsi->netdev,
5176                                    "the driver failed to link because an unqualified module was detected.");
5177         }
5178
5179         /* replay FDIR SB filters */
5180         if (vsi->type == I40E_VSI_FDIR) {
5181                 /* reset fd counters */
5182                 pf->fd_add_err = pf->fd_atr_cnt = 0;
5183                 if (pf->fd_tcp_rule > 0) {
5184                         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5185                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5186                                 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5187                         pf->fd_tcp_rule = 0;
5188                 }
5189                 i40e_fdir_filter_restore(vsi);
5190         }
5191
5192         /* On the next run of the service_task, notify any clients of the new
5193          * opened netdev
5194          */
5195         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5196         i40e_service_event_schedule(pf);
5197
5198         return 0;
5199 }
5200
5201 /**
5202  * i40e_vsi_reinit_locked - Reset the VSI
5203  * @vsi: the VSI being configured
5204  *
5205  * Rebuild the ring structs after some configuration
5206  * has changed, e.g. MTU size.
5207  **/
5208 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5209 {
5210         struct i40e_pf *pf = vsi->back;
5211
5212         WARN_ON(in_interrupt());
5213         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5214                 usleep_range(1000, 2000);
5215         i40e_down(vsi);
5216
5217         /* Give a VF some time to respond to the reset.  The
5218          * two second wait is based upon the watchdog cycle in
5219          * the VF driver.
5220          */
5221         if (vsi->type == I40E_VSI_SRIOV)
5222                 msleep(2000);
5223         i40e_up(vsi);
5224         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5225 }
5226
5227 /**
5228  * i40e_up - Bring the connection back up after being down
5229  * @vsi: the VSI being configured
5230  **/
5231 int i40e_up(struct i40e_vsi *vsi)
5232 {
5233         int err;
5234
5235         err = i40e_vsi_configure(vsi);
5236         if (!err)
5237                 err = i40e_up_complete(vsi);
5238
5239         return err;
5240 }
5241
5242 /**
5243  * i40e_down - Shutdown the connection processing
5244  * @vsi: the VSI being stopped
5245  **/
5246 void i40e_down(struct i40e_vsi *vsi)
5247 {
5248         int i;
5249
5250         /* It is assumed that the caller of this function
5251          * sets the vsi->state __I40E_DOWN bit.
5252          */
5253         if (vsi->netdev) {
5254                 netif_carrier_off(vsi->netdev);
5255                 netif_tx_disable(vsi->netdev);
5256         }
5257         i40e_vsi_disable_irq(vsi);
5258         i40e_vsi_control_rings(vsi, false);
5259         i40e_napi_disable_all(vsi);
5260
5261         for (i = 0; i < vsi->num_queue_pairs; i++) {
5262                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5263                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5264         }
5265 }
5266
5267 /**
5268  * i40e_setup_tc - configure multiple traffic classes
5269  * @netdev: net device to configure
5270  * @tc: number of traffic classes to enable
5271  **/
5272 #ifdef I40E_FCOE
5273 int i40e_setup_tc(struct net_device *netdev, u8 tc)
5274 #else
5275 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5276 #endif
5277 {
5278         struct i40e_netdev_priv *np = netdev_priv(netdev);
5279         struct i40e_vsi *vsi = np->vsi;
5280         struct i40e_pf *pf = vsi->back;
5281         u8 enabled_tc = 0;
5282         int ret = -EINVAL;
5283         int i;
5284
5285         /* Check if DCB enabled to continue */
5286         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5287                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5288                 goto exit;
5289         }
5290
5291         /* Check if MFP enabled */
5292         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5293                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5294                 goto exit;
5295         }
5296
5297         /* Check whether tc count is within enabled limit */
5298         if (tc > i40e_pf_get_num_tc(pf)) {
5299                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5300                 goto exit;
5301         }
5302
5303         /* Generate TC map for number of tc requested */
5304         for (i = 0; i < tc; i++)
5305                 enabled_tc |= BIT(i);
5306
5307         /* Requesting same TC configuration as already enabled */
5308         if (enabled_tc == vsi->tc_config.enabled_tc)
5309                 return 0;
5310
5311         /* Quiesce VSI queues */
5312         i40e_quiesce_vsi(vsi);
5313
5314         /* Configure VSI for enabled TCs */
5315         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5316         if (ret) {
5317                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5318                             vsi->seid);
5319                 goto exit;
5320         }
5321
5322         /* Unquiesce VSI */
5323         i40e_unquiesce_vsi(vsi);
5324
5325 exit:
5326         return ret;
5327 }
5328
5329 /**
5330  * i40e_open - Called when a network interface is made active
5331  * @netdev: network interface device structure
5332  *
5333  * The open entry point is called when a network interface is made
5334  * active by the system (IFF_UP).  At this point all resources needed
5335  * for transmit and receive operations are allocated, the interrupt
5336  * handler is registered with the OS, the netdev watchdog subtask is
5337  * enabled, and the stack is notified that the interface is ready.
5338  *
5339  * Returns 0 on success, negative value on failure
5340  **/
5341 int i40e_open(struct net_device *netdev)
5342 {
5343         struct i40e_netdev_priv *np = netdev_priv(netdev);
5344         struct i40e_vsi *vsi = np->vsi;
5345         struct i40e_pf *pf = vsi->back;
5346         int err;
5347
5348         /* disallow open during test or if eeprom is broken */
5349         if (test_bit(__I40E_TESTING, &pf->state) ||
5350             test_bit(__I40E_BAD_EEPROM, &pf->state))
5351                 return -EBUSY;
5352
5353         netif_carrier_off(netdev);
5354
5355         err = i40e_vsi_open(vsi);
5356         if (err)
5357                 return err;
5358
5359         /* configure global TSO hardware offload settings */
5360         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5361                                                        TCP_FLAG_FIN) >> 16);
5362         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5363                                                        TCP_FLAG_FIN |
5364                                                        TCP_FLAG_CWR) >> 16);
5365         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5366
5367 #ifdef CONFIG_I40E_VXLAN
5368         vxlan_get_rx_port(netdev);
5369 #endif
5370 #ifdef CONFIG_I40E_GENEVE
5371         geneve_get_rx_port(netdev);
5372 #endif
5373
5374         i40e_notify_client_of_netdev_open(vsi);
5375
5376         return 0;
5377 }
5378
5379 /**
5380  * i40e_vsi_open -
5381  * @vsi: the VSI to open
5382  *
5383  * Finish initialization of the VSI.
5384  *
5385  * Returns 0 on success, negative value on failure
5386  **/
5387 int i40e_vsi_open(struct i40e_vsi *vsi)
5388 {
5389         struct i40e_pf *pf = vsi->back;
5390         char int_name[I40E_INT_NAME_STR_LEN];
5391         int err;
5392
5393         /* allocate descriptors */
5394         err = i40e_vsi_setup_tx_resources(vsi);
5395         if (err)
5396                 goto err_setup_tx;
5397         err = i40e_vsi_setup_rx_resources(vsi);
5398         if (err)
5399                 goto err_setup_rx;
5400
5401         err = i40e_vsi_configure(vsi);
5402         if (err)
5403                 goto err_setup_rx;
5404
5405         if (vsi->netdev) {
5406                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5407                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5408                 err = i40e_vsi_request_irq(vsi, int_name);
5409                 if (err)
5410                         goto err_setup_rx;
5411
5412                 /* Notify the stack of the actual queue counts. */
5413                 err = netif_set_real_num_tx_queues(vsi->netdev,
5414                                                    vsi->num_queue_pairs);
5415                 if (err)
5416                         goto err_set_queues;
5417
5418                 err = netif_set_real_num_rx_queues(vsi->netdev,
5419                                                    vsi->num_queue_pairs);
5420                 if (err)
5421                         goto err_set_queues;
5422
5423         } else if (vsi->type == I40E_VSI_FDIR) {
5424                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5425                          dev_driver_string(&pf->pdev->dev),
5426                          dev_name(&pf->pdev->dev));
5427                 err = i40e_vsi_request_irq(vsi, int_name);
5428
5429         } else {
5430                 err = -EINVAL;
5431                 goto err_setup_rx;
5432         }
5433
5434         err = i40e_up_complete(vsi);
5435         if (err)
5436                 goto err_up_complete;
5437
5438         return 0;
5439
5440 err_up_complete:
5441         i40e_down(vsi);
5442 err_set_queues:
5443         i40e_vsi_free_irq(vsi);
5444 err_setup_rx:
5445         i40e_vsi_free_rx_resources(vsi);
5446 err_setup_tx:
5447         i40e_vsi_free_tx_resources(vsi);
5448         if (vsi == pf->vsi[pf->lan_vsi])
5449                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5450
5451         return err;
5452 }
5453
5454 /**
5455  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5456  * @pf: Pointer to PF
5457  *
5458  * This function destroys the hlist where all the Flow Director
5459  * filters were saved.
5460  **/
5461 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5462 {
5463         struct i40e_fdir_filter *filter;
5464         struct hlist_node *node2;
5465
5466         hlist_for_each_entry_safe(filter, node2,
5467                                   &pf->fdir_filter_list, fdir_node) {
5468                 hlist_del(&filter->fdir_node);
5469                 kfree(filter);
5470         }
5471         pf->fdir_pf_active_filters = 0;
5472 }
5473
5474 /**
5475  * i40e_close - Disables a network interface
5476  * @netdev: network interface device structure
5477  *
5478  * The close entry point is called when an interface is de-activated
5479  * by the OS.  The hardware is still under the driver's control, but
5480  * this netdev interface is disabled.
5481  *
5482  * Returns 0, this is not allowed to fail
5483  **/
5484 #ifdef I40E_FCOE
5485 int i40e_close(struct net_device *netdev)
5486 #else
5487 static int i40e_close(struct net_device *netdev)
5488 #endif
5489 {
5490         struct i40e_netdev_priv *np = netdev_priv(netdev);
5491         struct i40e_vsi *vsi = np->vsi;
5492
5493         i40e_vsi_close(vsi);
5494
5495         return 0;
5496 }
5497
5498 /**
5499  * i40e_do_reset - Start a PF or Core Reset sequence
5500  * @pf: board private structure
5501  * @reset_flags: which reset is requested
5502  *
5503  * The essential difference in resets is that the PF Reset
5504  * doesn't clear the packet buffers, doesn't reset the PE
5505  * firmware, and doesn't bother the other PFs on the chip.
5506  **/
5507 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5508 {
5509         u32 val;
5510
5511         WARN_ON(in_interrupt());
5512
5513         if (i40e_check_asq_alive(&pf->hw))
5514                 i40e_vc_notify_reset(pf);
5515
5516         /* do the biggest reset indicated */
5517         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5518
5519                 /* Request a Global Reset
5520                  *
5521                  * This will start the chip's countdown to the actual full
5522                  * chip reset event, and a warning interrupt to be sent
5523                  * to all PFs, including the requestor.  Our handler
5524                  * for the warning interrupt will deal with the shutdown
5525                  * and recovery of the switch setup.
5526                  */
5527                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5528                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5529                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5530                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5531
5532         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5533
5534                 /* Request a Core Reset
5535                  *
5536                  * Same as Global Reset, except does *not* include the MAC/PHY
5537                  */
5538                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5539                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5540                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5541                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5542                 i40e_flush(&pf->hw);
5543
5544         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5545
5546                 /* Request a PF Reset
5547                  *
5548                  * Resets only the PF-specific registers
5549                  *
5550                  * This goes directly to the tear-down and rebuild of
5551                  * the switch, since we need to do all the recovery as
5552                  * for the Core Reset.
5553                  */
5554                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5555                 i40e_handle_reset_warning(pf);
5556
5557         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5558                 int v;
5559
5560                 /* Find the VSI(s) that requested a re-init */
5561                 dev_info(&pf->pdev->dev,
5562                          "VSI reinit requested\n");
5563                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5564                         struct i40e_vsi *vsi = pf->vsi[v];
5565
5566                         if (vsi != NULL &&
5567                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5568                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5569                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5570                         }
5571                 }
5572         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5573                 int v;
5574
5575                 /* Find the VSI(s) that needs to be brought down */
5576                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5577                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5578                         struct i40e_vsi *vsi = pf->vsi[v];
5579
5580                         if (vsi != NULL &&
5581                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5582                                 set_bit(__I40E_DOWN, &vsi->state);
5583                                 i40e_down(vsi);
5584                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5585                         }
5586                 }
5587         } else {
5588                 dev_info(&pf->pdev->dev,
5589                          "bad reset request 0x%08x\n", reset_flags);
5590         }
5591 }
5592
5593 #ifdef CONFIG_I40E_DCB
5594 /**
5595  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5596  * @pf: board private structure
5597  * @old_cfg: current DCB config
5598  * @new_cfg: new DCB config
5599  **/
5600 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5601                             struct i40e_dcbx_config *old_cfg,
5602                             struct i40e_dcbx_config *new_cfg)
5603 {
5604         bool need_reconfig = false;
5605
5606         /* Check if ETS configuration has changed */
5607         if (memcmp(&new_cfg->etscfg,
5608                    &old_cfg->etscfg,
5609                    sizeof(new_cfg->etscfg))) {
5610                 /* If Priority Table has changed reconfig is needed */
5611                 if (memcmp(&new_cfg->etscfg.prioritytable,
5612                            &old_cfg->etscfg.prioritytable,
5613                            sizeof(new_cfg->etscfg.prioritytable))) {
5614                         need_reconfig = true;
5615                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5616                 }
5617
5618                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5619                            &old_cfg->etscfg.tcbwtable,
5620                            sizeof(new_cfg->etscfg.tcbwtable)))
5621                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5622
5623                 if (memcmp(&new_cfg->etscfg.tsatable,
5624                            &old_cfg->etscfg.tsatable,
5625                            sizeof(new_cfg->etscfg.tsatable)))
5626                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5627         }
5628
5629         /* Check if PFC configuration has changed */
5630         if (memcmp(&new_cfg->pfc,
5631                    &old_cfg->pfc,
5632                    sizeof(new_cfg->pfc))) {
5633                 need_reconfig = true;
5634                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5635         }
5636
5637         /* Check if APP Table has changed */
5638         if (memcmp(&new_cfg->app,
5639                    &old_cfg->app,
5640                    sizeof(new_cfg->app))) {
5641                 need_reconfig = true;
5642                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5643         }
5644
5645         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5646         return need_reconfig;
5647 }
5648
5649 /**
5650  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5651  * @pf: board private structure
5652  * @e: event info posted on ARQ
5653  **/
5654 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5655                                   struct i40e_arq_event_info *e)
5656 {
5657         struct i40e_aqc_lldp_get_mib *mib =
5658                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5659         struct i40e_hw *hw = &pf->hw;
5660         struct i40e_dcbx_config tmp_dcbx_cfg;
5661         bool need_reconfig = false;
5662         int ret = 0;
5663         u8 type;
5664
5665         /* Not DCB capable or capability disabled */
5666         if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5667                 return ret;
5668
5669         /* Ignore if event is not for Nearest Bridge */
5670         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5671                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5672         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5673         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5674                 return ret;
5675
5676         /* Check MIB Type and return if event for Remote MIB update */
5677         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5678         dev_dbg(&pf->pdev->dev,
5679                 "LLDP event mib type %s\n", type ? "remote" : "local");
5680         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5681                 /* Update the remote cached instance and return */
5682                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5683                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5684                                 &hw->remote_dcbx_config);
5685                 goto exit;
5686         }
5687
5688         /* Store the old configuration */
5689         tmp_dcbx_cfg = hw->local_dcbx_config;
5690
5691         /* Reset the old DCBx configuration data */
5692         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5693         /* Get updated DCBX data from firmware */
5694         ret = i40e_get_dcb_config(&pf->hw);
5695         if (ret) {
5696                 dev_info(&pf->pdev->dev,
5697                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5698                          i40e_stat_str(&pf->hw, ret),
5699                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5700                 goto exit;
5701         }
5702
5703         /* No change detected in DCBX configs */
5704         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5705                     sizeof(tmp_dcbx_cfg))) {
5706                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5707                 goto exit;
5708         }
5709
5710         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5711                                                &hw->local_dcbx_config);
5712
5713         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5714
5715         if (!need_reconfig)
5716                 goto exit;
5717
5718         /* Enable DCB tagging only when more than one TC */
5719         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5720                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5721         else
5722                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5723
5724         set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5725         /* Reconfiguration needed quiesce all VSIs */
5726         i40e_pf_quiesce_all_vsi(pf);
5727
5728         /* Changes in configuration update VEB/VSI */
5729         i40e_dcb_reconfigure(pf);
5730
5731         ret = i40e_resume_port_tx(pf);
5732
5733         clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5734         /* In case of error no point in resuming VSIs */
5735         if (ret)
5736                 goto exit;
5737
5738         /* Wait for the PF's Tx queues to be disabled */
5739         ret = i40e_pf_wait_txq_disabled(pf);
5740         if (ret) {
5741                 /* Schedule PF reset to recover */
5742                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5743                 i40e_service_event_schedule(pf);
5744         } else {
5745                 i40e_pf_unquiesce_all_vsi(pf);
5746         }
5747
5748 exit:
5749         return ret;
5750 }
5751 #endif /* CONFIG_I40E_DCB */
5752
5753 /**
5754  * i40e_do_reset_safe - Protected reset path for userland calls.
5755  * @pf: board private structure
5756  * @reset_flags: which reset is requested
5757  *
5758  **/
5759 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5760 {
5761         rtnl_lock();
5762         i40e_do_reset(pf, reset_flags);
5763         rtnl_unlock();
5764 }
5765
5766 /**
5767  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5768  * @pf: board private structure
5769  * @e: event info posted on ARQ
5770  *
5771  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5772  * and VF queues
5773  **/
5774 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5775                                            struct i40e_arq_event_info *e)
5776 {
5777         struct i40e_aqc_lan_overflow *data =
5778                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5779         u32 queue = le32_to_cpu(data->prtdcb_rupto);
5780         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5781         struct i40e_hw *hw = &pf->hw;
5782         struct i40e_vf *vf;
5783         u16 vf_id;
5784
5785         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5786                 queue, qtx_ctl);
5787
5788         /* Queue belongs to VF, find the VF and issue VF reset */
5789         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5790             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5791                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5792                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5793                 vf_id -= hw->func_caps.vf_base_id;
5794                 vf = &pf->vf[vf_id];
5795                 i40e_vc_notify_vf_reset(vf);
5796                 /* Allow VF to process pending reset notification */
5797                 msleep(20);
5798                 i40e_reset_vf(vf, false);
5799         }
5800 }
5801
5802 /**
5803  * i40e_service_event_complete - Finish up the service event
5804  * @pf: board private structure
5805  **/
5806 static void i40e_service_event_complete(struct i40e_pf *pf)
5807 {
5808         WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5809
5810         /* flush memory to make sure state is correct before next watchog */
5811         smp_mb__before_atomic();
5812         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5813 }
5814
5815 /**
5816  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5817  * @pf: board private structure
5818  **/
5819 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5820 {
5821         u32 val, fcnt_prog;
5822
5823         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5824         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5825         return fcnt_prog;
5826 }
5827
5828 /**
5829  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5830  * @pf: board private structure
5831  **/
5832 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5833 {
5834         u32 val, fcnt_prog;
5835
5836         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5837         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5838                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5839                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5840         return fcnt_prog;
5841 }
5842
5843 /**
5844  * i40e_get_global_fd_count - Get total FD filters programmed on device
5845  * @pf: board private structure
5846  **/
5847 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5848 {
5849         u32 val, fcnt_prog;
5850
5851         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5852         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5853                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5854                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5855         return fcnt_prog;
5856 }
5857
5858 /**
5859  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5860  * @pf: board private structure
5861  **/
5862 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5863 {
5864         struct i40e_fdir_filter *filter;
5865         u32 fcnt_prog, fcnt_avail;
5866         struct hlist_node *node;
5867
5868         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5869                 return;
5870
5871         /* Check if, FD SB or ATR was auto disabled and if there is enough room
5872          * to re-enable
5873          */
5874         fcnt_prog = i40e_get_global_fd_count(pf);
5875         fcnt_avail = pf->fdir_pf_filter_count;
5876         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5877             (pf->fd_add_err == 0) ||
5878             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5879                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5880                     (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5881                         pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5882                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5883                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5884                 }
5885         }
5886         /* Wait for some more space to be available to turn on ATR */
5887         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5888                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5889                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5890                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5891                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5892                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5893                 }
5894         }
5895
5896         /* if hw had a problem adding a filter, delete it */
5897         if (pf->fd_inv > 0) {
5898                 hlist_for_each_entry_safe(filter, node,
5899                                           &pf->fdir_filter_list, fdir_node) {
5900                         if (filter->fd_id == pf->fd_inv) {
5901                                 hlist_del(&filter->fdir_node);
5902                                 kfree(filter);
5903                                 pf->fdir_pf_active_filters--;
5904                         }
5905                 }
5906         }
5907 }
5908
5909 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5910 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5911 /**
5912  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5913  * @pf: board private structure
5914  **/
5915 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5916 {
5917         unsigned long min_flush_time;
5918         int flush_wait_retry = 50;
5919         bool disable_atr = false;
5920         int fd_room;
5921         int reg;
5922
5923         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5924                 return;
5925
5926         if (!time_after(jiffies, pf->fd_flush_timestamp +
5927                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5928                 return;
5929
5930         /* If the flush is happening too quick and we have mostly SB rules we
5931          * should not re-enable ATR for some time.
5932          */
5933         min_flush_time = pf->fd_flush_timestamp +
5934                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5935         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5936
5937         if (!(time_after(jiffies, min_flush_time)) &&
5938             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5939                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5940                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5941                 disable_atr = true;
5942         }
5943
5944         pf->fd_flush_timestamp = jiffies;
5945         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5946         /* flush all filters */
5947         wr32(&pf->hw, I40E_PFQF_CTL_1,
5948              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5949         i40e_flush(&pf->hw);
5950         pf->fd_flush_cnt++;
5951         pf->fd_add_err = 0;
5952         do {
5953                 /* Check FD flush status every 5-6msec */
5954                 usleep_range(5000, 6000);
5955                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5956                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5957                         break;
5958         } while (flush_wait_retry--);
5959         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5960                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5961         } else {
5962                 /* replay sideband filters */
5963                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5964                 if (!disable_atr)
5965                         pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5966                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5967                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5968                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5969         }
5970
5971 }
5972
5973 /**
5974  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5975  * @pf: board private structure
5976  **/
5977 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5978 {
5979         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5980 }
5981
5982 /* We can see up to 256 filter programming desc in transit if the filters are
5983  * being applied really fast; before we see the first
5984  * filter miss error on Rx queue 0. Accumulating enough error messages before
5985  * reacting will make sure we don't cause flush too often.
5986  */
5987 #define I40E_MAX_FD_PROGRAM_ERROR 256
5988
5989 /**
5990  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5991  * @pf: board private structure
5992  **/
5993 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5994 {
5995
5996         /* if interface is down do nothing */
5997         if (test_bit(__I40E_DOWN, &pf->state))
5998                 return;
5999
6000         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
6001                 return;
6002
6003         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6004                 i40e_fdir_flush_and_replay(pf);
6005
6006         i40e_fdir_check_and_reenable(pf);
6007
6008 }
6009
6010 /**
6011  * i40e_vsi_link_event - notify VSI of a link event
6012  * @vsi: vsi to be notified
6013  * @link_up: link up or down
6014  **/
6015 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6016 {
6017         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6018                 return;
6019
6020         switch (vsi->type) {
6021         case I40E_VSI_MAIN:
6022 #ifdef I40E_FCOE
6023         case I40E_VSI_FCOE:
6024 #endif
6025                 if (!vsi->netdev || !vsi->netdev_registered)
6026                         break;
6027
6028                 if (link_up) {
6029                         netif_carrier_on(vsi->netdev);
6030                         netif_tx_wake_all_queues(vsi->netdev);
6031                 } else {
6032                         netif_carrier_off(vsi->netdev);
6033                         netif_tx_stop_all_queues(vsi->netdev);
6034                 }
6035                 break;
6036
6037         case I40E_VSI_SRIOV:
6038         case I40E_VSI_VMDQ2:
6039         case I40E_VSI_CTRL:
6040         case I40E_VSI_IWARP:
6041         case I40E_VSI_MIRROR:
6042         default:
6043                 /* there is no notification for other VSIs */
6044                 break;
6045         }
6046 }
6047
6048 /**
6049  * i40e_veb_link_event - notify elements on the veb of a link event
6050  * @veb: veb to be notified
6051  * @link_up: link up or down
6052  **/
6053 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6054 {
6055         struct i40e_pf *pf;
6056         int i;
6057
6058         if (!veb || !veb->pf)
6059                 return;
6060         pf = veb->pf;
6061
6062         /* depth first... */
6063         for (i = 0; i < I40E_MAX_VEB; i++)
6064                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6065                         i40e_veb_link_event(pf->veb[i], link_up);
6066
6067         /* ... now the local VSIs */
6068         for (i = 0; i < pf->num_alloc_vsi; i++)
6069                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6070                         i40e_vsi_link_event(pf->vsi[i], link_up);
6071 }
6072
6073 /**
6074  * i40e_link_event - Update netif_carrier status
6075  * @pf: board private structure
6076  **/
6077 static void i40e_link_event(struct i40e_pf *pf)
6078 {
6079         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6080         u8 new_link_speed, old_link_speed;
6081         i40e_status status;
6082         bool new_link, old_link;
6083
6084         /* save off old link status information */
6085         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6086
6087         /* set this to force the get_link_status call to refresh state */
6088         pf->hw.phy.get_link_info = true;
6089
6090         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6091
6092         status = i40e_get_link_status(&pf->hw, &new_link);
6093         if (status) {
6094                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6095                         status);
6096                 return;
6097         }
6098
6099         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6100         new_link_speed = pf->hw.phy.link_info.link_speed;
6101
6102         if (new_link == old_link &&
6103             new_link_speed == old_link_speed &&
6104             (test_bit(__I40E_DOWN, &vsi->state) ||
6105              new_link == netif_carrier_ok(vsi->netdev)))
6106                 return;
6107
6108         if (!test_bit(__I40E_DOWN, &vsi->state))
6109                 i40e_print_link_message(vsi, new_link);
6110
6111         /* Notify the base of the switch tree connected to
6112          * the link.  Floating VEBs are not notified.
6113          */
6114         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6115                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6116         else
6117                 i40e_vsi_link_event(vsi, new_link);
6118
6119         if (pf->vf)
6120                 i40e_vc_notify_link_state(pf);
6121
6122         if (pf->flags & I40E_FLAG_PTP)
6123                 i40e_ptp_set_increment(pf);
6124 }
6125
6126 /**
6127  * i40e_watchdog_subtask - periodic checks not using event driven response
6128  * @pf: board private structure
6129  **/
6130 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6131 {
6132         int i;
6133
6134         /* if interface is down do nothing */
6135         if (test_bit(__I40E_DOWN, &pf->state) ||
6136             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6137                 return;
6138
6139         /* make sure we don't do these things too often */
6140         if (time_before(jiffies, (pf->service_timer_previous +
6141                                   pf->service_timer_period)))
6142                 return;
6143         pf->service_timer_previous = jiffies;
6144
6145         if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6146                 i40e_link_event(pf);
6147
6148         /* Update the stats for active netdevs so the network stack
6149          * can look at updated numbers whenever it cares to
6150          */
6151         for (i = 0; i < pf->num_alloc_vsi; i++)
6152                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6153                         i40e_update_stats(pf->vsi[i]);
6154
6155         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6156                 /* Update the stats for the active switching components */
6157                 for (i = 0; i < I40E_MAX_VEB; i++)
6158                         if (pf->veb[i])
6159                                 i40e_update_veb_stats(pf->veb[i]);
6160         }
6161
6162         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6163 }
6164
6165 /**
6166  * i40e_reset_subtask - Set up for resetting the device and driver
6167  * @pf: board private structure
6168  **/
6169 static void i40e_reset_subtask(struct i40e_pf *pf)
6170 {
6171         u32 reset_flags = 0;
6172
6173         rtnl_lock();
6174         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6175                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6176                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6177         }
6178         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6179                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6180                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6181         }
6182         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6183                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6184                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6185         }
6186         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6187                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6188                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6189         }
6190         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6191                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6192                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6193         }
6194
6195         /* If there's a recovery already waiting, it takes
6196          * precedence before starting a new reset sequence.
6197          */
6198         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6199                 i40e_handle_reset_warning(pf);
6200                 goto unlock;
6201         }
6202
6203         /* If we're already down or resetting, just bail */
6204         if (reset_flags &&
6205             !test_bit(__I40E_DOWN, &pf->state) &&
6206             !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6207                 i40e_do_reset(pf, reset_flags);
6208
6209 unlock:
6210         rtnl_unlock();
6211 }
6212
6213 /**
6214  * i40e_handle_link_event - Handle link event
6215  * @pf: board private structure
6216  * @e: event info posted on ARQ
6217  **/
6218 static void i40e_handle_link_event(struct i40e_pf *pf,
6219                                    struct i40e_arq_event_info *e)
6220 {
6221         struct i40e_aqc_get_link_status *status =
6222                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6223
6224         /* Do a new status request to re-enable LSE reporting
6225          * and load new status information into the hw struct
6226          * This completely ignores any state information
6227          * in the ARQ event info, instead choosing to always
6228          * issue the AQ update link status command.
6229          */
6230         i40e_link_event(pf);
6231
6232         /* check for unqualified module, if link is down */
6233         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6234             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6235             (!(status->link_info & I40E_AQ_LINK_UP)))
6236                 dev_err(&pf->pdev->dev,
6237                         "The driver failed to link because an unqualified module was detected.\n");
6238 }
6239
6240 /**
6241  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6242  * @pf: board private structure
6243  **/
6244 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6245 {
6246         struct i40e_arq_event_info event;
6247         struct i40e_hw *hw = &pf->hw;
6248         u16 pending, i = 0;
6249         i40e_status ret;
6250         u16 opcode;
6251         u32 oldval;
6252         u32 val;
6253
6254         /* Do not run clean AQ when PF reset fails */
6255         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6256                 return;
6257
6258         /* check for error indications */
6259         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6260         oldval = val;
6261         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6262                 if (hw->debug_mask & I40E_DEBUG_AQ)
6263                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6264                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6265         }
6266         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6267                 if (hw->debug_mask & I40E_DEBUG_AQ)
6268                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6269                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6270         }
6271         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6272                 if (hw->debug_mask & I40E_DEBUG_AQ)
6273                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6274                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6275         }
6276         if (oldval != val)
6277                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6278
6279         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6280         oldval = val;
6281         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6282                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6283                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6284                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6285         }
6286         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6287                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6288                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6289                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6290         }
6291         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6292                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6293                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6294                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6295         }
6296         if (oldval != val)
6297                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6298
6299         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6300         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6301         if (!event.msg_buf)
6302                 return;
6303
6304         do {
6305                 ret = i40e_clean_arq_element(hw, &event, &pending);
6306                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6307                         break;
6308                 else if (ret) {
6309                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6310                         break;
6311                 }
6312
6313                 opcode = le16_to_cpu(event.desc.opcode);
6314                 switch (opcode) {
6315
6316                 case i40e_aqc_opc_get_link_status:
6317                         i40e_handle_link_event(pf, &event);
6318                         break;
6319                 case i40e_aqc_opc_send_msg_to_pf:
6320                         ret = i40e_vc_process_vf_msg(pf,
6321                                         le16_to_cpu(event.desc.retval),
6322                                         le32_to_cpu(event.desc.cookie_high),
6323                                         le32_to_cpu(event.desc.cookie_low),
6324                                         event.msg_buf,
6325                                         event.msg_len);
6326                         break;
6327                 case i40e_aqc_opc_lldp_update_mib:
6328                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6329 #ifdef CONFIG_I40E_DCB
6330                         rtnl_lock();
6331                         ret = i40e_handle_lldp_event(pf, &event);
6332                         rtnl_unlock();
6333 #endif /* CONFIG_I40E_DCB */
6334                         break;
6335                 case i40e_aqc_opc_event_lan_overflow:
6336                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6337                         i40e_handle_lan_overflow_event(pf, &event);
6338                         break;
6339                 case i40e_aqc_opc_send_msg_to_peer:
6340                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6341                         break;
6342                 case i40e_aqc_opc_nvm_erase:
6343                 case i40e_aqc_opc_nvm_update:
6344                 case i40e_aqc_opc_oem_post_update:
6345                         i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
6346                         break;
6347                 default:
6348                         dev_info(&pf->pdev->dev,
6349                                  "ARQ Error: Unknown event 0x%04x received\n",
6350                                  opcode);
6351                         break;
6352                 }
6353         } while (pending && (i++ < pf->adminq_work_limit));
6354
6355         clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6356         /* re-enable Admin queue interrupt cause */
6357         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6358         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6359         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6360         i40e_flush(hw);
6361
6362         kfree(event.msg_buf);
6363 }
6364
6365 /**
6366  * i40e_verify_eeprom - make sure eeprom is good to use
6367  * @pf: board private structure
6368  **/
6369 static void i40e_verify_eeprom(struct i40e_pf *pf)
6370 {
6371         int err;
6372
6373         err = i40e_diag_eeprom_test(&pf->hw);
6374         if (err) {
6375                 /* retry in case of garbage read */
6376                 err = i40e_diag_eeprom_test(&pf->hw);
6377                 if (err) {
6378                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6379                                  err);
6380                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6381                 }
6382         }
6383
6384         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6385                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6386                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6387         }
6388 }
6389
6390 /**
6391  * i40e_enable_pf_switch_lb
6392  * @pf: pointer to the PF structure
6393  *
6394  * enable switch loop back or die - no point in a return value
6395  **/
6396 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6397 {
6398         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6399         struct i40e_vsi_context ctxt;
6400         int ret;
6401
6402         ctxt.seid = pf->main_vsi_seid;
6403         ctxt.pf_num = pf->hw.pf_id;
6404         ctxt.vf_num = 0;
6405         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6406         if (ret) {
6407                 dev_info(&pf->pdev->dev,
6408                          "couldn't get PF vsi config, err %s aq_err %s\n",
6409                          i40e_stat_str(&pf->hw, ret),
6410                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6411                 return;
6412         }
6413         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6414         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6415         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6416
6417         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6418         if (ret) {
6419                 dev_info(&pf->pdev->dev,
6420                          "update vsi switch failed, err %s aq_err %s\n",
6421                          i40e_stat_str(&pf->hw, ret),
6422                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6423         }
6424 }
6425
6426 /**
6427  * i40e_disable_pf_switch_lb
6428  * @pf: pointer to the PF structure
6429  *
6430  * disable switch loop back or die - no point in a return value
6431  **/
6432 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6433 {
6434         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6435         struct i40e_vsi_context ctxt;
6436         int ret;
6437
6438         ctxt.seid = pf->main_vsi_seid;
6439         ctxt.pf_num = pf->hw.pf_id;
6440         ctxt.vf_num = 0;
6441         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6442         if (ret) {
6443                 dev_info(&pf->pdev->dev,
6444                          "couldn't get PF vsi config, err %s aq_err %s\n",
6445                          i40e_stat_str(&pf->hw, ret),
6446                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6447                 return;
6448         }
6449         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6450         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6451         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6452
6453         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6454         if (ret) {
6455                 dev_info(&pf->pdev->dev,
6456                          "update vsi switch failed, err %s aq_err %s\n",
6457                          i40e_stat_str(&pf->hw, ret),
6458                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6459         }
6460 }
6461
6462 /**
6463  * i40e_config_bridge_mode - Configure the HW bridge mode
6464  * @veb: pointer to the bridge instance
6465  *
6466  * Configure the loop back mode for the LAN VSI that is downlink to the
6467  * specified HW bridge instance. It is expected this function is called
6468  * when a new HW bridge is instantiated.
6469  **/
6470 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6471 {
6472         struct i40e_pf *pf = veb->pf;
6473
6474         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6475                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6476                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6477         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6478                 i40e_disable_pf_switch_lb(pf);
6479         else
6480                 i40e_enable_pf_switch_lb(pf);
6481 }
6482
6483 /**
6484  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6485  * @veb: pointer to the VEB instance
6486  *
6487  * This is a recursive function that first builds the attached VSIs then
6488  * recurses in to build the next layer of VEB.  We track the connections
6489  * through our own index numbers because the seid's from the HW could
6490  * change across the reset.
6491  **/
6492 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6493 {
6494         struct i40e_vsi *ctl_vsi = NULL;
6495         struct i40e_pf *pf = veb->pf;
6496         int v, veb_idx;
6497         int ret;
6498
6499         /* build VSI that owns this VEB, temporarily attached to base VEB */
6500         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6501                 if (pf->vsi[v] &&
6502                     pf->vsi[v]->veb_idx == veb->idx &&
6503                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6504                         ctl_vsi = pf->vsi[v];
6505                         break;
6506                 }
6507         }
6508         if (!ctl_vsi) {
6509                 dev_info(&pf->pdev->dev,
6510                          "missing owner VSI for veb_idx %d\n", veb->idx);
6511                 ret = -ENOENT;
6512                 goto end_reconstitute;
6513         }
6514         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6515                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6516         ret = i40e_add_vsi(ctl_vsi);
6517         if (ret) {
6518                 dev_info(&pf->pdev->dev,
6519                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6520                          veb->idx, ret);
6521                 goto end_reconstitute;
6522         }
6523         i40e_vsi_reset_stats(ctl_vsi);
6524
6525         /* create the VEB in the switch and move the VSI onto the VEB */
6526         ret = i40e_add_veb(veb, ctl_vsi);
6527         if (ret)
6528                 goto end_reconstitute;
6529
6530         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6531                 veb->bridge_mode = BRIDGE_MODE_VEB;
6532         else
6533                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6534         i40e_config_bridge_mode(veb);
6535
6536         /* create the remaining VSIs attached to this VEB */
6537         for (v = 0; v < pf->num_alloc_vsi; v++) {
6538                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6539                         continue;
6540
6541                 if (pf->vsi[v]->veb_idx == veb->idx) {
6542                         struct i40e_vsi *vsi = pf->vsi[v];
6543
6544                         vsi->uplink_seid = veb->seid;
6545                         ret = i40e_add_vsi(vsi);
6546                         if (ret) {
6547                                 dev_info(&pf->pdev->dev,
6548                                          "rebuild of vsi_idx %d failed: %d\n",
6549                                          v, ret);
6550                                 goto end_reconstitute;
6551                         }
6552                         i40e_vsi_reset_stats(vsi);
6553                 }
6554         }
6555
6556         /* create any VEBs attached to this VEB - RECURSION */
6557         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6558                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6559                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6560                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6561                         if (ret)
6562                                 break;
6563                 }
6564         }
6565
6566 end_reconstitute:
6567         return ret;
6568 }
6569
6570 /**
6571  * i40e_get_capabilities - get info about the HW
6572  * @pf: the PF struct
6573  **/
6574 static int i40e_get_capabilities(struct i40e_pf *pf)
6575 {
6576         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6577         u16 data_size;
6578         int buf_len;
6579         int err;
6580
6581         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6582         do {
6583                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6584                 if (!cap_buf)
6585                         return -ENOMEM;
6586
6587                 /* this loads the data into the hw struct for us */
6588                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6589                                             &data_size,
6590                                             i40e_aqc_opc_list_func_capabilities,
6591                                             NULL);
6592                 /* data loaded, buffer no longer needed */
6593                 kfree(cap_buf);
6594
6595                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6596                         /* retry with a larger buffer */
6597                         buf_len = data_size;
6598                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6599                         dev_info(&pf->pdev->dev,
6600                                  "capability discovery failed, err %s aq_err %s\n",
6601                                  i40e_stat_str(&pf->hw, err),
6602                                  i40e_aq_str(&pf->hw,
6603                                              pf->hw.aq.asq_last_status));
6604                         return -ENODEV;
6605                 }
6606         } while (err);
6607
6608         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6609                 dev_info(&pf->pdev->dev,
6610                          "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",
6611                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6612                          pf->hw.func_caps.num_msix_vectors,
6613                          pf->hw.func_caps.num_msix_vectors_vf,
6614                          pf->hw.func_caps.fd_filters_guaranteed,
6615                          pf->hw.func_caps.fd_filters_best_effort,
6616                          pf->hw.func_caps.num_tx_qp,
6617                          pf->hw.func_caps.num_vsis);
6618
6619 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6620                        + pf->hw.func_caps.num_vfs)
6621         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6622                 dev_info(&pf->pdev->dev,
6623                          "got num_vsis %d, setting num_vsis to %d\n",
6624                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6625                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6626         }
6627
6628         return 0;
6629 }
6630
6631 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6632
6633 /**
6634  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6635  * @pf: board private structure
6636  **/
6637 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6638 {
6639         struct i40e_vsi *vsi;
6640         int i;
6641
6642         /* quick workaround for an NVM issue that leaves a critical register
6643          * uninitialized
6644          */
6645         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6646                 static const u32 hkey[] = {
6647                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6648                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6649                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6650                         0x95b3a76d};
6651
6652                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6653                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6654         }
6655
6656         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6657                 return;
6658
6659         /* find existing VSI and see if it needs configuring */
6660         vsi = NULL;
6661         for (i = 0; i < pf->num_alloc_vsi; i++) {
6662                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6663                         vsi = pf->vsi[i];
6664                         break;
6665                 }
6666         }
6667
6668         /* create a new VSI if none exists */
6669         if (!vsi) {
6670                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6671                                      pf->vsi[pf->lan_vsi]->seid, 0);
6672                 if (!vsi) {
6673                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6674                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6675                         return;
6676                 }
6677         }
6678
6679         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6680 }
6681
6682 /**
6683  * i40e_fdir_teardown - release the Flow Director resources
6684  * @pf: board private structure
6685  **/
6686 static void i40e_fdir_teardown(struct i40e_pf *pf)
6687 {
6688         int i;
6689
6690         i40e_fdir_filter_exit(pf);
6691         for (i = 0; i < pf->num_alloc_vsi; i++) {
6692                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6693                         i40e_vsi_release(pf->vsi[i]);
6694                         break;
6695                 }
6696         }
6697 }
6698
6699 /**
6700  * i40e_prep_for_reset - prep for the core to reset
6701  * @pf: board private structure
6702  *
6703  * Close up the VFs and other things in prep for PF Reset.
6704   **/
6705 static void i40e_prep_for_reset(struct i40e_pf *pf)
6706 {
6707         struct i40e_hw *hw = &pf->hw;
6708         i40e_status ret = 0;
6709         u32 v;
6710
6711         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6712         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6713                 return;
6714
6715         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6716
6717         /* quiesce the VSIs and their queues that are not already DOWN */
6718         i40e_pf_quiesce_all_vsi(pf);
6719
6720         for (v = 0; v < pf->num_alloc_vsi; v++) {
6721                 if (pf->vsi[v])
6722                         pf->vsi[v]->seid = 0;
6723         }
6724
6725         i40e_shutdown_adminq(&pf->hw);
6726
6727         /* call shutdown HMC */
6728         if (hw->hmc.hmc_obj) {
6729                 ret = i40e_shutdown_lan_hmc(hw);
6730                 if (ret)
6731                         dev_warn(&pf->pdev->dev,
6732                                  "shutdown_lan_hmc failed: %d\n", ret);
6733         }
6734 }
6735
6736 /**
6737  * i40e_send_version - update firmware with driver version
6738  * @pf: PF struct
6739  */
6740 static void i40e_send_version(struct i40e_pf *pf)
6741 {
6742         struct i40e_driver_version dv;
6743
6744         dv.major_version = DRV_VERSION_MAJOR;
6745         dv.minor_version = DRV_VERSION_MINOR;
6746         dv.build_version = DRV_VERSION_BUILD;
6747         dv.subbuild_version = 0;
6748         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6749         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6750 }
6751
6752 /**
6753  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6754  * @pf: board private structure
6755  * @reinit: if the Main VSI needs to re-initialized.
6756  **/
6757 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6758 {
6759         struct i40e_hw *hw = &pf->hw;
6760         u8 set_fc_aq_fail = 0;
6761         i40e_status ret;
6762         u32 val;
6763         u32 v;
6764
6765         /* Now we wait for GRST to settle out.
6766          * We don't have to delete the VEBs or VSIs from the hw switch
6767          * because the reset will make them disappear.
6768          */
6769         ret = i40e_pf_reset(hw);
6770         if (ret) {
6771                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6772                 set_bit(__I40E_RESET_FAILED, &pf->state);
6773                 goto clear_recovery;
6774         }
6775         pf->pfr_count++;
6776
6777         if (test_bit(__I40E_DOWN, &pf->state))
6778                 goto clear_recovery;
6779         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6780
6781         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6782         ret = i40e_init_adminq(&pf->hw);
6783         if (ret) {
6784                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6785                          i40e_stat_str(&pf->hw, ret),
6786                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6787                 goto clear_recovery;
6788         }
6789
6790         /* re-verify the eeprom if we just had an EMP reset */
6791         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6792                 i40e_verify_eeprom(pf);
6793
6794         i40e_clear_pxe_mode(hw);
6795         ret = i40e_get_capabilities(pf);
6796         if (ret)
6797                 goto end_core_reset;
6798
6799         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6800                                 hw->func_caps.num_rx_qp,
6801                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6802         if (ret) {
6803                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6804                 goto end_core_reset;
6805         }
6806         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6807         if (ret) {
6808                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6809                 goto end_core_reset;
6810         }
6811
6812 #ifdef CONFIG_I40E_DCB
6813         ret = i40e_init_pf_dcb(pf);
6814         if (ret) {
6815                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6816                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6817                 /* Continue without DCB enabled */
6818         }
6819 #endif /* CONFIG_I40E_DCB */
6820 #ifdef I40E_FCOE
6821         i40e_init_pf_fcoe(pf);
6822
6823 #endif
6824         /* do basic switch setup */
6825         ret = i40e_setup_pf_switch(pf, reinit);
6826         if (ret)
6827                 goto end_core_reset;
6828
6829         /* driver is only interested in link up/down and module qualification
6830          * reports from firmware
6831          */
6832         ret = i40e_aq_set_phy_int_mask(&pf->hw,
6833                                        I40E_AQ_EVENT_LINK_UPDOWN |
6834                                        I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6835         if (ret)
6836                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6837                          i40e_stat_str(&pf->hw, ret),
6838                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6839
6840         /* make sure our flow control settings are restored */
6841         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6842         if (ret)
6843                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6844                         i40e_stat_str(&pf->hw, ret),
6845                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6846
6847         /* Rebuild the VSIs and VEBs that existed before reset.
6848          * They are still in our local switch element arrays, so only
6849          * need to rebuild the switch model in the HW.
6850          *
6851          * If there were VEBs but the reconstitution failed, we'll try
6852          * try to recover minimal use by getting the basic PF VSI working.
6853          */
6854         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6855                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6856                 /* find the one VEB connected to the MAC, and find orphans */
6857                 for (v = 0; v < I40E_MAX_VEB; v++) {
6858                         if (!pf->veb[v])
6859                                 continue;
6860
6861                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6862                             pf->veb[v]->uplink_seid == 0) {
6863                                 ret = i40e_reconstitute_veb(pf->veb[v]);
6864
6865                                 if (!ret)
6866                                         continue;
6867
6868                                 /* If Main VEB failed, we're in deep doodoo,
6869                                  * so give up rebuilding the switch and set up
6870                                  * for minimal rebuild of PF VSI.
6871                                  * If orphan failed, we'll report the error
6872                                  * but try to keep going.
6873                                  */
6874                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6875                                         dev_info(&pf->pdev->dev,
6876                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6877                                                  ret);
6878                                         pf->vsi[pf->lan_vsi]->uplink_seid
6879                                                                 = pf->mac_seid;
6880                                         break;
6881                                 } else if (pf->veb[v]->uplink_seid == 0) {
6882                                         dev_info(&pf->pdev->dev,
6883                                                  "rebuild of orphan VEB failed: %d\n",
6884                                                  ret);
6885                                 }
6886                         }
6887                 }
6888         }
6889
6890         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6891                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6892                 /* no VEB, so rebuild only the Main VSI */
6893                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6894                 if (ret) {
6895                         dev_info(&pf->pdev->dev,
6896                                  "rebuild of Main VSI failed: %d\n", ret);
6897                         goto end_core_reset;
6898                 }
6899         }
6900
6901         /* Reconfigure hardware for allowing smaller MSS in the case
6902          * of TSO, so that we avoid the MDD being fired and causing
6903          * a reset in the case of small MSS+TSO.
6904          */
6905 #define I40E_REG_MSS          0x000E64DC
6906 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6907 #define I40E_64BYTE_MSS       0x400000
6908         val = rd32(hw, I40E_REG_MSS);
6909         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6910                 val &= ~I40E_REG_MSS_MIN_MASK;
6911                 val |= I40E_64BYTE_MSS;
6912                 wr32(hw, I40E_REG_MSS, val);
6913         }
6914
6915         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6916             (pf->hw.aq.fw_maj_ver < 4)) {
6917                 msleep(75);
6918                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6919                 if (ret)
6920                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6921                                  i40e_stat_str(&pf->hw, ret),
6922                                  i40e_aq_str(&pf->hw,
6923                                              pf->hw.aq.asq_last_status));
6924         }
6925         /* reinit the misc interrupt */
6926         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6927                 ret = i40e_setup_misc_vector(pf);
6928
6929         /* Add a filter to drop all Flow control frames from any VSI from being
6930          * transmitted. By doing so we stop a malicious VF from sending out
6931          * PAUSE or PFC frames and potentially controlling traffic for other
6932          * PF/VF VSIs.
6933          * The FW can still send Flow control frames if enabled.
6934          */
6935         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6936                                                        pf->main_vsi_seid);
6937
6938         /* restart the VSIs that were rebuilt and running before the reset */
6939         i40e_pf_unquiesce_all_vsi(pf);
6940
6941         if (pf->num_alloc_vfs) {
6942                 for (v = 0; v < pf->num_alloc_vfs; v++)
6943                         i40e_reset_vf(&pf->vf[v], true);
6944         }
6945
6946         /* tell the firmware that we're starting */
6947         i40e_send_version(pf);
6948
6949 end_core_reset:
6950         clear_bit(__I40E_RESET_FAILED, &pf->state);
6951 clear_recovery:
6952         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6953 }
6954
6955 /**
6956  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6957  * @pf: board private structure
6958  *
6959  * Close up the VFs and other things in prep for a Core Reset,
6960  * then get ready to rebuild the world.
6961  **/
6962 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6963 {
6964         i40e_prep_for_reset(pf);
6965         i40e_reset_and_rebuild(pf, false);
6966 }
6967
6968 /**
6969  * i40e_handle_mdd_event
6970  * @pf: pointer to the PF structure
6971  *
6972  * Called from the MDD irq handler to identify possibly malicious vfs
6973  **/
6974 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6975 {
6976         struct i40e_hw *hw = &pf->hw;
6977         bool mdd_detected = false;
6978         bool pf_mdd_detected = false;
6979         struct i40e_vf *vf;
6980         u32 reg;
6981         int i;
6982
6983         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6984                 return;
6985
6986         /* find what triggered the MDD event */
6987         reg = rd32(hw, I40E_GL_MDET_TX);
6988         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6989                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6990                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
6991                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6992                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
6993                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6994                                 I40E_GL_MDET_TX_EVENT_SHIFT;
6995                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6996                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
6997                                 pf->hw.func_caps.base_queue;
6998                 if (netif_msg_tx_err(pf))
6999                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7000                                  event, queue, pf_num, vf_num);
7001                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7002                 mdd_detected = true;
7003         }
7004         reg = rd32(hw, I40E_GL_MDET_RX);
7005         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7006                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7007                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7008                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7009                                 I40E_GL_MDET_RX_EVENT_SHIFT;
7010                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7011                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7012                                 pf->hw.func_caps.base_queue;
7013                 if (netif_msg_rx_err(pf))
7014                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7015                                  event, queue, func);
7016                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7017                 mdd_detected = true;
7018         }
7019
7020         if (mdd_detected) {
7021                 reg = rd32(hw, I40E_PF_MDET_TX);
7022                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7023                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7024                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7025                         pf_mdd_detected = true;
7026                 }
7027                 reg = rd32(hw, I40E_PF_MDET_RX);
7028                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7029                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7030                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7031                         pf_mdd_detected = true;
7032                 }
7033                 /* Queue belongs to the PF, initiate a reset */
7034                 if (pf_mdd_detected) {
7035                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7036                         i40e_service_event_schedule(pf);
7037                 }
7038         }
7039
7040         /* see if one of the VFs needs its hand slapped */
7041         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7042                 vf = &(pf->vf[i]);
7043                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7044                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7045                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7046                         vf->num_mdd_events++;
7047                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7048                                  i);
7049                 }
7050
7051                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7052                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7053                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7054                         vf->num_mdd_events++;
7055                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7056                                  i);
7057                 }
7058
7059                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7060                         dev_info(&pf->pdev->dev,
7061                                  "Too many MDD events on VF %d, disabled\n", i);
7062                         dev_info(&pf->pdev->dev,
7063                                  "Use PF Control I/F to re-enable the VF\n");
7064                         set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7065                 }
7066         }
7067
7068         /* re-enable mdd interrupt cause */
7069         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7070         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7071         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7072         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7073         i40e_flush(hw);
7074 }
7075
7076 /**
7077  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7078  * @pf: board private structure
7079  **/
7080 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7081 {
7082 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
7083         struct i40e_hw *hw = &pf->hw;
7084         i40e_status ret;
7085         __be16 port;
7086         int i;
7087
7088         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7089                 return;
7090
7091         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7092
7093         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7094                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7095                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7096                         port = pf->udp_ports[i].index;
7097                         if (port)
7098                                 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7099                                                      pf->udp_ports[i].type,
7100                                                      NULL, NULL);
7101                         else
7102                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7103
7104                         if (ret) {
7105                                 dev_info(&pf->pdev->dev,
7106                                          "%s vxlan port %d, index %d failed, err %s aq_err %s\n",
7107                                          port ? "add" : "delete",
7108                                          ntohs(port), i,
7109                                          i40e_stat_str(&pf->hw, ret),
7110                                          i40e_aq_str(&pf->hw,
7111                                                     pf->hw.aq.asq_last_status));
7112                                 pf->udp_ports[i].index = 0;
7113                         }
7114                 }
7115         }
7116 #endif
7117 }
7118
7119 /**
7120  * i40e_service_task - Run the driver's async subtasks
7121  * @work: pointer to work_struct containing our data
7122  **/
7123 static void i40e_service_task(struct work_struct *work)
7124 {
7125         struct i40e_pf *pf = container_of(work,
7126                                           struct i40e_pf,
7127                                           service_task);
7128         unsigned long start_time = jiffies;
7129
7130         /* don't bother with service tasks if a reset is in progress */
7131         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7132                 i40e_service_event_complete(pf);
7133                 return;
7134         }
7135
7136         i40e_detect_recover_hung(pf);
7137         i40e_reset_subtask(pf);
7138         i40e_handle_mdd_event(pf);
7139         i40e_vc_process_vflr_event(pf);
7140         i40e_watchdog_subtask(pf);
7141         i40e_fdir_reinit_subtask(pf);
7142         i40e_client_subtask(pf);
7143         i40e_sync_filters_subtask(pf);
7144         i40e_sync_udp_filters_subtask(pf);
7145         i40e_clean_adminq_subtask(pf);
7146
7147         i40e_service_event_complete(pf);
7148
7149         /* If the tasks have taken longer than one timer cycle or there
7150          * is more work to be done, reschedule the service task now
7151          * rather than wait for the timer to tick again.
7152          */
7153         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7154             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7155             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7156             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7157                 i40e_service_event_schedule(pf);
7158 }
7159
7160 /**
7161  * i40e_service_timer - timer callback
7162  * @data: pointer to PF struct
7163  **/
7164 static void i40e_service_timer(unsigned long data)
7165 {
7166         struct i40e_pf *pf = (struct i40e_pf *)data;
7167
7168         mod_timer(&pf->service_timer,
7169                   round_jiffies(jiffies + pf->service_timer_period));
7170         i40e_service_event_schedule(pf);
7171 }
7172
7173 /**
7174  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7175  * @vsi: the VSI being configured
7176  **/
7177 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7178 {
7179         struct i40e_pf *pf = vsi->back;
7180
7181         switch (vsi->type) {
7182         case I40E_VSI_MAIN:
7183                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7184                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7185                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7186                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7187                         vsi->num_q_vectors = pf->num_lan_msix;
7188                 else
7189                         vsi->num_q_vectors = 1;
7190
7191                 break;
7192
7193         case I40E_VSI_FDIR:
7194                 vsi->alloc_queue_pairs = 1;
7195                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7196                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7197                 vsi->num_q_vectors = 1;
7198                 break;
7199
7200         case I40E_VSI_VMDQ2:
7201                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7202                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7203                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7204                 vsi->num_q_vectors = pf->num_vmdq_msix;
7205                 break;
7206
7207         case I40E_VSI_SRIOV:
7208                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7209                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7210                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7211                 break;
7212
7213 #ifdef I40E_FCOE
7214         case I40E_VSI_FCOE:
7215                 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7216                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7217                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7218                 vsi->num_q_vectors = pf->num_fcoe_msix;
7219                 break;
7220
7221 #endif /* I40E_FCOE */
7222         default:
7223                 WARN_ON(1);
7224                 return -ENODATA;
7225         }
7226
7227         return 0;
7228 }
7229
7230 /**
7231  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7232  * @type: VSI pointer
7233  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7234  *
7235  * On error: returns error code (negative)
7236  * On success: returns 0
7237  **/
7238 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7239 {
7240         int size;
7241         int ret = 0;
7242
7243         /* allocate memory for both Tx and Rx ring pointers */
7244         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7245         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7246         if (!vsi->tx_rings)
7247                 return -ENOMEM;
7248         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7249
7250         if (alloc_qvectors) {
7251                 /* allocate memory for q_vector pointers */
7252                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7253                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7254                 if (!vsi->q_vectors) {
7255                         ret = -ENOMEM;
7256                         goto err_vectors;
7257                 }
7258         }
7259         return ret;
7260
7261 err_vectors:
7262         kfree(vsi->tx_rings);
7263         return ret;
7264 }
7265
7266 /**
7267  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7268  * @pf: board private structure
7269  * @type: type of VSI
7270  *
7271  * On error: returns error code (negative)
7272  * On success: returns vsi index in PF (positive)
7273  **/
7274 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7275 {
7276         int ret = -ENODEV;
7277         struct i40e_vsi *vsi;
7278         int vsi_idx;
7279         int i;
7280
7281         /* Need to protect the allocation of the VSIs at the PF level */
7282         mutex_lock(&pf->switch_mutex);
7283
7284         /* VSI list may be fragmented if VSI creation/destruction has
7285          * been happening.  We can afford to do a quick scan to look
7286          * for any free VSIs in the list.
7287          *
7288          * find next empty vsi slot, looping back around if necessary
7289          */
7290         i = pf->next_vsi;
7291         while (i < pf->num_alloc_vsi && pf->vsi[i])
7292                 i++;
7293         if (i >= pf->num_alloc_vsi) {
7294                 i = 0;
7295                 while (i < pf->next_vsi && pf->vsi[i])
7296                         i++;
7297         }
7298
7299         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7300                 vsi_idx = i;             /* Found one! */
7301         } else {
7302                 ret = -ENODEV;
7303                 goto unlock_pf;  /* out of VSI slots! */
7304         }
7305         pf->next_vsi = ++i;
7306
7307         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7308         if (!vsi) {
7309                 ret = -ENOMEM;
7310                 goto unlock_pf;
7311         }
7312         vsi->type = type;
7313         vsi->back = pf;
7314         set_bit(__I40E_DOWN, &vsi->state);
7315         vsi->flags = 0;
7316         vsi->idx = vsi_idx;
7317         vsi->rx_itr_setting = pf->rx_itr_default;
7318         vsi->tx_itr_setting = pf->tx_itr_default;
7319         vsi->int_rate_limit = 0;
7320         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7321                                 pf->rss_table_size : 64;
7322         vsi->netdev_registered = false;
7323         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7324         INIT_LIST_HEAD(&vsi->mac_filter_list);
7325         vsi->irqs_ready = false;
7326
7327         ret = i40e_set_num_rings_in_vsi(vsi);
7328         if (ret)
7329                 goto err_rings;
7330
7331         ret = i40e_vsi_alloc_arrays(vsi, true);
7332         if (ret)
7333                 goto err_rings;
7334
7335         /* Setup default MSIX irq handler for VSI */
7336         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7337
7338         /* Initialize VSI lock */
7339         spin_lock_init(&vsi->mac_filter_list_lock);
7340         pf->vsi[vsi_idx] = vsi;
7341         ret = vsi_idx;
7342         goto unlock_pf;
7343
7344 err_rings:
7345         pf->next_vsi = i - 1;
7346         kfree(vsi);
7347 unlock_pf:
7348         mutex_unlock(&pf->switch_mutex);
7349         return ret;
7350 }
7351
7352 /**
7353  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7354  * @type: VSI pointer
7355  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7356  *
7357  * On error: returns error code (negative)
7358  * On success: returns 0
7359  **/
7360 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7361 {
7362         /* free the ring and vector containers */
7363         if (free_qvectors) {
7364                 kfree(vsi->q_vectors);
7365                 vsi->q_vectors = NULL;
7366         }
7367         kfree(vsi->tx_rings);
7368         vsi->tx_rings = NULL;
7369         vsi->rx_rings = NULL;
7370 }
7371
7372 /**
7373  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7374  * and lookup table
7375  * @vsi: Pointer to VSI structure
7376  */
7377 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7378 {
7379         if (!vsi)
7380                 return;
7381
7382         kfree(vsi->rss_hkey_user);
7383         vsi->rss_hkey_user = NULL;
7384
7385         kfree(vsi->rss_lut_user);
7386         vsi->rss_lut_user = NULL;
7387 }
7388
7389 /**
7390  * i40e_vsi_clear - Deallocate the VSI provided
7391  * @vsi: the VSI being un-configured
7392  **/
7393 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7394 {
7395         struct i40e_pf *pf;
7396
7397         if (!vsi)
7398                 return 0;
7399
7400         if (!vsi->back)
7401                 goto free_vsi;
7402         pf = vsi->back;
7403
7404         mutex_lock(&pf->switch_mutex);
7405         if (!pf->vsi[vsi->idx]) {
7406                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7407                         vsi->idx, vsi->idx, vsi, vsi->type);
7408                 goto unlock_vsi;
7409         }
7410
7411         if (pf->vsi[vsi->idx] != vsi) {
7412                 dev_err(&pf->pdev->dev,
7413                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7414                         pf->vsi[vsi->idx]->idx,
7415                         pf->vsi[vsi->idx],
7416                         pf->vsi[vsi->idx]->type,
7417                         vsi->idx, vsi, vsi->type);
7418                 goto unlock_vsi;
7419         }
7420
7421         /* updates the PF for this cleared vsi */
7422         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7423         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7424
7425         i40e_vsi_free_arrays(vsi, true);
7426         i40e_clear_rss_config_user(vsi);
7427
7428         pf->vsi[vsi->idx] = NULL;
7429         if (vsi->idx < pf->next_vsi)
7430                 pf->next_vsi = vsi->idx;
7431
7432 unlock_vsi:
7433         mutex_unlock(&pf->switch_mutex);
7434 free_vsi:
7435         kfree(vsi);
7436
7437         return 0;
7438 }
7439
7440 /**
7441  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7442  * @vsi: the VSI being cleaned
7443  **/
7444 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7445 {
7446         int i;
7447
7448         if (vsi->tx_rings && vsi->tx_rings[0]) {
7449                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7450                         kfree_rcu(vsi->tx_rings[i], rcu);
7451                         vsi->tx_rings[i] = NULL;
7452                         vsi->rx_rings[i] = NULL;
7453                 }
7454         }
7455 }
7456
7457 /**
7458  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7459  * @vsi: the VSI being configured
7460  **/
7461 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7462 {
7463         struct i40e_ring *tx_ring, *rx_ring;
7464         struct i40e_pf *pf = vsi->back;
7465         int i;
7466
7467         /* Set basic values in the rings to be used later during open() */
7468         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7469                 /* allocate space for both Tx and Rx in one shot */
7470                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7471                 if (!tx_ring)
7472                         goto err_out;
7473
7474                 tx_ring->queue_index = i;
7475                 tx_ring->reg_idx = vsi->base_queue + i;
7476                 tx_ring->ring_active = false;
7477                 tx_ring->vsi = vsi;
7478                 tx_ring->netdev = vsi->netdev;
7479                 tx_ring->dev = &pf->pdev->dev;
7480                 tx_ring->count = vsi->num_desc;
7481                 tx_ring->size = 0;
7482                 tx_ring->dcb_tc = 0;
7483                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7484                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7485                 if (vsi->back->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE)
7486                         tx_ring->flags |= I40E_TXR_FLAGS_OUTER_UDP_CSUM;
7487                 vsi->tx_rings[i] = tx_ring;
7488
7489                 rx_ring = &tx_ring[1];
7490                 rx_ring->queue_index = i;
7491                 rx_ring->reg_idx = vsi->base_queue + i;
7492                 rx_ring->ring_active = false;
7493                 rx_ring->vsi = vsi;
7494                 rx_ring->netdev = vsi->netdev;
7495                 rx_ring->dev = &pf->pdev->dev;
7496                 rx_ring->count = vsi->num_desc;
7497                 rx_ring->size = 0;
7498                 rx_ring->dcb_tc = 0;
7499                 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7500                         set_ring_16byte_desc_enabled(rx_ring);
7501                 else
7502                         clear_ring_16byte_desc_enabled(rx_ring);
7503                 vsi->rx_rings[i] = rx_ring;
7504         }
7505
7506         return 0;
7507
7508 err_out:
7509         i40e_vsi_clear_rings(vsi);
7510         return -ENOMEM;
7511 }
7512
7513 /**
7514  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7515  * @pf: board private structure
7516  * @vectors: the number of MSI-X vectors to request
7517  *
7518  * Returns the number of vectors reserved, or error
7519  **/
7520 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7521 {
7522         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7523                                         I40E_MIN_MSIX, vectors);
7524         if (vectors < 0) {
7525                 dev_info(&pf->pdev->dev,
7526                          "MSI-X vector reservation failed: %d\n", vectors);
7527                 vectors = 0;
7528         }
7529
7530         return vectors;
7531 }
7532
7533 /**
7534  * i40e_init_msix - Setup the MSIX capability
7535  * @pf: board private structure
7536  *
7537  * Work with the OS to set up the MSIX vectors needed.
7538  *
7539  * Returns the number of vectors reserved or negative on failure
7540  **/
7541 static int i40e_init_msix(struct i40e_pf *pf)
7542 {
7543         struct i40e_hw *hw = &pf->hw;
7544         int vectors_left;
7545         int v_budget, i;
7546         int v_actual;
7547         int iwarp_requested = 0;
7548
7549         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7550                 return -ENODEV;
7551
7552         /* The number of vectors we'll request will be comprised of:
7553          *   - Add 1 for "other" cause for Admin Queue events, etc.
7554          *   - The number of LAN queue pairs
7555          *      - Queues being used for RSS.
7556          *              We don't need as many as max_rss_size vectors.
7557          *              use rss_size instead in the calculation since that
7558          *              is governed by number of cpus in the system.
7559          *      - assumes symmetric Tx/Rx pairing
7560          *   - The number of VMDq pairs
7561          *   - The CPU count within the NUMA node if iWARP is enabled
7562 #ifdef I40E_FCOE
7563          *   - The number of FCOE qps.
7564 #endif
7565          * Once we count this up, try the request.
7566          *
7567          * If we can't get what we want, we'll simplify to nearly nothing
7568          * and try again.  If that still fails, we punt.
7569          */
7570         vectors_left = hw->func_caps.num_msix_vectors;
7571         v_budget = 0;
7572
7573         /* reserve one vector for miscellaneous handler */
7574         if (vectors_left) {
7575                 v_budget++;
7576                 vectors_left--;
7577         }
7578
7579         /* reserve vectors for the main PF traffic queues */
7580         pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7581         vectors_left -= pf->num_lan_msix;
7582         v_budget += pf->num_lan_msix;
7583
7584         /* reserve one vector for sideband flow director */
7585         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7586                 if (vectors_left) {
7587                         v_budget++;
7588                         vectors_left--;
7589                 } else {
7590                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7591                 }
7592         }
7593
7594 #ifdef I40E_FCOE
7595         /* can we reserve enough for FCoE? */
7596         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7597                 if (!vectors_left)
7598                         pf->num_fcoe_msix = 0;
7599                 else if (vectors_left >= pf->num_fcoe_qps)
7600                         pf->num_fcoe_msix = pf->num_fcoe_qps;
7601                 else
7602                         pf->num_fcoe_msix = 1;
7603                 v_budget += pf->num_fcoe_msix;
7604                 vectors_left -= pf->num_fcoe_msix;
7605         }
7606
7607 #endif
7608         /* can we reserve enough for iWARP? */
7609         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7610                 if (!vectors_left)
7611                         pf->num_iwarp_msix = 0;
7612                 else if (vectors_left < pf->num_iwarp_msix)
7613                         pf->num_iwarp_msix = 1;
7614                 v_budget += pf->num_iwarp_msix;
7615                 vectors_left -= pf->num_iwarp_msix;
7616         }
7617
7618         /* any vectors left over go for VMDq support */
7619         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7620                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7621                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7622
7623                 /* if we're short on vectors for what's desired, we limit
7624                  * the queues per vmdq.  If this is still more than are
7625                  * available, the user will need to change the number of
7626                  * queues/vectors used by the PF later with the ethtool
7627                  * channels command
7628                  */
7629                 if (vmdq_vecs < vmdq_vecs_wanted)
7630                         pf->num_vmdq_qps = 1;
7631                 pf->num_vmdq_msix = pf->num_vmdq_qps;
7632
7633                 v_budget += vmdq_vecs;
7634                 vectors_left -= vmdq_vecs;
7635         }
7636
7637         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7638                                    GFP_KERNEL);
7639         if (!pf->msix_entries)
7640                 return -ENOMEM;
7641
7642         for (i = 0; i < v_budget; i++)
7643                 pf->msix_entries[i].entry = i;
7644         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7645
7646         if (v_actual != v_budget) {
7647                 /* If we have limited resources, we will start with no vectors
7648                  * for the special features and then allocate vectors to some
7649                  * of these features based on the policy and at the end disable
7650                  * the features that did not get any vectors.
7651                  */
7652                 iwarp_requested = pf->num_iwarp_msix;
7653                 pf->num_iwarp_msix = 0;
7654 #ifdef I40E_FCOE
7655                 pf->num_fcoe_qps = 0;
7656                 pf->num_fcoe_msix = 0;
7657 #endif
7658                 pf->num_vmdq_msix = 0;
7659         }
7660
7661         if (v_actual < I40E_MIN_MSIX) {
7662                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7663                 kfree(pf->msix_entries);
7664                 pf->msix_entries = NULL;
7665                 return -ENODEV;
7666
7667         } else if (v_actual == I40E_MIN_MSIX) {
7668                 /* Adjust for minimal MSIX use */
7669                 pf->num_vmdq_vsis = 0;
7670                 pf->num_vmdq_qps = 0;
7671                 pf->num_lan_qps = 1;
7672                 pf->num_lan_msix = 1;
7673
7674         } else if (v_actual != v_budget) {
7675                 int vec;
7676
7677                 /* reserve the misc vector */
7678                 vec = v_actual - 1;
7679
7680                 /* Scale vector usage down */
7681                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7682                 pf->num_vmdq_vsis = 1;
7683                 pf->num_vmdq_qps = 1;
7684                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7685
7686                 /* partition out the remaining vectors */
7687                 switch (vec) {
7688                 case 2:
7689                         pf->num_lan_msix = 1;
7690                         break;
7691                 case 3:
7692                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7693                                 pf->num_lan_msix = 1;
7694                                 pf->num_iwarp_msix = 1;
7695                         } else {
7696                                 pf->num_lan_msix = 2;
7697                         }
7698 #ifdef I40E_FCOE
7699                         /* give one vector to FCoE */
7700                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7701                                 pf->num_lan_msix = 1;
7702                                 pf->num_fcoe_msix = 1;
7703                         }
7704 #endif
7705                         break;
7706                 default:
7707                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7708                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7709                                                  iwarp_requested);
7710                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7711                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7712                         } else {
7713                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7714                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7715                         }
7716                         pf->num_lan_msix = min_t(int,
7717                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7718                                                               pf->num_lan_msix);
7719 #ifdef I40E_FCOE
7720                         /* give one vector to FCoE */
7721                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7722                                 pf->num_fcoe_msix = 1;
7723                                 vec--;
7724                         }
7725 #endif
7726                         break;
7727                 }
7728         }
7729
7730         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7731             (pf->num_vmdq_msix == 0)) {
7732                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7733                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7734         }
7735
7736         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7737             (pf->num_iwarp_msix == 0)) {
7738                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7739                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7740         }
7741 #ifdef I40E_FCOE
7742
7743         if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7744                 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7745                 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7746         }
7747 #endif
7748         return v_actual;
7749 }
7750
7751 /**
7752  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7753  * @vsi: the VSI being configured
7754  * @v_idx: index of the vector in the vsi struct
7755  *
7756  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7757  **/
7758 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7759 {
7760         struct i40e_q_vector *q_vector;
7761
7762         /* allocate q_vector */
7763         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7764         if (!q_vector)
7765                 return -ENOMEM;
7766
7767         q_vector->vsi = vsi;
7768         q_vector->v_idx = v_idx;
7769         cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7770         if (vsi->netdev)
7771                 netif_napi_add(vsi->netdev, &q_vector->napi,
7772                                i40e_napi_poll, NAPI_POLL_WEIGHT);
7773
7774         q_vector->rx.latency_range = I40E_LOW_LATENCY;
7775         q_vector->tx.latency_range = I40E_LOW_LATENCY;
7776
7777         /* tie q_vector and vsi together */
7778         vsi->q_vectors[v_idx] = q_vector;
7779
7780         return 0;
7781 }
7782
7783 /**
7784  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7785  * @vsi: the VSI being configured
7786  *
7787  * We allocate one q_vector per queue interrupt.  If allocation fails we
7788  * return -ENOMEM.
7789  **/
7790 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7791 {
7792         struct i40e_pf *pf = vsi->back;
7793         int v_idx, num_q_vectors;
7794         int err;
7795
7796         /* if not MSIX, give the one vector only to the LAN VSI */
7797         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7798                 num_q_vectors = vsi->num_q_vectors;
7799         else if (vsi == pf->vsi[pf->lan_vsi])
7800                 num_q_vectors = 1;
7801         else
7802                 return -EINVAL;
7803
7804         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7805                 err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7806                 if (err)
7807                         goto err_out;
7808         }
7809
7810         return 0;
7811
7812 err_out:
7813         while (v_idx--)
7814                 i40e_free_q_vector(vsi, v_idx);
7815
7816         return err;
7817 }
7818
7819 /**
7820  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7821  * @pf: board private structure to initialize
7822  **/
7823 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7824 {
7825         int vectors = 0;
7826         ssize_t size;
7827
7828         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7829                 vectors = i40e_init_msix(pf);
7830                 if (vectors < 0) {
7831                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
7832                                        I40E_FLAG_IWARP_ENABLED  |
7833 #ifdef I40E_FCOE
7834                                        I40E_FLAG_FCOE_ENABLED   |
7835 #endif
7836                                        I40E_FLAG_RSS_ENABLED    |
7837                                        I40E_FLAG_DCB_CAPABLE    |
7838                                        I40E_FLAG_SRIOV_ENABLED  |
7839                                        I40E_FLAG_FD_SB_ENABLED  |
7840                                        I40E_FLAG_FD_ATR_ENABLED |
7841                                        I40E_FLAG_VMDQ_ENABLED);
7842
7843                         /* rework the queue expectations without MSIX */
7844                         i40e_determine_queue_usage(pf);
7845                 }
7846         }
7847
7848         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7849             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7850                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7851                 vectors = pci_enable_msi(pf->pdev);
7852                 if (vectors < 0) {
7853                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7854                                  vectors);
7855                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7856                 }
7857                 vectors = 1;  /* one MSI or Legacy vector */
7858         }
7859
7860         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7861                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7862
7863         /* set up vector assignment tracking */
7864         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7865         pf->irq_pile = kzalloc(size, GFP_KERNEL);
7866         if (!pf->irq_pile) {
7867                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7868                 return -ENOMEM;
7869         }
7870         pf->irq_pile->num_entries = vectors;
7871         pf->irq_pile->search_hint = 0;
7872
7873         /* track first vector for misc interrupts, ignore return */
7874         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7875
7876         return 0;
7877 }
7878
7879 /**
7880  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7881  * @pf: board private structure
7882  *
7883  * This sets up the handler for MSIX 0, which is used to manage the
7884  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7885  * when in MSI or Legacy interrupt mode.
7886  **/
7887 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7888 {
7889         struct i40e_hw *hw = &pf->hw;
7890         int err = 0;
7891
7892         /* Only request the irq if this is the first time through, and
7893          * not when we're rebuilding after a Reset
7894          */
7895         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7896                 err = request_irq(pf->msix_entries[0].vector,
7897                                   i40e_intr, 0, pf->int_name, pf);
7898                 if (err) {
7899                         dev_info(&pf->pdev->dev,
7900                                  "request_irq for %s failed: %d\n",
7901                                  pf->int_name, err);
7902                         return -EFAULT;
7903                 }
7904         }
7905
7906         i40e_enable_misc_int_causes(pf);
7907
7908         /* associate no queues to the misc vector */
7909         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7910         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7911
7912         i40e_flush(hw);
7913
7914         i40e_irq_dynamic_enable_icr0(pf);
7915
7916         return err;
7917 }
7918
7919 /**
7920  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7921  * @vsi: vsi structure
7922  * @seed: RSS hash seed
7923  **/
7924 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7925                               u8 *lut, u16 lut_size)
7926 {
7927         struct i40e_aqc_get_set_rss_key_data rss_key;
7928         struct i40e_pf *pf = vsi->back;
7929         struct i40e_hw *hw = &pf->hw;
7930         bool pf_lut = false;
7931         u8 *rss_lut;
7932         int ret, i;
7933
7934         memset(&rss_key, 0, sizeof(rss_key));
7935         memcpy(&rss_key, seed, sizeof(rss_key));
7936
7937         rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7938         if (!rss_lut)
7939                 return -ENOMEM;
7940
7941         /* Populate the LUT with max no. of queues in round robin fashion */
7942         for (i = 0; i < vsi->rss_table_size; i++)
7943                 rss_lut[i] = i % vsi->rss_size;
7944
7945         ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7946         if (ret) {
7947                 dev_info(&pf->pdev->dev,
7948                          "Cannot set RSS key, err %s aq_err %s\n",
7949                          i40e_stat_str(&pf->hw, ret),
7950                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7951                 goto config_rss_aq_out;
7952         }
7953
7954         if (vsi->type == I40E_VSI_MAIN)
7955                 pf_lut = true;
7956
7957         ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7958                                   vsi->rss_table_size);
7959         if (ret)
7960                 dev_info(&pf->pdev->dev,
7961                          "Cannot set RSS lut, err %s aq_err %s\n",
7962                          i40e_stat_str(&pf->hw, ret),
7963                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7964
7965 config_rss_aq_out:
7966         kfree(rss_lut);
7967         return ret;
7968 }
7969
7970 /**
7971  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7972  * @vsi: VSI structure
7973  **/
7974 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7975 {
7976         u8 seed[I40E_HKEY_ARRAY_SIZE];
7977         struct i40e_pf *pf = vsi->back;
7978         u8 *lut;
7979         int ret;
7980
7981         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
7982                 return 0;
7983
7984         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
7985         if (!lut)
7986                 return -ENOMEM;
7987
7988         i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
7989         netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7990         vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
7991         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
7992         kfree(lut);
7993
7994         return ret;
7995 }
7996
7997 /**
7998  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
7999  * @vsi: Pointer to vsi structure
8000  * @seed: RSS hash seed
8001  * @lut: Lookup table
8002  * @lut_size: Lookup table size
8003  *
8004  * Returns 0 on success, negative on failure
8005  **/
8006 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8007                                const u8 *lut, u16 lut_size)
8008 {
8009         struct i40e_pf *pf = vsi->back;
8010         struct i40e_hw *hw = &pf->hw;
8011         u8 i;
8012
8013         /* Fill out hash function seed */
8014         if (seed) {
8015                 u32 *seed_dw = (u32 *)seed;
8016
8017                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8018                         wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
8019         }
8020
8021         if (lut) {
8022                 u32 *lut_dw = (u32 *)lut;
8023
8024                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8025                         return -EINVAL;
8026
8027                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8028                         wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8029         }
8030         i40e_flush(hw);
8031
8032         return 0;
8033 }
8034
8035 /**
8036  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8037  * @vsi: Pointer to VSI structure
8038  * @seed: Buffer to store the keys
8039  * @lut: Buffer to store the lookup table entries
8040  * @lut_size: Size of buffer to store the lookup table entries
8041  *
8042  * Returns 0 on success, negative on failure
8043  */
8044 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8045                             u8 *lut, u16 lut_size)
8046 {
8047         struct i40e_pf *pf = vsi->back;
8048         struct i40e_hw *hw = &pf->hw;
8049         u16 i;
8050
8051         if (seed) {
8052                 u32 *seed_dw = (u32 *)seed;
8053
8054                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8055                         seed_dw[i] = rd32(hw, I40E_PFQF_HKEY(i));
8056         }
8057         if (lut) {
8058                 u32 *lut_dw = (u32 *)lut;
8059
8060                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8061                         return -EINVAL;
8062                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8063                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8064         }
8065
8066         return 0;
8067 }
8068
8069 /**
8070  * i40e_config_rss - Configure RSS keys and lut
8071  * @vsi: Pointer to VSI structure
8072  * @seed: RSS hash seed
8073  * @lut: Lookup table
8074  * @lut_size: Lookup table size
8075  *
8076  * Returns 0 on success, negative on failure
8077  */
8078 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8079 {
8080         struct i40e_pf *pf = vsi->back;
8081
8082         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8083                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8084         else
8085                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8086 }
8087
8088 /**
8089  * i40e_get_rss - Get RSS keys and lut
8090  * @vsi: Pointer to VSI structure
8091  * @seed: Buffer to store the keys
8092  * @lut: Buffer to store the lookup table entries
8093  * lut_size: Size of buffer to store the lookup table entries
8094  *
8095  * Returns 0 on success, negative on failure
8096  */
8097 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8098 {
8099         return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8100 }
8101
8102 /**
8103  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8104  * @pf: Pointer to board private structure
8105  * @lut: Lookup table
8106  * @rss_table_size: Lookup table size
8107  * @rss_size: Range of queue number for hashing
8108  */
8109 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8110                               u16 rss_table_size, u16 rss_size)
8111 {
8112         u16 i;
8113
8114         for (i = 0; i < rss_table_size; i++)
8115                 lut[i] = i % rss_size;
8116 }
8117
8118 /**
8119  * i40e_pf_config_rss - Prepare for RSS if used
8120  * @pf: board private structure
8121  **/
8122 static int i40e_pf_config_rss(struct i40e_pf *pf)
8123 {
8124         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8125         u8 seed[I40E_HKEY_ARRAY_SIZE];
8126         u8 *lut;
8127         struct i40e_hw *hw = &pf->hw;
8128         u32 reg_val;
8129         u64 hena;
8130         int ret;
8131
8132         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8133         hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
8134                 ((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
8135         hena |= i40e_pf_get_default_rss_hena(pf);
8136
8137         wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
8138         wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8139
8140         /* Determine the RSS table size based on the hardware capabilities */
8141         reg_val = rd32(hw, I40E_PFQF_CTL_0);
8142         reg_val = (pf->rss_table_size == 512) ?
8143                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8144                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8145         wr32(hw, I40E_PFQF_CTL_0, reg_val);
8146
8147         /* Determine the RSS size of the VSI */
8148         if (!vsi->rss_size)
8149                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8150                                       vsi->num_queue_pairs);
8151
8152         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8153         if (!lut)
8154                 return -ENOMEM;
8155
8156         /* Use user configured lut if there is one, otherwise use default */
8157         if (vsi->rss_lut_user)
8158                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8159         else
8160                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8161
8162         /* Use user configured hash key if there is one, otherwise
8163          * use default.
8164          */
8165         if (vsi->rss_hkey_user)
8166                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8167         else
8168                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8169         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8170         kfree(lut);
8171
8172         return ret;
8173 }
8174
8175 /**
8176  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8177  * @pf: board private structure
8178  * @queue_count: the requested queue count for rss.
8179  *
8180  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8181  * count which may be different from the requested queue count.
8182  **/
8183 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8184 {
8185         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8186         int new_rss_size;
8187
8188         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8189                 return 0;
8190
8191         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8192
8193         if (queue_count != vsi->num_queue_pairs) {
8194                 vsi->req_queue_pairs = queue_count;
8195                 i40e_prep_for_reset(pf);
8196
8197                 pf->alloc_rss_size = new_rss_size;
8198
8199                 i40e_reset_and_rebuild(pf, true);
8200
8201                 /* Discard the user configured hash keys and lut, if less
8202                  * queues are enabled.
8203                  */
8204                 if (queue_count < vsi->rss_size) {
8205                         i40e_clear_rss_config_user(vsi);
8206                         dev_dbg(&pf->pdev->dev,
8207                                 "discard user configured hash keys and lut\n");
8208                 }
8209
8210                 /* Reset vsi->rss_size, as number of enabled queues changed */
8211                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8212                                       vsi->num_queue_pairs);
8213
8214                 i40e_pf_config_rss(pf);
8215         }
8216         dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8217                  pf->alloc_rss_size, pf->rss_size_max);
8218         return pf->alloc_rss_size;
8219 }
8220
8221 /**
8222  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8223  * @pf: board private structure
8224  **/
8225 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8226 {
8227         i40e_status status;
8228         bool min_valid, max_valid;
8229         u32 max_bw, min_bw;
8230
8231         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8232                                            &min_valid, &max_valid);
8233
8234         if (!status) {
8235                 if (min_valid)
8236                         pf->npar_min_bw = min_bw;
8237                 if (max_valid)
8238                         pf->npar_max_bw = max_bw;
8239         }
8240
8241         return status;
8242 }
8243
8244 /**
8245  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8246  * @pf: board private structure
8247  **/
8248 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8249 {
8250         struct i40e_aqc_configure_partition_bw_data bw_data;
8251         i40e_status status;
8252
8253         /* Set the valid bit for this PF */
8254         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8255         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8256         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8257
8258         /* Set the new bandwidths */
8259         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8260
8261         return status;
8262 }
8263
8264 /**
8265  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8266  * @pf: board private structure
8267  **/
8268 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8269 {
8270         /* Commit temporary BW setting to permanent NVM image */
8271         enum i40e_admin_queue_err last_aq_status;
8272         i40e_status ret;
8273         u16 nvm_word;
8274
8275         if (pf->hw.partition_id != 1) {
8276                 dev_info(&pf->pdev->dev,
8277                          "Commit BW only works on partition 1! This is partition %d",
8278                          pf->hw.partition_id);
8279                 ret = I40E_NOT_SUPPORTED;
8280                 goto bw_commit_out;
8281         }
8282
8283         /* Acquire NVM for read access */
8284         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8285         last_aq_status = pf->hw.aq.asq_last_status;
8286         if (ret) {
8287                 dev_info(&pf->pdev->dev,
8288                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8289                          i40e_stat_str(&pf->hw, ret),
8290                          i40e_aq_str(&pf->hw, last_aq_status));
8291                 goto bw_commit_out;
8292         }
8293
8294         /* Read word 0x10 of NVM - SW compatibility word 1 */
8295         ret = i40e_aq_read_nvm(&pf->hw,
8296                                I40E_SR_NVM_CONTROL_WORD,
8297                                0x10, sizeof(nvm_word), &nvm_word,
8298                                false, NULL);
8299         /* Save off last admin queue command status before releasing
8300          * the NVM
8301          */
8302         last_aq_status = pf->hw.aq.asq_last_status;
8303         i40e_release_nvm(&pf->hw);
8304         if (ret) {
8305                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8306                          i40e_stat_str(&pf->hw, ret),
8307                          i40e_aq_str(&pf->hw, last_aq_status));
8308                 goto bw_commit_out;
8309         }
8310
8311         /* Wait a bit for NVM release to complete */
8312         msleep(50);
8313
8314         /* Acquire NVM for write access */
8315         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8316         last_aq_status = pf->hw.aq.asq_last_status;
8317         if (ret) {
8318                 dev_info(&pf->pdev->dev,
8319                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8320                          i40e_stat_str(&pf->hw, ret),
8321                          i40e_aq_str(&pf->hw, last_aq_status));
8322                 goto bw_commit_out;
8323         }
8324         /* Write it back out unchanged to initiate update NVM,
8325          * which will force a write of the shadow (alt) RAM to
8326          * the NVM - thus storing the bandwidth values permanently.
8327          */
8328         ret = i40e_aq_update_nvm(&pf->hw,
8329                                  I40E_SR_NVM_CONTROL_WORD,
8330                                  0x10, sizeof(nvm_word),
8331                                  &nvm_word, true, NULL);
8332         /* Save off last admin queue command status before releasing
8333          * the NVM
8334          */
8335         last_aq_status = pf->hw.aq.asq_last_status;
8336         i40e_release_nvm(&pf->hw);
8337         if (ret)
8338                 dev_info(&pf->pdev->dev,
8339                          "BW settings NOT SAVED, err %s aq_err %s\n",
8340                          i40e_stat_str(&pf->hw, ret),
8341                          i40e_aq_str(&pf->hw, last_aq_status));
8342 bw_commit_out:
8343
8344         return ret;
8345 }
8346
8347 /**
8348  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8349  * @pf: board private structure to initialize
8350  *
8351  * i40e_sw_init initializes the Adapter private data structure.
8352  * Fields are initialized based on PCI device information and
8353  * OS network device settings (MTU size).
8354  **/
8355 static int i40e_sw_init(struct i40e_pf *pf)
8356 {
8357         int err = 0;
8358         int size;
8359
8360         pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8361                                 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8362         pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
8363         if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8364                 if (I40E_DEBUG_USER & debug)
8365                         pf->hw.debug_mask = debug;
8366                 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8367                                                 I40E_DEFAULT_MSG_ENABLE);
8368         }
8369
8370         /* Set default capability flags */
8371         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8372                     I40E_FLAG_MSI_ENABLED     |
8373                     I40E_FLAG_LINK_POLLING_ENABLED |
8374                     I40E_FLAG_MSIX_ENABLED;
8375
8376         if (iommu_present(&pci_bus_type))
8377                 pf->flags |= I40E_FLAG_RX_PS_ENABLED;
8378         else
8379                 pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
8380
8381         /* Set default ITR */
8382         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8383         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8384
8385         /* Depending on PF configurations, it is possible that the RSS
8386          * maximum might end up larger than the available queues
8387          */
8388         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8389         pf->alloc_rss_size = 1;
8390         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8391         pf->rss_size_max = min_t(int, pf->rss_size_max,
8392                                  pf->hw.func_caps.num_tx_qp);
8393         if (pf->hw.func_caps.rss) {
8394                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8395                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8396                                            num_online_cpus());
8397         }
8398
8399         /* MFP mode enabled */
8400         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8401                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8402                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8403                 if (i40e_get_npar_bw_setting(pf))
8404                         dev_warn(&pf->pdev->dev,
8405                                  "Could not get NPAR bw settings\n");
8406                 else
8407                         dev_info(&pf->pdev->dev,
8408                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8409                                  pf->npar_min_bw, pf->npar_max_bw);
8410         }
8411
8412         /* FW/NVM is not yet fixed in this regard */
8413         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8414             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8415                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8416                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8417                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8418                     pf->hw.num_partitions > 1)
8419                         dev_info(&pf->pdev->dev,
8420                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8421                 else
8422                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8423                 pf->fdir_pf_filter_count =
8424                                  pf->hw.func_caps.fd_filters_guaranteed;
8425                 pf->hw.fdir_shared_filter_count =
8426                                  pf->hw.func_caps.fd_filters_best_effort;
8427         }
8428
8429         if (pf->hw.func_caps.vmdq) {
8430                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8431                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8432                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8433         }
8434
8435         if (pf->hw.func_caps.iwarp) {
8436                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8437                 /* IWARP needs one extra vector for CQP just like MISC.*/
8438                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8439         }
8440
8441 #ifdef I40E_FCOE
8442         i40e_init_pf_fcoe(pf);
8443
8444 #endif /* I40E_FCOE */
8445 #ifdef CONFIG_PCI_IOV
8446         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8447                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8448                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8449                 pf->num_req_vfs = min_t(int,
8450                                         pf->hw.func_caps.num_vfs,
8451                                         I40E_MAX_VF_COUNT);
8452         }
8453 #endif /* CONFIG_PCI_IOV */
8454         if (pf->hw.mac.type == I40E_MAC_X722) {
8455                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8456                              I40E_FLAG_128_QP_RSS_CAPABLE |
8457                              I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8458                              I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8459                              I40E_FLAG_WB_ON_ITR_CAPABLE |
8460                              I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8461                              I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8462         }
8463         pf->eeprom_version = 0xDEAD;
8464         pf->lan_veb = I40E_NO_VEB;
8465         pf->lan_vsi = I40E_NO_VSI;
8466
8467         /* By default FW has this off for performance reasons */
8468         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8469
8470         /* set up queue assignment tracking */
8471         size = sizeof(struct i40e_lump_tracking)
8472                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8473         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8474         if (!pf->qp_pile) {
8475                 err = -ENOMEM;
8476                 goto sw_init_done;
8477         }
8478         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8479         pf->qp_pile->search_hint = 0;
8480
8481         pf->tx_timeout_recovery_level = 1;
8482
8483         mutex_init(&pf->switch_mutex);
8484
8485         /* If NPAR is enabled nudge the Tx scheduler */
8486         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8487                 i40e_set_npar_bw_setting(pf);
8488
8489 sw_init_done:
8490         return err;
8491 }
8492
8493 /**
8494  * i40e_set_ntuple - set the ntuple feature flag and take action
8495  * @pf: board private structure to initialize
8496  * @features: the feature set that the stack is suggesting
8497  *
8498  * returns a bool to indicate if reset needs to happen
8499  **/
8500 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8501 {
8502         bool need_reset = false;
8503
8504         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8505          * the state changed, we need to reset.
8506          */
8507         if (features & NETIF_F_NTUPLE) {
8508                 /* Enable filters and mark for reset */
8509                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8510                         need_reset = true;
8511                 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8512         } else {
8513                 /* turn off filters, mark for reset and clear SW filter list */
8514                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8515                         need_reset = true;
8516                         i40e_fdir_filter_exit(pf);
8517                 }
8518                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8519                 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8520                 /* reset fd counters */
8521                 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8522                 pf->fdir_pf_active_filters = 0;
8523                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8524                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8525                         dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8526                 /* if ATR was auto disabled it can be re-enabled. */
8527                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8528                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8529                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8530         }
8531         return need_reset;
8532 }
8533
8534 /**
8535  * i40e_set_features - set the netdev feature flags
8536  * @netdev: ptr to the netdev being adjusted
8537  * @features: the feature set that the stack is suggesting
8538  **/
8539 static int i40e_set_features(struct net_device *netdev,
8540                              netdev_features_t features)
8541 {
8542         struct i40e_netdev_priv *np = netdev_priv(netdev);
8543         struct i40e_vsi *vsi = np->vsi;
8544         struct i40e_pf *pf = vsi->back;
8545         bool need_reset;
8546
8547         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8548                 i40e_vlan_stripping_enable(vsi);
8549         else
8550                 i40e_vlan_stripping_disable(vsi);
8551
8552         need_reset = i40e_set_ntuple(pf, features);
8553
8554         if (need_reset)
8555                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8556
8557         return 0;
8558 }
8559
8560 #if IS_ENABLED(CONFIG_VXLAN) || IS_ENABLED(CONFIG_GENEVE)
8561 /**
8562  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8563  * @pf: board private structure
8564  * @port: The UDP port to look up
8565  *
8566  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8567  **/
8568 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8569 {
8570         u8 i;
8571
8572         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8573                 if (pf->udp_ports[i].index == port)
8574                         return i;
8575         }
8576
8577         return i;
8578 }
8579
8580 #endif
8581
8582 #if IS_ENABLED(CONFIG_VXLAN)
8583 /**
8584  * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8585  * @netdev: This physical port's netdev
8586  * @sa_family: Socket Family that VXLAN is notifying us about
8587  * @port: New UDP port number that VXLAN started listening to
8588  **/
8589 static void i40e_add_vxlan_port(struct net_device *netdev,
8590                                 sa_family_t sa_family, __be16 port)
8591 {
8592         struct i40e_netdev_priv *np = netdev_priv(netdev);
8593         struct i40e_vsi *vsi = np->vsi;
8594         struct i40e_pf *pf = vsi->back;
8595         u8 next_idx;
8596         u8 idx;
8597
8598         if (sa_family == AF_INET6)
8599                 return;
8600
8601         idx = i40e_get_udp_port_idx(pf, port);
8602
8603         /* Check if port already exists */
8604         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8605                 netdev_info(netdev, "vxlan port %d already offloaded\n",
8606                             ntohs(port));
8607                 return;
8608         }
8609
8610         /* Now check if there is space to add the new port */
8611         next_idx = i40e_get_udp_port_idx(pf, 0);
8612
8613         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8614                 netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8615                             ntohs(port));
8616                 return;
8617         }
8618
8619         /* New port: add it and mark its index in the bitmap */
8620         pf->udp_ports[next_idx].index = port;
8621         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8622         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8623         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8624 }
8625
8626 /**
8627  * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8628  * @netdev: This physical port's netdev
8629  * @sa_family: Socket Family that VXLAN is notifying us about
8630  * @port: UDP port number that VXLAN stopped listening to
8631  **/
8632 static void i40e_del_vxlan_port(struct net_device *netdev,
8633                                 sa_family_t sa_family, __be16 port)
8634 {
8635         struct i40e_netdev_priv *np = netdev_priv(netdev);
8636         struct i40e_vsi *vsi = np->vsi;
8637         struct i40e_pf *pf = vsi->back;
8638         u8 idx;
8639
8640         if (sa_family == AF_INET6)
8641                 return;
8642
8643         idx = i40e_get_udp_port_idx(pf, port);
8644
8645         /* Check if port already exists */
8646         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8647                 /* if port exists, set it to 0 (mark for deletion)
8648                  * and make it pending
8649                  */
8650                 pf->udp_ports[idx].index = 0;
8651                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8652                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8653         } else {
8654                 netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8655                             ntohs(port));
8656         }
8657 }
8658 #endif
8659
8660 #if IS_ENABLED(CONFIG_GENEVE)
8661 /**
8662  * i40e_add_geneve_port - Get notifications about GENEVE ports that come up
8663  * @netdev: This physical port's netdev
8664  * @sa_family: Socket Family that GENEVE is notifying us about
8665  * @port: New UDP port number that GENEVE started listening to
8666  **/
8667 static void i40e_add_geneve_port(struct net_device *netdev,
8668                                  sa_family_t sa_family, __be16 port)
8669 {
8670         struct i40e_netdev_priv *np = netdev_priv(netdev);
8671         struct i40e_vsi *vsi = np->vsi;
8672         struct i40e_pf *pf = vsi->back;
8673         u8 next_idx;
8674         u8 idx;
8675
8676         if (sa_family == AF_INET6)
8677                 return;
8678
8679         idx = i40e_get_udp_port_idx(pf, port);
8680
8681         /* Check if port already exists */
8682         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8683                 netdev_info(netdev, "udp port %d already offloaded\n",
8684                             ntohs(port));
8685                 return;
8686         }
8687
8688         /* Now check if there is space to add the new port */
8689         next_idx = i40e_get_udp_port_idx(pf, 0);
8690
8691         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8692                 netdev_info(netdev, "maximum number of UDP ports reached, not adding port %d\n",
8693                             ntohs(port));
8694                 return;
8695         }
8696
8697         /* New port: add it and mark its index in the bitmap */
8698         pf->udp_ports[next_idx].index = port;
8699         pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8700         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8701         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8702
8703         dev_info(&pf->pdev->dev, "adding geneve port %d\n", ntohs(port));
8704 }
8705
8706 /**
8707  * i40e_del_geneve_port - Get notifications about GENEVE ports that go away
8708  * @netdev: This physical port's netdev
8709  * @sa_family: Socket Family that GENEVE is notifying us about
8710  * @port: UDP port number that GENEVE stopped listening to
8711  **/
8712 static void i40e_del_geneve_port(struct net_device *netdev,
8713                                  sa_family_t sa_family, __be16 port)
8714 {
8715         struct i40e_netdev_priv *np = netdev_priv(netdev);
8716         struct i40e_vsi *vsi = np->vsi;
8717         struct i40e_pf *pf = vsi->back;
8718         u8 idx;
8719
8720         if (sa_family == AF_INET6)
8721                 return;
8722
8723         idx = i40e_get_udp_port_idx(pf, port);
8724
8725         /* Check if port already exists */
8726         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8727                 /* if port exists, set it to 0 (mark for deletion)
8728                  * and make it pending
8729                  */
8730                 pf->udp_ports[idx].index = 0;
8731                 pf->pending_udp_bitmap |= BIT_ULL(idx);
8732                 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8733
8734                 dev_info(&pf->pdev->dev, "deleting geneve port %d\n",
8735                          ntohs(port));
8736         } else {
8737                 netdev_warn(netdev, "geneve port %d was not found, not deleting\n",
8738                             ntohs(port));
8739         }
8740 }
8741 #endif
8742
8743 static int i40e_get_phys_port_id(struct net_device *netdev,
8744                                  struct netdev_phys_item_id *ppid)
8745 {
8746         struct i40e_netdev_priv *np = netdev_priv(netdev);
8747         struct i40e_pf *pf = np->vsi->back;
8748         struct i40e_hw *hw = &pf->hw;
8749
8750         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8751                 return -EOPNOTSUPP;
8752
8753         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8754         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8755
8756         return 0;
8757 }
8758
8759 /**
8760  * i40e_ndo_fdb_add - add an entry to the hardware database
8761  * @ndm: the input from the stack
8762  * @tb: pointer to array of nladdr (unused)
8763  * @dev: the net device pointer
8764  * @addr: the MAC address entry being added
8765  * @flags: instructions from stack about fdb operation
8766  */
8767 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8768                             struct net_device *dev,
8769                             const unsigned char *addr, u16 vid,
8770                             u16 flags)
8771 {
8772         struct i40e_netdev_priv *np = netdev_priv(dev);
8773         struct i40e_pf *pf = np->vsi->back;
8774         int err = 0;
8775
8776         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8777                 return -EOPNOTSUPP;
8778
8779         if (vid) {
8780                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8781                 return -EINVAL;
8782         }
8783
8784         /* Hardware does not support aging addresses so if a
8785          * ndm_state is given only allow permanent addresses
8786          */
8787         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8788                 netdev_info(dev, "FDB only supports static addresses\n");
8789                 return -EINVAL;
8790         }
8791
8792         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8793                 err = dev_uc_add_excl(dev, addr);
8794         else if (is_multicast_ether_addr(addr))
8795                 err = dev_mc_add_excl(dev, addr);
8796         else
8797                 err = -EINVAL;
8798
8799         /* Only return duplicate errors if NLM_F_EXCL is set */
8800         if (err == -EEXIST && !(flags & NLM_F_EXCL))
8801                 err = 0;
8802
8803         return err;
8804 }
8805
8806 /**
8807  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8808  * @dev: the netdev being configured
8809  * @nlh: RTNL message
8810  *
8811  * Inserts a new hardware bridge if not already created and
8812  * enables the bridging mode requested (VEB or VEPA). If the
8813  * hardware bridge has already been inserted and the request
8814  * is to change the mode then that requires a PF reset to
8815  * allow rebuild of the components with required hardware
8816  * bridge mode enabled.
8817  **/
8818 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8819                                    struct nlmsghdr *nlh,
8820                                    u16 flags)
8821 {
8822         struct i40e_netdev_priv *np = netdev_priv(dev);
8823         struct i40e_vsi *vsi = np->vsi;
8824         struct i40e_pf *pf = vsi->back;
8825         struct i40e_veb *veb = NULL;
8826         struct nlattr *attr, *br_spec;
8827         int i, rem;
8828
8829         /* Only for PF VSI for now */
8830         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8831                 return -EOPNOTSUPP;
8832
8833         /* Find the HW bridge for PF VSI */
8834         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8835                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8836                         veb = pf->veb[i];
8837         }
8838
8839         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8840
8841         nla_for_each_nested(attr, br_spec, rem) {
8842                 __u16 mode;
8843
8844                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8845                         continue;
8846
8847                 mode = nla_get_u16(attr);
8848                 if ((mode != BRIDGE_MODE_VEPA) &&
8849                     (mode != BRIDGE_MODE_VEB))
8850                         return -EINVAL;
8851
8852                 /* Insert a new HW bridge */
8853                 if (!veb) {
8854                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8855                                              vsi->tc_config.enabled_tc);
8856                         if (veb) {
8857                                 veb->bridge_mode = mode;
8858                                 i40e_config_bridge_mode(veb);
8859                         } else {
8860                                 /* No Bridge HW offload available */
8861                                 return -ENOENT;
8862                         }
8863                         break;
8864                 } else if (mode != veb->bridge_mode) {
8865                         /* Existing HW bridge but different mode needs reset */
8866                         veb->bridge_mode = mode;
8867                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8868                         if (mode == BRIDGE_MODE_VEB)
8869                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8870                         else
8871                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8872                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8873                         break;
8874                 }
8875         }
8876
8877         return 0;
8878 }
8879
8880 /**
8881  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8882  * @skb: skb buff
8883  * @pid: process id
8884  * @seq: RTNL message seq #
8885  * @dev: the netdev being configured
8886  * @filter_mask: unused
8887  * @nlflags: netlink flags passed in
8888  *
8889  * Return the mode in which the hardware bridge is operating in
8890  * i.e VEB or VEPA.
8891  **/
8892 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8893                                    struct net_device *dev,
8894                                    u32 __always_unused filter_mask,
8895                                    int nlflags)
8896 {
8897         struct i40e_netdev_priv *np = netdev_priv(dev);
8898         struct i40e_vsi *vsi = np->vsi;
8899         struct i40e_pf *pf = vsi->back;
8900         struct i40e_veb *veb = NULL;
8901         int i;
8902
8903         /* Only for PF VSI for now */
8904         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8905                 return -EOPNOTSUPP;
8906
8907         /* Find the HW bridge for the PF VSI */
8908         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8909                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8910                         veb = pf->veb[i];
8911         }
8912
8913         if (!veb)
8914                 return 0;
8915
8916         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8917                                        nlflags, 0, 0, filter_mask, NULL);
8918 }
8919
8920 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
8921  * inner mac plus all inner ethertypes.
8922  */
8923 #define I40E_MAX_TUNNEL_HDR_LEN 128
8924 /**
8925  * i40e_features_check - Validate encapsulated packet conforms to limits
8926  * @skb: skb buff
8927  * @dev: This physical port's netdev
8928  * @features: Offload features that the stack believes apply
8929  **/
8930 static netdev_features_t i40e_features_check(struct sk_buff *skb,
8931                                              struct net_device *dev,
8932                                              netdev_features_t features)
8933 {
8934         if (skb->encapsulation &&
8935             ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
8936              I40E_MAX_TUNNEL_HDR_LEN))
8937                 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8938
8939         return features;
8940 }
8941
8942 static const struct net_device_ops i40e_netdev_ops = {
8943         .ndo_open               = i40e_open,
8944         .ndo_stop               = i40e_close,
8945         .ndo_start_xmit         = i40e_lan_xmit_frame,
8946         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
8947         .ndo_set_rx_mode        = i40e_set_rx_mode,
8948         .ndo_validate_addr      = eth_validate_addr,
8949         .ndo_set_mac_address    = i40e_set_mac,
8950         .ndo_change_mtu         = i40e_change_mtu,
8951         .ndo_do_ioctl           = i40e_ioctl,
8952         .ndo_tx_timeout         = i40e_tx_timeout,
8953         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
8954         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
8955 #ifdef CONFIG_NET_POLL_CONTROLLER
8956         .ndo_poll_controller    = i40e_netpoll,
8957 #endif
8958         .ndo_setup_tc           = i40e_setup_tc,
8959 #ifdef I40E_FCOE
8960         .ndo_fcoe_enable        = i40e_fcoe_enable,
8961         .ndo_fcoe_disable       = i40e_fcoe_disable,
8962 #endif
8963         .ndo_set_features       = i40e_set_features,
8964         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
8965         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
8966         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
8967         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
8968         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
8969         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
8970 #if IS_ENABLED(CONFIG_VXLAN)
8971         .ndo_add_vxlan_port     = i40e_add_vxlan_port,
8972         .ndo_del_vxlan_port     = i40e_del_vxlan_port,
8973 #endif
8974 #if IS_ENABLED(CONFIG_GENEVE)
8975         .ndo_add_geneve_port    = i40e_add_geneve_port,
8976         .ndo_del_geneve_port    = i40e_del_geneve_port,
8977 #endif
8978         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
8979         .ndo_fdb_add            = i40e_ndo_fdb_add,
8980         .ndo_features_check     = i40e_features_check,
8981         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
8982         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
8983 };
8984
8985 /**
8986  * i40e_config_netdev - Setup the netdev flags
8987  * @vsi: the VSI being configured
8988  *
8989  * Returns 0 on success, negative value on failure
8990  **/
8991 static int i40e_config_netdev(struct i40e_vsi *vsi)
8992 {
8993         u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8994         struct i40e_pf *pf = vsi->back;
8995         struct i40e_hw *hw = &pf->hw;
8996         struct i40e_netdev_priv *np;
8997         struct net_device *netdev;
8998         u8 mac_addr[ETH_ALEN];
8999         int etherdev_size;
9000
9001         etherdev_size = sizeof(struct i40e_netdev_priv);
9002         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9003         if (!netdev)
9004                 return -ENOMEM;
9005
9006         vsi->netdev = netdev;
9007         np = netdev_priv(netdev);
9008         np->vsi = vsi;
9009
9010         netdev->hw_enc_features |= NETIF_F_IP_CSUM       |
9011                                   NETIF_F_RXCSUM         |
9012                                   NETIF_F_GSO_UDP_TUNNEL |
9013                                   NETIF_F_GSO_GRE        |
9014                                   NETIF_F_TSO;
9015
9016         netdev->features = NETIF_F_SG                  |
9017                            NETIF_F_IP_CSUM             |
9018                            NETIF_F_SCTP_CRC            |
9019                            NETIF_F_HIGHDMA             |
9020                            NETIF_F_GSO_UDP_TUNNEL      |
9021                            NETIF_F_GSO_GRE             |
9022                            NETIF_F_HW_VLAN_CTAG_TX     |
9023                            NETIF_F_HW_VLAN_CTAG_RX     |
9024                            NETIF_F_HW_VLAN_CTAG_FILTER |
9025                            NETIF_F_IPV6_CSUM           |
9026                            NETIF_F_TSO                 |
9027                            NETIF_F_TSO_ECN             |
9028                            NETIF_F_TSO6                |
9029                            NETIF_F_RXCSUM              |
9030                            NETIF_F_RXHASH              |
9031                            0;
9032
9033         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9034                 netdev->features |= NETIF_F_NTUPLE;
9035
9036         /* copy netdev features into list of user selectable features */
9037         netdev->hw_features |= netdev->features;
9038
9039         if (vsi->type == I40E_VSI_MAIN) {
9040                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9041                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9042                 /* The following steps are necessary to prevent reception
9043                  * of tagged packets - some older NVM configurations load a
9044                  * default a MAC-VLAN filter that accepts any tagged packet
9045                  * which must be replaced by a normal filter.
9046                  */
9047                 if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
9048                         spin_lock_bh(&vsi->mac_filter_list_lock);
9049                         i40e_add_filter(vsi, mac_addr,
9050                                         I40E_VLAN_ANY, false, true);
9051                         spin_unlock_bh(&vsi->mac_filter_list_lock);
9052                 }
9053         } else {
9054                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9055                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9056                          pf->vsi[pf->lan_vsi]->netdev->name);
9057                 random_ether_addr(mac_addr);
9058
9059                 spin_lock_bh(&vsi->mac_filter_list_lock);
9060                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9061                 spin_unlock_bh(&vsi->mac_filter_list_lock);
9062         }
9063
9064         spin_lock_bh(&vsi->mac_filter_list_lock);
9065         i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
9066         spin_unlock_bh(&vsi->mac_filter_list_lock);
9067
9068         ether_addr_copy(netdev->dev_addr, mac_addr);
9069         ether_addr_copy(netdev->perm_addr, mac_addr);
9070         /* vlan gets same features (except vlan offload)
9071          * after any tweaks for specific VSI types
9072          */
9073         netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
9074                                                      NETIF_F_HW_VLAN_CTAG_RX |
9075                                                    NETIF_F_HW_VLAN_CTAG_FILTER);
9076         netdev->priv_flags |= IFF_UNICAST_FLT;
9077         netdev->priv_flags |= IFF_SUPP_NOFCS;
9078         /* Setup netdev TC information */
9079         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9080
9081         netdev->netdev_ops = &i40e_netdev_ops;
9082         netdev->watchdog_timeo = 5 * HZ;
9083         i40e_set_ethtool_ops(netdev);
9084 #ifdef I40E_FCOE
9085         i40e_fcoe_config_netdev(netdev, vsi);
9086 #endif
9087
9088         return 0;
9089 }
9090
9091 /**
9092  * i40e_vsi_delete - Delete a VSI from the switch
9093  * @vsi: the VSI being removed
9094  *
9095  * Returns 0 on success, negative value on failure
9096  **/
9097 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9098 {
9099         /* remove default VSI is not allowed */
9100         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9101                 return;
9102
9103         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9104 }
9105
9106 /**
9107  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9108  * @vsi: the VSI being queried
9109  *
9110  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9111  **/
9112 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9113 {
9114         struct i40e_veb *veb;
9115         struct i40e_pf *pf = vsi->back;
9116
9117         /* Uplink is not a bridge so default to VEB */
9118         if (vsi->veb_idx == I40E_NO_VEB)
9119                 return 1;
9120
9121         veb = pf->veb[vsi->veb_idx];
9122         if (!veb) {
9123                 dev_info(&pf->pdev->dev,
9124                          "There is no veb associated with the bridge\n");
9125                 return -ENOENT;
9126         }
9127
9128         /* Uplink is a bridge in VEPA mode */
9129         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9130                 return 0;
9131         } else {
9132                 /* Uplink is a bridge in VEB mode */
9133                 return 1;
9134         }
9135
9136         /* VEPA is now default bridge, so return 0 */
9137         return 0;
9138 }
9139
9140 /**
9141  * i40e_add_vsi - Add a VSI to the switch
9142  * @vsi: the VSI being configured
9143  *
9144  * This initializes a VSI context depending on the VSI type to be added and
9145  * passes it down to the add_vsi aq command.
9146  **/
9147 static int i40e_add_vsi(struct i40e_vsi *vsi)
9148 {
9149         int ret = -ENODEV;
9150         u8 laa_macaddr[ETH_ALEN];
9151         bool found_laa_mac_filter = false;
9152         struct i40e_pf *pf = vsi->back;
9153         struct i40e_hw *hw = &pf->hw;
9154         struct i40e_vsi_context ctxt;
9155         struct i40e_mac_filter *f, *ftmp;
9156
9157         u8 enabled_tc = 0x1; /* TC0 enabled */
9158         int f_count = 0;
9159
9160         memset(&ctxt, 0, sizeof(ctxt));
9161         switch (vsi->type) {
9162         case I40E_VSI_MAIN:
9163                 /* The PF's main VSI is already setup as part of the
9164                  * device initialization, so we'll not bother with
9165                  * the add_vsi call, but we will retrieve the current
9166                  * VSI context.
9167                  */
9168                 ctxt.seid = pf->main_vsi_seid;
9169                 ctxt.pf_num = pf->hw.pf_id;
9170                 ctxt.vf_num = 0;
9171                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9172                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9173                 if (ret) {
9174                         dev_info(&pf->pdev->dev,
9175                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9176                                  i40e_stat_str(&pf->hw, ret),
9177                                  i40e_aq_str(&pf->hw,
9178                                              pf->hw.aq.asq_last_status));
9179                         return -ENOENT;
9180                 }
9181                 vsi->info = ctxt.info;
9182                 vsi->info.valid_sections = 0;
9183
9184                 vsi->seid = ctxt.seid;
9185                 vsi->id = ctxt.vsi_number;
9186
9187                 enabled_tc = i40e_pf_get_tc_map(pf);
9188
9189                 /* MFP mode setup queue map and update VSI */
9190                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9191                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9192                         memset(&ctxt, 0, sizeof(ctxt));
9193                         ctxt.seid = pf->main_vsi_seid;
9194                         ctxt.pf_num = pf->hw.pf_id;
9195                         ctxt.vf_num = 0;
9196                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9197                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9198                         if (ret) {
9199                                 dev_info(&pf->pdev->dev,
9200                                          "update vsi failed, err %s aq_err %s\n",
9201                                          i40e_stat_str(&pf->hw, ret),
9202                                          i40e_aq_str(&pf->hw,
9203                                                     pf->hw.aq.asq_last_status));
9204                                 ret = -ENOENT;
9205                                 goto err;
9206                         }
9207                         /* update the local VSI info queue map */
9208                         i40e_vsi_update_queue_map(vsi, &ctxt);
9209                         vsi->info.valid_sections = 0;
9210                 } else {
9211                         /* Default/Main VSI is only enabled for TC0
9212                          * reconfigure it to enable all TCs that are
9213                          * available on the port in SFP mode.
9214                          * For MFP case the iSCSI PF would use this
9215                          * flow to enable LAN+iSCSI TC.
9216                          */
9217                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9218                         if (ret) {
9219                                 dev_info(&pf->pdev->dev,
9220                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9221                                          enabled_tc,
9222                                          i40e_stat_str(&pf->hw, ret),
9223                                          i40e_aq_str(&pf->hw,
9224                                                     pf->hw.aq.asq_last_status));
9225                                 ret = -ENOENT;
9226                         }
9227                 }
9228                 break;
9229
9230         case I40E_VSI_FDIR:
9231                 ctxt.pf_num = hw->pf_id;
9232                 ctxt.vf_num = 0;
9233                 ctxt.uplink_seid = vsi->uplink_seid;
9234                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9235                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9236                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9237                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9238                         ctxt.info.valid_sections |=
9239                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9240                         ctxt.info.switch_id =
9241                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9242                 }
9243                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9244                 break;
9245
9246         case I40E_VSI_VMDQ2:
9247                 ctxt.pf_num = hw->pf_id;
9248                 ctxt.vf_num = 0;
9249                 ctxt.uplink_seid = vsi->uplink_seid;
9250                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9251                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9252
9253                 /* This VSI is connected to VEB so the switch_id
9254                  * should be set to zero by default.
9255                  */
9256                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9257                         ctxt.info.valid_sections |=
9258                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9259                         ctxt.info.switch_id =
9260                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9261                 }
9262
9263                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9264                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9265                 break;
9266
9267         case I40E_VSI_SRIOV:
9268                 ctxt.pf_num = hw->pf_id;
9269                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9270                 ctxt.uplink_seid = vsi->uplink_seid;
9271                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9272                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9273
9274                 /* This VSI is connected to VEB so the switch_id
9275                  * should be set to zero by default.
9276                  */
9277                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9278                         ctxt.info.valid_sections |=
9279                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9280                         ctxt.info.switch_id =
9281                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9282                 }
9283
9284                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9285                         ctxt.info.valid_sections |=
9286                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9287                         ctxt.info.queueing_opt_flags |=
9288                                                 I40E_AQ_VSI_QUE_OPT_TCP_ENA;
9289                 }
9290
9291                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9292                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9293                 if (pf->vf[vsi->vf_id].spoofchk) {
9294                         ctxt.info.valid_sections |=
9295                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9296                         ctxt.info.sec_flags |=
9297                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9298                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9299                 }
9300                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9301                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9302                 break;
9303
9304 #ifdef I40E_FCOE
9305         case I40E_VSI_FCOE:
9306                 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9307                 if (ret) {
9308                         dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9309                         return ret;
9310                 }
9311                 break;
9312
9313 #endif /* I40E_FCOE */
9314         case I40E_VSI_IWARP:
9315                 /* send down message to iWARP */
9316                 break;
9317
9318         default:
9319                 return -ENODEV;
9320         }
9321
9322         if (vsi->type != I40E_VSI_MAIN) {
9323                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9324                 if (ret) {
9325                         dev_info(&vsi->back->pdev->dev,
9326                                  "add vsi failed, err %s aq_err %s\n",
9327                                  i40e_stat_str(&pf->hw, ret),
9328                                  i40e_aq_str(&pf->hw,
9329                                              pf->hw.aq.asq_last_status));
9330                         ret = -ENOENT;
9331                         goto err;
9332                 }
9333                 vsi->info = ctxt.info;
9334                 vsi->info.valid_sections = 0;
9335                 vsi->seid = ctxt.seid;
9336                 vsi->id = ctxt.vsi_number;
9337         }
9338
9339         spin_lock_bh(&vsi->mac_filter_list_lock);
9340         /* If macvlan filters already exist, force them to get loaded */
9341         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9342                 f->changed = true;
9343                 f_count++;
9344
9345                 /* Expected to have only one MAC filter entry for LAA in list */
9346                 if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
9347                         ether_addr_copy(laa_macaddr, f->macaddr);
9348                         found_laa_mac_filter = true;
9349                 }
9350         }
9351         spin_unlock_bh(&vsi->mac_filter_list_lock);
9352
9353         if (found_laa_mac_filter) {
9354                 struct i40e_aqc_remove_macvlan_element_data element;
9355
9356                 memset(&element, 0, sizeof(element));
9357                 ether_addr_copy(element.mac_addr, laa_macaddr);
9358                 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
9359                 ret = i40e_aq_remove_macvlan(hw, vsi->seid,
9360                                              &element, 1, NULL);
9361                 if (ret) {
9362                         /* some older FW has a different default */
9363                         element.flags |=
9364                                        I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
9365                         i40e_aq_remove_macvlan(hw, vsi->seid,
9366                                                &element, 1, NULL);
9367                 }
9368
9369                 i40e_aq_mac_address_write(hw,
9370                                           I40E_AQC_WRITE_TYPE_LAA_WOL,
9371                                           laa_macaddr, NULL);
9372         }
9373
9374         if (f_count) {
9375                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9376                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9377         }
9378
9379         /* Update VSI BW information */
9380         ret = i40e_vsi_get_bw_info(vsi);
9381         if (ret) {
9382                 dev_info(&pf->pdev->dev,
9383                          "couldn't get vsi bw info, err %s aq_err %s\n",
9384                          i40e_stat_str(&pf->hw, ret),
9385                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9386                 /* VSI is already added so not tearing that up */
9387                 ret = 0;
9388         }
9389
9390 err:
9391         return ret;
9392 }
9393
9394 /**
9395  * i40e_vsi_release - Delete a VSI and free its resources
9396  * @vsi: the VSI being removed
9397  *
9398  * Returns 0 on success or < 0 on error
9399  **/
9400 int i40e_vsi_release(struct i40e_vsi *vsi)
9401 {
9402         struct i40e_mac_filter *f, *ftmp;
9403         struct i40e_veb *veb = NULL;
9404         struct i40e_pf *pf;
9405         u16 uplink_seid;
9406         int i, n;
9407
9408         pf = vsi->back;
9409
9410         /* release of a VEB-owner or last VSI is not allowed */
9411         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9412                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9413                          vsi->seid, vsi->uplink_seid);
9414                 return -ENODEV;
9415         }
9416         if (vsi == pf->vsi[pf->lan_vsi] &&
9417             !test_bit(__I40E_DOWN, &pf->state)) {
9418                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9419                 return -ENODEV;
9420         }
9421
9422         uplink_seid = vsi->uplink_seid;
9423         if (vsi->type != I40E_VSI_SRIOV) {
9424                 if (vsi->netdev_registered) {
9425                         vsi->netdev_registered = false;
9426                         if (vsi->netdev) {
9427                                 /* results in a call to i40e_close() */
9428                                 unregister_netdev(vsi->netdev);
9429                         }
9430                 } else {
9431                         i40e_vsi_close(vsi);
9432                 }
9433                 i40e_vsi_disable_irq(vsi);
9434         }
9435
9436         spin_lock_bh(&vsi->mac_filter_list_lock);
9437         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9438                 i40e_del_filter(vsi, f->macaddr, f->vlan,
9439                                 f->is_vf, f->is_netdev);
9440         spin_unlock_bh(&vsi->mac_filter_list_lock);
9441
9442         i40e_sync_vsi_filters(vsi);
9443
9444         i40e_vsi_delete(vsi);
9445         i40e_vsi_free_q_vectors(vsi);
9446         if (vsi->netdev) {
9447                 free_netdev(vsi->netdev);
9448                 vsi->netdev = NULL;
9449         }
9450         i40e_vsi_clear_rings(vsi);
9451         i40e_vsi_clear(vsi);
9452
9453         /* If this was the last thing on the VEB, except for the
9454          * controlling VSI, remove the VEB, which puts the controlling
9455          * VSI onto the next level down in the switch.
9456          *
9457          * Well, okay, there's one more exception here: don't remove
9458          * the orphan VEBs yet.  We'll wait for an explicit remove request
9459          * from up the network stack.
9460          */
9461         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9462                 if (pf->vsi[i] &&
9463                     pf->vsi[i]->uplink_seid == uplink_seid &&
9464                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9465                         n++;      /* count the VSIs */
9466                 }
9467         }
9468         for (i = 0; i < I40E_MAX_VEB; i++) {
9469                 if (!pf->veb[i])
9470                         continue;
9471                 if (pf->veb[i]->uplink_seid == uplink_seid)
9472                         n++;     /* count the VEBs */
9473                 if (pf->veb[i]->seid == uplink_seid)
9474                         veb = pf->veb[i];
9475         }
9476         if (n == 0 && veb && veb->uplink_seid != 0)
9477                 i40e_veb_release(veb);
9478
9479         return 0;
9480 }
9481
9482 /**
9483  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9484  * @vsi: ptr to the VSI
9485  *
9486  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9487  * corresponding SW VSI structure and initializes num_queue_pairs for the
9488  * newly allocated VSI.
9489  *
9490  * Returns 0 on success or negative on failure
9491  **/
9492 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9493 {
9494         int ret = -ENOENT;
9495         struct i40e_pf *pf = vsi->back;
9496
9497         if (vsi->q_vectors[0]) {
9498                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9499                          vsi->seid);
9500                 return -EEXIST;
9501         }
9502
9503         if (vsi->base_vector) {
9504                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9505                          vsi->seid, vsi->base_vector);
9506                 return -EEXIST;
9507         }
9508
9509         ret = i40e_vsi_alloc_q_vectors(vsi);
9510         if (ret) {
9511                 dev_info(&pf->pdev->dev,
9512                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9513                          vsi->num_q_vectors, vsi->seid, ret);
9514                 vsi->num_q_vectors = 0;
9515                 goto vector_setup_out;
9516         }
9517
9518         /* In Legacy mode, we do not have to get any other vector since we
9519          * piggyback on the misc/ICR0 for queue interrupts.
9520         */
9521         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9522                 return ret;
9523         if (vsi->num_q_vectors)
9524                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9525                                                  vsi->num_q_vectors, vsi->idx);
9526         if (vsi->base_vector < 0) {
9527                 dev_info(&pf->pdev->dev,
9528                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9529                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9530                 i40e_vsi_free_q_vectors(vsi);
9531                 ret = -ENOENT;
9532                 goto vector_setup_out;
9533         }
9534
9535 vector_setup_out:
9536         return ret;
9537 }
9538
9539 /**
9540  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9541  * @vsi: pointer to the vsi.
9542  *
9543  * This re-allocates a vsi's queue resources.
9544  *
9545  * Returns pointer to the successfully allocated and configured VSI sw struct
9546  * on success, otherwise returns NULL on failure.
9547  **/
9548 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9549 {
9550         struct i40e_pf *pf = vsi->back;
9551         u8 enabled_tc;
9552         int ret;
9553
9554         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9555         i40e_vsi_clear_rings(vsi);
9556
9557         i40e_vsi_free_arrays(vsi, false);
9558         i40e_set_num_rings_in_vsi(vsi);
9559         ret = i40e_vsi_alloc_arrays(vsi, false);
9560         if (ret)
9561                 goto err_vsi;
9562
9563         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9564         if (ret < 0) {
9565                 dev_info(&pf->pdev->dev,
9566                          "failed to get tracking for %d queues for VSI %d err %d\n",
9567                          vsi->alloc_queue_pairs, vsi->seid, ret);
9568                 goto err_vsi;
9569         }
9570         vsi->base_queue = ret;
9571
9572         /* Update the FW view of the VSI. Force a reset of TC and queue
9573          * layout configurations.
9574          */
9575         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9576         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9577         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9578         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9579
9580         /* assign it some queues */
9581         ret = i40e_alloc_rings(vsi);
9582         if (ret)
9583                 goto err_rings;
9584
9585         /* map all of the rings to the q_vectors */
9586         i40e_vsi_map_rings_to_vectors(vsi);
9587         return vsi;
9588
9589 err_rings:
9590         i40e_vsi_free_q_vectors(vsi);
9591         if (vsi->netdev_registered) {
9592                 vsi->netdev_registered = false;
9593                 unregister_netdev(vsi->netdev);
9594                 free_netdev(vsi->netdev);
9595                 vsi->netdev = NULL;
9596         }
9597         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9598 err_vsi:
9599         i40e_vsi_clear(vsi);
9600         return NULL;
9601 }
9602
9603 /**
9604  * i40e_macaddr_init - explicitly write the mac address filters.
9605  *
9606  * @vsi: pointer to the vsi.
9607  * @macaddr: the MAC address
9608  *
9609  * This is needed when the macaddr has been obtained by other
9610  * means than the default, e.g., from Open Firmware or IDPROM.
9611  * Returns 0 on success, negative on failure
9612  **/
9613 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
9614 {
9615         int ret;
9616         struct i40e_aqc_add_macvlan_element_data element;
9617
9618         ret = i40e_aq_mac_address_write(&vsi->back->hw,
9619                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
9620                                         macaddr, NULL);
9621         if (ret) {
9622                 dev_info(&vsi->back->pdev->dev,
9623                          "Addr change for VSI failed: %d\n", ret);
9624                 return -EADDRNOTAVAIL;
9625         }
9626
9627         memset(&element, 0, sizeof(element));
9628         ether_addr_copy(element.mac_addr, macaddr);
9629         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
9630         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
9631         if (ret) {
9632                 dev_info(&vsi->back->pdev->dev,
9633                          "add filter failed err %s aq_err %s\n",
9634                          i40e_stat_str(&vsi->back->hw, ret),
9635                          i40e_aq_str(&vsi->back->hw,
9636                                      vsi->back->hw.aq.asq_last_status));
9637         }
9638         return ret;
9639 }
9640
9641 /**
9642  * i40e_vsi_setup - Set up a VSI by a given type
9643  * @pf: board private structure
9644  * @type: VSI type
9645  * @uplink_seid: the switch element to link to
9646  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9647  *
9648  * This allocates the sw VSI structure and its queue resources, then add a VSI
9649  * to the identified VEB.
9650  *
9651  * Returns pointer to the successfully allocated and configure VSI sw struct on
9652  * success, otherwise returns NULL on failure.
9653  **/
9654 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9655                                 u16 uplink_seid, u32 param1)
9656 {
9657         struct i40e_vsi *vsi = NULL;
9658         struct i40e_veb *veb = NULL;
9659         int ret, i;
9660         int v_idx;
9661
9662         /* The requested uplink_seid must be either
9663          *     - the PF's port seid
9664          *              no VEB is needed because this is the PF
9665          *              or this is a Flow Director special case VSI
9666          *     - seid of an existing VEB
9667          *     - seid of a VSI that owns an existing VEB
9668          *     - seid of a VSI that doesn't own a VEB
9669          *              a new VEB is created and the VSI becomes the owner
9670          *     - seid of the PF VSI, which is what creates the first VEB
9671          *              this is a special case of the previous
9672          *
9673          * Find which uplink_seid we were given and create a new VEB if needed
9674          */
9675         for (i = 0; i < I40E_MAX_VEB; i++) {
9676                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9677                         veb = pf->veb[i];
9678                         break;
9679                 }
9680         }
9681
9682         if (!veb && uplink_seid != pf->mac_seid) {
9683
9684                 for (i = 0; i < pf->num_alloc_vsi; i++) {
9685                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9686                                 vsi = pf->vsi[i];
9687                                 break;
9688                         }
9689                 }
9690                 if (!vsi) {
9691                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9692                                  uplink_seid);
9693                         return NULL;
9694                 }
9695
9696                 if (vsi->uplink_seid == pf->mac_seid)
9697                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9698                                              vsi->tc_config.enabled_tc);
9699                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9700                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9701                                              vsi->tc_config.enabled_tc);
9702                 if (veb) {
9703                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9704                                 dev_info(&vsi->back->pdev->dev,
9705                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
9706                                 return NULL;
9707                         }
9708                         /* We come up by default in VEPA mode if SRIOV is not
9709                          * already enabled, in which case we can't force VEPA
9710                          * mode.
9711                          */
9712                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9713                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9714                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9715                         }
9716                         i40e_config_bridge_mode(veb);
9717                 }
9718                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9719                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9720                                 veb = pf->veb[i];
9721                 }
9722                 if (!veb) {
9723                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9724                         return NULL;
9725                 }
9726
9727                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9728                 uplink_seid = veb->seid;
9729         }
9730
9731         /* get vsi sw struct */
9732         v_idx = i40e_vsi_mem_alloc(pf, type);
9733         if (v_idx < 0)
9734                 goto err_alloc;
9735         vsi = pf->vsi[v_idx];
9736         if (!vsi)
9737                 goto err_alloc;
9738         vsi->type = type;
9739         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9740
9741         if (type == I40E_VSI_MAIN)
9742                 pf->lan_vsi = v_idx;
9743         else if (type == I40E_VSI_SRIOV)
9744                 vsi->vf_id = param1;
9745         /* assign it some queues */
9746         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9747                                 vsi->idx);
9748         if (ret < 0) {
9749                 dev_info(&pf->pdev->dev,
9750                          "failed to get tracking for %d queues for VSI %d err=%d\n",
9751                          vsi->alloc_queue_pairs, vsi->seid, ret);
9752                 goto err_vsi;
9753         }
9754         vsi->base_queue = ret;
9755
9756         /* get a VSI from the hardware */
9757         vsi->uplink_seid = uplink_seid;
9758         ret = i40e_add_vsi(vsi);
9759         if (ret)
9760                 goto err_vsi;
9761
9762         switch (vsi->type) {
9763         /* setup the netdev if needed */
9764         case I40E_VSI_MAIN:
9765                 /* Apply relevant filters if a platform-specific mac
9766                  * address was selected.
9767                  */
9768                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9769                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9770                         if (ret) {
9771                                 dev_warn(&pf->pdev->dev,
9772                                          "could not set up macaddr; err %d\n",
9773                                          ret);
9774                         }
9775                 }
9776         case I40E_VSI_VMDQ2:
9777         case I40E_VSI_FCOE:
9778                 ret = i40e_config_netdev(vsi);
9779                 if (ret)
9780                         goto err_netdev;
9781                 ret = register_netdev(vsi->netdev);
9782                 if (ret)
9783                         goto err_netdev;
9784                 vsi->netdev_registered = true;
9785                 netif_carrier_off(vsi->netdev);
9786 #ifdef CONFIG_I40E_DCB
9787                 /* Setup DCB netlink interface */
9788                 i40e_dcbnl_setup(vsi);
9789 #endif /* CONFIG_I40E_DCB */
9790                 /* fall through */
9791
9792         case I40E_VSI_FDIR:
9793                 /* set up vectors and rings if needed */
9794                 ret = i40e_vsi_setup_vectors(vsi);
9795                 if (ret)
9796                         goto err_msix;
9797
9798                 ret = i40e_alloc_rings(vsi);
9799                 if (ret)
9800                         goto err_rings;
9801
9802                 /* map all of the rings to the q_vectors */
9803                 i40e_vsi_map_rings_to_vectors(vsi);
9804
9805                 i40e_vsi_reset_stats(vsi);
9806                 break;
9807
9808         default:
9809                 /* no netdev or rings for the other VSI types */
9810                 break;
9811         }
9812
9813         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9814             (vsi->type == I40E_VSI_VMDQ2)) {
9815                 ret = i40e_vsi_config_rss(vsi);
9816         }
9817         return vsi;
9818
9819 err_rings:
9820         i40e_vsi_free_q_vectors(vsi);
9821 err_msix:
9822         if (vsi->netdev_registered) {
9823                 vsi->netdev_registered = false;
9824                 unregister_netdev(vsi->netdev);
9825                 free_netdev(vsi->netdev);
9826                 vsi->netdev = NULL;
9827         }
9828 err_netdev:
9829         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9830 err_vsi:
9831         i40e_vsi_clear(vsi);
9832 err_alloc:
9833         return NULL;
9834 }
9835
9836 /**
9837  * i40e_veb_get_bw_info - Query VEB BW information
9838  * @veb: the veb to query
9839  *
9840  * Query the Tx scheduler BW configuration data for given VEB
9841  **/
9842 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9843 {
9844         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9845         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9846         struct i40e_pf *pf = veb->pf;
9847         struct i40e_hw *hw = &pf->hw;
9848         u32 tc_bw_max;
9849         int ret = 0;
9850         int i;
9851
9852         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9853                                                   &bw_data, NULL);
9854         if (ret) {
9855                 dev_info(&pf->pdev->dev,
9856                          "query veb bw config failed, err %s aq_err %s\n",
9857                          i40e_stat_str(&pf->hw, ret),
9858                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9859                 goto out;
9860         }
9861
9862         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9863                                                    &ets_data, NULL);
9864         if (ret) {
9865                 dev_info(&pf->pdev->dev,
9866                          "query veb bw ets config failed, err %s aq_err %s\n",
9867                          i40e_stat_str(&pf->hw, ret),
9868                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9869                 goto out;
9870         }
9871
9872         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9873         veb->bw_max_quanta = ets_data.tc_bw_max;
9874         veb->is_abs_credits = bw_data.absolute_credits_enable;
9875         veb->enabled_tc = ets_data.tc_valid_bits;
9876         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9877                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9878         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9879                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9880                 veb->bw_tc_limit_credits[i] =
9881                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
9882                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9883         }
9884
9885 out:
9886         return ret;
9887 }
9888
9889 /**
9890  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9891  * @pf: board private structure
9892  *
9893  * On error: returns error code (negative)
9894  * On success: returns vsi index in PF (positive)
9895  **/
9896 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9897 {
9898         int ret = -ENOENT;
9899         struct i40e_veb *veb;
9900         int i;
9901
9902         /* Need to protect the allocation of switch elements at the PF level */
9903         mutex_lock(&pf->switch_mutex);
9904
9905         /* VEB list may be fragmented if VEB creation/destruction has
9906          * been happening.  We can afford to do a quick scan to look
9907          * for any free slots in the list.
9908          *
9909          * find next empty veb slot, looping back around if necessary
9910          */
9911         i = 0;
9912         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9913                 i++;
9914         if (i >= I40E_MAX_VEB) {
9915                 ret = -ENOMEM;
9916                 goto err_alloc_veb;  /* out of VEB slots! */
9917         }
9918
9919         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9920         if (!veb) {
9921                 ret = -ENOMEM;
9922                 goto err_alloc_veb;
9923         }
9924         veb->pf = pf;
9925         veb->idx = i;
9926         veb->enabled_tc = 1;
9927
9928         pf->veb[i] = veb;
9929         ret = i;
9930 err_alloc_veb:
9931         mutex_unlock(&pf->switch_mutex);
9932         return ret;
9933 }
9934
9935 /**
9936  * i40e_switch_branch_release - Delete a branch of the switch tree
9937  * @branch: where to start deleting
9938  *
9939  * This uses recursion to find the tips of the branch to be
9940  * removed, deleting until we get back to and can delete this VEB.
9941  **/
9942 static void i40e_switch_branch_release(struct i40e_veb *branch)
9943 {
9944         struct i40e_pf *pf = branch->pf;
9945         u16 branch_seid = branch->seid;
9946         u16 veb_idx = branch->idx;
9947         int i;
9948
9949         /* release any VEBs on this VEB - RECURSION */
9950         for (i = 0; i < I40E_MAX_VEB; i++) {
9951                 if (!pf->veb[i])
9952                         continue;
9953                 if (pf->veb[i]->uplink_seid == branch->seid)
9954                         i40e_switch_branch_release(pf->veb[i]);
9955         }
9956
9957         /* Release the VSIs on this VEB, but not the owner VSI.
9958          *
9959          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
9960          *       the VEB itself, so don't use (*branch) after this loop.
9961          */
9962         for (i = 0; i < pf->num_alloc_vsi; i++) {
9963                 if (!pf->vsi[i])
9964                         continue;
9965                 if (pf->vsi[i]->uplink_seid == branch_seid &&
9966                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9967                         i40e_vsi_release(pf->vsi[i]);
9968                 }
9969         }
9970
9971         /* There's one corner case where the VEB might not have been
9972          * removed, so double check it here and remove it if needed.
9973          * This case happens if the veb was created from the debugfs
9974          * commands and no VSIs were added to it.
9975          */
9976         if (pf->veb[veb_idx])
9977                 i40e_veb_release(pf->veb[veb_idx]);
9978 }
9979
9980 /**
9981  * i40e_veb_clear - remove veb struct
9982  * @veb: the veb to remove
9983  **/
9984 static void i40e_veb_clear(struct i40e_veb *veb)
9985 {
9986         if (!veb)
9987                 return;
9988
9989         if (veb->pf) {
9990                 struct i40e_pf *pf = veb->pf;
9991
9992                 mutex_lock(&pf->switch_mutex);
9993                 if (pf->veb[veb->idx] == veb)
9994                         pf->veb[veb->idx] = NULL;
9995                 mutex_unlock(&pf->switch_mutex);
9996         }
9997
9998         kfree(veb);
9999 }
10000
10001 /**
10002  * i40e_veb_release - Delete a VEB and free its resources
10003  * @veb: the VEB being removed
10004  **/
10005 void i40e_veb_release(struct i40e_veb *veb)
10006 {
10007         struct i40e_vsi *vsi = NULL;
10008         struct i40e_pf *pf;
10009         int i, n = 0;
10010
10011         pf = veb->pf;
10012
10013         /* find the remaining VSI and check for extras */
10014         for (i = 0; i < pf->num_alloc_vsi; i++) {
10015                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10016                         n++;
10017                         vsi = pf->vsi[i];
10018                 }
10019         }
10020         if (n != 1) {
10021                 dev_info(&pf->pdev->dev,
10022                          "can't remove VEB %d with %d VSIs left\n",
10023                          veb->seid, n);
10024                 return;
10025         }
10026
10027         /* move the remaining VSI to uplink veb */
10028         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10029         if (veb->uplink_seid) {
10030                 vsi->uplink_seid = veb->uplink_seid;
10031                 if (veb->uplink_seid == pf->mac_seid)
10032                         vsi->veb_idx = I40E_NO_VEB;
10033                 else
10034                         vsi->veb_idx = veb->veb_idx;
10035         } else {
10036                 /* floating VEB */
10037                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10038                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10039         }
10040
10041         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10042         i40e_veb_clear(veb);
10043 }
10044
10045 /**
10046  * i40e_add_veb - create the VEB in the switch
10047  * @veb: the VEB to be instantiated
10048  * @vsi: the controlling VSI
10049  **/
10050 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10051 {
10052         struct i40e_pf *pf = veb->pf;
10053         bool is_default = veb->pf->cur_promisc;
10054         bool is_cloud = false;
10055         int ret;
10056
10057         /* get a VEB from the hardware */
10058         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10059                               veb->enabled_tc, is_default,
10060                               is_cloud, &veb->seid, NULL);
10061         if (ret) {
10062                 dev_info(&pf->pdev->dev,
10063                          "couldn't add VEB, err %s aq_err %s\n",
10064                          i40e_stat_str(&pf->hw, ret),
10065                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10066                 return -EPERM;
10067         }
10068
10069         /* get statistics counter */
10070         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10071                                          &veb->stats_idx, NULL, NULL, NULL);
10072         if (ret) {
10073                 dev_info(&pf->pdev->dev,
10074                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10075                          i40e_stat_str(&pf->hw, ret),
10076                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10077                 return -EPERM;
10078         }
10079         ret = i40e_veb_get_bw_info(veb);
10080         if (ret) {
10081                 dev_info(&pf->pdev->dev,
10082                          "couldn't get VEB bw info, err %s aq_err %s\n",
10083                          i40e_stat_str(&pf->hw, ret),
10084                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10085                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10086                 return -ENOENT;
10087         }
10088
10089         vsi->uplink_seid = veb->seid;
10090         vsi->veb_idx = veb->idx;
10091         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10092
10093         return 0;
10094 }
10095
10096 /**
10097  * i40e_veb_setup - Set up a VEB
10098  * @pf: board private structure
10099  * @flags: VEB setup flags
10100  * @uplink_seid: the switch element to link to
10101  * @vsi_seid: the initial VSI seid
10102  * @enabled_tc: Enabled TC bit-map
10103  *
10104  * This allocates the sw VEB structure and links it into the switch
10105  * It is possible and legal for this to be a duplicate of an already
10106  * existing VEB.  It is also possible for both uplink and vsi seids
10107  * to be zero, in order to create a floating VEB.
10108  *
10109  * Returns pointer to the successfully allocated VEB sw struct on
10110  * success, otherwise returns NULL on failure.
10111  **/
10112 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10113                                 u16 uplink_seid, u16 vsi_seid,
10114                                 u8 enabled_tc)
10115 {
10116         struct i40e_veb *veb, *uplink_veb = NULL;
10117         int vsi_idx, veb_idx;
10118         int ret;
10119
10120         /* if one seid is 0, the other must be 0 to create a floating relay */
10121         if ((uplink_seid == 0 || vsi_seid == 0) &&
10122             (uplink_seid + vsi_seid != 0)) {
10123                 dev_info(&pf->pdev->dev,
10124                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10125                          uplink_seid, vsi_seid);
10126                 return NULL;
10127         }
10128
10129         /* make sure there is such a vsi and uplink */
10130         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10131                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10132                         break;
10133         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10134                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10135                          vsi_seid);
10136                 return NULL;
10137         }
10138
10139         if (uplink_seid && uplink_seid != pf->mac_seid) {
10140                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10141                         if (pf->veb[veb_idx] &&
10142                             pf->veb[veb_idx]->seid == uplink_seid) {
10143                                 uplink_veb = pf->veb[veb_idx];
10144                                 break;
10145                         }
10146                 }
10147                 if (!uplink_veb) {
10148                         dev_info(&pf->pdev->dev,
10149                                  "uplink seid %d not found\n", uplink_seid);
10150                         return NULL;
10151                 }
10152         }
10153
10154         /* get veb sw struct */
10155         veb_idx = i40e_veb_mem_alloc(pf);
10156         if (veb_idx < 0)
10157                 goto err_alloc;
10158         veb = pf->veb[veb_idx];
10159         veb->flags = flags;
10160         veb->uplink_seid = uplink_seid;
10161         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10162         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10163
10164         /* create the VEB in the switch */
10165         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10166         if (ret)
10167                 goto err_veb;
10168         if (vsi_idx == pf->lan_vsi)
10169                 pf->lan_veb = veb->idx;
10170
10171         return veb;
10172
10173 err_veb:
10174         i40e_veb_clear(veb);
10175 err_alloc:
10176         return NULL;
10177 }
10178
10179 /**
10180  * i40e_setup_pf_switch_element - set PF vars based on switch type
10181  * @pf: board private structure
10182  * @ele: element we are building info from
10183  * @num_reported: total number of elements
10184  * @printconfig: should we print the contents
10185  *
10186  * helper function to assist in extracting a few useful SEID values.
10187  **/
10188 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10189                                 struct i40e_aqc_switch_config_element_resp *ele,
10190                                 u16 num_reported, bool printconfig)
10191 {
10192         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10193         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10194         u8 element_type = ele->element_type;
10195         u16 seid = le16_to_cpu(ele->seid);
10196
10197         if (printconfig)
10198                 dev_info(&pf->pdev->dev,
10199                          "type=%d seid=%d uplink=%d downlink=%d\n",
10200                          element_type, seid, uplink_seid, downlink_seid);
10201
10202         switch (element_type) {
10203         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10204                 pf->mac_seid = seid;
10205                 break;
10206         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10207                 /* Main VEB? */
10208                 if (uplink_seid != pf->mac_seid)
10209                         break;
10210                 if (pf->lan_veb == I40E_NO_VEB) {
10211                         int v;
10212
10213                         /* find existing or else empty VEB */
10214                         for (v = 0; v < I40E_MAX_VEB; v++) {
10215                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10216                                         pf->lan_veb = v;
10217                                         break;
10218                                 }
10219                         }
10220                         if (pf->lan_veb == I40E_NO_VEB) {
10221                                 v = i40e_veb_mem_alloc(pf);
10222                                 if (v < 0)
10223                                         break;
10224                                 pf->lan_veb = v;
10225                         }
10226                 }
10227
10228                 pf->veb[pf->lan_veb]->seid = seid;
10229                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10230                 pf->veb[pf->lan_veb]->pf = pf;
10231                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10232                 break;
10233         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10234                 if (num_reported != 1)
10235                         break;
10236                 /* This is immediately after a reset so we can assume this is
10237                  * the PF's VSI
10238                  */
10239                 pf->mac_seid = uplink_seid;
10240                 pf->pf_seid = downlink_seid;
10241                 pf->main_vsi_seid = seid;
10242                 if (printconfig)
10243                         dev_info(&pf->pdev->dev,
10244                                  "pf_seid=%d main_vsi_seid=%d\n",
10245                                  pf->pf_seid, pf->main_vsi_seid);
10246                 break;
10247         case I40E_SWITCH_ELEMENT_TYPE_PF:
10248         case I40E_SWITCH_ELEMENT_TYPE_VF:
10249         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10250         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10251         case I40E_SWITCH_ELEMENT_TYPE_PE:
10252         case I40E_SWITCH_ELEMENT_TYPE_PA:
10253                 /* ignore these for now */
10254                 break;
10255         default:
10256                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10257                          element_type, seid);
10258                 break;
10259         }
10260 }
10261
10262 /**
10263  * i40e_fetch_switch_configuration - Get switch config from firmware
10264  * @pf: board private structure
10265  * @printconfig: should we print the contents
10266  *
10267  * Get the current switch configuration from the device and
10268  * extract a few useful SEID values.
10269  **/
10270 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10271 {
10272         struct i40e_aqc_get_switch_config_resp *sw_config;
10273         u16 next_seid = 0;
10274         int ret = 0;
10275         u8 *aq_buf;
10276         int i;
10277
10278         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10279         if (!aq_buf)
10280                 return -ENOMEM;
10281
10282         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10283         do {
10284                 u16 num_reported, num_total;
10285
10286                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10287                                                 I40E_AQ_LARGE_BUF,
10288                                                 &next_seid, NULL);
10289                 if (ret) {
10290                         dev_info(&pf->pdev->dev,
10291                                  "get switch config failed err %s aq_err %s\n",
10292                                  i40e_stat_str(&pf->hw, ret),
10293                                  i40e_aq_str(&pf->hw,
10294                                              pf->hw.aq.asq_last_status));
10295                         kfree(aq_buf);
10296                         return -ENOENT;
10297                 }
10298
10299                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10300                 num_total = le16_to_cpu(sw_config->header.num_total);
10301
10302                 if (printconfig)
10303                         dev_info(&pf->pdev->dev,
10304                                  "header: %d reported %d total\n",
10305                                  num_reported, num_total);
10306
10307                 for (i = 0; i < num_reported; i++) {
10308                         struct i40e_aqc_switch_config_element_resp *ele =
10309                                 &sw_config->element[i];
10310
10311                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10312                                                      printconfig);
10313                 }
10314         } while (next_seid != 0);
10315
10316         kfree(aq_buf);
10317         return ret;
10318 }
10319
10320 /**
10321  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10322  * @pf: board private structure
10323  * @reinit: if the Main VSI needs to re-initialized.
10324  *
10325  * Returns 0 on success, negative value on failure
10326  **/
10327 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10328 {
10329         int ret;
10330
10331         /* find out what's out there already */
10332         ret = i40e_fetch_switch_configuration(pf, false);
10333         if (ret) {
10334                 dev_info(&pf->pdev->dev,
10335                          "couldn't fetch switch config, err %s aq_err %s\n",
10336                          i40e_stat_str(&pf->hw, ret),
10337                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10338                 return ret;
10339         }
10340         i40e_pf_reset_stats(pf);
10341
10342         /* first time setup */
10343         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10344                 struct i40e_vsi *vsi = NULL;
10345                 u16 uplink_seid;
10346
10347                 /* Set up the PF VSI associated with the PF's main VSI
10348                  * that is already in the HW switch
10349                  */
10350                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10351                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10352                 else
10353                         uplink_seid = pf->mac_seid;
10354                 if (pf->lan_vsi == I40E_NO_VSI)
10355                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10356                 else if (reinit)
10357                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10358                 if (!vsi) {
10359                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10360                         i40e_fdir_teardown(pf);
10361                         return -EAGAIN;
10362                 }
10363         } else {
10364                 /* force a reset of TC and queue layout configurations */
10365                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10366
10367                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10368                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10369                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10370         }
10371         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10372
10373         i40e_fdir_sb_setup(pf);
10374
10375         /* Setup static PF queue filter control settings */
10376         ret = i40e_setup_pf_filter_control(pf);
10377         if (ret) {
10378                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10379                          ret);
10380                 /* Failure here should not stop continuing other steps */
10381         }
10382
10383         /* enable RSS in the HW, even for only one queue, as the stack can use
10384          * the hash
10385          */
10386         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10387                 i40e_pf_config_rss(pf);
10388
10389         /* fill in link information and enable LSE reporting */
10390         i40e_update_link_info(&pf->hw);
10391         i40e_link_event(pf);
10392
10393         /* Initialize user-specific link properties */
10394         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10395                                   I40E_AQ_AN_COMPLETED) ? true : false);
10396
10397         i40e_ptp_init(pf);
10398
10399         return ret;
10400 }
10401
10402 /**
10403  * i40e_determine_queue_usage - Work out queue distribution
10404  * @pf: board private structure
10405  **/
10406 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10407 {
10408         int queues_left;
10409
10410         pf->num_lan_qps = 0;
10411 #ifdef I40E_FCOE
10412         pf->num_fcoe_qps = 0;
10413 #endif
10414
10415         /* Find the max queues to be put into basic use.  We'll always be
10416          * using TC0, whether or not DCB is running, and TC0 will get the
10417          * big RSS set.
10418          */
10419         queues_left = pf->hw.func_caps.num_tx_qp;
10420
10421         if ((queues_left == 1) ||
10422             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10423                 /* one qp for PF, no queues for anything else */
10424                 queues_left = 0;
10425                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10426
10427                 /* make sure all the fancies are disabled */
10428                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10429                                I40E_FLAG_IWARP_ENABLED  |
10430 #ifdef I40E_FCOE
10431                                I40E_FLAG_FCOE_ENABLED   |
10432 #endif
10433                                I40E_FLAG_FD_SB_ENABLED  |
10434                                I40E_FLAG_FD_ATR_ENABLED |
10435                                I40E_FLAG_DCB_CAPABLE    |
10436                                I40E_FLAG_SRIOV_ENABLED  |
10437                                I40E_FLAG_VMDQ_ENABLED);
10438         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10439                                   I40E_FLAG_FD_SB_ENABLED |
10440                                   I40E_FLAG_FD_ATR_ENABLED |
10441                                   I40E_FLAG_DCB_CAPABLE))) {
10442                 /* one qp for PF */
10443                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10444                 queues_left -= pf->num_lan_qps;
10445
10446                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10447                                I40E_FLAG_IWARP_ENABLED  |
10448 #ifdef I40E_FCOE
10449                                I40E_FLAG_FCOE_ENABLED   |
10450 #endif
10451                                I40E_FLAG_FD_SB_ENABLED  |
10452                                I40E_FLAG_FD_ATR_ENABLED |
10453                                I40E_FLAG_DCB_ENABLED    |
10454                                I40E_FLAG_VMDQ_ENABLED);
10455         } else {
10456                 /* Not enough queues for all TCs */
10457                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10458                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10459                         pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10460                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10461                 }
10462                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10463                                         num_online_cpus());
10464                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10465                                         pf->hw.func_caps.num_tx_qp);
10466
10467                 queues_left -= pf->num_lan_qps;
10468         }
10469
10470 #ifdef I40E_FCOE
10471         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10472                 if (I40E_DEFAULT_FCOE <= queues_left) {
10473                         pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10474                 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10475                         pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10476                 } else {
10477                         pf->num_fcoe_qps = 0;
10478                         pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10479                         dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10480                 }
10481
10482                 queues_left -= pf->num_fcoe_qps;
10483         }
10484
10485 #endif
10486         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10487                 if (queues_left > 1) {
10488                         queues_left -= 1; /* save 1 queue for FD */
10489                 } else {
10490                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10491                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10492                 }
10493         }
10494
10495         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10496             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10497                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10498                                         (queues_left / pf->num_vf_qps));
10499                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10500         }
10501
10502         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10503             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10504                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10505                                           (queues_left / pf->num_vmdq_qps));
10506                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10507         }
10508
10509         pf->queues_left = queues_left;
10510         dev_dbg(&pf->pdev->dev,
10511                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10512                 pf->hw.func_caps.num_tx_qp,
10513                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10514                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10515                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10516                 queues_left);
10517 #ifdef I40E_FCOE
10518         dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10519 #endif
10520 }
10521
10522 /**
10523  * i40e_setup_pf_filter_control - Setup PF static filter control
10524  * @pf: PF to be setup
10525  *
10526  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10527  * settings. If PE/FCoE are enabled then it will also set the per PF
10528  * based filter sizes required for them. It also enables Flow director,
10529  * ethertype and macvlan type filter settings for the pf.
10530  *
10531  * Returns 0 on success, negative on failure
10532  **/
10533 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10534 {
10535         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10536
10537         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10538
10539         /* Flow Director is enabled */
10540         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10541                 settings->enable_fdir = true;
10542
10543         /* Ethtype and MACVLAN filters enabled for PF */
10544         settings->enable_ethtype = true;
10545         settings->enable_macvlan = true;
10546
10547         if (i40e_set_filter_control(&pf->hw, settings))
10548                 return -ENOENT;
10549
10550         return 0;
10551 }
10552
10553 #define INFO_STRING_LEN 255
10554 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10555 static void i40e_print_features(struct i40e_pf *pf)
10556 {
10557         struct i40e_hw *hw = &pf->hw;
10558         char *buf;
10559         int i;
10560
10561         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10562         if (!buf)
10563                 return;
10564
10565         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10566 #ifdef CONFIG_PCI_IOV
10567         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10568 #endif
10569         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d RX: %s",
10570                       pf->hw.func_caps.num_vsis,
10571                       pf->vsi[pf->lan_vsi]->num_queue_pairs,
10572                       pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
10573
10574         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10575                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10576         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10577                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10578         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10579                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10580                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10581         }
10582         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10583                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10584 #if IS_ENABLED(CONFIG_VXLAN)
10585         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10586 #endif
10587 #if IS_ENABLED(CONFIG_GENEVE)
10588         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10589 #endif
10590         if (pf->flags & I40E_FLAG_PTP)
10591                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10592 #ifdef I40E_FCOE
10593         if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10594                 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10595 #endif
10596         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10597                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10598         else
10599                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10600
10601         dev_info(&pf->pdev->dev, "%s\n", buf);
10602         kfree(buf);
10603         WARN_ON(i > INFO_STRING_LEN);
10604 }
10605
10606 /**
10607  * i40e_get_platform_mac_addr - get platform-specific MAC address
10608  *
10609  * @pdev: PCI device information struct
10610  * @pf: board private structure
10611  *
10612  * Look up the MAC address in Open Firmware  on systems that support it,
10613  * and use IDPROM on SPARC if no OF address is found. On return, the
10614  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10615  * has been selected.
10616  **/
10617 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10618 {
10619         struct device_node *dp = pci_device_to_OF_node(pdev);
10620         const unsigned char *addr;
10621         u8 *mac_addr = pf->hw.mac.addr;
10622
10623         pf->flags &= ~I40E_FLAG_PF_MAC;
10624         addr = of_get_mac_address(dp);
10625         if (addr) {
10626                 ether_addr_copy(mac_addr, addr);
10627                 pf->flags |= I40E_FLAG_PF_MAC;
10628 #ifdef CONFIG_SPARC
10629         } else {
10630                 ether_addr_copy(mac_addr, idprom->id_ethaddr);
10631                 pf->flags |= I40E_FLAG_PF_MAC;
10632 #endif /* CONFIG_SPARC */
10633         }
10634 }
10635
10636 /**
10637  * i40e_probe - Device initialization routine
10638  * @pdev: PCI device information struct
10639  * @ent: entry in i40e_pci_tbl
10640  *
10641  * i40e_probe initializes a PF identified by a pci_dev structure.
10642  * The OS initialization, configuring of the PF private structure,
10643  * and a hardware reset occur.
10644  *
10645  * Returns 0 on success, negative on failure
10646  **/
10647 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10648 {
10649         struct i40e_aq_get_phy_abilities_resp abilities;
10650         struct i40e_pf *pf;
10651         struct i40e_hw *hw;
10652         static u16 pfs_found;
10653         u16 wol_nvm_bits;
10654         u16 link_status;
10655         int err;
10656         u32 len;
10657         u32 val;
10658         u32 i;
10659         u8 set_fc_aq_fail;
10660
10661         err = pci_enable_device_mem(pdev);
10662         if (err)
10663                 return err;
10664
10665         /* set up for high or low dma */
10666         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10667         if (err) {
10668                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10669                 if (err) {
10670                         dev_err(&pdev->dev,
10671                                 "DMA configuration failed: 0x%x\n", err);
10672                         goto err_dma;
10673                 }
10674         }
10675
10676         /* set up pci connections */
10677         err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10678                                            IORESOURCE_MEM), i40e_driver_name);
10679         if (err) {
10680                 dev_info(&pdev->dev,
10681                          "pci_request_selected_regions failed %d\n", err);
10682                 goto err_pci_reg;
10683         }
10684
10685         pci_enable_pcie_error_reporting(pdev);
10686         pci_set_master(pdev);
10687
10688         /* Now that we have a PCI connection, we need to do the
10689          * low level device setup.  This is primarily setting up
10690          * the Admin Queue structures and then querying for the
10691          * device's current profile information.
10692          */
10693         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10694         if (!pf) {
10695                 err = -ENOMEM;
10696                 goto err_pf_alloc;
10697         }
10698         pf->next_vsi = 0;
10699         pf->pdev = pdev;
10700         set_bit(__I40E_DOWN, &pf->state);
10701
10702         hw = &pf->hw;
10703         hw->back = pf;
10704
10705         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10706                                 I40E_MAX_CSR_SPACE);
10707
10708         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10709         if (!hw->hw_addr) {
10710                 err = -EIO;
10711                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10712                          (unsigned int)pci_resource_start(pdev, 0),
10713                          pf->ioremap_len, err);
10714                 goto err_ioremap;
10715         }
10716         hw->vendor_id = pdev->vendor;
10717         hw->device_id = pdev->device;
10718         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10719         hw->subsystem_vendor_id = pdev->subsystem_vendor;
10720         hw->subsystem_device_id = pdev->subsystem_device;
10721         hw->bus.device = PCI_SLOT(pdev->devfn);
10722         hw->bus.func = PCI_FUNC(pdev->devfn);
10723         pf->instance = pfs_found;
10724
10725         if (debug != -1) {
10726                 pf->msg_enable = pf->hw.debug_mask;
10727                 pf->msg_enable = debug;
10728         }
10729
10730         /* do a special CORER for clearing PXE mode once at init */
10731         if (hw->revision_id == 0 &&
10732             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10733                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10734                 i40e_flush(hw);
10735                 msleep(200);
10736                 pf->corer_count++;
10737
10738                 i40e_clear_pxe_mode(hw);
10739         }
10740
10741         /* Reset here to make sure all is clean and to define PF 'n' */
10742         i40e_clear_hw(hw);
10743         err = i40e_pf_reset(hw);
10744         if (err) {
10745                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10746                 goto err_pf_reset;
10747         }
10748         pf->pfr_count++;
10749
10750         hw->aq.num_arq_entries = I40E_AQ_LEN;
10751         hw->aq.num_asq_entries = I40E_AQ_LEN;
10752         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10753         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10754         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10755
10756         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10757                  "%s-%s:misc",
10758                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10759
10760         err = i40e_init_shared_code(hw);
10761         if (err) {
10762                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10763                          err);
10764                 goto err_pf_reset;
10765         }
10766
10767         /* set up a default setting for link flow control */
10768         pf->hw.fc.requested_mode = I40E_FC_NONE;
10769
10770         /* set up the locks for the AQ, do this only once in probe
10771          * and destroy them only once in remove
10772          */
10773         mutex_init(&hw->aq.asq_mutex);
10774         mutex_init(&hw->aq.arq_mutex);
10775
10776         err = i40e_init_adminq(hw);
10777         if (err) {
10778                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10779                         dev_info(&pdev->dev,
10780                                  "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");
10781                 else
10782                         dev_info(&pdev->dev,
10783                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10784
10785                 goto err_pf_reset;
10786         }
10787
10788         /* provide nvm, fw, api versions */
10789         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10790                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10791                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
10792                  i40e_nvm_version_str(hw));
10793
10794         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10795             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10796                 dev_info(&pdev->dev,
10797                          "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");
10798         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10799                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10800                 dev_info(&pdev->dev,
10801                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10802
10803         i40e_verify_eeprom(pf);
10804
10805         /* Rev 0 hardware was never productized */
10806         if (hw->revision_id < 1)
10807                 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");
10808
10809         i40e_clear_pxe_mode(hw);
10810         err = i40e_get_capabilities(pf);
10811         if (err)
10812                 goto err_adminq_setup;
10813
10814         err = i40e_sw_init(pf);
10815         if (err) {
10816                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10817                 goto err_sw_init;
10818         }
10819
10820         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10821                                 hw->func_caps.num_rx_qp,
10822                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10823         if (err) {
10824                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10825                 goto err_init_lan_hmc;
10826         }
10827
10828         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10829         if (err) {
10830                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10831                 err = -ENOENT;
10832                 goto err_configure_lan_hmc;
10833         }
10834
10835         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10836          * Ignore error return codes because if it was already disabled via
10837          * hardware settings this will fail
10838          */
10839         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
10840             (pf->hw.aq.fw_maj_ver < 4)) {
10841                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10842                 i40e_aq_stop_lldp(hw, true, NULL);
10843         }
10844
10845         i40e_get_mac_addr(hw, hw->mac.addr);
10846         /* allow a platform config to override the HW addr */
10847         i40e_get_platform_mac_addr(pdev, pf);
10848         if (!is_valid_ether_addr(hw->mac.addr)) {
10849                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10850                 err = -EIO;
10851                 goto err_mac_addr;
10852         }
10853         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10854         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10855         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10856         if (is_valid_ether_addr(hw->mac.port_addr))
10857                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10858 #ifdef I40E_FCOE
10859         err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10860         if (err)
10861                 dev_info(&pdev->dev,
10862                          "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10863         if (!is_valid_ether_addr(hw->mac.san_addr)) {
10864                 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10865                          hw->mac.san_addr);
10866                 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10867         }
10868         dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10869 #endif /* I40E_FCOE */
10870
10871         pci_set_drvdata(pdev, pf);
10872         pci_save_state(pdev);
10873 #ifdef CONFIG_I40E_DCB
10874         err = i40e_init_pf_dcb(pf);
10875         if (err) {
10876                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10877                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10878                 /* Continue without DCB enabled */
10879         }
10880 #endif /* CONFIG_I40E_DCB */
10881
10882         /* set up periodic task facility */
10883         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10884         pf->service_timer_period = HZ;
10885
10886         INIT_WORK(&pf->service_task, i40e_service_task);
10887         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10888         pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10889
10890         /* NVM bit on means WoL disabled for the port */
10891         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10892         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10893                 pf->wol_en = false;
10894         else
10895                 pf->wol_en = true;
10896         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10897
10898         /* set up the main switch operations */
10899         i40e_determine_queue_usage(pf);
10900         err = i40e_init_interrupt_scheme(pf);
10901         if (err)
10902                 goto err_switch_setup;
10903
10904         /* The number of VSIs reported by the FW is the minimum guaranteed
10905          * to us; HW supports far more and we share the remaining pool with
10906          * the other PFs. We allocate space for more than the guarantee with
10907          * the understanding that we might not get them all later.
10908          */
10909         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10910                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10911         else
10912                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10913
10914         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10915         len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
10916         pf->vsi = kzalloc(len, GFP_KERNEL);
10917         if (!pf->vsi) {
10918                 err = -ENOMEM;
10919                 goto err_switch_setup;
10920         }
10921
10922 #ifdef CONFIG_PCI_IOV
10923         /* prep for VF support */
10924         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10925             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10926             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10927                 if (pci_num_vf(pdev))
10928                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10929         }
10930 #endif
10931         err = i40e_setup_pf_switch(pf, false);
10932         if (err) {
10933                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10934                 goto err_vsis;
10935         }
10936
10937         /* Make sure flow control is set according to current settings */
10938         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
10939         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
10940                 dev_dbg(&pf->pdev->dev,
10941                         "Set fc with err %s aq_err %s on get_phy_cap\n",
10942                         i40e_stat_str(hw, err),
10943                         i40e_aq_str(hw, hw->aq.asq_last_status));
10944         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
10945                 dev_dbg(&pf->pdev->dev,
10946                         "Set fc with err %s aq_err %s on set_phy_config\n",
10947                         i40e_stat_str(hw, err),
10948                         i40e_aq_str(hw, hw->aq.asq_last_status));
10949         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
10950                 dev_dbg(&pf->pdev->dev,
10951                         "Set fc with err %s aq_err %s on get_link_info\n",
10952                         i40e_stat_str(hw, err),
10953                         i40e_aq_str(hw, hw->aq.asq_last_status));
10954
10955         /* if FDIR VSI was set up, start it now */
10956         for (i = 0; i < pf->num_alloc_vsi; i++) {
10957                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
10958                         i40e_vsi_open(pf->vsi[i]);
10959                         break;
10960                 }
10961         }
10962
10963         /* driver is only interested in link up/down and module qualification
10964          * reports from firmware
10965          */
10966         err = i40e_aq_set_phy_int_mask(&pf->hw,
10967                                        I40E_AQ_EVENT_LINK_UPDOWN |
10968                                        I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
10969         if (err)
10970                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10971                          i40e_stat_str(&pf->hw, err),
10972                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10973
10974         /* Reconfigure hardware for allowing smaller MSS in the case
10975          * of TSO, so that we avoid the MDD being fired and causing
10976          * a reset in the case of small MSS+TSO.
10977          */
10978         val = rd32(hw, I40E_REG_MSS);
10979         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
10980                 val &= ~I40E_REG_MSS_MIN_MASK;
10981                 val |= I40E_64BYTE_MSS;
10982                 wr32(hw, I40E_REG_MSS, val);
10983         }
10984
10985         if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
10986             (pf->hw.aq.fw_maj_ver < 4)) {
10987                 msleep(75);
10988                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10989                 if (err)
10990                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10991                                  i40e_stat_str(&pf->hw, err),
10992                                  i40e_aq_str(&pf->hw,
10993                                              pf->hw.aq.asq_last_status));
10994         }
10995         /* The main driver is (mostly) up and happy. We need to set this state
10996          * before setting up the misc vector or we get a race and the vector
10997          * ends up disabled forever.
10998          */
10999         clear_bit(__I40E_DOWN, &pf->state);
11000
11001         /* In case of MSIX we are going to setup the misc vector right here
11002          * to handle admin queue events etc. In case of legacy and MSI
11003          * the misc functionality and queue processing is combined in
11004          * the same vector and that gets setup at open.
11005          */
11006         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11007                 err = i40e_setup_misc_vector(pf);
11008                 if (err) {
11009                         dev_info(&pdev->dev,
11010                                  "setup of misc vector failed: %d\n", err);
11011                         goto err_vsis;
11012                 }
11013         }
11014
11015 #ifdef CONFIG_PCI_IOV
11016         /* prep for VF support */
11017         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11018             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11019             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11020                 u32 val;
11021
11022                 /* disable link interrupts for VFs */
11023                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11024                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11025                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11026                 i40e_flush(hw);
11027
11028                 if (pci_num_vf(pdev)) {
11029                         dev_info(&pdev->dev,
11030                                  "Active VFs found, allocating resources.\n");
11031                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11032                         if (err)
11033                                 dev_info(&pdev->dev,
11034                                          "Error %d allocating resources for existing VFs\n",
11035                                          err);
11036                 }
11037         }
11038 #endif /* CONFIG_PCI_IOV */
11039
11040         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11041                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11042                                                       pf->num_iwarp_msix,
11043                                                       I40E_IWARP_IRQ_PILE_ID);
11044                 if (pf->iwarp_base_vector < 0) {
11045                         dev_info(&pdev->dev,
11046                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11047                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11048                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11049                 }
11050         }
11051
11052         i40e_dbg_pf_init(pf);
11053
11054         /* tell the firmware that we're starting */
11055         i40e_send_version(pf);
11056
11057         /* since everything's happy, start the service_task timer */
11058         mod_timer(&pf->service_timer,
11059                   round_jiffies(jiffies + pf->service_timer_period));
11060
11061         /* add this PF to client device list and launch a client service task */
11062         err = i40e_lan_add_device(pf);
11063         if (err)
11064                 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11065                          err);
11066
11067 #ifdef I40E_FCOE
11068         /* create FCoE interface */
11069         i40e_fcoe_vsi_setup(pf);
11070
11071 #endif
11072 #define PCI_SPEED_SIZE 8
11073 #define PCI_WIDTH_SIZE 8
11074         /* Devices on the IOSF bus do not have this information
11075          * and will report PCI Gen 1 x 1 by default so don't bother
11076          * checking them.
11077          */
11078         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11079                 char speed[PCI_SPEED_SIZE] = "Unknown";
11080                 char width[PCI_WIDTH_SIZE] = "Unknown";
11081
11082                 /* Get the negotiated link width and speed from PCI config
11083                  * space
11084                  */
11085                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11086                                           &link_status);
11087
11088                 i40e_set_pci_config_data(hw, link_status);
11089
11090                 switch (hw->bus.speed) {
11091                 case i40e_bus_speed_8000:
11092                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11093                 case i40e_bus_speed_5000:
11094                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11095                 case i40e_bus_speed_2500:
11096                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11097                 default:
11098                         break;
11099                 }
11100                 switch (hw->bus.width) {
11101                 case i40e_bus_width_pcie_x8:
11102                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11103                 case i40e_bus_width_pcie_x4:
11104                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11105                 case i40e_bus_width_pcie_x2:
11106                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11107                 case i40e_bus_width_pcie_x1:
11108                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11109                 default:
11110                         break;
11111                 }
11112
11113                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11114                          speed, width);
11115
11116                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11117                     hw->bus.speed < i40e_bus_speed_8000) {
11118                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11119                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11120                 }
11121         }
11122
11123         /* get the requested speeds from the fw */
11124         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11125         if (err)
11126                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11127                         i40e_stat_str(&pf->hw, err),
11128                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11129         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11130
11131         /* get the supported phy types from the fw */
11132         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11133         if (err)
11134                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11135                         i40e_stat_str(&pf->hw, err),
11136                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11137         pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11138
11139         /* Add a filter to drop all Flow control frames from any VSI from being
11140          * transmitted. By doing so we stop a malicious VF from sending out
11141          * PAUSE or PFC frames and potentially controlling traffic for other
11142          * PF/VF VSIs.
11143          * The FW can still send Flow control frames if enabled.
11144          */
11145         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11146                                                        pf->main_vsi_seid);
11147
11148         /* print a string summarizing features */
11149         i40e_print_features(pf);
11150
11151         return 0;
11152
11153         /* Unwind what we've done if something failed in the setup */
11154 err_vsis:
11155         set_bit(__I40E_DOWN, &pf->state);
11156         i40e_clear_interrupt_scheme(pf);
11157         kfree(pf->vsi);
11158 err_switch_setup:
11159         i40e_reset_interrupt_capability(pf);
11160         del_timer_sync(&pf->service_timer);
11161 err_mac_addr:
11162 err_configure_lan_hmc:
11163         (void)i40e_shutdown_lan_hmc(hw);
11164 err_init_lan_hmc:
11165         kfree(pf->qp_pile);
11166 err_sw_init:
11167 err_adminq_setup:
11168         (void)i40e_shutdown_adminq(hw);
11169 err_pf_reset:
11170         iounmap(hw->hw_addr);
11171 err_ioremap:
11172         kfree(pf);
11173 err_pf_alloc:
11174         pci_disable_pcie_error_reporting(pdev);
11175         pci_release_selected_regions(pdev,
11176                                      pci_select_bars(pdev, IORESOURCE_MEM));
11177 err_pci_reg:
11178 err_dma:
11179         pci_disable_device(pdev);
11180         return err;
11181 }
11182
11183 /**
11184  * i40e_remove - Device removal routine
11185  * @pdev: PCI device information struct
11186  *
11187  * i40e_remove is called by the PCI subsystem to alert the driver
11188  * that is should release a PCI device.  This could be caused by a
11189  * Hot-Plug event, or because the driver is going to be removed from
11190  * memory.
11191  **/
11192 static void i40e_remove(struct pci_dev *pdev)
11193 {
11194         struct i40e_pf *pf = pci_get_drvdata(pdev);
11195         struct i40e_hw *hw = &pf->hw;
11196         i40e_status ret_code;
11197         int i;
11198
11199         i40e_dbg_pf_exit(pf);
11200
11201         i40e_ptp_stop(pf);
11202
11203         /* Disable RSS in hw */
11204         wr32(hw, I40E_PFQF_HENA(0), 0);
11205         wr32(hw, I40E_PFQF_HENA(1), 0);
11206
11207         /* no more scheduling of any task */
11208         set_bit(__I40E_DOWN, &pf->state);
11209         del_timer_sync(&pf->service_timer);
11210         cancel_work_sync(&pf->service_task);
11211
11212         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11213                 i40e_free_vfs(pf);
11214                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11215         }
11216
11217         i40e_fdir_teardown(pf);
11218
11219         /* If there is a switch structure or any orphans, remove them.
11220          * This will leave only the PF's VSI remaining.
11221          */
11222         for (i = 0; i < I40E_MAX_VEB; i++) {
11223                 if (!pf->veb[i])
11224                         continue;
11225
11226                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11227                     pf->veb[i]->uplink_seid == 0)
11228                         i40e_switch_branch_release(pf->veb[i]);
11229         }
11230
11231         /* Now we can shutdown the PF's VSI, just before we kill
11232          * adminq and hmc.
11233          */
11234         if (pf->vsi[pf->lan_vsi])
11235                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11236
11237         /* remove attached clients */
11238         ret_code = i40e_lan_del_device(pf);
11239         if (ret_code) {
11240                 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11241                          ret_code);
11242         }
11243
11244         /* shutdown and destroy the HMC */
11245         if (pf->hw.hmc.hmc_obj) {
11246                 ret_code = i40e_shutdown_lan_hmc(&pf->hw);
11247                 if (ret_code)
11248                         dev_warn(&pdev->dev,
11249                                  "Failed to destroy the HMC resources: %d\n",
11250                                  ret_code);
11251         }
11252
11253         /* shutdown the adminq */
11254         ret_code = i40e_shutdown_adminq(&pf->hw);
11255         if (ret_code)
11256                 dev_warn(&pdev->dev,
11257                          "Failed to destroy the Admin Queue resources: %d\n",
11258                          ret_code);
11259
11260         /* destroy the locks only once, here */
11261         mutex_destroy(&hw->aq.arq_mutex);
11262         mutex_destroy(&hw->aq.asq_mutex);
11263
11264         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11265         i40e_clear_interrupt_scheme(pf);
11266         for (i = 0; i < pf->num_alloc_vsi; i++) {
11267                 if (pf->vsi[i]) {
11268                         i40e_vsi_clear_rings(pf->vsi[i]);
11269                         i40e_vsi_clear(pf->vsi[i]);
11270                         pf->vsi[i] = NULL;
11271                 }
11272         }
11273
11274         for (i = 0; i < I40E_MAX_VEB; i++) {
11275                 kfree(pf->veb[i]);
11276                 pf->veb[i] = NULL;
11277         }
11278
11279         kfree(pf->qp_pile);
11280         kfree(pf->vsi);
11281
11282         iounmap(pf->hw.hw_addr);
11283         kfree(pf);
11284         pci_release_selected_regions(pdev,
11285                                      pci_select_bars(pdev, IORESOURCE_MEM));
11286
11287         pci_disable_pcie_error_reporting(pdev);
11288         pci_disable_device(pdev);
11289 }
11290
11291 /**
11292  * i40e_pci_error_detected - warning that something funky happened in PCI land
11293  * @pdev: PCI device information struct
11294  *
11295  * Called to warn that something happened and the error handling steps
11296  * are in progress.  Allows the driver to quiesce things, be ready for
11297  * remediation.
11298  **/
11299 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11300                                                 enum pci_channel_state error)
11301 {
11302         struct i40e_pf *pf = pci_get_drvdata(pdev);
11303
11304         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11305
11306         /* shutdown all operations */
11307         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11308                 rtnl_lock();
11309                 i40e_prep_for_reset(pf);
11310                 rtnl_unlock();
11311         }
11312
11313         /* Request a slot reset */
11314         return PCI_ERS_RESULT_NEED_RESET;
11315 }
11316
11317 /**
11318  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11319  * @pdev: PCI device information struct
11320  *
11321  * Called to find if the driver can work with the device now that
11322  * the pci slot has been reset.  If a basic connection seems good
11323  * (registers are readable and have sane content) then return a
11324  * happy little PCI_ERS_RESULT_xxx.
11325  **/
11326 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11327 {
11328         struct i40e_pf *pf = pci_get_drvdata(pdev);
11329         pci_ers_result_t result;
11330         int err;
11331         u32 reg;
11332
11333         dev_dbg(&pdev->dev, "%s\n", __func__);
11334         if (pci_enable_device_mem(pdev)) {
11335                 dev_info(&pdev->dev,
11336                          "Cannot re-enable PCI device after reset.\n");
11337                 result = PCI_ERS_RESULT_DISCONNECT;
11338         } else {
11339                 pci_set_master(pdev);
11340                 pci_restore_state(pdev);
11341                 pci_save_state(pdev);
11342                 pci_wake_from_d3(pdev, false);
11343
11344                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11345                 if (reg == 0)
11346                         result = PCI_ERS_RESULT_RECOVERED;
11347                 else
11348                         result = PCI_ERS_RESULT_DISCONNECT;
11349         }
11350
11351         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11352         if (err) {
11353                 dev_info(&pdev->dev,
11354                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11355                          err);
11356                 /* non-fatal, continue */
11357         }
11358
11359         return result;
11360 }
11361
11362 /**
11363  * i40e_pci_error_resume - restart operations after PCI error recovery
11364  * @pdev: PCI device information struct
11365  *
11366  * Called to allow the driver to bring things back up after PCI error
11367  * and/or reset recovery has finished.
11368  **/
11369 static void i40e_pci_error_resume(struct pci_dev *pdev)
11370 {
11371         struct i40e_pf *pf = pci_get_drvdata(pdev);
11372
11373         dev_dbg(&pdev->dev, "%s\n", __func__);
11374         if (test_bit(__I40E_SUSPENDED, &pf->state))
11375                 return;
11376
11377         rtnl_lock();
11378         i40e_handle_reset_warning(pf);
11379         rtnl_unlock();
11380 }
11381
11382 /**
11383  * i40e_shutdown - PCI callback for shutting down
11384  * @pdev: PCI device information struct
11385  **/
11386 static void i40e_shutdown(struct pci_dev *pdev)
11387 {
11388         struct i40e_pf *pf = pci_get_drvdata(pdev);
11389         struct i40e_hw *hw = &pf->hw;
11390
11391         set_bit(__I40E_SUSPENDED, &pf->state);
11392         set_bit(__I40E_DOWN, &pf->state);
11393         rtnl_lock();
11394         i40e_prep_for_reset(pf);
11395         rtnl_unlock();
11396
11397         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11398         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11399
11400         del_timer_sync(&pf->service_timer);
11401         cancel_work_sync(&pf->service_task);
11402         i40e_fdir_teardown(pf);
11403
11404         rtnl_lock();
11405         i40e_prep_for_reset(pf);
11406         rtnl_unlock();
11407
11408         wr32(hw, I40E_PFPM_APM,
11409              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11410         wr32(hw, I40E_PFPM_WUFC,
11411              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11412
11413         i40e_clear_interrupt_scheme(pf);
11414
11415         if (system_state == SYSTEM_POWER_OFF) {
11416                 pci_wake_from_d3(pdev, pf->wol_en);
11417                 pci_set_power_state(pdev, PCI_D3hot);
11418         }
11419 }
11420
11421 #ifdef CONFIG_PM
11422 /**
11423  * i40e_suspend - PCI callback for moving to D3
11424  * @pdev: PCI device information struct
11425  **/
11426 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11427 {
11428         struct i40e_pf *pf = pci_get_drvdata(pdev);
11429         struct i40e_hw *hw = &pf->hw;
11430
11431         set_bit(__I40E_SUSPENDED, &pf->state);
11432         set_bit(__I40E_DOWN, &pf->state);
11433
11434         rtnl_lock();
11435         i40e_prep_for_reset(pf);
11436         rtnl_unlock();
11437
11438         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11439         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11440
11441         pci_wake_from_d3(pdev, pf->wol_en);
11442         pci_set_power_state(pdev, PCI_D3hot);
11443
11444         return 0;
11445 }
11446
11447 /**
11448  * i40e_resume - PCI callback for waking up from D3
11449  * @pdev: PCI device information struct
11450  **/
11451 static int i40e_resume(struct pci_dev *pdev)
11452 {
11453         struct i40e_pf *pf = pci_get_drvdata(pdev);
11454         u32 err;
11455
11456         pci_set_power_state(pdev, PCI_D0);
11457         pci_restore_state(pdev);
11458         /* pci_restore_state() clears dev->state_saves, so
11459          * call pci_save_state() again to restore it.
11460          */
11461         pci_save_state(pdev);
11462
11463         err = pci_enable_device_mem(pdev);
11464         if (err) {
11465                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11466                 return err;
11467         }
11468         pci_set_master(pdev);
11469
11470         /* no wakeup events while running */
11471         pci_wake_from_d3(pdev, false);
11472
11473         /* handling the reset will rebuild the device state */
11474         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11475                 clear_bit(__I40E_DOWN, &pf->state);
11476                 rtnl_lock();
11477                 i40e_reset_and_rebuild(pf, false);
11478                 rtnl_unlock();
11479         }
11480
11481         return 0;
11482 }
11483
11484 #endif
11485 static const struct pci_error_handlers i40e_err_handler = {
11486         .error_detected = i40e_pci_error_detected,
11487         .slot_reset = i40e_pci_error_slot_reset,
11488         .resume = i40e_pci_error_resume,
11489 };
11490
11491 static struct pci_driver i40e_driver = {
11492         .name     = i40e_driver_name,
11493         .id_table = i40e_pci_tbl,
11494         .probe    = i40e_probe,
11495         .remove   = i40e_remove,
11496 #ifdef CONFIG_PM
11497         .suspend  = i40e_suspend,
11498         .resume   = i40e_resume,
11499 #endif
11500         .shutdown = i40e_shutdown,
11501         .err_handler = &i40e_err_handler,
11502         .sriov_configure = i40e_pci_sriov_configure,
11503 };
11504
11505 /**
11506  * i40e_init_module - Driver registration routine
11507  *
11508  * i40e_init_module is the first routine called when the driver is
11509  * loaded. All it does is register with the PCI subsystem.
11510  **/
11511 static int __init i40e_init_module(void)
11512 {
11513         pr_info("%s: %s - version %s\n", i40e_driver_name,
11514                 i40e_driver_string, i40e_driver_version_str);
11515         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11516
11517         i40e_dbg_init();
11518         return pci_register_driver(&i40e_driver);
11519 }
11520 module_init(i40e_init_module);
11521
11522 /**
11523  * i40e_exit_module - Driver exit cleanup routine
11524  *
11525  * i40e_exit_module is called just before the driver is removed
11526  * from memory.
11527  **/
11528 static void __exit i40e_exit_module(void)
11529 {
11530         pci_unregister_driver(&i40e_driver);
11531         i40e_dbg_exit();
11532 }
11533 module_exit(i40e_exit_module);